]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - net/core/sock.c
projects / mirror_ubuntu-zesty-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
tcp: do not fake tcp headers in tcp_send_rcvq()
[mirror_ubuntu-zesty-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
c93bdd0e
MG		 328struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
329EXPORT_SYMBOL_GPL(memalloc_socks);
330
7cb02404
MG		 331/**
332 * sk_set_memalloc - sets %SOCK_MEMALLOC
333 * @sk: socket to set it on
334 *
335 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
336 * It's the responsibility of the admin to adjust min_free_kbytes
337 * to meet the requirements
338 */
339void sk_set_memalloc(struct sock *sk)
340{
341 sock_set_flag(sk, SOCK_MEMALLOC);
342 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 343 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 344}
345EXPORT_SYMBOL_GPL(sk_set_memalloc);
346
347void sk_clear_memalloc(struct sock *sk)
348{
349 sock_reset_flag(sk, SOCK_MEMALLOC);
350 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 351 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 352
353 /*
354 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
355 * progress of swapping. However, if SOCK_MEMALLOC is cleared while
356 * it has rmem allocations there is a risk that the user of the
357 * socket cannot make forward progress due to exceeding the rmem
358 * limits. By rights, sk_clear_memalloc() should only be called
359 * on sockets being torn down but warn and reset the accounting if
360 * that assumption breaks.
361 */
362 if (WARN_ON(sk->sk_forward_alloc))
363 sk_mem_reclaim(sk);
7cb02404
MG		 364}
365EXPORT_SYMBOL_GPL(sk_clear_memalloc);
366
b4b9e355
MG		 367int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
368{
369 int ret;
370 unsigned long pflags = current->flags;
371
372 /* these should have been dropped before queueing */
373 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
374
375 current->flags |= PF_MEMALLOC;
376 ret = sk->sk_backlog_rcv(sk, skb);
377 tsk_restore_flags(current, pflags, PF_MEMALLOC);
378
379 return ret;
380}
381EXPORT_SYMBOL(__sk_backlog_rcv);
382
1da177e4
LT		 383static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
384{
385 struct timeval tv;
386
387 if (optlen < sizeof(tv))
388 return -EINVAL;
389 if (copy_from_user(&tv, optval, sizeof(tv)))
390 return -EFAULT;
ba78073e
VA		 391 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
392 return -EDOM;
1da177e4 393
ba78073e 394 if (tv.tv_sec < 0) {
6f11df83
AM		 395 static int warned __read_mostly;
396
ba78073e 397 *timeo_p = 0;
50aab54f 398 if (warned < 10 && net_ratelimit()) {
ba78073e 399 warned++;
e005d193
JP		 400 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
401 __func__, current->comm, task_pid_nr(current));
50aab54f 402 }
ba78073e
VA		 403 return 0;
404 }
1da177e4
LT		 405 *timeo_p = MAX_SCHEDULE_TIMEOUT;
406 if (tv.tv_sec == 0 && tv.tv_usec == 0)
407 return 0;
408 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
409 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
410 return 0;
411}
412
413static void sock_warn_obsolete_bsdism(const char *name)
414{
415 static int warned;
416 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 417 if (strcmp(warncomm, current->comm) && warned < 5) {
418 strcpy(warncomm, current->comm);
e005d193
JP		 419 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
420 warncomm, name);
1da177e4
LT		 421 warned++;
422 }
423}
424
08e29af3
ED		 425#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
426
427static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 428{
08e29af3
ED		 429 if (sk->sk_flags & flags) {
430 sk->sk_flags &= ~flags;
431 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 432 net_disable_timestamp();
1da177e4
LT		 433 }
434}
435
436
f0088a50
DV		 437int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
438{
766e9037 439 int err;
3b885787
NH		 440 unsigned long flags;
441 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 442
0fd7bac6 443 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 444 atomic_inc(&sk->sk_drops);
3847ce32 445 trace_sock_rcvqueue_full(sk, skb);
766e9037 446 return -ENOMEM;
f0088a50
DV		 447 }
448
fda9ef5d 449 err = sk_filter(sk, skb);
f0088a50 450 if (err)
766e9037 451 return err;
f0088a50 452
c76562b6 453 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 454 atomic_inc(&sk->sk_drops);
455 return -ENOBUFS;
3ab224be
HA		 456 }
457
f0088a50
DV		 458 skb->dev = NULL;
459 skb_set_owner_r(skb, sk);
49ad9599 460
7fee226a
ED		 461 /* we escape from rcu protected region, make sure we dont leak
462 * a norefcounted dst
463 */
464 skb_dst_force(skb);
465
3b885787
NH		 466 spin_lock_irqsave(&list->lock, flags);
467 skb->dropcount = atomic_read(&sk->sk_drops);
468 __skb_queue_tail(list, skb);
469 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 470
471 if (!sock_flag(sk, SOCK_DEAD))
676d2369 472 sk->sk_data_ready(sk);
766e9037 473 return 0;
f0088a50
DV		 474}
475EXPORT_SYMBOL(sock_queue_rcv_skb);
476
58a5a7b9 477int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 478{
479 int rc = NET_RX_SUCCESS;
480
fda9ef5d 481 if (sk_filter(sk, skb))
f0088a50
DV		 482 goto discard_and_relse;
483
484 skb->dev = NULL;
485
274f482d 486 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 487 atomic_inc(&sk->sk_drops);
488 goto discard_and_relse;
489 }
58a5a7b9
ACM		 490 if (nested)
491 bh_lock_sock_nested(sk);
492 else
493 bh_lock_sock(sk);
a5b5bb9a
IM		 494 if (!sock_owned_by_user(sk)) {
495 /*
496 * trylock + unlock semantics:
497 */
498 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
499
c57943a1 500 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 501
502 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 503 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 504 bh_unlock_sock(sk);
505 atomic_inc(&sk->sk_drops);
506 goto discard_and_relse;
507 }
508
f0088a50
DV		 509 bh_unlock_sock(sk);
510out:
511 sock_put(sk);
512 return rc;
513discard_and_relse:
514 kfree_skb(skb);
515 goto out;
516}
517EXPORT_SYMBOL(sk_receive_skb);
518
519struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
520{
b6c6712a 521 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 522
523 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 524 sk_tx_queue_clear(sk);
a9b3cd7f 525 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 526 dst_release(dst);
527 return NULL;
528 }
529
530 return dst;
531}
532EXPORT_SYMBOL(__sk_dst_check);
533
534struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
535{
536 struct dst_entry *dst = sk_dst_get(sk);
537
538 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
539 sk_dst_reset(sk);
540 dst_release(dst);
541 return NULL;
542 }
543
544 return dst;
545}
546EXPORT_SYMBOL(sk_dst_check);
547
c91f6df2
BH		 548static int sock_setbindtodevice(struct sock *sk, char __user *optval,
549 int optlen)
4878809f
DM		 550{
551 int ret = -ENOPROTOOPT;
552#ifdef CONFIG_NETDEVICES
3b1e0a65 553 struct net *net = sock_net(sk);
4878809f
DM		 554 char devname[IFNAMSIZ];
555 int index;
556
557 /* Sorry... */
558 ret = -EPERM;
5e1fccc0 559 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 560 goto out;
561
562 ret = -EINVAL;
563 if (optlen < 0)
564 goto out;
565
566 /* Bind this socket to a particular device like "eth0",
567 * as specified in the passed interface name. If the
568 * name is "" or the option length is zero the socket
569 * is not bound.
570 */
571 if (optlen > IFNAMSIZ - 1)
572 optlen = IFNAMSIZ - 1;
573 memset(devname, 0, sizeof(devname));
574
575 ret = -EFAULT;
576 if (copy_from_user(devname, optval, optlen))
577 goto out;
578
000ba2e4
DM		 579 index = 0;
580 if (devname[0] != '\0') {
bf8e56bf 581 struct net_device *dev;
4878809f 582
bf8e56bf
ED		 583 rcu_read_lock();
584 dev = dev_get_by_name_rcu(net, devname);
585 if (dev)
586 index = dev->ifindex;
587 rcu_read_unlock();
4878809f
DM		 588 ret = -ENODEV;
589 if (!dev)
590 goto out;
4878809f
DM		 591 }
592
593 lock_sock(sk);
594 sk->sk_bound_dev_if = index;
595 sk_dst_reset(sk);
596 release_sock(sk);
597
598 ret = 0;
599
600out:
601#endif
602
603 return ret;
604}
605
c91f6df2
BH		 606static int sock_getbindtodevice(struct sock *sk, char __user *optval,
607 int __user *optlen, int len)
608{
609 int ret = -ENOPROTOOPT;
610#ifdef CONFIG_NETDEVICES
611 struct net *net = sock_net(sk);
c91f6df2 612 char devname[IFNAMSIZ];
c91f6df2
BH		 613
614 if (sk->sk_bound_dev_if == 0) {
615 len = 0;
616 goto zero;
617 }
618
619 ret = -EINVAL;
620 if (len < IFNAMSIZ)
621 goto out;
622
5dbe7c17
NS		 623 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
624 if (ret)
c91f6df2 625 goto out;
c91f6df2
BH		 626
627 len = strlen(devname) + 1;
628
629 ret = -EFAULT;
630 if (copy_to_user(optval, devname, len))
631 goto out;
632
633zero:
634 ret = -EFAULT;
635 if (put_user(len, optlen))
636 goto out;
637
638 ret = 0;
639
640out:
641#endif
642
643 return ret;
644}
645
c0ef877b
PE		 646static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
647{
648 if (valbool)
649 sock_set_flag(sk, bit);
650 else
651 sock_reset_flag(sk, bit);
652}
653
1da177e4
LT		 654/*
655 * This is meant for all protocols to use and covers goings on
656 * at the socket level. Everything here is generic.
