projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
static keys: Introduce 'struct static_key', static_key_true()/false() and static_key_...
[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>
118#include <asm/system.h>
119
120#include <linux/netdevice.h>
121#include <net/protocol.h>
122#include <linux/skbuff.h>
457c4cbc 123#include <net/net_namespace.h>
2e6599cb 124#include <net/request_sock.h>
1da177e4 125#include <net/sock.h>
20d49473 126#include <linux/net_tstamp.h>
1da177e4
LT		 127#include <net/xfrm.h>
128#include <linux/ipsec.h>
f8451725 129#include <net/cls_cgroup.h>
5bc1421e 130#include <net/netprio_cgroup.h>
1da177e4
LT		 131
132#include <linux/filter.h>
133
3847ce32
SM		 134#include <trace/events/sock.h>
135
1da177e4
LT		 136#ifdef CONFIG_INET
137#include <net/tcp.h>
138#endif
139
36b77a52 140static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 141static LIST_HEAD(proto_list);
142
143#ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
144int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
145{
146 struct proto *proto;
147 int ret = 0;
148
36b77a52 149 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 150 list_for_each_entry(proto, &proto_list, node) {
151 if (proto->init_cgroup) {
152 ret = proto->init_cgroup(cgrp, ss);
153 if (ret)
154 goto out;
155 }
156 }
157
36b77a52 158 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 159 return ret;
160out:
161 list_for_each_entry_continue_reverse(proto, &proto_list, node)
162 if (proto->destroy_cgroup)
163 proto->destroy_cgroup(cgrp, ss);
36b77a52 164 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 165 return ret;
166}
167
168void mem_cgroup_sockets_destroy(struct cgroup *cgrp, struct cgroup_subsys *ss)
169{
170 struct proto *proto;
171
36b77a52 172 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 173 list_for_each_entry_reverse(proto, &proto_list, node)
174 if (proto->destroy_cgroup)
175 proto->destroy_cgroup(cgrp, ss);
36b77a52 176 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 177}
178#endif
179
da21f24d
IM		 180/*
181 * Each address family might have different locking rules, so we have
182 * one slock key per address family:
183 */
a5b5bb9a
IM		 184static struct lock_class_key af_family_keys[AF_MAX];
185static struct lock_class_key af_family_slock_keys[AF_MAX];
186
c5905afb 187struct static_key memcg_socket_limit_enabled;
e1aab161
GC		 188EXPORT_SYMBOL(memcg_socket_limit_enabled);
189
a5b5bb9a
IM		 190/*
191 * Make lock validator output more readable. (we pre-construct these
192 * strings build-time, so that runtime initialization of socket
193 * locks is fast):
194 */
36cbd3dc 195static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 196 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
197 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
198 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
199 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
200 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
201 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
202 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 203 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 204 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 205 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 206 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 207 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 208 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
c7fe3b52 209 "sk_lock-AF_NFC" , "sk_lock-AF_MAX"
a5b5bb9a 210};
36cbd3dc 211static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 212 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
213 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
214 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
215 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
216 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
217 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
218 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 219 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 220 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 221 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 222 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 223 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 224 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
c7fe3b52 225 "slock-AF_NFC" , "slock-AF_MAX"
a5b5bb9a 226};
36cbd3dc 227static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 228 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
229 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
230 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
231 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
232 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
233 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
234 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 235 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 236 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 237 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 238 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 239 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 240 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
c7fe3b52 241 "clock-AF_NFC" , "clock-AF_MAX"
443aef0e 242};
da21f24d
IM		 243
244/*
245 * sk_callback_lock locking rules are per-address-family,
246 * so split the lock classes by using a per-AF key:
247 */
248static struct lock_class_key af_callback_keys[AF_MAX];
249
1da177e4
LT		 250/* Take into consideration the size of the struct sk_buff overhead in the
251 * determination of these values, since that is non-constant across
252 * platforms. This makes socket queueing behavior and performance
253 * not depend upon such differences.
254 */
255#define _SK_MEM_PACKETS 256
87fb4b7b 256#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 257#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
258#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
259
260/* Run time adjustable parameters. */
ab32ea5d
BH		 261__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
262__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
263__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
264__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 265
25985edc 266/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 267int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 268EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 269
5bc1421e
NH		 270#if defined(CONFIG_CGROUPS)
271#if !defined(CONFIG_NET_CLS_CGROUP)
f8451725
HX		 272int net_cls_subsys_id = -1;
273EXPORT_SYMBOL_GPL(net_cls_subsys_id);
274#endif
5bc1421e
NH		 275#if !defined(CONFIG_NETPRIO_CGROUP)
276int net_prio_subsys_id = -1;
277EXPORT_SYMBOL_GPL(net_prio_subsys_id);
278#endif
279#endif
f8451725 280
1da177e4
LT		 281static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
282{
283 struct timeval tv;
284
285 if (optlen < sizeof(tv))
286 return -EINVAL;
287 if (copy_from_user(&tv, optval, sizeof(tv)))
288 return -EFAULT;
ba78073e
VA		 289 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
290 return -EDOM;
1da177e4 291
ba78073e 292 if (tv.tv_sec < 0) {
6f11df83
AM		 293 static int warned __read_mostly;
294
ba78073e 295 *timeo_p = 0;
50aab54f 296 if (warned < 10 && net_ratelimit()) {
ba78073e
VA		 297 warned++;
298 printk(KERN_INFO "sock_set_timeout: `%s' (pid %d) "
299 "tries to set negative timeout\n",
ba25f9dc 300 current->comm, task_pid_nr(current));
50aab54f 301 }
ba78073e
VA		 302 return 0;
303 }
1da177e4
LT		 304 *timeo_p = MAX_SCHEDULE_TIMEOUT;
305 if (tv.tv_sec == 0 && tv.tv_usec == 0)
306 return 0;
307 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
308 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
309 return 0;
310}
311
312static void sock_warn_obsolete_bsdism(const char *name)
313{
314 static int warned;
315 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 316 if (strcmp(warncomm, current->comm) && warned < 5) {
317 strcpy(warncomm, current->comm);
1da177e4
LT		 318 printk(KERN_WARNING "process `%s' is using obsolete "
319 "%s SO_BSDCOMPAT\n", warncomm, name);
320 warned++;
321 }
322}
323
08e29af3
ED		 324#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
325
326static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 327{
08e29af3
ED		 328 if (sk->sk_flags & flags) {
329 sk->sk_flags &= ~flags;
330 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 331 net_disable_timestamp();
1da177e4
LT		 332 }
333}
334
335
f0088a50
DV		 336int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
337{
766e9037 338 int err;
f0088a50 339 int skb_len;
3b885787
NH		 340 unsigned long flags;
341 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 342
0fd7bac6 343 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 344 atomic_inc(&sk->sk_drops);
3847ce32 345 trace_sock_rcvqueue_full(sk, skb);
766e9037 346 return -ENOMEM;
f0088a50
DV		 347 }
348
fda9ef5d 349 err = sk_filter(sk, skb);
f0088a50 350 if (err)
766e9037 351 return err;
f0088a50 352
3ab224be 353 if (!sk_rmem_schedule(sk, skb->truesize)) {
766e9037
ED		 354 atomic_inc(&sk->sk_drops);
355 return -ENOBUFS;
3ab224be
HA		 356 }
357
f0088a50
DV		 358 skb->dev = NULL;
359 skb_set_owner_r(skb, sk);
49ad9599 360
f0088a50
DV		 361 /* Cache the SKB length before we tack it onto the receive
362 * queue. Once it is added it no longer belongs to us and
363 * may be freed by other threads of control pulling packets
364 * from the queue.
365 */
366 skb_len = skb->len;
367
7fee226a
ED		 368 /* we escape from rcu protected region, make sure we dont leak
369 * a norefcounted dst
370 */
371 skb_dst_force(skb);
372
3b885787
NH		 373 spin_lock_irqsave(&list->lock, flags);
374 skb->dropcount = atomic_read(&sk->sk_drops);
375 __skb_queue_tail(list, skb);
376 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 377
378 if (!sock_flag(sk, SOCK_DEAD))
379 sk->sk_data_ready(sk, skb_len);
766e9037 380 return 0;
f0088a50
DV		 381}
382EXPORT_SYMBOL(sock_queue_rcv_skb);
383
58a5a7b9 384int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 385{
386 int rc = NET_RX_SUCCESS;
387
fda9ef5d 388 if (sk_filter(sk, skb))
f0088a50
DV		 389 goto discard_and_relse;
390
391 skb->dev = NULL;
392
c377411f
ED		 393 if (sk_rcvqueues_full(sk, skb)) {
394 atomic_inc(&sk->sk_drops);
395 goto discard_and_relse;
396 }
58a5a7b9
ACM		 397 if (nested)
398 bh_lock_sock_nested(sk);
399 else
400 bh_lock_sock(sk);
a5b5bb9a
IM		 401 if (!sock_owned_by_user(sk)) {
402 /*
403 * trylock + unlock semantics:
404 */
405 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
406
c57943a1 407 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 408
409 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
a3a858ff 410 } else if (sk_add_backlog(sk, skb)) {
8eae939f
ZY		 411 bh_unlock_sock(sk);
412 atomic_inc(&sk->sk_drops);
413 goto discard_and_relse;
414 }
415
f0088a50
DV		 416 bh_unlock_sock(sk);
417out:
418 sock_put(sk);
419 return rc;
420discard_and_relse:
421 kfree_skb(skb);
422 goto out;
423}
424EXPORT_SYMBOL(sk_receive_skb);
425
ea94ff3b
KK		 426void sk_reset_txq(struct sock *sk)
427{
428 sk_tx_queue_clear(sk);
429}
430EXPORT_SYMBOL(sk_reset_txq);
431
f0088a50
DV		 432struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
433{
b6c6712a 434 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 435
436 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 437 sk_tx_queue_clear(sk);
a9b3cd7f 438 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 439 dst_release(dst);
440 return NULL;
441 }
442
443 return dst;
444}
445EXPORT_SYMBOL(__sk_dst_check);
446
447struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
448{
449 struct dst_entry *dst = sk_dst_get(sk);
450
451 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
452 sk_dst_reset(sk);
453 dst_release(dst);
454 return NULL;
455 }
456
457 return dst;
458}
459EXPORT_SYMBOL(sk_dst_check);
460
4878809f
DM		 461static int sock_bindtodevice(struct sock *sk, char __user *optval, int optlen)
462{
463 int ret = -ENOPROTOOPT;
464#ifdef CONFIG_NETDEVICES
3b1e0a65 465 struct net *net = sock_net(sk);
4878809f
DM		 466 char devname[IFNAMSIZ];
467 int index;
468
469 /* Sorry... */
470 ret = -EPERM;
471 if (!capable(CAP_NET_RAW))
472 goto out;
473
474 ret = -EINVAL;
475 if (optlen < 0)
476 goto out;
477
478 /* Bind this socket to a particular device like "eth0",
479 * as specified in the passed interface name. If the
480 * name is "" or the option length is zero the socket
481 * is not bound.
