import PULS_20180308
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / net / unix / af_unix.c
CommitLineData
1da177e4
LT
1/*
2 * NET4: Implementation of BSD Unix domain sockets.
3 *
113aa838 4 * Authors: Alan Cox, <alan@lxorguk.ukuu.org.uk>
1da177e4
LT
5 *
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 *
1da177e4
LT
11 * Fixes:
12 * Linus Torvalds : Assorted bug cures.
13 * Niibe Yutaka : async I/O support.
14 * Carsten Paeth : PF_UNIX check, address fixes.
15 * Alan Cox : Limit size of allocated blocks.
16 * Alan Cox : Fixed the stupid socketpair bug.
17 * Alan Cox : BSD compatibility fine tuning.
18 * Alan Cox : Fixed a bug in connect when interrupted.
19 * Alan Cox : Sorted out a proper draft version of
20 * file descriptor passing hacked up from
21 * Mike Shaver's work.
22 * Marty Leisner : Fixes to fd passing
23 * Nick Nevin : recvmsg bugfix.
24 * Alan Cox : Started proper garbage collector
25 * Heiko EiBfeldt : Missing verify_area check
26 * Alan Cox : Started POSIXisms
27 * Andreas Schwab : Replace inode by dentry for proper
28 * reference counting
29 * Kirk Petersen : Made this a module
30 * Christoph Rohland : Elegant non-blocking accept/connect algorithm.
31 * Lots of bug fixes.
32 * Alexey Kuznetosv : Repaired (I hope) bugs introduces
33 * by above two patches.
34 * Andrea Arcangeli : If possible we block in connect(2)
35 * if the max backlog of the listen socket
36 * is been reached. This won't break
37 * old apps and it will avoid huge amount
38 * of socks hashed (this for unix_gc()
39 * performances reasons).
40 * Security fix that limits the max
41 * number of socks to 2*max_files and
42 * the number of skb queueable in the
43 * dgram receiver.
44 * Artur Skawina : Hash function optimizations
45 * Alexey Kuznetsov : Full scale SMP. Lot of bugs are introduced 8)
46 * Malcolm Beattie : Set peercred for socketpair
47 * Michal Ostrowski : Module initialization cleanup.
48 * Arnaldo C. Melo : Remove MOD_{INC,DEC}_USE_COUNT,
49 * the core infrastructure is doing that
50 * for all net proto families now (2.5.69+)
51 *
52 *
53 * Known differences from reference BSD that was tested:
54 *
55 * [TO FIX]
56 * ECONNREFUSED is not returned from one end of a connected() socket to the
57 * other the moment one end closes.
58 * fstat() doesn't return st_dev=0, and give the blksize as high water mark
59 * and a fake inode identifier (nor the BSD first socket fstat twice bug).
60 * [NOT TO FIX]
61 * accept() returns a path name even if the connecting socket has closed
62 * in the meantime (BSD loses the path and gives up).
63 * accept() returns 0 length path for an unbound connector. BSD returns 16
64 * and a null first byte in the path (but not for gethost/peername - BSD bug ??)
65 * socketpair(...SOCK_RAW..) doesn't panic the kernel.
66 * BSD af_unix apparently has connect forgetting to block properly.
67 * (need to check this with the POSIX spec in detail)
68 *
69 * Differences from 2.0.0-11-... (ANK)
70 * Bug fixes and improvements.
71 * - client shutdown killed server socket.
72 * - removed all useless cli/sti pairs.
73 *
74 * Semantic changes/extensions.
75 * - generic control message passing.
76 * - SCM_CREDENTIALS control message.
77 * - "Abstract" (not FS based) socket bindings.
78 * Abstract names are sequences of bytes (not zero terminated)
79 * started by 0, so that this name space does not intersect
80 * with BSD names.
81 */
82
83#include <linux/module.h>
1da177e4 84#include <linux/kernel.h>
1da177e4
LT
85#include <linux/signal.h>
86#include <linux/sched.h>
87#include <linux/errno.h>
88#include <linux/string.h>
89#include <linux/stat.h>
90#include <linux/dcache.h>
91#include <linux/namei.h>
92#include <linux/socket.h>
93#include <linux/un.h>
94#include <linux/fcntl.h>
95#include <linux/termios.h>
96#include <linux/sockios.h>
97#include <linux/net.h>
98#include <linux/in.h>
99#include <linux/fs.h>
100#include <linux/slab.h>
101#include <asm/uaccess.h>
102#include <linux/skbuff.h>
103#include <linux/netdevice.h>
457c4cbc 104#include <net/net_namespace.h>
1da177e4 105#include <net/sock.h>
c752f073 106#include <net/tcp_states.h>
1da177e4
LT
107#include <net/af_unix.h>
108#include <linux/proc_fs.h>
109#include <linux/seq_file.h>
110#include <net/scm.h>
111#include <linux/init.h>
112#include <linux/poll.h>
1da177e4
LT
113#include <linux/rtnetlink.h>
114#include <linux/mount.h>
115#include <net/checksum.h>
116#include <linux/security.h>
6fa3eb70
S
117#include <linux/freezer.h>
118
119
120#include <linux/uio.h>
121#include <linux/blkdev.h>
122#include <linux/compat.h>
123#include <linux/rtc.h>
124#include <asm/kmap_types.h>
125#include <linux/device.h>
126
1da177e4 127
7123aaa3 128struct hlist_head unix_socket_table[2 * UNIX_HASH_SIZE];
fa7ff56f
PE
129EXPORT_SYMBOL_GPL(unix_socket_table);
130DEFINE_SPINLOCK(unix_table_lock);
131EXPORT_SYMBOL_GPL(unix_table_lock);
518de9b3 132static atomic_long_t unix_nr_socks;
1da177e4 133
1da177e4 134
7123aaa3
ED
135static struct hlist_head *unix_sockets_unbound(void *addr)
136{
137 unsigned long hash = (unsigned long)addr;
138
139 hash ^= hash >> 16;
140 hash ^= hash >> 8;
141 hash %= UNIX_HASH_SIZE;
142 return &unix_socket_table[UNIX_HASH_SIZE + hash];
143}
144
145#define UNIX_ABSTRACT(sk) (unix_sk(sk)->addr->hash < UNIX_HASH_SIZE)
1da177e4 146
6fa3eb70
S
147
148//for aee interface start
149#define __UNIX_SOCKET_OUTPUT_BUF_SIZE__ 3500
150static struct proc_dir_entry *gunix_socket_track_aee_entry = NULL;
151#define UNIX_SOCK_TRACK_AEE_PROCNAME "driver/usktrk_aee"
152#define UNIX_SOCK_TRACK_PROC_AEE_SIZE 3072
153
154static volatile unsigned int unix_sock_track_stop_flag = 0;
155#define unix_peer(sk) (unix_sk(sk)->peer)
156
157
877ce7c1 158#ifdef CONFIG_SECURITY_NETWORK
dc49c1f9 159static void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb)
877ce7c1 160{
dc49c1f9 161 memcpy(UNIXSID(skb), &scm->secid, sizeof(u32));
877ce7c1
CZ
162}
163
164static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb)
165{
dc49c1f9 166 scm->secid = *UNIXSID(skb);
877ce7c1
CZ
167}
168#else
dc49c1f9 169static inline void unix_get_secdata(struct scm_cookie *scm, struct sk_buff *skb)
877ce7c1
CZ
170{ }
171
172static inline void unix_set_secdata(struct scm_cookie *scm, struct sk_buff *skb)
173{ }
174#endif /* CONFIG_SECURITY_NETWORK */
175
1da177e4
LT
176/*
177 * SMP locking strategy:
fbe9cc4a 178 * hash table is protected with spinlock unix_table_lock
663717f6 179 * each socket state is protected by separate spin lock.
1da177e4
LT
180 */
181
95c96174 182static inline unsigned int unix_hash_fold(__wsum n)
1da177e4 183{
83bb80f4 184 unsigned int hash = (__force unsigned int)csum_fold(n);
95c96174 185
1da177e4
LT
186 hash ^= hash>>8;
187 return hash&(UNIX_HASH_SIZE-1);
188}
189
6fa3eb70 190
1da177e4
LT
191
192static inline int unix_our_peer(struct sock *sk, struct sock *osk)
193{
194 return unix_peer(osk) == sk;
195}
196
197static inline int unix_may_send(struct sock *sk, struct sock *osk)
198{
6eba6a37 199 return unix_peer(osk) == NULL || unix_our_peer(sk, osk);
1da177e4
LT
200}
201
3c73419c
RW
202static inline int unix_recvq_full(struct sock const *sk)
203{
204 return skb_queue_len(&sk->sk_receive_queue) > sk->sk_max_ack_backlog;
205}
206
fa7ff56f 207struct sock *unix_peer_get(struct sock *s)
1da177e4
LT
208{
209 struct sock *peer;
210
1c92b4e5 211 unix_state_lock(s);
1da177e4
LT
212 peer = unix_peer(s);
213 if (peer)
214 sock_hold(peer);
1c92b4e5 215 unix_state_unlock(s);
1da177e4
LT
216 return peer;
217}
fa7ff56f 218EXPORT_SYMBOL_GPL(unix_peer_get);
1da177e4
LT
219
220static inline void unix_release_addr(struct unix_address *addr)
221{
222 if (atomic_dec_and_test(&addr->refcnt))
223 kfree(addr);
224}
225
226/*
227 * Check unix socket name:
228 * - should be not zero length.
229 * - if started by not zero, should be NULL terminated (FS object)
230 * - if started by zero, it is abstract name.
231 */
ac7bfa62 232
95c96174 233static int unix_mkname(struct sockaddr_un *sunaddr, int len, unsigned int *hashp)
1da177e4
LT
234{
235 if (len <= sizeof(short) || len > sizeof(*sunaddr))
236 return -EINVAL;
237 if (!sunaddr || sunaddr->sun_family != AF_UNIX)
238 return -EINVAL;
239 if (sunaddr->sun_path[0]) {
240 /*
241 * This may look like an off by one error but it is a bit more
242 * subtle. 108 is the longest valid AF_UNIX path for a binding.
25985edc 243 * sun_path[108] doesn't as such exist. However in kernel space
1da177e4
LT
244 * we are guaranteed that it is a valid memory location in our
245 * kernel address buffer.
246 */
e27dfcea 247 ((char *)sunaddr)[len] = 0;
1da177e4
LT
248 len = strlen(sunaddr->sun_path)+1+sizeof(short);
249 return len;
250 }
251
07f0757a 252 *hashp = unix_hash_fold(csum_partial(sunaddr, len, 0));
1da177e4
LT
253 return len;
254}
255
256static void __unix_remove_socket(struct sock *sk)
257{
258 sk_del_node_init(sk);
259}
260
261static void __unix_insert_socket(struct hlist_head *list, struct sock *sk)
262{
547b792c 263 WARN_ON(!sk_unhashed(sk));
1da177e4
LT
264 sk_add_node(sk, list);
265}
266
267static inline void unix_remove_socket(struct sock *sk)
268{
fbe9cc4a 269 spin_lock(&unix_table_lock);
1da177e4 270 __unix_remove_socket(sk);
fbe9cc4a 271 spin_unlock(&unix_table_lock);
1da177e4
LT
272}
273
274static inline void unix_insert_socket(struct hlist_head *list, struct sock *sk)
275{
fbe9cc4a 276 spin_lock(&unix_table_lock);
1da177e4 277 __unix_insert_socket(list, sk);
fbe9cc4a 278 spin_unlock(&unix_table_lock);
1da177e4
LT
279}
280
097e66c5
DL
281static struct sock *__unix_find_socket_byname(struct net *net,
282 struct sockaddr_un *sunname,
95c96174 283 int len, int type, unsigned int hash)
1da177e4
LT
284{
285 struct sock *s;
1da177e4 286
b67bfe0d 287 sk_for_each(s, &unix_socket_table[hash ^ type]) {
1da177e4
LT
288 struct unix_sock *u = unix_sk(s);
289
878628fb 290 if (!net_eq(sock_net(s), net))
097e66c5
DL
291 continue;
292
1da177e4
LT
293 if (u->addr->len == len &&
294 !memcmp(u->addr->name, sunname, len))
295 goto found;
296 }
297 s = NULL;
298found:
299 return s;
300}
301
097e66c5
DL
302static inline struct sock *unix_find_socket_byname(struct net *net,
303 struct sockaddr_un *sunname,
1da177e4 304 int len, int type,
95c96174 305 unsigned int hash)
1da177e4
LT
306{
307 struct sock *s;
308
fbe9cc4a 309 spin_lock(&unix_table_lock);
097e66c5 310 s = __unix_find_socket_byname(net, sunname, len, type, hash);
1da177e4
LT
311 if (s)
312 sock_hold(s);
fbe9cc4a 313 spin_unlock(&unix_table_lock);
1da177e4
LT
314 return s;
315}
316
6616f788 317static struct sock *unix_find_socket_byinode(struct inode *i)
1da177e4
LT
318{
319 struct sock *s;
1da177e4 320
fbe9cc4a 321 spin_lock(&unix_table_lock);
b67bfe0d 322 sk_for_each(s,
1da177e4 323 &unix_socket_table[i->i_ino & (UNIX_HASH_SIZE - 1)]) {
40ffe67d 324 struct dentry *dentry = unix_sk(s)->path.dentry;
1da177e4 325
6eba6a37 326 if (dentry && dentry->d_inode == i) {
1da177e4
LT
327 sock_hold(s);
328 goto found;
329 }
330 }
331 s = NULL;
332found:
fbe9cc4a 333 spin_unlock(&unix_table_lock);
1da177e4
LT
334 return s;
335}
336
4b9e9796
S
337/* Support code for asymmetrically connected dgram sockets
338 *
339 * If a datagram socket is connected to a socket not itself connected
340 * to the first socket (eg, /dev/log), clients may only enqueue more
341 * messages if the present receive queue of the server socket is not
342 * "too large". This means there's a second writeability condition
343 * poll and sendmsg need to test. The dgram recv code will do a wake
344 * up on the peer_wait wait queue of a socket upon reception of a
345 * datagram which needs to be propagated to sleeping would-be writers
346 * since these might not have sent anything so far. This can't be
347 * accomplished via poll_wait because the lifetime of the server
348 * socket might be less than that of its clients if these break their
349 * association with it or if the server socket is closed while clients
350 * are still connected to it and there's no way to inform "a polling
351 * implementation" that it should let go of a certain wait queue
352 *
353 * In order to propagate a wake up, a wait_queue_t of the client
354 * socket is enqueued on the peer_wait queue of the server socket
355 * whose wake function does a wake_up on the ordinary client socket
356 * wait queue. This connection is established whenever a write (or
357 * poll for write) hit the flow control condition and broken when the
358 * association to the server socket is dissolved or after a wake up
359 * was relayed.
