blob: 735362c26c8e5a3329d3cc6517b330167b393db6 [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{
Stephen Smalley37a9a8d2015-06-10 08:44:59 -0400143 UNIXCB(skb).secid = scm->secid;
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700144}
145
146static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb)
147{
Stephen Smalley37a9a8d2015-06-10 08:44:59 -0400148 scm->secid = UNIXCB(skb).secid;
149}
150
151static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb)
152{
153 return (scm->secid == UNIXCB(skb).secid);
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700154}
155#else
Catherine Zhangdc49c1f2006-08-02 14:12:06 -0700156static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb)
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700157{ }
158
159static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb)
160{ }
Stephen Smalley37a9a8d2015-06-10 08:44:59 -0400161
162static inline bool unix_secdata_eq(struct scm_cookie *scm, struct sk_buff *skb)
163{
164 return true;
165}
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700166#endif /* CONFIG_SECURITY_NETWORK */
167
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168/*
169 * SMP locking strategy:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800170 * hash table is protected with spinlock unix_table_lock
Stephen Hemminger663717f2010-02-18 14:12:06 -0800171 * each socket state is protected by separate spin lock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172 */
173
Eric Dumazet95c96172012-04-15 05:58:06 +0000174static inline unsigned int unix_hash_fold(__wsum n)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175{
Anton Blanchard0a134042014-03-05 14:29:58 +1100176 unsigned int hash = (__force unsigned int)csum_fold(n);
Eric Dumazet95c96172012-04-15 05:58:06 +0000177
Linus Torvalds1da177e2005-04-16 15:20:36 -0700178 hash ^= hash>>8;
179 return hash&(UNIX_HASH_SIZE-1);
180}
181
182#define unix_peer(sk) (unix_sk(sk)->peer)
183
184static inline int unix_our_peer(struct sock *sk, struct sock *osk)
185{
186 return unix_peer(osk) == sk;
187}
188
189static inline int unix_may_send(struct sock *sk, struct sock *osk)
190{
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800191 return unix_peer(osk) == NULL || unix_our_peer(sk, osk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192}
193
Rainer Weikusat3c734192008-06-17 22:28:05 -0700194static inline int unix_recvq_full(struct sock const *sk)
195{
196 return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog;
197}
198
Pavel Emelyanovfa7ff562011-12-15 02:44:03 +0000199struct sock *unix_peer_get(struct sock *s)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200{
201 struct sock *peer;
202
David S. Miller1c92b4e2007-05-31 13:24:26 -0700203 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700204 peer = unix_peer(s);
205 if (peer)
206 sock_hold(peer);
David S. Miller1c92b4e2007-05-31 13:24:26 -0700207 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208 return peer;
209}
Pavel Emelyanovfa7ff562011-12-15 02:44:03 +0000210EXPORT_SYMBOL_GPL(unix_peer_get);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700211
212static inline void unix_release_addr(struct unix_address *addr)
213{
214 if (atomic_dec_and_test(&addr->refcnt))
215 kfree(addr);
216}
217
218/*
219 * Check unix socket name:
220 * - should be not zero length.
221 * - if started by not zero, should be NULL terminated (FS object)
222 * - if started by zero, it is abstract name.
223 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900224
Eric Dumazet95c96172012-04-15 05:58:06 +0000225static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226{
227 if (len <= sizeof(short) || len > sizeof(*sunaddr))
228 return -EINVAL;
229 if (!sunaddr || sunaddr->sun_family != AF_UNIX)
230 return -EINVAL;
231 if (sunaddr->sun_path[0]) {
232 /*
233 * This may look like an off by one error but it is a bit more
234 * subtle. 108 is the longest valid AF_UNIX path for a binding.
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300235 * sun_path[108] doesn't as such exist. However in kernel space
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236 * we are guaranteed that it is a valid memory location in our
237 * kernel address buffer.
238 */
Jianjun Konge27dfce2008-11-01 21:38:31 -0700239 ((char *)sunaddr)[len] = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 len = strlen(sunaddr->sun_path)+1+sizeof(short);
241 return len;
242 }
243
Joe Perches07f07572008-11-19 15:44:53 -0800244 *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245 return len;
246}
247
248static void __unix_remove_socket(struct sock *sk)
249{
250 sk_del_node_init(sk);
251}
252
253static void __unix_insert_socket(struct hlist_head *list, struct sock *sk)
254{
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700255 WARN_ON(!sk_unhashed(sk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256 sk_add_node(sk, list);
257}
258
259static inline void unix_remove_socket(struct sock *sk)
260{
David S. Millerfbe9cc42005-12-13 23:26:29 -0800261 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 __unix_remove_socket(sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800263 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264}
265
266static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk)
267{
David S. Millerfbe9cc42005-12-13 23:26:29 -0800268 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269 __unix_insert_socket(list, sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800270 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271}
272
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800273static struct sock *__unix_find_socket_byname(struct net *net,
274 struct sockaddr_un *sunname,
Eric Dumazet95c96172012-04-15 05:58:06 +0000275 int len, int type, unsigned int hash)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276{
277 struct sock *s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278
Sasha Levinb67bfe02013-02-27 17:06:00 -0800279 sk_for_each(s, &unix_socket_table[hash ^ type]) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 struct unix_sock *u = unix_sk(s);
281
YOSHIFUJI Hideaki878628f2008-03-26 03:57:35 +0900282 if (!net_eq(sock_net(s), net))
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800283 continue;
284
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285 if (u->addr->len == len &&
286 !memcmp(u->addr->name, sunname, len))
287 goto found;
288 }
289 s = NULL;
290found:
291 return s;
292}
293
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800294static inline struct sock *unix_find_socket_byname(struct net *net,
295 struct sockaddr_un *sunname,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700296 int len, int type,
Eric Dumazet95c96172012-04-15 05:58:06 +0000297 unsigned int hash)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298{
299 struct sock *s;
300
David S. Millerfbe9cc42005-12-13 23:26:29 -0800301 spin_lock(&unix_table_lock);
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800302 s = __unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 if (s)
304 sock_hold(s);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800305 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306 return s;
307}
308
Eric W. Biederman6616f782010-06-13 03:35:48 +0000309static struct sock *unix_find_socket_byinode(struct inode *i)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310{
311 struct sock *s;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312
David S. Millerfbe9cc42005-12-13 23:26:29 -0800313 spin_lock(&unix_table_lock);
Sasha Levinb67bfe02013-02-27 17:06:00 -0800314 sk_for_each(s,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315 &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) {
Al Viro40ffe672012-03-14 21:54:32 -0400316 struct dentry *dentry = unix_sk(s)->path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
Miklos Szeredieb0a4a42016-05-20 22:13:45 +0200318 if (dentry && d_real_inode(dentry) == i) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 sock_hold(s);
320 goto found;
321 }
322 }
323 s = NULL;
324found:
David S. Millerfbe9cc42005-12-13 23:26:29 -0800325 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326 return s;
327}
328
Rainer Weikusat7d267272015-11-20 22:07:23 +0000329/* Support code for asymmetrically connected dgram sockets
330 *
331 * If a datagram socket is connected to a socket not itself connected
332 * to the first socket (eg, /dev/log), clients may only enqueue more
333 * messages if the present receive queue of the server socket is not
334 * "too large". This means there's a second writeability condition
335 * poll and sendmsg need to test. The dgram recv code will do a wake
336 * up on the peer_wait wait queue of a socket upon reception of a
337 * datagram which needs to be propagated to sleeping would-be writers
338 * since these might not have sent anything so far. This can't be
339 * accomplished via poll_wait because the lifetime of the server
340 * socket might be less than that of its clients if these break their
341 * association with it or if the server socket is closed while clients
342 * are still connected to it and there's no way to inform "a polling
343 * implementation" that it should let go of a certain wait queue
344 *
345 * In order to propagate a wake up, a wait_queue_t of the client
346 * socket is enqueued on the peer_wait queue of the server socket
347 * whose wake function does a wake_up on the ordinary client socket
348 * wait queue. This connection is established whenever a write (or
349 * poll for write) hit the flow control condition and broken when the
350 * association to the server socket is dissolved or after a wake up
351 * was relayed.
352 */
353
354static int unix_dgram_peer_wake_relay(wait_queue_t *q, unsigned mode, int flags,
355 void *key)
356{
357 struct unix_sock *u;
358 wait_queue_head_t *u_sleep;
359
360 u = container_of(q, struct unix_sock, peer_wake);
361
362 __remove_wait_queue(&unix_sk(u->peer_wake.private)->peer_wait,
363 q);
364 u->peer_wake.private = NULL;
365
366 /* relaying can only happen while the wq still exists */
367 u_sleep = sk_sleep(&u->sk);
368 if (u_sleep)
369 wake_up_interruptible_poll(u_sleep, key);
370
371 return 0;
372}
373
374static int unix_dgram_peer_wake_connect(struct sock *sk, struct sock *other)
375{
376 struct unix_sock *u, *u_other;
377 int rc;
378
379 u = unix_sk(sk);
380 u_other = unix_sk(other);
381 rc = 0;
382 spin_lock(&u_other->peer_wait.lock);
383
384 if (!u->peer_wake.private) {
385 u->peer_wake.private = other;
386 __add_wait_queue(&u_other->peer_wait, &u->peer_wake);
387
388 rc = 1;
389 }
390
391 spin_unlock(&u_other->peer_wait.lock);
392 return rc;
393}
394
395static void unix_dgram_peer_wake_disconnect(struct sock *sk,
396 struct sock *other)
397{
398 struct unix_sock *u, *u_other;
399
400 u = unix_sk(sk);
401 u_other = unix_sk(other);
402 spin_lock(&u_other->peer_wait.lock);
403
404 if (u->peer_wake.private == other) {
405 __remove_wait_queue(&u_other->peer_wait, &u->peer_wake);
406 u->peer_wake.private = NULL;
407 }
408
409 spin_unlock(&u_other->peer_wait.lock);
410}
411
412static void unix_dgram_peer_wake_disconnect_wakeup(struct sock *sk,
413 struct sock *other)
414{
415 unix_dgram_peer_wake_disconnect(sk, other);
416 wake_up_interruptible_poll(sk_sleep(sk),
417 POLLOUT |
418 POLLWRNORM |
419 POLLWRBAND);
420}
421
422/* preconditions:
423 * - unix_peer(sk) == other
424 * - association is stable
425 */
426static int unix_dgram_peer_wake_me(struct sock *sk, struct sock *other)
427{
428 int connected;
429
430 connected = unix_dgram_peer_wake_connect(sk, other);
431
432 if (unix_recvq_full(other))
433 return 1;
434
435 if (connected)
436 unix_dgram_peer_wake_disconnect(sk, other);
437
438 return 0;
439}
440
Eric Dumazet1586a582015-10-23 10:59:16 -0700441static int unix_writable(const struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442{
Eric Dumazet1586a582015-10-23 10:59:16 -0700443 return sk->sk_state != TCP_LISTEN &&
444 (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445}
446
447static void unix_write_space(struct sock *sk)
448{
Eric Dumazet43815482010-04-29 11:01:49 +0000449 struct socket_wq *wq;
450
451 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 if (unix_writable(sk)) {
Eric Dumazet43815482010-04-29 11:01:49 +0000453 wq = rcu_dereference(sk->sk_wq);
Herbert Xu1ce0bf52015-11-26 13:55:39 +0800454 if (skwq_has_sleeper(wq))
Eric Dumazet67426b72010-10-29 20:44:44 +0000455 wake_up_interruptible_sync_poll(&wq->wait,
456 POLLOUT | POLLWRNORM | POLLWRBAND);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +0800457 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 }
Eric Dumazet43815482010-04-29 11:01:49 +0000459 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460}
461
462/* When dgram socket disconnects (or changes its peer), we clear its receive
463 * queue of packets arrived from previous peer. First, it allows to do
464 * flow control based only on wmem_alloc; second, sk connected to peer
465 * may receive messages only from that peer. */
466static void unix_dgram_disconnected(struct sock *sk, struct sock *other)
467{
David S. Millerb03efcf2005-07-08 14:57:23 -0700468 if (!skb_queue_empty(&sk->sk_receive_queue)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 skb_queue_purge(&sk->sk_receive_queue);
470 wake_up_interruptible_all(&unix_sk(sk)->peer_wait);
471
472 /* If one link of bidirectional dgram pipe is disconnected,
473 * we signal error. Messages are lost. Do not make this,
474 * when peer was not connected to us.
