projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
soreuseport: fast reuseport UDP socket selection
[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 874 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
ac5cc977
WC		 875 if (sk->sk_protocol == IPPROTO_TCP &&
876 sk->sk_type == SOCK_STREAM) {
4ed2d765
WB		 877 if (sk->sk_state != TCP_ESTABLISHED) {
878 ret = -EINVAL;
879 break;
880 }
881 sk->sk_tskey = tcp_sk(sk)->snd_una;
882 } else {
883 sk->sk_tskey = 0;
884 }
885 }
b9f40e21 886 sk->sk_tsflags = val;
20d49473
PO		 887 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
888 sock_enable_timestamp(sk,
889 SOCK_TIMESTAMPING_RX_SOFTWARE);
890 else
891 sock_disable_timestamp(sk,
08e29af3 892 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 893 break;
894
e71a4783
SH		 895 case SO_RCVLOWAT:
896 if (val < 0)
897 val = INT_MAX;
898 sk->sk_rcvlowat = val ? : 1;
899 break;
900
901 case SO_RCVTIMEO:
902 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
903 break;
904
905 case SO_SNDTIMEO:
906 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
907 break;
1da177e4 908
e71a4783
SH		 909 case SO_ATTACH_FILTER:
910 ret = -EINVAL;
911 if (optlen == sizeof(struct sock_fprog)) {
912 struct sock_fprog fprog;
1da177e4 913
e71a4783
SH		 914 ret = -EFAULT;
915 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 916 break;
e71a4783
SH		 917
918 ret = sk_attach_filter(&fprog, sk);
919 }
920 break;
921
89aa0758
AS		 922 case SO_ATTACH_BPF:
923 ret = -EINVAL;
924 if (optlen == sizeof(u32)) {
925 u32 ufd;
926
927 ret = -EFAULT;
928 if (copy_from_user(&ufd, optval, sizeof(ufd)))
929 break;
930
931 ret = sk_attach_bpf(ufd, sk);
932 }
933 break;
934
e71a4783 935 case SO_DETACH_FILTER:
55b33325 936 ret = sk_detach_filter(sk);
e71a4783 937 break;
1da177e4 938
d59577b6
VB		 939 case SO_LOCK_FILTER:
940 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
941 ret = -EPERM;
942 else
943 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
944 break;
945
e71a4783
SH		 946 case SO_PASSSEC:
947 if (valbool)
948 set_bit(SOCK_PASSSEC, &sock->flags);
949 else
950 clear_bit(SOCK_PASSSEC, &sock->flags);
951 break;
4a19ec58 952 case SO_MARK:
5e1fccc0 953 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 954 ret = -EPERM;
2a91525c 955 else
4a19ec58 956 sk->sk_mark = val;
4a19ec58 957 break;
877ce7c1 958
3b885787 959 case SO_RXQ_OVFL:
8083f0fc 960 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 961 break;
6e3e939f
JB		 962
963 case SO_WIFI_STATUS:
964 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
965 break;
966
ef64a54f
PE		 967 case SO_PEEK_OFF:
968 if (sock->ops->set_peek_off)
12663bfc 969 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 970 else
971 ret = -EOPNOTSUPP;
972 break;
3bdc0eba
BG		 973
974 case SO_NOFCS:
975 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
976 break;
977
7d4c04fc
KJ		 978 case SO_SELECT_ERR_QUEUE:
979 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
980 break;
981
e0d1095a 982#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 983 case SO_BUSY_POLL:
dafcc438
ET		 984 /* allow unprivileged users to decrease the value */
985 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
986 ret = -EPERM;
987 else {
988 if (val < 0)
989 ret = -EINVAL;
990 else
991 sk->sk_ll_usec = val;
992 }
993 break;
994#endif
62748f32
ED		 995
996 case SO_MAX_PACING_RATE:
997 sk->sk_max_pacing_rate = val;
998 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
999 sk->sk_max_pacing_rate);
1000 break;
1001
70da268b
ED		 1002 case SO_INCOMING_CPU:
1003 sk->sk_incoming_cpu = val;
1004 break;
1005
e71a4783
SH		 1006 default:
1007 ret = -ENOPROTOOPT;
1008 break;
4ec93edb 1009 }
1da177e4
LT		 1010 release_sock(sk);
1011 return ret;
1012}
2a91525c 1013EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1014
1015
8f09898b 1016static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1017 struct ucred *ucred)
3f551f94
EB		 1018{
1019 ucred->pid = pid_vnr(pid);
1020 ucred->uid = ucred->gid = -1;
1021 if (cred) {
1022 struct user_namespace *current_ns = current_user_ns();
1023
b2e4f544
EB		 1024 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1025 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1026 }
1027}
1028
1da177e4
LT		 1029int sock_getsockopt(struct socket *sock, int level, int optname,
1030 char __user *optval, int __user *optlen)
1031{
1032 struct sock *sk = sock->sk;
4ec93edb 1033
e71a4783 1034 union {
4ec93edb
YH		 1035 int val;
1036 struct linger ling;
1da177e4
LT		 1037 struct timeval tm;
1038 } v;
4ec93edb 1039
4d0392be 1040 int lv = sizeof(int);
1da177e4 1041 int len;
4ec93edb 1042
e71a4783 1043 if (get_user(len, optlen))
4ec93edb 1044 return -EFAULT;
e71a4783 1045 if (len < 0)
1da177e4 1046 return -EINVAL;
4ec93edb 1047
50fee1de 1048 memset(&v, 0, sizeof(v));
df0bca04 1049
2a91525c 1050 switch (optname) {
e71a4783
SH		 1051 case SO_DEBUG:
1052 v.val = sock_flag(sk, SOCK_DBG);
1053 break;
1054
1055 case SO_DONTROUTE:
1056 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1057 break;
1058
1059 case SO_BROADCAST:
1b23a5df 1060 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1061 break;
1062
1063 case SO_SNDBUF:
1064 v.val = sk->sk_sndbuf;
1065 break;
1066
1067 case SO_RCVBUF:
1068 v.val = sk->sk_rcvbuf;
1069 break;
1070
1071 case SO_REUSEADDR:
1072 v.val = sk->sk_reuse;
1073 break;
1074
055dc21a
TH		 1075 case SO_REUSEPORT:
1076 v.val = sk->sk_reuseport;
1077 break;
1078
e71a4783 1079 case SO_KEEPALIVE:
1b23a5df 1080 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1081 break;
1082
1083 case SO_TYPE:
1084 v.val = sk->sk_type;
1085 break;
1086
49c794e9
JE		 1087 case SO_PROTOCOL:
1088 v.val = sk->sk_protocol;
1089 break;
1090
0d6038ee
JE		 1091 case SO_DOMAIN:
1092 v.val = sk->sk_family;
1093 break;
1094
e71a4783
SH		 1095 case SO_ERROR:
1096 v.val = -sock_error(sk);
2a91525c 1097 if (v.val == 0)
e71a4783
SH		 1098 v.val = xchg(&sk->sk_err_soft, 0);
1099 break;
1100
1101 case SO_OOBINLINE:
1b23a5df 1102 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1103 break;
1104
1105 case SO_NO_CHECK:
28448b80 1106 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1107 break;
1108
1109 case SO_PRIORITY:
1110 v.val = sk->sk_priority;
1111 break;
1112
1113 case SO_LINGER:
1114 lv = sizeof(v.ling);
1b23a5df 1115 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1116 v.ling.l_linger = sk->sk_lingertime / HZ;
1117 break;
1118
1119 case SO_BSDCOMPAT:
1120 sock_warn_obsolete_bsdism("getsockopt");
1121 break;
1122
1123 case SO_TIMESTAMP:
92f37fd2
ED		 1124 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1125 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1126 break;
1127
1128 case SO_TIMESTAMPNS:
1129 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1130 break;
1131
20d49473 1132 case SO_TIMESTAMPING:
b9f40e21 1133 v.val = sk->sk_tsflags;
20d49473
PO		 1134 break;
1135
e71a4783 1136 case SO_RCVTIMEO:
2a91525c 1137 lv = sizeof(struct timeval);
e71a4783
SH		 1138 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1139 v.tm.tv_sec = 0;
1140 v.tm.tv_usec = 0;
1141 } else {
1142 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1143 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1144 }
1145 break;
1146
1147 case SO_SNDTIMEO:
2a91525c 1148 lv = sizeof(struct timeval);
e71a4783
SH		 1149 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1150 v.tm.tv_sec = 0;
