projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
userns: Convert sock_i_uid to return a kuid_t
[deliverable/linux.git] / net / core / sock.c
CommitLineData
1da177e4
LT		 1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Generic socket support routines. Memory allocators, socket lock/release
7 * handler for protocols to use and generic option handler.
8 *
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT		 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
14 *
15 * Fixes:
16 * Alan Cox : Numerous verify_area() problems
17 * Alan Cox : Connecting on a connecting socket
18 * now returns an error for tcp.
19 * Alan Cox : sock->protocol is set correctly.
20 * and is not sometimes left as 0.
21 * Alan Cox : connect handles icmp errors on a
22 * connect properly. Unfortunately there
23 * is a restart syscall nasty there. I
24 * can't match BSD without hacking the C
25 * library. Ideas urgently sought!
26 * Alan Cox : Disallow bind() to addresses that are
27 * not ours - especially broadcast ones!!
28 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost)
29 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets,
30 * instead they leave that for the DESTROY timer.
31 * Alan Cox : Clean up error flag in accept
32 * Alan Cox : TCP ack handling is buggy, the DESTROY timer
33 * was buggy. Put a remove_sock() in the handler
34 * for memory when we hit 0. Also altered the timer
4ec93edb 35 * code. The ACK stuff can wait and needs major
1da177e4
LT		 36 * TCP layer surgery.
37 * Alan Cox : Fixed TCP ack bug, removed remove sock
38 * and fixed timer/inet_bh race.
39 * Alan Cox : Added zapped flag for TCP
40 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code
41 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb
42 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources
43 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing.
44 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so...
45 * Rick Sladkey : Relaxed UDP rules for matching packets.
46 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support
47 * Pauline Middelink : identd support
48 * Alan Cox : Fixed connect() taking signals I think.
49 * Alan Cox : SO_LINGER supported
50 * Alan Cox : Error reporting fixes
51 * Anonymous : inet_create tidied up (sk->reuse setting)
52 * Alan Cox : inet sockets don't set sk->type!
53 * Alan Cox : Split socket option code
54 * Alan Cox : Callbacks
55 * Alan Cox : Nagle flag for Charles & Johannes stuff
56 * Alex : Removed restriction on inet fioctl
57 * Alan Cox : Splitting INET from NET core
58 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt()
59 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code
60 * Alan Cox : Split IP from generic code
61 * Alan Cox : New kfree_skbmem()
62 * Alan Cox : Make SO_DEBUG superuser only.
63 * Alan Cox : Allow anyone to clear SO_DEBUG
64 * (compatibility fix)
65 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput.
66 * Alan Cox : Allocator for a socket is settable.
67 * Alan Cox : SO_ERROR includes soft errors.
68 * Alan Cox : Allow NULL arguments on some SO_ opts
69 * Alan Cox : Generic socket allocation to make hooks
70 * easier (suggested by Craig Metz).
71 * Michael Pall : SO_ERROR returns positive errno again
72 * Steve Whitehouse: Added default destructor to free
73 * protocol private data.
74 * Steve Whitehouse: Added various other default routines
75 * common to several socket families.
76 * Chris Evans : Call suser() check last on F_SETOWN
77 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER.
78 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s()
79 * Andi Kleen : Fix write_space callback
80 * Chris Evans : Security fixes - signedness again
81 * Arnaldo C. Melo : cleanups, use skb_queue_purge
82 *
83 * To Fix:
84 *
85 *
86 * This program is free software; you can redistribute it and/or
87 * modify it under the terms of the GNU General Public License
88 * as published by the Free Software Foundation; either version
89 * 2 of the License, or (at your option) any later version.
90 */
91
e005d193
JP		 92#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
93
4fc268d2 94#include <linux/capability.h>
1da177e4
LT		 95#include <linux/errno.h>
96#include <linux/types.h>
97#include <linux/socket.h>
98#include <linux/in.h>
99#include <linux/kernel.h>
1da177e4
LT		 100#include <linux/module.h>
101#include <linux/proc_fs.h>
102#include <linux/seq_file.h>
103#include <linux/sched.h>
104#include <linux/timer.h>
105#include <linux/string.h>
106#include <linux/sockios.h>
107#include <linux/net.h>
108#include <linux/mm.h>
109#include <linux/slab.h>
110#include <linux/interrupt.h>
111#include <linux/poll.h>
112#include <linux/tcp.h>
113#include <linux/init.h>
a1f8e7f7 114#include <linux/highmem.h>
3f551f94 115#include <linux/user_namespace.h>
c5905afb 116#include <linux/static_key.h>
3969eb38 117#include <linux/memcontrol.h>
8c1ae10d 118#include <linux/prefetch.h>
1da177e4
LT		 119
120#include <asm/uaccess.h>
1da177e4
LT		 121
122#include <linux/netdevice.h>
123#include <net/protocol.h>
124#include <linux/skbuff.h>
457c4cbc 125#include <net/net_namespace.h>
2e6599cb 126#include <net/request_sock.h>
1da177e4 127#include <net/sock.h>
20d49473 128#include <linux/net_tstamp.h>
1da177e4
LT		 129#include <net/xfrm.h>
130#include <linux/ipsec.h>
f8451725 131#include <net/cls_cgroup.h>
5bc1421e 132#include <net/netprio_cgroup.h>
1da177e4
LT		 133
134#include <linux/filter.h>
135
3847ce32
SM		 136#include <trace/events/sock.h>
137
1da177e4
LT		 138#ifdef CONFIG_INET
139#include <net/tcp.h>
140#endif
141
36b77a52 142static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 143static LIST_HEAD(proto_list);
144
c255a458 145#ifdef CONFIG_MEMCG_KMEM
1d62e436 146int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
d1a4c0b3
GC		 147{
148 struct proto *proto;
149 int ret = 0;
150
36b77a52 151 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 152 list_for_each_entry(proto, &proto_list, node) {
153 if (proto->init_cgroup) {
1d62e436 154 ret = proto->init_cgroup(memcg, ss);
d1a4c0b3
GC		 155 if (ret)
156 goto out;
157 }
158 }
159
36b77a52 160 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 161 return ret;
162out:
163 list_for_each_entry_continue_reverse(proto, &proto_list, node)
164 if (proto->destroy_cgroup)
1d62e436 165 proto->destroy_cgroup(memcg);
36b77a52 166 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 167 return ret;
168}
169
1d62e436 170void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
d1a4c0b3
GC		 171{
172 struct proto *proto;
173
36b77a52 174 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 175 list_for_each_entry_reverse(proto, &proto_list, node)
176 if (proto->destroy_cgroup)
1d62e436 177 proto->destroy_cgroup(memcg);
36b77a52 178 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 179}
180#endif
181
da21f24d
IM		 182/*
183 * Each address family might have different locking rules, so we have
184 * one slock key per address family:
185 */
a5b5bb9a
IM		 186static struct lock_class_key af_family_keys[AF_MAX];
187static struct lock_class_key af_family_slock_keys[AF_MAX];
188
c5905afb 189struct static_key memcg_socket_limit_enabled;
e1aab161
GC		 190EXPORT_SYMBOL(memcg_socket_limit_enabled);
191
a5b5bb9a
IM		 192/*
193 * Make lock validator output more readable. (we pre-construct these
194 * strings build-time, so that runtime initialization of socket
195 * locks is fast):
196 */
36cbd3dc 197static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 198 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
199 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
200 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
201 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
202 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
203 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
204 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 205 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 206 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 207 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 208 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 209 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 210 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
c7fe3b52 211 "sk_lock-AF_NFC" , "sk_lock-AF_MAX"
a5b5bb9a 212};
36cbd3dc 213static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 214 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
215 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
216 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
217 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
218 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
219 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
220 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 221 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 222 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 223 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 224 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 225 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 226 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
c7fe3b52 227 "slock-AF_NFC" , "slock-AF_MAX"
a5b5bb9a 228};
36cbd3dc 229static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 230 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
231 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
232 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
233 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
234 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
235 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
236 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 237 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 238 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 239 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 240 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 241 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 242 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
c7fe3b52 243 "clock-AF_NFC" , "clock-AF_MAX"
443aef0e 244};
da21f24d
IM		 245
246/*
247 * sk_callback_lock locking rules are per-address-family,
248 * so split the lock classes by using a per-AF key:
249 */
250static struct lock_class_key af_callback_keys[AF_MAX];
251
1da177e4
LT		 252/* Take into consideration the size of the struct sk_buff overhead in the
253 * determination of these values, since that is non-constant across
254 * platforms. This makes socket queueing behavior and performance
255 * not depend upon such differences.
256 */
257#define _SK_MEM_PACKETS 256
87fb4b7b 258#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 259#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
260#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
261
262/* Run time adjustable parameters. */
ab32ea5d 263__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 264EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 265__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 266EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 267__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
268__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 269
25985edc 270/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 271int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 272EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 273
c93bdd0e
MG		 274struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
275EXPORT_SYMBOL_GPL(memalloc_socks);
276
7cb02404
MG		 277/**
278 * sk_set_memalloc - sets %SOCK_MEMALLOC
279 * @sk: socket to set it on
280 *
281 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
282 * It's the responsibility of the admin to adjust min_free_kbytes
283 * to meet the requirements
284 */
285void sk_set_memalloc(struct sock *sk)
286{
287 sock_set_flag(sk, SOCK_MEMALLOC);
288 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 289 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 290}
291EXPORT_SYMBOL_GPL(sk_set_memalloc);
292
293void sk_clear_memalloc(struct sock *sk)
294{
295 sock_reset_flag(sk, SOCK_MEMALLOC);
296 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 297 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 298
299 /*
300 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
301 * progress of swapping. However, if SOCK_MEMALLOC is cleared while
302 * it has rmem allocations there is a risk that the user of the
303 * socket cannot make forward progress due to exceeding the rmem
304 * limits. By rights, sk_clear_memalloc() should only be called
305 * on sockets being torn down but warn and reset the accounting if
306 * that assumption breaks.
