]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - net/core/sock.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
v2 e1000: Neaten e1000_dump function
[mirror_ubuntu-hirsute-kernel.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>
e1aab161 114#include <linux/jump_label.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
e1aab161
GC		 187struct jump_label_key memcg_socket_limit_enabled;
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
ef64a54f
PE		 796 case SO_PEEK_OFF:
797 if (sock->ops->set_peek_off)
798 sock->ops->set_peek_off(sk, val);
799 else
800 ret = -EOPNOTSUPP;
801 break;
3bdc0eba
BG		 802
803 case SO_NOFCS:
804 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
805 break;
806
e71a4783
SH		 807 default:
808 ret = -ENOPROTOOPT;
809 break;
4ec93edb 810 }
1da177e4
LT		 811 release_sock(sk);
812 return ret;
813}
2a91525c 814EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 815
816
3f551f94
EB		 817void cred_to_ucred(struct pid *pid, const struct cred *cred,
818 struct ucred *ucred)
819{
820 ucred->pid = pid_vnr(pid);
821 ucred->uid = ucred->gid = -1;
822 if (cred) {
823 struct user_namespace *current_ns = current_user_ns();
824
825 ucred->uid = user_ns_map_uid(current_ns, cred, cred->euid);
826 ucred->gid = user_ns_map_gid(current_ns, cred, cred->egid);
827 }
828}
3924773a 829EXPORT_SYMBOL_GPL(cred_to_ucred);
3f551f94 830
1da177e4
LT		 831int sock_getsockopt(struct socket *sock, int level, int optname,
832 char __user *optval, int __user *optlen)
833{
834 struct sock *sk = sock->sk;
4ec93edb 835
e71a4783 836 union {
4ec93edb
YH		 837 int val;
838 struct linger ling;
1da177e4
LT		 839 struct timeval tm;
840 } v;
4ec93edb 841
4d0392be 842 int lv = sizeof(int);
1da177e4 843 int len;
4ec93edb 844
e71a4783 845 if (get_user(len, optlen))
4ec93edb 846 return -EFAULT;
e71a4783 847 if (len < 0)
1da177e4 848 return -EINVAL;
4ec93edb 849
50fee1de 850 memset(&v, 0, sizeof(v));
df0bca04 851
2a91525c 852 switch (optname) {
e71a4783
SH		 853 case SO_DEBUG:
854 v.val = sock_flag(sk, SOCK_DBG);
855 break;
856
857 case SO_DONTROUTE:
858 v.val = sock_flag(sk, SOCK_LOCALROUTE);
859 break;
860
861 case SO_BROADCAST:
862 v.val = !!sock_flag(sk, SOCK_BROADCAST);
863 break;
864
865 case SO_SNDBUF:
866 v.val = sk->sk_sndbuf;
867 break;
868
869 case SO_RCVBUF:
870 v.val = sk->sk_rcvbuf;
871 break;
872
873 case SO_REUSEADDR:
874 v.val = sk->sk_reuse;
875 break;
876
877 case SO_KEEPALIVE:
878 v.val = !!sock_flag(sk, SOCK_KEEPOPEN);
879 break;
880
881 case SO_TYPE:
882 v.val = sk->sk_type;
883 break;
884
49c794e9
JE		 885 case SO_PROTOCOL:
886 v.val = sk->sk_protocol;
887 break;
888
0d6038ee
JE		 889 case SO_DOMAIN:
890 v.val = sk->sk_family;
891 break;
892
e71a4783
SH		 893 case SO_ERROR:
894 v.val = -sock_error(sk);
2a91525c 895 if (v.val == 0)
e71a4783
SH		 896 v.val = xchg(&sk->sk_err_soft, 0);
897 break;
898
899 case SO_OOBINLINE:
900 v.val = !!sock_flag(sk, SOCK_URGINLINE);
901 break;
902
903 case SO_NO_CHECK:
904 v.val = sk->sk_no_check;
905 break;
906
907 case SO_PRIORITY:
908 v.val = sk->sk_priority;
909 break;
910
911 case SO_LINGER:
912 lv = sizeof(v.ling);
913 v.ling.l_onoff = !!sock_flag(sk, SOCK_LINGER);
914 v.ling.l_linger = sk->sk_lingertime / HZ;
915 break;
916
917 case SO_BSDCOMPAT:
918 sock_warn_obsolete_bsdism("getsockopt");
919 break;
920
921 case SO_TIMESTAMP:
92f37fd2
ED		 922 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
923 !sock_flag(sk, SOCK_RCVTSTAMPNS);
924 break;
925
926 case SO_TIMESTAMPNS:
927 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 928 break;
929
20d49473
PO		 930 case SO_TIMESTAMPING:
931 v.val = 0;
932 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
933 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
934 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
935 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
936 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
937 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
938 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
939 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
940 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
941 v.val |= SOF_TIMESTAMPING_SOFTWARE;
942 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
943 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
944 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
945 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
946 break;
947
e71a4783 948 case SO_RCVTIMEO:
2a91525c 949 lv = sizeof(struct timeval);
e71a4783
SH		 950 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
951 v.tm.tv_sec = 0;
952 v.tm.tv_usec = 0;
953 } else {
954 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
955 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
956 }
957 break;
958
959 case SO_SNDTIMEO:
2a91525c 960 lv = sizeof(struct timeval);
e71a4783
SH		 961 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
962 v.tm.tv_sec = 0;
963 v.tm.tv_usec = 0;
964 } else {
965 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
966 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
967 }
968 break;
1da177e4 969
e71a4783
SH		 970 case SO_RCVLOWAT:
971 v.val = sk->sk_rcvlowat;
972 break;
1da177e4 973
e71a4783 974 case SO_SNDLOWAT:
2a91525c 975 v.val = 1;
e71a4783 976 break;
1da177e4 977
e71a4783
SH		 978 case SO_PASSCRED:
979 v.val = test_bit(SOCK_PASSCRED, &sock->flags) ? 1 : 0;
980 break;
1da177e4 981
e71a4783 982 case SO_PEERCRED:
109f6e39
EB		 983 {
984 struct ucred peercred;
985 if (len > sizeof(peercred))
986 len = sizeof(peercred);
987 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
988 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 989 return -EFAULT;
990 goto lenout;
109f6e39 991 }
1da177e4 992
e71a4783
SH		 993 case SO_PEERNAME:
994 {
995 char address[128];
996
997 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
998 return -ENOTCONN;
999 if (lv < len)
1000 return -EINVAL;
1001 if (copy_to_user(optval, address, len))
1002 return -EFAULT;
1003 goto lenout;
1004 }
1da177e4 1005
e71a4783
SH		 1006 /* Dubious BSD thing... Probably nobody even uses it, but
1007 * the UNIX standard wants it for whatever reason... -DaveM
1008 */
1009 case SO_ACCEPTCONN:
1010 v.val = sk->sk_state == TCP_LISTEN;
1011 break;
1da177e4 1012
e71a4783
SH		 1013 case SO_PASSSEC:
1014 v.val = test_bit(SOCK_PASSSEC, &sock->flags) ? 1 : 0;
1015 break;
877ce7c1 1016
e71a4783
SH		 1017 case SO_PEERSEC:
1018 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1019
4a19ec58
LAT		 1020 case SO_MARK:
1021 v.val = sk->sk_mark;
1022 break;
1023
3b885787
NH		 1024 case SO_RXQ_OVFL:
1025 v.val = !!sock_flag(sk, SOCK_RXQ_OVFL);
1026 break;
1027
6e3e939f
JB		 1028 case SO_WIFI_STATUS:
1029 v.val = !!sock_flag(sk, SOCK_WIFI_STATUS);
1030 break;
1031
ef64a54f
PE		 1032 case SO_PEEK_OFF:
1033 if (!sock->ops->set_peek_off)
1034 return -EOPNOTSUPP;
1035
1036 v.val = sk->sk_peek_off;
1037 break;
e71a4783
SH		 1038 default:
1039 return -ENOPROTOOPT;
1da177e4 1040 }
e71a4783 1041
1da177e4
LT		 1042 if (len > lv)
1043 len = lv;
1044 if (copy_to_user(optval, &v, len))
1045 return -EFAULT;
1046lenout:
4ec93edb
YH		 1047 if (put_user(len, optlen))
1048 return -EFAULT;
1049 return 0;
1da177e4
LT		 1050}
1051
a5b5bb9a
IM		 1052/*
1053 * Initialize an sk_lock.
