projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
netpoll: fix a missing dev refcounting
[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
1da177e4
LT		 329static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
330{
331 struct timeval tv;
332
333 if (optlen < sizeof(tv))
334 return -EINVAL;
335 if (copy_from_user(&tv, optval, sizeof(tv)))
336 return -EFAULT;
ba78073e
VA		 337 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
338 return -EDOM;
1da177e4 339
ba78073e 340 if (tv.tv_sec < 0) {
6f11df83
AM		 341 static int warned __read_mostly;
342
ba78073e 343 *timeo_p = 0;
50aab54f 344 if (warned < 10 && net_ratelimit()) {
ba78073e 345 warned++;
e005d193
JP		 346 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
347 __func__, current->comm, task_pid_nr(current));
50aab54f 348 }
ba78073e
VA		 349 return 0;
350 }
1da177e4
LT		 351 *timeo_p = MAX_SCHEDULE_TIMEOUT;
352 if (tv.tv_sec == 0 && tv.tv_usec == 0)
353 return 0;
354 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
355 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
356 return 0;
357}
358
359static void sock_warn_obsolete_bsdism(const char *name)
360{
361 static int warned;
362 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 363 if (strcmp(warncomm, current->comm) && warned < 5) {
364 strcpy(warncomm, current->comm);
e005d193
JP		 365 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
366 warncomm, name);
1da177e4
LT		 367 warned++;
368 }
369}
370
08e29af3
ED		 371#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
372
373static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 374{
08e29af3
ED		 375 if (sk->sk_flags & flags) {
376 sk->sk_flags &= ~flags;
377 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 378 net_disable_timestamp();
1da177e4
LT		 379 }
380}
381
382
f0088a50
DV		 383int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
384{
766e9037 385 int err;
f0088a50 386 int skb_len;
3b885787
NH		 387 unsigned long flags;
388 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 389
0fd7bac6 390 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 391 atomic_inc(&sk->sk_drops);
3847ce32 392 trace_sock_rcvqueue_full(sk, skb);
766e9037 393 return -ENOMEM;
f0088a50
DV		 394 }
395
fda9ef5d 396 err = sk_filter(sk, skb);
f0088a50 397 if (err)
766e9037 398 return err;
f0088a50 399
c76562b6 400 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 401 atomic_inc(&sk->sk_drops);
402 return -ENOBUFS;
3ab224be
HA		 403 }
404
f0088a50
DV		 405 skb->dev = NULL;
406 skb_set_owner_r(skb, sk);
49ad9599 407
f0088a50
DV		 408 /* Cache the SKB length before we tack it onto the receive
409 * queue. Once it is added it no longer belongs to us and
410 * may be freed by other threads of control pulling packets
411 * from the queue.
412 */
413 skb_len = skb->len;
414
7fee226a
ED		 415 /* we escape from rcu protected region, make sure we dont leak
416 * a norefcounted dst
417 */
418 skb_dst_force(skb);
419
3b885787
NH		 420 spin_lock_irqsave(&list->lock, flags);
421 skb->dropcount = atomic_read(&sk->sk_drops);
422 __skb_queue_tail(list, skb);
423 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 424
425 if (!sock_flag(sk, SOCK_DEAD))
426 sk->sk_data_ready(sk, skb_len);
766e9037 427 return 0;
f0088a50
DV		 428}
429EXPORT_SYMBOL(sock_queue_rcv_skb);
430
58a5a7b9 431int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 432{
433 int rc = NET_RX_SUCCESS;
434
fda9ef5d 435 if (sk_filter(sk, skb))
f0088a50
DV		 436 goto discard_and_relse;
437
438 skb->dev = NULL;
439
f545a38f 440 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) {
c377411f
ED		 441 atomic_inc(&sk->sk_drops);
442 goto discard_and_relse;
443 }
58a5a7b9
ACM		 444 if (nested)
445 bh_lock_sock_nested(sk);
446 else
447 bh_lock_sock(sk);
a5b5bb9a
IM		 448 if (!sock_owned_by_user(sk)) {
449 /*
450 * trylock + unlock semantics:
451 */
452 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
453
c57943a1 454 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 455
456 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 457 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 458 bh_unlock_sock(sk);
459 atomic_inc(&sk->sk_drops);
460 goto discard_and_relse;
461 }
462
f0088a50
DV		 463 bh_unlock_sock(sk);
464out:
465 sock_put(sk);
466 return rc;
467discard_and_relse:
468 kfree_skb(skb);
469 goto out;
470}
471EXPORT_SYMBOL(sk_receive_skb);
472
ea94ff3b
KK		 473void sk_reset_txq(struct sock *sk)
474{
475 sk_tx_queue_clear(sk);
476}
477EXPORT_SYMBOL(sk_reset_txq);
478
f0088a50
DV		 479struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
480{
b6c6712a 481 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 482
483 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 484 sk_tx_queue_clear(sk);
a9b3cd7f 485 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 486 dst_release(dst);
487 return NULL;
488 }
489
490 return dst;
491}
492EXPORT_SYMBOL(__sk_dst_check);
493
494struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
495{
496 struct dst_entry *dst = sk_dst_get(sk);
497
498 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
499 sk_dst_reset(sk);
500 dst_release(dst);
501 return NULL;
502 }
503
504 return dst;
505}
506EXPORT_SYMBOL(sk_dst_check);
507
c91f6df2
BH		 508static int sock_setbindtodevice(struct sock *sk, char __user *optval,
509 int optlen)
4878809f
DM		 510{
511 int ret = -ENOPROTOOPT;
512#ifdef CONFIG_NETDEVICES
3b1e0a65 513 struct net *net = sock_net(sk);
4878809f
DM		 514 char devname[IFNAMSIZ];
515 int index;
516
517 /* Sorry... */
518 ret = -EPERM;
5e1fccc0 519 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 520 goto out;
521
522 ret = -EINVAL;
523 if (optlen < 0)
524 goto out;
525
526 /* Bind this socket to a particular device like "eth0",
527 * as specified in the passed interface name. If the
528 * name is "" or the option length is zero the socket
529 * is not bound.
530 */
531 if (optlen > IFNAMSIZ - 1)
532 optlen = IFNAMSIZ - 1;
533 memset(devname, 0, sizeof(devname));
534
535 ret = -EFAULT;
536 if (copy_from_user(devname, optval, optlen))
537 goto out;
538
000ba2e4
DM		 539 index = 0;
540 if (devname[0] != '\0') {
bf8e56bf 541 struct net_device *dev;
4878809f 542
bf8e56bf
ED		 543 rcu_read_lock();
544 dev = dev_get_by_name_rcu(net, devname);
545 if (dev)
546 index = dev->ifindex;
547 rcu_read_unlock();
4878809f
DM		 548 ret = -ENODEV;
549 if (!dev)
550 goto out;
4878809f
DM		 551 }
552
553 lock_sock(sk);
554 sk->sk_bound_dev_if = index;
555 sk_dst_reset(sk);
556 release_sock(sk);
557
558 ret = 0;
559
560out:
561#endif
562
563 return ret;
564}
565
c91f6df2
BH		 566static int sock_getbindtodevice(struct sock *sk, char __user *optval,
567 int __user *optlen, int len)
568{
569 int ret = -ENOPROTOOPT;
570#ifdef CONFIG_NETDEVICES
571 struct net *net = sock_net(sk);
572 struct net_device *dev;
573 char devname[IFNAMSIZ];
574 unsigned seq;
575
576 if (sk->sk_bound_dev_if == 0) {
577 len = 0;
578 goto zero;
579 }
580
581 ret = -EINVAL;
582 if (len < IFNAMSIZ)
583 goto out;
584
585retry:
30e6c9fa 586 seq = read_seqcount_begin(&devnet_rename_seq);
c91f6df2
BH		 587 rcu_read_lock();
588 dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
589 ret = -ENODEV;
590 if (!dev) {
591 rcu_read_unlock();
592 goto out;
593 }
594
595 strcpy(devname, dev->name);
596 rcu_read_unlock();
30e6c9fa 597 if (read_seqcount_retry(&devnet_rename_seq, seq))
c91f6df2
BH		 598 goto retry;
599
600 len = strlen(devname) + 1;
601
602 ret = -EFAULT;
603 if (copy_to_user(optval, devname, len))
604 goto out;
605
606zero:
607 ret = -EFAULT;
608 if (put_user(len, optlen))
609 goto out;
610
611 ret = 0;
612
613out:
614#endif
615
616 return ret;
617}
618
c0ef877b
PE		 619static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
620{
621 if (valbool)
622 sock_set_flag(sk, bit);
623 else
624 sock_reset_flag(sk, bit);
625}
626
1da177e4
LT		 627/*
628 * This is meant for all protocols to use and covers goings on
629 * at the socket level. Everything here is generic.
