projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'skb_frags'
[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 95#include <linux/errno.h>
cb820f8e 96#include <linux/errqueue.h>
1da177e4
LT		 97#include <linux/types.h>
98#include <linux/socket.h>
99#include <linux/in.h>
100#include <linux/kernel.h>
1da177e4
LT		 101#include <linux/module.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
104#include <linux/sched.h>
105#include <linux/timer.h>
106#include <linux/string.h>
107#include <linux/sockios.h>
108#include <linux/net.h>
109#include <linux/mm.h>
110#include <linux/slab.h>
111#include <linux/interrupt.h>
112#include <linux/poll.h>
113#include <linux/tcp.h>
114#include <linux/init.h>
a1f8e7f7 115#include <linux/highmem.h>
3f551f94 116#include <linux/user_namespace.h>
c5905afb 117#include <linux/static_key.h>
3969eb38 118#include <linux/memcontrol.h>
8c1ae10d 119#include <linux/prefetch.h>
1da177e4
LT		 120
121#include <asm/uaccess.h>
1da177e4
LT		 122
123#include <linux/netdevice.h>
124#include <net/protocol.h>
125#include <linux/skbuff.h>
457c4cbc 126#include <net/net_namespace.h>
2e6599cb 127#include <net/request_sock.h>
1da177e4 128#include <net/sock.h>
20d49473 129#include <linux/net_tstamp.h>
1da177e4
LT		 130#include <net/xfrm.h>
131#include <linux/ipsec.h>
f8451725 132#include <net/cls_cgroup.h>
5bc1421e 133#include <net/netprio_cgroup.h>
1da177e4
LT		 134
135#include <linux/filter.h>
136
3847ce32
SM		 137#include <trace/events/sock.h>
138
1da177e4
LT		 139#ifdef CONFIG_INET
140#include <net/tcp.h>
141#endif
142
076bb0c8 143#include <net/busy_poll.h>
06021292 144
36b77a52 145static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 146static LIST_HEAD(proto_list);
147
c255a458 148#ifdef CONFIG_MEMCG_KMEM
1d62e436 149int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
d1a4c0b3
GC		 150{
151 struct proto *proto;
152 int ret = 0;
153
36b77a52 154 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 155 list_for_each_entry(proto, &proto_list, node) {
156 if (proto->init_cgroup) {
1d62e436 157 ret = proto->init_cgroup(memcg, ss);
d1a4c0b3
GC		 158 if (ret)
159 goto out;
160 }
161 }
162
36b77a52 163 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 164 return ret;
165out:
166 list_for_each_entry_continue_reverse(proto, &proto_list, node)
167 if (proto->destroy_cgroup)
1d62e436 168 proto->destroy_cgroup(memcg);
36b77a52 169 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 170 return ret;
171}
172
1d62e436 173void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
d1a4c0b3
GC		 174{
175 struct proto *proto;
176
36b77a52 177 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 178 list_for_each_entry_reverse(proto, &proto_list, node)
179 if (proto->destroy_cgroup)
1d62e436 180 proto->destroy_cgroup(memcg);
36b77a52 181 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 182}
183#endif
184
da21f24d
IM		 185/*
186 * Each address family might have different locking rules, so we have
187 * one slock key per address family:
188 */
a5b5bb9a
IM		 189static struct lock_class_key af_family_keys[AF_MAX];
190static struct lock_class_key af_family_slock_keys[AF_MAX];
191
cbda4eaf 192#if defined(CONFIG_MEMCG_KMEM)
c5905afb 193struct static_key memcg_socket_limit_enabled;
e1aab161 194EXPORT_SYMBOL(memcg_socket_limit_enabled);
cbda4eaf 195#endif
e1aab161 196
a5b5bb9a
IM		 197/*
198 * Make lock validator output more readable. (we pre-construct these
199 * strings build-time, so that runtime initialization of socket
200 * locks is fast):
201 */
36cbd3dc 202static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 203 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
204 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
205 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
206 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
207 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
208 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
209 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 210 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 211 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 212 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 213 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 214 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 215 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
456db6a4 216 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 217};
36cbd3dc 218static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 219 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
220 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
221 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
222 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
223 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
224 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
225 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 226 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 227 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 228 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 229 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 230 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 231 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 232 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 233};
36cbd3dc 234static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 235 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
236 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
237 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
238 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
239 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
240 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
241 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 242 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 243 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 244 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 245 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 246 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 247 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 248 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 249};
da21f24d
IM		 250
251/*
252 * sk_callback_lock locking rules are per-address-family,
253 * so split the lock classes by using a per-AF key:
254 */
255static struct lock_class_key af_callback_keys[AF_MAX];
256
1da177e4
LT		 257/* Take into consideration the size of the struct sk_buff overhead in the
258 * determination of these values, since that is non-constant across
259 * platforms. This makes socket queueing behavior and performance
260 * not depend upon such differences.
261 */
262#define _SK_MEM_PACKETS 256
87fb4b7b 263#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 264#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
265#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
266
267/* Run time adjustable parameters. */
ab32ea5d 268__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 269EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 270__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 271EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 272__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
273__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 274
25985edc 275/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 276int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 277EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 278
c93bdd0e
MG		 279struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
280EXPORT_SYMBOL_GPL(memalloc_socks);
281
7cb02404
MG		 282/**
283 * sk_set_memalloc - sets %SOCK_MEMALLOC
284 * @sk: socket to set it on
285 *
286 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
287 * It's the responsibility of the admin to adjust min_free_kbytes
288 * to meet the requirements
289 */
290void sk_set_memalloc(struct sock *sk)
291{
292 sock_set_flag(sk, SOCK_MEMALLOC);
293 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 294 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 295}
296EXPORT_SYMBOL_GPL(sk_set_memalloc);
297
298void sk_clear_memalloc(struct sock *sk)
299{
300 sock_reset_flag(sk, SOCK_MEMALLOC);
301 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 302 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 303
304 /*
305 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
306 * progress of swapping. However, if SOCK_MEMALLOC is cleared while
307 * it has rmem allocations there is a risk that the user of the
308 * socket cannot make forward progress due to exceeding the rmem
309 * limits. By rights, sk_clear_memalloc() should only be called
310 * on sockets being torn down but warn and reset the accounting if
311 * that assumption breaks.
312 */
313 if (WARN_ON(sk->sk_forward_alloc))
314 sk_mem_reclaim(sk);
7cb02404
MG		 315}
316EXPORT_SYMBOL_GPL(sk_clear_memalloc);
317
b4b9e355
MG		 318int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
319{
320 int ret;
321 unsigned long pflags = current->flags;
322
323 /* these should have been dropped before queueing */
324 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
325
326 current->flags |= PF_MEMALLOC;
327 ret = sk->sk_backlog_rcv(sk, skb);
328 tsk_restore_flags(current, pflags, PF_MEMALLOC);
329
330 return ret;
331}
332EXPORT_SYMBOL(__sk_backlog_rcv);
333
1da177e4
LT		 334static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
335{
336 struct timeval tv;
337
338 if (optlen < sizeof(tv))
339 return -EINVAL;
340 if (copy_from_user(&tv, optval, sizeof(tv)))
341 return -EFAULT;
ba78073e
VA		 342 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
343 return -EDOM;
1da177e4 344
ba78073e 345 if (tv.tv_sec < 0) {
6f11df83
AM		 346 static int warned __read_mostly;
347
ba78073e 348 *timeo_p = 0;
50aab54f 349 if (warned < 10 && net_ratelimit()) {
ba78073e 350 warned++;
e005d193
JP		 351 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
352 __func__, current->comm, task_pid_nr(current));
50aab54f 353 }
ba78073e
VA		 354 return 0;
355 }
1da177e4
LT		 356 *timeo_p = MAX_SCHEDULE_TIMEOUT;
357 if (tv.tv_sec == 0 && tv.tv_usec == 0)
358 return 0;
359 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
360 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
361 return 0;
362}
363
364static void sock_warn_obsolete_bsdism(const char *name)
365{
366 static int warned;
367 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 368 if (strcmp(warncomm, current->comm) && warned < 5) {
369 strcpy(warncomm, current->comm);
e005d193
JP		 370 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
371 warncomm, name);
1da177e4
LT		 372 warned++;
373 }
374}
375
08e29af3
ED		 376#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
377
378static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 379{
08e29af3
ED		 380 if (sk->sk_flags & flags) {
381 sk->sk_flags &= ~flags;
382 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 383 net_disable_timestamp();
1da177e4
LT		 384 }
385}
386
387
f0088a50
DV		 388int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
389{
766e9037 390 int err;
f0088a50 391 int skb_len;
3b885787
NH		 392 unsigned long flags;
393 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 394
0fd7bac6 395 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 396 atomic_inc(&sk->sk_drops);
3847ce32 397 trace_sock_rcvqueue_full(sk, skb);
766e9037 398 return -ENOMEM;
f0088a50
DV		 399 }
400
fda9ef5d 401 err = sk_filter(sk, skb);
f0088a50 402 if (err)
766e9037 403 return err;
f0088a50 404
c76562b6 405 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 406 atomic_inc(&sk->sk_drops);
407 return -ENOBUFS;
3ab224be
HA		 408 }
409
f0088a50
DV		 410 skb->dev = NULL;
411 skb_set_owner_r(skb, sk);
49ad9599 412
f0088a50
DV		 413 /* Cache the SKB length before we tack it onto the receive
414 * queue. Once it is added it no longer belongs to us and
415 * may be freed by other threads of control pulling packets
416 * from the queue.