1151 v.tm.tv_usec = 0;
1152 } else {
1153 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1154 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1155 }
1156 break;
1da177e4 1157
e71a4783
SH		 1158 case SO_RCVLOWAT:
1159 v.val = sk->sk_rcvlowat;
1160 break;
1da177e4 1161
e71a4783 1162 case SO_SNDLOWAT:
2a91525c 1163 v.val = 1;
e71a4783 1164 break;
1da177e4 1165
e71a4783 1166 case SO_PASSCRED:
82981930 1167 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1168 break;
1da177e4 1169
e71a4783 1170 case SO_PEERCRED:
109f6e39
EB		 1171 {
1172 struct ucred peercred;
1173 if (len > sizeof(peercred))
1174 len = sizeof(peercred);
1175 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1176 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1177 return -EFAULT;
1178 goto lenout;
109f6e39 1179 }
1da177e4 1180
e71a4783
SH		 1181 case SO_PEERNAME:
1182 {
1183 char address[128];
1184
1185 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1186 return -ENOTCONN;
1187 if (lv < len)
1188 return -EINVAL;
1189 if (copy_to_user(optval, address, len))
1190 return -EFAULT;
1191 goto lenout;
1192 }
1da177e4 1193
e71a4783
SH		 1194 /* Dubious BSD thing... Probably nobody even uses it, but
1195 * the UNIX standard wants it for whatever reason... -DaveM
1196 */
1197 case SO_ACCEPTCONN:
1198 v.val = sk->sk_state == TCP_LISTEN;
1199 break;
1da177e4 1200
e71a4783 1201 case SO_PASSSEC:
82981930 1202 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1203 break;
877ce7c1 1204
e71a4783
SH		 1205 case SO_PEERSEC:
1206 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1207
4a19ec58
LAT		 1208 case SO_MARK:
1209 v.val = sk->sk_mark;
1210 break;
1211
3b885787 1212 case SO_RXQ_OVFL:
1b23a5df 1213 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1214 break;
1215
6e3e939f 1216 case SO_WIFI_STATUS:
1b23a5df 1217 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1218 break;
1219
ef64a54f
PE		 1220 case SO_PEEK_OFF:
1221 if (!sock->ops->set_peek_off)
1222 return -EOPNOTSUPP;
1223
1224 v.val = sk->sk_peek_off;
1225 break;
bc2f7996 1226 case SO_NOFCS:
1b23a5df 1227 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1228 break;
c91f6df2 1229
f7b86bfe 1230 case SO_BINDTODEVICE:
c91f6df2
BH		 1231 return sock_getbindtodevice(sk, optval, optlen, len);
1232
a8fc9277
PE		 1233 case SO_GET_FILTER:
1234 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1235 if (len < 0)
1236 return len;
1237
1238 goto lenout;
c91f6df2 1239
d59577b6
VB		 1240 case SO_LOCK_FILTER:
1241 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1242 break;
1243
ea02f941
MS		 1244 case SO_BPF_EXTENSIONS:
1245 v.val = bpf_tell_extensions();
1246 break;
1247
7d4c04fc
KJ		 1248 case SO_SELECT_ERR_QUEUE:
1249 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1250 break;
1251
e0d1095a 1252#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1253 case SO_BUSY_POLL:
dafcc438
ET		 1254 v.val = sk->sk_ll_usec;
1255 break;
1256#endif
1257
62748f32
ED		 1258 case SO_MAX_PACING_RATE:
1259 v.val = sk->sk_max_pacing_rate;
1260 break;
1261
2c8c56e1
ED		 1262 case SO_INCOMING_CPU:
1263 v.val = sk->sk_incoming_cpu;
1264 break;
1265
e71a4783 1266 default:
443b5991
YH		 1267 /* We implement the SO_SNDLOWAT etc to not be settable
1268 * (1003.1g 7).
1269 */
e71a4783 1270 return -ENOPROTOOPT;
1da177e4 1271 }
e71a4783 1272
1da177e4
LT		 1273 if (len > lv)
1274 len = lv;
1275 if (copy_to_user(optval, &v, len))
1276 return -EFAULT;
1277lenout:
4ec93edb
YH		 1278 if (put_user(len, optlen))
1279 return -EFAULT;
1280 return 0;
1da177e4
LT		 1281}
1282
a5b5bb9a
IM		 1283/*
1284 * Initialize an sk_lock.
1285 *
1286 * (We also register the sk_lock with the lock validator.)
1287 */
b6f99a21 1288static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1289{
ed07536e
PZ		 1290 sock_lock_init_class_and_name(sk,
1291 af_family_slock_key_strings[sk->sk_family],
1292 af_family_slock_keys + sk->sk_family,
1293 af_family_key_strings[sk->sk_family],
1294 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1295}
1296
4dc6dc71
ED		 1297/*
1298 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1299 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1300 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1301 */
f1a6c4da
PE		 1302static void sock_copy(struct sock *nsk, const struct sock *osk)
1303{
1304#ifdef CONFIG_SECURITY_NETWORK
1305 void *sptr = nsk->sk_security;
1306#endif
68835aba
ED		 1307 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1308
1309 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1310 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1311
f1a6c4da
PE		 1312#ifdef CONFIG_SECURITY_NETWORK
1313 nsk->sk_security = sptr;
1314 security_sk_clone(osk, nsk);
1315#endif
1316}
1317
fcbdf09d
OP		 1318void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1319{
1320 unsigned long nulls1, nulls2;
1321
1322 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1323 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1324 if (nulls1 > nulls2)
1325 swap(nulls1, nulls2);
1326
1327 if (nulls1 != 0)
1328 memset((char *)sk, 0, nulls1);
1329 memset((char *)sk + nulls1 + sizeof(void *), 0,
1330 nulls2 - nulls1 - sizeof(void *));
1331 memset((char *)sk + nulls2 + sizeof(void *), 0,
1332 size - nulls2 - sizeof(void *));
1333}
1334EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1335
2e4afe7b
PE		 1336static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1337 int family)
c308c1b2
PE		 1338{
1339 struct sock *sk;
1340 struct kmem_cache *slab;
1341
1342 slab = prot->slab;
e912b114
ED		 1343 if (slab != NULL) {
1344 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1345 if (!sk)
1346 return sk;
1347 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1348 if (prot->clear_sk)
1349 prot->clear_sk(sk, prot->obj_size);
1350 else
1351 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1352 }
fcbdf09d 1353 } else
c308c1b2
PE		 1354 sk = kmalloc(prot->obj_size, priority);
1355
2e4afe7b 1356 if (sk != NULL) {
a98b65a3
VN		 1357 kmemcheck_annotate_bitfield(sk, flags);
1358
2e4afe7b
PE		 1359 if (security_sk_alloc(sk, family, priority))
1360 goto out_free;
1361
1362 if (!try_module_get(prot->owner))
1363 goto out_free_sec;
e022f0b4 1364 sk_tx_queue_clear(sk);
bd1060a1 1365 cgroup_sk_alloc(&sk->sk_cgrp_data);
2e4afe7b
PE		 1366 }
1367
c308c1b2 1368 return sk;
2e4afe7b
PE		 1369
1370out_free_sec:
1371 security_sk_free(sk);
1372out_free:
1373 if (slab != NULL)
1374 kmem_cache_free(slab, sk);
1375 else
1376 kfree(sk);
1377 return NULL;
c308c1b2
PE		 1378}
1379
1380static void sk_prot_free(struct proto *prot, struct sock *sk)
1381{
1382 struct kmem_cache *slab;
2e4afe7b 1383 struct module *owner;
c308c1b2 1384
2e4afe7b 1385 owner = prot->owner;
c308c1b2 1386 slab = prot->slab;
2e4afe7b 1387
bd1060a1 1388 cgroup_sk_free(&sk->sk_cgrp_data);
2e4afe7b 1389 security_sk_free(sk);
c308c1b2
PE		 1390 if (slab != NULL)
1391 kmem_cache_free(slab, sk);
1392 else
1393 kfree(sk);
2e4afe7b 1394 module_put(owner);
c308c1b2
PE		 1395}
1396
1da177e4
LT		 1397/**
1398 * sk_alloc - All socket objects are allocated here
c4ea43c5 1399 * @net: the applicable net namespace
4dc3b16b
PP		 1400 * @family: protocol family
1401 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1402 * @prot: struct proto associated with this new sock instance
11aa9c28 1403 * @kern: is this to be a kernel socket?