1054 *
1055 * (We also register the sk_lock with the lock validator.)
1056 */
b6f99a21 1057static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1058{
ed07536e
PZ		 1059 sock_lock_init_class_and_name(sk,
1060 af_family_slock_key_strings[sk->sk_family],
1061 af_family_slock_keys + sk->sk_family,
1062 af_family_key_strings[sk->sk_family],
1063 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1064}
1065
4dc6dc71
ED		 1066/*
1067 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1068 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1069 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1070 */
f1a6c4da
PE		 1071static void sock_copy(struct sock *nsk, const struct sock *osk)
1072{
1073#ifdef CONFIG_SECURITY_NETWORK
1074 void *sptr = nsk->sk_security;
1075#endif
68835aba
ED		 1076 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1077
1078 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1079 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1080
f1a6c4da
PE		 1081#ifdef CONFIG_SECURITY_NETWORK
1082 nsk->sk_security = sptr;
1083 security_sk_clone(osk, nsk);
1084#endif
1085}
1086
fcbdf09d
OP		 1087/*
1088 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
1089 * un-modified. Special care is taken when initializing object to zero.
1090 */
1091static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1092{
1093 if (offsetof(struct sock, sk_node.next) != 0)
1094 memset(sk, 0, offsetof(struct sock, sk_node.next));
1095 memset(&sk->sk_node.pprev, 0,
1096 size - offsetof(struct sock, sk_node.pprev));
1097}
1098
1099void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1100{
1101 unsigned long nulls1, nulls2;
1102
1103 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1104 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1105 if (nulls1 > nulls2)
1106 swap(nulls1, nulls2);
1107
1108 if (nulls1 != 0)
1109 memset((char *)sk, 0, nulls1);
1110 memset((char *)sk + nulls1 + sizeof(void *), 0,
1111 nulls2 - nulls1 - sizeof(void *));
1112 memset((char *)sk + nulls2 + sizeof(void *), 0,
1113 size - nulls2 - sizeof(void *));
1114}
1115EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1116
2e4afe7b
PE		 1117static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1118 int family)
c308c1b2
PE		 1119{
1120 struct sock *sk;
1121 struct kmem_cache *slab;
1122
1123 slab = prot->slab;
e912b114
ED		 1124 if (slab != NULL) {
1125 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1126 if (!sk)
1127 return sk;
1128 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1129 if (prot->clear_sk)
1130 prot->clear_sk(sk, prot->obj_size);
1131 else
1132 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1133 }
fcbdf09d 1134 } else
c308c1b2
PE		 1135 sk = kmalloc(prot->obj_size, priority);
1136
2e4afe7b 1137 if (sk != NULL) {
a98b65a3
VN		 1138 kmemcheck_annotate_bitfield(sk, flags);
1139
2e4afe7b
PE		 1140 if (security_sk_alloc(sk, family, priority))
1141 goto out_free;
1142
1143 if (!try_module_get(prot->owner))
1144 goto out_free_sec;
e022f0b4 1145 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1146 }
1147
c308c1b2 1148 return sk;
2e4afe7b
PE		 1149
1150out_free_sec:
1151 security_sk_free(sk);
1152out_free:
1153 if (slab != NULL)
1154 kmem_cache_free(slab, sk);
1155 else
1156 kfree(sk);
1157 return NULL;
c308c1b2
PE		 1158}
1159
1160static void sk_prot_free(struct proto *prot, struct sock *sk)
1161{
1162 struct kmem_cache *slab;
2e4afe7b 1163 struct module *owner;
c308c1b2 1164
2e4afe7b 1165 owner = prot->owner;
c308c1b2 1166 slab = prot->slab;
2e4afe7b
PE		 1167
1168 security_sk_free(sk);
c308c1b2
PE		 1169 if (slab != NULL)
1170 kmem_cache_free(slab, sk);
1171 else
1172 kfree(sk);
2e4afe7b 1173 module_put(owner);
c308c1b2
PE		 1174}
1175
f8451725
HX		 1176#ifdef CONFIG_CGROUPS
1177void sock_update_classid(struct sock *sk)
1178{
1144182a 1179 u32 classid;
f8451725 1180
1144182a
PM		 1181 rcu_read_lock(); /* doing current task, which cannot vanish. */
1182 classid = task_cls_classid(current);
1183 rcu_read_unlock();
f8451725
HX		 1184 if (classid && classid != sk->sk_classid)
1185 sk->sk_classid = classid;
1186}
82862742 1187EXPORT_SYMBOL(sock_update_classid);
5bc1421e
NH		 1188
1189void sock_update_netprioidx(struct sock *sk)
1190{
5bc1421e
NH		 1191 if (in_interrupt())
1192 return;
2b73bc65
NH		 1193
1194 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1195}
1196EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1197#endif
1198
1da177e4
LT		 1199/**
1200 * sk_alloc - All socket objects are allocated here
c4ea43c5 1201 * @net: the applicable net namespace
4dc3b16b
PP		 1202 * @family: protocol family
1203 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1204 * @prot: struct proto associated with this new sock instance
1da177e4 1205 */
1b8d7ae4 1206struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1207 struct proto *prot)
1da177e4 1208{
c308c1b2 1209 struct sock *sk;
1da177e4 1210
154adbc8 1211 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1212 if (sk) {
154adbc8
PE		 1213 sk->sk_family = family;
1214 /*
1215 * See comment in struct sock definition to understand
1216 * why we need sk_prot_creator -acme
1217 */
1218 sk->sk_prot = sk->sk_prot_creator = prot;
1219 sock_lock_init(sk);
3b1e0a65 1220 sock_net_set(sk, get_net(net));
d66ee058 1221 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725
HX		 1222
1223 sock_update_classid(sk);
5bc1421e 1224 sock_update_netprioidx(sk);
1da177e4 1225 }
a79af59e 1226
2e4afe7b 1227 return sk;
1da177e4 1228}
2a91525c 1229EXPORT_SYMBOL(sk_alloc);
1da177e4 1230
2b85a34e 1231static void __sk_free(struct sock *sk)
1da177e4
LT		 1232{
1233 struct sk_filter *filter;
1da177e4
LT		 1234
1235 if (sk->sk_destruct)
1236 sk->sk_destruct(sk);
1237
a898def2
PM		 1238 filter = rcu_dereference_check(sk->sk_filter,
1239 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1240 if (filter) {
309dd5fc 1241 sk_filter_uncharge(sk, filter);
a9b3cd7f 1242 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1243 }
1244
08e29af3 1245 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1246
1247 if (atomic_read(&sk->sk_omem_alloc))
1248 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n",
0dc47877 1249 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1250
109f6e39
EB		 1251 if (sk->sk_peer_cred)
1252 put_cred(sk->sk_peer_cred);
1253 put_pid(sk->sk_peer_pid);
3b1e0a65 1254 put_net(sock_net(sk));
c308c1b2 1255 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1256}
2b85a34e
ED		 1257
1258void sk_free(struct sock *sk)
1259{
1260 /*
25985edc 1261 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1262 * some packets are still in some tx queue.