482 */
483 if (optlen > IFNAMSIZ - 1)
484 optlen = IFNAMSIZ - 1;
485 memset(devname, 0, sizeof(devname));
486
487 ret = -EFAULT;
488 if (copy_from_user(devname, optval, optlen))
489 goto out;
490
000ba2e4
DM		 491 index = 0;
492 if (devname[0] != '\0') {
bf8e56bf 493 struct net_device *dev;
4878809f 494
bf8e56bf
ED		 495 rcu_read_lock();
496 dev = dev_get_by_name_rcu(net, devname);
497 if (dev)
498 index = dev->ifindex;
499 rcu_read_unlock();
4878809f
DM		 500 ret = -ENODEV;
501 if (!dev)
502 goto out;
4878809f
DM		 503 }
504
505 lock_sock(sk);
506 sk->sk_bound_dev_if = index;
507 sk_dst_reset(sk);
508 release_sock(sk);
509
510 ret = 0;
511
512out:
513#endif
514
515 return ret;
516}
517
c0ef877b
PE		 518static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
519{
520 if (valbool)
521 sock_set_flag(sk, bit);
522 else
523 sock_reset_flag(sk, bit);
524}
525
1da177e4
LT		 526/*
527 * This is meant for all protocols to use and covers goings on
528 * at the socket level. Everything here is generic.
529 */
530
531int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 532 char __user *optval, unsigned int optlen)
1da177e4 533{
2a91525c 534 struct sock *sk = sock->sk;
1da177e4
LT		 535 int val;
536 int valbool;
537 struct linger ling;
538 int ret = 0;
4ec93edb 539
1da177e4
LT		 540 /*
541 * Options without arguments
542 */
543
4878809f
DM		 544 if (optname == SO_BINDTODEVICE)
545 return sock_bindtodevice(sk, optval, optlen);
546
e71a4783
SH		 547 if (optlen < sizeof(int))
548 return -EINVAL;
4ec93edb 549
1da177e4
LT		 550 if (get_user(val, (int __user *)optval))
551 return -EFAULT;
4ec93edb 552
2a91525c 553 valbool = val ? 1 : 0;
1da177e4
LT		 554
555 lock_sock(sk);
556
2a91525c 557 switch (optname) {
e71a4783 558 case SO_DEBUG:
2a91525c 559 if (val && !capable(CAP_NET_ADMIN))
e71a4783 560 ret = -EACCES;
2a91525c 561 else
c0ef877b 562 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 563 break;
564 case SO_REUSEADDR:
565 sk->sk_reuse = valbool;
566 break;
567 case SO_TYPE:
49c794e9 568 case SO_PROTOCOL:
0d6038ee 569 case SO_DOMAIN:
e71a4783
SH		 570 case SO_ERROR:
571 ret = -ENOPROTOOPT;
572 break;
573 case SO_DONTROUTE:
c0ef877b 574 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 575 break;
576 case SO_BROADCAST:
577 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
578 break;
579 case SO_SNDBUF:
580 /* Don't error on this BSD doesn't and if you think
581 about it this is right. Otherwise apps have to
582 play 'guess the biggest size' games. RCVBUF/SNDBUF
583 are treated in BSD as hints */
584
585 if (val > sysctl_wmem_max)
586 val = sysctl_wmem_max;
b0573dea 587set_sndbuf:
e71a4783
SH		 588 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
589 if ((val * 2) < SOCK_MIN_SNDBUF)
590 sk->sk_sndbuf = SOCK_MIN_SNDBUF;
591 else
592 sk->sk_sndbuf = val * 2;
1da177e4 593
e71a4783
SH		 594 /*
595 * Wake up sending tasks if we
596 * upped the value.
597 */
598 sk->sk_write_space(sk);
599 break;
1da177e4 600
e71a4783
SH		 601 case SO_SNDBUFFORCE:
602 if (!capable(CAP_NET_ADMIN)) {
603 ret = -EPERM;
604 break;
605 }
606 goto set_sndbuf;
b0573dea 607
e71a4783
SH		 608 case SO_RCVBUF:
609 /* Don't error on this BSD doesn't and if you think
610 about it this is right. Otherwise apps have to
611 play 'guess the biggest size' games. RCVBUF/SNDBUF
612 are treated in BSD as hints */
4ec93edb 613
e71a4783
SH		 614 if (val > sysctl_rmem_max)
615 val = sysctl_rmem_max;
b0573dea 616set_rcvbuf:
e71a4783
SH		 617 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
618 /*
619 * We double it on the way in to account for
620 * "struct sk_buff" etc. overhead. Applications
621 * assume that the SO_RCVBUF setting they make will
622 * allow that much actual data to be received on that
623 * socket.
624 *
625 * Applications are unaware that "struct sk_buff" and
626 * other overheads allocate from the receive buffer
627 * during socket buffer allocation.
628 *
629 * And after considering the possible alternatives,
630 * returning the value we actually used in getsockopt
631 * is the most desirable behavior.
632 */
633 if ((val * 2) < SOCK_MIN_RCVBUF)
634 sk->sk_rcvbuf = SOCK_MIN_RCVBUF;
635 else
636 sk->sk_rcvbuf = val * 2;
637 break;
638
639 case SO_RCVBUFFORCE:
640 if (!capable(CAP_NET_ADMIN)) {
641 ret = -EPERM;
1da177e4 642 break;
e71a4783
SH		 643 }
644 goto set_rcvbuf;
1da177e4 645
e71a4783 646 case SO_KEEPALIVE:
1da177e4 647#ifdef CONFIG_INET
e71a4783
SH		 648 if (sk->sk_protocol == IPPROTO_TCP)
649 tcp_set_keepalive(sk, valbool);
1da177e4 650#endif
e71a4783
SH		 651 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
652 break;
653
654 case SO_OOBINLINE:
655 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
656 break;
657
658 case SO_NO_CHECK:
659 sk->sk_no_check = valbool;
660 break;
661
662 case SO_PRIORITY:
663 if ((val >= 0 && val <= 6) || capable(CAP_NET_ADMIN))
664 sk->sk_priority = val;
665 else
666 ret = -EPERM;
667 break;
668
669 case SO_LINGER:
670 if (optlen < sizeof(ling)) {
671 ret = -EINVAL; /* 1003.1g */
1da177e4 672 break;
e71a4783 673 }
2a91525c 674 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 675 ret = -EFAULT;
1da177e4 676 break;
e71a4783
SH		 677 }
678 if (!ling.l_onoff)
679 sock_reset_flag(sk, SOCK_LINGER);
680 else {
1da177e4 681#if (BITS_PER_LONG == 32)
e71a4783
SH		 682 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
683 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 684 else
e71a4783
SH		 685#endif
686 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
687 sock_set_flag(sk, SOCK_LINGER);
688 }
689 break;
690
691 case SO_BSDCOMPAT:
692 sock_warn_obsolete_bsdism("setsockopt");
693 break;
694
695 case SO_PASSCRED:
696 if (valbool)
697 set_bit(SOCK_PASSCRED, &sock->flags);
698 else
699 clear_bit(SOCK_PASSCRED, &sock->flags);
700 break;
701
702 case SO_TIMESTAMP:
92f37fd2 703 case SO_TIMESTAMPNS:
e71a4783 704 if (valbool) {
92f37fd2
ED		 705 if (optname == SO_TIMESTAMP)
706 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
707 else
708 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 709 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 710 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 711 } else {
e71a4783 712 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 713 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
714 }
e71a4783
SH		 715 break;
716
20d49473
PO		 717 case SO_TIMESTAMPING:
718 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 719 ret = -EINVAL;
20d49473
PO		 720 break;
721 }
722 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
723 val & SOF_TIMESTAMPING_TX_HARDWARE);
724 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
725 val & SOF_TIMESTAMPING_TX_SOFTWARE);
726 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
727 val & SOF_TIMESTAMPING_RX_HARDWARE);
728 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
729 sock_enable_timestamp(sk,
730 SOCK_TIMESTAMPING_RX_SOFTWARE);
731 else
732 sock_disable_timestamp(sk,
08e29af3 733 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 734 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
735 val & SOF_TIMESTAMPING_SOFTWARE);
736 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
737 val & SOF_TIMESTAMPING_SYS_HARDWARE);
738 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
739 val & SOF_TIMESTAMPING_RAW_HARDWARE);
740 break;
741
e71a4783
SH		 742 case SO_RCVLOWAT:
743 if (val < 0)
744 val = INT_MAX;
745 sk->sk_rcvlowat = val ? : 1;
746 break;
747
748 case SO_RCVTIMEO:
749 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
750 break;
751
752 case SO_SNDTIMEO:
753 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
754 break;
1da177e4 755
e71a4783
SH		 756 case SO_ATTACH_FILTER:
757 ret = -EINVAL;
758 if (optlen == sizeof(struct sock_fprog)) {
759 struct sock_fprog fprog;
1da177e4 760
e71a4783
SH		 761 ret = -EFAULT;
762 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 763 break;
e71a4783
SH		 764
765 ret = sk_attach_filter(&fprog, sk);
766 }
767 break;
768
769 case SO_DETACH_FILTER:
55b33325 770 ret = sk_detach_filter(sk);
e71a4783 771 break;
1da177e4 772
e71a4783
SH		 773 case SO_PASSSEC:
774 if (valbool)
775 set_bit(SOCK_PASSSEC, &sock->flags);
776 else