1da177e4 1404 */
1b8d7ae4 1405struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1406 struct proto *prot, int kern)
1da177e4 1407{
c308c1b2 1408 struct sock *sk;
1da177e4 1409
154adbc8 1410 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1411 if (sk) {
154adbc8
PE		 1412 sk->sk_family = family;
1413 /*
1414 * See comment in struct sock definition to understand
1415 * why we need sk_prot_creator -acme
1416 */
1417 sk->sk_prot = sk->sk_prot_creator = prot;
1418 sock_lock_init(sk);
26abe143
EB		 1419 sk->sk_net_refcnt = kern ? 0 : 1;
1420 if (likely(sk->sk_net_refcnt))
1421 get_net(net);
1422 sock_net_set(sk, net);
d66ee058 1423 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1424
2a56a1fe
TH		 1425 sock_update_classid(&sk->sk_cgrp_data);
1426 sock_update_netprioidx(&sk->sk_cgrp_data);
1da177e4 1427 }
a79af59e 1428
2e4afe7b 1429 return sk;
1da177e4 1430}
2a91525c 1431EXPORT_SYMBOL(sk_alloc);
1da177e4 1432
eb4cb008 1433void sk_destruct(struct sock *sk)
1da177e4
LT		 1434{
1435 struct sk_filter *filter;
1da177e4
LT		 1436
1437 if (sk->sk_destruct)
1438 sk->sk_destruct(sk);
1439
a898def2
PM		 1440 filter = rcu_dereference_check(sk->sk_filter,
1441 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1442 if (filter) {
309dd5fc 1443 sk_filter_uncharge(sk, filter);
a9b3cd7f 1444 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1445 }
1446
08e29af3 1447 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1448
1449 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1450 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1451 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1452
109f6e39
EB		 1453 if (sk->sk_peer_cred)
1454 put_cred(sk->sk_peer_cred);
1455 put_pid(sk->sk_peer_pid);
26abe143
EB		 1456 if (likely(sk->sk_net_refcnt))
1457 put_net(sock_net(sk));
c308c1b2 1458 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1459}
2b85a34e 1460
eb4cb008
CG		 1461static void __sk_free(struct sock *sk)
1462{
b922622e 1463 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1464 sock_diag_broadcast_destroy(sk);
1465 else
1466 sk_destruct(sk);
1467}
1468
2b85a34e
ED		 1469void sk_free(struct sock *sk)
1470{
1471 /*
25985edc 1472 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1473 * some packets are still in some tx queue.
1474 * If not null, sock_wfree() will call __sk_free(sk) later
1475 */
1476 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1477 __sk_free(sk);
1478}
2a91525c 1479EXPORT_SYMBOL(sk_free);
1da177e4 1480
475f1b52
SR		 1481static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1482{
1483 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1484 sock_update_memcg(newsk);
1485}
1486
e56c57d0
ED		 1487/**
1488 * sk_clone_lock - clone a socket, and lock its clone
1489 * @sk: the socket to clone
1490 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1491 *
1492 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1493 */
1494struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1495{
8fd1d178 1496 struct sock *newsk;
278571ba 1497 bool is_charged = true;
87d11ceb 1498
8fd1d178 1499 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1500 if (newsk != NULL) {
1501 struct sk_filter *filter;
1502
892c141e 1503 sock_copy(newsk, sk);
87d11ceb
ACM		 1504
1505 /* SANITY */
8a681736
SV		 1506 if (likely(newsk->sk_net_refcnt))
1507 get_net(sock_net(newsk));
87d11ceb
ACM		 1508 sk_node_init(&newsk->sk_node);
1509 sock_lock_init(newsk);
1510 bh_lock_sock(newsk);
fa438ccf 1511 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1512 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1513
1514 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1515 /*
1516 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1517 */
1518 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1519 atomic_set(&newsk->sk_omem_alloc, 0);
1520 skb_queue_head_init(&newsk->sk_receive_queue);
1521 skb_queue_head_init(&newsk->sk_write_queue);
1522
87d11ceb 1523 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1524 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1525 af_callback_keys + newsk->sk_family,
1526 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1527
1528 newsk->sk_dst_cache = NULL;
1529 newsk->sk_wmem_queued = 0;
1530 newsk->sk_forward_alloc = 0;
1531 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1532 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1533
1534 sock_reset_flag(newsk, SOCK_DONE);
1535 skb_queue_head_init(&newsk->sk_error_queue);
1536
0d7da9dd 1537 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1538 if (filter != NULL)
278571ba
AS		 1539 /* though it's an empty new sock, the charging may fail
1540 * if sysctl_optmem_max was changed between creation of
1541 * original socket and cloning
1542 */
1543 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1544
d188ba86 1545 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
87d11ceb
ACM		 1546 /* It is still raw copy of parent, so invalidate
1547 * destructor and make plain sk_free() */
1548 newsk->sk_destruct = NULL;
b0691c8e 1549 bh_unlock_sock(newsk);
87d11ceb
ACM		 1550 sk_free(newsk);
1551 newsk = NULL;
1552 goto out;
1553 }
1554
1555 newsk->sk_err = 0;
1556 newsk->sk_priority = 0;
2c8c56e1 1557 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1558 atomic64_set(&newsk->sk_cookie, 0);
4dc6dc71
ED		 1559 /*
1560 * Before updating sk_refcnt, we must commit prior changes to memory
1561 * (Documentation/RCU/rculist_nulls.txt for details)
1562 */
1563 smp_wmb();
87d11ceb
ACM		 1564 atomic_set(&newsk->sk_refcnt, 2);
1565
1566 /*
1567 * Increment the counter in the same struct proto as the master
1568 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1569 * is the same as sk->sk_prot->socks, as this field was copied
1570 * with memcpy).
1571 *
1572 * This _changes_ the previous behaviour, where
1573 * tcp_create_openreq_child always was incrementing the
1574 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1575 * to be taken into account in all callers. -acme
1576 */
1577 sk_refcnt_debug_inc(newsk);
972692e0 1578 sk_set_socket(newsk, NULL);
43815482 1579 newsk->sk_wq = NULL;
87d11ceb 1580
f3f511e1
GC		 1581 sk_update_clone(sk, newsk);
1582
87d11ceb 1583 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1584 sk_sockets_allocated_inc(newsk);
704da560 1585
080a270f
HFS		 1586 if (sock_needs_netstamp(sk) &&
1587 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1588 net_enable_timestamp();
87d11ceb
ACM		 1589 }
1590out:
1591 return newsk;
1592}
e56c57d0 1593EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1594
9958089a
AK		 1595void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1596{
d6a4e26a
ED		 1597 u32 max_segs = 1;
1598
6bd4f355 1599 sk_dst_set(sk, dst);
9958089a
AK		 1600 sk->sk_route_caps = dst->dev->features;
1601 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1602 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1603 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1604 if (sk_can_gso(sk)) {
82cc1a7a 1605 if (dst->header_len) {
9958089a 1606 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1607 } else {
9958089a 1608 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1609 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1610 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1611 }
9958089a 1612 }
d6a4e26a 1613 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1614}
1615EXPORT_SYMBOL_GPL(sk_setup_caps);
1616
1da177e4
LT		 1617/*
1618 * Simple resource managers for sockets.
1619 */
1620
1621
4ec93edb
YH		 1622/*
1623 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1624 */
1625void sock_wfree(struct sk_buff *skb)
1626{
1627 struct sock *sk = skb->sk;
d99927f4 1628 unsigned int len = skb->truesize;
1da177e4 1629
d99927f4
ED		 1630 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1631 /*
1632 * Keep a reference on sk_wmem_alloc, this will be released
1633 * after sk_write_space() call
1634 */
1635 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1636 sk->sk_write_space(sk);
d99927f4
ED		 1637 len = 1;
1638 }
2b85a34e 1639 /*
d99927f4
ED		 1640 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1641 * could not do because of in-flight packets
2b85a34e 1642 */
d99927f4 1643 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1644 __sk_free(sk);
1da177e4 1645}
2a91525c 1646EXPORT_SYMBOL(sock_wfree);
1da177e4 1647
9e17f8a4
ED		 1648void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1649{
1650 skb_orphan(skb);
1651 skb->sk = sk;
1652#ifdef CONFIG_INET
1653 if (unlikely(!sk_fullsock(sk))) {
1654 skb->destructor = sock_edemux;
1655 sock_hold(sk);
1656 return;
1657 }
1658#endif
1659 skb->destructor = sock_wfree;
1660 skb_set_hash_from_sk(skb, sk);
1661 /*
1662 * We used to take a refcount on sk, but following operation
1663 * is enough to guarantee sk_free() wont free this sock until
1664 * all in-flight packets are completed
1665 */
1666 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1667}
1668EXPORT_SYMBOL(skb_set_owner_w);
1669
f2f872f9
ED		 1670void skb_orphan_partial(struct sk_buff *skb)
1671{
1672 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1673 * so we do not completely orphan skb, but transfert all
1674 * accounted bytes but one, to avoid unexpected reorders.
