projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
tcp: fix recv with flags MSG_WAITALL | MSG_PEEK
[linux-2.6-block.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
08e29af3
ED		 425#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
426
427static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 428{
08e29af3
ED		 429 if (sk->sk_flags & flags) {
430 sk->sk_flags &= ~flags;
431 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 432 net_disable_timestamp();
1da177e4
LT		 433 }
434}
435
436
f0088a50
DV		 437int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
438{
766e9037 439 int err;
3b885787
NH		 440 unsigned long flags;
441 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 442
0fd7bac6 443 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 444 atomic_inc(&sk->sk_drops);
3847ce32 445 trace_sock_rcvqueue_full(sk, skb);
766e9037 446 return -ENOMEM;
f0088a50
DV		 447 }
448
fda9ef5d 449 err = sk_filter(sk, skb);
f0088a50 450 if (err)
766e9037 451 return err;
f0088a50 452
c76562b6 453 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 454 atomic_inc(&sk->sk_drops);
455 return -ENOBUFS;
3ab224be
HA		 456 }
457
f0088a50
DV		 458 skb->dev = NULL;
459 skb_set_owner_r(skb, sk);
49ad9599 460
7fee226a
ED		 461 /* we escape from rcu protected region, make sure we dont leak
462 * a norefcounted dst
463 */
464 skb_dst_force(skb);
465
3b885787 466 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 467 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 468 __skb_queue_tail(list, skb);
469 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 470
471 if (!sock_flag(sk, SOCK_DEAD))
676d2369 472 sk->sk_data_ready(sk);
766e9037 473 return 0;
f0088a50
DV		 474}
475EXPORT_SYMBOL(sock_queue_rcv_skb);
476
58a5a7b9 477int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 478{
479 int rc = NET_RX_SUCCESS;
480
fda9ef5d 481 if (sk_filter(sk, skb))
f0088a50
DV		 482 goto discard_and_relse;
483
484 skb->dev = NULL;
485
274f482d 486 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 487 atomic_inc(&sk->sk_drops);
488 goto discard_and_relse;
489 }
58a5a7b9
ACM		 490 if (nested)
491 bh_lock_sock_nested(sk);
492 else
493 bh_lock_sock(sk);
a5b5bb9a
IM		 494 if (!sock_owned_by_user(sk)) {
495 /*
496 * trylock + unlock semantics:
497 */
498 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
499
c57943a1 500 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 501
502 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 503 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 504 bh_unlock_sock(sk);
505 atomic_inc(&sk->sk_drops);
506 goto discard_and_relse;
507 }
508
f0088a50
DV		 509 bh_unlock_sock(sk);
510out:
511 sock_put(sk);
512 return rc;
513discard_and_relse:
514 kfree_skb(skb);
515 goto out;
516}
517EXPORT_SYMBOL(sk_receive_skb);
518
519struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
520{
b6c6712a 521 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 522
523 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 524 sk_tx_queue_clear(sk);
a9b3cd7f 525 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 526 dst_release(dst);
527 return NULL;
528 }
529
530 return dst;
531}
532EXPORT_SYMBOL(__sk_dst_check);
533
534struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
535{
536 struct dst_entry *dst = sk_dst_get(sk);
537
538 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
539 sk_dst_reset(sk);
540 dst_release(dst);
541 return NULL;
542 }
543
544 return dst;
545}
546EXPORT_SYMBOL(sk_dst_check);
547
c91f6df2
BH		 548static int sock_setbindtodevice(struct sock *sk, char __user *optval,
549 int optlen)
4878809f
DM		 550{
551 int ret = -ENOPROTOOPT;
552#ifdef CONFIG_NETDEVICES
3b1e0a65 553 struct net *net = sock_net(sk);
4878809f
DM		 554 char devname[IFNAMSIZ];
555 int index;
556
557 /* Sorry... */
558 ret = -EPERM;
5e1fccc0 559 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 560 goto out;
561
562 ret = -EINVAL;
563 if (optlen < 0)
564 goto out;
565
566 /* Bind this socket to a particular device like "eth0",
567 * as specified in the passed interface name. If the
568 * name is "" or the option length is zero the socket
569 * is not bound.
570 */
571 if (optlen > IFNAMSIZ - 1)
572 optlen = IFNAMSIZ - 1;
573 memset(devname, 0, sizeof(devname));
574
575 ret = -EFAULT;
576 if (copy_from_user(devname, optval, optlen))
577 goto out;
578
000ba2e4
DM		 579 index = 0;
580 if (devname[0] != '\0') {
bf8e56bf 581 struct net_device *dev;
4878809f 582
bf8e56bf
ED		 583 rcu_read_lock();
584 dev = dev_get_by_name_rcu(net, devname);
585 if (dev)
586 index = dev->ifindex;
587 rcu_read_unlock();
4878809f
DM		 588 ret = -ENODEV;
589 if (!dev)
590 goto out;
4878809f
DM		 591 }
592
593 lock_sock(sk);
594 sk->sk_bound_dev_if = index;
595 sk_dst_reset(sk);
596 release_sock(sk);
597
598 ret = 0;
599
600out:
601#endif
602
603 return ret;
604}
605
c91f6df2
BH		 606static int sock_getbindtodevice(struct sock *sk, char __user *optval,
607 int __user *optlen, int len)
608{
609 int ret = -ENOPROTOOPT;
610#ifdef CONFIG_NETDEVICES
611 struct net *net = sock_net(sk);
c91f6df2 612 char devname[IFNAMSIZ];
c91f6df2
BH		 613
614 if (sk->sk_bound_dev_if == 0) {
615 len = 0;
616 goto zero;
617 }
618
619 ret = -EINVAL;
620 if (len < IFNAMSIZ)
621 goto out;
622
5dbe7c17
NS		 623 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
624 if (ret)
c91f6df2 625 goto out;
c91f6df2
BH		 626
627 len = strlen(devname) + 1;
628
629 ret = -EFAULT;
630 if (copy_to_user(optval, devname, len))
631 goto out;
632
633zero:
634 ret = -EFAULT;
635 if (put_user(len, optlen))
636 goto out;
637
638 ret = 0;
639
640out:
641#endif
642
643 return ret;
644}
645
c0ef877b
PE		 646static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
647{
648 if (valbool)
649 sock_set_flag(sk, bit);
650 else
651 sock_reset_flag(sk, bit);
652}
653
f60e5990 654bool sk_mc_loop(struct sock *sk)
655{
656 if (dev_recursion_level())
657 return false;
658 if (!sk)
659 return true;
660 switch (sk->sk_family) {
661 case AF_INET:
662 return inet_sk(sk)->mc_loop;
663#if IS_ENABLED(CONFIG_IPV6)
664 case AF_INET6:
665 return inet6_sk(sk)->mc_loop;
666#endif
667 }
668 WARN_ON(1);
669 return true;
670}
671EXPORT_SYMBOL(sk_mc_loop);
672
1da177e4
LT		 673/*
674 * This is meant for all protocols to use and covers goings on
675 * at the socket level. Everything here is generic.
676 */
677
678int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 679 char __user *optval, unsigned int optlen)
1da177e4 680{
2a91525c 681 struct sock *sk = sock->sk;
1da177e4
LT		 682 int val;
683 int valbool;
684 struct linger ling;
685 int ret = 0;
4ec93edb 686
1da177e4
LT		 687 /*
688 * Options without arguments
689 */
690
4878809f 691 if (optname == SO_BINDTODEVICE)
c91f6df2 692 return sock_setbindtodevice(sk, optval, optlen);
4878809f 693
e71a4783
SH		 694 if (optlen < sizeof(int))
695 return -EINVAL;
4ec93edb 696
1da177e4
LT		 697 if (get_user(val, (int __user *)optval))
698 return -EFAULT;
4ec93edb 699
2a91525c 700 valbool = val ? 1 : 0;
1da177e4
LT		 701
702 lock_sock(sk);
703
2a91525c 704 switch (optname) {
e71a4783 705 case SO_DEBUG:
2a91525c 706 if (val && !capable(CAP_NET_ADMIN))
e71a4783 707 ret = -EACCES;
2a91525c 708 else
c0ef877b 709 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 710 break;
711 case SO_REUSEADDR:
4a17fd52 712 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 713 break;
055dc21a
TH		 714 case SO_REUSEPORT:
715 sk->sk_reuseport = valbool;
716 break;
e71a4783 717 case SO_TYPE:
49c794e9 718 case SO_PROTOCOL:
0d6038ee 719 case SO_DOMAIN:
e71a4783
SH		 720 case SO_ERROR:
721 ret = -ENOPROTOOPT;
722 break;
723 case SO_DONTROUTE:
c0ef877b 724 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 725 break;
726 case SO_BROADCAST:
727 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
728 break;
729 case SO_SNDBUF:
730 /* Don't error on this BSD doesn't and if you think
82981930
ED		 731 * about it this is right. Otherwise apps have to
732 * play 'guess the biggest size' games. RCVBUF/SNDBUF
733 * are treated in BSD as hints
734 */
735 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 736set_sndbuf:
e71a4783 737 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 738 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
739 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 740 sk->sk_write_space(sk);
741 break;
1da177e4 742
e71a4783
SH		 743 case SO_SNDBUFFORCE:
744 if (!capable(CAP_NET_ADMIN)) {
745 ret = -EPERM;
746 break;
747 }
748 goto set_sndbuf;
b0573dea 749
e71a4783
SH		 750 case SO_RCVBUF:
751 /* Don't error on this BSD doesn't and if you think
82981930
ED		 752 * about it this is right. Otherwise apps have to
753 * play 'guess the biggest size' games. RCVBUF/SNDBUF
754 * are treated in BSD as hints
755 */
756 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 757set_rcvbuf:
e71a4783
SH		 758 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
759 /*
760 * We double it on the way in to account for
761 * "struct sk_buff" etc. overhead. Applications
762 * assume that the SO_RCVBUF setting they make will
763 * allow that much actual data to be received on that
764 * socket.
