projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
[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
7d4c04fc
KJ		 1170 case SO_SELECT_ERR_QUEUE:
1171 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1172 break;
1173
e0d1095a 1174#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1175 case SO_BUSY_POLL:
dafcc438
ET		 1176 v.val = sk->sk_ll_usec;
1177 break;
1178#endif
1179
62748f32
ED		 1180 case SO_MAX_PACING_RATE:
1181 v.val = sk->sk_max_pacing_rate;
1182 break;
1183
e71a4783
SH		 1184 default:
1185 return -ENOPROTOOPT;
1da177e4 1186 }
e71a4783 1187
1da177e4
LT		 1188 if (len > lv)
1189 len = lv;
1190 if (copy_to_user(optval, &v, len))
1191 return -EFAULT;
1192lenout:
4ec93edb
YH		 1193 if (put_user(len, optlen))
1194 return -EFAULT;
1195 return 0;
1da177e4
LT		 1196}
1197
a5b5bb9a
IM		 1198/*
1199 * Initialize an sk_lock.
1200 *
1201 * (We also register the sk_lock with the lock validator.)
1202 */
b6f99a21 1203static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1204{
ed07536e
PZ		 1205 sock_lock_init_class_and_name(sk,
1206 af_family_slock_key_strings[sk->sk_family],
1207 af_family_slock_keys + sk->sk_family,
1208 af_family_key_strings[sk->sk_family],
1209 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1210}
1211
4dc6dc71
ED		 1212/*
1213 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1214 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1215 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1216 */
f1a6c4da
PE		 1217static void sock_copy(struct sock *nsk, const struct sock *osk)
1218{
1219#ifdef CONFIG_SECURITY_NETWORK
1220 void *sptr = nsk->sk_security;
1221#endif
68835aba
ED		 1222 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1223
1224 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1225 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1226
f1a6c4da
PE		 1227#ifdef CONFIG_SECURITY_NETWORK
1228 nsk->sk_security = sptr;
1229 security_sk_clone(osk, nsk);
1230#endif
1231}
1232
fcbdf09d
OP		 1233void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1234{
1235 unsigned long nulls1, nulls2;
1236
1237 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1238 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1239 if (nulls1 > nulls2)
1240 swap(nulls1, nulls2);
1241
1242 if (nulls1 != 0)
1243 memset((char *)sk, 0, nulls1);
1244 memset((char *)sk + nulls1 + sizeof(void *), 0,
1245 nulls2 - nulls1 - sizeof(void *));
1246 memset((char *)sk + nulls2 + sizeof(void *), 0,
1247 size - nulls2 - sizeof(void *));
1248}
1249EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1250
2e4afe7b
PE		 1251static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1252 int family)
c308c1b2
PE		 1253{
1254 struct sock *sk;
1255 struct kmem_cache *slab;
1256
1257 slab = prot->slab;
e912b114
ED		 1258 if (slab != NULL) {
1259 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1260 if (!sk)
1261 return sk;
1262 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1263 if (prot->clear_sk)
1264 prot->clear_sk(sk, prot->obj_size);
1265 else
1266 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1267 }
fcbdf09d 1268 } else
c308c1b2
PE		 1269 sk = kmalloc(prot->obj_size, priority);
1270
2e4afe7b 1271 if (sk != NULL) {
a98b65a3
VN		 1272 kmemcheck_annotate_bitfield(sk, flags);
1273
2e4afe7b
PE		 1274 if (security_sk_alloc(sk, family, priority))
1275 goto out_free;
1276
1277 if (!try_module_get(prot->owner))
1278 goto out_free_sec;
e022f0b4 1279 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1280 }
1281
c308c1b2 1282 return sk;
2e4afe7b
PE		 1283
1284out_free_sec:
1285 security_sk_free(sk);
1286out_free:
1287 if (slab != NULL)
1288 kmem_cache_free(slab, sk);
1289 else
1290 kfree(sk);
1291 return NULL;
c308c1b2
PE		 1292}
1293
1294static void sk_prot_free(struct proto *prot, struct sock *sk)
1295{
1296 struct kmem_cache *slab;
2e4afe7b 1297 struct module *owner;
c308c1b2 1298
2e4afe7b 1299 owner = prot->owner;
c308c1b2 1300 slab = prot->slab;
2e4afe7b
PE		 1301
1302 security_sk_free(sk);
c308c1b2
PE		 1303 if (slab != NULL)
1304 kmem_cache_free(slab, sk);
1305 else
1306 kfree(sk);
2e4afe7b 1307 module_put(owner);
c308c1b2
PE		 1308}
1309
86f8515f 1310#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1311void sock_update_netprioidx(struct sock *sk)
5bc1421e 1312{
5bc1421e
NH		 1313 if (in_interrupt())
1314 return;
2b73bc65 1315
6ffd4641 1316 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1317}
1318EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1319#endif
1320
1da177e4
LT		 1321/**
1322 * sk_alloc - All socket objects are allocated here
c4ea43c5 1323 * @net: the applicable net namespace
4dc3b16b
PP		 1324 * @family: protocol family
1325 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1326 * @prot: struct proto associated with this new sock instance
1da177e4 1327 */
1b8d7ae4 1328struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1329 struct proto *prot)
1da177e4 1330{
c308c1b2 1331 struct sock *sk;
1da177e4 1332
154adbc8 1333 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1334 if (sk) {
154adbc8
PE		 1335 sk->sk_family = family;
1336 /*
1337 * See comment in struct sock definition to understand
1338 * why we need sk_prot_creator -acme
1339 */
1340 sk->sk_prot = sk->sk_prot_creator = prot;
1341 sock_lock_init(sk);
3b1e0a65 1342 sock_net_set(sk, get_net(net));
d66ee058 1343 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1344
211d2f97 1345 sock_update_classid(sk);
6ffd4641 1346 sock_update_netprioidx(sk);
1da177e4 1347 }
a79af59e 1348
2e4afe7b 1349 return sk;
1da177e4 1350}
2a91525c 1351EXPORT_SYMBOL(sk_alloc);
1da177e4 1352
2b85a34e 1353static void __sk_free(struct sock *sk)
1da177e4
LT		 1354{
1355 struct sk_filter *filter;
1da177e4
LT		 1356
1357 if (sk->sk_destruct)
1358 sk->sk_destruct(sk);
1359
a898def2
PM		 1360 filter = rcu_dereference_check(sk->sk_filter,
1361 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1362 if (filter) {
309dd5fc 1363 sk_filter_uncharge(sk, filter);
a9b3cd7f 1364 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1365 }
1366
08e29af3 1367 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1368
1369 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1370 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1371 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1372
109f6e39
EB		 1373 if (sk->sk_peer_cred)
1374 put_cred(sk->sk_peer_cred);
1375 put_pid(sk->sk_peer_pid);
3b1e0a65 1376 put_net(sock_net(sk));
c308c1b2 1377 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1378}
2b85a34e
ED		 1379
1380void sk_free(struct sock *sk)
1381{
1382 /*
25985edc 1383 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1384 * some packets are still in some tx queue.