1675 */
1676 if (skb->destructor == sock_wfree
1677#ifdef CONFIG_INET
1678 || skb->destructor == tcp_wfree
1679#endif
1680 ) {
1681 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1682 skb->truesize = 1;
1683 } else {
1684 skb_orphan(skb);
1685 }
1686}
1687EXPORT_SYMBOL(skb_orphan_partial);
1688
4ec93edb
YH		 1689/*
1690 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1691 */
1692void sock_rfree(struct sk_buff *skb)
1693{
1694 struct sock *sk = skb->sk;
d361fd59 1695 unsigned int len = skb->truesize;
1da177e4 1696
d361fd59
ED		 1697 atomic_sub(len, &sk->sk_rmem_alloc);
1698 sk_mem_uncharge(sk, len);
1da177e4 1699}
2a91525c 1700EXPORT_SYMBOL(sock_rfree);
1da177e4 1701
7768eed8
OH		 1702/*
1703 * Buffer destructor for skbs that are not used directly in read or write
1704 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1705 */
62bccb8c
AD		 1706void sock_efree(struct sk_buff *skb)
1707{
1708 sock_put(skb->sk);
1709}
1710EXPORT_SYMBOL(sock_efree);
1711
976d0201 1712kuid_t sock_i_uid(struct sock *sk)
1da177e4 1713{
976d0201 1714 kuid_t uid;
1da177e4 1715
f064af1e 1716 read_lock_bh(&sk->sk_callback_lock);
976d0201 1717 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1718 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1719 return uid;
1720}
2a91525c 1721EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1722
1723unsigned long sock_i_ino(struct sock *sk)
1724{
1725 unsigned long ino;
1726
f064af1e 1727 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1728 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1729 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1730 return ino;
1731}
2a91525c 1732EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1733
1734/*
1735 * Allocate a skb from the socket's send buffer.
1736 */
86a76caf 1737struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1738 gfp_t priority)
1da177e4
LT		 1739{
1740 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1741 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1742 if (skb) {
1743 skb_set_owner_w(skb, sk);
1744 return skb;
1745 }
1746 }
1747 return NULL;
1748}
2a91525c 1749EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1750
4ec93edb 1751/*
1da177e4 1752 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1753 */
dd0fc66f 1754void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1755{
95c96174 1756 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1757 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1758 void *mem;
1759 /* First do the add, to avoid the race if kmalloc
4ec93edb 1760 * might sleep.
1da177e4
LT		 1761 */
1762 atomic_add(size, &sk->sk_omem_alloc);
1763 mem = kmalloc(size, priority);
1764 if (mem)
1765 return mem;
1766 atomic_sub(size, &sk->sk_omem_alloc);
1767 }
1768 return NULL;
1769}
2a91525c 1770EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1771
79e88659
DB		 1772/* Free an option memory block. Note, we actually want the inline
1773 * here as this allows gcc to detect the nullify and fold away the
1774 * condition entirely.
1da177e4 1775 */
79e88659
DB		 1776static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1777 const bool nullify)
1da177e4 1778{
e53da5fb
DM		 1779 if (WARN_ON_ONCE(!mem))
1780 return;
79e88659
DB		 1781 if (nullify)
1782 kzfree(mem);
1783 else
1784 kfree(mem);
1da177e4
LT		 1785 atomic_sub(size, &sk->sk_omem_alloc);
1786}
79e88659
DB		 1787
1788void sock_kfree_s(struct sock *sk, void *mem, int size)
1789{
1790 __sock_kfree_s(sk, mem, size, false);
1791}
2a91525c 1792EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1793
79e88659
DB		 1794void sock_kzfree_s(struct sock *sk, void *mem, int size)
1795{
1796 __sock_kfree_s(sk, mem, size, true);
1797}
1798EXPORT_SYMBOL(sock_kzfree_s);
1799
1da177e4
LT		 1800/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1801 I think, these locks should be removed for datagram sockets.
1802 */
2a91525c 1803static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1804{
1805 DEFINE_WAIT(wait);
1806
9cd3e072 1807 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1da177e4
LT		 1808 for (;;) {
1809 if (!timeo)
1810 break;
1811 if (signal_pending(current))
1812 break;
1813 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1814 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1815 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1816 break;
1817 if (sk->sk_shutdown & SEND_SHUTDOWN)
1818 break;
1819 if (sk->sk_err)
1820 break;
1821 timeo = schedule_timeout(timeo);
1822 }
aa395145 1823 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1824 return timeo;
1825}
1826
1827
1828/*
1829 * Generic send/receive buffer handlers
1830 */
1831
4cc7f68d
HX		 1832struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1833 unsigned long data_len, int noblock,
28d64271 1834 int *errcode, int max_page_order)
1da177e4 1835{
2e4e4410 1836 struct sk_buff *skb;
1da177e4
LT		 1837 long timeo;
1838 int err;
1839
1da177e4 1840 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1841 for (;;) {
1da177e4
LT		 1842 err = sock_error(sk);
1843 if (err != 0)
1844 goto failure;
1845
1846 err = -EPIPE;
1847 if (sk->sk_shutdown & SEND_SHUTDOWN)
1848 goto failure;
1849
2e4e4410
ED		 1850 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1851 break;
28d64271 1852
9cd3e072 1853 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2e4e4410
ED		 1854 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1855 err = -EAGAIN;
1856 if (!timeo)
1da177e4 1857 goto failure;
2e4e4410
ED		 1858 if (signal_pending(current))
1859 goto interrupted;
1860 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1861 }
2e4e4410
ED		 1862 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1863 errcode, sk->sk_allocation);
1864 if (skb)
1865 skb_set_owner_w(skb, sk);
1da177e4
LT		 1866 return skb;
1867
1868interrupted:
1869 err = sock_intr_errno(timeo);
1870failure:
1871 *errcode = err;
1872 return NULL;
1873}
4cc7f68d 1874EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1875
4ec93edb 1876struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1877 int noblock, int *errcode)
1878{
28d64271 1879 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1880}
2a91525c 1881EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1882
f28ea365
EJ		 1883int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1884 struct sockcm_cookie *sockc)
1885{
1886 struct cmsghdr *cmsg;
1887
1888 for_each_cmsghdr(cmsg, msg) {
1889 if (!CMSG_OK(msg, cmsg))
1890 return -EINVAL;
1891 if (cmsg->cmsg_level != SOL_SOCKET)
1892 continue;
1893 switch (cmsg->cmsg_type) {
1894 case SO_MARK:
1895 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1896 return -EPERM;
1897 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1898 return -EINVAL;
1899 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
1900 break;
1901 default:
1902 return -EINVAL;
1903 }
1904 }
1905 return 0;
1906}
1907EXPORT_SYMBOL(sock_cmsg_send);
1908
5640f768
ED		 1909/* On 32bit arches, an skb frag is limited to 2^15 */
1910#define SKB_FRAG_PAGE_ORDER get_order(32768)
1911
400dfd3a
ED		 1912/**
1913 * skb_page_frag_refill - check that a page_frag contains enough room
1914 * @sz: minimum size of the fragment we want to get
1915 * @pfrag: pointer to page_frag
82d5e2b8 1916 * @gfp: priority for memory allocation
400dfd3a
ED		 1917 *
1918 * Note: While this allocator tries to use high order pages, there is
1919 * no guarantee that allocations succeed. Therefore, @sz MUST be
1920 * less or equal than PAGE_SIZE.