417 */
418 skb_len = skb->len;
419
7fee226a
ED		 420 /* we escape from rcu protected region, make sure we dont leak
421 * a norefcounted dst
422 */
423 skb_dst_force(skb);
424
3b885787
NH		 425 spin_lock_irqsave(&list->lock, flags);
426 skb->dropcount = atomic_read(&sk->sk_drops);
427 __skb_queue_tail(list, skb);
428 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 429
430 if (!sock_flag(sk, SOCK_DEAD))
431 sk->sk_data_ready(sk, skb_len);
766e9037 432 return 0;
f0088a50
DV		 433}
434EXPORT_SYMBOL(sock_queue_rcv_skb);
435
58a5a7b9 436int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 437{
438 int rc = NET_RX_SUCCESS;
439
fda9ef5d 440 if (sk_filter(sk, skb))
f0088a50
DV		 441 goto discard_and_relse;
442
443 skb->dev = NULL;
444
f545a38f 445 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) {
c377411f
ED		 446 atomic_inc(&sk->sk_drops);
447 goto discard_and_relse;
448 }
58a5a7b9
ACM		 449 if (nested)
450 bh_lock_sock_nested(sk);
451 else
452 bh_lock_sock(sk);
a5b5bb9a
IM		 453 if (!sock_owned_by_user(sk)) {
454 /*
455 * trylock + unlock semantics:
456 */
457 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
458
c57943a1 459 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 460
461 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 462 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 463 bh_unlock_sock(sk);
464 atomic_inc(&sk->sk_drops);
465 goto discard_and_relse;
466 }
467
f0088a50
DV		 468 bh_unlock_sock(sk);
469out:
470 sock_put(sk);
471 return rc;
472discard_and_relse:
473 kfree_skb(skb);
474 goto out;
475}
476EXPORT_SYMBOL(sk_receive_skb);
477
478struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
479{
b6c6712a 480 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 481
482 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 483 sk_tx_queue_clear(sk);
a9b3cd7f 484 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 485 dst_release(dst);
486 return NULL;
487 }
488
489 return dst;
490}
491EXPORT_SYMBOL(__sk_dst_check);
492
493struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
494{
495 struct dst_entry *dst = sk_dst_get(sk);
496
497 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
498 sk_dst_reset(sk);
499 dst_release(dst);
500 return NULL;
501 }
502
503 return dst;
504}
505EXPORT_SYMBOL(sk_dst_check);
506
c91f6df2
BH		 507static int sock_setbindtodevice(struct sock *sk, char __user *optval,
508 int optlen)
4878809f
DM		 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
c91f6df2
BH		 565static int sock_getbindtodevice(struct sock *sk, char __user *optval,
566 int __user *optlen, int len)
567{
568 int ret = -ENOPROTOOPT;
569#ifdef CONFIG_NETDEVICES
570 struct net *net = sock_net(sk);
c91f6df2 571 char devname[IFNAMSIZ];
c91f6df2
BH		 572
573 if (sk->sk_bound_dev_if == 0) {
574 len = 0;
575 goto zero;
576 }
577
578 ret = -EINVAL;
579 if (len < IFNAMSIZ)
580 goto out;
581
5dbe7c17
NS		 582 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
583 if (ret)
c91f6df2 584 goto out;
c91f6df2
BH		 585
586 len = strlen(devname) + 1;
587
588 ret = -EFAULT;
589 if (copy_to_user(optval, devname, len))
590 goto out;
591
592zero:
593 ret = -EFAULT;
594 if (put_user(len, optlen))
595 goto out;
596
597 ret = 0;
598
599out:
600#endif
601
602 return ret;
603}
604
c0ef877b
PE		 605static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
606{
607 if (valbool)
608 sock_set_flag(sk, bit);
609 else
610 sock_reset_flag(sk, bit);
611}
612
1da177e4
LT		 613/*
614 * This is meant for all protocols to use and covers goings on
615 * at the socket level. Everything here is generic.
616 */
617
618int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 619 char __user *optval, unsigned int optlen)
1da177e4 620{
2a91525c 621 struct sock *sk = sock->sk;
1da177e4
LT		 622 int val;
623 int valbool;
624 struct linger ling;
625 int ret = 0;
4ec93edb 626
1da177e4
LT		 627 /*
628 * Options without arguments
629 */
630
4878809f 631 if (optname == SO_BINDTODEVICE)
c91f6df2 632 return sock_setbindtodevice(sk, optval, optlen);
4878809f 633
e71a4783
SH		 634 if (optlen < sizeof(int))
635 return -EINVAL;
4ec93edb 636
1da177e4
LT		 637 if (get_user(val, (int __user *)optval))
638 return -EFAULT;
4ec93edb 639
2a91525c 640 valbool = val ? 1 : 0;
1da177e4
LT		 641
642 lock_sock(sk);
643
2a91525c 644 switch (optname) {
e71a4783 645 case SO_DEBUG:
2a91525c 646 if (val && !capable(CAP_NET_ADMIN))
e71a4783 647 ret = -EACCES;
2a91525c 648 else
c0ef877b 649 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 650 break;
651 case SO_REUSEADDR:
4a17fd52 652 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 653 break;
055dc21a
TH		 654 case SO_REUSEPORT:
655 sk->sk_reuseport = valbool;
656 break;
e71a4783 657 case SO_TYPE:
49c794e9 658 case SO_PROTOCOL:
0d6038ee 659 case SO_DOMAIN:
e71a4783
SH		 660 case SO_ERROR:
661 ret = -ENOPROTOOPT;
662 break;
663 case SO_DONTROUTE:
c0ef877b 664 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 665 break;
666 case SO_BROADCAST:
667 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
668 break;
669 case SO_SNDBUF:
670 /* Don't error on this BSD doesn't and if you think
82981930
ED		 671 * about it this is right. Otherwise apps have to
672 * play 'guess the biggest size' games. RCVBUF/SNDBUF
673 * are treated in BSD as hints
674 */
675 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 676set_sndbuf:
e71a4783 677 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 678 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
679 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 680 sk->sk_write_space(sk);
681 break;
1da177e4 682
e71a4783
SH		 683 case SO_SNDBUFFORCE:
684 if (!capable(CAP_NET_ADMIN)) {
685 ret = -EPERM;
686 break;
687 }
688 goto set_sndbuf;
b0573dea 689
e71a4783
SH		 690 case SO_RCVBUF:
691 /* Don't error on this BSD doesn't and if you think
82981930
ED		 692 * about it this is right. Otherwise apps have to
693 * play 'guess the biggest size' games. RCVBUF/SNDBUF
694 * are treated in BSD as hints
695 */
696 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 697set_rcvbuf:
e71a4783
SH		 698 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
699 /*
700 * We double it on the way in to account for
701 * "struct sk_buff" etc. overhead. Applications
702 * assume that the SO_RCVBUF setting they make will
703 * allow that much actual data to be received on that
704 * socket.
705 *
706 * Applications are unaware that "struct sk_buff" and
707 * other overheads allocate from the receive buffer
708 * during socket buffer allocation.
709 *
710 * And after considering the possible alternatives,
711 * returning the value we actually used in getsockopt
712 * is the most desirable behavior.
713 */
82981930 714 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 715 break;
716
717 case SO_RCVBUFFORCE:
718 if (!capable(CAP_NET_ADMIN)) {
719 ret = -EPERM;
1da177e4 720 break;
e71a4783
SH		 721 }
722 goto set_rcvbuf;
1da177e4 723
e71a4783 724 case SO_KEEPALIVE:
1da177e4 725#ifdef CONFIG_INET
3e10986d
ED		 726 if (sk->sk_protocol == IPPROTO_TCP &&
727 sk->sk_type == SOCK_STREAM)
e71a4783 728 tcp_set_keepalive(sk, valbool);
1da177e4 729#endif
e71a4783
SH		 730 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
731 break;
732
733 case SO_OOBINLINE:
734 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
735 break;
736
737 case SO_NO_CHECK:
738 sk->sk_no_check = valbool;
739 break;
740
741 case SO_PRIORITY:
5e1fccc0
EB		 742 if ((val >= 0 && val <= 6) ||
743 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 744 sk->sk_priority = val;
745 else
746 ret = -EPERM;
747 break;
748
749 case SO_LINGER:
750 if (optlen < sizeof(ling)) {
751 ret = -EINVAL; /* 1003.1g */
1da177e4 752 break;
e71a4783 753 }
2a91525c 754 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 755 ret = -EFAULT;
1da177e4 756 break;
e71a4783
SH		 757 }
758 if (!ling.l_onoff)
759 sock_reset_flag(sk, SOCK_LINGER);
760 else {
1da177e4 761#if (BITS_PER_LONG == 32)
e71a4783
SH		 762 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
763 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 764 else
e71a4783
SH		 765#endif
766 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
767 sock_set_flag(sk, SOCK_LINGER);
768 }
769 break;
770
771 case SO_BSDCOMPAT:
772 sock_warn_obsolete_bsdism("setsockopt");
773 break;
774
775 case SO_PASSCRED:
776 if (valbool)
777 set_bit(SOCK_PASSCRED, &sock->flags);
778 else
779 clear_bit(SOCK_PASSCRED, &sock->flags);
780 break;
781
782 case SO_TIMESTAMP:
92f37fd2 783 case SO_TIMESTAMPNS:
e71a4783 784 if (valbool) {
92f37fd2
ED		 785 if (optname == SO_TIMESTAMP)
786 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
787 else
788 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 789 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 790 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 791 } else {
e71a4783 792 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 793 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
794 }
e71a4783
SH		 795 break;
796
20d49473
PO		 797 case SO_TIMESTAMPING:
798 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 799 ret = -EINVAL;
20d49473
PO		 800 break;
801 }
802 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
803 val & SOF_TIMESTAMPING_TX_HARDWARE);
804 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
805 val & SOF_TIMESTAMPING_TX_SOFTWARE);
806 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
807 val & SOF_TIMESTAMPING_RX_HARDWARE);
808 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
809 sock_enable_timestamp(sk,
810 SOCK_TIMESTAMPING_RX_SOFTWARE);
811 else
812 sock_disable_timestamp(sk,
08e29af3 813 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 814 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
815 val & SOF_TIMESTAMPING_SOFTWARE);
816 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
817 val & SOF_TIMESTAMPING_SYS_HARDWARE);
818 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
819 val & SOF_TIMESTAMPING_RAW_HARDWARE);
820 break;
821
e71a4783
SH		 822 case SO_RCVLOWAT:
823 if (val < 0)
824 val = INT_MAX;
825 sk->sk_rcvlowat = val ? : 1;
826 break;
827
828 case SO_RCVTIMEO:
829 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
830 break;
831
832 case SO_SNDTIMEO:
833 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
834 break;
1da177e4 835
e71a4783
SH		 836 case SO_ATTACH_FILTER:
837 ret = -EINVAL;
838 if (optlen == sizeof(struct sock_fprog)) {
839 struct sock_fprog fprog;
1da177e4 840
e71a4783
SH		 841 ret = -EFAULT;
842 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 843 break;
e71a4783
SH		 844
845 ret = sk_attach_filter(&fprog, sk);
846 }
847 break;
848
849 case SO_DETACH_FILTER:
55b33325 850 ret = sk_detach_filter(sk);
e71a4783 851 break;
1da177e4 852
d59577b6
VB		 853 case SO_LOCK_FILTER:
854 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
855 ret = -EPERM;
856 else
857 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
858 break;
859
e71a4783
SH		 860 case SO_PASSSEC:
861 if (valbool)
862 set_bit(SOCK_PASSSEC, &sock->flags);
863 else
864 clear_bit(SOCK_PASSSEC, &sock->flags);
865 break;
4a19ec58 866 case SO_MARK:
5e1fccc0 867 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 868 ret = -EPERM;
2a91525c 869 else
4a19ec58 870 sk->sk_mark = val;
4a19ec58 871 break;
877ce7c1 872
1da177e4
LT		 873 /* We implement the SO_SNDLOWAT etc to
874 not be settable (1003.1g 5.3) */
3b885787 875 case SO_RXQ_OVFL:
8083f0fc 876 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 877 break;
6e3e939f
JB		 878
879 case SO_WIFI_STATUS:
880 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
881 break;
882
ef64a54f
PE		 883 case SO_PEEK_OFF:
884 if (sock->ops->set_peek_off)
12663bfc 885 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 886 else
887 ret = -EOPNOTSUPP;
888 break;
3bdc0eba
BG		 889
890 case SO_NOFCS:
891 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
892 break;
893
7d4c04fc
KJ		 894 case SO_SELECT_ERR_QUEUE:
895 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
896 break;
897
e0d1095a 898#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 899 case SO_BUSY_POLL:
dafcc438
ET		 900 /* allow unprivileged users to decrease the value */
901 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
902 ret = -EPERM;
903 else {
904 if (val < 0)
905 ret = -EINVAL;
906 else
907 sk->sk_ll_usec = val;
908 }
909 break;
910#endif
62748f32
ED		 911
912 case SO_MAX_PACING_RATE:
913 sk->sk_max_pacing_rate = val;
914 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
915 sk->sk_max_pacing_rate);
916 break;
917
e71a4783
SH		 918 default:
919 ret = -ENOPROTOOPT;
920 break;
4ec93edb 921 }
1da177e4
LT		 922 release_sock(sk);
923 return ret;
924}
2a91525c 925EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 926
927
8f09898b 928static void cred_to_ucred(struct pid *pid, const struct cred *cred,
929 struct ucred *ucred)
3f551f94
EB		 930{
931 ucred->pid = pid_vnr(pid);
932 ucred->uid = ucred->gid = -1;
933 if (cred) {
934 struct user_namespace *current_ns = current_user_ns();
935
b2e4f544
EB		 936 ucred->uid = from_kuid_munged(current_ns, cred->euid);
937 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 938 }
939}
940
1da177e4
LT		 941int sock_getsockopt(struct socket *sock, int level, int optname,
942 char __user *optval, int __user *optlen)
943{
944 struct sock *sk = sock->sk;
4ec93edb 945
e71a4783 946 union {
4ec93edb
YH		 947 int val;
948 struct linger ling;
1da177e4
LT		 949 struct timeval tm;
950 } v;
4ec93edb 951
4d0392be 952 int lv = sizeof(int);
1da177e4 953 int len;
4ec93edb 954
e71a4783 955 if (get_user(len, optlen))
4ec93edb 956 return -EFAULT;
e71a4783 957 if (len < 0)
1da177e4 958 return -EINVAL;
4ec93edb 959
50fee1de 960 memset(&v, 0, sizeof(v));
df0bca04 961
2a91525c 962 switch (optname) {
e71a4783
SH		 963 case SO_DEBUG:
964 v.val = sock_flag(sk, SOCK_DBG);
965 break;
966
967 case SO_DONTROUTE:
968 v.val = sock_flag(sk, SOCK_LOCALROUTE);
969 break;
970
971 case SO_BROADCAST:
1b23a5df 972 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 973 break;
974
975 case SO_SNDBUF:
976 v.val = sk->sk_sndbuf;
977 break;
978
979 case SO_RCVBUF:
980 v.val = sk->sk_rcvbuf;
981 break;
982
983 case SO_REUSEADDR:
984 v.val = sk->sk_reuse;
985 break;
986
055dc21a
TH		 987 case SO_REUSEPORT:
988 v.val = sk->sk_reuseport;
989 break;
990
e71a4783 991 case SO_KEEPALIVE:
1b23a5df 992 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 993 break;
994
995 case SO_TYPE:
996 v.val = sk->sk_type;
997 break;
998
49c794e9
JE		 999 case SO_PROTOCOL:
1000 v.val = sk->sk_protocol;
1001 break;
1002
0d6038ee
JE		 1003 case SO_DOMAIN:
1004 v.val = sk->sk_family;
1005 break;
1006
e71a4783
SH		 1007 case SO_ERROR:
1008 v.val = -sock_error(sk);
2a91525c 1009 if (v.val == 0)
e71a4783
SH		 1010 v.val = xchg(&sk->sk_err_soft, 0);
1011 break;
1012
1013 case SO_OOBINLINE:
1b23a5df 1014 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1015 break;
1016
1017 case SO_NO_CHECK:
1018 v.val = sk->sk_no_check;
1019 break;
1020
1021 case SO_PRIORITY:
1022 v.val = sk->sk_priority;
1023 break;
1024
1025 case SO_LINGER:
1026 lv = sizeof(v.ling);
1b23a5df 1027 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1028 v.ling.l_linger = sk->sk_lingertime / HZ;
1029 break;
1030
1031 case SO_BSDCOMPAT:
1032 sock_warn_obsolete_bsdism("getsockopt");
1033 break;
1034
1035 case SO_TIMESTAMP:
92f37fd2
ED		 1036 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1037 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1038 break;
1039
1040 case SO_TIMESTAMPNS:
1041 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1042 break;
1043
20d49473
PO		 1044 case SO_TIMESTAMPING:
1045 v.val = 0;
1046 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
1047 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
1048 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
1049 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
1050 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
1051 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
1052 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
1053 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
1054 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
1055 v.val |= SOF_TIMESTAMPING_SOFTWARE;
1056 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
1057 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
1058 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
1059 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
1060 break;
1061
e71a4783 1062 case SO_RCVTIMEO:
2a91525c 1063 lv = sizeof(struct timeval);
e71a4783
SH		 1064 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1065 v.tm.tv_sec = 0;
1066 v.tm.tv_usec = 0;
1067 } else {
1068 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1069 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1070 }
1071 break;
1072
1073 case SO_SNDTIMEO:
2a91525c 1074 lv = sizeof(struct timeval);
e71a4783
SH		 1075 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1076 v.tm.tv_sec = 0;
1077 v.tm.tv_usec = 0;
1078 } else {
1079 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1080 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1081 }
1082 break;
1da177e4 1083
e71a4783
SH		 1084 case SO_RCVLOWAT:
1085 v.val = sk->sk_rcvlowat;
1086 break;
1da177e4 1087
e71a4783 1088 case SO_SNDLOWAT:
2a91525c 1089 v.val = 1;
e71a4783 1090 break;
1da177e4 1091
e71a4783 1092 case SO_PASSCRED:
82981930 1093 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1094 break;
1da177e4 1095
e71a4783 1096 case SO_PEERCRED:
109f6e39
EB		 1097 {
1098 struct ucred peercred;
1099 if (len > sizeof(peercred))
1100 len = sizeof(peercred);
1101 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1102 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1103 return -EFAULT;
1104 goto lenout;
109f6e39 1105 }
1da177e4 1106
e71a4783
SH		 1107 case SO_PEERNAME:
1108 {
1109 char address[128];
1110
1111 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1112 return -ENOTCONN;
1113 if (lv < len)
1114 return -EINVAL;
1115 if (copy_to_user(optval, address, len))
1116 return -EFAULT;
1117 goto lenout;
1118 }
1da177e4 1119
e71a4783
SH		 1120 /* Dubious BSD thing... Probably nobody even uses it, but
1121 * the UNIX standard wants it for whatever reason... -DaveM
1122 */
1123 case SO_ACCEPTCONN:
1124 v.val = sk->sk_state == TCP_LISTEN;
1125 break;
1da177e4 1126
e71a4783 1127 case SO_PASSSEC:
82981930 1128 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1129 break;
877ce7c1 1130
e71a4783
SH		 1131 case SO_PEERSEC:
1132 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1133
4a19ec58
LAT		 1134 case SO_MARK:
1135 v.val = sk->sk_mark;
1136 break;
1137
3b885787 1138 case SO_RXQ_OVFL:
1b23a5df 1139 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1140 break;
1141
6e3e939f 1142 case SO_WIFI_STATUS:
1b23a5df 1143 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1144 break;
1145
ef64a54f
PE		 1146 case SO_PEEK_OFF:
1147 if (!sock->ops->set_peek_off)
1148 return -EOPNOTSUPP;
1149
1150 v.val = sk->sk_peek_off;
1151 break;
bc2f7996 1152 case SO_NOFCS:
1b23a5df 1153 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1154 break;
c91f6df2 1155
f7b86bfe 1156 case SO_BINDTODEVICE:
c91f6df2
BH		 1157 return sock_getbindtodevice(sk, optval, optlen, len);
1158
a8fc9277
PE		 1159 case SO_GET_FILTER:
1160 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1161 if (len < 0)
1162 return len;
1163
1164 goto lenout;
c91f6df2 1165
d59577b6
VB		 1166 case SO_LOCK_FILTER:
1167 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1168 break;
1169
ea02f941
MS		 1170 case SO_BPF_EXTENSIONS:
1171 v.val = bpf_tell_extensions();
1172 break;
1173
7d4c04fc
KJ		 1174 case SO_SELECT_ERR_QUEUE:
1175 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1176 break;
1177
e0d1095a 1178#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1179 case SO_BUSY_POLL:
dafcc438
ET		 1180 v.val = sk->sk_ll_usec;
1181 break;
1182#endif
1183
62748f32
ED		 1184 case SO_MAX_PACING_RATE:
1185 v.val = sk->sk_max_pacing_rate;
1186 break;
1187
e71a4783
SH		 1188 default:
1189 return -ENOPROTOOPT;
1da177e4 1190 }
e71a4783 1191
1da177e4
LT		 1192 if (len > lv)
1193 len = lv;
1194 if (copy_to_user(optval, &v, len))
1195 return -EFAULT;
1196lenout:
4ec93edb
YH		 1197 if (put_user(len, optlen))
1198 return -EFAULT;
1199 return 0;
1da177e4
LT		 1200}
1201
a5b5bb9a
IM		 1202/*
1203 * Initialize an sk_lock.
1204 *
1205 * (We also register the sk_lock with the lock validator.)