360 */
361
362static int unix_dgram_peer_wake_relay(wait_queue_t *q, unsigned mode, int flags,
363 void *key)
364{
365 struct unix_sock *u;
366 wait_queue_head_t *u_sleep;
367
368 u = container_of(q, struct unix_sock, peer_wake);
369
370 __remove_wait_queue(&unix_sk(u->peer_wake.private)->peer_wait,
371 q);
372 u->peer_wake.private = NULL;
373
374 /* relaying can only happen while the wq still exists */
375 u_sleep = sk_sleep(&u->sk);
376 if (u_sleep)
377 wake_up_interruptible_poll(u_sleep, key);
378
379 return 0;
380}
381
382static int unix_dgram_peer_wake_connect(struct sock *sk, struct sock *other)
383{
384 struct unix_sock *u, *u_other;
385 int rc;
386
387 u = unix_sk(sk);
388 u_other = unix_sk(other);
389 rc = 0;
390 spin_lock(&u_other->peer_wait.lock);
391
392 if (!u->peer_wake.private) {
393 u->peer_wake.private = other;
394 __add_wait_queue(&u_other->peer_wait, &u->peer_wake);
395
396 rc = 1;
397 }
398
399 spin_unlock(&u_other->peer_wait.lock);
400 return rc;
401}
402
403static void unix_dgram_peer_wake_disconnect(struct sock *sk,
404 struct sock *other)
405{
406 struct unix_sock *u, *u_other;
407
408 u = unix_sk(sk);
409 u_other = unix_sk(other);
410 spin_lock(&u_other->peer_wait.lock);
411
412 if (u->peer_wake.private == other) {
413 __remove_wait_queue(&u_other->peer_wait, &u->peer_wake);
414 u->peer_wake.private = NULL;
415 }
416
417 spin_unlock(&u_other->peer_wait.lock);
418}
419
420static void unix_dgram_peer_wake_disconnect_wakeup(struct sock *sk,
421 struct sock *other)
422{
423 unix_dgram_peer_wake_disconnect(sk, other);
424 wake_up_interruptible_poll(sk_sleep(sk),
425 POLLOUT |
426 POLLWRNORM |
427 POLLWRBAND);
428}
429
430/* preconditions:
431 * - unix_peer(sk) == other
432 * - association is stable
433 */
434static int unix_dgram_peer_wake_me(struct sock *sk, struct sock *other)
435{
436 int connected;
437
438 connected = unix_dgram_peer_wake_connect(sk, other);
439
440 if (unix_recvq_full(other))
441 return 1;
442
443 if (connected)
444 unix_dgram_peer_wake_disconnect(sk, other);
445
446 return 0;
447}
448
1da177e4
LT
449static inline int unix_writable(struct sock *sk)
450{
451 return (atomic_read(&sk->sk_wmem_alloc) << 2) <= sk->sk_sndbuf;
452}
453
454static void unix_write_space(struct sock *sk)
455{
43815482
ED
456 struct socket_wq *wq;
457
458 rcu_read_lock();
1da177e4 459 if (unix_writable(sk)) {
43815482
ED
460 wq = rcu_dereference(sk->sk_wq);
461 if (wq_has_sleeper(wq))
67426b75
ED
462 wake_up_interruptible_sync_poll(&wq->wait,
463 POLLOUT | POLLWRNORM | POLLWRBAND);
8d8ad9d7 464 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4 465 }
43815482 466 rcu_read_unlock();
1da177e4
LT
467}
468
469/* When dgram socket disconnects (or changes its peer), we clear its receive
470 * queue of packets arrived from previous peer. First, it allows to do
471 * flow control based only on wmem_alloc; second, sk connected to peer
472 * may receive messages only from that peer. */
473static void unix_dgram_disconnected(struct sock *sk, struct sock *other)
474{
b03efcfb 475 if (!skb_queue_empty(&sk->sk_receive_queue)) {
1da177e4
LT
476 skb_queue_purge(&sk->sk_receive_queue);
477 wake_up_interruptible_all(&unix_sk(sk)->peer_wait);
478
479 /* If one link of bidirectional dgram pipe is disconnected,
480 * we signal error. Messages are lost. Do not make this,
481 * when peer was not connected to us.
482 */
483 if (!sock_flag(other, SOCK_DEAD) && unix_peer(other) == sk) {
484 other->sk_err = ECONNRESET;
485 other->sk_error_report(other);
486 }
487 }
488}
489
490static void unix_sock_destructor(struct sock *sk)
491{
492 struct unix_sock *u = unix_sk(sk);
493
494 skb_queue_purge(&sk->sk_receive_queue);
495
547b792c
IJ
496 WARN_ON(atomic_read(&sk->sk_wmem_alloc));
497 WARN_ON(!sk_unhashed(sk));
498 WARN_ON(sk->sk_socket);
1da177e4 499 if (!sock_flag(sk, SOCK_DEAD)) {
6fa3eb70
S
500 #ifdef CONFIG_MTK_NET_LOGGING
501 printk(KERN_INFO "[mtk_net][unix]Attempt to release alive unix socket: %p\n", sk);
502 #endif
1da177e4
LT
503 return;
504 }
505
506 if (u->addr)
507 unix_release_addr(u->addr);
508
518de9b3 509 atomic_long_dec(&unix_nr_socks);
6f756a8c 510 local_bh_disable();
a8076d8d 511 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
6f756a8c 512 local_bh_enable();
6fa3eb70
S
513 #ifdef UNIX_REFCNT_DEBUG
514 printk(KERN_DEBUG "[mtk_net][unix]UNIX %p is destroyed, %ld are still alive.\n", sk,
518de9b3 515 atomic_long_read(&unix_nr_socks));
6fa3eb70 516 #endif
1da177e4
LT
517}
518
ded34e0f 519static void unix_release_sock(struct sock *sk, int embrion)
1da177e4
LT
520{
521 struct unix_sock *u = unix_sk(sk);
40ffe67d 522 struct path path;
1da177e4
LT
523 struct sock *skpair;
524 struct sk_buff *skb;
525 int state;
526
527 unix_remove_socket(sk);
528
529 /* Clear state */
1c92b4e5 530 unix_state_lock(sk);
1da177e4
LT
531 sock_orphan(sk);
532 sk->sk_shutdown = SHUTDOWN_MASK;
40ffe67d
AV
533 path = u->path;
534 u->path.dentry = NULL;
535 u->path.mnt = NULL;
1da177e4
LT
536 state = sk->sk_state;
537 sk->sk_state = TCP_CLOSE;
1c92b4e5 538 unix_state_unlock(sk);
1da177e4
LT
539
540 wake_up_interruptible_all(&u->peer_wait);
541
e27dfcea 542 skpair = unix_peer(sk);
1da177e4 543
e27dfcea 544 if (skpair != NULL) {
1da177e4 545 if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) {
1c92b4e5 546 unix_state_lock(skpair);
1da177e4
LT
547 /* No more writes */
548 skpair->sk_shutdown = SHUTDOWN_MASK;
549 if (!skb_queue_empty(&sk->sk_receive_queue) || embrion)
550 skpair->sk_err = ECONNRESET;
1c92b4e5 551 unix_state_unlock(skpair);
1da177e4 552 skpair->sk_state_change(skpair);
8d8ad9d7 553 sk_wake_async(skpair, SOCK_WAKE_WAITD, POLL_HUP);
1da177e4 554 }
4b9e9796
S
555
556 unix_dgram_peer_wake_disconnect(sk, skpair);
1da177e4
LT
557 sock_put(skpair); /* It may now die */
558 unix_peer(sk) = NULL;
559 }
560
561 /* Try to flush out this socket. Throw out buffers at least */
562
563 while ((skb = skb_dequeue(&sk->sk_receive_queue)) != NULL) {
e27dfcea 564 if (state == TCP_LISTEN)
1da177e4
LT
565 unix_release_sock(skb->sk, 1);
566 /* passed fds are erased in the kfree_skb hook */
567 kfree_skb(skb);
568 }
569
40ffe67d
AV
570 if (path.dentry)
571 path_put(&path);
1da177e4
LT
572
573 sock_put(sk);
574
575 /* ---- Socket is dead now and most probably destroyed ---- */
576
577 /*
e04dae84 578 * Fixme: BSD difference: In BSD all sockets connected to us get
1da177e4
LT
579 * ECONNRESET and we die on the spot. In Linux we behave
580 * like files and pipes do and wait for the last
581 * dereference.
582 *
583 * Can't we simply set sock->err?
584 *
585 * What the above comment does talk about? --ANK(980817)
586 */
587
9305cfa4 588 if (unix_tot_inflight)
ac7bfa62 589 unix_gc(); /* Garbage collect fds */
1da177e4
LT
590}
591
109f6e39
EB
592static void init_peercred(struct sock *sk)
593{
594 put_pid(sk->sk_peer_pid);
595 if (sk->sk_peer_cred)
596 put_cred(sk->sk_peer_cred);
597 sk->sk_peer_pid = get_pid(task_tgid(current));
598 sk->sk_peer_cred = get_current_cred();
599}
600
601static void copy_peercred(struct sock *sk, struct sock *peersk)
602{
603 put_pid(sk->sk_peer_pid);
604 if (sk->sk_peer_cred)
605 put_cred(sk->sk_peer_cred);
606 sk->sk_peer_pid = get_pid(peersk->sk_peer_pid);
607 sk->sk_peer_cred = get_cred(peersk->sk_peer_cred);
608}
609
1da177e4
LT
610static int unix_listen(struct socket *sock, int backlog)
611{
612 int err;
613 struct sock *sk = sock->sk;
614 struct unix_sock *u = unix_sk(sk);
109f6e39 615 struct pid *old_pid = NULL;
1da177e4
LT
616
617 err = -EOPNOTSUPP;
6eba6a37
ED
618 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
619 goto out; /* Only stream/seqpacket sockets accept */
1da177e4
LT
620 err = -EINVAL;
621 if (!u->addr)
6eba6a37 622 goto out; /* No listens on an unbound socket */
1c92b4e5 623 unix_state_lock(sk);
1da177e4
LT
624 if (sk->sk_state != TCP_CLOSE && sk->sk_state != TCP_LISTEN)
625 goto out_unlock;
626 if (backlog > sk->sk_max_ack_backlog)
627 wake_up_interruptible_all(&u->peer_wait);
628 sk->sk_max_ack_backlog = backlog;
629 sk->sk_state = TCP_LISTEN;
630 /* set credentials so connect can copy them */
109f6e39 631 init_peercred(sk);
1da177e4
LT
632 err = 0;
633
634out_unlock:
1c92b4e5 635 unix_state_unlock(sk);
109f6e39 636 put_pid(old_pid);
1da177e4 637out:
6fa3eb70 638
1da177e4
LT
639 return err;
640}
641
642static int unix_release(struct socket *);
643static int unix_bind(struct socket *, struct sockaddr *, int);
644static int unix_stream_connect(struct socket *, struct sockaddr *,
645 int addr_len, int flags);
646static int unix_socketpair(struct socket *, struct socket *);
647static int unix_accept(struct socket *, struct socket *, int);
648static int unix_getname(struct socket *, struct sockaddr *, int *, int);
649static unsigned int unix_poll(struct file *, struct socket *, poll_table *);
ec0d215f
RW
650static unsigned int unix_dgram_poll(struct file *, struct socket *,
651 poll_table *);
1da177e4
LT
652static int unix_ioctl(struct socket *, unsigned int, unsigned long);
653static int unix_shutdown(struct socket *, int);
654static int unix_stream_sendmsg(struct kiocb *, struct socket *,
655 struct msghdr *, size_t);
656static int unix_stream_recvmsg(struct kiocb *, struct socket *,
657 struct msghdr *, size_t, int);
658static int unix_dgram_sendmsg(struct kiocb *, struct socket *,
659 struct msghdr *, size_t);
660static int unix_dgram_recvmsg(struct kiocb *, struct socket *,
661 struct msghdr *, size_t, int);