657 */
658
659int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 660 char __user *optval, unsigned int optlen)
1da177e4 661{
2a91525c 662 struct sock *sk = sock->sk;
1da177e4
LT		 663 int val;
664 int valbool;
665 struct linger ling;
666 int ret = 0;
4ec93edb 667
1da177e4
LT		 668 /*
669 * Options without arguments
670 */
671
4878809f 672 if (optname == SO_BINDTODEVICE)
c91f6df2 673 return sock_setbindtodevice(sk, optval, optlen);
4878809f 674
e71a4783
SH		 675 if (optlen < sizeof(int))
676 return -EINVAL;
4ec93edb 677
1da177e4
LT		 678 if (get_user(val, (int __user *)optval))
679 return -EFAULT;
4ec93edb 680
2a91525c 681 valbool = val ? 1 : 0;
1da177e4
LT		 682
683 lock_sock(sk);
684
2a91525c 685 switch (optname) {
e71a4783 686 case SO_DEBUG:
2a91525c 687 if (val && !capable(CAP_NET_ADMIN))
e71a4783 688 ret = -EACCES;
2a91525c 689 else
c0ef877b 690 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 691 break;
692 case SO_REUSEADDR:
4a17fd52 693 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 694 break;
055dc21a
TH		 695 case SO_REUSEPORT:
696 sk->sk_reuseport = valbool;
697 break;
e71a4783 698 case SO_TYPE:
49c794e9 699 case SO_PROTOCOL:
0d6038ee 700 case SO_DOMAIN:
e71a4783
SH		 701 case SO_ERROR:
702 ret = -ENOPROTOOPT;
703 break;
704 case SO_DONTROUTE:
c0ef877b 705 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 706 break;
707 case SO_BROADCAST:
708 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
709 break;
710 case SO_SNDBUF:
711 /* Don't error on this BSD doesn't and if you think
82981930
ED		 712 * about it this is right. Otherwise apps have to
713 * play 'guess the biggest size' games. RCVBUF/SNDBUF
714 * are treated in BSD as hints
715 */
716 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 717set_sndbuf:
e71a4783 718 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 719 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
720 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 721 sk->sk_write_space(sk);
722 break;
1da177e4 723
e71a4783
SH		 724 case SO_SNDBUFFORCE:
725 if (!capable(CAP_NET_ADMIN)) {
726 ret = -EPERM;
727 break;
728 }
729 goto set_sndbuf;
b0573dea 730
e71a4783
SH		 731 case SO_RCVBUF:
732 /* Don't error on this BSD doesn't and if you think
82981930
ED		 733 * about it this is right. Otherwise apps have to
734 * play 'guess the biggest size' games. RCVBUF/SNDBUF
735 * are treated in BSD as hints
736 */
737 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 738set_rcvbuf:
e71a4783
SH		 739 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
740 /*
741 * We double it on the way in to account for
742 * "struct sk_buff" etc. overhead. Applications
743 * assume that the SO_RCVBUF setting they make will
744 * allow that much actual data to be received on that
745 * socket.
746 *
747 * Applications are unaware that "struct sk_buff" and
748 * other overheads allocate from the receive buffer
749 * during socket buffer allocation.
750 *
751 * And after considering the possible alternatives,
752 * returning the value we actually used in getsockopt
753 * is the most desirable behavior.
754 */
82981930 755 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 756 break;
757
758 case SO_RCVBUFFORCE:
759 if (!capable(CAP_NET_ADMIN)) {
760 ret = -EPERM;
1da177e4 761 break;
e71a4783
SH		 762 }
763 goto set_rcvbuf;
1da177e4 764
e71a4783 765 case SO_KEEPALIVE:
1da177e4 766#ifdef CONFIG_INET
3e10986d
ED		 767 if (sk->sk_protocol == IPPROTO_TCP &&
768 sk->sk_type == SOCK_STREAM)
e71a4783 769 tcp_set_keepalive(sk, valbool);
1da177e4 770#endif
e71a4783
SH		 771 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
772 break;
773
774 case SO_OOBINLINE:
775 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
776 break;
777
778 case SO_NO_CHECK:
28448b80 779 sk->sk_no_check_tx = valbool;
e71a4783
SH		 780 break;
781
782 case SO_PRIORITY:
5e1fccc0
EB		 783 if ((val >= 0 && val <= 6) ||
784 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 785 sk->sk_priority = val;
786 else
787 ret = -EPERM;
788 break;
789
790 case SO_LINGER:
791 if (optlen < sizeof(ling)) {
792 ret = -EINVAL; /* 1003.1g */
1da177e4 793 break;
e71a4783 794 }
2a91525c 795 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 796 ret = -EFAULT;
1da177e4 797 break;
e71a4783
SH		 798 }
799 if (!ling.l_onoff)
800 sock_reset_flag(sk, SOCK_LINGER);
801 else {
1da177e4 802#if (BITS_PER_LONG == 32)
e71a4783
SH		 803 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
804 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 805 else
e71a4783
SH		 806#endif
807 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
808 sock_set_flag(sk, SOCK_LINGER);
809 }
810 break;
811
812 case SO_BSDCOMPAT:
813 sock_warn_obsolete_bsdism("setsockopt");
814 break;
815
816 case SO_PASSCRED:
817 if (valbool)
818 set_bit(SOCK_PASSCRED, &sock->flags);
819 else
820 clear_bit(SOCK_PASSCRED, &sock->flags);
821 break;
822
823 case SO_TIMESTAMP:
92f37fd2 824 case SO_TIMESTAMPNS:
e71a4783 825 if (valbool) {
92f37fd2
ED		 826 if (optname == SO_TIMESTAMP)
827 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
828 else
829 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 830 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 831 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 832 } else {
e71a4783 833 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 834 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
835 }
e71a4783
SH		 836 break;
837
20d49473
PO		 838 case SO_TIMESTAMPING:
839 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 840 ret = -EINVAL;
20d49473
PO		 841 break;
842 }
09c2d251 843 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765
WB		 844 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
845 if (sk->sk_protocol == IPPROTO_TCP) {
846 if (sk->sk_state != TCP_ESTABLISHED) {
847 ret = -EINVAL;
848 break;
849 }
850 sk->sk_tskey = tcp_sk(sk)->snd_una;
851 } else {
852 sk->sk_tskey = 0;
853 }
854 }
b9f40e21 855 sk->sk_tsflags = val;
20d49473
PO		 856 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
857 sock_enable_timestamp(sk,
858 SOCK_TIMESTAMPING_RX_SOFTWARE);
859 else
860 sock_disable_timestamp(sk,
08e29af3 861 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 862 break;
863
e71a4783
SH		 864 case SO_RCVLOWAT:
865 if (val < 0)
866 val = INT_MAX;
867 sk->sk_rcvlowat = val ? : 1;
868 break;
869
870 case SO_RCVTIMEO:
871 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
872 break;
873
874 case SO_SNDTIMEO:
875 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
876 break;
1da177e4 877
e71a4783
SH		 878 case SO_ATTACH_FILTER:
879 ret = -EINVAL;
880 if (optlen == sizeof(struct sock_fprog)) {
881 struct sock_fprog fprog;
1da177e4 882
e71a4783
SH		 883 ret = -EFAULT;
884 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 885 break;
e71a4783
SH		 886
887 ret = sk_attach_filter(&fprog, sk);
888 }
889 break;
890
891 case SO_DETACH_FILTER:
55b33325 892 ret = sk_detach_filter(sk);
e71a4783 893 break;
1da177e4 894
d59577b6
VB		 895 case SO_LOCK_FILTER:
896 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
897 ret = -EPERM;
898 else
899 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
900 break;
901
e71a4783
SH		 902 case SO_PASSSEC:
903 if (valbool)
904 set_bit(SOCK_PASSSEC, &sock->flags);
905 else
906 clear_bit(SOCK_PASSSEC, &sock->flags);
907 break;
4a19ec58 908 case SO_MARK:
5e1fccc0 909 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 910 ret = -EPERM;
2a91525c 911 else
4a19ec58 912 sk->sk_mark = val;
4a19ec58 913 break;
877ce7c1 914
1da177e4
LT		 915 /* We implement the SO_SNDLOWAT etc to
916 not be settable (1003.1g 5.3) */
3b885787 917 case SO_RXQ_OVFL:
8083f0fc 918 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 919 break;
6e3e939f
JB		 920
921 case SO_WIFI_STATUS:
922 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
923 break;
924
ef64a54f
PE		 925 case SO_PEEK_OFF:
926 if (sock->ops->set_peek_off)
12663bfc 927 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 928 else
929 ret = -EOPNOTSUPP;