765 *
766 * Applications are unaware that "struct sk_buff" and
767 * other overheads allocate from the receive buffer
768 * during socket buffer allocation.
769 *
770 * And after considering the possible alternatives,
771 * returning the value we actually used in getsockopt
772 * is the most desirable behavior.
773 */
82981930 774 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 775 break;
776
777 case SO_RCVBUFFORCE:
778 if (!capable(CAP_NET_ADMIN)) {
779 ret = -EPERM;
1da177e4 780 break;
e71a4783
SH		 781 }
782 goto set_rcvbuf;
1da177e4 783
e71a4783 784 case SO_KEEPALIVE:
1da177e4 785#ifdef CONFIG_INET
3e10986d
ED		 786 if (sk->sk_protocol == IPPROTO_TCP &&
787 sk->sk_type == SOCK_STREAM)
e71a4783 788 tcp_set_keepalive(sk, valbool);
1da177e4 789#endif
e71a4783
SH		 790 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
791 break;
792
793 case SO_OOBINLINE:
794 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
795 break;
796
797 case SO_NO_CHECK:
28448b80 798 sk->sk_no_check_tx = valbool;
e71a4783
SH		 799 break;
800
801 case SO_PRIORITY:
5e1fccc0
EB		 802 if ((val >= 0 && val <= 6) ||
803 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 804 sk->sk_priority = val;
805 else
806 ret = -EPERM;
807 break;
808
809 case SO_LINGER:
810 if (optlen < sizeof(ling)) {
811 ret = -EINVAL; /* 1003.1g */
1da177e4 812 break;
e71a4783 813 }
2a91525c 814 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 815 ret = -EFAULT;
1da177e4 816 break;
e71a4783
SH		 817 }
818 if (!ling.l_onoff)
819 sock_reset_flag(sk, SOCK_LINGER);
820 else {
1da177e4 821#if (BITS_PER_LONG == 32)
e71a4783
SH		 822 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
823 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 824 else
e71a4783
SH		 825#endif
826 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
827 sock_set_flag(sk, SOCK_LINGER);
828 }
829 break;
830
831 case SO_BSDCOMPAT:
832 sock_warn_obsolete_bsdism("setsockopt");
833 break;
834
835 case SO_PASSCRED:
836 if (valbool)
837 set_bit(SOCK_PASSCRED, &sock->flags);
838 else
839 clear_bit(SOCK_PASSCRED, &sock->flags);
840 break;
841
842 case SO_TIMESTAMP:
92f37fd2 843 case SO_TIMESTAMPNS:
e71a4783 844 if (valbool) {
92f37fd2
ED		 845 if (optname == SO_TIMESTAMP)
846 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
847 else
848 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 849 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 850 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 851 } else {
e71a4783 852 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 853 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
854 }
e71a4783
SH		 855 break;
856
20d49473
PO		 857 case SO_TIMESTAMPING:
858 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 859 ret = -EINVAL;
20d49473
PO		 860 break;
861 }
b245be1f 862
09c2d251 863 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765
WB		 864 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
865 if (sk->sk_protocol == IPPROTO_TCP) {
866 if (sk->sk_state != TCP_ESTABLISHED) {
867 ret = -EINVAL;
868 break;
869 }
870 sk->sk_tskey = tcp_sk(sk)->snd_una;
871 } else {
872 sk->sk_tskey = 0;
873 }
874 }
b9f40e21 875 sk->sk_tsflags = val;
20d49473
PO		 876 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
877 sock_enable_timestamp(sk,
878 SOCK_TIMESTAMPING_RX_SOFTWARE);
879 else
880 sock_disable_timestamp(sk,
08e29af3 881 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 882 break;
883
e71a4783
SH		 884 case SO_RCVLOWAT:
885 if (val < 0)
886 val = INT_MAX;
887 sk->sk_rcvlowat = val ? : 1;
888 break;
889
890 case SO_RCVTIMEO:
891 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
892 break;
893
894 case SO_SNDTIMEO:
895 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
896 break;
1da177e4 897
e71a4783
SH		 898 case SO_ATTACH_FILTER:
899 ret = -EINVAL;
900 if (optlen == sizeof(struct sock_fprog)) {
901 struct sock_fprog fprog;
1da177e4 902
e71a4783
SH		 903 ret = -EFAULT;
904 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 905 break;
e71a4783
SH		 906
907 ret = sk_attach_filter(&fprog, sk);
908 }
909 break;
910
89aa0758
AS		 911 case SO_ATTACH_BPF:
912 ret = -EINVAL;
913 if (optlen == sizeof(u32)) {
914 u32 ufd;
915
916 ret = -EFAULT;
917 if (copy_from_user(&ufd, optval, sizeof(ufd)))
918 break;
919
920 ret = sk_attach_bpf(ufd, sk);
921 }
922 break;
923
e71a4783 924 case SO_DETACH_FILTER:
55b33325 925 ret = sk_detach_filter(sk);
e71a4783 926 break;
1da177e4 927
d59577b6
VB		 928 case SO_LOCK_FILTER:
929 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
930 ret = -EPERM;
931 else
932 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
933 break;
934
e71a4783
SH		 935 case SO_PASSSEC:
936 if (valbool)
937 set_bit(SOCK_PASSSEC, &sock->flags);
938 else
939 clear_bit(SOCK_PASSSEC, &sock->flags);
940 break;
4a19ec58 941 case SO_MARK:
5e1fccc0 942 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 943 ret = -EPERM;
2a91525c 944 else
4a19ec58 945 sk->sk_mark = val;
4a19ec58 946 break;
877ce7c1 947
3b885787 948 case SO_RXQ_OVFL:
8083f0fc 949 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 950 break;
6e3e939f
JB		 951
952 case SO_WIFI_STATUS:
953 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
954 break;
955
ef64a54f
PE		 956 case SO_PEEK_OFF:
957 if (sock->ops->set_peek_off)
12663bfc 958 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 959 else
960 ret = -EOPNOTSUPP;
961 break;
3bdc0eba
BG		 962
963 case SO_NOFCS:
964 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
965 break;
966
7d4c04fc
KJ		 967 case SO_SELECT_ERR_QUEUE:
968 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
969 break;
970
e0d1095a 971#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 972 case SO_BUSY_POLL:
dafcc438
ET		 973 /* allow unprivileged users to decrease the value */
974 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
975 ret = -EPERM;
976 else {
977 if (val < 0)
978 ret = -EINVAL;
979 else
980 sk->sk_ll_usec = val;
981 }
982 break;
983#endif
62748f32
ED		 984
985 case SO_MAX_PACING_RATE:
986 sk->sk_max_pacing_rate = val;
987 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
988 sk->sk_max_pacing_rate);
989 break;
990
e71a4783
SH		 991 default:
992 ret = -ENOPROTOOPT;
993 break;
4ec93edb 994 }
1da177e4
LT		 995 release_sock(sk);
996 return ret;
997}
2a91525c 998EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 999
1000
8f09898b 1001static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1002 struct ucred *ucred)
3f551f94
EB		 1003{
1004 ucred->pid = pid_vnr(pid);
1005 ucred->uid = ucred->gid = -1;
1006 if (cred) {
1007 struct user_namespace *current_ns = current_user_ns();
1008
b2e4f544
EB		 1009 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1010 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1011 }
1012}
1013
1da177e4
LT		 1014int sock_getsockopt(struct socket *sock, int level, int optname,
1015 char __user *optval, int __user *optlen)
1016{
1017 struct sock *sk = sock->sk;
4ec93edb 1018
e71a4783 1019 union {
4ec93edb
YH		 1020 int val;
1021 struct linger ling;
1da177e4
LT		 1022 struct timeval tm;
1023 } v;
4ec93edb 1024
4d0392be 1025 int lv = sizeof(int);
1da177e4 1026 int len;
4ec93edb 1027
e71a4783 1028 if (get_user(len, optlen))
4ec93edb 1029 return -EFAULT;
e71a4783 1030 if (len < 0)
1da177e4 1031 return -EINVAL;
4ec93edb 1032
50fee1de 1033 memset(&v, 0, sizeof(v));
df0bca04 1034
2a91525c 1035 switch (optname) {
e71a4783
SH		 1036 case SO_DEBUG:
1037 v.val = sock_flag(sk, SOCK_DBG);
1038 break;
1039
1040 case SO_DONTROUTE:
1041 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1042 break;
1043
1044 case SO_BROADCAST:
1b23a5df 1045 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1046 break;
1047
1048 case SO_SNDBUF:
1049 v.val = sk->sk_sndbuf;
1050 break;
1051
1052 case SO_RCVBUF:
1053 v.val = sk->sk_rcvbuf;
1054 break;
1055
1056 case SO_REUSEADDR:
1057 v.val = sk->sk_reuse;
1058 break;
1059
055dc21a
TH		 1060 case SO_REUSEPORT:
1061 v.val = sk->sk_reuseport;
1062 break;
1063
e71a4783 1064 case SO_KEEPALIVE:
1b23a5df 1065 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1066 break;
1067
1068 case SO_TYPE:
1069 v.val = sk->sk_type;
1070 break;
1071
49c794e9
JE		 1072 case SO_PROTOCOL:
1073 v.val = sk->sk_protocol;
1074 break;
1075
0d6038ee
JE		 1076 case SO_DOMAIN:
1077 v.val = sk->sk_family;
1078 break;
1079
e71a4783
SH		 1080 case SO_ERROR:
1081 v.val = -sock_error(sk);
2a91525c 1082 if (v.val == 0)
e71a4783
SH		 1083 v.val = xchg(&sk->sk_err_soft, 0);
1084 break;
1085
1086 case SO_OOBINLINE:
1b23a5df 1087 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1088 break;
1089
1090 case SO_NO_CHECK:
28448b80 1091 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1092 break;
1093
1094 case SO_PRIORITY:
1095 v.val = sk->sk_priority;
