projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
drivers: net: xgene: fix Tx flow control
[deliverable/linux.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 95#include <linux/errno.h>
cb820f8e 96#include <linux/errqueue.h>
1da177e4
LT		 97#include <linux/types.h>
98#include <linux/socket.h>
99#include <linux/in.h>
100#include <linux/kernel.h>
1da177e4
LT		 101#include <linux/module.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
104#include <linux/sched.h>
105#include <linux/timer.h>
106#include <linux/string.h>
107#include <linux/sockios.h>
108#include <linux/net.h>
109#include <linux/mm.h>
110#include <linux/slab.h>
111#include <linux/interrupt.h>
112#include <linux/poll.h>
113#include <linux/tcp.h>
114#include <linux/init.h>
a1f8e7f7 115#include <linux/highmem.h>
3f551f94 116#include <linux/user_namespace.h>
c5905afb 117#include <linux/static_key.h>
3969eb38 118#include <linux/memcontrol.h>
8c1ae10d 119#include <linux/prefetch.h>
1da177e4
LT		 120
121#include <asm/uaccess.h>
1da177e4
LT		 122
123#include <linux/netdevice.h>
124#include <net/protocol.h>
125#include <linux/skbuff.h>
457c4cbc 126#include <net/net_namespace.h>
2e6599cb 127#include <net/request_sock.h>
1da177e4 128#include <net/sock.h>
20d49473 129#include <linux/net_tstamp.h>
1da177e4
LT		 130#include <net/xfrm.h>
131#include <linux/ipsec.h>
f8451725 132#include <net/cls_cgroup.h>
5bc1421e 133#include <net/netprio_cgroup.h>
eb4cb008 134#include <linux/sock_diag.h>
1da177e4
LT		 135
136#include <linux/filter.h>
137
3847ce32
SM		 138#include <trace/events/sock.h>
139
1da177e4
LT		 140#ifdef CONFIG_INET
141#include <net/tcp.h>
142#endif
143
076bb0c8 144#include <net/busy_poll.h>
06021292 145
36b77a52 146static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 147static LIST_HEAD(proto_list);
148
a3b299da
EB		 149/**
150 * sk_ns_capable - General socket capability test
151 * @sk: Socket to use a capability on or through
152 * @user_ns: The user namespace of the capability to use
153 * @cap: The capability to use
154 *
155 * Test to see if the opener of the socket had when the socket was
156 * created and the current process has the capability @cap in the user
157 * namespace @user_ns.
158 */
159bool sk_ns_capable(const struct sock *sk,
160 struct user_namespace *user_ns, int cap)
161{
162 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
163 ns_capable(user_ns, cap);
164}
165EXPORT_SYMBOL(sk_ns_capable);
166
167/**
168 * sk_capable - Socket global capability test
169 * @sk: Socket to use a capability on or through
e793c0f7 170 * @cap: The global capability to use
a3b299da
EB		 171 *
172 * Test to see if the opener of the socket had when the socket was
173 * created and the current process has the capability @cap in all user
174 * namespaces.
175 */
176bool sk_capable(const struct sock *sk, int cap)
177{
178 return sk_ns_capable(sk, &init_user_ns, cap);
179}
180EXPORT_SYMBOL(sk_capable);
181
182/**
183 * sk_net_capable - Network namespace socket capability test
184 * @sk: Socket to use a capability on or through
185 * @cap: The capability to use
186 *
e793c0f7 187 * Test to see if the opener of the socket had when the socket was created
a3b299da
EB		 188 * and the current process has the capability @cap over the network namespace
189 * the socket is a member of.
190 */
191bool sk_net_capable(const struct sock *sk, int cap)
192{
193 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
194}
195EXPORT_SYMBOL(sk_net_capable);
196
197
c255a458 198#ifdef CONFIG_MEMCG_KMEM
1d62e436 199int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
d1a4c0b3
GC		 200{
201 struct proto *proto;
202 int ret = 0;
203
36b77a52 204 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 205 list_for_each_entry(proto, &proto_list, node) {
206 if (proto->init_cgroup) {
1d62e436 207 ret = proto->init_cgroup(memcg, ss);
d1a4c0b3
GC		 208 if (ret)
209 goto out;
210 }
211 }
212
36b77a52 213 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 214 return ret;
215out:
216 list_for_each_entry_continue_reverse(proto, &proto_list, node)
217 if (proto->destroy_cgroup)
1d62e436 218 proto->destroy_cgroup(memcg);
36b77a52 219 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 220 return ret;
221}
222
1d62e436 223void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
d1a4c0b3
GC		 224{
225 struct proto *proto;
226
36b77a52 227 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 228 list_for_each_entry_reverse(proto, &proto_list, node)
229 if (proto->destroy_cgroup)
1d62e436 230 proto->destroy_cgroup(memcg);
36b77a52 231 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 232}
233#endif
234
da21f24d
IM		 235/*
236 * Each address family might have different locking rules, so we have
237 * one slock key per address family:
238 */
a5b5bb9a
IM		 239static struct lock_class_key af_family_keys[AF_MAX];
240static struct lock_class_key af_family_slock_keys[AF_MAX];
241
cbda4eaf 242#if defined(CONFIG_MEMCG_KMEM)
c5905afb 243struct static_key memcg_socket_limit_enabled;
e1aab161 244EXPORT_SYMBOL(memcg_socket_limit_enabled);
cbda4eaf 245#endif
e1aab161 246
a5b5bb9a
IM		 247/*
248 * Make lock validator output more readable. (we pre-construct these
249 * strings build-time, so that runtime initialization of socket
250 * locks is fast):
251 */
36cbd3dc 252static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 253 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
254 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
255 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
256 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
257 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
258 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
259 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 260 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 261 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 262 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 263 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 264 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 265 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
456db6a4 266 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 267};
36cbd3dc 268static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 269 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
270 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
271 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
272 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
273 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
274 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
275 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 276 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 277 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 278 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 279 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 280 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 281 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 282 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 283};
36cbd3dc 284static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 285 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
286 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
287 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
288 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
289 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
290 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
291 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 292 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 293 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 294 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 295 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 296 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 297 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 298 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 299};
da21f24d
IM		 300
301/*
302 * sk_callback_lock locking rules are per-address-family,
303 * so split the lock classes by using a per-AF key:
304 */
305static struct lock_class_key af_callback_keys[AF_MAX];
306
1da177e4
LT		 307/* Take into consideration the size of the struct sk_buff overhead in the
308 * determination of these values, since that is non-constant across
309 * platforms. This makes socket queueing behavior and performance
310 * not depend upon such differences.
311 */
312#define _SK_MEM_PACKETS 256
87fb4b7b 313#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 314#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
315#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
316
317/* Run time adjustable parameters. */
ab32ea5d 318__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 319EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 320__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 321EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 322__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
323__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 324
25985edc 325/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 326int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 327EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 328
b245be1f
WB		 329int sysctl_tstamp_allow_data __read_mostly = 1;
330
c93bdd0e
MG		 331struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
332EXPORT_SYMBOL_GPL(memalloc_socks);
333
7cb02404
MG		 334/**
335 * sk_set_memalloc - sets %SOCK_MEMALLOC
336 * @sk: socket to set it on
337 *
338 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
339 * It's the responsibility of the admin to adjust min_free_kbytes
340 * to meet the requirements
341 */
342void sk_set_memalloc(struct sock *sk)
343{
344 sock_set_flag(sk, SOCK_MEMALLOC);
345 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 346 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 347}
348EXPORT_SYMBOL_GPL(sk_set_memalloc);
349
350void sk_clear_memalloc(struct sock *sk)
351{
352 sock_reset_flag(sk, SOCK_MEMALLOC);
353 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 354 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 355
356 /*
357 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
5d753610
MG		 358 * progress of swapping. SOCK_MEMALLOC may be cleared while
359 * it has rmem allocations due to the last swapfile being deactivated
360 * but there is a risk that the socket is unusable due to exceeding
361 * the rmem limits. Reclaim the reserves and obey rmem limits again.
c76562b6 362 */
5d753610 363 sk_mem_reclaim(sk);
7cb02404
MG		 364}
365EXPORT_SYMBOL_GPL(sk_clear_memalloc);
366
b4b9e355
MG		 367int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
368{
369 int ret;
370 unsigned long pflags = current->flags;
371
372 /* these should have been dropped before queueing */
373 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
374
375 current->flags |= PF_MEMALLOC;
376 ret = sk->sk_backlog_rcv(sk, skb);
377 tsk_restore_flags(current, pflags, PF_MEMALLOC);
378
379 return ret;
380}
381EXPORT_SYMBOL(__sk_backlog_rcv);
382
1da177e4
LT		 383static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
384{
385 struct timeval tv;
386
387 if (optlen < sizeof(tv))
388 return -EINVAL;
389 if (copy_from_user(&tv, optval, sizeof(tv)))
390 return -EFAULT;
ba78073e
VA		 391 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
392 return -EDOM;
1da177e4 393
ba78073e 394 if (tv.tv_sec < 0) {
6f11df83
AM		 395 static int warned __read_mostly;
396
ba78073e 397 *timeo_p = 0;
50aab54f 398 if (warned < 10 && net_ratelimit()) {
ba78073e 399 warned++;
e005d193
JP		 400 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
401 __func__, current->comm, task_pid_nr(current));
50aab54f 402 }
ba78073e
VA		 403 return 0;
404 }
1da177e4
LT		 405 *timeo_p = MAX_SCHEDULE_TIMEOUT;
406 if (tv.tv_sec == 0 && tv.tv_usec == 0)
407 return 0;
408 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
409 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
410 return 0;
411}
412
413static void sock_warn_obsolete_bsdism(const char *name)
414{
415 static int warned;
416 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 417 if (strcmp(warncomm, current->comm) && warned < 5) {
418 strcpy(warncomm, current->comm);
e005d193
JP		 419 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
420 warncomm, name);
1da177e4
LT		 421 warned++;
422 }
423}
424
080a270f
HFS		 425static bool sock_needs_netstamp(const struct sock *sk)
426{
427 switch (sk->sk_family) {
428 case AF_UNSPEC:
429 case AF_UNIX:
430 return false;
431 default:
432 return true;
433 }
434}
435
08e29af3 436static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 437{
08e29af3
ED		 438 if (sk->sk_flags & flags) {
439 sk->sk_flags &= ~flags;
080a270f
HFS		 440 if (sock_needs_netstamp(sk) &&
441 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 442 net_disable_timestamp();
1da177e4
LT		 443 }
444}
445
446
f0088a50
DV		 447int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
448{
766e9037 449 int err;
3b885787
NH		 450 unsigned long flags;
451 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 452
0fd7bac6 453 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 454 atomic_inc(&sk->sk_drops);
3847ce32 455 trace_sock_rcvqueue_full(sk, skb);
766e9037 456 return -ENOMEM;
f0088a50
DV		 457 }
458
fda9ef5d 459 err = sk_filter(sk, skb);
f0088a50 460 if (err)
766e9037 461 return err;
f0088a50 462
c76562b6 463 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 464 atomic_inc(&sk->sk_drops);
465 return -ENOBUFS;
3ab224be
HA		 466 }
467
f0088a50
DV		 468 skb->dev = NULL;
469 skb_set_owner_r(skb, sk);
49ad9599 470
7fee226a
ED		 471 /* we escape from rcu protected region, make sure we dont leak
472 * a norefcounted dst
473 */
474 skb_dst_force(skb);
475
3b885787 476 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 477 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 478 __skb_queue_tail(list, skb);
479 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 480
481 if (!sock_flag(sk, SOCK_DEAD))
676d2369 482 sk->sk_data_ready(sk);
766e9037 483 return 0;
f0088a50
DV		 484}
485EXPORT_SYMBOL(sock_queue_rcv_skb);
486
58a5a7b9 487int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 488{
489 int rc = NET_RX_SUCCESS;
490
fda9ef5d 491 if (sk_filter(sk, skb))
f0088a50
DV		 492 goto discard_and_relse;
493
494 skb->dev = NULL;
495
274f482d 496 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 497 atomic_inc(&sk->sk_drops);
498 goto discard_and_relse;
499 }
58a5a7b9
ACM		 500 if (nested)
501 bh_lock_sock_nested(sk);
502 else
503 bh_lock_sock(sk);
a5b5bb9a
IM		 504 if (!sock_owned_by_user(sk)) {
505 /*
506 * trylock + unlock semantics:
507 */
508 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
509
c57943a1 510 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 511
512 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 513 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 514 bh_unlock_sock(sk);
515 atomic_inc(&sk->sk_drops);
516 goto discard_and_relse;
517 }
518
f0088a50
DV		 519 bh_unlock_sock(sk);
520out:
521 sock_put(sk);
522 return rc;
523discard_and_relse:
524 kfree_skb(skb);
525 goto out;
526}
527EXPORT_SYMBOL(sk_receive_skb);
528
529struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
530{
b6c6712a 531 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 532
533 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 534 sk_tx_queue_clear(sk);
a9b3cd7f 535 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 536 dst_release(dst);
537 return NULL;
538 }
539
540 return dst;
541}
542EXPORT_SYMBOL(__sk_dst_check);
543
544struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
545{
546 struct dst_entry *dst = sk_dst_get(sk);
547
548 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
549 sk_dst_reset(sk);
550 dst_release(dst);
551 return NULL;
552 }
553
554 return dst;
555}
556EXPORT_SYMBOL(sk_dst_check);
557
c91f6df2
BH		 558static int sock_setbindtodevice(struct sock *sk, char __user *optval,
559 int optlen)
4878809f
DM		 560{
561 int ret = -ENOPROTOOPT;
562#ifdef CONFIG_NETDEVICES
3b1e0a65 563 struct net *net = sock_net(sk);
4878809f
DM		 564 char devname[IFNAMSIZ];
565 int index;
566
567 /* Sorry... */
568 ret = -EPERM;
5e1fccc0 569 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 570 goto out;
571
572 ret = -EINVAL;
573 if (optlen < 0)
574 goto out;
575
576 /* Bind this socket to a particular device like "eth0",
577 * as specified in the passed interface name. If the
578 * name is "" or the option length is zero the socket
579 * is not bound.