630 */
631
632int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 633 char __user *optval, unsigned int optlen)
1da177e4 634{
2a91525c 635 struct sock *sk = sock->sk;
1da177e4
LT		 636 int val;
637 int valbool;
638 struct linger ling;
639 int ret = 0;
4ec93edb 640
1da177e4
LT		 641 /*
642 * Options without arguments
643 */
644
4878809f 645 if (optname == SO_BINDTODEVICE)
c91f6df2 646 return sock_setbindtodevice(sk, optval, optlen);
4878809f 647
e71a4783
SH		 648 if (optlen < sizeof(int))
649 return -EINVAL;
4ec93edb 650
1da177e4
LT		 651 if (get_user(val, (int __user *)optval))
652 return -EFAULT;
4ec93edb 653
2a91525c 654 valbool = val ? 1 : 0;
1da177e4
LT		 655
656 lock_sock(sk);
657
2a91525c 658 switch (optname) {
e71a4783 659 case SO_DEBUG:
2a91525c 660 if (val && !capable(CAP_NET_ADMIN))
e71a4783 661 ret = -EACCES;
2a91525c 662 else
c0ef877b 663 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 664 break;
665 case SO_REUSEADDR:
4a17fd52 666 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783
SH		 667 break;
668 case SO_TYPE:
49c794e9 669 case SO_PROTOCOL:
0d6038ee 670 case SO_DOMAIN:
e71a4783
SH		 671 case SO_ERROR:
672 ret = -ENOPROTOOPT;
673 break;
674 case SO_DONTROUTE:
c0ef877b 675 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 676 break;
677 case SO_BROADCAST:
678 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
679 break;
680 case SO_SNDBUF:
681 /* Don't error on this BSD doesn't and if you think
82981930
ED		 682 * about it this is right. Otherwise apps have to
683 * play 'guess the biggest size' games. RCVBUF/SNDBUF
684 * are treated in BSD as hints
685 */
686 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 687set_sndbuf:
e71a4783 688 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 689 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
690 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 691 sk->sk_write_space(sk);
692 break;
1da177e4 693
e71a4783
SH		 694 case SO_SNDBUFFORCE:
695 if (!capable(CAP_NET_ADMIN)) {
696 ret = -EPERM;
697 break;
698 }
699 goto set_sndbuf;
b0573dea 700
e71a4783
SH		 701 case SO_RCVBUF:
702 /* Don't error on this BSD doesn't and if you think
82981930
ED		 703 * about it this is right. Otherwise apps have to
704 * play 'guess the biggest size' games. RCVBUF/SNDBUF
705 * are treated in BSD as hints
706 */
707 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 708set_rcvbuf:
e71a4783
SH		 709 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
710 /*
711 * We double it on the way in to account for
712 * "struct sk_buff" etc. overhead. Applications
713 * assume that the SO_RCVBUF setting they make will
714 * allow that much actual data to be received on that
715 * socket.
716 *
717 * Applications are unaware that "struct sk_buff" and
718 * other overheads allocate from the receive buffer
719 * during socket buffer allocation.
720 *
721 * And after considering the possible alternatives,
722 * returning the value we actually used in getsockopt
723 * is the most desirable behavior.
724 */
82981930 725 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 726 break;
727
728 case SO_RCVBUFFORCE:
729 if (!capable(CAP_NET_ADMIN)) {
730 ret = -EPERM;
1da177e4 731 break;
e71a4783
SH		 732 }
733 goto set_rcvbuf;
1da177e4 734
e71a4783 735 case SO_KEEPALIVE:
1da177e4 736#ifdef CONFIG_INET
3e10986d
ED		 737 if (sk->sk_protocol == IPPROTO_TCP &&
738 sk->sk_type == SOCK_STREAM)
e71a4783 739 tcp_set_keepalive(sk, valbool);
1da177e4 740#endif
e71a4783
SH		 741 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
742 break;
743
744 case SO_OOBINLINE:
745 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
746 break;
747
748 case SO_NO_CHECK:
749 sk->sk_no_check = valbool;
750 break;
751
752 case SO_PRIORITY:
5e1fccc0
EB		 753 if ((val >= 0 && val <= 6) ||
754 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 755 sk->sk_priority = val;
756 else
757 ret = -EPERM;
758 break;
759
760 case SO_LINGER:
761 if (optlen < sizeof(ling)) {
762 ret = -EINVAL; /* 1003.1g */
1da177e4 763 break;
e71a4783 764 }
2a91525c 765 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 766 ret = -EFAULT;
1da177e4 767 break;
e71a4783
SH		 768 }
769 if (!ling.l_onoff)
770 sock_reset_flag(sk, SOCK_LINGER);
771 else {
1da177e4 772#if (BITS_PER_LONG == 32)
e71a4783
SH		 773 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
774 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 775 else
e71a4783
SH		 776#endif
777 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
778 sock_set_flag(sk, SOCK_LINGER);
779 }
780 break;
781
782 case SO_BSDCOMPAT:
783 sock_warn_obsolete_bsdism("setsockopt");
784 break;
785
786 case SO_PASSCRED:
787 if (valbool)
788 set_bit(SOCK_PASSCRED, &sock->flags);
789 else
790 clear_bit(SOCK_PASSCRED, &sock->flags);
791 break;
792
793 case SO_TIMESTAMP:
92f37fd2 794 case SO_TIMESTAMPNS:
e71a4783 795 if (valbool) {
92f37fd2
ED		 796 if (optname == SO_TIMESTAMP)
797 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
798 else
799 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 800 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 801 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 802 } else {
e71a4783 803 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 804 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
805 }
e71a4783
SH		 806 break;
807
20d49473
PO		 808 case SO_TIMESTAMPING:
809 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 810 ret = -EINVAL;
20d49473
PO		 811 break;
812 }
813 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
814 val & SOF_TIMESTAMPING_TX_HARDWARE);
815 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
816 val & SOF_TIMESTAMPING_TX_SOFTWARE);
817 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
818 val & SOF_TIMESTAMPING_RX_HARDWARE);
819 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
820 sock_enable_timestamp(sk,
821 SOCK_TIMESTAMPING_RX_SOFTWARE);
822 else
823 sock_disable_timestamp(sk,
08e29af3 824 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 825 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
826 val & SOF_TIMESTAMPING_SOFTWARE);
827 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
828 val & SOF_TIMESTAMPING_SYS_HARDWARE);
829 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
830 val & SOF_TIMESTAMPING_RAW_HARDWARE);
831 break;
832
e71a4783
SH		 833 case SO_RCVLOWAT:
834 if (val < 0)
835 val = INT_MAX;
836 sk->sk_rcvlowat = val ? : 1;
837 break;
838
839 case SO_RCVTIMEO:
840 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
841 break;
842
843 case SO_SNDTIMEO:
844 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
845 break;
1da177e4 846
e71a4783
SH		 847 case SO_ATTACH_FILTER:
848 ret = -EINVAL;
849 if (optlen == sizeof(struct sock_fprog)) {
850 struct sock_fprog fprog;
1da177e4 851
e71a4783
SH		 852 ret = -EFAULT;
853 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 854 break;
e71a4783
SH		 855
856 ret = sk_attach_filter(&fprog, sk);
857 }
858 break;
859
860 case SO_DETACH_FILTER:
55b33325 861 ret = sk_detach_filter(sk);
e71a4783 862 break;
1da177e4 863
e71a4783
SH		 864 case SO_PASSSEC:
865 if (valbool)
866 set_bit(SOCK_PASSSEC, &sock->flags);
867 else
868 clear_bit(SOCK_PASSSEC, &sock->flags);
869 break;
4a19ec58 870 case SO_MARK:
5e1fccc0 871 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 872 ret = -EPERM;
2a91525c 873 else
4a19ec58 874 sk->sk_mark = val;
4a19ec58 875 break;
877ce7c1 876
1da177e4
LT		 877 /* We implement the SO_SNDLOWAT etc to
878 not be settable (1003.1g 5.3) */
3b885787 879 case SO_RXQ_OVFL:
8083f0fc 880 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 881 break;
6e3e939f
JB		 882
883 case SO_WIFI_STATUS:
884 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
885 break;
886
ef64a54f
PE		 887 case SO_PEEK_OFF:
888 if (sock->ops->set_peek_off)
889 sock->ops->set_peek_off(sk, val);
890 else
891 ret = -EOPNOTSUPP;
892 break;
3bdc0eba
BG		 893
894 case SO_NOFCS:
895 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
896 break;
897
e71a4783
SH		 898 default:
899 ret = -ENOPROTOOPT;
900 break;
4ec93edb 901 }
1da177e4
LT		 902 release_sock(sk);
903 return ret;
904}
2a91525c 905EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 906
907
3f551f94
EB		 908void cred_to_ucred(struct pid *pid, const struct cred *cred,
909 struct ucred *ucred)
910{
911 ucred->pid = pid_vnr(pid);
912 ucred->uid = ucred->gid = -1;
913 if (cred) {
914 struct user_namespace *current_ns = current_user_ns();
915
b2e4f544
EB		 916 ucred->uid = from_kuid_munged(current_ns, cred->euid);
917 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 918 }
919}
3924773a 920EXPORT_SYMBOL_GPL(cred_to_ucred);
3f551f94 921
1da177e4
LT		 922int sock_getsockopt(struct socket *sock, int level, int optname,
923 char __user *optval, int __user *optlen)
924{
925 struct sock *sk = sock->sk;
4ec93edb 926
e71a4783 927 union {
4ec93edb
YH		 928 int val;
929 struct linger ling;
1da177e4
LT		 930 struct timeval tm;
931 } v;
4ec93edb 932
4d0392be 933 int lv = sizeof(int);
1da177e4 934 int len;
4ec93edb 935
e71a4783 936 if (get_user(len, optlen))
4ec93edb 937 return -EFAULT;
e71a4783 938 if (len < 0)
1da177e4 939 return -EINVAL;
4ec93edb 940
50fee1de 941 memset(&v, 0, sizeof(v));
df0bca04 942
2a91525c 943 switch (optname) {
e71a4783
SH		 944 case SO_DEBUG:
945 v.val = sock_flag(sk, SOCK_DBG);
946 break;
947
948 case SO_DONTROUTE:
949 v.val = sock_flag(sk, SOCK_LOCALROUTE);
950 break;
951
952 case SO_BROADCAST:
1b23a5df 953 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 954 break;
955
956 case SO_SNDBUF:
957 v.val = sk->sk_sndbuf;
958 break;
959
960 case SO_RCVBUF:
961 v.val = sk->sk_rcvbuf;
962 break;
963
964 case SO_REUSEADDR:
965 v.val = sk->sk_reuse;
966 break;
967
968 case SO_KEEPALIVE:
1b23a5df 969 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 970 break;
971
972 case SO_TYPE:
973 v.val = sk->sk_type;
974 break;
975
49c794e9
JE		 976 case SO_PROTOCOL:
977 v.val = sk->sk_protocol;