307 */
308 if (WARN_ON(sk->sk_forward_alloc))
309 sk_mem_reclaim(sk);
7cb02404
MG		 310}
311EXPORT_SYMBOL_GPL(sk_clear_memalloc);
312
b4b9e355
MG		 313int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
314{
315 int ret;
316 unsigned long pflags = current->flags;
317
318 /* these should have been dropped before queueing */
319 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
320
321 current->flags |= PF_MEMALLOC;
322 ret = sk->sk_backlog_rcv(sk, skb);
323 tsk_restore_flags(current, pflags, PF_MEMALLOC);
324
325 return ret;
326}
327EXPORT_SYMBOL(__sk_backlog_rcv);
328
5bc1421e
NH		 329#if defined(CONFIG_CGROUPS)
330#if !defined(CONFIG_NET_CLS_CGROUP)
f8451725
HX		 331int net_cls_subsys_id = -1;
332EXPORT_SYMBOL_GPL(net_cls_subsys_id);
333#endif
5bc1421e
NH		 334#if !defined(CONFIG_NETPRIO_CGROUP)
335int net_prio_subsys_id = -1;
336EXPORT_SYMBOL_GPL(net_prio_subsys_id);
337#endif
338#endif
f8451725 339
1da177e4
LT		 340static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
341{
342 struct timeval tv;
343
344 if (optlen < sizeof(tv))
345 return -EINVAL;
346 if (copy_from_user(&tv, optval, sizeof(tv)))
347 return -EFAULT;
ba78073e
VA		 348 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
349 return -EDOM;
1da177e4 350
ba78073e 351 if (tv.tv_sec < 0) {
6f11df83
AM		 352 static int warned __read_mostly;
353
ba78073e 354 *timeo_p = 0;
50aab54f 355 if (warned < 10 && net_ratelimit()) {
ba78073e 356 warned++;
e005d193
JP		 357 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
358 __func__, current->comm, task_pid_nr(current));
50aab54f 359 }
ba78073e
VA		 360 return 0;
361 }
1da177e4
LT		 362 *timeo_p = MAX_SCHEDULE_TIMEOUT;
363 if (tv.tv_sec == 0 && tv.tv_usec == 0)
364 return 0;
365 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
366 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
367 return 0;
368}
369
370static void sock_warn_obsolete_bsdism(const char *name)
371{
372 static int warned;
373 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 374 if (strcmp(warncomm, current->comm) && warned < 5) {
375 strcpy(warncomm, current->comm);
e005d193
JP		 376 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
377 warncomm, name);
1da177e4
LT		 378 warned++;
379 }
380}
381
08e29af3
ED		 382#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
383
384static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 385{
08e29af3
ED		 386 if (sk->sk_flags & flags) {
387 sk->sk_flags &= ~flags;
388 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 389 net_disable_timestamp();
1da177e4
LT		 390 }
391}
392
393
f0088a50
DV		 394int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
395{
766e9037 396 int err;
f0088a50 397 int skb_len;
3b885787
NH		 398 unsigned long flags;
399 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 400
0fd7bac6 401 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 402 atomic_inc(&sk->sk_drops);
3847ce32 403 trace_sock_rcvqueue_full(sk, skb);
766e9037 404 return -ENOMEM;
f0088a50
DV		 405 }
406
fda9ef5d 407 err = sk_filter(sk, skb);
f0088a50 408 if (err)
766e9037 409 return err;
f0088a50 410
c76562b6 411 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 412 atomic_inc(&sk->sk_drops);
413 return -ENOBUFS;
3ab224be
HA		 414 }
415
f0088a50
DV		 416 skb->dev = NULL;
417 skb_set_owner_r(skb, sk);
49ad9599 418
f0088a50
DV		 419 /* Cache the SKB length before we tack it onto the receive
420 * queue. Once it is added it no longer belongs to us and
421 * may be freed by other threads of control pulling packets
422 * from the queue.
423 */
424 skb_len = skb->len;
425
7fee226a
ED		 426 /* we escape from rcu protected region, make sure we dont leak
427 * a norefcounted dst
428 */
429 skb_dst_force(skb);
430
3b885787
NH		 431 spin_lock_irqsave(&list->lock, flags);
432 skb->dropcount = atomic_read(&sk->sk_drops);
433 __skb_queue_tail(list, skb);
434 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 435
436 if (!sock_flag(sk, SOCK_DEAD))
437 sk->sk_data_ready(sk, skb_len);
766e9037 438 return 0;
f0088a50
DV		 439}
440EXPORT_SYMBOL(sock_queue_rcv_skb);
441
58a5a7b9 442int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 443{
444 int rc = NET_RX_SUCCESS;
445
fda9ef5d 446 if (sk_filter(sk, skb))
f0088a50
DV		 447 goto discard_and_relse;
448
449 skb->dev = NULL;
450
f545a38f 451 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) {
c377411f
ED		 452 atomic_inc(&sk->sk_drops);
453 goto discard_and_relse;
454 }
58a5a7b9
ACM		 455 if (nested)
456 bh_lock_sock_nested(sk);
457 else
458 bh_lock_sock(sk);
a5b5bb9a
IM		 459 if (!sock_owned_by_user(sk)) {
460 /*
461 * trylock + unlock semantics:
462 */
463 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
464
c57943a1 465 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 466
467 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 468 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 469 bh_unlock_sock(sk);
470 atomic_inc(&sk->sk_drops);
471 goto discard_and_relse;
472 }
473
f0088a50
DV		 474 bh_unlock_sock(sk);
475out:
476 sock_put(sk);
477 return rc;
478discard_and_relse:
479 kfree_skb(skb);
480 goto out;
481}
482EXPORT_SYMBOL(sk_receive_skb);
483
ea94ff3b
KK		 484void sk_reset_txq(struct sock *sk)
485{
486 sk_tx_queue_clear(sk);
487}
488EXPORT_SYMBOL(sk_reset_txq);
489
f0088a50
DV		 490struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
491{
b6c6712a 492 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 493
494 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 495 sk_tx_queue_clear(sk);
a9b3cd7f 496 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 497 dst_release(dst);
498 return NULL;
499 }
500
501 return dst;
502}
503EXPORT_SYMBOL(__sk_dst_check);
504
505struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
506{
507 struct dst_entry *dst = sk_dst_get(sk);
508
509 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
510 sk_dst_reset(sk);
511 dst_release(dst);
512 return NULL;
513 }
514
515 return dst;
516}
517EXPORT_SYMBOL(sk_dst_check);
518
4878809f
DM		 519static int sock_bindtodevice(struct sock *sk, char __user *optval, int optlen)
520{
521 int ret = -ENOPROTOOPT;
522#ifdef CONFIG_NETDEVICES
3b1e0a65 523 struct net *net = sock_net(sk);
4878809f
DM		 524 char devname[IFNAMSIZ];
525 int index;
526
527 /* Sorry... */
528 ret = -EPERM;
529 if (!capable(CAP_NET_RAW))
530 goto out;
531
532 ret = -EINVAL;
533 if (optlen < 0)
534 goto out;
535
536 /* Bind this socket to a particular device like "eth0",
537 * as specified in the passed interface name. If the
538 * name is "" or the option length is zero the socket
539 * is not bound.
540 */
541 if (optlen > IFNAMSIZ - 1)
542 optlen = IFNAMSIZ - 1;
543 memset(devname, 0, sizeof(devname));
544
545 ret = -EFAULT;
546 if (copy_from_user(devname, optval, optlen))
547 goto out;
548
000ba2e4
DM		 549 index = 0;
550 if (devname[0] != '\0') {
bf8e56bf 551 struct net_device *dev;
4878809f 552
bf8e56bf
ED		 553 rcu_read_lock();
554 dev = dev_get_by_name_rcu(net, devname);
555 if (dev)
556 index = dev->ifindex;
557 rcu_read_unlock();
4878809f
DM		 558 ret = -ENODEV;
559 if (!dev)
560 goto out;
4878809f
DM		 561 }
562
563 lock_sock(sk);
564 sk->sk_bound_dev_if = index;
565 sk_dst_reset(sk);
566 release_sock(sk);
567
568 ret = 0;
569
570out:
571#endif
572
573 return ret;
574}
575
c0ef877b
PE		 576static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
577{
578 if (valbool)
579 sock_set_flag(sk, bit);
580 else
581 sock_reset_flag(sk, bit);
582}
583
1da177e4
LT		 584/*
585 * This is meant for all protocols to use and covers goings on
586 * at the socket level. Everything here is generic.
587 */
588
589int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 590 char __user *optval, unsigned int optlen)
1da177e4 591{
2a91525c 592 struct sock *sk = sock->sk;
1da177e4
LT		 593 int val;
594 int valbool;
595 struct linger ling;
596 int ret = 0;
4ec93edb 597
1da177e4
LT		 598 /*
599 * Options without arguments
600 */
601
4878809f
DM		 602 if (optname == SO_BINDTODEVICE)
603 return sock_bindtodevice(sk, optval, optlen);
604
e71a4783
SH		 605 if (optlen < sizeof(int))
606 return -EINVAL;
4ec93edb 607
1da177e4
LT		 608 if (get_user(val, (int __user *)optval))
609 return -EFAULT;
4ec93edb 610
2a91525c 611 valbool = val ? 1 : 0;
1da177e4
LT		 612
613 lock_sock(sk);
614
2a91525c 615 switch (optname) {
e71a4783 616 case SO_DEBUG:
2a91525c 617 if (val && !capable(CAP_NET_ADMIN))
e71a4783 618 ret = -EACCES;
2a91525c 619 else
c0ef877b 620 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 621 break;
622 case SO_REUSEADDR:
4a17fd52 623 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783
SH		 624 break;
625 case SO_TYPE:
49c794e9 626 case SO_PROTOCOL:
0d6038ee 627 case SO_DOMAIN:
e71a4783
SH		 628 case SO_ERROR:
629 ret = -ENOPROTOOPT;
630 break;
631 case SO_DONTROUTE:
c0ef877b 632 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 633 break;
634 case SO_BROADCAST:
635 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
636 break;
637 case SO_SNDBUF:
638 /* Don't error on this BSD doesn't and if you think
82981930
ED		 639 * about it this is right. Otherwise apps have to
640 * play 'guess the biggest size' games. RCVBUF/SNDBUF
641 * are treated in BSD as hints
642 */
643 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 644set_sndbuf:
e71a4783 645 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 646 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
647 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 648 sk->sk_write_space(sk);
649 break;
1da177e4 650
e71a4783
SH		 651 case SO_SNDBUFFORCE:
652 if (!capable(CAP_NET_ADMIN)) {
653 ret = -EPERM;
654 break;
655 }
656 goto set_sndbuf;
b0573dea 657
e71a4783
SH		 658 case SO_RCVBUF:
659 /* Don't error on this BSD doesn't and if you think
82981930
ED		 660 * about it this is right. Otherwise apps have to
661 * play 'guess the biggest size' games. RCVBUF/SNDBUF
662 * are treated in BSD as hints
663 */
664 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 665set_rcvbuf:
e71a4783
SH		 666 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
667 /*
668 * We double it on the way in to account for
669 * "struct sk_buff" etc. overhead. Applications
670 * assume that the SO_RCVBUF setting they make will
671 * allow that much actual data to be received on that
672 * socket.