475 */
476 if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) {
477 other->sk_err = ECONNRESET;
478 other->sk_error_report(other);
479 }
480 }
481}
482
483static void unix_sock_destructor(struct sock *sk)
484{
485 struct unix_sock *u = unix_sk(sk);
486
487 skb_queue_purge(&sk->sk_receive_queue);
488
Ilpo Järvinen547b7922008-07-25 21:43:18 -0700489 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
490 WARN_ON(!sk_unhashed(sk));
491 WARN_ON(sk->sk_socket);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 if (!sock_flag(sk, SOCK_DEAD)) {
wangweidong5cc208b2013-12-06 18:03:36 +0800493 pr_info("Attempt to release alive unix socket: %p\n", sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 return;
495 }
496
497 if (u->addr)
498 unix_release_addr(u->addr);
499
Eric Dumazet518de9b2010-10-26 14:22:44 -0700500 atomic_long_dec(&unix_nr_socks);
David S. Miller6f756a82008-11-23 17:34:03 -0800501 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800502 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
David S. Miller6f756a82008-11-23 17:34:03 -0800503 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504#ifdef UNIX_REFCNT_DEBUG
wangweidong5cc208b2013-12-06 18:03:36 +0800505 pr_debug("UNIX %p is destroyed, %ld are still alive.\n", sk,
Eric Dumazet518de9b2010-10-26 14:22:44 -0700506 atomic_long_read(&unix_nr_socks));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507#endif
508}
509
Paul Mooreded34e02013-03-25 03:18:33 +0000510static void unix_release_sock(struct sock *sk, int embrion)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511{
512 struct unix_sock *u = unix_sk(sk);
Al Viro40ffe672012-03-14 21:54:32 -0400513 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 struct sock *skpair;
515 struct sk_buff *skb;
516 int state;
517
518 unix_remove_socket(sk);
519
520 /* Clear state */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700521 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 sock_orphan(sk);
523 sk->sk_shutdown = SHUTDOWN_MASK;
Al Viro40ffe672012-03-14 21:54:32 -0400524 path = u->path;
525 u->path.dentry = NULL;
526 u->path.mnt = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 state = sk->sk_state;
528 sk->sk_state = TCP_CLOSE;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700529 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530
531 wake_up_interruptible_all(&u->peer_wait);
532
Jianjun Konge27dfce2008-11-01 21:38:31 -0700533 skpair = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534
Jianjun Konge27dfce2008-11-01 21:38:31 -0700535 if (skpair != NULL) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) {
David S. Miller1c92b4e2007-05-31 13:24:26 -0700537 unix_state_lock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 /* No more writes */
539 skpair->sk_shutdown = SHUTDOWN_MASK;
540 if (!skb_queue_empty(&sk->sk_receive_queue) || embrion)
541 skpair->sk_err = ECONNRESET;
David S. Miller1c92b4e2007-05-31 13:24:26 -0700542 unix_state_unlock(skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 skpair->sk_state_change(skpair);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +0800544 sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 }
Rainer Weikusat7d267272015-11-20 22:07:23 +0000546
547 unix_dgram_peer_wake_disconnect(sk, skpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548 sock_put(skpair); /* It may now die */
549 unix_peer(sk) = NULL;
550 }
551
552 /* Try to flush out this socket. Throw out buffers at least */
553
554 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
Jianjun Konge27dfce2008-11-01 21:38:31 -0700555 if (state == TCP_LISTEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 unix_release_sock(skb->sk, 1);
557 /* passed fds are erased in the kfree_skb hook */
Hannes Frederic Sowa73ed5d22015-11-10 16:23:15 +0100558 UNIXCB(skb).consumed = skb->len;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559 kfree_skb(skb);
560 }
561
Al Viro40ffe672012-03-14 21:54:32 -0400562 if (path.dentry)
563 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564
565 sock_put(sk);
566
567 /* ---- Socket is dead now and most probably destroyed ---- */
568
569 /*
Alan Coxe04dae82012-09-17 00:52:41 +0000570 * Fixme: BSD difference: In BSD all sockets connected to us get
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571 * ECONNRESET and we die on the spot. In Linux we behave
572 * like files and pipes do and wait for the last
573 * dereference.
574 *
575 * Can't we simply set sock->err?
576 *
577 * What the above comment does talk about? --ANK(980817)
578 */
579
Pavel Emelyanov9305cfa2007-11-10 22:06:01 -0800580 if (unix_tot_inflight)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900581 unix_gc(); /* Garbage collect fds */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582}
583
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000584static void init_peercred(struct sock *sk)
585{
586 put_pid(sk->sk_peer_pid);
587 if (sk->sk_peer_cred)
588 put_cred(sk->sk_peer_cred);
589 sk->sk_peer_pid = get_pid(task_tgid(current));
590 sk->sk_peer_cred = get_current_cred();
591}
592
593static void copy_peercred(struct sock *sk, struct sock *peersk)
594{
595 put_pid(sk->sk_peer_pid);
596 if (sk->sk_peer_cred)
597 put_cred(sk->sk_peer_cred);
598 sk->sk_peer_pid = get_pid(peersk->sk_peer_pid);
599 sk->sk_peer_cred = get_cred(peersk->sk_peer_cred);
600}
601
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602static int unix_listen(struct socket *sock, int backlog)
603{
604 int err;
605 struct sock *sk = sock->sk;
606 struct unix_sock *u = unix_sk(sk);
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000607 struct pid *old_pid = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608
609 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800610 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
611 goto out; /* Only stream/seqpacket sockets accept */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 err = -EINVAL;
613 if (!u->addr)
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800614 goto out; /* No listens on an unbound socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -0700615 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN)
617 goto out_unlock;
618 if (backlog > sk->sk_max_ack_backlog)
619 wake_up_interruptible_all(&u->peer_wait);
620 sk->sk_max_ack_backlog = backlog;
621 sk->sk_state = TCP_LISTEN;
622 /* set credentials so connect can copy them */
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000623 init_peercred(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624 err = 0;
625
626out_unlock:
David S. Miller1c92b4e2007-05-31 13:24:26 -0700627 unix_state_unlock(sk);
Eric W. Biederman109f6e32010-06-13 03:30:14 +0000628 put_pid(old_pid);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629out:
630 return err;
631}
632
633static int unix_release(struct socket *);
634static int unix_bind(struct socket *, struct sockaddr *, int);
635static int unix_stream_connect(struct socket *, struct sockaddr *,
636 int addr_len, int flags);
637static int unix_socketpair(struct socket *, struct socket *);
638static int unix_accept(struct socket *, struct socket *, int);
639static int unix_getname(struct socket *, struct sockaddr *, int *, int);
640static unsigned int unix_poll(struct file *, struct socket *, poll_table *);
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700641static unsigned int unix_dgram_poll(struct file *, struct socket *,
642 poll_table *);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643static int unix_ioctl(struct socket *, unsigned int, unsigned long);
644static int unix_shutdown(struct socket *, int);
Ying Xue1b784142015-03-02 15:37:48 +0800645static int unix_stream_sendmsg(struct socket *, struct msghdr *, size_t);
646static int unix_stream_recvmsg(struct socket *, struct msghdr *, size_t, int);
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +0200647static ssize_t unix_stream_sendpage(struct socket *, struct page *, int offset,
648 size_t size, int flags);
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +0200649static ssize_t unix_stream_splice_read(struct socket *, loff_t *ppos,
650 struct pipe_inode_info *, size_t size,
651 unsigned int flags);
Ying Xue1b784142015-03-02 15:37:48 +0800652static int unix_dgram_sendmsg(struct socket *, struct msghdr *, size_t);
653static int unix_dgram_recvmsg(struct socket *, struct msghdr *, size_t, int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654static int unix_dgram_connect(struct socket *, struct sockaddr *,
655 int, int);
Ying Xue1b784142015-03-02 15:37:48 +0800656static int unix_seqpacket_sendmsg(struct socket *, struct msghdr *, size_t);
657static int unix_seqpacket_recvmsg(struct socket *, struct msghdr *, size_t,
658 int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659
Sasha Levin12663bf2013-12-07 17:26:27 -0500660static int unix_set_peek_off(struct sock *sk, int val)
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000661{
662 struct unix_sock *u = unix_sk(sk);
663
Sasha Levin12663bf2013-12-07 17:26:27 -0500664 if (mutex_lock_interruptible(&u->readlock))
665 return -EINTR;
666
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000667 sk->sk_peek_off = val;
668 mutex_unlock(&u->readlock);
Sasha Levin12663bf2013-12-07 17:26:27 -0500669
670 return 0;
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000671}
672
673
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800674static const struct proto_ops unix_stream_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 .family = PF_UNIX,
676 .owner = THIS_MODULE,
677 .release = unix_release,
678 .bind = unix_bind,
679 .connect = unix_stream_connect,
680 .socketpair = unix_socketpair,
681 .accept = unix_accept,
682 .getname = unix_getname,
683 .poll = unix_poll,
684 .ioctl = unix_ioctl,
685 .listen = unix_listen,
686 .shutdown = unix_shutdown,
687 .setsockopt = sock_no_setsockopt,
688 .getsockopt = sock_no_getsockopt,
689 .sendmsg = unix_stream_sendmsg,
690 .recvmsg = unix_stream_recvmsg,
691 .mmap = sock_no_mmap,
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +0200692 .sendpage = unix_stream_sendpage,
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +0200693 .splice_read = unix_stream_splice_read,
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +0000694 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695};
696
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800697static const struct proto_ops unix_dgram_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700698 .family = PF_UNIX,
699 .owner = THIS_MODULE,
700 .release = unix_release,
701 .bind = unix_bind,
702 .connect = unix_dgram_connect,
703 .socketpair = unix_socketpair,
704 .accept = sock_no_accept,
705 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700706 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 .ioctl = unix_ioctl,
708 .listen = sock_no_listen,
709 .shutdown = unix_shutdown,
710 .setsockopt = sock_no_setsockopt,
711 .getsockopt = sock_no_getsockopt,
712 .sendmsg = unix_dgram_sendmsg,
713 .recvmsg = unix_dgram_recvmsg,
714 .mmap = sock_no_mmap,
715 .sendpage = sock_no_sendpage,
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000716 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700717};
718
Eric Dumazet90ddc4f2005-12-22 12:49:22 -0800719static const struct proto_ops unix_seqpacket_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720 .family = PF_UNIX,
721 .owner = THIS_MODULE,
722 .release = unix_release,
723 .bind = unix_bind,
724 .connect = unix_stream_connect,
725 .socketpair = unix_socketpair,
726 .accept = unix_accept,
727 .getname = unix_getname,
Rainer Weikusatec0d2152008-06-27 19:34:18 -0700728 .poll = unix_dgram_poll,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700729 .ioctl = unix_ioctl,
730 .listen = unix_listen,
731 .shutdown = unix_shutdown,
732 .setsockopt = sock_no_setsockopt,
733 .getsockopt = sock_no_getsockopt,
734 .sendmsg = unix_seqpacket_sendmsg,
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +0000735 .recvmsg = unix_seqpacket_recvmsg,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 .mmap = sock_no_mmap,
737 .sendpage = sock_no_sendpage,
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +0000738 .set_peek_off = unix_set_peek_off,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739};
740
741static struct proto unix_proto = {
Eric Dumazet248969a2008-11-17 00:00:30 -0800742 .name = "UNIX",
743 .owner = THIS_MODULE,
Eric Dumazet248969a2008-11-17 00:00:30 -0800744 .obj_size = sizeof(struct unix_sock),
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745};
746
Ingo Molnara09785a2006-07-03 00:25:12 -0700747/*
748 * AF_UNIX sockets do not interact with hardware, hence they
749 * dont trigger interrupts - so it's safe for them to have
750 * bh-unsafe locking for their sk_receive_queue.lock. Split off
751 * this special lock-class by reinitializing the spinlock key:
752 */
753static struct lock_class_key af_unix_sk_receive_queue_lock_key;
754
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500755static struct sock *unix_create1(struct net *net, struct socket *sock, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756{
757 struct sock *sk = NULL;
758 struct unix_sock *u;
759
Eric Dumazet518de9b2010-10-26 14:22:44 -0700760 atomic_long_inc(&unix_nr_socks);
761 if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files())
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762 goto out;
763
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500764 sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto, kern);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 if (!sk)
766 goto out;
767
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800768 sock_init_data(sock, sk);
Ingo Molnara09785a2006-07-03 00:25:12 -0700769 lockdep_set_class(&sk->sk_receive_queue.lock,
770 &af_unix_sk_receive_queue_lock_key);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771
772 sk->sk_write_space = unix_write_space;
Denis V. Luneva0a53c82007-12-11 04:19:17 -0800773 sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 sk->sk_destruct = unix_sock_destructor;
775 u = unix_sk(sk);
Al Viro40ffe672012-03-14 21:54:32 -0400776 u->path.dentry = NULL;
777 u->path.mnt = NULL;
Benjamin LaHaisefd19f322006-01-03 14:10:46 -0800778 spin_lock_init(&u->lock);
Al Viro516e0cc2008-07-26 00:39:17 -0400779 atomic_long_set(&u->inflight, 0);
Miklos Szeredi1fd05ba2007-07-11 14:22:39 -0700780 INIT_LIST_HEAD(&u->link);
Ingo Molnar57b47a52006-03-20 22:35:41 -0800781 mutex_init(&u->readlock); /* single task reading lock */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782 init_waitqueue_head(&u->peer_wait);
Rainer Weikusat7d267272015-11-20 22:07:23 +0000783 init_waitqueue_func_entry(&u->peer_wake, unix_dgram_peer_wake_relay);
Eric Dumazet7123aaa2012-06-08 05:03:21 +0000784 unix_insert_socket(unix_sockets_unbound(sk), sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700785out:
Pavel Emelyanov284b3272007-11-10 22:08:30 -0800786 if (sk == NULL)
Eric Dumazet518de9b2010-10-26 14:22:44 -0700787 atomic_long_dec(&unix_nr_socks);
Eric Dumazet920de802008-11-24 00:09:29 -0800788 else {
789 local_bh_disable();
Eric Dumazeta8076d82008-11-17 02:38:49 -0800790 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
Eric Dumazet920de802008-11-24 00:09:29 -0800791 local_bh_enable();
792 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 return sk;
794}
795
Eric Paris3f378b62009-11-05 22:18:14 -0800796static int unix_create(struct net *net, struct socket *sock, int protocol,
797 int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798{
799 if (protocol && protocol != PF_UNIX)
800 return -EPROTONOSUPPORT;
801
802 sock->state = SS_UNCONNECTED;
803
804 switch (sock->type) {
805 case SOCK_STREAM:
806 sock->ops = &unix_stream_ops;
807 break;
808 /*
809 * Believe it or not BSD has AF_UNIX, SOCK_RAW though
810 * nothing uses it.