777 clear_bit(SOCK_PASSSEC, &sock->flags);
778 break;
4a19ec58
LAT		 779 case SO_MARK:
780 if (!capable(CAP_NET_ADMIN))
781 ret = -EPERM;
2a91525c 782 else
4a19ec58 783 sk->sk_mark = val;
4a19ec58 784 break;
877ce7c1 785
1da177e4
LT		 786 /* We implement the SO_SNDLOWAT etc to
787 not be settable (1003.1g 5.3) */
3b885787 788 case SO_RXQ_OVFL:
8083f0fc 789 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 790 break;
6e3e939f
JB		 791
792 case SO_WIFI_STATUS:
793 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
794 break;
795
e71a4783
SH		 796 default:
797 ret = -ENOPROTOOPT;
798 break;
4ec93edb 799 }
1da177e4
LT		 800 release_sock(sk);
801 return ret;
802}
2a91525c 803EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 804
805
3f551f94
EB		 806void cred_to_ucred(struct pid *pid, const struct cred *cred,
807 struct ucred *ucred)
808{
809 ucred->pid = pid_vnr(pid);
810 ucred->uid = ucred->gid = -1;
811 if (cred) {
812 struct user_namespace *current_ns = current_user_ns();
813
814 ucred->uid = user_ns_map_uid(current_ns, cred, cred->euid);
815 ucred->gid = user_ns_map_gid(current_ns, cred, cred->egid);
816 }
817}
3924773a 818EXPORT_SYMBOL_GPL(cred_to_ucred);
3f551f94 819
1da177e4
LT		 820int sock_getsockopt(struct socket *sock, int level, int optname,
821 char __user *optval, int __user *optlen)
822{
823 struct sock *sk = sock->sk;
4ec93edb 824
e71a4783 825 union {
4ec93edb
YH		 826 int val;
827 struct linger ling;
1da177e4
LT		 828 struct timeval tm;
829 } v;
4ec93edb 830
4d0392be 831 int lv = sizeof(int);
1da177e4 832 int len;
4ec93edb 833
e71a4783 834 if (get_user(len, optlen))
4ec93edb 835 return -EFAULT;
e71a4783 836 if (len < 0)
1da177e4 837 return -EINVAL;
4ec93edb 838
50fee1de 839 memset(&v, 0, sizeof(v));
df0bca04 840
2a91525c 841 switch (optname) {
e71a4783
SH		 842 case SO_DEBUG:
843 v.val = sock_flag(sk, SOCK_DBG);
844 break;
845
846 case SO_DONTROUTE:
847 v.val = sock_flag(sk, SOCK_LOCALROUTE);
848 break;
849
850 case SO_BROADCAST:
851 v.val = !!sock_flag(sk, SOCK_BROADCAST);
852 break;
853
854 case SO_SNDBUF:
855 v.val = sk->sk_sndbuf;
856 break;
857
858 case SO_RCVBUF:
859 v.val = sk->sk_rcvbuf;
860 break;
861
862 case SO_REUSEADDR:
863 v.val = sk->sk_reuse;
864 break;
865
866 case SO_KEEPALIVE:
867 v.val = !!sock_flag(sk, SOCK_KEEPOPEN);
868 break;
869
870 case SO_TYPE:
871 v.val = sk->sk_type;
872 break;
873
49c794e9
JE		 874 case SO_PROTOCOL:
875 v.val = sk->sk_protocol;
876 break;
877
0d6038ee
JE		 878 case SO_DOMAIN:
879 v.val = sk->sk_family;
880 break;
881
e71a4783
SH		 882 case SO_ERROR:
883 v.val = -sock_error(sk);
2a91525c 884 if (v.val == 0)
e71a4783
SH		 885 v.val = xchg(&sk->sk_err_soft, 0);
886 break;
887
888 case SO_OOBINLINE:
889 v.val = !!sock_flag(sk, SOCK_URGINLINE);
890 break;
891
892 case SO_NO_CHECK:
893 v.val = sk->sk_no_check;
894 break;
895
896 case SO_PRIORITY:
897 v.val = sk->sk_priority;
898 break;
899
900 case SO_LINGER:
901 lv = sizeof(v.ling);
902 v.ling.l_onoff = !!sock_flag(sk, SOCK_LINGER);
903 v.ling.l_linger = sk->sk_lingertime / HZ;
904 break;
905
906 case SO_BSDCOMPAT:
907 sock_warn_obsolete_bsdism("getsockopt");
908 break;
909
910 case SO_TIMESTAMP:
92f37fd2
ED		 911 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
912 !sock_flag(sk, SOCK_RCVTSTAMPNS);
913 break;
914
915 case SO_TIMESTAMPNS:
916 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 917 break;
918
20d49473
PO		 919 case SO_TIMESTAMPING:
920 v.val = 0;
921 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
922 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
923 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
924 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
925 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
926 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
927 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
928 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
929 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
930 v.val |= SOF_TIMESTAMPING_SOFTWARE;
931 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
932 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
933 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
934 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
935 break;
936
e71a4783 937 case SO_RCVTIMEO:
2a91525c 938 lv = sizeof(struct timeval);
e71a4783
SH		 939 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
940 v.tm.tv_sec = 0;
941 v.tm.tv_usec = 0;
942 } else {
943 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
944 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
945 }
946 break;
947
948 case SO_SNDTIMEO:
2a91525c 949 lv = sizeof(struct timeval);
e71a4783
SH		 950 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
951 v.tm.tv_sec = 0;
952 v.tm.tv_usec = 0;
953 } else {
954 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
955 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
956 }
957 break;
1da177e4 958
e71a4783
SH		 959 case SO_RCVLOWAT:
960 v.val = sk->sk_rcvlowat;
961 break;
1da177e4 962
e71a4783 963 case SO_SNDLOWAT:
2a91525c 964 v.val = 1;
e71a4783 965 break;
1da177e4 966
e71a4783
SH		 967 case SO_PASSCRED:
968 v.val = test_bit(SOCK_PASSCRED, &sock->flags) ? 1 : 0;
969 break;
1da177e4 970
e71a4783 971 case SO_PEERCRED:
109f6e39
EB		 972 {
973 struct ucred peercred;
974 if (len > sizeof(peercred))
975 len = sizeof(peercred);
976 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
977 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 978 return -EFAULT;
979 goto lenout;
109f6e39 980 }
1da177e4 981
e71a4783
SH		 982 case SO_PEERNAME:
983 {
984 char address[128];
985
986 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
987 return -ENOTCONN;
988 if (lv < len)
989 return -EINVAL;
990 if (copy_to_user(optval, address, len))
991 return -EFAULT;
992 goto lenout;
993 }
1da177e4 994
e71a4783
SH		 995 /* Dubious BSD thing... Probably nobody even uses it, but
996 * the UNIX standard wants it for whatever reason... -DaveM
997 */
998 case SO_ACCEPTCONN:
999 v.val = sk->sk_state == TCP_LISTEN;
1000 break;
1da177e4 1001
e71a4783
SH		 1002 case SO_PASSSEC:
1003 v.val = test_bit(SOCK_PASSSEC, &sock->flags) ? 1 : 0;
1004 break;
877ce7c1 1005
e71a4783
SH		 1006 case SO_PEERSEC:
1007 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1008
4a19ec58
LAT		 1009 case SO_MARK:
1010 v.val = sk->sk_mark;
1011 break;
1012
3b885787
NH		 1013 case SO_RXQ_OVFL:
1014 v.val = !!sock_flag(sk, SOCK_RXQ_OVFL);
1015 break;
1016
6e3e939f
JB		 1017 case SO_WIFI_STATUS:
1018 v.val = !!sock_flag(sk, SOCK_WIFI_STATUS);
1019 break;
1020
e71a4783
SH		 1021 default:
1022 return -ENOPROTOOPT;
1da177e4 1023 }
e71a4783 1024
1da177e4
LT		 1025 if (len > lv)
1026 len = lv;
1027 if (copy_to_user(optval, &v, len))
1028 return -EFAULT;
1029lenout:
4ec93edb
YH		 1030 if (put_user(len, optlen))
1031 return -EFAULT;
1032 return 0;
1da177e4
LT		 1033}
1034
a5b5bb9a
IM		 1035/*
1036 * Initialize an sk_lock.
1037 *
1038 * (We also register the sk_lock with the lock validator.)
1039 */
b6f99a21 1040static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1041{
ed07536e
PZ		 1042 sock_lock_init_class_and_name(sk,
1043 af_family_slock_key_strings[sk->sk_family],
1044 af_family_slock_keys + sk->sk_family,
1045 af_family_key_strings[sk->sk_family],
1046 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1047}
1048
4dc6dc71
ED		 1049/*
1050 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1051 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1052 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1053 */
f1a6c4da
PE		 1054static void sock_copy(struct sock *nsk, const struct sock *osk)
1055{
1056#ifdef CONFIG_SECURITY_NETWORK
1057 void *sptr = nsk->sk_security;
1058#endif
68835aba
ED		 1059 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1060
1061 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1062 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1063
f1a6c4da
PE		 1064#ifdef CONFIG_SECURITY_NETWORK
1065 nsk->sk_security = sptr;
1066 security_sk_clone(osk, nsk);
1067#endif
1068}
1069
fcbdf09d
OP		 1070/*
1071 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
1072 * un-modified. Special care is taken when initializing object to zero.
