projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
cgroup: make css->refcnt clearing on cgroup removal optional
[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
c377411f
ED		 392 if (sk_rcvqueues_full(sk, skb)) {
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_);
a3a858ff 409 } else if (sk_add_backlog(sk, skb)) {
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:
564 sk->sk_reuse = valbool;
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
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 if (val > sysctl_wmem_max)
585 val = sysctl_wmem_max;
b0573dea 586set_sndbuf:
e71a4783
SH		 587 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
588 if ((val * 2) < SOCK_MIN_SNDBUF)
589 sk->sk_sndbuf = SOCK_MIN_SNDBUF;
590 else
591 sk->sk_sndbuf = val * 2;
1da177e4 592
e71a4783
SH		 593 /*
594 * Wake up sending tasks if we
595 * upped the value.
596 */
597 sk->sk_write_space(sk);
598 break;
1da177e4 599
e71a4783
SH		 600 case SO_SNDBUFFORCE:
601 if (!capable(CAP_NET_ADMIN)) {
602 ret = -EPERM;
603 break;
604 }
605 goto set_sndbuf;
b0573dea 606
e71a4783
SH		 607 case SO_RCVBUF:
608 /* Don't error on this BSD doesn't and if you think
609 about it this is right. Otherwise apps have to
610 play 'guess the biggest size' games. RCVBUF/SNDBUF
611 are treated in BSD as hints */
4ec93edb 612
e71a4783
SH		 613 if (val > sysctl_rmem_max)
614 val = sysctl_rmem_max;
b0573dea 615set_rcvbuf:
e71a4783
SH		 616 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
617 /*
618 * We double it on the way in to account for
619 * "struct sk_buff" etc. overhead. Applications
620 * assume that the SO_RCVBUF setting they make will
621 * allow that much actual data to be received on that
622 * socket.
623 *
624 * Applications are unaware that "struct sk_buff" and
625 * other overheads allocate from the receive buffer
626 * during socket buffer allocation.
627 *
628 * And after considering the possible alternatives,
629 * returning the value we actually used in getsockopt
630 * is the most desirable behavior.
631 */
632 if ((val * 2) < SOCK_MIN_RCVBUF)
633 sk->sk_rcvbuf = SOCK_MIN_RCVBUF;
634 else
635 sk->sk_rcvbuf = val * 2;
636 break;
637
638 case SO_RCVBUFFORCE:
639 if (!capable(CAP_NET_ADMIN)) {
640 ret = -EPERM;
1da177e4 641 break;
e71a4783
SH		 642 }
643 goto set_rcvbuf;
1da177e4 644
e71a4783 645 case SO_KEEPALIVE:
1da177e4 646#ifdef CONFIG_INET
e71a4783
SH		 647 if (sk->sk_protocol == IPPROTO_TCP)
648 tcp_set_keepalive(sk, valbool);
1da177e4 649#endif
e71a4783
SH		 650 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
651 break;
652
653 case SO_OOBINLINE:
654 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
655 break;
656
657 case SO_NO_CHECK:
658 sk->sk_no_check = valbool;
659 break;
660
661 case SO_PRIORITY:
662 if ((val >= 0 && val <= 6) || capable(CAP_NET_ADMIN))
663 sk->sk_priority = val;
664 else
665 ret = -EPERM;
666 break;
667
668 case SO_LINGER:
669 if (optlen < sizeof(ling)) {
670 ret = -EINVAL; /* 1003.1g */
1da177e4 671 break;
e71a4783 672 }
2a91525c 673 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 674 ret = -EFAULT;
1da177e4 675 break;
e71a4783
SH		 676 }
677 if (!ling.l_onoff)
678 sock_reset_flag(sk, SOCK_LINGER);
679 else {
1da177e4 680#if (BITS_PER_LONG == 32)
e71a4783
SH		 681 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
682 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 683 else
e71a4783
SH		 684#endif
685 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
686 sock_set_flag(sk, SOCK_LINGER);
687 }
688 break;
689
690 case SO_BSDCOMPAT:
691 sock_warn_obsolete_bsdism("setsockopt");
692 break;
693
694 case SO_PASSCRED:
695 if (valbool)
696 set_bit(SOCK_PASSCRED, &sock->flags);
697 else
698 clear_bit(SOCK_PASSCRED, &sock->flags);
699 break;
700
701 case SO_TIMESTAMP:
92f37fd2 702 case SO_TIMESTAMPNS:
e71a4783 703 if (valbool) {
92f37fd2
ED		 704 if (optname == SO_TIMESTAMP)
705 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
706 else
707 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 708 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 709 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 710 } else {
e71a4783 711 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 712 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
713 }
e71a4783
SH		 714 break;
715
20d49473
PO		 716 case SO_TIMESTAMPING:
717 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 718 ret = -EINVAL;
20d49473
PO		 719 break;
720 }
721 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
722 val & SOF_TIMESTAMPING_TX_HARDWARE);
723 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
724 val & SOF_TIMESTAMPING_TX_SOFTWARE);
725 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
726 val & SOF_TIMESTAMPING_RX_HARDWARE);
727 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
728 sock_enable_timestamp(sk,
729 SOCK_TIMESTAMPING_RX_SOFTWARE);
730 else
731 sock_disable_timestamp(sk,
08e29af3 732 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 733 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
734 val & SOF_TIMESTAMPING_SOFTWARE);
735 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
736 val & SOF_TIMESTAMPING_SYS_HARDWARE);
737 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
738 val & SOF_TIMESTAMPING_RAW_HARDWARE);
739 break;
740
e71a4783
SH		 741 case SO_RCVLOWAT:
742 if (val < 0)
743 val = INT_MAX;
744 sk->sk_rcvlowat = val ? : 1;
745 break;
746
747 case SO_RCVTIMEO:
748 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
749 break;
750
751 case SO_SNDTIMEO:
752 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
753 break;
1da177e4 754
e71a4783
SH		 755 case SO_ATTACH_FILTER:
756 ret = -EINVAL;
757 if (optlen == sizeof(struct sock_fprog)) {
758 struct sock_fprog fprog;
1da177e4 759
e71a4783
SH		 760 ret = -EFAULT;
761 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 762 break;
e71a4783
SH		 763
764 ret = sk_attach_filter(&fprog, sk);
765 }
766 break;
767
768 case SO_DETACH_FILTER:
55b33325 769 ret = sk_detach_filter(sk);
e71a4783 770 break;
1da177e4 771
e71a4783
SH		 772 case SO_PASSSEC:
773 if (valbool)
774 set_bit(SOCK_PASSSEC, &sock->flags);
775 else
776 clear_bit(SOCK_PASSSEC, &sock->flags);
777 break;
4a19ec58
LAT		 778 case SO_MARK:
779 if (!capable(CAP_NET_ADMIN))
780 ret = -EPERM;
2a91525c 781 else
4a19ec58 782 sk->sk_mark = val;
4a19ec58 783 break;
877ce7c1 784
1da177e4
LT		 785 /* We implement the SO_SNDLOWAT etc to
786 not be settable (1003.1g 5.3) */
3b885787 787 case SO_RXQ_OVFL:
8083f0fc 788 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 789 break;
6e3e939f
JB		 790
791 case SO_WIFI_STATUS:
792 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
793 break;
794
ef64a54f
PE		 795 case SO_PEEK_OFF:
796 if (sock->ops->set_peek_off)
797 sock->ops->set_peek_off(sk, val);
798 else
799 ret = -EOPNOTSUPP;
800 break;
3bdc0eba
BG		 801
802 case SO_NOFCS:
803 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
804 break;
805
e71a4783
SH		 806 default:
807 ret = -ENOPROTOOPT;
808 break;
4ec93edb 809 }
1da177e4
LT		 810 release_sock(sk);
811 return ret;
812}
2a91525c 813EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 814
815
3f551f94
EB		 816void cred_to_ucred(struct pid *pid, const struct cred *cred,
817 struct ucred *ucred)
818{
819 ucred->pid = pid_vnr(pid);
820 ucred->uid = ucred->gid = -1;
821 if (cred) {
822 struct user_namespace *current_ns = current_user_ns();
823