1096 break;
1097
1098 case SO_LINGER:
1099 lv = sizeof(v.ling);
1b23a5df 1100 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1101 v.ling.l_linger = sk->sk_lingertime / HZ;
1102 break;
1103
1104 case SO_BSDCOMPAT:
1105 sock_warn_obsolete_bsdism("getsockopt");
1106 break;
1107
1108 case SO_TIMESTAMP:
92f37fd2
ED		 1109 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1110 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1111 break;
1112
1113 case SO_TIMESTAMPNS:
1114 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1115 break;
1116
20d49473 1117 case SO_TIMESTAMPING:
b9f40e21 1118 v.val = sk->sk_tsflags;
20d49473
PO		 1119 break;
1120
e71a4783 1121 case SO_RCVTIMEO:
2a91525c 1122 lv = sizeof(struct timeval);
e71a4783
SH		 1123 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1124 v.tm.tv_sec = 0;
1125 v.tm.tv_usec = 0;
1126 } else {
1127 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1128 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1129 }
1130 break;
1131
1132 case SO_SNDTIMEO:
2a91525c 1133 lv = sizeof(struct timeval);
e71a4783
SH		 1134 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1135 v.tm.tv_sec = 0;
1136 v.tm.tv_usec = 0;
1137 } else {
1138 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1139 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1140 }
1141 break;
1da177e4 1142
e71a4783
SH		 1143 case SO_RCVLOWAT:
1144 v.val = sk->sk_rcvlowat;
1145 break;
1da177e4 1146
e71a4783 1147 case SO_SNDLOWAT:
2a91525c 1148 v.val = 1;
e71a4783 1149 break;
1da177e4 1150
e71a4783 1151 case SO_PASSCRED:
82981930 1152 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1153 break;
1da177e4 1154
e71a4783 1155 case SO_PEERCRED:
109f6e39
EB		 1156 {
1157 struct ucred peercred;
1158 if (len > sizeof(peercred))
1159 len = sizeof(peercred);
1160 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1161 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1162 return -EFAULT;
1163 goto lenout;
109f6e39 1164 }
1da177e4 1165
e71a4783
SH		 1166 case SO_PEERNAME:
1167 {
1168 char address[128];
1169
1170 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1171 return -ENOTCONN;
1172 if (lv < len)
1173 return -EINVAL;
1174 if (copy_to_user(optval, address, len))
1175 return -EFAULT;
1176 goto lenout;
1177 }
1da177e4 1178
e71a4783
SH		 1179 /* Dubious BSD thing... Probably nobody even uses it, but
1180 * the UNIX standard wants it for whatever reason... -DaveM
1181 */
1182 case SO_ACCEPTCONN:
1183 v.val = sk->sk_state == TCP_LISTEN;
1184 break;
1da177e4 1185
e71a4783 1186 case SO_PASSSEC:
82981930 1187 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1188 break;
877ce7c1 1189
e71a4783
SH		 1190 case SO_PEERSEC:
1191 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1192
4a19ec58
LAT		 1193 case SO_MARK:
1194 v.val = sk->sk_mark;
1195 break;
1196
3b885787 1197 case SO_RXQ_OVFL:
1b23a5df 1198 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1199 break;
1200
6e3e939f 1201 case SO_WIFI_STATUS:
1b23a5df 1202 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1203 break;
1204
ef64a54f
PE		 1205 case SO_PEEK_OFF:
1206 if (!sock->ops->set_peek_off)
1207 return -EOPNOTSUPP;
1208
1209 v.val = sk->sk_peek_off;
1210 break;
bc2f7996 1211 case SO_NOFCS:
1b23a5df 1212 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1213 break;
c91f6df2 1214
f7b86bfe 1215 case SO_BINDTODEVICE:
c91f6df2
BH		 1216 return sock_getbindtodevice(sk, optval, optlen, len);
1217
a8fc9277
PE		 1218 case SO_GET_FILTER:
1219 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1220 if (len < 0)
1221 return len;
1222
1223 goto lenout;
c91f6df2 1224
d59577b6
VB		 1225 case SO_LOCK_FILTER:
1226 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1227 break;
1228
ea02f941
MS		 1229 case SO_BPF_EXTENSIONS:
1230 v.val = bpf_tell_extensions();
1231 break;
1232
7d4c04fc
KJ		 1233 case SO_SELECT_ERR_QUEUE:
1234 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1235 break;
1236
e0d1095a 1237#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1238 case SO_BUSY_POLL:
dafcc438
ET		 1239 v.val = sk->sk_ll_usec;
1240 break;
1241#endif
1242
62748f32
ED		 1243 case SO_MAX_PACING_RATE:
1244 v.val = sk->sk_max_pacing_rate;
1245 break;
1246
2c8c56e1
ED		 1247 case SO_INCOMING_CPU:
1248 v.val = sk->sk_incoming_cpu;
1249 break;
1250
e71a4783 1251 default:
443b5991
YH		 1252 /* We implement the SO_SNDLOWAT etc to not be settable
1253 * (1003.1g 7).
1254 */
e71a4783 1255 return -ENOPROTOOPT;
1da177e4 1256 }
e71a4783 1257
1da177e4
LT		 1258 if (len > lv)
1259 len = lv;
1260 if (copy_to_user(optval, &v, len))
1261 return -EFAULT;
1262lenout:
4ec93edb
YH		 1263 if (put_user(len, optlen))
1264 return -EFAULT;
1265 return 0;
1da177e4
LT		 1266}
1267
a5b5bb9a
IM		 1268/*
1269 * Initialize an sk_lock.
1270 *
1271 * (We also register the sk_lock with the lock validator.)
1272 */
b6f99a21 1273static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1274{
ed07536e
PZ		 1275 sock_lock_init_class_and_name(sk,
1276 af_family_slock_key_strings[sk->sk_family],
1277 af_family_slock_keys + sk->sk_family,
1278 af_family_key_strings[sk->sk_family],
1279 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1280}
1281
4dc6dc71
ED		 1282/*
1283 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1284 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1285 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1286 */
f1a6c4da
PE		 1287static void sock_copy(struct sock *nsk, const struct sock *osk)
1288{
1289#ifdef CONFIG_SECURITY_NETWORK
1290 void *sptr = nsk->sk_security;
1291#endif
68835aba
ED		 1292 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1293
1294 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1295 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1296
f1a6c4da
PE		 1297#ifdef CONFIG_SECURITY_NETWORK
1298 nsk->sk_security = sptr;
1299 security_sk_clone(osk, nsk);
1300#endif
1301}
1302
fcbdf09d
OP		 1303void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1304{
1305 unsigned long nulls1, nulls2;
1306
1307 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1308 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1309 if (nulls1 > nulls2)
1310 swap(nulls1, nulls2);
1311
1312 if (nulls1 != 0)
1313 memset((char *)sk, 0, nulls1);
1314 memset((char *)sk + nulls1 + sizeof(void *), 0,
1315 nulls2 - nulls1 - sizeof(void *));
1316 memset((char *)sk + nulls2 + sizeof(void *), 0,
1317 size - nulls2 - sizeof(void *));
1318}
1319EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1320
2e4afe7b
PE		 1321static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1322 int family)
c308c1b2
PE		 1323{
1324 struct sock *sk;
1325 struct kmem_cache *slab;
1326
1327 slab = prot->slab;
e912b114
ED		 1328 if (slab != NULL) {
1329 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1330 if (!sk)
1331 return sk;
1332 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1333 if (prot->clear_sk)
1334 prot->clear_sk(sk, prot->obj_size);
1335 else
1336 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1337 }
fcbdf09d 1338 } else
c308c1b2
PE		 1339 sk = kmalloc(prot->obj_size, priority);
1340
2e4afe7b 1341 if (sk != NULL) {
a98b65a3
VN		 1342 kmemcheck_annotate_bitfield(sk, flags);
1343
2e4afe7b
PE		 1344 if (security_sk_alloc(sk, family, priority))
1345 goto out_free;
1346
1347 if (!try_module_get(prot->owner))
1348 goto out_free_sec;
e022f0b4 1349 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1350 }
1351
c308c1b2 1352 return sk;
2e4afe7b
PE		 1353
1354out_free_sec:
1355 security_sk_free(sk);
1356out_free:
1357 if (slab != NULL)
1358 kmem_cache_free(slab, sk);
1359 else
1360 kfree(sk);
1361 return NULL;
c308c1b2
PE		 1362}
1363
1364static void sk_prot_free(struct proto *prot, struct sock *sk)
1365{
1366 struct kmem_cache *slab;
2e4afe7b 1367 struct module *owner;
c308c1b2 1368
2e4afe7b 1369 owner = prot->owner;
c308c1b2 1370 slab = prot->slab;
2e4afe7b
PE		 1371
1372 security_sk_free(sk);
c308c1b2
PE		 1373 if (slab != NULL)
1374 kmem_cache_free(slab, sk);
1375 else
1376 kfree(sk);
2e4afe7b 1377 module_put(owner);
c308c1b2
PE		 1378}
1379
86f8515f 1380#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1381void sock_update_netprioidx(struct sock *sk)
5bc1421e 1382{
5bc1421e
NH		 1383 if (in_interrupt())
1384 return;
2b73bc65 1385
6ffd4641 1386 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1387}
1388EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1389#endif
1390
1da177e4
LT		 1391/**
1392 * sk_alloc - All socket objects are allocated here
c4ea43c5 1393 * @net: the applicable net namespace
4dc3b16b
PP		 1394 * @family: protocol family
1395 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1396 * @prot: struct proto associated with this new sock instance
11aa9c28 1397 * @kern: is this to be a kernel socket?