1073 */
1074static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1075{
1076 if (offsetof(struct sock, sk_node.next) != 0)
1077 memset(sk, 0, offsetof(struct sock, sk_node.next));
1078 memset(&sk->sk_node.pprev, 0,
1079 size - offsetof(struct sock, sk_node.pprev));
1080}
1081
1082void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1083{
1084 unsigned long nulls1, nulls2;
1085
1086 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1087 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1088 if (nulls1 > nulls2)
1089 swap(nulls1, nulls2);
1090
1091 if (nulls1 != 0)
1092 memset((char *)sk, 0, nulls1);
1093 memset((char *)sk + nulls1 + sizeof(void *), 0,
1094 nulls2 - nulls1 - sizeof(void *));
1095 memset((char *)sk + nulls2 + sizeof(void *), 0,
1096 size - nulls2 - sizeof(void *));
1097}
1098EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1099
2e4afe7b
PE		 1100static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1101 int family)
c308c1b2
PE		 1102{
1103 struct sock *sk;
1104 struct kmem_cache *slab;
1105
1106 slab = prot->slab;
e912b114
ED		 1107 if (slab != NULL) {
1108 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1109 if (!sk)
1110 return sk;
1111 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1112 if (prot->clear_sk)
1113 prot->clear_sk(sk, prot->obj_size);
1114 else
1115 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1116 }
fcbdf09d 1117 } else
c308c1b2
PE		 1118 sk = kmalloc(prot->obj_size, priority);
1119
2e4afe7b 1120 if (sk != NULL) {
a98b65a3
VN		 1121 kmemcheck_annotate_bitfield(sk, flags);
1122
2e4afe7b
PE		 1123 if (security_sk_alloc(sk, family, priority))
1124 goto out_free;
1125
1126 if (!try_module_get(prot->owner))
1127 goto out_free_sec;
e022f0b4 1128 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1129 }
1130
c308c1b2 1131 return sk;
2e4afe7b
PE		 1132
1133out_free_sec:
1134 security_sk_free(sk);
1135out_free:
1136 if (slab != NULL)
1137 kmem_cache_free(slab, sk);
1138 else
1139 kfree(sk);
1140 return NULL;
c308c1b2
PE		 1141}
1142
1143static void sk_prot_free(struct proto *prot, struct sock *sk)
1144{
1145 struct kmem_cache *slab;
2e4afe7b 1146 struct module *owner;
c308c1b2 1147
2e4afe7b 1148 owner = prot->owner;
c308c1b2 1149 slab = prot->slab;
2e4afe7b
PE		 1150
1151 security_sk_free(sk);
c308c1b2
PE		 1152 if (slab != NULL)
1153 kmem_cache_free(slab, sk);
1154 else
1155 kfree(sk);
2e4afe7b 1156 module_put(owner);
c308c1b2
PE		 1157}
1158
f8451725
HX		 1159#ifdef CONFIG_CGROUPS
1160void sock_update_classid(struct sock *sk)
1161{
1144182a 1162 u32 classid;
f8451725 1163
1144182a
PM		 1164 rcu_read_lock(); /* doing current task, which cannot vanish. */
1165 classid = task_cls_classid(current);
1166 rcu_read_unlock();
f8451725
HX		 1167 if (classid && classid != sk->sk_classid)
1168 sk->sk_classid = classid;
1169}
82862742 1170EXPORT_SYMBOL(sock_update_classid);
5bc1421e
NH		 1171
1172void sock_update_netprioidx(struct sock *sk)
1173{
1174 struct cgroup_netprio_state *state;
1175 if (in_interrupt())
1176 return;
1177 rcu_read_lock();
1178 state = task_netprio_state(current);
1179 sk->sk_cgrp_prioidx = state ? state->prioidx : 0;
1180 rcu_read_unlock();
1181}
1182EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1183#endif
1184
1da177e4
LT		 1185/**
1186 * sk_alloc - All socket objects are allocated here
c4ea43c5 1187 * @net: the applicable net namespace
4dc3b16b
PP		 1188 * @family: protocol family
1189 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1190 * @prot: struct proto associated with this new sock instance
1da177e4 1191 */
1b8d7ae4 1192struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1193 struct proto *prot)
1da177e4 1194{
c308c1b2 1195 struct sock *sk;
1da177e4 1196
154adbc8 1197 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1198 if (sk) {
154adbc8
PE		 1199 sk->sk_family = family;
1200 /*
1201 * See comment in struct sock definition to understand
1202 * why we need sk_prot_creator -acme
1203 */
1204 sk->sk_prot = sk->sk_prot_creator = prot;
1205 sock_lock_init(sk);
3b1e0a65 1206 sock_net_set(sk, get_net(net));
d66ee058 1207 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725
HX		 1208
1209 sock_update_classid(sk);
5bc1421e 1210 sock_update_netprioidx(sk);
1da177e4 1211 }
a79af59e 1212
2e4afe7b 1213 return sk;
1da177e4 1214}
2a91525c 1215EXPORT_SYMBOL(sk_alloc);
1da177e4 1216
2b85a34e 1217static void __sk_free(struct sock *sk)
1da177e4
LT		 1218{
1219 struct sk_filter *filter;
1da177e4
LT		 1220
1221 if (sk->sk_destruct)
1222 sk->sk_destruct(sk);
1223
a898def2
PM		 1224 filter = rcu_dereference_check(sk->sk_filter,
1225 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1226 if (filter) {
309dd5fc 1227 sk_filter_uncharge(sk, filter);
a9b3cd7f 1228 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1229 }
1230
08e29af3 1231 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1232
1233 if (atomic_read(&sk->sk_omem_alloc))
1234 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n",
0dc47877 1235 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1236
109f6e39
EB		 1237 if (sk->sk_peer_cred)
1238 put_cred(sk->sk_peer_cred);
1239 put_pid(sk->sk_peer_pid);
3b1e0a65 1240 put_net(sock_net(sk));
c308c1b2 1241 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1242}
2b85a34e
ED		 1243
1244void sk_free(struct sock *sk)
1245{
1246 /*
25985edc 1247 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1248 * some packets are still in some tx queue.
1249 * If not null, sock_wfree() will call __sk_free(sk) later
1250 */
1251 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1252 __sk_free(sk);
1253}
2a91525c 1254EXPORT_SYMBOL(sk_free);
1da177e4 1255
edf02087 1256/*
25985edc
LDM		 1257 * Last sock_put should drop reference to sk->sk_net. It has already
1258 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1259 * is not an option.
25985edc 1260 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1261 * destroy it in the context of init_net.
1262 */
1263void sk_release_kernel(struct sock *sk)
1264{
1265 if (sk == NULL || sk->sk_socket == NULL)
1266 return;
1267
1268 sock_hold(sk);
1269 sock_release(sk->sk_socket);
65a18ec5 1270 release_net(sock_net(sk));
3b1e0a65 1271 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1272 sock_put(sk);
1273}
45af1754 1274EXPORT_SYMBOL(sk_release_kernel);
edf02087 1275
475f1b52
SR		 1276static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1277{
1278 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1279 sock_update_memcg(newsk);
1280}
1281
e56c57d0
ED		 1282/**
1283 * sk_clone_lock - clone a socket, and lock its clone
1284 * @sk: the socket to clone
1285 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1286 *
1287 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1288 */
1289struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1290{
8fd1d178 1291 struct sock *newsk;
87d11ceb 1292
8fd1d178 1293 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1294 if (newsk != NULL) {
1295 struct sk_filter *filter;
1296
892c141e 1297 sock_copy(newsk, sk);
87d11ceb
ACM		 1298
1299 /* SANITY */
3b1e0a65 1300 get_net(sock_net(newsk));
87d11ceb
ACM		 1301 sk_node_init(&newsk->sk_node);
1302 sock_lock_init(newsk);
1303 bh_lock_sock(newsk);
fa438ccf 1304 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1305 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1306
1307 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1308 /*
1309 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1310 */
1311 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1312 atomic_set(&newsk->sk_omem_alloc, 0);
1313 skb_queue_head_init(&newsk->sk_receive_queue);
1314 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1315#ifdef CONFIG_NET_DMA
1316 skb_queue_head_init(&newsk->sk_async_wait_queue);
1317#endif
87d11ceb 1318
b6c6712a 1319 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1320 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1321 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1322 af_callback_keys + newsk->sk_family,
1323 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1324
1325 newsk->sk_dst_cache = NULL;
1326 newsk->sk_wmem_queued = 0;
1327 newsk->sk_forward_alloc = 0;
1328 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1329 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1330
1331 sock_reset_flag(newsk, SOCK_DONE);
1332 skb_queue_head_init(&newsk->sk_error_queue);
1333
0d7da9dd 1334 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1335 if (filter != NULL)
1336 sk_filter_charge(newsk, filter);
1337
1338 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1339 /* It is still raw copy of parent, so invalidate
1340 * destructor and make plain sk_free() */
1341 newsk->sk_destruct = NULL;
b0691c8e 1342 bh_unlock_sock(newsk);
87d11ceb
ACM		 1343 sk_free(newsk);
1344 newsk = NULL;
1345 goto out;
1346 }
1347
1348 newsk->sk_err = 0;
1349 newsk->sk_priority = 0;
4dc6dc71
ED		 1350 /*
1351 * Before updating sk_refcnt, we must commit prior changes to memory
1352 * (Documentation/RCU/rculist_nulls.txt for details)
1353 */
1354 smp_wmb();
87d11ceb
ACM		 1355 atomic_set(&newsk->sk_refcnt, 2);
1356
1357 /*
1358 * Increment the counter in the same struct proto as the master
1359 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1360 * is the same as sk->sk_prot->socks, as this field was copied
1361 * with memcpy).