1263 * If not null, sock_wfree() will call __sk_free(sk) later
1264 */
1265 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1266 __sk_free(sk);
1267}
2a91525c 1268EXPORT_SYMBOL(sk_free);
1da177e4 1269
edf02087 1270/*
25985edc
LDM		 1271 * Last sock_put should drop reference to sk->sk_net. It has already
1272 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1273 * is not an option.
25985edc 1274 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1275 * destroy it in the context of init_net.
1276 */
1277void sk_release_kernel(struct sock *sk)
1278{
1279 if (sk == NULL || sk->sk_socket == NULL)
1280 return;
1281
1282 sock_hold(sk);
1283 sock_release(sk->sk_socket);
65a18ec5 1284 release_net(sock_net(sk));
3b1e0a65 1285 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1286 sock_put(sk);
1287}
45af1754 1288EXPORT_SYMBOL(sk_release_kernel);
edf02087 1289
475f1b52
SR		 1290static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1291{
1292 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1293 sock_update_memcg(newsk);
1294}
1295
e56c57d0
ED		 1296/**
1297 * sk_clone_lock - clone a socket, and lock its clone
1298 * @sk: the socket to clone
1299 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1300 *
1301 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1302 */
1303struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1304{
8fd1d178 1305 struct sock *newsk;
87d11ceb 1306
8fd1d178 1307 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1308 if (newsk != NULL) {
1309 struct sk_filter *filter;
1310
892c141e 1311 sock_copy(newsk, sk);
87d11ceb
ACM		 1312
1313 /* SANITY */
3b1e0a65 1314 get_net(sock_net(newsk));
87d11ceb
ACM		 1315 sk_node_init(&newsk->sk_node);
1316 sock_lock_init(newsk);
1317 bh_lock_sock(newsk);
fa438ccf 1318 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1319 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1320
1321 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1322 /*
1323 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1324 */
1325 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1326 atomic_set(&newsk->sk_omem_alloc, 0);
1327 skb_queue_head_init(&newsk->sk_receive_queue);
1328 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1329#ifdef CONFIG_NET_DMA
1330 skb_queue_head_init(&newsk->sk_async_wait_queue);
1331#endif
87d11ceb 1332
b6c6712a 1333 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1334 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1335 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1336 af_callback_keys + newsk->sk_family,
1337 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1338
1339 newsk->sk_dst_cache = NULL;
1340 newsk->sk_wmem_queued = 0;
1341 newsk->sk_forward_alloc = 0;
1342 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1343 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1344
1345 sock_reset_flag(newsk, SOCK_DONE);
1346 skb_queue_head_init(&newsk->sk_error_queue);
1347
0d7da9dd 1348 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1349 if (filter != NULL)
1350 sk_filter_charge(newsk, filter);
1351
1352 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1353 /* It is still raw copy of parent, so invalidate
1354 * destructor and make plain sk_free() */
1355 newsk->sk_destruct = NULL;
b0691c8e 1356 bh_unlock_sock(newsk);
87d11ceb
ACM		 1357 sk_free(newsk);
1358 newsk = NULL;
1359 goto out;
1360 }
1361
1362 newsk->sk_err = 0;
1363 newsk->sk_priority = 0;
4dc6dc71
ED		 1364 /*
1365 * Before updating sk_refcnt, we must commit prior changes to memory
1366 * (Documentation/RCU/rculist_nulls.txt for details)
1367 */
1368 smp_wmb();
87d11ceb
ACM		 1369 atomic_set(&newsk->sk_refcnt, 2);
1370
1371 /*
1372 * Increment the counter in the same struct proto as the master
1373 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1374 * is the same as sk->sk_prot->socks, as this field was copied
1375 * with memcpy).
1376 *
1377 * This _changes_ the previous behaviour, where
1378 * tcp_create_openreq_child always was incrementing the
1379 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1380 * to be taken into account in all callers. -acme
1381 */
1382 sk_refcnt_debug_inc(newsk);
972692e0 1383 sk_set_socket(newsk, NULL);
43815482 1384 newsk->sk_wq = NULL;
87d11ceb 1385
f3f511e1
GC		 1386 sk_update_clone(sk, newsk);
1387
87d11ceb 1388 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1389 sk_sockets_allocated_inc(newsk);
704da560 1390
08e29af3 1391 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1392 net_enable_timestamp();
87d11ceb
ACM		 1393 }
1394out:
1395 return newsk;
1396}
e56c57d0 1397EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1398
9958089a
AK		 1399void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1400{
1401 __sk_dst_set(sk, dst);
1402 sk->sk_route_caps = dst->dev->features;
1403 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1404 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1405 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1406 if (sk_can_gso(sk)) {
82cc1a7a 1407 if (dst->header_len) {
9958089a 1408 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1409 } else {
9958089a 1410 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a
PWJ		 1411 sk->sk_gso_max_size = dst->dev->gso_max_size;
1412 }
9958089a
AK		 1413 }
1414}
1415EXPORT_SYMBOL_GPL(sk_setup_caps);
1416
1da177e4
LT		 1417void __init sk_init(void)
1418{
4481374c 1419 if (totalram_pages <= 4096) {
1da177e4
LT		 1420 sysctl_wmem_max = 32767;
1421 sysctl_rmem_max = 32767;
1422 sysctl_wmem_default = 32767;
1423 sysctl_rmem_default = 32767;
4481374c 1424 } else if (totalram_pages >= 131072) {
1da177e4
LT		 1425 sysctl_wmem_max = 131071;
1426 sysctl_rmem_max = 131071;
1427 }
1428}
1429
1430/*
1431 * Simple resource managers for sockets.
1432 */
1433
1434
4ec93edb
YH		 1435/*
1436 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1437 */
1438void sock_wfree(struct sk_buff *skb)
1439{
1440 struct sock *sk = skb->sk;
d99927f4 1441 unsigned int len = skb->truesize;
1da177e4 1442
d99927f4
ED		 1443 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1444 /*
1445 * Keep a reference on sk_wmem_alloc, this will be released
1446 * after sk_write_space() call
1447 */
1448 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1449 sk->sk_write_space(sk);
d99927f4
ED		 1450 len = 1;
1451 }
2b85a34e 1452 /*
d99927f4
ED		 1453 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1454 * could not do because of in-flight packets
2b85a34e 1455 */
d99927f4 1456 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1457 __sk_free(sk);
1da177e4 1458}
2a91525c 1459EXPORT_SYMBOL(sock_wfree);
1da177e4 1460
4ec93edb
YH		 1461/*
1462 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1463 */
1464void sock_rfree(struct sk_buff *skb)
1465{
1466 struct sock *sk = skb->sk;
d361fd59 1467 unsigned int len = skb->truesize;
1da177e4 1468
d361fd59
ED		 1469 atomic_sub(len, &sk->sk_rmem_alloc);
1470 sk_mem_uncharge(sk, len);
1da177e4 1471}
2a91525c 1472EXPORT_SYMBOL(sock_rfree);
1da177e4
LT		 1473
1474
1475int sock_i_uid(struct sock *sk)
1476{
1477 int uid;
1478
f064af1e 1479 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1480 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
f064af1e 1481 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1482 return uid;
1483}
2a91525c 1484EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1485
1486unsigned long sock_i_ino(struct sock *sk)
1487{
1488 unsigned long ino;
1489
f064af1e 1490 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1491 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1492 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1493 return ino;
1494}
2a91525c 1495EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1496
1497/*
1498 * Allocate a skb from the socket's send buffer.