824 ucred->uid = user_ns_map_uid(current_ns, cred, cred->euid);
825 ucred->gid = user_ns_map_gid(current_ns, cred, cred->egid);
826 }
827}
3924773a 828EXPORT_SYMBOL_GPL(cred_to_ucred);
3f551f94 829
1da177e4
LT		 830int sock_getsockopt(struct socket *sock, int level, int optname,
831 char __user *optval, int __user *optlen)
832{
833 struct sock *sk = sock->sk;
4ec93edb 834
e71a4783 835 union {
4ec93edb
YH		 836 int val;
837 struct linger ling;
1da177e4
LT		 838 struct timeval tm;
839 } v;
4ec93edb 840
4d0392be 841 int lv = sizeof(int);
1da177e4 842 int len;
4ec93edb 843
e71a4783 844 if (get_user(len, optlen))
4ec93edb 845 return -EFAULT;
e71a4783 846 if (len < 0)
1da177e4 847 return -EINVAL;
4ec93edb 848
50fee1de 849 memset(&v, 0, sizeof(v));
df0bca04 850
2a91525c 851 switch (optname) {
e71a4783
SH		 852 case SO_DEBUG:
853 v.val = sock_flag(sk, SOCK_DBG);
854 break;
855
856 case SO_DONTROUTE:
857 v.val = sock_flag(sk, SOCK_LOCALROUTE);
858 break;
859
860 case SO_BROADCAST:
861 v.val = !!sock_flag(sk, SOCK_BROADCAST);
862 break;
863
864 case SO_SNDBUF:
865 v.val = sk->sk_sndbuf;
866 break;
867
868 case SO_RCVBUF:
869 v.val = sk->sk_rcvbuf;
870 break;
871
872 case SO_REUSEADDR:
873 v.val = sk->sk_reuse;
874 break;
875
876 case SO_KEEPALIVE:
877 v.val = !!sock_flag(sk, SOCK_KEEPOPEN);
878 break;
879
880 case SO_TYPE:
881 v.val = sk->sk_type;
882 break;
883
49c794e9
JE		 884 case SO_PROTOCOL:
885 v.val = sk->sk_protocol;
886 break;
887
0d6038ee
JE		 888 case SO_DOMAIN:
889 v.val = sk->sk_family;
890 break;
891
e71a4783
SH		 892 case SO_ERROR:
893 v.val = -sock_error(sk);
2a91525c 894 if (v.val == 0)
e71a4783
SH		 895 v.val = xchg(&sk->sk_err_soft, 0);
896 break;
897
898 case SO_OOBINLINE:
899 v.val = !!sock_flag(sk, SOCK_URGINLINE);
900 break;
901
902 case SO_NO_CHECK:
903 v.val = sk->sk_no_check;
904 break;
905
906 case SO_PRIORITY:
907 v.val = sk->sk_priority;
908 break;
909
910 case SO_LINGER:
911 lv = sizeof(v.ling);
912 v.ling.l_onoff = !!sock_flag(sk, SOCK_LINGER);
913 v.ling.l_linger = sk->sk_lingertime / HZ;
914 break;
915
916 case SO_BSDCOMPAT:
917 sock_warn_obsolete_bsdism("getsockopt");
918 break;
919
920 case SO_TIMESTAMP:
92f37fd2
ED		 921 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
922 !sock_flag(sk, SOCK_RCVTSTAMPNS);
923 break;
924
925 case SO_TIMESTAMPNS:
926 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 927 break;
928
20d49473
PO		 929 case SO_TIMESTAMPING:
930 v.val = 0;
931 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
932 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
933 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
934 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
935 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
936 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
937 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
938 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
939 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
940 v.val |= SOF_TIMESTAMPING_SOFTWARE;
941 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
942 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
943 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
944 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
945 break;
946
e71a4783 947 case SO_RCVTIMEO:
2a91525c 948 lv = sizeof(struct timeval);
e71a4783
SH		 949 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
950 v.tm.tv_sec = 0;
951 v.tm.tv_usec = 0;
952 } else {
953 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
954 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
955 }
956 break;
957
958 case SO_SNDTIMEO:
2a91525c 959 lv = sizeof(struct timeval);
e71a4783
SH		 960 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
961 v.tm.tv_sec = 0;
962 v.tm.tv_usec = 0;
963 } else {
964 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
965 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
966 }
967 break;
1da177e4 968
e71a4783
SH		 969 case SO_RCVLOWAT:
970 v.val = sk->sk_rcvlowat;
971 break;
1da177e4 972
e71a4783 973 case SO_SNDLOWAT:
2a91525c 974 v.val = 1;
e71a4783 975 break;
1da177e4 976
e71a4783
SH		 977 case SO_PASSCRED:
978 v.val = test_bit(SOCK_PASSCRED, &sock->flags) ? 1 : 0;
979 break;
1da177e4 980
e71a4783 981 case SO_PEERCRED:
109f6e39
EB		 982 {
983 struct ucred peercred;
984 if (len > sizeof(peercred))
985 len = sizeof(peercred);
986 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
987 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 988 return -EFAULT;
989 goto lenout;
109f6e39 990 }
1da177e4 991
e71a4783
SH		 992 case SO_PEERNAME:
993 {
994 char address[128];
995
996 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
997 return -ENOTCONN;
998 if (lv < len)
999 return -EINVAL;
1000 if (copy_to_user(optval, address, len))
1001 return -EFAULT;
1002 goto lenout;
1003 }
1da177e4 1004
e71a4783
SH		 1005 /* Dubious BSD thing... Probably nobody even uses it, but
1006 * the UNIX standard wants it for whatever reason... -DaveM
1007 */
1008 case SO_ACCEPTCONN:
1009 v.val = sk->sk_state == TCP_LISTEN;
1010 break;
1da177e4 1011
e71a4783
SH		 1012 case SO_PASSSEC:
1013 v.val = test_bit(SOCK_PASSSEC, &sock->flags) ? 1 : 0;
1014 break;
877ce7c1 1015
e71a4783
SH		 1016 case SO_PEERSEC:
1017 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1018
4a19ec58
LAT		 1019 case SO_MARK:
1020 v.val = sk->sk_mark;
1021 break;
1022
3b885787
NH		 1023 case SO_RXQ_OVFL:
1024 v.val = !!sock_flag(sk, SOCK_RXQ_OVFL);
1025 break;
1026
6e3e939f
JB		 1027 case SO_WIFI_STATUS:
1028 v.val = !!sock_flag(sk, SOCK_WIFI_STATUS);
1029 break;
1030
ef64a54f
PE		 1031 case SO_PEEK_OFF:
1032 if (!sock->ops->set_peek_off)
1033 return -EOPNOTSUPP;
1034
1035 v.val = sk->sk_peek_off;
1036 break;
bc2f7996
DM		 1037 case SO_NOFCS:
1038 v.val = !!sock_flag(sk, SOCK_NOFCS);
1039 break;
e71a4783
SH		 1040 default:
1041 return -ENOPROTOOPT;
1da177e4 1042 }
e71a4783 1043
1da177e4
LT		 1044 if (len > lv)
1045 len = lv;
1046 if (copy_to_user(optval, &v, len))
1047 return -EFAULT;
1048lenout:
4ec93edb
YH		 1049 if (put_user(len, optlen))
1050 return -EFAULT;
1051 return 0;
1da177e4
LT		 1052}
1053
a5b5bb9a
IM		 1054/*
1055 * Initialize an sk_lock.
1056 *
1057 * (We also register the sk_lock with the lock validator.)
1058 */
b6f99a21 1059static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1060{
ed07536e
PZ		 1061 sock_lock_init_class_and_name(sk,
1062 af_family_slock_key_strings[sk->sk_family],
1063 af_family_slock_keys + sk->sk_family,
1064 af_family_key_strings[sk->sk_family],
1065 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1066}
1067
4dc6dc71
ED		 1068/*
1069 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1070 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1071 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1072 */
f1a6c4da
PE		 1073static void sock_copy(struct sock *nsk, const struct sock *osk)
1074{
1075#ifdef CONFIG_SECURITY_NETWORK
1076 void *sptr = nsk->sk_security;
1077#endif
68835aba
ED		 1078 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1079
1080 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1081 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1082
f1a6c4da
PE		 1083#ifdef CONFIG_SECURITY_NETWORK
1084 nsk->sk_security = sptr;
1085 security_sk_clone(osk, nsk);
1086#endif
1087}
1088
fcbdf09d
OP		 1089/*
1090 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
1091 * un-modified. Special care is taken when initializing object to zero.
