projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Linux 3.10.54
[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
e005d193
JP		 92#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
93
4fc268d2 94#include <linux/capability.h>
1da177e4
LT		 95#include <linux/errno.h>
96#include <linux/types.h>
97#include <linux/socket.h>
98#include <linux/in.h>
99#include <linux/kernel.h>
1da177e4
LT		 100#include <linux/module.h>
101#include <linux/proc_fs.h>
102#include <linux/seq_file.h>
103#include <linux/sched.h>
104#include <linux/timer.h>
105#include <linux/string.h>
106#include <linux/sockios.h>
107#include <linux/net.h>
108#include <linux/mm.h>
109#include <linux/slab.h>
110#include <linux/interrupt.h>
111#include <linux/poll.h>
112#include <linux/tcp.h>
113#include <linux/init.h>
a1f8e7f7 114#include <linux/highmem.h>
3f551f94 115#include <linux/user_namespace.h>
c5905afb 116#include <linux/static_key.h>
3969eb38 117#include <linux/memcontrol.h>
8c1ae10d 118#include <linux/prefetch.h>
1da177e4
LT		 119
120#include <asm/uaccess.h>
1da177e4
LT		 121
122#include <linux/netdevice.h>
123#include <net/protocol.h>
124#include <linux/skbuff.h>
457c4cbc 125#include <net/net_namespace.h>
2e6599cb 126#include <net/request_sock.h>
1da177e4 127#include <net/sock.h>
20d49473 128#include <linux/net_tstamp.h>
1da177e4
LT		 129#include <net/xfrm.h>
130#include <linux/ipsec.h>
f8451725 131#include <net/cls_cgroup.h>
5bc1421e 132#include <net/netprio_cgroup.h>
1da177e4
LT		 133
134#include <linux/filter.h>
135
3847ce32
SM		 136#include <trace/events/sock.h>
137
1da177e4
LT		 138#ifdef CONFIG_INET
139#include <net/tcp.h>
140#endif
141
36b77a52 142static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 143static LIST_HEAD(proto_list);
144
c35b4e28
EB		 145/**
146 * sk_ns_capable - General socket capability test
147 * @sk: Socket to use a capability on or through
148 * @user_ns: The user namespace of the capability to use
149 * @cap: The capability to use
150 *
151 * Test to see if the opener of the socket had when the socket was
152 * created and the current process has the capability @cap in the user
153 * namespace @user_ns.
154 */
155bool sk_ns_capable(const struct sock *sk,
156 struct user_namespace *user_ns, int cap)
157{
158 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
159 ns_capable(user_ns, cap);
160}
161EXPORT_SYMBOL(sk_ns_capable);
162
163/**
164 * sk_capable - Socket global capability test
165 * @sk: Socket to use a capability on or through
166 * @cap: The global capbility to use
167 *
168 * Test to see if the opener of the socket had when the socket was
169 * created and the current process has the capability @cap in all user
170 * namespaces.
171 */
172bool sk_capable(const struct sock *sk, int cap)
173{
174 return sk_ns_capable(sk, &init_user_ns, cap);
175}
176EXPORT_SYMBOL(sk_capable);
177
178/**
179 * sk_net_capable - Network namespace socket capability test
180 * @sk: Socket to use a capability on or through
181 * @cap: The capability to use
182 *
183 * Test to see if the opener of the socket had when the socke was created
184 * and the current process has the capability @cap over the network namespace
185 * the socket is a member of.
186 */
187bool sk_net_capable(const struct sock *sk, int cap)
188{
189 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
190}
191EXPORT_SYMBOL(sk_net_capable);
192
193
c255a458 194#ifdef CONFIG_MEMCG_KMEM
1d62e436 195int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
d1a4c0b3
GC		 196{
197 struct proto *proto;
198 int ret = 0;
199
36b77a52 200 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 201 list_for_each_entry(proto, &proto_list, node) {
202 if (proto->init_cgroup) {
1d62e436 203 ret = proto->init_cgroup(memcg, ss);
d1a4c0b3
GC		 204 if (ret)
205 goto out;
206 }
207 }
208
36b77a52 209 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 210 return ret;
211out:
212 list_for_each_entry_continue_reverse(proto, &proto_list, node)
213 if (proto->destroy_cgroup)
1d62e436 214 proto->destroy_cgroup(memcg);
36b77a52 215 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 216 return ret;
217}
218
1d62e436 219void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
d1a4c0b3
GC		 220{
221 struct proto *proto;
222
36b77a52 223 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 224 list_for_each_entry_reverse(proto, &proto_list, node)
225 if (proto->destroy_cgroup)
1d62e436 226 proto->destroy_cgroup(memcg);
36b77a52 227 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 228}
229#endif
230
da21f24d
IM		 231/*
232 * Each address family might have different locking rules, so we have
233 * one slock key per address family:
234 */
a5b5bb9a
IM		 235static struct lock_class_key af_family_keys[AF_MAX];
236static struct lock_class_key af_family_slock_keys[AF_MAX];
237
cbda4eaf 238#if defined(CONFIG_MEMCG_KMEM)
c5905afb 239struct static_key memcg_socket_limit_enabled;
e1aab161 240EXPORT_SYMBOL(memcg_socket_limit_enabled);
cbda4eaf 241#endif
e1aab161 242
a5b5bb9a
IM		 243/*
244 * Make lock validator output more readable. (we pre-construct these
245 * strings build-time, so that runtime initialization of socket
246 * locks is fast):
247 */
36cbd3dc 248static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 249 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
250 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
251 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
252 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
253 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
254 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
255 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 256 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 257 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 258 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 259 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 260 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 261 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
456db6a4 262 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 263};
36cbd3dc 264static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 265 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
266 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
267 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
268 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
269 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
270 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
271 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 272 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 273 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 274 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 275 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 276 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 277 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 278 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 279};
36cbd3dc 280static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 281 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
282 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
283 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
284 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
285 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
286 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
287 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 288 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 289 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 290 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 291 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 292 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 293 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 294 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 295};
da21f24d
IM		 296
297/*
298 * sk_callback_lock locking rules are per-address-family,
299 * so split the lock classes by using a per-AF key:
300 */
301static struct lock_class_key af_callback_keys[AF_MAX];
302
1da177e4
LT		 303/* Take into consideration the size of the struct sk_buff overhead in the
304 * determination of these values, since that is non-constant across
305 * platforms. This makes socket queueing behavior and performance
306 * not depend upon such differences.
307 */
308#define _SK_MEM_PACKETS 256
87fb4b7b 309#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 310#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
311#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
312
313/* Run time adjustable parameters. */
ab32ea5d 314__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 315EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 316__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 317EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 318__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
319__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 320
25985edc 321/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 322int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 323EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 324
c93bdd0e
MG		 325struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
326EXPORT_SYMBOL_GPL(memalloc_socks);
327
7cb02404
MG		 328/**
329 * sk_set_memalloc - sets %SOCK_MEMALLOC
330 * @sk: socket to set it on
331 *
332 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
333 * It's the responsibility of the admin to adjust min_free_kbytes
334 * to meet the requirements
335 */
336void sk_set_memalloc(struct sock *sk)
337{
338 sock_set_flag(sk, SOCK_MEMALLOC);
339 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 340 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 341}
342EXPORT_SYMBOL_GPL(sk_set_memalloc);
343
344void sk_clear_memalloc(struct sock *sk)
345{
346 sock_reset_flag(sk, SOCK_MEMALLOC);
347 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 348 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 349
350 /*
351 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
352 * progress of swapping. However, if SOCK_MEMALLOC is cleared while
353 * it has rmem allocations there is a risk that the user of the
354 * socket cannot make forward progress due to exceeding the rmem
355 * limits. By rights, sk_clear_memalloc() should only be called
356 * on sockets being torn down but warn and reset the accounting if
357 * that assumption breaks.
358 */
359 if (WARN_ON(sk->sk_forward_alloc))
360 sk_mem_reclaim(sk);
7cb02404
MG		 361}
362EXPORT_SYMBOL_GPL(sk_clear_memalloc);
363
b4b9e355
MG		 364int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
365{
366 int ret;
367 unsigned long pflags = current->flags;
368
369 /* these should have been dropped before queueing */
370 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
371
372 current->flags |= PF_MEMALLOC;
373 ret = sk->sk_backlog_rcv(sk, skb);
374 tsk_restore_flags(current, pflags, PF_MEMALLOC);
375
376 return ret;
377}
378EXPORT_SYMBOL(__sk_backlog_rcv);
379
1da177e4
LT		 380static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
381{
382 struct timeval tv;
383
384 if (optlen < sizeof(tv))
385 return -EINVAL;
386 if (copy_from_user(&tv, optval, sizeof(tv)))
387 return -EFAULT;
ba78073e
VA		 388 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
389 return -EDOM;
1da177e4 390
ba78073e 391 if (tv.tv_sec < 0) {
6f11df83
AM		 392 static int warned __read_mostly;
393
ba78073e 394 *timeo_p = 0;
50aab54f 395 if (warned < 10 && net_ratelimit()) {
ba78073e 396 warned++;
e005d193
JP		 397 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
398 __func__, current->comm, task_pid_nr(current));
50aab54f 399 }
ba78073e
VA		 400 return 0;
401 }
1da177e4
LT		 402 *timeo_p = MAX_SCHEDULE_TIMEOUT;
403 if (tv.tv_sec == 0 && tv.tv_usec == 0)
404 return 0;
405 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
406 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
407 return 0;
408}
409
410static void sock_warn_obsolete_bsdism(const char *name)
411{
412 static int warned;
413 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 414 if (strcmp(warncomm, current->comm) && warned < 5) {
415 strcpy(warncomm, current->comm);
e005d193
JP		 416 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
417 warncomm, name);
1da177e4
LT		 418 warned++;
419 }
420}
421
08e29af3
ED		 422#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
423
424static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 425{
08e29af3
ED		 426 if (sk->sk_flags & flags) {
427 sk->sk_flags &= ~flags;
428 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 429 net_disable_timestamp();
1da177e4
LT		 430 }
431}
432
433
f0088a50
DV		 434int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
435{
766e9037 436 int err;
f0088a50 437 int skb_len;
3b885787
NH		 438 unsigned long flags;
439 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 440
0fd7bac6 441 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 442 atomic_inc(&sk->sk_drops);
3847ce32 443 trace_sock_rcvqueue_full(sk, skb);
766e9037 444 return -ENOMEM;
f0088a50
DV		 445 }
446
fda9ef5d 447 err = sk_filter(sk, skb);
f0088a50 448 if (err)
766e9037 449 return err;
f0088a50 450
c76562b6 451 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 452 atomic_inc(&sk->sk_drops);
453 return -ENOBUFS;
3ab224be
HA		 454 }
455
f0088a50
DV		 456 skb->dev = NULL;
457 skb_set_owner_r(skb, sk);
49ad9599 458
f0088a50
DV		 459 /* Cache the SKB length before we tack it onto the receive
460 * queue. Once it is added it no longer belongs to us and
461 * may be freed by other threads of control pulling packets
462 * from the queue.