1206 */
b6f99a21 1207static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1208{
ed07536e
PZ		 1209 sock_lock_init_class_and_name(sk,
1210 af_family_slock_key_strings[sk->sk_family],
1211 af_family_slock_keys + sk->sk_family,
1212 af_family_key_strings[sk->sk_family],
1213 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1214}
1215
4dc6dc71
ED		 1216/*
1217 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1218 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1219 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1220 */
f1a6c4da
PE		 1221static void sock_copy(struct sock *nsk, const struct sock *osk)
1222{
1223#ifdef CONFIG_SECURITY_NETWORK
1224 void *sptr = nsk->sk_security;
1225#endif
68835aba
ED		 1226 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1227
1228 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1229 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1230
f1a6c4da
PE		 1231#ifdef CONFIG_SECURITY_NETWORK
1232 nsk->sk_security = sptr;
1233 security_sk_clone(osk, nsk);
1234#endif
1235}
1236
fcbdf09d
OP		 1237void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1238{
1239 unsigned long nulls1, nulls2;
1240
1241 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1242 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1243 if (nulls1 > nulls2)
1244 swap(nulls1, nulls2);
1245
1246 if (nulls1 != 0)
1247 memset((char *)sk, 0, nulls1);
1248 memset((char *)sk + nulls1 + sizeof(void *), 0,
1249 nulls2 - nulls1 - sizeof(void *));
1250 memset((char *)sk + nulls2 + sizeof(void *), 0,
1251 size - nulls2 - sizeof(void *));
1252}
1253EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1254
2e4afe7b
PE		 1255static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1256 int family)
c308c1b2
PE		 1257{
1258 struct sock *sk;
1259 struct kmem_cache *slab;
1260
1261 slab = prot->slab;
e912b114
ED		 1262 if (slab != NULL) {
1263 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1264 if (!sk)
1265 return sk;
1266 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1267 if (prot->clear_sk)
1268 prot->clear_sk(sk, prot->obj_size);
1269 else
1270 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1271 }
fcbdf09d 1272 } else
c308c1b2
PE		 1273 sk = kmalloc(prot->obj_size, priority);
1274
2e4afe7b 1275 if (sk != NULL) {
a98b65a3
VN		 1276 kmemcheck_annotate_bitfield(sk, flags);
1277
2e4afe7b
PE		 1278 if (security_sk_alloc(sk, family, priority))
1279 goto out_free;
1280
1281 if (!try_module_get(prot->owner))
1282 goto out_free_sec;
e022f0b4 1283 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1284 }
1285
c308c1b2 1286 return sk;
2e4afe7b
PE		 1287
1288out_free_sec:
1289 security_sk_free(sk);
1290out_free:
1291 if (slab != NULL)
1292 kmem_cache_free(slab, sk);
1293 else
1294 kfree(sk);
1295 return NULL;
c308c1b2
PE		 1296}
1297
1298static void sk_prot_free(struct proto *prot, struct sock *sk)
1299{
1300 struct kmem_cache *slab;
2e4afe7b 1301 struct module *owner;
c308c1b2 1302
2e4afe7b 1303 owner = prot->owner;
c308c1b2 1304 slab = prot->slab;
2e4afe7b
PE		 1305
1306 security_sk_free(sk);
c308c1b2
PE		 1307 if (slab != NULL)
1308 kmem_cache_free(slab, sk);
1309 else
1310 kfree(sk);
2e4afe7b 1311 module_put(owner);
c308c1b2
PE		 1312}
1313
86f8515f 1314#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1315void sock_update_netprioidx(struct sock *sk)
5bc1421e 1316{
5bc1421e
NH		 1317 if (in_interrupt())
1318 return;
2b73bc65 1319
6ffd4641 1320 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1321}
1322EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1323#endif
1324
1da177e4
LT		 1325/**
1326 * sk_alloc - All socket objects are allocated here
c4ea43c5 1327 * @net: the applicable net namespace
4dc3b16b
PP		 1328 * @family: protocol family
1329 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1330 * @prot: struct proto associated with this new sock instance
1da177e4 1331 */
1b8d7ae4 1332struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1333 struct proto *prot)
1da177e4 1334{
c308c1b2 1335 struct sock *sk;
1da177e4 1336
154adbc8 1337 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1338 if (sk) {
154adbc8
PE		 1339 sk->sk_family = family;
1340 /*
1341 * See comment in struct sock definition to understand
1342 * why we need sk_prot_creator -acme
1343 */
1344 sk->sk_prot = sk->sk_prot_creator = prot;
1345 sock_lock_init(sk);
3b1e0a65 1346 sock_net_set(sk, get_net(net));
d66ee058 1347 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1348
211d2f97 1349 sock_update_classid(sk);
6ffd4641 1350 sock_update_netprioidx(sk);
1da177e4 1351 }
a79af59e 1352
2e4afe7b 1353 return sk;
1da177e4 1354}
2a91525c 1355EXPORT_SYMBOL(sk_alloc);
1da177e4 1356
2b85a34e 1357static void __sk_free(struct sock *sk)
1da177e4
LT		 1358{
1359 struct sk_filter *filter;
1da177e4
LT		 1360
1361 if (sk->sk_destruct)
1362 sk->sk_destruct(sk);
1363
a898def2
PM		 1364 filter = rcu_dereference_check(sk->sk_filter,
1365 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1366 if (filter) {
309dd5fc 1367 sk_filter_uncharge(sk, filter);
a9b3cd7f 1368 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1369 }
1370
08e29af3 1371 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1372
1373 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1374 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1375 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1376
109f6e39
EB		 1377 if (sk->sk_peer_cred)
1378 put_cred(sk->sk_peer_cred);
1379 put_pid(sk->sk_peer_pid);
3b1e0a65 1380 put_net(sock_net(sk));
c308c1b2 1381 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1382}
2b85a34e
ED		 1383
1384void sk_free(struct sock *sk)
1385{
1386 /*
25985edc 1387 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1388 * some packets are still in some tx queue.
1389 * If not null, sock_wfree() will call __sk_free(sk) later
1390 */
1391 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1392 __sk_free(sk);
1393}
2a91525c 1394EXPORT_SYMBOL(sk_free);
1da177e4 1395
edf02087 1396/*
25985edc
LDM		 1397 * Last sock_put should drop reference to sk->sk_net. It has already
1398 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1399 * is not an option.
25985edc 1400 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1401 * destroy it in the context of init_net.
1402 */
1403void sk_release_kernel(struct sock *sk)
1404{
1405 if (sk == NULL || sk->sk_socket == NULL)
1406 return;
1407
1408 sock_hold(sk);
1409 sock_release(sk->sk_socket);
65a18ec5 1410 release_net(sock_net(sk));
3b1e0a65 1411 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1412 sock_put(sk);
1413}
45af1754 1414EXPORT_SYMBOL(sk_release_kernel);
edf02087 1415
475f1b52
SR		 1416static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1417{
1418 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1419 sock_update_memcg(newsk);
1420}
1421
e56c57d0
ED		 1422/**
1423 * sk_clone_lock - clone a socket, and lock its clone
1424 * @sk: the socket to clone
1425 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1426 *
1427 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1428 */
1429struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1430{
8fd1d178 1431 struct sock *newsk;
87d11ceb 1432
8fd1d178 1433 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1434 if (newsk != NULL) {
1435 struct sk_filter *filter;
1436
892c141e 1437 sock_copy(newsk, sk);
87d11ceb
ACM		 1438
1439 /* SANITY */
3b1e0a65 1440 get_net(sock_net(newsk));
87d11ceb
ACM		 1441 sk_node_init(&newsk->sk_node);
1442 sock_lock_init(newsk);
1443 bh_lock_sock(newsk);
fa438ccf 1444 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1445 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1446
1447 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1448 /*
1449 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1450 */
1451 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1452 atomic_set(&newsk->sk_omem_alloc, 0);
1453 skb_queue_head_init(&newsk->sk_receive_queue);
1454 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1455#ifdef CONFIG_NET_DMA
1456 skb_queue_head_init(&newsk->sk_async_wait_queue);
1457#endif
87d11ceb 1458
b6c6712a 1459 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1460 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1461 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1462 af_callback_keys + newsk->sk_family,
1463 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1464
1465 newsk->sk_dst_cache = NULL;
1466 newsk->sk_wmem_queued = 0;
1467 newsk->sk_forward_alloc = 0;
1468 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1469 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1470
1471 sock_reset_flag(newsk, SOCK_DONE);
1472 skb_queue_head_init(&newsk->sk_error_queue);
1473
0d7da9dd 1474 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1475 if (filter != NULL)
1476 sk_filter_charge(newsk, filter);
1477
1478 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1479 /* It is still raw copy of parent, so invalidate
1480 * destructor and make plain sk_free() */
1481 newsk->sk_destruct = NULL;
b0691c8e 1482 bh_unlock_sock(newsk);
87d11ceb
ACM		 1483 sk_free(newsk);
1484 newsk = NULL;
1485 goto out;
1486 }
1487
1488 newsk->sk_err = 0;
1489 newsk->sk_priority = 0;
4dc6dc71
ED		 1490 /*
1491 * Before updating sk_refcnt, we must commit prior changes to memory
1492 * (Documentation/RCU/rculist_nulls.txt for details)
1493 */
1494 smp_wmb();
87d11ceb
ACM		 1495 atomic_set(&newsk->sk_refcnt, 2);
1496
1497 /*
1498 * Increment the counter in the same struct proto as the master
1499 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1500 * is the same as sk->sk_prot->socks, as this field was copied
1501 * with memcpy).
1502 *
1503 * This _changes_ the previous behaviour, where
1504 * tcp_create_openreq_child always was incrementing the
1505 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1506 * to be taken into account in all callers. -acme
1507 */
1508 sk_refcnt_debug_inc(newsk);
972692e0 1509 sk_set_socket(newsk, NULL);
43815482 1510 newsk->sk_wq = NULL;
87d11ceb 1511
f3f511e1
GC		 1512 sk_update_clone(sk, newsk);
1513
87d11ceb 1514 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1515 sk_sockets_allocated_inc(newsk);
704da560 1516
08e29af3 1517 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1518 net_enable_timestamp();
87d11ceb
ACM		 1519 }
1520out:
1521 return newsk;
1522}
e56c57d0 1523EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1524
9958089a
AK		 1525void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1526{
1527 __sk_dst_set(sk, dst);
1528 sk->sk_route_caps = dst->dev->features;
1529 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1530 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1531 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1532 if (sk_can_gso(sk)) {
82cc1a7a 1533 if (dst->header_len) {
9958089a 1534 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1535 } else {
9958089a 1536 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1537 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1538 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1539 }
9958089a
AK		 1540 }
1541}
1542EXPORT_SYMBOL_GPL(sk_setup_caps);
1543
1da177e4
LT		 1544/*
1545 * Simple resource managers for sockets.
