]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blame - net/core/sock.c
projects / mirror_ubuntu-hirsute-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
netns: don't allocate an id for dead netns
[mirror_ubuntu-hirsute-kernel.git] / net / core / sock.c
CommitLineData
1da177e4
LT		 1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Generic socket support routines. Memory allocators, socket lock/release
7 * handler for protocols to use and generic option handler.
8 *
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT		 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
14 *
15 * Fixes:
16 * Alan Cox : Numerous verify_area() problems
17 * Alan Cox : Connecting on a connecting socket
18 * now returns an error for tcp.
19 * Alan Cox : sock->protocol is set correctly.
20 * and is not sometimes left as 0.
21 * Alan Cox : connect handles icmp errors on a
22 * connect properly. Unfortunately there
23 * is a restart syscall nasty there. I
24 * can't match BSD without hacking the C
25 * library. Ideas urgently sought!
26 * Alan Cox : Disallow bind() to addresses that are
27 * not ours - especially broadcast ones!!
28 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost)
29 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets,
30 * instead they leave that for the DESTROY timer.
31 * Alan Cox : Clean up error flag in accept
32 * Alan Cox : TCP ack handling is buggy, the DESTROY timer
33 * was buggy. Put a remove_sock() in the handler
34 * for memory when we hit 0. Also altered the timer
4ec93edb 35 * code. The ACK stuff can wait and needs major
1da177e4
LT		 36 * TCP layer surgery.
37 * Alan Cox : Fixed TCP ack bug, removed remove sock
38 * and fixed timer/inet_bh race.
39 * Alan Cox : Added zapped flag for TCP
40 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code
41 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb
42 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources
43 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing.
44 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so...
45 * Rick Sladkey : Relaxed UDP rules for matching packets.
46 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support
47 * Pauline Middelink : identd support
48 * Alan Cox : Fixed connect() taking signals I think.
49 * Alan Cox : SO_LINGER supported
50 * Alan Cox : Error reporting fixes
51 * Anonymous : inet_create tidied up (sk->reuse setting)
52 * Alan Cox : inet sockets don't set sk->type!
53 * Alan Cox : Split socket option code
54 * Alan Cox : Callbacks
55 * Alan Cox : Nagle flag for Charles & Johannes stuff
56 * Alex : Removed restriction on inet fioctl
57 * Alan Cox : Splitting INET from NET core
58 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt()
59 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code
60 * Alan Cox : Split IP from generic code
61 * Alan Cox : New kfree_skbmem()
62 * Alan Cox : Make SO_DEBUG superuser only.
63 * Alan Cox : Allow anyone to clear SO_DEBUG
64 * (compatibility fix)
65 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput.
66 * Alan Cox : Allocator for a socket is settable.
67 * Alan Cox : SO_ERROR includes soft errors.
68 * Alan Cox : Allow NULL arguments on some SO_ opts
69 * Alan Cox : Generic socket allocation to make hooks
70 * easier (suggested by Craig Metz).
71 * Michael Pall : SO_ERROR returns positive errno again
72 * Steve Whitehouse: Added default destructor to free
73 * protocol private data.
74 * Steve Whitehouse: Added various other default routines
75 * common to several socket families.
76 * Chris Evans : Call suser() check last on F_SETOWN
77 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER.
78 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s()
79 * Andi Kleen : Fix write_space callback
80 * Chris Evans : Security fixes - signedness again
81 * Arnaldo C. Melo : cleanups, use skb_queue_purge
82 *
83 * To Fix:
84 *
85 *
86 * This program is free software; you can redistribute it and/or
87 * modify it under the terms of the GNU General Public License
88 * as published by the Free Software Foundation; either version
89 * 2 of the License, or (at your option) any later version.
90 */
91
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>
1da177e4
LT		 134
135#include <linux/filter.h>
136
3847ce32
SM		 137#include <trace/events/sock.h>
138
1da177e4
LT		 139#ifdef CONFIG_INET
140#include <net/tcp.h>
141#endif
142
076bb0c8 143#include <net/busy_poll.h>
06021292 144
36b77a52 145static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 146static LIST_HEAD(proto_list);
147
a3b299da
EB		 148/**
149 * sk_ns_capable - General socket capability test
150 * @sk: Socket to use a capability on or through
151 * @user_ns: The user namespace of the capability to use
152 * @cap: The capability to use
153 *
154 * Test to see if the opener of the socket had when the socket was
155 * created and the current process has the capability @cap in the user
156 * namespace @user_ns.
157 */
158bool sk_ns_capable(const struct sock *sk,
159 struct user_namespace *user_ns, int cap)
160{
161 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
162 ns_capable(user_ns, cap);
163}
164EXPORT_SYMBOL(sk_ns_capable);
165
166/**
167 * sk_capable - Socket global capability test
168 * @sk: Socket to use a capability on or through
e793c0f7 169 * @cap: The global capability to use
a3b299da
EB		 170 *
171 * Test to see if the opener of the socket had when the socket was
172 * created and the current process has the capability @cap in all user
173 * namespaces.
174 */
175bool sk_capable(const struct sock *sk, int cap)
176{
177 return sk_ns_capable(sk, &init_user_ns, cap);
178}
179EXPORT_SYMBOL(sk_capable);
180
181/**
182 * sk_net_capable - Network namespace socket capability test
183 * @sk: Socket to use a capability on or through
184 * @cap: The capability to use
185 *
e793c0f7 186 * Test to see if the opener of the socket had when the socket was created
a3b299da
EB		 187 * and the current process has the capability @cap over the network namespace
188 * the socket is a member of.
189 */
190bool sk_net_capable(const struct sock *sk, int cap)
191{
192 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
193}
194EXPORT_SYMBOL(sk_net_capable);
195
196
c255a458 197#ifdef CONFIG_MEMCG_KMEM
1d62e436 198int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
d1a4c0b3
GC		 199{
200 struct proto *proto;
201 int ret = 0;
202
36b77a52 203 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 204 list_for_each_entry(proto, &proto_list, node) {
205 if (proto->init_cgroup) {
1d62e436 206 ret = proto->init_cgroup(memcg, ss);
d1a4c0b3
GC		 207 if (ret)
208 goto out;
209 }
210 }
211
36b77a52 212 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 213 return ret;
214out:
215 list_for_each_entry_continue_reverse(proto, &proto_list, node)
216 if (proto->destroy_cgroup)
1d62e436 217 proto->destroy_cgroup(memcg);
36b77a52 218 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 219 return ret;
220}
221
1d62e436 222void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
d1a4c0b3
GC		 223{
224 struct proto *proto;
225
36b77a52 226 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 227 list_for_each_entry_reverse(proto, &proto_list, node)
228 if (proto->destroy_cgroup)
1d62e436 229 proto->destroy_cgroup(memcg);
36b77a52 230 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 231}
232#endif
233
da21f24d
IM		 234/*
235 * Each address family might have different locking rules, so we have
236 * one slock key per address family:
237 */
a5b5bb9a
IM		 238static struct lock_class_key af_family_keys[AF_MAX];
239static struct lock_class_key af_family_slock_keys[AF_MAX];
240
cbda4eaf 241#if defined(CONFIG_MEMCG_KMEM)
c5905afb 242struct static_key memcg_socket_limit_enabled;
e1aab161 243EXPORT_SYMBOL(memcg_socket_limit_enabled);
cbda4eaf 244#endif
e1aab161 245
a5b5bb9a
IM		 246/*
247 * Make lock validator output more readable. (we pre-construct these
248 * strings build-time, so that runtime initialization of socket
249 * locks is fast):
250 */
36cbd3dc 251static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 252 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
253 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
254 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
255 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
256 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
257 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
258 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 259 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 260 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 261 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 262 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 263 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 264 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
456db6a4 265 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 266};
36cbd3dc 267static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 268 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
269 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
270 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
271 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
272 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
273 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
274 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 275 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 276 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 277 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 278 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 279 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 280 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 281 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 282};
36cbd3dc 283static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 284 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
285 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
286 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
287 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
288 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
289 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
290 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 291 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 292 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 293 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 294 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 295 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 296 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 297 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 298};
da21f24d
IM		 299
300/*
301 * sk_callback_lock locking rules are per-address-family,
302 * so split the lock classes by using a per-AF key:
303 */
304static struct lock_class_key af_callback_keys[AF_MAX];
305
1da177e4
LT		 306/* Take into consideration the size of the struct sk_buff overhead in the
307 * determination of these values, since that is non-constant across
308 * platforms. This makes socket queueing behavior and performance
309 * not depend upon such differences.
310 */
311#define _SK_MEM_PACKETS 256
87fb4b7b 312#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 313#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
314#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
315
316/* Run time adjustable parameters. */
ab32ea5d 317__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 318EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 319__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 320EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 321__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
322__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 323
25985edc 324/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 325int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 326EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 327
b245be1f
WB		 328int sysctl_tstamp_allow_data __read_mostly = 1;
329
c93bdd0e
MG		 330struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
331EXPORT_SYMBOL_GPL(memalloc_socks);
332
7cb02404
MG		 333/**
334 * sk_set_memalloc - sets %SOCK_MEMALLOC
335 * @sk: socket to set it on
336 *
337 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
338 * It's the responsibility of the admin to adjust min_free_kbytes
339 * to meet the requirements
340 */
341void sk_set_memalloc(struct sock *sk)
342{
343 sock_set_flag(sk, SOCK_MEMALLOC);
344 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 345 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 346}
347EXPORT_SYMBOL_GPL(sk_set_memalloc);
348
349void sk_clear_memalloc(struct sock *sk)
350{
351 sock_reset_flag(sk, SOCK_MEMALLOC);
352 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 353 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 354
355 /*
356 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
357 * progress of swapping. However, if SOCK_MEMALLOC is cleared while
358 * it has rmem allocations there is a risk that the user of the
359 * socket cannot make forward progress due to exceeding the rmem
360 * limits. By rights, sk_clear_memalloc() should only be called
361 * on sockets being torn down but warn and reset the accounting if
362 * that assumption breaks.