978 break;
979
0d6038ee
JE		 980 case SO_DOMAIN:
981 v.val = sk->sk_family;
982 break;
983
e71a4783
SH		 984 case SO_ERROR:
985 v.val = -sock_error(sk);
2a91525c 986 if (v.val == 0)
e71a4783
SH		 987 v.val = xchg(&sk->sk_err_soft, 0);
988 break;
989
990 case SO_OOBINLINE:
1b23a5df 991 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 992 break;
993
994 case SO_NO_CHECK:
995 v.val = sk->sk_no_check;
996 break;
997
998 case SO_PRIORITY:
999 v.val = sk->sk_priority;
1000 break;
1001
1002 case SO_LINGER:
1003 lv = sizeof(v.ling);
1b23a5df 1004 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1005 v.ling.l_linger = sk->sk_lingertime / HZ;
1006 break;
1007
1008 case SO_BSDCOMPAT:
1009 sock_warn_obsolete_bsdism("getsockopt");
1010 break;
1011
1012 case SO_TIMESTAMP:
92f37fd2
ED		 1013 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1014 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1015 break;
1016
1017 case SO_TIMESTAMPNS:
1018 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1019 break;
1020
20d49473
PO		 1021 case SO_TIMESTAMPING:
1022 v.val = 0;
1023 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
1024 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
1025 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
1026 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
1027 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
1028 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
1029 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
1030 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
1031 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
1032 v.val |= SOF_TIMESTAMPING_SOFTWARE;
1033 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
1034 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
1035 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
1036 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
1037 break;
1038
e71a4783 1039 case SO_RCVTIMEO:
2a91525c 1040 lv = sizeof(struct timeval);
e71a4783
SH		 1041 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1042 v.tm.tv_sec = 0;
1043 v.tm.tv_usec = 0;
1044 } else {
1045 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1046 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1047 }
1048 break;
1049
1050 case SO_SNDTIMEO:
2a91525c 1051 lv = sizeof(struct timeval);
e71a4783
SH		 1052 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1053 v.tm.tv_sec = 0;
1054 v.tm.tv_usec = 0;
1055 } else {
1056 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1057 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1058 }
1059 break;
1da177e4 1060
e71a4783
SH		 1061 case SO_RCVLOWAT:
1062 v.val = sk->sk_rcvlowat;
1063 break;
1da177e4 1064
e71a4783 1065 case SO_SNDLOWAT:
2a91525c 1066 v.val = 1;
e71a4783 1067 break;
1da177e4 1068
e71a4783 1069 case SO_PASSCRED:
82981930 1070 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1071 break;
1da177e4 1072
e71a4783 1073 case SO_PEERCRED:
109f6e39
EB		 1074 {
1075 struct ucred peercred;
1076 if (len > sizeof(peercred))
1077 len = sizeof(peercred);
1078 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1079 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1080 return -EFAULT;
1081 goto lenout;
109f6e39 1082 }
1da177e4 1083
e71a4783
SH		 1084 case SO_PEERNAME:
1085 {
1086 char address[128];
1087
1088 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1089 return -ENOTCONN;
1090 if (lv < len)
1091 return -EINVAL;
1092 if (copy_to_user(optval, address, len))
1093 return -EFAULT;
1094 goto lenout;
1095 }
1da177e4 1096
e71a4783
SH		 1097 /* Dubious BSD thing... Probably nobody even uses it, but
1098 * the UNIX standard wants it for whatever reason... -DaveM
1099 */
1100 case SO_ACCEPTCONN:
1101 v.val = sk->sk_state == TCP_LISTEN;
1102 break;
1da177e4 1103
e71a4783 1104 case SO_PASSSEC:
82981930 1105 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1106 break;
877ce7c1 1107
e71a4783
SH		 1108 case SO_PEERSEC:
1109 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1110
4a19ec58
LAT		 1111 case SO_MARK:
1112 v.val = sk->sk_mark;
1113 break;
1114
3b885787 1115 case SO_RXQ_OVFL:
1b23a5df 1116 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1117 break;
1118
6e3e939f 1119 case SO_WIFI_STATUS:
1b23a5df 1120 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1121 break;
1122
ef64a54f
PE		 1123 case SO_PEEK_OFF:
1124 if (!sock->ops->set_peek_off)
1125 return -EOPNOTSUPP;
1126
1127 v.val = sk->sk_peek_off;
1128 break;
bc2f7996 1129 case SO_NOFCS:
1b23a5df 1130 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1131 break;
c91f6df2 1132
f7b86bfe 1133 case SO_BINDTODEVICE:
c91f6df2
BH		 1134 return sock_getbindtodevice(sk, optval, optlen, len);
1135
a8fc9277
PE		 1136 case SO_GET_FILTER:
1137 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1138 if (len < 0)
1139 return len;
1140
1141 goto lenout;
c91f6df2 1142
e71a4783
SH		 1143 default:
1144 return -ENOPROTOOPT;
1da177e4 1145 }
e71a4783 1146
1da177e4
LT		 1147 if (len > lv)
1148 len = lv;
1149 if (copy_to_user(optval, &v, len))
1150 return -EFAULT;
1151lenout:
4ec93edb
YH		 1152 if (put_user(len, optlen))
1153 return -EFAULT;
1154 return 0;
1da177e4
LT		 1155}
1156
a5b5bb9a
IM		 1157/*
1158 * Initialize an sk_lock.
1159 *
1160 * (We also register the sk_lock with the lock validator.)
1161 */
b6f99a21 1162static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1163{
ed07536e
PZ		 1164 sock_lock_init_class_and_name(sk,
1165 af_family_slock_key_strings[sk->sk_family],
1166 af_family_slock_keys + sk->sk_family,
1167 af_family_key_strings[sk->sk_family],
1168 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1169}
1170
4dc6dc71
ED		 1171/*
1172 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1173 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1174 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1175 */
f1a6c4da
PE		 1176static void sock_copy(struct sock *nsk, const struct sock *osk)
1177{
1178#ifdef CONFIG_SECURITY_NETWORK
1179 void *sptr = nsk->sk_security;
1180#endif
68835aba
ED		 1181 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1182
1183 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1184 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1185
f1a6c4da
PE		 1186#ifdef CONFIG_SECURITY_NETWORK
1187 nsk->sk_security = sptr;
1188 security_sk_clone(osk, nsk);
1189#endif
1190}
1191
fcbdf09d
OP		 1192/*
1193 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
1194 * un-modified. Special care is taken when initializing object to zero.
1195 */
1196static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1197{
1198 if (offsetof(struct sock, sk_node.next) != 0)
1199 memset(sk, 0, offsetof(struct sock, sk_node.next));
1200 memset(&sk->sk_node.pprev, 0,
1201 size - offsetof(struct sock, sk_node.pprev));
1202}
1203
1204void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1205{
1206 unsigned long nulls1, nulls2;
1207
1208 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1209 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1210 if (nulls1 > nulls2)
1211 swap(nulls1, nulls2);
1212
1213 if (nulls1 != 0)
1214 memset((char *)sk, 0, nulls1);
1215 memset((char *)sk + nulls1 + sizeof(void *), 0,
1216 nulls2 - nulls1 - sizeof(void *));
1217 memset((char *)sk + nulls2 + sizeof(void *), 0,
1218 size - nulls2 - sizeof(void *));
1219}
1220EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1221
2e4afe7b
PE		 1222static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1223 int family)
c308c1b2
PE		 1224{
1225 struct sock *sk;
1226 struct kmem_cache *slab;
1227
1228 slab = prot->slab;
e912b114
ED		 1229 if (slab != NULL) {
1230 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1231 if (!sk)
1232 return sk;
1233 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1234 if (prot->clear_sk)
1235 prot->clear_sk(sk, prot->obj_size);
1236 else
1237 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1238 }
fcbdf09d 1239 } else
c308c1b2
PE		 1240 sk = kmalloc(prot->obj_size, priority);
1241
2e4afe7b 1242 if (sk != NULL) {
a98b65a3
VN		 1243 kmemcheck_annotate_bitfield(sk, flags);
1244
2e4afe7b
PE		 1245 if (security_sk_alloc(sk, family, priority))
1246 goto out_free;
1247
1248 if (!try_module_get(prot->owner))
1249 goto out_free_sec;
e022f0b4 1250 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1251 }
1252
c308c1b2 1253 return sk;
2e4afe7b
PE		 1254
1255out_free_sec:
1256 security_sk_free(sk);
1257out_free:
1258 if (slab != NULL)
1259 kmem_cache_free(slab, sk);
1260 else
1261 kfree(sk);
1262 return NULL;
c308c1b2
PE		 1263}
1264
1265static void sk_prot_free(struct proto *prot, struct sock *sk)
1266{
1267 struct kmem_cache *slab;
2e4afe7b 1268 struct module *owner;
c308c1b2 1269
2e4afe7b 1270 owner = prot->owner;
c308c1b2 1271 slab = prot->slab;
2e4afe7b
PE		 1272
1273 security_sk_free(sk);
c308c1b2
PE		 1274 if (slab != NULL)
1275 kmem_cache_free(slab, sk);
1276 else
1277 kfree(sk);
2e4afe7b 1278 module_put(owner);
c308c1b2
PE		 1279}
1280
f8451725 1281#ifdef CONFIG_CGROUPS
8fb974c9 1282#if IS_ENABLED(CONFIG_NET_CLS_CGROUP)
fd9a08a7 1283void sock_update_classid(struct sock *sk, struct task_struct *task)
f8451725 1284{
1144182a 1285 u32 classid;
f8451725 1286
fd9a08a7 1287 classid = task_cls_classid(task);
3afa6d00 1288 if (classid != sk->sk_classid)
f8451725
HX		 1289 sk->sk_classid = classid;
1290}
82862742 1291EXPORT_SYMBOL(sock_update_classid);
8fb974c9 1292#endif
5bc1421e 1293
51e4e7fa 1294#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
406a3c63 1295void sock_update_netprioidx(struct sock *sk, struct task_struct *task)
5bc1421e 1296{
5bc1421e
NH		 1297 if (in_interrupt())
1298 return;
2b73bc65 1299
406a3c63 1300 sk->sk_cgrp_prioidx = task_netprioidx(task);
5bc1421e
NH		 1301}
1302EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725 1303#endif
51e4e7fa 1304#endif
f8451725 1305
1da177e4
LT		 1306/**
1307 * sk_alloc - All socket objects are allocated here
c4ea43c5 1308 * @net: the applicable net namespace