1362 *
1363 * This _changes_ the previous behaviour, where
1364 * tcp_create_openreq_child always was incrementing the
1365 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1366 * to be taken into account in all callers. -acme
1367 */
1368 sk_refcnt_debug_inc(newsk);
972692e0 1369 sk_set_socket(newsk, NULL);
43815482 1370 newsk->sk_wq = NULL;
87d11ceb 1371
f3f511e1
GC		 1372 sk_update_clone(sk, newsk);
1373
87d11ceb 1374 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1375 sk_sockets_allocated_inc(newsk);
704da560 1376
08e29af3 1377 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1378 net_enable_timestamp();
87d11ceb
ACM		 1379 }
1380out:
1381 return newsk;
1382}
e56c57d0 1383EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1384
9958089a
AK		 1385void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1386{
1387 __sk_dst_set(sk, dst);
1388 sk->sk_route_caps = dst->dev->features;
1389 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1390 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1391 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1392 if (sk_can_gso(sk)) {
82cc1a7a 1393 if (dst->header_len) {
9958089a 1394 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1395 } else {
9958089a 1396 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a
PWJ		 1397 sk->sk_gso_max_size = dst->dev->gso_max_size;
1398 }
9958089a
AK		 1399 }
1400}
1401EXPORT_SYMBOL_GPL(sk_setup_caps);
1402
1da177e4
LT		 1403void __init sk_init(void)
1404{
4481374c 1405 if (totalram_pages <= 4096) {
1da177e4
LT		 1406 sysctl_wmem_max = 32767;
1407 sysctl_rmem_max = 32767;
1408 sysctl_wmem_default = 32767;
1409 sysctl_rmem_default = 32767;
4481374c 1410 } else if (totalram_pages >= 131072) {
1da177e4
LT		 1411 sysctl_wmem_max = 131071;
1412 sysctl_rmem_max = 131071;
1413 }
1414}
1415
1416/*
1417 * Simple resource managers for sockets.
1418 */
1419
1420
4ec93edb
YH		 1421/*
1422 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1423 */
1424void sock_wfree(struct sk_buff *skb)
1425{
1426 struct sock *sk = skb->sk;
d99927f4 1427 unsigned int len = skb->truesize;
1da177e4 1428
d99927f4
ED		 1429 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1430 /*
1431 * Keep a reference on sk_wmem_alloc, this will be released
1432 * after sk_write_space() call
1433 */
1434 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1435 sk->sk_write_space(sk);
d99927f4
ED		 1436 len = 1;
1437 }
2b85a34e 1438 /*
d99927f4
ED		 1439 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1440 * could not do because of in-flight packets
2b85a34e 1441 */
d99927f4 1442 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1443 __sk_free(sk);
1da177e4 1444}
2a91525c 1445EXPORT_SYMBOL(sock_wfree);
1da177e4 1446
4ec93edb
YH		 1447/*
1448 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1449 */
1450void sock_rfree(struct sk_buff *skb)
1451{
1452 struct sock *sk = skb->sk;
d361fd59 1453 unsigned int len = skb->truesize;
1da177e4 1454
d361fd59
ED		 1455 atomic_sub(len, &sk->sk_rmem_alloc);
1456 sk_mem_uncharge(sk, len);
1da177e4 1457}
2a91525c 1458EXPORT_SYMBOL(sock_rfree);
1da177e4
LT		 1459
1460
1461int sock_i_uid(struct sock *sk)
1462{
1463 int uid;
1464
f064af1e 1465 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1466 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
f064af1e 1467 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1468 return uid;
1469}
2a91525c 1470EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1471
1472unsigned long sock_i_ino(struct sock *sk)
1473{
1474 unsigned long ino;
1475
f064af1e 1476 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1477 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1478 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1479 return ino;
1480}
2a91525c 1481EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1482
1483/*
1484 * Allocate a skb from the socket's send buffer.
1485 */
86a76caf 1486struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1487 gfp_t priority)
1da177e4
LT		 1488{
1489 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1490 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1491 if (skb) {
1492 skb_set_owner_w(skb, sk);
1493 return skb;
1494 }
1495 }
1496 return NULL;
1497}
2a91525c 1498EXPORT_SYMBOL(sock_wmalloc);
1da177e4
LT		 1499
1500/*
1501 * Allocate a skb from the socket's receive buffer.
4ec93edb 1502 */
86a76caf 1503struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1504 gfp_t priority)
1da177e4
LT		 1505{
1506 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1507 struct sk_buff *skb = alloc_skb(size, priority);
1508 if (skb) {
1509 skb_set_owner_r(skb, sk);
1510 return skb;
1511 }
1512 }
1513 return NULL;
1514}
1515
4ec93edb 1516/*
1da177e4 1517 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1518 */
dd0fc66f 1519void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4
LT		 1520{
1521 if ((unsigned)size <= sysctl_optmem_max &&
1522 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1523 void *mem;
1524 /* First do the add, to avoid the race if kmalloc
4ec93edb 1525 * might sleep.
1da177e4
LT		 1526 */
1527 atomic_add(size, &sk->sk_omem_alloc);
1528 mem = kmalloc(size, priority);
1529 if (mem)
1530 return mem;
1531 atomic_sub(size, &sk->sk_omem_alloc);
1532 }
1533 return NULL;
1534}
2a91525c 1535EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1536
1537/*
1538 * Free an option memory block.
1539 */
1540void sock_kfree_s(struct sock *sk, void *mem, int size)
1541{
1542 kfree(mem);
1543 atomic_sub(size, &sk->sk_omem_alloc);
1544}
2a91525c 1545EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1546
1547/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1548 I think, these locks should be removed for datagram sockets.
1549 */
2a91525c 1550static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1551{
1552 DEFINE_WAIT(wait);
1553
1554 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1555 for (;;) {
1556 if (!timeo)
1557 break;
1558 if (signal_pending(current))
1559 break;
1560 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1561 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1562 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1563 break;
1564 if (sk->sk_shutdown & SEND_SHUTDOWN)
1565 break;
1566 if (sk->sk_err)
1567 break;
1568 timeo = schedule_timeout(timeo);
1569 }
aa395145 1570 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1571 return timeo;
1572}
1573
1574
1575/*
1576 * Generic send/receive buffer handlers
1577 */
1578
4cc7f68d
HX		 1579struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1580 unsigned long data_len, int noblock,
1581 int *errcode)
1da177e4
LT		 1582{
1583 struct sk_buff *skb;
7d877f3b 1584 gfp_t gfp_mask;
1da177e4
LT		 1585 long timeo;
1586 int err;
1587
1588 gfp_mask = sk->sk_allocation;
1589 if (gfp_mask & __GFP_WAIT)
1590 gfp_mask |= __GFP_REPEAT;
1591
1592 timeo = sock_sndtimeo(sk, noblock);
1593 while (1) {
1594 err = sock_error(sk);
1595 if (err != 0)
1596 goto failure;
1597
1598 err = -EPIPE;
1599 if (sk->sk_shutdown & SEND_SHUTDOWN)
1600 goto failure;
1601
1602 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1603 skb = alloc_skb(header_len, gfp_mask);
1da177e4
LT		 1604 if (skb) {
1605 int npages;
1606 int i;
1607
1608 /* No pages, we're done... */
1609 if (!data_len)
1610 break;
1611
1612 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1613 skb->truesize += data_len;
1614 skb_shinfo(skb)->nr_frags = npages;
1615 for (i = 0; i < npages; i++) {
1616 struct page *page;
1da177e4
LT		 1617
1618 page = alloc_pages(sk->sk_allocation, 0);
1619 if (!page) {
1620 err = -ENOBUFS;
1621 skb_shinfo(skb)->nr_frags = i;
1622 kfree_skb(skb);
1623 goto failure;
1624 }
1625
ea2ab693
IC		 1626 __skb_fill_page_desc(skb, i,
1627 page, 0,
1628 (data_len >= PAGE_SIZE ?
1629 PAGE_SIZE :
1630 data_len));
1da177e4
LT		 1631 data_len -= PAGE_SIZE;
1632 }
1633
1634 /* Full success... */
1635 break;
1636 }
1637 err = -ENOBUFS;
1638 goto failure;
1639 }
1640 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1641 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1642 err = -EAGAIN;
1643 if (!timeo)
1644 goto failure;
1645 if (signal_pending(current))
1646 goto interrupted;
1647 timeo = sock_wait_for_wmem(sk, timeo);
1648 }
1649
1650 skb_set_owner_w(skb, sk);
1651 return skb;
1652
1653interrupted:
1654 err = sock_intr_errno(timeo);
1655failure:
1656 *errcode = err;
1657 return NULL;
1658}
4cc7f68d 1659EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1660
4ec93edb 1661struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1662 int noblock, int *errcode)
1663{
1664 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1665}
2a91525c 1666EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4
LT		 1667
1668static void __lock_sock(struct sock *sk)
f39234d6
NK		 1669 __releases(&sk->sk_lock.slock)
1670 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1671{
1672 DEFINE_WAIT(wait);
1673
e71a4783 1674 for (;;) {
1da177e4
LT		 1675 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1676 TASK_UNINTERRUPTIBLE);
1677 spin_unlock_bh(&sk->sk_lock.slock);
1678 schedule();
1679 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1680 if (!sock_owned_by_user(sk))
1da177e4
LT		 1681 break;
1682 }
1683 finish_wait(&sk->sk_lock.wq, &wait);
1684}
1685
1686static void __release_sock(struct sock *sk)
f39234d6
NK		 1687 __releases(&sk->sk_lock.slock)
1688 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1689{
1690 struct sk_buff *skb = sk->sk_backlog.head;
1691
1692 do {
1693 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1694 bh_unlock_sock(sk);
1695
1696 do {
1697 struct sk_buff *next = skb->next;
1698
7fee226a 1699 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1700 skb->next = NULL;
c57943a1 1701 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1702
1703 /*
1704 * We are in process context here with softirqs
1705 * disabled, use cond_resched_softirq() to preempt.