1092 */
1093static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1094{
1095 if (offsetof(struct sock, sk_node.next) != 0)
1096 memset(sk, 0, offsetof(struct sock, sk_node.next));
1097 memset(&sk->sk_node.pprev, 0,
1098 size - offsetof(struct sock, sk_node.pprev));
1099}
1100
1101void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1102{
1103 unsigned long nulls1, nulls2;
1104
1105 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1106 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1107 if (nulls1 > nulls2)
1108 swap(nulls1, nulls2);
1109
1110 if (nulls1 != 0)
1111 memset((char *)sk, 0, nulls1);
1112 memset((char *)sk + nulls1 + sizeof(void *), 0,
1113 nulls2 - nulls1 - sizeof(void *));
1114 memset((char *)sk + nulls2 + sizeof(void *), 0,
1115 size - nulls2 - sizeof(void *));
1116}
1117EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1118
2e4afe7b
PE		 1119static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1120 int family)
c308c1b2
PE		 1121{
1122 struct sock *sk;
1123 struct kmem_cache *slab;
1124
1125 slab = prot->slab;
e912b114
ED		 1126 if (slab != NULL) {
1127 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1128 if (!sk)
1129 return sk;
1130 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1131 if (prot->clear_sk)
1132 prot->clear_sk(sk, prot->obj_size);
1133 else
1134 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1135 }
fcbdf09d 1136 } else
c308c1b2
PE		 1137 sk = kmalloc(prot->obj_size, priority);
1138
2e4afe7b 1139 if (sk != NULL) {
a98b65a3
VN		 1140 kmemcheck_annotate_bitfield(sk, flags);
1141
2e4afe7b
PE		 1142 if (security_sk_alloc(sk, family, priority))
1143 goto out_free;
1144
1145 if (!try_module_get(prot->owner))
1146 goto out_free_sec;
e022f0b4 1147 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1148 }
1149
c308c1b2 1150 return sk;
2e4afe7b
PE		 1151
1152out_free_sec:
1153 security_sk_free(sk);
1154out_free:
1155 if (slab != NULL)
1156 kmem_cache_free(slab, sk);
1157 else
1158 kfree(sk);
1159 return NULL;
c308c1b2
PE		 1160}
1161
1162static void sk_prot_free(struct proto *prot, struct sock *sk)
1163{
1164 struct kmem_cache *slab;
2e4afe7b 1165 struct module *owner;
c308c1b2 1166
2e4afe7b 1167 owner = prot->owner;
c308c1b2 1168 slab = prot->slab;
2e4afe7b
PE		 1169
1170 security_sk_free(sk);
c308c1b2
PE		 1171 if (slab != NULL)
1172 kmem_cache_free(slab, sk);
1173 else
1174 kfree(sk);
2e4afe7b 1175 module_put(owner);
c308c1b2
PE		 1176}
1177
f8451725
HX		 1178#ifdef CONFIG_CGROUPS
1179void sock_update_classid(struct sock *sk)
1180{
1144182a 1181 u32 classid;
f8451725 1182
1144182a
PM		 1183 rcu_read_lock(); /* doing current task, which cannot vanish. */
1184 classid = task_cls_classid(current);
1185 rcu_read_unlock();
f8451725
HX		 1186 if (classid && classid != sk->sk_classid)
1187 sk->sk_classid = classid;
1188}
82862742 1189EXPORT_SYMBOL(sock_update_classid);
5bc1421e
NH		 1190
1191void sock_update_netprioidx(struct sock *sk)
1192{
5bc1421e
NH		 1193 if (in_interrupt())
1194 return;
2b73bc65
NH		 1195
1196 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1197}
1198EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1199#endif
1200
1da177e4
LT		 1201/**
1202 * sk_alloc - All socket objects are allocated here
c4ea43c5 1203 * @net: the applicable net namespace
4dc3b16b
PP		 1204 * @family: protocol family
1205 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1206 * @prot: struct proto associated with this new sock instance
1da177e4 1207 */
1b8d7ae4 1208struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1209 struct proto *prot)
1da177e4 1210{
c308c1b2 1211 struct sock *sk;
1da177e4 1212
154adbc8 1213 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1214 if (sk) {
154adbc8
PE		 1215 sk->sk_family = family;
1216 /*
1217 * See comment in struct sock definition to understand
1218 * why we need sk_prot_creator -acme
1219 */
1220 sk->sk_prot = sk->sk_prot_creator = prot;
1221 sock_lock_init(sk);
3b1e0a65 1222 sock_net_set(sk, get_net(net));
d66ee058 1223 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725
HX		 1224
1225 sock_update_classid(sk);
5bc1421e 1226 sock_update_netprioidx(sk);
1da177e4 1227 }
a79af59e 1228
2e4afe7b 1229 return sk;
1da177e4 1230}
2a91525c 1231EXPORT_SYMBOL(sk_alloc);
1da177e4 1232
2b85a34e 1233static void __sk_free(struct sock *sk)
1da177e4
LT		 1234{
1235 struct sk_filter *filter;
1da177e4
LT		 1236
1237 if (sk->sk_destruct)
1238 sk->sk_destruct(sk);
1239
a898def2
PM		 1240 filter = rcu_dereference_check(sk->sk_filter,
1241 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1242 if (filter) {
309dd5fc 1243 sk_filter_uncharge(sk, filter);
a9b3cd7f 1244 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1245 }
1246
08e29af3 1247 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1248
1249 if (atomic_read(&sk->sk_omem_alloc))
1250 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n",
0dc47877 1251 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1252
109f6e39
EB		 1253 if (sk->sk_peer_cred)
1254 put_cred(sk->sk_peer_cred);
1255 put_pid(sk->sk_peer_pid);
3b1e0a65 1256 put_net(sock_net(sk));
c308c1b2 1257 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1258}
2b85a34e
ED		 1259
1260void sk_free(struct sock *sk)
1261{
1262 /*
25985edc 1263 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1264 * some packets are still in some tx queue.
1265 * If not null, sock_wfree() will call __sk_free(sk) later
1266 */
1267 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1268 __sk_free(sk);
1269}
2a91525c 1270EXPORT_SYMBOL(sk_free);
1da177e4 1271
edf02087 1272/*
25985edc
LDM		 1273 * Last sock_put should drop reference to sk->sk_net. It has already
1274 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1275 * is not an option.
25985edc 1276 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1277 * destroy it in the context of init_net.
1278 */
1279void sk_release_kernel(struct sock *sk)
1280{
1281 if (sk == NULL || sk->sk_socket == NULL)
1282 return;
1283
1284 sock_hold(sk);
1285 sock_release(sk->sk_socket);
65a18ec5 1286 release_net(sock_net(sk));
3b1e0a65 1287 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1288 sock_put(sk);
1289}
45af1754 1290EXPORT_SYMBOL(sk_release_kernel);
edf02087 1291
475f1b52
SR		 1292static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1293{
1294 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1295 sock_update_memcg(newsk);
1296}
1297
e56c57d0
ED		 1298/**
1299 * sk_clone_lock - clone a socket, and lock its clone
1300 * @sk: the socket to clone
1301 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1302 *
1303 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1304 */
1305struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1306{
8fd1d178 1307 struct sock *newsk;
87d11ceb 1308
8fd1d178 1309 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1310 if (newsk != NULL) {
1311 struct sk_filter *filter;
1312
892c141e 1313 sock_copy(newsk, sk);
87d11ceb
ACM		 1314
1315 /* SANITY */
3b1e0a65 1316 get_net(sock_net(newsk));
87d11ceb
ACM		 1317 sk_node_init(&newsk->sk_node);
1318 sock_lock_init(newsk);
1319 bh_lock_sock(newsk);
fa438ccf 1320 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1321 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1322
1323 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1324 /*
1325 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1326 */
1327 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1328 atomic_set(&newsk->sk_omem_alloc, 0);
1329 skb_queue_head_init(&newsk->sk_receive_queue);
1330 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1331#ifdef CONFIG_NET_DMA
1332 skb_queue_head_init(&newsk->sk_async_wait_queue);
1333#endif
87d11ceb 1334
b6c6712a 1335 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1336 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1337 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1338 af_callback_keys + newsk->sk_family,
1339 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1340
1341 newsk->sk_dst_cache = NULL;
1342 newsk->sk_wmem_queued = 0;
1343 newsk->sk_forward_alloc = 0;
1344 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1345 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1346
1347 sock_reset_flag(newsk, SOCK_DONE);
1348 skb_queue_head_init(&newsk->sk_error_queue);
1349
0d7da9dd 1350 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1351 if (filter != NULL)
1352 sk_filter_charge(newsk, filter);
1353
1354 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1355 /* It is still raw copy of parent, so invalidate
1356 * destructor and make plain sk_free() */
1357 newsk->sk_destruct = NULL;
b0691c8e 1358 bh_unlock_sock(newsk);
87d11ceb
ACM		 1359 sk_free(newsk);
1360 newsk = NULL;
1361 goto out;
1362 }
1363
1364 newsk->sk_err = 0;
1365 newsk->sk_priority = 0;
4dc6dc71
ED		 1366 /*
1367 * Before updating sk_refcnt, we must commit prior changes to memory
1368 * (Documentation/RCU/rculist_nulls.txt for details)
1369 */
1370 smp_wmb();
87d11ceb
ACM		 1371 atomic_set(&newsk->sk_refcnt, 2);
1372
1373 /*
1374 * Increment the counter in the same struct proto as the master
1375 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1376 * is the same as sk->sk_prot->socks, as this field was copied
1377 * with memcpy).