580 */
581 if (optlen > IFNAMSIZ - 1)
582 optlen = IFNAMSIZ - 1;
583 memset(devname, 0, sizeof(devname));
584
585 ret = -EFAULT;
586 if (copy_from_user(devname, optval, optlen))
587 goto out;
588
000ba2e4
DM		 589 index = 0;
590 if (devname[0] != '\0') {
bf8e56bf 591 struct net_device *dev;
4878809f 592
bf8e56bf
ED		 593 rcu_read_lock();
594 dev = dev_get_by_name_rcu(net, devname);
595 if (dev)
596 index = dev->ifindex;
597 rcu_read_unlock();
4878809f
DM		 598 ret = -ENODEV;
599 if (!dev)
600 goto out;
4878809f
DM		 601 }
602
603 lock_sock(sk);
604 sk->sk_bound_dev_if = index;
605 sk_dst_reset(sk);
606 release_sock(sk);
607
608 ret = 0;
609
610out:
611#endif
612
613 return ret;
614}
615
c91f6df2
BH		 616static int sock_getbindtodevice(struct sock *sk, char __user *optval,
617 int __user *optlen, int len)
618{
619 int ret = -ENOPROTOOPT;
620#ifdef CONFIG_NETDEVICES
621 struct net *net = sock_net(sk);
c91f6df2 622 char devname[IFNAMSIZ];
c91f6df2
BH		 623
624 if (sk->sk_bound_dev_if == 0) {
625 len = 0;
626 goto zero;
627 }
628
629 ret = -EINVAL;
630 if (len < IFNAMSIZ)
631 goto out;
632
5dbe7c17
NS		 633 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
634 if (ret)
c91f6df2 635 goto out;
c91f6df2
BH		 636
637 len = strlen(devname) + 1;
638
639 ret = -EFAULT;
640 if (copy_to_user(optval, devname, len))
641 goto out;
642
643zero:
644 ret = -EFAULT;
645 if (put_user(len, optlen))
646 goto out;
647
648 ret = 0;
649
650out:
651#endif
652
653 return ret;
654}
655
c0ef877b
PE		 656static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
657{
658 if (valbool)
659 sock_set_flag(sk, bit);
660 else
661 sock_reset_flag(sk, bit);
662}
663
f60e5990 664bool sk_mc_loop(struct sock *sk)
665{
666 if (dev_recursion_level())
667 return false;
668 if (!sk)
669 return true;
670 switch (sk->sk_family) {
671 case AF_INET:
672 return inet_sk(sk)->mc_loop;
673#if IS_ENABLED(CONFIG_IPV6)
674 case AF_INET6:
675 return inet6_sk(sk)->mc_loop;
676#endif
677 }
678 WARN_ON(1);
679 return true;
680}
681EXPORT_SYMBOL(sk_mc_loop);
682
1da177e4
LT		 683/*
684 * This is meant for all protocols to use and covers goings on
685 * at the socket level. Everything here is generic.
686 */
687
688int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 689 char __user *optval, unsigned int optlen)
1da177e4 690{
2a91525c 691 struct sock *sk = sock->sk;
1da177e4
LT		 692 int val;
693 int valbool;
694 struct linger ling;
695 int ret = 0;
4ec93edb 696
1da177e4
LT		 697 /*
698 * Options without arguments
699 */
700
4878809f 701 if (optname == SO_BINDTODEVICE)
c91f6df2 702 return sock_setbindtodevice(sk, optval, optlen);
4878809f 703
e71a4783
SH		 704 if (optlen < sizeof(int))
705 return -EINVAL;
4ec93edb 706
1da177e4
LT		 707 if (get_user(val, (int __user *)optval))
708 return -EFAULT;
4ec93edb 709
2a91525c 710 valbool = val ? 1 : 0;
1da177e4
LT		 711
712 lock_sock(sk);
713
2a91525c 714 switch (optname) {
e71a4783 715 case SO_DEBUG:
2a91525c 716 if (val && !capable(CAP_NET_ADMIN))
e71a4783 717 ret = -EACCES;
2a91525c 718 else
c0ef877b 719 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 720 break;
721 case SO_REUSEADDR:
4a17fd52 722 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 723 break;
055dc21a
TH		 724 case SO_REUSEPORT:
725 sk->sk_reuseport = valbool;
726 break;
e71a4783 727 case SO_TYPE:
49c794e9 728 case SO_PROTOCOL:
0d6038ee 729 case SO_DOMAIN:
e71a4783
SH		 730 case SO_ERROR:
731 ret = -ENOPROTOOPT;
732 break;
733 case SO_DONTROUTE:
c0ef877b 734 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 735 break;
736 case SO_BROADCAST:
737 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
738 break;
739 case SO_SNDBUF:
740 /* Don't error on this BSD doesn't and if you think
82981930
ED		 741 * about it this is right. Otherwise apps have to
742 * play 'guess the biggest size' games. RCVBUF/SNDBUF
743 * are treated in BSD as hints
744 */
745 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 746set_sndbuf:
e71a4783 747 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 748 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
749 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 750 sk->sk_write_space(sk);
751 break;
1da177e4 752
e71a4783
SH		 753 case SO_SNDBUFFORCE:
754 if (!capable(CAP_NET_ADMIN)) {
755 ret = -EPERM;
756 break;
757 }
758 goto set_sndbuf;
b0573dea 759
e71a4783
SH		 760 case SO_RCVBUF:
761 /* Don't error on this BSD doesn't and if you think
82981930
ED		 762 * about it this is right. Otherwise apps have to
763 * play 'guess the biggest size' games. RCVBUF/SNDBUF
764 * are treated in BSD as hints
765 */
766 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 767set_rcvbuf:
e71a4783
SH		 768 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
769 /*
770 * We double it on the way in to account for
771 * "struct sk_buff" etc. overhead. Applications
772 * assume that the SO_RCVBUF setting they make will
773 * allow that much actual data to be received on that
774 * socket.
775 *
776 * Applications are unaware that "struct sk_buff" and
777 * other overheads allocate from the receive buffer
778 * during socket buffer allocation.
779 *
780 * And after considering the possible alternatives,
781 * returning the value we actually used in getsockopt
782 * is the most desirable behavior.