1706 * This is safe to do because we've taken the backlog
1707 * queue private:
1708 */
1709 cond_resched_softirq();
1710
1711 skb = next;
1712 } while (skb != NULL);
1713
1714 bh_lock_sock(sk);
e71a4783 1715 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1716
1717 /*
1718 * Doing the zeroing here guarantee we can not loop forever
1719 * while a wild producer attempts to flood us.
1720 */
1721 sk->sk_backlog.len = 0;
1da177e4
LT		 1722}
1723
1724/**
1725 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1726 * @sk: sock to wait on
1727 * @timeo: for how long
1da177e4
LT		 1728 *
1729 * Now socket state including sk->sk_err is changed only under lock,
1730 * hence we may omit checks after joining wait queue.
1731 * We check receive queue before schedule() only as optimization;
1732 * it is very likely that release_sock() added new data.
1733 */
1734int sk_wait_data(struct sock *sk, long *timeo)
1735{
1736 int rc;
1737 DEFINE_WAIT(wait);
1738
aa395145 1739 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1740 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1741 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1742 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1743 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1744 return rc;
1745}
1da177e4
LT		 1746EXPORT_SYMBOL(sk_wait_data);
1747
3ab224be
HA		 1748/**
1749 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1750 * @sk: socket
1751 * @size: memory size to allocate
1752 * @kind: allocation type
1753 *
1754 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1755 * rmem allocation. This function assumes that protocols which have
1756 * memory_pressure use sk_wmem_queued as write buffer accounting.
1757 */
1758int __sk_mem_schedule(struct sock *sk, int size, int kind)
1759{
1760 struct proto *prot = sk->sk_prot;
1761 int amt = sk_mem_pages(size);
8d987e5c 1762 long allocated;
e1aab161 1763 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1764
1765 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1766
e1aab161 1767 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1768
1769 /* Under limit. */
e1aab161
GC		 1770 if (parent_status == UNDER_LIMIT &&
1771 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1772 sk_leave_memory_pressure(sk);
3ab224be
HA		 1773 return 1;
1774 }
1775
e1aab161
GC		 1776 /* Under pressure. (we or our parents) */
1777 if ((parent_status > SOFT_LIMIT) ||
1778 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1779 sk_enter_memory_pressure(sk);
3ab224be 1780
e1aab161
GC		 1781 /* Over hard limit (we or our parents) */
1782 if ((parent_status == OVER_LIMIT) ||
1783 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1784 goto suppress_allocation;
1785
1786 /* guarantee minimum buffer size under pressure */
1787 if (kind == SK_MEM_RECV) {
1788 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1789 return 1;
180d8cd9 1790
3ab224be
HA		 1791 } else { /* SK_MEM_SEND */
1792 if (sk->sk_type == SOCK_STREAM) {
1793 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1794 return 1;
1795 } else if (atomic_read(&sk->sk_wmem_alloc) <
1796 prot->sysctl_wmem[0])
1797 return 1;
1798 }
1799
180d8cd9 1800 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 1801 int alloc;
1802
180d8cd9 1803 if (!sk_under_memory_pressure(sk))
1748376b 1804 return 1;
180d8cd9
GC		 1805 alloc = sk_sockets_allocated_read_positive(sk);
1806 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 1807 sk_mem_pages(sk->sk_wmem_queued +
1808 atomic_read(&sk->sk_rmem_alloc) +
1809 sk->sk_forward_alloc))
1810 return 1;
1811 }
1812
1813suppress_allocation:
1814
1815 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1816 sk_stream_moderate_sndbuf(sk);
1817
1818 /* Fail only if socket is _under_ its sndbuf.
1819 * In this case we cannot block, so that we have to fail.
1820 */
1821 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1822 return 1;
1823 }
1824
3847ce32
SM		 1825 trace_sock_exceed_buf_limit(sk, prot, allocated);
1826
3ab224be
HA		 1827 /* Alas. Undo changes. */
1828 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 1829
0e90b31f 1830 sk_memory_allocated_sub(sk, amt);
180d8cd9 1831
3ab224be
HA		 1832 return 0;
1833}
3ab224be
HA		 1834EXPORT_SYMBOL(__sk_mem_schedule);
1835
1836/**
1837 * __sk_reclaim - reclaim memory_allocated
1838 * @sk: socket
1839 */
1840void __sk_mem_reclaim(struct sock *sk)
1841{
180d8cd9 1842 sk_memory_allocated_sub(sk,
0e90b31f 1843 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 1844 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
1845
180d8cd9
GC		 1846 if (sk_under_memory_pressure(sk) &&
1847 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
1848 sk_leave_memory_pressure(sk);
3ab224be 1849}
3ab224be
HA		 1850EXPORT_SYMBOL(__sk_mem_reclaim);
1851
1852
1da177e4
LT		 1853/*
1854 * Set of default routines for initialising struct proto_ops when
1855 * the protocol does not support a particular function. In certain
1856 * cases where it makes no sense for a protocol to have a "do nothing"
1857 * function, some default processing is provided.
1858 */
1859
1860int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1861{
1862 return -EOPNOTSUPP;
1863}
2a91525c 1864EXPORT_SYMBOL(sock_no_bind);
1da177e4 1865
4ec93edb 1866int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1867 int len, int flags)
1868{
1869 return -EOPNOTSUPP;
1870}
2a91525c 1871EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 1872
1873int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1874{
1875 return -EOPNOTSUPP;
1876}
2a91525c 1877EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 1878
1879int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1880{
1881 return -EOPNOTSUPP;
1882}
2a91525c 1883EXPORT_SYMBOL(sock_no_accept);
1da177e4 1884
4ec93edb 1885int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1886 int *len, int peer)
1887{
1888 return -EOPNOTSUPP;
1889}
2a91525c 1890EXPORT_SYMBOL(sock_no_getname);
1da177e4 1891
2a91525c 1892unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 1893{
1894 return 0;
1895}
2a91525c 1896EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 1897
1898int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1899{
1900 return -EOPNOTSUPP;
1901}
2a91525c 1902EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 1903
1904int sock_no_listen(struct socket *sock, int backlog)
1905{
1906 return -EOPNOTSUPP;
1907}
2a91525c 1908EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 1909
1910int sock_no_shutdown(struct socket *sock, int how)
1911{
1912 return -EOPNOTSUPP;
1913}
2a91525c 1914EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 1915
1916int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 1917 char __user *optval, unsigned int optlen)
1da177e4
LT		 1918{
1919 return -EOPNOTSUPP;
1920}
2a91525c 1921EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 1922
1923int sock_no_getsockopt(struct socket *sock, int level, int optname,
1924 char __user *optval, int __user *optlen)
1925{
1926 return -EOPNOTSUPP;
1927}
2a91525c 1928EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 1929
1930int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1931 size_t len)
1932{
1933 return -EOPNOTSUPP;
1934}
2a91525c 1935EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 1936
1937int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1938 size_t len, int flags)
1939{
1940 return -EOPNOTSUPP;
1941}
2a91525c 1942EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 1943
1944int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
1945{
1946 /* Mirror missing mmap method error code */
1947 return -ENODEV;
1948}
2a91525c 1949EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 1950
1951ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
1952{
1953 ssize_t res;
1954 struct msghdr msg = {.msg_flags = flags};
1955 struct kvec iov;
1956 char *kaddr = kmap(page);
1957 iov.iov_base = kaddr + offset;
1958 iov.iov_len = size;
1959 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
1960 kunmap(page);
1961 return res;
1962}
2a91525c 1963EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 1964
1965/*
1966 * Default Socket Callbacks
1967 */
1968
1969static void sock_def_wakeup(struct sock *sk)
1970{
43815482
ED		 1971 struct socket_wq *wq;
1972
1973 rcu_read_lock();
1974 wq = rcu_dereference(sk->sk_wq);
1975 if (wq_has_sleeper(wq))
1976 wake_up_interruptible_all(&wq->wait);
1977 rcu_read_unlock();
1da177e4
LT		 1978}
1979
1980static void sock_def_error_report(struct sock *sk)
1981{
43815482
ED		 1982 struct socket_wq *wq;
1983
1984 rcu_read_lock();
1985 wq = rcu_dereference(sk->sk_wq);
1986 if (wq_has_sleeper(wq))
1987 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 1988 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 1989 rcu_read_unlock();
1da177e4
LT		 1990}
1991
1992static void sock_def_readable(struct sock *sk, int len)
1993{
43815482
ED		 1994 struct socket_wq *wq;
1995
1996 rcu_read_lock();
1997 wq = rcu_dereference(sk->sk_wq);
1998 if (wq_has_sleeper(wq))
2c6607c6 1999 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2000 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2001 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2002 rcu_read_unlock();
1da177e4
LT		 2003}
2004
2005static void sock_def_write_space(struct sock *sk)
2006{
43815482
ED		 2007 struct socket_wq *wq;
2008
2009 rcu_read_lock();
1da177e4
LT		 2010
2011 /* Do not wake up a writer until he can make "significant"
2012 * progress. --DaveM
2013 */
e71a4783 2014 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2015 wq = rcu_dereference(sk->sk_wq);
2016 if (wq_has_sleeper(wq))
2017 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2018 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2019
2020 /* Should agree with poll, otherwise some programs break */
2021 if (sock_writeable(sk))
8d8ad9d7 2022 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2023 }
2024
43815482 2025 rcu_read_unlock();
1da177e4
LT		 2026}
2027