1385 * If not null, sock_wfree() will call __sk_free(sk) later
1386 */
1387 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1388 __sk_free(sk);
1389}
2a91525c 1390EXPORT_SYMBOL(sk_free);
1da177e4 1391
edf02087 1392/*
25985edc
LDM		 1393 * Last sock_put should drop reference to sk->sk_net. It has already
1394 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1395 * is not an option.
25985edc 1396 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1397 * destroy it in the context of init_net.
1398 */
1399void sk_release_kernel(struct sock *sk)
1400{
1401 if (sk == NULL || sk->sk_socket == NULL)
1402 return;
1403
1404 sock_hold(sk);
1405 sock_release(sk->sk_socket);
65a18ec5 1406 release_net(sock_net(sk));
3b1e0a65 1407 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1408 sock_put(sk);
1409}
45af1754 1410EXPORT_SYMBOL(sk_release_kernel);
edf02087 1411
475f1b52
SR		 1412static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1413{
1414 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1415 sock_update_memcg(newsk);
1416}
1417
e56c57d0
ED		 1418/**
1419 * sk_clone_lock - clone a socket, and lock its clone
1420 * @sk: the socket to clone
1421 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1422 *
1423 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1424 */
1425struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1426{
8fd1d178 1427 struct sock *newsk;
87d11ceb 1428
8fd1d178 1429 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1430 if (newsk != NULL) {
1431 struct sk_filter *filter;
1432
892c141e 1433 sock_copy(newsk, sk);
87d11ceb
ACM		 1434
1435 /* SANITY */
3b1e0a65 1436 get_net(sock_net(newsk));
87d11ceb
ACM		 1437 sk_node_init(&newsk->sk_node);
1438 sock_lock_init(newsk);
1439 bh_lock_sock(newsk);
fa438ccf 1440 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1441 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1442
1443 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1444 /*
1445 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1446 */
1447 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1448 atomic_set(&newsk->sk_omem_alloc, 0);
1449 skb_queue_head_init(&newsk->sk_receive_queue);
1450 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1451#ifdef CONFIG_NET_DMA
1452 skb_queue_head_init(&newsk->sk_async_wait_queue);
1453#endif
87d11ceb 1454
b6c6712a 1455 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1456 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1457 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1458 af_callback_keys + newsk->sk_family,
1459 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1460
1461 newsk->sk_dst_cache = NULL;
1462 newsk->sk_wmem_queued = 0;
1463 newsk->sk_forward_alloc = 0;
1464 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1465 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1466
1467 sock_reset_flag(newsk, SOCK_DONE);
1468 skb_queue_head_init(&newsk->sk_error_queue);
1469
0d7da9dd 1470 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1471 if (filter != NULL)
1472 sk_filter_charge(newsk, filter);
1473
1474 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1475 /* It is still raw copy of parent, so invalidate
1476 * destructor and make plain sk_free() */
1477 newsk->sk_destruct = NULL;
b0691c8e 1478 bh_unlock_sock(newsk);
87d11ceb
ACM		 1479 sk_free(newsk);
1480 newsk = NULL;
1481 goto out;
1482 }
1483
1484 newsk->sk_err = 0;
1485 newsk->sk_priority = 0;
4dc6dc71
ED		 1486 /*
1487 * Before updating sk_refcnt, we must commit prior changes to memory
1488 * (Documentation/RCU/rculist_nulls.txt for details)
1489 */
1490 smp_wmb();
87d11ceb
ACM		 1491 atomic_set(&newsk->sk_refcnt, 2);
1492
1493 /*
1494 * Increment the counter in the same struct proto as the master
1495 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1496 * is the same as sk->sk_prot->socks, as this field was copied
1497 * with memcpy).
1498 *
1499 * This _changes_ the previous behaviour, where
1500 * tcp_create_openreq_child always was incrementing the
1501 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1502 * to be taken into account in all callers. -acme
1503 */
1504 sk_refcnt_debug_inc(newsk);
972692e0 1505 sk_set_socket(newsk, NULL);
43815482 1506 newsk->sk_wq = NULL;
87d11ceb 1507
f3f511e1
GC		 1508 sk_update_clone(sk, newsk);
1509
87d11ceb 1510 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1511 sk_sockets_allocated_inc(newsk);
704da560 1512
08e29af3 1513 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1514 net_enable_timestamp();
87d11ceb
ACM		 1515 }
1516out:
1517 return newsk;
1518}
e56c57d0 1519EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1520
9958089a
AK		 1521void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1522{
1523 __sk_dst_set(sk, dst);
1524 sk->sk_route_caps = dst->dev->features;
1525 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1526 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1527 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1528 if (sk_can_gso(sk)) {
82cc1a7a 1529 if (dst->header_len) {
9958089a 1530 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1531 } else {
9958089a 1532 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1533 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1534 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1535 }
9958089a
AK		 1536 }
1537}
1538EXPORT_SYMBOL_GPL(sk_setup_caps);
1539
1da177e4
LT		 1540/*
1541 * Simple resource managers for sockets.
1542 */
1543
1544
4ec93edb
YH		 1545/*
1546 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1547 */
1548void sock_wfree(struct sk_buff *skb)
1549{
1550 struct sock *sk = skb->sk;
d99927f4 1551 unsigned int len = skb->truesize;
1da177e4 1552
d99927f4
ED		 1553 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1554 /*
1555 * Keep a reference on sk_wmem_alloc, this will be released
1556 * after sk_write_space() call
1557 */
1558 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1559 sk->sk_write_space(sk);
d99927f4
ED		 1560 len = 1;
1561 }
2b85a34e 1562 /*
d99927f4
ED		 1563 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1564 * could not do because of in-flight packets
2b85a34e 1565 */
d99927f4 1566 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1567 __sk_free(sk);
1da177e4 1568}
2a91525c 1569EXPORT_SYMBOL(sock_wfree);
1da177e4 1570
f2f872f9
ED		 1571void skb_orphan_partial(struct sk_buff *skb)
1572{
1573 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1574 * so we do not completely orphan skb, but transfert all
1575 * accounted bytes but one, to avoid unexpected reorders.
