projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: Stop decapitating clones that have a head_frag
[GitHub/mt8127/android_kernel_alcatel_ttab.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
4fc268d2 92#include <linux/capability.h>
1da177e4
LT		 93#include <linux/errno.h>
94#include <linux/types.h>
95#include <linux/socket.h>
96#include <linux/in.h>
97#include <linux/kernel.h>
1da177e4
LT		 98#include <linux/module.h>
99#include <linux/proc_fs.h>
100#include <linux/seq_file.h>
101#include <linux/sched.h>
102#include <linux/timer.h>
103#include <linux/string.h>
104#include <linux/sockios.h>
105#include <linux/net.h>
106#include <linux/mm.h>
107#include <linux/slab.h>
108#include <linux/interrupt.h>
109#include <linux/poll.h>
110#include <linux/tcp.h>
111#include <linux/init.h>
a1f8e7f7 112#include <linux/highmem.h>
3f551f94 113#include <linux/user_namespace.h>
c5905afb 114#include <linux/static_key.h>
3969eb38 115#include <linux/memcontrol.h>
1da177e4
LT		 116
117#include <asm/uaccess.h>
1da177e4
LT		 118
119#include <linux/netdevice.h>
120#include <net/protocol.h>
121#include <linux/skbuff.h>
457c4cbc 122#include <net/net_namespace.h>
2e6599cb 123#include <net/request_sock.h>
1da177e4 124#include <net/sock.h>
20d49473 125#include <linux/net_tstamp.h>
1da177e4
LT		 126#include <net/xfrm.h>
127#include <linux/ipsec.h>
f8451725 128#include <net/cls_cgroup.h>
5bc1421e 129#include <net/netprio_cgroup.h>
1da177e4
LT		 130
131#include <linux/filter.h>
132
3847ce32
SM		 133#include <trace/events/sock.h>
134
1da177e4
LT		 135#ifdef CONFIG_INET
136#include <net/tcp.h>
137#endif
138
36b77a52 139static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 140static LIST_HEAD(proto_list);
141
142#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
143int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
144{
145 struct proto *proto;
146 int ret = 0;
147
36b77a52 148 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 149 list_for_each_entry(proto, &proto_list, node) {
150 if (proto->init_cgroup) {
151 ret = proto->init_cgroup(cgrp, ss);
152 if (ret)
153 goto out;
154 }
155 }
156
36b77a52 157 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 158 return ret;
159out:
160 list_for_each_entry_continue_reverse(proto, &proto_list, node)
161 if (proto->destroy_cgroup)
761b3ef5 162 proto->destroy_cgroup(cgrp);
36b77a52 163 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 164 return ret;
165}
166
761b3ef5 167void mem_cgroup_sockets_destroy(struct cgroup *cgrp)
d1a4c0b3
GC		 168{
169 struct proto *proto;
170
36b77a52 171 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 172 list_for_each_entry_reverse(proto, &proto_list, node)
173 if (proto->destroy_cgroup)
761b3ef5 174 proto->destroy_cgroup(cgrp);
36b77a52 175 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 176}
177#endif
178
da21f24d
IM		 179/*
180 * Each address family might have different locking rules, so we have
181 * one slock key per address family:
182 */
a5b5bb9a
IM		 183static struct lock_class_key af_family_keys[AF_MAX];
184static struct lock_class_key af_family_slock_keys[AF_MAX];
185
c5905afb 186struct static_key memcg_socket_limit_enabled;
e1aab161
GC		 187EXPORT_SYMBOL(memcg_socket_limit_enabled);
188
a5b5bb9a
IM		 189/*
190 * Make lock validator output more readable. (we pre-construct these
191 * strings build-time, so that runtime initialization of socket
192 * locks is fast):
193 */
36cbd3dc 194static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 195 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
196 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
197 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
198 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
199 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
200 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
201 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 202 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 203 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 204 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 205 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 206 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 207 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
c7fe3b52 208 "sk_lock-AF_NFC" , "sk_lock-AF_MAX"
a5b5bb9a 209};
36cbd3dc 210static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 211 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
212 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
213 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
214 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
215 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
216 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
217 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 218 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 219 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 220 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 221 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 222 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 223 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
c7fe3b52 224 "slock-AF_NFC" , "slock-AF_MAX"
a5b5bb9a 225};
36cbd3dc 226static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 227 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
228 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
229 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
230 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
231 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
232 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
233 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 234 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 235 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 236 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 237 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 238 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 239 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
c7fe3b52 240 "clock-AF_NFC" , "clock-AF_MAX"
443aef0e 241};
da21f24d
IM		 242
243/*
244 * sk_callback_lock locking rules are per-address-family,
245 * so split the lock classes by using a per-AF key:
246 */
247static struct lock_class_key af_callback_keys[AF_MAX];
248
1da177e4
LT		 249/* Take into consideration the size of the struct sk_buff overhead in the
250 * determination of these values, since that is non-constant across
251 * platforms. This makes socket queueing behavior and performance
252 * not depend upon such differences.
253 */
254#define _SK_MEM_PACKETS 256
87fb4b7b 255#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 256#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
257#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
258
259/* Run time adjustable parameters. */
ab32ea5d
BH		 260__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
261__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
262__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
263__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 264
25985edc 265/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 266int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 267EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 268
5bc1421e
NH		 269#if defined(CONFIG_CGROUPS)
270#if !defined(CONFIG_NET_CLS_CGROUP)
f8451725
HX		 271int net_cls_subsys_id = -1;
272EXPORT_SYMBOL_GPL(net_cls_subsys_id);
273#endif
5bc1421e
NH		 274#if !defined(CONFIG_NETPRIO_CGROUP)
275int net_prio_subsys_id = -1;
276EXPORT_SYMBOL_GPL(net_prio_subsys_id);
277#endif
278#endif
f8451725 279
1da177e4
LT		 280static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
281{
282 struct timeval tv;
283
284 if (optlen < sizeof(tv))
285 return -EINVAL;
286 if (copy_from_user(&tv, optval, sizeof(tv)))
287 return -EFAULT;
ba78073e
VA		 288 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
289 return -EDOM;
1da177e4 290
ba78073e 291 if (tv.tv_sec < 0) {
6f11df83
AM		 292 static int warned __read_mostly;
293
ba78073e 294 *timeo_p = 0;
50aab54f 295 if (warned < 10 && net_ratelimit()) {
ba78073e
VA		 296 warned++;
297 printk(KERN_INFO "sock_set_timeout: `%s' (pid %d) "
298 "tries to set negative timeout\n",
ba25f9dc 299 current->comm, task_pid_nr(current));
50aab54f 300 }
ba78073e
VA		 301 return 0;
302 }
1da177e4
LT		 303 *timeo_p = MAX_SCHEDULE_TIMEOUT;
304 if (tv.tv_sec == 0 && tv.tv_usec == 0)
305 return 0;
306 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
307 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
308 return 0;
309}
310
311static void sock_warn_obsolete_bsdism(const char *name)
312{
313 static int warned;
314 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 315 if (strcmp(warncomm, current->comm) && warned < 5) {
316 strcpy(warncomm, current->comm);
1da177e4
LT		 317 printk(KERN_WARNING "process `%s' is using obsolete "
318 "%s SO_BSDCOMPAT\n", warncomm, name);
319 warned++;
320 }
321}
322
08e29af3
ED		 323#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
324
325static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 326{
08e29af3
ED		 327 if (sk->sk_flags & flags) {
328 sk->sk_flags &= ~flags;
329 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 330 net_disable_timestamp();
1da177e4
LT		 331 }
332}
333
334
f0088a50
DV		 335int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
336{
766e9037 337 int err;
f0088a50 338 int skb_len;
3b885787
NH		 339 unsigned long flags;
340 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 341
0fd7bac6 342 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 343 atomic_inc(&sk->sk_drops);
3847ce32 344 trace_sock_rcvqueue_full(sk, skb);
766e9037 345 return -ENOMEM;
f0088a50
DV		 346 }
347
fda9ef5d 348 err = sk_filter(sk, skb);
f0088a50 349 if (err)
766e9037 350 return err;
f0088a50 351
3ab224be 352 if (!sk_rmem_schedule(sk, skb->truesize)) {
766e9037
ED		 353 atomic_inc(&sk->sk_drops);
354 return -ENOBUFS;
3ab224be
HA		 355 }
356
f0088a50
DV		 357 skb->dev = NULL;
358 skb_set_owner_r(skb, sk);
49ad9599 359
f0088a50
DV		 360 /* Cache the SKB length before we tack it onto the receive
361 * queue. Once it is added it no longer belongs to us and
362 * may be freed by other threads of control pulling packets
363 * from the queue.