1499 */
86a76caf 1500struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1501 gfp_t priority)
1da177e4
LT		 1502{
1503 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1504 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1505 if (skb) {
1506 skb_set_owner_w(skb, sk);
1507 return skb;
1508 }
1509 }
1510 return NULL;
1511}
2a91525c 1512EXPORT_SYMBOL(sock_wmalloc);
1da177e4
LT		 1513
1514/*
1515 * Allocate a skb from the socket's receive buffer.
4ec93edb 1516 */
86a76caf 1517struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1518 gfp_t priority)
1da177e4
LT		 1519{
1520 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1521 struct sk_buff *skb = alloc_skb(size, priority);
1522 if (skb) {
1523 skb_set_owner_r(skb, sk);
1524 return skb;
1525 }
1526 }
1527 return NULL;
1528}
1529
4ec93edb 1530/*
1da177e4 1531 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1532 */
dd0fc66f 1533void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4
LT		 1534{
1535 if ((unsigned)size <= sysctl_optmem_max &&
1536 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1537 void *mem;
1538 /* First do the add, to avoid the race if kmalloc
4ec93edb 1539 * might sleep.
1da177e4
LT		 1540 */
1541 atomic_add(size, &sk->sk_omem_alloc);
1542 mem = kmalloc(size, priority);
1543 if (mem)
1544 return mem;
1545 atomic_sub(size, &sk->sk_omem_alloc);
1546 }
1547 return NULL;
1548}
2a91525c 1549EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1550
1551/*
1552 * Free an option memory block.
1553 */
1554void sock_kfree_s(struct sock *sk, void *mem, int size)
1555{
1556 kfree(mem);
1557 atomic_sub(size, &sk->sk_omem_alloc);
1558}
2a91525c 1559EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1560
1561/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1562 I think, these locks should be removed for datagram sockets.
1563 */
2a91525c 1564static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1565{
1566 DEFINE_WAIT(wait);
1567
1568 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1569 for (;;) {
1570 if (!timeo)
1571 break;
1572 if (signal_pending(current))
1573 break;
1574 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1575 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1576 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1577 break;
1578 if (sk->sk_shutdown & SEND_SHUTDOWN)
1579 break;
1580 if (sk->sk_err)
1581 break;
1582 timeo = schedule_timeout(timeo);
1583 }
aa395145 1584 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1585 return timeo;
1586}
1587
1588
1589/*
1590 * Generic send/receive buffer handlers
1591 */
1592
4cc7f68d
HX		 1593struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1594 unsigned long data_len, int noblock,
1595 int *errcode)
1da177e4
LT		 1596{
1597 struct sk_buff *skb;
7d877f3b 1598 gfp_t gfp_mask;
1da177e4
LT		 1599 long timeo;
1600 int err;
1601
1602 gfp_mask = sk->sk_allocation;
1603 if (gfp_mask & __GFP_WAIT)
1604 gfp_mask |= __GFP_REPEAT;
1605
1606 timeo = sock_sndtimeo(sk, noblock);
1607 while (1) {
1608 err = sock_error(sk);
1609 if (err != 0)
1610 goto failure;
1611
1612 err = -EPIPE;
1613 if (sk->sk_shutdown & SEND_SHUTDOWN)
1614 goto failure;
1615
1616 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1617 skb = alloc_skb(header_len, gfp_mask);
1da177e4
LT		 1618 if (skb) {
1619 int npages;
1620 int i;
1621
1622 /* No pages, we're done... */
1623 if (!data_len)
1624 break;
1625
1626 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1627 skb->truesize += data_len;
1628 skb_shinfo(skb)->nr_frags = npages;
1629 for (i = 0; i < npages; i++) {
1630 struct page *page;
1da177e4
LT		 1631
1632 page = alloc_pages(sk->sk_allocation, 0);
1633 if (!page) {
1634 err = -ENOBUFS;
1635 skb_shinfo(skb)->nr_frags = i;
1636 kfree_skb(skb);
1637 goto failure;
1638 }
1639
ea2ab693
IC		 1640 __skb_fill_page_desc(skb, i,
1641 page, 0,
1642 (data_len >= PAGE_SIZE ?
1643 PAGE_SIZE :
1644 data_len));
1da177e4
LT		 1645 data_len -= PAGE_SIZE;
1646 }
1647
1648 /* Full success... */
1649 break;
1650 }
1651 err = -ENOBUFS;
1652 goto failure;
1653 }
1654 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1655 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1656 err = -EAGAIN;
1657 if (!timeo)
1658 goto failure;
1659 if (signal_pending(current))
1660 goto interrupted;
1661 timeo = sock_wait_for_wmem(sk, timeo);
1662 }
1663
1664 skb_set_owner_w(skb, sk);
1665 return skb;
1666
1667interrupted:
1668 err = sock_intr_errno(timeo);
1669failure:
1670 *errcode = err;
1671 return NULL;
1672}
4cc7f68d 1673EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1674
4ec93edb 1675struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1676 int noblock, int *errcode)
1677{
1678 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1679}
2a91525c 1680EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4
LT		 1681
1682static void __lock_sock(struct sock *sk)
f39234d6
NK		 1683 __releases(&sk->sk_lock.slock)
1684 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1685{
1686 DEFINE_WAIT(wait);
1687
e71a4783 1688 for (;;) {
1da177e4
LT		 1689 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1690 TASK_UNINTERRUPTIBLE);
1691 spin_unlock_bh(&sk->sk_lock.slock);
1692 schedule();
1693 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1694 if (!sock_owned_by_user(sk))
1da177e4
LT		 1695 break;
1696 }
1697 finish_wait(&sk->sk_lock.wq, &wait);
1698}
1699
1700static void __release_sock(struct sock *sk)
f39234d6
NK		 1701 __releases(&sk->sk_lock.slock)
1702 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1703{
1704 struct sk_buff *skb = sk->sk_backlog.head;
1705
1706 do {
1707 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1708 bh_unlock_sock(sk);
1709
1710 do {
1711 struct sk_buff *next = skb->next;
1712
7fee226a 1713 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1714 skb->next = NULL;
c57943a1 1715 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1716
1717 /*
1718 * We are in process context here with softirqs
1719 * disabled, use cond_resched_softirq() to preempt.
1720 * This is safe to do because we've taken the backlog
1721 * queue private:
1722 */
1723 cond_resched_softirq();
1724
1725 skb = next;
1726 } while (skb != NULL);
1727
1728 bh_lock_sock(sk);
e71a4783 1729 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1730
1731 /*
1732 * Doing the zeroing here guarantee we can not loop forever
1733 * while a wild producer attempts to flood us.