662static int unix_dgram_connect(struct socket *, struct sockaddr *,
663 int, int);
664static int unix_seqpacket_sendmsg(struct kiocb *, struct socket *,
665 struct msghdr *, size_t);
a05d2ad1
EB
666static int unix_seqpacket_recvmsg(struct kiocb *, struct socket *,
667 struct msghdr *, size_t, int);
1da177e4 668
d90d9ff6 669static int unix_set_peek_off(struct sock *sk, int val)
f55bb7f9
PE
670{
671 struct unix_sock *u = unix_sk(sk);
672
d90d9ff6
SL
673 if (mutex_lock_interruptible(&u->readlock))
674 return -EINTR;
675
f55bb7f9
PE
676 sk->sk_peek_off = val;
677 mutex_unlock(&u->readlock);
d90d9ff6
SL
678
679 return 0;
f55bb7f9
PE
680}
681
682
90ddc4f0 683static const struct proto_ops unix_stream_ops = {
1da177e4
LT
684 .family = PF_UNIX,
685 .owner = THIS_MODULE,
686 .release = unix_release,
687 .bind = unix_bind,
688 .connect = unix_stream_connect,
689 .socketpair = unix_socketpair,
690 .accept = unix_accept,
691 .getname = unix_getname,
692 .poll = unix_poll,
693 .ioctl = unix_ioctl,
694 .listen = unix_listen,
695 .shutdown = unix_shutdown,
696 .setsockopt = sock_no_setsockopt,
697 .getsockopt = sock_no_getsockopt,
698 .sendmsg = unix_stream_sendmsg,
699 .recvmsg = unix_stream_recvmsg,
700 .mmap = sock_no_mmap,
701 .sendpage = sock_no_sendpage,
fc0d7536 702 .set_peek_off = unix_set_peek_off,
1da177e4
LT
703};
704
90ddc4f0 705static const struct proto_ops unix_dgram_ops = {
1da177e4
LT
706 .family = PF_UNIX,
707 .owner = THIS_MODULE,
708 .release = unix_release,
709 .bind = unix_bind,
710 .connect = unix_dgram_connect,
711 .socketpair = unix_socketpair,
712 .accept = sock_no_accept,
713 .getname = unix_getname,
ec0d215f 714 .poll = unix_dgram_poll,
1da177e4
LT
715 .ioctl = unix_ioctl,
716 .listen = sock_no_listen,
717 .shutdown = unix_shutdown,
718 .setsockopt = sock_no_setsockopt,
719 .getsockopt = sock_no_getsockopt,
720 .sendmsg = unix_dgram_sendmsg,
721 .recvmsg = unix_dgram_recvmsg,
722 .mmap = sock_no_mmap,
723 .sendpage = sock_no_sendpage,
f55bb7f9 724 .set_peek_off = unix_set_peek_off,
1da177e4
LT
725};
726
90ddc4f0 727static const struct proto_ops unix_seqpacket_ops = {
1da177e4
LT
728 .family = PF_UNIX,
729 .owner = THIS_MODULE,
730 .release = unix_release,
731 .bind = unix_bind,
732 .connect = unix_stream_connect,
733 .socketpair = unix_socketpair,
734 .accept = unix_accept,
735 .getname = unix_getname,
ec0d215f 736 .poll = unix_dgram_poll,
1da177e4
LT
737 .ioctl = unix_ioctl,
738 .listen = unix_listen,
739 .shutdown = unix_shutdown,
740 .setsockopt = sock_no_setsockopt,
741 .getsockopt = sock_no_getsockopt,
742 .sendmsg = unix_seqpacket_sendmsg,
a05d2ad1 743 .recvmsg = unix_seqpacket_recvmsg,
1da177e4
LT
744 .mmap = sock_no_mmap,
745 .sendpage = sock_no_sendpage,
f55bb7f9 746 .set_peek_off = unix_set_peek_off,
1da177e4
LT
747};
748
749static struct proto unix_proto = {
248969ae
ED
750 .name = "UNIX",
751 .owner = THIS_MODULE,
248969ae 752 .obj_size = sizeof(struct unix_sock),
1da177e4
LT
753};
754
a09785a2
IM
755/*
756 * AF_UNIX sockets do not interact with hardware, hence they
757 * dont trigger interrupts - so it's safe for them to have
758 * bh-unsafe locking for their sk_receive_queue.lock. Split off
759 * this special lock-class by reinitializing the spinlock key:
760 */
761static struct lock_class_key af_unix_sk_receive_queue_lock_key;
762
6eba6a37 763static struct sock *unix_create1(struct net *net, struct socket *sock)
1da177e4
LT
764{
765 struct sock *sk = NULL;
766 struct unix_sock *u;
767
518de9b3
ED
768 atomic_long_inc(&unix_nr_socks);
769 if (atomic_long_read(&unix_nr_socks) > 2 * get_max_files())
1da177e4
LT
770 goto out;
771
6257ff21 772 sk = sk_alloc(net, PF_UNIX, GFP_KERNEL, &unix_proto);
1da177e4
LT
773 if (!sk)
774 goto out;
775
6eba6a37 776 sock_init_data(sock, sk);
a09785a2
IM
777 lockdep_set_class(&sk->sk_receive_queue.lock,
778 &af_unix_sk_receive_queue_lock_key);
1da177e4
LT
779
780 sk->sk_write_space = unix_write_space;
a0a53c8b 781 sk->sk_max_ack_backlog = net->unx.sysctl_max_dgram_qlen;
1da177e4
LT
782 sk->sk_destruct = unix_sock_destructor;
783 u = unix_sk(sk);
40ffe67d
AV
784 u->path.dentry = NULL;
785 u->path.mnt = NULL;
fd19f329 786 spin_lock_init(&u->lock);
516e0cc5 787 atomic_long_set(&u->inflight, 0);
1fd05ba5 788 INIT_LIST_HEAD(&u->link);
57b47a53 789 mutex_init(&u->readlock); /* single task reading lock */
1da177e4 790 init_waitqueue_head(&u->peer_wait);
4b9e9796 791 init_waitqueue_func_entry(&u->peer_wake, unix_dgram_peer_wake_relay);
7123aaa3 792 unix_insert_socket(unix_sockets_unbound(sk), sk);
1da177e4 793out:
284b327b 794 if (sk == NULL)
518de9b3 795 atomic_long_dec(&unix_nr_socks);
920de804
ED
796 else {
797 local_bh_disable();
a8076d8d 798 sock_prot_inuse_add(sock_net(sk), sk->sk_prot, 1);
920de804
ED
799 local_bh_enable();
800 }
1da177e4
LT
801 return sk;
802}
803
3f378b68
EP
804static int unix_create(struct net *net, struct socket *sock, int protocol,
805 int kern)
1da177e4
LT
806{
807 if (protocol && protocol != PF_UNIX)
808 return -EPROTONOSUPPORT;
809
810 sock->state = SS_UNCONNECTED;
811
812 switch (sock->type) {
813 case SOCK_STREAM:
814 sock->ops = &unix_stream_ops;
815 break;
816 /*
817 * Believe it or not BSD has AF_UNIX, SOCK_RAW though
818 * nothing uses it.
819 */
820 case SOCK_RAW:
e27dfcea 821 sock->type = SOCK_DGRAM;
1da177e4
LT
822 case SOCK_DGRAM:
823 sock->ops = &unix_dgram_ops;
824 break;
825 case SOCK_SEQPACKET:
826 sock->ops = &unix_seqpacket_ops;
827 break;
828 default:
829 return -ESOCKTNOSUPPORT;
830 }
831
1b8d7ae4 832 return unix_create1(net, sock) ? 0 : -ENOMEM;
1da177e4
LT
833}
834
835static int unix_release(struct socket *sock)
836{
837 struct sock *sk = sock->sk;
838
839 if (!sk)
840 return 0;
841
ded34e0f 842 unix_release_sock(sk, 0);
1da177e4
LT
843 sock->sk = NULL;
844
ded34e0f 845 return 0;
1da177e4
LT
846}
847
848static int unix_autobind(struct socket *sock)
849{
850 struct sock *sk = sock->sk;
3b1e0a65 851 struct net *net = sock_net(sk);
1da177e4
LT
852 struct unix_sock *u = unix_sk(sk);
853 static u32 ordernum = 1;
6eba6a37 854 struct unix_address *addr;
1da177e4 855 int err;
8df73ff9 856 unsigned int retries = 0;
1da177e4 857
57bc52eb
SL
858 err = mutex_lock_interruptible(&u->readlock);
859 if (err)
860 return err;
1da177e4
LT
861
862 err = 0;
863 if (u->addr)
864 goto out;
865
866 err = -ENOMEM;
0da974f4 867 addr = kzalloc(sizeof(*addr) + sizeof(short) + 16, GFP_KERNEL);
1da177e4
LT
868 if (!addr)
869 goto out;
870
1da177e4
LT
871 addr->name->sun_family = AF_UNIX;
872 atomic_set(&addr->refcnt, 1);
873
874retry:
875 addr->len = sprintf(addr->name->sun_path+1, "%05x", ordernum) + 1 + sizeof(short);
07f0757a 876 addr->hash = unix_hash_fold(csum_partial(addr->name, addr->len, 0));
1da177e4 877
fbe9cc4a 878 spin_lock(&unix_table_lock);
1da177e4
LT
879 ordernum = (ordernum+1)&0xFFFFF;
880
097e66c5 881 if (__unix_find_socket_byname(net, addr->name, addr->len, sock->type,
1da177e4 882 addr->hash)) {
fbe9cc4a 883 spin_unlock(&unix_table_lock);
8df73ff9
TH
884 /*
885 * __unix_find_socket_byname() may take long time if many names
886 * are already in use.
887 */
888 cond_resched();
889 /* Give up if all names seems to be in use. */
890 if (retries++ == 0xFFFFF) {
891 err = -ENOSPC;
892 kfree(addr);
893 goto out;
894 }
1da177e4
LT
895 goto retry;
896 }
897 addr->hash ^= sk->sk_type;
898
899 __unix_remove_socket(sk);
900 u->addr = addr;
901 __unix_insert_socket(&unix_socket_table[addr->hash], sk);
fbe9cc4a 902 spin_unlock(&unix_table_lock);
1da177e4
LT
903 err = 0;
904
57b47a53 905out: mutex_unlock(&u->readlock);
1da177e4
LT
906 return err;
907}
908
097e66c5
DL
909static struct sock *unix_find_other(struct net *net,
910 struct sockaddr_un *sunname, int len,
95c96174 911 int type, unsigned int hash, int *error)
1da177e4
LT
912{
913 struct sock *u;
421748ec 914 struct path path;
1da177e4 915 int err = 0;
ac7bfa62 916
1da177e4 917 if (sunname->sun_path[0]) {
421748ec
AV
918 struct inode *inode;
919 err = kern_path(sunname->sun_path, LOOKUP_FOLLOW, &path);
1da177e4
LT
920 if (err)
921 goto fail;
421748ec
AV
922 inode = path.dentry->d_inode;
923 err = inode_permission(inode, MAY_WRITE);
1da177e4
LT
924 if (err)
925 goto put_fail;
926
927 err = -ECONNREFUSED;
421748ec 928 if (!S_ISSOCK(inode->i_mode))
1da177e4 929 goto put_fail;
6616f788 930 u = unix_find_socket_byinode(inode);
1da177e4
LT
931 if (!u)
932 goto put_fail;
933
934 if (u->sk_type == type)
68ac1234 935 touch_atime(&path);
1da177e4 936
421748ec 937 path_put(&path);
1da177e4 938
e27dfcea 939 err = -EPROTOTYPE;
1da177e4
LT
940 if (u->sk_type != type) {
941 sock_put(u);
942 goto fail;
943 }
944 } else {
945 err = -ECONNREFUSED;
e27dfcea 946 u = unix_find_socket_byname(net, sunname, len, type, hash);
1da177e4
LT
947 if (u) {
948 struct dentry *dentry;
40ffe67d 949 dentry = unix_sk(u)->path.dentry;
1da177e4 950 if (dentry)
68ac1234 951 touch_atime(&unix_sk(u)->path);
1da177e4
LT
952 } else
953 goto fail;
954 }
955 return u;
956
957put_fail:
421748ec 958 path_put(&path);
1da177e4 959fail:
e27dfcea 960 *error = err;
1da177e4
LT
961 return NULL;
962}
963
faf02010
AV
964static int unix_mknod(const char *sun_path, umode_t mode, struct path *res)
965{
966 struct dentry *dentry;
967 struct path path;
968 int err = 0;
969 /*
970 * Get the parent directory, calculate the hash for last
971 * component.
972 */
973 dentry = kern_path_create(AT_FDCWD, sun_path, &path, 0);
974 err = PTR_ERR(dentry);
975 if (IS_ERR(dentry))
976 return err;
977
978 /*
979 * All right, let's create it.