811 */
812 case SOCK_RAW:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700813 sock->type = SOCK_DGRAM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 case SOCK_DGRAM:
815 sock->ops = &unix_dgram_ops;
816 break;
817 case SOCK_SEQPACKET:
818 sock->ops = &unix_seqpacket_ops;
819 break;
820 default:
821 return -ESOCKTNOSUPPORT;
822 }
823
Eric W. Biederman11aa9c22015-05-08 21:09:13 -0500824 return unix_create1(net, sock, kern) ? 0 : -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825}
826
827static int unix_release(struct socket *sock)
828{
829 struct sock *sk = sock->sk;
830
831 if (!sk)
832 return 0;
833
Paul Mooreded34e02013-03-25 03:18:33 +0000834 unix_release_sock(sk, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835 sock->sk = NULL;
836
Paul Mooreded34e02013-03-25 03:18:33 +0000837 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838}
839
840static int unix_autobind(struct socket *sock)
841{
842 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900843 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 struct unix_sock *u = unix_sk(sk);
845 static u32 ordernum = 1;
Eric Dumazet6eba6a32008-11-16 22:58:44 -0800846 struct unix_address *addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 int err;
Tetsuo Handa8df73ff2010-09-04 01:34:28 +0000848 unsigned int retries = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849
Sasha Levin37ab4fa2013-12-13 10:54:22 -0500850 err = mutex_lock_interruptible(&u->readlock);
851 if (err)
852 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853
854 err = 0;
855 if (u->addr)
856 goto out;
857
858 err = -ENOMEM;
Panagiotis Issaris0da974f2006-07-21 14:51:30 -0700859 addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 if (!addr)
861 goto out;
862
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 addr->name->sun_family = AF_UNIX;
864 atomic_set(&addr->refcnt, 1);
865
866retry:
867 addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short);
Joe Perches07f07572008-11-19 15:44:53 -0800868 addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869
David S. Millerfbe9cc42005-12-13 23:26:29 -0800870 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 ordernum = (ordernum+1)&0xFFFFF;
872
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800873 if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 addr->hash)) {
David S. Millerfbe9cc42005-12-13 23:26:29 -0800875 spin_unlock(&unix_table_lock);
Tetsuo Handa8df73ff2010-09-04 01:34:28 +0000876 /*
877 * __unix_find_socket_byname() may take long time if many names
878 * are already in use.
879 */
880 cond_resched();
881 /* Give up if all names seems to be in use. */
882 if (retries++ == 0xFFFFF) {
883 err = -ENOSPC;
884 kfree(addr);
885 goto out;
886 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 goto retry;
888 }
889 addr->hash ^= sk->sk_type;
890
891 __unix_remove_socket(sk);
892 u->addr = addr;
893 __unix_insert_socket(&unix_socket_table[addr->hash], sk);
David S. Millerfbe9cc42005-12-13 23:26:29 -0800894 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 err = 0;
896
Ingo Molnar57b47a52006-03-20 22:35:41 -0800897out: mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 return err;
899}
900
Denis V. Lunev097e66c2007-11-19 22:29:30 -0800901static struct sock *unix_find_other(struct net *net,
902 struct sockaddr_un *sunname, int len,
Eric Dumazet95c96172012-04-15 05:58:06 +0000903 int type, unsigned int hash, int *error)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904{
905 struct sock *u;
Al Viro421748e2008-08-02 01:04:36 -0400906 struct path path;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 int err = 0;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +0900908
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 if (sunname->sun_path[0]) {
Al Viro421748e2008-08-02 01:04:36 -0400910 struct inode *inode;
911 err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 if (err)
913 goto fail;
Miklos Szeredieb0a4a42016-05-20 22:13:45 +0200914 inode = d_real_inode(path.dentry);
Al Viro421748e2008-08-02 01:04:36 -0400915 err = inode_permission(inode, MAY_WRITE);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916 if (err)
917 goto put_fail;
918
919 err = -ECONNREFUSED;
Al Viro421748e2008-08-02 01:04:36 -0400920 if (!S_ISSOCK(inode->i_mode))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 goto put_fail;
Eric W. Biederman6616f782010-06-13 03:35:48 +0000922 u = unix_find_socket_byinode(inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 if (!u)
924 goto put_fail;
925
926 if (u->sk_type == type)
Al Viro68ac1232012-03-15 08:21:57 -0400927 touch_atime(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928
Al Viro421748e2008-08-02 01:04:36 -0400929 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930
Jianjun Konge27dfce2008-11-01 21:38:31 -0700931 err = -EPROTOTYPE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700932 if (u->sk_type != type) {
933 sock_put(u);
934 goto fail;
935 }
936 } else {
937 err = -ECONNREFUSED;
Jianjun Konge27dfce2008-11-01 21:38:31 -0700938 u = unix_find_socket_byname(net, sunname, len, type, hash);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 if (u) {
940 struct dentry *dentry;
Al Viro40ffe672012-03-14 21:54:32 -0400941 dentry = unix_sk(u)->path.dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 if (dentry)
Al Viro68ac1232012-03-15 08:21:57 -0400943 touch_atime(&unix_sk(u)->path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 } else
945 goto fail;
946 }
947 return u;
948
949put_fail:
Al Viro421748e2008-08-02 01:04:36 -0400950 path_put(&path);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951fail:
Jianjun Konge27dfce2008-11-01 21:38:31 -0700952 *error = err;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953 return NULL;
954}
955
Al Virod3607752016-03-25 15:21:09 -0400956static int unix_mknod(struct dentry *dentry, const struct path *path, umode_t mode,
Rainer Weikusatc845acb2016-01-03 18:56:38 +0000957 struct path *res)
Al Virofaf02012012-07-20 02:37:29 +0400958{
Rainer Weikusatc845acb2016-01-03 18:56:38 +0000959 int err;
Al Virofaf02012012-07-20 02:37:29 +0400960
Rainer Weikusatc845acb2016-01-03 18:56:38 +0000961 err = security_path_mknod(path, dentry, mode, 0);
Al Virofaf02012012-07-20 02:37:29 +0400962 if (!err) {
Rainer Weikusatc845acb2016-01-03 18:56:38 +0000963 err = vfs_mknod(d_inode(path->dentry), dentry, mode, 0);
Al Virofaf02012012-07-20 02:37:29 +0400964 if (!err) {
Rainer Weikusatc845acb2016-01-03 18:56:38 +0000965 res->mnt = mntget(path->mnt);
Al Virofaf02012012-07-20 02:37:29 +0400966 res->dentry = dget(dentry);
967 }
968 }
Rainer Weikusatc845acb2016-01-03 18:56:38 +0000969
Al Virofaf02012012-07-20 02:37:29 +0400970 return err;
971}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972
973static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
974{
975 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900976 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 struct unix_sock *u = unix_sk(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -0700978 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Al Virodae6ad82011-06-26 11:50:15 -0400979 char *sun_path = sunaddr->sun_path;
Rainer Weikusatc845acb2016-01-03 18:56:38 +0000980 int err, name_err;
Eric Dumazet95c96172012-04-15 05:58:06 +0000981 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700982 struct unix_address *addr;
983 struct hlist_head *list;
Rainer Weikusatc845acb2016-01-03 18:56:38 +0000984 struct path path;
985 struct dentry *dentry;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986
987 err = -EINVAL;
988 if (sunaddr->sun_family != AF_UNIX)
989 goto out;
990
Jianjun Konge27dfce2008-11-01 21:38:31 -0700991 if (addr_len == sizeof(short)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992 err = unix_autobind(sock);
993 goto out;
994 }
995
996 err = unix_mkname(sunaddr, addr_len, &hash);
997 if (err < 0)
998 goto out;
999 addr_len = err;
1000
Rainer Weikusatc845acb2016-01-03 18:56:38 +00001001 name_err = 0;
1002 dentry = NULL;
1003 if (sun_path[0]) {
1004 /* Get the parent directory, calculate the hash for last
1005 * component.
1006 */
1007 dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0);
1008
1009 if (IS_ERR(dentry)) {
1010 /* delay report until after 'already bound' check */
1011 name_err = PTR_ERR(dentry);
1012 dentry = NULL;
1013 }
1014 }
1015
Sasha Levin37ab4fa2013-12-13 10:54:22 -05001016 err = mutex_lock_interruptible(&u->readlock);
1017 if (err)
Rainer Weikusatc845acb2016-01-03 18:56:38 +00001018 goto out_path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019
1020 err = -EINVAL;
1021 if (u->addr)
1022 goto out_up;
1023
Rainer Weikusatc845acb2016-01-03 18:56:38 +00001024 if (name_err) {
1025 err = name_err == -EEXIST ? -EADDRINUSE : name_err;
1026 goto out_up;
1027 }
1028
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029 err = -ENOMEM;
1030 addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL);
1031 if (!addr)
1032 goto out_up;
1033
1034 memcpy(addr->name, sunaddr, addr_len);
1035 addr->len = addr_len;
1036 addr->hash = hash ^ sk->sk_type;
1037 atomic_set(&addr->refcnt, 1);
1038
Rainer Weikusatc845acb2016-01-03 18:56:38 +00001039 if (dentry) {
1040 struct path u_path;
Al Virofaf02012012-07-20 02:37:29 +04001041 umode_t mode = S_IFSOCK |
Al Viroce3b0f82009-03-29 19:08:22 -04001042 (SOCK_INODE(sock)->i_mode & ~current_umask());
Rainer Weikusatc845acb2016-01-03 18:56:38 +00001043 err = unix_mknod(dentry, &path, mode, &u_path);
Al Virofaf02012012-07-20 02:37:29 +04001044 if (err) {
1045 if (err == -EEXIST)
1046 err = -EADDRINUSE;
1047 unix_release_addr(addr);
1048 goto out_up;
1049 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 addr->hash = UNIX_HASH_SIZE;
Miklos Szeredieb0a4a42016-05-20 22:13:45 +02001051 hash = d_real_inode(dentry)->i_ino & (UNIX_HASH_SIZE - 1);
Al Virofaf02012012-07-20 02:37:29 +04001052 spin_lock(&unix_table_lock);
Rainer Weikusatc845acb2016-01-03 18:56:38 +00001053 u->path = u_path;
Al Virofaf02012012-07-20 02:37:29 +04001054 list = &unix_socket_table[hash];
1055 } else {
1056 spin_lock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 err = -EADDRINUSE;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001058 if (__unix_find_socket_byname(net, sunaddr, addr_len,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059 sk->sk_type, hash)) {
1060 unix_release_addr(addr);
1061 goto out_unlock;
1062 }
1063
1064 list = &unix_socket_table[addr->hash];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065 }
1066
1067 err = 0;
1068 __unix_remove_socket(sk);
1069 u->addr = addr;
1070 __unix_insert_socket(list, sk);
1071
1072out_unlock:
David S. Millerfbe9cc42005-12-13 23:26:29 -08001073 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074out_up:
Ingo Molnar57b47a52006-03-20 22:35:41 -08001075 mutex_unlock(&u->readlock);
Rainer Weikusatc845acb2016-01-03 18:56:38 +00001076out_path:
1077 if (dentry)
1078 done_path_create(&path, dentry);
1079
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080out:
1081 return err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082}
1083
David S. Miller278a3de2007-05-31 15:19:20 -07001084static void unix_state_double_lock(struct sock *sk1, struct sock *sk2)
1085{
1086 if (unlikely(sk1 == sk2) || !sk2) {
1087 unix_state_lock(sk1);
1088 return;
1089 }
1090 if (sk1 < sk2) {
1091 unix_state_lock(sk1);
1092 unix_state_lock_nested(sk2);
1093 } else {
1094 unix_state_lock(sk2);
1095 unix_state_lock_nested(sk1);
1096 }
1097}
1098
1099static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2)
1100{
1101 if (unlikely(sk1 == sk2) || !sk2) {
1102 unix_state_unlock(sk1);
1103 return;
1104 }
1105 unix_state_unlock(sk1);
1106 unix_state_unlock(sk2);
1107}
1108
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr,
1110 int alen, int flags)
1111{
1112 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001113 struct net *net = sock_net(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001114 struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115 struct sock *other;
Eric Dumazet95c96172012-04-15 05:58:06 +00001116 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001117 int err;
1118
1119 if (addr->sa_family != AF_UNSPEC) {
1120 err = unix_mkname(sunaddr, alen, &hash);
1121 if (err < 0)
1122 goto out;
1123 alen = err;
1124
1125 if (test_bit(SOCK_PASSCRED, &sock->flags) &&
1126 !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0)
1127 goto out;
1128
David S. Miller278a3de2007-05-31 15:19:20 -07001129restart:
Jianjun Konge27dfce2008-11-01 21:38:31 -07001130 other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 if (!other)
1132 goto out;
1133
David S. Miller278a3de2007-05-31 15:19:20 -07001134 unix_state_double_lock(sk, other);
1135
1136 /* Apparently VFS overslept socket death. Retry. */
1137 if (sock_flag(other, SOCK_DEAD)) {
1138 unix_state_double_unlock(sk, other);
1139 sock_put(other);
1140 goto restart;
1141 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142
1143 err = -EPERM;
1144 if (!unix_may_send(sk, other))
1145 goto out_unlock;
1146
1147 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1148 if (err)
1149 goto out_unlock;
1150
1151 } else {
1152 /*
1153 * 1003.1g breaking connected state with AF_UNSPEC
1154 */
1155 other = NULL;
David S. Miller278a3de2007-05-31 15:19:20 -07001156 unix_state_double_lock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 }
1158
1159 /*
1160 * If it was connected, reconnect.
1161 */
1162 if (unix_peer(sk)) {
1163 struct sock *old_peer = unix_peer(sk);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001164 unix_peer(sk) = other;
Rainer Weikusat7d267272015-11-20 22:07:23 +00001165 unix_dgram_peer_wake_disconnect_wakeup(sk, old_peer);
1166
David S. Miller278a3de2007-05-31 15:19:20 -07001167 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168
1169 if (other != old_peer)
1170 unix_dgram_disconnected(sk, old_peer);
1171 sock_put(old_peer);
1172 } else {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001173 unix_peer(sk) = other;
David S. Miller278a3de2007-05-31 15:19:20 -07001174 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 }
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001176 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177
1178out_unlock:
David S. Miller278a3de2007-05-31 15:19:20 -07001179 unix_state_double_unlock(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 sock_put(other);
1181out:
1182 return err;
1183}
1184
1185static long unix_wait_for_peer(struct sock *other, long timeo)
1186{
1187 struct unix_sock *u = unix_sk(other);
1188 int sched;
1189 DEFINE_WAIT(wait);
1190
1191 prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE);
1192
1193 sched = !sock_flag(other, SOCK_DEAD) &&
1194 !(other->sk_shutdown & RCV_SHUTDOWN) &&
Rainer Weikusat3c734192008-06-17 22:28:05 -07001195 unix_recvq_full(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196
David S. Miller1c92b4e2007-05-31 13:24:26 -07001197 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198
1199 if (sched)
1200 timeo = schedule_timeout(timeo);
1201
1202 finish_wait(&u->peer_wait, &wait);
1203 return timeo;
1204}
1205
1206static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr,
1207 int addr_len, int flags)
1208{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001209 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001211 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001212 struct unix_sock *u = unix_sk(sk), *newu, *otheru;
1213 struct sock *newsk = NULL;
1214 struct sock *other = NULL;
1215 struct sk_buff *skb = NULL;
Eric Dumazet95c96172012-04-15 05:58:06 +00001216 unsigned int hash;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 int st;
1218 int err;
1219 long timeo;
1220
1221 err = unix_mkname(sunaddr, addr_len, &hash);
1222 if (err < 0)
1223 goto out;
1224 addr_len = err;
1225
Joe Perchesf64f9e72009-11-29 16:55:45 -08001226 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr &&
1227 (err = unix_autobind(sock)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228 goto out;
1229
1230 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1231
1232 /* First of all allocate resources.