1734 */
1735 sk->sk_backlog.len = 0;
1da177e4
LT		 1736}
1737
1738/**
1739 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1740 * @sk: sock to wait on
1741 * @timeo: for how long
1da177e4
LT		 1742 *
1743 * Now socket state including sk->sk_err is changed only under lock,
1744 * hence we may omit checks after joining wait queue.
1745 * We check receive queue before schedule() only as optimization;
1746 * it is very likely that release_sock() added new data.
1747 */
1748int sk_wait_data(struct sock *sk, long *timeo)
1749{
1750 int rc;
1751 DEFINE_WAIT(wait);
1752
aa395145 1753 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1754 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1755 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1756 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1757 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1758 return rc;
1759}
1da177e4
LT		 1760EXPORT_SYMBOL(sk_wait_data);
1761
3ab224be
HA		 1762/**
1763 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1764 * @sk: socket
1765 * @size: memory size to allocate
1766 * @kind: allocation type
1767 *
1768 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1769 * rmem allocation. This function assumes that protocols which have
1770 * memory_pressure use sk_wmem_queued as write buffer accounting.
1771 */
1772int __sk_mem_schedule(struct sock *sk, int size, int kind)
1773{
1774 struct proto *prot = sk->sk_prot;
1775 int amt = sk_mem_pages(size);
8d987e5c 1776 long allocated;
e1aab161 1777 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1778
1779 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1780
e1aab161 1781 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1782
1783 /* Under limit. */
e1aab161
GC		 1784 if (parent_status == UNDER_LIMIT &&
1785 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1786 sk_leave_memory_pressure(sk);
3ab224be
HA		 1787 return 1;
1788 }
1789
e1aab161
GC		 1790 /* Under pressure. (we or our parents) */
1791 if ((parent_status > SOFT_LIMIT) ||
1792 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1793 sk_enter_memory_pressure(sk);
3ab224be 1794
e1aab161
GC		 1795 /* Over hard limit (we or our parents) */
1796 if ((parent_status == OVER_LIMIT) ||
1797 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1798 goto suppress_allocation;
1799
1800 /* guarantee minimum buffer size under pressure */
1801 if (kind == SK_MEM_RECV) {
1802 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1803 return 1;
180d8cd9 1804
3ab224be
HA		 1805 } else { /* SK_MEM_SEND */
1806 if (sk->sk_type == SOCK_STREAM) {
1807 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1808 return 1;
1809 } else if (atomic_read(&sk->sk_wmem_alloc) <
1810 prot->sysctl_wmem[0])
1811 return 1;
1812 }
1813
180d8cd9 1814 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 1815 int alloc;
1816
180d8cd9 1817 if (!sk_under_memory_pressure(sk))
1748376b 1818 return 1;
180d8cd9
GC		 1819 alloc = sk_sockets_allocated_read_positive(sk);
1820 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 1821 sk_mem_pages(sk->sk_wmem_queued +
1822 atomic_read(&sk->sk_rmem_alloc) +
1823 sk->sk_forward_alloc))
1824 return 1;
1825 }
1826
1827suppress_allocation:
1828
1829 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1830 sk_stream_moderate_sndbuf(sk);
1831
1832 /* Fail only if socket is _under_ its sndbuf.
1833 * In this case we cannot block, so that we have to fail.
1834 */
1835 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1836 return 1;
1837 }
1838
3847ce32
SM		 1839 trace_sock_exceed_buf_limit(sk, prot, allocated);
1840
3ab224be
HA		 1841 /* Alas. Undo changes. */
1842 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 1843
0e90b31f 1844 sk_memory_allocated_sub(sk, amt);
180d8cd9 1845
3ab224be
HA		 1846 return 0;
1847}
3ab224be
HA		 1848EXPORT_SYMBOL(__sk_mem_schedule);
1849
1850/**
1851 * __sk_reclaim - reclaim memory_allocated
1852 * @sk: socket
1853 */
1854void __sk_mem_reclaim(struct sock *sk)
1855{
180d8cd9 1856 sk_memory_allocated_sub(sk,
0e90b31f 1857 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 1858 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
1859
180d8cd9
GC		 1860 if (sk_under_memory_pressure(sk) &&
1861 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
1862 sk_leave_memory_pressure(sk);
3ab224be 1863}
3ab224be
HA		 1864EXPORT_SYMBOL(__sk_mem_reclaim);
1865
1866
1da177e4
LT		 1867/*
1868 * Set of default routines for initialising struct proto_ops when
1869 * the protocol does not support a particular function. In certain
1870 * cases where it makes no sense for a protocol to have a "do nothing"
1871 * function, some default processing is provided.
1872 */
1873
1874int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1875{
1876 return -EOPNOTSUPP;
1877}
2a91525c 1878EXPORT_SYMBOL(sock_no_bind);
1da177e4 1879
4ec93edb 1880int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1881 int len, int flags)
1882{
1883 return -EOPNOTSUPP;
1884}
2a91525c 1885EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 1886
1887int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1888{
1889 return -EOPNOTSUPP;
1890}
2a91525c 1891EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 1892
1893int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1894{
1895 return -EOPNOTSUPP;
1896}
2a91525c 1897EXPORT_SYMBOL(sock_no_accept);
1da177e4 1898
4ec93edb 1899int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1900 int *len, int peer)
1901{
1902 return -EOPNOTSUPP;
1903}
2a91525c 1904EXPORT_SYMBOL(sock_no_getname);
1da177e4 1905
2a91525c 1906unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 1907{
1908 return 0;
1909}
2a91525c 1910EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 1911
1912int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1913{
1914 return -EOPNOTSUPP;
1915}
2a91525c 1916EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 1917
1918int sock_no_listen(struct socket *sock, int backlog)
1919{
1920 return -EOPNOTSUPP;
1921}
2a91525c 1922EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 1923
1924int sock_no_shutdown(struct socket *sock, int how)
1925{
1926 return -EOPNOTSUPP;
1927}
2a91525c 1928EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 1929
1930int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 1931 char __user *optval, unsigned int optlen)
1da177e4
LT		 1932{
1933 return -EOPNOTSUPP;
1934}
2a91525c 1935EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 1936
1937int sock_no_getsockopt(struct socket *sock, int level, int optname,
1938 char __user *optval, int __user *optlen)
1939{
1940 return -EOPNOTSUPP;
1941}
2a91525c 1942EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 1943
1944int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1945 size_t len)
1946{
1947 return -EOPNOTSUPP;
1948}
2a91525c 1949EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 1950
1951int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1952 size_t len, int flags)
1953{
1954 return -EOPNOTSUPP;
1955}
2a91525c 1956EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 1957
1958int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
1959{
1960 /* Mirror missing mmap method error code */
1961 return -ENODEV;
1962}
2a91525c 1963EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 1964
1965ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
1966{
1967 ssize_t res;
1968 struct msghdr msg = {.msg_flags = flags};
1969 struct kvec iov;
1970 char *kaddr = kmap(page);
1971 iov.iov_base = kaddr + offset;
1972 iov.iov_len = size;
1973 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
1974 kunmap(page);
1975 return res;
1976}
2a91525c 1977EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 1978
1979/*
1980 * Default Socket Callbacks
1981 */
1982
1983static void sock_def_wakeup(struct sock *sk)
1984{
43815482
ED		 1985 struct socket_wq *wq;
1986
1987 rcu_read_lock();
1988 wq = rcu_dereference(sk->sk_wq);
1989 if (wq_has_sleeper(wq))
1990 wake_up_interruptible_all(&wq->wait);
1991 rcu_read_unlock();
1da177e4
LT		 1992}
1993
1994static void sock_def_error_report(struct sock *sk)
1995{
43815482
ED		 1996 struct socket_wq *wq;
1997