2028static void sock_def_destruct(struct sock *sk)
2029{
a51482bd 2030 kfree(sk->sk_protinfo);
1da177e4
LT		 2031}
2032
2033void sk_send_sigurg(struct sock *sk)
2034{
2035 if (sk->sk_socket && sk->sk_socket->file)
2036 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2037 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2038}
2a91525c 2039EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2040
2041void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2042 unsigned long expires)
2043{
2044 if (!mod_timer(timer, expires))
2045 sock_hold(sk);
2046}
1da177e4
LT		 2047EXPORT_SYMBOL(sk_reset_timer);
2048
2049void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2050{
2051 if (timer_pending(timer) && del_timer(timer))
2052 __sock_put(sk);
2053}
1da177e4
LT		 2054EXPORT_SYMBOL(sk_stop_timer);
2055
2056void sock_init_data(struct socket *sock, struct sock *sk)
2057{
2058 skb_queue_head_init(&sk->sk_receive_queue);
2059 skb_queue_head_init(&sk->sk_write_queue);
2060 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2061#ifdef CONFIG_NET_DMA
2062 skb_queue_head_init(&sk->sk_async_wait_queue);
2063#endif
1da177e4
LT		 2064
2065 sk->sk_send_head = NULL;
2066
2067 init_timer(&sk->sk_timer);
4ec93edb 2068
1da177e4
LT		 2069 sk->sk_allocation = GFP_KERNEL;
2070 sk->sk_rcvbuf = sysctl_rmem_default;
2071 sk->sk_sndbuf = sysctl_wmem_default;
2072 sk->sk_state = TCP_CLOSE;
972692e0 2073 sk_set_socket(sk, sock);
1da177e4
LT		 2074
2075 sock_set_flag(sk, SOCK_ZAPPED);
2076
e71a4783 2077 if (sock) {
1da177e4 2078 sk->sk_type = sock->type;
43815482 2079 sk->sk_wq = sock->wq;
1da177e4
LT		 2080 sock->sk = sk;
2081 } else
43815482 2082 sk->sk_wq = NULL;
1da177e4 2083
b6c6712a 2084 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2085 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2086 lockdep_set_class_and_name(&sk->sk_callback_lock,
2087 af_callback_keys + sk->sk_family,
2088 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2089
2090 sk->sk_state_change = sock_def_wakeup;
2091 sk->sk_data_ready = sock_def_readable;
2092 sk->sk_write_space = sock_def_write_space;
2093 sk->sk_error_report = sock_def_error_report;
2094 sk->sk_destruct = sock_def_destruct;
2095
2096 sk->sk_sndmsg_page = NULL;
2097 sk->sk_sndmsg_off = 0;
2098
109f6e39
EB		 2099 sk->sk_peer_pid = NULL;
2100 sk->sk_peer_cred = NULL;
1da177e4
LT		 2101 sk->sk_write_pending = 0;
2102 sk->sk_rcvlowat = 1;
2103 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2104 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2105
f37f0afb 2106 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2107
4dc6dc71
ED		 2108 /*
2109 * Before updating sk_refcnt, we must commit prior changes to memory
2110 * (Documentation/RCU/rculist_nulls.txt for details)
2111 */
2112 smp_wmb();
1da177e4 2113 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2114 atomic_set(&sk->sk_drops, 0);
1da177e4 2115}
2a91525c 2116EXPORT_SYMBOL(sock_init_data);
1da177e4 2117
b5606c2d 2118void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2119{
2120 might_sleep();
a5b5bb9a 2121 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2122 if (sk->sk_lock.owned)
1da177e4 2123 __lock_sock(sk);
d2e9117c 2124 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2125 spin_unlock(&sk->sk_lock.slock);
2126 /*
2127 * The sk_lock has mutex_lock() semantics here:
2128 */
fcc70d5f 2129 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2130 local_bh_enable();
1da177e4 2131}
fcc70d5f 2132EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2133
b5606c2d 2134void release_sock(struct sock *sk)
1da177e4 2135{
a5b5bb9a
IM		 2136 /*
2137 * The sk_lock has mutex_unlock() semantics:
2138 */
2139 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2140
2141 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2142 if (sk->sk_backlog.tail)
2143 __release_sock(sk);
d2e9117c 2144 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2145 if (waitqueue_active(&sk->sk_lock.wq))
2146 wake_up(&sk->sk_lock.wq);
2147 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2148}
2149EXPORT_SYMBOL(release_sock);
2150
8a74ad60
ED		 2151/**
2152 * lock_sock_fast - fast version of lock_sock
2153 * @sk: socket
2154 *
2155 * This version should be used for very small section, where process wont block
2156 * return false if fast path is taken
2157 * sk_lock.slock locked, owned = 0, BH disabled
2158 * return true if slow path is taken
2159 * sk_lock.slock unlocked, owned = 1, BH enabled
2160 */
2161bool lock_sock_fast(struct sock *sk)
2162{
2163 might_sleep();
2164 spin_lock_bh(&sk->sk_lock.slock);
2165
2166 if (!sk->sk_lock.owned)
2167 /*
2168 * Note : We must disable BH
2169 */
2170 return false;
2171
2172 __lock_sock(sk);
2173 sk->sk_lock.owned = 1;
2174 spin_unlock(&sk->sk_lock.slock);
2175 /*
2176 * The sk_lock has mutex_lock() semantics here:
2177 */
2178 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2179 local_bh_enable();
2180 return true;
2181}
2182EXPORT_SYMBOL(lock_sock_fast);
2183
1da177e4 2184int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2185{
b7aa0bf7 2186 struct timeval tv;
1da177e4 2187 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2188 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2189 tv = ktime_to_timeval(sk->sk_stamp);
2190 if (tv.tv_sec == -1)
1da177e4 2191 return -ENOENT;
b7aa0bf7
ED		 2192 if (tv.tv_sec == 0) {
2193 sk->sk_stamp = ktime_get_real();
2194 tv = ktime_to_timeval(sk->sk_stamp);
2195 }
2196 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2197}
1da177e4
LT		 2198EXPORT_SYMBOL(sock_get_timestamp);
2199
ae40eb1e
ED		 2200int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2201{
2202 struct timespec ts;
2203 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2204 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2205 ts = ktime_to_timespec(sk->sk_stamp);
2206 if (ts.tv_sec == -1)
2207 return -ENOENT;
2208 if (ts.tv_sec == 0) {
2209 sk->sk_stamp = ktime_get_real();
2210 ts = ktime_to_timespec(sk->sk_stamp);
2211 }
2212 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2213}
2214EXPORT_SYMBOL(sock_get_timestampns);
2215
20d49473 2216void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2217{
20d49473 2218 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2219 unsigned long previous_flags = sk->sk_flags;
2220
20d49473
PO		 2221 sock_set_flag(sk, flag);
2222 /*
2223 * we just set one of the two flags which require net
2224 * time stamping, but time stamping might have been on
2225 * already because of the other one
2226 */
08e29af3 2227 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2228 net_enable_timestamp();
1da177e4
LT		 2229 }
2230}
1da177e4
LT		 2231
2232/*
2233 * Get a socket option on an socket.
2234 *
2235 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2236 * asynchronous errors should be reported by getsockopt. We assume
2237 * this means if you specify SO_ERROR (otherwise whats the point of it).
2238 */
2239int sock_common_getsockopt(struct socket *sock, int level, int optname,
2240 char __user *optval, int __user *optlen)
2241{
2242 struct sock *sk = sock->sk;
2243
2244 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2245}
1da177e4
LT		 2246EXPORT_SYMBOL(sock_common_getsockopt);
2247
3fdadf7d 2248#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2249int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2250 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2251{
2252 struct sock *sk = sock->sk;
2253
1e51f951 2254 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2255 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2256 optval, optlen);
3fdadf7d
DM		 2257 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2258}
2259EXPORT_SYMBOL(compat_sock_common_getsockopt);
2260#endif
2261
1da177e4
LT		 2262int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2263 struct msghdr *msg, size_t size, int flags)
2264{
2265 struct sock *sk = sock->sk;
2266 int addr_len = 0;
2267 int err;
2268
2269 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2270 flags & ~MSG_DONTWAIT, &addr_len);
2271 if (err >= 0)
2272 msg->msg_namelen = addr_len;
2273 return err;
2274}
1da177e4
LT		 2275EXPORT_SYMBOL(sock_common_recvmsg);
2276
2277/*
2278 * Set socket options on an inet socket.
2279 */
2280int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2281 char __user *optval, unsigned int optlen)
1da177e4
LT		 2282{
2283 struct sock *sk = sock->sk;
2284
2285 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2286}
1da177e4
LT		 2287EXPORT_SYMBOL(sock_common_setsockopt);
2288
3fdadf7d 2289#ifdef CONFIG_COMPAT
543d9cfe 2290int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2291 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2292{
2293 struct sock *sk = sock->sk;
2294
543d9cfe
ACM		 2295 if (sk->sk_prot->compat_setsockopt != NULL)
2296 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2297 optval, optlen);
3fdadf7d
DM		 2298 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2299}
2300EXPORT_SYMBOL(compat_sock_common_setsockopt);
2301#endif
2302
1da177e4
LT		 2303void sk_common_release(struct sock *sk)
2304{
2305 if (sk->sk_prot->destroy)
2306 sk->sk_prot->destroy(sk);
2307
2308 /*
2309 * Observation: when sock_common_release is called, processes have
2310 * no access to socket. But net still has.
2311 * Step one, detach it from networking:
2312 *
2313 * A. Remove from hash tables.
2314 */
2315
2316 sk->sk_prot->unhash(sk);
2317
2318 /*
2319 * In this point socket cannot receive new packets, but it is possible
2320 * that some packets are in flight because some CPU runs receiver and
2321 * did hash table lookup before we unhashed socket. They will achieve
2322 * receive queue and will be purged by socket destructor.
2323 *
2324 * Also we still have packets pending on receive queue and probably,
2325 * our own packets waiting in device queues. sock_destroy will drain
2326 * receive queue, but transmitted packets will delay socket destruction
2327 * until the last reference will be released.