673 *
674 * Applications are unaware that "struct sk_buff" and
675 * other overheads allocate from the receive buffer
676 * during socket buffer allocation.
677 *
678 * And after considering the possible alternatives,
679 * returning the value we actually used in getsockopt
680 * is the most desirable behavior.
681 */
82981930 682 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 683 break;
684
685 case SO_RCVBUFFORCE:
686 if (!capable(CAP_NET_ADMIN)) {
687 ret = -EPERM;
1da177e4 688 break;
e71a4783
SH		 689 }
690 goto set_rcvbuf;
1da177e4 691
e71a4783 692 case SO_KEEPALIVE:
1da177e4 693#ifdef CONFIG_INET
e71a4783
SH		 694 if (sk->sk_protocol == IPPROTO_TCP)
695 tcp_set_keepalive(sk, valbool);
1da177e4 696#endif
e71a4783
SH		 697 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
698 break;
699
700 case SO_OOBINLINE:
701 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
702 break;
703
704 case SO_NO_CHECK:
705 sk->sk_no_check = valbool;
706 break;
707
708 case SO_PRIORITY:
709 if ((val >= 0 && val <= 6) || capable(CAP_NET_ADMIN))
710 sk->sk_priority = val;
711 else
712 ret = -EPERM;
713 break;
714
715 case SO_LINGER:
716 if (optlen < sizeof(ling)) {
717 ret = -EINVAL; /* 1003.1g */
1da177e4 718 break;
e71a4783 719 }
2a91525c 720 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 721 ret = -EFAULT;
1da177e4 722 break;
e71a4783
SH		 723 }
724 if (!ling.l_onoff)
725 sock_reset_flag(sk, SOCK_LINGER);
726 else {
1da177e4 727#if (BITS_PER_LONG == 32)
e71a4783
SH		 728 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
729 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 730 else
e71a4783
SH		 731#endif
732 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
733 sock_set_flag(sk, SOCK_LINGER);
734 }
735 break;
736
737 case SO_BSDCOMPAT:
738 sock_warn_obsolete_bsdism("setsockopt");
739 break;
740
741 case SO_PASSCRED:
742 if (valbool)
743 set_bit(SOCK_PASSCRED, &sock->flags);
744 else
745 clear_bit(SOCK_PASSCRED, &sock->flags);
746 break;
747
748 case SO_TIMESTAMP:
92f37fd2 749 case SO_TIMESTAMPNS:
e71a4783 750 if (valbool) {
92f37fd2
ED		 751 if (optname == SO_TIMESTAMP)
752 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
753 else
754 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 755 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 756 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 757 } else {
e71a4783 758 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 759 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
760 }
e71a4783
SH		 761 break;
762
20d49473
PO		 763 case SO_TIMESTAMPING:
764 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 765 ret = -EINVAL;
20d49473
PO		 766 break;
767 }
768 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
769 val & SOF_TIMESTAMPING_TX_HARDWARE);
770 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
771 val & SOF_TIMESTAMPING_TX_SOFTWARE);
772 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
773 val & SOF_TIMESTAMPING_RX_HARDWARE);
774 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
775 sock_enable_timestamp(sk,
776 SOCK_TIMESTAMPING_RX_SOFTWARE);
777 else
778 sock_disable_timestamp(sk,
08e29af3 779 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 780 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
781 val & SOF_TIMESTAMPING_SOFTWARE);
782 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
783 val & SOF_TIMESTAMPING_SYS_HARDWARE);
784 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
785 val & SOF_TIMESTAMPING_RAW_HARDWARE);
786 break;
787
e71a4783
SH		 788 case SO_RCVLOWAT:
789 if (val < 0)
790 val = INT_MAX;
791 sk->sk_rcvlowat = val ? : 1;
792 break;
793
794 case SO_RCVTIMEO:
795 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
796 break;
797
798 case SO_SNDTIMEO:
799 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
800 break;
1da177e4 801
e71a4783
SH		 802 case SO_ATTACH_FILTER:
803 ret = -EINVAL;
804 if (optlen == sizeof(struct sock_fprog)) {
805 struct sock_fprog fprog;
1da177e4 806
e71a4783
SH		 807 ret = -EFAULT;
808 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 809 break;
e71a4783
SH		 810
811 ret = sk_attach_filter(&fprog, sk);
812 }
813 break;
814
815 case SO_DETACH_FILTER:
55b33325 816 ret = sk_detach_filter(sk);
e71a4783 817 break;
1da177e4 818
e71a4783
SH		 819 case SO_PASSSEC:
820 if (valbool)
821 set_bit(SOCK_PASSSEC, &sock->flags);
822 else
823 clear_bit(SOCK_PASSSEC, &sock->flags);
824 break;
4a19ec58
LAT		 825 case SO_MARK:
826 if (!capable(CAP_NET_ADMIN))
827 ret = -EPERM;
2a91525c 828 else
4a19ec58 829 sk->sk_mark = val;
4a19ec58 830 break;
877ce7c1 831
1da177e4
LT		 832 /* We implement the SO_SNDLOWAT etc to
833 not be settable (1003.1g 5.3) */
3b885787 834 case SO_RXQ_OVFL:
8083f0fc 835 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 836 break;
6e3e939f
JB		 837
838 case SO_WIFI_STATUS:
839 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
840 break;
841
ef64a54f
PE		 842 case SO_PEEK_OFF:
843 if (sock->ops->set_peek_off)
844 sock->ops->set_peek_off(sk, val);
845 else
846 ret = -EOPNOTSUPP;
847 break;
3bdc0eba
BG		 848
849 case SO_NOFCS:
850 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
851 break;
852
e71a4783
SH		 853 default:
854 ret = -ENOPROTOOPT;
855 break;
4ec93edb 856 }
1da177e4
LT		 857 release_sock(sk);
858 return ret;
859}
2a91525c 860EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 861
862
3f551f94
EB		 863void cred_to_ucred(struct pid *pid, const struct cred *cred,
864 struct ucred *ucred)
865{
866 ucred->pid = pid_vnr(pid);
867 ucred->uid = ucred->gid = -1;
868 if (cred) {
869 struct user_namespace *current_ns = current_user_ns();
870
b2e4f544
EB		 871 ucred->uid = from_kuid_munged(current_ns, cred->euid);
872 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 873 }
874}
3924773a 875EXPORT_SYMBOL_GPL(cred_to_ucred);
3f551f94 876
1da177e4
LT		 877int sock_getsockopt(struct socket *sock, int level, int optname,
878 char __user *optval, int __user *optlen)
879{
880 struct sock *sk = sock->sk;
4ec93edb 881
e71a4783 882 union {
4ec93edb
YH		 883 int val;
884 struct linger ling;
1da177e4
LT		 885 struct timeval tm;
886 } v;
4ec93edb 887
4d0392be 888 int lv = sizeof(int);
1da177e4 889 int len;
4ec93edb 890
e71a4783 891 if (get_user(len, optlen))
4ec93edb 892 return -EFAULT;
e71a4783 893 if (len < 0)
1da177e4 894 return -EINVAL;
4ec93edb 895
50fee1de 896 memset(&v, 0, sizeof(v));
df0bca04 897
2a91525c 898 switch (optname) {
e71a4783
SH		 899 case SO_DEBUG:
900 v.val = sock_flag(sk, SOCK_DBG);
901 break;
902
903 case SO_DONTROUTE:
904 v.val = sock_flag(sk, SOCK_LOCALROUTE);
905 break;
906
907 case SO_BROADCAST:
1b23a5df 908 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 909 break;
910
911 case SO_SNDBUF:
912 v.val = sk->sk_sndbuf;
913 break;
914
915 case SO_RCVBUF:
916 v.val = sk->sk_rcvbuf;
917 break;
918
919 case SO_REUSEADDR:
920 v.val = sk->sk_reuse;
921 break;
922
923 case SO_KEEPALIVE:
1b23a5df 924 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 925 break;
926
927 case SO_TYPE:
928 v.val = sk->sk_type;
929 break;
930
49c794e9
JE		 931 case SO_PROTOCOL:
932 v.val = sk->sk_protocol;
933 break;
934
0d6038ee
JE		 935 case SO_DOMAIN:
936 v.val = sk->sk_family;
937 break;
938
e71a4783
SH		 939 case SO_ERROR:
940 v.val = -sock_error(sk);
2a91525c 941 if (v.val == 0)
e71a4783
SH		 942 v.val = xchg(&sk->sk_err_soft, 0);
943 break;
944
945 case SO_OOBINLINE:
1b23a5df 946 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 947 break;
948
949 case SO_NO_CHECK:
950 v.val = sk->sk_no_check;
951 break;
952
953 case SO_PRIORITY:
954 v.val = sk->sk_priority;
955 break;
956
957 case SO_LINGER:
958 lv = sizeof(v.ling);
1b23a5df 959 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 960 v.ling.l_linger = sk->sk_lingertime / HZ;
961 break;
962
963 case SO_BSDCOMPAT:
964 sock_warn_obsolete_bsdism("getsockopt");
965 break;
966
967 case SO_TIMESTAMP:
92f37fd2
ED		 968 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
969 !sock_flag(sk, SOCK_RCVTSTAMPNS);
970 break;
971
972 case SO_TIMESTAMPNS:
973 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 974 break;
975
20d49473
PO		 976 case SO_TIMESTAMPING:
977 v.val = 0;
978 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
979 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
980 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
981 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
982 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
983 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
984 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
985 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
986 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
987 v.val |= SOF_TIMESTAMPING_SOFTWARE;
988 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
989 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
990 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
991 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
992 break;
993
e71a4783 994 case SO_RCVTIMEO:
2a91525c 995 lv = sizeof(struct timeval);
e71a4783
SH		 996 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
997 v.tm.tv_sec = 0;
998 v.tm.tv_usec = 0;
999 } else {
1000 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1001 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1002 }
1003 break;
1004
1005 case SO_SNDTIMEO:
2a91525c 1006 lv = sizeof(struct timeval);
e71a4783
SH		 1007 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1008 v.tm.tv_sec = 0;
1009 v.tm.tv_usec = 0;
1010 } else {
1011 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1012 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1013 }
1014 break;
1da177e4 1015
e71a4783
SH		 1016 case SO_RCVLOWAT:
1017 v.val = sk->sk_rcvlowat;
1018 break;
1da177e4 1019
e71a4783 1020 case SO_SNDLOWAT:
2a91525c 1021 v.val = 1;
e71a4783 1022 break;
1da177e4 1023
e71a4783 1024 case SO_PASSCRED:
82981930 1025 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1026 break;
1da177e4 1027
e71a4783 1028 case SO_PEERCRED:
109f6e39
EB		 1029 {
1030 struct ucred peercred;
1031 if (len > sizeof(peercred))
1032 len = sizeof(peercred);
1033 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1034 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1035 return -EFAULT;
1036 goto lenout;
109f6e39 1037 }
1da177e4 1038
e71a4783
SH		 1039 case SO_PEERNAME:
1040 {
1041 char address[128];
1042
1043 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1044 return -ENOTCONN;
1045 if (lv < len)
1046 return -EINVAL;
1047 if (copy_to_user(optval, address, len))
1048 return -EFAULT;
1049 goto lenout;
1050 }
1da177e4 1051
e71a4783
SH		 1052 /* Dubious BSD thing... Probably nobody even uses it, but
1053 * the UNIX standard wants it for whatever reason... -DaveM
1054 */
1055 case SO_ACCEPTCONN:
1056 v.val = sk->sk_state == TCP_LISTEN;
1057 break;
1da177e4 1058
e71a4783 1059 case SO_PASSSEC:
82981930 1060 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1061 break;
877ce7c1 1062
e71a4783
SH		 1063 case SO_PEERSEC:
1064 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1065
4a19ec58
LAT		 1066 case SO_MARK:
1067 v.val = sk->sk_mark;
1068 break;
1069
3b885787 1070 case SO_RXQ_OVFL:
1b23a5df 1071 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1072 break;
1073
6e3e939f 1074 case SO_WIFI_STATUS:
1b23a5df 1075 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1076 break;
1077
ef64a54f
PE		 1078 case SO_PEEK_OFF:
1079 if (!sock->ops->set_peek_off)
1080 return -EOPNOTSUPP;
1081
1082 v.val = sk->sk_peek_off;
1083 break;
bc2f7996 1084 case SO_NOFCS:
1b23a5df 1085 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1086 break;
e71a4783
SH		 1087 default:
1088 return -ENOPROTOOPT;
1da177e4 1089 }
e71a4783 1090
1da177e4
LT		 1091 if (len > lv)
1092 len = lv;
1093 if (copy_to_user(optval, &v, len))
1094 return -EFAULT;
1095lenout:
4ec93edb
YH		 1096 if (put_user(len, optlen))
1097 return -EFAULT;
1098 return 0;
1da177e4
LT		 1099}
1100
a5b5bb9a
IM		 1101/*
1102 * Initialize an sk_lock.