364 */
365 skb_len = skb->len;
366
7fee226a
ED		 367 /* we escape from rcu protected region, make sure we dont leak
368 * a norefcounted dst
369 */
370 skb_dst_force(skb);
371
3b885787
NH		 372 spin_lock_irqsave(&list->lock, flags);
373 skb->dropcount = atomic_read(&sk->sk_drops);
374 __skb_queue_tail(list, skb);
375 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 376
377 if (!sock_flag(sk, SOCK_DEAD))
378 sk->sk_data_ready(sk, skb_len);
766e9037 379 return 0;
f0088a50
DV		 380}
381EXPORT_SYMBOL(sock_queue_rcv_skb);
382
58a5a7b9 383int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 384{
385 int rc = NET_RX_SUCCESS;
386
fda9ef5d 387 if (sk_filter(sk, skb))
f0088a50
DV		 388 goto discard_and_relse;
389
390 skb->dev = NULL;
391
f545a38f 392 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) {
c377411f
ED		 393 atomic_inc(&sk->sk_drops);
394 goto discard_and_relse;
395 }
58a5a7b9
ACM		 396 if (nested)
397 bh_lock_sock_nested(sk);
398 else
399 bh_lock_sock(sk);
a5b5bb9a
IM		 400 if (!sock_owned_by_user(sk)) {
401 /*
402 * trylock + unlock semantics:
403 */
404 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
405
c57943a1 406 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 407
408 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 409 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 410 bh_unlock_sock(sk);
411 atomic_inc(&sk->sk_drops);
412 goto discard_and_relse;
413 }
414
f0088a50
DV		 415 bh_unlock_sock(sk);
416out:
417 sock_put(sk);
418 return rc;
419discard_and_relse:
420 kfree_skb(skb);
421 goto out;
422}
423EXPORT_SYMBOL(sk_receive_skb);
424
ea94ff3b
KK		 425void sk_reset_txq(struct sock *sk)
426{
427 sk_tx_queue_clear(sk);
428}
429EXPORT_SYMBOL(sk_reset_txq);
430
f0088a50
DV		 431struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
432{
b6c6712a 433 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 434
435 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 436 sk_tx_queue_clear(sk);
a9b3cd7f 437 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 438 dst_release(dst);
439 return NULL;
440 }
441
442 return dst;
443}
444EXPORT_SYMBOL(__sk_dst_check);
445
446struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
447{
448 struct dst_entry *dst = sk_dst_get(sk);
449
450 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
451 sk_dst_reset(sk);
452 dst_release(dst);
453 return NULL;
454 }
455
456 return dst;
457}
458EXPORT_SYMBOL(sk_dst_check);
459
4878809f
DM		 460static int sock_bindtodevice(struct sock *sk, char __user *optval, int optlen)
461{
462 int ret = -ENOPROTOOPT;
463#ifdef CONFIG_NETDEVICES
3b1e0a65 464 struct net *net = sock_net(sk);
4878809f
DM		 465 char devname[IFNAMSIZ];
466 int index;
467
468 /* Sorry... */
469 ret = -EPERM;
470 if (!capable(CAP_NET_RAW))
471 goto out;
472
473 ret = -EINVAL;
474 if (optlen < 0)
475 goto out;
476
477 /* Bind this socket to a particular device like "eth0",
478 * as specified in the passed interface name. If the
479 * name is "" or the option length is zero the socket
480 * is not bound.
481 */
482 if (optlen > IFNAMSIZ - 1)
483 optlen = IFNAMSIZ - 1;
484 memset(devname, 0, sizeof(devname));
485
486 ret = -EFAULT;
487 if (copy_from_user(devname, optval, optlen))
488 goto out;
489
000ba2e4
DM		 490 index = 0;
491 if (devname[0] != '\0') {
bf8e56bf 492 struct net_device *dev;
4878809f 493
bf8e56bf
ED		 494 rcu_read_lock();
495 dev = dev_get_by_name_rcu(net, devname);
496 if (dev)
497 index = dev->ifindex;
498 rcu_read_unlock();
4878809f
DM		 499 ret = -ENODEV;
500 if (!dev)
501 goto out;
4878809f
DM		 502 }
503
504 lock_sock(sk);
505 sk->sk_bound_dev_if = index;
506 sk_dst_reset(sk);
507 release_sock(sk);
508
509 ret = 0;
510
511out:
512#endif
513
514 return ret;
515}
516
c0ef877b
PE		 517static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
518{
519 if (valbool)
520 sock_set_flag(sk, bit);
521 else
522 sock_reset_flag(sk, bit);
523}
524
1da177e4
LT		 525/*
526 * This is meant for all protocols to use and covers goings on
527 * at the socket level. Everything here is generic.
528 */
529
530int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 531 char __user *optval, unsigned int optlen)
1da177e4 532{
2a91525c 533 struct sock *sk = sock->sk;
1da177e4
LT		 534 int val;
535 int valbool;
536 struct linger ling;
537 int ret = 0;
4ec93edb 538
1da177e4
LT		 539 /*
540 * Options without arguments
541 */
542
4878809f
DM		 543 if (optname == SO_BINDTODEVICE)
544 return sock_bindtodevice(sk, optval, optlen);
545
e71a4783
SH		 546 if (optlen < sizeof(int))
547 return -EINVAL;
4ec93edb 548
1da177e4
LT		 549 if (get_user(val, (int __user *)optval))
550 return -EFAULT;
4ec93edb 551
2a91525c 552 valbool = val ? 1 : 0;
1da177e4
LT		 553
554 lock_sock(sk);
555
2a91525c 556 switch (optname) {
e71a4783 557 case SO_DEBUG:
2a91525c 558 if (val && !capable(CAP_NET_ADMIN))
e71a4783 559 ret = -EACCES;
2a91525c 560 else
c0ef877b 561 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 562 break;
563 case SO_REUSEADDR:
4a17fd52 564 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783
SH		 565 break;
566 case SO_TYPE:
49c794e9 567 case SO_PROTOCOL:
0d6038ee 568 case SO_DOMAIN:
e71a4783
SH		 569 case SO_ERROR:
570 ret = -ENOPROTOOPT;
571 break;
572 case SO_DONTROUTE:
c0ef877b 573 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 574 break;
575 case SO_BROADCAST:
576 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
577 break;
578 case SO_SNDBUF:
579 /* Don't error on this BSD doesn't and if you think
82981930
ED		 580 * about it this is right. Otherwise apps have to
581 * play 'guess the biggest size' games. RCVBUF/SNDBUF
582 * are treated in BSD as hints
583 */
584 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 585set_sndbuf:
e71a4783 586 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 587 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
588 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 589 sk->sk_write_space(sk);
590 break;
1da177e4 591
e71a4783
SH		 592 case SO_SNDBUFFORCE:
593 if (!capable(CAP_NET_ADMIN)) {
594 ret = -EPERM;
595 break;
596 }
597 goto set_sndbuf;
b0573dea 598
e71a4783
SH		 599 case SO_RCVBUF:
600 /* Don't error on this BSD doesn't and if you think
82981930
ED		 601 * about it this is right. Otherwise apps have to
602 * play 'guess the biggest size' games. RCVBUF/SNDBUF
603 * are treated in BSD as hints
604 */
605 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 606set_rcvbuf:
e71a4783
SH		 607 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
608 /*
609 * We double it on the way in to account for
610 * "struct sk_buff" etc. overhead. Applications
611 * assume that the SO_RCVBUF setting they make will
612 * allow that much actual data to be received on that
613 * socket.
614 *
615 * Applications are unaware that "struct sk_buff" and
616 * other overheads allocate from the receive buffer
617 * during socket buffer allocation.
618 *
619 * And after considering the possible alternatives,
620 * returning the value we actually used in getsockopt
621 * is the most desirable behavior.
622 */
82981930 623 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 624 break;
625
626 case SO_RCVBUFFORCE:
627 if (!capable(CAP_NET_ADMIN)) {
628 ret = -EPERM;
1da177e4 629 break;
e71a4783
SH		 630 }
631 goto set_rcvbuf;
1da177e4 632
e71a4783 633 case SO_KEEPALIVE:
1da177e4 634#ifdef CONFIG_INET
e71a4783
SH		 635 if (sk->sk_protocol == IPPROTO_TCP)
636 tcp_set_keepalive(sk, valbool);
1da177e4 637#endif
e71a4783
SH		 638 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
639 break;
640
641 case SO_OOBINLINE:
642 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
643 break;
644
645 case SO_NO_CHECK:
646 sk->sk_no_check = valbool;
647 break;
648
649 case SO_PRIORITY:
650 if ((val >= 0 && val <= 6) || capable(CAP_NET_ADMIN))
651 sk->sk_priority = val;
652 else
653 ret = -EPERM;
654 break;
655
656 case SO_LINGER:
657 if (optlen < sizeof(ling)) {
658 ret = -EINVAL; /* 1003.1g */
1da177e4 659 break;
e71a4783 660 }
2a91525c 661 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 662 ret = -EFAULT;
1da177e4 663 break;
e71a4783
SH		 664 }
665 if (!ling.l_onoff)
666 sock_reset_flag(sk, SOCK_LINGER);
667 else {
1da177e4 668#if (BITS_PER_LONG == 32)
e71a4783
SH		 669 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
670 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 671 else
e71a4783
SH		 672#endif
673 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
674 sock_set_flag(sk, SOCK_LINGER);
675 }
676 break;
677
678 case SO_BSDCOMPAT:
679 sock_warn_obsolete_bsdism("setsockopt");
680 break;
681
682 case SO_PASSCRED:
683 if (valbool)
684 set_bit(SOCK_PASSCRED, &sock->flags);
685 else
686 clear_bit(SOCK_PASSCRED, &sock->flags);
687 break;
688
689 case SO_TIMESTAMP:
92f37fd2 690 case SO_TIMESTAMPNS:
e71a4783 691 if (valbool) {
92f37fd2
ED		 692 if (optname == SO_TIMESTAMP)
693 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
694 else
695 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 696 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 697 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 698 } else {
e71a4783 699 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 700 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
701 }
e71a4783
SH		 702 break;
703
20d49473
PO		 704 case SO_TIMESTAMPING:
705 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 706 ret = -EINVAL;
20d49473
PO		 707 break;
708 }
709 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
710 val & SOF_TIMESTAMPING_TX_HARDWARE);
711 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
712 val & SOF_TIMESTAMPING_TX_SOFTWARE);
713 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
714 val & SOF_TIMESTAMPING_RX_HARDWARE);
715 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
716 sock_enable_timestamp(sk,
717 SOCK_TIMESTAMPING_RX_SOFTWARE);
718 else
719 sock_disable_timestamp(sk,
08e29af3 720 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 721 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
722 val & SOF_TIMESTAMPING_SOFTWARE);
723 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
724 val & SOF_TIMESTAMPING_SYS_HARDWARE);
725 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
726 val & SOF_TIMESTAMPING_RAW_HARDWARE);
727 break;
728
e71a4783
SH		 729 case SO_RCVLOWAT:
730 if (val < 0)
731 val = INT_MAX;
732 sk->sk_rcvlowat = val ? : 1;
733 break;
734
735 case SO_RCVTIMEO:
736 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
737 break;
738
739 case SO_SNDTIMEO:
740 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
741 break;
1da177e4 742
e71a4783
SH		 743 case SO_ATTACH_FILTER:
744 ret = -EINVAL;
745 if (optlen == sizeof(struct sock_fprog)) {
746 struct sock_fprog fprog;
1da177e4 747
e71a4783
SH		 748 ret = -EFAULT;
749 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 750 break;
e71a4783
SH		 751
752 ret = sk_attach_filter(&fprog, sk);
753 }
754 break;
755
756 case SO_DETACH_FILTER:
55b33325 757 ret = sk_detach_filter(sk);
e71a4783 758 break;
1da177e4 759
e71a4783
SH		 760 case SO_PASSSEC:
761 if (valbool)
762 set_bit(SOCK_PASSSEC, &sock->flags);
763 else
764 clear_bit(SOCK_PASSSEC, &sock->flags);
765 break;
4a19ec58
LAT		 766 case SO_MARK:
767 if (!capable(CAP_NET_ADMIN))
768 ret = -EPERM;
2a91525c 769 else
4a19ec58 770 sk->sk_mark = val;
4a19ec58 771 break;
877ce7c1 772
1da177e4
LT		 773 /* We implement the SO_SNDLOWAT etc to
774 not be settable (1003.1g 5.3) */
3b885787 775 case SO_RXQ_OVFL:
8083f0fc 776 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 777 break;
6e3e939f
JB		 778
779 case SO_WIFI_STATUS:
780 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
781 break;
782
ef64a54f
PE		 783 case SO_PEEK_OFF:
784 if (sock->ops->set_peek_off)
785 sock->ops->set_peek_off(sk, val);
786 else
787 ret = -EOPNOTSUPP;
788 break;
3bdc0eba
BG		 789
790 case SO_NOFCS:
791 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
792 break;
793
e71a4783
SH		 794 default:
795 ret = -ENOPROTOOPT;
796 break;
4ec93edb 797 }
1da177e4
LT		 798 release_sock(sk);
799 return ret;
800}
2a91525c 801EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 802
803
3f551f94
EB		 804void cred_to_ucred(struct pid *pid, const struct cred *cred,
805 struct ucred *ucred)
806{
807 ucred->pid = pid_vnr(pid);
808 ucred->uid = ucred->gid = -1;
809 if (cred) {
810 struct user_namespace *current_ns = current_user_ns();
811
812 ucred->uid = user_ns_map_uid(current_ns, cred, cred->euid);
813 ucred->gid = user_ns_map_gid(current_ns, cred, cred->egid);
814 }
815}
3924773a 816EXPORT_SYMBOL_GPL(cred_to_ucred);
3f551f94 817
1da177e4
LT		 818int sock_getsockopt(struct socket *sock, int level, int optname,
819 char __user *optval, int __user *optlen)
820{
821 struct sock *sk = sock->sk;
4ec93edb 822
e71a4783 823 union {
4ec93edb
YH		 824 int val;
825 struct linger ling;
1da177e4
LT		 826 struct timeval tm;
827 } v;
4ec93edb 828
4d0392be 829 int lv = sizeof(int);
1da177e4 830 int len;
4ec93edb 831
e71a4783 832 if (get_user(len, optlen))
4ec93edb 833 return -EFAULT;
e71a4783 834 if (len < 0)
1da177e4 835 return -EINVAL;
4ec93edb 836
50fee1de 837 memset(&v, 0, sizeof(v));
df0bca04 838
2a91525c 839 switch (optname) {
e71a4783
SH		 840 case SO_DEBUG:
841 v.val = sock_flag(sk, SOCK_DBG);
842 break;
843
844 case SO_DONTROUTE:
845 v.val = sock_flag(sk, SOCK_LOCALROUTE);
846 break;
847
848 case SO_BROADCAST:
849 v.val = !!sock_flag(sk, SOCK_BROADCAST);
850 break;
851
852 case SO_SNDBUF:
853 v.val = sk->sk_sndbuf;
854 break;
855
856 case SO_RCVBUF:
857 v.val = sk->sk_rcvbuf;
858 break;
859
860 case SO_REUSEADDR:
861 v.val = sk->sk_reuse;
862 break;
863
864 case SO_KEEPALIVE:
865 v.val = !!sock_flag(sk, SOCK_KEEPOPEN);
866 break;
867
868 case SO_TYPE:
869 v.val = sk->sk_type;
870 break;
871
49c794e9
JE		 872 case SO_PROTOCOL:
873 v.val = sk->sk_protocol;
874 break;
875
0d6038ee
JE		 876 case SO_DOMAIN:
877 v.val = sk->sk_family;
878 break;
879
e71a4783
SH		 880 case SO_ERROR:
881 v.val = -sock_error(sk);
2a91525c 882 if (v.val == 0)
e71a4783
SH		 883 v.val = xchg(&sk->sk_err_soft, 0);
884 break;
885
886 case SO_OOBINLINE:
887 v.val = !!sock_flag(sk, SOCK_URGINLINE);
888 break;
889
890 case SO_NO_CHECK:
891 v.val = sk->sk_no_check;
892 break;
893
894 case SO_PRIORITY:
895 v.val = sk->sk_priority;
896 break;
897
898 case SO_LINGER:
899 lv = sizeof(v.ling);
900 v.ling.l_onoff = !!sock_flag(sk, SOCK_LINGER);
901 v.ling.l_linger = sk->sk_lingertime / HZ;
902 break;
903
904 case SO_BSDCOMPAT:
905 sock_warn_obsolete_bsdism("getsockopt");
906 break;
907
908 case SO_TIMESTAMP:
92f37fd2
ED		 909 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
910 !sock_flag(sk, SOCK_RCVTSTAMPNS);
911 break;
912
913 case SO_TIMESTAMPNS:
914 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 915 break;
916
20d49473
PO		 917 case SO_TIMESTAMPING:
918 v.val = 0;
919 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
920 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
921 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
922 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
923 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
924 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
925 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
926 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
927 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
928 v.val |= SOF_TIMESTAMPING_SOFTWARE;
929 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
930 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
931 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
932 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
933 break;
934
e71a4783 935 case SO_RCVTIMEO:
2a91525c 936 lv = sizeof(struct timeval);
e71a4783
SH		 937 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
938 v.tm.tv_sec = 0;
939 v.tm.tv_usec = 0;
940 } else {
941 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
942 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
943 }
944 break;
945
946 case SO_SNDTIMEO:
2a91525c 947 lv = sizeof(struct timeval);
e71a4783
SH		 948 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
949 v.tm.tv_sec = 0;
950 v.tm.tv_usec = 0;
951 } else {
952 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
953 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
954 }
955 break;
1da177e4 956
e71a4783
SH		 957 case SO_RCVLOWAT:
958 v.val = sk->sk_rcvlowat;
959 break;
1da177e4 960
e71a4783 961 case SO_SNDLOWAT:
2a91525c 962 v.val = 1;
e71a4783 963 break;
1da177e4 964
e71a4783 965 case SO_PASSCRED:
82981930 966 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 967 break;
1da177e4 968
e71a4783 969 case SO_PEERCRED:
109f6e39
EB		 970 {
971 struct ucred peercred;
972 if (len > sizeof(peercred))
973 len = sizeof(peercred);
974 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
975 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 976 return -EFAULT;
977 goto lenout;
109f6e39 978 }
1da177e4 979
e71a4783
SH		 980 case SO_PEERNAME:
981 {
982 char address[128];
983
984 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
985 return -ENOTCONN;
986 if (lv < len)
987 return -EINVAL;
988 if (copy_to_user(optval, address, len))
989 return -EFAULT;
990 goto lenout;
991 }
1da177e4 992
e71a4783
SH		 993 /* Dubious BSD thing... Probably nobody even uses it, but
994 * the UNIX standard wants it for whatever reason... -DaveM
995 */
996 case SO_ACCEPTCONN:
997 v.val = sk->sk_state == TCP_LISTEN;
998 break;
1da177e4 999
e71a4783 1000 case SO_PASSSEC:
82981930 1001 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1002 break;
877ce7c1 1003
e71a4783
SH		 1004 case SO_PEERSEC:
1005 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1006
4a19ec58
LAT		 1007 case SO_MARK:
1008 v.val = sk->sk_mark;
1009 break;
1010
3b885787
NH		 1011 case SO_RXQ_OVFL:
1012 v.val = !!sock_flag(sk, SOCK_RXQ_OVFL);
1013 break;
1014
6e3e939f
JB		 1015 case SO_WIFI_STATUS:
1016 v.val = !!sock_flag(sk, SOCK_WIFI_STATUS);
1017 break;
1018
ef64a54f
PE		 1019 case SO_PEEK_OFF:
1020 if (!sock->ops->set_peek_off)
1021 return -EOPNOTSUPP;
1022
1023 v.val = sk->sk_peek_off;
1024 break;
bc2f7996
DM		 1025 case SO_NOFCS:
1026 v.val = !!sock_flag(sk, SOCK_NOFCS);
1027 break;
e71a4783
SH		 1028 default:
1029 return -ENOPROTOOPT;
1da177e4 1030 }
e71a4783 1031
1da177e4
LT		 1032 if (len > lv)
1033 len = lv;
1034 if (copy_to_user(optval, &v, len))
1035 return -EFAULT;
1036lenout:
4ec93edb
YH		 1037 if (put_user(len, optlen))
1038 return -EFAULT;
1039 return 0;
1da177e4
LT		 1040}
1041
a5b5bb9a
IM		 1042/*
1043 * Initialize an sk_lock.
1044 *
1045 * (We also register the sk_lock with the lock validator.)