1378 *
1379 * This _changes_ the previous behaviour, where
1380 * tcp_create_openreq_child always was incrementing the
1381 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1382 * to be taken into account in all callers. -acme
1383 */
1384 sk_refcnt_debug_inc(newsk);
972692e0 1385 sk_set_socket(newsk, NULL);
43815482 1386 newsk->sk_wq = NULL;
87d11ceb 1387
f3f511e1
GC		 1388 sk_update_clone(sk, newsk);
1389
87d11ceb 1390 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1391 sk_sockets_allocated_inc(newsk);
704da560 1392
08e29af3 1393 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1394 net_enable_timestamp();
87d11ceb
ACM		 1395 }
1396out:
1397 return newsk;
1398}
e56c57d0 1399EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1400
9958089a
AK		 1401void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1402{
1403 __sk_dst_set(sk, dst);
1404 sk->sk_route_caps = dst->dev->features;
1405 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1406 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1407 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1408 if (sk_can_gso(sk)) {
82cc1a7a 1409 if (dst->header_len) {
9958089a 1410 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1411 } else {
9958089a 1412 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a
PWJ		 1413 sk->sk_gso_max_size = dst->dev->gso_max_size;
1414 }
9958089a
AK		 1415 }
1416}
1417EXPORT_SYMBOL_GPL(sk_setup_caps);
1418
1da177e4
LT		 1419void __init sk_init(void)
1420{
4481374c 1421 if (totalram_pages <= 4096) {
1da177e4
LT		 1422 sysctl_wmem_max = 32767;
1423 sysctl_rmem_max = 32767;
1424 sysctl_wmem_default = 32767;
1425 sysctl_rmem_default = 32767;
4481374c 1426 } else if (totalram_pages >= 131072) {
1da177e4
LT		 1427 sysctl_wmem_max = 131071;
1428 sysctl_rmem_max = 131071;
1429 }
1430}
1431
1432/*
1433 * Simple resource managers for sockets.
1434 */
1435
1436
4ec93edb
YH		 1437/*
1438 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1439 */
1440void sock_wfree(struct sk_buff *skb)
1441{
1442 struct sock *sk = skb->sk;
d99927f4 1443 unsigned int len = skb->truesize;
1da177e4 1444
d99927f4
ED		 1445 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1446 /*
1447 * Keep a reference on sk_wmem_alloc, this will be released
1448 * after sk_write_space() call
1449 */
1450 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1451 sk->sk_write_space(sk);
d99927f4
ED		 1452 len = 1;
1453 }
2b85a34e 1454 /*
d99927f4
ED		 1455 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1456 * could not do because of in-flight packets
2b85a34e 1457 */
d99927f4 1458 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1459 __sk_free(sk);
1da177e4 1460}
2a91525c 1461EXPORT_SYMBOL(sock_wfree);
1da177e4 1462
4ec93edb
YH		 1463/*
1464 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1465 */
1466void sock_rfree(struct sk_buff *skb)
1467{
1468 struct sock *sk = skb->sk;
d361fd59 1469 unsigned int len = skb->truesize;
1da177e4 1470
d361fd59
ED		 1471 atomic_sub(len, &sk->sk_rmem_alloc);
1472 sk_mem_uncharge(sk, len);
1da177e4 1473}
2a91525c 1474EXPORT_SYMBOL(sock_rfree);
1da177e4
LT		 1475
1476
1477int sock_i_uid(struct sock *sk)
1478{
1479 int uid;
1480
f064af1e 1481 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1482 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
f064af1e 1483 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1484 return uid;
1485}
2a91525c 1486EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1487
1488unsigned long sock_i_ino(struct sock *sk)
1489{
1490 unsigned long ino;
1491
f064af1e 1492 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1493 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1494 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1495 return ino;
1496}
2a91525c 1497EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1498
1499/*
1500 * Allocate a skb from the socket's send buffer.
1501 */
86a76caf 1502struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1503 gfp_t priority)
1da177e4
LT		 1504{
1505 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1506 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1507 if (skb) {
1508 skb_set_owner_w(skb, sk);
1509 return skb;
1510 }
1511 }
1512 return NULL;
1513}
2a91525c 1514EXPORT_SYMBOL(sock_wmalloc);
1da177e4
LT		 1515
1516/*
1517 * Allocate a skb from the socket's receive buffer.
4ec93edb 1518 */
86a76caf 1519struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1520 gfp_t priority)
1da177e4
LT		 1521{
1522 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1523 struct sk_buff *skb = alloc_skb(size, priority);
1524 if (skb) {
1525 skb_set_owner_r(skb, sk);
1526 return skb;
1527 }
1528 }
1529 return NULL;
1530}
1531
4ec93edb 1532/*
1da177e4 1533 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1534 */
dd0fc66f 1535void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4
LT		 1536{
1537 if ((unsigned)size <= sysctl_optmem_max &&
1538 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1539 void *mem;
1540 /* First do the add, to avoid the race if kmalloc
4ec93edb 1541 * might sleep.
1da177e4
LT		 1542 */
1543 atomic_add(size, &sk->sk_omem_alloc);
1544 mem = kmalloc(size, priority);
1545 if (mem)
1546 return mem;
1547 atomic_sub(size, &sk->sk_omem_alloc);
1548 }
1549 return NULL;
1550}
2a91525c 1551EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1552
1553/*
1554 * Free an option memory block.
1555 */
1556void sock_kfree_s(struct sock *sk, void *mem, int size)
1557{
1558 kfree(mem);
1559 atomic_sub(size, &sk->sk_omem_alloc);
1560}
2a91525c 1561EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1562
1563/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1564 I think, these locks should be removed for datagram sockets.