1103 *
1104 * (We also register the sk_lock with the lock validator.)
1105 */
b6f99a21 1106static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1107{
ed07536e
PZ		 1108 sock_lock_init_class_and_name(sk,
1109 af_family_slock_key_strings[sk->sk_family],
1110 af_family_slock_keys + sk->sk_family,
1111 af_family_key_strings[sk->sk_family],
1112 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1113}
1114
4dc6dc71
ED		 1115/*
1116 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1117 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1118 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1119 */
f1a6c4da
PE		 1120static void sock_copy(struct sock *nsk, const struct sock *osk)
1121{
1122#ifdef CONFIG_SECURITY_NETWORK
1123 void *sptr = nsk->sk_security;
1124#endif
68835aba
ED		 1125 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1126
1127 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1128 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1129
f1a6c4da
PE		 1130#ifdef CONFIG_SECURITY_NETWORK
1131 nsk->sk_security = sptr;
1132 security_sk_clone(osk, nsk);
1133#endif
1134}
1135
fcbdf09d
OP		 1136/*
1137 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
1138 * un-modified. Special care is taken when initializing object to zero.
1139 */
1140static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1141{
1142 if (offsetof(struct sock, sk_node.next) != 0)
1143 memset(sk, 0, offsetof(struct sock, sk_node.next));
1144 memset(&sk->sk_node.pprev, 0,
1145 size - offsetof(struct sock, sk_node.pprev));
1146}
1147
1148void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1149{
1150 unsigned long nulls1, nulls2;
1151
1152 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1153 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1154 if (nulls1 > nulls2)
1155 swap(nulls1, nulls2);
1156
1157 if (nulls1 != 0)
1158 memset((char *)sk, 0, nulls1);
1159 memset((char *)sk + nulls1 + sizeof(void *), 0,
1160 nulls2 - nulls1 - sizeof(void *));
1161 memset((char *)sk + nulls2 + sizeof(void *), 0,
1162 size - nulls2 - sizeof(void *));
1163}
1164EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1165
2e4afe7b
PE		 1166static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1167 int family)
c308c1b2
PE		 1168{
1169 struct sock *sk;
1170 struct kmem_cache *slab;
1171
1172 slab = prot->slab;
e912b114
ED		 1173 if (slab != NULL) {
1174 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1175 if (!sk)
1176 return sk;
1177 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1178 if (prot->clear_sk)
1179 prot->clear_sk(sk, prot->obj_size);
1180 else
1181 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1182 }
fcbdf09d 1183 } else
c308c1b2
PE		 1184 sk = kmalloc(prot->obj_size, priority);
1185
2e4afe7b 1186 if (sk != NULL) {
a98b65a3
VN		 1187 kmemcheck_annotate_bitfield(sk, flags);
1188
2e4afe7b
PE		 1189 if (security_sk_alloc(sk, family, priority))
1190 goto out_free;
1191
1192 if (!try_module_get(prot->owner))
1193 goto out_free_sec;
e022f0b4 1194 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1195 }
1196
c308c1b2 1197 return sk;
2e4afe7b
PE		 1198
1199out_free_sec:
1200 security_sk_free(sk);
1201out_free:
1202 if (slab != NULL)
1203 kmem_cache_free(slab, sk);
1204 else
1205 kfree(sk);
1206 return NULL;
c308c1b2
PE		 1207}
1208
1209static void sk_prot_free(struct proto *prot, struct sock *sk)
1210{
1211 struct kmem_cache *slab;
2e4afe7b 1212 struct module *owner;
c308c1b2 1213
2e4afe7b 1214 owner = prot->owner;
c308c1b2 1215 slab = prot->slab;
2e4afe7b
PE		 1216
1217 security_sk_free(sk);
c308c1b2
PE		 1218 if (slab != NULL)
1219 kmem_cache_free(slab, sk);
1220 else
1221 kfree(sk);
2e4afe7b 1222 module_put(owner);
c308c1b2
PE		 1223}
1224
f8451725
HX		 1225#ifdef CONFIG_CGROUPS
1226void sock_update_classid(struct sock *sk)
1227{
1144182a 1228 u32 classid;
f8451725 1229
1144182a
PM		 1230 rcu_read_lock(); /* doing current task, which cannot vanish. */
1231 classid = task_cls_classid(current);
1232 rcu_read_unlock();
f8451725
HX		 1233 if (classid && classid != sk->sk_classid)
1234 sk->sk_classid = classid;
1235}
82862742 1236EXPORT_SYMBOL(sock_update_classid);
5bc1421e 1237
406a3c63 1238void sock_update_netprioidx(struct sock *sk, struct task_struct *task)
5bc1421e 1239{
5bc1421e
NH		 1240 if (in_interrupt())
1241 return;
2b73bc65 1242
406a3c63 1243 sk->sk_cgrp_prioidx = task_netprioidx(task);
5bc1421e
NH		 1244}
1245EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1246#endif
1247
1da177e4
LT		 1248/**
1249 * sk_alloc - All socket objects are allocated here
c4ea43c5 1250 * @net: the applicable net namespace
4dc3b16b
PP		 1251 * @family: protocol family
1252 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1253 * @prot: struct proto associated with this new sock instance
1da177e4 1254 */
1b8d7ae4 1255struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1256 struct proto *prot)
1da177e4 1257{
c308c1b2 1258 struct sock *sk;
1da177e4 1259
154adbc8 1260 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1261 if (sk) {
154adbc8
PE		 1262 sk->sk_family = family;
1263 /*
1264 * See comment in struct sock definition to understand
1265 * why we need sk_prot_creator -acme
1266 */
1267 sk->sk_prot = sk->sk_prot_creator = prot;
1268 sock_lock_init(sk);
3b1e0a65 1269 sock_net_set(sk, get_net(net));
d66ee058 1270 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725
HX		 1271
1272 sock_update_classid(sk);
406a3c63 1273 sock_update_netprioidx(sk, current);
1da177e4 1274 }
a79af59e 1275
2e4afe7b 1276 return sk;
1da177e4 1277}
2a91525c 1278EXPORT_SYMBOL(sk_alloc);
1da177e4 1279
2b85a34e 1280static void __sk_free(struct sock *sk)
1da177e4
LT		 1281{
1282 struct sk_filter *filter;
1da177e4
LT		 1283
1284 if (sk->sk_destruct)
1285 sk->sk_destruct(sk);
1286
a898def2
PM		 1287 filter = rcu_dereference_check(sk->sk_filter,
1288 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1289 if (filter) {
309dd5fc 1290 sk_filter_uncharge(sk, filter);
a9b3cd7f 1291 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1292 }
1293
08e29af3 1294 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1295
1296 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1297 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1298 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1299
109f6e39
EB		 1300 if (sk->sk_peer_cred)
1301 put_cred(sk->sk_peer_cred);
1302 put_pid(sk->sk_peer_pid);
3b1e0a65 1303 put_net(sock_net(sk));
c308c1b2 1304 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1305}
2b85a34e
ED		 1306
1307void sk_free(struct sock *sk)
1308{
1309 /*
25985edc 1310 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1311 * some packets are still in some tx queue.
1312 * If not null, sock_wfree() will call __sk_free(sk) later
1313 */
1314 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1315 __sk_free(sk);
1316}
2a91525c 1317EXPORT_SYMBOL(sk_free);
1da177e4 1318
edf02087 1319/*
25985edc
LDM		 1320 * Last sock_put should drop reference to sk->sk_net. It has already
1321 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1322 * is not an option.
25985edc 1323 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1324 * destroy it in the context of init_net.