1576 */
1577 if (skb->destructor == sock_wfree
1578#ifdef CONFIG_INET
1579 || skb->destructor == tcp_wfree
1580#endif
1581 ) {
1582 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1583 skb->truesize = 1;
1584 } else {
1585 skb_orphan(skb);
1586 }
1587}
1588EXPORT_SYMBOL(skb_orphan_partial);
1589
4ec93edb
YH		 1590/*
1591 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1592 */
1593void sock_rfree(struct sk_buff *skb)
1594{
1595 struct sock *sk = skb->sk;
d361fd59 1596 unsigned int len = skb->truesize;
1da177e4 1597
d361fd59
ED		 1598 atomic_sub(len, &sk->sk_rmem_alloc);
1599 sk_mem_uncharge(sk, len);
1da177e4 1600}
2a91525c 1601EXPORT_SYMBOL(sock_rfree);
1da177e4 1602
41063e9d
DM		 1603void sock_edemux(struct sk_buff *skb)
1604{
e812347c
ED		 1605 struct sock *sk = skb->sk;
1606
1c463e57 1607#ifdef CONFIG_INET
e812347c
ED		 1608 if (sk->sk_state == TCP_TIME_WAIT)
1609 inet_twsk_put(inet_twsk(sk));
1610 else
1c463e57 1611#endif
e812347c 1612 sock_put(sk);
41063e9d
DM		 1613}
1614EXPORT_SYMBOL(sock_edemux);
1da177e4 1615
976d0201 1616kuid_t sock_i_uid(struct sock *sk)
1da177e4 1617{
976d0201 1618 kuid_t uid;
1da177e4 1619
f064af1e 1620 read_lock_bh(&sk->sk_callback_lock);
976d0201 1621 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1622 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1623 return uid;
1624}
2a91525c 1625EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1626
1627unsigned long sock_i_ino(struct sock *sk)
1628{
1629 unsigned long ino;
1630
f064af1e 1631 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1632 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1633 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1634 return ino;
1635}
2a91525c 1636EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1637
1638/*
1639 * Allocate a skb from the socket's send buffer.
1640 */
86a76caf 1641struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1642 gfp_t priority)
1da177e4
LT		 1643{
1644 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1645 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1646 if (skb) {
1647 skb_set_owner_w(skb, sk);
1648 return skb;
1649 }
1650 }
1651 return NULL;
1652}
2a91525c 1653EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1654
4ec93edb 1655/*
1da177e4 1656 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1657 */
dd0fc66f 1658void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1659{
95c96174 1660 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1661 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1662 void *mem;
1663 /* First do the add, to avoid the race if kmalloc
4ec93edb 1664 * might sleep.
1da177e4
LT		 1665 */
1666 atomic_add(size, &sk->sk_omem_alloc);
1667 mem = kmalloc(size, priority);
1668 if (mem)
1669 return mem;
1670 atomic_sub(size, &sk->sk_omem_alloc);
1671 }
1672 return NULL;
1673}
2a91525c 1674EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1675
1676/*
1677 * Free an option memory block.
1678 */
1679void sock_kfree_s(struct sock *sk, void *mem, int size)
1680{
1681 kfree(mem);
1682 atomic_sub(size, &sk->sk_omem_alloc);
1683}
2a91525c 1684EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1685
1686/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1687 I think, these locks should be removed for datagram sockets.
1688 */
2a91525c 1689static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1690{
1691 DEFINE_WAIT(wait);
1692
1693 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1694 for (;;) {
1695 if (!timeo)
1696 break;
1697 if (signal_pending(current))
1698 break;
1699 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1700 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1701 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1702 break;
1703 if (sk->sk_shutdown & SEND_SHUTDOWN)
1704 break;
1705 if (sk->sk_err)
1706 break;
1707 timeo = schedule_timeout(timeo);
1708 }
aa395145 1709 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1710 return timeo;
1711}
1712
1713
1714/*
1715 * Generic send/receive buffer handlers
1716 */
1717
4cc7f68d
HX		 1718struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1719 unsigned long data_len, int noblock,
28d64271 1720 int *errcode, int max_page_order)
1da177e4 1721{
28d64271
ED		 1722 struct sk_buff *skb = NULL;
1723 unsigned long chunk;
7d877f3b 1724 gfp_t gfp_mask;
1da177e4
LT		 1725 long timeo;
1726 int err;
cc9b17ad 1727 int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
28d64271
ED		 1728 struct page *page;
1729 int i;
cc9b17ad
JW		 1730
1731 err = -EMSGSIZE;
1732 if (npages > MAX_SKB_FRAGS)
1733 goto failure;
1da177e4 1734
1da177e4 1735 timeo = sock_sndtimeo(sk, noblock);
28d64271 1736 while (!skb) {
1da177e4
LT		 1737 err = sock_error(sk);
1738 if (err != 0)
1739 goto failure;
1740
1741 err = -EPIPE;
1742 if (sk->sk_shutdown & SEND_SHUTDOWN)
1743 goto failure;
1744
28d64271
ED		 1745 if (atomic_read(&sk->sk_wmem_alloc) >= sk->sk_sndbuf) {
1746 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1747 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1748 err = -EAGAIN;
1749 if (!timeo)
1750 goto failure;
1751 if (signal_pending(current))
1752 goto interrupted;
1753 timeo = sock_wait_for_wmem(sk, timeo);
1754 continue;
1755 }
1da177e4 1756
28d64271
ED		 1757 err = -ENOBUFS;
1758 gfp_mask = sk->sk_allocation;
1759 if (gfp_mask & __GFP_WAIT)
1760 gfp_mask |= __GFP_REPEAT;
1761
1762 skb = alloc_skb(header_len, gfp_mask);
1763 if (!skb)
1da177e4 1764 goto failure;
28d64271
ED		 1765
1766 skb->truesize += data_len;
1767
1768 for (i = 0; npages > 0; i++) {
1769 int order = max_page_order;
1770
1771 while (order) {
1772 if (npages >= 1 << order) {
1773 page = alloc_pages(sk->sk_allocation |
1774 __GFP_COMP | __GFP_NOWARN,
1775 order);
1776 if (page)
1777 goto fill_page;
1778 }
1779 order--;
1780 }
1781 page = alloc_page(sk->sk_allocation);
1782 if (!page)
1783 goto failure;
1784fill_page:
1785 chunk = min_t(unsigned long, data_len,
1786 PAGE_SIZE << order);
1787 skb_fill_page_desc(skb, i, page, 0, chunk);
1788 data_len -= chunk;
1789 npages -= 1 << order;
1da177e4 1790 }
1da177e4
LT		 1791 }
1792
1793 skb_set_owner_w(skb, sk);
1794 return skb;
1795
1796interrupted:
1797 err = sock_intr_errno(timeo);
1798failure:
28d64271 1799 kfree_skb(skb);
1da177e4
LT		 1800 *errcode = err;
1801 return NULL;
1802}
4cc7f68d 1803EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1804
4ec93edb 1805struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1806 int noblock, int *errcode)
1807{
28d64271 1808 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1809}
2a91525c 1810EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1811
5640f768
ED		 1812/* On 32bit arches, an skb frag is limited to 2^15 */
1813#define SKB_FRAG_PAGE_ORDER get_order(32768)
1814
400dfd3a
ED		 1815/**
1816 * skb_page_frag_refill - check that a page_frag contains enough room
1817 * @sz: minimum size of the fragment we want to get
1818 * @pfrag: pointer to page_frag
1819 * @prio: priority for memory allocation
1820 *
1821 * Note: While this allocator tries to use high order pages, there is
1822 * no guarantee that allocations succeed. Therefore, @sz MUST be
1823 * less or equal than PAGE_SIZE.