1546 */
1547
1548
4ec93edb
YH		 1549/*
1550 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1551 */
1552void sock_wfree(struct sk_buff *skb)
1553{
1554 struct sock *sk = skb->sk;
d99927f4 1555 unsigned int len = skb->truesize;
1da177e4 1556
d99927f4
ED		 1557 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1558 /*
1559 * Keep a reference on sk_wmem_alloc, this will be released
1560 * after sk_write_space() call
1561 */
1562 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1563 sk->sk_write_space(sk);
d99927f4
ED		 1564 len = 1;
1565 }
2b85a34e 1566 /*
d99927f4
ED		 1567 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1568 * could not do because of in-flight packets
2b85a34e 1569 */
d99927f4 1570 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1571 __sk_free(sk);
1da177e4 1572}
2a91525c 1573EXPORT_SYMBOL(sock_wfree);
1da177e4 1574
f2f872f9
ED		 1575void skb_orphan_partial(struct sk_buff *skb)
1576{
1577 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1578 * so we do not completely orphan skb, but transfert all
1579 * accounted bytes but one, to avoid unexpected reorders.
1580 */
1581 if (skb->destructor == sock_wfree
1582#ifdef CONFIG_INET
1583 || skb->destructor == tcp_wfree
1584#endif
1585 ) {
1586 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1587 skb->truesize = 1;
1588 } else {
1589 skb_orphan(skb);
1590 }
1591}
1592EXPORT_SYMBOL(skb_orphan_partial);
1593
4ec93edb
YH		 1594/*
1595 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1596 */
1597void sock_rfree(struct sk_buff *skb)
1598{
1599 struct sock *sk = skb->sk;
d361fd59 1600 unsigned int len = skb->truesize;
1da177e4 1601
d361fd59
ED		 1602 atomic_sub(len, &sk->sk_rmem_alloc);
1603 sk_mem_uncharge(sk, len);
1da177e4 1604}
2a91525c 1605EXPORT_SYMBOL(sock_rfree);
1da177e4 1606
41063e9d
DM		 1607void sock_edemux(struct sk_buff *skb)
1608{
e812347c
ED		 1609 struct sock *sk = skb->sk;
1610
1c463e57 1611#ifdef CONFIG_INET
e812347c
ED		 1612 if (sk->sk_state == TCP_TIME_WAIT)
1613 inet_twsk_put(inet_twsk(sk));
1614 else
1c463e57 1615#endif
e812347c 1616 sock_put(sk);
41063e9d
DM		 1617}
1618EXPORT_SYMBOL(sock_edemux);
1da177e4 1619
976d0201 1620kuid_t sock_i_uid(struct sock *sk)
1da177e4 1621{
976d0201 1622 kuid_t uid;
1da177e4 1623
f064af1e 1624 read_lock_bh(&sk->sk_callback_lock);
976d0201 1625 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1626 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1627 return uid;
1628}
2a91525c 1629EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1630
1631unsigned long sock_i_ino(struct sock *sk)
1632{
1633 unsigned long ino;
1634
f064af1e 1635 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1636 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1637 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1638 return ino;
1639}
2a91525c 1640EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1641
1642/*
1643 * Allocate a skb from the socket's send buffer.
1644 */
86a76caf 1645struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1646 gfp_t priority)
1da177e4
LT		 1647{
1648 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1649 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1650 if (skb) {
1651 skb_set_owner_w(skb, sk);
1652 return skb;
1653 }
1654 }
1655 return NULL;
1656}
2a91525c 1657EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1658
4ec93edb 1659/*
1da177e4 1660 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1661 */
dd0fc66f 1662void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1663{
95c96174 1664 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1665 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1666 void *mem;
1667 /* First do the add, to avoid the race if kmalloc
4ec93edb 1668 * might sleep.
1da177e4
LT		 1669 */
1670 atomic_add(size, &sk->sk_omem_alloc);
1671 mem = kmalloc(size, priority);
1672 if (mem)
1673 return mem;
1674 atomic_sub(size, &sk->sk_omem_alloc);
1675 }
1676 return NULL;
1677}
2a91525c 1678EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1679
1680/*
1681 * Free an option memory block.
1682 */
1683void sock_kfree_s(struct sock *sk, void *mem, int size)
1684{
1685 kfree(mem);
1686 atomic_sub(size, &sk->sk_omem_alloc);
1687}
2a91525c 1688EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1689
1690/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1691 I think, these locks should be removed for datagram sockets.
1692 */
2a91525c 1693static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1694{
1695 DEFINE_WAIT(wait);
1696
1697 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1698 for (;;) {
1699 if (!timeo)
1700 break;
1701 if (signal_pending(current))
1702 break;
1703 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1704 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1705 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1706 break;
1707 if (sk->sk_shutdown & SEND_SHUTDOWN)
1708 break;
1709 if (sk->sk_err)
1710 break;
1711 timeo = schedule_timeout(timeo);
1712 }
aa395145 1713 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1714 return timeo;
1715}
1716
1717
1718/*
1719 * Generic send/receive buffer handlers
1720 */
1721
4cc7f68d
HX		 1722struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1723 unsigned long data_len, int noblock,
28d64271 1724 int *errcode, int max_page_order)
1da177e4 1725{
28d64271
ED		 1726 struct sk_buff *skb = NULL;
1727 unsigned long chunk;
7d877f3b 1728 gfp_t gfp_mask;
1da177e4
LT		 1729 long timeo;
1730 int err;
cc9b17ad 1731 int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
28d64271
ED		 1732 struct page *page;
1733 int i;
cc9b17ad
JW		 1734
1735 err = -EMSGSIZE;
1736 if (npages > MAX_SKB_FRAGS)
1737 goto failure;
1da177e4 1738
1da177e4 1739 timeo = sock_sndtimeo(sk, noblock);
28d64271 1740 while (!skb) {
1da177e4
LT		 1741 err = sock_error(sk);
1742 if (err != 0)
1743 goto failure;
1744
1745 err = -EPIPE;
1746 if (sk->sk_shutdown & SEND_SHUTDOWN)
1747 goto failure;
1748
28d64271
ED		 1749 if (atomic_read(&sk->sk_wmem_alloc) >= sk->sk_sndbuf) {
1750 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1751 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1752 err = -EAGAIN;
1753 if (!timeo)
1754 goto failure;
1755 if (signal_pending(current))
1756 goto interrupted;
1757 timeo = sock_wait_for_wmem(sk, timeo);
1758 continue;
1759 }
1da177e4 1760
28d64271
ED		 1761 err = -ENOBUFS;
1762 gfp_mask = sk->sk_allocation;
1763 if (gfp_mask & __GFP_WAIT)
1764 gfp_mask |= __GFP_REPEAT;
1765
1766 skb = alloc_skb(header_len, gfp_mask);
1767 if (!skb)
1da177e4 1768 goto failure;
28d64271
ED		 1769
1770 skb->truesize += data_len;
1771
1772 for (i = 0; npages > 0; i++) {
1773 int order = max_page_order;
1774
1775 while (order) {
1776 if (npages >= 1 << order) {
1777 page = alloc_pages(sk->sk_allocation |
ed98df33
ED		 1778 __GFP_COMP |
1779 __GFP_NOWARN |
1780 __GFP_NORETRY,
28d64271
ED		 1781 order);
1782 if (page)
1783 goto fill_page;
1784 }
1785 order--;
1786 }
1787 page = alloc_page(sk->sk_allocation);
1788 if (!page)
1789 goto failure;
1790fill_page:
1791 chunk = min_t(unsigned long, data_len,
1792 PAGE_SIZE << order);
1793 skb_fill_page_desc(skb, i, page, 0, chunk);
1794 data_len -= chunk;
1795 npages -= 1 << order;
1da177e4 1796 }
1da177e4
LT		 1797 }
1798
1799 skb_set_owner_w(skb, sk);
1800 return skb;
1801
1802interrupted:
1803 err = sock_intr_errno(timeo);
1804failure:
28d64271 1805 kfree_skb(skb);
1da177e4
LT		 1806 *errcode = err;
1807 return NULL;
1808}
4cc7f68d 1809EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1810
4ec93edb 1811struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1812 int noblock, int *errcode)
1813{
28d64271 1814 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1815}
2a91525c 1816EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1817
5640f768
ED		 1818/* On 32bit arches, an skb frag is limited to 2^15 */
1819#define SKB_FRAG_PAGE_ORDER get_order(32768)
1820
400dfd3a
ED		 1821/**
1822 * skb_page_frag_refill - check that a page_frag contains enough room
1823 * @sz: minimum size of the fragment we want to get
1824 * @pfrag: pointer to page_frag
1825 * @prio: priority for memory allocation
1826 *
1827 * Note: While this allocator tries to use high order pages, there is
1828 * no guarantee that allocations succeed. Therefore, @sz MUST be
1829 * less or equal than PAGE_SIZE.
1830 */
1831bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio)
5640f768
ED		 1832{
1833 int order;
1834
1835 if (pfrag->page) {
1836 if (atomic_read(&pfrag->page->_count) == 1) {
1837 pfrag->offset = 0;
1838 return true;
1839 }
400dfd3a 1840 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1841 return true;
1842 put_page(pfrag->page);
1843 }
1844
097b4f19 1845 order = SKB_FRAG_PAGE_ORDER;
5640f768 1846 do {
400dfd3a 1847 gfp_t gfp = prio;
5640f768
ED		 1848
1849 if (order)
ed98df33 1850 gfp |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY;
5640f768
ED		 1851 pfrag->page = alloc_pages(gfp, order);
1852 if (likely(pfrag->page)) {
1853 pfrag->offset = 0;
1854 pfrag->size = PAGE_SIZE << order;
1855 return true;
1856 }
1857 } while (--order >= 0);
1858
400dfd3a
ED		 1859 return false;
1860}
1861EXPORT_SYMBOL(skb_page_frag_refill);
1862
1863bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1864{
1865 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1866 return true;
1867
5640f768
ED		 1868 sk_enter_memory_pressure(sk);
1869 sk_stream_moderate_sndbuf(sk);
1870 return false;
1871}
1872EXPORT_SYMBOL(sk_page_frag_refill);
1873
1da177e4 1874static void __lock_sock(struct sock *sk)
f39234d6
NK		 1875 __releases(&sk->sk_lock.slock)
1876 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1877{
1878 DEFINE_WAIT(wait);
1879
e71a4783 1880 for (;;) {
1da177e4
LT		 1881 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1882 TASK_UNINTERRUPTIBLE);
1883 spin_unlock_bh(&sk->sk_lock.slock);
1884 schedule();
1885 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1886 if (!sock_owned_by_user(sk))
1da177e4
LT		 1887 break;
1888 }
1889 finish_wait(&sk->sk_lock.wq, &wait);
1890}
1891
1892static void __release_sock(struct sock *sk)
f39234d6
NK		 1893 __releases(&sk->sk_lock.slock)
1894 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1895{
1896 struct sk_buff *skb = sk->sk_backlog.head;
1897
1898 do {
1899 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1900 bh_unlock_sock(sk);
1901
1902 do {
1903 struct sk_buff *next = skb->next;
1904
e4cbb02a 1905 prefetch(next);
7fee226a 1906 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1907 skb->next = NULL;
c57943a1 1908 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1909
1910 /*
1911 * We are in process context here with softirqs
1912 * disabled, use cond_resched_softirq() to preempt.