980 */
981 err = security_path_mknod(&path, dentry, mode, 0);
982 if (!err) {
983 err = vfs_mknod(path.dentry->d_inode, dentry, mode, 0);
984 if (!err) {
985 res->mnt = mntget(path.mnt);
986 res->dentry = dget(dentry);
987 }
988 }
989 done_path_create(&path, dentry);
990 return err;
991}
1da177e4
LT
992
993static int unix_bind(struct socket *sock, struct sockaddr *uaddr, int addr_len)
994{
995 struct sock *sk = sock->sk;
3b1e0a65 996 struct net *net = sock_net(sk);
1da177e4 997 struct unix_sock *u = unix_sk(sk);
e27dfcea 998 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
dae6ad8f 999 char *sun_path = sunaddr->sun_path;
1da177e4 1000 int err;
95c96174 1001 unsigned int hash;
1da177e4
LT
1002 struct unix_address *addr;
1003 struct hlist_head *list;
1004
1005 err = -EINVAL;
1006 if (sunaddr->sun_family != AF_UNIX)
1007 goto out;
1008
e27dfcea 1009 if (addr_len == sizeof(short)) {
1da177e4
LT
1010 err = unix_autobind(sock);
1011 goto out;
1012 }
1013
1014 err = unix_mkname(sunaddr, addr_len, &hash);
1015 if (err < 0)
1016 goto out;
1017 addr_len = err;
1018
57bc52eb
SL
1019 err = mutex_lock_interruptible(&u->readlock);
1020 if (err)
1021 goto out;
1da177e4
LT
1022
1023 err = -EINVAL;
1024 if (u->addr)
1025 goto out_up;
1026
1027 err = -ENOMEM;
1028 addr = kmalloc(sizeof(*addr)+addr_len, GFP_KERNEL);
1029 if (!addr)
1030 goto out_up;
1031
1032 memcpy(addr->name, sunaddr, addr_len);
1033 addr->len = addr_len;
1034 addr->hash = hash ^ sk->sk_type;
1035 atomic_set(&addr->refcnt, 1);
1036
dae6ad8f 1037 if (sun_path[0]) {
6fa3eb70
S
1038 struct path path;
1039
faf02010 1040 umode_t mode = S_IFSOCK |
ce3b0f8d 1041 (SOCK_INODE(sock)->i_mode & ~current_umask());
faf02010
AV
1042 err = unix_mknod(sun_path, mode, &path);
1043 if (err) {
1044 if (err == -EEXIST)
1045 err = -EADDRINUSE;
1046 unix_release_addr(addr);
1047 goto out_up;
1048 }
1da177e4 1049 addr->hash = UNIX_HASH_SIZE;
faf02010
AV
1050 hash = path.dentry->d_inode->i_ino & (UNIX_HASH_SIZE-1);
1051 spin_lock(&unix_table_lock);
1052 u->path = path;
1053 list = &unix_socket_table[hash];
1054 } else {
1055 spin_lock(&unix_table_lock);
1da177e4 1056 err = -EADDRINUSE;
097e66c5 1057 if (__unix_find_socket_byname(net, sunaddr, addr_len,
1da177e4
LT
1058 sk->sk_type, hash)) {
1059 unix_release_addr(addr);
1060 goto out_unlock;
1061 }
1062
1063 list = &unix_socket_table[addr->hash];
1da177e4
LT
1064 }
1065
1066 err = 0;
1067 __unix_remove_socket(sk);
1068 u->addr = addr;
1069 __unix_insert_socket(list, sk);
1070
1071out_unlock:
fbe9cc4a 1072 spin_unlock(&unix_table_lock);
1da177e4 1073out_up:
57b47a53 1074 mutex_unlock(&u->readlock);
1da177e4 1075out:
6fa3eb70 1076
1da177e4 1077 return err;
1da177e4
LT
1078}
1079
278a3de5
DM
1080static void unix_state_double_lock(struct sock *sk1, struct sock *sk2)
1081{
1082 if (unlikely(sk1 == sk2) || !sk2) {
1083 unix_state_lock(sk1);
1084 return;
1085 }
1086 if (sk1 < sk2) {
1087 unix_state_lock(sk1);
1088 unix_state_lock_nested(sk2);
1089 } else {
1090 unix_state_lock(sk2);
1091 unix_state_lock_nested(sk1);
1092 }
1093}
1094
1095static void unix_state_double_unlock(struct sock *sk1, struct sock *sk2)
1096{
1097 if (unlikely(sk1 == sk2) || !sk2) {
1098 unix_state_unlock(sk1);
1099 return;
1100 }
1101 unix_state_unlock(sk1);
1102 unix_state_unlock(sk2);
1103}
1104
1da177e4
LT
1105static int unix_dgram_connect(struct socket *sock, struct sockaddr *addr,
1106 int alen, int flags)
1107{
1108 struct sock *sk = sock->sk;
3b1e0a65 1109 struct net *net = sock_net(sk);
e27dfcea 1110 struct sockaddr_un *sunaddr = (struct sockaddr_un *)addr;
1da177e4 1111 struct sock *other;
95c96174 1112 unsigned int hash;
1da177e4
LT
1113 int err;
1114
1115 if (addr->sa_family != AF_UNSPEC) {
6fa3eb70 1116
1da177e4
LT
1117 err = unix_mkname(sunaddr, alen, &hash);
1118 if (err < 0)
1119 goto out;
1120 alen = err;
1121
1122 if (test_bit(SOCK_PASSCRED, &sock->flags) &&
1123 !unix_sk(sk)->addr && (err = unix_autobind(sock)) != 0)
1124 goto out;
1125
278a3de5 1126restart:
e27dfcea 1127 other = unix_find_other(net, sunaddr, alen, sock->type, hash, &err);
1da177e4
LT
1128 if (!other)
1129 goto out;
1130
278a3de5
DM
1131 unix_state_double_lock(sk, other);
1132
1133 /* Apparently VFS overslept socket death. Retry. */
1134 if (sock_flag(other, SOCK_DEAD)) {
1135 unix_state_double_unlock(sk, other);
1136 sock_put(other);
1137 goto restart;
1138 }
1da177e4
LT
1139
1140 err = -EPERM;
1141 if (!unix_may_send(sk, other))
1142 goto out_unlock;
1143
1144 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1145 if (err)
1146 goto out_unlock;
1147
1148 } else {
1149 /*
1150 * 1003.1g breaking connected state with AF_UNSPEC
1151 */
1152 other = NULL;
278a3de5 1153 unix_state_double_lock(sk, other);
1da177e4
LT
1154 }
1155
1156 /*
1157 * If it was connected, reconnect.
1158 */
1159 if (unix_peer(sk)) {
1160 struct sock *old_peer = unix_peer(sk);
e27dfcea 1161 unix_peer(sk) = other;
4b9e9796
S
1162 unix_dgram_peer_wake_disconnect_wakeup(sk, old_peer);
1163
278a3de5 1164 unix_state_double_unlock(sk, other);
1da177e4
LT
1165
1166 if (other != old_peer)
1167 unix_dgram_disconnected(sk, old_peer);
1168 sock_put(old_peer);
1169 } else {
e27dfcea 1170 unix_peer(sk) = other;
278a3de5 1171 unix_state_double_unlock(sk, other);
1da177e4 1172 }
6fa3eb70
S
1173
1174#ifdef CONFIG_MTK_NET_LOGGING
1175 if((SOCK_INODE(sock)!= NULL) && (sunaddr != NULL) && (other->sk_socket != NULL) && (SOCK_INODE(other->sk_socket) != NULL))
1176 {
1177 printk(KERN_INFO "[mtk_net][socket]unix_dgram_connect[%lu]:connect [%s] other[%lu]\n",SOCK_INODE(sock)->i_ino,sunaddr->sun_path,SOCK_INODE(other->sk_socket)->i_ino);
1178 }
1179#endif
1180
ac7bfa62 1181 return 0;
1da177e4
LT
1182
1183out_unlock:
278a3de5 1184 unix_state_double_unlock(sk, other);
1da177e4
LT
1185 sock_put(other);
1186out:
6fa3eb70 1187
1da177e4
LT
1188 return err;
1189}
1190
1191static long unix_wait_for_peer(struct sock *other, long timeo)
1192{
1193 struct unix_sock *u = unix_sk(other);
1194 int sched;
1195 DEFINE_WAIT(wait);
1196
1197 prepare_to_wait_exclusive(&u->peer_wait, &wait, TASK_INTERRUPTIBLE);
1198
1199 sched = !sock_flag(other, SOCK_DEAD) &&
1200 !(other->sk_shutdown & RCV_SHUTDOWN) &&
3c73419c 1201 unix_recvq_full(other);
1da177e4 1202
1c92b4e5 1203 unix_state_unlock(other);
1da177e4
LT
1204
1205 if (sched)
1206 timeo = schedule_timeout(timeo);
1207
1208 finish_wait(&u->peer_wait, &wait);
1209 return timeo;
1210}
1211
1212static int unix_stream_connect(struct socket *sock, struct sockaddr *uaddr,
1213 int addr_len, int flags)
1214{
e27dfcea 1215 struct sockaddr_un *sunaddr = (struct sockaddr_un *)uaddr;
1da177e4 1216 struct sock *sk = sock->sk;
3b1e0a65 1217 struct net *net = sock_net(sk);
1da177e4
LT
1218 struct unix_sock *u = unix_sk(sk), *newu, *otheru;
1219 struct sock *newsk = NULL;
1220 struct sock *other = NULL;
1221 struct sk_buff *skb = NULL;
95c96174 1222 unsigned int hash;
1da177e4
LT
1223 int st;
1224 int err;
1225 long timeo;
1226
1227 err = unix_mkname(sunaddr, addr_len, &hash);
1228 if (err < 0)
1229 goto out;
1230 addr_len = err;
1231
f64f9e71
JP
1232 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr &&
1233 (err = unix_autobind(sock)) != 0)
1da177e4
LT
1234 goto out;
1235
1236 timeo = sock_sndtimeo(sk, flags & O_NONBLOCK);
1237
1238 /* First of all allocate resources.
1239 If we will make it after state is locked,
1240 we will have to recheck all again in any case.
1241 */
1242
1243 err = -ENOMEM;
1244
1245 /* create new sock for complete connection */
3b1e0a65 1246 newsk = unix_create1(sock_net(sk), NULL);
1da177e4
LT
1247 if (newsk == NULL)
1248 goto out;
1249
1250 /* Allocate skb for sending to listening sock */
1251 skb = sock_wmalloc(newsk, 1, 0, GFP_KERNEL);
1252 if (skb == NULL)
1253 goto out;
1254
1255restart:
1256 /* Find listening sock. */
097e66c5 1257 other = unix_find_other(net, sunaddr, addr_len, sk->sk_type, hash, &err);
1da177e4
LT
1258 if (!other)
1259 goto out;
1260
1261 /* Latch state of peer */
1c92b4e5 1262 unix_state_lock(other);
1da177e4
LT
1263
1264 /* Apparently VFS overslept socket death. Retry. */
1265 if (sock_flag(other, SOCK_DEAD)) {
1c92b4e5 1266 unix_state_unlock(other);
1da177e4
LT
1267 sock_put(other);
1268 goto restart;
1269 }
1270
1271 err = -ECONNREFUSED;
1272 if (other->sk_state != TCP_LISTEN)
1273 goto out_unlock;
77238f2b
TS
1274 if (other->sk_shutdown & RCV_SHUTDOWN)
1275 goto out_unlock;
1da177e4 1276
3c73419c 1277 if (unix_recvq_full(other)) {
1da177e4
LT
1278 err = -EAGAIN;
1279 if (!timeo)
1280 goto out_unlock;
1281
1282 timeo = unix_wait_for_peer(other, timeo);
1283
1284 err = sock_intr_errno(timeo);
1285 if (signal_pending(current))
1286 goto out;
1287 sock_put(other);
1288 goto restart;
ac7bfa62 1289 }
1da177e4
LT
1290
1291 /* Latch our state.
1292
e5537bfc 1293 It is tricky place. We need to grab our state lock and cannot
1da177e4
LT
1294 drop lock on peer. It is dangerous because deadlock is
1295 possible. Connect to self case and simultaneous
1296 attempt to connect are eliminated by checking socket
1297 state. other is TCP_LISTEN, if sk is TCP_LISTEN we
1298 check this before attempt to grab lock.
1299
1300 Well, and we have to recheck the state after socket locked.
1301 */
1302 st = sk->sk_state;
1303
1304 switch (st) {
1305 case TCP_CLOSE:
1306 /* This is ok... continue with connect */
1307 break;
1308 case TCP_ESTABLISHED:
1309 /* Socket is already connected */
1310 err = -EISCONN;
1311 goto out_unlock;
1312 default:
1313 err = -EINVAL;
1314 goto out_unlock;
1315 }
1316
1c92b4e5 1317 unix_state_lock_nested(sk);
1da177e4
LT
1318
1319 if (sk->sk_state != st) {
1c92b4e5
DM
1320 unix_state_unlock(sk);
1321 unix_state_unlock(other);
1da177e4
LT
1322 sock_put(other);
1323 goto restart;
1324 }
1325
3610cda5 1326 err = security_unix_stream_connect(sk, other, newsk);
1da177e4 1327 if (err) {
1c92b4e5 1328 unix_state_unlock(sk);
1da177e4
LT
1329 goto out_unlock;
1330 }
1331
1332 /* The way is open! Fastly set all the necessary fields... */
1333
1334 sock_hold(sk);
1335 unix_peer(newsk) = sk;
1336 newsk->sk_state = TCP_ESTABLISHED;
1337 newsk->sk_type = sk->sk_type;
109f6e39 1338 init_peercred(newsk);
1da177e4 1339 newu = unix_sk(newsk);
eaefd110 1340 RCU_INIT_POINTER(newsk->sk_wq, &newu->peer_wq);
1da177e4
LT
1341 otheru = unix_sk(other);
1342
1343 /* copy address information from listening to new sock*/
1344 if (otheru->addr) {
1345 atomic_inc(&otheru->addr->refcnt);
1346 newu->addr = otheru->addr;
1347 }
40ffe67d
AV
1348 if (otheru->path.dentry) {
1349 path_get(&otheru->path);
1350 newu->path = otheru->path;
1da177e4
LT
1351 }
1352
1353 /* Set credentials */
109f6e39 1354 copy_peercred(sk, other);
1da177e4 1355
1da177e4
LT
1356 sock->state = SS_CONNECTED;
1357 sk->sk_state = TCP_ESTABLISHED;
830a1e5c
BL
1358 sock_hold(newsk);
1359
1360 smp_mb__after_atomic_inc(); /* sock_hold() does an atomic_inc() */
1361 unix_peer(sk) = newsk;
1da177e4 1362
1c92b4e5 1363 unix_state_unlock(sk);
1da177e4
LT
1364
1365 /* take ten and and send info to listening sock */
1366 spin_lock(&other->sk_receive_queue.lock);
1367 __skb_queue_tail(&other->sk_receive_queue, skb);
1da177e4 1368 spin_unlock(&other->sk_receive_queue.lock);
1c92b4e5 1369 unix_state_unlock(other);
6fa3eb70
S
1370
1371 #ifdef CONFIG_MTK_NET_LOGGING
1372 if((SOCK_INODE(sock)!= NULL) && (sunaddr != NULL) && (other->sk_socket != NULL) && (SOCK_INODE(other->sk_socket) != NULL))
1373 {
1374 printk(KERN_INFO "[mtk_net][socket]unix_stream_connect[%lu ]: connect [%s] other[%lu] \n",SOCK_INODE(sock)->i_ino,sunaddr->sun_path,SOCK_INODE(other->sk_socket)->i_ino);
1375 }
1376 #endif
1377
1da177e4
LT
1378 other->sk_data_ready(other, 0);
1379 sock_put(other);
6fa3eb70 1380
1da177e4
LT
1381 return 0;
1382
1383out_unlock:
1384 if (other)
1c92b4e5 1385 unix_state_unlock(other);
1da177e4
LT
1386
1387out:
40d44446 1388 kfree_skb(skb);
1da177e4
LT
1389 if (newsk)
1390 unix_release_sock(newsk, 0);
1391 if (other)
1392 sock_put(other);
6fa3eb70 1393
1da177e4
LT
1394 return err;
1395}
1396
1397static int unix_socketpair(struct socket *socka, struct socket *sockb)
1398{
e27dfcea 1399 struct sock *ska = socka->sk, *skb = sockb->sk;
1da177e4
LT
1400
1401 /* Join our sockets back to back */
1402 sock_hold(ska);
1403 sock_hold(skb);
e27dfcea
JK
1404 unix_peer(ska) = skb;
1405 unix_peer(skb) = ska;
109f6e39
EB
1406 init_peercred(ska);
1407 init_peercred(skb);
1da177e4
LT
1408
1409 if (ska->sk_type != SOCK_DGRAM) {
1410 ska->sk_state = TCP_ESTABLISHED;
1411 skb->sk_state = TCP_ESTABLISHED;
1412 socka->state = SS_CONNECTED;
1413 sockb->state = SS_CONNECTED;
1414 }
1415 return 0;
1416}
1417
a769ad65
DB
1418static void unix_sock_inherit_flags(const struct socket *old,
1419 struct socket *new)
1420{
1421 if (test_bit(SOCK_PASSCRED, &old->flags))
1422 set_bit(SOCK_PASSCRED, &new->flags);
1423 if (test_bit(SOCK_PASSSEC, &old->flags))
1424 set_bit(SOCK_PASSSEC, &new->flags);
1425}
1426
1da177e4
LT
1427static int unix_accept(struct socket *sock, struct socket *newsock, int flags)
1428{
1429 struct sock *sk = sock->sk;
1430 struct sock *tsk;
1431 struct sk_buff *skb;
1432 int err;
1433
1434 err = -EOPNOTSUPP;
6eba6a37 1435 if (sock->type != SOCK_STREAM && sock->type != SOCK_SEQPACKET)
1da177e4
LT
1436 goto out;
1437
1438 err = -EINVAL;
1439 if (sk->sk_state != TCP_LISTEN)
1440 goto out;
1441
1442 /* If socket state is TCP_LISTEN it cannot change (for now...),
1443 * so that no locks are necessary.