1824 */
1825bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio)
5640f768
ED		 1826{
1827 int order;
1828
1829 if (pfrag->page) {
1830 if (atomic_read(&pfrag->page->_count) == 1) {
1831 pfrag->offset = 0;
1832 return true;
1833 }
400dfd3a 1834 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1835 return true;
1836 put_page(pfrag->page);
1837 }
1838
097b4f19 1839 order = SKB_FRAG_PAGE_ORDER;
5640f768 1840 do {
400dfd3a 1841 gfp_t gfp = prio;
5640f768
ED		 1842
1843 if (order)
1844 gfp |= __GFP_COMP | __GFP_NOWARN;
1845 pfrag->page = alloc_pages(gfp, order);
1846 if (likely(pfrag->page)) {
1847 pfrag->offset = 0;
1848 pfrag->size = PAGE_SIZE << order;
1849 return true;
1850 }
1851 } while (--order >= 0);
1852
400dfd3a
ED		 1853 return false;
1854}
1855EXPORT_SYMBOL(skb_page_frag_refill);
1856
1857bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1858{
1859 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1860 return true;
1861
5640f768
ED		 1862 sk_enter_memory_pressure(sk);
1863 sk_stream_moderate_sndbuf(sk);
1864 return false;
1865}
1866EXPORT_SYMBOL(sk_page_frag_refill);
1867
1da177e4 1868static void __lock_sock(struct sock *sk)
f39234d6
NK		 1869 __releases(&sk->sk_lock.slock)
1870 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1871{
1872 DEFINE_WAIT(wait);
1873
e71a4783 1874 for (;;) {
1da177e4
LT		 1875 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1876 TASK_UNINTERRUPTIBLE);
1877 spin_unlock_bh(&sk->sk_lock.slock);
1878 schedule();
1879 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1880 if (!sock_owned_by_user(sk))
1da177e4
LT		 1881 break;
1882 }
1883 finish_wait(&sk->sk_lock.wq, &wait);
1884}
1885
1886static void __release_sock(struct sock *sk)
f39234d6
NK		 1887 __releases(&sk->sk_lock.slock)
1888 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1889{
1890 struct sk_buff *skb = sk->sk_backlog.head;
1891
1892 do {
1893 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1894 bh_unlock_sock(sk);
1895
1896 do {
1897 struct sk_buff *next = skb->next;
1898
e4cbb02a 1899 prefetch(next);
7fee226a 1900 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1901 skb->next = NULL;
c57943a1 1902 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1903
1904 /*
1905 * We are in process context here with softirqs
1906 * disabled, use cond_resched_softirq() to preempt.
1907 * This is safe to do because we've taken the backlog
1908 * queue private:
1909 */
1910 cond_resched_softirq();
1911
1912 skb = next;
1913 } while (skb != NULL);
1914
1915 bh_lock_sock(sk);
e71a4783 1916 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1917
1918 /*
1919 * Doing the zeroing here guarantee we can not loop forever
1920 * while a wild producer attempts to flood us.
1921 */
1922 sk->sk_backlog.len = 0;
1da177e4
LT		 1923}
1924
1925/**
1926 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1927 * @sk: sock to wait on
1928 * @timeo: for how long
1da177e4
LT		 1929 *
1930 * Now socket state including sk->sk_err is changed only under lock,
1931 * hence we may omit checks after joining wait queue.
1932 * We check receive queue before schedule() only as optimization;
1933 * it is very likely that release_sock() added new data.
1934 */
1935int sk_wait_data(struct sock *sk, long *timeo)
1936{
1937 int rc;
1938 DEFINE_WAIT(wait);
1939
aa395145 1940 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1941 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1942 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1943 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1944 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1945 return rc;
1946}
1da177e4
LT		 1947EXPORT_SYMBOL(sk_wait_data);
1948
3ab224be
HA		 1949/**
1950 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1951 * @sk: socket
1952 * @size: memory size to allocate
1953 * @kind: allocation type
1954 *
1955 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1956 * rmem allocation. This function assumes that protocols which have
1957 * memory_pressure use sk_wmem_queued as write buffer accounting.
1958 */
1959int __sk_mem_schedule(struct sock *sk, int size, int kind)
1960{
1961 struct proto *prot = sk->sk_prot;
1962 int amt = sk_mem_pages(size);
8d987e5c 1963 long allocated;
e1aab161 1964 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1965
1966 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1967
e1aab161 1968 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1969
1970 /* Under limit. */
e1aab161
GC		 1971 if (parent_status == UNDER_LIMIT &&
1972 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1973 sk_leave_memory_pressure(sk);
3ab224be
HA		 1974 return 1;
1975 }
1976
e1aab161
GC		 1977 /* Under pressure. (we or our parents) */
1978 if ((parent_status > SOFT_LIMIT) ||
1979 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1980 sk_enter_memory_pressure(sk);
3ab224be 1981
e1aab161
GC		 1982 /* Over hard limit (we or our parents) */
1983 if ((parent_status == OVER_LIMIT) ||
1984 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1985 goto suppress_allocation;
1986
1987 /* guarantee minimum buffer size under pressure */
1988 if (kind == SK_MEM_RECV) {
1989 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1990 return 1;
180d8cd9 1991
3ab224be
HA		 1992 } else { /* SK_MEM_SEND */
1993 if (sk->sk_type == SOCK_STREAM) {
1994 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1995 return 1;
1996 } else if (atomic_read(&sk->sk_wmem_alloc) <
1997 prot->sysctl_wmem[0])
1998 return 1;
1999 }
2000
180d8cd9 2001 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2002 int alloc;
2003
180d8cd9 2004 if (!sk_under_memory_pressure(sk))
1748376b 2005 return 1;
180d8cd9
GC		 2006 alloc = sk_sockets_allocated_read_positive(sk);
2007 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2008 sk_mem_pages(sk->sk_wmem_queued +
2009 atomic_read(&sk->sk_rmem_alloc) +
2010 sk->sk_forward_alloc))
2011 return 1;
2012 }
2013
2014suppress_allocation:
2015
2016 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2017 sk_stream_moderate_sndbuf(sk);
2018
2019 /* Fail only if socket is _under_ its sndbuf.
2020 * In this case we cannot block, so that we have to fail.