1565 */
2a91525c 1566static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1567{
1568 DEFINE_WAIT(wait);
1569
1570 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1571 for (;;) {
1572 if (!timeo)
1573 break;
1574 if (signal_pending(current))
1575 break;
1576 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1577 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1578 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1579 break;
1580 if (sk->sk_shutdown & SEND_SHUTDOWN)
1581 break;
1582 if (sk->sk_err)
1583 break;
1584 timeo = schedule_timeout(timeo);
1585 }
aa395145 1586 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1587 return timeo;
1588}
1589
1590
1591/*
1592 * Generic send/receive buffer handlers
1593 */
1594
4cc7f68d
HX		 1595struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1596 unsigned long data_len, int noblock,
1597 int *errcode)
1da177e4
LT		 1598{
1599 struct sk_buff *skb;
7d877f3b 1600 gfp_t gfp_mask;
1da177e4
LT		 1601 long timeo;
1602 int err;
1603
1604 gfp_mask = sk->sk_allocation;
1605 if (gfp_mask & __GFP_WAIT)
1606 gfp_mask |= __GFP_REPEAT;
1607
1608 timeo = sock_sndtimeo(sk, noblock);
1609 while (1) {
1610 err = sock_error(sk);
1611 if (err != 0)
1612 goto failure;
1613
1614 err = -EPIPE;
1615 if (sk->sk_shutdown & SEND_SHUTDOWN)
1616 goto failure;
1617
1618 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1619 skb = alloc_skb(header_len, gfp_mask);
1da177e4
LT		 1620 if (skb) {
1621 int npages;
1622 int i;
1623
1624 /* No pages, we're done... */
1625 if (!data_len)
1626 break;
1627
1628 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1629 skb->truesize += data_len;
1630 skb_shinfo(skb)->nr_frags = npages;
1631 for (i = 0; i < npages; i++) {
1632 struct page *page;
1da177e4
LT		 1633
1634 page = alloc_pages(sk->sk_allocation, 0);
1635 if (!page) {
1636 err = -ENOBUFS;
1637 skb_shinfo(skb)->nr_frags = i;
1638 kfree_skb(skb);
1639 goto failure;
1640 }
1641
ea2ab693
IC		 1642 __skb_fill_page_desc(skb, i,
1643 page, 0,
1644 (data_len >= PAGE_SIZE ?
1645 PAGE_SIZE :
1646 data_len));
1da177e4
LT		 1647 data_len -= PAGE_SIZE;
1648 }
1649
1650 /* Full success... */
1651 break;
1652 }
1653 err = -ENOBUFS;
1654 goto failure;
1655 }
1656 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1657 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1658 err = -EAGAIN;
1659 if (!timeo)
1660 goto failure;
1661 if (signal_pending(current))
1662 goto interrupted;
1663 timeo = sock_wait_for_wmem(sk, timeo);
1664 }
1665
1666 skb_set_owner_w(skb, sk);
1667 return skb;
1668
1669interrupted:
1670 err = sock_intr_errno(timeo);
1671failure:
1672 *errcode = err;
1673 return NULL;
1674}
4cc7f68d 1675EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1676
4ec93edb 1677struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1678 int noblock, int *errcode)
1679{
1680 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1681}
2a91525c 1682EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4
LT		 1683
1684static void __lock_sock(struct sock *sk)
f39234d6
NK		 1685 __releases(&sk->sk_lock.slock)
1686 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1687{
1688 DEFINE_WAIT(wait);
1689
e71a4783 1690 for (;;) {
1da177e4
LT		 1691 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1692 TASK_UNINTERRUPTIBLE);
1693 spin_unlock_bh(&sk->sk_lock.slock);
1694 schedule();
1695 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1696 if (!sock_owned_by_user(sk))
1da177e4
LT		 1697 break;
1698 }
1699 finish_wait(&sk->sk_lock.wq, &wait);
1700}
1701
1702static void __release_sock(struct sock *sk)
f39234d6
NK		 1703 __releases(&sk->sk_lock.slock)
1704 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1705{
1706 struct sk_buff *skb = sk->sk_backlog.head;
1707
1708 do {
1709 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1710 bh_unlock_sock(sk);
1711
1712 do {
1713 struct sk_buff *next = skb->next;
1714
7fee226a 1715 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1716 skb->next = NULL;
c57943a1 1717 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1718
1719 /*
1720 * We are in process context here with softirqs
1721 * disabled, use cond_resched_softirq() to preempt.
1722 * This is safe to do because we've taken the backlog
1723 * queue private:
1724 */
1725 cond_resched_softirq();
1726
1727 skb = next;
1728 } while (skb != NULL);
1729
1730 bh_lock_sock(sk);
e71a4783 1731 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1732
1733 /*
1734 * Doing the zeroing here guarantee we can not loop forever
1735 * while a wild producer attempts to flood us.
1736 */
1737 sk->sk_backlog.len = 0;
1da177e4
LT		 1738}
1739
1740/**
1741 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1742 * @sk: sock to wait on
1743 * @timeo: for how long
1da177e4
LT		 1744 *
1745 * Now socket state including sk->sk_err is changed only under lock,
1746 * hence we may omit checks after joining wait queue.
1747 * We check receive queue before schedule() only as optimization;
1748 * it is very likely that release_sock() added new data.
1749 */
1750int sk_wait_data(struct sock *sk, long *timeo)
1751{
1752 int rc;
1753 DEFINE_WAIT(wait);
1754
aa395145 1755 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1756 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1757 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1758 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1759 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1760 return rc;
1761}
1da177e4
LT		 1762EXPORT_SYMBOL(sk_wait_data);
1763
3ab224be
HA		 1764/**
1765 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1766 * @sk: socket
1767 * @size: memory size to allocate
1768 * @kind: allocation type
1769 *
1770 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1771 * rmem allocation. This function assumes that protocols which have
1772 * memory_pressure use sk_wmem_queued as write buffer accounting.
1773 */
1774int __sk_mem_schedule(struct sock *sk, int size, int kind)
1775{
1776 struct proto *prot = sk->sk_prot;
1777 int amt = sk_mem_pages(size);
8d987e5c 1778 long allocated;
e1aab161 1779 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1780
1781 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1782
e1aab161 1783 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1784
1785 /* Under limit. */
e1aab161
GC		 1786 if (parent_status == UNDER_LIMIT &&
1787 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1788 sk_leave_memory_pressure(sk);
3ab224be
HA		 1789 return 1;
1790 }
1791
e1aab161
GC		 1792 /* Under pressure. (we or our parents) */
1793 if ((parent_status > SOFT_LIMIT) ||
1794 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1795 sk_enter_memory_pressure(sk);
3ab224be 1796
e1aab161
GC		 1797 /* Over hard limit (we or our parents) */
1798 if ((parent_status == OVER_LIMIT) ||
1799 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1800 goto suppress_allocation;
1801
1802 /* guarantee minimum buffer size under pressure */
1803 if (kind == SK_MEM_RECV) {
1804 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1805 return 1;
180d8cd9 1806
3ab224be
HA		 1807 } else { /* SK_MEM_SEND */
1808 if (sk->sk_type == SOCK_STREAM) {
1809 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1810 return 1;
1811 } else if (atomic_read(&sk->sk_wmem_alloc) <
1812 prot->sysctl_wmem[0])
1813 return 1;
1814 }
1815
180d8cd9 1816 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 1817 int alloc;
1818
180d8cd9 1819 if (!sk_under_memory_pressure(sk))
1748376b 1820 return 1;
180d8cd9
GC		 1821 alloc = sk_sockets_allocated_read_positive(sk);
1822 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 1823 sk_mem_pages(sk->sk_wmem_queued +
1824 atomic_read(&sk->sk_rmem_alloc) +
1825 sk->sk_forward_alloc))
1826 return 1;
1827 }
1828
1829suppress_allocation:
1830
1831 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1832 sk_stream_moderate_sndbuf(sk);
1833
1834 /* Fail only if socket is _under_ its sndbuf.
1835 * In this case we cannot block, so that we have to fail.
1836 */
1837 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1838 return 1;
1839 }
1840
3847ce32
SM		 1841 trace_sock_exceed_buf_limit(sk, prot, allocated);
1842
3ab224be
HA		 1843 /* Alas. Undo changes. */
1844 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 1845
0e90b31f 1846 sk_memory_allocated_sub(sk, amt);
180d8cd9 1847
3ab224be
HA		 1848 return 0;
1849}
3ab224be
HA		 1850EXPORT_SYMBOL(__sk_mem_schedule);
1851
1852/**
1853 * __sk_reclaim - reclaim memory_allocated
1854 * @sk: socket
1855 */
1856void __sk_mem_reclaim(struct sock *sk)
1857{
180d8cd9 1858 sk_memory_allocated_sub(sk,
0e90b31f 1859 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 1860 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
1861
180d8cd9
GC		 1862 if (sk_under_memory_pressure(sk) &&
1863 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
1864 sk_leave_memory_pressure(sk);
3ab224be 1865}
3ab224be
HA		 1866EXPORT_SYMBOL(__sk_mem_reclaim);
1867
1868
1da177e4
LT		 1869/*
1870 * Set of default routines for initialising struct proto_ops when
1871 * the protocol does not support a particular function. In certain
1872 * cases where it makes no sense for a protocol to have a "do nothing"
1873 * function, some default processing is provided.
