projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
cpts: add missing kconfig dependency
[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
4878809f
DM		 508static int sock_bindtodevice(struct sock *sk, char __user *optval, int optlen)
509{
510 int ret = -ENOPROTOOPT;
511#ifdef CONFIG_NETDEVICES
3b1e0a65 512 struct net *net = sock_net(sk);
4878809f
DM		 513 char devname[IFNAMSIZ];
514 int index;
515
516 /* Sorry... */
517 ret = -EPERM;
5e1fccc0 518 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 519 goto out;
520
521 ret = -EINVAL;
522 if (optlen < 0)
523 goto out;
524
525 /* Bind this socket to a particular device like "eth0",
526 * as specified in the passed interface name. If the
527 * name is "" or the option length is zero the socket
528 * is not bound.
529 */
530 if (optlen > IFNAMSIZ - 1)
531 optlen = IFNAMSIZ - 1;
532 memset(devname, 0, sizeof(devname));
533
534 ret = -EFAULT;
535 if (copy_from_user(devname, optval, optlen))
536 goto out;
537
000ba2e4
DM		 538 index = 0;
539 if (devname[0] != '\0') {
bf8e56bf 540 struct net_device *dev;
4878809f 541
bf8e56bf
ED		 542 rcu_read_lock();
543 dev = dev_get_by_name_rcu(net, devname);
544 if (dev)
545 index = dev->ifindex;
546 rcu_read_unlock();
4878809f
DM		 547 ret = -ENODEV;
548 if (!dev)
549 goto out;
4878809f
DM		 550 }
551
552 lock_sock(sk);
553 sk->sk_bound_dev_if = index;
554 sk_dst_reset(sk);
555 release_sock(sk);
556
557 ret = 0;
558
559out:
560#endif
561
562 return ret;
563}
564
c0ef877b
PE		 565static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
566{
567 if (valbool)
568 sock_set_flag(sk, bit);
569 else
570 sock_reset_flag(sk, bit);
571}
572
1da177e4
LT		 573/*
574 * This is meant for all protocols to use and covers goings on
575 * at the socket level. Everything here is generic.
576 */
577
578int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 579 char __user *optval, unsigned int optlen)
1da177e4 580{
2a91525c 581 struct sock *sk = sock->sk;
1da177e4
LT		 582 int val;
583 int valbool;
584 struct linger ling;
585 int ret = 0;
4ec93edb 586
1da177e4
LT		 587 /*
588 * Options without arguments
589 */
590
4878809f
DM		 591 if (optname == SO_BINDTODEVICE)
592 return sock_bindtodevice(sk, optval, optlen);
593
e71a4783
SH		 594 if (optlen < sizeof(int))
595 return -EINVAL;
4ec93edb 596
1da177e4
LT		 597 if (get_user(val, (int __user *)optval))
598 return -EFAULT;
4ec93edb 599
2a91525c 600 valbool = val ? 1 : 0;
1da177e4
LT		 601
602 lock_sock(sk);
603
2a91525c 604 switch (optname) {
e71a4783 605 case SO_DEBUG:
2a91525c 606 if (val && !capable(CAP_NET_ADMIN))
e71a4783 607 ret = -EACCES;
2a91525c 608 else
c0ef877b 609 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 610 break;
611 case SO_REUSEADDR:
4a17fd52 612 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783
SH		 613 break;
614 case SO_TYPE:
49c794e9 615 case SO_PROTOCOL:
0d6038ee 616 case SO_DOMAIN:
e71a4783
SH		 617 case SO_ERROR:
618 ret = -ENOPROTOOPT;
619 break;
620 case SO_DONTROUTE:
c0ef877b 621 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 622 break;
623 case SO_BROADCAST:
624 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
625 break;
626 case SO_SNDBUF:
627 /* Don't error on this BSD doesn't and if you think
82981930
ED		 628 * about it this is right. Otherwise apps have to
629 * play 'guess the biggest size' games. RCVBUF/SNDBUF
630 * are treated in BSD as hints
631 */
632 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 633set_sndbuf:
e71a4783 634 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 635 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
636 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 637 sk->sk_write_space(sk);
638 break;
1da177e4 639
e71a4783
SH		 640 case SO_SNDBUFFORCE:
641 if (!capable(CAP_NET_ADMIN)) {
642 ret = -EPERM;
643 break;
644 }
645 goto set_sndbuf;
b0573dea 646
e71a4783
SH		 647 case SO_RCVBUF:
648 /* Don't error on this BSD doesn't and if you think
82981930
ED		 649 * about it this is right. Otherwise apps have to
650 * play 'guess the biggest size' games. RCVBUF/SNDBUF
651 * are treated in BSD as hints
652 */
653 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 654set_rcvbuf:
e71a4783
SH		 655 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
656 /*
657 * We double it on the way in to account for
658 * "struct sk_buff" etc. overhead. Applications
659 * assume that the SO_RCVBUF setting they make will
660 * allow that much actual data to be received on that
661 * socket.
662 *
663 * Applications are unaware that "struct sk_buff" and
664 * other overheads allocate from the receive buffer
665 * during socket buffer allocation.
666 *
667 * And after considering the possible alternatives,
668 * returning the value we actually used in getsockopt
669 * is the most desirable behavior.
670 */
82981930 671 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 672 break;
673
674 case SO_RCVBUFFORCE:
675 if (!capable(CAP_NET_ADMIN)) {
676 ret = -EPERM;
1da177e4 677 break;
e71a4783
SH		 678 }
679 goto set_rcvbuf;
1da177e4 680
e71a4783 681 case SO_KEEPALIVE:
1da177e4 682#ifdef CONFIG_INET
3e10986d
ED		 683 if (sk->sk_protocol == IPPROTO_TCP &&
684 sk->sk_type == SOCK_STREAM)
e71a4783 685 tcp_set_keepalive(sk, valbool);
1da177e4 686#endif
e71a4783
SH		 687 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
688 break;
689
690 case SO_OOBINLINE:
691 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
692 break;
693
694 case SO_NO_CHECK:
695 sk->sk_no_check = valbool;
696 break;
697
698 case SO_PRIORITY:
5e1fccc0
EB		 699 if ((val >= 0 && val <= 6) ||
700 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 701 sk->sk_priority = val;
702 else
703 ret = -EPERM;
704 break;
705
706 case SO_LINGER:
707 if (optlen < sizeof(ling)) {
708 ret = -EINVAL; /* 1003.1g */
1da177e4 709 break;
e71a4783 710 }
2a91525c 711 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 712 ret = -EFAULT;
1da177e4 713 break;
e71a4783
SH		 714 }
715 if (!ling.l_onoff)
716 sock_reset_flag(sk, SOCK_LINGER);
717 else {
1da177e4 718#if (BITS_PER_LONG == 32)
e71a4783
SH		 719 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
720 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 721 else
e71a4783
SH		 722#endif
723 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
724 sock_set_flag(sk, SOCK_LINGER);
725 }
726 break;
727
728 case SO_BSDCOMPAT:
729 sock_warn_obsolete_bsdism("setsockopt");
730 break;
731
732 case SO_PASSCRED:
733 if (valbool)
734 set_bit(SOCK_PASSCRED, &sock->flags);
735 else
736 clear_bit(SOCK_PASSCRED, &sock->flags);
737 break;
738
739 case SO_TIMESTAMP:
92f37fd2 740 case SO_TIMESTAMPNS:
e71a4783 741 if (valbool) {
92f37fd2
ED		 742 if (optname == SO_TIMESTAMP)
743 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
744 else
745 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 746 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 747 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 748 } else {
e71a4783 749 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 750 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
751 }
e71a4783
SH		 752 break;
753
20d49473
PO		 754 case SO_TIMESTAMPING:
755 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 756 ret = -EINVAL;
20d49473
PO		 757 break;
758 }
759 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
760 val & SOF_TIMESTAMPING_TX_HARDWARE);
761 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
762 val & SOF_TIMESTAMPING_TX_SOFTWARE);
763 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
764 val & SOF_TIMESTAMPING_RX_HARDWARE);
765 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
766 sock_enable_timestamp(sk,
767 SOCK_TIMESTAMPING_RX_SOFTWARE);
768 else
769 sock_disable_timestamp(sk,
08e29af3 770 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 771 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