1325 */
1326void sk_release_kernel(struct sock *sk)
1327{
1328 if (sk == NULL || sk->sk_socket == NULL)
1329 return;
1330
1331 sock_hold(sk);
1332 sock_release(sk->sk_socket);
65a18ec5 1333 release_net(sock_net(sk));
3b1e0a65 1334 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1335 sock_put(sk);
1336}
45af1754 1337EXPORT_SYMBOL(sk_release_kernel);
edf02087 1338
475f1b52
SR		 1339static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1340{
1341 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1342 sock_update_memcg(newsk);
1343}
1344
e56c57d0
ED		 1345/**
1346 * sk_clone_lock - clone a socket, and lock its clone
1347 * @sk: the socket to clone
1348 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1349 *
1350 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1351 */
1352struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1353{
8fd1d178 1354 struct sock *newsk;
87d11ceb 1355
8fd1d178 1356 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1357 if (newsk != NULL) {
1358 struct sk_filter *filter;
1359
892c141e 1360 sock_copy(newsk, sk);
87d11ceb
ACM		 1361
1362 /* SANITY */
3b1e0a65 1363 get_net(sock_net(newsk));
87d11ceb
ACM		 1364 sk_node_init(&newsk->sk_node);
1365 sock_lock_init(newsk);
1366 bh_lock_sock(newsk);
fa438ccf 1367 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1368 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1369
1370 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1371 /*
1372 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1373 */
1374 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1375 atomic_set(&newsk->sk_omem_alloc, 0);
1376 skb_queue_head_init(&newsk->sk_receive_queue);
1377 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1378#ifdef CONFIG_NET_DMA
1379 skb_queue_head_init(&newsk->sk_async_wait_queue);
1380#endif
87d11ceb 1381
b6c6712a 1382 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1383 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1384 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1385 af_callback_keys + newsk->sk_family,
1386 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1387
1388 newsk->sk_dst_cache = NULL;
1389 newsk->sk_wmem_queued = 0;
1390 newsk->sk_forward_alloc = 0;
1391 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1392 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1393
1394 sock_reset_flag(newsk, SOCK_DONE);
1395 skb_queue_head_init(&newsk->sk_error_queue);
1396
0d7da9dd 1397 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1398 if (filter != NULL)
1399 sk_filter_charge(newsk, filter);
1400
1401 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1402 /* It is still raw copy of parent, so invalidate
1403 * destructor and make plain sk_free() */
1404 newsk->sk_destruct = NULL;
b0691c8e 1405 bh_unlock_sock(newsk);
87d11ceb
ACM		 1406 sk_free(newsk);
1407 newsk = NULL;
1408 goto out;
1409 }
1410
1411 newsk->sk_err = 0;
1412 newsk->sk_priority = 0;
4dc6dc71
ED		 1413 /*
1414 * Before updating sk_refcnt, we must commit prior changes to memory
1415 * (Documentation/RCU/rculist_nulls.txt for details)
1416 */
1417 smp_wmb();
87d11ceb
ACM		 1418 atomic_set(&newsk->sk_refcnt, 2);
1419
1420 /*
1421 * Increment the counter in the same struct proto as the master
1422 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1423 * is the same as sk->sk_prot->socks, as this field was copied
1424 * with memcpy).
1425 *
1426 * This _changes_ the previous behaviour, where
1427 * tcp_create_openreq_child always was incrementing the
1428 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1429 * to be taken into account in all callers. -acme
1430 */
1431 sk_refcnt_debug_inc(newsk);
972692e0 1432 sk_set_socket(newsk, NULL);
43815482 1433 newsk->sk_wq = NULL;
87d11ceb 1434
f3f511e1
GC		 1435 sk_update_clone(sk, newsk);
1436
87d11ceb 1437 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1438 sk_sockets_allocated_inc(newsk);
704da560 1439
08e29af3 1440 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1441 net_enable_timestamp();
87d11ceb
ACM		 1442 }
1443out:
1444 return newsk;
1445}
e56c57d0 1446EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1447
9958089a
AK		 1448void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1449{
1450 __sk_dst_set(sk, dst);
1451 sk->sk_route_caps = dst->dev->features;
1452 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1453 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1454 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1455 if (sk_can_gso(sk)) {
82cc1a7a 1456 if (dst->header_len) {
9958089a 1457 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1458 } else {
9958089a 1459 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a
PWJ		 1460 sk->sk_gso_max_size = dst->dev->gso_max_size;
1461 }
9958089a
AK		 1462 }
1463}
1464EXPORT_SYMBOL_GPL(sk_setup_caps);
1465
1da177e4
LT		 1466void __init sk_init(void)
1467{
4481374c 1468 if (totalram_pages <= 4096) {
1da177e4
LT		 1469 sysctl_wmem_max = 32767;
1470 sysctl_rmem_max = 32767;
1471 sysctl_wmem_default = 32767;
1472 sysctl_rmem_default = 32767;
4481374c 1473 } else if (totalram_pages >= 131072) {
1da177e4
LT		 1474 sysctl_wmem_max = 131071;
1475 sysctl_rmem_max = 131071;
1476 }
1477}
1478
1479/*
1480 * Simple resource managers for sockets.
1481 */
1482
1483
4ec93edb
YH		 1484/*
1485 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1486 */
1487void sock_wfree(struct sk_buff *skb)
1488{
1489 struct sock *sk = skb->sk;
d99927f4 1490 unsigned int len = skb->truesize;
1da177e4 1491
d99927f4
ED		 1492 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1493 /*
1494 * Keep a reference on sk_wmem_alloc, this will be released
1495 * after sk_write_space() call
1496 */
1497 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1498 sk->sk_write_space(sk);
d99927f4
ED		 1499 len = 1;
1500 }
2b85a34e 1501 /*
d99927f4
ED		 1502 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1503 * could not do because of in-flight packets
2b85a34e 1504 */
d99927f4 1505 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1506 __sk_free(sk);
1da177e4 1507}
2a91525c 1508EXPORT_SYMBOL(sock_wfree);
1da177e4 1509
4ec93edb
YH		 1510/*
1511 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1512 */
1513void sock_rfree(struct sk_buff *skb)
1514{
1515 struct sock *sk = skb->sk;
d361fd59 1516 unsigned int len = skb->truesize;
1da177e4 1517
d361fd59
ED		 1518 atomic_sub(len, &sk->sk_rmem_alloc);
1519 sk_mem_uncharge(sk, len);
1da177e4 1520}
2a91525c 1521EXPORT_SYMBOL(sock_rfree);
1da177e4 1522
41063e9d
DM		 1523void sock_edemux(struct sk_buff *skb)
1524{
1525 sock_put(skb->sk);
1526}
1527EXPORT_SYMBOL(sock_edemux);
1da177e4 1528
976d0201 1529kuid_t sock_i_uid(struct sock *sk)
1da177e4 1530{
976d0201 1531 kuid_t uid;
1da177e4 1532
f064af1e 1533 read_lock_bh(&sk->sk_callback_lock);
976d0201 1534 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1535 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1536 return uid;
1537}
2a91525c 1538EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1539
1540unsigned long sock_i_ino(struct sock *sk)
1541{
1542 unsigned long ino;
1543
f064af1e 1544 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1545 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1546 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1547 return ino;
1548}
2a91525c 1549EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1550
1551/*
1552 * Allocate a skb from the socket's send buffer.
1553 */
86a76caf 1554struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1555 gfp_t priority)
1da177e4
LT		 1556{
1557 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1558 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1559 if (skb) {
1560 skb_set_owner_w(skb, sk);
1561 return skb;
1562 }
1563 }
1564 return NULL;
1565}
2a91525c 1566EXPORT_SYMBOL(sock_wmalloc);
1da177e4
LT		 1567
1568/*
1569 * Allocate a skb from the socket's receive buffer.
4ec93edb 1570 */
86a76caf 1571struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1572 gfp_t priority)
1da177e4
LT		 1573{
1574 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1575 struct sk_buff *skb = alloc_skb(size, priority);
1576 if (skb) {
1577 skb_set_owner_r(skb, sk);
1578 return skb;
1579 }
1580 }
1581 return NULL;
1582}
1583
4ec93edb 1584/*
1da177e4 1585 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1586 */
dd0fc66f 1587void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1588{
95c96174 1589 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1590 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1591 void *mem;
1592 /* First do the add, to avoid the race if kmalloc
4ec93edb 1593 * might sleep.
1da177e4
LT		 1594 */
1595 atomic_add(size, &sk->sk_omem_alloc);
1596 mem = kmalloc(size, priority);
1597 if (mem)
1598 return mem;
1599 atomic_sub(size, &sk->sk_omem_alloc);
1600 }
1601 return NULL;
1602}
2a91525c 1603EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1604
1605/*
1606 * Free an option memory block.
1607 */
1608void sock_kfree_s(struct sock *sk, void *mem, int size)
1609{
1610 kfree(mem);
1611 atomic_sub(size, &sk->sk_omem_alloc);
1612}
2a91525c 1613EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1614
1615/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1616 I think, these locks should be removed for datagram sockets.
1617 */
2a91525c 1618static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1619{
1620 DEFINE_WAIT(wait);
1621
1622 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1623 for (;;) {
1624 if (!timeo)
1625 break;
1626 if (signal_pending(current))
1627 break;
1628 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1629 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1630 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1631 break;
1632 if (sk->sk_shutdown & SEND_SHUTDOWN)
1633 break;
1634 if (sk->sk_err)
1635 break;
1636 timeo = schedule_timeout(timeo);
1637 }
aa395145 1638 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1639 return timeo;
1640}
1641
1642
1643/*
1644 * Generic send/receive buffer handlers
1645 */
1646
4cc7f68d
HX		 1647struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1648 unsigned long data_len, int noblock,
1649 int *errcode)
1da177e4
LT		 1650{
1651 struct sk_buff *skb;
7d877f3b 1652 gfp_t gfp_mask;
1da177e4
LT		 1653 long timeo;
1654 int err;
cc9b17ad
JW		 1655 int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1656
1657 err = -EMSGSIZE;
1658 if (npages > MAX_SKB_FRAGS)
1659 goto failure;
1da177e4
LT		 1660
1661 gfp_mask = sk->sk_allocation;
1662 if (gfp_mask & __GFP_WAIT)
1663 gfp_mask |= __GFP_REPEAT;
1664
1665 timeo = sock_sndtimeo(sk, noblock);
1666 while (1) {
1667 err = sock_error(sk);
1668 if (err != 0)
1669 goto failure;
1670
1671 err = -EPIPE;
1672 if (sk->sk_shutdown & SEND_SHUTDOWN)
1673 goto failure;
1674
1675 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1676 skb = alloc_skb(header_len, gfp_mask);
1da177e4 1677 if (skb) {
1da177e4
LT		 1678 int i;
1679
1680 /* No pages, we're done... */
1681 if (!data_len)
1682 break;
1683
1da177e4
LT		 1684 skb->truesize += data_len;
1685 skb_shinfo(skb)->nr_frags = npages;
1686 for (i = 0; i < npages; i++) {
1687 struct page *page;
1da177e4
LT		 1688
1689 page = alloc_pages(sk->sk_allocation, 0);
1690 if (!page) {
1691 err = -ENOBUFS;
1692 skb_shinfo(skb)->nr_frags = i;
1693 kfree_skb(skb);
1694 goto failure;
1695 }
1696
ea2ab693
IC		 1697 __skb_fill_page_desc(skb, i,
1698 page, 0,
1699 (data_len >= PAGE_SIZE ?