463 */
464 skb_len = skb->len;
465
7fee226a
ED		 466 /* we escape from rcu protected region, make sure we dont leak
467 * a norefcounted dst
468 */
469 skb_dst_force(skb);
470
3b885787
NH		 471 spin_lock_irqsave(&list->lock, flags);
472 skb->dropcount = atomic_read(&sk->sk_drops);
473 __skb_queue_tail(list, skb);
474 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 475
476 if (!sock_flag(sk, SOCK_DEAD))
477 sk->sk_data_ready(sk, skb_len);
766e9037 478 return 0;
f0088a50
DV		 479}
480EXPORT_SYMBOL(sock_queue_rcv_skb);
481
58a5a7b9 482int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 483{
484 int rc = NET_RX_SUCCESS;
485
fda9ef5d 486 if (sk_filter(sk, skb))
f0088a50
DV		 487 goto discard_and_relse;
488
489 skb->dev = NULL;
490
f545a38f 491 if (sk_rcvqueues_full(sk, skb, sk->sk_rcvbuf)) {
c377411f
ED		 492 atomic_inc(&sk->sk_drops);
493 goto discard_and_relse;
494 }
58a5a7b9
ACM		 495 if (nested)
496 bh_lock_sock_nested(sk);
497 else
498 bh_lock_sock(sk);
a5b5bb9a
IM		 499 if (!sock_owned_by_user(sk)) {
500 /*
501 * trylock + unlock semantics:
502 */
503 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
504
c57943a1 505 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 506
507 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 508 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 509 bh_unlock_sock(sk);
510 atomic_inc(&sk->sk_drops);
511 goto discard_and_relse;
512 }
513
f0088a50
DV		 514 bh_unlock_sock(sk);
515out:
516 sock_put(sk);
517 return rc;
518discard_and_relse:
519 kfree_skb(skb);
520 goto out;
521}
522EXPORT_SYMBOL(sk_receive_skb);
523
ea94ff3b
KK		 524void sk_reset_txq(struct sock *sk)
525{
526 sk_tx_queue_clear(sk);
527}
528EXPORT_SYMBOL(sk_reset_txq);
529
f0088a50
DV		 530struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
531{
b6c6712a 532 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 533
534 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 535 sk_tx_queue_clear(sk);
a9b3cd7f 536 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 537 dst_release(dst);
538 return NULL;
539 }
540
541 return dst;
542}
543EXPORT_SYMBOL(__sk_dst_check);
544
545struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
546{
547 struct dst_entry *dst = sk_dst_get(sk);
548
549 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
550 sk_dst_reset(sk);
551 dst_release(dst);
552 return NULL;
553 }
554
555 return dst;
556}
557EXPORT_SYMBOL(sk_dst_check);
558
c91f6df2
BH		 559static int sock_setbindtodevice(struct sock *sk, char __user *optval,
560 int optlen)
4878809f
DM		 561{
562 int ret = -ENOPROTOOPT;
563#ifdef CONFIG_NETDEVICES
3b1e0a65 564 struct net *net = sock_net(sk);
4878809f
DM		 565 char devname[IFNAMSIZ];
566 int index;
567
568 /* Sorry... */
569 ret = -EPERM;
5e1fccc0 570 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 571 goto out;
572
573 ret = -EINVAL;
574 if (optlen < 0)
575 goto out;
576
577 /* Bind this socket to a particular device like "eth0",
578 * as specified in the passed interface name. If the
579 * name is "" or the option length is zero the socket
580 * is not bound.
581 */
582 if (optlen > IFNAMSIZ - 1)
583 optlen = IFNAMSIZ - 1;
584 memset(devname, 0, sizeof(devname));
585
586 ret = -EFAULT;
587 if (copy_from_user(devname, optval, optlen))
588 goto out;
589
000ba2e4
DM		 590 index = 0;
591 if (devname[0] != '\0') {
bf8e56bf 592 struct net_device *dev;
4878809f 593
bf8e56bf
ED		 594 rcu_read_lock();
595 dev = dev_get_by_name_rcu(net, devname);
596 if (dev)
597 index = dev->ifindex;
598 rcu_read_unlock();
4878809f
DM		 599 ret = -ENODEV;
600 if (!dev)
601 goto out;
4878809f
DM		 602 }
603
604 lock_sock(sk);
605 sk->sk_bound_dev_if = index;
606 sk_dst_reset(sk);
607 release_sock(sk);
608
609 ret = 0;
610
611out:
612#endif
613
614 return ret;
615}
616
c91f6df2
BH		 617static int sock_getbindtodevice(struct sock *sk, char __user *optval,
618 int __user *optlen, int len)
619{
620 int ret = -ENOPROTOOPT;
621#ifdef CONFIG_NETDEVICES
622 struct net *net = sock_net(sk);
c91f6df2 623 char devname[IFNAMSIZ];
c91f6df2
BH		 624
625 if (sk->sk_bound_dev_if == 0) {
626 len = 0;
627 goto zero;
628 }
629
630 ret = -EINVAL;
631 if (len < IFNAMSIZ)
632 goto out;
633
5dbe7c17
NS		 634 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
635 if (ret)
c91f6df2 636 goto out;
c91f6df2
BH		 637
638 len = strlen(devname) + 1;
639
640 ret = -EFAULT;
641 if (copy_to_user(optval, devname, len))
642 goto out;
643
644zero:
645 ret = -EFAULT;
646 if (put_user(len, optlen))
647 goto out;
648
649 ret = 0;
650
651out:
652#endif
653
654 return ret;
655}
656
c0ef877b
PE		 657static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
658{
659 if (valbool)
660 sock_set_flag(sk, bit);
661 else
662 sock_reset_flag(sk, bit);
663}
664
1da177e4
LT		 665/*
666 * This is meant for all protocols to use and covers goings on
667 * at the socket level. Everything here is generic.
668 */
669
670int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 671 char __user *optval, unsigned int optlen)
1da177e4 672{
2a91525c 673 struct sock *sk = sock->sk;
1da177e4
LT		 674 int val;
675 int valbool;
676 struct linger ling;
677 int ret = 0;
4ec93edb 678
1da177e4
LT		 679 /*
680 * Options without arguments
681 */
682
4878809f 683 if (optname == SO_BINDTODEVICE)
c91f6df2 684 return sock_setbindtodevice(sk, optval, optlen);
4878809f 685
e71a4783
SH		 686 if (optlen < sizeof(int))
687 return -EINVAL;
4ec93edb 688
1da177e4
LT		 689 if (get_user(val, (int __user *)optval))
690 return -EFAULT;
4ec93edb 691
2a91525c 692 valbool = val ? 1 : 0;
1da177e4
LT		 693
694 lock_sock(sk);
695
2a91525c 696 switch (optname) {
e71a4783 697 case SO_DEBUG:
2a91525c 698 if (val && !capable(CAP_NET_ADMIN))
e71a4783 699 ret = -EACCES;
2a91525c 700 else
c0ef877b 701 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 702 break;
703 case SO_REUSEADDR:
4a17fd52 704 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 705 break;
055dc21a
TH		 706 case SO_REUSEPORT:
707 sk->sk_reuseport = valbool;
708 break;
e71a4783 709 case SO_TYPE:
49c794e9 710 case SO_PROTOCOL:
0d6038ee 711 case SO_DOMAIN:
e71a4783
SH		 712 case SO_ERROR:
713 ret = -ENOPROTOOPT;
714 break;
715 case SO_DONTROUTE:
c0ef877b 716 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 717 break;
718 case SO_BROADCAST:
719 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
720 break;
721 case SO_SNDBUF:
722 /* Don't error on this BSD doesn't and if you think
82981930
ED		 723 * about it this is right. Otherwise apps have to
724 * play 'guess the biggest size' games. RCVBUF/SNDBUF
725 * are treated in BSD as hints
726 */
727 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 728set_sndbuf:
e71a4783 729 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 730 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
731 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 732 sk->sk_write_space(sk);
733 break;
1da177e4 734
e71a4783
SH		 735 case SO_SNDBUFFORCE:
736 if (!capable(CAP_NET_ADMIN)) {
737 ret = -EPERM;
738 break;
739 }
740 goto set_sndbuf;
b0573dea 741
e71a4783
SH		 742 case SO_RCVBUF:
743 /* Don't error on this BSD doesn't and if you think
82981930
ED		 744 * about it this is right. Otherwise apps have to
745 * play 'guess the biggest size' games. RCVBUF/SNDBUF
746 * are treated in BSD as hints
747 */
748 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 749set_rcvbuf:
e71a4783
SH		 750 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
751 /*
752 * We double it on the way in to account for
753 * "struct sk_buff" etc. overhead. Applications
754 * assume that the SO_RCVBUF setting they make will
755 * allow that much actual data to be received on that
756 * socket.
757 *
758 * Applications are unaware that "struct sk_buff" and
759 * other overheads allocate from the receive buffer
760 * during socket buffer allocation.
761 *
762 * And after considering the possible alternatives,
763 * returning the value we actually used in getsockopt
764 * is the most desirable behavior.