363 */
364 if (WARN_ON(sk->sk_forward_alloc))
365 sk_mem_reclaim(sk);
7cb02404
MG		 366}
367EXPORT_SYMBOL_GPL(sk_clear_memalloc);
368
b4b9e355
MG		 369int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
370{
371 int ret;
372 unsigned long pflags = current->flags;
373
374 /* these should have been dropped before queueing */
375 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
376
377 current->flags |= PF_MEMALLOC;
378 ret = sk->sk_backlog_rcv(sk, skb);
379 tsk_restore_flags(current, pflags, PF_MEMALLOC);
380
381 return ret;
382}
383EXPORT_SYMBOL(__sk_backlog_rcv);
384
1da177e4
LT		 385static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
386{
387 struct timeval tv;
388
389 if (optlen < sizeof(tv))
390 return -EINVAL;
391 if (copy_from_user(&tv, optval, sizeof(tv)))
392 return -EFAULT;
ba78073e
VA		 393 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
394 return -EDOM;
1da177e4 395
ba78073e 396 if (tv.tv_sec < 0) {
6f11df83
AM		 397 static int warned __read_mostly;
398
ba78073e 399 *timeo_p = 0;
50aab54f 400 if (warned < 10 && net_ratelimit()) {
ba78073e 401 warned++;
e005d193
JP		 402 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
403 __func__, current->comm, task_pid_nr(current));
50aab54f 404 }
ba78073e
VA		 405 return 0;
406 }
1da177e4
LT		 407 *timeo_p = MAX_SCHEDULE_TIMEOUT;
408 if (tv.tv_sec == 0 && tv.tv_usec == 0)
409 return 0;
410 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
411 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
412 return 0;
413}
414
415static void sock_warn_obsolete_bsdism(const char *name)
416{
417 static int warned;
418 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 419 if (strcmp(warncomm, current->comm) && warned < 5) {
420 strcpy(warncomm, current->comm);
e005d193
JP		 421 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
422 warncomm, name);
1da177e4
LT		 423 warned++;
424 }
425}
426
08e29af3
ED		 427#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
428
429static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 430{
08e29af3
ED		 431 if (sk->sk_flags & flags) {
432 sk->sk_flags &= ~flags;
433 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 434 net_disable_timestamp();
1da177e4
LT		 435 }
436}
437
438
f0088a50
DV		 439int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
440{
766e9037 441 int err;
3b885787
NH		 442 unsigned long flags;
443 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 444
0fd7bac6 445 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 446 atomic_inc(&sk->sk_drops);
3847ce32 447 trace_sock_rcvqueue_full(sk, skb);
766e9037 448 return -ENOMEM;
f0088a50
DV		 449 }
450
fda9ef5d 451 err = sk_filter(sk, skb);
f0088a50 452 if (err)
766e9037 453 return err;
f0088a50 454
c76562b6 455 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 456 atomic_inc(&sk->sk_drops);
457 return -ENOBUFS;
3ab224be
HA		 458 }
459
f0088a50
DV		 460 skb->dev = NULL;
461 skb_set_owner_r(skb, sk);
49ad9599 462
7fee226a
ED		 463 /* we escape from rcu protected region, make sure we dont leak
464 * a norefcounted dst
465 */
466 skb_dst_force(skb);
467
3b885787
NH		 468 spin_lock_irqsave(&list->lock, flags);
469 skb->dropcount = atomic_read(&sk->sk_drops);
470 __skb_queue_tail(list, skb);
471 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 472
473 if (!sock_flag(sk, SOCK_DEAD))
676d2369 474 sk->sk_data_ready(sk);
766e9037 475 return 0;
f0088a50
DV		 476}
477EXPORT_SYMBOL(sock_queue_rcv_skb);
478
58a5a7b9 479int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 480{
481 int rc = NET_RX_SUCCESS;
482
fda9ef5d 483 if (sk_filter(sk, skb))
f0088a50
DV		 484 goto discard_and_relse;
485
486 skb->dev = NULL;
487
274f482d 488 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 489 atomic_inc(&sk->sk_drops);
490 goto discard_and_relse;
491 }
58a5a7b9
ACM		 492 if (nested)
493 bh_lock_sock_nested(sk);
494 else
495 bh_lock_sock(sk);
a5b5bb9a
IM		 496 if (!sock_owned_by_user(sk)) {
497 /*
498 * trylock + unlock semantics:
499 */
500 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
501
c57943a1 502 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 503
504 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 505 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 506 bh_unlock_sock(sk);
507 atomic_inc(&sk->sk_drops);
508 goto discard_and_relse;
509 }
510
f0088a50
DV		 511 bh_unlock_sock(sk);
512out:
513 sock_put(sk);
514 return rc;
515discard_and_relse:
516 kfree_skb(skb);
517 goto out;
518}
519EXPORT_SYMBOL(sk_receive_skb);
520
521struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
522{
b6c6712a 523 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 524
525 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 526 sk_tx_queue_clear(sk);
a9b3cd7f 527 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 528 dst_release(dst);
529 return NULL;
530 }
531
532 return dst;
533}
534EXPORT_SYMBOL(__sk_dst_check);
535
536struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
537{
538 struct dst_entry *dst = sk_dst_get(sk);
539
540 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
541 sk_dst_reset(sk);
542 dst_release(dst);
543 return NULL;
544 }
545
546 return dst;
547}
548EXPORT_SYMBOL(sk_dst_check);
549
c91f6df2
BH		 550static int sock_setbindtodevice(struct sock *sk, char __user *optval,
551 int optlen)
4878809f
DM		 552{
553 int ret = -ENOPROTOOPT;
554#ifdef CONFIG_NETDEVICES
3b1e0a65 555 struct net *net = sock_net(sk);
4878809f
DM		 556 char devname[IFNAMSIZ];
557 int index;
558
559 /* Sorry... */
560 ret = -EPERM;
5e1fccc0 561 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 562 goto out;
563
564 ret = -EINVAL;
565 if (optlen < 0)
566 goto out;
567
568 /* Bind this socket to a particular device like "eth0",
569 * as specified in the passed interface name. If the
570 * name is "" or the option length is zero the socket
571 * is not bound.
572 */
573 if (optlen > IFNAMSIZ - 1)
574 optlen = IFNAMSIZ - 1;
575 memset(devname, 0, sizeof(devname));
576
577 ret = -EFAULT;
578 if (copy_from_user(devname, optval, optlen))
579 goto out;
580
000ba2e4
DM		 581 index = 0;
582 if (devname[0] != '\0') {
bf8e56bf 583 struct net_device *dev;
4878809f 584
bf8e56bf
ED		 585 rcu_read_lock();
586 dev = dev_get_by_name_rcu(net, devname);
587 if (dev)
588 index = dev->ifindex;
589 rcu_read_unlock();
4878809f
DM		 590 ret = -ENODEV;
591 if (!dev)
592 goto out;
4878809f
DM		 593 }
594
595 lock_sock(sk);
596 sk->sk_bound_dev_if = index;
597 sk_dst_reset(sk);
598 release_sock(sk);
599
600 ret = 0;
601
602out:
603#endif
604
605 return ret;
606}
607
c91f6df2
BH		 608static int sock_getbindtodevice(struct sock *sk, char __user *optval,
609 int __user *optlen, int len)
610{
611 int ret = -ENOPROTOOPT;
612#ifdef CONFIG_NETDEVICES
613 struct net *net = sock_net(sk);
c91f6df2 614 char devname[IFNAMSIZ];
c91f6df2
BH		 615
616 if (sk->sk_bound_dev_if == 0) {
617 len = 0;
618 goto zero;
619 }
620
621 ret = -EINVAL;
622 if (len < IFNAMSIZ)
623 goto out;
624
5dbe7c17
NS		 625 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
626 if (ret)
c91f6df2 627 goto out;
c91f6df2
BH		 628
629 len = strlen(devname) + 1;
630
631 ret = -EFAULT;
632 if (copy_to_user(optval, devname, len))
633 goto out;
634
635zero:
636 ret = -EFAULT;
637 if (put_user(len, optlen))
638 goto out;
639
640 ret = 0;
641
642out:
643#endif
644
645 return ret;
646}
647
c0ef877b
PE		 648static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
649{
650 if (valbool)
651 sock_set_flag(sk, bit);
652 else
653 sock_reset_flag(sk, bit);
654}
655
1da177e4
LT		 656/*
657 * This is meant for all protocols to use and covers goings on
658 * at the socket level. Everything here is generic.
659 */
660
661int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 662 char __user *optval, unsigned int optlen)
1da177e4 663{
2a91525c 664 struct sock *sk = sock->sk;
1da177e4
LT		 665 int val;
666 int valbool;
667 struct linger ling;
668 int ret = 0;
4ec93edb 669
1da177e4
LT		 670 /*
671 * Options without arguments
672 */
673
4878809f 674 if (optname == SO_BINDTODEVICE)
c91f6df2 675 return sock_setbindtodevice(sk, optval, optlen);
4878809f 676
e71a4783
SH		 677 if (optlen < sizeof(int))
678 return -EINVAL;
4ec93edb 679
1da177e4
LT		 680 if (get_user(val, (int __user *)optval))
681 return -EFAULT;
4ec93edb 682
2a91525c 683 valbool = val ? 1 : 0;
1da177e4
LT		 684
685 lock_sock(sk);
686
2a91525c 687 switch (optname) {
e71a4783 688 case SO_DEBUG:
2a91525c 689 if (val && !capable(CAP_NET_ADMIN))
e71a4783 690 ret = -EACCES;
2a91525c 691 else
c0ef877b 692 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 693 break;
694 case SO_REUSEADDR:
4a17fd52 695 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 696 break;
055dc21a
TH		 697 case SO_REUSEPORT:
698 sk->sk_reuseport = valbool;
699 break;
e71a4783 700 case SO_TYPE:
49c794e9 701 case SO_PROTOCOL:
0d6038ee 702 case SO_DOMAIN:
e71a4783
SH		 703 case SO_ERROR:
704 ret = -ENOPROTOOPT;
705 break;
706 case SO_DONTROUTE:
c0ef877b 707 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 708 break;
709 case SO_BROADCAST:
710 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
711 break;
712 case SO_SNDBUF:
713 /* Don't error on this BSD doesn't and if you think
82981930
ED		 714 * about it this is right. Otherwise apps have to
715 * play 'guess the biggest size' games. RCVBUF/SNDBUF
716 * are treated in BSD as hints
717 */
718 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 719set_sndbuf:
e71a4783 720 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 721 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
722 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 723 sk->sk_write_space(sk);
724 break;
1da177e4 725
e71a4783
SH		 726 case SO_SNDBUFFORCE:
727 if (!capable(CAP_NET_ADMIN)) {
728 ret = -EPERM;
729 break;
730 }
731 goto set_sndbuf;
b0573dea 732
e71a4783
SH		 733 case SO_RCVBUF:
734 /* Don't error on this BSD doesn't and if you think
82981930
ED		 735 * about it this is right. Otherwise apps have to
736 * play 'guess the biggest size' games. RCVBUF/SNDBUF
737 * are treated in BSD as hints
738 */
739 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 740set_rcvbuf:
e71a4783
SH		 741 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
742 /*
743 * We double it on the way in to account for
744 * "struct sk_buff" etc. overhead. Applications
745 * assume that the SO_RCVBUF setting they make will
746 * allow that much actual data to be received on that
747 * socket.
748 *
749 * Applications are unaware that "struct sk_buff" and
750 * other overheads allocate from the receive buffer
751 * during socket buffer allocation.
752 *
753 * And after considering the possible alternatives,
754 * returning the value we actually used in getsockopt
755 * is the most desirable behavior.