1046 */
b6f99a21 1047static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1048{
ed07536e
PZ		 1049 sock_lock_init_class_and_name(sk,
1050 af_family_slock_key_strings[sk->sk_family],
1051 af_family_slock_keys + sk->sk_family,
1052 af_family_key_strings[sk->sk_family],
1053 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1054}
1055
4dc6dc71
ED		 1056/*
1057 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1058 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1059 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1060 */
f1a6c4da
PE		 1061static void sock_copy(struct sock *nsk, const struct sock *osk)
1062{
1063#ifdef CONFIG_SECURITY_NETWORK
1064 void *sptr = nsk->sk_security;
1065#endif
68835aba
ED		 1066 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1067
1068 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1069 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1070
f1a6c4da
PE		 1071#ifdef CONFIG_SECURITY_NETWORK
1072 nsk->sk_security = sptr;
1073 security_sk_clone(osk, nsk);
1074#endif
1075}
1076
fcbdf09d
OP		 1077/*
1078 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
1079 * un-modified. Special care is taken when initializing object to zero.
1080 */
1081static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1082{
1083 if (offsetof(struct sock, sk_node.next) != 0)
1084 memset(sk, 0, offsetof(struct sock, sk_node.next));
1085 memset(&sk->sk_node.pprev, 0,
1086 size - offsetof(struct sock, sk_node.pprev));
1087}
1088
1089void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1090{
1091 unsigned long nulls1, nulls2;
1092
1093 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1094 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1095 if (nulls1 > nulls2)
1096 swap(nulls1, nulls2);
1097
1098 if (nulls1 != 0)
1099 memset((char *)sk, 0, nulls1);
1100 memset((char *)sk + nulls1 + sizeof(void *), 0,
1101 nulls2 - nulls1 - sizeof(void *));
1102 memset((char *)sk + nulls2 + sizeof(void *), 0,
1103 size - nulls2 - sizeof(void *));
1104}
1105EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1106
2e4afe7b
PE		 1107static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1108 int family)
c308c1b2
PE		 1109{
1110 struct sock *sk;
1111 struct kmem_cache *slab;
1112
1113 slab = prot->slab;
e912b114
ED		 1114 if (slab != NULL) {
1115 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1116 if (!sk)
1117 return sk;
1118 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1119 if (prot->clear_sk)
1120 prot->clear_sk(sk, prot->obj_size);
1121 else
1122 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1123 }
fcbdf09d 1124 } else
c308c1b2
PE		 1125 sk = kmalloc(prot->obj_size, priority);
1126
2e4afe7b 1127 if (sk != NULL) {
a98b65a3
VN		 1128 kmemcheck_annotate_bitfield(sk, flags);
1129
2e4afe7b
PE		 1130 if (security_sk_alloc(sk, family, priority))
1131 goto out_free;
1132
1133 if (!try_module_get(prot->owner))
1134 goto out_free_sec;
e022f0b4 1135 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1136 }
1137
c308c1b2 1138 return sk;
2e4afe7b
PE		 1139
1140out_free_sec:
1141 security_sk_free(sk);
1142out_free:
1143 if (slab != NULL)
1144 kmem_cache_free(slab, sk);
1145 else
1146 kfree(sk);
1147 return NULL;
c308c1b2
PE		 1148}
1149
1150static void sk_prot_free(struct proto *prot, struct sock *sk)
1151{
1152 struct kmem_cache *slab;
2e4afe7b 1153 struct module *owner;
c308c1b2 1154
2e4afe7b 1155 owner = prot->owner;
c308c1b2 1156 slab = prot->slab;
2e4afe7b
PE		 1157
1158 security_sk_free(sk);
c308c1b2
PE		 1159 if (slab != NULL)
1160 kmem_cache_free(slab, sk);
1161 else
1162 kfree(sk);
2e4afe7b 1163 module_put(owner);
c308c1b2
PE		 1164}
1165
f8451725
HX		 1166#ifdef CONFIG_CGROUPS
1167void sock_update_classid(struct sock *sk)
1168{
1144182a 1169 u32 classid;
f8451725 1170
1144182a
PM		 1171 rcu_read_lock(); /* doing current task, which cannot vanish. */
1172 classid = task_cls_classid(current);
1173 rcu_read_unlock();
f8451725
HX		 1174 if (classid && classid != sk->sk_classid)
1175 sk->sk_classid = classid;
1176}
82862742 1177EXPORT_SYMBOL(sock_update_classid);
5bc1421e
NH		 1178
1179void sock_update_netprioidx(struct sock *sk)
1180{
5bc1421e
NH		 1181 if (in_interrupt())
1182 return;
2b73bc65
NH		 1183
1184 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1185}
1186EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1187#endif
1188
1da177e4
LT		 1189/**
1190 * sk_alloc - All socket objects are allocated here
c4ea43c5 1191 * @net: the applicable net namespace
4dc3b16b
PP		 1192 * @family: protocol family
1193 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1194 * @prot: struct proto associated with this new sock instance
1da177e4 1195 */
1b8d7ae4 1196struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1197 struct proto *prot)
1da177e4 1198{
c308c1b2 1199 struct sock *sk;
1da177e4 1200
154adbc8 1201 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1202 if (sk) {
154adbc8
PE		 1203 sk->sk_family = family;
1204 /*
1205 * See comment in struct sock definition to understand
1206 * why we need sk_prot_creator -acme
1207 */
1208 sk->sk_prot = sk->sk_prot_creator = prot;
1209 sock_lock_init(sk);
3b1e0a65 1210 sock_net_set(sk, get_net(net));
d66ee058 1211 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725
HX		 1212
1213 sock_update_classid(sk);
5bc1421e 1214 sock_update_netprioidx(sk);
1da177e4 1215 }
a79af59e 1216
2e4afe7b 1217 return sk;
1da177e4 1218}
2a91525c 1219EXPORT_SYMBOL(sk_alloc);
1da177e4 1220
2b85a34e 1221static void __sk_free(struct sock *sk)
1da177e4
LT		 1222{
1223 struct sk_filter *filter;
1da177e4
LT		 1224
1225 if (sk->sk_destruct)
1226 sk->sk_destruct(sk);
1227
a898def2
PM		 1228 filter = rcu_dereference_check(sk->sk_filter,
1229 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1230 if (filter) {
309dd5fc 1231 sk_filter_uncharge(sk, filter);
a9b3cd7f 1232 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1233 }
1234
08e29af3 1235 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1236
1237 if (atomic_read(&sk->sk_omem_alloc))
1238 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n",
0dc47877 1239 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1240
109f6e39
EB		 1241 if (sk->sk_peer_cred)
1242 put_cred(sk->sk_peer_cred);
1243 put_pid(sk->sk_peer_pid);
3b1e0a65 1244 put_net(sock_net(sk));
c308c1b2 1245 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1246}
2b85a34e
ED		 1247
1248void sk_free(struct sock *sk)
1249{
1250 /*
25985edc 1251 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1252 * some packets are still in some tx queue.
1253 * If not null, sock_wfree() will call __sk_free(sk) later
1254 */
1255 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1256 __sk_free(sk);
1257}
2a91525c 1258EXPORT_SYMBOL(sk_free);
1da177e4 1259
edf02087 1260/*
25985edc
LDM		 1261 * Last sock_put should drop reference to sk->sk_net. It has already
1262 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1263 * is not an option.
25985edc 1264 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1265 * destroy it in the context of init_net.
1266 */
1267void sk_release_kernel(struct sock *sk)
1268{
1269 if (sk == NULL || sk->sk_socket == NULL)
1270 return;
1271
1272 sock_hold(sk);
1273 sock_release(sk->sk_socket);
65a18ec5 1274 release_net(sock_net(sk));
3b1e0a65 1275 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1276 sock_put(sk);
1277}
45af1754 1278EXPORT_SYMBOL(sk_release_kernel);
edf02087 1279
475f1b52
SR		 1280static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1281{
1282 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1283 sock_update_memcg(newsk);
1284}
1285
e56c57d0
ED		 1286/**
1287 * sk_clone_lock - clone a socket, and lock its clone
1288 * @sk: the socket to clone
1289 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1290 *
1291 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1292 */
1293struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1294{
8fd1d178 1295 struct sock *newsk;
87d11ceb 1296
8fd1d178 1297 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1298 if (newsk != NULL) {
1299 struct sk_filter *filter;
1300
892c141e 1301 sock_copy(newsk, sk);
87d11ceb
ACM		 1302
1303 /* SANITY */
3b1e0a65 1304 get_net(sock_net(newsk));
87d11ceb
ACM		 1305 sk_node_init(&newsk->sk_node);
1306 sock_lock_init(newsk);
1307 bh_lock_sock(newsk);
fa438ccf 1308 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1309 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1310
1311 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1312 /*
1313 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1314 */
1315 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1316 atomic_set(&newsk->sk_omem_alloc, 0);
1317 skb_queue_head_init(&newsk->sk_receive_queue);
1318 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1319#ifdef CONFIG_NET_DMA
1320 skb_queue_head_init(&newsk->sk_async_wait_queue);
1321#endif
87d11ceb 1322
b6c6712a 1323 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1324 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1325 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1326 af_callback_keys + newsk->sk_family,
1327 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1328
1329 newsk->sk_dst_cache = NULL;
1330 newsk->sk_wmem_queued = 0;
1331 newsk->sk_forward_alloc = 0;
1332 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1333 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1334
1335 sock_reset_flag(newsk, SOCK_DONE);
1336 skb_queue_head_init(&newsk->sk_error_queue);
1337
0d7da9dd 1338 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1339 if (filter != NULL)
1340 sk_filter_charge(newsk, filter);
1341
1342 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1343 /* It is still raw copy of parent, so invalidate
1344 * destructor and make plain sk_free() */
1345 newsk->sk_destruct = NULL;
b0691c8e 1346 bh_unlock_sock(newsk);
87d11ceb
ACM		 1347 sk_free(newsk);
1348 newsk = NULL;
1349 goto out;
1350 }
1351
1352 newsk->sk_err = 0;
1353 newsk->sk_priority = 0;
4dc6dc71
ED		 1354 /*
1355 * Before updating sk_refcnt, we must commit prior changes to memory
1356 * (Documentation/RCU/rculist_nulls.txt for details)
1357 */
1358 smp_wmb();
87d11ceb
ACM		 1359 atomic_set(&newsk->sk_refcnt, 2);
1360
1361 /*
1362 * Increment the counter in the same struct proto as the master
1363 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1364 * is the same as sk->sk_prot->socks, as this field was copied
1365 * with memcpy).