1874 */
1875
1876int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1877{
1878 return -EOPNOTSUPP;
1879}
2a91525c 1880EXPORT_SYMBOL(sock_no_bind);
1da177e4 1881
4ec93edb 1882int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1883 int len, int flags)
1884{
1885 return -EOPNOTSUPP;
1886}
2a91525c 1887EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 1888
1889int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1890{
1891 return -EOPNOTSUPP;
1892}
2a91525c 1893EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 1894
1895int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1896{
1897 return -EOPNOTSUPP;
1898}
2a91525c 1899EXPORT_SYMBOL(sock_no_accept);
1da177e4 1900
4ec93edb 1901int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1902 int *len, int peer)
1903{
1904 return -EOPNOTSUPP;
1905}
2a91525c 1906EXPORT_SYMBOL(sock_no_getname);
1da177e4 1907
2a91525c 1908unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 1909{
1910 return 0;
1911}
2a91525c 1912EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 1913
1914int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1915{
1916 return -EOPNOTSUPP;
1917}
2a91525c 1918EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 1919
1920int sock_no_listen(struct socket *sock, int backlog)
1921{
1922 return -EOPNOTSUPP;
1923}
2a91525c 1924EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 1925
1926int sock_no_shutdown(struct socket *sock, int how)
1927{
1928 return -EOPNOTSUPP;
1929}
2a91525c 1930EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 1931
1932int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 1933 char __user *optval, unsigned int optlen)
1da177e4
LT		 1934{
1935 return -EOPNOTSUPP;
1936}
2a91525c 1937EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 1938
1939int sock_no_getsockopt(struct socket *sock, int level, int optname,
1940 char __user *optval, int __user *optlen)
1941{
1942 return -EOPNOTSUPP;
1943}
2a91525c 1944EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 1945
1946int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1947 size_t len)
1948{
1949 return -EOPNOTSUPP;
1950}
2a91525c 1951EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 1952
1953int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1954 size_t len, int flags)
1955{
1956 return -EOPNOTSUPP;
1957}
2a91525c 1958EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 1959
1960int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
1961{
1962 /* Mirror missing mmap method error code */
1963 return -ENODEV;
1964}
2a91525c 1965EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 1966
1967ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
1968{
1969 ssize_t res;
1970 struct msghdr msg = {.msg_flags = flags};
1971 struct kvec iov;
1972 char *kaddr = kmap(page);
1973 iov.iov_base = kaddr + offset;
1974 iov.iov_len = size;
1975 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
1976 kunmap(page);
1977 return res;
1978}
2a91525c 1979EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 1980
1981/*
1982 * Default Socket Callbacks
1983 */
1984
1985static void sock_def_wakeup(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_all(&wq->wait);
1993 rcu_read_unlock();
1da177e4
LT		 1994}
1995
1996static void sock_def_error_report(struct sock *sk)
1997{
43815482
ED		 1998 struct socket_wq *wq;
1999
2000 rcu_read_lock();
2001 wq = rcu_dereference(sk->sk_wq);
2002 if (wq_has_sleeper(wq))
2003 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2004 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2005 rcu_read_unlock();
1da177e4
LT		 2006}
2007
2008static void sock_def_readable(struct sock *sk, int len)
2009{
43815482
ED		 2010 struct socket_wq *wq;
2011
2012 rcu_read_lock();
2013 wq = rcu_dereference(sk->sk_wq);
2014 if (wq_has_sleeper(wq))
2c6607c6 2015 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2016 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2017 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2018 rcu_read_unlock();
1da177e4
LT		 2019}
2020
2021static void sock_def_write_space(struct sock *sk)
2022{
43815482
ED		 2023 struct socket_wq *wq;
2024
2025 rcu_read_lock();
1da177e4
LT		 2026
2027 /* Do not wake up a writer until he can make "significant"
2028 * progress. --DaveM
2029 */
e71a4783 2030 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2031 wq = rcu_dereference(sk->sk_wq);
2032 if (wq_has_sleeper(wq))
2033 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2034 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2035
2036 /* Should agree with poll, otherwise some programs break */
2037 if (sock_writeable(sk))
8d8ad9d7 2038 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2039 }
2040
43815482 2041 rcu_read_unlock();
1da177e4
LT		 2042}
2043
2044static void sock_def_destruct(struct sock *sk)
2045{
a51482bd 2046 kfree(sk->sk_protinfo);
1da177e4
LT		 2047}
2048
2049void sk_send_sigurg(struct sock *sk)
2050{
2051 if (sk->sk_socket && sk->sk_socket->file)
2052 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2053 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2054}
2a91525c 2055EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2056
2057void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2058 unsigned long expires)
2059{
2060 if (!mod_timer(timer, expires))
2061 sock_hold(sk);
2062}
1da177e4
LT		 2063EXPORT_SYMBOL(sk_reset_timer);
2064
2065void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2066{
2067 if (timer_pending(timer) && del_timer(timer))
2068 __sock_put(sk);
2069}
1da177e4
LT		 2070EXPORT_SYMBOL(sk_stop_timer);
2071
2072void sock_init_data(struct socket *sock, struct sock *sk)
2073{
2074 skb_queue_head_init(&sk->sk_receive_queue);
2075 skb_queue_head_init(&sk->sk_write_queue);
2076 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2077#ifdef CONFIG_NET_DMA
2078 skb_queue_head_init(&sk->sk_async_wait_queue);
2079#endif
1da177e4
LT		 2080
2081 sk->sk_send_head = NULL;
2082
2083 init_timer(&sk->sk_timer);
4ec93edb 2084
1da177e4
LT		 2085 sk->sk_allocation = GFP_KERNEL;
2086 sk->sk_rcvbuf = sysctl_rmem_default;
2087 sk->sk_sndbuf = sysctl_wmem_default;
2088 sk->sk_state = TCP_CLOSE;
972692e0 2089 sk_set_socket(sk, sock);
1da177e4
LT		 2090
2091 sock_set_flag(sk, SOCK_ZAPPED);
2092
e71a4783 2093 if (sock) {
1da177e4 2094 sk->sk_type = sock->type;
43815482 2095 sk->sk_wq = sock->wq;
1da177e4
LT		 2096 sock->sk = sk;
2097 } else
43815482 2098 sk->sk_wq = NULL;
1da177e4 2099
b6c6712a 2100 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2101 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2102 lockdep_set_class_and_name(&sk->sk_callback_lock,
2103 af_callback_keys + sk->sk_family,
2104 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2105
2106 sk->sk_state_change = sock_def_wakeup;
2107 sk->sk_data_ready = sock_def_readable;
2108 sk->sk_write_space = sock_def_write_space;
2109 sk->sk_error_report = sock_def_error_report;
2110 sk->sk_destruct = sock_def_destruct;
2111
2112 sk->sk_sndmsg_page = NULL;
2113 sk->sk_sndmsg_off = 0;
ef64a54f 2114 sk->sk_peek_off = -1;
1da177e4 2115
109f6e39
EB		 2116 sk->sk_peer_pid = NULL;
2117 sk->sk_peer_cred = NULL;
1da177e4
LT		 2118 sk->sk_write_pending = 0;
2119 sk->sk_rcvlowat = 1;
2120 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2121 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2122
f37f0afb 2123 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2124