756 */
82981930 757 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 758 break;
759
760 case SO_RCVBUFFORCE:
761 if (!capable(CAP_NET_ADMIN)) {
762 ret = -EPERM;
1da177e4 763 break;
e71a4783
SH		 764 }
765 goto set_rcvbuf;
1da177e4 766
e71a4783 767 case SO_KEEPALIVE:
1da177e4 768#ifdef CONFIG_INET
3e10986d
ED		 769 if (sk->sk_protocol == IPPROTO_TCP &&
770 sk->sk_type == SOCK_STREAM)
e71a4783 771 tcp_set_keepalive(sk, valbool);
1da177e4 772#endif
e71a4783
SH		 773 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
774 break;
775
776 case SO_OOBINLINE:
777 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
778 break;
779
780 case SO_NO_CHECK:
28448b80 781 sk->sk_no_check_tx = valbool;
e71a4783
SH		 782 break;
783
784 case SO_PRIORITY:
5e1fccc0
EB		 785 if ((val >= 0 && val <= 6) ||
786 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 787 sk->sk_priority = val;
788 else
789 ret = -EPERM;
790 break;
791
792 case SO_LINGER:
793 if (optlen < sizeof(ling)) {
794 ret = -EINVAL; /* 1003.1g */
1da177e4 795 break;
e71a4783 796 }
2a91525c 797 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 798 ret = -EFAULT;
1da177e4 799 break;
e71a4783
SH		 800 }
801 if (!ling.l_onoff)
802 sock_reset_flag(sk, SOCK_LINGER);
803 else {
1da177e4 804#if (BITS_PER_LONG == 32)
e71a4783
SH		 805 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
806 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 807 else
e71a4783
SH		 808#endif
809 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
810 sock_set_flag(sk, SOCK_LINGER);
811 }
812 break;
813
814 case SO_BSDCOMPAT:
815 sock_warn_obsolete_bsdism("setsockopt");
816 break;
817
818 case SO_PASSCRED:
819 if (valbool)
820 set_bit(SOCK_PASSCRED, &sock->flags);
821 else
822 clear_bit(SOCK_PASSCRED, &sock->flags);
823 break;
824
825 case SO_TIMESTAMP:
92f37fd2 826 case SO_TIMESTAMPNS:
e71a4783 827 if (valbool) {
92f37fd2
ED		 828 if (optname == SO_TIMESTAMP)
829 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
830 else
831 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 832 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 833 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 834 } else {
e71a4783 835 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 836 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
837 }
e71a4783
SH		 838 break;
839
20d49473
PO		 840 case SO_TIMESTAMPING:
841 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 842 ret = -EINVAL;
20d49473
PO		 843 break;
844 }
b245be1f 845
09c2d251 846 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765
WB		 847 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
848 if (sk->sk_protocol == IPPROTO_TCP) {
849 if (sk->sk_state != TCP_ESTABLISHED) {
850 ret = -EINVAL;
851 break;
852 }
853 sk->sk_tskey = tcp_sk(sk)->snd_una;
854 } else {
855 sk->sk_tskey = 0;
856 }
857 }
b9f40e21 858 sk->sk_tsflags = val;
20d49473
PO		 859 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
860 sock_enable_timestamp(sk,
861 SOCK_TIMESTAMPING_RX_SOFTWARE);
862 else
863 sock_disable_timestamp(sk,
08e29af3 864 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 865 break;
866
e71a4783
SH		 867 case SO_RCVLOWAT:
868 if (val < 0)
869 val = INT_MAX;
870 sk->sk_rcvlowat = val ? : 1;
871 break;
872
873 case SO_RCVTIMEO:
874 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
875 break;
876
877 case SO_SNDTIMEO:
878 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
879 break;
1da177e4 880
e71a4783
SH		 881 case SO_ATTACH_FILTER:
882 ret = -EINVAL;
883 if (optlen == sizeof(struct sock_fprog)) {
884 struct sock_fprog fprog;
1da177e4 885
e71a4783
SH		 886 ret = -EFAULT;
887 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 888 break;
e71a4783
SH		 889
890 ret = sk_attach_filter(&fprog, sk);
891 }
892 break;
893
89aa0758
AS		 894 case SO_ATTACH_BPF:
895 ret = -EINVAL;
896 if (optlen == sizeof(u32)) {
897 u32 ufd;
898
899 ret = -EFAULT;
900 if (copy_from_user(&ufd, optval, sizeof(ufd)))
901 break;
902
903 ret = sk_attach_bpf(ufd, sk);
904 }
905 break;
906
e71a4783 907 case SO_DETACH_FILTER:
55b33325 908 ret = sk_detach_filter(sk);
e71a4783 909 break;
1da177e4 910
d59577b6
VB		 911 case SO_LOCK_FILTER:
912 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
913 ret = -EPERM;
914 else
915 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
916 break;
917
e71a4783
SH		 918 case SO_PASSSEC:
919 if (valbool)
920 set_bit(SOCK_PASSSEC, &sock->flags);
921 else
922 clear_bit(SOCK_PASSSEC, &sock->flags);
923 break;
4a19ec58 924 case SO_MARK:
5e1fccc0 925 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 926 ret = -EPERM;
2a91525c 927 else
4a19ec58 928 sk->sk_mark = val;
4a19ec58 929 break;
877ce7c1 930
1da177e4
LT		 931 /* We implement the SO_SNDLOWAT etc to
932 not be settable (1003.1g 5.3) */
3b885787 933 case SO_RXQ_OVFL:
8083f0fc 934 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 935 break;
6e3e939f
JB		 936
937 case SO_WIFI_STATUS:
938 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
939 break;
940
ef64a54f
PE		 941 case SO_PEEK_OFF:
942 if (sock->ops->set_peek_off)
12663bfc 943 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 944 else
945 ret = -EOPNOTSUPP;
946 break;
3bdc0eba
BG		 947
948 case SO_NOFCS:
949 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
950 break;
951
7d4c04fc
KJ		 952 case SO_SELECT_ERR_QUEUE:
953 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
954 break;
955
e0d1095a 956#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 957 case SO_BUSY_POLL:
dafcc438
ET		 958 /* allow unprivileged users to decrease the value */
959 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
960 ret = -EPERM;
961 else {
962 if (val < 0)
963 ret = -EINVAL;
964 else
965 sk->sk_ll_usec = val;
966 }
967 break;
968#endif
62748f32
ED		 969
970 case SO_MAX_PACING_RATE:
971 sk->sk_max_pacing_rate = val;
972 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
973 sk->sk_max_pacing_rate);
974 break;
975
e71a4783
SH		 976 default:
977 ret = -ENOPROTOOPT;
978 break;
4ec93edb 979 }
1da177e4
LT		 980 release_sock(sk);
981 return ret;
982}
2a91525c 983EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 984
985
8f09898b 986static void cred_to_ucred(struct pid *pid, const struct cred *cred,
987 struct ucred *ucred)
3f551f94
EB		 988{
989 ucred->pid = pid_vnr(pid);
990 ucred->uid = ucred->gid = -1;
991 if (cred) {
992 struct user_namespace *current_ns = current_user_ns();
993
b2e4f544
EB		 994 ucred->uid = from_kuid_munged(current_ns, cred->euid);
995 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 996 }
997}
998
1da177e4
LT		 999int sock_getsockopt(struct socket *sock, int level, int optname,
1000 char __user *optval, int __user *optlen)
1001{
1002 struct sock *sk = sock->sk;
4ec93edb 1003
e71a4783 1004 union {
4ec93edb
YH		 1005 int val;
1006 struct linger ling;
1da177e4
LT		 1007 struct timeval tm;
1008 } v;
4ec93edb 1009
4d0392be 1010 int lv = sizeof(int);
1da177e4 1011 int len;
4ec93edb 1012
e71a4783 1013 if (get_user(len, optlen))
4ec93edb 1014 return -EFAULT;
e71a4783 1015 if (len < 0)
1da177e4 1016 return -EINVAL;
4ec93edb 1017
50fee1de 1018 memset(&v, 0, sizeof(v));
df0bca04 1019
2a91525c 1020 switch (optname) {
e71a4783
SH		 1021 case SO_DEBUG:
1022 v.val = sock_flag(sk, SOCK_DBG);
1023 break;
1024
1025 case SO_DONTROUTE:
1026 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1027 break;
1028
1029 case SO_BROADCAST:
1b23a5df 1030 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1031 break;
1032
1033 case SO_SNDBUF:
1034 v.val = sk->sk_sndbuf;
1035 break;
1036
1037 case SO_RCVBUF:
1038 v.val = sk->sk_rcvbuf;
1039 break;
1040
1041 case SO_REUSEADDR:
1042 v.val = sk->sk_reuse;
1043 break;
1044
055dc21a
TH		 1045 case SO_REUSEPORT:
1046 v.val = sk->sk_reuseport;
1047 break;
1048
e71a4783 1049 case SO_KEEPALIVE:
1b23a5df 1050 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1051 break;
1052
1053 case SO_TYPE:
1054 v.val = sk->sk_type;
1055 break;
1056
49c794e9
JE		 1057 case SO_PROTOCOL:
1058 v.val = sk->sk_protocol;
1059 break;
1060
0d6038ee
JE		 1061 case SO_DOMAIN:
1062 v.val = sk->sk_family;
1063 break;
1064
e71a4783
SH		 1065 case SO_ERROR:
1066 v.val = -sock_error(sk);
2a91525c 1067 if (v.val == 0)
e71a4783
SH		 1068 v.val = xchg(&sk->sk_err_soft, 0);
1069 break;
1070
1071 case SO_OOBINLINE:
1b23a5df 1072 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1073 break;
1074
1075 case SO_NO_CHECK:
28448b80 1076 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1077 break;
1078
1079 case SO_PRIORITY:
1080 v.val = sk->sk_priority;
1081 break;
1082
1083 case SO_LINGER:
1084 lv = sizeof(v.ling);
1b23a5df 1085 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1086 v.ling.l_linger = sk->sk_lingertime / HZ;
1087 break;
1088
1089 case SO_BSDCOMPAT:
1090 sock_warn_obsolete_bsdism("getsockopt");
1091 break;
1092
1093 case SO_TIMESTAMP:
92f37fd2
ED		 1094 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1095 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1096 break;
1097
1098 case SO_TIMESTAMPNS:
1099 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1100 break;
1101
20d49473 1102 case SO_TIMESTAMPING:
b9f40e21 1103 v.val = sk->sk_tsflags;
20d49473
PO		 1104 break;
1105
e71a4783 1106 case SO_RCVTIMEO:
2a91525c 1107 lv = sizeof(struct timeval);
e71a4783
SH		 1108 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1109 v.tm.tv_sec = 0;
1110 v.tm.tv_usec = 0;
1111 } else {
1112 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1113 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1114 }
1115 break;
1116
1117 case SO_SNDTIMEO:
2a91525c 1118 lv = sizeof(struct timeval);
e71a4783
SH		 1119 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1120 v.tm.tv_sec = 0;
1121 v.tm.tv_usec = 0;
1122 } else {
1123 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1124 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1125 }
1126 break;
1da177e4 1127
e71a4783
SH		 1128 case SO_RCVLOWAT:
1129 v.val = sk->sk_rcvlowat;
1130 break;
1da177e4 1131
e71a4783 1132 case SO_SNDLOWAT:
2a91525c 1133 v.val = 1;
e71a4783 1134 break;
1da177e4 1135
e71a4783 1136 case SO_PASSCRED:
82981930 1137 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1138 break;
1da177e4 1139
e71a4783 1140 case SO_PEERCRED:
109f6e39
EB		 1141 {
1142 struct ucred peercred;
1143 if (len > sizeof(peercred))
1144 len = sizeof(peercred);
1145 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1146 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1147 return -EFAULT;
1148 goto lenout;
109f6e39 1149 }
1da177e4 1150
e71a4783
SH		 1151 case SO_PEERNAME:
1152 {
1153 char address[128];
1154
1155 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1156 return -ENOTCONN;
1157 if (lv < len)
1158 return -EINVAL;
1159 if (copy_to_user(optval, address, len))
1160 return -EFAULT;
1161 goto lenout;
1162 }
1da177e4 1163
e71a4783
SH		 1164 /* Dubious BSD thing... Probably nobody even uses it, but
1165 * the UNIX standard wants it for whatever reason... -DaveM
1166 */
1167 case SO_ACCEPTCONN:
1168 v.val = sk->sk_state == TCP_LISTEN;
1169 break;
1da177e4 1170
e71a4783 1171 case SO_PASSSEC:
82981930 1172 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1173 break;
877ce7c1 1174
e71a4783
SH		 1175 case SO_PEERSEC:
1176 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1177
4a19ec58
LAT		 1178 case SO_MARK:
1179 v.val = sk->sk_mark;
1180 break;
1181
3b885787 1182 case SO_RXQ_OVFL:
1b23a5df 1183 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1184 break;
1185
6e3e939f 1186 case SO_WIFI_STATUS:
1b23a5df 1187 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1188 break;
1189
ef64a54f
PE		 1190 case SO_PEEK_OFF:
1191 if (!sock->ops->set_peek_off)
1192 return -EOPNOTSUPP;
1193
1194 v.val = sk->sk_peek_off;
1195 break;
bc2f7996 1196 case SO_NOFCS:
1b23a5df 1197 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1198 break;
c91f6df2 1199
f7b86bfe 1200 case SO_BINDTODEVICE:
c91f6df2
BH		 1201 return sock_getbindtodevice(sk, optval, optlen, len);
1202
a8fc9277
PE		 1203 case SO_GET_FILTER:
1204 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1205 if (len < 0)
1206 return len;
1207
1208 goto lenout;
c91f6df2 1209
d59577b6
VB		 1210 case SO_LOCK_FILTER:
1211 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1212 break;
1213
ea02f941
MS		 1214 case SO_BPF_EXTENSIONS:
1215 v.val = bpf_tell_extensions();
1216 break;
1217
7d4c04fc
KJ		 1218 case SO_SELECT_ERR_QUEUE:
1219 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1220 break;
1221
e0d1095a 1222#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1223 case SO_BUSY_POLL:
dafcc438
ET		 1224 v.val = sk->sk_ll_usec;
1225 break;
1226#endif
1227
62748f32
ED		 1228 case SO_MAX_PACING_RATE:
1229 v.val = sk->sk_max_pacing_rate;
1230 break;
1231
2c8c56e1
ED		 1232 case SO_INCOMING_CPU:
1233 v.val = sk->sk_incoming_cpu;
1234 break;
1235
e71a4783
SH		 1236 default:
1237 return -ENOPROTOOPT;
1da177e4 1238 }
e71a4783 1239
1da177e4
LT		 1240 if (len > lv)
1241 len = lv;
1242 if (copy_to_user(optval, &v, len))
1243 return -EFAULT;
1244lenout:
4ec93edb
YH		 1245 if (put_user(len, optlen))
1246 return -EFAULT;
1247 return 0;
1da177e4
LT		 1248}
1249
a5b5bb9a
IM		 1250/*
1251 * Initialize an sk_lock.