4dc3b16b
PP		 1309 * @family: protocol family
1310 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1311 * @prot: struct proto associated with this new sock instance
1da177e4 1312 */
1b8d7ae4 1313struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1314 struct proto *prot)
1da177e4 1315{
c308c1b2 1316 struct sock *sk;
1da177e4 1317
154adbc8 1318 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1319 if (sk) {
154adbc8
PE		 1320 sk->sk_family = family;
1321 /*
1322 * See comment in struct sock definition to understand
1323 * why we need sk_prot_creator -acme
1324 */
1325 sk->sk_prot = sk->sk_prot_creator = prot;
1326 sock_lock_init(sk);
3b1e0a65 1327 sock_net_set(sk, get_net(net));
d66ee058 1328 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1329
fd9a08a7 1330 sock_update_classid(sk, current);
406a3c63 1331 sock_update_netprioidx(sk, current);
1da177e4 1332 }
a79af59e 1333
2e4afe7b 1334 return sk;
1da177e4 1335}
2a91525c 1336EXPORT_SYMBOL(sk_alloc);
1da177e4 1337
2b85a34e 1338static void __sk_free(struct sock *sk)
1da177e4
LT		 1339{
1340 struct sk_filter *filter;
1da177e4
LT		 1341
1342 if (sk->sk_destruct)
1343 sk->sk_destruct(sk);
1344
a898def2
PM		 1345 filter = rcu_dereference_check(sk->sk_filter,
1346 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1347 if (filter) {
309dd5fc 1348 sk_filter_uncharge(sk, filter);
a9b3cd7f 1349 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1350 }
1351
08e29af3 1352 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1353
1354 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1355 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1356 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1357
109f6e39
EB		 1358 if (sk->sk_peer_cred)
1359 put_cred(sk->sk_peer_cred);
1360 put_pid(sk->sk_peer_pid);
3b1e0a65 1361 put_net(sock_net(sk));
c308c1b2 1362 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1363}
2b85a34e
ED		 1364
1365void sk_free(struct sock *sk)
1366{
1367 /*
25985edc 1368 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1369 * some packets are still in some tx queue.
1370 * If not null, sock_wfree() will call __sk_free(sk) later
1371 */
1372 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1373 __sk_free(sk);
1374}
2a91525c 1375EXPORT_SYMBOL(sk_free);
1da177e4 1376
edf02087 1377/*
25985edc
LDM		 1378 * Last sock_put should drop reference to sk->sk_net. It has already
1379 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1380 * is not an option.
25985edc 1381 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1382 * destroy it in the context of init_net.
1383 */
1384void sk_release_kernel(struct sock *sk)
1385{
1386 if (sk == NULL || sk->sk_socket == NULL)
1387 return;
1388
1389 sock_hold(sk);
1390 sock_release(sk->sk_socket);
65a18ec5 1391 release_net(sock_net(sk));
3b1e0a65 1392 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1393 sock_put(sk);
1394}
45af1754 1395EXPORT_SYMBOL(sk_release_kernel);
edf02087 1396
475f1b52
SR		 1397static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1398{
1399 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1400 sock_update_memcg(newsk);
1401}
1402
e56c57d0
ED		 1403/**
1404 * sk_clone_lock - clone a socket, and lock its clone
1405 * @sk: the socket to clone
1406 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1407 *
1408 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1409 */
1410struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1411{
8fd1d178 1412 struct sock *newsk;
87d11ceb 1413
8fd1d178 1414 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1415 if (newsk != NULL) {
1416 struct sk_filter *filter;
1417
892c141e 1418 sock_copy(newsk, sk);
87d11ceb
ACM		 1419
1420 /* SANITY */
3b1e0a65 1421 get_net(sock_net(newsk));
87d11ceb
ACM		 1422 sk_node_init(&newsk->sk_node);
1423 sock_lock_init(newsk);
1424 bh_lock_sock(newsk);
fa438ccf 1425 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1426 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1427
1428 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1429 /*
1430 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1431 */
1432 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1433 atomic_set(&newsk->sk_omem_alloc, 0);
1434 skb_queue_head_init(&newsk->sk_receive_queue);
1435 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1436#ifdef CONFIG_NET_DMA
1437 skb_queue_head_init(&newsk->sk_async_wait_queue);
1438#endif
87d11ceb 1439
b6c6712a 1440 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1441 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1442 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1443 af_callback_keys + newsk->sk_family,
1444 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1445
1446 newsk->sk_dst_cache = NULL;
1447 newsk->sk_wmem_queued = 0;
1448 newsk->sk_forward_alloc = 0;
1449 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1450 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1451
1452 sock_reset_flag(newsk, SOCK_DONE);
1453 skb_queue_head_init(&newsk->sk_error_queue);
1454
0d7da9dd 1455 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1456 if (filter != NULL)
1457 sk_filter_charge(newsk, filter);
1458
1459 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1460 /* It is still raw copy of parent, so invalidate
1461 * destructor and make plain sk_free() */
1462 newsk->sk_destruct = NULL;
b0691c8e 1463 bh_unlock_sock(newsk);
87d11ceb
ACM		 1464 sk_free(newsk);
1465 newsk = NULL;
1466 goto out;
1467 }
1468
1469 newsk->sk_err = 0;
1470 newsk->sk_priority = 0;
4dc6dc71
ED		 1471 /*
1472 * Before updating sk_refcnt, we must commit prior changes to memory
1473 * (Documentation/RCU/rculist_nulls.txt for details)
1474 */
1475 smp_wmb();
87d11ceb
ACM		 1476 atomic_set(&newsk->sk_refcnt, 2);
1477
1478 /*
1479 * Increment the counter in the same struct proto as the master
1480 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1481 * is the same as sk->sk_prot->socks, as this field was copied
1482 * with memcpy).
1483 *
1484 * This _changes_ the previous behaviour, where
1485 * tcp_create_openreq_child always was incrementing the
1486 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1487 * to be taken into account in all callers. -acme
1488 */
1489 sk_refcnt_debug_inc(newsk);
972692e0 1490 sk_set_socket(newsk, NULL);
43815482 1491 newsk->sk_wq = NULL;
87d11ceb 1492
f3f511e1
GC		 1493 sk_update_clone(sk, newsk);
1494
87d11ceb 1495 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1496 sk_sockets_allocated_inc(newsk);
704da560 1497
08e29af3 1498 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1499 net_enable_timestamp();
87d11ceb
ACM		 1500 }
1501out:
1502 return newsk;
1503}
e56c57d0 1504EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1505
9958089a
AK		 1506void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1507{
1508 __sk_dst_set(sk, dst);
1509 sk->sk_route_caps = dst->dev->features;
1510 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1511 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1512 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1513 if (sk_can_gso(sk)) {
82cc1a7a 1514 if (dst->header_len) {
9958089a 1515 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1516 } else {
9958089a 1517 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1518 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1519 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1520 }
9958089a
AK		 1521 }
1522}
1523EXPORT_SYMBOL_GPL(sk_setup_caps);
1524
1da177e4
LT		 1525/*
1526 * Simple resource managers for sockets.
1527 */
1528
1529
4ec93edb
YH		 1530/*
1531 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1532 */
1533void sock_wfree(struct sk_buff *skb)
1534{
1535 struct sock *sk = skb->sk;
d99927f4 1536 unsigned int len = skb->truesize;
1da177e4 1537
d99927f4
ED		 1538 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1539 /*
1540 * Keep a reference on sk_wmem_alloc, this will be released
1541 * after sk_write_space() call
1542 */
1543 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1544 sk->sk_write_space(sk);
d99927f4
ED		 1545 len = 1;
1546 }
2b85a34e 1547 /*
d99927f4
ED		 1548 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1549 * could not do because of in-flight packets
2b85a34e 1550 */
d99927f4 1551 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1552 __sk_free(sk);
1da177e4 1553}
2a91525c 1554EXPORT_SYMBOL(sock_wfree);
1da177e4 1555
4ec93edb
YH		 1556/*
1557 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1558 */
1559void sock_rfree(struct sk_buff *skb)
1560{
1561 struct sock *sk = skb->sk;
d361fd59 1562 unsigned int len = skb->truesize;
1da177e4 1563
d361fd59
ED		 1564 atomic_sub(len, &sk->sk_rmem_alloc);
1565 sk_mem_uncharge(sk, len);
1da177e4 1566}
2a91525c 1567EXPORT_SYMBOL(sock_rfree);
1da177e4 1568
41063e9d
DM		 1569void sock_edemux(struct sk_buff *skb)
1570{
e812347c
ED		 1571 struct sock *sk = skb->sk;
1572
1c463e57 1573#ifdef CONFIG_INET
e812347c
ED		 1574 if (sk->sk_state == TCP_TIME_WAIT)
1575 inet_twsk_put(inet_twsk(sk));
1576 else
1c463e57 1577#endif
e812347c 1578 sock_put(sk);
41063e9d
DM		 1579}
1580EXPORT_SYMBOL(sock_edemux);
1da177e4 1581
976d0201 1582kuid_t sock_i_uid(struct sock *sk)
1da177e4 1583{
976d0201 1584 kuid_t uid;
1da177e4 1585
f064af1e 1586 read_lock_bh(&sk->sk_callback_lock);
976d0201 1587 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1588 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1589 return uid;
1590}
2a91525c 1591EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1592
1593unsigned long sock_i_ino(struct sock *sk)
1594{
1595 unsigned long ino;
1596
f064af1e 1597 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1598 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1599 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1600 return ino;
1601}
2a91525c 1602EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1603
1604/*
1605 * Allocate a skb from the socket's send buffer.