4dc6dc71
ED		 2125 /*
2126 * Before updating sk_refcnt, we must commit prior changes to memory
2127 * (Documentation/RCU/rculist_nulls.txt for details)
2128 */
2129 smp_wmb();
1da177e4 2130 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2131 atomic_set(&sk->sk_drops, 0);
1da177e4 2132}
2a91525c 2133EXPORT_SYMBOL(sock_init_data);
1da177e4 2134
b5606c2d 2135void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2136{
2137 might_sleep();
a5b5bb9a 2138 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2139 if (sk->sk_lock.owned)
1da177e4 2140 __lock_sock(sk);
d2e9117c 2141 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2142 spin_unlock(&sk->sk_lock.slock);
2143 /*
2144 * The sk_lock has mutex_lock() semantics here:
2145 */
fcc70d5f 2146 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2147 local_bh_enable();
1da177e4 2148}
fcc70d5f 2149EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2150
b5606c2d 2151void release_sock(struct sock *sk)
1da177e4 2152{
a5b5bb9a
IM		 2153 /*
2154 * The sk_lock has mutex_unlock() semantics:
2155 */
2156 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2157
2158 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2159 if (sk->sk_backlog.tail)
2160 __release_sock(sk);
d2e9117c 2161 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2162 if (waitqueue_active(&sk->sk_lock.wq))
2163 wake_up(&sk->sk_lock.wq);
2164 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2165}
2166EXPORT_SYMBOL(release_sock);
2167
8a74ad60
ED		 2168/**
2169 * lock_sock_fast - fast version of lock_sock
2170 * @sk: socket
2171 *
2172 * This version should be used for very small section, where process wont block
2173 * return false if fast path is taken
2174 * sk_lock.slock locked, owned = 0, BH disabled
2175 * return true if slow path is taken
2176 * sk_lock.slock unlocked, owned = 1, BH enabled
2177 */
2178bool lock_sock_fast(struct sock *sk)
2179{
2180 might_sleep();
2181 spin_lock_bh(&sk->sk_lock.slock);
2182
2183 if (!sk->sk_lock.owned)
2184 /*
2185 * Note : We must disable BH
2186 */
2187 return false;
2188
2189 __lock_sock(sk);
2190 sk->sk_lock.owned = 1;
2191 spin_unlock(&sk->sk_lock.slock);
2192 /*
2193 * The sk_lock has mutex_lock() semantics here:
2194 */
2195 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2196 local_bh_enable();
2197 return true;
2198}
2199EXPORT_SYMBOL(lock_sock_fast);
2200
1da177e4 2201int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2202{
b7aa0bf7 2203 struct timeval tv;
1da177e4 2204 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2205 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2206 tv = ktime_to_timeval(sk->sk_stamp);
2207 if (tv.tv_sec == -1)
1da177e4 2208 return -ENOENT;
b7aa0bf7
ED		 2209 if (tv.tv_sec == 0) {
2210 sk->sk_stamp = ktime_get_real();
2211 tv = ktime_to_timeval(sk->sk_stamp);
2212 }
2213 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2214}
1da177e4
LT		 2215EXPORT_SYMBOL(sock_get_timestamp);
2216
ae40eb1e
ED		 2217int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2218{
2219 struct timespec ts;
2220 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2221 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2222 ts = ktime_to_timespec(sk->sk_stamp);
2223 if (ts.tv_sec == -1)
2224 return -ENOENT;
2225 if (ts.tv_sec == 0) {
2226 sk->sk_stamp = ktime_get_real();
2227 ts = ktime_to_timespec(sk->sk_stamp);
2228 }
2229 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2230}
2231EXPORT_SYMBOL(sock_get_timestampns);
2232
20d49473 2233void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2234{
20d49473 2235 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2236 unsigned long previous_flags = sk->sk_flags;
2237
20d49473
PO		 2238 sock_set_flag(sk, flag);
2239 /*
2240 * we just set one of the two flags which require net
2241 * time stamping, but time stamping might have been on
2242 * already because of the other one
2243 */
08e29af3 2244 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2245 net_enable_timestamp();
1da177e4
LT		 2246 }
2247}
1da177e4
LT		 2248
2249/*
2250 * Get a socket option on an socket.
2251 *
2252 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2253 * asynchronous errors should be reported by getsockopt. We assume
2254 * this means if you specify SO_ERROR (otherwise whats the point of it).
2255 */
2256int sock_common_getsockopt(struct socket *sock, int level, int optname,
2257 char __user *optval, int __user *optlen)
2258{
2259 struct sock *sk = sock->sk;
2260
2261 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2262}
1da177e4
LT		 2263EXPORT_SYMBOL(sock_common_getsockopt);
2264
3fdadf7d 2265#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2266int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2267 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2268{
2269 struct sock *sk = sock->sk;
2270
1e51f951 2271 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2272 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2273 optval, optlen);
3fdadf7d
DM		 2274 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2275}
2276EXPORT_SYMBOL(compat_sock_common_getsockopt);
2277#endif
2278
1da177e4
LT		 2279int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2280 struct msghdr *msg, size_t size, int flags)
2281{
2282 struct sock *sk = sock->sk;
2283 int addr_len = 0;
2284 int err;
2285
2286 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2287 flags & ~MSG_DONTWAIT, &addr_len);
2288 if (err >= 0)
2289 msg->msg_namelen = addr_len;
2290 return err;
2291}
1da177e4
LT		 2292EXPORT_SYMBOL(sock_common_recvmsg);
2293
2294/*
2295 * Set socket options on an inet socket.
2296 */
2297int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2298 char __user *optval, unsigned int optlen)
1da177e4
LT		 2299{
2300 struct sock *sk = sock->sk;
2301
2302 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2303}
1da177e4
LT		 2304EXPORT_SYMBOL(sock_common_setsockopt);
2305
3fdadf7d 2306#ifdef CONFIG_COMPAT
543d9cfe 2307int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2308 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2309{
2310 struct sock *sk = sock->sk;
2311
543d9cfe
ACM		 2312 if (sk->sk_prot->compat_setsockopt != NULL)
2313 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2314 optval, optlen);
3fdadf7d
DM		 2315 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2316}
2317EXPORT_SYMBOL(compat_sock_common_setsockopt);
2318#endif
2319
1da177e4
LT		 2320void sk_common_release(struct sock *sk)
2321{
2322 if (sk->sk_prot->destroy)
2323 sk->sk_prot->destroy(sk);
2324
2325 /*
2326 * Observation: when sock_common_release is called, processes have
2327 * no access to socket. But net still has.
2328 * Step one, detach it from networking:
2329 *
2330 * A. Remove from hash tables.
2331 */
2332
2333 sk->sk_prot->unhash(sk);
2334
2335 /*
2336 * In this point socket cannot receive new packets, but it is possible
2337 * that some packets are in flight because some CPU runs receiver and
2338 * did hash table lookup before we unhashed socket. They will achieve
2339 * receive queue and will be purged by socket destructor.
2340 *
2341 * Also we still have packets pending on receive queue and probably,
2342 * our own packets waiting in device queues. sock_destroy will drain
2343 * receive queue, but transmitted packets will delay socket destruction
2344 * until the last reference will be released.