1366 *
1367 * This _changes_ the previous behaviour, where
1368 * tcp_create_openreq_child always was incrementing the
1369 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1370 * to be taken into account in all callers. -acme
1371 */
1372 sk_refcnt_debug_inc(newsk);
972692e0 1373 sk_set_socket(newsk, NULL);
43815482 1374 newsk->sk_wq = NULL;
87d11ceb 1375
f3f511e1
GC		 1376 sk_update_clone(sk, newsk);
1377
87d11ceb 1378 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1379 sk_sockets_allocated_inc(newsk);
704da560 1380
08e29af3 1381 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1382 net_enable_timestamp();
87d11ceb
ACM		 1383 }
1384out:
1385 return newsk;
1386}
e56c57d0 1387EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1388
9958089a
AK		 1389void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1390{
1391 __sk_dst_set(sk, dst);
1392 sk->sk_route_caps = dst->dev->features;
1393 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1394 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1395 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1396 if (sk_can_gso(sk)) {
82cc1a7a 1397 if (dst->header_len) {
9958089a 1398 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1399 } else {
9958089a 1400 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a
PWJ		 1401 sk->sk_gso_max_size = dst->dev->gso_max_size;
1402 }
9958089a
AK		 1403 }
1404}
1405EXPORT_SYMBOL_GPL(sk_setup_caps);
1406
1da177e4
LT		 1407void __init sk_init(void)
1408{
4481374c 1409 if (totalram_pages <= 4096) {
1da177e4
LT		 1410 sysctl_wmem_max = 32767;
1411 sysctl_rmem_max = 32767;
1412 sysctl_wmem_default = 32767;
1413 sysctl_rmem_default = 32767;
4481374c 1414 } else if (totalram_pages >= 131072) {
1da177e4
LT		 1415 sysctl_wmem_max = 131071;
1416 sysctl_rmem_max = 131071;
1417 }
1418}
1419
1420/*
1421 * Simple resource managers for sockets.
1422 */
1423
1424
4ec93edb
YH		 1425/*
1426 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1427 */
1428void sock_wfree(struct sk_buff *skb)
1429{
1430 struct sock *sk = skb->sk;
d99927f4 1431 unsigned int len = skb->truesize;
1da177e4 1432
d99927f4
ED		 1433 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1434 /*
1435 * Keep a reference on sk_wmem_alloc, this will be released
1436 * after sk_write_space() call
1437 */
1438 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1439 sk->sk_write_space(sk);
d99927f4
ED		 1440 len = 1;
1441 }
2b85a34e 1442 /*
d99927f4
ED		 1443 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1444 * could not do because of in-flight packets
2b85a34e 1445 */
d99927f4 1446 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1447 __sk_free(sk);
1da177e4 1448}
2a91525c 1449EXPORT_SYMBOL(sock_wfree);
1da177e4 1450
4ec93edb
YH		 1451/*
1452 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1453 */
1454void sock_rfree(struct sk_buff *skb)
1455{
1456 struct sock *sk = skb->sk;
d361fd59 1457 unsigned int len = skb->truesize;
1da177e4 1458
d361fd59
ED		 1459 atomic_sub(len, &sk->sk_rmem_alloc);
1460 sk_mem_uncharge(sk, len);
1da177e4 1461}
2a91525c 1462EXPORT_SYMBOL(sock_rfree);
1da177e4
LT		 1463
1464
1465int sock_i_uid(struct sock *sk)
1466{
1467 int uid;
1468
f064af1e 1469 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1470 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
f064af1e 1471 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1472 return uid;
1473}
2a91525c 1474EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1475
1476unsigned long sock_i_ino(struct sock *sk)
1477{
1478 unsigned long ino;
1479
f064af1e 1480 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1481 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1482 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1483 return ino;
1484}
2a91525c 1485EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1486
1487/*
1488 * Allocate a skb from the socket's send buffer.
1489 */
86a76caf 1490struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1491 gfp_t priority)
1da177e4
LT		 1492{
1493 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1494 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1495 if (skb) {
1496 skb_set_owner_w(skb, sk);
1497 return skb;
1498 }
1499 }
1500 return NULL;
1501}
2a91525c 1502EXPORT_SYMBOL(sock_wmalloc);
1da177e4
LT		 1503
1504/*
1505 * Allocate a skb from the socket's receive buffer.
4ec93edb 1506 */
86a76caf 1507struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1508 gfp_t priority)
1da177e4
LT		 1509{
1510 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1511 struct sk_buff *skb = alloc_skb(size, priority);
1512 if (skb) {
1513 skb_set_owner_r(skb, sk);
1514 return skb;
1515 }
1516 }
1517 return NULL;
1518}
1519
4ec93edb 1520/*
1da177e4 1521 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1522 */
dd0fc66f 1523void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1524{
95c96174 1525 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1526 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1527 void *mem;
1528 /* First do the add, to avoid the race if kmalloc
4ec93edb 1529 * might sleep.
1da177e4
LT		 1530 */
1531 atomic_add(size, &sk->sk_omem_alloc);
1532 mem = kmalloc(size, priority);
1533 if (mem)
1534 return mem;
1535 atomic_sub(size, &sk->sk_omem_alloc);
1536 }
1537 return NULL;
1538}
2a91525c 1539EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1540
1541/*
1542 * Free an option memory block.
1543 */
1544void sock_kfree_s(struct sock *sk, void *mem, int size)
1545{
1546 kfree(mem);
1547 atomic_sub(size, &sk->sk_omem_alloc);
1548}
2a91525c 1549EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1550
1551/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1552 I think, these locks should be removed for datagram sockets.
1553 */
2a91525c 1554static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1555{
1556 DEFINE_WAIT(wait);
1557
1558 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1559 for (;;) {
1560 if (!timeo)
1561 break;
1562 if (signal_pending(current))
1563 break;
1564 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1565 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1566 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1567 break;
1568 if (sk->sk_shutdown & SEND_SHUTDOWN)
1569 break;
1570 if (sk->sk_err)
1571 break;
1572 timeo = schedule_timeout(timeo);
1573 }
aa395145 1574 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1575 return timeo;
1576}
1577
1578
1579/*
1580 * Generic send/receive buffer handlers
1581 */
1582
4cc7f68d
HX		 1583struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1584 unsigned long data_len, int noblock,
1585 int *errcode)
1da177e4
LT		 1586{
1587 struct sk_buff *skb;
7d877f3b 1588 gfp_t gfp_mask;
1da177e4
LT		 1589 long timeo;
1590 int err;
1591
1592 gfp_mask = sk->sk_allocation;
1593 if (gfp_mask & __GFP_WAIT)
1594 gfp_mask |= __GFP_REPEAT;
1595
1596 timeo = sock_sndtimeo(sk, noblock);
1597 while (1) {
1598 err = sock_error(sk);
1599 if (err != 0)
1600 goto failure;
1601
1602 err = -EPIPE;
1603 if (sk->sk_shutdown & SEND_SHUTDOWN)
1604 goto failure;
1605
1606 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1607 skb = alloc_skb(header_len, gfp_mask);
1da177e4
LT		 1608 if (skb) {
1609 int npages;
1610 int i;
1611
1612 /* No pages, we're done... */
1613 if (!data_len)
1614 break;
1615
1616 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1617 skb->truesize += data_len;
1618 skb_shinfo(skb)->nr_frags = npages;
1619 for (i = 0; i < npages; i++) {
1620 struct page *page;
1da177e4
LT		 1621
1622 page = alloc_pages(sk->sk_allocation, 0);
1623 if (!page) {
1624 err = -ENOBUFS;
1625 skb_shinfo(skb)->nr_frags = i;
1626 kfree_skb(skb);
1627 goto failure;
1628 }
1629
ea2ab693
IC		 1630 __skb_fill_page_desc(skb, i,
1631 page, 0,
1632 (data_len >= PAGE_SIZE ?
1633 PAGE_SIZE :
1634 data_len));
1da177e4
LT		 1635 data_len -= PAGE_SIZE;
1636 }
1637
1638 /* Full success... */
1639 break;
1640 }
1641 err = -ENOBUFS;
1642 goto failure;
1643 }
1644 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1645 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1646 err = -EAGAIN;
1647 if (!timeo)
1648 goto failure;
1649 if (signal_pending(current))
1650 goto interrupted;
1651 timeo = sock_wait_for_wmem(sk, timeo);
1652 }
1653
1654 skb_set_owner_w(skb, sk);
1655 return skb;
1656
1657interrupted:
1658 err = sock_intr_errno(timeo);
1659failure:
1660 *errcode = err;
1661 return NULL;
1662}
4cc7f68d 1663EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1664
4ec93edb 1665struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1666 int noblock, int *errcode)
1667{
1668 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1669}
2a91525c 1670EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4
LT		 1671
1672static void __lock_sock(struct sock *sk)
f39234d6
NK		 1673 __releases(&sk->sk_lock.slock)
1674 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1675{
1676 DEFINE_WAIT(wait);
1677
e71a4783 1678 for (;;) {
1da177e4
LT		 1679 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1680 TASK_UNINTERRUPTIBLE);
1681 spin_unlock_bh(&sk->sk_lock.slock);
1682 schedule();
1683 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1684 if (!sock_owned_by_user(sk))
1da177e4
LT		 1685 break;
1686 }
1687 finish_wait(&sk->sk_lock.wq, &wait);
1688}
1689
1690static void __release_sock(struct sock *sk)
f39234d6
NK		 1691 __releases(&sk->sk_lock.slock)
1692 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1693{
1694 struct sk_buff *skb = sk->sk_backlog.head;
1695
1696 do {
1697 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1698 bh_unlock_sock(sk);
1699
1700 do {
1701 struct sk_buff *next = skb->next;
1702
e4cbb02a 1703 prefetch(next);
7fee226a 1704 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1705 skb->next = NULL;
c57943a1 1706 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1707
1708 /*
1709 * We are in process context here with softirqs
1710 * disabled, use cond_resched_softirq() to preempt.