1233 If we will make it after state is locked,
1234 we will have to recheck all again in any case.
1235 */
1236
1237 err = -ENOMEM;
1238
1239 /* create new sock for complete connection */
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05001240 newsk = unix_create1(sock_net(sk), NULL, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 if (newsk == NULL)
1242 goto out;
1243
1244 /* Allocate skb for sending to listening sock */
1245 skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL);
1246 if (skb == NULL)
1247 goto out;
1248
1249restart:
1250 /* Find listening sock. */
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001251 other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252 if (!other)
1253 goto out;
1254
1255 /* Latch state of peer */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001256 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257
1258 /* Apparently VFS overslept socket death. Retry. */
1259 if (sock_flag(other, SOCK_DEAD)) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001260 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 sock_put(other);
1262 goto restart;
1263 }
1264
1265 err = -ECONNREFUSED;
1266 if (other->sk_state != TCP_LISTEN)
1267 goto out_unlock;
Tomoki Sekiyama77238f22009-10-18 23:17:37 -07001268 if (other->sk_shutdown & RCV_SHUTDOWN)
1269 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270
Rainer Weikusat3c734192008-06-17 22:28:05 -07001271 if (unix_recvq_full(other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 err = -EAGAIN;
1273 if (!timeo)
1274 goto out_unlock;
1275
1276 timeo = unix_wait_for_peer(other, timeo);
1277
1278 err = sock_intr_errno(timeo);
1279 if (signal_pending(current))
1280 goto out;
1281 sock_put(other);
1282 goto restart;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001283 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284
1285 /* Latch our state.
1286
Daniel Balutae5537bf2011-03-14 15:25:33 -07001287 It is tricky place. We need to grab our state lock and cannot
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 drop lock on peer. It is dangerous because deadlock is
1289 possible. Connect to self case and simultaneous
1290 attempt to connect are eliminated by checking socket
1291 state. other is TCP_LISTEN, if sk is TCP_LISTEN we
1292 check this before attempt to grab lock.
1293
1294 Well, and we have to recheck the state after socket locked.
1295 */
1296 st = sk->sk_state;
1297
1298 switch (st) {
1299 case TCP_CLOSE:
1300 /* This is ok... continue with connect */
1301 break;
1302 case TCP_ESTABLISHED:
1303 /* Socket is already connected */
1304 err = -EISCONN;
1305 goto out_unlock;
1306 default:
1307 err = -EINVAL;
1308 goto out_unlock;
1309 }
1310
David S. Miller1c92b4e2007-05-31 13:24:26 -07001311 unix_state_lock_nested(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312
1313 if (sk->sk_state != st) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001314 unix_state_unlock(sk);
1315 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 sock_put(other);
1317 goto restart;
1318 }
1319
David S. Miller3610cda2011-01-05 15:38:53 -08001320 err = security_unix_stream_connect(sk, other, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321 if (err) {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001322 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323 goto out_unlock;
1324 }
1325
1326 /* The way is open! Fastly set all the necessary fields... */
1327
1328 sock_hold(sk);
1329 unix_peer(newsk) = sk;
1330 newsk->sk_state = TCP_ESTABLISHED;
1331 newsk->sk_type = sk->sk_type;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001332 init_peercred(newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 newu = unix_sk(newsk);
Eric Dumazeteaefd112011-02-18 03:26:36 +00001334 RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 otheru = unix_sk(other);
1336
1337 /* copy address information from listening to new sock*/
1338 if (otheru->addr) {
1339 atomic_inc(&otheru->addr->refcnt);
1340 newu->addr = otheru->addr;
1341 }
Al Viro40ffe672012-03-14 21:54:32 -04001342 if (otheru->path.dentry) {
1343 path_get(&otheru->path);
1344 newu->path = otheru->path;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 }
1346
1347 /* Set credentials */
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001348 copy_peercred(sk, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350 sock->state = SS_CONNECTED;
1351 sk->sk_state = TCP_ESTABLISHED;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001352 sock_hold(newsk);
1353
Peter Zijlstra4e857c52014-03-17 18:06:10 +01001354 smp_mb__after_atomic(); /* sock_hold() does an atomic_inc() */
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001355 unix_peer(sk) = newsk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356
David S. Miller1c92b4e2007-05-31 13:24:26 -07001357 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358
1359 /* take ten and and send info to listening sock */
1360 spin_lock(&other->sk_receive_queue.lock);
1361 __skb_queue_tail(&other->sk_receive_queue, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 spin_unlock(&other->sk_receive_queue.lock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001363 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001364 other->sk_data_ready(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 sock_put(other);
1366 return 0;
1367
1368out_unlock:
1369 if (other)
David S. Miller1c92b4e2007-05-31 13:24:26 -07001370 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371
1372out:
Wei Yongjun40d44442009-02-25 00:32:45 +00001373 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374 if (newsk)
1375 unix_release_sock(newsk, 0);
1376 if (other)
1377 sock_put(other);
1378 return err;
1379}
1380
1381static int unix_socketpair(struct socket *socka, struct socket *sockb)
1382{
Jianjun Konge27dfce2008-11-01 21:38:31 -07001383 struct sock *ska = socka->sk, *skb = sockb->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384
1385 /* Join our sockets back to back */
1386 sock_hold(ska);
1387 sock_hold(skb);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001388 unix_peer(ska) = skb;
1389 unix_peer(skb) = ska;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001390 init_peercred(ska);
1391 init_peercred(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392
1393 if (ska->sk_type != SOCK_DGRAM) {
1394 ska->sk_state = TCP_ESTABLISHED;
1395 skb->sk_state = TCP_ESTABLISHED;
1396 socka->state = SS_CONNECTED;
1397 sockb->state = SS_CONNECTED;
1398 }
1399 return 0;
1400}
1401
Daniel Borkmann90c6bd32013-10-17 22:51:31 +02001402static void unix_sock_inherit_flags(const struct socket *old,
1403 struct socket *new)
1404{
1405 if (test_bit(SOCK_PASSCRED, &old->flags))
1406 set_bit(SOCK_PASSCRED, &new->flags);
1407 if (test_bit(SOCK_PASSSEC, &old->flags))
1408 set_bit(SOCK_PASSSEC, &new->flags);
1409}
1410
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411static int unix_accept(struct socket *sock, struct socket *newsock, int flags)
1412{
1413 struct sock *sk = sock->sk;
1414 struct sock *tsk;
1415 struct sk_buff *skb;
1416 int err;
1417
1418 err = -EOPNOTSUPP;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001419 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420 goto out;
1421
1422 err = -EINVAL;
1423 if (sk->sk_state != TCP_LISTEN)
1424 goto out;
1425
1426 /* If socket state is TCP_LISTEN it cannot change (for now...),
1427 * so that no locks are necessary.
1428 */
1429
1430 skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err);
1431 if (!skb) {
1432 /* This means receive shutdown. */
1433 if (err == 0)
1434 err = -EINVAL;
1435 goto out;
1436 }
1437
1438 tsk = skb->sk;
1439 skb_free_datagram(sk, skb);
1440 wake_up_interruptible(&unix_sk(sk)->peer_wait);
1441
1442 /* attach accepted sock to socket */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001443 unix_state_lock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444 newsock->state = SS_CONNECTED;
Daniel Borkmann90c6bd32013-10-17 22:51:31 +02001445 unix_sock_inherit_flags(sock, newsock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 sock_graft(tsk, newsock);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001447 unix_state_unlock(tsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448 return 0;
1449
1450out:
1451 return err;
1452}
1453
1454
1455static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer)
1456{
1457 struct sock *sk = sock->sk;
1458 struct unix_sock *u;
Cyrill Gorcunov13cfa972009-11-08 05:51:19 +00001459 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460 int err = 0;
1461
1462 if (peer) {
1463 sk = unix_peer_get(sk);
1464
1465 err = -ENOTCONN;
1466 if (!sk)
1467 goto out;
1468 err = 0;
1469 } else {
1470 sock_hold(sk);
1471 }
1472
1473 u = unix_sk(sk);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001474 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 if (!u->addr) {
1476 sunaddr->sun_family = AF_UNIX;
1477 sunaddr->sun_path[0] = 0;
1478 *uaddr_len = sizeof(short);
1479 } else {
1480 struct unix_address *addr = u->addr;
1481
1482 *uaddr_len = addr->len;
1483 memcpy(sunaddr, addr->name, *uaddr_len);
1484 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07001485 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 sock_put(sk);
1487out:
1488 return err;
1489}
1490
1491static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1492{
1493 int i;
1494
1495 scm->fp = UNIXCB(skb).fp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496 UNIXCB(skb).fp = NULL;
1497
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001498 for (i = scm->fp->count-1; i >= 0; i--)
Hannes Frederic Sowa415e3d32016-02-03 02:11:03 +01001499 unix_notinflight(scm->fp->user, scm->fp->fp[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500}
1501
Eric W. Biederman7361c362010-06-13 03:34:33 +00001502static void unix_destruct_scm(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503{
1504 struct scm_cookie scm;
1505 memset(&scm, 0, sizeof(scm));
Eric W. Biederman7361c362010-06-13 03:34:33 +00001506 scm.pid = UNIXCB(skb).pid;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001507 if (UNIXCB(skb).fp)
1508 unix_detach_fds(&scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509
1510 /* Alas, it calls VFS */
1511 /* So fscking what? fput() had been SMP-safe since the last Summer */
1512 scm_destroy(&scm);
1513 sock_wfree(skb);
1514}
1515
willy tarreau712f4aa2016-01-10 07:54:56 +01001516/*
1517 * The "user->unix_inflight" variable is protected by the garbage
1518 * collection lock, and we just read it locklessly here. If you go
1519 * over the limit, there might be a tiny race in actually noticing
1520 * it across threads. Tough.
1521 */
1522static inline bool too_many_unix_fds(struct task_struct *p)
1523{
1524 struct user_struct *user = current_user();
1525
1526 if (unlikely(user->unix_inflight > task_rlimit(p, RLIMIT_NOFILE)))
1527 return !capable(CAP_SYS_RESOURCE) && !capable(CAP_SYS_ADMIN);
1528 return false;
1529}
1530
Eric Dumazet25888e32010-11-25 04:11:39 +00001531#define MAX_RECURSION_LEVEL 4
1532
Miklos Szeredi62093442008-11-09 15:23:57 +01001533static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534{
1535 int i;
Eric Dumazet25888e32010-11-25 04:11:39 +00001536 unsigned char max_level = 0;
Eric Dumazet25888e32010-11-25 04:11:39 +00001537
willy tarreau712f4aa2016-01-10 07:54:56 +01001538 if (too_many_unix_fds(current))
1539 return -ETOOMANYREFS;
1540
Eric Dumazet25888e32010-11-25 04:11:39 +00001541 for (i = scm->fp->count - 1; i >= 0; i--) {
1542 struct sock *sk = unix_get_socket(scm->fp->fp[i]);
1543
David Herrmann3575dbf2016-02-02 18:17:54 +01001544 if (sk)
Eric Dumazet25888e32010-11-25 04:11:39 +00001545 max_level = max(max_level,
1546 unix_sk(sk)->recursion_level);
Eric Dumazet25888e32010-11-25 04:11:39 +00001547 }
1548 if (unlikely(max_level > MAX_RECURSION_LEVEL))
1549 return -ETOOMANYREFS;
Miklos Szeredi62093442008-11-09 15:23:57 +01001550
1551 /*
1552 * Need to duplicate file references for the sake of garbage
1553 * collection. Otherwise a socket in the fps might become a
1554 * candidate for GC while the skb is not yet queued.
1555 */
1556 UNIXCB(skb).fp = scm_fp_dup(scm->fp);
1557 if (!UNIXCB(skb).fp)
1558 return -ENOMEM;
1559
willy tarreau712f4aa2016-01-10 07:54:56 +01001560 for (i = scm->fp->count - 1; i >= 0; i--)
Hannes Frederic Sowa415e3d32016-02-03 02:11:03 +01001561 unix_inflight(scm->fp->user, scm->fp->fp[i]);
Eric Dumazet25888e32010-11-25 04:11:39 +00001562 return max_level;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563}
1564
David S. Millerf78a5fd2011-09-16 19:34:00 -04001565static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds)
Eric W. Biederman7361c362010-06-13 03:34:33 +00001566{
1567 int err = 0;
Eric Dumazet16e57262011-09-19 05:52:27 +00001568
David S. Millerf78a5fd2011-09-16 19:34:00 -04001569 UNIXCB(skb).pid = get_pid(scm->pid);
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00001570 UNIXCB(skb).uid = scm->creds.uid;
1571 UNIXCB(skb).gid = scm->creds.gid;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001572 UNIXCB(skb).fp = NULL;
Stephen Smalley37a9a8d2015-06-10 08:44:59 -04001573 unix_get_secdata(scm, skb);
Eric W. Biederman7361c362010-06-13 03:34:33 +00001574 if (scm->fp && send_fds)
1575 err = unix_attach_fds(scm, skb);
1576
1577 skb->destructor = unix_destruct_scm;
1578 return err;
1579}
1580
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01001581static bool unix_passcred_enabled(const struct socket *sock,
1582 const struct sock *other)
1583{
1584 return test_bit(SOCK_PASSCRED, &sock->flags) ||
1585 !other->sk_socket ||
1586 test_bit(SOCK_PASSCRED, &other->sk_socket->flags);
1587}
1588
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589/*
Eric Dumazet16e57262011-09-19 05:52:27 +00001590 * Some apps rely on write() giving SCM_CREDENTIALS
1591 * We include credentials if source or destination socket
1592 * asserted SOCK_PASSCRED.