930 break;
3bdc0eba
BG		 931
932 case SO_NOFCS:
933 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
934 break;
935
7d4c04fc
KJ		 936 case SO_SELECT_ERR_QUEUE:
937 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
938 break;
939
e0d1095a 940#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 941 case SO_BUSY_POLL:
dafcc438
ET		 942 /* allow unprivileged users to decrease the value */
943 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
944 ret = -EPERM;
945 else {
946 if (val < 0)
947 ret = -EINVAL;
948 else
949 sk->sk_ll_usec = val;
950 }
951 break;
952#endif
62748f32
ED		 953
954 case SO_MAX_PACING_RATE:
955 sk->sk_max_pacing_rate = val;
956 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
957 sk->sk_max_pacing_rate);
958 break;
959
e71a4783
SH		 960 default:
961 ret = -ENOPROTOOPT;
962 break;
4ec93edb 963 }
1da177e4
LT		 964 release_sock(sk);
965 return ret;
966}
2a91525c 967EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 968
969
8f09898b 970static void cred_to_ucred(struct pid *pid, const struct cred *cred,
971 struct ucred *ucred)
3f551f94
EB		 972{
973 ucred->pid = pid_vnr(pid);
974 ucred->uid = ucred->gid = -1;
975 if (cred) {
976 struct user_namespace *current_ns = current_user_ns();
977
b2e4f544
EB		 978 ucred->uid = from_kuid_munged(current_ns, cred->euid);
979 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 980 }
981}
982
1da177e4
LT		 983int sock_getsockopt(struct socket *sock, int level, int optname,
984 char __user *optval, int __user *optlen)
985{
986 struct sock *sk = sock->sk;
4ec93edb 987
e71a4783 988 union {
4ec93edb
YH		 989 int val;
990 struct linger ling;
1da177e4
LT		 991 struct timeval tm;
992 } v;
4ec93edb 993
4d0392be 994 int lv = sizeof(int);
1da177e4 995 int len;
4ec93edb 996
e71a4783 997 if (get_user(len, optlen))
4ec93edb 998 return -EFAULT;
e71a4783 999 if (len < 0)
1da177e4 1000 return -EINVAL;
4ec93edb 1001
50fee1de 1002 memset(&v, 0, sizeof(v));
df0bca04 1003
2a91525c 1004 switch (optname) {
e71a4783
SH		 1005 case SO_DEBUG:
1006 v.val = sock_flag(sk, SOCK_DBG);
1007 break;
1008
1009 case SO_DONTROUTE:
1010 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1011 break;
1012
1013 case SO_BROADCAST:
1b23a5df 1014 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1015 break;
1016
1017 case SO_SNDBUF:
1018 v.val = sk->sk_sndbuf;
1019 break;
1020
1021 case SO_RCVBUF:
1022 v.val = sk->sk_rcvbuf;
1023 break;
1024
1025 case SO_REUSEADDR:
1026 v.val = sk->sk_reuse;
1027 break;
1028
055dc21a
TH		 1029 case SO_REUSEPORT:
1030 v.val = sk->sk_reuseport;
1031 break;
1032
e71a4783 1033 case SO_KEEPALIVE:
1b23a5df 1034 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1035 break;
1036
1037 case SO_TYPE:
1038 v.val = sk->sk_type;
1039 break;
1040
49c794e9
JE		 1041 case SO_PROTOCOL:
1042 v.val = sk->sk_protocol;
1043 break;
1044
0d6038ee
JE		 1045 case SO_DOMAIN:
1046 v.val = sk->sk_family;
1047 break;
1048
e71a4783
SH		 1049 case SO_ERROR:
1050 v.val = -sock_error(sk);
2a91525c 1051 if (v.val == 0)
e71a4783
SH		 1052 v.val = xchg(&sk->sk_err_soft, 0);
1053 break;
1054
1055 case SO_OOBINLINE:
1b23a5df 1056 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1057 break;
1058
1059 case SO_NO_CHECK:
28448b80 1060 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1061 break;
1062
1063 case SO_PRIORITY:
1064 v.val = sk->sk_priority;
1065 break;
1066
1067 case SO_LINGER:
1068 lv = sizeof(v.ling);
1b23a5df 1069 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1070 v.ling.l_linger = sk->sk_lingertime / HZ;
1071 break;
1072
1073 case SO_BSDCOMPAT:
1074 sock_warn_obsolete_bsdism("getsockopt");
1075 break;
1076
1077 case SO_TIMESTAMP:
92f37fd2
ED		 1078 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1079 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1080 break;
1081
1082 case SO_TIMESTAMPNS:
1083 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1084 break;
1085
20d49473 1086 case SO_TIMESTAMPING:
b9f40e21 1087 v.val = sk->sk_tsflags;
20d49473
PO		 1088 break;
1089
e71a4783 1090 case SO_RCVTIMEO:
2a91525c 1091 lv = sizeof(struct timeval);
e71a4783
SH		 1092 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1093 v.tm.tv_sec = 0;
1094 v.tm.tv_usec = 0;
1095 } else {
1096 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1097 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1098 }
1099 break;
1100
1101 case SO_SNDTIMEO:
2a91525c 1102 lv = sizeof(struct timeval);
e71a4783
SH		 1103 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1104 v.tm.tv_sec = 0;
1105 v.tm.tv_usec = 0;
1106 } else {
1107 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1108 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1109 }
1110 break;
1da177e4 1111
e71a4783
SH		 1112 case SO_RCVLOWAT:
1113 v.val = sk->sk_rcvlowat;
1114 break;
1da177e4 1115
e71a4783 1116 case SO_SNDLOWAT:
2a91525c 1117 v.val = 1;
e71a4783 1118 break;
1da177e4 1119
e71a4783 1120 case SO_PASSCRED:
82981930 1121 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1122 break;
1da177e4 1123
e71a4783 1124 case SO_PEERCRED:
109f6e39
EB		 1125 {
1126 struct ucred peercred;
1127 if (len > sizeof(peercred))
1128 len = sizeof(peercred);
1129 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1130 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1131 return -EFAULT;
1132 goto lenout;
109f6e39 1133 }
1da177e4 1134
e71a4783
SH		 1135 case SO_PEERNAME:
1136 {
1137 char address[128];
1138
1139 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1140 return -ENOTCONN;
1141 if (lv < len)
1142 return -EINVAL;
1143 if (copy_to_user(optval, address, len))
1144 return -EFAULT;
1145 goto lenout;
1146 }
1da177e4 1147
e71a4783
SH		 1148 /* Dubious BSD thing... Probably nobody even uses it, but
1149 * the UNIX standard wants it for whatever reason... -DaveM
1150 */
1151 case SO_ACCEPTCONN:
1152 v.val = sk->sk_state == TCP_LISTEN;
1153 break;
1da177e4 1154
e71a4783 1155 case SO_PASSSEC:
82981930 1156 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1157 break;
877ce7c1 1158
e71a4783
SH		 1159 case SO_PEERSEC:
1160 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1161
4a19ec58
LAT		 1162 case SO_MARK:
1163 v.val = sk->sk_mark;
1164 break;
1165
3b885787 1166 case SO_RXQ_OVFL:
1b23a5df 1167 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1168 break;
1169
6e3e939f 1170 case SO_WIFI_STATUS:
1b23a5df 1171 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1172 break;
1173
ef64a54f
PE		 1174 case SO_PEEK_OFF:
1175 if (!sock->ops->set_peek_off)
1176 return -EOPNOTSUPP;
1177
1178 v.val = sk->sk_peek_off;
1179 break;
bc2f7996 1180 case SO_NOFCS:
1b23a5df 1181 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1182 break;
c91f6df2 1183
f7b86bfe 1184 case SO_BINDTODEVICE:
c91f6df2
BH		 1185 return sock_getbindtodevice(sk, optval, optlen, len);
1186
a8fc9277
PE		 1187 case SO_GET_FILTER:
1188 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1189 if (len < 0)
1190 return len;
1191
1192 goto lenout;
c91f6df2 1193
d59577b6
VB		 1194 case SO_LOCK_FILTER:
1195 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1196 break;
1197
ea02f941
MS		 1198 case SO_BPF_EXTENSIONS:
1199 v.val = bpf_tell_extensions();
1200 break;
1201
7d4c04fc
KJ		 1202 case SO_SELECT_ERR_QUEUE:
1203 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1204 break;
1205
e0d1095a 1206#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1207 case SO_BUSY_POLL:
dafcc438
ET		 1208 v.val = sk->sk_ll_usec;
1209 break;
1210#endif
1211
62748f32
ED		 1212 case SO_MAX_PACING_RATE:
1213 v.val = sk->sk_max_pacing_rate;
1214 break;
1215
e71a4783
SH		 1216 default:
1217 return -ENOPROTOOPT;
1da177e4 1218 }
e71a4783 1219
1da177e4
LT		 1220 if (len > lv)
1221 len = lv;
1222 if (copy_to_user(optval, &v, len))
1223 return -EFAULT;
1224lenout:
4ec93edb
YH		 1225 if (put_user(len, optlen))
1226 return -EFAULT;
1227 return 0;
1da177e4
LT		 1228}
1229
a5b5bb9a
IM		 1230/*
1231 * Initialize an sk_lock.
1232 *
1233 * (We also register the sk_lock with the lock validator.)