772 val & SOF_TIMESTAMPING_SOFTWARE);
773 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
774 val & SOF_TIMESTAMPING_SYS_HARDWARE);
775 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
776 val & SOF_TIMESTAMPING_RAW_HARDWARE);
777 break;
778
e71a4783
SH		 779 case SO_RCVLOWAT:
780 if (val < 0)
781 val = INT_MAX;
782 sk->sk_rcvlowat = val ? : 1;
783 break;
784
785 case SO_RCVTIMEO:
786 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
787 break;
788
789 case SO_SNDTIMEO:
790 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
791 break;
1da177e4 792
e71a4783
SH		 793 case SO_ATTACH_FILTER:
794 ret = -EINVAL;
795 if (optlen == sizeof(struct sock_fprog)) {
796 struct sock_fprog fprog;
1da177e4 797
e71a4783
SH		 798 ret = -EFAULT;
799 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 800 break;
e71a4783
SH		 801
802 ret = sk_attach_filter(&fprog, sk);
803 }
804 break;
805
806 case SO_DETACH_FILTER:
55b33325 807 ret = sk_detach_filter(sk);
e71a4783 808 break;
1da177e4 809
e71a4783
SH		 810 case SO_PASSSEC:
811 if (valbool)
812 set_bit(SOCK_PASSSEC, &sock->flags);
813 else
814 clear_bit(SOCK_PASSSEC, &sock->flags);
815 break;
4a19ec58 816 case SO_MARK:
5e1fccc0 817 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 818 ret = -EPERM;
2a91525c 819 else
4a19ec58 820 sk->sk_mark = val;
4a19ec58 821 break;
877ce7c1 822
1da177e4
LT		 823 /* We implement the SO_SNDLOWAT etc to
824 not be settable (1003.1g 5.3) */
3b885787 825 case SO_RXQ_OVFL:
8083f0fc 826 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 827 break;
6e3e939f
JB		 828
829 case SO_WIFI_STATUS:
830 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
831 break;
832
ef64a54f
PE		 833 case SO_PEEK_OFF:
834 if (sock->ops->set_peek_off)
835 sock->ops->set_peek_off(sk, val);
836 else
837 ret = -EOPNOTSUPP;
838 break;
3bdc0eba
BG		 839
840 case SO_NOFCS:
841 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
842 break;
843
e71a4783
SH		 844 default:
845 ret = -ENOPROTOOPT;
846 break;
4ec93edb 847 }
1da177e4
LT		 848 release_sock(sk);
849 return ret;
850}
2a91525c 851EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 852
853
3f551f94
EB		 854void cred_to_ucred(struct pid *pid, const struct cred *cred,
855 struct ucred *ucred)
856{
857 ucred->pid = pid_vnr(pid);
858 ucred->uid = ucred->gid = -1;
859 if (cred) {
860 struct user_namespace *current_ns = current_user_ns();
861
b2e4f544
EB		 862 ucred->uid = from_kuid_munged(current_ns, cred->euid);
863 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 864 }
865}
3924773a 866EXPORT_SYMBOL_GPL(cred_to_ucred);
3f551f94 867
1da177e4
LT		 868int sock_getsockopt(struct socket *sock, int level, int optname,
869 char __user *optval, int __user *optlen)
870{
871 struct sock *sk = sock->sk;
4ec93edb 872
e71a4783 873 union {
4ec93edb
YH		 874 int val;
875 struct linger ling;
1da177e4
LT		 876 struct timeval tm;
877 } v;
4ec93edb 878
4d0392be 879 int lv = sizeof(int);
1da177e4 880 int len;
4ec93edb 881
e71a4783 882 if (get_user(len, optlen))
4ec93edb 883 return -EFAULT;
e71a4783 884 if (len < 0)
1da177e4 885 return -EINVAL;
4ec93edb 886
50fee1de 887 memset(&v, 0, sizeof(v));
df0bca04 888
2a91525c 889 switch (optname) {
e71a4783
SH		 890 case SO_DEBUG:
891 v.val = sock_flag(sk, SOCK_DBG);
892 break;
893
894 case SO_DONTROUTE:
895 v.val = sock_flag(sk, SOCK_LOCALROUTE);
896 break;
897
898 case SO_BROADCAST:
1b23a5df 899 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 900 break;
901
902 case SO_SNDBUF:
903 v.val = sk->sk_sndbuf;
904 break;
905
906 case SO_RCVBUF:
907 v.val = sk->sk_rcvbuf;
908 break;
909
910 case SO_REUSEADDR:
911 v.val = sk->sk_reuse;
912 break;
913
914 case SO_KEEPALIVE:
1b23a5df 915 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 916 break;
917
918 case SO_TYPE:
919 v.val = sk->sk_type;
920 break;
921
49c794e9
JE		 922 case SO_PROTOCOL:
923 v.val = sk->sk_protocol;
924 break;
925
0d6038ee
JE		 926 case SO_DOMAIN:
927 v.val = sk->sk_family;
928 break;
929
e71a4783
SH		 930 case SO_ERROR:
931 v.val = -sock_error(sk);
2a91525c 932 if (v.val == 0)
e71a4783
SH		 933 v.val = xchg(&sk->sk_err_soft, 0);
934 break;
935
936 case SO_OOBINLINE:
1b23a5df 937 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 938 break;
939
940 case SO_NO_CHECK:
941 v.val = sk->sk_no_check;
942 break;
943
944 case SO_PRIORITY:
945 v.val = sk->sk_priority;
946 break;
947
948 case SO_LINGER:
949 lv = sizeof(v.ling);
1b23a5df 950 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 951 v.ling.l_linger = sk->sk_lingertime / HZ;
952 break;
953
954 case SO_BSDCOMPAT:
955 sock_warn_obsolete_bsdism("getsockopt");
956 break;
957
958 case SO_TIMESTAMP:
92f37fd2
ED		 959 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
960 !sock_flag(sk, SOCK_RCVTSTAMPNS);
961 break;
962
963 case SO_TIMESTAMPNS:
964 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 965 break;
966
20d49473
PO		 967 case SO_TIMESTAMPING:
968 v.val = 0;
969 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
970 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
971 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
972 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
973 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
974 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
975 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
976 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
977 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
978 v.val |= SOF_TIMESTAMPING_SOFTWARE;
979 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
980 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
981 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
982 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
983 break;
984
e71a4783 985 case SO_RCVTIMEO:
2a91525c 986 lv = sizeof(struct timeval);
e71a4783
SH		 987 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
988 v.tm.tv_sec = 0;
989 v.tm.tv_usec = 0;
990 } else {
991 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
992 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
993 }
994 break;
995
996 case SO_SNDTIMEO:
2a91525c 997 lv = sizeof(struct timeval);
e71a4783
SH		 998 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
999 v.tm.tv_sec = 0;
1000 v.tm.tv_usec = 0;
1001 } else {
1002 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1003 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1004 }
1005 break;
1da177e4 1006
e71a4783
SH		 1007 case SO_RCVLOWAT:
1008 v.val = sk->sk_rcvlowat;
1009 break;
1da177e4 1010
e71a4783 1011 case SO_SNDLOWAT:
2a91525c 1012 v.val = 1;
e71a4783 1013 break;
1da177e4 1014
e71a4783 1015 case SO_PASSCRED:
82981930 1016 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1017 break;
1da177e4 1018
e71a4783 1019 case SO_PEERCRED:
109f6e39
EB		 1020 {
1021 struct ucred peercred;
1022 if (len > sizeof(peercred))
1023 len = sizeof(peercred);
1024 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1025 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1026 return -EFAULT;
1027 goto lenout;
109f6e39 1028 }
1da177e4 1029
e71a4783
SH		 1030 case SO_PEERNAME:
1031 {
1032 char address[128];
1033
1034 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1035 return -ENOTCONN;
1036 if (lv < len)
1037 return -EINVAL;
1038 if (copy_to_user(optval, address, len))
1039 return -EFAULT;
1040 goto lenout;
1041 }
1da177e4 1042
e71a4783
SH		 1043 /* Dubious BSD thing... Probably nobody even uses it, but
1044 * the UNIX standard wants it for whatever reason... -DaveM
1045 */
1046 case SO_ACCEPTCONN:
1047 v.val = sk->sk_state == TCP_LISTEN;
1048 break;
1da177e4 1049
e71a4783 1050 case SO_PASSSEC:
82981930 1051 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1052 break;
877ce7c1 1053
e71a4783
SH		 1054 case SO_PEERSEC:
1055 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1056