765 */
82981930 766 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 767 break;
768
769 case SO_RCVBUFFORCE:
770 if (!capable(CAP_NET_ADMIN)) {
771 ret = -EPERM;
1da177e4 772 break;
e71a4783
SH		 773 }
774 goto set_rcvbuf;
1da177e4 775
e71a4783 776 case SO_KEEPALIVE:
1da177e4 777#ifdef CONFIG_INET
3e10986d
ED		 778 if (sk->sk_protocol == IPPROTO_TCP &&
779 sk->sk_type == SOCK_STREAM)
e71a4783 780 tcp_set_keepalive(sk, valbool);
1da177e4 781#endif
e71a4783
SH		 782 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
783 break;
784
785 case SO_OOBINLINE:
786 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
787 break;
788
789 case SO_NO_CHECK:
790 sk->sk_no_check = valbool;
791 break;
792
793 case SO_PRIORITY:
5e1fccc0
EB		 794 if ((val >= 0 && val <= 6) ||
795 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 796 sk->sk_priority = val;
797 else
798 ret = -EPERM;
799 break;
800
801 case SO_LINGER:
802 if (optlen < sizeof(ling)) {
803 ret = -EINVAL; /* 1003.1g */
1da177e4 804 break;
e71a4783 805 }
2a91525c 806 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 807 ret = -EFAULT;
1da177e4 808 break;
e71a4783
SH		 809 }
810 if (!ling.l_onoff)
811 sock_reset_flag(sk, SOCK_LINGER);
812 else {
1da177e4 813#if (BITS_PER_LONG == 32)
e71a4783
SH		 814 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
815 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 816 else
e71a4783
SH		 817#endif
818 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
819 sock_set_flag(sk, SOCK_LINGER);
820 }
821 break;
822
823 case SO_BSDCOMPAT:
824 sock_warn_obsolete_bsdism("setsockopt");
825 break;
826
827 case SO_PASSCRED:
828 if (valbool)
829 set_bit(SOCK_PASSCRED, &sock->flags);
830 else
831 clear_bit(SOCK_PASSCRED, &sock->flags);
832 break;
833
834 case SO_TIMESTAMP:
92f37fd2 835 case SO_TIMESTAMPNS:
e71a4783 836 if (valbool) {
92f37fd2
ED		 837 if (optname == SO_TIMESTAMP)
838 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
839 else
840 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 841 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 842 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 843 } else {
e71a4783 844 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 845 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
846 }
e71a4783
SH		 847 break;
848
20d49473
PO		 849 case SO_TIMESTAMPING:
850 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 851 ret = -EINVAL;
20d49473
PO		 852 break;
853 }
854 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
855 val & SOF_TIMESTAMPING_TX_HARDWARE);
856 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
857 val & SOF_TIMESTAMPING_TX_SOFTWARE);
858 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
859 val & SOF_TIMESTAMPING_RX_HARDWARE);
860 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
861 sock_enable_timestamp(sk,
862 SOCK_TIMESTAMPING_RX_SOFTWARE);
863 else
864 sock_disable_timestamp(sk,
08e29af3 865 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 866 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
867 val & SOF_TIMESTAMPING_SOFTWARE);
868 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
869 val & SOF_TIMESTAMPING_SYS_HARDWARE);
870 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
871 val & SOF_TIMESTAMPING_RAW_HARDWARE);
872 break;
873
e71a4783
SH		 874 case SO_RCVLOWAT:
875 if (val < 0)
876 val = INT_MAX;
877 sk->sk_rcvlowat = val ? : 1;
878 break;
879
880 case SO_RCVTIMEO:
881 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
882 break;
883
884 case SO_SNDTIMEO:
885 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
886 break;
1da177e4 887
e71a4783
SH		 888 case SO_ATTACH_FILTER:
889 ret = -EINVAL;
890 if (optlen == sizeof(struct sock_fprog)) {
891 struct sock_fprog fprog;
1da177e4 892
e71a4783
SH		 893 ret = -EFAULT;
894 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 895 break;
e71a4783
SH		 896
897 ret = sk_attach_filter(&fprog, sk);
898 }
899 break;
900
901 case SO_DETACH_FILTER:
55b33325 902 ret = sk_detach_filter(sk);
e71a4783 903 break;
1da177e4 904
d59577b6
VB		 905 case SO_LOCK_FILTER:
906 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
907 ret = -EPERM;
908 else
909 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
910 break;
911
e71a4783
SH		 912 case SO_PASSSEC:
913 if (valbool)
914 set_bit(SOCK_PASSSEC, &sock->flags);
915 else
916 clear_bit(SOCK_PASSSEC, &sock->flags);
917 break;
4a19ec58 918 case SO_MARK:
5e1fccc0 919 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 920 ret = -EPERM;
2a91525c 921 else
4a19ec58 922 sk->sk_mark = val;
4a19ec58 923 break;
877ce7c1 924
1da177e4
LT		 925 /* We implement the SO_SNDLOWAT etc to
926 not be settable (1003.1g 5.3) */
3b885787 927 case SO_RXQ_OVFL:
8083f0fc 928 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 929 break;
6e3e939f
JB		 930
931 case SO_WIFI_STATUS:
932 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
933 break;
934
ef64a54f
PE		 935 case SO_PEEK_OFF:
936 if (sock->ops->set_peek_off)
d90d9ff6 937 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 938 else
939 ret = -EOPNOTSUPP;
940 break;
3bdc0eba
BG		 941
942 case SO_NOFCS:
943 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
944 break;
945
7d4c04fc
KJ		 946 case SO_SELECT_ERR_QUEUE:
947 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
948 break;
949
e71a4783
SH		 950 default:
951 ret = -ENOPROTOOPT;
952 break;
4ec93edb 953 }
1da177e4
LT		 954 release_sock(sk);
955 return ret;
956}
2a91525c 957EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 958
959
3f551f94
EB		 960void cred_to_ucred(struct pid *pid, const struct cred *cred,
961 struct ucred *ucred)
962{
963 ucred->pid = pid_vnr(pid);
964 ucred->uid = ucred->gid = -1;
965 if (cred) {
966 struct user_namespace *current_ns = current_user_ns();
967
b2e4f544
EB		 968 ucred->uid = from_kuid_munged(current_ns, cred->euid);
969 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 970 }
971}
3924773a 972EXPORT_SYMBOL_GPL(cred_to_ucred);
3f551f94 973
1da177e4
LT		 974int sock_getsockopt(struct socket *sock, int level, int optname,
975 char __user *optval, int __user *optlen)
976{
977 struct sock *sk = sock->sk;
4ec93edb 978
e71a4783 979 union {
4ec93edb
YH		 980 int val;
981 struct linger ling;
1da177e4
LT		 982 struct timeval tm;
983 } v;
4ec93edb 984
4d0392be 985 int lv = sizeof(int);
1da177e4 986 int len;
4ec93edb 987
e71a4783 988 if (get_user(len, optlen))
4ec93edb 989 return -EFAULT;
e71a4783 990 if (len < 0)
1da177e4 991 return -EINVAL;
4ec93edb 992
50fee1de 993 memset(&v, 0, sizeof(v));
df0bca04 994
2a91525c 995 switch (optname) {
e71a4783
SH		 996 case SO_DEBUG:
997 v.val = sock_flag(sk, SOCK_DBG);
998 break;
999
1000 case SO_DONTROUTE:
1001 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1002 break;
1003
1004 case SO_BROADCAST:
1b23a5df 1005 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1006 break;
1007
1008 case SO_SNDBUF:
1009 v.val = sk->sk_sndbuf;
1010 break;
1011
1012 case SO_RCVBUF:
1013 v.val = sk->sk_rcvbuf;
1014 break;
1015
1016 case SO_REUSEADDR:
1017 v.val = sk->sk_reuse;
1018 break;
1019
055dc21a
TH		 1020 case SO_REUSEPORT:
1021 v.val = sk->sk_reuseport;
1022 break;
1023
e71a4783 1024 case SO_KEEPALIVE:
1b23a5df 1025 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1026 break;
1027
1028 case SO_TYPE:
1029 v.val = sk->sk_type;
1030 break;
1031
49c794e9
JE		 1032 case SO_PROTOCOL:
1033 v.val = sk->sk_protocol;
1034 break;
1035
0d6038ee
JE		 1036 case SO_DOMAIN:
1037 v.val = sk->sk_family;
1038 break;
1039
e71a4783
SH		 1040 case SO_ERROR:
1041 v.val = -sock_error(sk);
2a91525c 1042 if (v.val == 0)
e71a4783
SH		 1043 v.val = xchg(&sk->sk_err_soft, 0);
1044 break;
1045
1046 case SO_OOBINLINE:
1b23a5df 1047 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1048 break;
1049
1050 case SO_NO_CHECK:
1051 v.val = sk->sk_no_check;
1052 break;
1053
1054 case SO_PRIORITY:
1055 v.val = sk->sk_priority;
1056 break;
1057
1058 case SO_LINGER:
1059 lv = sizeof(v.ling);
1b23a5df 1060 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1061 v.ling.l_linger = sk->sk_lingertime / HZ;
1062 break;
1063
1064 case SO_BSDCOMPAT:
1065 sock_warn_obsolete_bsdism("getsockopt");
1066 break;
1067
1068 case SO_TIMESTAMP:
92f37fd2
ED		 1069 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1070 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1071 break;
1072
1073 case SO_TIMESTAMPNS:
1074 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1075 break;
1076
20d49473
PO		 1077 case SO_TIMESTAMPING:
1078 v.val = 0;
1079 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
1080 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
1081 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
1082 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
1083 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
1084 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
1085 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
1086 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
1087 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
1088 v.val |= SOF_TIMESTAMPING_SOFTWARE;
1089 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
1090 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
1091 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
1092 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
1093 break;
1094
e71a4783 1095 case SO_RCVTIMEO:
2a91525c 1096 lv = sizeof(struct timeval);
e71a4783
SH		 1097 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1098 v.tm.tv_sec = 0;
1099 v.tm.tv_usec = 0;
1100 } else {
1101 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1102 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1103 }
1104 break;
1105
1106 case SO_SNDTIMEO:
2a91525c 1107 lv = sizeof(struct timeval);
e71a4783
SH		 1108 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1109 v.tm.tv_sec = 0;
1110 v.tm.tv_usec = 0;
1111 } else {
1112 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1113 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1114 }
1115 break;
1da177e4 1116
e71a4783
SH		 1117 case SO_RCVLOWAT:
1118 v.val = sk->sk_rcvlowat;
1119 break;
1da177e4 1120
e71a4783 1121 case SO_SNDLOWAT:
2a91525c 1122 v.val = 1;
e71a4783 1123 break;
1da177e4 1124
e71a4783 1125 case SO_PASSCRED:
82981930 1126 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1127 break;
1da177e4 1128
e71a4783 1129 case SO_PEERCRED:
109f6e39
EB		 1130 {
1131 struct ucred peercred;
1132 if (len > sizeof(peercred))
1133 len = sizeof(peercred);
1134 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1135 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1136 return -EFAULT;
1137 goto lenout;
109f6e39 1138 }
1da177e4 1139
e71a4783
SH		 1140 case SO_PEERNAME:
1141 {
1142 char address[128];
1143
1144 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1145 return -ENOTCONN;
1146 if (lv < len)
1147 return -EINVAL;
1148 if (copy_to_user(optval, address, len))
1149 return -EFAULT;
1150 goto lenout;
1151 }
1da177e4 1152
e71a4783
SH		 1153 /* Dubious BSD thing... Probably nobody even uses it, but
1154 * the UNIX standard wants it for whatever reason... -DaveM
1155 */
1156 case SO_ACCEPTCONN:
1157 v.val = sk->sk_state == TCP_LISTEN;
1158 break;
1da177e4 1159
e71a4783 1160 case SO_PASSSEC:
82981930 1161 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1162 break;
877ce7c1 1163
e71a4783
SH		 1164 case SO_PEERSEC:
1165 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1166
4a19ec58
LAT		 1167 case SO_MARK:
1168 v.val = sk->sk_mark;
1169 break;
1170
3b885787 1171 case SO_RXQ_OVFL:
1b23a5df 1172 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1173 break;
1174
6e3e939f 1175 case SO_WIFI_STATUS:
1b23a5df 1176 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1177 break;
1178
ef64a54f
PE		 1179 case SO_PEEK_OFF:
1180 if (!sock->ops->set_peek_off)
1181 return -EOPNOTSUPP;
1182
1183 v.val = sk->sk_peek_off;
1184 break;
bc2f7996 1185 case SO_NOFCS:
1b23a5df 1186 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1187 break;
c91f6df2 1188
f7b86bfe 1189 case SO_BINDTODEVICE:
c91f6df2
BH		 1190 return sock_getbindtodevice(sk, optval, optlen, len);
1191
a8fc9277
PE		 1192 case SO_GET_FILTER:
1193 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1194 if (len < 0)
1195 return len;
1196
1197 goto lenout;
c91f6df2 1198
d59577b6
VB		 1199 case SO_LOCK_FILTER:
1200 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1201 break;
1202
7d4c04fc
KJ		 1203 case SO_SELECT_ERR_QUEUE:
1204 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1205 break;
1206
e71a4783
SH		 1207 default:
1208 return -ENOPROTOOPT;
1da177e4 1209 }
e71a4783 1210
1da177e4
LT		 1211 if (len > lv)
1212 len = lv;
1213 if (copy_to_user(optval, &v, len))
1214 return -EFAULT;
1215lenout:
4ec93edb
YH		 1216 if (put_user(len, optlen))
1217 return -EFAULT;
1218 return 0;
1da177e4
LT		 1219}
1220
a5b5bb9a
IM		 1221/*
1222 * Initialize an sk_lock.