1913 * This is safe to do because we've taken the backlog
1914 * queue private:
1915 */
1916 cond_resched_softirq();
1917
1918 skb = next;
1919 } while (skb != NULL);
1920
1921 bh_lock_sock(sk);
e71a4783 1922 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1923
1924 /*
1925 * Doing the zeroing here guarantee we can not loop forever
1926 * while a wild producer attempts to flood us.
1927 */
1928 sk->sk_backlog.len = 0;
1da177e4
LT		 1929}
1930
1931/**
1932 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1933 * @sk: sock to wait on
1934 * @timeo: for how long
1da177e4
LT		 1935 *
1936 * Now socket state including sk->sk_err is changed only under lock,
1937 * hence we may omit checks after joining wait queue.
1938 * We check receive queue before schedule() only as optimization;
1939 * it is very likely that release_sock() added new data.
1940 */
1941int sk_wait_data(struct sock *sk, long *timeo)
1942{
1943 int rc;
1944 DEFINE_WAIT(wait);
1945
aa395145 1946 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1947 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1948 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1949 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1950 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1951 return rc;
1952}
1da177e4
LT		 1953EXPORT_SYMBOL(sk_wait_data);
1954
3ab224be
HA		 1955/**
1956 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1957 * @sk: socket
1958 * @size: memory size to allocate
1959 * @kind: allocation type
1960 *
1961 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1962 * rmem allocation. This function assumes that protocols which have
1963 * memory_pressure use sk_wmem_queued as write buffer accounting.
1964 */
1965int __sk_mem_schedule(struct sock *sk, int size, int kind)
1966{
1967 struct proto *prot = sk->sk_prot;
1968 int amt = sk_mem_pages(size);
8d987e5c 1969 long allocated;
e1aab161 1970 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1971
1972 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1973
e1aab161 1974 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1975
1976 /* Under limit. */
e1aab161
GC		 1977 if (parent_status == UNDER_LIMIT &&
1978 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1979 sk_leave_memory_pressure(sk);
3ab224be
HA		 1980 return 1;
1981 }
1982
e1aab161
GC		 1983 /* Under pressure. (we or our parents) */
1984 if ((parent_status > SOFT_LIMIT) ||
1985 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1986 sk_enter_memory_pressure(sk);
3ab224be 1987
e1aab161
GC		 1988 /* Over hard limit (we or our parents) */
1989 if ((parent_status == OVER_LIMIT) ||
1990 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1991 goto suppress_allocation;
1992
1993 /* guarantee minimum buffer size under pressure */
1994 if (kind == SK_MEM_RECV) {
1995 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1996 return 1;
180d8cd9 1997
3ab224be
HA		 1998 } else { /* SK_MEM_SEND */
1999 if (sk->sk_type == SOCK_STREAM) {
2000 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2001 return 1;
2002 } else if (atomic_read(&sk->sk_wmem_alloc) <
2003 prot->sysctl_wmem[0])
2004 return 1;
2005 }
2006
180d8cd9 2007 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2008 int alloc;
2009
180d8cd9 2010 if (!sk_under_memory_pressure(sk))
1748376b 2011 return 1;
180d8cd9
GC		 2012 alloc = sk_sockets_allocated_read_positive(sk);
2013 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2014 sk_mem_pages(sk->sk_wmem_queued +
2015 atomic_read(&sk->sk_rmem_alloc) +
2016 sk->sk_forward_alloc))
2017 return 1;
2018 }
2019
2020suppress_allocation:
2021
2022 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2023 sk_stream_moderate_sndbuf(sk);
2024
2025 /* Fail only if socket is _under_ its sndbuf.
2026 * In this case we cannot block, so that we have to fail.
2027 */
2028 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2029 return 1;
2030 }
2031
3847ce32
SM		 2032 trace_sock_exceed_buf_limit(sk, prot, allocated);
2033
3ab224be
HA		 2034 /* Alas. Undo changes. */
2035 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2036
0e90b31f 2037 sk_memory_allocated_sub(sk, amt);
180d8cd9 2038
3ab224be
HA		 2039 return 0;
2040}
3ab224be
HA		 2041EXPORT_SYMBOL(__sk_mem_schedule);
2042
2043/**
2044 * __sk_reclaim - reclaim memory_allocated
2045 * @sk: socket
2046 */
2047void __sk_mem_reclaim(struct sock *sk)
2048{
180d8cd9 2049 sk_memory_allocated_sub(sk,
0e90b31f 2050 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 2051 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
2052
180d8cd9
GC		 2053 if (sk_under_memory_pressure(sk) &&
2054 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2055 sk_leave_memory_pressure(sk);
3ab224be 2056}
3ab224be
HA		 2057EXPORT_SYMBOL(__sk_mem_reclaim);
2058
2059
1da177e4
LT		 2060/*
2061 * Set of default routines for initialising struct proto_ops when
2062 * the protocol does not support a particular function. In certain
2063 * cases where it makes no sense for a protocol to have a "do nothing"
2064 * function, some default processing is provided.
2065 */
2066
2067int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2068{
2069 return -EOPNOTSUPP;
2070}
2a91525c 2071EXPORT_SYMBOL(sock_no_bind);
1da177e4 2072
4ec93edb 2073int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2074 int len, int flags)
2075{
2076 return -EOPNOTSUPP;
2077}
2a91525c 2078EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2079
2080int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2081{
2082 return -EOPNOTSUPP;
2083}
2a91525c 2084EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2085
2086int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2087{
2088 return -EOPNOTSUPP;
2089}
2a91525c 2090EXPORT_SYMBOL(sock_no_accept);
1da177e4 2091
4ec93edb 2092int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2093 int *len, int peer)
2094{
2095 return -EOPNOTSUPP;
2096}
2a91525c 2097EXPORT_SYMBOL(sock_no_getname);
1da177e4 2098
2a91525c 2099unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2100{
2101 return 0;
2102}
2a91525c 2103EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2104
2105int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2106{
2107 return -EOPNOTSUPP;
2108}
2a91525c 2109EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2110
2111int sock_no_listen(struct socket *sock, int backlog)
2112{
2113 return -EOPNOTSUPP;
2114}
2a91525c 2115EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2116
2117int sock_no_shutdown(struct socket *sock, int how)
2118{
2119 return -EOPNOTSUPP;
2120}
2a91525c 2121EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2122
2123int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2124 char __user *optval, unsigned int optlen)
1da177e4
LT		 2125{
2126 return -EOPNOTSUPP;
2127}
2a91525c 2128EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2129
2130int sock_no_getsockopt(struct socket *sock, int level, int optname,
2131 char __user *optval, int __user *optlen)
2132{
2133 return -EOPNOTSUPP;
2134}
2a91525c 2135EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2136
2137int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2138 size_t len)
2139{
2140 return -EOPNOTSUPP;
2141}
2a91525c 2142EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2143
2144int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2145 size_t len, int flags)
2146{
2147 return -EOPNOTSUPP;
2148}
2a91525c 2149EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2150
2151int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2152{
2153 /* Mirror missing mmap method error code */
2154 return -ENODEV;
2155}
2a91525c 2156EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2157
2158ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2159{
2160 ssize_t res;
2161 struct msghdr msg = {.msg_flags = flags};
2162 struct kvec iov;
2163 char *kaddr = kmap(page);
2164 iov.iov_base = kaddr + offset;
2165 iov.iov_len = size;
2166 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2167 kunmap(page);
2168 return res;
2169}
2a91525c 2170EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2171
2172/*
2173 * Default Socket Callbacks
2174 */
2175
2176static void sock_def_wakeup(struct sock *sk)
2177{
43815482
ED		 2178 struct socket_wq *wq;
2179
2180 rcu_read_lock();
2181 wq = rcu_dereference(sk->sk_wq);
2182 if (wq_has_sleeper(wq))
2183 wake_up_interruptible_all(&wq->wait);
2184 rcu_read_unlock();
1da177e4
LT		 2185}
2186
2187static void sock_def_error_report(struct sock *sk)
2188{
43815482
ED		 2189 struct socket_wq *wq;
2190
2191 rcu_read_lock();
2192 wq = rcu_dereference(sk->sk_wq);
2193 if (wq_has_sleeper(wq))
2194 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2195 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2196 rcu_read_unlock();
1da177e4
LT		 2197}
2198
2199static void sock_def_readable(struct sock *sk, int len)
2200{
43815482
ED		 2201 struct socket_wq *wq;
2202
2203 rcu_read_lock();
2204 wq = rcu_dereference(sk->sk_wq);
2205 if (wq_has_sleeper(wq))
2c6607c6 2206 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2207 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2208 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2209 rcu_read_unlock();
1da177e4
LT		 2210}
2211
2212static void sock_def_write_space(struct sock *sk)
2213{
43815482
ED		 2214 struct socket_wq *wq;
2215
2216 rcu_read_lock();
1da177e4
LT		 2217
2218 /* Do not wake up a writer until he can make "significant"
2219 * progress. --DaveM
2220 */
e71a4783 2221 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2222 wq = rcu_dereference(sk->sk_wq);
2223 if (wq_has_sleeper(wq))
2224 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2225 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2226
2227 /* Should agree with poll, otherwise some programs break */
2228 if (sock_writeable(sk))
8d8ad9d7 2229 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2230 }
2231
43815482 2232 rcu_read_unlock();
1da177e4
LT		 2233}
2234
2235static void sock_def_destruct(struct sock *sk)
2236{
a51482bd 2237 kfree(sk->sk_protinfo);
1da177e4
LT		 2238}
2239
2240void sk_send_sigurg(struct sock *sk)
2241{
2242 if (sk->sk_socket && sk->sk_socket->file)
2243 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2244 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2245}
2a91525c 2246EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2247
2248void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2249 unsigned long expires)
2250{
2251 if (!mod_timer(timer, expires))
2252 sock_hold(sk);
2253}
1da177e4
LT		 2254EXPORT_SYMBOL(sk_reset_timer);
2255
2256void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2257{