2021 */
2022 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2023 return 1;
2024 }
2025
3847ce32
SM		 2026 trace_sock_exceed_buf_limit(sk, prot, allocated);
2027
3ab224be
HA		 2028 /* Alas. Undo changes. */
2029 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2030
0e90b31f 2031 sk_memory_allocated_sub(sk, amt);
180d8cd9 2032
3ab224be
HA		 2033 return 0;
2034}
3ab224be
HA		 2035EXPORT_SYMBOL(__sk_mem_schedule);
2036
2037/**
2038 * __sk_reclaim - reclaim memory_allocated
2039 * @sk: socket
2040 */
2041void __sk_mem_reclaim(struct sock *sk)
2042{
180d8cd9 2043 sk_memory_allocated_sub(sk,
0e90b31f 2044 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 2045 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
2046
180d8cd9
GC		 2047 if (sk_under_memory_pressure(sk) &&
2048 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2049 sk_leave_memory_pressure(sk);
3ab224be 2050}
3ab224be
HA		 2051EXPORT_SYMBOL(__sk_mem_reclaim);
2052
2053
1da177e4
LT		 2054/*
2055 * Set of default routines for initialising struct proto_ops when
2056 * the protocol does not support a particular function. In certain
2057 * cases where it makes no sense for a protocol to have a "do nothing"
2058 * function, some default processing is provided.
2059 */
2060
2061int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2062{
2063 return -EOPNOTSUPP;
2064}
2a91525c 2065EXPORT_SYMBOL(sock_no_bind);
1da177e4 2066
4ec93edb 2067int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2068 int len, int flags)
2069{
2070 return -EOPNOTSUPP;
2071}
2a91525c 2072EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2073
2074int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2075{
2076 return -EOPNOTSUPP;
2077}
2a91525c 2078EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2079
2080int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2081{
2082 return -EOPNOTSUPP;
2083}
2a91525c 2084EXPORT_SYMBOL(sock_no_accept);
1da177e4 2085
4ec93edb 2086int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2087 int *len, int peer)
2088{
2089 return -EOPNOTSUPP;
2090}
2a91525c 2091EXPORT_SYMBOL(sock_no_getname);
1da177e4 2092
2a91525c 2093unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2094{
2095 return 0;
2096}
2a91525c 2097EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2098
2099int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2100{
2101 return -EOPNOTSUPP;
2102}
2a91525c 2103EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2104
2105int sock_no_listen(struct socket *sock, int backlog)
2106{
2107 return -EOPNOTSUPP;
2108}
2a91525c 2109EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2110
2111int sock_no_shutdown(struct socket *sock, int how)
2112{
2113 return -EOPNOTSUPP;
2114}
2a91525c 2115EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2116
2117int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2118 char __user *optval, unsigned int optlen)
1da177e4
LT		 2119{
2120 return -EOPNOTSUPP;
2121}
2a91525c 2122EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2123
2124int sock_no_getsockopt(struct socket *sock, int level, int optname,
2125 char __user *optval, int __user *optlen)
2126{
2127 return -EOPNOTSUPP;
2128}
2a91525c 2129EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2130
2131int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2132 size_t len)
2133{
2134 return -EOPNOTSUPP;
2135}
2a91525c 2136EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2137
2138int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2139 size_t len, int flags)
2140{
2141 return -EOPNOTSUPP;
2142}
2a91525c 2143EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2144
2145int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2146{
2147 /* Mirror missing mmap method error code */
2148 return -ENODEV;
2149}
2a91525c 2150EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2151
2152ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2153{
2154 ssize_t res;
2155 struct msghdr msg = {.msg_flags = flags};
2156 struct kvec iov;
2157 char *kaddr = kmap(page);
2158 iov.iov_base = kaddr + offset;
2159 iov.iov_len = size;
2160 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2161 kunmap(page);
2162 return res;
2163}
2a91525c 2164EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2165
2166/*
2167 * Default Socket Callbacks
2168 */
2169
2170static void sock_def_wakeup(struct sock *sk)
2171{
43815482
ED		 2172 struct socket_wq *wq;
2173
2174 rcu_read_lock();
2175 wq = rcu_dereference(sk->sk_wq);
2176 if (wq_has_sleeper(wq))
2177 wake_up_interruptible_all(&wq->wait);
2178 rcu_read_unlock();
1da177e4
LT		 2179}
2180
2181static void sock_def_error_report(struct sock *sk)
2182{
43815482
ED		 2183 struct socket_wq *wq;
2184
2185 rcu_read_lock();
2186 wq = rcu_dereference(sk->sk_wq);
2187 if (wq_has_sleeper(wq))
2188 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2189 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2190 rcu_read_unlock();
1da177e4
LT		 2191}
2192
2193static void sock_def_readable(struct sock *sk, int len)
2194{
43815482
ED		 2195 struct socket_wq *wq;
2196
2197 rcu_read_lock();
2198 wq = rcu_dereference(sk->sk_wq);
2199 if (wq_has_sleeper(wq))
2c6607c6 2200 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2201 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2202 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2203 rcu_read_unlock();
1da177e4
LT		 2204}
2205
2206static void sock_def_write_space(struct sock *sk)
2207{
43815482
ED		 2208 struct socket_wq *wq;
2209
2210 rcu_read_lock();
1da177e4
LT		 2211
2212 /* Do not wake up a writer until he can make "significant"
2213 * progress. --DaveM
2214 */
e71a4783 2215 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2216 wq = rcu_dereference(sk->sk_wq);
2217 if (wq_has_sleeper(wq))
2218 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2219 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2220
2221 /* Should agree with poll, otherwise some programs break */
2222 if (sock_writeable(sk))
8d8ad9d7 2223 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2224 }
2225
43815482 2226 rcu_read_unlock();
1da177e4
LT		 2227}
2228
2229static void sock_def_destruct(struct sock *sk)
2230{
a51482bd 2231 kfree(sk->sk_protinfo);
1da177e4
LT		 2232}
2233
2234void sk_send_sigurg(struct sock *sk)
2235{
2236 if (sk->sk_socket && sk->sk_socket->file)
2237 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2238 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2239}
2a91525c 2240EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2241
2242void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2243 unsigned long expires)
2244{
2245 if (!mod_timer(timer, expires))
2246 sock_hold(sk);
2247}
1da177e4
LT		 2248EXPORT_SYMBOL(sk_reset_timer);
2249
2250void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2251{
25cc4ae9 2252 if (del_timer(timer))
1da177e4
LT		 2253 __sock_put(sk);
2254}
1da177e4
LT		 2255EXPORT_SYMBOL(sk_stop_timer);
2256
2257void sock_init_data(struct socket *sock, struct sock *sk)
2258{
2259 skb_queue_head_init(&sk->sk_receive_queue);
2260 skb_queue_head_init(&sk->sk_write_queue);
2261 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2262#ifdef CONFIG_NET_DMA
2263 skb_queue_head_init(&sk->sk_async_wait_queue);