783 */
82981930 784 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 785 break;
786
787 case SO_RCVBUFFORCE:
788 if (!capable(CAP_NET_ADMIN)) {
789 ret = -EPERM;
1da177e4 790 break;
e71a4783
SH		 791 }
792 goto set_rcvbuf;
1da177e4 793
e71a4783 794 case SO_KEEPALIVE:
1da177e4 795#ifdef CONFIG_INET
3e10986d
ED		 796 if (sk->sk_protocol == IPPROTO_TCP &&
797 sk->sk_type == SOCK_STREAM)
e71a4783 798 tcp_set_keepalive(sk, valbool);
1da177e4 799#endif
e71a4783
SH		 800 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
801 break;
802
803 case SO_OOBINLINE:
804 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
805 break;
806
807 case SO_NO_CHECK:
28448b80 808 sk->sk_no_check_tx = valbool;
e71a4783
SH		 809 break;
810
811 case SO_PRIORITY:
5e1fccc0
EB		 812 if ((val >= 0 && val <= 6) ||
813 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 814 sk->sk_priority = val;
815 else
816 ret = -EPERM;
817 break;
818
819 case SO_LINGER:
820 if (optlen < sizeof(ling)) {
821 ret = -EINVAL; /* 1003.1g */
1da177e4 822 break;
e71a4783 823 }
2a91525c 824 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 825 ret = -EFAULT;
1da177e4 826 break;
e71a4783
SH		 827 }
828 if (!ling.l_onoff)
829 sock_reset_flag(sk, SOCK_LINGER);
830 else {
1da177e4 831#if (BITS_PER_LONG == 32)
e71a4783
SH		 832 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
833 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 834 else
e71a4783
SH		 835#endif
836 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
837 sock_set_flag(sk, SOCK_LINGER);
838 }
839 break;
840
841 case SO_BSDCOMPAT:
842 sock_warn_obsolete_bsdism("setsockopt");
843 break;
844
845 case SO_PASSCRED:
846 if (valbool)
847 set_bit(SOCK_PASSCRED, &sock->flags);
848 else
849 clear_bit(SOCK_PASSCRED, &sock->flags);
850 break;
851
852 case SO_TIMESTAMP:
92f37fd2 853 case SO_TIMESTAMPNS:
e71a4783 854 if (valbool) {
92f37fd2
ED		 855 if (optname == SO_TIMESTAMP)
856 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
857 else
858 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 859 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 860 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 861 } else {
e71a4783 862 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 863 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
864 }
e71a4783
SH		 865 break;
866
20d49473
PO		 867 case SO_TIMESTAMPING:
868 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 869 ret = -EINVAL;
20d49473
PO		 870 break;
871 }
b245be1f 872
09c2d251 873 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765
WB		 874 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
875 if (sk->sk_protocol == IPPROTO_TCP) {
876 if (sk->sk_state != TCP_ESTABLISHED) {
877 ret = -EINVAL;
878 break;
879 }
880 sk->sk_tskey = tcp_sk(sk)->snd_una;
881 } else {
882 sk->sk_tskey = 0;
883 }
884 }
b9f40e21 885 sk->sk_tsflags = val;
20d49473
PO		 886 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
887 sock_enable_timestamp(sk,
888 SOCK_TIMESTAMPING_RX_SOFTWARE);
889 else
890 sock_disable_timestamp(sk,
08e29af3 891 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 892 break;
893
e71a4783
SH		 894 case SO_RCVLOWAT:
895 if (val < 0)
896 val = INT_MAX;
897 sk->sk_rcvlowat = val ? : 1;
898 break;
899
900 case SO_RCVTIMEO:
901 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
902 break;
903
904 case SO_SNDTIMEO:
905 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
906 break;
1da177e4 907
e71a4783
SH		 908 case SO_ATTACH_FILTER:
909 ret = -EINVAL;
910 if (optlen == sizeof(struct sock_fprog)) {
911 struct sock_fprog fprog;
1da177e4 912
e71a4783
SH		 913 ret = -EFAULT;
914 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 915 break;
e71a4783
SH		 916
917 ret = sk_attach_filter(&fprog, sk);
918 }
919 break;
920
89aa0758
AS		 921 case SO_ATTACH_BPF:
922 ret = -EINVAL;
923 if (optlen == sizeof(u32)) {
924 u32 ufd;
925
926 ret = -EFAULT;
927 if (copy_from_user(&ufd, optval, sizeof(ufd)))
928 break;
929
930 ret = sk_attach_bpf(ufd, sk);
931 }
932 break;
933
e71a4783 934 case SO_DETACH_FILTER:
55b33325 935 ret = sk_detach_filter(sk);
e71a4783 936 break;
1da177e4 937
d59577b6
VB		 938 case SO_LOCK_FILTER:
939 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
940 ret = -EPERM;
941 else
942 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
943 break;
944
e71a4783
SH		 945 case SO_PASSSEC:
946 if (valbool)
947 set_bit(SOCK_PASSSEC, &sock->flags);
948 else
949 clear_bit(SOCK_PASSSEC, &sock->flags);
950 break;
4a19ec58 951 case SO_MARK:
5e1fccc0 952 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 953 ret = -EPERM;
2a91525c 954 else
4a19ec58 955 sk->sk_mark = val;
4a19ec58 956 break;
877ce7c1 957
3b885787 958 case SO_RXQ_OVFL:
8083f0fc 959 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 960 break;
6e3e939f
JB		 961
962 case SO_WIFI_STATUS:
963 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
964 break;
965
ef64a54f
PE		 966 case SO_PEEK_OFF:
967 if (sock->ops->set_peek_off)
12663bfc 968 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 969 else
970 ret = -EOPNOTSUPP;
971 break;
3bdc0eba
BG		 972
973 case SO_NOFCS:
974 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
975 break;
976
7d4c04fc
KJ		 977 case SO_SELECT_ERR_QUEUE:
978 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
979 break;
980
e0d1095a 981#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 982 case SO_BUSY_POLL:
dafcc438
ET		 983 /* allow unprivileged users to decrease the value */
984 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
985 ret = -EPERM;
986 else {
987 if (val < 0)
988 ret = -EINVAL;
989 else
990 sk->sk_ll_usec = val;
991 }
992 break;
993#endif
62748f32
ED		 994
995 case SO_MAX_PACING_RATE:
996 sk->sk_max_pacing_rate = val;
997 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
998 sk->sk_max_pacing_rate);
999 break;
1000
70da268b
ED		 1001 case SO_INCOMING_CPU:
1002 sk->sk_incoming_cpu = val;
1003 break;
1004
e71a4783
SH		 1005 default:
1006 ret = -ENOPROTOOPT;
1007 break;
4ec93edb 1008 }
1da177e4
LT		 1009 release_sock(sk);
1010 return ret;
1011}
2a91525c 1012EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1013
1014
8f09898b 1015static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1016 struct ucred *ucred)
3f551f94
EB		 1017{
1018 ucred->pid = pid_vnr(pid);
1019 ucred->uid = ucred->gid = -1;
1020 if (cred) {
1021 struct user_namespace *current_ns = current_user_ns();
1022
b2e4f544
EB		 1023 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1024 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1025 }
1026}
1027
1da177e4
LT		 1028int sock_getsockopt(struct socket *sock, int level, int optname,
1029 char __user *optval, int __user *optlen)
1030{
1031 struct sock *sk = sock->sk;
4ec93edb 1032
e71a4783 1033 union {
4ec93edb
YH		 1034 int val;
1035 struct linger ling;
1da177e4
LT		 1036 struct timeval tm;
1037 } v;
4ec93edb 1038
4d0392be 1039 int lv = sizeof(int);
1da177e4 1040 int len;
4ec93edb 1041
e71a4783 1042 if (get_user(len, optlen))
4ec93edb 1043 return -EFAULT;
e71a4783 1044 if (len < 0)
1da177e4 1045 return -EINVAL;
4ec93edb 1046
50fee1de 1047 memset(&v, 0, sizeof(v));
df0bca04 1048
2a91525c 1049 switch (optname) {
e71a4783
SH		 1050 case SO_DEBUG:
1051 v.val = sock_flag(sk, SOCK_DBG);
1052 break;
1053
1054 case SO_DONTROUTE:
1055 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1056 break;
1057
1058 case SO_BROADCAST:
1b23a5df 1059 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1060 break;
1061
1062 case SO_SNDBUF:
1063 v.val = sk->sk_sndbuf;
1064 break;
1065
1066 case SO_RCVBUF:
1067 v.val = sk->sk_rcvbuf;
1068 break;
1069
1070 case SO_REUSEADDR:
1071 v.val = sk->sk_reuse;
1072 break;
1073
055dc21a
TH		 1074 case SO_REUSEPORT:
1075 v.val = sk->sk_reuseport;
1076 break;
1077
e71a4783 1078 case SO_KEEPALIVE:
1b23a5df 1079 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1080 break;
1081
1082 case SO_TYPE:
1083 v.val = sk->sk_type;
1084 break;
1085
49c794e9
JE		 1086 case SO_PROTOCOL:
1087 v.val = sk->sk_protocol;
1088 break;
1089
0d6038ee
JE		 1090 case SO_DOMAIN:
1091 v.val = sk->sk_family;
1092 break;
1093
e71a4783
SH		 1094 case SO_ERROR:
1095 v.val = -sock_error(sk);
2a91525c 1096 if (v.val == 0)
e71a4783
SH		 1097 v.val = xchg(&sk->sk_err_soft, 0);
1098 break;
1099
1100 case SO_OOBINLINE:
1b23a5df 1101 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1102 break;
1103
1104 case SO_NO_CHECK:
28448b80 1105 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1106 break;
1107
1108 case SO_PRIORITY:
1109 v.val = sk->sk_priority;
1110 break;
1111
1112 case SO_LINGER:
1113 lv = sizeof(v.ling);
1b23a5df 1114 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1115 v.ling.l_linger = sk->sk_lingertime / HZ;
1116 break;
1117
1118 case SO_BSDCOMPAT:
1119 sock_warn_obsolete_bsdism("getsockopt");
1120 break;
1121
1122 case SO_TIMESTAMP:
92f37fd2
ED		 1123 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1124 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1125 break;
1126
1127 case SO_TIMESTAMPNS:
1128 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1129 break;
1130
20d49473 1131 case SO_TIMESTAMPING:
b9f40e21 1132 v.val = sk->sk_tsflags;
20d49473
PO		 1133 break;
1134
e71a4783 1135 case SO_RCVTIMEO:
2a91525c 1136 lv = sizeof(struct timeval);
e71a4783
SH		 1137 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1138 v.tm.tv_sec = 0;
1139 v.tm.tv_usec = 0;
1140 } else {
1141 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1142 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1143 }
1144 break;
1145
1146 case SO_SNDTIMEO:
2a91525c 1147 lv = sizeof(struct timeval);
e71a4783
SH		 1148 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1149 v.tm.tv_sec = 0;
1150 v.tm.tv_usec = 0;
1151 } else {
1152 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1153 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1154 }
1155 break;
1da177e4 1156
e71a4783
SH		 1157 case SO_RCVLOWAT:
1158 v.val = sk->sk_rcvlowat;
1159 break;
1da177e4 1160
e71a4783 1161 case SO_SNDLOWAT:
2a91525c 1162 v.val = 1;
e71a4783 1163 break;
1da177e4 1164
e71a4783 1165 case SO_PASSCRED:
82981930 1166 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1167 break;
1da177e4 1168
e71a4783 1169 case SO_PEERCRED:
109f6e39
EB		 1170 {
1171 struct ucred peercred;
1172 if (len > sizeof(peercred))
1173 len = sizeof(peercred);
1174 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1175 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1176 return -EFAULT;
1177 goto lenout;
109f6e39 1178 }
1da177e4 1179
e71a4783
SH		 1180 case SO_PEERNAME:
1181 {
1182 char address[128];
1183
1184 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1185 return -ENOTCONN;
1186 if (lv < len)
1187 return -EINVAL;
1188 if (copy_to_user(optval, address, len))
1189 return -EFAULT;
1190 goto lenout;
1191 }
1da177e4 1192
e71a4783
SH		 1193 /* Dubious BSD thing... Probably nobody even uses it, but
1194 * the UNIX standard wants it for whatever reason... -DaveM
1195 */
1196 case SO_ACCEPTCONN:
1197 v.val = sk->sk_state == TCP_LISTEN;
1198 break;
1da177e4 1199
e71a4783 1200 case SO_PASSSEC:
82981930 1201 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1202 break;
877ce7c1 1203
e71a4783
SH		 1204 case SO_PEERSEC:
1205 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1206
4a19ec58
LAT		 1207 case SO_MARK:
1208 v.val = sk->sk_mark;
1209 break;
1210
3b885787 1211 case SO_RXQ_OVFL:
1b23a5df 1212 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1213 break;
1214
6e3e939f 1215 case SO_WIFI_STATUS:
1b23a5df 1216 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1217 break;
1218
ef64a54f
PE		 1219 case SO_PEEK_OFF:
1220 if (!sock->ops->set_peek_off)
1221 return -EOPNOTSUPP;
1222
1223 v.val = sk->sk_peek_off;
1224 break;
bc2f7996 1225 case SO_NOFCS:
1b23a5df 1226 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1227 break;
c91f6df2 1228
f7b86bfe 1229 case SO_BINDTODEVICE:
c91f6df2
BH		 1230 return sock_getbindtodevice(sk, optval, optlen, len);
1231
a8fc9277
PE		 1232 case SO_GET_FILTER:
1233 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1234 if (len < 0)
1235 return len;
1236
1237 goto lenout;
c91f6df2 1238
d59577b6
VB		 1239 case SO_LOCK_FILTER:
1240 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1241 break;
1242
ea02f941
MS		 1243 case SO_BPF_EXTENSIONS:
1244 v.val = bpf_tell_extensions();
1245 break;
1246
7d4c04fc
KJ		 1247 case SO_SELECT_ERR_QUEUE:
1248 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1249 break;
1250
e0d1095a 1251#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1252 case SO_BUSY_POLL:
dafcc438
ET		 1253 v.val = sk->sk_ll_usec;
1254 break;
1255#endif
1256
62748f32
ED		 1257 case SO_MAX_PACING_RATE:
1258 v.val = sk->sk_max_pacing_rate;
1259 break;
1260
2c8c56e1
ED		 1261 case SO_INCOMING_CPU:
1262 v.val = sk->sk_incoming_cpu;
1263 break;
1264
e71a4783 1265 default:
443b5991
YH		 1266 /* We implement the SO_SNDLOWAT etc to not be settable
1267 * (1003.1g 7).