1234 */
b6f99a21 1235static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1236{
ed07536e
PZ		 1237 sock_lock_init_class_and_name(sk,
1238 af_family_slock_key_strings[sk->sk_family],
1239 af_family_slock_keys + sk->sk_family,
1240 af_family_key_strings[sk->sk_family],
1241 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1242}
1243
4dc6dc71
ED		 1244/*
1245 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1246 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1247 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1248 */
f1a6c4da
PE		 1249static void sock_copy(struct sock *nsk, const struct sock *osk)
1250{
1251#ifdef CONFIG_SECURITY_NETWORK
1252 void *sptr = nsk->sk_security;
1253#endif
68835aba
ED		 1254 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1255
1256 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1257 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1258
f1a6c4da
PE		 1259#ifdef CONFIG_SECURITY_NETWORK
1260 nsk->sk_security = sptr;
1261 security_sk_clone(osk, nsk);
1262#endif
1263}
1264
fcbdf09d
OP		 1265void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1266{
1267 unsigned long nulls1, nulls2;
1268
1269 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1270 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1271 if (nulls1 > nulls2)
1272 swap(nulls1, nulls2);
1273
1274 if (nulls1 != 0)
1275 memset((char *)sk, 0, nulls1);
1276 memset((char *)sk + nulls1 + sizeof(void *), 0,
1277 nulls2 - nulls1 - sizeof(void *));
1278 memset((char *)sk + nulls2 + sizeof(void *), 0,
1279 size - nulls2 - sizeof(void *));
1280}
1281EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1282
2e4afe7b
PE		 1283static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1284 int family)
c308c1b2
PE		 1285{
1286 struct sock *sk;
1287 struct kmem_cache *slab;
1288
1289 slab = prot->slab;
e912b114
ED		 1290 if (slab != NULL) {
1291 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1292 if (!sk)
1293 return sk;
1294 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1295 if (prot->clear_sk)
1296 prot->clear_sk(sk, prot->obj_size);
1297 else
1298 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1299 }
fcbdf09d 1300 } else
c308c1b2
PE		 1301 sk = kmalloc(prot->obj_size, priority);
1302
2e4afe7b 1303 if (sk != NULL) {
a98b65a3
VN		 1304 kmemcheck_annotate_bitfield(sk, flags);
1305
2e4afe7b
PE		 1306 if (security_sk_alloc(sk, family, priority))
1307 goto out_free;
1308
1309 if (!try_module_get(prot->owner))
1310 goto out_free_sec;
e022f0b4 1311 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1312 }
1313
c308c1b2 1314 return sk;
2e4afe7b
PE		 1315
1316out_free_sec:
1317 security_sk_free(sk);
1318out_free:
1319 if (slab != NULL)
1320 kmem_cache_free(slab, sk);
1321 else
1322 kfree(sk);
1323 return NULL;
c308c1b2
PE		 1324}
1325
1326static void sk_prot_free(struct proto *prot, struct sock *sk)
1327{
1328 struct kmem_cache *slab;
2e4afe7b 1329 struct module *owner;
c308c1b2 1330
2e4afe7b 1331 owner = prot->owner;
c308c1b2 1332 slab = prot->slab;
2e4afe7b
PE		 1333
1334 security_sk_free(sk);
c308c1b2
PE		 1335 if (slab != NULL)
1336 kmem_cache_free(slab, sk);
1337 else
1338 kfree(sk);
2e4afe7b 1339 module_put(owner);
c308c1b2
PE		 1340}
1341
86f8515f 1342#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1343void sock_update_netprioidx(struct sock *sk)
5bc1421e 1344{
5bc1421e
NH		 1345 if (in_interrupt())
1346 return;
2b73bc65 1347
6ffd4641 1348 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1349}
1350EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1351#endif
1352
1da177e4
LT		 1353/**
1354 * sk_alloc - All socket objects are allocated here
c4ea43c5 1355 * @net: the applicable net namespace
4dc3b16b
PP		 1356 * @family: protocol family
1357 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1358 * @prot: struct proto associated with this new sock instance
1da177e4 1359 */
1b8d7ae4 1360struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1361 struct proto *prot)
1da177e4 1362{
c308c1b2 1363 struct sock *sk;
1da177e4 1364
154adbc8 1365 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1366 if (sk) {
154adbc8
PE		 1367 sk->sk_family = family;
1368 /*
1369 * See comment in struct sock definition to understand
1370 * why we need sk_prot_creator -acme
1371 */
1372 sk->sk_prot = sk->sk_prot_creator = prot;
1373 sock_lock_init(sk);
3b1e0a65 1374 sock_net_set(sk, get_net(net));
d66ee058 1375 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1376
211d2f97 1377 sock_update_classid(sk);
6ffd4641 1378 sock_update_netprioidx(sk);
1da177e4 1379 }
a79af59e 1380
2e4afe7b 1381 return sk;
1da177e4 1382}
2a91525c 1383EXPORT_SYMBOL(sk_alloc);
1da177e4 1384
2b85a34e 1385static void __sk_free(struct sock *sk)
1da177e4
LT		 1386{
1387 struct sk_filter *filter;
1da177e4
LT		 1388
1389 if (sk->sk_destruct)
1390 sk->sk_destruct(sk);
1391
a898def2
PM		 1392 filter = rcu_dereference_check(sk->sk_filter,
1393 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1394 if (filter) {
309dd5fc 1395 sk_filter_uncharge(sk, filter);
a9b3cd7f 1396 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1397 }
1398
08e29af3 1399 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1400
1401 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1402 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1403 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1404
109f6e39
EB		 1405 if (sk->sk_peer_cred)
1406 put_cred(sk->sk_peer_cred);
1407 put_pid(sk->sk_peer_pid);
3b1e0a65 1408 put_net(sock_net(sk));
c308c1b2 1409 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1410}
2b85a34e
ED		 1411
1412void sk_free(struct sock *sk)
1413{
1414 /*
25985edc 1415 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1416 * some packets are still in some tx queue.
1417 * If not null, sock_wfree() will call __sk_free(sk) later
1418 */
1419 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1420 __sk_free(sk);
1421}
2a91525c 1422EXPORT_SYMBOL(sk_free);
1da177e4 1423
edf02087 1424/*
25985edc
LDM		 1425 * Last sock_put should drop reference to sk->sk_net. It has already
1426 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1427 * is not an option.
25985edc 1428 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1429 * destroy it in the context of init_net.
1430 */
1431void sk_release_kernel(struct sock *sk)
1432{
1433 if (sk == NULL || sk->sk_socket == NULL)
1434 return;
1435
1436 sock_hold(sk);
1437 sock_release(sk->sk_socket);
65a18ec5 1438 release_net(sock_net(sk));
3b1e0a65 1439 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1440 sock_put(sk);
1441}
45af1754 1442EXPORT_SYMBOL(sk_release_kernel);
edf02087 1443
475f1b52
SR		 1444static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1445{
1446 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1447 sock_update_memcg(newsk);
1448}
1449
e56c57d0
ED		 1450/**
1451 * sk_clone_lock - clone a socket, and lock its clone
1452 * @sk: the socket to clone
1453 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1454 *
1455 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1456 */
1457struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1458{
8fd1d178 1459 struct sock *newsk;
278571ba 1460 bool is_charged = true;
87d11ceb 1461
8fd1d178 1462 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1463 if (newsk != NULL) {
1464 struct sk_filter *filter;
1465
892c141e 1466 sock_copy(newsk, sk);
87d11ceb
ACM		 1467
1468 /* SANITY */
3b1e0a65 1469 get_net(sock_net(newsk));
87d11ceb
ACM		 1470 sk_node_init(&newsk->sk_node);
1471 sock_lock_init(newsk);
1472 bh_lock_sock(newsk);
fa438ccf 1473 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1474 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1475
1476 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1477 /*
1478 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1479 */
1480 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1481 atomic_set(&newsk->sk_omem_alloc, 0);
1482 skb_queue_head_init(&newsk->sk_receive_queue);
1483 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1484#ifdef CONFIG_NET_DMA
1485 skb_queue_head_init(&newsk->sk_async_wait_queue);
1486#endif
87d11ceb 1487
b6c6712a 1488 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1489 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1490 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1491 af_callback_keys + newsk->sk_family,
1492 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1493
1494 newsk->sk_dst_cache = NULL;
1495 newsk->sk_wmem_queued = 0;
1496 newsk->sk_forward_alloc = 0;
1497 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1498 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1499
1500 sock_reset_flag(newsk, SOCK_DONE);
1501 skb_queue_head_init(&newsk->sk_error_queue);
1502
0d7da9dd 1503 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1504 if (filter != NULL)
278571ba
AS		 1505 /* though it's an empty new sock, the charging may fail
1506 * if sysctl_optmem_max was changed between creation of
1507 * original socket and cloning
1508 */
1509 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1510
278571ba 1511 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk))) {
87d11ceb
ACM		 1512 /* It is still raw copy of parent, so invalidate
1513 * destructor and make plain sk_free() */
1514 newsk->sk_destruct = NULL;
b0691c8e 1515 bh_unlock_sock(newsk);
87d11ceb
ACM		 1516 sk_free(newsk);
1517 newsk = NULL;
1518 goto out;
1519 }
1520
1521 newsk->sk_err = 0;
1522 newsk->sk_priority = 0;
4dc6dc71
ED		 1523 /*
1524 * Before updating sk_refcnt, we must commit prior changes to memory
1525 * (Documentation/RCU/rculist_nulls.txt for details)
1526 */
1527 smp_wmb();
87d11ceb
ACM		 1528 atomic_set(&newsk->sk_refcnt, 2);
1529
1530 /*
1531 * Increment the counter in the same struct proto as the master
1532 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1533 * is the same as sk->sk_prot->socks, as this field was copied
1534 * with memcpy).