1252 *
1253 * (We also register the sk_lock with the lock validator.)
1254 */
b6f99a21 1255static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1256{
ed07536e
PZ		 1257 sock_lock_init_class_and_name(sk,
1258 af_family_slock_key_strings[sk->sk_family],
1259 af_family_slock_keys + sk->sk_family,
1260 af_family_key_strings[sk->sk_family],
1261 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1262}
1263
4dc6dc71
ED		 1264/*
1265 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1266 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1267 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1268 */
f1a6c4da
PE		 1269static void sock_copy(struct sock *nsk, const struct sock *osk)
1270{
1271#ifdef CONFIG_SECURITY_NETWORK
1272 void *sptr = nsk->sk_security;
1273#endif
68835aba
ED		 1274 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1275
1276 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1277 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1278
f1a6c4da
PE		 1279#ifdef CONFIG_SECURITY_NETWORK
1280 nsk->sk_security = sptr;
1281 security_sk_clone(osk, nsk);
1282#endif
1283}
1284
fcbdf09d
OP		 1285void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1286{
1287 unsigned long nulls1, nulls2;
1288
1289 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1290 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1291 if (nulls1 > nulls2)
1292 swap(nulls1, nulls2);
1293
1294 if (nulls1 != 0)
1295 memset((char *)sk, 0, nulls1);
1296 memset((char *)sk + nulls1 + sizeof(void *), 0,
1297 nulls2 - nulls1 - sizeof(void *));
1298 memset((char *)sk + nulls2 + sizeof(void *), 0,
1299 size - nulls2 - sizeof(void *));
1300}
1301EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1302
2e4afe7b
PE		 1303static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1304 int family)
c308c1b2
PE		 1305{
1306 struct sock *sk;
1307 struct kmem_cache *slab;
1308
1309 slab = prot->slab;
e912b114
ED		 1310 if (slab != NULL) {
1311 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1312 if (!sk)
1313 return sk;
1314 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1315 if (prot->clear_sk)
1316 prot->clear_sk(sk, prot->obj_size);
1317 else
1318 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1319 }
fcbdf09d 1320 } else
c308c1b2
PE		 1321 sk = kmalloc(prot->obj_size, priority);
1322
2e4afe7b 1323 if (sk != NULL) {
a98b65a3
VN		 1324 kmemcheck_annotate_bitfield(sk, flags);
1325
2e4afe7b
PE		 1326 if (security_sk_alloc(sk, family, priority))
1327 goto out_free;
1328
1329 if (!try_module_get(prot->owner))
1330 goto out_free_sec;
e022f0b4 1331 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1332 }
1333
c308c1b2 1334 return sk;
2e4afe7b
PE		 1335
1336out_free_sec:
1337 security_sk_free(sk);
1338out_free:
1339 if (slab != NULL)
1340 kmem_cache_free(slab, sk);
1341 else
1342 kfree(sk);
1343 return NULL;
c308c1b2
PE		 1344}
1345
1346static void sk_prot_free(struct proto *prot, struct sock *sk)
1347{
1348 struct kmem_cache *slab;
2e4afe7b 1349 struct module *owner;
c308c1b2 1350
2e4afe7b 1351 owner = prot->owner;
c308c1b2 1352 slab = prot->slab;
2e4afe7b
PE		 1353
1354 security_sk_free(sk);
c308c1b2
PE		 1355 if (slab != NULL)
1356 kmem_cache_free(slab, sk);
1357 else
1358 kfree(sk);
2e4afe7b 1359 module_put(owner);
c308c1b2
PE		 1360}
1361
86f8515f 1362#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1363void sock_update_netprioidx(struct sock *sk)
5bc1421e 1364{
5bc1421e
NH		 1365 if (in_interrupt())
1366 return;
2b73bc65 1367
6ffd4641 1368 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1369}
1370EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1371#endif
1372
1da177e4
LT		 1373/**
1374 * sk_alloc - All socket objects are allocated here
c4ea43c5 1375 * @net: the applicable net namespace
4dc3b16b
PP		 1376 * @family: protocol family
1377 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1378 * @prot: struct proto associated with this new sock instance
1da177e4 1379 */
1b8d7ae4 1380struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1381 struct proto *prot)
1da177e4 1382{
c308c1b2 1383 struct sock *sk;
1da177e4 1384
154adbc8 1385 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1386 if (sk) {
154adbc8
PE		 1387 sk->sk_family = family;
1388 /*
1389 * See comment in struct sock definition to understand
1390 * why we need sk_prot_creator -acme
1391 */
1392 sk->sk_prot = sk->sk_prot_creator = prot;
1393 sock_lock_init(sk);
3b1e0a65 1394 sock_net_set(sk, get_net(net));
d66ee058 1395 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1396
211d2f97 1397 sock_update_classid(sk);
6ffd4641 1398 sock_update_netprioidx(sk);
1da177e4 1399 }
a79af59e 1400
2e4afe7b 1401 return sk;
1da177e4 1402}
2a91525c 1403EXPORT_SYMBOL(sk_alloc);
1da177e4 1404
2b85a34e 1405static void __sk_free(struct sock *sk)
1da177e4
LT		 1406{
1407 struct sk_filter *filter;
1da177e4
LT		 1408
1409 if (sk->sk_destruct)
1410 sk->sk_destruct(sk);
1411
a898def2
PM		 1412 filter = rcu_dereference_check(sk->sk_filter,
1413 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1414 if (filter) {
309dd5fc 1415 sk_filter_uncharge(sk, filter);
a9b3cd7f 1416 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1417 }
1418
08e29af3 1419 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1420
1421 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1422 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1423 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1424
109f6e39
EB		 1425 if (sk->sk_peer_cred)
1426 put_cred(sk->sk_peer_cred);
1427 put_pid(sk->sk_peer_pid);
3b1e0a65 1428 put_net(sock_net(sk));
c308c1b2 1429 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1430}
2b85a34e
ED		 1431
1432void sk_free(struct sock *sk)
1433{
1434 /*
25985edc 1435 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1436 * some packets are still in some tx queue.
1437 * If not null, sock_wfree() will call __sk_free(sk) later
1438 */
1439 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1440 __sk_free(sk);
1441}
2a91525c 1442EXPORT_SYMBOL(sk_free);
1da177e4 1443
edf02087 1444/*
25985edc
LDM		 1445 * Last sock_put should drop reference to sk->sk_net. It has already
1446 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1447 * is not an option.
25985edc 1448 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1449 * destroy it in the context of init_net.
1450 */
1451void sk_release_kernel(struct sock *sk)
1452{
1453 if (sk == NULL || sk->sk_socket == NULL)
1454 return;
1455
1456 sock_hold(sk);
1457 sock_release(sk->sk_socket);
65a18ec5 1458 release_net(sock_net(sk));
3b1e0a65 1459 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1460 sock_put(sk);
1461}
45af1754 1462EXPORT_SYMBOL(sk_release_kernel);
edf02087 1463
475f1b52
SR		 1464static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1465{
1466 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1467 sock_update_memcg(newsk);
1468}
1469
e56c57d0
ED		 1470/**
1471 * sk_clone_lock - clone a socket, and lock its clone
1472 * @sk: the socket to clone
1473 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1474 *
1475 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1476 */
1477struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1478{
8fd1d178 1479 struct sock *newsk;
278571ba 1480 bool is_charged = true;
87d11ceb 1481
8fd1d178 1482 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1483 if (newsk != NULL) {
1484 struct sk_filter *filter;
1485
892c141e 1486 sock_copy(newsk, sk);
87d11ceb
ACM		 1487
1488 /* SANITY */
3b1e0a65 1489 get_net(sock_net(newsk));
87d11ceb
ACM		 1490 sk_node_init(&newsk->sk_node);
1491 sock_lock_init(newsk);
1492 bh_lock_sock(newsk);
fa438ccf 1493 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1494 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1495
1496 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1497 /*
1498 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1499 */
1500 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1501 atomic_set(&newsk->sk_omem_alloc, 0);
1502 skb_queue_head_init(&newsk->sk_receive_queue);
1503 skb_queue_head_init(&newsk->sk_write_queue);
1504
b6c6712a 1505 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1506 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1507 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1508 af_callback_keys + newsk->sk_family,
1509 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1510
1511 newsk->sk_dst_cache = NULL;
1512 newsk->sk_wmem_queued = 0;
1513 newsk->sk_forward_alloc = 0;
1514 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1515 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1516
1517 sock_reset_flag(newsk, SOCK_DONE);
1518 skb_queue_head_init(&newsk->sk_error_queue);
1519
0d7da9dd 1520 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1521 if (filter != NULL)
278571ba
AS		 1522 /* though it's an empty new sock, the charging may fail
1523 * if sysctl_optmem_max was changed between creation of
1524 * original socket and cloning
1525 */
1526 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1527
278571ba 1528 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk))) {
87d11ceb
ACM		 1529 /* It is still raw copy of parent, so invalidate
1530 * destructor and make plain sk_free() */
1531 newsk->sk_destruct = NULL;
b0691c8e 1532 bh_unlock_sock(newsk);
87d11ceb
ACM		 1533 sk_free(newsk);
1534 newsk = NULL;
1535 goto out;
1536 }
1537
1538 newsk->sk_err = 0;
1539 newsk->sk_priority = 0;
2c8c56e1 1540 newsk->sk_incoming_cpu = raw_smp_processor_id();
4dc6dc71
ED		 1541 /*
1542 * Before updating sk_refcnt, we must commit prior changes to memory
1543 * (Documentation/RCU/rculist_nulls.txt for details)
1544 */
1545 smp_wmb();
87d11ceb
ACM		 1546 atomic_set(&newsk->sk_refcnt, 2);
1547
1548 /*
1549 * Increment the counter in the same struct proto as the master
1550 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1551 * is the same as sk->sk_prot->socks, as this field was copied
1552 * with memcpy).