1593 */
1594static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock,
1595 const struct sock *other)
1596{
Eric W. Biederman6b0ee8c02013-04-03 17:28:16 +00001597 if (UNIXCB(skb).pid)
Eric Dumazet16e57262011-09-19 05:52:27 +00001598 return;
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01001599 if (unix_passcred_enabled(sock, other)) {
Eric Dumazet16e57262011-09-19 05:52:27 +00001600 UNIXCB(skb).pid = get_pid(task_tgid(current));
David S. Miller6e0895c2013-04-22 20:32:51 -04001601 current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid);
Eric Dumazet16e57262011-09-19 05:52:27 +00001602 }
1603}
1604
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01001605static int maybe_init_creds(struct scm_cookie *scm,
1606 struct socket *socket,
1607 const struct sock *other)
1608{
1609 int err;
1610 struct msghdr msg = { .msg_controllen = 0 };
1611
1612 err = scm_send(socket, &msg, scm, false);
1613 if (err)
1614 return err;
1615
1616 if (unix_passcred_enabled(socket, other)) {
1617 scm->pid = get_pid(task_tgid(current));
1618 current_uid_gid(&scm->creds.uid, &scm->creds.gid);
1619 }
1620 return err;
1621}
1622
1623static bool unix_skb_scm_eq(struct sk_buff *skb,
1624 struct scm_cookie *scm)
1625{
1626 const struct unix_skb_parms *u = &UNIXCB(skb);
1627
1628 return u->pid == scm->pid &&
1629 uid_eq(u->uid, scm->creds.uid) &&
1630 gid_eq(u->gid, scm->creds.gid) &&
1631 unix_secdata_eq(scm, skb);
1632}
1633
Eric Dumazet16e57262011-09-19 05:52:27 +00001634/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 * Send AF_UNIX data.
1636 */
1637
Ying Xue1b784142015-03-02 15:37:48 +08001638static int unix_dgram_sendmsg(struct socket *sock, struct msghdr *msg,
1639 size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001642 struct net *net = sock_net(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643 struct unix_sock *u = unix_sk(sk);
Steffen Hurrle342dfc32014-01-17 22:53:15 +01001644 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, msg->msg_name);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 struct sock *other = NULL;
1646 int namelen = 0; /* fake GCC */
1647 int err;
Eric Dumazet95c96172012-04-15 05:58:06 +00001648 unsigned int hash;
David S. Millerf78a5fd2011-09-16 19:34:00 -04001649 struct sk_buff *skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 long timeo;
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001651 struct scm_cookie scm;
Eric Dumazet25888e32010-11-25 04:11:39 +00001652 int max_level;
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001653 int data_len = 0;
Rainer Weikusat7d267272015-11-20 22:07:23 +00001654 int sk_locked;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655
dann frazier5f23b732008-11-26 15:32:27 -08001656 wait_for_unix_gc();
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001657 err = scm_send(sock, msg, &scm, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 if (err < 0)
1659 return err;
1660
1661 err = -EOPNOTSUPP;
1662 if (msg->msg_flags&MSG_OOB)
1663 goto out;
1664
1665 if (msg->msg_namelen) {
1666 err = unix_mkname(sunaddr, msg->msg_namelen, &hash);
1667 if (err < 0)
1668 goto out;
1669 namelen = err;
1670 } else {
1671 sunaddr = NULL;
1672 err = -ENOTCONN;
1673 other = unix_peer_get(sk);
1674 if (!other)
1675 goto out;
1676 }
1677
Joe Perchesf64f9e72009-11-29 16:55:45 -08001678 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr
1679 && (err = unix_autobind(sock)) != 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680 goto out;
1681
1682 err = -EMSGSIZE;
1683 if (len > sk->sk_sndbuf - 32)
1684 goto out;
1685
Kirill Tkhai31ff6aa2014-05-15 19:56:28 +04001686 if (len > SKB_MAX_ALLOC) {
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001687 data_len = min_t(size_t,
1688 len - SKB_MAX_ALLOC,
1689 MAX_SKB_FRAGS * PAGE_SIZE);
Kirill Tkhai31ff6aa2014-05-15 19:56:28 +04001690 data_len = PAGE_ALIGN(data_len);
1691
1692 BUILD_BUG_ON(SKB_MAX_ALLOC < PAGE_SIZE);
1693 }
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001694
1695 skb = sock_alloc_send_pskb(sk, len - data_len, data_len,
Eric Dumazet28d64272013-08-08 14:38:47 -07001696 msg->msg_flags & MSG_DONTWAIT, &err,
1697 PAGE_ALLOC_COSTLY_ORDER);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001698 if (skb == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699 goto out;
1700
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001701 err = unix_scm_to_skb(&scm, skb, true);
Eric Dumazet25888e32010-11-25 04:11:39 +00001702 if (err < 0)
Eric W. Biederman7361c362010-06-13 03:34:33 +00001703 goto out_free;
Eric Dumazet25888e32010-11-25 04:11:39 +00001704 max_level = err + 1;
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001705
Eric Dumazeteb6a2482012-04-03 05:28:28 +00001706 skb_put(skb, len - data_len);
1707 skb->data_len = data_len;
1708 skb->len = len;
Al Viroc0371da2014-11-24 10:42:55 -05001709 err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 if (err)
1711 goto out_free;
1712
1713 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1714
1715restart:
1716 if (!other) {
1717 err = -ECONNRESET;
1718 if (sunaddr == NULL)
1719 goto out_free;
1720
Denis V. Lunev097e66c2007-11-19 22:29:30 -08001721 other = unix_find_other(net, sunaddr, namelen, sk->sk_type,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 hash, &err);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001723 if (other == NULL)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 goto out_free;
1725 }
1726
Alban Crequyd6ae3ba2011-01-18 06:39:15 +00001727 if (sk_filter(other, skb) < 0) {
1728 /* Toss the packet but do not return any error to the sender */
1729 err = len;
1730 goto out_free;
1731 }
1732
Rainer Weikusat7d267272015-11-20 22:07:23 +00001733 sk_locked = 0;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001734 unix_state_lock(other);
Rainer Weikusat7d267272015-11-20 22:07:23 +00001735restart_locked:
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 err = -EPERM;
1737 if (!unix_may_send(sk, other))
1738 goto out_unlock;
1739
Rainer Weikusat7d267272015-11-20 22:07:23 +00001740 if (unlikely(sock_flag(other, SOCK_DEAD))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001741 /*
1742 * Check with 1003.1g - what should
1743 * datagram error
1744 */
David S. Miller1c92b4e2007-05-31 13:24:26 -07001745 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746 sock_put(other);
1747
Rainer Weikusat7d267272015-11-20 22:07:23 +00001748 if (!sk_locked)
1749 unix_state_lock(sk);
1750
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751 err = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 if (unix_peer(sk) == other) {
Jianjun Konge27dfce2008-11-01 21:38:31 -07001753 unix_peer(sk) = NULL;
Rainer Weikusat7d267272015-11-20 22:07:23 +00001754 unix_dgram_peer_wake_disconnect_wakeup(sk, other);
1755
David S. Miller1c92b4e2007-05-31 13:24:26 -07001756 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757
1758 unix_dgram_disconnected(sk, other);
1759 sock_put(other);
1760 err = -ECONNREFUSED;
1761 } else {
David S. Miller1c92b4e2007-05-31 13:24:26 -07001762 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763 }
1764
1765 other = NULL;
1766 if (err)
1767 goto out_free;
1768 goto restart;
1769 }
1770
1771 err = -EPIPE;
1772 if (other->sk_shutdown & RCV_SHUTDOWN)
1773 goto out_unlock;
1774
1775 if (sk->sk_type != SOCK_SEQPACKET) {
1776 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1777 if (err)
1778 goto out_unlock;
1779 }
1780
Rainer Weikusata5527dd2016-02-11 19:37:27 +00001781 /* other == sk && unix_peer(other) != sk if
1782 * - unix_peer(sk) == NULL, destination address bound to sk
1783 * - unix_peer(sk) == sk by time of get but disconnected before lock
1784 */
1785 if (other != sk &&
1786 unlikely(unix_peer(other) != sk && unix_recvq_full(other))) {
Rainer Weikusat7d267272015-11-20 22:07:23 +00001787 if (timeo) {
1788 timeo = unix_wait_for_peer(other, timeo);
1789
1790 err = sock_intr_errno(timeo);
1791 if (signal_pending(current))
1792 goto out_free;
1793
1794 goto restart;
1795 }
1796
1797 if (!sk_locked) {
1798 unix_state_unlock(other);
1799 unix_state_double_lock(sk, other);
1800 }
1801
1802 if (unix_peer(sk) != other ||
1803 unix_dgram_peer_wake_me(sk, other)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804 err = -EAGAIN;
Rainer Weikusat7d267272015-11-20 22:07:23 +00001805 sk_locked = 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 goto out_unlock;
1807 }
1808
Rainer Weikusat7d267272015-11-20 22:07:23 +00001809 if (!sk_locked) {
1810 sk_locked = 1;
1811 goto restart_locked;
1812 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813 }
1814
Rainer Weikusat7d267272015-11-20 22:07:23 +00001815 if (unlikely(sk_locked))
1816 unix_state_unlock(sk);
1817
Alban Crequy3f661162010-10-04 08:48:28 +00001818 if (sock_flag(other, SOCK_RCVTSTAMP))
1819 __net_timestamp(skb);
Eric Dumazet16e57262011-09-19 05:52:27 +00001820 maybe_add_creds(skb, sock, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 skb_queue_tail(&other->sk_receive_queue, skb);
Eric Dumazet25888e32010-11-25 04:11:39 +00001822 if (max_level > unix_sk(other)->recursion_level)
1823 unix_sk(other)->recursion_level = max_level;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001824 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001825 other->sk_data_ready(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001826 sock_put(other);
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001827 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828 return len;
1829
1830out_unlock:
Rainer Weikusat7d267272015-11-20 22:07:23 +00001831 if (sk_locked)
1832 unix_state_unlock(sk);
David S. Miller1c92b4e2007-05-31 13:24:26 -07001833 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834out_free:
1835 kfree_skb(skb);
1836out:
1837 if (other)
1838 sock_put(other);
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001839 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840 return err;
1841}
1842
Eric Dumazete370a722013-08-08 14:37:32 -07001843/* We use paged skbs for stream sockets, and limit occupancy to 32768
1844 * bytes, and a minimun of a full page.