4a19ec58
LAT		 1057 case SO_MARK:
1058 v.val = sk->sk_mark;
1059 break;
1060
3b885787 1061 case SO_RXQ_OVFL:
1b23a5df 1062 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1063 break;
1064
6e3e939f 1065 case SO_WIFI_STATUS:
1b23a5df 1066 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1067 break;
1068
ef64a54f
PE		 1069 case SO_PEEK_OFF:
1070 if (!sock->ops->set_peek_off)
1071 return -EOPNOTSUPP;
1072
1073 v.val = sk->sk_peek_off;
1074 break;
bc2f7996 1075 case SO_NOFCS:
1b23a5df 1076 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1077 break;
f7b86bfe
PE		 1078 case SO_BINDTODEVICE:
1079 v.val = sk->sk_bound_dev_if;
1080 break;
a8fc9277
PE		 1081 case SO_GET_FILTER:
1082 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1083 if (len < 0)
1084 return len;
1085
1086 goto lenout;
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 1225#ifdef CONFIG_CGROUPS
8fb974c9 1226#if IS_ENABLED(CONFIG_NET_CLS_CGROUP)
fd9a08a7 1227void sock_update_classid(struct sock *sk, struct task_struct *task)
f8451725 1228{
1144182a 1229 u32 classid;
f8451725 1230
fd9a08a7 1231 classid = task_cls_classid(task);
3afa6d00 1232 if (classid != sk->sk_classid)
f8451725
HX		 1233 sk->sk_classid = classid;
1234}
82862742 1235EXPORT_SYMBOL(sock_update_classid);
8fb974c9 1236#endif
5bc1421e 1237
51e4e7fa 1238#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
406a3c63 1239void sock_update_netprioidx(struct sock *sk, struct task_struct *task)
5bc1421e 1240{
5bc1421e
NH		 1241 if (in_interrupt())
1242 return;
2b73bc65 1243
406a3c63 1244 sk->sk_cgrp_prioidx = task_netprioidx(task);
5bc1421e
NH		 1245}
1246EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725 1247#endif
51e4e7fa 1248#endif
f8451725 1249
1da177e4
LT		 1250/**
1251 * sk_alloc - All socket objects are allocated here
c4ea43c5 1252 * @net: the applicable net namespace
4dc3b16b
PP		 1253 * @family: protocol family
1254 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1255 * @prot: struct proto associated with this new sock instance
1da177e4 1256 */
1b8d7ae4 1257struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1258 struct proto *prot)
1da177e4 1259{
c308c1b2 1260 struct sock *sk;
1da177e4 1261
154adbc8 1262 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1263 if (sk) {
154adbc8
PE		 1264 sk->sk_family = family;
1265 /*
1266 * See comment in struct sock definition to understand
1267 * why we need sk_prot_creator -acme
1268 */
1269 sk->sk_prot = sk->sk_prot_creator = prot;
1270 sock_lock_init(sk);
3b1e0a65 1271 sock_net_set(sk, get_net(net));
d66ee058 1272 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1273
fd9a08a7 1274 sock_update_classid(sk, current);
406a3c63 1275 sock_update_netprioidx(sk, current);
1da177e4 1276 }
a79af59e 1277
2e4afe7b 1278 return sk;
1da177e4 1279}
2a91525c 1280EXPORT_SYMBOL(sk_alloc);
1da177e4 1281
2b85a34e 1282static void __sk_free(struct sock *sk)
1da177e4
LT		 1283{
1284 struct sk_filter *filter;
1da177e4
LT		 1285
1286 if (sk->sk_destruct)
1287 sk->sk_destruct(sk);
1288
a898def2
PM		 1289 filter = rcu_dereference_check(sk->sk_filter,
1290 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1291 if (filter) {
309dd5fc 1292 sk_filter_uncharge(sk, filter);
a9b3cd7f 1293 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1294 }
1295
08e29af3 1296 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1297
1298 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1299 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1300 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1301
109f6e39
EB		 1302 if (sk->sk_peer_cred)
1303 put_cred(sk->sk_peer_cred);
1304 put_pid(sk->sk_peer_pid);
3b1e0a65 1305 put_net(sock_net(sk));
c308c1b2 1306 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1307}
2b85a34e
ED		 1308
1309void sk_free(struct sock *sk)
1310{
1311 /*
25985edc 1312 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1313 * some packets are still in some tx queue.
1314 * If not null, sock_wfree() will call __sk_free(sk) later
1315 */
1316 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1317 __sk_free(sk);
1318}
2a91525c 1319EXPORT_SYMBOL(sk_free);
1da177e4 1320
edf02087 1321/*
25985edc
LDM		 1322 * Last sock_put should drop reference to sk->sk_net. It has already
1323 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1324 * is not an option.
25985edc 1325 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1326 * destroy it in the context of init_net.
1327 */
1328void sk_release_kernel(struct sock *sk)
1329{
1330 if (sk == NULL || sk->sk_socket == NULL)
1331 return;
1332
1333 sock_hold(sk);
1334 sock_release(sk->sk_socket);
65a18ec5 1335 release_net(sock_net(sk));
3b1e0a65 1336 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1337 sock_put(sk);
1338}
45af1754 1339EXPORT_SYMBOL(sk_release_kernel);
edf02087 1340
475f1b52
SR		 1341static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1342{
1343 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1344 sock_update_memcg(newsk);
1345}
1346
e56c57d0
ED		 1347/**
1348 * sk_clone_lock - clone a socket, and lock its clone
1349 * @sk: the socket to clone
1350 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1351 *
1352 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1353 */
1354struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1355{
8fd1d178 1356 struct sock *newsk;
87d11ceb 1357
8fd1d178 1358 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1359 if (newsk != NULL) {
1360 struct sk_filter *filter;
1361
892c141e 1362 sock_copy(newsk, sk);
87d11ceb
ACM		 1363
1364 /* SANITY */
3b1e0a65 1365 get_net(sock_net(newsk));
87d11ceb
ACM		 1366 sk_node_init(&newsk->sk_node);
1367 sock_lock_init(newsk);
1368 bh_lock_sock(newsk);
fa438ccf 1369 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1370 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1371
1372 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1373 /*
1374 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1375 */
1376 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1377 atomic_set(&newsk->sk_omem_alloc, 0);
1378 skb_queue_head_init(&newsk->sk_receive_queue);
1379 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1380#ifdef CONFIG_NET_DMA
1381 skb_queue_head_init(&newsk->sk_async_wait_queue);
1382#endif
87d11ceb 1383
b6c6712a 1384 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1385 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1386 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1387 af_callback_keys + newsk->sk_family,
1388 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1389
1390 newsk->sk_dst_cache = NULL;
1391 newsk->sk_wmem_queued = 0;
1392 newsk->sk_forward_alloc = 0;
1393 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1394 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1395
1396 sock_reset_flag(newsk, SOCK_DONE);
1397 skb_queue_head_init(&newsk->sk_error_queue);
1398
0d7da9dd 1399 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1400 if (filter != NULL)
1401 sk_filter_charge(newsk, filter);
1402
1403 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1404 /* It is still raw copy of parent, so invalidate
1405 * destructor and make plain sk_free() */
1406 newsk->sk_destruct = NULL;
b0691c8e 1407 bh_unlock_sock(newsk);
87d11ceb
ACM		 1408 sk_free(newsk);
1409 newsk = NULL;
1410 goto out;
1411 }
1412
1413 newsk->sk_err = 0;
1414 newsk->sk_priority = 0;
4dc6dc71
ED		 1415 /*
1416 * Before updating sk_refcnt, we must commit prior changes to memory
1417 * (Documentation/RCU/rculist_nulls.txt for details)
1418 */
1419 smp_wmb();
87d11ceb
ACM		 1420 atomic_set(&newsk->sk_refcnt, 2);
1421
1422 /*
1423 * Increment the counter in the same struct proto as the master
1424 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1425 * is the same as sk->sk_prot->socks, as this field was copied
1426 * with memcpy).