1da177e4 1398 */
1b8d7ae4 1399struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1400 struct proto *prot, int kern)
1da177e4 1401{
c308c1b2 1402 struct sock *sk;
1da177e4 1403
154adbc8 1404 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1405 if (sk) {
154adbc8
PE		 1406 sk->sk_family = family;
1407 /*
1408 * See comment in struct sock definition to understand
1409 * why we need sk_prot_creator -acme
1410 */
1411 sk->sk_prot = sk->sk_prot_creator = prot;
1412 sock_lock_init(sk);
26abe143
EB		 1413 sk->sk_net_refcnt = kern ? 0 : 1;
1414 if (likely(sk->sk_net_refcnt))
1415 get_net(net);
1416 sock_net_set(sk, net);
d66ee058 1417 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1418
211d2f97 1419 sock_update_classid(sk);
6ffd4641 1420 sock_update_netprioidx(sk);
1da177e4 1421 }
a79af59e 1422
2e4afe7b 1423 return sk;
1da177e4 1424}
2a91525c 1425EXPORT_SYMBOL(sk_alloc);
1da177e4 1426
eb4cb008 1427void sk_destruct(struct sock *sk)
1da177e4
LT		 1428{
1429 struct sk_filter *filter;
1da177e4
LT		 1430
1431 if (sk->sk_destruct)
1432 sk->sk_destruct(sk);
1433
a898def2
PM		 1434 filter = rcu_dereference_check(sk->sk_filter,
1435 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1436 if (filter) {
309dd5fc 1437 sk_filter_uncharge(sk, filter);
a9b3cd7f 1438 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1439 }
1440
08e29af3 1441 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1442
1443 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1444 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1445 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1446
109f6e39
EB		 1447 if (sk->sk_peer_cred)
1448 put_cred(sk->sk_peer_cred);
1449 put_pid(sk->sk_peer_pid);
26abe143
EB		 1450 if (likely(sk->sk_net_refcnt))
1451 put_net(sock_net(sk));
c308c1b2 1452 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1453}
2b85a34e 1454
eb4cb008
CG		 1455static void __sk_free(struct sock *sk)
1456{
b922622e 1457 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1458 sock_diag_broadcast_destroy(sk);
1459 else
1460 sk_destruct(sk);
1461}
1462
2b85a34e
ED		 1463void sk_free(struct sock *sk)
1464{
1465 /*
25985edc 1466 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1467 * some packets are still in some tx queue.
1468 * If not null, sock_wfree() will call __sk_free(sk) later
1469 */
1470 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1471 __sk_free(sk);
1472}
2a91525c 1473EXPORT_SYMBOL(sk_free);
1da177e4 1474
475f1b52
SR		 1475static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1476{
1477 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1478 sock_update_memcg(newsk);
1479}
1480
e56c57d0
ED		 1481/**
1482 * sk_clone_lock - clone a socket, and lock its clone
1483 * @sk: the socket to clone
1484 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1485 *
1486 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1487 */
1488struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1489{
8fd1d178 1490 struct sock *newsk;
278571ba 1491 bool is_charged = true;
87d11ceb 1492
8fd1d178 1493 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1494 if (newsk != NULL) {
1495 struct sk_filter *filter;
1496
892c141e 1497 sock_copy(newsk, sk);
87d11ceb
ACM		 1498
1499 /* SANITY */
3b1e0a65 1500 get_net(sock_net(newsk));
87d11ceb
ACM		 1501 sk_node_init(&newsk->sk_node);
1502 sock_lock_init(newsk);
1503 bh_lock_sock(newsk);
fa438ccf 1504 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1505 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1506
1507 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1508 /*
1509 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1510 */
1511 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1512 atomic_set(&newsk->sk_omem_alloc, 0);
1513 skb_queue_head_init(&newsk->sk_receive_queue);
1514 skb_queue_head_init(&newsk->sk_write_queue);
1515
b6c6712a 1516 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1517 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1518 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1519 af_callback_keys + newsk->sk_family,
1520 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1521
1522 newsk->sk_dst_cache = NULL;
1523 newsk->sk_wmem_queued = 0;
1524 newsk->sk_forward_alloc = 0;
1525 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1526 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1527
1528 sock_reset_flag(newsk, SOCK_DONE);
1529 skb_queue_head_init(&newsk->sk_error_queue);
1530
0d7da9dd 1531 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1532 if (filter != NULL)
278571ba
AS		 1533 /* though it's an empty new sock, the charging may fail
1534 * if sysctl_optmem_max was changed between creation of
1535 * original socket and cloning
1536 */
1537 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1538
278571ba 1539 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk))) {
87d11ceb
ACM		 1540 /* It is still raw copy of parent, so invalidate
1541 * destructor and make plain sk_free() */
1542 newsk->sk_destruct = NULL;
b0691c8e 1543 bh_unlock_sock(newsk);
87d11ceb
ACM		 1544 sk_free(newsk);
1545 newsk = NULL;
1546 goto out;
1547 }
1548
1549 newsk->sk_err = 0;
1550 newsk->sk_priority = 0;
2c8c56e1 1551 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1552 atomic64_set(&newsk->sk_cookie, 0);
4dc6dc71
ED		 1553 /*
1554 * Before updating sk_refcnt, we must commit prior changes to memory
1555 * (Documentation/RCU/rculist_nulls.txt for details)
1556 */
1557 smp_wmb();
87d11ceb
ACM		 1558 atomic_set(&newsk->sk_refcnt, 2);
1559
1560 /*
1561 * Increment the counter in the same struct proto as the master
1562 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1563 * is the same as sk->sk_prot->socks, as this field was copied
1564 * with memcpy).
1565 *
1566 * This _changes_ the previous behaviour, where
1567 * tcp_create_openreq_child always was incrementing the
1568 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1569 * to be taken into account in all callers. -acme
1570 */
1571 sk_refcnt_debug_inc(newsk);
972692e0 1572 sk_set_socket(newsk, NULL);
43815482 1573 newsk->sk_wq = NULL;
87d11ceb 1574
f3f511e1
GC		 1575 sk_update_clone(sk, newsk);
1576
87d11ceb 1577 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1578 sk_sockets_allocated_inc(newsk);
704da560 1579
08e29af3 1580 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1581 net_enable_timestamp();
87d11ceb
ACM		 1582 }
1583out:
1584 return newsk;
1585}
e56c57d0 1586EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1587
9958089a
AK		 1588void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1589{
d6a4e26a
ED		 1590 u32 max_segs = 1;
1591
9958089a
AK		 1592 __sk_dst_set(sk, dst);
1593 sk->sk_route_caps = dst->dev->features;
1594 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1595 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1596 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1597 if (sk_can_gso(sk)) {
82cc1a7a 1598 if (dst->header_len) {
9958089a 1599 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1600 } else {
9958089a 1601 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1602 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1603 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1604 }
9958089a 1605 }
d6a4e26a 1606 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1607}
1608EXPORT_SYMBOL_GPL(sk_setup_caps);
1609
1da177e4
LT		 1610/*
1611 * Simple resource managers for sockets.
1612 */
1613
1614
4ec93edb
YH		 1615/*
1616 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1617 */
1618void sock_wfree(struct sk_buff *skb)
1619{
1620 struct sock *sk = skb->sk;
d99927f4 1621 unsigned int len = skb->truesize;
1da177e4 1622
d99927f4
ED		 1623 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1624 /*
1625 * Keep a reference on sk_wmem_alloc, this will be released
1626 * after sk_write_space() call
1627 */
1628 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1629 sk->sk_write_space(sk);
d99927f4
ED		 1630 len = 1;
1631 }
2b85a34e 1632 /*
d99927f4
ED		 1633 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1634 * could not do because of in-flight packets
2b85a34e 1635 */
d99927f4 1636 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1637 __sk_free(sk);
1da177e4 1638}
2a91525c 1639EXPORT_SYMBOL(sock_wfree);
1da177e4 1640
f2f872f9
ED		 1641void skb_orphan_partial(struct sk_buff *skb)
1642{
1643 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1644 * so we do not completely orphan skb, but transfert all
1645 * accounted bytes but one, to avoid unexpected reorders.
1646 */
1647 if (skb->destructor == sock_wfree
1648#ifdef CONFIG_INET
1649 || skb->destructor == tcp_wfree
1650#endif
1651 ) {
1652 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1653 skb->truesize = 1;
1654 } else {
1655 skb_orphan(skb);
1656 }
1657}
1658EXPORT_SYMBOL(skb_orphan_partial);
1659
4ec93edb
YH		 1660/*
1661 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1662 */
1663void sock_rfree(struct sk_buff *skb)
1664{
1665 struct sock *sk = skb->sk;
d361fd59 1666 unsigned int len = skb->truesize;
1da177e4 1667
d361fd59
ED		 1668 atomic_sub(len, &sk->sk_rmem_alloc);
1669 sk_mem_uncharge(sk, len);
1da177e4 1670}
2a91525c 1671EXPORT_SYMBOL(sock_rfree);
1da177e4 1672
7768eed8
OH		 1673/*
1674 * Buffer destructor for skbs that are not used directly in read or write
1675 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1676 */
62bccb8c
AD		 1677void sock_efree(struct sk_buff *skb)
1678{
1679 sock_put(skb->sk);
1680}
1681EXPORT_SYMBOL(sock_efree);
1682
976d0201 1683kuid_t sock_i_uid(struct sock *sk)
1da177e4 1684{
976d0201 1685 kuid_t uid;
1da177e4 1686
f064af1e 1687 read_lock_bh(&sk->sk_callback_lock);
976d0201 1688 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1689 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1690 return uid;
1691}
2a91525c 1692EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1693
1694unsigned long sock_i_ino(struct sock *sk)
1695{
1696 unsigned long ino;
1697
f064af1e 1698 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1699 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1700 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1701 return ino;
1702}
2a91525c 1703EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1704
1705/*
1706 * Allocate a skb from the socket's send buffer.