1444 */
6fa3eb70 1445
1da177e4
LT
1446 skb = skb_recv_datagram(sk, 0, flags&O_NONBLOCK, &err);
1447 if (!skb) {
1448 /* This means receive shutdown. */
1449 if (err == 0)
1450 err = -EINVAL;
1451 goto out;
1452 }
1453
1454 tsk = skb->sk;
1455 skb_free_datagram(sk, skb);
1456 wake_up_interruptible(&unix_sk(sk)->peer_wait);
1457
1458 /* attach accepted sock to socket */
1c92b4e5 1459 unix_state_lock(tsk);
1da177e4 1460 newsock->state = SS_CONNECTED;
a769ad65 1461 unix_sock_inherit_flags(sock, newsock);
1da177e4 1462 sock_graft(tsk, newsock);
1c92b4e5 1463 unix_state_unlock(tsk);
6fa3eb70 1464
1da177e4
LT
1465 return 0;
1466
1467out:
6fa3eb70 1468
1da177e4
LT
1469 return err;
1470}
1471
1472
1473static int unix_getname(struct socket *sock, struct sockaddr *uaddr, int *uaddr_len, int peer)
1474{
1475 struct sock *sk = sock->sk;
1476 struct unix_sock *u;
13cfa97b 1477 DECLARE_SOCKADDR(struct sockaddr_un *, sunaddr, uaddr);
1da177e4
LT
1478 int err = 0;
1479
1480 if (peer) {
1481 sk = unix_peer_get(sk);
1482
1483 err = -ENOTCONN;
1484 if (!sk)
1485 goto out;
1486 err = 0;
1487 } else {
1488 sock_hold(sk);
1489 }
1490
1491 u = unix_sk(sk);
1c92b4e5 1492 unix_state_lock(sk);
1da177e4
LT
1493 if (!u->addr) {
1494 sunaddr->sun_family = AF_UNIX;
1495 sunaddr->sun_path[0] = 0;
1496 *uaddr_len = sizeof(short);
1497 } else {
1498 struct unix_address *addr = u->addr;
1499
1500 *uaddr_len = addr->len;
1501 memcpy(sunaddr, addr->name, *uaddr_len);
1502 }
1c92b4e5 1503 unix_state_unlock(sk);
1da177e4
LT
1504 sock_put(sk);
1505out:
1506 return err;
1507}
1508
1509static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1510{
1511 int i;
1512
1513 scm->fp = UNIXCB(skb).fp;
1da177e4
LT
1514 UNIXCB(skb).fp = NULL;
1515
6eba6a37 1516 for (i = scm->fp->count-1; i >= 0; i--)
1da177e4
LT
1517 unix_notinflight(scm->fp->fp[i]);
1518}
1519
7361c36c 1520static void unix_destruct_scm(struct sk_buff *skb)
1da177e4
LT
1521{
1522 struct scm_cookie scm;
1523 memset(&scm, 0, sizeof(scm));
7361c36c 1524 scm.pid = UNIXCB(skb).pid;
7361c36c
EB
1525 if (UNIXCB(skb).fp)
1526 unix_detach_fds(&scm, skb);
1da177e4
LT
1527
1528 /* Alas, it calls VFS */
1529 /* So fscking what? fput() had been SMP-safe since the last Summer */
1530 scm_destroy(&scm);
1531 sock_wfree(skb);
1532}
1533
25888e30
ED
1534#define MAX_RECURSION_LEVEL 4
1535
6209344f 1536static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
1da177e4
LT
1537{
1538 int i;
25888e30
ED
1539 unsigned char max_level = 0;
1540 int unix_sock_count = 0;
1541
1542 for (i = scm->fp->count - 1; i >= 0; i--) {
1543 struct sock *sk = unix_get_socket(scm->fp->fp[i]);
1544
1545 if (sk) {
1546 unix_sock_count++;
1547 max_level = max(max_level,
1548 unix_sk(sk)->recursion_level);
1549 }
1550 }
1551 if (unlikely(max_level > MAX_RECURSION_LEVEL))
1552 return -ETOOMANYREFS;
6209344f
MS
1553
1554 /*
1555 * Need to duplicate file references for the sake of garbage
1556 * collection. Otherwise a socket in the fps might become a
1557 * candidate for GC while the skb is not yet queued.
1558 */
1559 UNIXCB(skb).fp = scm_fp_dup(scm->fp);
1560 if (!UNIXCB(skb).fp)
1561 return -ENOMEM;
1562
25888e30
ED
1563 if (unix_sock_count) {
1564 for (i = scm->fp->count - 1; i >= 0; i--)
1565 unix_inflight(scm->fp->fp[i]);
1566 }
1567 return max_level;
1da177e4
LT
1568}
1569
f78a5fda 1570static int unix_scm_to_skb(struct scm_cookie *scm, struct sk_buff *skb, bool send_fds)
7361c36c
EB
1571{
1572 int err = 0;
16e57262 1573
f78a5fda 1574 UNIXCB(skb).pid = get_pid(scm->pid);
6b0ee8c0
EB
1575 UNIXCB(skb).uid = scm->creds.uid;
1576 UNIXCB(skb).gid = scm->creds.gid;
7361c36c
EB
1577 UNIXCB(skb).fp = NULL;
1578 if (scm->fp && send_fds)
1579 err = unix_attach_fds(scm, skb);
1580
1581 skb->destructor = unix_destruct_scm;
1582 return err;
1583}
1584
16e57262
ED
1585/*
1586 * Some apps rely on write() giving SCM_CREDENTIALS
1587 * We include credentials if source or destination socket
1588 * asserted SOCK_PASSCRED.
1589 */
1590static void maybe_add_creds(struct sk_buff *skb, const struct socket *sock,
1591 const struct sock *other)
1592{
6b0ee8c0 1593 if (UNIXCB(skb).pid)
16e57262
ED
1594 return;
1595 if (test_bit(SOCK_PASSCRED, &sock->flags) ||
25da0e3e
EB
1596 !other->sk_socket ||
1597 test_bit(SOCK_PASSCRED, &other->sk_socket->flags)) {
16e57262 1598 UNIXCB(skb).pid = get_pid(task_tgid(current));
6e0895c2 1599 current_uid_gid(&UNIXCB(skb).uid, &UNIXCB(skb).gid);
16e57262
ED
1600 }
1601}
1602
1da177e4
LT
1603/*
1604 * Send AF_UNIX data.
1605 */
1606
1607static int unix_dgram_sendmsg(struct kiocb *kiocb, struct socket *sock,
1608 struct msghdr *msg, size_t len)
1609{
1610 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1611 struct sock *sk = sock->sk;
3b1e0a65 1612 struct net *net = sock_net(sk);
1da177e4 1613 struct unix_sock *u = unix_sk(sk);
e27dfcea 1614 struct sockaddr_un *sunaddr = msg->msg_name;
1da177e4
LT
1615 struct sock *other = NULL;
1616 int namelen = 0; /* fake GCC */
1617 int err;
95c96174 1618 unsigned int hash;
f78a5fda 1619 struct sk_buff *skb;
1da177e4
LT
1620 long timeo;
1621 struct scm_cookie tmp_scm;
25888e30 1622 int max_level;
eb6a2481 1623 int data_len = 0;
4b9e9796 1624 int sk_locked;
6fa3eb70 1625
1da177e4
LT
1626 if (NULL == siocb->scm)
1627 siocb->scm = &tmp_scm;
5f23b734 1628 wait_for_unix_gc();
e0e3cea4 1629 err = scm_send(sock, msg, siocb->scm, false);
1da177e4
LT
1630 if (err < 0)
1631 return err;
1632
1633 err = -EOPNOTSUPP;
1634 if (msg->msg_flags&MSG_OOB)
1635 goto out;
1636
1637 if (msg->msg_namelen) {
1638 err = unix_mkname(sunaddr, msg->msg_namelen, &hash);
1639 if (err < 0)
1640 goto out;
1641 namelen = err;
1642 } else {
1643 sunaddr = NULL;
1644 err = -ENOTCONN;
1645 other = unix_peer_get(sk);
1646 if (!other)
1647 goto out;
1648 }
1649
f64f9e71
JP
1650 if (test_bit(SOCK_PASSCRED, &sock->flags) && !u->addr
1651 && (err = unix_autobind(sock)) != 0)
1da177e4
LT
1652 goto out;
1653
1654 err = -EMSGSIZE;
1655 if (len > sk->sk_sndbuf - 32)
1656 goto out;
1657
eb6a2481
ED
1658 if (len > SKB_MAX_ALLOC)
1659 data_len = min_t(size_t,
1660 len - SKB_MAX_ALLOC,
1661 MAX_SKB_FRAGS * PAGE_SIZE);
1662
1663 skb = sock_alloc_send_pskb(sk, len - data_len, data_len,
1664 msg->msg_flags & MSG_DONTWAIT, &err);
e27dfcea 1665 if (skb == NULL)
1da177e4
LT
1666 goto out;
1667
f78a5fda 1668 err = unix_scm_to_skb(siocb->scm, skb, true);
25888e30 1669 if (err < 0)
7361c36c 1670 goto out_free;
25888e30 1671 max_level = err + 1;
dc49c1f9 1672 unix_get_secdata(siocb->scm, skb);
877ce7c1 1673
eb6a2481
ED
1674 skb_put(skb, len - data_len);
1675 skb->data_len = data_len;
1676 skb->len = len;
1677 err = skb_copy_datagram_from_iovec(skb, 0, msg->msg_iov, 0, len);
1da177e4
LT
1678 if (err)
1679 goto out_free;
1680
1681 timeo = sock_sndtimeo(sk, msg->msg_flags & MSG_DONTWAIT);
1682
1683restart:
1684 if (!other) {
1685 err = -ECONNRESET;
1686 if (sunaddr == NULL)
1687 goto out_free;
1688
097e66c5 1689 other = unix_find_other(net, sunaddr, namelen, sk->sk_type,
1da177e4 1690 hash, &err);
e27dfcea 1691 if (other == NULL)
1da177e4
LT
1692 goto out_free;
1693 }
1694
d6ae3bae
AC
1695 if (sk_filter(other, skb) < 0) {
1696 /* Toss the packet but do not return any error to the sender */
1697 err = len;
1698 goto out_free;
1699 }
1700
4b9e9796 1701 sk_locked = 0;
1c92b4e5 1702 unix_state_lock(other);
4b9e9796 1703restart_locked:
1da177e4
LT
1704 err = -EPERM;
1705 if (!unix_may_send(sk, other))
1706 goto out_unlock;
1707
4b9e9796 1708 if (unlikely(sock_flag(other, SOCK_DEAD))) {
1da177e4
LT
1709 /*
1710 * Check with 1003.1g - what should
1711 * datagram error
1712 */
1c92b4e5 1713 unix_state_unlock(other);
1da177e4
LT
1714 sock_put(other);
1715
4b9e9796 1716 if (!sk_locked)
1c92b4e5 1717 unix_state_lock(sk);
1da177e4 1718 if (unix_peer(sk) == other) {
e27dfcea 1719 unix_peer(sk) = NULL;
4b9e9796
S
1720 unix_dgram_peer_wake_disconnect_wakeup(sk, other);
1721
1c92b4e5 1722 unix_state_unlock(sk);
1da177e4
LT
1723
1724 unix_dgram_disconnected(sk, other);
1725 sock_put(other);
1726 err = -ECONNREFUSED;
1727 } else {
1c92b4e5 1728 unix_state_unlock(sk);
1da177e4
LT
1729 }
1730
1731 other = NULL;
1732 if (err)
1733 goto out_free;
1734 goto restart;
1735 }
1736
1737 err = -EPIPE;
1738 if (other->sk_shutdown & RCV_SHUTDOWN)
1739 goto out_unlock;
1740
1741 if (sk->sk_type != SOCK_SEQPACKET) {
1742 err = security_unix_may_send(sk->sk_socket, other->sk_socket);
1743 if (err)
1744 goto out_unlock;
1745 }
1746
4b9e9796
S
1747 /* other == sk && unix_peer(other) != sk if
1748 * - unix_peer(sk) == NULL, destination address bound to sk
1749 * - unix_peer(sk) == sk by time of get but disconnected before lock
1750 */
1751 if (other != sk &&
1752 unlikely(unix_peer(other) != sk && unix_recvq_full(other))) {
1753 if (timeo) {
1754 timeo = unix_wait_for_peer(other, timeo);
1755
1756 err = sock_intr_errno(timeo);
1757 if (signal_pending(current))
1758 goto out_free;
1759
1760 goto restart;
1da177e4
LT
1761 }
1762
4b9e9796
S
1763 if (!sk_locked) {
1764 unix_state_unlock(other);
1765 unix_state_double_lock(sk, other);
1766 }
1da177e4 1767
4b9e9796
S
1768 if (unix_peer(sk) != other ||
1769 unix_dgram_peer_wake_me(sk, other)) {
1770 err = -EAGAIN;
1771 sk_locked = 1;