2264#endif
1da177e4
LT		 2265
2266 sk->sk_send_head = NULL;
2267
2268 init_timer(&sk->sk_timer);
4ec93edb 2269
1da177e4
LT		 2270 sk->sk_allocation = GFP_KERNEL;
2271 sk->sk_rcvbuf = sysctl_rmem_default;
2272 sk->sk_sndbuf = sysctl_wmem_default;
2273 sk->sk_state = TCP_CLOSE;
972692e0 2274 sk_set_socket(sk, sock);
1da177e4
LT		 2275
2276 sock_set_flag(sk, SOCK_ZAPPED);
2277
e71a4783 2278 if (sock) {
1da177e4 2279 sk->sk_type = sock->type;
43815482 2280 sk->sk_wq = sock->wq;
1da177e4
LT		 2281 sock->sk = sk;
2282 } else
43815482 2283 sk->sk_wq = NULL;
1da177e4 2284
b6c6712a 2285 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2286 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2287 lockdep_set_class_and_name(&sk->sk_callback_lock,
2288 af_callback_keys + sk->sk_family,
2289 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2290
2291 sk->sk_state_change = sock_def_wakeup;
2292 sk->sk_data_ready = sock_def_readable;
2293 sk->sk_write_space = sock_def_write_space;
2294 sk->sk_error_report = sock_def_error_report;
2295 sk->sk_destruct = sock_def_destruct;
2296
5640f768
ED		 2297 sk->sk_frag.page = NULL;
2298 sk->sk_frag.offset = 0;
ef64a54f 2299 sk->sk_peek_off = -1;
1da177e4 2300
109f6e39
EB		 2301 sk->sk_peer_pid = NULL;
2302 sk->sk_peer_cred = NULL;
1da177e4
LT		 2303 sk->sk_write_pending = 0;
2304 sk->sk_rcvlowat = 1;
2305 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2306 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2307
f37f0afb 2308 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2309
e0d1095a 2310#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2311 sk->sk_napi_id = 0;
64b0dc51 2312 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2313#endif
2314
62748f32 2315 sk->sk_max_pacing_rate = ~0U;
7eec4174 2316 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2317 /*
2318 * Before updating sk_refcnt, we must commit prior changes to memory
2319 * (Documentation/RCU/rculist_nulls.txt for details)
2320 */
2321 smp_wmb();
1da177e4 2322 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2323 atomic_set(&sk->sk_drops, 0);
1da177e4 2324}
2a91525c 2325EXPORT_SYMBOL(sock_init_data);
1da177e4 2326
b5606c2d 2327void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2328{
2329 might_sleep();
a5b5bb9a 2330 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2331 if (sk->sk_lock.owned)
1da177e4 2332 __lock_sock(sk);
d2e9117c 2333 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2334 spin_unlock(&sk->sk_lock.slock);
2335 /*
2336 * The sk_lock has mutex_lock() semantics here:
2337 */
fcc70d5f 2338 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2339 local_bh_enable();
1da177e4 2340}
fcc70d5f 2341EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2342
b5606c2d 2343void release_sock(struct sock *sk)
1da177e4 2344{
a5b5bb9a
IM		 2345 /*
2346 * The sk_lock has mutex_unlock() semantics:
2347 */
2348 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2349
2350 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2351 if (sk->sk_backlog.tail)
2352 __release_sock(sk);
46d3ceab
ED		 2353
2354 if (sk->sk_prot->release_cb)
2355 sk->sk_prot->release_cb(sk);
2356
d2e9117c 2357 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2358 if (waitqueue_active(&sk->sk_lock.wq))
2359 wake_up(&sk->sk_lock.wq);
2360 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2361}
2362EXPORT_SYMBOL(release_sock);
2363
8a74ad60
ED		 2364/**
2365 * lock_sock_fast - fast version of lock_sock
2366 * @sk: socket
2367 *
2368 * This version should be used for very small section, where process wont block
2369 * return false if fast path is taken
2370 * sk_lock.slock locked, owned = 0, BH disabled
2371 * return true if slow path is taken
2372 * sk_lock.slock unlocked, owned = 1, BH enabled
2373 */
2374bool lock_sock_fast(struct sock *sk)
2375{
2376 might_sleep();
2377 spin_lock_bh(&sk->sk_lock.slock);
2378
2379 if (!sk->sk_lock.owned)
2380 /*
2381 * Note : We must disable BH
2382 */
2383 return false;
2384
2385 __lock_sock(sk);
2386 sk->sk_lock.owned = 1;
2387 spin_unlock(&sk->sk_lock.slock);
2388 /*
2389 * The sk_lock has mutex_lock() semantics here:
2390 */
2391 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2392 local_bh_enable();
2393 return true;
2394}
2395EXPORT_SYMBOL(lock_sock_fast);
2396
1da177e4 2397int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2398{
b7aa0bf7 2399 struct timeval tv;
1da177e4 2400 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2401 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2402 tv = ktime_to_timeval(sk->sk_stamp);
2403 if (tv.tv_sec == -1)
1da177e4 2404 return -ENOENT;
b7aa0bf7
ED		 2405 if (tv.tv_sec == 0) {
2406 sk->sk_stamp = ktime_get_real();
2407 tv = ktime_to_timeval(sk->sk_stamp);
2408 }
2409 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2410}
1da177e4
LT		 2411EXPORT_SYMBOL(sock_get_timestamp);
2412
ae40eb1e
ED		 2413int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2414{
2415 struct timespec ts;
2416 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2417 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2418 ts = ktime_to_timespec(sk->sk_stamp);
2419 if (ts.tv_sec == -1)
2420 return -ENOENT;
2421 if (ts.tv_sec == 0) {
2422 sk->sk_stamp = ktime_get_real();
2423 ts = ktime_to_timespec(sk->sk_stamp);
2424 }
2425 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2426}
2427EXPORT_SYMBOL(sock_get_timestampns);
2428
20d49473 2429void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2430{
20d49473 2431 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2432 unsigned long previous_flags = sk->sk_flags;
2433
20d49473
PO		 2434 sock_set_flag(sk, flag);
2435 /*
2436 * we just set one of the two flags which require net
2437 * time stamping, but time stamping might have been on
2438 * already because of the other one
2439 */
08e29af3 2440 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2441 net_enable_timestamp();
1da177e4
LT		 2442 }
2443}
1da177e4 2444
cb820f8e
RC		 2445int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2446 int level, int type)
2447{
2448 struct sock_exterr_skb *serr;
2449 struct sk_buff *skb, *skb2;
2450 int copied, err;
2451
2452 err = -EAGAIN;
2453 skb = skb_dequeue(&sk->sk_error_queue);
2454 if (skb == NULL)
2455 goto out;
2456
2457 copied = skb->len;
2458 if (copied > len) {
2459 msg->msg_flags |= MSG_TRUNC;
2460 copied = len;
2461 }
2462 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2463 if (err)
2464 goto out_free_skb;
2465
2466 sock_recv_timestamp(msg, sk, skb);
2467
2468 serr = SKB_EXT_ERR(skb);
2469 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2470
2471 msg->msg_flags |= MSG_ERRQUEUE;
2472 err = copied;
2473
2474 /* Reset and regenerate socket error */
2475 spin_lock_bh(&sk->sk_error_queue.lock);
2476 sk->sk_err = 0;
2477 if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) {
2478 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
2479 spin_unlock_bh(&sk->sk_error_queue.lock);
2480 sk->sk_error_report(sk);
2481 } else
2482 spin_unlock_bh(&sk->sk_error_queue.lock);
2483
2484out_free_skb:
2485 kfree_skb(skb);
2486out:
2487 return err;
2488}
2489EXPORT_SYMBOL(sock_recv_errqueue);
2490
1da177e4
LT		 2491/*
2492 * Get a socket option on an socket.