1606 */
86a76caf 1607struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1608 gfp_t priority)
1da177e4
LT		 1609{
1610 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1611 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1612 if (skb) {
1613 skb_set_owner_w(skb, sk);
1614 return skb;
1615 }
1616 }
1617 return NULL;
1618}
2a91525c 1619EXPORT_SYMBOL(sock_wmalloc);
1da177e4
LT		 1620
1621/*
1622 * Allocate a skb from the socket's receive buffer.
4ec93edb 1623 */
86a76caf 1624struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1625 gfp_t priority)
1da177e4
LT		 1626{
1627 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1628 struct sk_buff *skb = alloc_skb(size, priority);
1629 if (skb) {
1630 skb_set_owner_r(skb, sk);
1631 return skb;
1632 }
1633 }
1634 return NULL;
1635}
1636
4ec93edb 1637/*
1da177e4 1638 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1639 */
dd0fc66f 1640void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1641{
95c96174 1642 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1643 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1644 void *mem;
1645 /* First do the add, to avoid the race if kmalloc
4ec93edb 1646 * might sleep.
1da177e4
LT		 1647 */
1648 atomic_add(size, &sk->sk_omem_alloc);
1649 mem = kmalloc(size, priority);
1650 if (mem)
1651 return mem;
1652 atomic_sub(size, &sk->sk_omem_alloc);
1653 }
1654 return NULL;
1655}
2a91525c 1656EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1657
1658/*
1659 * Free an option memory block.
1660 */
1661void sock_kfree_s(struct sock *sk, void *mem, int size)
1662{
1663 kfree(mem);
1664 atomic_sub(size, &sk->sk_omem_alloc);
1665}
2a91525c 1666EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1667
1668/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1669 I think, these locks should be removed for datagram sockets.
1670 */
2a91525c 1671static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1672{
1673 DEFINE_WAIT(wait);
1674
1675 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1676 for (;;) {
1677 if (!timeo)
1678 break;
1679 if (signal_pending(current))
1680 break;
1681 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1682 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1683 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1684 break;
1685 if (sk->sk_shutdown & SEND_SHUTDOWN)
1686 break;
1687 if (sk->sk_err)
1688 break;
1689 timeo = schedule_timeout(timeo);
1690 }
aa395145 1691 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1692 return timeo;
1693}
1694
1695
1696/*
1697 * Generic send/receive buffer handlers
1698 */
1699
4cc7f68d
HX		 1700struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1701 unsigned long data_len, int noblock,
1702 int *errcode)
1da177e4
LT		 1703{
1704 struct sk_buff *skb;
7d877f3b 1705 gfp_t gfp_mask;
1da177e4
LT		 1706 long timeo;
1707 int err;
cc9b17ad
JW		 1708 int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1709
1710 err = -EMSGSIZE;
1711 if (npages > MAX_SKB_FRAGS)
1712 goto failure;
1da177e4
LT		 1713
1714 gfp_mask = sk->sk_allocation;
1715 if (gfp_mask & __GFP_WAIT)
1716 gfp_mask |= __GFP_REPEAT;
1717
1718 timeo = sock_sndtimeo(sk, noblock);
1719 while (1) {
1720 err = sock_error(sk);
1721 if (err != 0)
1722 goto failure;
1723
1724 err = -EPIPE;
1725 if (sk->sk_shutdown & SEND_SHUTDOWN)
1726 goto failure;
1727
1728 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1729 skb = alloc_skb(header_len, gfp_mask);
1da177e4 1730 if (skb) {
1da177e4
LT		 1731 int i;
1732
1733 /* No pages, we're done... */
1734 if (!data_len)
1735 break;
1736
1da177e4
LT		 1737 skb->truesize += data_len;
1738 skb_shinfo(skb)->nr_frags = npages;
1739 for (i = 0; i < npages; i++) {
1740 struct page *page;
1da177e4
LT		 1741
1742 page = alloc_pages(sk->sk_allocation, 0);
1743 if (!page) {
1744 err = -ENOBUFS;
1745 skb_shinfo(skb)->nr_frags = i;
1746 kfree_skb(skb);
1747 goto failure;
1748 }
1749
ea2ab693
IC		 1750 __skb_fill_page_desc(skb, i,
1751 page, 0,
1752 (data_len >= PAGE_SIZE ?
1753 PAGE_SIZE :
1754 data_len));
1da177e4
LT		 1755 data_len -= PAGE_SIZE;
1756 }
1757
1758 /* Full success... */
1759 break;
1760 }
1761 err = -ENOBUFS;
1762 goto failure;
1763 }
1764 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1765 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1766 err = -EAGAIN;
1767 if (!timeo)
1768 goto failure;
1769 if (signal_pending(current))
1770 goto interrupted;
1771 timeo = sock_wait_for_wmem(sk, timeo);
1772 }
1773
1774 skb_set_owner_w(skb, sk);
1775 return skb;
1776
1777interrupted:
1778 err = sock_intr_errno(timeo);
1779failure:
1780 *errcode = err;
1781 return NULL;
1782}
4cc7f68d 1783EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1784
4ec93edb 1785struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1786 int noblock, int *errcode)
1787{
1788 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1789}
2a91525c 1790EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1791
5640f768
ED		 1792/* On 32bit arches, an skb frag is limited to 2^15 */
1793#define SKB_FRAG_PAGE_ORDER get_order(32768)
1794
1795bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1796{
1797 int order;
1798
1799 if (pfrag->page) {
1800 if (atomic_read(&pfrag->page->_count) == 1) {
1801 pfrag->offset = 0;
1802 return true;
1803 }
1804 if (pfrag->offset < pfrag->size)
1805 return true;
1806 put_page(pfrag->page);
1807 }
1808
1809 /* We restrict high order allocations to users that can afford to wait */
1810 order = (sk->sk_allocation & __GFP_WAIT) ? SKB_FRAG_PAGE_ORDER : 0;
1811
1812 do {
1813 gfp_t gfp = sk->sk_allocation;
1814
1815 if (order)
1816 gfp |= __GFP_COMP | __GFP_NOWARN;
1817 pfrag->page = alloc_pages(gfp, order);
1818 if (likely(pfrag->page)) {
1819 pfrag->offset = 0;
1820 pfrag->size = PAGE_SIZE << order;
1821 return true;
1822 }
1823 } while (--order >= 0);
1824
1825 sk_enter_memory_pressure(sk);
1826 sk_stream_moderate_sndbuf(sk);
1827 return false;
1828}
1829EXPORT_SYMBOL(sk_page_frag_refill);
1830
1da177e4 1831static void __lock_sock(struct sock *sk)
f39234d6
NK		 1832 __releases(&sk->sk_lock.slock)
1833 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1834{
1835 DEFINE_WAIT(wait);
1836
e71a4783 1837 for (;;) {
1da177e4
LT		 1838 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1839 TASK_UNINTERRUPTIBLE);
1840 spin_unlock_bh(&sk->sk_lock.slock);
1841 schedule();
1842 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1843 if (!sock_owned_by_user(sk))
1da177e4
LT		 1844 break;
1845 }
1846 finish_wait(&sk->sk_lock.wq, &wait);
1847}
1848
1849static void __release_sock(struct sock *sk)
f39234d6
NK		 1850 __releases(&sk->sk_lock.slock)
1851 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1852{
1853 struct sk_buff *skb = sk->sk_backlog.head;
1854
1855 do {
1856 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1857 bh_unlock_sock(sk);
1858
1859 do {
1860 struct sk_buff *next = skb->next;
1861
e4cbb02a 1862 prefetch(next);
7fee226a 1863 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1864 skb->next = NULL;
c57943a1 1865 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1866
1867 /*
1868 * We are in process context here with softirqs
1869 * disabled, use cond_resched_softirq() to preempt.
1870 * This is safe to do because we've taken the backlog
1871 * queue private:
1872 */
1873 cond_resched_softirq();
1874
1875 skb = next;
1876 } while (skb != NULL);
1877
1878 bh_lock_sock(sk);
e71a4783 1879 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1880
1881 /*
1882 * Doing the zeroing here guarantee we can not loop forever
1883 * while a wild producer attempts to flood us.
1884 */
1885 sk->sk_backlog.len = 0;
1da177e4
LT		 1886}
1887
1888/**
1889 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1890 * @sk: sock to wait on
1891 * @timeo: for how long
1da177e4
LT		 1892 *
1893 * Now socket state including sk->sk_err is changed only under lock,
1894 * hence we may omit checks after joining wait queue.
1895 * We check receive queue before schedule() only as optimization;
1896 * it is very likely that release_sock() added new data.