1921 */
d9b2938a 1922bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1923{
5640f768
ED		 1924 if (pfrag->page) {
1925 if (atomic_read(&pfrag->page->_count) == 1) {
1926 pfrag->offset = 0;
1927 return true;
1928 }
400dfd3a 1929 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1930 return true;
1931 put_page(pfrag->page);
1932 }
1933
d9b2938a
ED		 1934 pfrag->offset = 0;
1935 if (SKB_FRAG_PAGE_ORDER) {
d0164adc
MG		 1936 /* Avoid direct reclaim but allow kswapd to wake */
1937 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
1938 __GFP_COMP | __GFP_NOWARN |
1939 __GFP_NORETRY,
d9b2938a 1940 SKB_FRAG_PAGE_ORDER);
5640f768 1941 if (likely(pfrag->page)) {
d9b2938a 1942 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1943 return true;
1944 }
d9b2938a
ED		 1945 }
1946 pfrag->page = alloc_page(gfp);
1947 if (likely(pfrag->page)) {
1948 pfrag->size = PAGE_SIZE;
1949 return true;
1950 }
400dfd3a
ED		 1951 return false;
1952}
1953EXPORT_SYMBOL(skb_page_frag_refill);
1954
1955bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1956{
1957 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1958 return true;
1959
5640f768
ED		 1960 sk_enter_memory_pressure(sk);
1961 sk_stream_moderate_sndbuf(sk);
1962 return false;
1963}
1964EXPORT_SYMBOL(sk_page_frag_refill);
1965
1da177e4 1966static void __lock_sock(struct sock *sk)
f39234d6
NK		 1967 __releases(&sk->sk_lock.slock)
1968 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1969{
1970 DEFINE_WAIT(wait);
1971
e71a4783 1972 for (;;) {
1da177e4
LT		 1973 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1974 TASK_UNINTERRUPTIBLE);
1975 spin_unlock_bh(&sk->sk_lock.slock);
1976 schedule();
1977 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1978 if (!sock_owned_by_user(sk))
1da177e4
LT		 1979 break;
1980 }
1981 finish_wait(&sk->sk_lock.wq, &wait);
1982}
1983
1984static void __release_sock(struct sock *sk)
f39234d6
NK		 1985 __releases(&sk->sk_lock.slock)
1986 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1987{
1988 struct sk_buff *skb = sk->sk_backlog.head;
1989
1990 do {
1991 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1992 bh_unlock_sock(sk);
1993
1994 do {
1995 struct sk_buff *next = skb->next;
1996
e4cbb02a 1997 prefetch(next);
7fee226a 1998 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1999 skb->next = NULL;
c57943a1 2000 sk_backlog_rcv(sk, skb);
1da177e4
LT		 2001
2002 /*
2003 * We are in process context here with softirqs
2004 * disabled, use cond_resched_softirq() to preempt.
2005 * This is safe to do because we've taken the backlog
2006 * queue private:
2007 */
2008 cond_resched_softirq();
2009
2010 skb = next;
2011 } while (skb != NULL);
2012
2013 bh_lock_sock(sk);
e71a4783 2014 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 2015
2016 /*
2017 * Doing the zeroing here guarantee we can not loop forever
2018 * while a wild producer attempts to flood us.
2019 */
2020 sk->sk_backlog.len = 0;
1da177e4
LT		 2021}
2022
2023/**
2024 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 2025 * @sk: sock to wait on
2026 * @timeo: for how long
dfbafc99 2027 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 2028 *
2029 * Now socket state including sk->sk_err is changed only under lock,
2030 * hence we may omit checks after joining wait queue.
2031 * We check receive queue before schedule() only as optimization;
2032 * it is very likely that release_sock() added new data.
2033 */
dfbafc99 2034int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4
LT		 2035{
2036 int rc;
2037 DEFINE_WAIT(wait);
2038
aa395145 2039 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
9cd3e072 2040 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
dfbafc99 2041 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);
9cd3e072 2042 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
aa395145 2043 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2044 return rc;
2045}
1da177e4
LT		 2046EXPORT_SYMBOL(sk_wait_data);
2047
3ab224be
HA		 2048/**
2049 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2050 * @sk: socket
2051 * @size: memory size to allocate
2052 * @kind: allocation type
2053 *
2054 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2055 * rmem allocation. This function assumes that protocols which have
2056 * memory_pressure use sk_wmem_queued as write buffer accounting.
2057 */
2058int __sk_mem_schedule(struct sock *sk, int size, int kind)
2059{
2060 struct proto *prot = sk->sk_prot;
2061 int amt = sk_mem_pages(size);
8d987e5c 2062 long allocated;
e1aab161 2063 int parent_status = UNDER_LIMIT;
3ab224be
HA		 2064
2065 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2066
e1aab161 2067 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 2068
2069 /* Under limit. */
e1aab161
GC		 2070 if (parent_status == UNDER_LIMIT &&
2071 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2072 sk_leave_memory_pressure(sk);
3ab224be
HA		 2073 return 1;
2074 }
2075
e1aab161
GC		 2076 /* Under pressure. (we or our parents) */
2077 if ((parent_status > SOFT_LIMIT) ||
2078 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2079 sk_enter_memory_pressure(sk);
3ab224be 2080
e1aab161
GC		 2081 /* Over hard limit (we or our parents) */
2082 if ((parent_status == OVER_LIMIT) ||
2083 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 2084 goto suppress_allocation;
2085
2086 /* guarantee minimum buffer size under pressure */
2087 if (kind == SK_MEM_RECV) {
2088 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2089 return 1;
180d8cd9 2090
3ab224be
HA		 2091 } else { /* SK_MEM_SEND */
2092 if (sk->sk_type == SOCK_STREAM) {
2093 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2094 return 1;
2095 } else if (atomic_read(&sk->sk_wmem_alloc) <
2096 prot->sysctl_wmem[0])
2097 return 1;
2098 }
2099
180d8cd9 2100 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2101 int alloc;
2102
180d8cd9 2103 if (!sk_under_memory_pressure(sk))
1748376b 2104 return 1;
180d8cd9
GC		 2105 alloc = sk_sockets_allocated_read_positive(sk);
2106 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2107 sk_mem_pages(sk->sk_wmem_queued +
2108 atomic_read(&sk->sk_rmem_alloc) +
2109 sk->sk_forward_alloc))
2110 return 1;
2111 }
2112
2113suppress_allocation:
2114
2115 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2116 sk_stream_moderate_sndbuf(sk);
2117
2118 /* Fail only if socket is _under_ its sndbuf.
2119 * In this case we cannot block, so that we have to fail.
2120 */
2121 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2122 return 1;
2123 }
2124
3847ce32
SM		 2125 trace_sock_exceed_buf_limit(sk, prot, allocated);
2126
3ab224be
HA		 2127 /* Alas. Undo changes. */
2128 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2129
0e90b31f 2130 sk_memory_allocated_sub(sk, amt);
180d8cd9 2131
3ab224be
HA		 2132 return 0;
2133}
3ab224be
HA		 2134EXPORT_SYMBOL(__sk_mem_schedule);
2135
2136/**
69dba9bb 2137 * __sk_mem_reclaim - reclaim memory_allocated
3ab224be 2138 * @sk: socket
1a24e04e 2139 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
3ab224be 2140 */
1a24e04e 2141void __sk_mem_reclaim(struct sock *sk, int amount)
3ab224be 2142{
1a24e04e
ED		 2143 amount >>= SK_MEM_QUANTUM_SHIFT;
2144 sk_memory_allocated_sub(sk, amount);
2145 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
3ab224be 2146
180d8cd9
GC		 2147 if (sk_under_memory_pressure(sk) &&
2148 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2149 sk_leave_memory_pressure(sk);
3ab224be 2150}
3ab224be
HA		 2151EXPORT_SYMBOL(__sk_mem_reclaim);
2152
2153
1da177e4
LT		 2154/*
2155 * Set of default routines for initialising struct proto_ops when
2156 * the protocol does not support a particular function. In certain
2157 * cases where it makes no sense for a protocol to have a "do nothing"
2158 * function, some default processing is provided.