1707 */
86a76caf 1708struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1709 gfp_t priority)
1da177e4
LT		 1710{
1711 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1712 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1713 if (skb) {
1714 skb_set_owner_w(skb, sk);
1715 return skb;
1716 }
1717 }
1718 return NULL;
1719}
2a91525c 1720EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1721
4ec93edb 1722/*
1da177e4 1723 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1724 */
dd0fc66f 1725void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1726{
95c96174 1727 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1728 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1729 void *mem;
1730 /* First do the add, to avoid the race if kmalloc
4ec93edb 1731 * might sleep.
1da177e4
LT		 1732 */
1733 atomic_add(size, &sk->sk_omem_alloc);
1734 mem = kmalloc(size, priority);
1735 if (mem)
1736 return mem;
1737 atomic_sub(size, &sk->sk_omem_alloc);
1738 }
1739 return NULL;
1740}
2a91525c 1741EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1742
79e88659
DB		 1743/* Free an option memory block. Note, we actually want the inline
1744 * here as this allows gcc to detect the nullify and fold away the
1745 * condition entirely.
1da177e4 1746 */
79e88659
DB		 1747static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1748 const bool nullify)
1da177e4 1749{
e53da5fb
DM		 1750 if (WARN_ON_ONCE(!mem))
1751 return;
79e88659
DB		 1752 if (nullify)
1753 kzfree(mem);
1754 else
1755 kfree(mem);
1da177e4
LT		 1756 atomic_sub(size, &sk->sk_omem_alloc);
1757}
79e88659
DB		 1758
1759void sock_kfree_s(struct sock *sk, void *mem, int size)
1760{
1761 __sock_kfree_s(sk, mem, size, false);
1762}
2a91525c 1763EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1764
79e88659
DB		 1765void sock_kzfree_s(struct sock *sk, void *mem, int size)
1766{
1767 __sock_kfree_s(sk, mem, size, true);
1768}
1769EXPORT_SYMBOL(sock_kzfree_s);
1770
1da177e4
LT		 1771/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1772 I think, these locks should be removed for datagram sockets.
1773 */
2a91525c 1774static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1775{
1776 DEFINE_WAIT(wait);
1777
1778 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1779 for (;;) {
1780 if (!timeo)
1781 break;
1782 if (signal_pending(current))
1783 break;
1784 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1785 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1786 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1787 break;
1788 if (sk->sk_shutdown & SEND_SHUTDOWN)
1789 break;
1790 if (sk->sk_err)
1791 break;
1792 timeo = schedule_timeout(timeo);
1793 }
aa395145 1794 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1795 return timeo;
1796}
1797
1798
1799/*
1800 * Generic send/receive buffer handlers
1801 */
1802
4cc7f68d
HX		 1803struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1804 unsigned long data_len, int noblock,
28d64271 1805 int *errcode, int max_page_order)
1da177e4 1806{
2e4e4410 1807 struct sk_buff *skb;
1da177e4
LT		 1808 long timeo;
1809 int err;
1810
1da177e4 1811 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1812 for (;;) {
1da177e4
LT		 1813 err = sock_error(sk);
1814 if (err != 0)
1815 goto failure;
1816
1817 err = -EPIPE;
1818 if (sk->sk_shutdown & SEND_SHUTDOWN)
1819 goto failure;
1820
2e4e4410
ED		 1821 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1822 break;
28d64271 1823
2e4e4410
ED		 1824 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1825 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1826 err = -EAGAIN;
1827 if (!timeo)
1da177e4 1828 goto failure;
2e4e4410
ED		 1829 if (signal_pending(current))
1830 goto interrupted;
1831 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1832 }
2e4e4410
ED		 1833 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1834 errcode, sk->sk_allocation);
1835 if (skb)
1836 skb_set_owner_w(skb, sk);
1da177e4
LT		 1837 return skb;
1838
1839interrupted:
1840 err = sock_intr_errno(timeo);
1841failure:
1842 *errcode = err;
1843 return NULL;
1844}
4cc7f68d 1845EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1846
4ec93edb 1847struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1848 int noblock, int *errcode)
1849{
28d64271 1850 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1851}
2a91525c 1852EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1853
5640f768
ED		 1854/* On 32bit arches, an skb frag is limited to 2^15 */
1855#define SKB_FRAG_PAGE_ORDER get_order(32768)
1856
400dfd3a
ED		 1857/**
1858 * skb_page_frag_refill - check that a page_frag contains enough room
1859 * @sz: minimum size of the fragment we want to get
1860 * @pfrag: pointer to page_frag
82d5e2b8 1861 * @gfp: priority for memory allocation
400dfd3a
ED		 1862 *
1863 * Note: While this allocator tries to use high order pages, there is
1864 * no guarantee that allocations succeed. Therefore, @sz MUST be
1865 * less or equal than PAGE_SIZE.
1866 */
d9b2938a 1867bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1868{
5640f768
ED		 1869 if (pfrag->page) {
1870 if (atomic_read(&pfrag->page->_count) == 1) {
1871 pfrag->offset = 0;
1872 return true;
1873 }
400dfd3a 1874 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1875 return true;
1876 put_page(pfrag->page);
1877 }
1878
d9b2938a
ED		 1879 pfrag->offset = 0;
1880 if (SKB_FRAG_PAGE_ORDER) {
fb05e7a8 1881 pfrag->page = alloc_pages((gfp & ~__GFP_WAIT) | __GFP_COMP |
d9b2938a
ED		 1882 __GFP_NOWARN | __GFP_NORETRY,
1883 SKB_FRAG_PAGE_ORDER);
5640f768 1884 if (likely(pfrag->page)) {
d9b2938a 1885 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1886 return true;
1887 }
d9b2938a
ED		 1888 }
1889 pfrag->page = alloc_page(gfp);
1890 if (likely(pfrag->page)) {
1891 pfrag->size = PAGE_SIZE;
1892 return true;
1893 }
400dfd3a
ED		 1894 return false;
1895}
1896EXPORT_SYMBOL(skb_page_frag_refill);
1897
1898bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1899{
1900 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1901 return true;
1902
5640f768
ED		 1903 sk_enter_memory_pressure(sk);
1904 sk_stream_moderate_sndbuf(sk);
1905 return false;
1906}
1907EXPORT_SYMBOL(sk_page_frag_refill);
1908
1da177e4 1909static void __lock_sock(struct sock *sk)
f39234d6
NK		 1910 __releases(&sk->sk_lock.slock)
1911 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1912{
1913 DEFINE_WAIT(wait);
1914
e71a4783 1915 for (;;) {
1da177e4
LT		 1916 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1917 TASK_UNINTERRUPTIBLE);
1918 spin_unlock_bh(&sk->sk_lock.slock);
1919 schedule();
1920 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1921 if (!sock_owned_by_user(sk))
1da177e4
LT		 1922 break;
1923 }
1924 finish_wait(&sk->sk_lock.wq, &wait);
1925}
1926
1927static void __release_sock(struct sock *sk)
f39234d6
NK		 1928 __releases(&sk->sk_lock.slock)
1929 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1930{
1931 struct sk_buff *skb = sk->sk_backlog.head;
1932
1933 do {
1934 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1935 bh_unlock_sock(sk);
1936
1937 do {
1938 struct sk_buff *next = skb->next;
1939
e4cbb02a 1940 prefetch(next);
7fee226a 1941 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1942 skb->next = NULL;
c57943a1 1943 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1944
1945 /*
1946 * We are in process context here with softirqs
1947 * disabled, use cond_resched_softirq() to preempt.
1948 * This is safe to do because we've taken the backlog
1949 * queue private:
1950 */
1951 cond_resched_softirq();
1952
1953 skb = next;
1954 } while (skb != NULL);
1955
1956 bh_lock_sock(sk);
e71a4783 1957 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1958
1959 /*
1960 * Doing the zeroing here guarantee we can not loop forever
1961 * while a wild producer attempts to flood us.
1962 */
1963 sk->sk_backlog.len = 0;
1da177e4
LT		 1964}
1965
1966/**
1967 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1968 * @sk: sock to wait on
1969 * @timeo: for how long
dfbafc99 1970 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 1971 *
1972 * Now socket state including sk->sk_err is changed only under lock,
1973 * hence we may omit checks after joining wait queue.
1974 * We check receive queue before schedule() only as optimization;
1975 * it is very likely that release_sock() added new data.
1976 */
dfbafc99 1977int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4
LT		 1978{
1979 int rc;
1980 DEFINE_WAIT(wait);
1981
aa395145 1982 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4 1983 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
dfbafc99 1984 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);
1da177e4 1985 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1986 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1987 return rc;
1988}
1da177e4
LT		 1989EXPORT_SYMBOL(sk_wait_data);
1990
3ab224be
HA		 1991/**
1992 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1993 * @sk: socket
1994 * @size: memory size to allocate
1995 * @kind: allocation type
1996 *
1997 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1998 * rmem allocation. This function assumes that protocols which have
1999 * memory_pressure use sk_wmem_queued as write buffer accounting.