1427 *
1428 * This _changes_ the previous behaviour, where
1429 * tcp_create_openreq_child always was incrementing the
1430 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1431 * to be taken into account in all callers. -acme
1432 */
1433 sk_refcnt_debug_inc(newsk);
972692e0 1434 sk_set_socket(newsk, NULL);
43815482 1435 newsk->sk_wq = NULL;
87d11ceb 1436
f3f511e1
GC		 1437 sk_update_clone(sk, newsk);
1438
87d11ceb 1439 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1440 sk_sockets_allocated_inc(newsk);
704da560 1441
08e29af3 1442 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1443 net_enable_timestamp();
87d11ceb
ACM		 1444 }
1445out:
1446 return newsk;
1447}
e56c57d0 1448EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1449
9958089a
AK		 1450void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1451{
1452 __sk_dst_set(sk, dst);
1453 sk->sk_route_caps = dst->dev->features;
1454 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1455 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1456 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1457 if (sk_can_gso(sk)) {
82cc1a7a 1458 if (dst->header_len) {
9958089a 1459 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1460 } else {
9958089a 1461 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1462 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1463 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1464 }
9958089a
AK		 1465 }
1466}
1467EXPORT_SYMBOL_GPL(sk_setup_caps);
1468
1da177e4
LT		 1469/*
1470 * Simple resource managers for sockets.
1471 */
1472
1473
4ec93edb
YH		 1474/*
1475 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1476 */
1477void sock_wfree(struct sk_buff *skb)
1478{
1479 struct sock *sk = skb->sk;
d99927f4 1480 unsigned int len = skb->truesize;
1da177e4 1481
d99927f4
ED		 1482 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1483 /*
1484 * Keep a reference on sk_wmem_alloc, this will be released
1485 * after sk_write_space() call
1486 */
1487 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1488 sk->sk_write_space(sk);
d99927f4
ED		 1489 len = 1;
1490 }
2b85a34e 1491 /*
d99927f4
ED		 1492 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1493 * could not do because of in-flight packets
2b85a34e 1494 */
d99927f4 1495 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1496 __sk_free(sk);
1da177e4 1497}
2a91525c 1498EXPORT_SYMBOL(sock_wfree);
1da177e4 1499
4ec93edb
YH		 1500/*
1501 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1502 */
1503void sock_rfree(struct sk_buff *skb)
1504{
1505 struct sock *sk = skb->sk;
d361fd59 1506 unsigned int len = skb->truesize;
1da177e4 1507
d361fd59
ED		 1508 atomic_sub(len, &sk->sk_rmem_alloc);
1509 sk_mem_uncharge(sk, len);
1da177e4 1510}
2a91525c 1511EXPORT_SYMBOL(sock_rfree);
1da177e4 1512
41063e9d
DM		 1513void sock_edemux(struct sk_buff *skb)
1514{
e812347c
ED		 1515 struct sock *sk = skb->sk;
1516
1c463e57 1517#ifdef CONFIG_INET
e812347c
ED		 1518 if (sk->sk_state == TCP_TIME_WAIT)
1519 inet_twsk_put(inet_twsk(sk));
1520 else
1c463e57 1521#endif
e812347c 1522 sock_put(sk);
41063e9d
DM		 1523}
1524EXPORT_SYMBOL(sock_edemux);
1da177e4 1525
976d0201 1526kuid_t sock_i_uid(struct sock *sk)
1da177e4 1527{
976d0201 1528 kuid_t uid;
1da177e4 1529
f064af1e 1530 read_lock_bh(&sk->sk_callback_lock);
976d0201 1531 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1532 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1533 return uid;
1534}
2a91525c 1535EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1536
1537unsigned long sock_i_ino(struct sock *sk)
1538{
1539 unsigned long ino;
1540
f064af1e 1541 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1542 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1543 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1544 return ino;
1545}
2a91525c 1546EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1547
1548/*
1549 * Allocate a skb from the socket's send buffer.
1550 */
86a76caf 1551struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1552 gfp_t priority)
1da177e4
LT		 1553{
1554 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1555 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1556 if (skb) {
1557 skb_set_owner_w(skb, sk);
1558 return skb;
1559 }
1560 }
1561 return NULL;
1562}
2a91525c 1563EXPORT_SYMBOL(sock_wmalloc);
1da177e4
LT		 1564
1565/*
1566 * Allocate a skb from the socket's receive buffer.
4ec93edb 1567 */
86a76caf 1568struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1569 gfp_t priority)
1da177e4
LT		 1570{
1571 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1572 struct sk_buff *skb = alloc_skb(size, priority);
1573 if (skb) {
1574 skb_set_owner_r(skb, sk);
1575 return skb;
1576 }
1577 }
1578 return NULL;
1579}
1580
4ec93edb 1581/*
1da177e4 1582 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1583 */
dd0fc66f 1584void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1585{
95c96174 1586 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1587 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1588 void *mem;
1589 /* First do the add, to avoid the race if kmalloc
4ec93edb 1590 * might sleep.
1da177e4
LT		 1591 */
1592 atomic_add(size, &sk->sk_omem_alloc);
1593 mem = kmalloc(size, priority);
1594 if (mem)
1595 return mem;
1596 atomic_sub(size, &sk->sk_omem_alloc);
1597 }
1598 return NULL;
1599}
2a91525c 1600EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1601
1602/*
1603 * Free an option memory block.
1604 */
1605void sock_kfree_s(struct sock *sk, void *mem, int size)
1606{
1607 kfree(mem);
1608 atomic_sub(size, &sk->sk_omem_alloc);
1609}
2a91525c 1610EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1611
1612/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1613 I think, these locks should be removed for datagram sockets.
1614 */
2a91525c 1615static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1616{
1617 DEFINE_WAIT(wait);
1618
1619 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1620 for (;;) {
1621 if (!timeo)
1622 break;
1623 if (signal_pending(current))
1624 break;
1625 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1626 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1627 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1628 break;
1629 if (sk->sk_shutdown & SEND_SHUTDOWN)
1630 break;
1631 if (sk->sk_err)
1632 break;
1633 timeo = schedule_timeout(timeo);
1634 }
aa395145 1635 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1636 return timeo;
1637}
1638
1639
1640/*
1641 * Generic send/receive buffer handlers
1642 */
1643
4cc7f68d
HX		 1644struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1645 unsigned long data_len, int noblock,
1646 int *errcode)
1da177e4
LT		 1647{
1648 struct sk_buff *skb;
7d877f3b 1649 gfp_t gfp_mask;
1da177e4
LT		 1650 long timeo;
1651 int err;
cc9b17ad
JW		 1652 int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1653
1654 err = -EMSGSIZE;
1655 if (npages > MAX_SKB_FRAGS)
1656 goto failure;
1da177e4
LT		 1657
1658 gfp_mask = sk->sk_allocation;
1659 if (gfp_mask & __GFP_WAIT)
1660 gfp_mask |= __GFP_REPEAT;
1661
1662 timeo = sock_sndtimeo(sk, noblock);
1663 while (1) {
1664 err = sock_error(sk);
1665 if (err != 0)
1666 goto failure;
1667
1668 err = -EPIPE;
1669 if (sk->sk_shutdown & SEND_SHUTDOWN)
1670 goto failure;
1671
1672 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1673 skb = alloc_skb(header_len, gfp_mask);
1da177e4 1674 if (skb) {
1da177e4
LT		 1675 int i;
1676
1677 /* No pages, we're done... */
1678 if (!data_len)
1679 break;
1680
1da177e4
LT		 1681 skb->truesize += data_len;
1682 skb_shinfo(skb)->nr_frags = npages;
1683 for (i = 0; i < npages; i++) {
1684 struct page *page;
1da177e4
LT		 1685
1686 page = alloc_pages(sk->sk_allocation, 0);
1687 if (!page) {
1688 err = -ENOBUFS;
1689 skb_shinfo(skb)->nr_frags = i;
1690 kfree_skb(skb);
1691 goto failure;
1692 }
1693
ea2ab693
IC		 1694 __skb_fill_page_desc(skb, i,
1695 page, 0,
1696 (data_len >= PAGE_SIZE ?