2159 */
2160
2161int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2162{
2163 return -EOPNOTSUPP;
2164}
2a91525c 2165EXPORT_SYMBOL(sock_no_bind);
1da177e4 2166
4ec93edb 2167int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2168 int len, int flags)
2169{
2170 return -EOPNOTSUPP;
2171}
2a91525c 2172EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2173
2174int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2175{
2176 return -EOPNOTSUPP;
2177}
2a91525c 2178EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2179
2180int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2181{
2182 return -EOPNOTSUPP;
2183}
2a91525c 2184EXPORT_SYMBOL(sock_no_accept);
1da177e4 2185
4ec93edb 2186int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2187 int *len, int peer)
2188{
2189 return -EOPNOTSUPP;
2190}
2a91525c 2191EXPORT_SYMBOL(sock_no_getname);
1da177e4 2192
2a91525c 2193unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2194{
2195 return 0;
2196}
2a91525c 2197EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2198
2199int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2200{
2201 return -EOPNOTSUPP;
2202}
2a91525c 2203EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2204
2205int sock_no_listen(struct socket *sock, int backlog)
2206{
2207 return -EOPNOTSUPP;
2208}
2a91525c 2209EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2210
2211int sock_no_shutdown(struct socket *sock, int how)
2212{
2213 return -EOPNOTSUPP;
2214}
2a91525c 2215EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2216
2217int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2218 char __user *optval, unsigned int optlen)
1da177e4
LT		 2219{
2220 return -EOPNOTSUPP;
2221}
2a91525c 2222EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2223
2224int sock_no_getsockopt(struct socket *sock, int level, int optname,
2225 char __user *optval, int __user *optlen)
2226{
2227 return -EOPNOTSUPP;
2228}
2a91525c 2229EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2230
1b784140 2231int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2232{
2233 return -EOPNOTSUPP;
2234}
2a91525c 2235EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2236
1b784140
YX		 2237int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2238 int flags)
1da177e4
LT		 2239{
2240 return -EOPNOTSUPP;
2241}
2a91525c 2242EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2243
2244int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2245{
2246 /* Mirror missing mmap method error code */
2247 return -ENODEV;
2248}
2a91525c 2249EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2250
2251ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2252{
2253 ssize_t res;
2254 struct msghdr msg = {.msg_flags = flags};
2255 struct kvec iov;
2256 char *kaddr = kmap(page);
2257 iov.iov_base = kaddr + offset;
2258 iov.iov_len = size;
2259 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2260 kunmap(page);
2261 return res;
2262}
2a91525c 2263EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2264
2265/*
2266 * Default Socket Callbacks
2267 */
2268
2269static void sock_def_wakeup(struct sock *sk)
2270{
43815482
ED		 2271 struct socket_wq *wq;
2272
2273 rcu_read_lock();
2274 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2275 if (skwq_has_sleeper(wq))
43815482
ED		 2276 wake_up_interruptible_all(&wq->wait);
2277 rcu_read_unlock();
1da177e4
LT		 2278}
2279
2280static void sock_def_error_report(struct sock *sk)
2281{
43815482
ED		 2282 struct socket_wq *wq;
2283
2284 rcu_read_lock();
2285 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2286 if (skwq_has_sleeper(wq))
43815482 2287 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2288 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2289 rcu_read_unlock();
1da177e4
LT		 2290}
2291
676d2369 2292static void sock_def_readable(struct sock *sk)
1da177e4 2293{
43815482
ED		 2294 struct socket_wq *wq;
2295
2296 rcu_read_lock();
2297 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2298 if (skwq_has_sleeper(wq))
2c6607c6 2299 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2300 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2301 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2302 rcu_read_unlock();
1da177e4
LT		 2303}
2304
2305static void sock_def_write_space(struct sock *sk)
2306{
43815482
ED		 2307 struct socket_wq *wq;
2308
2309 rcu_read_lock();
1da177e4
LT		 2310
2311 /* Do not wake up a writer until he can make "significant"
2312 * progress. --DaveM
2313 */
e71a4783 2314 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482 2315 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2316 if (skwq_has_sleeper(wq))
43815482 2317 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2318 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2319
2320 /* Should agree with poll, otherwise some programs break */
2321 if (sock_writeable(sk))
8d8ad9d7 2322 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2323 }
2324
43815482 2325 rcu_read_unlock();
1da177e4
LT		 2326}
2327
2328static void sock_def_destruct(struct sock *sk)
2329{
1da177e4
LT		 2330}
2331
2332void sk_send_sigurg(struct sock *sk)
2333{
2334 if (sk->sk_socket && sk->sk_socket->file)
2335 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2336 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2337}
2a91525c 2338EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2339
2340void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2341 unsigned long expires)
2342{
2343 if (!mod_timer(timer, expires))
2344 sock_hold(sk);
2345}
1da177e4
LT		 2346EXPORT_SYMBOL(sk_reset_timer);
2347
2348void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2349{
25cc4ae9 2350 if (del_timer(timer))
1da177e4
LT		 2351 __sock_put(sk);
2352}
1da177e4
LT		 2353EXPORT_SYMBOL(sk_stop_timer);
2354
2355void sock_init_data(struct socket *sock, struct sock *sk)
2356{
2357 skb_queue_head_init(&sk->sk_receive_queue);
2358 skb_queue_head_init(&sk->sk_write_queue);
2359 skb_queue_head_init(&sk->sk_error_queue);
2360
2361 sk->sk_send_head = NULL;
2362
2363 init_timer(&sk->sk_timer);
4ec93edb 2364
1da177e4
LT		 2365 sk->sk_allocation = GFP_KERNEL;
2366 sk->sk_rcvbuf = sysctl_rmem_default;
2367 sk->sk_sndbuf = sysctl_wmem_default;
2368 sk->sk_state = TCP_CLOSE;
972692e0 2369 sk_set_socket(sk, sock);
1da177e4
LT		 2370
2371 sock_set_flag(sk, SOCK_ZAPPED);
2372
e71a4783 2373 if (sock) {
1da177e4 2374 sk->sk_type = sock->type;
43815482 2375 sk->sk_wq = sock->wq;
1da177e4
LT		 2376 sock->sk = sk;
2377 } else
43815482 2378 sk->sk_wq = NULL;
1da177e4 2379
1da177e4 2380 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2381 lockdep_set_class_and_name(&sk->sk_callback_lock,
2382 af_callback_keys + sk->sk_family,
2383 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2384
2385 sk->sk_state_change = sock_def_wakeup;
2386 sk->sk_data_ready = sock_def_readable;
2387 sk->sk_write_space = sock_def_write_space;
2388 sk->sk_error_report = sock_def_error_report;
2389 sk->sk_destruct = sock_def_destruct;
2390
5640f768
ED		 2391 sk->sk_frag.page = NULL;
2392 sk->sk_frag.offset = 0;
ef64a54f 2393 sk->sk_peek_off = -1;
1da177e4 2394
109f6e39
EB		 2395 sk->sk_peer_pid = NULL;
2396 sk->sk_peer_cred = NULL;
1da177e4
LT		 2397 sk->sk_write_pending = 0;
2398 sk->sk_rcvlowat = 1;
2399 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2400 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2401
f37f0afb 2402 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2403
e0d1095a 2404#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2405 sk->sk_napi_id = 0;
64b0dc51 2406 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2407#endif
2408
62748f32 2409 sk->sk_max_pacing_rate = ~0U;
7eec4174 2410 sk->sk_pacing_rate = ~0U;
70da268b 2411 sk->sk_incoming_cpu = -1;
4dc6dc71
ED		 2412 /*
2413 * Before updating sk_refcnt, we must commit prior changes to memory
2414 * (Documentation/RCU/rculist_nulls.txt for details)
2415 */
2416 smp_wmb();
1da177e4 2417 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2418 atomic_set(&sk->sk_drops, 0);
1da177e4 2419}
2a91525c 2420EXPORT_SYMBOL(sock_init_data);
1da177e4 2421
b5606c2d 2422void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2423{
2424 might_sleep();
a5b5bb9a 2425 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2426 if (sk->sk_lock.owned)
1da177e4 2427 __lock_sock(sk);
d2e9117c 2428 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2429 spin_unlock(&sk->sk_lock.slock);
2430 /*
2431 * The sk_lock has mutex_lock() semantics here:
2432 */
fcc70d5f 2433 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2434 local_bh_enable();
1da177e4 2435}
fcc70d5f 2436EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2437
b5606c2d 2438void release_sock(struct sock *sk)
1da177e4 2439{
a5b5bb9a
IM		 2440 /*
2441 * The sk_lock has mutex_unlock() semantics:
2442 */
2443 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2444
2445 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2446 if (sk->sk_backlog.tail)
2447 __release_sock(sk);
46d3ceab 2448
c3f9b018
ED		 2449 /* Warning : release_cb() might need to release sk ownership,
2450 * ie call sock_release_ownership(sk) before us.