1223 *
1224 * (We also register the sk_lock with the lock validator.)
1225 */
b6f99a21 1226static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1227{
ed07536e
PZ		 1228 sock_lock_init_class_and_name(sk,
1229 af_family_slock_key_strings[sk->sk_family],
1230 af_family_slock_keys + sk->sk_family,
1231 af_family_key_strings[sk->sk_family],
1232 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1233}
1234
4dc6dc71
ED		 1235/*
1236 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1237 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1238 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1239 */
f1a6c4da
PE		 1240static void sock_copy(struct sock *nsk, const struct sock *osk)
1241{
1242#ifdef CONFIG_SECURITY_NETWORK
1243 void *sptr = nsk->sk_security;
1244#endif
68835aba
ED		 1245 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1246
1247 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1248 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1249
f1a6c4da
PE		 1250#ifdef CONFIG_SECURITY_NETWORK
1251 nsk->sk_security = sptr;
1252 security_sk_clone(osk, nsk);
1253#endif
1254}
1255
fcbdf09d
OP		 1256void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1257{
1258 unsigned long nulls1, nulls2;
1259
1260 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1261 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1262 if (nulls1 > nulls2)
1263 swap(nulls1, nulls2);
1264
1265 if (nulls1 != 0)
1266 memset((char *)sk, 0, nulls1);
1267 memset((char *)sk + nulls1 + sizeof(void *), 0,
1268 nulls2 - nulls1 - sizeof(void *));
1269 memset((char *)sk + nulls2 + sizeof(void *), 0,
1270 size - nulls2 - sizeof(void *));
1271}
1272EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1273
2e4afe7b
PE		 1274static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1275 int family)
c308c1b2
PE		 1276{
1277 struct sock *sk;
1278 struct kmem_cache *slab;
1279
1280 slab = prot->slab;
e912b114
ED		 1281 if (slab != NULL) {
1282 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1283 if (!sk)
1284 return sk;
1285 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1286 if (prot->clear_sk)
1287 prot->clear_sk(sk, prot->obj_size);
1288 else
1289 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1290 }
fcbdf09d 1291 } else
c308c1b2
PE		 1292 sk = kmalloc(prot->obj_size, priority);
1293
2e4afe7b 1294 if (sk != NULL) {
a98b65a3
VN		 1295 kmemcheck_annotate_bitfield(sk, flags);
1296
2e4afe7b
PE		 1297 if (security_sk_alloc(sk, family, priority))
1298 goto out_free;
1299
1300 if (!try_module_get(prot->owner))
1301 goto out_free_sec;
e022f0b4 1302 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1303 }
1304
c308c1b2 1305 return sk;
2e4afe7b
PE		 1306
1307out_free_sec:
1308 security_sk_free(sk);
1309out_free:
1310 if (slab != NULL)
1311 kmem_cache_free(slab, sk);
1312 else
1313 kfree(sk);
1314 return NULL;
c308c1b2
PE		 1315}
1316
1317static void sk_prot_free(struct proto *prot, struct sock *sk)
1318{
1319 struct kmem_cache *slab;
2e4afe7b 1320 struct module *owner;
c308c1b2 1321
2e4afe7b 1322 owner = prot->owner;
c308c1b2 1323 slab = prot->slab;
2e4afe7b
PE		 1324
1325 security_sk_free(sk);
c308c1b2
PE		 1326 if (slab != NULL)
1327 kmem_cache_free(slab, sk);
1328 else
1329 kfree(sk);
2e4afe7b 1330 module_put(owner);
c308c1b2
PE		 1331}
1332
8fb974c9 1333#if IS_ENABLED(CONFIG_NET_CLS_CGROUP)
211d2f97 1334void sock_update_classid(struct sock *sk)
f8451725 1335{
1144182a 1336 u32 classid;
f8451725 1337
211d2f97 1338 classid = task_cls_classid(current);
3afa6d00 1339 if (classid != sk->sk_classid)
f8451725
HX		 1340 sk->sk_classid = classid;
1341}
82862742 1342EXPORT_SYMBOL(sock_update_classid);
8fb974c9 1343#endif
5bc1421e 1344
51e4e7fa 1345#if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
6ffd4641 1346void sock_update_netprioidx(struct sock *sk)
5bc1421e 1347{
5bc1421e
NH		 1348 if (in_interrupt())
1349 return;
2b73bc65 1350
6ffd4641 1351 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1352}
1353EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1354#endif
1355
1da177e4
LT		 1356/**
1357 * sk_alloc - All socket objects are allocated here
c4ea43c5 1358 * @net: the applicable net namespace
4dc3b16b
PP		 1359 * @family: protocol family
1360 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1361 * @prot: struct proto associated with this new sock instance
1da177e4 1362 */
1b8d7ae4 1363struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1364 struct proto *prot)
1da177e4 1365{
c308c1b2 1366 struct sock *sk;
1da177e4 1367
154adbc8 1368 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1369 if (sk) {
154adbc8
PE		 1370 sk->sk_family = family;
1371 /*
1372 * See comment in struct sock definition to understand
1373 * why we need sk_prot_creator -acme
1374 */
1375 sk->sk_prot = sk->sk_prot_creator = prot;
1376 sock_lock_init(sk);
3b1e0a65 1377 sock_net_set(sk, get_net(net));
d66ee058 1378 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1379
211d2f97 1380 sock_update_classid(sk);
6ffd4641 1381 sock_update_netprioidx(sk);
1da177e4 1382 }
a79af59e 1383
2e4afe7b 1384 return sk;
1da177e4 1385}
2a91525c 1386EXPORT_SYMBOL(sk_alloc);
1da177e4 1387
2b85a34e 1388static void __sk_free(struct sock *sk)
1da177e4
LT		 1389{
1390 struct sk_filter *filter;
1da177e4
LT		 1391
1392 if (sk->sk_destruct)
1393 sk->sk_destruct(sk);
1394
a898def2
PM		 1395 filter = rcu_dereference_check(sk->sk_filter,
1396 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1397 if (filter) {
309dd5fc 1398 sk_filter_uncharge(sk, filter);
a9b3cd7f 1399 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1400 }
1401
08e29af3 1402 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1403
1404 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1405 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1406 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1407
109f6e39
EB		 1408 if (sk->sk_peer_cred)
1409 put_cred(sk->sk_peer_cred);
1410 put_pid(sk->sk_peer_pid);
3b1e0a65 1411 put_net(sock_net(sk));
c308c1b2 1412 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1413}
2b85a34e
ED		 1414
1415void sk_free(struct sock *sk)
1416{
1417 /*
25985edc 1418 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1419 * some packets are still in some tx queue.
1420 * If not null, sock_wfree() will call __sk_free(sk) later
1421 */
1422 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1423 __sk_free(sk);
1424}
2a91525c 1425EXPORT_SYMBOL(sk_free);
1da177e4 1426
edf02087 1427/*
25985edc
LDM		 1428 * Last sock_put should drop reference to sk->sk_net. It has already
1429 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1430 * is not an option.
25985edc 1431 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1432 * destroy it in the context of init_net.
1433 */
1434void sk_release_kernel(struct sock *sk)
1435{
1436 if (sk == NULL || sk->sk_socket == NULL)
1437 return;
1438
1439 sock_hold(sk);
1440 sock_release(sk->sk_socket);
65a18ec5 1441 release_net(sock_net(sk));
3b1e0a65 1442 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1443 sock_put(sk);
1444}
45af1754 1445EXPORT_SYMBOL(sk_release_kernel);
edf02087 1446
475f1b52
SR		 1447static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1448{
1449 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1450 sock_update_memcg(newsk);
1451}
1452
e56c57d0
ED		 1453/**
1454 * sk_clone_lock - clone a socket, and lock its clone
1455 * @sk: the socket to clone
1456 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1457 *
1458 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1459 */
1460struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1461{
8fd1d178 1462 struct sock *newsk;
87d11ceb 1463
8fd1d178 1464 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1465 if (newsk != NULL) {
1466 struct sk_filter *filter;
1467
892c141e 1468 sock_copy(newsk, sk);
87d11ceb
ACM		 1469
1470 /* SANITY */
3b1e0a65 1471 get_net(sock_net(newsk));
87d11ceb
ACM		 1472 sk_node_init(&newsk->sk_node);
1473 sock_lock_init(newsk);
1474 bh_lock_sock(newsk);
fa438ccf 1475 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1476 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1477
1478 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1479 /*
1480 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1481 */
1482 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1483 atomic_set(&newsk->sk_omem_alloc, 0);
1484 skb_queue_head_init(&newsk->sk_receive_queue);
1485 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1486#ifdef CONFIG_NET_DMA
1487 skb_queue_head_init(&newsk->sk_async_wait_queue);
1488#endif
87d11ceb 1489
b6c6712a 1490 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1491 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1492 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1493 af_callback_keys + newsk->sk_family,
1494 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1495
1496 newsk->sk_dst_cache = NULL;
1497 newsk->sk_wmem_queued = 0;
1498 newsk->sk_forward_alloc = 0;
1499 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1500 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1501
1502 sock_reset_flag(newsk, SOCK_DONE);
1503 skb_queue_head_init(&newsk->sk_error_queue);
1504
0d7da9dd 1505 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb
ACM		 1506 if (filter != NULL)
1507 sk_filter_charge(newsk, filter);
1508
1509 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1510 /* It is still raw copy of parent, so invalidate
1511 * destructor and make plain sk_free() */
1512 newsk->sk_destruct = NULL;
b0691c8e 1513 bh_unlock_sock(newsk);
87d11ceb
ACM		 1514 sk_free(newsk);
1515 newsk = NULL;
1516 goto out;
1517 }
1518
1519 newsk->sk_err = 0;
1520 newsk->sk_priority = 0;
4dc6dc71
ED		 1521 /*
1522 * Before updating sk_refcnt, we must commit prior changes to memory
1523 * (Documentation/RCU/rculist_nulls.txt for details)
1524 */
1525 smp_wmb();
87d11ceb
ACM		 1526 atomic_set(&newsk->sk_refcnt, 2);
1527
1528 /*
1529 * Increment the counter in the same struct proto as the master
1530 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1531 * is the same as sk->sk_prot->socks, as this field was copied
1532 * with memcpy).