1553 *
1554 * This _changes_ the previous behaviour, where
1555 * tcp_create_openreq_child always was incrementing the
1556 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1557 * to be taken into account in all callers. -acme
1558 */
1559 sk_refcnt_debug_inc(newsk);
972692e0 1560 sk_set_socket(newsk, NULL);
43815482 1561 newsk->sk_wq = NULL;
87d11ceb 1562
f3f511e1
GC		 1563 sk_update_clone(sk, newsk);
1564
87d11ceb 1565 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1566 sk_sockets_allocated_inc(newsk);
704da560 1567
08e29af3 1568 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1569 net_enable_timestamp();
87d11ceb
ACM		 1570 }
1571out:
1572 return newsk;
1573}
e56c57d0 1574EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1575
9958089a
AK		 1576void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1577{
1578 __sk_dst_set(sk, dst);
1579 sk->sk_route_caps = dst->dev->features;
1580 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1581 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1582 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1583 if (sk_can_gso(sk)) {
82cc1a7a 1584 if (dst->header_len) {
9958089a 1585 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1586 } else {
9958089a 1587 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1588 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1589 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1590 }
9958089a
AK		 1591 }
1592}
1593EXPORT_SYMBOL_GPL(sk_setup_caps);
1594
1da177e4
LT		 1595/*
1596 * Simple resource managers for sockets.
1597 */
1598
1599
4ec93edb
YH		 1600/*
1601 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1602 */
1603void sock_wfree(struct sk_buff *skb)
1604{
1605 struct sock *sk = skb->sk;
d99927f4 1606 unsigned int len = skb->truesize;
1da177e4 1607
d99927f4
ED		 1608 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1609 /*
1610 * Keep a reference on sk_wmem_alloc, this will be released
1611 * after sk_write_space() call
1612 */
1613 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1614 sk->sk_write_space(sk);
d99927f4
ED		 1615 len = 1;
1616 }
2b85a34e 1617 /*
d99927f4
ED		 1618 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1619 * could not do because of in-flight packets
2b85a34e 1620 */
d99927f4 1621 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1622 __sk_free(sk);
1da177e4 1623}
2a91525c 1624EXPORT_SYMBOL(sock_wfree);
1da177e4 1625
f2f872f9
ED		 1626void skb_orphan_partial(struct sk_buff *skb)
1627{
1628 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1629 * so we do not completely orphan skb, but transfert all
1630 * accounted bytes but one, to avoid unexpected reorders.
1631 */
1632 if (skb->destructor == sock_wfree
1633#ifdef CONFIG_INET
1634 || skb->destructor == tcp_wfree
1635#endif
1636 ) {
1637 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1638 skb->truesize = 1;
1639 } else {
1640 skb_orphan(skb);
1641 }
1642}
1643EXPORT_SYMBOL(skb_orphan_partial);
1644
4ec93edb
YH		 1645/*
1646 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1647 */
1648void sock_rfree(struct sk_buff *skb)
1649{
1650 struct sock *sk = skb->sk;
d361fd59 1651 unsigned int len = skb->truesize;
1da177e4 1652
d361fd59
ED		 1653 atomic_sub(len, &sk->sk_rmem_alloc);
1654 sk_mem_uncharge(sk, len);
1da177e4 1655}
2a91525c 1656EXPORT_SYMBOL(sock_rfree);
1da177e4 1657
7768eed8
OH		 1658/*
1659 * Buffer destructor for skbs that are not used directly in read or write
1660 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1661 */
62bccb8c
AD		 1662void sock_efree(struct sk_buff *skb)
1663{
1664 sock_put(skb->sk);
1665}
1666EXPORT_SYMBOL(sock_efree);
1667
82eabd9e 1668#ifdef CONFIG_INET
41063e9d
DM		 1669void sock_edemux(struct sk_buff *skb)
1670{
e812347c
ED		 1671 struct sock *sk = skb->sk;
1672
1673 if (sk->sk_state == TCP_TIME_WAIT)
1674 inet_twsk_put(inet_twsk(sk));
1675 else
1676 sock_put(sk);
41063e9d
DM		 1677}
1678EXPORT_SYMBOL(sock_edemux);
82eabd9e 1679#endif
1da177e4 1680
976d0201 1681kuid_t sock_i_uid(struct sock *sk)
1da177e4 1682{
976d0201 1683 kuid_t uid;
1da177e4 1684
f064af1e 1685 read_lock_bh(&sk->sk_callback_lock);
976d0201 1686 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1687 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1688 return uid;
1689}
2a91525c 1690EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1691
1692unsigned long sock_i_ino(struct sock *sk)
1693{
1694 unsigned long ino;
1695
f064af1e 1696 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1697 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1698 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1699 return ino;
1700}
2a91525c 1701EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1702
1703/*
1704 * Allocate a skb from the socket's send buffer.
1705 */
86a76caf 1706struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1707 gfp_t priority)
1da177e4
LT		 1708{
1709 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1710 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1711 if (skb) {
1712 skb_set_owner_w(skb, sk);
1713 return skb;
1714 }
1715 }
1716 return NULL;
1717}
2a91525c 1718EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1719
4ec93edb 1720/*
1da177e4 1721 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1722 */
dd0fc66f 1723void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1724{
95c96174 1725 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1726 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1727 void *mem;
1728 /* First do the add, to avoid the race if kmalloc
4ec93edb 1729 * might sleep.
1da177e4
LT		 1730 */
1731 atomic_add(size, &sk->sk_omem_alloc);
1732 mem = kmalloc(size, priority);
1733 if (mem)
1734 return mem;
1735 atomic_sub(size, &sk->sk_omem_alloc);
1736 }
1737 return NULL;
1738}
2a91525c 1739EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1740
79e88659
DB		 1741/* Free an option memory block. Note, we actually want the inline
1742 * here as this allows gcc to detect the nullify and fold away the
1743 * condition entirely.
1da177e4 1744 */
79e88659
DB		 1745static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1746 const bool nullify)
1da177e4 1747{
e53da5fb
DM		 1748 if (WARN_ON_ONCE(!mem))
1749 return;
79e88659
DB		 1750 if (nullify)
1751 kzfree(mem);
1752 else
1753 kfree(mem);
1da177e4
LT		 1754 atomic_sub(size, &sk->sk_omem_alloc);
1755}
79e88659
DB		 1756
1757void sock_kfree_s(struct sock *sk, void *mem, int size)
1758{
1759 __sock_kfree_s(sk, mem, size, false);
1760}
2a91525c 1761EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1762
79e88659
DB		 1763void sock_kzfree_s(struct sock *sk, void *mem, int size)
1764{
1765 __sock_kfree_s(sk, mem, size, true);
1766}
1767EXPORT_SYMBOL(sock_kzfree_s);
1768
1da177e4
LT		 1769/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1770 I think, these locks should be removed for datagram sockets.
1771 */
2a91525c 1772static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1773{
1774 DEFINE_WAIT(wait);
1775
1776 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1777 for (;;) {
1778 if (!timeo)
1779 break;
1780 if (signal_pending(current))
1781 break;
1782 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1783 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1784 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1785 break;
1786 if (sk->sk_shutdown & SEND_SHUTDOWN)
1787 break;
1788 if (sk->sk_err)
1789 break;
1790 timeo = schedule_timeout(timeo);
1791 }
aa395145 1792 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1793 return timeo;
1794}
1795
1796
1797/*
1798 * Generic send/receive buffer handlers
1799 */
1800
4cc7f68d
HX		 1801struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1802 unsigned long data_len, int noblock,
28d64271 1803 int *errcode, int max_page_order)
1da177e4 1804{
2e4e4410 1805 struct sk_buff *skb;
1da177e4
LT		 1806 long timeo;
1807 int err;
1808
1da177e4 1809 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1810 for (;;) {
1da177e4
LT		 1811 err = sock_error(sk);
1812 if (err != 0)
1813 goto failure;
1814
1815 err = -EPIPE;
1816 if (sk->sk_shutdown & SEND_SHUTDOWN)
1817 goto failure;
1818
2e4e4410
ED		 1819 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1820 break;
28d64271 1821
2e4e4410
ED		 1822 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1823 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1824 err = -EAGAIN;
1825 if (!timeo)
1da177e4 1826 goto failure;
2e4e4410
ED		 1827 if (signal_pending(current))
1828 goto interrupted;
1829 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1830 }
2e4e4410
ED		 1831 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1832 errcode, sk->sk_allocation);
1833 if (skb)
1834 skb_set_owner_w(skb, sk);
1da177e4
LT		 1835 return skb;
1836
1837interrupted:
1838 err = sock_intr_errno(timeo);
1839failure:
1840 *errcode = err;
1841 return NULL;
1842}
4cc7f68d 1843EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1844
4ec93edb 1845struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1846 int noblock, int *errcode)
1847{
28d64271 1848 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1849}
2a91525c 1850EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1851
5640f768
ED		 1852/* On 32bit arches, an skb frag is limited to 2^15 */
1853#define SKB_FRAG_PAGE_ORDER get_order(32768)
1854
400dfd3a
ED		 1855/**
1856 * skb_page_frag_refill - check that a page_frag contains enough room
1857 * @sz: minimum size of the fragment we want to get
1858 * @pfrag: pointer to page_frag
82d5e2b8 1859 * @gfp: priority for memory allocation
400dfd3a
ED		 1860 *
1861 * Note: While this allocator tries to use high order pages, there is
1862 * no guarantee that allocations succeed. Therefore, @sz MUST be
1863 * less or equal than PAGE_SIZE.