2451 */
46d3ceab
ED		 2452 if (sk->sk_prot->release_cb)
2453 sk->sk_prot->release_cb(sk);
2454
c3f9b018 2455 sock_release_ownership(sk);
a5b5bb9a
IM		 2456 if (waitqueue_active(&sk->sk_lock.wq))
2457 wake_up(&sk->sk_lock.wq);
2458 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2459}
2460EXPORT_SYMBOL(release_sock);
2461
8a74ad60
ED		 2462/**
2463 * lock_sock_fast - fast version of lock_sock
2464 * @sk: socket
2465 *
2466 * This version should be used for very small section, where process wont block
2467 * return false if fast path is taken
2468 * sk_lock.slock locked, owned = 0, BH disabled
2469 * return true if slow path is taken
2470 * sk_lock.slock unlocked, owned = 1, BH enabled
2471 */
2472bool lock_sock_fast(struct sock *sk)
2473{
2474 might_sleep();
2475 spin_lock_bh(&sk->sk_lock.slock);
2476
2477 if (!sk->sk_lock.owned)
2478 /*
2479 * Note : We must disable BH
2480 */
2481 return false;
2482
2483 __lock_sock(sk);
2484 sk->sk_lock.owned = 1;
2485 spin_unlock(&sk->sk_lock.slock);
2486 /*
2487 * The sk_lock has mutex_lock() semantics here:
2488 */
2489 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2490 local_bh_enable();
2491 return true;
2492}
2493EXPORT_SYMBOL(lock_sock_fast);
2494
1da177e4 2495int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2496{
b7aa0bf7 2497 struct timeval tv;
1da177e4 2498 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2499 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2500 tv = ktime_to_timeval(sk->sk_stamp);
2501 if (tv.tv_sec == -1)
1da177e4 2502 return -ENOENT;
b7aa0bf7
ED		 2503 if (tv.tv_sec == 0) {
2504 sk->sk_stamp = ktime_get_real();
2505 tv = ktime_to_timeval(sk->sk_stamp);
2506 }
2507 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2508}
1da177e4
LT		 2509EXPORT_SYMBOL(sock_get_timestamp);
2510
ae40eb1e
ED		 2511int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2512{
2513 struct timespec ts;
2514 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2515 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2516 ts = ktime_to_timespec(sk->sk_stamp);
2517 if (ts.tv_sec == -1)
2518 return -ENOENT;
2519 if (ts.tv_sec == 0) {
2520 sk->sk_stamp = ktime_get_real();
2521 ts = ktime_to_timespec(sk->sk_stamp);
2522 }
2523 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2524}
2525EXPORT_SYMBOL(sock_get_timestampns);
2526
20d49473 2527void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2528{
20d49473 2529 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2530 unsigned long previous_flags = sk->sk_flags;
2531
20d49473
PO		 2532 sock_set_flag(sk, flag);
2533 /*
2534 * we just set one of the two flags which require net
2535 * time stamping, but time stamping might have been on
2536 * already because of the other one
2537 */
080a270f
HFS		 2538 if (sock_needs_netstamp(sk) &&
2539 !(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2540 net_enable_timestamp();
1da177e4
LT		 2541 }
2542}
1da177e4 2543
cb820f8e
RC		 2544int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2545 int level, int type)
2546{
2547 struct sock_exterr_skb *serr;
364a9e93 2548 struct sk_buff *skb;
cb820f8e
RC		 2549 int copied, err;
2550
2551 err = -EAGAIN;
364a9e93 2552 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2553 if (skb == NULL)
2554 goto out;
2555
2556 copied = skb->len;
2557 if (copied > len) {
2558 msg->msg_flags |= MSG_TRUNC;
2559 copied = len;
2560 }
51f3d02b 2561 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2562 if (err)
2563 goto out_free_skb;
2564
2565 sock_recv_timestamp(msg, sk, skb);
2566
2567 serr = SKB_EXT_ERR(skb);
2568 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2569
2570 msg->msg_flags |= MSG_ERRQUEUE;
2571 err = copied;
2572
cb820f8e
RC		 2573out_free_skb:
2574 kfree_skb(skb);
2575out:
2576 return err;
2577}
2578EXPORT_SYMBOL(sock_recv_errqueue);
2579
1da177e4
LT		 2580/*
2581 * Get a socket option on an socket.
2582 *
2583 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2584 * asynchronous errors should be reported by getsockopt. We assume
2585 * this means if you specify SO_ERROR (otherwise whats the point of it).
2586 */
2587int sock_common_getsockopt(struct socket *sock, int level, int optname,
2588 char __user *optval, int __user *optlen)
2589{
2590 struct sock *sk = sock->sk;
2591
2592 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2593}
1da177e4
LT		 2594EXPORT_SYMBOL(sock_common_getsockopt);
2595
3fdadf7d 2596#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2597int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2598 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2599{
2600 struct sock *sk = sock->sk;
2601
1e51f951 2602 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2603 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2604 optval, optlen);
3fdadf7d
DM		 2605 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2606}
2607EXPORT_SYMBOL(compat_sock_common_getsockopt);
2608#endif
2609
1b784140
YX		 2610int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2611 int flags)
1da177e4
LT		 2612{
2613 struct sock *sk = sock->sk;
2614 int addr_len = 0;
2615 int err;
2616
1b784140 2617 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2618 flags & ~MSG_DONTWAIT, &addr_len);
2619 if (err >= 0)
2620 msg->msg_namelen = addr_len;
2621 return err;
2622}
1da177e4
LT		 2623EXPORT_SYMBOL(sock_common_recvmsg);
2624
2625/*
2626 * Set socket options on an inet socket.
2627 */
2628int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2629 char __user *optval, unsigned int optlen)
1da177e4
LT		 2630{
2631 struct sock *sk = sock->sk;
2632
2633 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2634}
1da177e4
LT		 2635EXPORT_SYMBOL(sock_common_setsockopt);
2636
3fdadf7d 2637#ifdef CONFIG_COMPAT
543d9cfe 2638int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2639 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2640{
2641 struct sock *sk = sock->sk;
2642
543d9cfe
ACM		 2643 if (sk->sk_prot->compat_setsockopt != NULL)
2644 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2645 optval, optlen);
3fdadf7d
DM		 2646 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2647}
2648EXPORT_SYMBOL(compat_sock_common_setsockopt);
2649#endif
2650
1da177e4
LT		 2651void sk_common_release(struct sock *sk)
2652{
2653 if (sk->sk_prot->destroy)
2654 sk->sk_prot->destroy(sk);
2655
2656 /*
2657 * Observation: when sock_common_release is called, processes have
2658 * no access to socket. But net still has.
2659 * Step one, detach it from networking:
2660 *
2661 * A. Remove from hash tables.
2662 */
2663
2664 sk->sk_prot->unhash(sk);
2665
2666 /*
2667 * In this point socket cannot receive new packets, but it is possible
2668 * that some packets are in flight because some CPU runs receiver and
2669 * did hash table lookup before we unhashed socket. They will achieve
2670 * receive queue and will be purged by socket destructor.
2671 *
2672 * Also we still have packets pending on receive queue and probably,
2673 * our own packets waiting in device queues. sock_destroy will drain
2674 * receive queue, but transmitted packets will delay socket destruction
2675 * until the last reference will be released.