2345 */
2346
2347 sock_orphan(sk);
2348
2349 xfrm_sk_free_policy(sk);
2350
e6848976 2351 sk_refcnt_debug_release(sk);
1da177e4
LT		 2352 sock_put(sk);
2353}
1da177e4
LT		 2354EXPORT_SYMBOL(sk_common_release);
2355
13ff3d6f
PE		 2356#ifdef CONFIG_PROC_FS
2357#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2358struct prot_inuse {
2359 int val[PROTO_INUSE_NR];
2360};
13ff3d6f
PE		 2361
2362static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2363
2364#ifdef CONFIG_NET_NS
2365void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2366{
d6d9ca0f 2367 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2368}
2369EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2370
2371int sock_prot_inuse_get(struct net *net, struct proto *prot)
2372{
2373 int cpu, idx = prot->inuse_idx;
2374 int res = 0;
2375
2376 for_each_possible_cpu(cpu)
2377 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2378
2379 return res >= 0 ? res : 0;
2380}
2381EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2382
2c8c1e72 2383static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2384{
2385 net->core.inuse = alloc_percpu(struct prot_inuse);
2386 return net->core.inuse ? 0 : -ENOMEM;
2387}
2388
2c8c1e72 2389static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2390{
2391 free_percpu(net->core.inuse);
2392}
2393
2394static struct pernet_operations net_inuse_ops = {
2395 .init = sock_inuse_init_net,
2396 .exit = sock_inuse_exit_net,
2397};
2398
2399static __init int net_inuse_init(void)
2400{
2401 if (register_pernet_subsys(&net_inuse_ops))
2402 panic("Cannot initialize net inuse counters");
2403
2404 return 0;
2405}
2406
2407core_initcall(net_inuse_init);
2408#else
1338d466
PE		 2409static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2410
c29a0bc4 2411void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2412{
d6d9ca0f 2413 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2414}
2415EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2416
c29a0bc4 2417int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2418{
2419 int cpu, idx = prot->inuse_idx;
2420 int res = 0;
2421
2422 for_each_possible_cpu(cpu)
2423 res += per_cpu(prot_inuse, cpu).val[idx];
2424
2425 return res >= 0 ? res : 0;
2426}
2427EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2428#endif
13ff3d6f
PE		 2429
2430static void assign_proto_idx(struct proto *prot)
2431{
2432 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2433
2434 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
2435 printk(KERN_ERR "PROTO_INUSE_NR exhausted\n");
2436 return;
2437 }
2438
2439 set_bit(prot->inuse_idx, proto_inuse_idx);
2440}
2441
2442static void release_proto_idx(struct proto *prot)
2443{
2444 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2445 clear_bit(prot->inuse_idx, proto_inuse_idx);
2446}
2447#else
2448static inline void assign_proto_idx(struct proto *prot)
2449{
2450}
2451
2452static inline void release_proto_idx(struct proto *prot)
2453{
2454}
2455#endif
2456
b733c007
PE		 2457int proto_register(struct proto *prot, int alloc_slab)
2458{
1da177e4
LT		 2459 if (alloc_slab) {
2460 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2461 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2462 NULL);
1da177e4
LT		 2463
2464 if (prot->slab == NULL) {
2465 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n",
2466 prot->name);
60e7663d 2467 goto out;
1da177e4 2468 }
2e6599cb
ACM		 2469
2470 if (prot->rsk_prot != NULL) {
faf23422 2471 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2472 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2473 goto out_free_sock_slab;
2474
7e56b5d6 2475 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2476 prot->rsk_prot->obj_size, 0,
20c2df83 2477 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2478
2479 if (prot->rsk_prot->slab == NULL) {
2480 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n",
2481 prot->name);
2482 goto out_free_request_sock_slab_name;
2483 }
2484 }
8feaf0c0 2485
6d6ee43e 2486 if (prot->twsk_prot != NULL) {
faf23422 2487 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2488
7e56b5d6 2489 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2490 goto out_free_request_sock_slab;
2491
6d6ee43e 2492 prot->twsk_prot->twsk_slab =
7e56b5d6 2493 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2494 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2495 0,
2496 SLAB_HWCACHE_ALIGN |
2497 prot->slab_flags,
20c2df83 2498 NULL);
6d6ee43e 2499 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2500 goto out_free_timewait_sock_slab_name;
2501 }
1da177e4
LT		 2502 }
2503
36b77a52 2504 mutex_lock(&proto_list_mutex);
1da177e4 2505 list_add(&prot->node, &proto_list);
13ff3d6f 2506 assign_proto_idx(prot);
36b77a52 2507 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2508 return 0;
2509
8feaf0c0 2510out_free_timewait_sock_slab_name:
7e56b5d6 2511 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2512out_free_request_sock_slab:
2513 if (prot->rsk_prot && prot->rsk_prot->slab) {
2514 kmem_cache_destroy(prot->rsk_prot->slab);
2515 prot->rsk_prot->slab = NULL;
2516 }
2e6599cb 2517out_free_request_sock_slab_name:
72150e9b
DC		 2518 if (prot->rsk_prot)
2519 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2520out_free_sock_slab:
2521 kmem_cache_destroy(prot->slab);
2522 prot->slab = NULL;
b733c007
PE		 2523out:
2524 return -ENOBUFS;
1da177e4 2525}
1da177e4
LT		 2526EXPORT_SYMBOL(proto_register);
2527
2528void proto_unregister(struct proto *prot)
2529{
36b77a52 2530 mutex_lock(&proto_list_mutex);
13ff3d6f 2531 release_proto_idx(prot);
0a3f4358 2532 list_del(&prot->node);
36b77a52 2533 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2534
2535 if (prot->slab != NULL) {
2536 kmem_cache_destroy(prot->slab);
2537 prot->slab = NULL;
2538 }
2539
2e6599cb 2540 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2541 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2542 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2543 prot->rsk_prot->slab = NULL;
2544 }
2545
6d6ee43e 2546 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2547 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2548 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2549 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2550 }
1da177e4 2551}
1da177e4
LT		 2552EXPORT_SYMBOL(proto_unregister);
2553
2554#ifdef CONFIG_PROC_FS
1da177e4 2555static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2556 __acquires(proto_list_mutex)
1da177e4 2557{
36b77a52 2558 mutex_lock(&proto_list_mutex);
60f0438a 2559 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2560}
2561
2562static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2563{
60f0438a 2564 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2565}
2566
2567static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2568 __releases(proto_list_mutex)
1da177e4 2569{
36b77a52 2570 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2571}
2572
2573static char proto_method_implemented(const void *method)
2574{
2575 return method == NULL ? 'n' : 'y';
2576}
180d8cd9
GC		 2577static long sock_prot_memory_allocated(struct proto *proto)
2578{
2579 return proto->memory_allocated != NULL ? proto_memory_allocated(proto): -1L;
2580}
2581
2582static char *sock_prot_memory_pressure(struct proto *proto)
2583{
2584 return proto->memory_pressure != NULL ?
2585 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2586}
1da177e4
LT		 2587
2588static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2589{
180d8cd9 2590
8d987e5c 2591 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2592 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2593 proto->name,
2594 proto->obj_size,
14e943db 2595 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2596 sock_prot_memory_allocated(proto),
2597 sock_prot_memory_pressure(proto),
1da177e4
LT		 2598 proto->max_header,
2599 proto->slab == NULL ? "no" : "yes",
2600 module_name(proto->owner),
2601 proto_method_implemented(proto->close),
2602 proto_method_implemented(proto->connect),
2603 proto_method_implemented(proto->disconnect),
2604 proto_method_implemented(proto->accept),
2605 proto_method_implemented(proto->ioctl),
2606 proto_method_implemented(proto->init),
2607 proto_method_implemented(proto->destroy),
2608 proto_method_implemented(proto->shutdown),
2609 proto_method_implemented(proto->setsockopt),
2610 proto_method_implemented(proto->getsockopt),
2611 proto_method_implemented(proto->sendmsg),
2612 proto_method_implemented(proto->recvmsg),
2613 proto_method_implemented(proto->sendpage),
2614 proto_method_implemented(proto->bind),
2615 proto_method_implemented(proto->backlog_rcv),
2616 proto_method_implemented(proto->hash),
2617 proto_method_implemented(proto->unhash),
2618 proto_method_implemented(proto->get_port),
2619 proto_method_implemented(proto->enter_memory_pressure));
2620}
2621
2622static int proto_seq_show(struct seq_file *seq, void *v)
2623{
60f0438a 2624 if (v == &proto_list)
1da177e4
LT		 2625 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2626 "protocol",
2627 "size",
2628 "sockets",
2629 "memory",
2630 "press",
2631 "maxhdr",
2632 "slab",
2633 "module",
2634 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2635 else
60f0438a 2636 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2637 return 0;
2638}
2639
f690808e 2640static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2641 .start = proto_seq_start,
2642 .next = proto_seq_next,
2643 .stop = proto_seq_stop,
2644 .show = proto_seq_show,
2645};
2646
2647static int proto_seq_open(struct inode *inode, struct file *file)
2648{
14e943db
ED		 2649 return seq_open_net(inode, file, &proto_seq_ops,
2650 sizeof(struct seq_net_private));
1da177e4
LT		 2651}
2652
9a32144e 2653static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2654 .owner = THIS_MODULE,
2655 .open = proto_seq_open,
2656 .read = seq_read,
2657 .llseek = seq_lseek,
14e943db
ED		 2658 .release = seq_release_net,
2659};
2660
2661static __net_init int proto_init_net(struct net *net)
2662{
2663 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2664 return -ENOMEM;
2665
2666 return 0;
2667}
2668
2669static __net_exit void proto_exit_net(struct net *net)
2670{
2671 proc_net_remove(net, "protocols");
2672}
2673
2674
2675static __net_initdata struct pernet_operations proto_net_ops = {
2676 .init = proto_init_net,
2677 .exit = proto_exit_net,
1da177e4
LT		 2678};
2679
2680static int __init proto_init(void)
2681{
14e943db 2682 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2683}
2684
2685subsys_initcall(proto_init);
2686
2687#endif /* PROC_FS */
kernel source for telekom puls (alcatel/mediatek)