2328 */
2329
2330 sock_orphan(sk);
2331
2332 xfrm_sk_free_policy(sk);
2333
e6848976 2334 sk_refcnt_debug_release(sk);
1da177e4
LT		 2335 sock_put(sk);
2336}
1da177e4
LT		 2337EXPORT_SYMBOL(sk_common_release);
2338
13ff3d6f
PE		 2339#ifdef CONFIG_PROC_FS
2340#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2341struct prot_inuse {
2342 int val[PROTO_INUSE_NR];
2343};
13ff3d6f
PE		 2344
2345static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2346
2347#ifdef CONFIG_NET_NS
2348void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2349{
d6d9ca0f 2350 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2351}
2352EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2353
2354int sock_prot_inuse_get(struct net *net, struct proto *prot)
2355{
2356 int cpu, idx = prot->inuse_idx;
2357 int res = 0;
2358
2359 for_each_possible_cpu(cpu)
2360 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2361
2362 return res >= 0 ? res : 0;
2363}
2364EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2365
2c8c1e72 2366static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2367{
2368 net->core.inuse = alloc_percpu(struct prot_inuse);
2369 return net->core.inuse ? 0 : -ENOMEM;
2370}
2371
2c8c1e72 2372static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2373{
2374 free_percpu(net->core.inuse);
2375}
2376
2377static struct pernet_operations net_inuse_ops = {
2378 .init = sock_inuse_init_net,
2379 .exit = sock_inuse_exit_net,
2380};
2381
2382static __init int net_inuse_init(void)
2383{
2384 if (register_pernet_subsys(&net_inuse_ops))
2385 panic("Cannot initialize net inuse counters");
2386
2387 return 0;
2388}
2389
2390core_initcall(net_inuse_init);
2391#else
1338d466
PE		 2392static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2393
c29a0bc4 2394void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2395{
d6d9ca0f 2396 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2397}
2398EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2399
c29a0bc4 2400int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2401{
2402 int cpu, idx = prot->inuse_idx;
2403 int res = 0;
2404
2405 for_each_possible_cpu(cpu)
2406 res += per_cpu(prot_inuse, cpu).val[idx];
2407
2408 return res >= 0 ? res : 0;
2409}
2410EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2411#endif
13ff3d6f
PE		 2412
2413static void assign_proto_idx(struct proto *prot)
2414{
2415 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2416
2417 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
2418 printk(KERN_ERR "PROTO_INUSE_NR exhausted\n");
2419 return;
2420 }
2421
2422 set_bit(prot->inuse_idx, proto_inuse_idx);
2423}
2424
2425static void release_proto_idx(struct proto *prot)
2426{
2427 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2428 clear_bit(prot->inuse_idx, proto_inuse_idx);
2429}
2430#else
2431static inline void assign_proto_idx(struct proto *prot)
2432{
2433}
2434
2435static inline void release_proto_idx(struct proto *prot)
2436{
2437}
2438#endif
2439
b733c007
PE		 2440int proto_register(struct proto *prot, int alloc_slab)
2441{
1da177e4
LT		 2442 if (alloc_slab) {
2443 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2444 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2445 NULL);
1da177e4
LT		 2446
2447 if (prot->slab == NULL) {
2448 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n",
2449 prot->name);
60e7663d 2450 goto out;
1da177e4 2451 }
2e6599cb
ACM		 2452
2453 if (prot->rsk_prot != NULL) {
faf23422 2454 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2455 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2456 goto out_free_sock_slab;
2457
7e56b5d6 2458 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2459 prot->rsk_prot->obj_size, 0,
20c2df83 2460 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2461
2462 if (prot->rsk_prot->slab == NULL) {
2463 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n",
2464 prot->name);
2465 goto out_free_request_sock_slab_name;
2466 }
2467 }
8feaf0c0 2468
6d6ee43e 2469 if (prot->twsk_prot != NULL) {
faf23422 2470 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2471
7e56b5d6 2472 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2473 goto out_free_request_sock_slab;
2474
6d6ee43e 2475 prot->twsk_prot->twsk_slab =
7e56b5d6 2476 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2477 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2478 0,
2479 SLAB_HWCACHE_ALIGN |
2480 prot->slab_flags,
20c2df83 2481 NULL);
6d6ee43e 2482 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2483 goto out_free_timewait_sock_slab_name;
2484 }
1da177e4
LT		 2485 }
2486
36b77a52 2487 mutex_lock(&proto_list_mutex);
1da177e4 2488 list_add(&prot->node, &proto_list);
13ff3d6f 2489 assign_proto_idx(prot);
36b77a52 2490 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2491 return 0;
2492
8feaf0c0 2493out_free_timewait_sock_slab_name:
7e56b5d6 2494 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2495out_free_request_sock_slab:
2496 if (prot->rsk_prot && prot->rsk_prot->slab) {
2497 kmem_cache_destroy(prot->rsk_prot->slab);
2498 prot->rsk_prot->slab = NULL;
2499 }
2e6599cb 2500out_free_request_sock_slab_name:
72150e9b
DC		 2501 if (prot->rsk_prot)
2502 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2503out_free_sock_slab:
2504 kmem_cache_destroy(prot->slab);
2505 prot->slab = NULL;
b733c007
PE		 2506out:
2507 return -ENOBUFS;
1da177e4 2508}
1da177e4
LT		 2509EXPORT_SYMBOL(proto_register);
2510
2511void proto_unregister(struct proto *prot)
2512{
36b77a52 2513 mutex_lock(&proto_list_mutex);
13ff3d6f 2514 release_proto_idx(prot);
0a3f4358 2515 list_del(&prot->node);
36b77a52 2516 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2517
2518 if (prot->slab != NULL) {
2519 kmem_cache_destroy(prot->slab);
2520 prot->slab = NULL;
2521 }
2522
2e6599cb 2523 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2524 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2525 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2526 prot->rsk_prot->slab = NULL;
2527 }
2528
6d6ee43e 2529 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2530 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2531 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2532 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2533 }
1da177e4 2534}
1da177e4
LT		 2535EXPORT_SYMBOL(proto_unregister);
2536
2537#ifdef CONFIG_PROC_FS
1da177e4 2538static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2539 __acquires(proto_list_mutex)
1da177e4 2540{
36b77a52 2541 mutex_lock(&proto_list_mutex);
60f0438a 2542 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2543}
2544
2545static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2546{
60f0438a 2547 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2548}
2549
2550static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2551 __releases(proto_list_mutex)
1da177e4 2552{
36b77a52 2553 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2554}
2555
2556static char proto_method_implemented(const void *method)
2557{
2558 return method == NULL ? 'n' : 'y';
2559}
180d8cd9
GC		 2560static long sock_prot_memory_allocated(struct proto *proto)
2561{
2562 return proto->memory_allocated != NULL ? proto_memory_allocated(proto): -1L;
2563}
2564
2565static char *sock_prot_memory_pressure(struct proto *proto)
2566{
2567 return proto->memory_pressure != NULL ?
2568 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2569}
1da177e4
LT		 2570
2571static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2572{
180d8cd9 2573
8d987e5c 2574 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2575 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2576 proto->name,
2577 proto->obj_size,
14e943db 2578 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2579 sock_prot_memory_allocated(proto),
2580 sock_prot_memory_pressure(proto),
1da177e4
LT		 2581 proto->max_header,
2582 proto->slab == NULL ? "no" : "yes",
2583 module_name(proto->owner),
2584 proto_method_implemented(proto->close),
2585 proto_method_implemented(proto->connect),
2586 proto_method_implemented(proto->disconnect),
2587 proto_method_implemented(proto->accept),
2588 proto_method_implemented(proto->ioctl),
2589 proto_method_implemented(proto->init),
2590 proto_method_implemented(proto->destroy),
2591 proto_method_implemented(proto->shutdown),
2592 proto_method_implemented(proto->setsockopt),
2593 proto_method_implemented(proto->getsockopt),
2594 proto_method_implemented(proto->sendmsg),
2595 proto_method_implemented(proto->recvmsg),
2596 proto_method_implemented(proto->sendpage),
2597 proto_method_implemented(proto->bind),
2598 proto_method_implemented(proto->backlog_rcv),
2599 proto_method_implemented(proto->hash),
2600 proto_method_implemented(proto->unhash),
2601 proto_method_implemented(proto->get_port),
2602 proto_method_implemented(proto->enter_memory_pressure));
2603}
2604
2605static int proto_seq_show(struct seq_file *seq, void *v)
2606{
60f0438a 2607 if (v == &proto_list)
1da177e4
LT		 2608 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2609 "protocol",
2610 "size",
2611 "sockets",
2612 "memory",
2613 "press",
2614 "maxhdr",
2615 "slab",
2616 "module",
2617 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2618 else
60f0438a 2619 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2620 return 0;
2621}
2622
f690808e 2623static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2624 .start = proto_seq_start,
2625 .next = proto_seq_next,
2626 .stop = proto_seq_stop,
2627 .show = proto_seq_show,
2628};
2629
2630static int proto_seq_open(struct inode *inode, struct file *file)
2631{
14e943db
ED		 2632 return seq_open_net(inode, file, &proto_seq_ops,
2633 sizeof(struct seq_net_private));
1da177e4
LT		 2634}
2635
9a32144e 2636static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2637 .owner = THIS_MODULE,
2638 .open = proto_seq_open,
2639 .read = seq_read,
2640 .llseek = seq_lseek,
14e943db
ED		 2641 .release = seq_release_net,
2642};
2643
2644static __net_init int proto_init_net(struct net *net)
2645{
2646 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2647 return -ENOMEM;
2648
2649 return 0;
2650}
2651
2652static __net_exit void proto_exit_net(struct net *net)
2653{
2654 proc_net_remove(net, "protocols");
2655}
2656
2657
2658static __net_initdata struct pernet_operations proto_net_ops = {
2659 .init = proto_init_net,
2660 .exit = proto_exit_net,
1da177e4
LT		 2661};
2662
2663static int __init proto_init(void)
2664{
14e943db 2665 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2666}
2667
2668subsys_initcall(proto_init);
2669
2670#endif /* PROC_FS */
kernel source for telekom puls (alcatel/mediatek)