1864 */
d9b2938a 1865bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1866{
5640f768
ED		 1867 if (pfrag->page) {
1868 if (atomic_read(&pfrag->page->_count) == 1) {
1869 pfrag->offset = 0;
1870 return true;
1871 }
400dfd3a 1872 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1873 return true;
1874 put_page(pfrag->page);
1875 }
1876
d9b2938a
ED		 1877 pfrag->offset = 0;
1878 if (SKB_FRAG_PAGE_ORDER) {
1879 pfrag->page = alloc_pages(gfp | __GFP_COMP |
1880 __GFP_NOWARN | __GFP_NORETRY,
1881 SKB_FRAG_PAGE_ORDER);
5640f768 1882 if (likely(pfrag->page)) {
d9b2938a 1883 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1884 return true;
1885 }
d9b2938a
ED		 1886 }
1887 pfrag->page = alloc_page(gfp);
1888 if (likely(pfrag->page)) {
1889 pfrag->size = PAGE_SIZE;
1890 return true;
1891 }
400dfd3a
ED		 1892 return false;
1893}
1894EXPORT_SYMBOL(skb_page_frag_refill);
1895
1896bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1897{
1898 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1899 return true;
1900
5640f768
ED		 1901 sk_enter_memory_pressure(sk);
1902 sk_stream_moderate_sndbuf(sk);
1903 return false;
1904}
1905EXPORT_SYMBOL(sk_page_frag_refill);
1906
1da177e4 1907static void __lock_sock(struct sock *sk)
f39234d6
NK		 1908 __releases(&sk->sk_lock.slock)
1909 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1910{
1911 DEFINE_WAIT(wait);
1912
e71a4783 1913 for (;;) {
1da177e4
LT		 1914 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1915 TASK_UNINTERRUPTIBLE);
1916 spin_unlock_bh(&sk->sk_lock.slock);
1917 schedule();
1918 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1919 if (!sock_owned_by_user(sk))
1da177e4
LT		 1920 break;
1921 }
1922 finish_wait(&sk->sk_lock.wq, &wait);
1923}
1924
1925static void __release_sock(struct sock *sk)
f39234d6
NK		 1926 __releases(&sk->sk_lock.slock)
1927 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1928{
1929 struct sk_buff *skb = sk->sk_backlog.head;
1930
1931 do {
1932 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1933 bh_unlock_sock(sk);
1934
1935 do {
1936 struct sk_buff *next = skb->next;
1937
e4cbb02a 1938 prefetch(next);
7fee226a 1939 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1940 skb->next = NULL;
c57943a1 1941 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1942
1943 /*
1944 * We are in process context here with softirqs
1945 * disabled, use cond_resched_softirq() to preempt.
1946 * This is safe to do because we've taken the backlog
1947 * queue private:
1948 */
1949 cond_resched_softirq();
1950
1951 skb = next;
1952 } while (skb != NULL);
1953
1954 bh_lock_sock(sk);
e71a4783 1955 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1956
1957 /*
1958 * Doing the zeroing here guarantee we can not loop forever
1959 * while a wild producer attempts to flood us.
1960 */
1961 sk->sk_backlog.len = 0;
1da177e4
LT		 1962}
1963
1964/**
1965 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1966 * @sk: sock to wait on
1967 * @timeo: for how long
1da177e4
LT		 1968 *
1969 * Now socket state including sk->sk_err is changed only under lock,
1970 * hence we may omit checks after joining wait queue.
1971 * We check receive queue before schedule() only as optimization;
1972 * it is very likely that release_sock() added new data.
1973 */
1974int sk_wait_data(struct sock *sk, long *timeo)
1975{
1976 int rc;
1977 DEFINE_WAIT(wait);
1978
aa395145 1979 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1980 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1981 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1982 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1983 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1984 return rc;
1985}
1da177e4
LT		 1986EXPORT_SYMBOL(sk_wait_data);
1987
3ab224be
HA		 1988/**
1989 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1990 * @sk: socket
1991 * @size: memory size to allocate
1992 * @kind: allocation type
1993 *
1994 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1995 * rmem allocation. This function assumes that protocols which have
1996 * memory_pressure use sk_wmem_queued as write buffer accounting.
1997 */
1998int __sk_mem_schedule(struct sock *sk, int size, int kind)
1999{
2000 struct proto *prot = sk->sk_prot;
2001 int amt = sk_mem_pages(size);
8d987e5c 2002 long allocated;
e1aab161 2003 int parent_status = UNDER_LIMIT;
3ab224be
HA		 2004
2005 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2006
e1aab161 2007 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 2008
2009 /* Under limit. */
e1aab161
GC		 2010 if (parent_status == UNDER_LIMIT &&
2011 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2012 sk_leave_memory_pressure(sk);
3ab224be
HA		 2013 return 1;
2014 }
2015
e1aab161
GC		 2016 /* Under pressure. (we or our parents) */
2017 if ((parent_status > SOFT_LIMIT) ||
2018 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2019 sk_enter_memory_pressure(sk);
3ab224be 2020
e1aab161
GC		 2021 /* Over hard limit (we or our parents) */
2022 if ((parent_status == OVER_LIMIT) ||
2023 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 2024 goto suppress_allocation;
2025
2026 /* guarantee minimum buffer size under pressure */
2027 if (kind == SK_MEM_RECV) {
2028 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2029 return 1;
180d8cd9 2030
3ab224be
HA		 2031 } else { /* SK_MEM_SEND */
2032 if (sk->sk_type == SOCK_STREAM) {
2033 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2034 return 1;
2035 } else if (atomic_read(&sk->sk_wmem_alloc) <
2036 prot->sysctl_wmem[0])
2037 return 1;
2038 }
2039
180d8cd9 2040 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2041 int alloc;
2042
180d8cd9 2043 if (!sk_under_memory_pressure(sk))
1748376b 2044 return 1;
180d8cd9
GC		 2045 alloc = sk_sockets_allocated_read_positive(sk);
2046 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2047 sk_mem_pages(sk->sk_wmem_queued +
2048 atomic_read(&sk->sk_rmem_alloc) +
2049 sk->sk_forward_alloc))
2050 return 1;
2051 }
2052
2053suppress_allocation:
2054
2055 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2056 sk_stream_moderate_sndbuf(sk);
2057
2058 /* Fail only if socket is _under_ its sndbuf.
2059 * In this case we cannot block, so that we have to fail.
2060 */
2061 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2062 return 1;
2063 }
2064
3847ce32
SM		 2065 trace_sock_exceed_buf_limit(sk, prot, allocated);
2066
3ab224be
HA		 2067 /* Alas. Undo changes. */
2068 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2069
0e90b31f 2070 sk_memory_allocated_sub(sk, amt);
180d8cd9 2071
3ab224be
HA		 2072 return 0;
2073}
3ab224be
HA		 2074EXPORT_SYMBOL(__sk_mem_schedule);
2075
2076/**
2077 * __sk_reclaim - reclaim memory_allocated
2078 * @sk: socket
2079 */
2080void __sk_mem_reclaim(struct sock *sk)
2081{
180d8cd9 2082 sk_memory_allocated_sub(sk,
0e90b31f 2083 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 2084 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
2085
180d8cd9
GC		 2086 if (sk_under_memory_pressure(sk) &&
2087 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2088 sk_leave_memory_pressure(sk);
3ab224be 2089}
3ab224be
HA		 2090EXPORT_SYMBOL(__sk_mem_reclaim);
2091
2092
1da177e4
LT		 2093/*
2094 * Set of default routines for initialising struct proto_ops when
2095 * the protocol does not support a particular function. In certain
2096 * cases where it makes no sense for a protocol to have a "do nothing"
2097 * function, some default processing is provided.
2098 */
2099
2100int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2101{
2102 return -EOPNOTSUPP;
2103}
2a91525c 2104EXPORT_SYMBOL(sock_no_bind);
1da177e4 2105
4ec93edb 2106int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2107 int len, int flags)
2108{
2109 return -EOPNOTSUPP;
2110}
2a91525c 2111EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2112
2113int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2114{
2115 return -EOPNOTSUPP;
2116}
2a91525c 2117EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2118
2119int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2120{
2121 return -EOPNOTSUPP;
2122}
2a91525c 2123EXPORT_SYMBOL(sock_no_accept);
1da177e4 2124
4ec93edb 2125int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2126 int *len, int peer)
2127{
2128 return -EOPNOTSUPP;
2129}
2a91525c 2130EXPORT_SYMBOL(sock_no_getname);
1da177e4 2131
2a91525c 2132unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2133{
2134 return 0;
2135}
2a91525c 2136EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2137
2138int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2139{
2140 return -EOPNOTSUPP;
2141}
2a91525c 2142EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2143
2144int sock_no_listen(struct socket *sock, int backlog)
2145{
2146 return -EOPNOTSUPP;
2147}
2a91525c 2148EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2149
2150int sock_no_shutdown(struct socket *sock, int how)
2151{
2152 return -EOPNOTSUPP;
2153}
2a91525c 2154EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2155
2156int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2157 char __user *optval, unsigned int optlen)
1da177e4
LT		 2158{
2159 return -EOPNOTSUPP;
2160}
2a91525c 2161EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2162
2163int sock_no_getsockopt(struct socket *sock, int level, int optname,
2164 char __user *optval, int __user *optlen)
2165{
2166 return -EOPNOTSUPP;
2167}
2a91525c 2168EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2169
2170int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2171 size_t len)
2172{
2173 return -EOPNOTSUPP;
2174}
2a91525c 2175EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2176
2177int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2178 size_t len, int flags)
2179{
2180 return -EOPNOTSUPP;
2181}
2a91525c 2182EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2183
2184int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2185{
2186 /* Mirror missing mmap method error code */
2187 return -ENODEV;
2188}
2a91525c 2189EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2190
2191ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2192{
2193 ssize_t res;
2194 struct msghdr msg = {.msg_flags = flags};
2195 struct kvec iov;
2196 char *kaddr = kmap(page);
2197 iov.iov_base = kaddr + offset;
2198 iov.iov_len = size;
2199 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2200 kunmap(page);
2201 return res;
2202}
2a91525c 2203EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2204
2205/*
2206 * Default Socket Callbacks
2207 */
2208
2209static void sock_def_wakeup(struct sock *sk)
2210{
43815482
ED		 2211 struct socket_wq *wq;
2212
2213 rcu_read_lock();
2214 wq = rcu_dereference(sk->sk_wq);
2215 if (wq_has_sleeper(wq))
2216 wake_up_interruptible_all(&wq->wait);
2217 rcu_read_unlock();
1da177e4
LT		 2218}
2219
2220static void sock_def_error_report(struct sock *sk)
2221{
43815482
ED		 2222 struct socket_wq *wq;
2223
2224 rcu_read_lock();
2225 wq = rcu_dereference(sk->sk_wq);
2226 if (wq_has_sleeper(wq))
2227 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2228 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2229 rcu_read_unlock();
1da177e4
LT		 2230}
2231
676d2369 2232static void sock_def_readable(struct sock *sk)
1da177e4 2233{
43815482
ED		 2234 struct socket_wq *wq;
2235
2236 rcu_read_lock();
2237 wq = rcu_dereference(sk->sk_wq);
2238 if (wq_has_sleeper(wq))