2000 */
2001int __sk_mem_schedule(struct sock *sk, int size, int kind)
2002{
2003 struct proto *prot = sk->sk_prot;
2004 int amt = sk_mem_pages(size);
8d987e5c 2005 long allocated;
e1aab161 2006 int parent_status = UNDER_LIMIT;
3ab224be
HA		 2007
2008 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2009
e1aab161 2010 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 2011
2012 /* Under limit. */
e1aab161
GC		 2013 if (parent_status == UNDER_LIMIT &&
2014 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2015 sk_leave_memory_pressure(sk);
3ab224be
HA		 2016 return 1;
2017 }
2018
e1aab161
GC		 2019 /* Under pressure. (we or our parents) */
2020 if ((parent_status > SOFT_LIMIT) ||
2021 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2022 sk_enter_memory_pressure(sk);
3ab224be 2023
e1aab161
GC		 2024 /* Over hard limit (we or our parents) */
2025 if ((parent_status == OVER_LIMIT) ||
2026 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 2027 goto suppress_allocation;
2028
2029 /* guarantee minimum buffer size under pressure */
2030 if (kind == SK_MEM_RECV) {
2031 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2032 return 1;
180d8cd9 2033
3ab224be
HA		 2034 } else { /* SK_MEM_SEND */
2035 if (sk->sk_type == SOCK_STREAM) {
2036 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2037 return 1;
2038 } else if (atomic_read(&sk->sk_wmem_alloc) <
2039 prot->sysctl_wmem[0])
2040 return 1;
2041 }
2042
180d8cd9 2043 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2044 int alloc;
2045
180d8cd9 2046 if (!sk_under_memory_pressure(sk))
1748376b 2047 return 1;
180d8cd9
GC		 2048 alloc = sk_sockets_allocated_read_positive(sk);
2049 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2050 sk_mem_pages(sk->sk_wmem_queued +
2051 atomic_read(&sk->sk_rmem_alloc) +
2052 sk->sk_forward_alloc))
2053 return 1;
2054 }
2055
2056suppress_allocation:
2057
2058 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2059 sk_stream_moderate_sndbuf(sk);
2060
2061 /* Fail only if socket is _under_ its sndbuf.
2062 * In this case we cannot block, so that we have to fail.
2063 */
2064 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2065 return 1;
2066 }
2067
3847ce32
SM		 2068 trace_sock_exceed_buf_limit(sk, prot, allocated);
2069
3ab224be
HA		 2070 /* Alas. Undo changes. */
2071 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2072
0e90b31f 2073 sk_memory_allocated_sub(sk, amt);
180d8cd9 2074
3ab224be
HA		 2075 return 0;
2076}
3ab224be
HA		 2077EXPORT_SYMBOL(__sk_mem_schedule);
2078
2079/**
2080 * __sk_reclaim - reclaim memory_allocated
2081 * @sk: socket
1a24e04e 2082 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
3ab224be 2083 */
1a24e04e 2084void __sk_mem_reclaim(struct sock *sk, int amount)
3ab224be 2085{
1a24e04e
ED		 2086 amount >>= SK_MEM_QUANTUM_SHIFT;
2087 sk_memory_allocated_sub(sk, amount);
2088 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
3ab224be 2089
180d8cd9
GC		 2090 if (sk_under_memory_pressure(sk) &&
2091 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2092 sk_leave_memory_pressure(sk);
3ab224be 2093}
3ab224be
HA		 2094EXPORT_SYMBOL(__sk_mem_reclaim);
2095
2096
1da177e4
LT		 2097/*
2098 * Set of default routines for initialising struct proto_ops when
2099 * the protocol does not support a particular function. In certain
2100 * cases where it makes no sense for a protocol to have a "do nothing"
2101 * function, some default processing is provided.
2102 */
2103
2104int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2105{
2106 return -EOPNOTSUPP;
2107}
2a91525c 2108EXPORT_SYMBOL(sock_no_bind);
1da177e4 2109
4ec93edb 2110int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2111 int len, int flags)
2112{
2113 return -EOPNOTSUPP;
2114}
2a91525c 2115EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2116
2117int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2118{
2119 return -EOPNOTSUPP;
2120}
2a91525c 2121EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2122
2123int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2124{
2125 return -EOPNOTSUPP;
2126}
2a91525c 2127EXPORT_SYMBOL(sock_no_accept);
1da177e4 2128
4ec93edb 2129int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2130 int *len, int peer)
2131{
2132 return -EOPNOTSUPP;
2133}
2a91525c 2134EXPORT_SYMBOL(sock_no_getname);
1da177e4 2135
2a91525c 2136unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2137{
2138 return 0;
2139}
2a91525c 2140EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2141
2142int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2143{
2144 return -EOPNOTSUPP;
2145}
2a91525c 2146EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2147
2148int sock_no_listen(struct socket *sock, int backlog)
2149{
2150 return -EOPNOTSUPP;
2151}
2a91525c 2152EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2153
2154int sock_no_shutdown(struct socket *sock, int how)
2155{
2156 return -EOPNOTSUPP;
2157}
2a91525c 2158EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2159
2160int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2161 char __user *optval, unsigned int optlen)
1da177e4
LT		 2162{
2163 return -EOPNOTSUPP;
2164}
2a91525c 2165EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2166
2167int sock_no_getsockopt(struct socket *sock, int level, int optname,
2168 char __user *optval, int __user *optlen)
2169{
2170 return -EOPNOTSUPP;
2171}
2a91525c 2172EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2173
1b784140 2174int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2175{
2176 return -EOPNOTSUPP;
2177}
2a91525c 2178EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2179
1b784140
YX		 2180int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2181 int flags)
1da177e4
LT		 2182{
2183 return -EOPNOTSUPP;
2184}
2a91525c 2185EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2186
2187int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2188{
2189 /* Mirror missing mmap method error code */
2190 return -ENODEV;
2191}
2a91525c 2192EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2193
2194ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2195{
2196 ssize_t res;
2197 struct msghdr msg = {.msg_flags = flags};
2198 struct kvec iov;
2199 char *kaddr = kmap(page);
2200 iov.iov_base = kaddr + offset;
2201 iov.iov_len = size;
2202 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2203 kunmap(page);
2204 return res;
2205}
2a91525c 2206EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2207
2208/*
2209 * Default Socket Callbacks
2210 */
2211
2212static void sock_def_wakeup(struct sock *sk)
2213{
43815482
ED		 2214 struct socket_wq *wq;
2215
2216 rcu_read_lock();
2217 wq = rcu_dereference(sk->sk_wq);
2218 if (wq_has_sleeper(wq))
2219 wake_up_interruptible_all(&wq->wait);
2220 rcu_read_unlock();
1da177e4
LT		 2221}
2222
2223static void sock_def_error_report(struct sock *sk)
2224{
43815482
ED		 2225 struct socket_wq *wq;
2226
2227 rcu_read_lock();
2228 wq = rcu_dereference(sk->sk_wq);
2229 if (wq_has_sleeper(wq))
2230 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2231 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2232 rcu_read_unlock();
1da177e4
LT		 2233}
2234
676d2369 2235static void sock_def_readable(struct sock *sk)
1da177e4 2236{
43815482
ED		 2237 struct socket_wq *wq;
2238
2239 rcu_read_lock();
2240 wq = rcu_dereference(sk->sk_wq);
2241 if (wq_has_sleeper(wq))
2c6607c6 2242 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2243 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2244 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2245 rcu_read_unlock();
1da177e4
LT		 2246}
2247
2248static void sock_def_write_space(struct sock *sk)
2249{
43815482
ED		 2250 struct socket_wq *wq;
2251
2252 rcu_read_lock();
1da177e4
LT		 2253
2254 /* Do not wake up a writer until he can make "significant"
2255 * progress. --DaveM
2256 */
e71a4783 2257 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2258 wq = rcu_dereference(sk->sk_wq);
2259 if (wq_has_sleeper(wq))
2260 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2261 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2262
2263 /* Should agree with poll, otherwise some programs break */
2264 if (sock_writeable(sk))
8d8ad9d7 2265 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2266 }
2267
43815482 2268 rcu_read_unlock();
1da177e4
LT		 2269}
2270
2271static void sock_def_destruct(struct sock *sk)
2272{
1da177e4
LT		 2273}
2274
2275void sk_send_sigurg(struct sock *sk)
2276{
2277 if (sk->sk_socket && sk->sk_socket->file)
2278 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2279 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2280}
2a91525c 2281EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2282
2283void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2284 unsigned long expires)
2285{
2286 if (!mod_timer(timer, expires))
2287 sock_hold(sk);
2288}
1da177e4
LT		 2289EXPORT_SYMBOL(sk_reset_timer);
2290
2291void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2292{
25cc4ae9 2293 if (del_timer(timer))
1da177e4
LT		 2294 __sock_put(sk);
2295}
1da177e4
LT		 2296EXPORT_SYMBOL(sk_stop_timer);
2297
2298void sock_init_data(struct socket *sock, struct sock *sk)
2299{
2300 skb_queue_head_init(&sk->sk_receive_queue);
2301 skb_queue_head_init(&sk->sk_write_queue);
2302 skb_queue_head_init(&sk->sk_error_queue);
2303
2304 sk->sk_send_head = NULL;
2305
2306 init_timer(&sk->sk_timer);
4ec93edb 2307
1da177e4
LT		 2308 sk->sk_allocation = GFP_KERNEL;
2309 sk->sk_rcvbuf = sysctl_rmem_default;
2310 sk->sk_sndbuf = sysctl_wmem_default;
2311 sk->sk_state = TCP_CLOSE;
972692e0 2312 sk_set_socket(sk, sock);
1da177e4
LT		 2313
2314 sock_set_flag(sk, SOCK_ZAPPED);
2315
e71a4783 2316 if (sock) {
1da177e4 2317 sk->sk_type = sock->type;
43815482 2318 sk->sk_wq = sock->wq;
1da177e4
LT		 2319 sock->sk = sk;
2320 } else
43815482 2321 sk->sk_wq = NULL;
1da177e4 2322
b6c6712a 2323 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2324 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2325 lockdep_set_class_and_name(&sk->sk_callback_lock,
2326 af_callback_keys + sk->sk_family,
2327 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2328
2329 sk->sk_state_change = sock_def_wakeup;
2330 sk->sk_data_ready = sock_def_readable;
2331 sk->sk_write_space = sock_def_write_space;
2332 sk->sk_error_report = sock_def_error_report;
2333 sk->sk_destruct = sock_def_destruct;
2334
5640f768
ED		 2335 sk->sk_frag.page = NULL;
2336 sk->sk_frag.offset = 0;
ef64a54f 2337 sk->sk_peek_off = -1;
1da177e4 2338
109f6e39
EB		 2339 sk->sk_peer_pid = NULL;
2340 sk->sk_peer_cred = NULL;
1da177e4
LT		 2341 sk->sk_write_pending = 0;
2342 sk->sk_rcvlowat = 1;
2343 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2344 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2345
f37f0afb 2346 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2347
e0d1095a 2348#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2349 sk->sk_napi_id = 0;
64b0dc51 2350 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2351#endif
2352
62748f32 2353 sk->sk_max_pacing_rate = ~0U;
7eec4174 2354 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2355 /*
2356 * Before updating sk_refcnt, we must commit prior changes to memory
2357 * (Documentation/RCU/rculist_nulls.txt for details)
2358 */
2359 smp_wmb();
1da177e4 2360 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2361 atomic_set(&sk->sk_drops, 0);
1da177e4 2362}
2a91525c 2363EXPORT_SYMBOL(sock_init_data);
1da177e4 2364
b5606c2d 2365void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2366{
2367 might_sleep();
a5b5bb9a 2368 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2369 if (sk->sk_lock.owned)
1da177e4 2370 __lock_sock(sk);
d2e9117c 2371 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2372 spin_unlock(&sk->sk_lock.slock);
2373 /*
2374 * The sk_lock has mutex_lock() semantics here:
2375 */
fcc70d5f 2376 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2377 local_bh_enable();
1da177e4 2378}
fcc70d5f 2379EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2380
b5606c2d 2381void release_sock(struct sock *sk)
1da177e4 2382{
a5b5bb9a
IM		 2383 /*
2384 * The sk_lock has mutex_unlock() semantics:
2385 */
2386 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2387
2388 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2389 if (sk->sk_backlog.tail)
2390 __release_sock(sk);
46d3ceab 2391
c3f9b018
ED		 2392 /* Warning : release_cb() might need to release sk ownership,
2393 * ie call sock_release_ownership(sk) before us.