1772 goto out_unlock;
1773 }
1da177e4 1774
4b9e9796
S
1775 if (!sk_locked) {
1776 sk_locked = 1;
1777 goto restart_locked;
1778 }
1da177e4
LT
1779 }
1780
4b9e9796
S
1781 if (unlikely(sk_locked))
1782 unix_state_unlock(sk);
1783
3f66116e
AC
1784 if (sock_flag(other, SOCK_RCVTSTAMP))
1785 __net_timestamp(skb);
16e57262 1786 maybe_add_creds(skb, sock, other);
1da177e4 1787 skb_queue_tail(&other->sk_receive_queue, skb);
25888e30
ED
1788 if (max_level > unix_sk(other)->recursion_level)
1789 unix_sk(other)->recursion_level = max_level;
1c92b4e5 1790 unix_state_unlock(other);
1da177e4
LT
1791 other->sk_data_ready(other, len);
1792 sock_put(other);
f78a5fda 1793 scm_destroy(siocb->scm);
6fa3eb70 1794
1da177e4
LT
1795 return len;
1796
1797out_unlock:
4b9e9796
S
1798 if (sk_locked)
1799 unix_state_unlock(sk);
1c92b4e5 1800 unix_state_unlock(other);
1da177e4
LT
1801out_free:
1802 kfree_skb(skb);
1803out:
1804 if (other)
1805 sock_put(other);
f78a5fda 1806 scm_destroy(siocb->scm);
6fa3eb70 1807
1da177e4
LT
1808 return err;
1809}
1810
ac7bfa62 1811
1da177e4
LT
1812static int unix_stream_sendmsg(struct kiocb *kiocb, struct socket *sock,
1813 struct msghdr *msg, size_t len)
1814{
1815 struct sock_iocb *siocb = kiocb_to_siocb(kiocb);
1816 struct sock *sk = sock->sk;
1817 struct sock *other = NULL;
6eba6a37 1818 int err, size;
f78a5fda 1819 struct sk_buff *skb;
e27dfcea 1820 int sent = 0;
1da177e4 1821 struct scm_cookie tmp_scm;
8ba69ba6 1822 bool fds_sent = false;
25888e30 1823 int max_level;
1da177e4
LT
1824
1825 if (NULL == siocb->scm)
1826 siocb->scm = &tmp_scm;
6fa3eb70 1827
5f23b734 1828 wait_for_unix_gc();
e0e3cea4 1829 err = scm_send(sock, msg, siocb->scm, false);
1da177e4
LT
1830 if (err < 0)
1831 return err;
1832
1833 err = -EOPNOTSUPP;
1834 if (msg->msg_flags&MSG_OOB)
1835 goto out_err;
1836
1837 if (msg->msg_namelen) {
1838 err = sk->sk_state == TCP_ESTABLISHED ? -EISCONN : -EOPNOTSUPP;
1839 goto out_err;
1840 } else {
1da177e4 1841 err = -ENOTCONN;
830a1e5c 1842 other = unix_peer(sk);
1da177e4
LT
1843 if (!other)
1844 goto out_err;
1845 }
1846
1847 if (sk->sk_shutdown & SEND_SHUTDOWN)
1848 goto pipe_err;
1849
6eba6a37 1850 while (sent < len) {
1da177e4 1851 /*
e9df7d7f
BL
1852 * Optimisation for the fact that under 0.01% of X
1853 * messages typically need breaking up.
1da177e4
LT
1854 */
1855
e9df7d7f 1856 size = len-sent;
1da177e4
LT
1857
1858 /* Keep two messages in the pipe so it schedules better */
e9df7d7f
BL
1859 if (size > ((sk->sk_sndbuf >> 1) - 64))
1860 size = (sk->sk_sndbuf >> 1) - 64;
1da177e4
LT
1861
1862 if (size > SKB_MAX_ALLOC)
1863 size = SKB_MAX_ALLOC;
ac7bfa62 1864
1da177e4
LT
1865 /*
1866 * Grab a buffer
1867 */
ac7bfa62 1868
6eba6a37
ED
1869 skb = sock_alloc_send_skb(sk, size, msg->msg_flags&MSG_DONTWAIT,
1870 &err);
6fa3eb70 1871
1da177e4 1872
e27dfcea 1873 if (skb == NULL)
1da177e4
LT
1874 goto out_err;
1875
1876 /*
1877 * If you pass two values to the sock_alloc_send_skb
1878 * it tries to grab the large buffer with GFP_NOFS
1879 * (which can fail easily), and if it fails grab the
1880 * fallback size buffer which is under a page and will
1881 * succeed. [Alan]
1882 */
1883 size = min_t(int, size, skb_tailroom(skb));
1884
7361c36c 1885
f78a5fda
DM
1886 /* Only send the fds in the first buffer */
1887 err = unix_scm_to_skb(siocb->scm, skb, !fds_sent);
25888e30 1888 if (err < 0) {
7361c36c 1889 kfree_skb(skb);
f78a5fda 1890 goto out_err;
6209344f 1891 }
25888e30 1892 max_level = err + 1;
7361c36c 1893 fds_sent = true;
1da177e4 1894
6eba6a37
ED
1895 err = memcpy_fromiovec(skb_put(skb, size), msg->msg_iov, size);
1896 if (err) {
1da177e4 1897 kfree_skb(skb);
f78a5fda 1898 goto out_err;
1da177e4
LT
1899 }
1900
1c92b4e5 1901 unix_state_lock(other);
1da177e4
LT
1902
1903 if (sock_flag(other, SOCK_DEAD) ||
1904 (other->sk_shutdown & RCV_SHUTDOWN))
6fa3eb70
S
1905 {
1906 if( other->sk_socket )
1907 {
1908 if(sk->sk_socket)
1909 {
1910
1911 #ifdef CONFIG_MTK_NET_LOGGING
1912 printk(KERN_INFO " [mtk_net][unix]: sendmsg[%lu:%lu]:peer close\n" ,SOCK_INODE(sk->sk_socket)->i_ino,SOCK_INODE(other->sk_socket)->i_ino);
1913 #endif
1914 }
1915 else{
1916 #ifdef CONFIG_MTK_NET_LOGGING
1917 printk(KERN_INFO " [mtk_net][unix]: sendmsg[null:%lu]:peer close\n" ,SOCK_INODE(other->sk_socket)->i_ino);
1918 #endif
1919 }
1920
1921 }
1922 else
1923 {
1924 #ifdef CONFIG_MTK_NET_LOGGING
1925 printk(KERN_INFO " [mtk_net][unix]: sendmsg:peer close \n" );
1926 #endif
1927 }
1928
1929
1da177e4 1930 goto pipe_err_free;
6fa3eb70 1931 }
1da177e4 1932
16e57262 1933 maybe_add_creds(skb, sock, other);
1da177e4 1934 skb_queue_tail(&other->sk_receive_queue, skb);
25888e30
ED
1935 if (max_level > unix_sk(other)->recursion_level)
1936 unix_sk(other)->recursion_level = max_level;
1c92b4e5 1937 unix_state_unlock(other);
1da177e4 1938 other->sk_data_ready(other, size);
e27dfcea 1939 sent += size;
1da177e4 1940 }
1da177e4 1941
f78a5fda 1942 scm_destroy(siocb->scm);
1da177e4
LT
1943 siocb->scm = NULL;
1944
1945 return sent;
1946
1947pipe_err_free:
1c92b4e5 1948 unix_state_unlock(other);
1da177e4
LT
1949 kfree_skb(skb);
1950pipe_err:
6eba6a37
ED
1951 if (sent == 0 && !(msg->msg_flags&MSG_NOSIGNAL))
1952 send_sig(SIGPIPE, current, 0);
1da177e4
LT
1953 err = -EPIPE;
1954out_err:
f78a5fda 1955 scm_destroy(siocb->scm);
1da177e4 1956 siocb->scm = NULL;
6fa3eb70 1957
1da177e4
LT
1958 return sent ? : err;
1959}
1960
1961static int unix_seqpacket_sendmsg(struct kiocb *kiocb, struct socket *sock,
1962 struct msghdr *msg, size_t len)
1963{
1964 int err;
1965 struct sock *sk = sock->sk;
ac7bfa62 1966
1da177e4
LT
1967 err = sock_error(sk);
1968 if (err)
1969 return err;
1970
1971 if (sk->sk_state != TCP_ESTABLISHED)
1972 return -ENOTCONN;
1973
1974 if (msg->msg_namelen)
1975 msg->msg_namelen = 0;
1976
1977 return unix_dgram_sendmsg(kiocb, sock, msg, len);
1978}
ac7bfa62 1979
a05d2ad1
EB
1980static int unix_seqpacket_recvmsg(struct kiocb *iocb, struct socket *sock,
1981 struct msghdr *msg, size_t size,
1982 int flags)
1983{
1984 struct sock *sk = sock->sk;
1985
1986 if (sk->sk_state != TCP_ESTABLISHED)
1987 return -ENOTCONN;
1988
1989 return unix_dgram_recvmsg(iocb, sock, msg, size, flags);
1990}
1991
1da177e4
LT
1992static void unix_copy_addr(struct msghdr *msg, struct sock *sk)
1993{
1994 struct unix_sock *u = unix_sk(sk);
1995
1da177e4
LT
1996 if (u->addr) {
1997 msg->msg_namelen = u->addr->len;
1998 memcpy(msg->msg_name, u->addr->name, u->addr->len);
1999 }
2000}
2001
2002static int unix_dgram_recvmsg(struct kiocb *iocb, struct socket *sock,
2003 struct msghdr *msg, size_t size,
2004 int flags)
2005{
2006 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
2007 struct scm_cookie tmp_scm;
2008 struct sock *sk = sock->sk;
2009 struct unix_sock *u = unix_sk(sk);
2010 int noblock = flags & MSG_DONTWAIT;
2011 struct sk_buff *skb;
2012 int err;
f55bb7f9 2013 int peeked, skip;
1da177e4
LT
2014
2015 err = -EOPNOTSUPP;
2016 if (flags&MSG_OOB)
2017 goto out;
2018
b3ca9b02 2019 err = mutex_lock_interruptible(&u->readlock);
f423fefe
ED
2020 if (unlikely(err)) {
2021 /* recvmsg() in non blocking mode is supposed to return -EAGAIN
2022 * sk_rcvtimeo is not honored by mutex_lock_interruptible()
2023 */
2024 err = noblock ? -EAGAIN : -ERESTARTSYS;
b3ca9b02
RW
2025 goto out;
2026 }
1da177e4 2027
f55bb7f9
PE
2028 skip = sk_peek_offset(sk, flags);
2029
2030 skb = __skb_recv_datagram(sk, flags, &peeked, &skip, &err);
0a112258
FZ
2031 if (!skb) {
2032 unix_state_lock(sk);
2033 /* Signal EOF on disconnected non-blocking SEQPACKET socket. */
2034 if (sk->sk_type == SOCK_SEQPACKET && err == -EAGAIN &&
2035 (sk->sk_shutdown & RCV_SHUTDOWN))
2036 err = 0;
2037 unix_state_unlock(sk);
1da177e4 2038 goto out_unlock;
0a112258 2039 }
1da177e4 2040
67426b75
ED
2041 wake_up_interruptible_sync_poll(&u->peer_wait,
2042 POLLOUT | POLLWRNORM | POLLWRBAND);
1da177e4
LT
2043
2044 if (msg->msg_name)
2045 unix_copy_addr(msg, skb->sk);
2046
f55bb7f9
PE
2047 if (size > skb->len - skip)
2048 size = skb->len - skip;
2049 else if (size < skb->len - skip)
1da177e4
LT
2050 msg->msg_flags |= MSG_TRUNC;
2051
f55bb7f9 2052 err = skb_copy_datagram_iovec(skb, skip, msg->msg_iov, size);
1da177e4
LT
2053 if (err)
2054 goto out_free;
2055
3f66116e
AC
2056 if (sock_flag(sk, SOCK_RCVTSTAMP))
2057 __sock_recv_timestamp(msg, sk, skb);
2058
1da177e4
LT
2059 if (!siocb->scm) {
2060 siocb->scm = &tmp_scm;
2061 memset(&tmp_scm, 0, sizeof(tmp_scm));
2062 }
6b0ee8c0 2063 scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid);
877ce7c1 2064 unix_set_secdata(siocb->scm, skb);
1da177e4 2065
6eba6a37 2066 if (!(flags & MSG_PEEK)) {
1da177e4
LT
2067 if (UNIXCB(skb).fp)
2068 unix_detach_fds(siocb->scm, skb);
f55bb7f9
PE
2069
2070 sk_peek_offset_bwd(sk, skb->len);
6eba6a37 2071 } else {
1da177e4
LT
2072 /* It is questionable: on PEEK we could:
2073 - do not return fds - good, but too simple 8)
2074 - return fds, and do not return them on read (old strategy,
2075 apparently wrong)
2076 - clone fds (I chose it for now, it is the most universal
2077 solution)
ac7bfa62
YH
2078
2079 POSIX 1003.1g does not actually define this clearly
2080 at all. POSIX 1003.1g doesn't define a lot of things
2081 clearly however!