1998 rcu_read_lock();
1999 wq = rcu_dereference(sk->sk_wq);
2000 if (wq_has_sleeper(wq))
2001 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2002 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2003 rcu_read_unlock();
1da177e4
LT		 2004}
2005
2006static void sock_def_readable(struct sock *sk, int len)
2007{
43815482
ED		 2008 struct socket_wq *wq;
2009
2010 rcu_read_lock();
2011 wq = rcu_dereference(sk->sk_wq);
2012 if (wq_has_sleeper(wq))
2c6607c6 2013 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2014 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2015 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2016 rcu_read_unlock();
1da177e4
LT		 2017}
2018
2019static void sock_def_write_space(struct sock *sk)
2020{
43815482
ED		 2021 struct socket_wq *wq;
2022
2023 rcu_read_lock();
1da177e4
LT		 2024
2025 /* Do not wake up a writer until he can make "significant"
2026 * progress. --DaveM
2027 */
e71a4783 2028 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2029 wq = rcu_dereference(sk->sk_wq);
2030 if (wq_has_sleeper(wq))
2031 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2032 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2033
2034 /* Should agree with poll, otherwise some programs break */
2035 if (sock_writeable(sk))
8d8ad9d7 2036 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2037 }
2038
43815482 2039 rcu_read_unlock();
1da177e4
LT		 2040}
2041
2042static void sock_def_destruct(struct sock *sk)
2043{
a51482bd 2044 kfree(sk->sk_protinfo);
1da177e4
LT		 2045}
2046
2047void sk_send_sigurg(struct sock *sk)
2048{
2049 if (sk->sk_socket && sk->sk_socket->file)
2050 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2051 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2052}
2a91525c 2053EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2054
2055void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2056 unsigned long expires)
2057{
2058 if (!mod_timer(timer, expires))
2059 sock_hold(sk);
2060}
1da177e4
LT		 2061EXPORT_SYMBOL(sk_reset_timer);
2062
2063void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2064{
2065 if (timer_pending(timer) && del_timer(timer))
2066 __sock_put(sk);
2067}
1da177e4
LT		 2068EXPORT_SYMBOL(sk_stop_timer);
2069
2070void sock_init_data(struct socket *sock, struct sock *sk)
2071{
2072 skb_queue_head_init(&sk->sk_receive_queue);
2073 skb_queue_head_init(&sk->sk_write_queue);
2074 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2075#ifdef CONFIG_NET_DMA
2076 skb_queue_head_init(&sk->sk_async_wait_queue);
2077#endif
1da177e4
LT		 2078
2079 sk->sk_send_head = NULL;
2080
2081 init_timer(&sk->sk_timer);
4ec93edb 2082
1da177e4
LT		 2083 sk->sk_allocation = GFP_KERNEL;
2084 sk->sk_rcvbuf = sysctl_rmem_default;
2085 sk->sk_sndbuf = sysctl_wmem_default;
2086 sk->sk_state = TCP_CLOSE;
972692e0 2087 sk_set_socket(sk, sock);
1da177e4
LT		 2088
2089 sock_set_flag(sk, SOCK_ZAPPED);
2090
e71a4783 2091 if (sock) {
1da177e4 2092 sk->sk_type = sock->type;
43815482 2093 sk->sk_wq = sock->wq;
1da177e4
LT		 2094 sock->sk = sk;
2095 } else
43815482 2096 sk->sk_wq = NULL;
1da177e4 2097
b6c6712a 2098 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2099 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2100 lockdep_set_class_and_name(&sk->sk_callback_lock,
2101 af_callback_keys + sk->sk_family,
2102 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2103
2104 sk->sk_state_change = sock_def_wakeup;
2105 sk->sk_data_ready = sock_def_readable;
2106 sk->sk_write_space = sock_def_write_space;
2107 sk->sk_error_report = sock_def_error_report;
2108 sk->sk_destruct = sock_def_destruct;
2109
2110 sk->sk_sndmsg_page = NULL;
2111 sk->sk_sndmsg_off = 0;
ef64a54f 2112 sk->sk_peek_off = -1;
1da177e4 2113
109f6e39
EB		 2114 sk->sk_peer_pid = NULL;
2115 sk->sk_peer_cred = NULL;
1da177e4
LT		 2116 sk->sk_write_pending = 0;
2117 sk->sk_rcvlowat = 1;
2118 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2119 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2120
f37f0afb 2121 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2122
4dc6dc71
ED		 2123 /*
2124 * Before updating sk_refcnt, we must commit prior changes to memory
2125 * (Documentation/RCU/rculist_nulls.txt for details)
2126 */
2127 smp_wmb();
1da177e4 2128 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2129 atomic_set(&sk->sk_drops, 0);
1da177e4 2130}
2a91525c 2131EXPORT_SYMBOL(sock_init_data);
1da177e4 2132
b5606c2d 2133void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2134{
2135 might_sleep();
a5b5bb9a 2136 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2137 if (sk->sk_lock.owned)
1da177e4 2138 __lock_sock(sk);
d2e9117c 2139 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2140 spin_unlock(&sk->sk_lock.slock);
2141 /*
2142 * The sk_lock has mutex_lock() semantics here:
2143 */
fcc70d5f 2144 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2145 local_bh_enable();
1da177e4 2146}
fcc70d5f 2147EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2148
b5606c2d 2149void release_sock(struct sock *sk)
1da177e4 2150{
a5b5bb9a
IM		 2151 /*
2152 * The sk_lock has mutex_unlock() semantics:
2153 */
2154 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2155
2156 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2157 if (sk->sk_backlog.tail)
2158 __release_sock(sk);
d2e9117c 2159 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 2160 if (waitqueue_active(&sk->sk_lock.wq))
2161 wake_up(&sk->sk_lock.wq);
2162 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2163}
2164EXPORT_SYMBOL(release_sock);
2165
8a74ad60
ED		 2166/**
2167 * lock_sock_fast - fast version of lock_sock
2168 * @sk: socket
2169 *
2170 * This version should be used for very small section, where process wont block
2171 * return false if fast path is taken
2172 * sk_lock.slock locked, owned = 0, BH disabled
2173 * return true if slow path is taken
2174 * sk_lock.slock unlocked, owned = 1, BH enabled
2175 */
2176bool lock_sock_fast(struct sock *sk)
2177{
2178 might_sleep();
2179 spin_lock_bh(&sk->sk_lock.slock);
2180
2181 if (!sk->sk_lock.owned)
2182 /*
2183 * Note : We must disable BH
2184 */
2185 return false;
2186
2187 __lock_sock(sk);
2188 sk->sk_lock.owned = 1;
2189 spin_unlock(&sk->sk_lock.slock);
2190 /*
2191 * The sk_lock has mutex_lock() semantics here:
2192 */
2193 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2194 local_bh_enable();
2195 return true;
2196}
2197EXPORT_SYMBOL(lock_sock_fast);
2198
1da177e4 2199int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2200{
b7aa0bf7 2201 struct timeval tv;
1da177e4 2202 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2203 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2204 tv = ktime_to_timeval(sk->sk_stamp);
2205 if (tv.tv_sec == -1)
1da177e4 2206 return -ENOENT;
b7aa0bf7
ED		 2207 if (tv.tv_sec == 0) {
2208 sk->sk_stamp = ktime_get_real();
2209 tv = ktime_to_timeval(sk->sk_stamp);
2210 }
2211 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2212}
1da177e4
LT		 2213EXPORT_SYMBOL(sock_get_timestamp);
2214
ae40eb1e
ED		 2215int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2216{
2217 struct timespec ts;
2218 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2219 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2220 ts = ktime_to_timespec(sk->sk_stamp);
2221 if (ts.tv_sec == -1)
2222 return -ENOENT;
2223 if (ts.tv_sec == 0) {
2224 sk->sk_stamp = ktime_get_real();
2225 ts = ktime_to_timespec(sk->sk_stamp);
2226 }
2227 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2228}
2229EXPORT_SYMBOL(sock_get_timestampns);
2230
20d49473 2231void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2232{
20d49473 2233 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2234 unsigned long previous_flags = sk->sk_flags;
2235
20d49473
PO		 2236 sock_set_flag(sk, flag);
2237 /*
2238 * we just set one of the two flags which require net
2239 * time stamping, but time stamping might have been on
2240 * already because of the other one
2241 */
08e29af3 2242 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2243 net_enable_timestamp();
1da177e4
LT		 2244 }
2245}
1da177e4
LT		 2246
2247/*
2248 * Get a socket option on an socket.