2493 *
2494 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2495 * asynchronous errors should be reported by getsockopt. We assume
2496 * this means if you specify SO_ERROR (otherwise whats the point of it).
2497 */
2498int sock_common_getsockopt(struct socket *sock, int level, int optname,
2499 char __user *optval, int __user *optlen)
2500{
2501 struct sock *sk = sock->sk;
2502
2503 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2504}
1da177e4
LT		 2505EXPORT_SYMBOL(sock_common_getsockopt);
2506
3fdadf7d 2507#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2508int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2509 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2510{
2511 struct sock *sk = sock->sk;
2512
1e51f951 2513 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2514 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2515 optval, optlen);
3fdadf7d
DM		 2516 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2517}
2518EXPORT_SYMBOL(compat_sock_common_getsockopt);
2519#endif
2520
1da177e4
LT		 2521int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2522 struct msghdr *msg, size_t size, int flags)
2523{
2524 struct sock *sk = sock->sk;
2525 int addr_len = 0;
2526 int err;
2527
2528 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2529 flags & ~MSG_DONTWAIT, &addr_len);
2530 if (err >= 0)
2531 msg->msg_namelen = addr_len;
2532 return err;
2533}
1da177e4
LT		 2534EXPORT_SYMBOL(sock_common_recvmsg);
2535
2536/*
2537 * Set socket options on an inet socket.
2538 */
2539int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2540 char __user *optval, unsigned int optlen)
1da177e4
LT		 2541{
2542 struct sock *sk = sock->sk;
2543
2544 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2545}
1da177e4
LT		 2546EXPORT_SYMBOL(sock_common_setsockopt);
2547
3fdadf7d 2548#ifdef CONFIG_COMPAT
543d9cfe 2549int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2550 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2551{
2552 struct sock *sk = sock->sk;
2553
543d9cfe
ACM		 2554 if (sk->sk_prot->compat_setsockopt != NULL)
2555 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2556 optval, optlen);
3fdadf7d
DM		 2557 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2558}
2559EXPORT_SYMBOL(compat_sock_common_setsockopt);
2560#endif
2561
1da177e4
LT		 2562void sk_common_release(struct sock *sk)
2563{
2564 if (sk->sk_prot->destroy)
2565 sk->sk_prot->destroy(sk);
2566
2567 /*
2568 * Observation: when sock_common_release is called, processes have
2569 * no access to socket. But net still has.
2570 * Step one, detach it from networking:
2571 *
2572 * A. Remove from hash tables.
2573 */
2574
2575 sk->sk_prot->unhash(sk);
2576
2577 /*
2578 * In this point socket cannot receive new packets, but it is possible
2579 * that some packets are in flight because some CPU runs receiver and
2580 * did hash table lookup before we unhashed socket. They will achieve
2581 * receive queue and will be purged by socket destructor.
2582 *
2583 * Also we still have packets pending on receive queue and probably,
2584 * our own packets waiting in device queues. sock_destroy will drain
2585 * receive queue, but transmitted packets will delay socket destruction
2586 * until the last reference will be released.
2587 */
2588
2589 sock_orphan(sk);
2590
2591 xfrm_sk_free_policy(sk);
2592
e6848976 2593 sk_refcnt_debug_release(sk);
5640f768
ED		 2594
2595 if (sk->sk_frag.page) {
2596 put_page(sk->sk_frag.page);
2597 sk->sk_frag.page = NULL;
2598 }
2599
1da177e4
LT		 2600 sock_put(sk);
2601}
1da177e4
LT		 2602EXPORT_SYMBOL(sk_common_release);
2603
13ff3d6f
PE		 2604#ifdef CONFIG_PROC_FS
2605#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2606struct prot_inuse {
2607 int val[PROTO_INUSE_NR];
2608};
13ff3d6f
PE		 2609
2610static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2611
2612#ifdef CONFIG_NET_NS
2613void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2614{
d6d9ca0f 2615 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2616}
2617EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2618
2619int sock_prot_inuse_get(struct net *net, struct proto *prot)
2620{
2621 int cpu, idx = prot->inuse_idx;
2622 int res = 0;
2623
2624 for_each_possible_cpu(cpu)
2625 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2626
2627 return res >= 0 ? res : 0;
2628}
2629EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2630
2c8c1e72 2631static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2632{
2633 net->core.inuse = alloc_percpu(struct prot_inuse);
2634 return net->core.inuse ? 0 : -ENOMEM;
2635}
2636
2c8c1e72 2637static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2638{
2639 free_percpu(net->core.inuse);
2640}
2641
2642static struct pernet_operations net_inuse_ops = {
2643 .init = sock_inuse_init_net,
2644 .exit = sock_inuse_exit_net,
2645};
2646
2647static __init int net_inuse_init(void)
2648{
2649 if (register_pernet_subsys(&net_inuse_ops))
2650 panic("Cannot initialize net inuse counters");
2651
2652 return 0;
2653}
2654
2655core_initcall(net_inuse_init);
2656#else
1338d466
PE		 2657static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2658
c29a0bc4 2659void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2660{
d6d9ca0f 2661 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2662}
2663EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2664
c29a0bc4 2665int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2666{
2667 int cpu, idx = prot->inuse_idx;
2668 int res = 0;
2669
2670 for_each_possible_cpu(cpu)
2671 res += per_cpu(prot_inuse, cpu).val[idx];
2672
2673 return res >= 0 ? res : 0;
2674}
2675EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2676#endif
13ff3d6f
PE		 2677
2678static void assign_proto_idx(struct proto *prot)
2679{
2680 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2681
2682 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2683 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2684 return;
2685 }
2686
2687 set_bit(prot->inuse_idx, proto_inuse_idx);
2688}
2689
2690static void release_proto_idx(struct proto *prot)
2691{
2692 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2693 clear_bit(prot->inuse_idx, proto_inuse_idx);
2694}
2695#else
2696static inline void assign_proto_idx(struct proto *prot)
2697{
2698}
2699
2700static inline void release_proto_idx(struct proto *prot)
2701{
2702}
2703#endif
2704
b733c007
PE		 2705int proto_register(struct proto *prot, int alloc_slab)
2706{
1da177e4
LT		 2707 if (alloc_slab) {
2708 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2709 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2710 NULL);
1da177e4
LT		 2711
2712 if (prot->slab == NULL) {
e005d193
JP		 2713 pr_crit("%s: Can't create sock SLAB cache!\n",
2714 prot->name);
60e7663d 2715 goto out;
1da177e4 2716 }
2e6599cb
ACM		 2717
2718 if (prot->rsk_prot != NULL) {
faf23422 2719 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2720 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2721 goto out_free_sock_slab;
2722