2082
1da177e4 2083 */
f55bb7f9
PE
2084
2085 sk_peek_offset_fwd(sk, size);
2086
1da177e4
LT
2087 if (UNIXCB(skb).fp)
2088 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
2089 }
9f6f9af7 2090 err = (flags & MSG_TRUNC) ? skb->len - skip : size;
1da177e4
LT
2091
2092 scm_recv(sock, msg, siocb->scm, flags);
2093
2094out_free:
6eba6a37 2095 skb_free_datagram(sk, skb);
1da177e4 2096out_unlock:
57b47a53 2097 mutex_unlock(&u->readlock);
1da177e4 2098out:
6fa3eb70 2099
1da177e4
LT
2100 return err;
2101}
2102
2103/*
79f632c7 2104 * Sleep until more data has arrived. But check for races..
1da177e4 2105 */
79f632c7
BP
2106static long unix_stream_data_wait(struct sock *sk, long timeo,
2107 struct sk_buff *last)
1da177e4
LT
2108{
2109 DEFINE_WAIT(wait);
2110
1c92b4e5 2111 unix_state_lock(sk);
1da177e4
LT
2112
2113 for (;;) {
aa395145 2114 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4 2115
79f632c7 2116 if (skb_peek_tail(&sk->sk_receive_queue) != last ||
1da177e4
LT
2117 sk->sk_err ||
2118 (sk->sk_shutdown & RCV_SHUTDOWN) ||
2119 signal_pending(current) ||
2120 !timeo)
2121 break;
2122
2123 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1c92b4e5 2124 unix_state_unlock(sk);
6fa3eb70 2125 timeo = freezable_schedule_timeout(timeo);
1c92b4e5 2126 unix_state_lock(sk);
1da177e4
LT
2127 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
2128 }
2129
aa395145 2130 finish_wait(sk_sleep(sk), &wait);
1c92b4e5 2131 unix_state_unlock(sk);
1da177e4
LT
2132 return timeo;
2133}
2134
1da177e4
LT
2135static int unix_stream_recvmsg(struct kiocb *iocb, struct socket *sock,
2136 struct msghdr *msg, size_t size,
2137 int flags)
2138{
2139 struct sock_iocb *siocb = kiocb_to_siocb(iocb);
2140 struct scm_cookie tmp_scm;
2141 struct sock *sk = sock->sk;
2142 struct unix_sock *u = unix_sk(sk);
e27dfcea 2143 struct sockaddr_un *sunaddr = msg->msg_name;
1da177e4 2144 int copied = 0;
f423fefe 2145 int noblock = flags & MSG_DONTWAIT;
1da177e4
LT
2146 int check_creds = 0;
2147 int target;
2148 int err = 0;
2149 long timeo;
fc0d7536 2150 int skip;
6fa3eb70 2151 struct sock * other = unix_peer(sk);
1da177e4
LT
2152
2153 err = -EINVAL;
2154 if (sk->sk_state != TCP_ESTABLISHED)
2155 goto out;
2156
2157 err = -EOPNOTSUPP;
2158 if (flags&MSG_OOB)
2159 goto out;
2160
2161 target = sock_rcvlowat(sk, flags&MSG_WAITALL, size);
f423fefe 2162 timeo = sock_rcvtimeo(sk, noblock);
1da177e4 2163
1da177e4
LT
2164 /* Lock the socket to prevent queue disordering
2165 * while sleeps in memcpy_tomsg
2166 */
2167
2168 if (!siocb->scm) {
2169 siocb->scm = &tmp_scm;
2170 memset(&tmp_scm, 0, sizeof(tmp_scm));
2171 }
2172
b3ca9b02 2173 err = mutex_lock_interruptible(&u->readlock);
f423fefe
ED
2174 if (unlikely(err)) {
2175 /* recvmsg() in non blocking mode is supposed to return -EAGAIN
2176 * sk_rcvtimeo is not honored by mutex_lock_interruptible()
2177 */
2178 err = noblock ? -EAGAIN : -ERESTARTSYS;
b3ca9b02
RW
2179 goto out;
2180 }
1da177e4 2181
6eba6a37 2182 do {
1da177e4 2183 int chunk;
79f632c7 2184 struct sk_buff *skb, *last;
1da177e4 2185
3c0d2f37 2186 unix_state_lock(sk);
79f632c7 2187 last = skb = skb_peek(&sk->sk_receive_queue);
fc0d7536 2188again:
6eba6a37 2189 if (skb == NULL) {
25888e30 2190 unix_sk(sk)->recursion_level = 0;
1da177e4 2191 if (copied >= target)
3c0d2f37 2192 goto unlock;
1da177e4
LT
2193
2194 /*
2195 * POSIX 1003.1g mandates this order.
2196 */
ac7bfa62 2197
6eba6a37
ED
2198 err = sock_error(sk);
2199 if (err)
3c0d2f37 2200 goto unlock;
1da177e4 2201 if (sk->sk_shutdown & RCV_SHUTDOWN)
6fa3eb70
S
2202 {
2203 if(sk && sk->sk_socket )
2204 {
2205 if(other && other->sk_socket ){
2206 #ifdef CONFIG_MTK_NET_LOGGING
2207
2208 printk(KERN_INFO " [mtk_net][unix]: recvmsg[%lu:%lu]:exit read due to peer shutdown \n" ,SOCK_INODE(sk->sk_socket)->i_ino,SOCK_INODE(other->sk_socket)->i_ino);
2209 #endif
2210 }else{
2211 #ifdef CONFIG_MTK_NET_LOGGING
2212 printk(KERN_INFO "[mtk_net][unix]: recvmsg[%lu:null]:exit read due to peer shutdown \n" ,SOCK_INODE(sk->sk_socket)->i_ino);
2213 #endif
2214 }
2215 }
2216 else{
2217 #ifdef CONFIG_MTK_NET_LOGGING
2218 printk(KERN_INFO " [mtk_net][unix]: recvmsg: exit read due to peer shutdown \n" );
2219 #endif
2220 }
3c0d2f37 2221 goto unlock;
6fa3eb70 2222 }
3c0d2f37 2223 unix_state_unlock(sk);
1da177e4
LT
2224 err = -EAGAIN;
2225 if (!timeo)
2226 break;
57b47a53 2227 mutex_unlock(&u->readlock);
1da177e4 2228
79f632c7 2229 timeo = unix_stream_data_wait(sk, timeo, last);
6fa3eb70
S
2230 if (!timeo)
2231 {
2232 if(sk && sk->sk_socket )
2233 {
2234 if(other && other->sk_socket ){
2235 #ifdef CONFIG_MTK_NET_LOGGING
2236 printk(KERN_INFO " [mtk_net][unix]: recvmsg[%lu:%lu]:exit read due to timeout \n" ,SOCK_INODE(sk->sk_socket)->i_ino,SOCK_INODE(other->sk_socket)->i_ino);
2237 #endif
2238 }else{
2239 #ifdef CONFIG_MTK_NET_LOGGING
2240 printk(KERN_INFO " [mtk_net][unix]: recvmsg[%lu:null]:exit read due to timeout \n" ,SOCK_INODE(sk->sk_socket)->i_ino);
2241 #endif
2242 }
2243 }
2244 else
2245 {
2246 #ifdef CONFIG_MTK_NET_LOGGING
2247 printk(KERN_INFO " [mtk_net][unix]: recvmsg:exit read due to timeout \n" );
2248 #endif
2249 }
2250
2251 }
1da177e4 2252
b3ca9b02
RW
2253 if (signal_pending(current)
2254 || mutex_lock_interruptible(&u->readlock)) {
1da177e4
LT
2255 err = sock_intr_errno(timeo);
2256 goto out;
2257 }
b3ca9b02 2258
1da177e4 2259 continue;
3c0d2f37
MS
2260 unlock:
2261 unix_state_unlock(sk);
2262 break;
1da177e4 2263 }
fc0d7536 2264
79f632c7
BP
2265 skip = sk_peek_offset(sk, flags);
2266 while (skip >= skb->len) {
fc0d7536 2267 skip -= skb->len;
79f632c7 2268 last = skb;
fc0d7536 2269 skb = skb_peek_next(skb, &sk->sk_receive_queue);
79f632c7
BP
2270 if (!skb)
2271 goto again;
fc0d7536
PE
2272 }
2273
3c0d2f37 2274 unix_state_unlock(sk);
1da177e4
LT
2275
2276 if (check_creds) {
2277 /* Never glue messages from different writers */
7361c36c 2278 if ((UNIXCB(skb).pid != siocb->scm->pid) ||
6b0ee8c0
EB
2279 !uid_eq(UNIXCB(skb).uid, siocb->scm->creds.uid) ||
2280 !gid_eq(UNIXCB(skb).gid, siocb->scm->creds.gid))
1da177e4 2281 break;
0e82e7f6 2282 } else if (test_bit(SOCK_PASSCRED, &sock->flags)) {
1da177e4 2283 /* Copy credentials */
6b0ee8c0 2284 scm_set_cred(siocb->scm, UNIXCB(skb).pid, UNIXCB(skb).uid, UNIXCB(skb).gid);
1da177e4
LT
2285 check_creds = 1;
2286 }
2287
2288 /* Copy address just once */
6eba6a37 2289 if (sunaddr) {
1da177e4
LT
2290 unix_copy_addr(msg, skb->sk);
2291 sunaddr = NULL;
2292 }
2293
fc0d7536
PE
2294 chunk = min_t(unsigned int, skb->len - skip, size);
2295 if (memcpy_toiovec(msg->msg_iov, skb->data + skip, chunk)) {
1da177e4
LT
2296 if (copied == 0)
2297 copied = -EFAULT;
2298 break;
2299 }
2300 copied += chunk;
2301 size -= chunk;
2302
2303 /* Mark read part of skb as used */
6eba6a37 2304 if (!(flags & MSG_PEEK)) {
1da177e4
LT
2305 skb_pull(skb, chunk);
2306
fc0d7536
PE
2307 sk_peek_offset_bwd(sk, chunk);
2308
1da177e4
LT
2309 if (UNIXCB(skb).fp)
2310 unix_detach_fds(siocb->scm, skb);
2311
6f01fd6e 2312 if (skb->len)
1da177e4 2313 break;
1da177e4 2314
6f01fd6e 2315 skb_unlink(skb, &sk->sk_receive_queue);
70d4bf6d 2316 consume_skb(skb);
1da177e4
LT
2317
2318 if (siocb->scm->fp)
2319 break;
6eba6a37 2320 } else {
1da177e4
LT
2321 /* It is questionable, see note in unix_dgram_recvmsg.
2322 */
2323 if (UNIXCB(skb).fp)
2324 siocb->scm->fp = scm_fp_dup(UNIXCB(skb).fp);
2325
fc0d7536
PE
2326 sk_peek_offset_fwd(sk, chunk);
2327
1da177e4
LT
2328 break;
2329 }
2330 } while (size);
2331
57b47a53 2332 mutex_unlock(&u->readlock);
1da177e4
LT
2333 scm_recv(sock, msg, siocb->scm, flags);
2334out:
6fa3eb70 2335
1da177e4
LT
2336 return copied ? : err;
2337}
2338
2339static int unix_shutdown(struct socket *sock, int mode)
2340{
2341 struct sock *sk = sock->sk;
2342 struct sock *other;
2343
fc61b928
XW
2344 if (mode < SHUT_RD || mode > SHUT_RDWR)
2345 return -EINVAL;
2346 /* This maps:
2347 * SHUT_RD (0) -> RCV_SHUTDOWN (1)
2348 * SHUT_WR (1) -> SEND_SHUTDOWN (2)
2349 * SHUT_RDWR (2) -> SHUTDOWN_MASK (3)
2350 */
2351 ++mode;
7180a031
AC
2352
2353 unix_state_lock(sk);
2354 sk->sk_shutdown |= mode;
2355 other = unix_peer(sk);
2356 if (other)
2357 sock_hold(other);
2358 unix_state_unlock(sk);
2359 sk->sk_state_change(sk);
2360
2361 if (other &&
2362 (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET)) {
2363
2364 int peer_mode = 0;
2365
2366 if (mode&RCV_SHUTDOWN)
2367 peer_mode |= SEND_SHUTDOWN;
2368 if (mode&SEND_SHUTDOWN)
2369 peer_mode |= RCV_SHUTDOWN;
2370 unix_state_lock(other);
2371 other->sk_shutdown |= peer_mode;
2372 unix_state_unlock(other);
2373 other->sk_state_change(other);
2374 if (peer_mode == SHUTDOWN_MASK)
2375 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_HUP);
2376 else if (peer_mode & RCV_SHUTDOWN)
2377 sk_wake_async(other, SOCK_WAKE_WAITD, POLL_IN);
1da177e4 2378 }
7180a031
AC
2379 if (other)
2380 sock_put(other);
2381
1da177e4
LT
2382 return 0;
2383}
2384
885ee74d
PE
2385long unix_inq_len(struct sock *sk)
2386{
2387 struct sk_buff *skb;
2388 long amount = 0;
2389
2390 if (sk->sk_state == TCP_LISTEN)
2391 return -EINVAL;
2392
2393 spin_lock(&sk->sk_receive_queue.lock);
2394 if (sk->sk_type == SOCK_STREAM ||
2395 sk->sk_type == SOCK_SEQPACKET) {
2396 skb_queue_walk(&sk->sk_receive_queue, skb)
2397 amount += skb->len;
2398 } else {
2399 skb = skb_peek(&sk->sk_receive_queue);
2400 if (skb)
2401 amount = skb->len;
2402 }
2403 spin_unlock(&sk->sk_receive_queue.lock);
2404
2405 return amount;
2406}
2407EXPORT_SYMBOL_GPL(unix_inq_len);
2408
2409long unix_outq_len(struct sock *sk)
2410{
2411 return sk_wmem_alloc_get(sk);
2412}
2413EXPORT_SYMBOL_GPL(unix_outq_len);
2414
1da177e4
LT
2415static int unix_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2416{
2417 struct sock *sk = sock->sk;
e27dfcea 2418 long amount = 0;
1da177e4
LT
2419 int err;
2420
6eba6a37
ED
2421 switch (cmd) {
2422 case SIOCOUTQ:
885ee74d 2423 amount = unix_outq_len(sk);
6eba6a37
ED
2424 err = put_user(amount, (int __user *)arg);
2425 break;
2426 case SIOCINQ:
885ee74d
PE
2427 amount = unix_inq_len(sk);
2428 if (amount < 0)
2429 err = amount;
2430 else
1da177e4 2431 err = put_user(amount, (int __user *)arg);
885ee74d 2432 break;
6eba6a37
ED
2433 default:
2434 err = -ENOIOCTLCMD;
2435 break;
1da177e4
LT
2436 }
2437 return err;
2438}
2439
6eba6a37 2440static unsigned int unix_poll(struct file *file, struct socket *sock, poll_table *wait)
1da177e4
LT
2441{
2442 struct sock *sk = sock->sk;
2443 unsigned int mask;
2444
aa395145 2445 sock_poll_wait(file, sk_sleep(sk), wait);
1da177e4
LT
2446 mask = 0;
2447
2448 /* exceptional events? */
2449 if (sk->sk_err)
2450 mask |= POLLERR;
2451 if (sk->sk_shutdown == SHUTDOWN_MASK)
2452 mask |= POLLHUP;
f348d70a 2453 if (sk->sk_shutdown & RCV_SHUTDOWN)
db40980f 2454 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
1da177e4
LT
2455
2456 /* readable? */
db40980f 2457 if (!skb_queue_empty(&sk->sk_receive_queue))
1da177e4
LT
2458 mask |= POLLIN | POLLRDNORM;
2459
2460 /* Connection-based need to check for termination and startup */
6eba6a37
ED
2461 if ((sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) &&
2462 sk->sk_state == TCP_CLOSE)
1da177e4
LT
2463 mask |= POLLHUP;
2464
2465 /*
2466 * we set writable also when the other side has shut down the
2467 * connection. This prevents stuck sockets.