2249 *
2250 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2251 * asynchronous errors should be reported by getsockopt. We assume
2252 * this means if you specify SO_ERROR (otherwise whats the point of it).
2253 */
2254int sock_common_getsockopt(struct socket *sock, int level, int optname,
2255 char __user *optval, int __user *optlen)
2256{
2257 struct sock *sk = sock->sk;
2258
2259 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2260}
1da177e4
LT		 2261EXPORT_SYMBOL(sock_common_getsockopt);
2262
3fdadf7d 2263#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2264int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2265 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2266{
2267 struct sock *sk = sock->sk;
2268
1e51f951 2269 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2270 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2271 optval, optlen);
3fdadf7d
DM		 2272 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2273}
2274EXPORT_SYMBOL(compat_sock_common_getsockopt);
2275#endif
2276
1da177e4
LT		 2277int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2278 struct msghdr *msg, size_t size, int flags)
2279{
2280 struct sock *sk = sock->sk;
2281 int addr_len = 0;
2282 int err;
2283
2284 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2285 flags & ~MSG_DONTWAIT, &addr_len);
2286 if (err >= 0)
2287 msg->msg_namelen = addr_len;
2288 return err;
2289}
1da177e4
LT		 2290EXPORT_SYMBOL(sock_common_recvmsg);
2291
2292/*
2293 * Set socket options on an inet socket.
2294 */
2295int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2296 char __user *optval, unsigned int optlen)
1da177e4
LT		 2297{
2298 struct sock *sk = sock->sk;
2299
2300 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2301}
1da177e4
LT		 2302EXPORT_SYMBOL(sock_common_setsockopt);
2303
3fdadf7d 2304#ifdef CONFIG_COMPAT
543d9cfe 2305int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2306 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2307{
2308 struct sock *sk = sock->sk;
2309
543d9cfe
ACM		 2310 if (sk->sk_prot->compat_setsockopt != NULL)
2311 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2312 optval, optlen);
3fdadf7d
DM		 2313 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2314}
2315EXPORT_SYMBOL(compat_sock_common_setsockopt);
2316#endif
2317
1da177e4
LT		 2318void sk_common_release(struct sock *sk)
2319{
2320 if (sk->sk_prot->destroy)
2321 sk->sk_prot->destroy(sk);
2322
2323 /*
2324 * Observation: when sock_common_release is called, processes have
2325 * no access to socket. But net still has.
2326 * Step one, detach it from networking:
2327 *
2328 * A. Remove from hash tables.
2329 */
2330
2331 sk->sk_prot->unhash(sk);
2332
2333 /*
2334 * In this point socket cannot receive new packets, but it is possible
2335 * that some packets are in flight because some CPU runs receiver and
2336 * did hash table lookup before we unhashed socket. They will achieve
2337 * receive queue and will be purged by socket destructor.
2338 *
2339 * Also we still have packets pending on receive queue and probably,
2340 * our own packets waiting in device queues. sock_destroy will drain
2341 * receive queue, but transmitted packets will delay socket destruction
2342 * until the last reference will be released.
2343 */
2344
2345 sock_orphan(sk);
2346
2347 xfrm_sk_free_policy(sk);
2348
e6848976 2349 sk_refcnt_debug_release(sk);
1da177e4
LT		 2350 sock_put(sk);
2351}
1da177e4
LT		 2352EXPORT_SYMBOL(sk_common_release);
2353
13ff3d6f
PE		 2354#ifdef CONFIG_PROC_FS
2355#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2356struct prot_inuse {
2357 int val[PROTO_INUSE_NR];
2358};
13ff3d6f
PE		 2359
2360static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2361
2362#ifdef CONFIG_NET_NS
2363void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2364{
d6d9ca0f 2365 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2366}
2367EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2368
2369int sock_prot_inuse_get(struct net *net, struct proto *prot)
2370{
2371 int cpu, idx = prot->inuse_idx;
2372 int res = 0;
2373
2374 for_each_possible_cpu(cpu)
2375 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2376
2377 return res >= 0 ? res : 0;
2378}
2379EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2380
2c8c1e72 2381static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2382{
2383 net->core.inuse = alloc_percpu(struct prot_inuse);
2384 return net->core.inuse ? 0 : -ENOMEM;
2385}
2386
2c8c1e72 2387static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2388{
2389 free_percpu(net->core.inuse);
2390}
2391
2392static struct pernet_operations net_inuse_ops = {
2393 .init = sock_inuse_init_net,
2394 .exit = sock_inuse_exit_net,
2395};
2396
2397static __init int net_inuse_init(void)
2398{
2399 if (register_pernet_subsys(&net_inuse_ops))
2400 panic("Cannot initialize net inuse counters");
2401
2402 return 0;
2403}
2404
2405core_initcall(net_inuse_init);
2406#else
1338d466
PE		 2407static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2408
c29a0bc4 2409void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2410{
d6d9ca0f 2411 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2412}
2413EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2414
c29a0bc4 2415int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2416{
2417 int cpu, idx = prot->inuse_idx;
2418 int res = 0;
2419
2420 for_each_possible_cpu(cpu)
2421 res += per_cpu(prot_inuse, cpu).val[idx];
2422
2423 return res >= 0 ? res : 0;
2424}
2425EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2426#endif
13ff3d6f
PE		 2427
2428static void assign_proto_idx(struct proto *prot)
2429{
2430 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2431
2432 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
2433 printk(KERN_ERR "PROTO_INUSE_NR exhausted\n");
2434 return;
2435 }
2436
2437 set_bit(prot->inuse_idx, proto_inuse_idx);
2438}
2439
2440static void release_proto_idx(struct proto *prot)
2441{
2442 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2443 clear_bit(prot->inuse_idx, proto_inuse_idx);
2444}
2445#else
2446static inline void assign_proto_idx(struct proto *prot)
2447{
2448}
2449
2450static inline void release_proto_idx(struct proto *prot)
2451{
2452}
2453#endif
2454
b733c007
PE		 2455int proto_register(struct proto *prot, int alloc_slab)
2456{
1da177e4
LT		 2457 if (alloc_slab) {
2458 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2459 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2460 NULL);
1da177e4
LT		 2461
2462 if (prot->slab == NULL) {
2463 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n",
2464 prot->name);
60e7663d 2465 goto out;
1da177e4 2466 }
2e6599cb
ACM		 2467
2468 if (prot->rsk_prot != NULL) {
faf23422 2469 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2470 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2471 goto out_free_sock_slab;
2472
7e56b5d6 2473 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2474 prot->rsk_prot->obj_size, 0,