1535 *
1536 * This _changes_ the previous behaviour, where
1537 * tcp_create_openreq_child always was incrementing the
1538 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1539 * to be taken into account in all callers. -acme
1540 */
1541 sk_refcnt_debug_inc(newsk);
972692e0 1542 sk_set_socket(newsk, NULL);
43815482 1543 newsk->sk_wq = NULL;
87d11ceb 1544
f3f511e1
GC		 1545 sk_update_clone(sk, newsk);
1546
87d11ceb 1547 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1548 sk_sockets_allocated_inc(newsk);
704da560 1549
08e29af3 1550 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1551 net_enable_timestamp();
87d11ceb
ACM		 1552 }
1553out:
1554 return newsk;
1555}
e56c57d0 1556EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1557
9958089a
AK		 1558void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1559{
1560 __sk_dst_set(sk, dst);
1561 sk->sk_route_caps = dst->dev->features;
1562 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1563 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1564 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1565 if (sk_can_gso(sk)) {
82cc1a7a 1566 if (dst->header_len) {
9958089a 1567 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1568 } else {
9958089a 1569 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1570 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1571 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1572 }
9958089a
AK		 1573 }
1574}
1575EXPORT_SYMBOL_GPL(sk_setup_caps);
1576
1da177e4
LT		 1577/*
1578 * Simple resource managers for sockets.
1579 */
1580
1581
4ec93edb
YH		 1582/*
1583 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1584 */
1585void sock_wfree(struct sk_buff *skb)
1586{
1587 struct sock *sk = skb->sk;
d99927f4 1588 unsigned int len = skb->truesize;
1da177e4 1589
d99927f4
ED		 1590 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1591 /*
1592 * Keep a reference on sk_wmem_alloc, this will be released
1593 * after sk_write_space() call
1594 */
1595 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1596 sk->sk_write_space(sk);
d99927f4
ED		 1597 len = 1;
1598 }
2b85a34e 1599 /*
d99927f4
ED		 1600 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1601 * could not do because of in-flight packets
2b85a34e 1602 */
d99927f4 1603 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1604 __sk_free(sk);
1da177e4 1605}
2a91525c 1606EXPORT_SYMBOL(sock_wfree);
1da177e4 1607
f2f872f9
ED		 1608void skb_orphan_partial(struct sk_buff *skb)
1609{
1610 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1611 * so we do not completely orphan skb, but transfert all
1612 * accounted bytes but one, to avoid unexpected reorders.
1613 */
1614 if (skb->destructor == sock_wfree
1615#ifdef CONFIG_INET
1616 || skb->destructor == tcp_wfree
1617#endif
1618 ) {
1619 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1620 skb->truesize = 1;
1621 } else {
1622 skb_orphan(skb);
1623 }
1624}
1625EXPORT_SYMBOL(skb_orphan_partial);
1626
4ec93edb
YH		 1627/*
1628 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1629 */
1630void sock_rfree(struct sk_buff *skb)
1631{
1632 struct sock *sk = skb->sk;
d361fd59 1633 unsigned int len = skb->truesize;
1da177e4 1634
d361fd59
ED		 1635 atomic_sub(len, &sk->sk_rmem_alloc);
1636 sk_mem_uncharge(sk, len);
1da177e4 1637}
2a91525c 1638EXPORT_SYMBOL(sock_rfree);
1da177e4 1639
62bccb8c
AD		 1640void sock_efree(struct sk_buff *skb)
1641{
1642 sock_put(skb->sk);
1643}
1644EXPORT_SYMBOL(sock_efree);
1645
82eabd9e 1646#ifdef CONFIG_INET
41063e9d
DM		 1647void sock_edemux(struct sk_buff *skb)
1648{
e812347c
ED		 1649 struct sock *sk = skb->sk;
1650
1651 if (sk->sk_state == TCP_TIME_WAIT)
1652 inet_twsk_put(inet_twsk(sk));
1653 else
1654 sock_put(sk);
41063e9d
DM		 1655}
1656EXPORT_SYMBOL(sock_edemux);
82eabd9e 1657#endif
1da177e4 1658
976d0201 1659kuid_t sock_i_uid(struct sock *sk)
1da177e4 1660{
976d0201 1661 kuid_t uid;
1da177e4 1662
f064af1e 1663 read_lock_bh(&sk->sk_callback_lock);
976d0201 1664 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1665 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1666 return uid;
1667}
2a91525c 1668EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1669
1670unsigned long sock_i_ino(struct sock *sk)
1671{
1672 unsigned long ino;
1673
f064af1e 1674 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1675 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1676 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1677 return ino;
1678}
2a91525c 1679EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1680
1681/*
1682 * Allocate a skb from the socket's send buffer.
1683 */
86a76caf 1684struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1685 gfp_t priority)
1da177e4
LT		 1686{
1687 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1688 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1689 if (skb) {
1690 skb_set_owner_w(skb, sk);
1691 return skb;
1692 }
1693 }
1694 return NULL;
1695}
2a91525c 1696EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1697
4ec93edb 1698/*
1da177e4 1699 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1700 */
dd0fc66f 1701void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1702{
95c96174 1703 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1704 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1705 void *mem;
1706 /* First do the add, to avoid the race if kmalloc
4ec93edb 1707 * might sleep.
1da177e4
LT		 1708 */
1709 atomic_add(size, &sk->sk_omem_alloc);
1710 mem = kmalloc(size, priority);
1711 if (mem)
1712 return mem;
1713 atomic_sub(size, &sk->sk_omem_alloc);
1714 }
1715 return NULL;
1716}
2a91525c 1717EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1718
1719/*
1720 * Free an option memory block.
1721 */
1722void sock_kfree_s(struct sock *sk, void *mem, int size)
1723{
1724 kfree(mem);
1725 atomic_sub(size, &sk->sk_omem_alloc);
1726}
2a91525c 1727EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1728
1729/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1730 I think, these locks should be removed for datagram sockets.
1731 */
2a91525c 1732static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1733{
1734 DEFINE_WAIT(wait);
1735
1736 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1737 for (;;) {
1738 if (!timeo)
1739 break;
1740 if (signal_pending(current))
1741 break;
1742 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1743 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1744 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1745 break;
1746 if (sk->sk_shutdown & SEND_SHUTDOWN)
1747 break;
1748 if (sk->sk_err)
1749 break;
1750 timeo = schedule_timeout(timeo);
1751 }
aa395145 1752 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1753 return timeo;
1754}
1755
1756
1757/*
1758 * Generic send/receive buffer handlers
1759 */
1760
4cc7f68d
HX		 1761struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1762 unsigned long data_len, int noblock,
28d64271 1763 int *errcode, int max_page_order)
1da177e4 1764{
28d64271
ED		 1765 struct sk_buff *skb = NULL;
1766 unsigned long chunk;
7d877f3b 1767 gfp_t gfp_mask;
1da177e4
LT		 1768 long timeo;
1769 int err;
cc9b17ad 1770 int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
28d64271
ED		 1771 struct page *page;
1772 int i;
cc9b17ad
JW		 1773
1774 err = -EMSGSIZE;
1775 if (npages > MAX_SKB_FRAGS)
1776 goto failure;
1da177e4 1777
1da177e4 1778 timeo = sock_sndtimeo(sk, noblock);
28d64271 1779 while (!skb) {
1da177e4
LT		 1780 err = sock_error(sk);
1781 if (err != 0)
1782 goto failure;
1783
1784 err = -EPIPE;
1785 if (sk->sk_shutdown & SEND_SHUTDOWN)
1786 goto failure;
1787
28d64271
ED		 1788 if (atomic_read(&sk->sk_wmem_alloc) >= sk->sk_sndbuf) {
1789 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1790 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1791 err = -EAGAIN;
1792 if (!timeo)
1793 goto failure;
1794 if (signal_pending(current))
1795 goto interrupted;
1796 timeo = sock_wait_for_wmem(sk, timeo);
1797 continue;
1798 }
1da177e4 1799
28d64271
ED		 1800 err = -ENOBUFS;
1801 gfp_mask = sk->sk_allocation;
1802 if (gfp_mask & __GFP_WAIT)
1803 gfp_mask |= __GFP_REPEAT;
1804
1805 skb = alloc_skb(header_len, gfp_mask);
1806 if (!skb)
1da177e4 1807 goto failure;
28d64271
ED		 1808
1809 skb->truesize += data_len;
1810
1811 for (i = 0; npages > 0; i++) {
1812 int order = max_page_order;
1813
1814 while (order) {
1815 if (npages >= 1 << order) {
1816 page = alloc_pages(sk->sk_allocation |
ed98df33
ED		 1817 __GFP_COMP |
1818 __GFP_NOWARN |
1819 __GFP_NORETRY,
28d64271
ED		 1820 order);
1821 if (page)
1822 goto fill_page;
d9b2938a
ED		 1823 /* Do not retry other high order allocations */
1824 order = 1;
1825 max_page_order = 0;
28d64271
ED		 1826 }
1827 order--;
1828 }
1829 page = alloc_page(sk->sk_allocation);
1830 if (!page)
1831 goto failure;
1832fill_page:
1833 chunk = min_t(unsigned long, data_len,
1834 PAGE_SIZE << order);
1835 skb_fill_page_desc(skb, i, page, 0, chunk);
1836 data_len -= chunk;
1837 npages -= 1 << order;
1da177e4 1838 }
1da177e4
LT		 1839 }
1840
1841 skb_set_owner_w(skb, sk);
1842 return skb;
1843
1844interrupted:
1845 err = sock_intr_errno(timeo);
1846failure:
28d64271 1847 kfree_skb(skb);
1da177e4
LT		 1848 *errcode = err;
1849 return NULL;
1850}
4cc7f68d 1851EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1852
4ec93edb 1853struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1854 int noblock, int *errcode)
1855{
28d64271 1856 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1857}
2a91525c 1858EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1859
5640f768
ED		 1860/* On 32bit arches, an skb frag is limited to 2^15 */
1861#define SKB_FRAG_PAGE_ORDER get_order(32768)
1862
400dfd3a
ED		 1863/**
1864 * skb_page_frag_refill - check that a page_frag contains enough room
1865 * @sz: minimum size of the fragment we want to get
1866 * @pfrag: pointer to page_frag
1867 * @prio: priority for memory allocation
1868 *
1869 * Note: While this allocator tries to use high order pages, there is
1870 * no guarantee that allocations succeed. Therefore, @sz MUST be
1871 * less or equal than PAGE_SIZE.