7e56b5d6 2723 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2724 prot->rsk_prot->obj_size, 0,
20c2df83 2725 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2726
2727 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2728 pr_crit("%s: Can't create request sock SLAB cache!\n",
2729 prot->name);
2e6599cb
ACM		 2730 goto out_free_request_sock_slab_name;
2731 }
2732 }
8feaf0c0 2733
6d6ee43e 2734 if (prot->twsk_prot != NULL) {
faf23422 2735 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2736
7e56b5d6 2737 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2738 goto out_free_request_sock_slab;
2739
6d6ee43e 2740 prot->twsk_prot->twsk_slab =
7e56b5d6 2741 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2742 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2743 0,
2744 SLAB_HWCACHE_ALIGN |
2745 prot->slab_flags,
20c2df83 2746 NULL);
6d6ee43e 2747 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2748 goto out_free_timewait_sock_slab_name;
2749 }
1da177e4
LT		 2750 }
2751
36b77a52 2752 mutex_lock(&proto_list_mutex);
1da177e4 2753 list_add(&prot->node, &proto_list);
13ff3d6f 2754 assign_proto_idx(prot);
36b77a52 2755 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2756 return 0;
2757
8feaf0c0 2758out_free_timewait_sock_slab_name:
7e56b5d6 2759 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2760out_free_request_sock_slab:
2761 if (prot->rsk_prot && prot->rsk_prot->slab) {
2762 kmem_cache_destroy(prot->rsk_prot->slab);
2763 prot->rsk_prot->slab = NULL;
2764 }
2e6599cb 2765out_free_request_sock_slab_name:
72150e9b
DC		 2766 if (prot->rsk_prot)
2767 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2768out_free_sock_slab:
2769 kmem_cache_destroy(prot->slab);
2770 prot->slab = NULL;
b733c007
PE		 2771out:
2772 return -ENOBUFS;
1da177e4 2773}
1da177e4
LT		 2774EXPORT_SYMBOL(proto_register);
2775
2776void proto_unregister(struct proto *prot)
2777{
36b77a52 2778 mutex_lock(&proto_list_mutex);
13ff3d6f 2779 release_proto_idx(prot);
0a3f4358 2780 list_del(&prot->node);
36b77a52 2781 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2782
2783 if (prot->slab != NULL) {
2784 kmem_cache_destroy(prot->slab);
2785 prot->slab = NULL;
2786 }
2787
2e6599cb 2788 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2789 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2790 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2791 prot->rsk_prot->slab = NULL;
2792 }
2793
6d6ee43e 2794 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2795 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2796 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2797 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2798 }
1da177e4 2799}
1da177e4
LT		 2800EXPORT_SYMBOL(proto_unregister);
2801
2802#ifdef CONFIG_PROC_FS
1da177e4 2803static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2804 __acquires(proto_list_mutex)
1da177e4 2805{
36b77a52 2806 mutex_lock(&proto_list_mutex);
60f0438a 2807 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2808}
2809
2810static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2811{
60f0438a 2812 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2813}
2814
2815static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2816 __releases(proto_list_mutex)
1da177e4 2817{
36b77a52 2818 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2819}
2820
2821static char proto_method_implemented(const void *method)
2822{
2823 return method == NULL ? 'n' : 'y';
2824}
180d8cd9
GC		 2825static long sock_prot_memory_allocated(struct proto *proto)
2826{
cb75a36c 2827 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2828}
2829
2830static char *sock_prot_memory_pressure(struct proto *proto)
2831{
2832 return proto->memory_pressure != NULL ?
2833 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2834}
1da177e4
LT		 2835
2836static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2837{
180d8cd9 2838
8d987e5c 2839 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2840 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2841 proto->name,
2842 proto->obj_size,
14e943db 2843 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2844 sock_prot_memory_allocated(proto),
2845 sock_prot_memory_pressure(proto),
1da177e4
LT		 2846 proto->max_header,
2847 proto->slab == NULL ? "no" : "yes",
2848 module_name(proto->owner),
2849 proto_method_implemented(proto->close),
2850 proto_method_implemented(proto->connect),
2851 proto_method_implemented(proto->disconnect),
2852 proto_method_implemented(proto->accept),
2853 proto_method_implemented(proto->ioctl),
2854 proto_method_implemented(proto->init),
2855 proto_method_implemented(proto->destroy),
2856 proto_method_implemented(proto->shutdown),
2857 proto_method_implemented(proto->setsockopt),
2858 proto_method_implemented(proto->getsockopt),
2859 proto_method_implemented(proto->sendmsg),
2860 proto_method_implemented(proto->recvmsg),
2861 proto_method_implemented(proto->sendpage),
2862 proto_method_implemented(proto->bind),
2863 proto_method_implemented(proto->backlog_rcv),
2864 proto_method_implemented(proto->hash),
2865 proto_method_implemented(proto->unhash),
2866 proto_method_implemented(proto->get_port),
2867 proto_method_implemented(proto->enter_memory_pressure));
2868}
2869
2870static int proto_seq_show(struct seq_file *seq, void *v)
2871{
60f0438a 2872 if (v == &proto_list)
1da177e4
LT		 2873 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2874 "protocol",
2875 "size",
2876 "sockets",
2877 "memory",
2878 "press",
2879 "maxhdr",
2880 "slab",
2881 "module",
2882 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2883 else
60f0438a 2884 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2885 return 0;
2886}
2887
f690808e 2888static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2889 .start = proto_seq_start,
2890 .next = proto_seq_next,
2891 .stop = proto_seq_stop,
2892 .show = proto_seq_show,
2893};
2894
2895static int proto_seq_open(struct inode *inode, struct file *file)
2896{
14e943db
ED		 2897 return seq_open_net(inode, file, &proto_seq_ops,
2898 sizeof(struct seq_net_private));
1da177e4
LT		 2899}
2900
9a32144e 2901static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2902 .owner = THIS_MODULE,
2903 .open = proto_seq_open,
2904 .read = seq_read,
2905 .llseek = seq_lseek,
14e943db
ED		 2906 .release = seq_release_net,
2907};
2908
2909static __net_init int proto_init_net(struct net *net)
2910{
d4beaa66 2911 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2912 return -ENOMEM;
2913
2914 return 0;
2915}
2916
2917static __net_exit void proto_exit_net(struct net *net)
2918{
ece31ffd 2919 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2920}
2921
2922
2923static __net_initdata struct pernet_operations proto_net_ops = {
2924 .init = proto_init_net,
2925 .exit = proto_exit_net,
1da177e4
LT		 2926};
2927
2928static int __init proto_init(void)
2929{
14e943db 2930 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2931}
2932
2933subsys_initcall(proto_init);
2934
2935#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 0.955611 seconds and 5 git commands to generate.