2468 */
2469 if (unix_writable(sk))
2470 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2471
2472 return mask;
2473}
2474
ec0d215f
RW
2475static unsigned int unix_dgram_poll(struct file *file, struct socket *sock,
2476 poll_table *wait)
3c73419c 2477{
ec0d215f
RW
2478 struct sock *sk = sock->sk, *other;
2479 unsigned int mask, writable;
3c73419c 2480
aa395145 2481 sock_poll_wait(file, sk_sleep(sk), wait);
3c73419c
RW
2482 mask = 0;
2483
2484 /* exceptional events? */
2485 if (sk->sk_err || !skb_queue_empty(&sk->sk_error_queue))
7d4c04fc 2486 mask |= POLLERR |
8facd5fb 2487 (sock_flag(sk, SOCK_SELECT_ERR_QUEUE) ? POLLPRI : 0);
7d4c04fc 2488
3c73419c 2489 if (sk->sk_shutdown & RCV_SHUTDOWN)
5456f09a 2490 mask |= POLLRDHUP | POLLIN | POLLRDNORM;
3c73419c
RW
2491 if (sk->sk_shutdown == SHUTDOWN_MASK)
2492 mask |= POLLHUP;
2493
2494 /* readable? */
5456f09a 2495 if (!skb_queue_empty(&sk->sk_receive_queue))
3c73419c
RW
2496 mask |= POLLIN | POLLRDNORM;
2497
2498 /* Connection-based need to check for termination and startup */
2499 if (sk->sk_type == SOCK_SEQPACKET) {
2500 if (sk->sk_state == TCP_CLOSE)
2501 mask |= POLLHUP;
2502 /* connection hasn't started yet? */
2503 if (sk->sk_state == TCP_SYN_SENT)
6fa3eb70
S
2504 {
2505
3c73419c 2506 return mask;
6fa3eb70
S
2507 }
2508 }
3c73419c 2509
973a34aa 2510 /* No write status requested, avoid expensive OUT tests. */
626cf236 2511 if (!(poll_requested_events(wait) & (POLLWRBAND|POLLWRNORM|POLLOUT)))
6fa3eb70 2512 {
973a34aa 2513 return mask;
6fa3eb70 2514 }
973a34aa 2515
ec0d215f 2516 writable = unix_writable(sk);
4b9e9796
S
2517 if (writable) {
2518 unix_state_lock(sk);
2519
2520 other = unix_peer(sk);
2521 if (other && unix_peer(other) != sk &&
2522 unix_recvq_full(other) &&
2523 unix_dgram_peer_wake_me(sk, other))
2524 writable = 0;
2525
2526 unix_state_unlock(sk);
ec0d215f
RW
2527 }
2528
2529 if (writable)
3c73419c
RW
2530 mask |= POLLOUT | POLLWRNORM | POLLWRBAND;
2531 else
2532 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
2533
3c73419c
RW
2534 return mask;
2535}
1da177e4
LT
2536
2537#ifdef CONFIG_PROC_FS
a53eb3fe 2538
7123aaa3
ED
2539#define BUCKET_SPACE (BITS_PER_LONG - (UNIX_HASH_BITS + 1) - 1)
2540
2541#define get_bucket(x) ((x) >> BUCKET_SPACE)
2542#define get_offset(x) ((x) & ((1L << BUCKET_SPACE) - 1))
2543#define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o))
a53eb3fe 2544
7123aaa3 2545static struct sock *unix_from_bucket(struct seq_file *seq, loff_t *pos)
1da177e4 2546{
7123aaa3
ED
2547 unsigned long offset = get_offset(*pos);
2548 unsigned long bucket = get_bucket(*pos);
2549 struct sock *sk;
2550 unsigned long count = 0;
1da177e4 2551
7123aaa3
ED
2552 for (sk = sk_head(&unix_socket_table[bucket]); sk; sk = sk_next(sk)) {
2553 if (sock_net(sk) != seq_file_net(seq))
097e66c5 2554 continue;
7123aaa3
ED
2555 if (++count == offset)
2556 break;
2557 }
2558
2559 return sk;
2560}
2561
2562static struct sock *unix_next_socket(struct seq_file *seq,
2563 struct sock *sk,
2564 loff_t *pos)
2565{
2566 unsigned long bucket;
2567
2568 while (sk > (struct sock *)SEQ_START_TOKEN) {
2569 sk = sk_next(sk);
2570 if (!sk)
2571 goto next_bucket;
2572 if (sock_net(sk) == seq_file_net(seq))
2573 return sk;
1da177e4 2574 }
7123aaa3
ED
2575
2576 do {
2577 sk = unix_from_bucket(seq, pos);
2578 if (sk)
2579 return sk;
2580
2581next_bucket:
2582 bucket = get_bucket(*pos) + 1;
2583 *pos = set_bucket_offset(bucket, 1);
2584 } while (bucket < ARRAY_SIZE(unix_socket_table));
2585
1da177e4
LT
2586 return NULL;
2587}
2588
1da177e4 2589static void *unix_seq_start(struct seq_file *seq, loff_t *pos)
9a429c49 2590 __acquires(unix_table_lock)
1da177e4 2591{
fbe9cc4a 2592 spin_lock(&unix_table_lock);
7123aaa3
ED
2593
2594 if (!*pos)
2595 return SEQ_START_TOKEN;
2596
2597 if (get_bucket(*pos) >= ARRAY_SIZE(unix_socket_table))
2598 return NULL;
2599
2600 return unix_next_socket(seq, NULL, pos);
1da177e4
LT
2601}
2602
2603static void *unix_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2604{
2605 ++*pos;
7123aaa3 2606 return unix_next_socket(seq, v, pos);
1da177e4
LT
2607}
2608
2609static void unix_seq_stop(struct seq_file *seq, void *v)
9a429c49 2610 __releases(unix_table_lock)
1da177e4 2611{
fbe9cc4a 2612 spin_unlock(&unix_table_lock);
1da177e4
LT
2613}
2614
2615static int unix_seq_show(struct seq_file *seq, void *v)
2616{
ac7bfa62 2617
b9f3124f 2618 if (v == SEQ_START_TOKEN)
1da177e4
LT
2619 seq_puts(seq, "Num RefCount Protocol Flags Type St "
2620 "Inode Path\n");
2621 else {
2622 struct sock *s = v;
2623 struct unix_sock *u = unix_sk(s);
1c92b4e5 2624 unix_state_lock(s);
1da177e4 2625
71338aa7 2626 seq_printf(seq, "%pK: %08X %08X %08X %04X %02X %5lu",
1da177e4
LT
2627 s,
2628 atomic_read(&s->sk_refcnt),
2629 0,
2630 s->sk_state == TCP_LISTEN ? __SO_ACCEPTCON : 0,
2631 s->sk_type,
2632 s->sk_socket ?
2633 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTED : SS_UNCONNECTED) :
2634 (s->sk_state == TCP_ESTABLISHED ? SS_CONNECTING : SS_DISCONNECTING),
2635 sock_i_ino(s));
2636
2637 if (u->addr) {
2638 int i, len;
2639 seq_putc(seq, ' ');
2640
2641 i = 0;
2642 len = u->addr->len - sizeof(short);
2643 if (!UNIX_ABSTRACT(s))
2644 len--;
2645 else {
2646 seq_putc(seq, '@');
2647 i++;
2648 }
2649 for ( ; i < len; i++)
2650 seq_putc(seq, u->addr->name->sun_path[i]);
2651 }
1c92b4e5 2652 unix_state_unlock(s);
1da177e4
LT
2653 seq_putc(seq, '\n');
2654 }
2655
2656 return 0;
2657}
2658
56b3d975 2659static const struct seq_operations unix_seq_ops = {
1da177e4
LT
2660 .start = unix_seq_start,
2661 .next = unix_seq_next,
2662 .stop = unix_seq_stop,
2663 .show = unix_seq_show,
2664};
2665
1da177e4
LT
2666static int unix_seq_open(struct inode *inode, struct file *file)
2667{
e372c414 2668 return seq_open_net(inode, file, &unix_seq_ops,
8b51b064 2669 sizeof(struct seq_net_private));
1da177e4
LT
2670}
2671
da7071d7 2672static const struct file_operations unix_seq_fops = {
1da177e4
LT
2673 .owner = THIS_MODULE,
2674 .open = unix_seq_open,
2675 .read = seq_read,
2676 .llseek = seq_lseek,
e372c414 2677 .release = seq_release_net,
1da177e4
LT
2678};
2679
2680#endif
2681
ec1b4cf7 2682static const struct net_proto_family unix_family_ops = {
1da177e4
LT
2683 .family = PF_UNIX,
2684 .create = unix_create,
2685 .owner = THIS_MODULE,
2686};
2687
097e66c5 2688
2c8c1e72 2689static int __net_init unix_net_init(struct net *net)
097e66c5
DL
2690{
2691 int error = -ENOMEM;
2692
a0a53c8b 2693 net->unx.sysctl_max_dgram_qlen = 10;
1597fbc0
PE
2694 if (unix_sysctl_register(net))
2695 goto out;
d392e497 2696
097e66c5 2697#ifdef CONFIG_PROC_FS
d4beaa66 2698 if (!proc_create("unix", 0, net->proc_net, &unix_seq_fops)) {
1597fbc0 2699 unix_sysctl_unregister(net);
097e66c5 2700 goto out;
1597fbc0 2701 }
097e66c5
DL
2702#endif
2703 error = 0;
2704out:
48dcc33e 2705 return error;
097e66c5
DL
2706}
2707
2c8c1e72 2708static void __net_exit unix_net_exit(struct net *net)
097e66c5 2709{
1597fbc0 2710 unix_sysctl_unregister(net);
ece31ffd 2711 remove_proc_entry("unix", net->proc_net);
097e66c5
DL
2712}
2713
2714static struct pernet_operations unix_net_ops = {
2715 .init = unix_net_init,
2716 .exit = unix_net_exit,
2717};
2718
1da177e4
LT
2719static int __init af_unix_init(void)
2720{
2721 int rc = -1;
1da177e4 2722
b4fff5f8 2723 BUILD_BUG_ON(sizeof(struct unix_skb_parms) > FIELD_SIZEOF(struct sk_buff, cb));
1da177e4
LT
2724
2725 rc = proto_register(&unix_proto, 1);
ac7bfa62
YH
2726 if (rc != 0) {
2727 printk(KERN_CRIT "%s: Cannot create unix_sock SLAB cache!\n",
0dc47877 2728 __func__);
1da177e4
LT
2729 goto out;
2730 }
2731
2732 sock_register(&unix_family_ops);
097e66c5 2733 register_pernet_subsys(&unix_net_ops);
1da177e4
LT
2734out:
2735 return rc;
2736}
2737
2738static void __exit af_unix_exit(void)
2739{
2740 sock_unregister(PF_UNIX);
1da177e4 2741 proto_unregister(&unix_proto);
097e66c5 2742 unregister_pernet_subsys(&unix_net_ops);
1da177e4
LT
2743}
2744
3d366960
DW
2745/* Earlier than device_initcall() so that other drivers invoking
2746 request_module() don't end up in a loop when modprobe tries
2747 to use a UNIX socket. But later than subsys_initcall() because
2748 we depend on stuff initialised there */
2749fs_initcall(af_unix_init);
1da177e4
LT
2750module_exit(af_unix_exit);
2751
2752MODULE_LICENSE("GPL");
2753MODULE_ALIAS_NETPROTO(PF_UNIX);