1533 *
1534 * This _changes_ the previous behaviour, where
1535 * tcp_create_openreq_child always was incrementing the
1536 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1537 * to be taken into account in all callers. -acme
1538 */
1539 sk_refcnt_debug_inc(newsk);
972692e0 1540 sk_set_socket(newsk, NULL);
43815482 1541 newsk->sk_wq = NULL;
87d11ceb 1542
f3f511e1
GC		 1543 sk_update_clone(sk, newsk);
1544
87d11ceb 1545 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1546 sk_sockets_allocated_inc(newsk);
704da560 1547
08e29af3 1548 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1549 net_enable_timestamp();
87d11ceb
ACM		 1550 }
1551out:
1552 return newsk;
1553}
e56c57d0 1554EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1555
9958089a
AK		 1556void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1557{
1558 __sk_dst_set(sk, dst);
1559 sk->sk_route_caps = dst->dev->features;
1560 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1561 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1562 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1563 if (sk_can_gso(sk)) {
82cc1a7a 1564 if (dst->header_len) {
9958089a 1565 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1566 } else {
9958089a 1567 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1568 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1569 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1570 }
9958089a
AK		 1571 }
1572}
1573EXPORT_SYMBOL_GPL(sk_setup_caps);
1574
1da177e4
LT		 1575/*
1576 * Simple resource managers for sockets.
1577 */
1578
1579
4ec93edb
YH		 1580/*
1581 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1582 */
1583void sock_wfree(struct sk_buff *skb)
1584{
1585 struct sock *sk = skb->sk;
d99927f4 1586 unsigned int len = skb->truesize;
1da177e4 1587
d99927f4
ED		 1588 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1589 /*
1590 * Keep a reference on sk_wmem_alloc, this will be released
1591 * after sk_write_space() call
1592 */
1593 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1594 sk->sk_write_space(sk);
d99927f4
ED		 1595 len = 1;
1596 }
2b85a34e 1597 /*
d99927f4
ED		 1598 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1599 * could not do because of in-flight packets
2b85a34e 1600 */
d99927f4 1601 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1602 __sk_free(sk);
1da177e4 1603}
2a91525c 1604EXPORT_SYMBOL(sock_wfree);
1da177e4 1605
4ec93edb
YH		 1606/*
1607 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1608 */
1609void sock_rfree(struct sk_buff *skb)
1610{
1611 struct sock *sk = skb->sk;
d361fd59 1612 unsigned int len = skb->truesize;
1da177e4 1613
d361fd59
ED		 1614 atomic_sub(len, &sk->sk_rmem_alloc);
1615 sk_mem_uncharge(sk, len);
1da177e4 1616}
2a91525c 1617EXPORT_SYMBOL(sock_rfree);
1da177e4 1618
41063e9d
DM		 1619void sock_edemux(struct sk_buff *skb)
1620{
e812347c
ED		 1621 struct sock *sk = skb->sk;
1622
1c463e57 1623#ifdef CONFIG_INET
e812347c
ED		 1624 if (sk->sk_state == TCP_TIME_WAIT)
1625 inet_twsk_put(inet_twsk(sk));
1626 else
1c463e57 1627#endif
e812347c 1628 sock_put(sk);
41063e9d
DM		 1629}
1630EXPORT_SYMBOL(sock_edemux);
1da177e4 1631
976d0201 1632kuid_t sock_i_uid(struct sock *sk)
1da177e4 1633{
976d0201 1634 kuid_t uid;
1da177e4 1635
f064af1e 1636 read_lock_bh(&sk->sk_callback_lock);
976d0201 1637 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1638 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1639 return uid;
1640}
2a91525c 1641EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1642
1643unsigned long sock_i_ino(struct sock *sk)
1644{
1645 unsigned long ino;
1646
f064af1e 1647 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1648 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1649 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1650 return ino;
1651}
2a91525c 1652EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1653
1654/*
1655 * Allocate a skb from the socket's send buffer.
1656 */
86a76caf 1657struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1658 gfp_t priority)
1da177e4
LT		 1659{
1660 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1661 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1662 if (skb) {
1663 skb_set_owner_w(skb, sk);
1664 return skb;
1665 }
1666 }
1667 return NULL;
1668}
2a91525c 1669EXPORT_SYMBOL(sock_wmalloc);
1da177e4
LT		 1670
1671/*
1672 * Allocate a skb from the socket's receive buffer.
4ec93edb 1673 */
86a76caf 1674struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1675 gfp_t priority)
1da177e4
LT		 1676{
1677 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1678 struct sk_buff *skb = alloc_skb(size, priority);
1679 if (skb) {
1680 skb_set_owner_r(skb, sk);
1681 return skb;
1682 }
1683 }
1684 return NULL;
1685}
1686
4ec93edb 1687/*
1da177e4 1688 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1689 */
dd0fc66f 1690void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1691{
95c96174 1692 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1693 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1694 void *mem;
1695 /* First do the add, to avoid the race if kmalloc
4ec93edb 1696 * might sleep.
1da177e4
LT		 1697 */
1698 atomic_add(size, &sk->sk_omem_alloc);
1699 mem = kmalloc(size, priority);
1700 if (mem)
1701 return mem;
1702 atomic_sub(size, &sk->sk_omem_alloc);
1703 }
1704 return NULL;
1705}
2a91525c 1706EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1707
1708/*
1709 * Free an option memory block.
1710 */
1711void sock_kfree_s(struct sock *sk, void *mem, int size)
1712{
1713 kfree(mem);
1714 atomic_sub(size, &sk->sk_omem_alloc);
1715}
2a91525c 1716EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1717
1718/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1719 I think, these locks should be removed for datagram sockets.
1720 */
2a91525c 1721static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1722{
1723 DEFINE_WAIT(wait);
1724
1725 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1726 for (;;) {
1727 if (!timeo)
1728 break;
1729 if (signal_pending(current))
1730 break;
1731 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1732 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1733 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1734 break;
1735 if (sk->sk_shutdown & SEND_SHUTDOWN)
1736 break;
1737 if (sk->sk_err)
1738 break;
1739 timeo = schedule_timeout(timeo);
1740 }
aa395145 1741 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1742 return timeo;
1743}
1744
1745
1746/*
1747 * Generic send/receive buffer handlers
1748 */
1749
4cc7f68d
HX		 1750struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1751 unsigned long data_len, int noblock,
1752 int *errcode)
1da177e4
LT		 1753{
1754 struct sk_buff *skb;
7d877f3b 1755 gfp_t gfp_mask;
1da177e4
LT		 1756 long timeo;
1757 int err;
cc9b17ad
JW		 1758 int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1759
1760 err = -EMSGSIZE;
1761 if (npages > MAX_SKB_FRAGS)
1762 goto failure;
1da177e4
LT		 1763
1764 gfp_mask = sk->sk_allocation;
1765 if (gfp_mask & __GFP_WAIT)
1766 gfp_mask |= __GFP_REPEAT;
1767
1768 timeo = sock_sndtimeo(sk, noblock);
1769 while (1) {
1770 err = sock_error(sk);
1771 if (err != 0)
1772 goto failure;
1773
1774 err = -EPIPE;
1775 if (sk->sk_shutdown & SEND_SHUTDOWN)
1776 goto failure;
1777
1778 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1779 skb = alloc_skb(header_len, gfp_mask);
1da177e4 1780 if (skb) {
1da177e4
LT		 1781 int i;
1782
1783 /* No pages, we're done... */
1784 if (!data_len)
1785 break;
1786
1da177e4
LT		 1787 skb->truesize += data_len;
1788 skb_shinfo(skb)->nr_frags = npages;
1789 for (i = 0; i < npages; i++) {
1790 struct page *page;
1da177e4
LT		 1791
1792 page = alloc_pages(sk->sk_allocation, 0);
1793 if (!page) {
1794 err = -ENOBUFS;
1795 skb_shinfo(skb)->nr_frags = i;
1796 kfree_skb(skb);
1797 goto failure;
1798 }
1799
ea2ab693
IC		 1800 __skb_fill_page_desc(skb, i,
1801 page, 0,
1802 (data_len >= PAGE_SIZE ?