1711 * This is safe to do because we've taken the backlog
1712 * queue private:
1713 */
1714 cond_resched_softirq();
1715
1716 skb = next;
1717 } while (skb != NULL);
1718
1719 bh_lock_sock(sk);
e71a4783 1720 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1721
1722 /*
1723 * Doing the zeroing here guarantee we can not loop forever
1724 * while a wild producer attempts to flood us.
1725 */
1726 sk->sk_backlog.len = 0;
1da177e4
LT		 1727}
1728
1729/**
1730 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1731 * @sk: sock to wait on
1732 * @timeo: for how long
1da177e4
LT		 1733 *
1734 * Now socket state including sk->sk_err is changed only under lock,
1735 * hence we may omit checks after joining wait queue.
1736 * We check receive queue before schedule() only as optimization;
1737 * it is very likely that release_sock() added new data.
1738 */
1739int sk_wait_data(struct sock *sk, long *timeo)
1740{
1741 int rc;
1742 DEFINE_WAIT(wait);
1743
aa395145 1744 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1745 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1746 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1747 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1748 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1749 return rc;
1750}
1da177e4
LT		 1751EXPORT_SYMBOL(sk_wait_data);
1752
3ab224be
HA		 1753/**
1754 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1755 * @sk: socket
1756 * @size: memory size to allocate
1757 * @kind: allocation type
1758 *
1759 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1760 * rmem allocation. This function assumes that protocols which have
1761 * memory_pressure use sk_wmem_queued as write buffer accounting.
1762 */
1763int __sk_mem_schedule(struct sock *sk, int size, int kind)
1764{
1765 struct proto *prot = sk->sk_prot;
1766 int amt = sk_mem_pages(size);
8d987e5c 1767 long allocated;
e1aab161 1768 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1769
1770 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1771
e1aab161 1772 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1773
1774 /* Under limit. */
e1aab161
GC		 1775 if (parent_status == UNDER_LIMIT &&
1776 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1777 sk_leave_memory_pressure(sk);
3ab224be
HA		 1778 return 1;
1779 }
1780
e1aab161
GC		 1781 /* Under pressure. (we or our parents) */
1782 if ((parent_status > SOFT_LIMIT) ||
1783 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1784 sk_enter_memory_pressure(sk);
3ab224be 1785
e1aab161
GC		 1786 /* Over hard limit (we or our parents) */
1787 if ((parent_status == OVER_LIMIT) ||
1788 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1789 goto suppress_allocation;
1790
1791 /* guarantee minimum buffer size under pressure */
1792 if (kind == SK_MEM_RECV) {
1793 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1794 return 1;
180d8cd9 1795
3ab224be
HA		 1796 } else { /* SK_MEM_SEND */
1797 if (sk->sk_type == SOCK_STREAM) {
1798 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1799 return 1;
1800 } else if (atomic_read(&sk->sk_wmem_alloc) <
1801 prot->sysctl_wmem[0])
1802 return 1;
1803 }
1804
180d8cd9 1805 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 1806 int alloc;
1807
180d8cd9 1808 if (!sk_under_memory_pressure(sk))
1748376b 1809 return 1;
180d8cd9
GC		 1810 alloc = sk_sockets_allocated_read_positive(sk);
1811 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 1812 sk_mem_pages(sk->sk_wmem_queued +
1813 atomic_read(&sk->sk_rmem_alloc) +
1814 sk->sk_forward_alloc))
1815 return 1;
1816 }
1817
1818suppress_allocation:
1819
1820 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1821 sk_stream_moderate_sndbuf(sk);
1822
1823 /* Fail only if socket is _under_ its sndbuf.
1824 * In this case we cannot block, so that we have to fail.
1825 */
1826 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1827 return 1;
1828 }
1829
3847ce32
SM		 1830 trace_sock_exceed_buf_limit(sk, prot, allocated);
1831
3ab224be
HA		 1832 /* Alas. Undo changes. */
1833 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 1834
0e90b31f 1835 sk_memory_allocated_sub(sk, amt);
180d8cd9 1836
3ab224be
HA		 1837 return 0;
1838}
3ab224be
HA		 1839EXPORT_SYMBOL(__sk_mem_schedule);
1840
1841/**
1842 * __sk_reclaim - reclaim memory_allocated
1843 * @sk: socket
1844 */
1845void __sk_mem_reclaim(struct sock *sk)
1846{
180d8cd9 1847 sk_memory_allocated_sub(sk,
0e90b31f 1848 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 1849 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
1850
180d8cd9
GC		 1851 if (sk_under_memory_pressure(sk) &&
1852 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
1853 sk_leave_memory_pressure(sk);
3ab224be 1854}
3ab224be
HA		 1855EXPORT_SYMBOL(__sk_mem_reclaim);
1856
1857
1da177e4
LT		 1858/*
1859 * Set of default routines for initialising struct proto_ops when
1860 * the protocol does not support a particular function. In certain
1861 * cases where it makes no sense for a protocol to have a "do nothing"
1862 * function, some default processing is provided.
1863 */
1864
1865int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1866{
1867 return -EOPNOTSUPP;
1868}
2a91525c 1869EXPORT_SYMBOL(sock_no_bind);
1da177e4 1870
4ec93edb 1871int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1872 int len, int flags)
1873{
1874 return -EOPNOTSUPP;
1875}
2a91525c 1876EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 1877
1878int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1879{
1880 return -EOPNOTSUPP;
1881}
2a91525c 1882EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 1883
1884int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1885{
1886 return -EOPNOTSUPP;
1887}
2a91525c 1888EXPORT_SYMBOL(sock_no_accept);
1da177e4 1889
4ec93edb 1890int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1891 int *len, int peer)
1892{
1893 return -EOPNOTSUPP;
1894}
2a91525c 1895EXPORT_SYMBOL(sock_no_getname);
1da177e4 1896
2a91525c 1897unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 1898{
1899 return 0;
1900}
2a91525c 1901EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 1902
1903int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1904{
1905 return -EOPNOTSUPP;
1906}
2a91525c 1907EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 1908
1909int sock_no_listen(struct socket *sock, int backlog)
1910{
1911 return -EOPNOTSUPP;
1912}
2a91525c 1913EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 1914
1915int sock_no_shutdown(struct socket *sock, int how)
1916{
1917 return -EOPNOTSUPP;
1918}
2a91525c 1919EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 1920
1921int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 1922 char __user *optval, unsigned int optlen)
1da177e4
LT		 1923{
1924 return -EOPNOTSUPP;
1925}
2a91525c 1926EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 1927
1928int sock_no_getsockopt(struct socket *sock, int level, int optname,
1929 char __user *optval, int __user *optlen)
1930{
1931 return -EOPNOTSUPP;
1932}
2a91525c 1933EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 1934
1935int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1936 size_t len)
1937{
1938 return -EOPNOTSUPP;
1939}
2a91525c 1940EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 1941
1942int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1943 size_t len, int flags)
1944{
1945 return -EOPNOTSUPP;
1946}
2a91525c 1947EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 1948
1949int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
1950{
1951 /* Mirror missing mmap method error code */
1952 return -ENODEV;
1953}
2a91525c 1954EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 1955
1956ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
1957{
1958 ssize_t res;
1959 struct msghdr msg = {.msg_flags = flags};
1960 struct kvec iov;
1961 char *kaddr = kmap(page);
1962 iov.iov_base = kaddr + offset;
1963 iov.iov_len = size;
1964 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
1965 kunmap(page);
1966 return res;
1967}
2a91525c 1968EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 1969
1970/*
1971 * Default Socket Callbacks
1972 */
1973
1974static void sock_def_wakeup(struct sock *sk)
1975{
43815482
ED		 1976 struct socket_wq *wq;
1977
1978 rcu_read_lock();
1979 wq = rcu_dereference(sk->sk_wq);
1980 if (wq_has_sleeper(wq))
1981 wake_up_interruptible_all(&wq->wait);
1982 rcu_read_unlock();
1da177e4
LT		 1983}
1984
1985static void sock_def_error_report(struct sock *sk)
1986{
43815482
ED		 1987 struct socket_wq *wq;
1988
1989 rcu_read_lock();
1990 wq = rcu_dereference(sk->sk_wq);
1991 if (wq_has_sleeper(wq))
1992 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 1993 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 1994 rcu_read_unlock();
1da177e4
LT		 1995}
1996
1997static void sock_def_readable(struct sock *sk, int len)
1998{
43815482
ED		 1999 struct socket_wq *wq;
2000
2001 rcu_read_lock();
2002 wq = rcu_dereference(sk->sk_wq);
2003 if (wq_has_sleeper(wq))
2c6607c6 2004 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2005 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2006 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2007 rcu_read_unlock();
1da177e4
LT		 2008}
2009
2010static void sock_def_write_space(struct sock *sk)
2011{
43815482
ED		 2012 struct socket_wq *wq;
2013
2014 rcu_read_lock();
1da177e4
LT		 2015
2016 /* Do not wake up a writer until he can make "significant"
2017 * progress. --DaveM
2018 */
e71a4783 2019 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2020 wq = rcu_dereference(sk->sk_wq);
2021 if (wq_has_sleeper(wq))
2022 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2023 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2024
2025 /* Should agree with poll, otherwise some programs break */
2026 if (sock_writeable(sk))
8d8ad9d7 2027 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2028 }
2029
43815482 2030 rcu_read_unlock();
1da177e4
LT		 2031}
2032
2033static void sock_def_destruct(struct sock *sk)