1697 PAGE_SIZE :
1698 data_len));
1da177e4
LT		 1699 data_len -= PAGE_SIZE;
1700 }
1701
1702 /* Full success... */
1703 break;
1704 }
1705 err = -ENOBUFS;
1706 goto failure;
1707 }
1708 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1709 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1710 err = -EAGAIN;
1711 if (!timeo)
1712 goto failure;
1713 if (signal_pending(current))
1714 goto interrupted;
1715 timeo = sock_wait_for_wmem(sk, timeo);
1716 }
1717
1718 skb_set_owner_w(skb, sk);
1719 return skb;
1720
1721interrupted:
1722 err = sock_intr_errno(timeo);
1723failure:
1724 *errcode = err;
1725 return NULL;
1726}
4cc7f68d 1727EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1728
4ec93edb 1729struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1730 int noblock, int *errcode)
1731{
1732 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1733}
2a91525c 1734EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1735
5640f768
ED		 1736/* On 32bit arches, an skb frag is limited to 2^15 */
1737#define SKB_FRAG_PAGE_ORDER get_order(32768)
1738
1739bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1740{
1741 int order;
1742
1743 if (pfrag->page) {
1744 if (atomic_read(&pfrag->page->_count) == 1) {
1745 pfrag->offset = 0;
1746 return true;
1747 }
1748 if (pfrag->offset < pfrag->size)
1749 return true;
1750 put_page(pfrag->page);
1751 }
1752
1753 /* We restrict high order allocations to users that can afford to wait */
1754 order = (sk->sk_allocation & __GFP_WAIT) ? SKB_FRAG_PAGE_ORDER : 0;
1755
1756 do {
1757 gfp_t gfp = sk->sk_allocation;
1758
1759 if (order)
1760 gfp |= __GFP_COMP | __GFP_NOWARN;
1761 pfrag->page = alloc_pages(gfp, order);
1762 if (likely(pfrag->page)) {
1763 pfrag->offset = 0;
1764 pfrag->size = PAGE_SIZE << order;
1765 return true;
1766 }
1767 } while (--order >= 0);
1768
1769 sk_enter_memory_pressure(sk);
1770 sk_stream_moderate_sndbuf(sk);
1771 return false;
1772}
1773EXPORT_SYMBOL(sk_page_frag_refill);
1774
1da177e4 1775static void __lock_sock(struct sock *sk)
f39234d6
NK		 1776 __releases(&sk->sk_lock.slock)
1777 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1778{
1779 DEFINE_WAIT(wait);
1780
e71a4783 1781 for (;;) {
1da177e4
LT		 1782 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1783 TASK_UNINTERRUPTIBLE);
1784 spin_unlock_bh(&sk->sk_lock.slock);
1785 schedule();
1786 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1787 if (!sock_owned_by_user(sk))
1da177e4
LT		 1788 break;
1789 }
1790 finish_wait(&sk->sk_lock.wq, &wait);
1791}
1792
1793static void __release_sock(struct sock *sk)
f39234d6
NK		 1794 __releases(&sk->sk_lock.slock)
1795 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1796{
1797 struct sk_buff *skb = sk->sk_backlog.head;
1798
1799 do {
1800 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1801 bh_unlock_sock(sk);
1802
1803 do {
1804 struct sk_buff *next = skb->next;
1805
e4cbb02a 1806 prefetch(next);
7fee226a 1807 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1808 skb->next = NULL;
c57943a1 1809 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1810
1811 /*
1812 * We are in process context here with softirqs
1813 * disabled, use cond_resched_softirq() to preempt.
1814 * This is safe to do because we've taken the backlog
1815 * queue private:
1816 */
1817 cond_resched_softirq();
1818
1819 skb = next;
1820 } while (skb != NULL);
1821
1822 bh_lock_sock(sk);
e71a4783 1823 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1824
1825 /*
1826 * Doing the zeroing here guarantee we can not loop forever
1827 * while a wild producer attempts to flood us.
1828 */
1829 sk->sk_backlog.len = 0;
1da177e4
LT		 1830}
1831
1832/**
1833 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1834 * @sk: sock to wait on
1835 * @timeo: for how long
1da177e4
LT		 1836 *
1837 * Now socket state including sk->sk_err is changed only under lock,
1838 * hence we may omit checks after joining wait queue.
1839 * We check receive queue before schedule() only as optimization;
1840 * it is very likely that release_sock() added new data.
1841 */
1842int sk_wait_data(struct sock *sk, long *timeo)
1843{
1844 int rc;
1845 DEFINE_WAIT(wait);
1846
aa395145 1847 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1848 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1849 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1850 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1851 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1852 return rc;
1853}
1da177e4
LT		 1854EXPORT_SYMBOL(sk_wait_data);
1855
3ab224be
HA		 1856/**
1857 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1858 * @sk: socket
1859 * @size: memory size to allocate
1860 * @kind: allocation type
1861 *
1862 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1863 * rmem allocation. This function assumes that protocols which have
1864 * memory_pressure use sk_wmem_queued as write buffer accounting.
1865 */
1866int __sk_mem_schedule(struct sock *sk, int size, int kind)
1867{
1868 struct proto *prot = sk->sk_prot;
1869 int amt = sk_mem_pages(size);
8d987e5c 1870 long allocated;
e1aab161 1871 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1872
1873 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1874
e1aab161 1875 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1876
1877 /* Under limit. */
e1aab161
GC		 1878 if (parent_status == UNDER_LIMIT &&
1879 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1880 sk_leave_memory_pressure(sk);
3ab224be
HA		 1881 return 1;
1882 }
1883
e1aab161
GC		 1884 /* Under pressure. (we or our parents) */
1885 if ((parent_status > SOFT_LIMIT) ||
1886 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1887 sk_enter_memory_pressure(sk);
3ab224be 1888
e1aab161
GC		 1889 /* Over hard limit (we or our parents) */
1890 if ((parent_status == OVER_LIMIT) ||
1891 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1892 goto suppress_allocation;
1893
1894 /* guarantee minimum buffer size under pressure */
1895 if (kind == SK_MEM_RECV) {
1896 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1897 return 1;
180d8cd9 1898
3ab224be
HA		 1899 } else { /* SK_MEM_SEND */
1900 if (sk->sk_type == SOCK_STREAM) {
1901 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1902 return 1;
1903 } else if (atomic_read(&sk->sk_wmem_alloc) <
1904 prot->sysctl_wmem[0])
1905 return 1;
1906 }
1907
180d8cd9 1908 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 1909 int alloc;
1910
180d8cd9 1911 if (!sk_under_memory_pressure(sk))
1748376b 1912 return 1;
180d8cd9
GC		 1913 alloc = sk_sockets_allocated_read_positive(sk);
1914 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 1915 sk_mem_pages(sk->sk_wmem_queued +
1916 atomic_read(&sk->sk_rmem_alloc) +
1917 sk->sk_forward_alloc))
1918 return 1;
1919 }
1920
1921suppress_allocation:
1922
1923 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1924 sk_stream_moderate_sndbuf(sk);
1925
1926 /* Fail only if socket is _under_ its sndbuf.
1927 * In this case we cannot block, so that we have to fail.
1928 */
1929 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1930 return 1;
1931 }
1932
3847ce32
SM		 1933 trace_sock_exceed_buf_limit(sk, prot, allocated);
1934
3ab224be
HA		 1935 /* Alas. Undo changes. */
1936 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 1937
0e90b31f 1938 sk_memory_allocated_sub(sk, amt);
180d8cd9 1939
3ab224be
HA		 1940 return 0;
1941}
3ab224be
HA		 1942EXPORT_SYMBOL(__sk_mem_schedule);
1943
1944/**
1945 * __sk_reclaim - reclaim memory_allocated
1946 * @sk: socket
1947 */
1948void __sk_mem_reclaim(struct sock *sk)
1949{
180d8cd9 1950 sk_memory_allocated_sub(sk,
0e90b31f 1951 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 1952 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
1953
180d8cd9
GC		 1954 if (sk_under_memory_pressure(sk) &&
1955 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
1956 sk_leave_memory_pressure(sk);
3ab224be 1957}
3ab224be
HA		 1958EXPORT_SYMBOL(__sk_mem_reclaim);
1959
1960
1da177e4
LT		 1961/*
1962 * Set of default routines for initialising struct proto_ops when
1963 * the protocol does not support a particular function. In certain
1964 * cases where it makes no sense for a protocol to have a "do nothing"
1965 * function, some default processing is provided.