1845 */
1846#define UNIX_SKB_FRAGS_SZ (PAGE_SIZE << get_order(32768))
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001847
Ying Xue1b784142015-03-02 15:37:48 +08001848static int unix_stream_sendmsg(struct socket *sock, struct msghdr *msg,
1849 size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850{
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851 struct sock *sk = sock->sk;
1852 struct sock *other = NULL;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001853 int err, size;
David S. Millerf78a5fd2011-09-16 19:34:00 -04001854 struct sk_buff *skb;
Jianjun Konge27dfce2008-11-01 21:38:31 -07001855 int sent = 0;
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001856 struct scm_cookie scm;
Miklos Szeredi8ba69ba2009-09-11 11:31:45 -07001857 bool fds_sent = false;
Eric Dumazet25888e32010-11-25 04:11:39 +00001858 int max_level;
Eric Dumazete370a722013-08-08 14:37:32 -07001859 int data_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860
dann frazier5f23b732008-11-26 15:32:27 -08001861 wait_for_unix_gc();
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001862 err = scm_send(sock, msg, &scm, false);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 if (err < 0)
1864 return err;
1865
1866 err = -EOPNOTSUPP;
1867 if (msg->msg_flags&MSG_OOB)
1868 goto out_err;
1869
1870 if (msg->msg_namelen) {
1871 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1872 goto out_err;
1873 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 err = -ENOTCONN;
Benjamin LaHaise830a1e52005-12-13 23:22:32 -08001875 other = unix_peer(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001876 if (!other)
1877 goto out_err;
1878 }
1879
1880 if (sk->sk_shutdown & SEND_SHUTDOWN)
1881 goto pipe_err;
1882
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001883 while (sent < len) {
Eric Dumazete370a722013-08-08 14:37:32 -07001884 size = len - sent;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885
1886 /* Keep two messages in the pipe so it schedules better */
Eric Dumazete370a722013-08-08 14:37:32 -07001887 size = min_t(int, size, (sk->sk_sndbuf >> 1) - 64);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888
Eric Dumazete370a722013-08-08 14:37:32 -07001889 /* allow fallback to order-0 allocations */
1890 size = min_t(int, size, SKB_MAX_HEAD(0) + UNIX_SKB_FRAGS_SZ);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001891
Eric Dumazete370a722013-08-08 14:37:32 -07001892 data_len = max_t(int, 0, size - SKB_MAX_HEAD(0));
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09001893
Kirill Tkhai31ff6aa2014-05-15 19:56:28 +04001894 data_len = min_t(size_t, size, PAGE_ALIGN(data_len));
1895
Eric Dumazete370a722013-08-08 14:37:32 -07001896 skb = sock_alloc_send_pskb(sk, size - data_len, data_len,
Eric Dumazet28d64272013-08-08 14:38:47 -07001897 msg->msg_flags & MSG_DONTWAIT, &err,
1898 get_order(UNIX_SKB_FRAGS_SZ));
Eric Dumazete370a722013-08-08 14:37:32 -07001899 if (!skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 goto out_err;
1901
David S. Millerf78a5fd2011-09-16 19:34:00 -04001902 /* Only send the fds in the first buffer */
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001903 err = unix_scm_to_skb(&scm, skb, !fds_sent);
Eric Dumazet25888e32010-11-25 04:11:39 +00001904 if (err < 0) {
Eric W. Biederman7361c362010-06-13 03:34:33 +00001905 kfree_skb(skb);
David S. Millerf78a5fd2011-09-16 19:34:00 -04001906 goto out_err;
Miklos Szeredi62093442008-11-09 15:23:57 +01001907 }
Eric Dumazet25888e32010-11-25 04:11:39 +00001908 max_level = err + 1;
Eric W. Biederman7361c362010-06-13 03:34:33 +00001909 fds_sent = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910
Eric Dumazete370a722013-08-08 14:37:32 -07001911 skb_put(skb, size - data_len);
1912 skb->data_len = data_len;
1913 skb->len = size;
Al Viroc0371da2014-11-24 10:42:55 -05001914 err = skb_copy_datagram_from_iter(skb, 0, &msg->msg_iter, size);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001915 if (err) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 kfree_skb(skb);
David S. Millerf78a5fd2011-09-16 19:34:00 -04001917 goto out_err;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 }
1919
David S. Miller1c92b4e2007-05-31 13:24:26 -07001920 unix_state_lock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
1922 if (sock_flag(other, SOCK_DEAD) ||
1923 (other->sk_shutdown & RCV_SHUTDOWN))
1924 goto pipe_err_free;
1925
Eric Dumazet16e57262011-09-19 05:52:27 +00001926 maybe_add_creds(skb, sock, other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927 skb_queue_tail(&other->sk_receive_queue, skb);
Eric Dumazet25888e32010-11-25 04:11:39 +00001928 if (max_level > unix_sk(other)->recursion_level)
1929 unix_sk(other)->recursion_level = max_level;
David S. Miller1c92b4e2007-05-31 13:24:26 -07001930 unix_state_unlock(other);
David S. Miller676d2362014-04-11 16:15:36 -04001931 other->sk_data_ready(other);
Jianjun Konge27dfce2008-11-01 21:38:31 -07001932 sent += size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001935 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
1937 return sent;
1938
1939pipe_err_free:
David S. Miller1c92b4e2007-05-31 13:24:26 -07001940 unix_state_unlock(other);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 kfree_skb(skb);
1942pipe_err:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08001943 if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL))
1944 send_sig(SIGPIPE, current, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 err = -EPIPE;
1946out_err:
Christoph Hellwig7cc05662015-01-28 18:04:53 +01001947 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 return sent ? : err;
1949}
1950
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02001951static ssize_t unix_stream_sendpage(struct socket *socket, struct page *page,
1952 int offset, size_t size, int flags)
1953{
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01001954 int err;
1955 bool send_sigpipe = false;
1956 bool init_scm = true;
1957 struct scm_cookie scm;
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02001958 struct sock *other, *sk = socket->sk;
1959 struct sk_buff *skb, *newskb = NULL, *tail = NULL;
1960
1961 if (flags & MSG_OOB)
1962 return -EOPNOTSUPP;
1963
1964 other = unix_peer(sk);
1965 if (!other || sk->sk_state != TCP_ESTABLISHED)
1966 return -ENOTCONN;
1967
1968 if (false) {
1969alloc_skb:
1970 unix_state_unlock(other);
1971 mutex_unlock(&unix_sk(other)->readlock);
1972 newskb = sock_alloc_send_pskb(sk, 0, 0, flags & MSG_DONTWAIT,
1973 &err, 0);
1974 if (!newskb)
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01001975 goto err;
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02001976 }
1977
1978 /* we must acquire readlock as we modify already present
1979 * skbs in the sk_receive_queue and mess with skb->len
1980 */
1981 err = mutex_lock_interruptible(&unix_sk(other)->readlock);
1982 if (err) {
1983 err = flags & MSG_DONTWAIT ? -EAGAIN : -ERESTARTSYS;
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02001984 goto err;
1985 }
1986
1987 if (sk->sk_shutdown & SEND_SHUTDOWN) {
1988 err = -EPIPE;
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01001989 send_sigpipe = true;
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02001990 goto err_unlock;
1991 }
1992
1993 unix_state_lock(other);
1994
1995 if (sock_flag(other, SOCK_DEAD) ||
1996 other->sk_shutdown & RCV_SHUTDOWN) {
1997 err = -EPIPE;
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01001998 send_sigpipe = true;
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02001999 goto err_state_unlock;
2000 }
2001
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01002002 if (init_scm) {
2003 err = maybe_init_creds(&scm, socket, other);
2004 if (err)
2005 goto err_state_unlock;
2006 init_scm = false;
2007 }
2008
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002009 skb = skb_peek_tail(&other->sk_receive_queue);
2010 if (tail && tail == skb) {
2011 skb = newskb;
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01002012 } else if (!skb || !unix_skb_scm_eq(skb, &scm)) {
2013 if (newskb) {
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002014 skb = newskb;
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01002015 } else {
2016 tail = skb;
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002017 goto alloc_skb;
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01002018 }
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002019 } else if (newskb) {
2020 /* this is fast path, we don't necessarily need to
2021 * call to kfree_skb even though with newskb == NULL
2022 * this - does no harm
2023 */
2024 consume_skb(newskb);
Hannes Frederic Sowa8844f972015-11-16 16:25:56 +01002025 newskb = NULL;
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002026 }
2027
2028 if (skb_append_pagefrags(skb, page, offset, size)) {
2029 tail = skb;
2030 goto alloc_skb;
2031 }
2032
2033 skb->len += size;
2034 skb->data_len += size;
2035 skb->truesize += size;
2036 atomic_add(size, &sk->sk_wmem_alloc);
2037
Hannes Frederic Sowaa3a116e2015-11-17 15:10:59 +01002038 if (newskb) {
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01002039 err = unix_scm_to_skb(&scm, skb, false);
2040 if (err)
2041 goto err_state_unlock;
Hannes Frederic Sowaa3a116e2015-11-17 15:10:59 +01002042 spin_lock(&other->sk_receive_queue.lock);
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002043 __skb_queue_tail(&other->sk_receive_queue, newskb);
Hannes Frederic Sowaa3a116e2015-11-17 15:10:59 +01002044 spin_unlock(&other->sk_receive_queue.lock);
2045 }
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002046
2047 unix_state_unlock(other);
2048 mutex_unlock(&unix_sk(other)->readlock);
2049
2050 other->sk_data_ready(other);
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01002051 scm_destroy(&scm);
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002052 return size;
2053
2054err_state_unlock:
2055 unix_state_unlock(other);
2056err_unlock:
2057 mutex_unlock(&unix_sk(other)->readlock);
2058err:
2059 kfree_skb(newskb);
2060 if (send_sigpipe && !(flags & MSG_NOSIGNAL))
2061 send_sig(SIGPIPE, current, 0);
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01002062 if (!init_scm)
2063 scm_destroy(&scm);
Hannes Frederic Sowa869e7c62015-05-21 16:59:59 +02002064 return err;
2065}
2066
Ying Xue1b784142015-03-02 15:37:48 +08002067static int unix_seqpacket_sendmsg(struct socket *sock, struct msghdr *msg,
2068 size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069{
2070 int err;
2071 struct sock *sk = sock->sk;
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002072
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073 err = sock_error(sk);
2074 if (err)
2075 return err;
2076
2077 if (sk->sk_state != TCP_ESTABLISHED)
2078 return -ENOTCONN;
2079
2080 if (msg->msg_namelen)
2081 msg->msg_namelen = 0;
2082
Ying Xue1b784142015-03-02 15:37:48 +08002083 return unix_dgram_sendmsg(sock, msg, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002084}
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002085
Ying Xue1b784142015-03-02 15:37:48 +08002086static int unix_seqpacket_recvmsg(struct socket *sock, struct msghdr *msg,
2087 size_t size, int flags)
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +00002088{
2089 struct sock *sk = sock->sk;
2090
2091 if (sk->sk_state != TCP_ESTABLISHED)
2092 return -ENOTCONN;
2093
Ying Xue1b784142015-03-02 15:37:48 +08002094 return unix_dgram_recvmsg(sock, msg, size, flags);
Eric W. Biedermana05d2ad2011-04-24 01:54:57 +00002095}
2096
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097static void unix_copy_addr(struct msghdr *msg, struct sock *sk)
2098{
2099 struct unix_sock *u = unix_sk(sk);
2100
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101 if (u->addr) {
2102 msg->msg_namelen = u->addr->len;
2103 memcpy(msg->msg_name, u->addr->name, u->addr->len);
2104 }
2105}
2106
Ying Xue1b784142015-03-02 15:37:48 +08002107static int unix_dgram_recvmsg(struct socket *sock, struct msghdr *msg,
2108 size_t size, int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109{
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002110 struct scm_cookie scm;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111 struct sock *sk = sock->sk;
2112 struct unix_sock *u = unix_sk(sk);
Rainer Weikusat64874282015-12-06 21:11:38 +00002113 struct sk_buff *skb, *last;
2114 long timeo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115 int err;
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00002116 int peeked, skip;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117
2118 err = -EOPNOTSUPP;
2119 if (flags&MSG_OOB)
2120 goto out;
2121
Rainer Weikusat64874282015-12-06 21:11:38 +00002122 timeo = sock_rcvtimeo(sk, flags & MSG_DONTWAIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123
Rainer Weikusat64874282015-12-06 21:11:38 +00002124 do {
2125 mutex_lock(&u->readlock);
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00002126
Rainer Weikusat64874282015-12-06 21:11:38 +00002127 skip = sk_peek_offset(sk, flags);
2128 skb = __skb_try_recv_datagram(sk, flags, &peeked, &skip, &err,
2129 &last);
2130 if (skb)
2131 break;
2132
2133 mutex_unlock(&u->readlock);
2134
2135 if (err != -EAGAIN)
2136 break;
2137 } while (timeo &&
2138 !__skb_wait_for_more_packets(sk, &err, &timeo, last));
2139
2140 if (!skb) { /* implies readlock unlocked */
Florian Zumbiehl0a112252007-11-29 23:19:23 +11002141 unix_state_lock(sk);
2142 /* Signal EOF on disconnected non-blocking SEQPACKET socket. */
2143 if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN &&
2144 (sk->sk_shutdown & RCV_SHUTDOWN))
2145 err = 0;
2146 unix_state_unlock(sk);
Rainer Weikusat64874282015-12-06 21:11:38 +00002147 goto out;
Florian Zumbiehl0a112252007-11-29 23:19:23 +11002148 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149
Rainer Weikusat77b75f42015-11-26 19:23:15 +00002150 if (wq_has_sleeper(&u->peer_wait))
2151 wake_up_interruptible_sync_poll(&u->peer_wait,
2152 POLLOUT | POLLWRNORM |
2153 POLLWRBAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154
2155 if (msg->msg_name)
2156 unix_copy_addr(msg, skb->sk);
2157
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00002158 if (size > skb->len - skip)
2159 size = skb->len - skip;
2160 else if (size < skb->len - skip)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002161 msg->msg_flags |= MSG_TRUNC;
2162
David S. Miller51f3d022014-11-05 16:46:40 -05002163 err = skb_copy_datagram_msg(skb, skip, msg, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002164 if (err)
2165 goto out_free;
2166
Alban Crequy3f661162010-10-04 08:48:28 +00002167 if (sock_flag(sk, SOCK_RCVTSTAMP))
2168 __sock_recv_timestamp(msg, sk, skb);
2169
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002170 memset(&scm, 0, sizeof(scm));
2171
2172 scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid);
2173 unix_set_secdata(&scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002175 if (!(flags & MSG_PEEK)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176 if (UNIXCB(skb).fp)
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002177 unix_detach_fds(&scm, skb);
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00002178
2179 sk_peek_offset_bwd(sk, skb->len);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002180 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181 /* It is questionable: on PEEK we could:
2182 - do not return fds - good, but too simple 8)
2183 - return fds, and do not return them on read (old strategy,
2184 apparently wrong)
2185 - clone fds (I chose it for now, it is the most universal
2186 solution)
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002187
2188 POSIX 1003.1g does not actually define this clearly
2189 at all. POSIX 1003.1g doesn't define a lot of things
2190 clearly however!
2191
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192 */
Pavel Emelyanovf55bb7f2012-02-21 07:31:51 +00002193
2194 sk_peek_offset_fwd(sk, size);
2195
Linus Torvalds1da177e2005-04-16 15:20:36 -07002196 if (UNIXCB(skb).fp)
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002197 scm.fp = scm_fp_dup(UNIXCB(skb).fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002198 }
Eric Dumazet9f6f9af2012-02-21 23:24:55 +00002199 err = (flags & MSG_TRUNC) ? skb->len - skip : size;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002201 scm_recv(sock, msg, &scm, flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202
2203out_free:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002204 skb_free_datagram(sk, skb);
Ingo Molnar57b47a52006-03-20 22:35:41 -08002205 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206out:
2207 return err;
2208}
2209
2210/*
Benjamin Poirier79f632c2013-04-29 11:42:14 +00002211 * Sleep until more data has arrived. But check for races..