2034{
a51482bd 2035 kfree(sk->sk_protinfo);
1da177e4
LT		 2036}
2037
2038void sk_send_sigurg(struct sock *sk)
2039{
2040 if (sk->sk_socket && sk->sk_socket->file)
2041 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2042 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2043}
2a91525c 2044EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2045
2046void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2047 unsigned long expires)
2048{
2049 if (!mod_timer(timer, expires))
2050 sock_hold(sk);
2051}
1da177e4
LT		 2052EXPORT_SYMBOL(sk_reset_timer);
2053
2054void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2055{
2056 if (timer_pending(timer) && del_timer(timer))
2057 __sock_put(sk);
2058}
1da177e4
LT		 2059EXPORT_SYMBOL(sk_stop_timer);
2060
2061void sock_init_data(struct socket *sock, struct sock *sk)
2062{
2063 skb_queue_head_init(&sk->sk_receive_queue);
2064 skb_queue_head_init(&sk->sk_write_queue);
2065 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2066#ifdef CONFIG_NET_DMA
2067 skb_queue_head_init(&sk->sk_async_wait_queue);
2068#endif
1da177e4
LT		 2069
2070 sk->sk_send_head = NULL;
2071
2072 init_timer(&sk->sk_timer);
4ec93edb 2073
1da177e4
LT		 2074 sk->sk_allocation = GFP_KERNEL;
2075 sk->sk_rcvbuf = sysctl_rmem_default;
2076 sk->sk_sndbuf = sysctl_wmem_default;
2077 sk->sk_state = TCP_CLOSE;
972692e0 2078 sk_set_socket(sk, sock);
1da177e4
LT		 2079
2080 sock_set_flag(sk, SOCK_ZAPPED);
2081
e71a4783 2082 if (sock) {
1da177e4 2083 sk->sk_type = sock->type;
43815482 2084 sk->sk_wq = sock->wq;
1da177e4
LT		 2085 sock->sk = sk;
2086 } else
43815482 2087 sk->sk_wq = NULL;
1da177e4 2088
b6c6712a 2089 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2090 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2091 lockdep_set_class_and_name(&sk->sk_callback_lock,
2092 af_callback_keys + sk->sk_family,
2093 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2094
2095 sk->sk_state_change = sock_def_wakeup;
2096 sk->sk_data_ready = sock_def_readable;
2097 sk->sk_write_space = sock_def_write_space;
2098 sk->sk_error_report = sock_def_error_report;
2099 sk->sk_destruct = sock_def_destruct;
2100
2101 sk->sk_sndmsg_page = NULL;
2102 sk->sk_sndmsg_off = 0;
ef64a54f 2103 sk->sk_peek_off = -1;
1da177e4 2104
109f6e39
EB		 2105 sk->sk_peer_pid = NULL;
2106 sk->sk_peer_cred = NULL;
1da177e4
LT		 2107 sk->sk_write_pending = 0;
2108 sk->sk_rcvlowat = 1;
2109 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2110 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2111
f37f0afb 2112 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2113
4dc6dc71
ED		 2114 /*
2115 * Before updating sk_refcnt, we must commit prior changes to memory
2116 * (Documentation/RCU/rculist_nulls.txt for details)
2117 */
2118 smp_wmb();
1da177e4 2119 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2120 atomic_set(&sk->sk_drops, 0);
1da177e4 2121}
2a91525c 2122EXPORT_SYMBOL(sock_init_data);
1da177e4 2123
b5606c2d 2124void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2125{
2126 might_sleep();
a5b5bb9a 2127 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2128 if (sk->sk_lock.owned)
1da177e4 2129 __lock_sock(sk);
d2e9117c 2130 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2131 spin_unlock(&sk->sk_lock.slock);
2132 /*
2133 * The sk_lock has mutex_lock() semantics here:
2134 */
fcc70d5f 2135 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2136 local_bh_enable();
1da177e4 2137}
fcc70d5f 2138EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2139
b5606c2d 2140void release_sock(struct sock *sk)
1da177e4 2141{
a5b5bb9a
IM		 2142 /*
2143 * The sk_lock has mutex_unlock() semantics:
2144 */
2145 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2146
2147 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2148 if (sk->sk_backlog.tail)
2149 __release_sock(sk);
d2e9117c 2150 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2151 if (waitqueue_active(&sk->sk_lock.wq))
2152 wake_up(&sk->sk_lock.wq);
2153 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2154}
2155EXPORT_SYMBOL(release_sock);
2156
8a74ad60
ED		 2157/**
2158 * lock_sock_fast - fast version of lock_sock
2159 * @sk: socket
2160 *
2161 * This version should be used for very small section, where process wont block
2162 * return false if fast path is taken
2163 * sk_lock.slock locked, owned = 0, BH disabled
2164 * return true if slow path is taken
2165 * sk_lock.slock unlocked, owned = 1, BH enabled
2166 */
2167bool lock_sock_fast(struct sock *sk)
2168{
2169 might_sleep();
2170 spin_lock_bh(&sk->sk_lock.slock);
2171
2172 if (!sk->sk_lock.owned)
2173 /*
2174 * Note : We must disable BH
2175 */
2176 return false;
2177
2178 __lock_sock(sk);
2179 sk->sk_lock.owned = 1;
2180 spin_unlock(&sk->sk_lock.slock);
2181 /*
2182 * The sk_lock has mutex_lock() semantics here:
2183 */
2184 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2185 local_bh_enable();
2186 return true;
2187}
2188EXPORT_SYMBOL(lock_sock_fast);
2189
1da177e4 2190int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2191{
b7aa0bf7 2192 struct timeval tv;
1da177e4 2193 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2194 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2195 tv = ktime_to_timeval(sk->sk_stamp);
2196 if (tv.tv_sec == -1)
1da177e4 2197 return -ENOENT;
b7aa0bf7
ED		 2198 if (tv.tv_sec == 0) {
2199 sk->sk_stamp = ktime_get_real();
2200 tv = ktime_to_timeval(sk->sk_stamp);
2201 }
2202 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2203}
1da177e4
LT		 2204EXPORT_SYMBOL(sock_get_timestamp);
2205
ae40eb1e
ED		 2206int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2207{
2208 struct timespec ts;
2209 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2210 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2211 ts = ktime_to_timespec(sk->sk_stamp);
2212 if (ts.tv_sec == -1)
2213 return -ENOENT;
2214 if (ts.tv_sec == 0) {
2215 sk->sk_stamp = ktime_get_real();
2216 ts = ktime_to_timespec(sk->sk_stamp);
2217 }
2218 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2219}
2220EXPORT_SYMBOL(sock_get_timestampns);
2221
20d49473 2222void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2223{
20d49473 2224 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2225 unsigned long previous_flags = sk->sk_flags;
2226
20d49473
PO		 2227 sock_set_flag(sk, flag);
2228 /*
2229 * we just set one of the two flags which require net
2230 * time stamping, but time stamping might have been on
2231 * already because of the other one
2232 */
08e29af3 2233 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2234 net_enable_timestamp();
1da177e4
LT		 2235 }
2236}
1da177e4
LT		 2237
2238/*
2239 * Get a socket option on an socket.
2240 *
2241 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2242 * asynchronous errors should be reported by getsockopt. We assume
2243 * this means if you specify SO_ERROR (otherwise whats the point of it).
2244 */
2245int sock_common_getsockopt(struct socket *sock, int level, int optname,
2246 char __user *optval, int __user *optlen)
2247{
2248 struct sock *sk = sock->sk;
2249
2250 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2251}
1da177e4
LT		 2252EXPORT_SYMBOL(sock_common_getsockopt);
2253
3fdadf7d 2254#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2255int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2256 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2257{
2258 struct sock *sk = sock->sk;
2259
1e51f951 2260 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2261 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2262 optval, optlen);
3fdadf7d
DM		 2263 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2264}
2265EXPORT_SYMBOL(compat_sock_common_getsockopt);
2266#endif
2267
1da177e4
LT		 2268int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2269 struct msghdr *msg, size_t size, int flags)
2270{
2271 struct sock *sk = sock->sk;
2272 int addr_len = 0;
2273 int err;
2274
2275 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2276 flags & ~MSG_DONTWAIT, &addr_len);
2277 if (err >= 0)
2278 msg->msg_namelen = addr_len;
2279 return err;
2280}
1da177e4
LT		 2281EXPORT_SYMBOL(sock_common_recvmsg);
2282
2283/*
2284 * Set socket options on an inet socket.
2285 */
2286int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2287 char __user *optval, unsigned int optlen)
1da177e4
LT		 2288{
2289 struct sock *sk = sock->sk;
2290
2291 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2292}
1da177e4
LT		 2293EXPORT_SYMBOL(sock_common_setsockopt);
2294
3fdadf7d 2295#ifdef CONFIG_COMPAT
543d9cfe 2296int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2297 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2298{
2299 struct sock *sk = sock->sk;
2300
543d9cfe
ACM		 2301 if (sk->sk_prot->compat_setsockopt != NULL)
2302 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2303 optval, optlen);
3fdadf7d
DM		 2304 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2305}
2306EXPORT_SYMBOL(compat_sock_common_setsockopt);
2307#endif
2308
1da177e4
LT		 2309void sk_common_release(struct sock *sk)
2310{
2311 if (sk->sk_prot->destroy)
2312 sk->sk_prot->destroy(sk);
2313
2314 /*
2315 * Observation: when sock_common_release is called, processes have
2316 * no access to socket. But net still has.
2317 * Step one, detach it from networking:
2318 *
2319 * A. Remove from hash tables.
2320 */
2321
2322 sk->sk_prot->unhash(sk);
2323
2324 /*
2325 * In this point socket cannot receive new packets, but it is possible
2326 * that some packets are in flight because some CPU runs receiver and
2327 * did hash table lookup before we unhashed socket. They will achieve
2328 * receive queue and will be purged by socket destructor.
2329 *
2330 * Also we still have packets pending on receive queue and probably,
2331 * our own packets waiting in device queues. sock_destroy will drain
2332 * receive queue, but transmitted packets will delay socket destruction
2333 * until the last reference will be released.