1268 */
e71a4783 1269 return -ENOPROTOOPT;
1da177e4 1270 }
e71a4783 1271
1da177e4
LT		 1272 if (len > lv)
1273 len = lv;
1274 if (copy_to_user(optval, &v, len))
1275 return -EFAULT;
1276lenout:
4ec93edb
YH		 1277 if (put_user(len, optlen))
1278 return -EFAULT;
1279 return 0;
1da177e4
LT		 1280}
1281
a5b5bb9a
IM		 1282/*
1283 * Initialize an sk_lock.
1284 *
1285 * (We also register the sk_lock with the lock validator.)
1286 */
b6f99a21 1287static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1288{
ed07536e
PZ		 1289 sock_lock_init_class_and_name(sk,
1290 af_family_slock_key_strings[sk->sk_family],
1291 af_family_slock_keys + sk->sk_family,
1292 af_family_key_strings[sk->sk_family],
1293 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1294}
1295
4dc6dc71
ED		 1296/*
1297 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1298 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1299 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1300 */
f1a6c4da
PE		 1301static void sock_copy(struct sock *nsk, const struct sock *osk)
1302{
1303#ifdef CONFIG_SECURITY_NETWORK
1304 void *sptr = nsk->sk_security;
1305#endif
68835aba
ED		 1306 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1307
1308 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1309 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1310
f1a6c4da
PE		 1311#ifdef CONFIG_SECURITY_NETWORK
1312 nsk->sk_security = sptr;
1313 security_sk_clone(osk, nsk);
1314#endif
1315}
1316
fcbdf09d
OP		 1317void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1318{
1319 unsigned long nulls1, nulls2;
1320
1321 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1322 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1323 if (nulls1 > nulls2)
1324 swap(nulls1, nulls2);
1325
1326 if (nulls1 != 0)
1327 memset((char *)sk, 0, nulls1);
1328 memset((char *)sk + nulls1 + sizeof(void *), 0,
1329 nulls2 - nulls1 - sizeof(void *));
1330 memset((char *)sk + nulls2 + sizeof(void *), 0,
1331 size - nulls2 - sizeof(void *));
1332}
1333EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1334
2e4afe7b
PE		 1335static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1336 int family)
c308c1b2
PE		 1337{
1338 struct sock *sk;
1339 struct kmem_cache *slab;
1340
1341 slab = prot->slab;
e912b114
ED		 1342 if (slab != NULL) {
1343 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1344 if (!sk)
1345 return sk;
1346 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1347 if (prot->clear_sk)
1348 prot->clear_sk(sk, prot->obj_size);
1349 else
1350 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1351 }
fcbdf09d 1352 } else
c308c1b2
PE		 1353 sk = kmalloc(prot->obj_size, priority);
1354
2e4afe7b 1355 if (sk != NULL) {
a98b65a3
VN		 1356 kmemcheck_annotate_bitfield(sk, flags);
1357
2e4afe7b
PE		 1358 if (security_sk_alloc(sk, family, priority))
1359 goto out_free;
1360
1361 if (!try_module_get(prot->owner))
1362 goto out_free_sec;
e022f0b4 1363 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1364 }
1365
c308c1b2 1366 return sk;
2e4afe7b
PE		 1367
1368out_free_sec:
1369 security_sk_free(sk);
1370out_free:
1371 if (slab != NULL)
1372 kmem_cache_free(slab, sk);
1373 else
1374 kfree(sk);
1375 return NULL;
c308c1b2
PE		 1376}
1377
1378static void sk_prot_free(struct proto *prot, struct sock *sk)
1379{
1380 struct kmem_cache *slab;
2e4afe7b 1381 struct module *owner;
c308c1b2 1382
2e4afe7b 1383 owner = prot->owner;
c308c1b2 1384 slab = prot->slab;
2e4afe7b
PE		 1385
1386 security_sk_free(sk);
c308c1b2
PE		 1387 if (slab != NULL)
1388 kmem_cache_free(slab, sk);
1389 else
1390 kfree(sk);
2e4afe7b 1391 module_put(owner);
c308c1b2
PE		 1392}
1393
86f8515f 1394#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1395void sock_update_netprioidx(struct sock *sk)
5bc1421e 1396{
5bc1421e
NH		 1397 if (in_interrupt())
1398 return;
2b73bc65 1399
6ffd4641 1400 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1401}
1402EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1403#endif
1404
1da177e4
LT		 1405/**
1406 * sk_alloc - All socket objects are allocated here
c4ea43c5 1407 * @net: the applicable net namespace
4dc3b16b
PP		 1408 * @family: protocol family
1409 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1410 * @prot: struct proto associated with this new sock instance
11aa9c28 1411 * @kern: is this to be a kernel socket?
1da177e4 1412 */
1b8d7ae4 1413struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1414 struct proto *prot, int kern)
1da177e4 1415{
c308c1b2 1416 struct sock *sk;
1da177e4 1417
154adbc8 1418 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1419 if (sk) {
154adbc8
PE		 1420 sk->sk_family = family;
1421 /*
1422 * See comment in struct sock definition to understand
1423 * why we need sk_prot_creator -acme
1424 */
1425 sk->sk_prot = sk->sk_prot_creator = prot;
1426 sock_lock_init(sk);
26abe143
EB		 1427 sk->sk_net_refcnt = kern ? 0 : 1;
1428 if (likely(sk->sk_net_refcnt))
1429 get_net(net);
1430 sock_net_set(sk, net);
d66ee058 1431 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1432
211d2f97 1433 sock_update_classid(sk);
6ffd4641 1434 sock_update_netprioidx(sk);
1da177e4 1435 }
a79af59e 1436
2e4afe7b 1437 return sk;
1da177e4 1438}
2a91525c 1439EXPORT_SYMBOL(sk_alloc);
1da177e4 1440
eb4cb008 1441void sk_destruct(struct sock *sk)
1da177e4
LT		 1442{
1443 struct sk_filter *filter;
1da177e4
LT		 1444
1445 if (sk->sk_destruct)
1446 sk->sk_destruct(sk);
1447
a898def2
PM		 1448 filter = rcu_dereference_check(sk->sk_filter,
1449 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1450 if (filter) {
309dd5fc 1451 sk_filter_uncharge(sk, filter);
a9b3cd7f 1452 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1453 }
1454
08e29af3 1455 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1456
1457 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1458 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1459 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1460
109f6e39
EB		 1461 if (sk->sk_peer_cred)
1462 put_cred(sk->sk_peer_cred);
1463 put_pid(sk->sk_peer_pid);
26abe143
EB		 1464 if (likely(sk->sk_net_refcnt))
1465 put_net(sock_net(sk));
c308c1b2 1466 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1467}
2b85a34e 1468
eb4cb008
CG		 1469static void __sk_free(struct sock *sk)
1470{
b922622e 1471 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1472 sock_diag_broadcast_destroy(sk);
1473 else
1474 sk_destruct(sk);
1475}
1476
2b85a34e
ED		 1477void sk_free(struct sock *sk)
1478{
1479 /*
25985edc 1480 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1481 * some packets are still in some tx queue.
1482 * If not null, sock_wfree() will call __sk_free(sk) later
1483 */
1484 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1485 __sk_free(sk);
1486}
2a91525c 1487EXPORT_SYMBOL(sk_free);
1da177e4 1488
475f1b52
SR		 1489static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1490{
1491 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1492 sock_update_memcg(newsk);
1493}
1494
e56c57d0
ED		 1495/**
1496 * sk_clone_lock - clone a socket, and lock its clone
1497 * @sk: the socket to clone
1498 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1499 *
1500 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1501 */
1502struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1503{
8fd1d178 1504 struct sock *newsk;
278571ba 1505 bool is_charged = true;
87d11ceb 1506
8fd1d178 1507 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1508 if (newsk != NULL) {
1509 struct sk_filter *filter;
1510
892c141e 1511 sock_copy(newsk, sk);
87d11ceb
ACM		 1512
1513 /* SANITY */
8a681736
SV		 1514 if (likely(newsk->sk_net_refcnt))
1515 get_net(sock_net(newsk));
87d11ceb
ACM		 1516 sk_node_init(&newsk->sk_node);
1517 sock_lock_init(newsk);
1518 bh_lock_sock(newsk);
fa438ccf 1519 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1520 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1521
1522 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1523 /*
1524 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1525 */
1526 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1527 atomic_set(&newsk->sk_omem_alloc, 0);
1528 skb_queue_head_init(&newsk->sk_receive_queue);
1529 skb_queue_head_init(&newsk->sk_write_queue);
1530
87d11ceb 1531 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1532 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1533 af_callback_keys + newsk->sk_family,
1534 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1535
1536 newsk->sk_dst_cache = NULL;
1537 newsk->sk_wmem_queued = 0;
1538 newsk->sk_forward_alloc = 0;
1539 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1540 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1541
1542 sock_reset_flag(newsk, SOCK_DONE);
1543 skb_queue_head_init(&newsk->sk_error_queue);
1544
0d7da9dd 1545 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1546 if (filter != NULL)
278571ba
AS		 1547 /* though it's an empty new sock, the charging may fail
1548 * if sysctl_optmem_max was changed between creation of
1549 * original socket and cloning
1550 */
1551 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1552
d188ba86 1553 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
87d11ceb
ACM		 1554 /* It is still raw copy of parent, so invalidate
1555 * destructor and make plain sk_free() */
1556 newsk->sk_destruct = NULL;
b0691c8e 1557 bh_unlock_sock(newsk);
87d11ceb
ACM		 1558 sk_free(newsk);
1559 newsk = NULL;
1560 goto out;
1561 }
1562
1563 newsk->sk_err = 0;
1564 newsk->sk_priority = 0;
2c8c56e1 1565 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1566 atomic64_set(&newsk->sk_cookie, 0);
4dc6dc71
ED		 1567 /*
1568 * Before updating sk_refcnt, we must commit prior changes to memory
1569 * (Documentation/RCU/rculist_nulls.txt for details)
1570 */
1571 smp_wmb();
87d11ceb
ACM		 1572 atomic_set(&newsk->sk_refcnt, 2);
1573
1574 /*
1575 * Increment the counter in the same struct proto as the master
1576 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1577 * is the same as sk->sk_prot->socks, as this field was copied
1578 * with memcpy).