1803 PAGE_SIZE :
1804 data_len));
1da177e4
LT		 1805 data_len -= PAGE_SIZE;
1806 }
1807
1808 /* Full success... */
1809 break;
1810 }
1811 err = -ENOBUFS;
1812 goto failure;
1813 }
1814 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1815 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1816 err = -EAGAIN;
1817 if (!timeo)
1818 goto failure;
1819 if (signal_pending(current))
1820 goto interrupted;
1821 timeo = sock_wait_for_wmem(sk, timeo);
1822 }
1823
1824 skb_set_owner_w(skb, sk);
1825 return skb;
1826
1827interrupted:
1828 err = sock_intr_errno(timeo);
1829failure:
1830 *errcode = err;
1831 return NULL;
1832}
4cc7f68d 1833EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1834
4ec93edb 1835struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1836 int noblock, int *errcode)
1837{
1838 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1839}
2a91525c 1840EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1841
5640f768
ED		 1842/* On 32bit arches, an skb frag is limited to 2^15 */
1843#define SKB_FRAG_PAGE_ORDER get_order(32768)
1844
1845bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1846{
1847 int order;
1848
1849 if (pfrag->page) {
1850 if (atomic_read(&pfrag->page->_count) == 1) {
1851 pfrag->offset = 0;
1852 return true;
1853 }
1854 if (pfrag->offset < pfrag->size)
1855 return true;
1856 put_page(pfrag->page);
1857 }
1858
1859 /* We restrict high order allocations to users that can afford to wait */
1860 order = (sk->sk_allocation & __GFP_WAIT) ? SKB_FRAG_PAGE_ORDER : 0;
1861
1862 do {
1863 gfp_t gfp = sk->sk_allocation;
1864
1865 if (order)
a9e3d789 1866 gfp |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY;
5640f768
ED		 1867 pfrag->page = alloc_pages(gfp, order);
1868 if (likely(pfrag->page)) {
1869 pfrag->offset = 0;
1870 pfrag->size = PAGE_SIZE << order;
1871 return true;
1872 }
1873 } while (--order >= 0);
1874
1875 sk_enter_memory_pressure(sk);
1876 sk_stream_moderate_sndbuf(sk);
1877 return false;
1878}
1879EXPORT_SYMBOL(sk_page_frag_refill);
1880
1da177e4 1881static void __lock_sock(struct sock *sk)
f39234d6
NK		 1882 __releases(&sk->sk_lock.slock)
1883 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1884{
1885 DEFINE_WAIT(wait);
1886
e71a4783 1887 for (;;) {
1da177e4
LT		 1888 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1889 TASK_UNINTERRUPTIBLE);
1890 spin_unlock_bh(&sk->sk_lock.slock);
1891 schedule();
1892 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1893 if (!sock_owned_by_user(sk))
1da177e4
LT		 1894 break;
1895 }
1896 finish_wait(&sk->sk_lock.wq, &wait);
1897}
1898
1899static void __release_sock(struct sock *sk)
f39234d6
NK		 1900 __releases(&sk->sk_lock.slock)
1901 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1902{
1903 struct sk_buff *skb = sk->sk_backlog.head;
1904
1905 do {
1906 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1907 bh_unlock_sock(sk);
1908
1909 do {
1910 struct sk_buff *next = skb->next;
1911
e4cbb02a 1912 prefetch(next);
7fee226a 1913 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1914 skb->next = NULL;
c57943a1 1915 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1916
1917 /*
1918 * We are in process context here with softirqs
1919 * disabled, use cond_resched_softirq() to preempt.
1920 * This is safe to do because we've taken the backlog
1921 * queue private:
1922 */
1923 cond_resched_softirq();
1924
1925 skb = next;
1926 } while (skb != NULL);
1927
1928 bh_lock_sock(sk);
e71a4783 1929 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1930
1931 /*
1932 * Doing the zeroing here guarantee we can not loop forever
1933 * while a wild producer attempts to flood us.
1934 */
1935 sk->sk_backlog.len = 0;
1da177e4
LT		 1936}
1937
1938/**
1939 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1940 * @sk: sock to wait on
1941 * @timeo: for how long
1da177e4
LT		 1942 *
1943 * Now socket state including sk->sk_err is changed only under lock,
1944 * hence we may omit checks after joining wait queue.
1945 * We check receive queue before schedule() only as optimization;
1946 * it is very likely that release_sock() added new data.
1947 */
1948int sk_wait_data(struct sock *sk, long *timeo)
1949{
1950 int rc;
1951 DEFINE_WAIT(wait);
1952
aa395145 1953 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1954 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1955 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1956 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1957 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1958 return rc;
1959}
1da177e4
LT		 1960EXPORT_SYMBOL(sk_wait_data);
1961
3ab224be
HA		 1962/**
1963 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1964 * @sk: socket
1965 * @size: memory size to allocate
1966 * @kind: allocation type
1967 *
1968 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1969 * rmem allocation. This function assumes that protocols which have
1970 * memory_pressure use sk_wmem_queued as write buffer accounting.
1971 */
1972int __sk_mem_schedule(struct sock *sk, int size, int kind)
1973{
1974 struct proto *prot = sk->sk_prot;
1975 int amt = sk_mem_pages(size);
8d987e5c 1976 long allocated;
e1aab161 1977 int parent_status = UNDER_LIMIT;
3ab224be
HA		 1978
1979 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 1980
e1aab161 1981 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 1982
1983 /* Under limit. */
e1aab161
GC		 1984 if (parent_status == UNDER_LIMIT &&
1985 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 1986 sk_leave_memory_pressure(sk);
3ab224be
HA		 1987 return 1;
1988 }
1989
e1aab161
GC		 1990 /* Under pressure. (we or our parents) */
1991 if ((parent_status > SOFT_LIMIT) ||
1992 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 1993 sk_enter_memory_pressure(sk);
3ab224be 1994
e1aab161
GC		 1995 /* Over hard limit (we or our parents) */
1996 if ((parent_status == OVER_LIMIT) ||
1997 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 1998 goto suppress_allocation;
1999
2000 /* guarantee minimum buffer size under pressure */
2001 if (kind == SK_MEM_RECV) {
2002 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2003 return 1;
180d8cd9 2004
3ab224be
HA		 2005 } else { /* SK_MEM_SEND */
2006 if (sk->sk_type == SOCK_STREAM) {
2007 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2008 return 1;
2009 } else if (atomic_read(&sk->sk_wmem_alloc) <
2010 prot->sysctl_wmem[0])
2011 return 1;
2012 }
2013
180d8cd9 2014 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2015 int alloc;
2016
180d8cd9 2017 if (!sk_under_memory_pressure(sk))
1748376b 2018 return 1;
180d8cd9
GC		 2019 alloc = sk_sockets_allocated_read_positive(sk);
2020 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2021 sk_mem_pages(sk->sk_wmem_queued +
2022 atomic_read(&sk->sk_rmem_alloc) +
2023 sk->sk_forward_alloc))
2024 return 1;
2025 }
2026
2027suppress_allocation:
2028
2029 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2030 sk_stream_moderate_sndbuf(sk);
2031
2032 /* Fail only if socket is _under_ its sndbuf.
2033 * In this case we cannot block, so that we have to fail.
2034 */
2035 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2036 return 1;
2037 }
2038
3847ce32
SM		 2039 trace_sock_exceed_buf_limit(sk, prot, allocated);
2040
3ab224be
HA		 2041 /* Alas. Undo changes. */
2042 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2043
0e90b31f 2044 sk_memory_allocated_sub(sk, amt);
180d8cd9 2045
3ab224be
HA		 2046 return 0;
2047}
3ab224be
HA		 2048EXPORT_SYMBOL(__sk_mem_schedule);
2049
2050/**
2051 * __sk_reclaim - reclaim memory_allocated
2052 * @sk: socket
2053 */
2054void __sk_mem_reclaim(struct sock *sk)
2055{
180d8cd9 2056 sk_memory_allocated_sub(sk,
0e90b31f 2057 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 2058 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
2059
180d8cd9
GC		 2060 if (sk_under_memory_pressure(sk) &&
2061 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2062 sk_leave_memory_pressure(sk);
3ab224be 2063}
3ab224be
HA		 2064EXPORT_SYMBOL(__sk_mem_reclaim);
2065
2066
1da177e4
LT		 2067/*
2068 * Set of default routines for initialising struct proto_ops when
2069 * the protocol does not support a particular function. In certain
2070 * cases where it makes no sense for a protocol to have a "do nothing"
2071 * function, some default processing is provided.
2072 */
2073
2074int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2075{
2076 return -EOPNOTSUPP;
2077}
2a91525c 2078EXPORT_SYMBOL(sock_no_bind);
1da177e4 2079
4ec93edb 2080int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2081 int len, int flags)
2082{
2083 return -EOPNOTSUPP;
2084}
2a91525c 2085EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2086
2087int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2088{
2089 return -EOPNOTSUPP;
2090}
2a91525c 2091EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2092
2093int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2094{
2095 return -EOPNOTSUPP;
2096}
2a91525c 2097EXPORT_SYMBOL(sock_no_accept);
1da177e4 2098
4ec93edb 2099int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2100 int *len, int peer)
2101{
2102 return -EOPNOTSUPP;
2103}
2a91525c 2104EXPORT_SYMBOL(sock_no_getname);
1da177e4 2105
2a91525c 2106unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2107{
2108 return 0;
2109}
2a91525c 2110EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2111
2112int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2113{
2114 return -EOPNOTSUPP;
2115}
2a91525c 2116EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2117
2118int sock_no_listen(struct socket *sock, int backlog)
2119{
2120 return -EOPNOTSUPP;
2121}
2a91525c 2122EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2123
2124int sock_no_shutdown(struct socket *sock, int how)
2125{
2126 return -EOPNOTSUPP;
2127}
2a91525c 2128EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2129
2130int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2131 char __user *optval, unsigned int optlen)
1da177e4
LT		 2132{
2133 return -EOPNOTSUPP;
2134}
2a91525c 2135EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2136
2137int sock_no_getsockopt(struct socket *sock, int level, int optname,
2138 char __user *optval, int __user *optlen)
2139{
2140 return -EOPNOTSUPP;
2141}
2a91525c 2142EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2143
2144int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2145 size_t len)
2146{
2147 return -EOPNOTSUPP;
2148}
2a91525c 2149EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2150
2151int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2152 size_t len, int flags)
2153{
2154 return -EOPNOTSUPP;
2155}
2a91525c 2156EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2157
2158int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2159{
2160 /* Mirror missing mmap method error code */
2161 return -ENODEV;
2162}
2a91525c 2163EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2164
2165ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2166{
2167 ssize_t res;
2168 struct msghdr msg = {.msg_flags = flags};
2169 struct kvec iov;
2170 char *kaddr = kmap(page);
2171 iov.iov_base = kaddr + offset;
2172 iov.iov_len = size;
2173 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2174 kunmap(page);
2175 return res;
2176}
2a91525c 2177EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2178
2179/*
2180 * Default Socket Callbacks
2181 */
2182
2183static void sock_def_wakeup(struct sock *sk)
2184{