1700 PAGE_SIZE :
1701 data_len));
1da177e4
LT		 1702 data_len -= PAGE_SIZE;
1703 }
1704
1705 /* Full success... */
1706 break;
1707 }
1708 err = -ENOBUFS;
1709 goto failure;
1710 }
1711 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1712 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1713 err = -EAGAIN;
1714 if (!timeo)
1715 goto failure;
1716 if (signal_pending(current))
1717 goto interrupted;
1718 timeo = sock_wait_for_wmem(sk, timeo);
1719 }
1720
1721 skb_set_owner_w(skb, sk);
1722 return skb;
1723
1724interrupted:
1725 err = sock_intr_errno(timeo);
1726failure:
1727 *errcode = err;
1728 return NULL;
1729}
4cc7f68d 1730EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1731
4ec93edb 1732struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1733 int noblock, int *errcode)
1734{
1735 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1736}
2a91525c 1737EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4
LT		 1738
1739static void __lock_sock(struct sock *sk)
f39234d6
NK		 1740 __releases(&sk->sk_lock.slock)
1741 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1742{
1743 DEFINE_WAIT(wait);
1744
e71a4783 1745 for (;;) {
1da177e4
LT		 1746 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1747 TASK_UNINTERRUPTIBLE);
1748 spin_unlock_bh(&sk->sk_lock.slock);
1749 schedule();
1750 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1751 if (!sock_owned_by_user(sk))
1da177e4
LT		 1752 break;
1753 }
1754 finish_wait(&sk->sk_lock.wq, &wait);
1755}
1756
1757static void __release_sock(struct sock *sk)
f39234d6
NK		 1758 __releases(&sk->sk_lock.slock)
1759 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1760{
1761 struct sk_buff *skb = sk->sk_backlog.head;
1762
1763 do {
1764 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1765 bh_unlock_sock(sk);
1766
1767 do {
1768 struct sk_buff *next = skb->next;
1769
e4cbb02a 1770 prefetch(next);
7fee226a 1771 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1772 skb->next = NULL;
c57943a1 1773 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1774
1775 /*
1776 * We are in process context here with softirqs
1777 * disabled, use cond_resched_softirq() to preempt.
1778 * This is safe to do because we've taken the backlog
1779 * queue private:
1780 */
1781 cond_resched_softirq();
1782
1783 skb = next;
1784 } while (skb != NULL);
1785
1786 bh_lock_sock(sk);
e71a4783 1787 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1788
1789 /*
1790 * Doing the zeroing here guarantee we can not loop forever
1791 * while a wild producer attempts to flood us.
1792 */
1793 sk->sk_backlog.len = 0;
1da177e4
LT		 1794}
1795
1796/**
1797 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1798 * @sk: sock to wait on
1799 * @timeo: for how long
1da177e4
LT		 1800 *
1801 * Now socket state including sk->sk_err is changed only under lock,
1802 * hence we may omit checks after joining wait queue.
1803 * We check receive queue before schedule() only as optimization;
1804 * it is very likely that release_sock() added new data.
1805 */
1806int sk_wait_data(struct sock *sk, long *timeo)
1807{
1808 int rc;
1809 DEFINE_WAIT(wait);
1810
aa395145 1811 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1812 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1813 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1814 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1815 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1816 return rc;
1817}
1da177e4
LT		 1818EXPORT_SYMBOL(sk_wait_data);
1819
3ab224be
HA		 1820/**
1821 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1822 * @sk: socket
1823 * @size: memory size to allocate
1824 * @kind: allocation type
1825 *
1826 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1827 * rmem allocation. This function assumes that protocols which have
1828 * memory_pressure use sk_wmem_queued as write buffer accounting.
1829 */
1830int __sk_mem_schedule(struct sock *sk, int size, int kind)
1831{
1832 struct proto *prot = sk->sk_prot;
1833 int amt = sk_mem_pages(size);
8d987e5c 1834 long allocated;
e1aab161 1835 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1836
1837 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1838
e1aab161 1839 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1840
1841 /* Under limit. */
e1aab161
GC		 1842 if (parent_status == UNDER_LIMIT &&
1843 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1844 sk_leave_memory_pressure(sk);
3ab224be
HA		 1845 return 1;
1846 }
1847
e1aab161
GC		 1848 /* Under pressure. (we or our parents) */
1849 if ((parent_status > SOFT_LIMIT) ||
1850 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1851 sk_enter_memory_pressure(sk);
3ab224be 1852
e1aab161
GC		 1853 /* Over hard limit (we or our parents) */
1854 if ((parent_status == OVER_LIMIT) ||
1855 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1856 goto suppress_allocation;
1857
1858 /* guarantee minimum buffer size under pressure */
1859 if (kind == SK_MEM_RECV) {
1860 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1861 return 1;
180d8cd9 1862
3ab224be
HA		 1863 } else { /* SK_MEM_SEND */
1864 if (sk->sk_type == SOCK_STREAM) {
1865 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1866 return 1;
1867 } else if (atomic_read(&sk->sk_wmem_alloc) <
1868 prot->sysctl_wmem[0])
1869 return 1;
1870 }
1871
180d8cd9 1872 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 1873 int alloc;
1874
180d8cd9 1875 if (!sk_under_memory_pressure(sk))
1748376b 1876 return 1;
180d8cd9
GC		 1877 alloc = sk_sockets_allocated_read_positive(sk);
1878 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 1879 sk_mem_pages(sk->sk_wmem_queued +
1880 atomic_read(&sk->sk_rmem_alloc) +
1881 sk->sk_forward_alloc))
1882 return 1;
1883 }
1884
1885suppress_allocation:
1886
1887 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1888 sk_stream_moderate_sndbuf(sk);
1889
1890 /* Fail only if socket is _under_ its sndbuf.
1891 * In this case we cannot block, so that we have to fail.
1892 */
1893 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1894 return 1;
1895 }
1896
3847ce32
SM		 1897 trace_sock_exceed_buf_limit(sk, prot, allocated);
1898
3ab224be
HA		 1899 /* Alas. Undo changes. */
1900 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 1901
0e90b31f 1902 sk_memory_allocated_sub(sk, amt);
180d8cd9 1903
3ab224be
HA		 1904 return 0;
1905}
3ab224be
HA		 1906EXPORT_SYMBOL(__sk_mem_schedule);
1907
1908/**
1909 * __sk_reclaim - reclaim memory_allocated
1910 * @sk: socket
1911 */
1912void __sk_mem_reclaim(struct sock *sk)
1913{
180d8cd9 1914 sk_memory_allocated_sub(sk,
0e90b31f 1915 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 1916 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
1917
180d8cd9
GC		 1918 if (sk_under_memory_pressure(sk) &&
1919 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
1920 sk_leave_memory_pressure(sk);
3ab224be 1921}
3ab224be
HA		 1922EXPORT_SYMBOL(__sk_mem_reclaim);
1923
1924
1da177e4
LT		 1925/*
1926 * Set of default routines for initialising struct proto_ops when
1927 * the protocol does not support a particular function. In certain
1928 * cases where it makes no sense for a protocol to have a "do nothing"
1929 * function, some default processing is provided.
1930 */
1931
1932int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1933{
1934 return -EOPNOTSUPP;
1935}
2a91525c 1936EXPORT_SYMBOL(sock_no_bind);
1da177e4 1937
4ec93edb 1938int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1939 int len, int flags)
1940{
1941 return -EOPNOTSUPP;
1942}
2a91525c 1943EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 1944
1945int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1946{
1947 return -EOPNOTSUPP;
1948}
2a91525c 1949EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 1950
1951int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1952{
1953 return -EOPNOTSUPP;
1954}
2a91525c 1955EXPORT_SYMBOL(sock_no_accept);
1da177e4 1956
4ec93edb 1957int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1958 int *len, int peer)
1959{
1960 return -EOPNOTSUPP;
1961}
2a91525c 1962EXPORT_SYMBOL(sock_no_getname);
1da177e4 1963
2a91525c 1964unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 1965{
1966 return 0;
1967}
2a91525c 1968EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 1969
1970int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1971{
1972 return -EOPNOTSUPP;
1973}
2a91525c 1974EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 1975
1976int sock_no_listen(struct socket *sock, int backlog)
1977{
1978 return -EOPNOTSUPP;
1979}
2a91525c 1980EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 1981
1982int sock_no_shutdown(struct socket *sock, int how)
1983{
1984 return -EOPNOTSUPP;
1985}
2a91525c 1986EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 1987
1988int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 1989 char __user *optval, unsigned int optlen)
1da177e4
LT		 1990{
1991 return -EOPNOTSUPP;
1992}
2a91525c 1993EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 1994
1995int sock_no_getsockopt(struct socket *sock, int level, int optname,
1996 char __user *optval, int __user *optlen)
1997{
1998 return -EOPNOTSUPP;
1999}
2a91525c 2000EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2001
2002int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2003 size_t len)
2004{
2005 return -EOPNOTSUPP;
2006}
2a91525c 2007EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2008
2009int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2010 size_t len, int flags)
2011{
2012 return -EOPNOTSUPP;
2013}
2a91525c 2014EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2015
2016int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2017{
2018 /* Mirror missing mmap method error code */
2019 return -ENODEV;
2020}
2a91525c 2021EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2022
2023ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2024{
2025 ssize_t res;
2026 struct msghdr msg = {.msg_flags = flags};
2027 struct kvec iov;
2028 char *kaddr = kmap(page);
2029 iov.iov_base = kaddr + offset;
2030 iov.iov_len = size;
2031 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2032 kunmap(page);
2033 return res;
2034}
2a91525c 2035EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2036
2037/*
2038 * Default Socket Callbacks
2039 */
2040
2041static void sock_def_wakeup(struct sock *sk)
2042{
43815482
ED		 2043 struct socket_wq *wq;
2044
2045 rcu_read_lock();
2046 wq = rcu_dereference(sk->sk_wq);
2047 if (wq_has_sleeper(wq))
2048 wake_up_interruptible_all(&wq->wait);
2049 rcu_read_unlock();
1da177e4
LT		 2050}
2051
2052static void sock_def_error_report(struct sock *sk)
2053{
43815482
ED		 2054 struct socket_wq *wq;
2055
2056 rcu_read_lock();
2057 wq = rcu_dereference(sk->sk_wq);
2058 if (wq_has_sleeper(wq))