1579 *
1580 * This _changes_ the previous behaviour, where
1581 * tcp_create_openreq_child always was incrementing the
1582 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1583 * to be taken into account in all callers. -acme
1584 */
1585 sk_refcnt_debug_inc(newsk);
972692e0 1586 sk_set_socket(newsk, NULL);
43815482 1587 newsk->sk_wq = NULL;
87d11ceb 1588
f3f511e1
GC		 1589 sk_update_clone(sk, newsk);
1590
87d11ceb 1591 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1592 sk_sockets_allocated_inc(newsk);
704da560 1593
080a270f
HFS		 1594 if (sock_needs_netstamp(sk) &&
1595 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1596 net_enable_timestamp();
87d11ceb
ACM		 1597 }
1598out:
1599 return newsk;
1600}
e56c57d0 1601EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1602
9958089a
AK		 1603void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1604{
d6a4e26a
ED		 1605 u32 max_segs = 1;
1606
6bd4f355 1607 sk_dst_set(sk, dst);
9958089a
AK		 1608 sk->sk_route_caps = dst->dev->features;
1609 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1610 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1611 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1612 if (sk_can_gso(sk)) {
82cc1a7a 1613 if (dst->header_len) {
9958089a 1614 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1615 } else {
9958089a 1616 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1617 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1618 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1619 }
9958089a 1620 }
d6a4e26a 1621 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1622}
1623EXPORT_SYMBOL_GPL(sk_setup_caps);
1624
1da177e4
LT		 1625/*
1626 * Simple resource managers for sockets.
1627 */
1628
1629
4ec93edb
YH		 1630/*
1631 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1632 */
1633void sock_wfree(struct sk_buff *skb)
1634{
1635 struct sock *sk = skb->sk;
d99927f4 1636 unsigned int len = skb->truesize;
1da177e4 1637
d99927f4
ED		 1638 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1639 /*
1640 * Keep a reference on sk_wmem_alloc, this will be released
1641 * after sk_write_space() call
1642 */
1643 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1644 sk->sk_write_space(sk);
d99927f4
ED		 1645 len = 1;
1646 }
2b85a34e 1647 /*
d99927f4
ED		 1648 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1649 * could not do because of in-flight packets
2b85a34e 1650 */
d99927f4 1651 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1652 __sk_free(sk);
1da177e4 1653}
2a91525c 1654EXPORT_SYMBOL(sock_wfree);
1da177e4 1655
9e17f8a4
ED		 1656void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1657{
1658 skb_orphan(skb);
1659 skb->sk = sk;
1660#ifdef CONFIG_INET
1661 if (unlikely(!sk_fullsock(sk))) {
1662 skb->destructor = sock_edemux;
1663 sock_hold(sk);
1664 return;
1665 }
1666#endif
1667 skb->destructor = sock_wfree;
1668 skb_set_hash_from_sk(skb, sk);
1669 /*
1670 * We used to take a refcount on sk, but following operation
1671 * is enough to guarantee sk_free() wont free this sock until
1672 * all in-flight packets are completed
1673 */
1674 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1675}
1676EXPORT_SYMBOL(skb_set_owner_w);
1677
f2f872f9
ED		 1678void skb_orphan_partial(struct sk_buff *skb)
1679{
1680 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1681 * so we do not completely orphan skb, but transfert all
1682 * accounted bytes but one, to avoid unexpected reorders.
1683 */
1684 if (skb->destructor == sock_wfree
1685#ifdef CONFIG_INET
1686 || skb->destructor == tcp_wfree
1687#endif
1688 ) {
1689 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1690 skb->truesize = 1;
1691 } else {
1692 skb_orphan(skb);
1693 }
1694}
1695EXPORT_SYMBOL(skb_orphan_partial);
1696
4ec93edb
YH		 1697/*
1698 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1699 */
1700void sock_rfree(struct sk_buff *skb)
1701{
1702 struct sock *sk = skb->sk;
d361fd59 1703 unsigned int len = skb->truesize;
1da177e4 1704
d361fd59
ED		 1705 atomic_sub(len, &sk->sk_rmem_alloc);
1706 sk_mem_uncharge(sk, len);
1da177e4 1707}
2a91525c 1708EXPORT_SYMBOL(sock_rfree);
1da177e4 1709
7768eed8
OH		 1710/*
1711 * Buffer destructor for skbs that are not used directly in read or write
1712 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1713 */
62bccb8c
AD		 1714void sock_efree(struct sk_buff *skb)
1715{
1716 sock_put(skb->sk);
1717}
1718EXPORT_SYMBOL(sock_efree);
1719
976d0201 1720kuid_t sock_i_uid(struct sock *sk)
1da177e4 1721{
976d0201 1722 kuid_t uid;
1da177e4 1723
f064af1e 1724 read_lock_bh(&sk->sk_callback_lock);
976d0201 1725 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1726 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1727 return uid;
1728}
2a91525c 1729EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1730
1731unsigned long sock_i_ino(struct sock *sk)
1732{
1733 unsigned long ino;
1734
f064af1e 1735 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1736 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1737 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1738 return ino;
1739}
2a91525c 1740EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1741
1742/*
1743 * Allocate a skb from the socket's send buffer.
1744 */
86a76caf 1745struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1746 gfp_t priority)
1da177e4
LT		 1747{
1748 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1749 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1750 if (skb) {
1751 skb_set_owner_w(skb, sk);
1752 return skb;
1753 }
1754 }
1755 return NULL;
1756}
2a91525c 1757EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1758
4ec93edb 1759/*
1da177e4 1760 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1761 */
dd0fc66f 1762void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1763{
95c96174 1764 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1765 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1766 void *mem;
1767 /* First do the add, to avoid the race if kmalloc
4ec93edb 1768 * might sleep.
1da177e4
LT		 1769 */
1770 atomic_add(size, &sk->sk_omem_alloc);
1771 mem = kmalloc(size, priority);
1772 if (mem)
1773 return mem;
1774 atomic_sub(size, &sk->sk_omem_alloc);
1775 }
1776 return NULL;
1777}
2a91525c 1778EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1779
79e88659
DB		 1780/* Free an option memory block. Note, we actually want the inline
1781 * here as this allows gcc to detect the nullify and fold away the
1782 * condition entirely.
1da177e4 1783 */
79e88659
DB		 1784static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1785 const bool nullify)
1da177e4 1786{
e53da5fb
DM		 1787 if (WARN_ON_ONCE(!mem))
1788 return;
79e88659
DB		 1789 if (nullify)
1790 kzfree(mem);
1791 else
1792 kfree(mem);
1da177e4
LT		 1793 atomic_sub(size, &sk->sk_omem_alloc);
1794}
79e88659
DB		 1795
1796void sock_kfree_s(struct sock *sk, void *mem, int size)
1797{
1798 __sock_kfree_s(sk, mem, size, false);
1799}
2a91525c 1800EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1801
79e88659
DB		 1802void sock_kzfree_s(struct sock *sk, void *mem, int size)
1803{
1804 __sock_kfree_s(sk, mem, size, true);
1805}
1806EXPORT_SYMBOL(sock_kzfree_s);
1807
1da177e4
LT		 1808/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1809 I think, these locks should be removed for datagram sockets.
1810 */
2a91525c 1811static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1812{
1813 DEFINE_WAIT(wait);
1814
9cd3e072 1815 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1da177e4
LT		 1816 for (;;) {
1817 if (!timeo)
1818 break;
1819 if (signal_pending(current))
1820 break;
1821 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1822 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1823 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1824 break;
1825 if (sk->sk_shutdown & SEND_SHUTDOWN)
1826 break;
1827 if (sk->sk_err)
1828 break;
1829 timeo = schedule_timeout(timeo);
1830 }
aa395145 1831 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1832 return timeo;
1833}
1834
1835
1836/*
1837 * Generic send/receive buffer handlers
1838 */
1839
4cc7f68d
HX		 1840struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1841 unsigned long data_len, int noblock,
28d64271 1842 int *errcode, int max_page_order)
1da177e4 1843{
2e4e4410 1844 struct sk_buff *skb;
1da177e4
LT		 1845 long timeo;
1846 int err;
1847
1da177e4 1848 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1849 for (;;) {
1da177e4
LT		 1850 err = sock_error(sk);
1851 if (err != 0)
1852 goto failure;
1853
1854 err = -EPIPE;
1855 if (sk->sk_shutdown & SEND_SHUTDOWN)
1856 goto failure;
1857
2e4e4410
ED		 1858 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1859 break;
28d64271 1860
9cd3e072 1861 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2e4e4410
ED		 1862 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1863 err = -EAGAIN;
1864 if (!timeo)
1da177e4 1865 goto failure;
2e4e4410
ED		 1866 if (signal_pending(current))
1867 goto interrupted;
1868 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1869 }
2e4e4410
ED		 1870 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1871 errcode, sk->sk_allocation);
1872 if (skb)
1873 skb_set_owner_w(skb, sk);
1da177e4
LT		 1874 return skb;
1875
1876interrupted:
1877 err = sock_intr_errno(timeo);
1878failure:
1879 *errcode = err;
1880 return NULL;
1881}
4cc7f68d 1882EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1883
4ec93edb 1884struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1885 int noblock, int *errcode)
1886{
28d64271 1887 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1888}
2a91525c 1889EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1890
f28ea365
EJ		 1891int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1892 struct sockcm_cookie *sockc)
1893{
1894 struct cmsghdr *cmsg;
1895
1896 for_each_cmsghdr(cmsg, msg) {
1897 if (!CMSG_OK(msg, cmsg))
1898 return -EINVAL;
1899 if (cmsg->cmsg_level != SOL_SOCKET)
1900 continue;
1901 switch (cmsg->cmsg_type) {
1902 case SO_MARK:
1903 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1904 return -EPERM;
1905 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1906 return -EINVAL;
1907 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
1908 break;
1909 default:
1910 return -EINVAL;
1911 }
1912 }
1913 return 0;
1914}
1915EXPORT_SYMBOL(sock_cmsg_send);
1916
5640f768
ED		 1917/* On 32bit arches, an skb frag is limited to 2^15 */
1918#define SKB_FRAG_PAGE_ORDER get_order(32768)
1919
400dfd3a
ED		 1920/**
1921 * skb_page_frag_refill - check that a page_frag contains enough room
1922 * @sz: minimum size of the fragment we want to get
1923 * @pfrag: pointer to page_frag
82d5e2b8 1924 * @gfp: priority for memory allocation
400dfd3a
ED		 1925 *
1926 * Note: While this allocator tries to use high order pages, there is
1927 * no guarantee that allocations succeed. Therefore, @sz MUST be
1928 * less or equal than PAGE_SIZE.