2676 */
2677
2678 sock_orphan(sk);
2679
2680 xfrm_sk_free_policy(sk);
2681
e6848976 2682 sk_refcnt_debug_release(sk);
5640f768
ED		 2683
2684 if (sk->sk_frag.page) {
2685 put_page(sk->sk_frag.page);
2686 sk->sk_frag.page = NULL;
2687 }
2688
1da177e4
LT		 2689 sock_put(sk);
2690}
1da177e4
LT		 2691EXPORT_SYMBOL(sk_common_release);
2692
13ff3d6f
PE		 2693#ifdef CONFIG_PROC_FS
2694#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2695struct prot_inuse {
2696 int val[PROTO_INUSE_NR];
2697};
13ff3d6f
PE		 2698
2699static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2700
2701#ifdef CONFIG_NET_NS
2702void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2703{
d6d9ca0f 2704 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2705}
2706EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2707
2708int sock_prot_inuse_get(struct net *net, struct proto *prot)
2709{
2710 int cpu, idx = prot->inuse_idx;
2711 int res = 0;
2712
2713 for_each_possible_cpu(cpu)
2714 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2715
2716 return res >= 0 ? res : 0;
2717}
2718EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2719
2c8c1e72 2720static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2721{
2722 net->core.inuse = alloc_percpu(struct prot_inuse);
2723 return net->core.inuse ? 0 : -ENOMEM;
2724}
2725
2c8c1e72 2726static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2727{
2728 free_percpu(net->core.inuse);
2729}
2730
2731static struct pernet_operations net_inuse_ops = {
2732 .init = sock_inuse_init_net,
2733 .exit = sock_inuse_exit_net,
2734};
2735
2736static __init int net_inuse_init(void)
2737{
2738 if (register_pernet_subsys(&net_inuse_ops))
2739 panic("Cannot initialize net inuse counters");
2740
2741 return 0;
2742}
2743
2744core_initcall(net_inuse_init);
2745#else
1338d466
PE		 2746static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2747
c29a0bc4 2748void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2749{
d6d9ca0f 2750 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2751}
2752EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2753
c29a0bc4 2754int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2755{
2756 int cpu, idx = prot->inuse_idx;
2757 int res = 0;
2758
2759 for_each_possible_cpu(cpu)
2760 res += per_cpu(prot_inuse, cpu).val[idx];
2761
2762 return res >= 0 ? res : 0;
2763}
2764EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2765#endif
13ff3d6f
PE		 2766
2767static void assign_proto_idx(struct proto *prot)
2768{
2769 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2770
2771 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2772 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2773 return;
2774 }
2775
2776 set_bit(prot->inuse_idx, proto_inuse_idx);
2777}
2778
2779static void release_proto_idx(struct proto *prot)
2780{
2781 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2782 clear_bit(prot->inuse_idx, proto_inuse_idx);
2783}
2784#else
2785static inline void assign_proto_idx(struct proto *prot)
2786{
2787}
2788
2789static inline void release_proto_idx(struct proto *prot)
2790{
2791}
2792#endif
2793
0159dfd3
ED		 2794static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2795{
2796 if (!rsk_prot)
2797 return;
2798 kfree(rsk_prot->slab_name);
2799 rsk_prot->slab_name = NULL;
adf78eda
JL		 2800 kmem_cache_destroy(rsk_prot->slab);
2801 rsk_prot->slab = NULL;
0159dfd3
ED		 2802}
2803
2804static int req_prot_init(const struct proto *prot)
2805{
2806 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2807
2808 if (!rsk_prot)
2809 return 0;
2810
2811 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2812 prot->name);
2813 if (!rsk_prot->slab_name)
2814 return -ENOMEM;
2815
2816 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2817 rsk_prot->obj_size, 0,
e96f78ab 2818 prot->slab_flags, NULL);
0159dfd3
ED		 2819
2820 if (!rsk_prot->slab) {
2821 pr_crit("%s: Can't create request sock SLAB cache!\n",
2822 prot->name);
2823 return -ENOMEM;
2824 }
2825 return 0;
2826}
2827
b733c007
PE		 2828int proto_register(struct proto *prot, int alloc_slab)
2829{
1da177e4
LT		 2830 if (alloc_slab) {
2831 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2832 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2833 NULL);
1da177e4
LT		 2834
2835 if (prot->slab == NULL) {
e005d193
JP		 2836 pr_crit("%s: Can't create sock SLAB cache!\n",
2837 prot->name);
60e7663d 2838 goto out;
1da177e4 2839 }
2e6599cb 2840
0159dfd3
ED		 2841 if (req_prot_init(prot))
2842 goto out_free_request_sock_slab;
8feaf0c0 2843
6d6ee43e 2844 if (prot->twsk_prot != NULL) {
faf23422 2845 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2846
7e56b5d6 2847 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2848 goto out_free_request_sock_slab;
2849
6d6ee43e 2850 prot->twsk_prot->twsk_slab =
7e56b5d6 2851 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2852 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2853 0,
52db70dc 2854 prot->slab_flags,
20c2df83 2855 NULL);
6d6ee43e 2856 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2857 goto out_free_timewait_sock_slab_name;
2858 }
1da177e4
LT		 2859 }
2860
36b77a52 2861 mutex_lock(&proto_list_mutex);
1da177e4 2862 list_add(&prot->node, &proto_list);
13ff3d6f 2863 assign_proto_idx(prot);
36b77a52 2864 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2865 return 0;
2866
8feaf0c0 2867out_free_timewait_sock_slab_name:
7e56b5d6 2868 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2869out_free_request_sock_slab:
0159dfd3
ED		 2870 req_prot_cleanup(prot->rsk_prot);
2871
2e6599cb
ACM		 2872 kmem_cache_destroy(prot->slab);
2873 prot->slab = NULL;
b733c007
PE		 2874out:
2875 return -ENOBUFS;
1da177e4 2876}
1da177e4
LT		 2877EXPORT_SYMBOL(proto_register);
2878
2879void proto_unregister(struct proto *prot)
2880{
36b77a52 2881 mutex_lock(&proto_list_mutex);
13ff3d6f 2882 release_proto_idx(prot);
0a3f4358 2883 list_del(&prot->node);
36b77a52 2884 mutex_unlock(&proto_list_mutex);
1da177e4 2885
adf78eda
JL		 2886 kmem_cache_destroy(prot->slab);
2887 prot->slab = NULL;
1da177e4 2888
0159dfd3 2889 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2890
6d6ee43e 2891 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2892 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2893 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2894 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2895 }
1da177e4 2896}
1da177e4
LT		 2897EXPORT_SYMBOL(proto_unregister);
2898
2899#ifdef CONFIG_PROC_FS
1da177e4 2900static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2901 __acquires(proto_list_mutex)
1da177e4 2902{
36b77a52 2903 mutex_lock(&proto_list_mutex);
60f0438a 2904 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2905}
2906
2907static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2908{
60f0438a 2909 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2910}
2911
2912static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2913 __releases(proto_list_mutex)
1da177e4 2914{
36b77a52 2915 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2916}
2917
2918static char proto_method_implemented(const void *method)
2919{
2920 return method == NULL ? 'n' : 'y';
2921}
180d8cd9
GC		 2922static long sock_prot_memory_allocated(struct proto *proto)
2923{
cb75a36c 2924 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2925}
2926
2927static char *sock_prot_memory_pressure(struct proto *proto)
2928{
2929 return proto->memory_pressure != NULL ?
2930 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2931}
1da177e4
LT		 2932
2933static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2934{
180d8cd9 2935
8d987e5c 2936 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2937 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2938 proto->name,
2939 proto->obj_size,
14e943db 2940 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2941 sock_prot_memory_allocated(proto),
2942 sock_prot_memory_pressure(proto),
1da177e4
LT		 2943 proto->max_header,
2944 proto->slab == NULL ? "no" : "yes",
2945 module_name(proto->owner),
2946 proto_method_implemented(proto->close),
2947 proto_method_implemented(proto->connect),
2948 proto_method_implemented(proto->disconnect),
2949 proto_method_implemented(proto->accept),
2950 proto_method_implemented(proto->ioctl),
2951 proto_method_implemented(proto->init),
2952 proto_method_implemented(proto->destroy),
2953 proto_method_implemented(proto->shutdown),
2954 proto_method_implemented(proto->setsockopt),
2955 proto_method_implemented(proto->getsockopt),
2956 proto_method_implemented(proto->sendmsg),
2957 proto_method_implemented(proto->recvmsg),
2958 proto_method_implemented(proto->sendpage),
2959 proto_method_implemented(proto->bind),
2960 proto_method_implemented(proto->backlog_rcv),
2961 proto_method_implemented(proto->hash),
2962 proto_method_implemented(proto->unhash),
2963 proto_method_implemented(proto->get_port),
2964 proto_method_implemented(proto->enter_memory_pressure));
2965}
2966
2967static int proto_seq_show(struct seq_file *seq, void *v)
2968{
60f0438a 2969 if (v == &proto_list)
1da177e4
LT		 2970 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2971 "protocol",
2972 "size",
2973 "sockets",
2974 "memory",
2975 "press",
2976 "maxhdr",
2977 "slab",
2978 "module",
2979 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2980 else
60f0438a 2981 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2982 return 0;
2983}
2984
f690808e 2985static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2986 .start = proto_seq_start,
2987 .next = proto_seq_next,
2988 .stop = proto_seq_stop,
2989 .show = proto_seq_show,
2990};
2991
2992static int proto_seq_open(struct inode *inode, struct file *file)
2993{
14e943db
ED		 2994 return seq_open_net(inode, file, &proto_seq_ops,
2995 sizeof(struct seq_net_private));
1da177e4
LT		 2996}
2997
9a32144e 2998static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2999 .owner = THIS_MODULE,
3000 .open = proto_seq_open,
3001 .read = seq_read,
3002 .llseek = seq_lseek,
14e943db
ED		 3003 .release = seq_release_net,
3004};
3005
3006static __net_init int proto_init_net(struct net *net)
3007{
d4beaa66 3008 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 3009 return -ENOMEM;
3010
3011 return 0;
3012}
3013
3014static __net_exit void proto_exit_net(struct net *net)
3015{
ece31ffd 3016 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 3017}
3018
3019
3020static __net_initdata struct pernet_operations proto_net_ops = {
3021 .init = proto_init_net,
3022 .exit = proto_exit_net,
1da177e4
LT		 3023};
3024
3025static int __init proto_init(void)
3026{
14e943db 3027 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 3028}
3029
3030subsys_initcall(proto_init);
3031
3032#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 1.728112 seconds and 5 git commands to generate.