1966 */
1967
1968int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1969{
1970 return -EOPNOTSUPP;
1971}
2a91525c 1972EXPORT_SYMBOL(sock_no_bind);
1da177e4 1973
4ec93edb 1974int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1975 int len, int flags)
1976{
1977 return -EOPNOTSUPP;
1978}
2a91525c 1979EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 1980
1981int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1982{
1983 return -EOPNOTSUPP;
1984}
2a91525c 1985EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 1986
1987int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1988{
1989 return -EOPNOTSUPP;
1990}
2a91525c 1991EXPORT_SYMBOL(sock_no_accept);
1da177e4 1992
4ec93edb 1993int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1994 int *len, int peer)
1995{
1996 return -EOPNOTSUPP;
1997}
2a91525c 1998EXPORT_SYMBOL(sock_no_getname);
1da177e4 1999
2a91525c 2000unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2001{
2002 return 0;
2003}
2a91525c 2004EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2005
2006int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2007{
2008 return -EOPNOTSUPP;
2009}
2a91525c 2010EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2011
2012int sock_no_listen(struct socket *sock, int backlog)
2013{
2014 return -EOPNOTSUPP;
2015}
2a91525c 2016EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2017
2018int sock_no_shutdown(struct socket *sock, int how)
2019{
2020 return -EOPNOTSUPP;
2021}
2a91525c 2022EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2023
2024int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2025 char __user *optval, unsigned int optlen)
1da177e4
LT		 2026{
2027 return -EOPNOTSUPP;
2028}
2a91525c 2029EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2030
2031int sock_no_getsockopt(struct socket *sock, int level, int optname,
2032 char __user *optval, int __user *optlen)
2033{
2034 return -EOPNOTSUPP;
2035}
2a91525c 2036EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2037
2038int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2039 size_t len)
2040{
2041 return -EOPNOTSUPP;
2042}
2a91525c 2043EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2044
2045int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2046 size_t len, int flags)
2047{
2048 return -EOPNOTSUPP;
2049}
2a91525c 2050EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2051
2052int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2053{
2054 /* Mirror missing mmap method error code */
2055 return -ENODEV;
2056}
2a91525c 2057EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2058
2059ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2060{
2061 ssize_t res;
2062 struct msghdr msg = {.msg_flags = flags};
2063 struct kvec iov;
2064 char *kaddr = kmap(page);
2065 iov.iov_base = kaddr + offset;
2066 iov.iov_len = size;
2067 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2068 kunmap(page);
2069 return res;
2070}
2a91525c 2071EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2072
2073/*
2074 * Default Socket Callbacks
2075 */
2076
2077static void sock_def_wakeup(struct sock *sk)
2078{
43815482
ED		 2079 struct socket_wq *wq;
2080
2081 rcu_read_lock();
2082 wq = rcu_dereference(sk->sk_wq);
2083 if (wq_has_sleeper(wq))
2084 wake_up_interruptible_all(&wq->wait);
2085 rcu_read_unlock();
1da177e4
LT		 2086}
2087
2088static void sock_def_error_report(struct sock *sk)
2089{
43815482
ED		 2090 struct socket_wq *wq;
2091
2092 rcu_read_lock();
2093 wq = rcu_dereference(sk->sk_wq);
2094 if (wq_has_sleeper(wq))
2095 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2096 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2097 rcu_read_unlock();
1da177e4
LT		 2098}
2099
2100static void sock_def_readable(struct sock *sk, int len)
2101{
43815482
ED		 2102 struct socket_wq *wq;
2103
2104 rcu_read_lock();
2105 wq = rcu_dereference(sk->sk_wq);
2106 if (wq_has_sleeper(wq))
2c6607c6 2107 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2108 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2109 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2110 rcu_read_unlock();
1da177e4
LT		 2111}
2112
2113static void sock_def_write_space(struct sock *sk)
2114{
43815482
ED		 2115 struct socket_wq *wq;
2116
2117 rcu_read_lock();
1da177e4
LT		 2118
2119 /* Do not wake up a writer until he can make "significant"
2120 * progress. --DaveM
2121 */
e71a4783 2122 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2123 wq = rcu_dereference(sk->sk_wq);
2124 if (wq_has_sleeper(wq))
2125 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2126 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2127
2128 /* Should agree with poll, otherwise some programs break */
2129 if (sock_writeable(sk))
8d8ad9d7 2130 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2131 }
2132
43815482 2133 rcu_read_unlock();
1da177e4
LT		 2134}
2135
2136static void sock_def_destruct(struct sock *sk)
2137{
a51482bd 2138 kfree(sk->sk_protinfo);
1da177e4
LT		 2139}
2140
2141void sk_send_sigurg(struct sock *sk)
2142{
2143 if (sk->sk_socket && sk->sk_socket->file)
2144 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2145 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2146}
2a91525c 2147EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2148
2149void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2150 unsigned long expires)
2151{
2152 if (!mod_timer(timer, expires))
2153 sock_hold(sk);
2154}
1da177e4
LT		 2155EXPORT_SYMBOL(sk_reset_timer);
2156
2157void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2158{
2159 if (timer_pending(timer) && del_timer(timer))
2160 __sock_put(sk);
2161}
1da177e4
LT		 2162EXPORT_SYMBOL(sk_stop_timer);
2163
2164void sock_init_data(struct socket *sock, struct sock *sk)
2165{
2166 skb_queue_head_init(&sk->sk_receive_queue);
2167 skb_queue_head_init(&sk->sk_write_queue);
2168 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2169#ifdef CONFIG_NET_DMA
2170 skb_queue_head_init(&sk->sk_async_wait_queue);
2171#endif
1da177e4
LT		 2172
2173 sk->sk_send_head = NULL;
2174
2175 init_timer(&sk->sk_timer);
4ec93edb 2176
1da177e4
LT		 2177 sk->sk_allocation = GFP_KERNEL;
2178 sk->sk_rcvbuf = sysctl_rmem_default;
2179 sk->sk_sndbuf = sysctl_wmem_default;
2180 sk->sk_state = TCP_CLOSE;
972692e0 2181 sk_set_socket(sk, sock);
1da177e4
LT		 2182
2183 sock_set_flag(sk, SOCK_ZAPPED);
2184
e71a4783 2185 if (sock) {
1da177e4 2186 sk->sk_type = sock->type;
43815482 2187 sk->sk_wq = sock->wq;
1da177e4
LT		 2188 sock->sk = sk;
2189 } else
43815482 2190 sk->sk_wq = NULL;
1da177e4 2191
b6c6712a 2192 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2193 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2194 lockdep_set_class_and_name(&sk->sk_callback_lock,
2195 af_callback_keys + sk->sk_family,
2196 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2197
2198 sk->sk_state_change = sock_def_wakeup;
2199 sk->sk_data_ready = sock_def_readable;
2200 sk->sk_write_space = sock_def_write_space;
2201 sk->sk_error_report = sock_def_error_report;
2202 sk->sk_destruct = sock_def_destruct;
2203
5640f768
ED		 2204 sk->sk_frag.page = NULL;
2205 sk->sk_frag.offset = 0;
ef64a54f 2206 sk->sk_peek_off = -1;
1da177e4 2207
109f6e39
EB		 2208 sk->sk_peer_pid = NULL;
2209 sk->sk_peer_cred = NULL;
1da177e4
LT		 2210 sk->sk_write_pending = 0;
2211 sk->sk_rcvlowat = 1;
2212 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2213 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2214
f37f0afb 2215 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2216