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 */
Benjamin Poirier79f632c2013-04-29 11:42:14 +00002213static long unix_stream_data_wait(struct sock *sk, long timeo,
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002214 struct sk_buff *last, unsigned int last_len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215{
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002216 struct sk_buff *tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217 DEFINE_WAIT(wait);
2218
David S. Miller1c92b4e2007-05-31 13:24:26 -07002219 unix_state_lock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220
2221 for (;;) {
Eric Dumazetaa395142010-04-20 13:03:51 +00002222 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002224 tail = skb_peek_tail(&sk->sk_receive_queue);
2225 if (tail != last ||
2226 (tail && tail->len != last_len) ||
Linus Torvalds1da177e2005-04-16 15:20:36 -07002227 sk->sk_err ||
2228 (sk->sk_shutdown & RCV_SHUTDOWN) ||
2229 signal_pending(current) ||
2230 !timeo)
2231 break;
2232
Eric Dumazet9cd3e072015-11-29 20:03:10 -08002233 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002234 unix_state_unlock(sk);
Colin Cross2b15af62013-05-06 23:50:21 +00002235 timeo = freezable_schedule_timeout(timeo);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002236 unix_state_lock(sk);
Mark Salyzynb48732e2015-05-26 08:22:19 -07002237
2238 if (sock_flag(sk, SOCK_DEAD))
2239 break;
2240
Eric Dumazet9cd3e072015-11-29 20:03:10 -08002241 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242 }
2243
Eric Dumazetaa395142010-04-20 13:03:51 +00002244 finish_wait(sk_sleep(sk), &wait);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002245 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246 return timeo;
2247}
2248
Eric Dumazete370a722013-08-08 14:37:32 -07002249static unsigned int unix_skb_len(const struct sk_buff *skb)
2250{
2251 return skb->len - UNIXCB(skb).consumed;
2252}
2253
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002254struct unix_stream_read_state {
2255 int (*recv_actor)(struct sk_buff *, int, int,
2256 struct unix_stream_read_state *);
2257 struct socket *socket;
2258 struct msghdr *msg;
2259 struct pipe_inode_info *pipe;
2260 size_t size;
2261 int flags;
2262 unsigned int splice_flags;
2263};
2264
2265static int unix_stream_read_generic(struct unix_stream_read_state *state)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266{
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002267 struct scm_cookie scm;
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002268 struct socket *sock = state->socket;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002269 struct sock *sk = sock->sk;
2270 struct unix_sock *u = unix_sk(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271 int copied = 0;
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002272 int flags = state->flags;
Eric Dumazetde144392014-03-25 18:42:27 -07002273 int noblock = flags & MSG_DONTWAIT;
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002274 bool check_creds = false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275 int target;
2276 int err = 0;
2277 long timeo;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002278 int skip;
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002279 size_t size = state->size;
2280 unsigned int last_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281
Rainer Weikusat1b92ee32016-02-08 18:47:19 +00002282 if (unlikely(sk->sk_state != TCP_ESTABLISHED)) {
2283 err = -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 goto out;
Rainer Weikusat1b92ee32016-02-08 18:47:19 +00002285 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286
Rainer Weikusat1b92ee32016-02-08 18:47:19 +00002287 if (unlikely(flags & MSG_OOB)) {
2288 err = -EOPNOTSUPP;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289 goto out;
Rainer Weikusat1b92ee32016-02-08 18:47:19 +00002290 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002292 target = sock_rcvlowat(sk, flags & MSG_WAITALL, size);
Eric Dumazetde144392014-03-25 18:42:27 -07002293 timeo = sock_rcvtimeo(sk, noblock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002295 memset(&scm, 0, sizeof(scm));
2296
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297 /* Lock the socket to prevent queue disordering
2298 * while sleeps in memcpy_tomsg
2299 */
Rainer Weikusat3822b5c2015-12-16 20:09:25 +00002300 mutex_lock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301
Andrey Vagine9193d62015-10-02 00:05:36 +03002302 if (flags & MSG_PEEK)
2303 skip = sk_peek_offset(sk, flags);
2304 else
2305 skip = 0;
2306
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002307 do {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308 int chunk;
Hannes Frederic Sowa73ed5d22015-11-10 16:23:15 +01002309 bool drop_skb;
Benjamin Poirier79f632c2013-04-29 11:42:14 +00002310 struct sk_buff *skb, *last;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311
Rainer Weikusat18eceb82016-02-18 12:39:46 +00002312redo:
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07002313 unix_state_lock(sk);
Mark Salyzynb48732e2015-05-26 08:22:19 -07002314 if (sock_flag(sk, SOCK_DEAD)) {
2315 err = -ECONNRESET;
2316 goto unlock;
2317 }
Benjamin Poirier79f632c2013-04-29 11:42:14 +00002318 last = skb = skb_peek(&sk->sk_receive_queue);
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002319 last_len = last ? last->len : 0;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002320again:
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002321 if (skb == NULL) {
Eric Dumazet25888e32010-11-25 04:11:39 +00002322 unix_sk(sk)->recursion_level = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 if (copied >= target)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07002324 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325
2326 /*
2327 * POSIX 1003.1g mandates this order.
2328 */
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002329
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002330 err = sock_error(sk);
2331 if (err)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07002332 goto unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333 if (sk->sk_shutdown & RCV_SHUTDOWN)
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07002334 goto unlock;
2335
2336 unix_state_unlock(sk);
Rainer Weikusat1b92ee32016-02-08 18:47:19 +00002337 if (!timeo) {
2338 err = -EAGAIN;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002339 break;
Rainer Weikusat1b92ee32016-02-08 18:47:19 +00002340 }
2341
Ingo Molnar57b47a52006-03-20 22:35:41 -08002342 mutex_unlock(&u->readlock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002344 timeo = unix_stream_data_wait(sk, timeo, last,
2345 last_len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002346
Rainer Weikusat3822b5c2015-12-16 20:09:25 +00002347 if (signal_pending(current)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 err = sock_intr_errno(timeo);
Eric Dumazetfa0dc042016-01-24 13:53:50 -08002349 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350 goto out;
2351 }
Rainer Weikusatb3ca9b02011-02-28 04:50:55 +00002352
Rainer Weikusat3822b5c2015-12-16 20:09:25 +00002353 mutex_lock(&u->readlock);
Rainer Weikusat18eceb82016-02-18 12:39:46 +00002354 goto redo;
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002355unlock:
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07002356 unix_state_unlock(sk);
2357 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358 }
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002359
Eric Dumazete370a722013-08-08 14:37:32 -07002360 while (skip >= unix_skb_len(skb)) {
2361 skip -= unix_skb_len(skb);
Benjamin Poirier79f632c2013-04-29 11:42:14 +00002362 last = skb;
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002363 last_len = skb->len;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002364 skb = skb_peek_next(skb, &sk->sk_receive_queue);
Benjamin Poirier79f632c2013-04-29 11:42:14 +00002365 if (!skb)
2366 goto again;
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002367 }
2368
Miklos Szeredi3c0d2f32007-06-05 13:10:29 -07002369 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002370
2371 if (check_creds) {
2372 /* Never glue messages from different writers */
Hannes Frederic Sowa9490f882015-11-26 12:08:18 +01002373 if (!unix_skb_scm_eq(skb, &scm))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374 break;
Eric W. Biederman0e82e7f6d2013-04-03 16:14:47 +00002375 } else if (test_bit(SOCK_PASSCRED, &sock->flags)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376 /* Copy credentials */
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002377 scm_set_cred(&scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid);
Stephen Smalley37a9a8d2015-06-10 08:44:59 -04002378 unix_set_secdata(&scm, skb);
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002379 check_creds = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380 }
2381
2382 /* Copy address just once */
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002383 if (state->msg && state->msg->msg_name) {
2384 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr,
2385 state->msg->msg_name);
2386 unix_copy_addr(state->msg, skb->sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387 sunaddr = NULL;
2388 }
2389
Eric Dumazete370a722013-08-08 14:37:32 -07002390 chunk = min_t(unsigned int, unix_skb_len(skb) - skip, size);
Hannes Frederic Sowa73ed5d22015-11-10 16:23:15 +01002391 skb_get(skb);
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002392 chunk = state->recv_actor(skb, skip, chunk, state);
Hannes Frederic Sowa73ed5d22015-11-10 16:23:15 +01002393 drop_skb = !unix_skb_len(skb);
2394 /* skb is only safe to use if !drop_skb */
2395 consume_skb(skb);
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002396 if (chunk < 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397 if (copied == 0)
2398 copied = -EFAULT;
2399 break;
2400 }
2401 copied += chunk;
2402 size -= chunk;
2403
Hannes Frederic Sowa73ed5d22015-11-10 16:23:15 +01002404 if (drop_skb) {
2405 /* the skb was touched by a concurrent reader;
2406 * we should not expect anything from this skb
2407 * anymore and assume it invalid - we can be
2408 * sure it was dropped from the socket queue
2409 *
2410 * let's report a short read
2411 */
2412 err = 0;
2413 break;
2414 }
2415
Linus Torvalds1da177e2005-04-16 15:20:36 -07002416 /* Mark read part of skb as used */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002417 if (!(flags & MSG_PEEK)) {
Eric Dumazete370a722013-08-08 14:37:32 -07002418 UNIXCB(skb).consumed += chunk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002419
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002420 sk_peek_offset_bwd(sk, chunk);
2421
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422 if (UNIXCB(skb).fp)
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002423 unix_detach_fds(&scm, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002424
Eric Dumazete370a722013-08-08 14:37:32 -07002425 if (unix_skb_len(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002427
Eric Dumazet6f01fd62012-01-28 16:11:03 +00002428 skb_unlink(skb, &sk->sk_receive_queue);
Neil Horman70d4bf62010-07-20 06:45:56 +00002429 consume_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002430
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002431 if (scm.fp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002432 break;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002433 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434 /* It is questionable, see note in unix_dgram_recvmsg.
2435 */
2436 if (UNIXCB(skb).fp)
Christoph Hellwig7cc05662015-01-28 18:04:53 +01002437 scm.fp = scm_fp_dup(UNIXCB(skb).fp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438
Andrey Vagine9193d62015-10-02 00:05:36 +03002439 sk_peek_offset_fwd(sk, chunk);
Pavel Emelyanovfc0d7532012-02-21 07:32:06 +00002440
Aaron Conole9f389e32015-09-26 18:50:43 -04002441 if (UNIXCB(skb).fp)
2442 break;
2443
Andrey Vagine9193d62015-10-02 00:05:36 +03002444 skip = 0;
Aaron Conole9f389e32015-09-26 18:50:43 -04002445 last = skb;
2446 last_len = skb->len;
2447 unix_state_lock(sk);
2448 skb = skb_peek_next(skb, &sk->sk_receive_queue);
2449 if (skb)
2450 goto again;
2451 unix_state_unlock(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 break;
2453 }
2454 } while (size);
2455
Ingo Molnar57b47a52006-03-20 22:35:41 -08002456 mutex_unlock(&u->readlock);
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002457 if (state->msg)
2458 scm_recv(sock, state->msg, &scm, flags);
2459 else
2460 scm_destroy(&scm);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002461out:
2462 return copied ? : err;
2463}
2464
Hannes Frederic Sowa2b514572015-05-21 17:00:01 +02002465static int unix_stream_read_actor(struct sk_buff *skb,
2466 int skip, int chunk,
2467 struct unix_stream_read_state *state)
2468{
2469 int ret;
2470
2471 ret = skb_copy_datagram_msg(skb, UNIXCB(skb).consumed + skip,
2472 state->msg, chunk);
2473 return ret ?: chunk;
2474}
2475
2476static int unix_stream_recvmsg(struct socket *sock, struct msghdr *msg,
2477 size_t size, int flags)
2478{
2479 struct unix_stream_read_state state = {
2480 .recv_actor = unix_stream_read_actor,
2481 .socket = sock,
2482 .msg = msg,
2483 .size = size,
2484 .flags = flags
2485 };
2486
2487 return unix_stream_read_generic(&state);
2488}
2489
2490static ssize_t skb_unix_socket_splice(struct sock *sk,
2491 struct pipe_inode_info *pipe,
2492 struct splice_pipe_desc *spd)
2493{
2494 int ret;
2495 struct unix_sock *u = unix_sk(sk);
2496
2497 mutex_unlock(&u->readlock);
2498 ret = splice_to_pipe(pipe, spd);
2499 mutex_lock(&u->readlock);
2500
2501 return ret;
2502}
2503
2504static int unix_stream_splice_actor(struct sk_buff *skb,
2505 int skip, int chunk,
2506 struct unix_stream_read_state *state)
2507{
2508 return skb_splice_bits(skb, state->socket->sk,
2509 UNIXCB(skb).consumed + skip,
2510 state->pipe, chunk, state->splice_flags,
2511 skb_unix_socket_splice);
2512}
2513
2514static ssize_t unix_stream_splice_read(struct socket *sock, loff_t *ppos,
2515 struct pipe_inode_info *pipe,
2516 size_t size, unsigned int flags)
2517{
2518 struct unix_stream_read_state state = {
2519 .recv_actor = unix_stream_splice_actor,
2520 .socket = sock,
2521 .pipe = pipe,
2522 .size = size,
2523 .splice_flags = flags,
2524 };
2525
2526 if (unlikely(*ppos))
2527 return -ESPIPE;
2528
2529 if (sock->file->f_flags & O_NONBLOCK ||
2530 flags & SPLICE_F_NONBLOCK)
2531 state.flags = MSG_DONTWAIT;
2532
2533 return unix_stream_read_generic(&state);
2534}
2535
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536static int unix_shutdown(struct socket *sock, int mode)
2537{
2538 struct sock *sk = sock->sk;
2539 struct sock *other;
2540
Xi Wangfc61b922012-08-26 16:47:13 +00002541 if (mode < SHUT_RD || mode > SHUT_RDWR)
2542 return -EINVAL;
2543 /* This maps:
2544 * SHUT_RD (0) -> RCV_SHUTDOWN (1)
2545 * SHUT_WR (1) -> SEND_SHUTDOWN (2)
2546 * SHUT_RDWR (2) -> SHUTDOWN_MASK (3)
2547 */
2548 ++mode;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549
Alban Crequy7180a032011-01-19 04:56:36 +00002550 unix_state_lock(sk);
2551 sk->sk_shutdown |= mode;
2552 other = unix_peer(sk);
2553 if (other)
2554 sock_hold(other);
2555 unix_state_unlock(sk);
2556 sk->sk_state_change(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557
Alban Crequy7180a032011-01-19 04:56:36 +00002558 if (other &&
2559 (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002560
Alban Crequy7180a032011-01-19 04:56:36 +00002561 int peer_mode = 0;
2562
2563 if (mode&RCV_SHUTDOWN)
2564 peer_mode |= SEND_SHUTDOWN;
2565 if (mode&SEND_SHUTDOWN)
2566 peer_mode |= RCV_SHUTDOWN;
2567 unix_state_lock(other);
2568 other->sk_shutdown |= peer_mode;
2569 unix_state_unlock(other);
2570 other->sk_state_change(other);
2571 if (peer_mode == SHUTDOWN_MASK)
2572 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP);
2573 else if (peer_mode & RCV_SHUTDOWN)
2574 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575 }
Alban Crequy7180a032011-01-19 04:56:36 +00002576 if (other)
2577 sock_put(other);
2578
Linus Torvalds1da177e2005-04-16 15:20:36 -07002579 return 0;
2580}
2581
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002582long unix_inq_len(struct sock *sk)
2583{
2584 struct sk_buff *skb;
2585 long amount = 0;
2586
2587 if (sk->sk_state == TCP_LISTEN)
2588 return -EINVAL;
2589
2590 spin_lock(&sk->sk_receive_queue.lock);
2591 if (sk->sk_type == SOCK_STREAM ||
2592 sk->sk_type == SOCK_SEQPACKET) {
2593 skb_queue_walk(&sk->sk_receive_queue, skb)
Eric Dumazete370a722013-08-08 14:37:32 -07002594 amount += unix_skb_len(skb);
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002595 } else {
2596 skb = skb_peek(&sk->sk_receive_queue);
2597 if (skb)
2598 amount = skb->len;
2599 }
2600 spin_unlock(&sk->sk_receive_queue.lock);
2601
2602 return amount;
2603}
2604EXPORT_SYMBOL_GPL(unix_inq_len);
2605
2606long unix_outq_len(struct sock *sk)
2607{
2608 return sk_wmem_alloc_get(sk);
2609}
2610EXPORT_SYMBOL_GPL(unix_outq_len);
2611
Linus Torvalds1da177e2005-04-16 15:20:36 -07002612static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2613{
2614 struct sock *sk = sock->sk;
Jianjun Konge27dfce2008-11-01 21:38:31 -07002615 long amount = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002616 int err;
2617
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002618 switch (cmd) {
2619 case SIOCOUTQ:
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002620 amount = unix_outq_len(sk);
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002621 err = put_user(amount, (int __user *)arg);
2622 break;
2623 case SIOCINQ:
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002624 amount = unix_inq_len(sk);
2625 if (amount < 0)
2626 err = amount;
2627 else
Linus Torvalds1da177e2005-04-16 15:20:36 -07002628 err = put_user(amount, (int __user *)arg);
Pavel Emelyanov885ee742011-12-30 00:54:11 +00002629 break;
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002630 default:
2631 err = -ENOIOCTLCMD;
2632 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002633 }
2634 return err;
2635}
2636
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002637static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002638{
2639 struct sock *sk = sock->sk;
2640 unsigned int mask;
2641
Eric Dumazetaa395142010-04-20 13:03:51 +00002642 sock_poll_wait(file, sk_sleep(sk), wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002643 mask = 0;
2644
2645 /* exceptional events? */
2646 if (sk->sk_err)
2647 mask |= POLLERR;
2648 if (sk->sk_shutdown == SHUTDOWN_MASK)
2649 mask |= POLLHUP;
Davide Libenzif348d702006-03-25 03:07:39 -08002650 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazetdb409802010-09-06 11:13:50 +00002651 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652
2653 /* readable? */
Eric Dumazetdb409802010-09-06 11:13:50 +00002654 if (!skb_queue_empty(&sk->sk_receive_queue))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002655 mask |= POLLIN | POLLRDNORM;
2656
2657 /* Connection-based need to check for termination and startup */
Eric Dumazet6eba6a32008-11-16 22:58:44 -08002658 if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) &&
2659 sk->sk_state == TCP_CLOSE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002660 mask |= POLLHUP;
2661
2662 /*
2663 * we set writable also when the other side has shut down the
2664 * connection. This prevents stuck sockets.