2334 */
2335
2336 sock_orphan(sk);
2337
2338 xfrm_sk_free_policy(sk);
2339
e6848976 2340 sk_refcnt_debug_release(sk);
1da177e4
LT		 2341 sock_put(sk);
2342}
1da177e4
LT		 2343EXPORT_SYMBOL(sk_common_release);
2344
13ff3d6f
PE		 2345#ifdef CONFIG_PROC_FS
2346#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2347struct prot_inuse {
2348 int val[PROTO_INUSE_NR];
2349};
13ff3d6f
PE		 2350
2351static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2352
2353#ifdef CONFIG_NET_NS
2354void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2355{
d6d9ca0f 2356 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2357}
2358EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2359
2360int sock_prot_inuse_get(struct net *net, struct proto *prot)
2361{
2362 int cpu, idx = prot->inuse_idx;
2363 int res = 0;
2364
2365 for_each_possible_cpu(cpu)
2366 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2367
2368 return res >= 0 ? res : 0;
2369}
2370EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2371
2c8c1e72 2372static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2373{
2374 net->core.inuse = alloc_percpu(struct prot_inuse);
2375 return net->core.inuse ? 0 : -ENOMEM;
2376}
2377
2c8c1e72 2378static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2379{
2380 free_percpu(net->core.inuse);
2381}
2382
2383static struct pernet_operations net_inuse_ops = {
2384 .init = sock_inuse_init_net,
2385 .exit = sock_inuse_exit_net,
2386};
2387
2388static __init int net_inuse_init(void)
2389{
2390 if (register_pernet_subsys(&net_inuse_ops))
2391 panic("Cannot initialize net inuse counters");
2392
2393 return 0;
2394}
2395
2396core_initcall(net_inuse_init);
2397#else
1338d466
PE		 2398static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2399
c29a0bc4 2400void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2401{
d6d9ca0f 2402 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2403}
2404EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2405
c29a0bc4 2406int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2407{
2408 int cpu, idx = prot->inuse_idx;
2409 int res = 0;
2410
2411 for_each_possible_cpu(cpu)
2412 res += per_cpu(prot_inuse, cpu).val[idx];
2413
2414 return res >= 0 ? res : 0;
2415}
2416EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2417#endif
13ff3d6f
PE		 2418
2419static void assign_proto_idx(struct proto *prot)
2420{
2421 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2422
2423 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
2424 printk(KERN_ERR "PROTO_INUSE_NR exhausted\n");
2425 return;
2426 }
2427
2428 set_bit(prot->inuse_idx, proto_inuse_idx);
2429}
2430
2431static void release_proto_idx(struct proto *prot)
2432{
2433 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2434 clear_bit(prot->inuse_idx, proto_inuse_idx);
2435}
2436#else
2437static inline void assign_proto_idx(struct proto *prot)
2438{
2439}
2440
2441static inline void release_proto_idx(struct proto *prot)
2442{
2443}
2444#endif
2445
b733c007
PE		 2446int proto_register(struct proto *prot, int alloc_slab)
2447{
1da177e4
LT		 2448 if (alloc_slab) {
2449 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2450 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2451 NULL);
1da177e4
LT		 2452
2453 if (prot->slab == NULL) {
2454 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n",
2455 prot->name);
60e7663d 2456 goto out;
1da177e4 2457 }
2e6599cb
ACM		 2458
2459 if (prot->rsk_prot != NULL) {
faf23422 2460 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2461 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2462 goto out_free_sock_slab;
2463
7e56b5d6 2464 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2465 prot->rsk_prot->obj_size, 0,
20c2df83 2466 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2467
2468 if (prot->rsk_prot->slab == NULL) {
2469 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n",
2470 prot->name);
2471 goto out_free_request_sock_slab_name;
2472 }
2473 }
8feaf0c0 2474
6d6ee43e 2475 if (prot->twsk_prot != NULL) {
faf23422 2476 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2477
7e56b5d6 2478 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2479 goto out_free_request_sock_slab;
2480
6d6ee43e 2481 prot->twsk_prot->twsk_slab =
7e56b5d6 2482 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2483 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2484 0,
2485 SLAB_HWCACHE_ALIGN |
2486 prot->slab_flags,
20c2df83 2487 NULL);
6d6ee43e 2488 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2489 goto out_free_timewait_sock_slab_name;
2490 }
1da177e4
LT		 2491 }
2492
36b77a52 2493 mutex_lock(&proto_list_mutex);
1da177e4 2494 list_add(&prot->node, &proto_list);
13ff3d6f 2495 assign_proto_idx(prot);
36b77a52 2496 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2497 return 0;
2498
8feaf0c0 2499out_free_timewait_sock_slab_name:
7e56b5d6 2500 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2501out_free_request_sock_slab:
2502 if (prot->rsk_prot && prot->rsk_prot->slab) {
2503 kmem_cache_destroy(prot->rsk_prot->slab);
2504 prot->rsk_prot->slab = NULL;
2505 }
2e6599cb 2506out_free_request_sock_slab_name:
72150e9b
DC		 2507 if (prot->rsk_prot)
2508 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2509out_free_sock_slab:
2510 kmem_cache_destroy(prot->slab);
2511 prot->slab = NULL;
b733c007
PE		 2512out:
2513 return -ENOBUFS;
1da177e4 2514}
1da177e4
LT		 2515EXPORT_SYMBOL(proto_register);
2516
2517void proto_unregister(struct proto *prot)
2518{
36b77a52 2519 mutex_lock(&proto_list_mutex);
13ff3d6f 2520 release_proto_idx(prot);
0a3f4358 2521 list_del(&prot->node);
36b77a52 2522 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2523
2524 if (prot->slab != NULL) {
2525 kmem_cache_destroy(prot->slab);
2526 prot->slab = NULL;
2527 }
2528
2e6599cb 2529 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2530 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2531 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2532 prot->rsk_prot->slab = NULL;
2533 }
2534
6d6ee43e 2535 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2536 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2537 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2538 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2539 }
1da177e4 2540}
1da177e4
LT		 2541EXPORT_SYMBOL(proto_unregister);
2542
2543#ifdef CONFIG_PROC_FS
1da177e4 2544static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2545 __acquires(proto_list_mutex)
1da177e4 2546{
36b77a52 2547 mutex_lock(&proto_list_mutex);
60f0438a 2548 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2549}
2550
2551static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2552{
60f0438a 2553 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2554}
2555
2556static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2557 __releases(proto_list_mutex)
1da177e4 2558{
36b77a52 2559 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2560}
2561
2562static char proto_method_implemented(const void *method)
2563{
2564 return method == NULL ? 'n' : 'y';
2565}
180d8cd9
GC		 2566static long sock_prot_memory_allocated(struct proto *proto)
2567{
cb75a36c 2568 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2569}
2570
2571static char *sock_prot_memory_pressure(struct proto *proto)
2572{
2573 return proto->memory_pressure != NULL ?
2574 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2575}
1da177e4
LT		 2576
2577static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2578{
180d8cd9 2579
8d987e5c 2580 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2581 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2582 proto->name,
2583 proto->obj_size,
14e943db 2584 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2585 sock_prot_memory_allocated(proto),
2586 sock_prot_memory_pressure(proto),
1da177e4
LT		 2587 proto->max_header,
2588 proto->slab == NULL ? "no" : "yes",
2589 module_name(proto->owner),
2590 proto_method_implemented(proto->close),
2591 proto_method_implemented(proto->connect),
2592 proto_method_implemented(proto->disconnect),
2593 proto_method_implemented(proto->accept),
2594 proto_method_implemented(proto->ioctl),
2595 proto_method_implemented(proto->init),
2596 proto_method_implemented(proto->destroy),
2597 proto_method_implemented(proto->shutdown),
2598 proto_method_implemented(proto->setsockopt),
2599 proto_method_implemented(proto->getsockopt),
2600 proto_method_implemented(proto->sendmsg),
2601 proto_method_implemented(proto->recvmsg),
2602 proto_method_implemented(proto->sendpage),
2603 proto_method_implemented(proto->bind),
2604 proto_method_implemented(proto->backlog_rcv),
2605 proto_method_implemented(proto->hash),
2606 proto_method_implemented(proto->unhash),
2607 proto_method_implemented(proto->get_port),
2608 proto_method_implemented(proto->enter_memory_pressure));
2609}
2610
2611static int proto_seq_show(struct seq_file *seq, void *v)
2612{
60f0438a 2613 if (v == &proto_list)
1da177e4
LT		 2614 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2615 "protocol",
2616 "size",
2617 "sockets",
2618 "memory",
2619 "press",
2620 "maxhdr",
2621 "slab",
2622 "module",
2623 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2624 else
60f0438a 2625 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2626 return 0;
2627}
2628
f690808e 2629static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2630 .start = proto_seq_start,
2631 .next = proto_seq_next,
2632 .stop = proto_seq_stop,
2633 .show = proto_seq_show,
2634};
2635
2636static int proto_seq_open(struct inode *inode, struct file *file)
2637{
14e943db
ED		 2638 return seq_open_net(inode, file, &proto_seq_ops,
2639 sizeof(struct seq_net_private));
1da177e4
LT		 2640}
2641
9a32144e 2642static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2643 .owner = THIS_MODULE,
2644 .open = proto_seq_open,
2645 .read = seq_read,
2646 .llseek = seq_lseek,
14e943db
ED		 2647 .release = seq_release_net,
2648};
2649
2650static __net_init int proto_init_net(struct net *net)
2651{
2652 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2653 return -ENOMEM;
2654
2655 return 0;
2656}
2657
2658static __net_exit void proto_exit_net(struct net *net)
2659{
2660 proc_net_remove(net, "protocols");
2661}
2662
2663
2664static __net_initdata struct pernet_operations proto_net_ops = {
2665 .init = proto_init_net,
2666 .exit = proto_exit_net,
1da177e4
LT		 2667};
2668
2669static int __init proto_init(void)
2670{
14e943db 2671 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2672}
2673
2674subsys_initcall(proto_init);
2675
2676#endif /* PROC_FS */
kernel source for telekom puls (alcatel/mediatek)