2059 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2060 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2061 rcu_read_unlock();
1da177e4
LT		 2062}
2063
2064static void sock_def_readable(struct sock *sk, int len)
2065{
43815482
ED		 2066 struct socket_wq *wq;
2067
2068 rcu_read_lock();
2069 wq = rcu_dereference(sk->sk_wq);
2070 if (wq_has_sleeper(wq))
2c6607c6 2071 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2072 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2073 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2074 rcu_read_unlock();
1da177e4
LT		 2075}
2076
2077static void sock_def_write_space(struct sock *sk)
2078{
43815482
ED		 2079 struct socket_wq *wq;
2080
2081 rcu_read_lock();
1da177e4
LT		 2082
2083 /* Do not wake up a writer until he can make "significant"
2084 * progress. --DaveM
2085 */
e71a4783 2086 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2087 wq = rcu_dereference(sk->sk_wq);
2088 if (wq_has_sleeper(wq))
2089 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2090 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2091
2092 /* Should agree with poll, otherwise some programs break */
2093 if (sock_writeable(sk))
8d8ad9d7 2094 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2095 }
2096
43815482 2097 rcu_read_unlock();
1da177e4
LT		 2098}
2099
2100static void sock_def_destruct(struct sock *sk)
2101{
a51482bd 2102 kfree(sk->sk_protinfo);
1da177e4
LT		 2103}
2104
2105void sk_send_sigurg(struct sock *sk)
2106{
2107 if (sk->sk_socket && sk->sk_socket->file)
2108 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2109 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2110}
2a91525c 2111EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2112
2113void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2114 unsigned long expires)
2115{
2116 if (!mod_timer(timer, expires))
2117 sock_hold(sk);
2118}
1da177e4
LT		 2119EXPORT_SYMBOL(sk_reset_timer);
2120
2121void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2122{
2123 if (timer_pending(timer) && del_timer(timer))
2124 __sock_put(sk);
2125}
1da177e4
LT		 2126EXPORT_SYMBOL(sk_stop_timer);
2127
2128void sock_init_data(struct socket *sock, struct sock *sk)
2129{
2130 skb_queue_head_init(&sk->sk_receive_queue);
2131 skb_queue_head_init(&sk->sk_write_queue);
2132 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2133#ifdef CONFIG_NET_DMA
2134 skb_queue_head_init(&sk->sk_async_wait_queue);
2135#endif
1da177e4
LT		 2136
2137 sk->sk_send_head = NULL;
2138
2139 init_timer(&sk->sk_timer);
4ec93edb 2140
1da177e4
LT		 2141 sk->sk_allocation = GFP_KERNEL;
2142 sk->sk_rcvbuf = sysctl_rmem_default;
2143 sk->sk_sndbuf = sysctl_wmem_default;
2144 sk->sk_state = TCP_CLOSE;
972692e0 2145 sk_set_socket(sk, sock);
1da177e4
LT		 2146
2147 sock_set_flag(sk, SOCK_ZAPPED);
2148
e71a4783 2149 if (sock) {
1da177e4 2150 sk->sk_type = sock->type;
43815482 2151 sk->sk_wq = sock->wq;
1da177e4
LT		 2152 sock->sk = sk;
2153 } else
43815482 2154 sk->sk_wq = NULL;
1da177e4 2155
b6c6712a 2156 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2157 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2158 lockdep_set_class_and_name(&sk->sk_callback_lock,
2159 af_callback_keys + sk->sk_family,
2160 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2161
2162 sk->sk_state_change = sock_def_wakeup;
2163 sk->sk_data_ready = sock_def_readable;
2164 sk->sk_write_space = sock_def_write_space;
2165 sk->sk_error_report = sock_def_error_report;
2166 sk->sk_destruct = sock_def_destruct;
2167
2168 sk->sk_sndmsg_page = NULL;
2169 sk->sk_sndmsg_off = 0;
ef64a54f 2170 sk->sk_peek_off = -1;
1da177e4 2171
109f6e39
EB		 2172 sk->sk_peer_pid = NULL;
2173 sk->sk_peer_cred = NULL;
1da177e4
LT		 2174 sk->sk_write_pending = 0;
2175 sk->sk_rcvlowat = 1;
2176 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2177 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2178
f37f0afb 2179 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2180
4dc6dc71
ED		 2181 /*
2182 * Before updating sk_refcnt, we must commit prior changes to memory
2183 * (Documentation/RCU/rculist_nulls.txt for details)
2184 */
2185 smp_wmb();
1da177e4 2186 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2187 atomic_set(&sk->sk_drops, 0);
1da177e4 2188}
2a91525c 2189EXPORT_SYMBOL(sock_init_data);
1da177e4 2190
b5606c2d 2191void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2192{
2193 might_sleep();
a5b5bb9a 2194 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2195 if (sk->sk_lock.owned)
1da177e4 2196 __lock_sock(sk);
d2e9117c 2197 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2198 spin_unlock(&sk->sk_lock.slock);
2199 /*
2200 * The sk_lock has mutex_lock() semantics here:
2201 */
fcc70d5f 2202 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2203 local_bh_enable();
1da177e4 2204}
fcc70d5f 2205EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2206
b5606c2d 2207void release_sock(struct sock *sk)
1da177e4 2208{
a5b5bb9a
IM		 2209 /*
2210 * The sk_lock has mutex_unlock() semantics:
2211 */
2212 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2213
2214 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2215 if (sk->sk_backlog.tail)
2216 __release_sock(sk);
46d3ceab
ED		 2217
2218 if (sk->sk_prot->release_cb)
2219 sk->sk_prot->release_cb(sk);
2220
d2e9117c 2221 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2222 if (waitqueue_active(&sk->sk_lock.wq))
2223 wake_up(&sk->sk_lock.wq);
2224 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2225}
2226EXPORT_SYMBOL(release_sock);
2227
8a74ad60
ED		 2228/**
2229 * lock_sock_fast - fast version of lock_sock
2230 * @sk: socket
2231 *
2232 * This version should be used for very small section, where process wont block
2233 * return false if fast path is taken
2234 * sk_lock.slock locked, owned = 0, BH disabled
2235 * return true if slow path is taken
2236 * sk_lock.slock unlocked, owned = 1, BH enabled
2237 */
2238bool lock_sock_fast(struct sock *sk)
2239{
2240 might_sleep();
2241 spin_lock_bh(&sk->sk_lock.slock);
2242
2243 if (!sk->sk_lock.owned)
2244 /*
2245 * Note : We must disable BH
2246 */
2247 return false;
2248
2249 __lock_sock(sk);
2250 sk->sk_lock.owned = 1;
2251 spin_unlock(&sk->sk_lock.slock);
2252 /*
2253 * The sk_lock has mutex_lock() semantics here:
2254 */
2255 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2256 local_bh_enable();
2257 return true;
2258}
2259EXPORT_SYMBOL(lock_sock_fast);
2260
1da177e4 2261int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2262{
b7aa0bf7 2263 struct timeval tv;
1da177e4 2264 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2265 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2266 tv = ktime_to_timeval(sk->sk_stamp);
2267 if (tv.tv_sec == -1)
1da177e4 2268 return -ENOENT;
b7aa0bf7
ED		 2269 if (tv.tv_sec == 0) {
2270 sk->sk_stamp = ktime_get_real();
2271 tv = ktime_to_timeval(sk->sk_stamp);
2272 }
2273 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2274}
1da177e4
LT		 2275EXPORT_SYMBOL(sock_get_timestamp);
2276
ae40eb1e
ED		 2277int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2278{
2279 struct timespec ts;
2280 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2281 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2282 ts = ktime_to_timespec(sk->sk_stamp);
2283 if (ts.tv_sec == -1)
2284 return -ENOENT;
2285 if (ts.tv_sec == 0) {
2286 sk->sk_stamp = ktime_get_real();
2287 ts = ktime_to_timespec(sk->sk_stamp);
2288 }
2289 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2290}
2291EXPORT_SYMBOL(sock_get_timestampns);
2292
20d49473 2293void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2294{
20d49473 2295 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2296 unsigned long previous_flags = sk->sk_flags;
2297
20d49473
PO		 2298 sock_set_flag(sk, flag);
2299 /*
2300 * we just set one of the two flags which require net
2301 * time stamping, but time stamping might have been on
2302 * already because of the other one
2303 */
08e29af3 2304 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2305 net_enable_timestamp();
1da177e4
LT		 2306 }
2307}
1da177e4
LT		 2308
2309/*
2310 * Get a socket option on an socket.
2311 *
2312 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2313 * asynchronous errors should be reported by getsockopt. We assume
2314 * this means if you specify SO_ERROR (otherwise whats the point of it).
2315 */
2316int sock_common_getsockopt(struct socket *sock, int level, int optname,
2317 char __user *optval, int __user *optlen)
2318{
2319 struct sock *sk = sock->sk;
2320
2321 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2322}
1da177e4
LT		 2323EXPORT_SYMBOL(sock_common_getsockopt);
2324
3fdadf7d 2325#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2326int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2327 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2328{
2329 struct sock *sk = sock->sk;
2330
1e51f951 2331 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2332 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2333 optval, optlen);
3fdadf7d
DM		 2334 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2335}
2336EXPORT_SYMBOL(compat_sock_common_getsockopt);
2337#endif
2338
1da177e4
LT		 2339int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2340 struct msghdr *msg, size_t size, int flags)
2341{
2342 struct sock *sk = sock->sk;
2343 int addr_len = 0;
2344 int err;
2345
2346 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2347 flags & ~MSG_DONTWAIT, &addr_len);
2348 if (err >= 0)
2349 msg->msg_namelen = addr_len;
2350 return err;
2351}
1da177e4
LT		 2352EXPORT_SYMBOL(sock_common_recvmsg);
2353
2354/*
2355 * Set socket options on an inet socket.
2356 */
2357int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2358 char __user *optval, unsigned int optlen)
1da177e4
LT		 2359{
2360 struct sock *sk = sock->sk;
2361
2362 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2363}
1da177e4
LT		 2364EXPORT_SYMBOL(sock_common_setsockopt);
2365
3fdadf7d 2366#ifdef CONFIG_COMPAT
543d9cfe 2367int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2368 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2369{
2370 struct sock *sk = sock->sk;
2371
543d9cfe
ACM		 2372 if (sk->sk_prot->compat_setsockopt != NULL)
2373 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2374 optval, optlen);
3fdadf7d
DM		 2375 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2376}
2377EXPORT_SYMBOL(compat_sock_common_setsockopt);
2378#endif
2379
1da177e4
LT		 2380void sk_common_release(struct sock *sk)
2381{
2382 if (sk->sk_prot->destroy)
2383 sk->sk_prot->destroy(sk);
2384
2385 /*
2386 * Observation: when sock_common_release is called, processes have
2387 * no access to socket. But net still has.
2388 * Step one, detach it from networking:
2389 *
2390 * A. Remove from hash tables.