2665 */
2666 if (unix_writable(sk))
2667 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2668
2669 return mask;
2670}
2671
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002672static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
2673 poll_table *wait)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002674{
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002675 struct sock *sk = sock->sk, *other;
2676 unsigned int mask, writable;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002677
Eric Dumazetaa395142010-04-20 13:03:51 +00002678 sock_poll_wait(file, sk_sleep(sk), wait);
Rainer Weikusat3c734192008-06-17 22:28:05 -07002679 mask = 0;
2680
2681 /* exceptional events? */
2682 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +00002683 mask |= POLLERR |
Jacob Keller8facd5f2013-04-02 13:55:40 -07002684 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +00002685
Rainer Weikusat3c734192008-06-17 22:28:05 -07002686 if (sk->sk_shutdown & RCV_SHUTDOWN)
Eric Dumazet5456f092010-10-31 05:36:23 +00002687 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
Rainer Weikusat3c734192008-06-17 22:28:05 -07002688 if (sk->sk_shutdown == SHUTDOWN_MASK)
2689 mask |= POLLHUP;
2690
2691 /* readable? */
Eric Dumazet5456f092010-10-31 05:36:23 +00002692 if (!skb_queue_empty(&sk->sk_receive_queue))
Rainer Weikusat3c734192008-06-17 22:28:05 -07002693 mask |= POLLIN | POLLRDNORM;
2694
2695 /* Connection-based need to check for termination and startup */
2696 if (sk->sk_type == SOCK_SEQPACKET) {
2697 if (sk->sk_state == TCP_CLOSE)
2698 mask |= POLLHUP;
2699 /* connection hasn't started yet? */
2700 if (sk->sk_state == TCP_SYN_SENT)
2701 return mask;
2702 }
2703
Eric Dumazet973a34a2010-10-31 05:38:25 +00002704 /* No write status requested, avoid expensive OUT tests. */
Hans Verkuil626cf232012-03-23 15:02:27 -07002705 if (!(poll_requested_events(wait) & (POLLWRBAND|POLLWRNORM|POLLOUT)))
Eric Dumazet973a34a2010-10-31 05:38:25 +00002706 return mask;
2707
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002708 writable = unix_writable(sk);
Rainer Weikusat7d267272015-11-20 22:07:23 +00002709 if (writable) {
2710 unix_state_lock(sk);
2711
2712 other = unix_peer(sk);
2713 if (other && unix_peer(other) != sk &&
2714 unix_recvq_full(other) &&
2715 unix_dgram_peer_wake_me(sk, other))
2716 writable = 0;
2717
2718 unix_state_unlock(sk);
Rainer Weikusatec0d2152008-06-27 19:34:18 -07002719 }
2720
2721 if (writable)
Rainer Weikusat3c734192008-06-17 22:28:05 -07002722 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2723 else
Eric Dumazet9cd3e072015-11-29 20:03:10 -08002724 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
Rainer Weikusat3c734192008-06-17 22:28:05 -07002725
Rainer Weikusat3c734192008-06-17 22:28:05 -07002726 return mask;
2727}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002728
2729#ifdef CONFIG_PROC_FS
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002730
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002731#define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1)
2732
2733#define get_bucket(x) ((x) >> BUCKET_SPACE)
2734#define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1))
2735#define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o))
Pavel Emelyanova53eb3f2007-11-23 20:30:01 +08002736
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002737static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002738{
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002739 unsigned long offset = get_offset(*pos);
2740 unsigned long bucket = get_bucket(*pos);
2741 struct sock *sk;
2742 unsigned long count = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002743
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002744 for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) {
2745 if (sock_net(sk) != seq_file_net(seq))
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002746 continue;
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002747 if (++count == offset)
2748 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749 }
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002750
2751 return sk;
2752}
2753
2754static struct sock *unix_next_socket(struct seq_file *seq,
2755 struct sock *sk,
2756 loff_t *pos)
2757{
2758 unsigned long bucket;
2759
2760 while (sk > (struct sock *)SEQ_START_TOKEN) {
2761 sk = sk_next(sk);
2762 if (!sk)
2763 goto next_bucket;
2764 if (sock_net(sk) == seq_file_net(seq))
2765 return sk;
2766 }
2767
2768 do {
2769 sk = unix_from_bucket(seq, pos);
2770 if (sk)
2771 return sk;
2772
2773next_bucket:
2774 bucket = get_bucket(*pos) + 1;
2775 *pos = set_bucket_offset(bucket, 1);
2776 } while (bucket < ARRAY_SIZE(unix_socket_table));
2777
Linus Torvalds1da177e2005-04-16 15:20:36 -07002778 return NULL;
2779}
2780
Linus Torvalds1da177e2005-04-16 15:20:36 -07002781static void *unix_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002782 __acquires(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002783{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002784 spin_lock(&unix_table_lock);
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002785
2786 if (!*pos)
2787 return SEQ_START_TOKEN;
2788
2789 if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table))
2790 return NULL;
2791
2792 return unix_next_socket(seq, NULL, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002793}
2794
2795static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2796{
2797 ++*pos;
Eric Dumazet7123aaa2012-06-08 05:03:21 +00002798 return unix_next_socket(seq, v, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002799}
2800
2801static void unix_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002802 __releases(unix_table_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002803{
David S. Millerfbe9cc42005-12-13 23:26:29 -08002804 spin_unlock(&unix_table_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002805}
2806
2807static int unix_seq_show(struct seq_file *seq, void *v)
2808{
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002809
Joe Perchesb9f31242008-04-12 19:04:38 -07002810 if (v == SEQ_START_TOKEN)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002811 seq_puts(seq, "Num RefCount Protocol Flags Type St "
2812 "Inode Path\n");
2813 else {
2814 struct sock *s = v;
2815 struct unix_sock *u = unix_sk(s);
David S. Miller1c92b4e2007-05-31 13:24:26 -07002816 unix_state_lock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002817
Dan Rosenberg71338aa2011-05-23 12:17:35 +00002818 seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu",
Linus Torvalds1da177e2005-04-16 15:20:36 -07002819 s,
2820 atomic_read(&s->sk_refcnt),
2821 0,
2822 s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0,
2823 s->sk_type,
2824 s->sk_socket ?
2825 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) :
2826 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING),
2827 sock_i_ino(s));
2828
2829 if (u->addr) {
2830 int i, len;
2831 seq_putc(seq, ' ');
2832
2833 i = 0;
2834 len = u->addr->len - sizeof(short);
2835 if (!UNIX_ABSTRACT(s))
2836 len--;
2837 else {
2838 seq_putc(seq, '@');
2839 i++;
2840 }
2841 for ( ; i < len; i++)
2842 seq_putc(seq, u->addr->name->sun_path[i]);
2843 }
David S. Miller1c92b4e2007-05-31 13:24:26 -07002844 unix_state_unlock(s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002845 seq_putc(seq, '\n');
2846 }
2847
2848 return 0;
2849}
2850
Philippe De Muyter56b3d972007-07-10 23:07:31 -07002851static const struct seq_operations unix_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002852 .start = unix_seq_start,
2853 .next = unix_seq_next,
2854 .stop = unix_seq_stop,
2855 .show = unix_seq_show,
2856};
2857
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858static int unix_seq_open(struct inode *inode, struct file *file)
2859{
Denis V. Luneve372c412007-11-19 22:31:54 -08002860 return seq_open_net(inode, file, &unix_seq_ops,
Eric Dumazet8b51b062012-06-08 22:10:20 +00002861 sizeof(struct seq_net_private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002862}
2863
Arjan van de Venda7071d2007-02-12 00:55:36 -08002864static const struct file_operations unix_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865 .owner = THIS_MODULE,
2866 .open = unix_seq_open,
2867 .read = seq_read,
2868 .llseek = seq_lseek,
Denis V. Luneve372c412007-11-19 22:31:54 -08002869 .release = seq_release_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002870};
2871
2872#endif
2873
Stephen Hemmingerec1b4cf2009-10-05 05:58:39 +00002874static const struct net_proto_family unix_family_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002875 .family = PF_UNIX,
2876 .create = unix_create,
2877 .owner = THIS_MODULE,
2878};
2879
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002880
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002881static int __net_init unix_net_init(struct net *net)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002882{
2883 int error = -ENOMEM;
2884
Denis V. Luneva0a53c82007-12-11 04:19:17 -08002885 net->unx.sysctl_max_dgram_qlen = 10;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002886 if (unix_sysctl_register(net))
2887 goto out;
Pavel Emelyanovd392e492007-12-01 23:44:15 +11002888
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002889#ifdef CONFIG_PROC_FS
Gao fengd4beaa62013-02-18 01:34:54 +00002890 if (!proc_create("unix", 0, net->proc_net, &unix_seq_fops)) {
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002891 unix_sysctl_unregister(net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002892 goto out;
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002893 }
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002894#endif
2895 error = 0;
2896out:
Jianjun Kong48dcc33e2008-11-01 21:37:27 -07002897 return error;
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002898}
2899
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002900static void __net_exit unix_net_exit(struct net *net)
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002901{
Pavel Emelyanov1597fbc2007-12-01 23:51:01 +11002902 unix_sysctl_unregister(net);
Gao fengece31ff2013-02-18 01:34:56 +00002903 remove_proc_entry("unix", net->proc_net);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002904}
2905
2906static struct pernet_operations unix_net_ops = {
2907 .init = unix_net_init,
2908 .exit = unix_net_exit,
2909};
2910
Linus Torvalds1da177e2005-04-16 15:20:36 -07002911static int __init af_unix_init(void)
2912{
2913 int rc = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002914
YOSHIFUJI Hideaki / 吉藤英明b4fff5f2013-01-09 07:20:07 +00002915 BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002916
2917 rc = proto_register(&unix_proto, 1);
YOSHIFUJI Hideakiac7bfa62007-02-09 23:25:23 +09002918 if (rc != 0) {
wangweidong5cc208b2013-12-06 18:03:36 +08002919 pr_crit("%s: Cannot create unix_sock SLAB cache!\n", __func__);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002920 goto out;
2921 }
2922
2923 sock_register(&unix_family_ops);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002924 register_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002925out:
2926 return rc;
2927}
2928
2929static void __exit af_unix_exit(void)
2930{
2931 sock_unregister(PF_UNIX);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002932 proto_unregister(&unix_proto);
Denis V. Lunev097e66c2007-11-19 22:29:30 -08002933 unregister_pernet_subsys(&unix_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002934}
2935
David Woodhouse3d366962008-04-24 00:59:25 -07002936/* Earlier than device_initcall() so that other drivers invoking
2937 request_module() don't end up in a loop when modprobe tries
2938 to use a UNIX socket. But later than subsys_initcall() because
2939 we depend on stuff initialised there */
2940fs_initcall(af_unix_init);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002941module_exit(af_unix_exit);
2942
2943MODULE_LICENSE("GPL");
2944MODULE_ALIAS_NETPROTO(PF_UNIX);