1929 */
d9b2938a 1930bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1931{
5640f768
ED		 1932 if (pfrag->page) {
1933 if (atomic_read(&pfrag->page->_count) == 1) {
1934 pfrag->offset = 0;
1935 return true;
1936 }
400dfd3a 1937 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1938 return true;
1939 put_page(pfrag->page);
1940 }
1941
d9b2938a
ED		 1942 pfrag->offset = 0;
1943 if (SKB_FRAG_PAGE_ORDER) {
d0164adc
MG		 1944 /* Avoid direct reclaim but allow kswapd to wake */
1945 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
1946 __GFP_COMP | __GFP_NOWARN |
1947 __GFP_NORETRY,
d9b2938a 1948 SKB_FRAG_PAGE_ORDER);
5640f768 1949 if (likely(pfrag->page)) {
d9b2938a 1950 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1951 return true;
1952 }
d9b2938a
ED		 1953 }
1954 pfrag->page = alloc_page(gfp);
1955 if (likely(pfrag->page)) {
1956 pfrag->size = PAGE_SIZE;
1957 return true;
1958 }
400dfd3a
ED		 1959 return false;
1960}
1961EXPORT_SYMBOL(skb_page_frag_refill);
1962
1963bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1964{
1965 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1966 return true;
1967
5640f768
ED		 1968 sk_enter_memory_pressure(sk);
1969 sk_stream_moderate_sndbuf(sk);
1970 return false;
1971}
1972EXPORT_SYMBOL(sk_page_frag_refill);
1973
1da177e4 1974static void __lock_sock(struct sock *sk)
f39234d6
NK		 1975 __releases(&sk->sk_lock.slock)
1976 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1977{
1978 DEFINE_WAIT(wait);
1979
e71a4783 1980 for (;;) {
1da177e4
LT		 1981 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1982 TASK_UNINTERRUPTIBLE);
1983 spin_unlock_bh(&sk->sk_lock.slock);
1984 schedule();
1985 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1986 if (!sock_owned_by_user(sk))
1da177e4
LT		 1987 break;
1988 }
1989 finish_wait(&sk->sk_lock.wq, &wait);
1990}
1991
1992static void __release_sock(struct sock *sk)
f39234d6
NK		 1993 __releases(&sk->sk_lock.slock)
1994 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1995{
1996 struct sk_buff *skb = sk->sk_backlog.head;
1997
1998 do {
1999 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
2000 bh_unlock_sock(sk);
2001
2002 do {
2003 struct sk_buff *next = skb->next;
2004
e4cbb02a 2005 prefetch(next);
7fee226a 2006 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 2007 skb->next = NULL;
c57943a1 2008 sk_backlog_rcv(sk, skb);
1da177e4
LT		 2009
2010 /*
2011 * We are in process context here with softirqs
2012 * disabled, use cond_resched_softirq() to preempt.
2013 * This is safe to do because we've taken the backlog
2014 * queue private:
2015 */
2016 cond_resched_softirq();
2017
2018 skb = next;
2019 } while (skb != NULL);
2020
2021 bh_lock_sock(sk);
e71a4783 2022 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 2023
2024 /*
2025 * Doing the zeroing here guarantee we can not loop forever
2026 * while a wild producer attempts to flood us.
2027 */
2028 sk->sk_backlog.len = 0;
1da177e4
LT		 2029}
2030
2031/**
2032 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 2033 * @sk: sock to wait on
2034 * @timeo: for how long
dfbafc99 2035 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 2036 *
2037 * Now socket state including sk->sk_err is changed only under lock,
2038 * hence we may omit checks after joining wait queue.
2039 * We check receive queue before schedule() only as optimization;
2040 * it is very likely that release_sock() added new data.
2041 */
dfbafc99 2042int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4
LT		 2043{
2044 int rc;
2045 DEFINE_WAIT(wait);
2046
aa395145 2047 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
9cd3e072 2048 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
dfbafc99 2049 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);
9cd3e072 2050 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
aa395145 2051 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2052 return rc;
2053}
1da177e4
LT		 2054EXPORT_SYMBOL(sk_wait_data);
2055
3ab224be
HA		 2056/**
2057 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2058 * @sk: socket
2059 * @size: memory size to allocate
2060 * @kind: allocation type
2061 *
2062 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2063 * rmem allocation. This function assumes that protocols which have
2064 * memory_pressure use sk_wmem_queued as write buffer accounting.
2065 */
2066int __sk_mem_schedule(struct sock *sk, int size, int kind)
2067{
2068 struct proto *prot = sk->sk_prot;
2069 int amt = sk_mem_pages(size);
8d987e5c 2070 long allocated;
e1aab161 2071 int parent_status = UNDER_LIMIT;
3ab224be
HA		 2072
2073 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2074
e1aab161 2075 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 2076
2077 /* Under limit. */
e1aab161
GC		 2078 if (parent_status == UNDER_LIMIT &&
2079 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2080 sk_leave_memory_pressure(sk);
3ab224be
HA		 2081 return 1;
2082 }
2083
e1aab161
GC		 2084 /* Under pressure. (we or our parents) */
2085 if ((parent_status > SOFT_LIMIT) ||
2086 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2087 sk_enter_memory_pressure(sk);
3ab224be 2088
e1aab161
GC		 2089 /* Over hard limit (we or our parents) */
2090 if ((parent_status == OVER_LIMIT) ||
2091 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 2092 goto suppress_allocation;
2093
2094 /* guarantee minimum buffer size under pressure */
2095 if (kind == SK_MEM_RECV) {
2096 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2097 return 1;
180d8cd9 2098
3ab224be
HA		 2099 } else { /* SK_MEM_SEND */
2100 if (sk->sk_type == SOCK_STREAM) {
2101 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2102 return 1;
2103 } else if (atomic_read(&sk->sk_wmem_alloc) <
2104 prot->sysctl_wmem[0])
2105 return 1;
2106 }
2107
180d8cd9 2108 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2109 int alloc;
2110
180d8cd9 2111 if (!sk_under_memory_pressure(sk))
1748376b 2112 return 1;
180d8cd9
GC		 2113 alloc = sk_sockets_allocated_read_positive(sk);
2114 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2115 sk_mem_pages(sk->sk_wmem_queued +
2116 atomic_read(&sk->sk_rmem_alloc) +
2117 sk->sk_forward_alloc))
2118 return 1;
2119 }
2120
2121suppress_allocation:
2122
2123 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2124 sk_stream_moderate_sndbuf(sk);
2125
2126 /* Fail only if socket is _under_ its sndbuf.
2127 * In this case we cannot block, so that we have to fail.
2128 */
2129 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2130 return 1;
2131 }
2132
3847ce32
SM		 2133 trace_sock_exceed_buf_limit(sk, prot, allocated);
2134
3ab224be
HA		 2135 /* Alas. Undo changes. */
2136 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2137
0e90b31f 2138 sk_memory_allocated_sub(sk, amt);
180d8cd9 2139
3ab224be
HA		 2140 return 0;
2141}
3ab224be
HA		 2142EXPORT_SYMBOL(__sk_mem_schedule);
2143
2144/**
69dba9bb 2145 * __sk_mem_reclaim - reclaim memory_allocated
3ab224be 2146 * @sk: socket
1a24e04e 2147 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
3ab224be 2148 */
1a24e04e 2149void __sk_mem_reclaim(struct sock *sk, int amount)
3ab224be 2150{
1a24e04e
ED		 2151 amount >>= SK_MEM_QUANTUM_SHIFT;
2152 sk_memory_allocated_sub(sk, amount);
2153 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
3ab224be 2154
180d8cd9
GC		 2155 if (sk_under_memory_pressure(sk) &&
2156 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2157 sk_leave_memory_pressure(sk);
3ab224be 2158}
3ab224be
HA		 2159EXPORT_SYMBOL(__sk_mem_reclaim);
2160
2161
1da177e4
LT		 2162/*
2163 * Set of default routines for initialising struct proto_ops when
2164 * the protocol does not support a particular function. In certain
2165 * cases where it makes no sense for a protocol to have a "do nothing"
2166 * function, some default processing is provided.
2167 */
2168
2169int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2170{
2171 return -EOPNOTSUPP;
2172}
2a91525c 2173EXPORT_SYMBOL(sock_no_bind);
1da177e4 2174
4ec93edb 2175int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2176 int len, int flags)
2177{
2178 return -EOPNOTSUPP;
2179}
2a91525c 2180EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2181
2182int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2183{
2184 return -EOPNOTSUPP;
2185}
2a91525c 2186EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2187
2188int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2189{
2190 return -EOPNOTSUPP;
2191}
2a91525c 2192EXPORT_SYMBOL(sock_no_accept);
1da177e4 2193
4ec93edb 2194int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2195 int *len, int peer)
2196{
2197 return -EOPNOTSUPP;
2198}
2a91525c 2199EXPORT_SYMBOL(sock_no_getname);
1da177e4 2200
2a91525c 2201unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2202{
2203 return 0;
2204}
2a91525c 2205EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2206
2207int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2208{
2209 return -EOPNOTSUPP;
2210}
2a91525c 2211EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2212
2213int sock_no_listen(struct socket *sock, int backlog)
2214{
2215 return -EOPNOTSUPP;
2216}
2a91525c 2217EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2218
2219int sock_no_shutdown(struct socket *sock, int how)
2220{
2221 return -EOPNOTSUPP;
2222}
2a91525c 2223EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2224
2225int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2226 char __user *optval, unsigned int optlen)
1da177e4
LT		 2227{
2228 return -EOPNOTSUPP;
2229}
2a91525c 2230EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2231
2232int sock_no_getsockopt(struct socket *sock, int level, int optname,
2233 char __user *optval, int __user *optlen)
2234{
2235 return -EOPNOTSUPP;
2236}
2a91525c 2237EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2238
1b784140 2239int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2240{
2241 return -EOPNOTSUPP;
2242}
2a91525c 2243EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2244
1b784140
YX		 2245int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2246 int flags)
1da177e4
LT		 2247{
2248 return -EOPNOTSUPP;
2249}
2a91525c 2250EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2251
2252int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2253{
2254 /* Mirror missing mmap method error code */
2255 return -ENODEV;
2256}
2a91525c 2257EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2258
2259ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2260{
2261 ssize_t res;
2262 struct msghdr msg = {.msg_flags = flags};
2263 struct kvec iov;
2264 char *kaddr = kmap(page);
2265 iov.iov_base = kaddr + offset;
2266 iov.iov_len = size;
2267 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2268 kunmap(page);
2269 return res;
2270}
2a91525c 2271EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2272
2273/*
2274 * Default Socket Callbacks
2275 */
2276
2277static void sock_def_wakeup(struct sock *sk)
2278{
43815482
ED		 2279 struct socket_wq *wq;
2280
2281 rcu_read_lock();
2282 wq = rcu_dereference(sk->sk_wq);
2283 if (wq_has_sleeper(wq))
2284 wake_up_interruptible_all(&wq->wait);
2285 rcu_read_unlock();
1da177e4
LT		 2286}
2287
2288static void sock_def_error_report(struct sock *sk)
2289{
43815482
ED		 2290 struct socket_wq *wq;
2291
2292 rcu_read_lock();
2293 wq = rcu_dereference(sk->sk_wq);
2294 if (wq_has_sleeper(wq))
2295 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2296 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2297 rcu_read_unlock();
1da177e4
LT		 2298}
2299
676d2369 2300static void sock_def_readable(struct sock *sk)
1da177e4 2301{
43815482
ED		 2302 struct socket_wq *wq;