2394 */
46d3ceab
ED		 2395 if (sk->sk_prot->release_cb)
2396 sk->sk_prot->release_cb(sk);
2397
c3f9b018 2398 sock_release_ownership(sk);
a5b5bb9a
IM		 2399 if (waitqueue_active(&sk->sk_lock.wq))
2400 wake_up(&sk->sk_lock.wq);
2401 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2402}
2403EXPORT_SYMBOL(release_sock);
2404
8a74ad60
ED		 2405/**
2406 * lock_sock_fast - fast version of lock_sock
2407 * @sk: socket
2408 *
2409 * This version should be used for very small section, where process wont block
2410 * return false if fast path is taken
2411 * sk_lock.slock locked, owned = 0, BH disabled
2412 * return true if slow path is taken
2413 * sk_lock.slock unlocked, owned = 1, BH enabled
2414 */
2415bool lock_sock_fast(struct sock *sk)
2416{
2417 might_sleep();
2418 spin_lock_bh(&sk->sk_lock.slock);
2419
2420 if (!sk->sk_lock.owned)
2421 /*
2422 * Note : We must disable BH
2423 */
2424 return false;
2425
2426 __lock_sock(sk);
2427 sk->sk_lock.owned = 1;
2428 spin_unlock(&sk->sk_lock.slock);
2429 /*
2430 * The sk_lock has mutex_lock() semantics here:
2431 */
2432 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2433 local_bh_enable();
2434 return true;
2435}
2436EXPORT_SYMBOL(lock_sock_fast);
2437
1da177e4 2438int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2439{
b7aa0bf7 2440 struct timeval tv;
1da177e4 2441 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2442 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2443 tv = ktime_to_timeval(sk->sk_stamp);
2444 if (tv.tv_sec == -1)
1da177e4 2445 return -ENOENT;
b7aa0bf7
ED		 2446 if (tv.tv_sec == 0) {
2447 sk->sk_stamp = ktime_get_real();
2448 tv = ktime_to_timeval(sk->sk_stamp);
2449 }
2450 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2451}
1da177e4
LT		 2452EXPORT_SYMBOL(sock_get_timestamp);
2453
ae40eb1e
ED		 2454int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2455{
2456 struct timespec ts;
2457 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2458 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2459 ts = ktime_to_timespec(sk->sk_stamp);
2460 if (ts.tv_sec == -1)
2461 return -ENOENT;
2462 if (ts.tv_sec == 0) {
2463 sk->sk_stamp = ktime_get_real();
2464 ts = ktime_to_timespec(sk->sk_stamp);
2465 }
2466 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2467}
2468EXPORT_SYMBOL(sock_get_timestampns);
2469
20d49473 2470void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2471{
20d49473 2472 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2473 unsigned long previous_flags = sk->sk_flags;
2474
20d49473
PO		 2475 sock_set_flag(sk, flag);
2476 /*
2477 * we just set one of the two flags which require net
2478 * time stamping, but time stamping might have been on
2479 * already because of the other one
2480 */
08e29af3 2481 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2482 net_enable_timestamp();
1da177e4
LT		 2483 }
2484}
1da177e4 2485
cb820f8e
RC		 2486int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2487 int level, int type)
2488{
2489 struct sock_exterr_skb *serr;
364a9e93 2490 struct sk_buff *skb;
cb820f8e
RC		 2491 int copied, err;
2492
2493 err = -EAGAIN;
364a9e93 2494 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2495 if (skb == NULL)
2496 goto out;
2497
2498 copied = skb->len;
2499 if (copied > len) {
2500 msg->msg_flags |= MSG_TRUNC;
2501 copied = len;
2502 }
51f3d02b 2503 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2504 if (err)
2505 goto out_free_skb;
2506
2507 sock_recv_timestamp(msg, sk, skb);
2508
2509 serr = SKB_EXT_ERR(skb);
2510 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2511
2512 msg->msg_flags |= MSG_ERRQUEUE;
2513 err = copied;
2514
cb820f8e
RC		 2515out_free_skb:
2516 kfree_skb(skb);
2517out:
2518 return err;
2519}
2520EXPORT_SYMBOL(sock_recv_errqueue);
2521
1da177e4
LT		 2522/*
2523 * Get a socket option on an socket.
2524 *
2525 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2526 * asynchronous errors should be reported by getsockopt. We assume
2527 * this means if you specify SO_ERROR (otherwise whats the point of it).
2528 */
2529int sock_common_getsockopt(struct socket *sock, int level, int optname,
2530 char __user *optval, int __user *optlen)
2531{
2532 struct sock *sk = sock->sk;
2533
2534 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2535}
1da177e4
LT		 2536EXPORT_SYMBOL(sock_common_getsockopt);
2537
3fdadf7d 2538#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2539int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2540 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2541{
2542 struct sock *sk = sock->sk;
2543
1e51f951 2544 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2545 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2546 optval, optlen);
3fdadf7d
DM		 2547 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2548}
2549EXPORT_SYMBOL(compat_sock_common_getsockopt);
2550#endif
2551
1b784140
YX		 2552int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2553 int flags)
1da177e4
LT		 2554{
2555 struct sock *sk = sock->sk;
2556 int addr_len = 0;
2557 int err;
2558
1b784140 2559 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2560 flags & ~MSG_DONTWAIT, &addr_len);
2561 if (err >= 0)
2562 msg->msg_namelen = addr_len;
2563 return err;
2564}
1da177e4
LT		 2565EXPORT_SYMBOL(sock_common_recvmsg);
2566
2567/*
2568 * Set socket options on an inet socket.
2569 */
2570int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2571 char __user *optval, unsigned int optlen)
1da177e4
LT		 2572{
2573 struct sock *sk = sock->sk;
2574
2575 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2576}
1da177e4
LT		 2577EXPORT_SYMBOL(sock_common_setsockopt);
2578
3fdadf7d 2579#ifdef CONFIG_COMPAT
543d9cfe 2580int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2581 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2582{
2583 struct sock *sk = sock->sk;
2584
543d9cfe
ACM		 2585 if (sk->sk_prot->compat_setsockopt != NULL)
2586 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2587 optval, optlen);
3fdadf7d
DM		 2588 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2589}
2590EXPORT_SYMBOL(compat_sock_common_setsockopt);
2591#endif
2592
1da177e4
LT		 2593void sk_common_release(struct sock *sk)
2594{
2595 if (sk->sk_prot->destroy)
2596 sk->sk_prot->destroy(sk);
2597
2598 /*
2599 * Observation: when sock_common_release is called, processes have
2600 * no access to socket. But net still has.
2601 * Step one, detach it from networking:
2602 *
2603 * A. Remove from hash tables.
2604 */
2605
2606 sk->sk_prot->unhash(sk);
2607
2608 /*
2609 * In this point socket cannot receive new packets, but it is possible
2610 * that some packets are in flight because some CPU runs receiver and
2611 * did hash table lookup before we unhashed socket. They will achieve
2612 * receive queue and will be purged by socket destructor.
2613 *
2614 * Also we still have packets pending on receive queue and probably,
2615 * our own packets waiting in device queues. sock_destroy will drain
2616 * receive queue, but transmitted packets will delay socket destruction
2617 * until the last reference will be released.