2391 */
2392
2393 sk->sk_prot->unhash(sk);
2394
2395 /*
2396 * In this point socket cannot receive new packets, but it is possible
2397 * that some packets are in flight because some CPU runs receiver and
2398 * did hash table lookup before we unhashed socket. They will achieve
2399 * receive queue and will be purged by socket destructor.
2400 *
2401 * Also we still have packets pending on receive queue and probably,
2402 * our own packets waiting in device queues. sock_destroy will drain
2403 * receive queue, but transmitted packets will delay socket destruction
2404 * until the last reference will be released.
2405 */
2406
2407 sock_orphan(sk);
2408
2409 xfrm_sk_free_policy(sk);
2410
e6848976 2411 sk_refcnt_debug_release(sk);
1da177e4
LT		 2412 sock_put(sk);
2413}
1da177e4
LT		 2414EXPORT_SYMBOL(sk_common_release);
2415
13ff3d6f
PE		 2416#ifdef CONFIG_PROC_FS
2417#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2418struct prot_inuse {
2419 int val[PROTO_INUSE_NR];
2420};
13ff3d6f
PE		 2421
2422static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2423
2424#ifdef CONFIG_NET_NS
2425void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2426{
d6d9ca0f 2427 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2428}
2429EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2430
2431int sock_prot_inuse_get(struct net *net, struct proto *prot)
2432{
2433 int cpu, idx = prot->inuse_idx;
2434 int res = 0;
2435
2436 for_each_possible_cpu(cpu)
2437 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2438
2439 return res >= 0 ? res : 0;
2440}
2441EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2442
2c8c1e72 2443static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2444{
2445 net->core.inuse = alloc_percpu(struct prot_inuse);
2446 return net->core.inuse ? 0 : -ENOMEM;
2447}
2448
2c8c1e72 2449static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2450{
2451 free_percpu(net->core.inuse);
2452}
2453
2454static struct pernet_operations net_inuse_ops = {
2455 .init = sock_inuse_init_net,
2456 .exit = sock_inuse_exit_net,
2457};
2458
2459static __init int net_inuse_init(void)
2460{
2461 if (register_pernet_subsys(&net_inuse_ops))
2462 panic("Cannot initialize net inuse counters");
2463
2464 return 0;
2465}
2466
2467core_initcall(net_inuse_init);
2468#else
1338d466
PE		 2469static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2470
c29a0bc4 2471void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2472{
d6d9ca0f 2473 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2474}
2475EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2476
c29a0bc4 2477int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2478{
2479 int cpu, idx = prot->inuse_idx;
2480 int res = 0;
2481
2482 for_each_possible_cpu(cpu)
2483 res += per_cpu(prot_inuse, cpu).val[idx];
2484
2485 return res >= 0 ? res : 0;
2486}
2487EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2488#endif
13ff3d6f
PE		 2489
2490static void assign_proto_idx(struct proto *prot)
2491{
2492 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2493
2494 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2495 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2496 return;
2497 }
2498
2499 set_bit(prot->inuse_idx, proto_inuse_idx);
2500}
2501
2502static void release_proto_idx(struct proto *prot)
2503{
2504 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2505 clear_bit(prot->inuse_idx, proto_inuse_idx);
2506}
2507#else
2508static inline void assign_proto_idx(struct proto *prot)
2509{
2510}
2511
2512static inline void release_proto_idx(struct proto *prot)
2513{
2514}
2515#endif
2516
b733c007
PE		 2517int proto_register(struct proto *prot, int alloc_slab)
2518{
1da177e4
LT		 2519 if (alloc_slab) {
2520 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2521 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2522 NULL);
1da177e4
LT		 2523
2524 if (prot->slab == NULL) {
e005d193
JP		 2525 pr_crit("%s: Can't create sock SLAB cache!\n",
2526 prot->name);
60e7663d 2527 goto out;
1da177e4 2528 }
2e6599cb
ACM		 2529
2530 if (prot->rsk_prot != NULL) {
faf23422 2531 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2532 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2533 goto out_free_sock_slab;
2534
7e56b5d6 2535 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2536 prot->rsk_prot->obj_size, 0,
20c2df83 2537 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2538
2539 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2540 pr_crit("%s: Can't create request sock SLAB cache!\n",
2541 prot->name);
2e6599cb
ACM		 2542 goto out_free_request_sock_slab_name;
2543 }
2544 }
8feaf0c0 2545
6d6ee43e 2546 if (prot->twsk_prot != NULL) {
faf23422 2547 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2548
7e56b5d6 2549 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2550 goto out_free_request_sock_slab;
2551
6d6ee43e 2552 prot->twsk_prot->twsk_slab =
7e56b5d6 2553 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2554 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2555 0,
2556 SLAB_HWCACHE_ALIGN |
2557 prot->slab_flags,
20c2df83 2558 NULL);
6d6ee43e 2559 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2560 goto out_free_timewait_sock_slab_name;
2561 }
1da177e4
LT		 2562 }
2563
36b77a52 2564 mutex_lock(&proto_list_mutex);
1da177e4 2565 list_add(&prot->node, &proto_list);
13ff3d6f 2566 assign_proto_idx(prot);
36b77a52 2567 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2568 return 0;
2569
8feaf0c0 2570out_free_timewait_sock_slab_name:
7e56b5d6 2571 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2572out_free_request_sock_slab:
2573 if (prot->rsk_prot && prot->rsk_prot->slab) {
2574 kmem_cache_destroy(prot->rsk_prot->slab);
2575 prot->rsk_prot->slab = NULL;
2576 }
2e6599cb 2577out_free_request_sock_slab_name:
72150e9b
DC		 2578 if (prot->rsk_prot)
2579 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2580out_free_sock_slab:
2581 kmem_cache_destroy(prot->slab);
2582 prot->slab = NULL;
b733c007
PE		 2583out:
2584 return -ENOBUFS;
1da177e4 2585}
1da177e4
LT		 2586EXPORT_SYMBOL(proto_register);
2587
2588void proto_unregister(struct proto *prot)
2589{
36b77a52 2590 mutex_lock(&proto_list_mutex);
13ff3d6f 2591 release_proto_idx(prot);
0a3f4358 2592 list_del(&prot->node);
36b77a52 2593 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2594
2595 if (prot->slab != NULL) {
2596 kmem_cache_destroy(prot->slab);
2597 prot->slab = NULL;
2598 }
2599
2e6599cb 2600 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2601 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2602 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2603 prot->rsk_prot->slab = NULL;
2604 }
2605
6d6ee43e 2606 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2607 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2608 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2609 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2610 }
1da177e4 2611}
1da177e4
LT		 2612EXPORT_SYMBOL(proto_unregister);
2613
2614#ifdef CONFIG_PROC_FS
1da177e4 2615static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2616 __acquires(proto_list_mutex)
1da177e4 2617{
36b77a52 2618 mutex_lock(&proto_list_mutex);
60f0438a 2619 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2620}
2621
2622static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2623{
60f0438a 2624 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2625}
2626
2627static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2628 __releases(proto_list_mutex)
1da177e4 2629{
36b77a52 2630 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2631}
2632
2633static char proto_method_implemented(const void *method)
2634{
2635 return method == NULL ? 'n' : 'y';
2636}
180d8cd9
GC		 2637static long sock_prot_memory_allocated(struct proto *proto)
2638{
cb75a36c 2639 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2640}
2641
2642static char *sock_prot_memory_pressure(struct proto *proto)
2643{
2644 return proto->memory_pressure != NULL ?
2645 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2646}
1da177e4
LT		 2647
2648static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2649{
180d8cd9 2650
8d987e5c 2651 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2652 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2653 proto->name,
2654 proto->obj_size,
14e943db 2655 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2656 sock_prot_memory_allocated(proto),
2657 sock_prot_memory_pressure(proto),
1da177e4
LT		 2658 proto->max_header,
2659 proto->slab == NULL ? "no" : "yes",
2660 module_name(proto->owner),
2661 proto_method_implemented(proto->close),
2662 proto_method_implemented(proto->connect),
2663 proto_method_implemented(proto->disconnect),
2664 proto_method_implemented(proto->accept),
2665 proto_method_implemented(proto->ioctl),
2666 proto_method_implemented(proto->init),
2667 proto_method_implemented(proto->destroy),
2668 proto_method_implemented(proto->shutdown),
2669 proto_method_implemented(proto->setsockopt),
2670 proto_method_implemented(proto->getsockopt),
2671 proto_method_implemented(proto->sendmsg),
2672 proto_method_implemented(proto->recvmsg),
2673 proto_method_implemented(proto->sendpage),
2674 proto_method_implemented(proto->bind),
2675 proto_method_implemented(proto->backlog_rcv),
2676 proto_method_implemented(proto->hash),
2677 proto_method_implemented(proto->unhash),
2678 proto_method_implemented(proto->get_port),
2679 proto_method_implemented(proto->enter_memory_pressure));
2680}
2681
2682static int proto_seq_show(struct seq_file *seq, void *v)
2683{
60f0438a 2684 if (v == &proto_list)
1da177e4
LT		 2685 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2686 "protocol",
2687 "size",
2688 "sockets",
2689 "memory",
2690 "press",
2691 "maxhdr",
2692 "slab",
2693 "module",
2694 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2695 else
60f0438a 2696 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2697 return 0;
2698}
2699
f690808e 2700static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2701 .start = proto_seq_start,
2702 .next = proto_seq_next,
2703 .stop = proto_seq_stop,
2704 .show = proto_seq_show,
2705};
2706
2707static int proto_seq_open(struct inode *inode, struct file *file)
2708{
14e943db
ED		 2709 return seq_open_net(inode, file, &proto_seq_ops,
2710 sizeof(struct seq_net_private));
1da177e4
LT		 2711}
2712
9a32144e 2713static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2714 .owner = THIS_MODULE,
2715 .open = proto_seq_open,
2716 .read = seq_read,
2717 .llseek = seq_lseek,
14e943db
ED		 2718 .release = seq_release_net,
2719};
2720
2721static __net_init int proto_init_net(struct net *net)
2722{
2723 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2724 return -ENOMEM;
2725
2726 return 0;
2727}
2728
2729static __net_exit void proto_exit_net(struct net *net)
2730{
2731 proc_net_remove(net, "protocols");
2732}
2733
2734
2735static __net_initdata struct pernet_operations proto_net_ops = {
2736 .init = proto_init_net,
2737 .exit = proto_exit_net,
1da177e4
LT		 2738};
2739
2740static int __init proto_init(void)
2741{
14e943db 2742 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2743}
2744
2745subsys_initcall(proto_init);
2746
2747#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 1.38852 seconds and 5 git commands to generate.