1897 */
1898int sk_wait_data(struct sock *sk, long *timeo)
1899{
1900 int rc;
1901 DEFINE_WAIT(wait);
1902
aa395145 1903 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1904 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1905 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1906 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1907 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1908 return rc;
1909}
1da177e4
LT		 1910EXPORT_SYMBOL(sk_wait_data);
1911
3ab224be
HA		 1912/**
1913 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1914 * @sk: socket
1915 * @size: memory size to allocate
1916 * @kind: allocation type
1917 *
1918 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1919 * rmem allocation. This function assumes that protocols which have
1920 * memory_pressure use sk_wmem_queued as write buffer accounting.
1921 */
1922int __sk_mem_schedule(struct sock *sk, int size, int kind)
1923{
1924 struct proto *prot = sk->sk_prot;
1925 int amt = sk_mem_pages(size);
8d987e5c 1926 long allocated;
e1aab161 1927 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1928
1929 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1930
e1aab161 1931 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1932
1933 /* Under limit. */
e1aab161
GC		 1934 if (parent_status == UNDER_LIMIT &&
1935 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1936 sk_leave_memory_pressure(sk);
3ab224be
HA		 1937 return 1;
1938 }
1939
e1aab161
GC		 1940 /* Under pressure. (we or our parents) */
1941 if ((parent_status > SOFT_LIMIT) ||
1942 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1943 sk_enter_memory_pressure(sk);
3ab224be 1944
e1aab161
GC		 1945 /* Over hard limit (we or our parents) */
1946 if ((parent_status == OVER_LIMIT) ||
1947 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1948 goto suppress_allocation;
1949
1950 /* guarantee minimum buffer size under pressure */
1951 if (kind == SK_MEM_RECV) {
1952 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1953 return 1;
180d8cd9 1954
3ab224be
HA		 1955 } else { /* SK_MEM_SEND */
1956 if (sk->sk_type == SOCK_STREAM) {
1957 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1958 return 1;
1959 } else if (atomic_read(&sk->sk_wmem_alloc) <
1960 prot->sysctl_wmem[0])
1961 return 1;
1962 }
1963
180d8cd9 1964 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 1965 int alloc;
1966
180d8cd9 1967 if (!sk_under_memory_pressure(sk))
1748376b 1968 return 1;
180d8cd9
GC		 1969 alloc = sk_sockets_allocated_read_positive(sk);
1970 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 1971 sk_mem_pages(sk->sk_wmem_queued +
1972 atomic_read(&sk->sk_rmem_alloc) +
1973 sk->sk_forward_alloc))
1974 return 1;
1975 }
1976
1977suppress_allocation:
1978
1979 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1980 sk_stream_moderate_sndbuf(sk);
1981
1982 /* Fail only if socket is _under_ its sndbuf.
1983 * In this case we cannot block, so that we have to fail.
1984 */
1985 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1986 return 1;
1987 }
1988
3847ce32
SM		 1989 trace_sock_exceed_buf_limit(sk, prot, allocated);
1990
3ab224be
HA		 1991 /* Alas. Undo changes. */
1992 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 1993
0e90b31f 1994 sk_memory_allocated_sub(sk, amt);
180d8cd9 1995
3ab224be
HA		 1996 return 0;
1997}
3ab224be
HA		 1998EXPORT_SYMBOL(__sk_mem_schedule);
1999
2000/**
2001 * __sk_reclaim - reclaim memory_allocated
2002 * @sk: socket
2003 */
2004void __sk_mem_reclaim(struct sock *sk)
2005{
180d8cd9 2006 sk_memory_allocated_sub(sk,
0e90b31f 2007 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 2008 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
2009
180d8cd9
GC		 2010 if (sk_under_memory_pressure(sk) &&
2011 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2012 sk_leave_memory_pressure(sk);
3ab224be 2013}
3ab224be
HA		 2014EXPORT_SYMBOL(__sk_mem_reclaim);
2015
2016
1da177e4
LT		 2017/*
2018 * Set of default routines for initialising struct proto_ops when
2019 * the protocol does not support a particular function. In certain
2020 * cases where it makes no sense for a protocol to have a "do nothing"
2021 * function, some default processing is provided.
2022 */
2023
2024int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2025{
2026 return -EOPNOTSUPP;
2027}
2a91525c 2028EXPORT_SYMBOL(sock_no_bind);
1da177e4 2029
4ec93edb 2030int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2031 int len, int flags)
2032{
2033 return -EOPNOTSUPP;
2034}
2a91525c 2035EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2036
2037int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2038{
2039 return -EOPNOTSUPP;
2040}
2a91525c 2041EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2042
2043int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2044{
2045 return -EOPNOTSUPP;
2046}
2a91525c 2047EXPORT_SYMBOL(sock_no_accept);
1da177e4 2048
4ec93edb 2049int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2050 int *len, int peer)
2051{
2052 return -EOPNOTSUPP;
2053}
2a91525c 2054EXPORT_SYMBOL(sock_no_getname);
1da177e4 2055
2a91525c 2056unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2057{
2058 return 0;
2059}
2a91525c 2060EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2061
2062int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2063{
2064 return -EOPNOTSUPP;
2065}
2a91525c 2066EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2067
2068int sock_no_listen(struct socket *sock, int backlog)
2069{
2070 return -EOPNOTSUPP;
2071}
2a91525c 2072EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2073
2074int sock_no_shutdown(struct socket *sock, int how)
2075{
2076 return -EOPNOTSUPP;
2077}
2a91525c 2078EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2079
2080int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2081 char __user *optval, unsigned int optlen)
1da177e4
LT		 2082{
2083 return -EOPNOTSUPP;
2084}
2a91525c 2085EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2086
2087int sock_no_getsockopt(struct socket *sock, int level, int optname,
2088 char __user *optval, int __user *optlen)
2089{
2090 return -EOPNOTSUPP;
2091}
2a91525c 2092EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2093
2094int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2095 size_t len)
2096{
2097 return -EOPNOTSUPP;
2098}
2a91525c 2099EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2100
2101int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2102 size_t len, int flags)
2103{
2104 return -EOPNOTSUPP;
2105}
2a91525c 2106EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2107
2108int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2109{
2110 /* Mirror missing mmap method error code */
2111 return -ENODEV;
2112}
2a91525c 2113EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2114
2115ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2116{
2117 ssize_t res;
2118 struct msghdr msg = {.msg_flags = flags};
2119 struct kvec iov;
2120 char *kaddr = kmap(page);
2121 iov.iov_base = kaddr + offset;
2122 iov.iov_len = size;
2123 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2124 kunmap(page);
2125 return res;
2126}
2a91525c 2127EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2128
2129/*
2130 * Default Socket Callbacks
2131 */
2132
2133static void sock_def_wakeup(struct sock *sk)
2134{
43815482
ED		 2135 struct socket_wq *wq;
2136
2137 rcu_read_lock();
2138 wq = rcu_dereference(sk->sk_wq);
2139 if (wq_has_sleeper(wq))
2140 wake_up_interruptible_all(&wq->wait);
2141 rcu_read_unlock();
1da177e4
LT		 2142}
2143
2144static void sock_def_error_report(struct sock *sk)
2145{
43815482
ED		 2146 struct socket_wq *wq;
2147
2148 rcu_read_lock();
2149 wq = rcu_dereference(sk->sk_wq);
2150 if (wq_has_sleeper(wq))
2151 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2152 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2153 rcu_read_unlock();
1da177e4
LT		 2154}
2155
2156static void sock_def_readable(struct sock *sk, int len)
2157{
43815482
ED		 2158 struct socket_wq *wq;
2159
2160 rcu_read_lock();
2161 wq = rcu_dereference(sk->sk_wq);
2162 if (wq_has_sleeper(wq))
2c6607c6 2163 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2164 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2165 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2166 rcu_read_unlock();
1da177e4
LT		 2167}
2168
2169static void sock_def_write_space(struct sock *sk)
2170{
43815482
ED		 2171 struct socket_wq *wq;
2172
2173 rcu_read_lock();
1da177e4
LT		 2174
2175 /* Do not wake up a writer until he can make "significant"
2176 * progress. --DaveM
2177 */
e71a4783 2178 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2179 wq = rcu_dereference(sk->sk_wq);
2180 if (wq_has_sleeper(wq))
2181 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2182 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2183
2184 /* Should agree with poll, otherwise some programs break */
2185 if (sock_writeable(sk))
8d8ad9d7 2186 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2187 }
2188
43815482 2189 rcu_read_unlock();
1da177e4
LT		 2190}
2191
2192static void sock_def_destruct(struct sock *sk)
2193{
a51482bd 2194 kfree(sk->sk_protinfo);
1da177e4
LT		 2195}
2196
2197void sk_send_sigurg(struct sock *sk)
2198{
2199 if (sk->sk_socket && sk->sk_socket->file)
2200 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2201 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2202}
2a91525c 2203EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2204
2205void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2206 unsigned long expires)
2207{
2208 if (!mod_timer(timer, expires))