2618 */
2619
2620 sock_orphan(sk);
2621
2622 xfrm_sk_free_policy(sk);
2623
e6848976 2624 sk_refcnt_debug_release(sk);
5640f768
ED		 2625
2626 if (sk->sk_frag.page) {
2627 put_page(sk->sk_frag.page);
2628 sk->sk_frag.page = NULL;
2629 }
2630
1da177e4
LT		 2631 sock_put(sk);
2632}
1da177e4
LT		 2633EXPORT_SYMBOL(sk_common_release);
2634
13ff3d6f
PE		 2635#ifdef CONFIG_PROC_FS
2636#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2637struct prot_inuse {
2638 int val[PROTO_INUSE_NR];
2639};
13ff3d6f
PE		 2640
2641static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2642
2643#ifdef CONFIG_NET_NS
2644void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2645{
d6d9ca0f 2646 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2647}
2648EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2649
2650int sock_prot_inuse_get(struct net *net, struct proto *prot)
2651{
2652 int cpu, idx = prot->inuse_idx;
2653 int res = 0;
2654
2655 for_each_possible_cpu(cpu)
2656 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2657
2658 return res >= 0 ? res : 0;
2659}
2660EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2661
2c8c1e72 2662static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2663{
2664 net->core.inuse = alloc_percpu(struct prot_inuse);
2665 return net->core.inuse ? 0 : -ENOMEM;
2666}
2667
2c8c1e72 2668static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2669{
2670 free_percpu(net->core.inuse);
2671}
2672
2673static struct pernet_operations net_inuse_ops = {
2674 .init = sock_inuse_init_net,
2675 .exit = sock_inuse_exit_net,
2676};
2677
2678static __init int net_inuse_init(void)
2679{
2680 if (register_pernet_subsys(&net_inuse_ops))
2681 panic("Cannot initialize net inuse counters");
2682
2683 return 0;
2684}
2685
2686core_initcall(net_inuse_init);
2687#else
1338d466
PE		 2688static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2689
c29a0bc4 2690void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2691{
d6d9ca0f 2692 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2693}
2694EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2695
c29a0bc4 2696int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2697{
2698 int cpu, idx = prot->inuse_idx;
2699 int res = 0;
2700
2701 for_each_possible_cpu(cpu)
2702 res += per_cpu(prot_inuse, cpu).val[idx];
2703
2704 return res >= 0 ? res : 0;
2705}
2706EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2707#endif
13ff3d6f
PE		 2708
2709static void assign_proto_idx(struct proto *prot)
2710{
2711 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2712
2713 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2714 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2715 return;
2716 }
2717
2718 set_bit(prot->inuse_idx, proto_inuse_idx);
2719}
2720
2721static void release_proto_idx(struct proto *prot)
2722{
2723 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2724 clear_bit(prot->inuse_idx, proto_inuse_idx);
2725}
2726#else
2727static inline void assign_proto_idx(struct proto *prot)
2728{
2729}
2730
2731static inline void release_proto_idx(struct proto *prot)
2732{
2733}
2734#endif
2735
0159dfd3
ED		 2736static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2737{
2738 if (!rsk_prot)
2739 return;
2740 kfree(rsk_prot->slab_name);
2741 rsk_prot->slab_name = NULL;
2742 if (rsk_prot->slab) {
2743 kmem_cache_destroy(rsk_prot->slab);
2744 rsk_prot->slab = NULL;
2745 }
2746}
2747
2748static int req_prot_init(const struct proto *prot)
2749{
2750 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2751
2752 if (!rsk_prot)
2753 return 0;
2754
2755 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2756 prot->name);
2757 if (!rsk_prot->slab_name)
2758 return -ENOMEM;
2759
2760 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2761 rsk_prot->obj_size, 0,
fa76ce73 2762 0, NULL);
0159dfd3
ED		 2763
2764 if (!rsk_prot->slab) {
2765 pr_crit("%s: Can't create request sock SLAB cache!\n",
2766 prot->name);
2767 return -ENOMEM;
2768 }
2769 return 0;
2770}
2771
b733c007
PE		 2772int proto_register(struct proto *prot, int alloc_slab)
2773{
1da177e4
LT		 2774 if (alloc_slab) {
2775 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2776 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2777 NULL);
1da177e4
LT		 2778
2779 if (prot->slab == NULL) {
e005d193
JP		 2780 pr_crit("%s: Can't create sock SLAB cache!\n",
2781 prot->name);
60e7663d 2782 goto out;
1da177e4 2783 }
2e6599cb 2784
0159dfd3
ED		 2785 if (req_prot_init(prot))
2786 goto out_free_request_sock_slab;
8feaf0c0 2787
6d6ee43e 2788 if (prot->twsk_prot != NULL) {
faf23422 2789 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2790
7e56b5d6 2791 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2792 goto out_free_request_sock_slab;
2793
6d6ee43e 2794 prot->twsk_prot->twsk_slab =
7e56b5d6 2795 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2796 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2797 0,
52db70dc 2798 prot->slab_flags,
20c2df83 2799 NULL);
6d6ee43e 2800 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2801 goto out_free_timewait_sock_slab_name;
2802 }
1da177e4
LT		 2803 }
2804
36b77a52 2805 mutex_lock(&proto_list_mutex);
1da177e4 2806 list_add(&prot->node, &proto_list);
13ff3d6f 2807 assign_proto_idx(prot);
36b77a52 2808 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2809 return 0;
2810
8feaf0c0 2811out_free_timewait_sock_slab_name:
7e56b5d6 2812 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2813out_free_request_sock_slab:
0159dfd3
ED		 2814 req_prot_cleanup(prot->rsk_prot);
2815
2e6599cb
ACM		 2816 kmem_cache_destroy(prot->slab);
2817 prot->slab = NULL;
b733c007
PE		 2818out:
2819 return -ENOBUFS;
1da177e4 2820}
1da177e4
LT		 2821EXPORT_SYMBOL(proto_register);
2822
2823void proto_unregister(struct proto *prot)
2824{
36b77a52 2825 mutex_lock(&proto_list_mutex);
13ff3d6f 2826 release_proto_idx(prot);
0a3f4358 2827 list_del(&prot->node);
36b77a52 2828 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2829
2830 if (prot->slab != NULL) {
2831 kmem_cache_destroy(prot->slab);
2832 prot->slab = NULL;
2833 }
2834
0159dfd3 2835 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2836
6d6ee43e 2837 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2838 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2839 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2840 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2841 }
1da177e4 2842}
1da177e4
LT		 2843EXPORT_SYMBOL(proto_unregister);
2844
2845#ifdef CONFIG_PROC_FS
1da177e4 2846static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2847 __acquires(proto_list_mutex)
1da177e4 2848{
36b77a52 2849 mutex_lock(&proto_list_mutex);
60f0438a 2850 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2851}
2852
2853static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2854{
60f0438a 2855 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2856}
2857
2858static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2859 __releases(proto_list_mutex)
1da177e4 2860{
36b77a52 2861 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2862}
2863
2864static char proto_method_implemented(const void *method)
2865{
2866 return method == NULL ? 'n' : 'y';
2867}
180d8cd9
GC		 2868static long sock_prot_memory_allocated(struct proto *proto)
2869{
cb75a36c 2870 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2871}
2872
2873static char *sock_prot_memory_pressure(struct proto *proto)
2874{
2875 return proto->memory_pressure != NULL ?
2876 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2877}
1da177e4
LT		 2878
2879static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2880{
180d8cd9 2881
8d987e5c 2882 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2883 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2884 proto->name,
2885 proto->obj_size,
14e943db 2886 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2887 sock_prot_memory_allocated(proto),
2888 sock_prot_memory_pressure(proto),
1da177e4
LT		 2889 proto->max_header,
2890 proto->slab == NULL ? "no" : "yes",
2891 module_name(proto->owner),
2892 proto_method_implemented(proto->close),
2893 proto_method_implemented(proto->connect),
2894 proto_method_implemented(proto->disconnect),
2895 proto_method_implemented(proto->accept),
2896 proto_method_implemented(proto->ioctl),
2897 proto_method_implemented(proto->init),
2898 proto_method_implemented(proto->destroy),
2899 proto_method_implemented(proto->shutdown),
2900 proto_method_implemented(proto->setsockopt),
2901 proto_method_implemented(proto->getsockopt),
2902 proto_method_implemented(proto->sendmsg),
2903 proto_method_implemented(proto->recvmsg),
2904 proto_method_implemented(proto->sendpage),
2905 proto_method_implemented(proto->bind),
2906 proto_method_implemented(proto->backlog_rcv),
2907 proto_method_implemented(proto->hash),
2908 proto_method_implemented(proto->unhash),
2909 proto_method_implemented(proto->get_port),
2910 proto_method_implemented(proto->enter_memory_pressure));
2911}
2912
2913static int proto_seq_show(struct seq_file *seq, void *v)
2914{
60f0438a 2915 if (v == &proto_list)
1da177e4
LT		 2916 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2917 "protocol",
2918 "size",
2919 "sockets",
2920 "memory",
2921 "press",
2922 "maxhdr",
2923 "slab",
2924 "module",
2925 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2926 else
60f0438a 2927 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2928 return 0;
2929}
2930
f690808e 2931static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2932 .start = proto_seq_start,
2933 .next = proto_seq_next,
2934 .stop = proto_seq_stop,
2935 .show = proto_seq_show,
2936};
2937
2938static int proto_seq_open(struct inode *inode, struct file *file)
2939{
14e943db
ED		 2940 return seq_open_net(inode, file, &proto_seq_ops,
2941 sizeof(struct seq_net_private));
1da177e4
LT		 2942}
2943
9a32144e 2944static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2945 .owner = THIS_MODULE,
2946 .open = proto_seq_open,
2947 .read = seq_read,
2948 .llseek = seq_lseek,
14e943db
ED		 2949 .release = seq_release_net,
2950};
2951
2952static __net_init int proto_init_net(struct net *net)
2953{
d4beaa66 2954 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2955 return -ENOMEM;
2956
2957 return 0;
2958}
2959
2960static __net_exit void proto_exit_net(struct net *net)
2961{
ece31ffd 2962 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2963}
2964
2965
2966static __net_initdata struct pernet_operations proto_net_ops = {
2967 .init = proto_init_net,
2968 .exit = proto_exit_net,
1da177e4
LT		 2969};
2970
2971static int __init proto_init(void)
2972{
14e943db 2973 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2974}
2975
2976subsys_initcall(proto_init);
2977
2978#endif /* PROC_FS */
Linux 6.x block layer and io_uring tree(s)