43815482
ED		 2185 struct socket_wq *wq;
2186
2187 rcu_read_lock();
2188 wq = rcu_dereference(sk->sk_wq);
2189 if (wq_has_sleeper(wq))
2190 wake_up_interruptible_all(&wq->wait);
2191 rcu_read_unlock();
1da177e4
LT		 2192}
2193
2194static void sock_def_error_report(struct sock *sk)
2195{
43815482
ED		 2196 struct socket_wq *wq;
2197
2198 rcu_read_lock();
2199 wq = rcu_dereference(sk->sk_wq);
2200 if (wq_has_sleeper(wq))
2201 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2202 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2203 rcu_read_unlock();
1da177e4
LT		 2204}
2205
2206static void sock_def_readable(struct sock *sk, int len)
2207{
43815482
ED		 2208 struct socket_wq *wq;
2209
2210 rcu_read_lock();
2211 wq = rcu_dereference(sk->sk_wq);
2212 if (wq_has_sleeper(wq))
2c6607c6 2213 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2214 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2215 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2216 rcu_read_unlock();
1da177e4
LT		 2217}
2218
2219static void sock_def_write_space(struct sock *sk)
2220{
43815482
ED		 2221 struct socket_wq *wq;
2222
2223 rcu_read_lock();
1da177e4
LT		 2224
2225 /* Do not wake up a writer until he can make "significant"
2226 * progress. --DaveM
2227 */
e71a4783 2228 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2229 wq = rcu_dereference(sk->sk_wq);
2230 if (wq_has_sleeper(wq))
2231 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2232 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2233
2234 /* Should agree with poll, otherwise some programs break */
2235 if (sock_writeable(sk))
8d8ad9d7 2236 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2237 }
2238
43815482 2239 rcu_read_unlock();
1da177e4
LT		 2240}
2241
2242static void sock_def_destruct(struct sock *sk)
2243{
a51482bd 2244 kfree(sk->sk_protinfo);
1da177e4
LT		 2245}
2246
2247void sk_send_sigurg(struct sock *sk)
2248{
2249 if (sk->sk_socket && sk->sk_socket->file)
2250 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2251 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2252}
2a91525c 2253EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2254
2255void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2256 unsigned long expires)
2257{
2258 if (!mod_timer(timer, expires))
2259 sock_hold(sk);
2260}
1da177e4
LT		 2261EXPORT_SYMBOL(sk_reset_timer);
2262
2263void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2264{
25cc4ae9 2265 if (del_timer(timer))
1da177e4
LT		 2266 __sock_put(sk);
2267}
1da177e4
LT		 2268EXPORT_SYMBOL(sk_stop_timer);
2269
2270void sock_init_data(struct socket *sock, struct sock *sk)
2271{
2272 skb_queue_head_init(&sk->sk_receive_queue);
2273 skb_queue_head_init(&sk->sk_write_queue);
2274 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2275#ifdef CONFIG_NET_DMA
2276 skb_queue_head_init(&sk->sk_async_wait_queue);
2277#endif
1da177e4
LT		 2278
2279 sk->sk_send_head = NULL;
2280
2281 init_timer(&sk->sk_timer);
4ec93edb 2282
1da177e4
LT		 2283 sk->sk_allocation = GFP_KERNEL;
2284 sk->sk_rcvbuf = sysctl_rmem_default;
2285 sk->sk_sndbuf = sysctl_wmem_default;
2286 sk->sk_state = TCP_CLOSE;
972692e0 2287 sk_set_socket(sk, sock);
1da177e4
LT		 2288
2289 sock_set_flag(sk, SOCK_ZAPPED);
2290
e71a4783 2291 if (sock) {
1da177e4 2292 sk->sk_type = sock->type;
43815482 2293 sk->sk_wq = sock->wq;
1da177e4
LT		 2294 sock->sk = sk;
2295 } else
43815482 2296 sk->sk_wq = NULL;
1da177e4 2297
b6c6712a 2298 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2299 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2300 lockdep_set_class_and_name(&sk->sk_callback_lock,
2301 af_callback_keys + sk->sk_family,
2302 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2303
2304 sk->sk_state_change = sock_def_wakeup;
2305 sk->sk_data_ready = sock_def_readable;
2306 sk->sk_write_space = sock_def_write_space;
2307 sk->sk_error_report = sock_def_error_report;
2308 sk->sk_destruct = sock_def_destruct;
2309
5640f768
ED		 2310 sk->sk_frag.page = NULL;
2311 sk->sk_frag.offset = 0;
ef64a54f 2312 sk->sk_peek_off = -1;
1da177e4 2313
109f6e39
EB		 2314 sk->sk_peer_pid = NULL;
2315 sk->sk_peer_cred = NULL;
1da177e4
LT		 2316 sk->sk_write_pending = 0;
2317 sk->sk_rcvlowat = 1;
2318 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2319 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2320
f37f0afb 2321 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2322
5e25ba50 2323 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2324 /*
2325 * Before updating sk_refcnt, we must commit prior changes to memory
2326 * (Documentation/RCU/rculist_nulls.txt for details)
2327 */
2328 smp_wmb();
1da177e4 2329 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2330 atomic_set(&sk->sk_drops, 0);
1da177e4 2331}
2a91525c 2332EXPORT_SYMBOL(sock_init_data);
1da177e4 2333
b5606c2d 2334void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2335{
2336 might_sleep();
a5b5bb9a 2337 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2338 if (sk->sk_lock.owned)
1da177e4 2339 __lock_sock(sk);
d2e9117c 2340 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2341 spin_unlock(&sk->sk_lock.slock);
2342 /*
2343 * The sk_lock has mutex_lock() semantics here:
2344 */
fcc70d5f 2345 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2346 local_bh_enable();
1da177e4 2347}
fcc70d5f 2348EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2349
b5606c2d 2350void release_sock(struct sock *sk)
1da177e4 2351{
a5b5bb9a
IM		 2352 /*
2353 * The sk_lock has mutex_unlock() semantics:
2354 */
2355 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2356
2357 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2358 if (sk->sk_backlog.tail)
2359 __release_sock(sk);
46d3ceab 2360
cbbb5a25
ED		 2361 /* Warning : release_cb() might need to release sk ownership,
2362 * ie call sock_release_ownership(sk) before us.
2363 */
46d3ceab
ED		 2364 if (sk->sk_prot->release_cb)
2365 sk->sk_prot->release_cb(sk);
2366
cbbb5a25 2367 sock_release_ownership(sk);
a5b5bb9a
IM		 2368 if (waitqueue_active(&sk->sk_lock.wq))
2369 wake_up(&sk->sk_lock.wq);
2370 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2371}
2372EXPORT_SYMBOL(release_sock);
2373
8a74ad60
ED		 2374/**
2375 * lock_sock_fast - fast version of lock_sock
2376 * @sk: socket
2377 *
2378 * This version should be used for very small section, where process wont block
2379 * return false if fast path is taken
2380 * sk_lock.slock locked, owned = 0, BH disabled
2381 * return true if slow path is taken
2382 * sk_lock.slock unlocked, owned = 1, BH enabled
2383 */
2384bool lock_sock_fast(struct sock *sk)
2385{
2386 might_sleep();
2387 spin_lock_bh(&sk->sk_lock.slock);
2388
2389 if (!sk->sk_lock.owned)
2390 /*
2391 * Note : We must disable BH
2392 */
2393 return false;
2394
2395 __lock_sock(sk);
2396 sk->sk_lock.owned = 1;
2397 spin_unlock(&sk->sk_lock.slock);
2398 /*
2399 * The sk_lock has mutex_lock() semantics here:
2400 */
2401 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2402 local_bh_enable();
2403 return true;
2404}
2405EXPORT_SYMBOL(lock_sock_fast);
2406
1da177e4 2407int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2408{
b7aa0bf7 2409 struct timeval tv;
1da177e4 2410 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2411 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2412 tv = ktime_to_timeval(sk->sk_stamp);
2413 if (tv.tv_sec == -1)
1da177e4 2414 return -ENOENT;
b7aa0bf7
ED		 2415 if (tv.tv_sec == 0) {
2416 sk->sk_stamp = ktime_get_real();
2417 tv = ktime_to_timeval(sk->sk_stamp);
2418 }
2419 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2420}
1da177e4
LT		 2421EXPORT_SYMBOL(sock_get_timestamp);
2422
ae40eb1e
ED		 2423int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2424{
2425 struct timespec ts;
2426 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2427 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2428 ts = ktime_to_timespec(sk->sk_stamp);
2429 if (ts.tv_sec == -1)
2430 return -ENOENT;
2431 if (ts.tv_sec == 0) {
2432 sk->sk_stamp = ktime_get_real();
2433 ts = ktime_to_timespec(sk->sk_stamp);
2434 }
2435 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2436}
2437EXPORT_SYMBOL(sock_get_timestampns);
2438
20d49473 2439void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2440{
20d49473 2441 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2442 unsigned long previous_flags = sk->sk_flags;
2443
20d49473
PO		 2444 sock_set_flag(sk, flag);
2445 /*
2446 * we just set one of the two flags which require net
2447 * time stamping, but time stamping might have been on
2448 * already because of the other one
2449 */
08e29af3 2450 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2451 net_enable_timestamp();
1da177e4
LT		 2452 }
2453}
1da177e4
LT		 2454
2455/*
2456 * Get a socket option on an socket.
2457 *
2458 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2459 * asynchronous errors should be reported by getsockopt. We assume
2460 * this means if you specify SO_ERROR (otherwise whats the point of it).
2461 */
2462int sock_common_getsockopt(struct socket *sock, int level, int optname,
2463 char __user *optval, int __user *optlen)
2464{
2465 struct sock *sk = sock->sk;
2466
2467 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2468}
1da177e4
LT		 2469EXPORT_SYMBOL(sock_common_getsockopt);
2470
3fdadf7d 2471#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2472int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2473 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2474{
2475 struct sock *sk = sock->sk;
2476
1e51f951 2477 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2478 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2479 optval, optlen);
3fdadf7d
DM		 2480 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2481}
2482EXPORT_SYMBOL(compat_sock_common_getsockopt);
2483#endif
2484
1da177e4
LT		 2485int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2486 struct msghdr *msg, size_t size, int flags)
2487{
2488 struct sock *sk = sock->sk;
2489 int addr_len = 0;
2490 int err;
2491
2492 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2493 flags & ~MSG_DONTWAIT, &addr_len);
2494 if (err >= 0)
2495 msg->msg_namelen = addr_len;
2496 return err;
2497}
1da177e4
LT		 2498EXPORT_SYMBOL(sock_common_recvmsg);
2499
2500/*
2501 * Set socket options on an inet socket.
2502 */
2503int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2504 char __user *optval, unsigned int optlen)
1da177e4
LT		 2505{
2506 struct sock *sk = sock->sk;
2507
2508 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2509}
1da177e4
LT		 2510EXPORT_SYMBOL(sock_common_setsockopt);
2511
3fdadf7d 2512#ifdef CONFIG_COMPAT
543d9cfe 2513int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2514 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2515{
2516 struct sock *sk = sock->sk;
2517
543d9cfe
ACM		 2518 if (sk->sk_prot->compat_setsockopt != NULL)
2519 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2520 optval, optlen);
3fdadf7d
DM		 2521 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2522}
2523EXPORT_SYMBOL(compat_sock_common_setsockopt);
2524#endif
2525
1da177e4
LT		 2526void sk_common_release(struct sock *sk)
2527{
2528 if (sk->sk_prot->destroy)
2529 sk->sk_prot->destroy(sk);
2530
2531 /*
2532 * Observation: when sock_common_release is called, processes have
2533 * no access to socket. But net still has.
2534 * Step one, detach it from networking:
2535 *
2536 * A. Remove from hash tables.