2c6607c6 2239 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2240 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2241 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2242 rcu_read_unlock();
1da177e4
LT		 2243}
2244
2245static void sock_def_write_space(struct sock *sk)
2246{
43815482
ED		 2247 struct socket_wq *wq;
2248
2249 rcu_read_lock();
1da177e4
LT		 2250
2251 /* Do not wake up a writer until he can make "significant"
2252 * progress. --DaveM
2253 */
e71a4783 2254 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2255 wq = rcu_dereference(sk->sk_wq);
2256 if (wq_has_sleeper(wq))
2257 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2258 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2259
2260 /* Should agree with poll, otherwise some programs break */
2261 if (sock_writeable(sk))
8d8ad9d7 2262 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2263 }
2264
43815482 2265 rcu_read_unlock();
1da177e4
LT		 2266}
2267
2268static void sock_def_destruct(struct sock *sk)
2269{
a51482bd 2270 kfree(sk->sk_protinfo);
1da177e4
LT		 2271}
2272
2273void sk_send_sigurg(struct sock *sk)
2274{
2275 if (sk->sk_socket && sk->sk_socket->file)
2276 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2277 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2278}
2a91525c 2279EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2280
2281void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2282 unsigned long expires)
2283{
2284 if (!mod_timer(timer, expires))
2285 sock_hold(sk);
2286}
1da177e4
LT		 2287EXPORT_SYMBOL(sk_reset_timer);
2288
2289void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2290{
25cc4ae9 2291 if (del_timer(timer))
1da177e4
LT		 2292 __sock_put(sk);
2293}
1da177e4
LT		 2294EXPORT_SYMBOL(sk_stop_timer);
2295
2296void sock_init_data(struct socket *sock, struct sock *sk)
2297{
2298 skb_queue_head_init(&sk->sk_receive_queue);
2299 skb_queue_head_init(&sk->sk_write_queue);
2300 skb_queue_head_init(&sk->sk_error_queue);
2301
2302 sk->sk_send_head = NULL;
2303
2304 init_timer(&sk->sk_timer);
4ec93edb 2305
1da177e4
LT		 2306 sk->sk_allocation = GFP_KERNEL;
2307 sk->sk_rcvbuf = sysctl_rmem_default;
2308 sk->sk_sndbuf = sysctl_wmem_default;
2309 sk->sk_state = TCP_CLOSE;
972692e0 2310 sk_set_socket(sk, sock);
1da177e4
LT		 2311
2312 sock_set_flag(sk, SOCK_ZAPPED);
2313
e71a4783 2314 if (sock) {
1da177e4 2315 sk->sk_type = sock->type;
43815482 2316 sk->sk_wq = sock->wq;
1da177e4
LT		 2317 sock->sk = sk;
2318 } else
43815482 2319 sk->sk_wq = NULL;
1da177e4 2320
b6c6712a 2321 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2322 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2323 lockdep_set_class_and_name(&sk->sk_callback_lock,
2324 af_callback_keys + sk->sk_family,
2325 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2326
2327 sk->sk_state_change = sock_def_wakeup;
2328 sk->sk_data_ready = sock_def_readable;
2329 sk->sk_write_space = sock_def_write_space;
2330 sk->sk_error_report = sock_def_error_report;
2331 sk->sk_destruct = sock_def_destruct;
2332
5640f768
ED		 2333 sk->sk_frag.page = NULL;
2334 sk->sk_frag.offset = 0;
ef64a54f 2335 sk->sk_peek_off = -1;
1da177e4 2336
109f6e39
EB		 2337 sk->sk_peer_pid = NULL;
2338 sk->sk_peer_cred = NULL;
1da177e4
LT		 2339 sk->sk_write_pending = 0;
2340 sk->sk_rcvlowat = 1;
2341 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2342 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2343
f37f0afb 2344 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2345
e0d1095a 2346#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2347 sk->sk_napi_id = 0;
64b0dc51 2348 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2349#endif
2350
62748f32 2351 sk->sk_max_pacing_rate = ~0U;
7eec4174 2352 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2353 /*
2354 * Before updating sk_refcnt, we must commit prior changes to memory
2355 * (Documentation/RCU/rculist_nulls.txt for details)
2356 */
2357 smp_wmb();
1da177e4 2358 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2359 atomic_set(&sk->sk_drops, 0);
1da177e4 2360}
2a91525c 2361EXPORT_SYMBOL(sock_init_data);
1da177e4 2362
b5606c2d 2363void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2364{
2365 might_sleep();
a5b5bb9a 2366 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2367 if (sk->sk_lock.owned)
1da177e4 2368 __lock_sock(sk);
d2e9117c 2369 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2370 spin_unlock(&sk->sk_lock.slock);
2371 /*
2372 * The sk_lock has mutex_lock() semantics here:
2373 */
fcc70d5f 2374 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2375 local_bh_enable();
1da177e4 2376}
fcc70d5f 2377EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2378
b5606c2d 2379void release_sock(struct sock *sk)
1da177e4 2380{
a5b5bb9a
IM		 2381 /*
2382 * The sk_lock has mutex_unlock() semantics:
2383 */
2384 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2385
2386 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2387 if (sk->sk_backlog.tail)
2388 __release_sock(sk);
46d3ceab 2389
c3f9b018
ED		 2390 /* Warning : release_cb() might need to release sk ownership,
2391 * ie call sock_release_ownership(sk) before us.
2392 */
46d3ceab
ED		 2393 if (sk->sk_prot->release_cb)
2394 sk->sk_prot->release_cb(sk);
2395
c3f9b018 2396 sock_release_ownership(sk);
a5b5bb9a
IM		 2397 if (waitqueue_active(&sk->sk_lock.wq))
2398 wake_up(&sk->sk_lock.wq);
2399 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2400}
2401EXPORT_SYMBOL(release_sock);
2402
8a74ad60
ED		 2403/**
2404 * lock_sock_fast - fast version of lock_sock
2405 * @sk: socket
2406 *
2407 * This version should be used for very small section, where process wont block
2408 * return false if fast path is taken
2409 * sk_lock.slock locked, owned = 0, BH disabled
2410 * return true if slow path is taken
2411 * sk_lock.slock unlocked, owned = 1, BH enabled
2412 */
2413bool lock_sock_fast(struct sock *sk)
2414{
2415 might_sleep();
2416 spin_lock_bh(&sk->sk_lock.slock);
2417
2418 if (!sk->sk_lock.owned)
2419 /*
2420 * Note : We must disable BH
2421 */
2422 return false;
2423
2424 __lock_sock(sk);
2425 sk->sk_lock.owned = 1;
2426 spin_unlock(&sk->sk_lock.slock);
2427 /*
2428 * The sk_lock has mutex_lock() semantics here:
2429 */
2430 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2431 local_bh_enable();
2432 return true;
2433}
2434EXPORT_SYMBOL(lock_sock_fast);
2435
1da177e4 2436int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2437{
b7aa0bf7 2438 struct timeval tv;
1da177e4 2439 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2440 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2441 tv = ktime_to_timeval(sk->sk_stamp);
2442 if (tv.tv_sec == -1)
1da177e4 2443 return -ENOENT;
b7aa0bf7
ED		 2444 if (tv.tv_sec == 0) {
2445 sk->sk_stamp = ktime_get_real();
2446 tv = ktime_to_timeval(sk->sk_stamp);
2447 }
2448 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2449}
1da177e4
LT		 2450EXPORT_SYMBOL(sock_get_timestamp);
2451
ae40eb1e
ED		 2452int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2453{
2454 struct timespec ts;
2455 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2456 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2457 ts = ktime_to_timespec(sk->sk_stamp);
2458 if (ts.tv_sec == -1)
2459 return -ENOENT;
2460 if (ts.tv_sec == 0) {
2461 sk->sk_stamp = ktime_get_real();
2462 ts = ktime_to_timespec(sk->sk_stamp);
2463 }
2464 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2465}
2466EXPORT_SYMBOL(sock_get_timestampns);
2467
20d49473 2468void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2469{
20d49473 2470 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2471 unsigned long previous_flags = sk->sk_flags;
2472
20d49473
PO		 2473 sock_set_flag(sk, flag);
2474 /*
2475 * we just set one of the two flags which require net
2476 * time stamping, but time stamping might have been on
2477 * already because of the other one
2478 */
08e29af3 2479 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2480 net_enable_timestamp();
1da177e4
LT		 2481 }
2482}
1da177e4 2483
cb820f8e
RC		 2484int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2485 int level, int type)
2486{
2487 struct sock_exterr_skb *serr;
364a9e93 2488 struct sk_buff *skb;
cb820f8e
RC		 2489 int copied, err;
2490
2491 err = -EAGAIN;
364a9e93 2492 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2493 if (skb == NULL)
2494 goto out;
2495
2496 copied = skb->len;
2497 if (copied > len) {
2498 msg->msg_flags |= MSG_TRUNC;
2499 copied = len;
2500 }
51f3d02b 2501 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2502 if (err)
2503 goto out_free_skb;
2504
2505 sock_recv_timestamp(msg, sk, skb);
2506
2507 serr = SKB_EXT_ERR(skb);
2508 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2509
2510 msg->msg_flags |= MSG_ERRQUEUE;
2511 err = copied;
2512
cb820f8e
RC		 2513out_free_skb:
2514 kfree_skb(skb);
2515out:
2516 return err;
2517}
2518EXPORT_SYMBOL(sock_recv_errqueue);
2519
1da177e4
LT		 2520/*
2521 * Get a socket option on an socket.
2522 *
2523 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2524 * asynchronous errors should be reported by getsockopt. We assume
2525 * this means if you specify SO_ERROR (otherwise whats the point of it).
2526 */
2527int sock_common_getsockopt(struct socket *sock, int level, int optname,
2528 char __user *optval, int __user *optlen)
2529{
2530 struct sock *sk = sock->sk;
2531
2532 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2533}
1da177e4
LT		 2534EXPORT_SYMBOL(sock_common_getsockopt);
2535
3fdadf7d 2536#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2537int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2538 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2539{
2540 struct sock *sk = sock->sk;
2541
1e51f951 2542 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2543 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2544 optval, optlen);
3fdadf7d
DM		 2545 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2546}
2547EXPORT_SYMBOL(compat_sock_common_getsockopt);
2548#endif
2549
1da177e4
LT		 2550int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2551 struct msghdr *msg, size_t size, int flags)
2552{
2553 struct sock *sk = sock->sk;
2554 int addr_len = 0;
2555 int err;
2556
2557 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2558 flags & ~MSG_DONTWAIT, &addr_len);
2559 if (err >= 0)
2560 msg->msg_namelen = addr_len;
2561 return err;
2562}
1da177e4
LT		 2563EXPORT_SYMBOL(sock_common_recvmsg);
2564
2565/*
2566 * Set socket options on an inet socket.
2567 */
2568int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2569 char __user *optval, unsigned int optlen)
1da177e4
LT		 2570{
2571 struct sock *sk = sock->sk;
2572
2573 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2574}
1da177e4
LT		 2575EXPORT_SYMBOL(sock_common_setsockopt);
2576
3fdadf7d 2577#ifdef CONFIG_COMPAT
543d9cfe 2578int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2579 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2580{
2581 struct sock *sk = sock->sk;
2582
543d9cfe
ACM		 2583 if (sk->sk_prot->compat_setsockopt != NULL)
2584 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2585 optval, optlen);
3fdadf7d
DM		 2586 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2587}
2588EXPORT_SYMBOL(compat_sock_common_setsockopt);
2589#endif
2590
1da177e4
LT		 2591void sk_common_release(struct sock *sk)
2592{
2593 if (sk->sk_prot->destroy)
2594 sk->sk_prot->destroy(sk);
2595
2596 /*
2597 * Observation: when sock_common_release is called, processes have
2598 * no access to socket. But net still has.