1872 */
d9b2938a 1873bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1874{
5640f768
ED		 1875 if (pfrag->page) {
1876 if (atomic_read(&pfrag->page->_count) == 1) {
1877 pfrag->offset = 0;
1878 return true;
1879 }
400dfd3a 1880 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1881 return true;
1882 put_page(pfrag->page);
1883 }
1884
d9b2938a
ED		 1885 pfrag->offset = 0;
1886 if (SKB_FRAG_PAGE_ORDER) {
1887 pfrag->page = alloc_pages(gfp | __GFP_COMP |
1888 __GFP_NOWARN | __GFP_NORETRY,
1889 SKB_FRAG_PAGE_ORDER);
5640f768 1890 if (likely(pfrag->page)) {
d9b2938a 1891 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1892 return true;
1893 }
d9b2938a
ED		 1894 }
1895 pfrag->page = alloc_page(gfp);
1896 if (likely(pfrag->page)) {
1897 pfrag->size = PAGE_SIZE;
1898 return true;
1899 }
400dfd3a
ED		 1900 return false;
1901}
1902EXPORT_SYMBOL(skb_page_frag_refill);
1903
1904bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1905{
1906 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1907 return true;
1908
5640f768
ED		 1909 sk_enter_memory_pressure(sk);
1910 sk_stream_moderate_sndbuf(sk);
1911 return false;
1912}
1913EXPORT_SYMBOL(sk_page_frag_refill);
1914
1da177e4 1915static void __lock_sock(struct sock *sk)
f39234d6
NK		 1916 __releases(&sk->sk_lock.slock)
1917 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1918{
1919 DEFINE_WAIT(wait);
1920
e71a4783 1921 for (;;) {
1da177e4
LT		 1922 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1923 TASK_UNINTERRUPTIBLE);
1924 spin_unlock_bh(&sk->sk_lock.slock);
1925 schedule();
1926 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1927 if (!sock_owned_by_user(sk))
1da177e4
LT		 1928 break;
1929 }
1930 finish_wait(&sk->sk_lock.wq, &wait);
1931}
1932
1933static void __release_sock(struct sock *sk)
f39234d6
NK		 1934 __releases(&sk->sk_lock.slock)
1935 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1936{
1937 struct sk_buff *skb = sk->sk_backlog.head;
1938
1939 do {
1940 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1941 bh_unlock_sock(sk);
1942
1943 do {
1944 struct sk_buff *next = skb->next;
1945
e4cbb02a 1946 prefetch(next);
7fee226a 1947 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1948 skb->next = NULL;
c57943a1 1949 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1950
1951 /*
1952 * We are in process context here with softirqs
1953 * disabled, use cond_resched_softirq() to preempt.
1954 * This is safe to do because we've taken the backlog
1955 * queue private:
1956 */
1957 cond_resched_softirq();
1958
1959 skb = next;
1960 } while (skb != NULL);
1961
1962 bh_lock_sock(sk);
e71a4783 1963 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1964
1965 /*
1966 * Doing the zeroing here guarantee we can not loop forever
1967 * while a wild producer attempts to flood us.
1968 */
1969 sk->sk_backlog.len = 0;
1da177e4
LT		 1970}
1971
1972/**
1973 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1974 * @sk: sock to wait on
1975 * @timeo: for how long
1da177e4
LT		 1976 *
1977 * Now socket state including sk->sk_err is changed only under lock,
1978 * hence we may omit checks after joining wait queue.
1979 * We check receive queue before schedule() only as optimization;
1980 * it is very likely that release_sock() added new data.
1981 */
1982int sk_wait_data(struct sock *sk, long *timeo)
1983{
1984 int rc;
1985 DEFINE_WAIT(wait);
1986
aa395145 1987 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1988 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1989 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1990 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 1991 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1992 return rc;
1993}
1da177e4
LT		 1994EXPORT_SYMBOL(sk_wait_data);
1995
3ab224be
HA		 1996/**
1997 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1998 * @sk: socket
1999 * @size: memory size to allocate
2000 * @kind: allocation type
2001 *
2002 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2003 * rmem allocation. This function assumes that protocols which have
2004 * memory_pressure use sk_wmem_queued as write buffer accounting.
2005 */
2006int __sk_mem_schedule(struct sock *sk, int size, int kind)
2007{
2008 struct proto *prot = sk->sk_prot;
2009 int amt = sk_mem_pages(size);
8d987e5c 2010 long allocated;
e1aab161 2011 int parent_status = UNDER_LIMIT;
3ab224be
HA		 2012
2013 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2014
e1aab161 2015 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 2016
2017 /* Under limit. */
e1aab161
GC		 2018 if (parent_status == UNDER_LIMIT &&
2019 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2020 sk_leave_memory_pressure(sk);
3ab224be
HA		 2021 return 1;
2022 }
2023
e1aab161
GC		 2024 /* Under pressure. (we or our parents) */
2025 if ((parent_status > SOFT_LIMIT) ||
2026 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2027 sk_enter_memory_pressure(sk);
3ab224be 2028
e1aab161
GC		 2029 /* Over hard limit (we or our parents) */
2030 if ((parent_status == OVER_LIMIT) ||
2031 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 2032 goto suppress_allocation;
2033
2034 /* guarantee minimum buffer size under pressure */
2035 if (kind == SK_MEM_RECV) {
2036 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2037 return 1;
180d8cd9 2038
3ab224be
HA		 2039 } else { /* SK_MEM_SEND */
2040 if (sk->sk_type == SOCK_STREAM) {
2041 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2042 return 1;
2043 } else if (atomic_read(&sk->sk_wmem_alloc) <
2044 prot->sysctl_wmem[0])
2045 return 1;
2046 }
2047
180d8cd9 2048 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2049 int alloc;
2050
180d8cd9 2051 if (!sk_under_memory_pressure(sk))
1748376b 2052 return 1;
180d8cd9
GC		 2053 alloc = sk_sockets_allocated_read_positive(sk);
2054 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2055 sk_mem_pages(sk->sk_wmem_queued +
2056 atomic_read(&sk->sk_rmem_alloc) +
2057 sk->sk_forward_alloc))
2058 return 1;
2059 }
2060
2061suppress_allocation:
2062
2063 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2064 sk_stream_moderate_sndbuf(sk);
2065
2066 /* Fail only if socket is _under_ its sndbuf.
2067 * In this case we cannot block, so that we have to fail.
2068 */
2069 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2070 return 1;
2071 }
2072
3847ce32
SM		 2073 trace_sock_exceed_buf_limit(sk, prot, allocated);
2074
3ab224be
HA		 2075 /* Alas. Undo changes. */
2076 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2077
0e90b31f 2078 sk_memory_allocated_sub(sk, amt);
180d8cd9 2079
3ab224be
HA		 2080 return 0;
2081}
3ab224be
HA		 2082EXPORT_SYMBOL(__sk_mem_schedule);
2083
2084/**
2085 * __sk_reclaim - reclaim memory_allocated
2086 * @sk: socket
2087 */
2088void __sk_mem_reclaim(struct sock *sk)
2089{
180d8cd9 2090 sk_memory_allocated_sub(sk,
0e90b31f 2091 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 2092 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
2093
180d8cd9
GC		 2094 if (sk_under_memory_pressure(sk) &&
2095 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2096 sk_leave_memory_pressure(sk);
3ab224be 2097}
3ab224be
HA		 2098EXPORT_SYMBOL(__sk_mem_reclaim);
2099
2100
1da177e4
LT		 2101/*
2102 * Set of default routines for initialising struct proto_ops when
2103 * the protocol does not support a particular function. In certain
2104 * cases where it makes no sense for a protocol to have a "do nothing"
2105 * function, some default processing is provided.