2537 */
2538
2539 sk->sk_prot->unhash(sk);
2540
2541 /*
2542 * In this point socket cannot receive new packets, but it is possible
2543 * that some packets are in flight because some CPU runs receiver and
2544 * did hash table lookup before we unhashed socket. They will achieve
2545 * receive queue and will be purged by socket destructor.
2546 *
2547 * Also we still have packets pending on receive queue and probably,
2548 * our own packets waiting in device queues. sock_destroy will drain
2549 * receive queue, but transmitted packets will delay socket destruction
2550 * until the last reference will be released.
2551 */
2552
2553 sock_orphan(sk);
2554
2555 xfrm_sk_free_policy(sk);
2556
e6848976 2557 sk_refcnt_debug_release(sk);
5640f768
ED		 2558
2559 if (sk->sk_frag.page) {
2560 put_page(sk->sk_frag.page);
2561 sk->sk_frag.page = NULL;
2562 }
2563
1da177e4
LT		 2564 sock_put(sk);
2565}
1da177e4
LT		 2566EXPORT_SYMBOL(sk_common_release);
2567
13ff3d6f
PE		 2568#ifdef CONFIG_PROC_FS
2569#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2570struct prot_inuse {
2571 int val[PROTO_INUSE_NR];
2572};
13ff3d6f
PE		 2573
2574static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2575
2576#ifdef CONFIG_NET_NS
2577void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2578{
d6d9ca0f 2579 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2580}
2581EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2582
2583int sock_prot_inuse_get(struct net *net, struct proto *prot)
2584{
2585 int cpu, idx = prot->inuse_idx;
2586 int res = 0;
2587
2588 for_each_possible_cpu(cpu)
2589 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2590
2591 return res >= 0 ? res : 0;
2592}
2593EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2594
2c8c1e72 2595static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2596{
2597 net->core.inuse = alloc_percpu(struct prot_inuse);
2598 return net->core.inuse ? 0 : -ENOMEM;
2599}
2600
2c8c1e72 2601static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2602{
2603 free_percpu(net->core.inuse);
2604}
2605
2606static struct pernet_operations net_inuse_ops = {
2607 .init = sock_inuse_init_net,
2608 .exit = sock_inuse_exit_net,
2609};
2610
2611static __init int net_inuse_init(void)
2612{
2613 if (register_pernet_subsys(&net_inuse_ops))
2614 panic("Cannot initialize net inuse counters");
2615
2616 return 0;
2617}
2618
2619core_initcall(net_inuse_init);
2620#else
1338d466
PE		 2621static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2622
c29a0bc4 2623void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2624{
d6d9ca0f 2625 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2626}
2627EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2628
c29a0bc4 2629int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2630{
2631 int cpu, idx = prot->inuse_idx;
2632 int res = 0;
2633
2634 for_each_possible_cpu(cpu)
2635 res += per_cpu(prot_inuse, cpu).val[idx];
2636
2637 return res >= 0 ? res : 0;
2638}
2639EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2640#endif
13ff3d6f
PE		 2641
2642static void assign_proto_idx(struct proto *prot)
2643{
2644 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2645
2646 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2647 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2648 return;
2649 }
2650
2651 set_bit(prot->inuse_idx, proto_inuse_idx);
2652}
2653
2654static void release_proto_idx(struct proto *prot)
2655{
2656 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2657 clear_bit(prot->inuse_idx, proto_inuse_idx);
2658}
2659#else
2660static inline void assign_proto_idx(struct proto *prot)
2661{
2662}
2663
2664static inline void release_proto_idx(struct proto *prot)
2665{
2666}
2667#endif
2668
b733c007
PE		 2669int proto_register(struct proto *prot, int alloc_slab)
2670{
1da177e4
LT		 2671 if (alloc_slab) {
2672 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2673 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2674 NULL);
1da177e4
LT		 2675
2676 if (prot->slab == NULL) {
e005d193
JP		 2677 pr_crit("%s: Can't create sock SLAB cache!\n",
2678 prot->name);
60e7663d 2679 goto out;
1da177e4 2680 }
2e6599cb
ACM		 2681
2682 if (prot->rsk_prot != NULL) {
faf23422 2683 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2684 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2685 goto out_free_sock_slab;
2686
7e56b5d6 2687 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2688 prot->rsk_prot->obj_size, 0,
20c2df83 2689 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2690
2691 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2692 pr_crit("%s: Can't create request sock SLAB cache!\n",
2693 prot->name);
2e6599cb
ACM		 2694 goto out_free_request_sock_slab_name;
2695 }
2696 }
8feaf0c0 2697
6d6ee43e 2698 if (prot->twsk_prot != NULL) {
faf23422 2699 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2700
7e56b5d6 2701 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2702 goto out_free_request_sock_slab;
2703
6d6ee43e 2704 prot->twsk_prot->twsk_slab =
7e56b5d6 2705 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2706 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2707 0,
2708 SLAB_HWCACHE_ALIGN |
2709 prot->slab_flags,
20c2df83 2710 NULL);
6d6ee43e 2711 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2712 goto out_free_timewait_sock_slab_name;
2713 }
1da177e4
LT		 2714 }
2715
36b77a52 2716 mutex_lock(&proto_list_mutex);
1da177e4 2717 list_add(&prot->node, &proto_list);
13ff3d6f 2718 assign_proto_idx(prot);
36b77a52 2719 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2720 return 0;
2721
8feaf0c0 2722out_free_timewait_sock_slab_name:
7e56b5d6 2723 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2724out_free_request_sock_slab:
2725 if (prot->rsk_prot && prot->rsk_prot->slab) {
2726 kmem_cache_destroy(prot->rsk_prot->slab);
2727 prot->rsk_prot->slab = NULL;
2728 }
2e6599cb 2729out_free_request_sock_slab_name:
72150e9b
DC		 2730 if (prot->rsk_prot)
2731 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2732out_free_sock_slab:
2733 kmem_cache_destroy(prot->slab);
2734 prot->slab = NULL;
b733c007
PE		 2735out:
2736 return -ENOBUFS;
1da177e4 2737}
1da177e4
LT		 2738EXPORT_SYMBOL(proto_register);
2739
2740void proto_unregister(struct proto *prot)
2741{
36b77a52 2742 mutex_lock(&proto_list_mutex);
13ff3d6f 2743 release_proto_idx(prot);
0a3f4358 2744 list_del(&prot->node);
36b77a52 2745 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2746
2747 if (prot->slab != NULL) {
2748 kmem_cache_destroy(prot->slab);
2749 prot->slab = NULL;
2750 }
2751
2e6599cb 2752 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2753 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2754 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2755 prot->rsk_prot->slab = NULL;
2756 }
2757
6d6ee43e 2758 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2759 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2760 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2761 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2762 }
1da177e4 2763}
1da177e4
LT		 2764EXPORT_SYMBOL(proto_unregister);
2765
2766#ifdef CONFIG_PROC_FS
1da177e4 2767static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2768 __acquires(proto_list_mutex)
1da177e4 2769{
36b77a52 2770 mutex_lock(&proto_list_mutex);
60f0438a 2771 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2772}
2773
2774static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2775{
60f0438a 2776 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2777}
2778
2779static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2780 __releases(proto_list_mutex)
1da177e4 2781{
36b77a52 2782 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2783}
2784
2785static char proto_method_implemented(const void *method)
2786{
2787 return method == NULL ? 'n' : 'y';
2788}
180d8cd9
GC		 2789static long sock_prot_memory_allocated(struct proto *proto)
2790{
cb75a36c 2791 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2792}
2793
2794static char *sock_prot_memory_pressure(struct proto *proto)
2795{
2796 return proto->memory_pressure != NULL ?
2797 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2798}
1da177e4
LT		 2799
2800static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2801{
180d8cd9 2802
8d987e5c 2803 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2804 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2805 proto->name,
2806 proto->obj_size,
14e943db 2807 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2808 sock_prot_memory_allocated(proto),
2809 sock_prot_memory_pressure(proto),
1da177e4
LT		 2810 proto->max_header,
2811 proto->slab == NULL ? "no" : "yes",
2812 module_name(proto->owner),
2813 proto_method_implemented(proto->close),
2814 proto_method_implemented(proto->connect),
2815 proto_method_implemented(proto->disconnect),
2816 proto_method_implemented(proto->accept),
2817 proto_method_implemented(proto->ioctl),
2818 proto_method_implemented(proto->init),
2819 proto_method_implemented(proto->destroy),
2820 proto_method_implemented(proto->shutdown),
2821 proto_method_implemented(proto->setsockopt),
2822 proto_method_implemented(proto->getsockopt),
2823 proto_method_implemented(proto->sendmsg),
2824 proto_method_implemented(proto->recvmsg),
2825 proto_method_implemented(proto->sendpage),
2826 proto_method_implemented(proto->bind),
2827 proto_method_implemented(proto->backlog_rcv),
2828 proto_method_implemented(proto->hash),
2829 proto_method_implemented(proto->unhash),
2830 proto_method_implemented(proto->get_port),
2831 proto_method_implemented(proto->enter_memory_pressure));
2832}
2833
2834static int proto_seq_show(struct seq_file *seq, void *v)
2835{
60f0438a 2836 if (v == &proto_list)
1da177e4
LT		 2837 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2838 "protocol",
2839 "size",
2840 "sockets",
2841 "memory",
2842 "press",
2843 "maxhdr",
2844 "slab",
2845 "module",
2846 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2847 else
60f0438a 2848 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2849 return 0;
2850}
2851
f690808e 2852static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2853 .start = proto_seq_start,
2854 .next = proto_seq_next,
2855 .stop = proto_seq_stop,
2856 .show = proto_seq_show,
2857};
2858
2859static int proto_seq_open(struct inode *inode, struct file *file)
2860{
14e943db
ED		 2861 return seq_open_net(inode, file, &proto_seq_ops,
2862 sizeof(struct seq_net_private));
1da177e4
LT		 2863}
2864
9a32144e 2865static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2866 .owner = THIS_MODULE,
2867 .open = proto_seq_open,
2868 .read = seq_read,
2869 .llseek = seq_lseek,
14e943db
ED		 2870 .release = seq_release_net,
2871};
2872
2873static __net_init int proto_init_net(struct net *net)
2874{
d4beaa66 2875 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2876 return -ENOMEM;
2877
2878 return 0;
2879}
2880
2881static __net_exit void proto_exit_net(struct net *net)
2882{
ece31ffd 2883 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2884}
2885
2886
2887static __net_initdata struct pernet_operations proto_net_ops = {
2888 .init = proto_init_net,
2889 .exit = proto_exit_net,
1da177e4
LT		 2890};
2891
2892static int __init proto_init(void)
2893{
14e943db 2894 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2895}
2896
2897subsys_initcall(proto_init);
2898
2899#endif /* PROC_FS */
kernel source for telekom puls (alcatel/mediatek)