4dc6dc71
ED		 2217 /*
2218 * Before updating sk_refcnt, we must commit prior changes to memory
2219 * (Documentation/RCU/rculist_nulls.txt for details)
2220 */
2221 smp_wmb();
1da177e4 2222 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2223 atomic_set(&sk->sk_drops, 0);
1da177e4 2224}
2a91525c 2225EXPORT_SYMBOL(sock_init_data);
1da177e4 2226
b5606c2d 2227void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2228{
2229 might_sleep();
a5b5bb9a 2230 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2231 if (sk->sk_lock.owned)
1da177e4 2232 __lock_sock(sk);
d2e9117c 2233 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2234 spin_unlock(&sk->sk_lock.slock);
2235 /*
2236 * The sk_lock has mutex_lock() semantics here:
2237 */
fcc70d5f 2238 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2239 local_bh_enable();
1da177e4 2240}
fcc70d5f 2241EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2242
b5606c2d 2243void release_sock(struct sock *sk)
1da177e4 2244{
a5b5bb9a
IM		 2245 /*
2246 * The sk_lock has mutex_unlock() semantics:
2247 */
2248 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2249
2250 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2251 if (sk->sk_backlog.tail)
2252 __release_sock(sk);
46d3ceab
ED		 2253
2254 if (sk->sk_prot->release_cb)
2255 sk->sk_prot->release_cb(sk);
2256
d2e9117c 2257 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2258 if (waitqueue_active(&sk->sk_lock.wq))
2259 wake_up(&sk->sk_lock.wq);
2260 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2261}
2262EXPORT_SYMBOL(release_sock);
2263
8a74ad60
ED		 2264/**
2265 * lock_sock_fast - fast version of lock_sock
2266 * @sk: socket
2267 *
2268 * This version should be used for very small section, where process wont block
2269 * return false if fast path is taken
2270 * sk_lock.slock locked, owned = 0, BH disabled
2271 * return true if slow path is taken
2272 * sk_lock.slock unlocked, owned = 1, BH enabled
2273 */
2274bool lock_sock_fast(struct sock *sk)
2275{
2276 might_sleep();
2277 spin_lock_bh(&sk->sk_lock.slock);
2278
2279 if (!sk->sk_lock.owned)
2280 /*
2281 * Note : We must disable BH
2282 */
2283 return false;
2284
2285 __lock_sock(sk);
2286 sk->sk_lock.owned = 1;
2287 spin_unlock(&sk->sk_lock.slock);
2288 /*
2289 * The sk_lock has mutex_lock() semantics here:
2290 */
2291 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2292 local_bh_enable();
2293 return true;
2294}
2295EXPORT_SYMBOL(lock_sock_fast);
2296
1da177e4 2297int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2298{
b7aa0bf7 2299 struct timeval tv;
1da177e4 2300 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2301 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2302 tv = ktime_to_timeval(sk->sk_stamp);
2303 if (tv.tv_sec == -1)
1da177e4 2304 return -ENOENT;
b7aa0bf7
ED		 2305 if (tv.tv_sec == 0) {
2306 sk->sk_stamp = ktime_get_real();
2307 tv = ktime_to_timeval(sk->sk_stamp);
2308 }
2309 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2310}
1da177e4
LT		 2311EXPORT_SYMBOL(sock_get_timestamp);
2312
ae40eb1e
ED		 2313int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2314{
2315 struct timespec ts;
2316 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2317 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2318 ts = ktime_to_timespec(sk->sk_stamp);
2319 if (ts.tv_sec == -1)
2320 return -ENOENT;
2321 if (ts.tv_sec == 0) {
2322 sk->sk_stamp = ktime_get_real();
2323 ts = ktime_to_timespec(sk->sk_stamp);
2324 }
2325 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2326}
2327EXPORT_SYMBOL(sock_get_timestampns);
2328
20d49473 2329void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2330{
20d49473 2331 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2332 unsigned long previous_flags = sk->sk_flags;
2333
20d49473
PO		 2334 sock_set_flag(sk, flag);
2335 /*
2336 * we just set one of the two flags which require net
2337 * time stamping, but time stamping might have been on
2338 * already because of the other one
2339 */
08e29af3 2340 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2341 net_enable_timestamp();
1da177e4
LT		 2342 }
2343}
1da177e4
LT		 2344
2345/*
2346 * Get a socket option on an socket.
2347 *
2348 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2349 * asynchronous errors should be reported by getsockopt. We assume
2350 * this means if you specify SO_ERROR (otherwise whats the point of it).
2351 */
2352int sock_common_getsockopt(struct socket *sock, int level, int optname,
2353 char __user *optval, int __user *optlen)
2354{
2355 struct sock *sk = sock->sk;
2356
2357 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2358}
1da177e4
LT		 2359EXPORT_SYMBOL(sock_common_getsockopt);
2360
3fdadf7d 2361#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2362int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2363 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2364{
2365 struct sock *sk = sock->sk;
2366
1e51f951 2367 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2368 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2369 optval, optlen);
3fdadf7d
DM		 2370 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2371}
2372EXPORT_SYMBOL(compat_sock_common_getsockopt);
2373#endif
2374
1da177e4
LT		 2375int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2376 struct msghdr *msg, size_t size, int flags)
2377{
2378 struct sock *sk = sock->sk;
2379 int addr_len = 0;
2380 int err;
2381
2382 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2383 flags & ~MSG_DONTWAIT, &addr_len);
2384 if (err >= 0)
2385 msg->msg_namelen = addr_len;
2386 return err;
2387}
1da177e4
LT		 2388EXPORT_SYMBOL(sock_common_recvmsg);
2389
2390/*
2391 * Set socket options on an inet socket.
2392 */
2393int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2394 char __user *optval, unsigned int optlen)
1da177e4
LT		 2395{
2396 struct sock *sk = sock->sk;
2397
2398 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2399}
1da177e4
LT		 2400EXPORT_SYMBOL(sock_common_setsockopt);
2401
3fdadf7d 2402#ifdef CONFIG_COMPAT
543d9cfe 2403int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2404 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2405{
2406 struct sock *sk = sock->sk;
2407
543d9cfe
ACM		 2408 if (sk->sk_prot->compat_setsockopt != NULL)
2409 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2410 optval, optlen);
3fdadf7d
DM		 2411 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2412}
2413EXPORT_SYMBOL(compat_sock_common_setsockopt);
2414#endif
2415
1da177e4
LT		 2416void sk_common_release(struct sock *sk)
2417{
2418 if (sk->sk_prot->destroy)
2419 sk->sk_prot->destroy(sk);
2420
2421 /*
2422 * Observation: when sock_common_release is called, processes have
2423 * no access to socket. But net still has.
2424 * Step one, detach it from networking:
2425 *
2426 * A. Remove from hash tables.
2427 */
2428
2429 sk->sk_prot->unhash(sk);
2430
2431 /*
2432 * In this point socket cannot receive new packets, but it is possible
2433 * that some packets are in flight because some CPU runs receiver and
2434 * did hash table lookup before we unhashed socket. They will achieve
2435 * receive queue and will be purged by socket destructor.
2436 *
2437 * Also we still have packets pending on receive queue and probably,
2438 * our own packets waiting in device queues. sock_destroy will drain
2439 * receive queue, but transmitted packets will delay socket destruction
2440 * until the last reference will be released.