2209 sock_hold(sk);
2210}
1da177e4
LT		 2211EXPORT_SYMBOL(sk_reset_timer);
2212
2213void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2214{
2215 if (timer_pending(timer) && del_timer(timer))
2216 __sock_put(sk);
2217}
1da177e4
LT		 2218EXPORT_SYMBOL(sk_stop_timer);
2219
2220void sock_init_data(struct socket *sock, struct sock *sk)
2221{
2222 skb_queue_head_init(&sk->sk_receive_queue);
2223 skb_queue_head_init(&sk->sk_write_queue);
2224 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2225#ifdef CONFIG_NET_DMA
2226 skb_queue_head_init(&sk->sk_async_wait_queue);
2227#endif
1da177e4
LT		 2228
2229 sk->sk_send_head = NULL;
2230
2231 init_timer(&sk->sk_timer);
4ec93edb 2232
1da177e4
LT		 2233 sk->sk_allocation = GFP_KERNEL;
2234 sk->sk_rcvbuf = sysctl_rmem_default;
2235 sk->sk_sndbuf = sysctl_wmem_default;
2236 sk->sk_state = TCP_CLOSE;
972692e0 2237 sk_set_socket(sk, sock);
1da177e4
LT		 2238
2239 sock_set_flag(sk, SOCK_ZAPPED);
2240
e71a4783 2241 if (sock) {
1da177e4 2242 sk->sk_type = sock->type;
43815482 2243 sk->sk_wq = sock->wq;
1da177e4
LT		 2244 sock->sk = sk;
2245 } else
43815482 2246 sk->sk_wq = NULL;
1da177e4 2247
b6c6712a 2248 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2249 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2250 lockdep_set_class_and_name(&sk->sk_callback_lock,
2251 af_callback_keys + sk->sk_family,
2252 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2253
2254 sk->sk_state_change = sock_def_wakeup;
2255 sk->sk_data_ready = sock_def_readable;
2256 sk->sk_write_space = sock_def_write_space;
2257 sk->sk_error_report = sock_def_error_report;
2258 sk->sk_destruct = sock_def_destruct;
2259
5640f768
ED		 2260 sk->sk_frag.page = NULL;
2261 sk->sk_frag.offset = 0;
ef64a54f 2262 sk->sk_peek_off = -1;
1da177e4 2263
109f6e39
EB		 2264 sk->sk_peer_pid = NULL;
2265 sk->sk_peer_cred = NULL;
1da177e4
LT		 2266 sk->sk_write_pending = 0;
2267 sk->sk_rcvlowat = 1;
2268 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2269 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2270
f37f0afb 2271 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2272
4dc6dc71
ED		 2273 /*
2274 * Before updating sk_refcnt, we must commit prior changes to memory
2275 * (Documentation/RCU/rculist_nulls.txt for details)
2276 */
2277 smp_wmb();
1da177e4 2278 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2279 atomic_set(&sk->sk_drops, 0);
1da177e4 2280}
2a91525c 2281EXPORT_SYMBOL(sock_init_data);
1da177e4 2282
b5606c2d 2283void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2284{
2285 might_sleep();
a5b5bb9a 2286 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2287 if (sk->sk_lock.owned)
1da177e4 2288 __lock_sock(sk);
d2e9117c 2289 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2290 spin_unlock(&sk->sk_lock.slock);
2291 /*
2292 * The sk_lock has mutex_lock() semantics here:
2293 */
fcc70d5f 2294 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2295 local_bh_enable();
1da177e4 2296}
fcc70d5f 2297EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2298
b5606c2d 2299void release_sock(struct sock *sk)
1da177e4 2300{
a5b5bb9a
IM		 2301 /*
2302 * The sk_lock has mutex_unlock() semantics:
2303 */
2304 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2305
2306 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2307 if (sk->sk_backlog.tail)
2308 __release_sock(sk);
46d3ceab
ED		 2309
2310 if (sk->sk_prot->release_cb)
2311 sk->sk_prot->release_cb(sk);
2312
d2e9117c 2313 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2314 if (waitqueue_active(&sk->sk_lock.wq))
2315 wake_up(&sk->sk_lock.wq);
2316 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2317}
2318EXPORT_SYMBOL(release_sock);
2319
8a74ad60
ED		 2320/**
2321 * lock_sock_fast - fast version of lock_sock
2322 * @sk: socket
2323 *
2324 * This version should be used for very small section, where process wont block
2325 * return false if fast path is taken
2326 * sk_lock.slock locked, owned = 0, BH disabled
2327 * return true if slow path is taken
2328 * sk_lock.slock unlocked, owned = 1, BH enabled
2329 */
2330bool lock_sock_fast(struct sock *sk)
2331{
2332 might_sleep();
2333 spin_lock_bh(&sk->sk_lock.slock);
2334
2335 if (!sk->sk_lock.owned)
2336 /*
2337 * Note : We must disable BH
2338 */
2339 return false;
2340
2341 __lock_sock(sk);
2342 sk->sk_lock.owned = 1;
2343 spin_unlock(&sk->sk_lock.slock);
2344 /*
2345 * The sk_lock has mutex_lock() semantics here:
2346 */
2347 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2348 local_bh_enable();
2349 return true;
2350}
2351EXPORT_SYMBOL(lock_sock_fast);
2352
1da177e4 2353int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2354{
b7aa0bf7 2355 struct timeval tv;
1da177e4 2356 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2357 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2358 tv = ktime_to_timeval(sk->sk_stamp);
2359 if (tv.tv_sec == -1)
1da177e4 2360 return -ENOENT;
b7aa0bf7
ED		 2361 if (tv.tv_sec == 0) {
2362 sk->sk_stamp = ktime_get_real();
2363 tv = ktime_to_timeval(sk->sk_stamp);
2364 }
2365 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2366}
1da177e4
LT		 2367EXPORT_SYMBOL(sock_get_timestamp);
2368
ae40eb1e
ED		 2369int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2370{
2371 struct timespec ts;
2372 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2373 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2374 ts = ktime_to_timespec(sk->sk_stamp);
2375 if (ts.tv_sec == -1)
2376 return -ENOENT;
2377 if (ts.tv_sec == 0) {
2378 sk->sk_stamp = ktime_get_real();
2379 ts = ktime_to_timespec(sk->sk_stamp);
2380 }
2381 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2382}
2383EXPORT_SYMBOL(sock_get_timestampns);
2384
20d49473 2385void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2386{
20d49473 2387 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2388 unsigned long previous_flags = sk->sk_flags;
2389
20d49473
PO		 2390 sock_set_flag(sk, flag);
2391 /*
2392 * we just set one of the two flags which require net
2393 * time stamping, but time stamping might have been on
2394 * already because of the other one
2395 */
08e29af3 2396 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2397 net_enable_timestamp();
1da177e4
LT		 2398 }
2399}
1da177e4
LT		 2400
2401/*
2402 * Get a socket option on an socket.
2403 *
2404 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2405 * asynchronous errors should be reported by getsockopt. We assume
2406 * this means if you specify SO_ERROR (otherwise whats the point of it).
2407 */
2408int sock_common_getsockopt(struct socket *sock, int level, int optname,
2409 char __user *optval, int __user *optlen)
2410{
2411 struct sock *sk = sock->sk;
2412
2413 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2414}
1da177e4
LT		 2415EXPORT_SYMBOL(sock_common_getsockopt);
2416
3fdadf7d 2417#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2418int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2419 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2420{
2421 struct sock *sk = sock->sk;
2422
1e51f951 2423 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2424 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2425 optval, optlen);
3fdadf7d
DM		 2426 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2427}
2428EXPORT_SYMBOL(compat_sock_common_getsockopt);
2429#endif
2430
1da177e4
LT		 2431int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2432 struct msghdr *msg, size_t size, int flags)
2433{
2434 struct sock *sk = sock->sk;
2435 int addr_len = 0;
2436 int err;
2437
2438 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2439 flags & ~MSG_DONTWAIT, &addr_len);
2440 if (err >= 0)
2441 msg->msg_namelen = addr_len;
2442 return err;
2443}
1da177e4
LT		 2444EXPORT_SYMBOL(sock_common_recvmsg);
2445
2446/*
2447 * Set socket options on an inet socket.
2448 */
2449int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2450 char __user *optval, unsigned int optlen)
1da177e4
LT		 2451{
2452 struct sock *sk = sock->sk;
2453
2454 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2455}
1da177e4
LT		 2456EXPORT_SYMBOL(sock_common_setsockopt);
2457
3fdadf7d 2458#ifdef CONFIG_COMPAT
543d9cfe 2459int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2460 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2461{
2462 struct sock *sk = sock->sk;
2463
543d9cfe
ACM		 2464 if (sk->sk_prot->compat_setsockopt != NULL)
2465 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2466 optval, optlen);
3fdadf7d
DM		 2467 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2468}
2469EXPORT_SYMBOL(compat_sock_common_setsockopt);
2470#endif
2471
1da177e4
LT		 2472void sk_common_release(struct sock *sk)
2473{
2474 if (sk->sk_prot->destroy)
2475 sk->sk_prot->destroy(sk);
2476
2477 /*
2478 * Observation: when sock_common_release is called, processes have
2479 * no access to socket. But net still has.
2480 * Step one, detach it from networking:
2481 *
2482 * A. Remove from hash tables.
2483 */
2484
2485 sk->sk_prot->unhash(sk);
2486
2487 /*
2488 * In this point socket cannot receive new packets, but it is possible
2489 * that some packets are in flight because some CPU runs receiver and
2490 * did hash table lookup before we unhashed socket. They will achieve
2491 * receive queue and will be purged by socket destructor.
2492 *
2493 * Also we still have packets pending on receive queue and probably,
2494 * our own packets waiting in device queues. sock_destroy will drain
2495 * receive queue, but transmitted packets will delay socket destruction
2496 * until the last reference will be released.