2106 */
2107
2108int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2109{
2110 return -EOPNOTSUPP;
2111}
2a91525c 2112EXPORT_SYMBOL(sock_no_bind);
1da177e4 2113
4ec93edb 2114int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2115 int len, int flags)
2116{
2117 return -EOPNOTSUPP;
2118}
2a91525c 2119EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2120
2121int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2122{
2123 return -EOPNOTSUPP;
2124}
2a91525c 2125EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2126
2127int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2128{
2129 return -EOPNOTSUPP;
2130}
2a91525c 2131EXPORT_SYMBOL(sock_no_accept);
1da177e4 2132
4ec93edb 2133int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2134 int *len, int peer)
2135{
2136 return -EOPNOTSUPP;
2137}
2a91525c 2138EXPORT_SYMBOL(sock_no_getname);
1da177e4 2139
2a91525c 2140unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2141{
2142 return 0;
2143}
2a91525c 2144EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2145
2146int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2147{
2148 return -EOPNOTSUPP;
2149}
2a91525c 2150EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2151
2152int sock_no_listen(struct socket *sock, int backlog)
2153{
2154 return -EOPNOTSUPP;
2155}
2a91525c 2156EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2157
2158int sock_no_shutdown(struct socket *sock, int how)
2159{
2160 return -EOPNOTSUPP;
2161}
2a91525c 2162EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2163
2164int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2165 char __user *optval, unsigned int optlen)
1da177e4
LT		 2166{
2167 return -EOPNOTSUPP;
2168}
2a91525c 2169EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2170
2171int sock_no_getsockopt(struct socket *sock, int level, int optname,
2172 char __user *optval, int __user *optlen)
2173{
2174 return -EOPNOTSUPP;
2175}
2a91525c 2176EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2177
2178int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2179 size_t len)
2180{
2181 return -EOPNOTSUPP;
2182}
2a91525c 2183EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2184
2185int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2186 size_t len, int flags)
2187{
2188 return -EOPNOTSUPP;
2189}
2a91525c 2190EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2191
2192int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2193{
2194 /* Mirror missing mmap method error code */
2195 return -ENODEV;
2196}
2a91525c 2197EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2198
2199ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2200{
2201 ssize_t res;
2202 struct msghdr msg = {.msg_flags = flags};
2203 struct kvec iov;
2204 char *kaddr = kmap(page);
2205 iov.iov_base = kaddr + offset;
2206 iov.iov_len = size;
2207 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2208 kunmap(page);
2209 return res;
2210}
2a91525c 2211EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2212
2213/*
2214 * Default Socket Callbacks
2215 */
2216
2217static void sock_def_wakeup(struct sock *sk)
2218{
43815482
ED		 2219 struct socket_wq *wq;
2220
2221 rcu_read_lock();
2222 wq = rcu_dereference(sk->sk_wq);
2223 if (wq_has_sleeper(wq))
2224 wake_up_interruptible_all(&wq->wait);
2225 rcu_read_unlock();
1da177e4
LT		 2226}
2227
2228static void sock_def_error_report(struct sock *sk)
2229{
43815482
ED		 2230 struct socket_wq *wq;
2231
2232 rcu_read_lock();
2233 wq = rcu_dereference(sk->sk_wq);
2234 if (wq_has_sleeper(wq))
2235 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2236 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2237 rcu_read_unlock();
1da177e4
LT		 2238}
2239
676d2369 2240static void sock_def_readable(struct sock *sk)
1da177e4 2241{
43815482
ED		 2242 struct socket_wq *wq;
2243
2244 rcu_read_lock();
2245 wq = rcu_dereference(sk->sk_wq);
2246 if (wq_has_sleeper(wq))
2c6607c6 2247 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2248 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2249 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2250 rcu_read_unlock();
1da177e4
LT		 2251}
2252
2253static void sock_def_write_space(struct sock *sk)
2254{
43815482
ED		 2255 struct socket_wq *wq;
2256
2257 rcu_read_lock();
1da177e4
LT		 2258
2259 /* Do not wake up a writer until he can make "significant"
2260 * progress. --DaveM
2261 */
e71a4783 2262 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2263 wq = rcu_dereference(sk->sk_wq);
2264 if (wq_has_sleeper(wq))
2265 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2266 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2267
2268 /* Should agree with poll, otherwise some programs break */
2269 if (sock_writeable(sk))
8d8ad9d7 2270 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2271 }
2272
43815482 2273 rcu_read_unlock();
1da177e4
LT		 2274}
2275
2276static void sock_def_destruct(struct sock *sk)
2277{
a51482bd 2278 kfree(sk->sk_protinfo);
1da177e4
LT		 2279}
2280
2281void sk_send_sigurg(struct sock *sk)
2282{
2283 if (sk->sk_socket && sk->sk_socket->file)
2284 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2285 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2286}
2a91525c 2287EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2288
2289void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2290 unsigned long expires)
2291{
2292 if (!mod_timer(timer, expires))
2293 sock_hold(sk);
2294}
1da177e4
LT		 2295EXPORT_SYMBOL(sk_reset_timer);
2296
2297void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2298{
25cc4ae9 2299 if (del_timer(timer))
1da177e4
LT		 2300 __sock_put(sk);
2301}
1da177e4
LT		 2302EXPORT_SYMBOL(sk_stop_timer);
2303
2304void sock_init_data(struct socket *sock, struct sock *sk)
2305{
2306 skb_queue_head_init(&sk->sk_receive_queue);
2307 skb_queue_head_init(&sk->sk_write_queue);
2308 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2309#ifdef CONFIG_NET_DMA
2310 skb_queue_head_init(&sk->sk_async_wait_queue);
2311#endif
1da177e4
LT		 2312
2313 sk->sk_send_head = NULL;
2314
2315 init_timer(&sk->sk_timer);
4ec93edb 2316
1da177e4
LT		 2317 sk->sk_allocation = GFP_KERNEL;
2318 sk->sk_rcvbuf = sysctl_rmem_default;
2319 sk->sk_sndbuf = sysctl_wmem_default;
2320 sk->sk_state = TCP_CLOSE;
972692e0 2321 sk_set_socket(sk, sock);
1da177e4
LT		 2322
2323 sock_set_flag(sk, SOCK_ZAPPED);
2324
e71a4783 2325 if (sock) {
1da177e4 2326 sk->sk_type = sock->type;
43815482 2327 sk->sk_wq = sock->wq;
1da177e4
LT		 2328 sock->sk = sk;
2329 } else
43815482 2330 sk->sk_wq = NULL;
1da177e4 2331
b6c6712a 2332 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2333 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2334 lockdep_set_class_and_name(&sk->sk_callback_lock,
2335 af_callback_keys + sk->sk_family,
2336 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2337
2338 sk->sk_state_change = sock_def_wakeup;
2339 sk->sk_data_ready = sock_def_readable;
2340 sk->sk_write_space = sock_def_write_space;
2341 sk->sk_error_report = sock_def_error_report;
2342 sk->sk_destruct = sock_def_destruct;
2343
5640f768
ED		 2344 sk->sk_frag.page = NULL;
2345 sk->sk_frag.offset = 0;
ef64a54f 2346 sk->sk_peek_off = -1;
1da177e4 2347
109f6e39
EB		 2348 sk->sk_peer_pid = NULL;
2349 sk->sk_peer_cred = NULL;
1da177e4
LT		 2350 sk->sk_write_pending = 0;
2351 sk->sk_rcvlowat = 1;
2352 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2353 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2354
f37f0afb 2355 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2356
e0d1095a 2357#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2358 sk->sk_napi_id = 0;
64b0dc51 2359 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2360#endif
2361
62748f32 2362 sk->sk_max_pacing_rate = ~0U;
7eec4174 2363 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2364 /*
2365 * Before updating sk_refcnt, we must commit prior changes to memory
2366 * (Documentation/RCU/rculist_nulls.txt for details)
2367 */
2368 smp_wmb();
1da177e4 2369 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2370 atomic_set(&sk->sk_drops, 0);
1da177e4 2371}
2a91525c 2372EXPORT_SYMBOL(sock_init_data);
1da177e4 2373
b5606c2d 2374void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2375{
2376 might_sleep();
a5b5bb9a 2377 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2378 if (sk->sk_lock.owned)
1da177e4 2379 __lock_sock(sk);
d2e9117c 2380 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2381 spin_unlock(&sk->sk_lock.slock);
2382 /*
2383 * The sk_lock has mutex_lock() semantics here:
2384 */
fcc70d5f 2385 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2386 local_bh_enable();
1da177e4 2387}
fcc70d5f 2388EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2389
b5606c2d 2390void release_sock(struct sock *sk)
1da177e4 2391{
a5b5bb9a
IM		 2392 /*
2393 * The sk_lock has mutex_unlock() semantics:
2394 */
2395 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2396
2397 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2398 if (sk->sk_backlog.tail)
2399 __release_sock(sk);
46d3ceab 2400
c3f9b018
ED		 2401 /* Warning : release_cb() might need to release sk ownership,
2402 * ie call sock_release_ownership(sk) before us.