2441 */
2442
2443 sock_orphan(sk);
2444
2445 xfrm_sk_free_policy(sk);
2446
e6848976 2447 sk_refcnt_debug_release(sk);
5640f768
ED		 2448
2449 if (sk->sk_frag.page) {
2450 put_page(sk->sk_frag.page);
2451 sk->sk_frag.page = NULL;
2452 }
2453
1da177e4
LT		 2454 sock_put(sk);
2455}
1da177e4
LT		 2456EXPORT_SYMBOL(sk_common_release);
2457
13ff3d6f
PE		 2458#ifdef CONFIG_PROC_FS
2459#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2460struct prot_inuse {
2461 int val[PROTO_INUSE_NR];
2462};
13ff3d6f
PE		 2463
2464static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2465
2466#ifdef CONFIG_NET_NS
2467void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2468{
d6d9ca0f 2469 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2470}
2471EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2472
2473int sock_prot_inuse_get(struct net *net, struct proto *prot)
2474{
2475 int cpu, idx = prot->inuse_idx;
2476 int res = 0;
2477
2478 for_each_possible_cpu(cpu)
2479 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2480
2481 return res >= 0 ? res : 0;
2482}
2483EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2484
2c8c1e72 2485static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2486{
2487 net->core.inuse = alloc_percpu(struct prot_inuse);
2488 return net->core.inuse ? 0 : -ENOMEM;
2489}
2490
2c8c1e72 2491static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2492{
2493 free_percpu(net->core.inuse);
2494}
2495
2496static struct pernet_operations net_inuse_ops = {
2497 .init = sock_inuse_init_net,
2498 .exit = sock_inuse_exit_net,
2499};
2500
2501static __init int net_inuse_init(void)
2502{
2503 if (register_pernet_subsys(&net_inuse_ops))
2504 panic("Cannot initialize net inuse counters");
2505
2506 return 0;
2507}
2508
2509core_initcall(net_inuse_init);
2510#else
1338d466
PE		 2511static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2512
c29a0bc4 2513void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2514{
d6d9ca0f 2515 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2516}
2517EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2518
c29a0bc4 2519int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2520{
2521 int cpu, idx = prot->inuse_idx;
2522 int res = 0;
2523
2524 for_each_possible_cpu(cpu)
2525 res += per_cpu(prot_inuse, cpu).val[idx];
2526
2527 return res >= 0 ? res : 0;
2528}
2529EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2530#endif
13ff3d6f
PE		 2531
2532static void assign_proto_idx(struct proto *prot)
2533{
2534 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2535
2536 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2537 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2538 return;
2539 }
2540
2541 set_bit(prot->inuse_idx, proto_inuse_idx);
2542}
2543
2544static void release_proto_idx(struct proto *prot)
2545{
2546 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2547 clear_bit(prot->inuse_idx, proto_inuse_idx);
2548}
2549#else
2550static inline void assign_proto_idx(struct proto *prot)
2551{
2552}
2553
2554static inline void release_proto_idx(struct proto *prot)
2555{
2556}
2557#endif
2558
b733c007
PE		 2559int proto_register(struct proto *prot, int alloc_slab)
2560{
1da177e4
LT		 2561 if (alloc_slab) {
2562 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2563 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2564 NULL);
1da177e4
LT		 2565
2566 if (prot->slab == NULL) {
e005d193
JP		 2567 pr_crit("%s: Can't create sock SLAB cache!\n",
2568 prot->name);
60e7663d 2569 goto out;
1da177e4 2570 }
2e6599cb
ACM		 2571
2572 if (prot->rsk_prot != NULL) {
faf23422 2573 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2574 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2575 goto out_free_sock_slab;
2576
7e56b5d6 2577 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2578 prot->rsk_prot->obj_size, 0,
20c2df83 2579 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2580
2581 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2582 pr_crit("%s: Can't create request sock SLAB cache!\n",
2583 prot->name);
2e6599cb
ACM		 2584 goto out_free_request_sock_slab_name;
2585 }
2586 }
8feaf0c0 2587
6d6ee43e 2588 if (prot->twsk_prot != NULL) {
faf23422 2589 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2590
7e56b5d6 2591 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2592 goto out_free_request_sock_slab;
2593
6d6ee43e 2594 prot->twsk_prot->twsk_slab =
7e56b5d6 2595 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2596 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2597 0,
2598 SLAB_HWCACHE_ALIGN |
2599 prot->slab_flags,
20c2df83 2600 NULL);
6d6ee43e 2601 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2602 goto out_free_timewait_sock_slab_name;
2603 }
1da177e4
LT		 2604 }
2605
36b77a52 2606 mutex_lock(&proto_list_mutex);
1da177e4 2607 list_add(&prot->node, &proto_list);
13ff3d6f 2608 assign_proto_idx(prot);
36b77a52 2609 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2610 return 0;
2611
8feaf0c0 2612out_free_timewait_sock_slab_name:
7e56b5d6 2613 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2614out_free_request_sock_slab:
2615 if (prot->rsk_prot && prot->rsk_prot->slab) {
2616 kmem_cache_destroy(prot->rsk_prot->slab);
2617 prot->rsk_prot->slab = NULL;
2618 }
2e6599cb 2619out_free_request_sock_slab_name:
72150e9b
DC		 2620 if (prot->rsk_prot)
2621 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2622out_free_sock_slab:
2623 kmem_cache_destroy(prot->slab);
2624 prot->slab = NULL;
b733c007
PE		 2625out:
2626 return -ENOBUFS;
1da177e4 2627}
1da177e4
LT		 2628EXPORT_SYMBOL(proto_register);
2629
2630void proto_unregister(struct proto *prot)
2631{
36b77a52 2632 mutex_lock(&proto_list_mutex);
13ff3d6f 2633 release_proto_idx(prot);
0a3f4358 2634 list_del(&prot->node);
36b77a52 2635 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2636
2637 if (prot->slab != NULL) {
2638 kmem_cache_destroy(prot->slab);
2639 prot->slab = NULL;
2640 }
2641
2e6599cb 2642 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2643 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2644 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2645 prot->rsk_prot->slab = NULL;
2646 }
2647
6d6ee43e 2648 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2649 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2650 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2651 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2652 }
1da177e4 2653}
1da177e4
LT		 2654EXPORT_SYMBOL(proto_unregister);
2655
2656#ifdef CONFIG_PROC_FS
1da177e4 2657static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2658 __acquires(proto_list_mutex)
1da177e4 2659{
36b77a52 2660 mutex_lock(&proto_list_mutex);
60f0438a 2661 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2662}
2663
2664static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2665{
60f0438a 2666 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2667}
2668
2669static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2670 __releases(proto_list_mutex)
1da177e4 2671{
36b77a52 2672 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2673}
2674
2675static char proto_method_implemented(const void *method)
2676{
2677 return method == NULL ? 'n' : 'y';
2678}
180d8cd9
GC		 2679static long sock_prot_memory_allocated(struct proto *proto)
2680{
cb75a36c 2681 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2682}
2683
2684static char *sock_prot_memory_pressure(struct proto *proto)
2685{
2686 return proto->memory_pressure != NULL ?
2687 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2688}
1da177e4
LT		 2689
2690static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2691{
180d8cd9 2692
8d987e5c 2693 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2694 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2695 proto->name,
2696 proto->obj_size,
14e943db 2697 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2698 sock_prot_memory_allocated(proto),
2699 sock_prot_memory_pressure(proto),
1da177e4
LT		 2700 proto->max_header,
2701 proto->slab == NULL ? "no" : "yes",
2702 module_name(proto->owner),
2703 proto_method_implemented(proto->close),
2704 proto_method_implemented(proto->connect),
2705 proto_method_implemented(proto->disconnect),
2706 proto_method_implemented(proto->accept),
2707 proto_method_implemented(proto->ioctl),
2708 proto_method_implemented(proto->init),
2709 proto_method_implemented(proto->destroy),
2710 proto_method_implemented(proto->shutdown),
2711 proto_method_implemented(proto->setsockopt),
2712 proto_method_implemented(proto->getsockopt),
2713 proto_method_implemented(proto->sendmsg),
2714 proto_method_implemented(proto->recvmsg),
2715 proto_method_implemented(proto->sendpage),
2716 proto_method_implemented(proto->bind),
2717 proto_method_implemented(proto->backlog_rcv),
2718 proto_method_implemented(proto->hash),
2719 proto_method_implemented(proto->unhash),
2720 proto_method_implemented(proto->get_port),
2721 proto_method_implemented(proto->enter_memory_pressure));
2722}
2723
2724static int proto_seq_show(struct seq_file *seq, void *v)
2725{
60f0438a 2726 if (v == &proto_list)
1da177e4
LT		 2727 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2728 "protocol",
2729 "size",
2730 "sockets",
2731 "memory",
2732 "press",
2733 "maxhdr",
2734 "slab",
2735 "module",
2736 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2737 else
60f0438a 2738 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2739 return 0;
2740}
2741
f690808e 2742static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2743 .start = proto_seq_start,
2744 .next = proto_seq_next,
2745 .stop = proto_seq_stop,
2746 .show = proto_seq_show,
2747};
2748
2749static int proto_seq_open(struct inode *inode, struct file *file)
2750{
14e943db
ED		 2751 return seq_open_net(inode, file, &proto_seq_ops,
2752 sizeof(struct seq_net_private));
1da177e4
LT		 2753}
2754
9a32144e 2755static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2756 .owner = THIS_MODULE,
2757 .open = proto_seq_open,
2758 .read = seq_read,
2759 .llseek = seq_lseek,
14e943db
ED		 2760 .release = seq_release_net,
2761};
2762
2763static __net_init int proto_init_net(struct net *net)
2764{
2765 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2766 return -ENOMEM;
2767
2768 return 0;
2769}
2770
2771static __net_exit void proto_exit_net(struct net *net)
2772{
2773 proc_net_remove(net, "protocols");
2774}
2775
2776
2777static __net_initdata struct pernet_operations proto_net_ops = {
2778 .init = proto_init_net,
2779 .exit = proto_exit_net,
1da177e4
LT		 2780};
2781
2782static int __init proto_init(void)
2783{
14e943db 2784 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2785}
2786
2787subsys_initcall(proto_init);
2788
2789#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 0.885062 seconds and 5 git commands to generate.