2497 */
2498
2499 sock_orphan(sk);
2500
2501 xfrm_sk_free_policy(sk);
2502
e6848976 2503 sk_refcnt_debug_release(sk);
5640f768
ED		 2504
2505 if (sk->sk_frag.page) {
2506 put_page(sk->sk_frag.page);
2507 sk->sk_frag.page = NULL;
2508 }
2509
1da177e4
LT		 2510 sock_put(sk);
2511}
1da177e4
LT		 2512EXPORT_SYMBOL(sk_common_release);
2513
13ff3d6f
PE		 2514#ifdef CONFIG_PROC_FS
2515#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2516struct prot_inuse {
2517 int val[PROTO_INUSE_NR];
2518};
13ff3d6f
PE		 2519
2520static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2521
2522#ifdef CONFIG_NET_NS
2523void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2524{
d6d9ca0f 2525 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2526}
2527EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2528
2529int sock_prot_inuse_get(struct net *net, struct proto *prot)
2530{
2531 int cpu, idx = prot->inuse_idx;
2532 int res = 0;
2533
2534 for_each_possible_cpu(cpu)
2535 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2536
2537 return res >= 0 ? res : 0;
2538}
2539EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2540
2c8c1e72 2541static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2542{
2543 net->core.inuse = alloc_percpu(struct prot_inuse);
2544 return net->core.inuse ? 0 : -ENOMEM;
2545}
2546
2c8c1e72 2547static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2548{
2549 free_percpu(net->core.inuse);
2550}
2551
2552static struct pernet_operations net_inuse_ops = {
2553 .init = sock_inuse_init_net,
2554 .exit = sock_inuse_exit_net,
2555};
2556
2557static __init int net_inuse_init(void)
2558{
2559 if (register_pernet_subsys(&net_inuse_ops))
2560 panic("Cannot initialize net inuse counters");
2561
2562 return 0;
2563}
2564
2565core_initcall(net_inuse_init);
2566#else
1338d466
PE		 2567static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2568
c29a0bc4 2569void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2570{
d6d9ca0f 2571 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2572}
2573EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2574
c29a0bc4 2575int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2576{
2577 int cpu, idx = prot->inuse_idx;
2578 int res = 0;
2579
2580 for_each_possible_cpu(cpu)
2581 res += per_cpu(prot_inuse, cpu).val[idx];
2582
2583 return res >= 0 ? res : 0;
2584}
2585EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2586#endif
13ff3d6f
PE		 2587
2588static void assign_proto_idx(struct proto *prot)
2589{
2590 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2591
2592 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2593 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2594 return;
2595 }
2596
2597 set_bit(prot->inuse_idx, proto_inuse_idx);
2598}
2599
2600static void release_proto_idx(struct proto *prot)
2601{
2602 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2603 clear_bit(prot->inuse_idx, proto_inuse_idx);
2604}
2605#else
2606static inline void assign_proto_idx(struct proto *prot)
2607{
2608}
2609
2610static inline void release_proto_idx(struct proto *prot)
2611{
2612}
2613#endif
2614
b733c007
PE		 2615int proto_register(struct proto *prot, int alloc_slab)
2616{
1da177e4
LT		 2617 if (alloc_slab) {
2618 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2619 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2620 NULL);
1da177e4
LT		 2621
2622 if (prot->slab == NULL) {
e005d193
JP		 2623 pr_crit("%s: Can't create sock SLAB cache!\n",
2624 prot->name);
60e7663d 2625 goto out;
1da177e4 2626 }
2e6599cb
ACM		 2627
2628 if (prot->rsk_prot != NULL) {
faf23422 2629 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2630 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2631 goto out_free_sock_slab;
2632
7e56b5d6 2633 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2634 prot->rsk_prot->obj_size, 0,
20c2df83 2635 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2636
2637 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2638 pr_crit("%s: Can't create request sock SLAB cache!\n",
2639 prot->name);
2e6599cb
ACM		 2640 goto out_free_request_sock_slab_name;
2641 }
2642 }
8feaf0c0 2643
6d6ee43e 2644 if (prot->twsk_prot != NULL) {
faf23422 2645 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2646
7e56b5d6 2647 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2648 goto out_free_request_sock_slab;
2649
6d6ee43e 2650 prot->twsk_prot->twsk_slab =
7e56b5d6 2651 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2652 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2653 0,
2654 SLAB_HWCACHE_ALIGN |
2655 prot->slab_flags,
20c2df83 2656 NULL);
6d6ee43e 2657 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2658 goto out_free_timewait_sock_slab_name;
2659 }
1da177e4
LT		 2660 }
2661
36b77a52 2662 mutex_lock(&proto_list_mutex);
1da177e4 2663 list_add(&prot->node, &proto_list);
13ff3d6f 2664 assign_proto_idx(prot);
36b77a52 2665 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2666 return 0;
2667
8feaf0c0 2668out_free_timewait_sock_slab_name:
7e56b5d6 2669 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2670out_free_request_sock_slab:
2671 if (prot->rsk_prot && prot->rsk_prot->slab) {
2672 kmem_cache_destroy(prot->rsk_prot->slab);
2673 prot->rsk_prot->slab = NULL;
2674 }
2e6599cb 2675out_free_request_sock_slab_name:
72150e9b
DC		 2676 if (prot->rsk_prot)
2677 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2678out_free_sock_slab:
2679 kmem_cache_destroy(prot->slab);
2680 prot->slab = NULL;
b733c007
PE		 2681out:
2682 return -ENOBUFS;
1da177e4 2683}
1da177e4
LT		 2684EXPORT_SYMBOL(proto_register);
2685
2686void proto_unregister(struct proto *prot)
2687{
36b77a52 2688 mutex_lock(&proto_list_mutex);
13ff3d6f 2689 release_proto_idx(prot);
0a3f4358 2690 list_del(&prot->node);
36b77a52 2691 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2692
2693 if (prot->slab != NULL) {
2694 kmem_cache_destroy(prot->slab);
2695 prot->slab = NULL;
2696 }
2697
2e6599cb 2698 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2699 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2700 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2701 prot->rsk_prot->slab = NULL;
2702 }
2703
6d6ee43e 2704 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2705 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2706 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2707 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2708 }
1da177e4 2709}
1da177e4
LT		 2710EXPORT_SYMBOL(proto_unregister);
2711
2712#ifdef CONFIG_PROC_FS
1da177e4 2713static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2714 __acquires(proto_list_mutex)
1da177e4 2715{
36b77a52 2716 mutex_lock(&proto_list_mutex);
60f0438a 2717 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2718}
2719
2720static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2721{
60f0438a 2722 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2723}
2724
2725static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2726 __releases(proto_list_mutex)
1da177e4 2727{
36b77a52 2728 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2729}
2730
2731static char proto_method_implemented(const void *method)
2732{
2733 return method == NULL ? 'n' : 'y';
2734}
180d8cd9
GC		 2735static long sock_prot_memory_allocated(struct proto *proto)
2736{
cb75a36c 2737 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2738}
2739
2740static char *sock_prot_memory_pressure(struct proto *proto)
2741{
2742 return proto->memory_pressure != NULL ?
2743 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2744}
1da177e4
LT		 2745
2746static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2747{
180d8cd9 2748
8d987e5c 2749 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2750 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2751 proto->name,
2752 proto->obj_size,
14e943db 2753 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2754 sock_prot_memory_allocated(proto),
2755 sock_prot_memory_pressure(proto),
1da177e4
LT		 2756 proto->max_header,
2757 proto->slab == NULL ? "no" : "yes",
2758 module_name(proto->owner),
2759 proto_method_implemented(proto->close),
2760 proto_method_implemented(proto->connect),
2761 proto_method_implemented(proto->disconnect),
2762 proto_method_implemented(proto->accept),
2763 proto_method_implemented(proto->ioctl),
2764 proto_method_implemented(proto->init),
2765 proto_method_implemented(proto->destroy),
2766 proto_method_implemented(proto->shutdown),
2767 proto_method_implemented(proto->setsockopt),
2768 proto_method_implemented(proto->getsockopt),
2769 proto_method_implemented(proto->sendmsg),
2770 proto_method_implemented(proto->recvmsg),
2771 proto_method_implemented(proto->sendpage),
2772 proto_method_implemented(proto->bind),
2773 proto_method_implemented(proto->backlog_rcv),
2774 proto_method_implemented(proto->hash),
2775 proto_method_implemented(proto->unhash),
2776 proto_method_implemented(proto->get_port),
2777 proto_method_implemented(proto->enter_memory_pressure));
2778}
2779
2780static int proto_seq_show(struct seq_file *seq, void *v)
2781{
60f0438a 2782 if (v == &proto_list)
1da177e4
LT		 2783 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2784 "protocol",
2785 "size",
2786 "sockets",
2787 "memory",
2788 "press",
2789 "maxhdr",
2790 "slab",
2791 "module",
2792 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2793 else
60f0438a 2794 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2795 return 0;
2796}
2797
f690808e 2798static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2799 .start = proto_seq_start,
2800 .next = proto_seq_next,
2801 .stop = proto_seq_stop,
2802 .show = proto_seq_show,
2803};
2804
2805static int proto_seq_open(struct inode *inode, struct file *file)
2806{
14e943db
ED		 2807 return seq_open_net(inode, file, &proto_seq_ops,
2808 sizeof(struct seq_net_private));
1da177e4
LT		 2809}
2810
9a32144e 2811static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2812 .owner = THIS_MODULE,
2813 .open = proto_seq_open,
2814 .read = seq_read,
2815 .llseek = seq_lseek,
14e943db
ED		 2816 .release = seq_release_net,
2817};
2818
2819static __net_init int proto_init_net(struct net *net)
2820{
2821 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2822 return -ENOMEM;
2823
2824 return 0;
2825}
2826
2827static __net_exit void proto_exit_net(struct net *net)
2828{
2829 proc_net_remove(net, "protocols");
2830}
2831
2832
2833static __net_initdata struct pernet_operations proto_net_ops = {
2834 .init = proto_init_net,
2835 .exit = proto_exit_net,
1da177e4
LT		 2836};
2837
2838static int __init proto_init(void)
2839{
14e943db 2840 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2841}
2842
2843subsys_initcall(proto_init);
2844
2845#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 0.942156 seconds and 5 git commands to generate.