25cc4ae9 2258 if (del_timer(timer))
1da177e4
LT		 2259 __sock_put(sk);
2260}
1da177e4
LT		 2261EXPORT_SYMBOL(sk_stop_timer);
2262
2263void sock_init_data(struct socket *sock, struct sock *sk)
2264{
2265 skb_queue_head_init(&sk->sk_receive_queue);
2266 skb_queue_head_init(&sk->sk_write_queue);
2267 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2268#ifdef CONFIG_NET_DMA
2269 skb_queue_head_init(&sk->sk_async_wait_queue);
2270#endif
1da177e4
LT		 2271
2272 sk->sk_send_head = NULL;
2273
2274 init_timer(&sk->sk_timer);
4ec93edb 2275
1da177e4
LT		 2276 sk->sk_allocation = GFP_KERNEL;
2277 sk->sk_rcvbuf = sysctl_rmem_default;
2278 sk->sk_sndbuf = sysctl_wmem_default;
2279 sk->sk_state = TCP_CLOSE;
972692e0 2280 sk_set_socket(sk, sock);
1da177e4
LT		 2281
2282 sock_set_flag(sk, SOCK_ZAPPED);
2283
e71a4783 2284 if (sock) {
1da177e4 2285 sk->sk_type = sock->type;
43815482 2286 sk->sk_wq = sock->wq;
1da177e4
LT		 2287 sock->sk = sk;
2288 } else
43815482 2289 sk->sk_wq = NULL;
1da177e4 2290
b6c6712a 2291 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2292 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2293 lockdep_set_class_and_name(&sk->sk_callback_lock,
2294 af_callback_keys + sk->sk_family,
2295 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2296
2297 sk->sk_state_change = sock_def_wakeup;
2298 sk->sk_data_ready = sock_def_readable;
2299 sk->sk_write_space = sock_def_write_space;
2300 sk->sk_error_report = sock_def_error_report;
2301 sk->sk_destruct = sock_def_destruct;
2302
5640f768
ED		 2303 sk->sk_frag.page = NULL;
2304 sk->sk_frag.offset = 0;
ef64a54f 2305 sk->sk_peek_off = -1;
1da177e4 2306
109f6e39
EB		 2307 sk->sk_peer_pid = NULL;
2308 sk->sk_peer_cred = NULL;
1da177e4
LT		 2309 sk->sk_write_pending = 0;
2310 sk->sk_rcvlowat = 1;
2311 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2312 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2313
f37f0afb 2314 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2315
e0d1095a 2316#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2317 sk->sk_napi_id = 0;
64b0dc51 2318 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2319#endif
2320
62748f32 2321 sk->sk_max_pacing_rate = ~0U;
7eec4174 2322 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2323 /*
2324 * Before updating sk_refcnt, we must commit prior changes to memory
2325 * (Documentation/RCU/rculist_nulls.txt for details)
2326 */
2327 smp_wmb();
1da177e4 2328 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2329 atomic_set(&sk->sk_drops, 0);
1da177e4 2330}
2a91525c 2331EXPORT_SYMBOL(sock_init_data);
1da177e4 2332
b5606c2d 2333void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2334{
2335 might_sleep();
a5b5bb9a 2336 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2337 if (sk->sk_lock.owned)
1da177e4 2338 __lock_sock(sk);
d2e9117c 2339 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2340 spin_unlock(&sk->sk_lock.slock);
2341 /*
2342 * The sk_lock has mutex_lock() semantics here:
2343 */
fcc70d5f 2344 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2345 local_bh_enable();
1da177e4 2346}
fcc70d5f 2347EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2348
b5606c2d 2349void release_sock(struct sock *sk)
1da177e4 2350{
a5b5bb9a
IM		 2351 /*
2352 * The sk_lock has mutex_unlock() semantics:
2353 */
2354 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2355
2356 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2357 if (sk->sk_backlog.tail)
2358 __release_sock(sk);
46d3ceab
ED		 2359
2360 if (sk->sk_prot->release_cb)
2361 sk->sk_prot->release_cb(sk);
2362
d2e9117c 2363 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2364 if (waitqueue_active(&sk->sk_lock.wq))
2365 wake_up(&sk->sk_lock.wq);
2366 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2367}
2368EXPORT_SYMBOL(release_sock);
2369
8a74ad60
ED		 2370/**
2371 * lock_sock_fast - fast version of lock_sock
2372 * @sk: socket
2373 *
2374 * This version should be used for very small section, where process wont block
2375 * return false if fast path is taken
2376 * sk_lock.slock locked, owned = 0, BH disabled
2377 * return true if slow path is taken
2378 * sk_lock.slock unlocked, owned = 1, BH enabled
2379 */
2380bool lock_sock_fast(struct sock *sk)
2381{
2382 might_sleep();
2383 spin_lock_bh(&sk->sk_lock.slock);
2384
2385 if (!sk->sk_lock.owned)
2386 /*
2387 * Note : We must disable BH
2388 */
2389 return false;
2390
2391 __lock_sock(sk);
2392 sk->sk_lock.owned = 1;
2393 spin_unlock(&sk->sk_lock.slock);
2394 /*
2395 * The sk_lock has mutex_lock() semantics here:
2396 */
2397 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2398 local_bh_enable();
2399 return true;
2400}
2401EXPORT_SYMBOL(lock_sock_fast);
2402
1da177e4 2403int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2404{
b7aa0bf7 2405 struct timeval tv;
1da177e4 2406 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2407 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2408 tv = ktime_to_timeval(sk->sk_stamp);
2409 if (tv.tv_sec == -1)
1da177e4 2410 return -ENOENT;
b7aa0bf7
ED		 2411 if (tv.tv_sec == 0) {
2412 sk->sk_stamp = ktime_get_real();
2413 tv = ktime_to_timeval(sk->sk_stamp);
2414 }
2415 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2416}
1da177e4
LT		 2417EXPORT_SYMBOL(sock_get_timestamp);
2418
ae40eb1e
ED		 2419int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2420{
2421 struct timespec ts;
2422 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2423 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2424 ts = ktime_to_timespec(sk->sk_stamp);
2425 if (ts.tv_sec == -1)
2426 return -ENOENT;
2427 if (ts.tv_sec == 0) {
2428 sk->sk_stamp = ktime_get_real();
2429 ts = ktime_to_timespec(sk->sk_stamp);
2430 }
2431 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2432}
2433EXPORT_SYMBOL(sock_get_timestampns);
2434
20d49473 2435void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2436{
20d49473 2437 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2438 unsigned long previous_flags = sk->sk_flags;
2439
20d49473
PO		 2440 sock_set_flag(sk, flag);
2441 /*
2442 * we just set one of the two flags which require net
2443 * time stamping, but time stamping might have been on
2444 * already because of the other one
2445 */
08e29af3 2446 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2447 net_enable_timestamp();
1da177e4
LT		 2448 }
2449}
1da177e4 2450
cb820f8e
RC		 2451int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2452 int level, int type)
2453{
2454 struct sock_exterr_skb *serr;
2455 struct sk_buff *skb, *skb2;
2456 int copied, err;
2457
2458 err = -EAGAIN;
2459 skb = skb_dequeue(&sk->sk_error_queue);
2460 if (skb == NULL)
2461 goto out;
2462
2463 copied = skb->len;
2464 if (copied > len) {
2465 msg->msg_flags |= MSG_TRUNC;
2466 copied = len;
2467 }
2468 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2469 if (err)
2470 goto out_free_skb;
2471
2472 sock_recv_timestamp(msg, sk, skb);
2473
2474 serr = SKB_EXT_ERR(skb);
2475 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2476
2477 msg->msg_flags |= MSG_ERRQUEUE;
2478 err = copied;
2479
2480 /* Reset and regenerate socket error */
2481 spin_lock_bh(&sk->sk_error_queue.lock);
2482 sk->sk_err = 0;
2483 if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) {
2484 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
2485 spin_unlock_bh(&sk->sk_error_queue.lock);
2486 sk->sk_error_report(sk);
2487 } else
2488 spin_unlock_bh(&sk->sk_error_queue.lock);
2489
2490out_free_skb:
2491 kfree_skb(skb);
2492out:
2493 return err;
2494}
2495EXPORT_SYMBOL(sock_recv_errqueue);
2496
1da177e4
LT		 2497/*
2498 * Get a socket option on an socket.
2499 *
2500 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2501 * asynchronous errors should be reported by getsockopt. We assume
2502 * this means if you specify SO_ERROR (otherwise whats the point of it).
2503 */
2504int sock_common_getsockopt(struct socket *sock, int level, int optname,
2505 char __user *optval, int __user *optlen)
2506{
2507 struct sock *sk = sock->sk;
2508
2509 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2510}
1da177e4
LT		 2511EXPORT_SYMBOL(sock_common_getsockopt);
2512
3fdadf7d 2513#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2514int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2515 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2516{
2517 struct sock *sk = sock->sk;
2518
1e51f951 2519 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2520 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2521 optval, optlen);
3fdadf7d
DM		 2522 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2523}
2524EXPORT_SYMBOL(compat_sock_common_getsockopt);
2525#endif
2526
1da177e4
LT		 2527int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2528 struct msghdr *msg, size_t size, int flags)
2529{
2530 struct sock *sk = sock->sk;
2531 int addr_len = 0;
2532 int err;
2533
2534 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2535 flags & ~MSG_DONTWAIT, &addr_len);
2536 if (err >= 0)
2537 msg->msg_namelen = addr_len;
2538 return err;
2539}
1da177e4
LT		 2540EXPORT_SYMBOL(sock_common_recvmsg);
2541
2542/*
2543 * Set socket options on an inet socket.
2544 */
2545int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2546 char __user *optval, unsigned int optlen)
1da177e4
LT		 2547{
2548 struct sock *sk = sock->sk;
2549
2550 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2551}
1da177e4
LT		 2552EXPORT_SYMBOL(sock_common_setsockopt);
2553
3fdadf7d 2554#ifdef CONFIG_COMPAT
543d9cfe 2555int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2556 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2557{
2558 struct sock *sk = sock->sk;
2559
543d9cfe
ACM		 2560 if (sk->sk_prot->compat_setsockopt != NULL)
2561 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2562 optval, optlen);
3fdadf7d
DM		 2563 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2564}
2565EXPORT_SYMBOL(compat_sock_common_setsockopt);
2566#endif
2567
1da177e4
LT		 2568void sk_common_release(struct sock *sk)
2569{
2570 if (sk->sk_prot->destroy)
2571 sk->sk_prot->destroy(sk);
2572
2573 /*
2574 * Observation: when sock_common_release is called, processes have
2575 * no access to socket. But net still has.
2576 * Step one, detach it from networking:
2577 *
2578 * A. Remove from hash tables.
2579 */
2580
2581 sk->sk_prot->unhash(sk);
2582
2583 /*
2584 * In this point socket cannot receive new packets, but it is possible
2585 * that some packets are in flight because some CPU runs receiver and
2586 * did hash table lookup before we unhashed socket. They will achieve
2587 * receive queue and will be purged by socket destructor.