2303
2304 rcu_read_lock();
2305 wq = rcu_dereference(sk->sk_wq);
2306 if (wq_has_sleeper(wq))
2c6607c6 2307 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2308 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2309 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2310 rcu_read_unlock();
1da177e4
LT		 2311}
2312
2313static void sock_def_write_space(struct sock *sk)
2314{
43815482
ED		 2315 struct socket_wq *wq;
2316
2317 rcu_read_lock();
1da177e4
LT		 2318
2319 /* Do not wake up a writer until he can make "significant"
2320 * progress. --DaveM
2321 */
e71a4783 2322 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2323 wq = rcu_dereference(sk->sk_wq);
2324 if (wq_has_sleeper(wq))
2325 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2326 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2327
2328 /* Should agree with poll, otherwise some programs break */
2329 if (sock_writeable(sk))
8d8ad9d7 2330 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2331 }
2332
43815482 2333 rcu_read_unlock();
1da177e4
LT		 2334}
2335
2336static void sock_def_destruct(struct sock *sk)
2337{
1da177e4
LT		 2338}
2339
2340void sk_send_sigurg(struct sock *sk)
2341{
2342 if (sk->sk_socket && sk->sk_socket->file)
2343 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2344 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2345}
2a91525c 2346EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2347
2348void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2349 unsigned long expires)
2350{
2351 if (!mod_timer(timer, expires))
2352 sock_hold(sk);
2353}
1da177e4
LT		 2354EXPORT_SYMBOL(sk_reset_timer);
2355
2356void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2357{
25cc4ae9 2358 if (del_timer(timer))
1da177e4
LT		 2359 __sock_put(sk);
2360}
1da177e4
LT		 2361EXPORT_SYMBOL(sk_stop_timer);
2362
2363void sock_init_data(struct socket *sock, struct sock *sk)
2364{
2365 skb_queue_head_init(&sk->sk_receive_queue);
2366 skb_queue_head_init(&sk->sk_write_queue);
2367 skb_queue_head_init(&sk->sk_error_queue);
2368
2369 sk->sk_send_head = NULL;
2370
2371 init_timer(&sk->sk_timer);
4ec93edb 2372
1da177e4
LT		 2373 sk->sk_allocation = GFP_KERNEL;
2374 sk->sk_rcvbuf = sysctl_rmem_default;
2375 sk->sk_sndbuf = sysctl_wmem_default;
2376 sk->sk_state = TCP_CLOSE;
972692e0 2377 sk_set_socket(sk, sock);
1da177e4
LT		 2378
2379 sock_set_flag(sk, SOCK_ZAPPED);
2380
e71a4783 2381 if (sock) {
1da177e4 2382 sk->sk_type = sock->type;
43815482 2383 sk->sk_wq = sock->wq;
1da177e4
LT		 2384 sock->sk = sk;
2385 } else
43815482 2386 sk->sk_wq = NULL;
1da177e4 2387
1da177e4 2388 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2389 lockdep_set_class_and_name(&sk->sk_callback_lock,
2390 af_callback_keys + sk->sk_family,
2391 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2392
2393 sk->sk_state_change = sock_def_wakeup;
2394 sk->sk_data_ready = sock_def_readable;
2395 sk->sk_write_space = sock_def_write_space;
2396 sk->sk_error_report = sock_def_error_report;
2397 sk->sk_destruct = sock_def_destruct;
2398
5640f768
ED		 2399 sk->sk_frag.page = NULL;
2400 sk->sk_frag.offset = 0;
ef64a54f 2401 sk->sk_peek_off = -1;
1da177e4 2402
109f6e39
EB		 2403 sk->sk_peer_pid = NULL;
2404 sk->sk_peer_cred = NULL;
1da177e4
LT		 2405 sk->sk_write_pending = 0;
2406 sk->sk_rcvlowat = 1;
2407 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2408 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2409
f37f0afb 2410 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2411
e0d1095a 2412#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2413 sk->sk_napi_id = 0;
64b0dc51 2414 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2415#endif
2416
62748f32 2417 sk->sk_max_pacing_rate = ~0U;
7eec4174 2418 sk->sk_pacing_rate = ~0U;
70da268b 2419 sk->sk_incoming_cpu = -1;
4dc6dc71
ED		 2420 /*
2421 * Before updating sk_refcnt, we must commit prior changes to memory
2422 * (Documentation/RCU/rculist_nulls.txt for details)
2423 */
2424 smp_wmb();
1da177e4 2425 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2426 atomic_set(&sk->sk_drops, 0);
1da177e4 2427}
2a91525c 2428EXPORT_SYMBOL(sock_init_data);
1da177e4 2429
b5606c2d 2430void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2431{
2432 might_sleep();
a5b5bb9a 2433 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2434 if (sk->sk_lock.owned)
1da177e4 2435 __lock_sock(sk);
d2e9117c 2436 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2437 spin_unlock(&sk->sk_lock.slock);
2438 /*
2439 * The sk_lock has mutex_lock() semantics here:
2440 */
fcc70d5f 2441 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2442 local_bh_enable();
1da177e4 2443}
fcc70d5f 2444EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2445
b5606c2d 2446void release_sock(struct sock *sk)
1da177e4 2447{
a5b5bb9a
IM		 2448 /*
2449 * The sk_lock has mutex_unlock() semantics:
2450 */
2451 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2452
2453 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2454 if (sk->sk_backlog.tail)
2455 __release_sock(sk);
46d3ceab 2456
c3f9b018
ED		 2457 /* Warning : release_cb() might need to release sk ownership,
2458 * ie call sock_release_ownership(sk) before us.
2459 */
46d3ceab
ED		 2460 if (sk->sk_prot->release_cb)
2461 sk->sk_prot->release_cb(sk);
2462
c3f9b018 2463 sock_release_ownership(sk);
a5b5bb9a
IM		 2464 if (waitqueue_active(&sk->sk_lock.wq))
2465 wake_up(&sk->sk_lock.wq);
2466 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2467}
2468EXPORT_SYMBOL(release_sock);
2469
8a74ad60
ED		 2470/**
2471 * lock_sock_fast - fast version of lock_sock
2472 * @sk: socket
2473 *
2474 * This version should be used for very small section, where process wont block
2475 * return false if fast path is taken
2476 * sk_lock.slock locked, owned = 0, BH disabled
2477 * return true if slow path is taken
2478 * sk_lock.slock unlocked, owned = 1, BH enabled
2479 */
2480bool lock_sock_fast(struct sock *sk)
2481{
2482 might_sleep();
2483 spin_lock_bh(&sk->sk_lock.slock);
2484
2485 if (!sk->sk_lock.owned)
2486 /*
2487 * Note : We must disable BH
2488 */
2489 return false;
2490
2491 __lock_sock(sk);
2492 sk->sk_lock.owned = 1;
2493 spin_unlock(&sk->sk_lock.slock);
2494 /*
2495 * The sk_lock has mutex_lock() semantics here:
2496 */
2497 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2498 local_bh_enable();
2499 return true;
2500}
2501EXPORT_SYMBOL(lock_sock_fast);
2502
1da177e4 2503int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2504{
b7aa0bf7 2505 struct timeval tv;
1da177e4 2506 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2507 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2508 tv = ktime_to_timeval(sk->sk_stamp);
2509 if (tv.tv_sec == -1)
1da177e4 2510 return -ENOENT;
b7aa0bf7
ED		 2511 if (tv.tv_sec == 0) {
2512 sk->sk_stamp = ktime_get_real();
2513 tv = ktime_to_timeval(sk->sk_stamp);
2514 }
2515 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2516}
1da177e4
LT		 2517EXPORT_SYMBOL(sock_get_timestamp);
2518
ae40eb1e
ED		 2519int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2520{
2521 struct timespec ts;
2522 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2523 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2524 ts = ktime_to_timespec(sk->sk_stamp);
2525 if (ts.tv_sec == -1)
2526 return -ENOENT;
2527 if (ts.tv_sec == 0) {
2528 sk->sk_stamp = ktime_get_real();
2529 ts = ktime_to_timespec(sk->sk_stamp);
2530 }
2531 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2532}
2533EXPORT_SYMBOL(sock_get_timestampns);
2534
20d49473 2535void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2536{
20d49473 2537 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2538 unsigned long previous_flags = sk->sk_flags;
2539
20d49473
PO		 2540 sock_set_flag(sk, flag);
2541 /*
2542 * we just set one of the two flags which require net
2543 * time stamping, but time stamping might have been on
2544 * already because of the other one
2545 */
080a270f
HFS		 2546 if (sock_needs_netstamp(sk) &&
2547 !(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2548 net_enable_timestamp();
1da177e4
LT		 2549 }
2550}
1da177e4 2551
cb820f8e
RC		 2552int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2553 int level, int type)
2554{
2555 struct sock_exterr_skb *serr;
364a9e93 2556 struct sk_buff *skb;
cb820f8e
RC		 2557 int copied, err;
2558
2559 err = -EAGAIN;
364a9e93 2560 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2561 if (skb == NULL)
2562 goto out;
2563
2564 copied = skb->len;
2565 if (copied > len) {
2566 msg->msg_flags |= MSG_TRUNC;
2567 copied = len;
2568 }
51f3d02b 2569 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2570 if (err)
2571 goto out_free_skb;
2572
2573 sock_recv_timestamp(msg, sk, skb);
2574
2575 serr = SKB_EXT_ERR(skb);
2576 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2577
2578 msg->msg_flags |= MSG_ERRQUEUE;
2579 err = copied;
2580
cb820f8e
RC		 2581out_free_skb:
2582 kfree_skb(skb);
2583out:
2584 return err;
2585}
2586EXPORT_SYMBOL(sock_recv_errqueue);
2587
1da177e4
LT		 2588/*
2589 * Get a socket option on an socket.
2590 *
2591 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2592 * asynchronous errors should be reported by getsockopt. We assume
2593 * this means if you specify SO_ERROR (otherwise whats the point of it).
2594 */
2595int sock_common_getsockopt(struct socket *sock, int level, int optname,
2596 char __user *optval, int __user *optlen)
2597{
2598 struct sock *sk = sock->sk;
2599
2600 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2601}
1da177e4
LT		 2602EXPORT_SYMBOL(sock_common_getsockopt);
2603
3fdadf7d 2604#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2605int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2606 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2607{
2608 struct sock *sk = sock->sk;
2609
1e51f951 2610 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2611 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2612 optval, optlen);
3fdadf7d
DM		 2613 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2614}
2615EXPORT_SYMBOL(compat_sock_common_getsockopt);
2616#endif
2617
1b784140
YX		 2618int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2619 int flags)
1da177e4
LT		 2620{
2621 struct sock *sk = sock->sk;
2622 int addr_len = 0;
2623 int err;
2624
1b784140 2625 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2626 flags & ~MSG_DONTWAIT, &addr_len);
2627 if (err >= 0)
2628 msg->msg_namelen = addr_len;
2629 return err;
2630}
1da177e4
LT		 2631EXPORT_SYMBOL(sock_common_recvmsg);
2632
2633/*
2634 * Set socket options on an inet socket.
2635 */
2636int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2637 char __user *optval, unsigned int optlen)
1da177e4
LT		 2638{
2639 struct sock *sk = sock->sk;
2640
2641 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2642}
1da177e4
LT		 2643EXPORT_SYMBOL(sock_common_setsockopt);
2644
3fdadf7d 2645#ifdef CONFIG_COMPAT
543d9cfe 2646int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2647 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2648{
2649 struct sock *sk = sock->sk;
2650
543d9cfe
ACM		 2651 if (sk->sk_prot->compat_setsockopt != NULL)
2652 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2653 optval, optlen);
3fdadf7d
DM		 2654 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2655}
2656EXPORT_SYMBOL(compat_sock_common_setsockopt);
2657#endif
2658
1da177e4
LT		 2659void sk_common_release(struct sock *sk)
2660{
2661 if (sk->sk_prot->destroy)
2662 sk->sk_prot->destroy(sk);
2663
2664 /*
2665 * Observation: when sock_common_release is called, processes have
2666 * no access to socket. But net still has.