2599 * Step one, detach it from networking:
2600 *
2601 * A. Remove from hash tables.
2602 */
2603
2604 sk->sk_prot->unhash(sk);
2605
2606 /*
2607 * In this point socket cannot receive new packets, but it is possible
2608 * that some packets are in flight because some CPU runs receiver and
2609 * did hash table lookup before we unhashed socket. They will achieve
2610 * receive queue and will be purged by socket destructor.
2611 *
2612 * Also we still have packets pending on receive queue and probably,
2613 * our own packets waiting in device queues. sock_destroy will drain
2614 * receive queue, but transmitted packets will delay socket destruction
2615 * until the last reference will be released.
2616 */
2617
2618 sock_orphan(sk);
2619
2620 xfrm_sk_free_policy(sk);
2621
e6848976 2622 sk_refcnt_debug_release(sk);
5640f768
ED		 2623
2624 if (sk->sk_frag.page) {
2625 put_page(sk->sk_frag.page);
2626 sk->sk_frag.page = NULL;
2627 }
2628
1da177e4
LT		 2629 sock_put(sk);
2630}
1da177e4
LT		 2631EXPORT_SYMBOL(sk_common_release);
2632
13ff3d6f
PE		 2633#ifdef CONFIG_PROC_FS
2634#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2635struct prot_inuse {
2636 int val[PROTO_INUSE_NR];
2637};
13ff3d6f
PE		 2638
2639static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2640
2641#ifdef CONFIG_NET_NS
2642void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2643{
d6d9ca0f 2644 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2645}
2646EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2647
2648int sock_prot_inuse_get(struct net *net, struct proto *prot)
2649{
2650 int cpu, idx = prot->inuse_idx;
2651 int res = 0;
2652
2653 for_each_possible_cpu(cpu)
2654 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2655
2656 return res >= 0 ? res : 0;
2657}
2658EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2659
2c8c1e72 2660static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2661{
2662 net->core.inuse = alloc_percpu(struct prot_inuse);
2663 return net->core.inuse ? 0 : -ENOMEM;
2664}
2665
2c8c1e72 2666static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2667{
2668 free_percpu(net->core.inuse);
2669}
2670
2671static struct pernet_operations net_inuse_ops = {
2672 .init = sock_inuse_init_net,
2673 .exit = sock_inuse_exit_net,
2674};
2675
2676static __init int net_inuse_init(void)
2677{
2678 if (register_pernet_subsys(&net_inuse_ops))
2679 panic("Cannot initialize net inuse counters");
2680
2681 return 0;
2682}
2683
2684core_initcall(net_inuse_init);
2685#else
1338d466
PE		 2686static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2687
c29a0bc4 2688void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2689{
d6d9ca0f 2690 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2691}
2692EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2693
c29a0bc4 2694int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2695{
2696 int cpu, idx = prot->inuse_idx;
2697 int res = 0;
2698
2699 for_each_possible_cpu(cpu)
2700 res += per_cpu(prot_inuse, cpu).val[idx];
2701
2702 return res >= 0 ? res : 0;
2703}
2704EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2705#endif
13ff3d6f
PE		 2706
2707static void assign_proto_idx(struct proto *prot)
2708{
2709 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2710
2711 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2712 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2713 return;
2714 }
2715
2716 set_bit(prot->inuse_idx, proto_inuse_idx);
2717}
2718
2719static void release_proto_idx(struct proto *prot)
2720{
2721 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2722 clear_bit(prot->inuse_idx, proto_inuse_idx);
2723}
2724#else
2725static inline void assign_proto_idx(struct proto *prot)
2726{
2727}
2728
2729static inline void release_proto_idx(struct proto *prot)
2730{
2731}
2732#endif
2733
b733c007
PE		 2734int proto_register(struct proto *prot, int alloc_slab)
2735{
1da177e4
LT		 2736 if (alloc_slab) {
2737 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2738 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2739 NULL);
1da177e4
LT		 2740
2741 if (prot->slab == NULL) {
e005d193
JP		 2742 pr_crit("%s: Can't create sock SLAB cache!\n",
2743 prot->name);
60e7663d 2744 goto out;
1da177e4 2745 }
2e6599cb
ACM		 2746
2747 if (prot->rsk_prot != NULL) {
faf23422 2748 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2749 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2750 goto out_free_sock_slab;
2751
7e56b5d6 2752 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2753 prot->rsk_prot->obj_size, 0,
20c2df83 2754 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2755
2756 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2757 pr_crit("%s: Can't create request sock SLAB cache!\n",
2758 prot->name);
2e6599cb
ACM		 2759 goto out_free_request_sock_slab_name;
2760 }
2761 }
8feaf0c0 2762
6d6ee43e 2763 if (prot->twsk_prot != NULL) {
faf23422 2764 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2765
7e56b5d6 2766 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2767 goto out_free_request_sock_slab;
2768
6d6ee43e 2769 prot->twsk_prot->twsk_slab =
7e56b5d6 2770 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2771 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2772 0,
2773 SLAB_HWCACHE_ALIGN |
2774 prot->slab_flags,
20c2df83 2775 NULL);
6d6ee43e 2776 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2777 goto out_free_timewait_sock_slab_name;
2778 }
1da177e4
LT		 2779 }
2780
36b77a52 2781 mutex_lock(&proto_list_mutex);
1da177e4 2782 list_add(&prot->node, &proto_list);
13ff3d6f 2783 assign_proto_idx(prot);
36b77a52 2784 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2785 return 0;
2786
8feaf0c0 2787out_free_timewait_sock_slab_name:
7e56b5d6 2788 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2789out_free_request_sock_slab:
2790 if (prot->rsk_prot && prot->rsk_prot->slab) {
2791 kmem_cache_destroy(prot->rsk_prot->slab);
2792 prot->rsk_prot->slab = NULL;
2793 }
2e6599cb 2794out_free_request_sock_slab_name:
72150e9b
DC		 2795 if (prot->rsk_prot)
2796 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2797out_free_sock_slab:
2798 kmem_cache_destroy(prot->slab);
2799 prot->slab = NULL;
b733c007
PE		 2800out:
2801 return -ENOBUFS;
1da177e4 2802}
1da177e4
LT		 2803EXPORT_SYMBOL(proto_register);
2804
2805void proto_unregister(struct proto *prot)
2806{
36b77a52 2807 mutex_lock(&proto_list_mutex);
13ff3d6f 2808 release_proto_idx(prot);
0a3f4358 2809 list_del(&prot->node);
36b77a52 2810 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2811
2812 if (prot->slab != NULL) {
2813 kmem_cache_destroy(prot->slab);
2814 prot->slab = NULL;
2815 }
2816
2e6599cb 2817 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2818 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2819 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2820 prot->rsk_prot->slab = NULL;
2821 }
2822
6d6ee43e 2823 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2824 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2825 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2826 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2827 }
1da177e4 2828}
1da177e4
LT		 2829EXPORT_SYMBOL(proto_unregister);
2830
2831#ifdef CONFIG_PROC_FS
1da177e4 2832static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2833 __acquires(proto_list_mutex)
1da177e4 2834{
36b77a52 2835 mutex_lock(&proto_list_mutex);
60f0438a 2836 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2837}
2838
2839static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2840{
60f0438a 2841 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2842}
2843
2844static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2845 __releases(proto_list_mutex)
1da177e4 2846{
36b77a52 2847 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2848}
2849
2850static char proto_method_implemented(const void *method)
2851{
2852 return method == NULL ? 'n' : 'y';
2853}
180d8cd9
GC		 2854static long sock_prot_memory_allocated(struct proto *proto)
2855{
cb75a36c 2856 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2857}
2858
2859static char *sock_prot_memory_pressure(struct proto *proto)
2860{
2861 return proto->memory_pressure != NULL ?
2862 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2863}
1da177e4
LT		 2864
2865static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2866{
180d8cd9 2867
8d987e5c 2868 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2869 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2870 proto->name,
2871 proto->obj_size,
14e943db 2872 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2873 sock_prot_memory_allocated(proto),
2874 sock_prot_memory_pressure(proto),
1da177e4
LT		 2875 proto->max_header,
2876 proto->slab == NULL ? "no" : "yes",
2877 module_name(proto->owner),
2878 proto_method_implemented(proto->close),
2879 proto_method_implemented(proto->connect),
2880 proto_method_implemented(proto->disconnect),
2881 proto_method_implemented(proto->accept),
2882 proto_method_implemented(proto->ioctl),
2883 proto_method_implemented(proto->init),
2884 proto_method_implemented(proto->destroy),
2885 proto_method_implemented(proto->shutdown),
2886 proto_method_implemented(proto->setsockopt),
2887 proto_method_implemented(proto->getsockopt),
2888 proto_method_implemented(proto->sendmsg),
2889 proto_method_implemented(proto->recvmsg),
2890 proto_method_implemented(proto->sendpage),
2891 proto_method_implemented(proto->bind),
2892 proto_method_implemented(proto->backlog_rcv),
2893 proto_method_implemented(proto->hash),
2894 proto_method_implemented(proto->unhash),
2895 proto_method_implemented(proto->get_port),
2896 proto_method_implemented(proto->enter_memory_pressure));
2897}
2898
2899static int proto_seq_show(struct seq_file *seq, void *v)
2900{
60f0438a 2901 if (v == &proto_list)
1da177e4
LT		 2902 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2903 "protocol",
2904 "size",
2905 "sockets",
2906 "memory",
2907 "press",
2908 "maxhdr",
2909 "slab",
2910 "module",
2911 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2912 else
60f0438a 2913 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2914 return 0;
2915}
2916
f690808e 2917static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2918 .start = proto_seq_start,
2919 .next = proto_seq_next,
2920 .stop = proto_seq_stop,
2921 .show = proto_seq_show,
2922};
2923
2924static int proto_seq_open(struct inode *inode, struct file *file)
2925{
14e943db
ED		 2926 return seq_open_net(inode, file, &proto_seq_ops,
2927 sizeof(struct seq_net_private));
1da177e4
LT		 2928}
2929
9a32144e 2930static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2931 .owner = THIS_MODULE,
2932 .open = proto_seq_open,
2933 .read = seq_read,
2934 .llseek = seq_lseek,
14e943db
ED		 2935 .release = seq_release_net,
2936};
2937
2938static __net_init int proto_init_net(struct net *net)
2939{
d4beaa66 2940 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2941 return -ENOMEM;
2942
2943 return 0;
2944}
2945
2946static __net_exit void proto_exit_net(struct net *net)
2947{
ece31ffd 2948 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2949}
2950
2951
2952static __net_initdata struct pernet_operations proto_net_ops = {
2953 .init = proto_init_net,
2954 .exit = proto_exit_net,
1da177e4
LT		 2955};
2956
2957static int __init proto_init(void)
2958{
14e943db 2959 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2960}
2961
2962subsys_initcall(proto_init);
2963
2964#endif /* PROC_FS */
Ubuntu Hirsute Linux kernel mirror