2588 *
2589 * Also we still have packets pending on receive queue and probably,
2590 * our own packets waiting in device queues. sock_destroy will drain
2591 * receive queue, but transmitted packets will delay socket destruction
2592 * until the last reference will be released.
2593 */
2594
2595 sock_orphan(sk);
2596
2597 xfrm_sk_free_policy(sk);
2598
e6848976 2599 sk_refcnt_debug_release(sk);
5640f768
ED		 2600
2601 if (sk->sk_frag.page) {
2602 put_page(sk->sk_frag.page);
2603 sk->sk_frag.page = NULL;
2604 }
2605
1da177e4
LT		 2606 sock_put(sk);
2607}
1da177e4
LT		 2608EXPORT_SYMBOL(sk_common_release);
2609
13ff3d6f
PE		 2610#ifdef CONFIG_PROC_FS
2611#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2612struct prot_inuse {
2613 int val[PROTO_INUSE_NR];
2614};
13ff3d6f
PE		 2615
2616static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2617
2618#ifdef CONFIG_NET_NS
2619void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2620{
d6d9ca0f 2621 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2622}
2623EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2624
2625int sock_prot_inuse_get(struct net *net, struct proto *prot)
2626{
2627 int cpu, idx = prot->inuse_idx;
2628 int res = 0;
2629
2630 for_each_possible_cpu(cpu)
2631 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2632
2633 return res >= 0 ? res : 0;
2634}
2635EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2636
2c8c1e72 2637static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2638{
2639 net->core.inuse = alloc_percpu(struct prot_inuse);
2640 return net->core.inuse ? 0 : -ENOMEM;
2641}
2642
2c8c1e72 2643static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2644{
2645 free_percpu(net->core.inuse);
2646}
2647
2648static struct pernet_operations net_inuse_ops = {
2649 .init = sock_inuse_init_net,
2650 .exit = sock_inuse_exit_net,
2651};
2652
2653static __init int net_inuse_init(void)
2654{
2655 if (register_pernet_subsys(&net_inuse_ops))
2656 panic("Cannot initialize net inuse counters");
2657
2658 return 0;
2659}
2660
2661core_initcall(net_inuse_init);
2662#else
1338d466
PE		 2663static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2664
c29a0bc4 2665void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2666{
d6d9ca0f 2667 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2668}
2669EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2670
c29a0bc4 2671int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2672{
2673 int cpu, idx = prot->inuse_idx;
2674 int res = 0;
2675
2676 for_each_possible_cpu(cpu)
2677 res += per_cpu(prot_inuse, cpu).val[idx];
2678
2679 return res >= 0 ? res : 0;
2680}
2681EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2682#endif
13ff3d6f
PE		 2683
2684static void assign_proto_idx(struct proto *prot)
2685{
2686 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2687
2688 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2689 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2690 return;
2691 }
2692
2693 set_bit(prot->inuse_idx, proto_inuse_idx);
2694}
2695
2696static void release_proto_idx(struct proto *prot)
2697{
2698 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2699 clear_bit(prot->inuse_idx, proto_inuse_idx);
2700}
2701#else
2702static inline void assign_proto_idx(struct proto *prot)
2703{
2704}
2705
2706static inline void release_proto_idx(struct proto *prot)
2707{
2708}
2709#endif
2710
b733c007
PE		 2711int proto_register(struct proto *prot, int alloc_slab)
2712{
1da177e4
LT		 2713 if (alloc_slab) {
2714 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2715 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2716 NULL);
1da177e4
LT		 2717
2718 if (prot->slab == NULL) {
e005d193
JP		 2719 pr_crit("%s: Can't create sock SLAB cache!\n",
2720 prot->name);
60e7663d 2721 goto out;
1da177e4 2722 }
2e6599cb
ACM		 2723
2724 if (prot->rsk_prot != NULL) {
faf23422 2725 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2726 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2727 goto out_free_sock_slab;
2728
7e56b5d6 2729 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2730 prot->rsk_prot->obj_size, 0,
20c2df83 2731 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2732
2733 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2734 pr_crit("%s: Can't create request sock SLAB cache!\n",
2735 prot->name);
2e6599cb
ACM		 2736 goto out_free_request_sock_slab_name;
2737 }
2738 }
8feaf0c0 2739
6d6ee43e 2740 if (prot->twsk_prot != NULL) {
faf23422 2741 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2742
7e56b5d6 2743 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2744 goto out_free_request_sock_slab;
2745
6d6ee43e 2746 prot->twsk_prot->twsk_slab =
7e56b5d6 2747 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2748 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2749 0,
2750 SLAB_HWCACHE_ALIGN |
2751 prot->slab_flags,
20c2df83 2752 NULL);
6d6ee43e 2753 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2754 goto out_free_timewait_sock_slab_name;
2755 }
1da177e4
LT		 2756 }
2757
36b77a52 2758 mutex_lock(&proto_list_mutex);
1da177e4 2759 list_add(&prot->node, &proto_list);
13ff3d6f 2760 assign_proto_idx(prot);
36b77a52 2761 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2762 return 0;
2763
8feaf0c0 2764out_free_timewait_sock_slab_name:
7e56b5d6 2765 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2766out_free_request_sock_slab:
2767 if (prot->rsk_prot && prot->rsk_prot->slab) {
2768 kmem_cache_destroy(prot->rsk_prot->slab);
2769 prot->rsk_prot->slab = NULL;
2770 }
2e6599cb 2771out_free_request_sock_slab_name:
72150e9b
DC		 2772 if (prot->rsk_prot)
2773 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2774out_free_sock_slab:
2775 kmem_cache_destroy(prot->slab);
2776 prot->slab = NULL;
b733c007
PE		 2777out:
2778 return -ENOBUFS;
1da177e4 2779}
1da177e4
LT		 2780EXPORT_SYMBOL(proto_register);
2781
2782void proto_unregister(struct proto *prot)
2783{
36b77a52 2784 mutex_lock(&proto_list_mutex);
13ff3d6f 2785 release_proto_idx(prot);
0a3f4358 2786 list_del(&prot->node);
36b77a52 2787 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2788
2789 if (prot->slab != NULL) {
2790 kmem_cache_destroy(prot->slab);
2791 prot->slab = NULL;
2792 }
2793
2e6599cb 2794 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2795 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2796 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2797 prot->rsk_prot->slab = NULL;
2798 }
2799
6d6ee43e 2800 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2801 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2802 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2803 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2804 }
1da177e4 2805}
1da177e4
LT		 2806EXPORT_SYMBOL(proto_unregister);
2807
2808#ifdef CONFIG_PROC_FS
1da177e4 2809static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2810 __acquires(proto_list_mutex)
1da177e4 2811{
36b77a52 2812 mutex_lock(&proto_list_mutex);
60f0438a 2813 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2814}
2815
2816static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2817{
60f0438a 2818 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2819}
2820
2821static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2822 __releases(proto_list_mutex)
1da177e4 2823{
36b77a52 2824 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2825}
2826
2827static char proto_method_implemented(const void *method)
2828{
2829 return method == NULL ? 'n' : 'y';
2830}
180d8cd9
GC		 2831static long sock_prot_memory_allocated(struct proto *proto)
2832{
cb75a36c 2833 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2834}
2835
2836static char *sock_prot_memory_pressure(struct proto *proto)
2837{
2838 return proto->memory_pressure != NULL ?
2839 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2840}
1da177e4
LT		 2841
2842static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2843{
180d8cd9 2844
8d987e5c 2845 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2846 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2847 proto->name,
2848 proto->obj_size,
14e943db 2849 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2850 sock_prot_memory_allocated(proto),
2851 sock_prot_memory_pressure(proto),
1da177e4
LT		 2852 proto->max_header,
2853 proto->slab == NULL ? "no" : "yes",
2854 module_name(proto->owner),
2855 proto_method_implemented(proto->close),
2856 proto_method_implemented(proto->connect),
2857 proto_method_implemented(proto->disconnect),
2858 proto_method_implemented(proto->accept),
2859 proto_method_implemented(proto->ioctl),
2860 proto_method_implemented(proto->init),
2861 proto_method_implemented(proto->destroy),
2862 proto_method_implemented(proto->shutdown),
2863 proto_method_implemented(proto->setsockopt),
2864 proto_method_implemented(proto->getsockopt),
2865 proto_method_implemented(proto->sendmsg),
2866 proto_method_implemented(proto->recvmsg),
2867 proto_method_implemented(proto->sendpage),
2868 proto_method_implemented(proto->bind),
2869 proto_method_implemented(proto->backlog_rcv),
2870 proto_method_implemented(proto->hash),
2871 proto_method_implemented(proto->unhash),
2872 proto_method_implemented(proto->get_port),
2873 proto_method_implemented(proto->enter_memory_pressure));
2874}
2875
2876static int proto_seq_show(struct seq_file *seq, void *v)
2877{
60f0438a 2878 if (v == &proto_list)
1da177e4
LT		 2879 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2880 "protocol",
2881 "size",
2882 "sockets",
2883 "memory",
2884 "press",
2885 "maxhdr",
2886 "slab",
2887 "module",
2888 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2889 else
60f0438a 2890 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2891 return 0;
2892}
2893
f690808e 2894static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2895 .start = proto_seq_start,
2896 .next = proto_seq_next,
2897 .stop = proto_seq_stop,
2898 .show = proto_seq_show,
2899};
2900
2901static int proto_seq_open(struct inode *inode, struct file *file)
2902{
14e943db
ED		 2903 return seq_open_net(inode, file, &proto_seq_ops,
2904 sizeof(struct seq_net_private));
1da177e4
LT		 2905}
2906
9a32144e 2907static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2908 .owner = THIS_MODULE,
2909 .open = proto_seq_open,
2910 .read = seq_read,
2911 .llseek = seq_lseek,
14e943db
ED		 2912 .release = seq_release_net,
2913};
2914
2915static __net_init int proto_init_net(struct net *net)
2916{
d4beaa66 2917 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2918 return -ENOMEM;
2919
2920 return 0;
2921}
2922
2923static __net_exit void proto_exit_net(struct net *net)
2924{
ece31ffd 2925 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2926}
2927
2928
2929static __net_initdata struct pernet_operations proto_net_ops = {
2930 .init = proto_init_net,
2931 .exit = proto_exit_net,
1da177e4
LT		 2932};
2933
2934static int __init proto_init(void)
2935{
14e943db 2936 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2937}
2938
2939subsys_initcall(proto_init);
2940
2941#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 1.326902 seconds and 5 git commands to generate.