20c2df83 2475 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2476
2477 if (prot->rsk_prot->slab == NULL) {
2478 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n",
2479 prot->name);
2480 goto out_free_request_sock_slab_name;
2481 }
2482 }
8feaf0c0 2483
6d6ee43e 2484 if (prot->twsk_prot != NULL) {
faf23422 2485 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2486
7e56b5d6 2487 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2488 goto out_free_request_sock_slab;
2489
6d6ee43e 2490 prot->twsk_prot->twsk_slab =
7e56b5d6 2491 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2492 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2493 0,
2494 SLAB_HWCACHE_ALIGN |
2495 prot->slab_flags,
20c2df83 2496 NULL);
6d6ee43e 2497 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2498 goto out_free_timewait_sock_slab_name;
2499 }
1da177e4
LT		 2500 }
2501
36b77a52 2502 mutex_lock(&proto_list_mutex);
1da177e4 2503 list_add(&prot->node, &proto_list);
13ff3d6f 2504 assign_proto_idx(prot);
36b77a52 2505 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2506 return 0;
2507
8feaf0c0 2508out_free_timewait_sock_slab_name:
7e56b5d6 2509 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2510out_free_request_sock_slab:
2511 if (prot->rsk_prot && prot->rsk_prot->slab) {
2512 kmem_cache_destroy(prot->rsk_prot->slab);
2513 prot->rsk_prot->slab = NULL;
2514 }
2e6599cb 2515out_free_request_sock_slab_name:
72150e9b
DC		 2516 if (prot->rsk_prot)
2517 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2518out_free_sock_slab:
2519 kmem_cache_destroy(prot->slab);
2520 prot->slab = NULL;
b733c007
PE		 2521out:
2522 return -ENOBUFS;
1da177e4 2523}
1da177e4
LT		 2524EXPORT_SYMBOL(proto_register);
2525
2526void proto_unregister(struct proto *prot)
2527{
36b77a52 2528 mutex_lock(&proto_list_mutex);
13ff3d6f 2529 release_proto_idx(prot);
0a3f4358 2530 list_del(&prot->node);
36b77a52 2531 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2532
2533 if (prot->slab != NULL) {
2534 kmem_cache_destroy(prot->slab);
2535 prot->slab = NULL;
2536 }
2537
2e6599cb 2538 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2539 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2540 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2541 prot->rsk_prot->slab = NULL;
2542 }
2543
6d6ee43e 2544 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2545 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2546 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2547 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2548 }
1da177e4 2549}
1da177e4
LT		 2550EXPORT_SYMBOL(proto_unregister);
2551
2552#ifdef CONFIG_PROC_FS
1da177e4 2553static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2554 __acquires(proto_list_mutex)
1da177e4 2555{
36b77a52 2556 mutex_lock(&proto_list_mutex);
60f0438a 2557 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2558}
2559
2560static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2561{
60f0438a 2562 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2563}
2564
2565static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2566 __releases(proto_list_mutex)
1da177e4 2567{
36b77a52 2568 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2569}
2570
2571static char proto_method_implemented(const void *method)
2572{
2573 return method == NULL ? 'n' : 'y';
2574}
180d8cd9
GC		 2575static long sock_prot_memory_allocated(struct proto *proto)
2576{
2577 return proto->memory_allocated != NULL ? proto_memory_allocated(proto): -1L;
2578}
2579
2580static char *sock_prot_memory_pressure(struct proto *proto)
2581{
2582 return proto->memory_pressure != NULL ?
2583 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2584}
1da177e4
LT		 2585
2586static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2587{
180d8cd9 2588
8d987e5c 2589 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2590 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2591 proto->name,
2592 proto->obj_size,
14e943db 2593 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2594 sock_prot_memory_allocated(proto),
2595 sock_prot_memory_pressure(proto),
1da177e4
LT		 2596 proto->max_header,
2597 proto->slab == NULL ? "no" : "yes",
2598 module_name(proto->owner),
2599 proto_method_implemented(proto->close),
2600 proto_method_implemented(proto->connect),
2601 proto_method_implemented(proto->disconnect),
2602 proto_method_implemented(proto->accept),
2603 proto_method_implemented(proto->ioctl),
2604 proto_method_implemented(proto->init),
2605 proto_method_implemented(proto->destroy),
2606 proto_method_implemented(proto->shutdown),
2607 proto_method_implemented(proto->setsockopt),
2608 proto_method_implemented(proto->getsockopt),
2609 proto_method_implemented(proto->sendmsg),
2610 proto_method_implemented(proto->recvmsg),
2611 proto_method_implemented(proto->sendpage),
2612 proto_method_implemented(proto->bind),
2613 proto_method_implemented(proto->backlog_rcv),
2614 proto_method_implemented(proto->hash),
2615 proto_method_implemented(proto->unhash),
2616 proto_method_implemented(proto->get_port),
2617 proto_method_implemented(proto->enter_memory_pressure));
2618}
2619
2620static int proto_seq_show(struct seq_file *seq, void *v)
2621{
60f0438a 2622 if (v == &proto_list)
1da177e4
LT		 2623 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2624 "protocol",
2625 "size",
2626 "sockets",
2627 "memory",
2628 "press",
2629 "maxhdr",
2630 "slab",
2631 "module",
2632 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2633 else
60f0438a 2634 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2635 return 0;
2636}
2637
f690808e 2638static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2639 .start = proto_seq_start,
2640 .next = proto_seq_next,
2641 .stop = proto_seq_stop,
2642 .show = proto_seq_show,
2643};
2644
2645static int proto_seq_open(struct inode *inode, struct file *file)
2646{
14e943db
ED		 2647 return seq_open_net(inode, file, &proto_seq_ops,
2648 sizeof(struct seq_net_private));
1da177e4
LT		 2649}
2650
9a32144e 2651static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2652 .owner = THIS_MODULE,
2653 .open = proto_seq_open,
2654 .read = seq_read,
2655 .llseek = seq_lseek,
14e943db
ED		 2656 .release = seq_release_net,
2657};
2658
2659static __net_init int proto_init_net(struct net *net)
2660{
2661 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2662 return -ENOMEM;
2663
2664 return 0;
2665}
2666
2667static __net_exit void proto_exit_net(struct net *net)
2668{
2669 proc_net_remove(net, "protocols");
2670}
2671
2672
2673static __net_initdata struct pernet_operations proto_net_ops = {
2674 .init = proto_init_net,
2675 .exit = proto_exit_net,
1da177e4
LT		 2676};
2677
2678static int __init proto_init(void)
2679{
14e943db 2680 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2681}
2682
2683subsys_initcall(proto_init);
2684
2685#endif /* PROC_FS */
Ubuntu Hirsute Linux kernel mirror