2667 * Step one, detach it from networking:
2668 *
2669 * A. Remove from hash tables.
2670 */
2671
2672 sk->sk_prot->unhash(sk);
2673
2674 /*
2675 * In this point socket cannot receive new packets, but it is possible
2676 * that some packets are in flight because some CPU runs receiver and
2677 * did hash table lookup before we unhashed socket. They will achieve
2678 * receive queue and will be purged by socket destructor.
2679 *
2680 * Also we still have packets pending on receive queue and probably,
2681 * our own packets waiting in device queues. sock_destroy will drain
2682 * receive queue, but transmitted packets will delay socket destruction
2683 * until the last reference will be released.
2684 */
2685
2686 sock_orphan(sk);
2687
2688 xfrm_sk_free_policy(sk);
2689
e6848976 2690 sk_refcnt_debug_release(sk);
5640f768
ED		 2691
2692 if (sk->sk_frag.page) {
2693 put_page(sk->sk_frag.page);
2694 sk->sk_frag.page = NULL;
2695 }
2696
1da177e4
LT		 2697 sock_put(sk);
2698}
1da177e4
LT		 2699EXPORT_SYMBOL(sk_common_release);
2700
13ff3d6f
PE		 2701#ifdef CONFIG_PROC_FS
2702#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2703struct prot_inuse {
2704 int val[PROTO_INUSE_NR];
2705};
13ff3d6f
PE		 2706
2707static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2708
2709#ifdef CONFIG_NET_NS
2710void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2711{
d6d9ca0f 2712 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2713}
2714EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2715
2716int sock_prot_inuse_get(struct net *net, struct proto *prot)
2717{
2718 int cpu, idx = prot->inuse_idx;
2719 int res = 0;
2720
2721 for_each_possible_cpu(cpu)
2722 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2723
2724 return res >= 0 ? res : 0;
2725}
2726EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2727
2c8c1e72 2728static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2729{
2730 net->core.inuse = alloc_percpu(struct prot_inuse);
2731 return net->core.inuse ? 0 : -ENOMEM;
2732}
2733
2c8c1e72 2734static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2735{
2736 free_percpu(net->core.inuse);
2737}
2738
2739static struct pernet_operations net_inuse_ops = {
2740 .init = sock_inuse_init_net,
2741 .exit = sock_inuse_exit_net,
2742};
2743
2744static __init int net_inuse_init(void)
2745{
2746 if (register_pernet_subsys(&net_inuse_ops))
2747 panic("Cannot initialize net inuse counters");
2748
2749 return 0;
2750}
2751
2752core_initcall(net_inuse_init);
2753#else
1338d466
PE		 2754static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2755
c29a0bc4 2756void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2757{
d6d9ca0f 2758 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2759}
2760EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2761
c29a0bc4 2762int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2763{
2764 int cpu, idx = prot->inuse_idx;
2765 int res = 0;
2766
2767 for_each_possible_cpu(cpu)
2768 res += per_cpu(prot_inuse, cpu).val[idx];
2769
2770 return res >= 0 ? res : 0;
2771}
2772EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2773#endif
13ff3d6f
PE		 2774
2775static void assign_proto_idx(struct proto *prot)
2776{
2777 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2778
2779 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2780 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2781 return;
2782 }
2783
2784 set_bit(prot->inuse_idx, proto_inuse_idx);
2785}
2786
2787static void release_proto_idx(struct proto *prot)
2788{
2789 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2790 clear_bit(prot->inuse_idx, proto_inuse_idx);
2791}
2792#else
2793static inline void assign_proto_idx(struct proto *prot)
2794{
2795}
2796
2797static inline void release_proto_idx(struct proto *prot)
2798{
2799}
2800#endif
2801
0159dfd3
ED		 2802static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2803{
2804 if (!rsk_prot)
2805 return;
2806 kfree(rsk_prot->slab_name);
2807 rsk_prot->slab_name = NULL;
adf78eda
JL		 2808 kmem_cache_destroy(rsk_prot->slab);
2809 rsk_prot->slab = NULL;
0159dfd3
ED		 2810}
2811
2812static int req_prot_init(const struct proto *prot)
2813{
2814 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2815
2816 if (!rsk_prot)
2817 return 0;
2818
2819 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2820 prot->name);
2821 if (!rsk_prot->slab_name)
2822 return -ENOMEM;
2823
2824 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2825 rsk_prot->obj_size, 0,
e96f78ab 2826 prot->slab_flags, NULL);
0159dfd3
ED		 2827
2828 if (!rsk_prot->slab) {
2829 pr_crit("%s: Can't create request sock SLAB cache!\n",
2830 prot->name);
2831 return -ENOMEM;
2832 }
2833 return 0;
2834}
2835
b733c007
PE		 2836int proto_register(struct proto *prot, int alloc_slab)
2837{
1da177e4
LT		 2838 if (alloc_slab) {
2839 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2840 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2841 NULL);
1da177e4
LT		 2842
2843 if (prot->slab == NULL) {
e005d193
JP		 2844 pr_crit("%s: Can't create sock SLAB cache!\n",
2845 prot->name);
60e7663d 2846 goto out;
1da177e4 2847 }
2e6599cb 2848
0159dfd3
ED		 2849 if (req_prot_init(prot))
2850 goto out_free_request_sock_slab;
8feaf0c0 2851
6d6ee43e 2852 if (prot->twsk_prot != NULL) {
faf23422 2853 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2854
7e56b5d6 2855 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2856 goto out_free_request_sock_slab;
2857
6d6ee43e 2858 prot->twsk_prot->twsk_slab =
7e56b5d6 2859 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2860 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2861 0,
52db70dc 2862 prot->slab_flags,
20c2df83 2863 NULL);
6d6ee43e 2864 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2865 goto out_free_timewait_sock_slab_name;
2866 }
1da177e4
LT		 2867 }
2868
36b77a52 2869 mutex_lock(&proto_list_mutex);
1da177e4 2870 list_add(&prot->node, &proto_list);
13ff3d6f 2871 assign_proto_idx(prot);
36b77a52 2872 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2873 return 0;
2874
8feaf0c0 2875out_free_timewait_sock_slab_name:
7e56b5d6 2876 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2877out_free_request_sock_slab:
0159dfd3
ED		 2878 req_prot_cleanup(prot->rsk_prot);
2879
2e6599cb
ACM		 2880 kmem_cache_destroy(prot->slab);
2881 prot->slab = NULL;
b733c007
PE		 2882out:
2883 return -ENOBUFS;
1da177e4 2884}
1da177e4
LT		 2885EXPORT_SYMBOL(proto_register);
2886
2887void proto_unregister(struct proto *prot)
2888{
36b77a52 2889 mutex_lock(&proto_list_mutex);
13ff3d6f 2890 release_proto_idx(prot);
0a3f4358 2891 list_del(&prot->node);
36b77a52 2892 mutex_unlock(&proto_list_mutex);
1da177e4 2893
adf78eda
JL		 2894 kmem_cache_destroy(prot->slab);
2895 prot->slab = NULL;
1da177e4 2896
0159dfd3 2897 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2898
6d6ee43e 2899 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2900 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2901 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2902 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2903 }
1da177e4 2904}
1da177e4
LT		 2905EXPORT_SYMBOL(proto_unregister);
2906
2907#ifdef CONFIG_PROC_FS
1da177e4 2908static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2909 __acquires(proto_list_mutex)
1da177e4 2910{
36b77a52 2911 mutex_lock(&proto_list_mutex);
60f0438a 2912 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2913}
2914
2915static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2916{
60f0438a 2917 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2918}
2919
2920static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2921 __releases(proto_list_mutex)
1da177e4 2922{
36b77a52 2923 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2924}
2925
2926static char proto_method_implemented(const void *method)
2927{
2928 return method == NULL ? 'n' : 'y';
2929}
180d8cd9
GC		 2930static long sock_prot_memory_allocated(struct proto *proto)
2931{
cb75a36c 2932 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2933}
2934
2935static char *sock_prot_memory_pressure(struct proto *proto)
2936{
2937 return proto->memory_pressure != NULL ?
2938 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2939}
1da177e4
LT		 2940
2941static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2942{
180d8cd9 2943
8d987e5c 2944 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2945 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2946 proto->name,
2947 proto->obj_size,
14e943db 2948 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2949 sock_prot_memory_allocated(proto),
2950 sock_prot_memory_pressure(proto),
1da177e4
LT		 2951 proto->max_header,
2952 proto->slab == NULL ? "no" : "yes",
2953 module_name(proto->owner),
2954 proto_method_implemented(proto->close),
2955 proto_method_implemented(proto->connect),
2956 proto_method_implemented(proto->disconnect),
2957 proto_method_implemented(proto->accept),
2958 proto_method_implemented(proto->ioctl),
2959 proto_method_implemented(proto->init),
2960 proto_method_implemented(proto->destroy),
2961 proto_method_implemented(proto->shutdown),
2962 proto_method_implemented(proto->setsockopt),
2963 proto_method_implemented(proto->getsockopt),
2964 proto_method_implemented(proto->sendmsg),
2965 proto_method_implemented(proto->recvmsg),
2966 proto_method_implemented(proto->sendpage),
2967 proto_method_implemented(proto->bind),
2968 proto_method_implemented(proto->backlog_rcv),
2969 proto_method_implemented(proto->hash),
2970 proto_method_implemented(proto->unhash),
2971 proto_method_implemented(proto->get_port),
2972 proto_method_implemented(proto->enter_memory_pressure));
2973}
2974
2975static int proto_seq_show(struct seq_file *seq, void *v)
2976{
60f0438a 2977 if (v == &proto_list)
1da177e4
LT		 2978 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2979 "protocol",
2980 "size",
2981 "sockets",
2982 "memory",
2983 "press",
2984 "maxhdr",
2985 "slab",
2986 "module",
2987 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2988 else
60f0438a 2989 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2990 return 0;
2991}
2992
f690808e 2993static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2994 .start = proto_seq_start,
2995 .next = proto_seq_next,
2996 .stop = proto_seq_stop,
2997 .show = proto_seq_show,
2998};
2999
3000static int proto_seq_open(struct inode *inode, struct file *file)
3001{
14e943db
ED		 3002 return seq_open_net(inode, file, &proto_seq_ops,
3003 sizeof(struct seq_net_private));
1da177e4
LT		 3004}
3005
9a32144e 3006static const struct file_operations proto_seq_fops = {
1da177e4
LT		 3007 .owner = THIS_MODULE,
3008 .open = proto_seq_open,
3009 .read = seq_read,
3010 .llseek = seq_lseek,
14e943db
ED		 3011 .release = seq_release_net,
3012};
3013
3014static __net_init int proto_init_net(struct net *net)
3015{
d4beaa66 3016 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 3017 return -ENOMEM;
3018
3019 return 0;
3020}
3021
3022static __net_exit void proto_exit_net(struct net *net)
3023{
ece31ffd 3024 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 3025}
3026
3027
3028static __net_initdata struct pernet_operations proto_net_ops = {
3029 .init = proto_init_net,
3030 .exit = proto_exit_net,
1da177e4
LT		 3031};
3032
3033static int __init proto_init(void)
3034{
14e943db 3035 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 3036}
3037
3038subsys_initcall(proto_init);
3039
3040#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 1.130534 seconds and 5 git commands to generate.