2403 */
46d3ceab
ED		 2404 if (sk->sk_prot->release_cb)
2405 sk->sk_prot->release_cb(sk);
2406
c3f9b018 2407 sock_release_ownership(sk);
a5b5bb9a
IM		 2408 if (waitqueue_active(&sk->sk_lock.wq))
2409 wake_up(&sk->sk_lock.wq);
2410 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2411}
2412EXPORT_SYMBOL(release_sock);
2413
8a74ad60
ED		 2414/**
2415 * lock_sock_fast - fast version of lock_sock
2416 * @sk: socket
2417 *
2418 * This version should be used for very small section, where process wont block
2419 * return false if fast path is taken
2420 * sk_lock.slock locked, owned = 0, BH disabled
2421 * return true if slow path is taken
2422 * sk_lock.slock unlocked, owned = 1, BH enabled
2423 */
2424bool lock_sock_fast(struct sock *sk)
2425{
2426 might_sleep();
2427 spin_lock_bh(&sk->sk_lock.slock);
2428
2429 if (!sk->sk_lock.owned)
2430 /*
2431 * Note : We must disable BH
2432 */
2433 return false;
2434
2435 __lock_sock(sk);
2436 sk->sk_lock.owned = 1;
2437 spin_unlock(&sk->sk_lock.slock);
2438 /*
2439 * The sk_lock has mutex_lock() semantics here:
2440 */
2441 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2442 local_bh_enable();
2443 return true;
2444}
2445EXPORT_SYMBOL(lock_sock_fast);
2446
1da177e4 2447int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2448{
b7aa0bf7 2449 struct timeval tv;
1da177e4 2450 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2451 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2452 tv = ktime_to_timeval(sk->sk_stamp);
2453 if (tv.tv_sec == -1)
1da177e4 2454 return -ENOENT;
b7aa0bf7
ED		 2455 if (tv.tv_sec == 0) {
2456 sk->sk_stamp = ktime_get_real();
2457 tv = ktime_to_timeval(sk->sk_stamp);
2458 }
2459 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2460}
1da177e4
LT		 2461EXPORT_SYMBOL(sock_get_timestamp);
2462
ae40eb1e
ED		 2463int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2464{
2465 struct timespec ts;
2466 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2467 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2468 ts = ktime_to_timespec(sk->sk_stamp);
2469 if (ts.tv_sec == -1)
2470 return -ENOENT;
2471 if (ts.tv_sec == 0) {
2472 sk->sk_stamp = ktime_get_real();
2473 ts = ktime_to_timespec(sk->sk_stamp);
2474 }
2475 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2476}
2477EXPORT_SYMBOL(sock_get_timestampns);
2478
20d49473 2479void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2480{
20d49473 2481 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2482 unsigned long previous_flags = sk->sk_flags;
2483
20d49473
PO		 2484 sock_set_flag(sk, flag);
2485 /*
2486 * we just set one of the two flags which require net
2487 * time stamping, but time stamping might have been on
2488 * already because of the other one
2489 */
08e29af3 2490 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2491 net_enable_timestamp();
1da177e4
LT		 2492 }
2493}
1da177e4 2494
cb820f8e
RC		 2495int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2496 int level, int type)
2497{
2498 struct sock_exterr_skb *serr;
364a9e93 2499 struct sk_buff *skb;
cb820f8e
RC		 2500 int copied, err;
2501
2502 err = -EAGAIN;
364a9e93 2503 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2504 if (skb == NULL)
2505 goto out;
2506
2507 copied = skb->len;
2508 if (copied > len) {
2509 msg->msg_flags |= MSG_TRUNC;
2510 copied = len;
2511 }
2512 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2513 if (err)
2514 goto out_free_skb;
2515
2516 sock_recv_timestamp(msg, sk, skb);
2517
2518 serr = SKB_EXT_ERR(skb);
2519 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2520
2521 msg->msg_flags |= MSG_ERRQUEUE;
2522 err = copied;
2523
cb820f8e
RC		 2524out_free_skb:
2525 kfree_skb(skb);
2526out:
2527 return err;
2528}
2529EXPORT_SYMBOL(sock_recv_errqueue);
2530
1da177e4
LT		 2531/*
2532 * Get a socket option on an socket.
2533 *
2534 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2535 * asynchronous errors should be reported by getsockopt. We assume
2536 * this means if you specify SO_ERROR (otherwise whats the point of it).
2537 */
2538int sock_common_getsockopt(struct socket *sock, int level, int optname,
2539 char __user *optval, int __user *optlen)
2540{
2541 struct sock *sk = sock->sk;
2542
2543 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2544}
1da177e4
LT		 2545EXPORT_SYMBOL(sock_common_getsockopt);
2546
3fdadf7d 2547#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2548int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2549 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2550{
2551 struct sock *sk = sock->sk;
2552
1e51f951 2553 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2554 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2555 optval, optlen);
3fdadf7d
DM		 2556 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2557}
2558EXPORT_SYMBOL(compat_sock_common_getsockopt);
2559#endif
2560
1da177e4
LT		 2561int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2562 struct msghdr *msg, size_t size, int flags)
2563{
2564 struct sock *sk = sock->sk;
2565 int addr_len = 0;
2566 int err;
2567
2568 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2569 flags & ~MSG_DONTWAIT, &addr_len);
2570 if (err >= 0)
2571 msg->msg_namelen = addr_len;
2572 return err;
2573}
1da177e4
LT		 2574EXPORT_SYMBOL(sock_common_recvmsg);
2575
2576/*
2577 * Set socket options on an inet socket.
2578 */
2579int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2580 char __user *optval, unsigned int optlen)
1da177e4
LT		 2581{
2582 struct sock *sk = sock->sk;
2583
2584 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2585}
1da177e4
LT		 2586EXPORT_SYMBOL(sock_common_setsockopt);
2587
3fdadf7d 2588#ifdef CONFIG_COMPAT
543d9cfe 2589int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2590 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2591{
2592 struct sock *sk = sock->sk;
2593
543d9cfe
ACM		 2594 if (sk->sk_prot->compat_setsockopt != NULL)
2595 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2596 optval, optlen);
3fdadf7d
DM		 2597 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2598}
2599EXPORT_SYMBOL(compat_sock_common_setsockopt);
2600#endif
2601
1da177e4
LT		 2602void sk_common_release(struct sock *sk)
2603{
2604 if (sk->sk_prot->destroy)
2605 sk->sk_prot->destroy(sk);
2606
2607 /*
2608 * Observation: when sock_common_release is called, processes have
2609 * no access to socket. But net still has.
2610 * Step one, detach it from networking:
2611 *
2612 * A. Remove from hash tables.
2613 */
2614
2615 sk->sk_prot->unhash(sk);
2616
2617 /*
2618 * In this point socket cannot receive new packets, but it is possible
2619 * that some packets are in flight because some CPU runs receiver and
2620 * did hash table lookup before we unhashed socket. They will achieve
2621 * receive queue and will be purged by socket destructor.
2622 *
2623 * Also we still have packets pending on receive queue and probably,
2624 * our own packets waiting in device queues. sock_destroy will drain
2625 * receive queue, but transmitted packets will delay socket destruction
2626 * until the last reference will be released.
2627 */
2628
2629 sock_orphan(sk);
2630
2631 xfrm_sk_free_policy(sk);
2632
e6848976 2633 sk_refcnt_debug_release(sk);
5640f768
ED		 2634
2635 if (sk->sk_frag.page) {
2636 put_page(sk->sk_frag.page);
2637 sk->sk_frag.page = NULL;
2638 }
2639
1da177e4
LT		 2640 sock_put(sk);
2641}
1da177e4
LT		 2642EXPORT_SYMBOL(sk_common_release);
2643
13ff3d6f
PE		 2644#ifdef CONFIG_PROC_FS
2645#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2646struct prot_inuse {
2647 int val[PROTO_INUSE_NR];
2648};
13ff3d6f
PE		 2649
2650static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2651
2652#ifdef CONFIG_NET_NS
2653void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2654{
d6d9ca0f 2655 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2656}
2657EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2658
2659int sock_prot_inuse_get(struct net *net, struct proto *prot)
2660{
2661 int cpu, idx = prot->inuse_idx;
2662 int res = 0;
2663
2664 for_each_possible_cpu(cpu)
2665 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2666
2667 return res >= 0 ? res : 0;
2668}
2669EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2670
2c8c1e72 2671static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2672{
2673 net->core.inuse = alloc_percpu(struct prot_inuse);
2674 return net->core.inuse ? 0 : -ENOMEM;
2675}
2676
2c8c1e72 2677static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2678{
2679 free_percpu(net->core.inuse);
2680}
2681
2682static struct pernet_operations net_inuse_ops = {
2683 .init = sock_inuse_init_net,
2684 .exit = sock_inuse_exit_net,
2685};
2686
2687static __init int net_inuse_init(void)
2688{
2689 if (register_pernet_subsys(&net_inuse_ops))
2690 panic("Cannot initialize net inuse counters");
2691
2692 return 0;
2693}
2694
2695core_initcall(net_inuse_init);
2696#else
1338d466
PE		 2697static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2698
c29a0bc4 2699void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2700{
d6d9ca0f 2701 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2702}
2703EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2704
c29a0bc4 2705int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2706{
2707 int cpu, idx = prot->inuse_idx;
2708 int res = 0;
2709
2710 for_each_possible_cpu(cpu)
2711 res += per_cpu(prot_inuse, cpu).val[idx];
2712
2713 return res >= 0 ? res : 0;
2714}
2715EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2716#endif
13ff3d6f
PE		 2717
2718static void assign_proto_idx(struct proto *prot)
2719{
2720 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2721
2722 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2723 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2724 return;
2725 }
2726
2727 set_bit(prot->inuse_idx, proto_inuse_idx);
2728}
2729
2730static void release_proto_idx(struct proto *prot)
2731{
2732 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2733 clear_bit(prot->inuse_idx, proto_inuse_idx);
2734}
2735#else
2736static inline void assign_proto_idx(struct proto *prot)
2737{
2738}
2739
2740static inline void release_proto_idx(struct proto *prot)
2741{
2742}
2743#endif
2744
b733c007
PE		 2745int proto_register(struct proto *prot, int alloc_slab)
2746{
1da177e4
LT		 2747 if (alloc_slab) {
2748 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2749 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2750 NULL);
1da177e4
LT		 2751
2752 if (prot->slab == NULL) {
e005d193
JP		 2753 pr_crit("%s: Can't create sock SLAB cache!\n",
2754 prot->name);
60e7663d 2755 goto out;
1da177e4 2756 }
2e6599cb
ACM		 2757
2758 if (prot->rsk_prot != NULL) {
faf23422 2759 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2760 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2761 goto out_free_sock_slab;
2762
7e56b5d6 2763 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2764 prot->rsk_prot->obj_size, 0,
20c2df83 2765 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2766
2767 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2768 pr_crit("%s: Can't create request sock SLAB cache!\n",
2769 prot->name);
2e6599cb
ACM		 2770 goto out_free_request_sock_slab_name;
2771 }
2772 }
8feaf0c0 2773
6d6ee43e 2774 if (prot->twsk_prot != NULL) {
faf23422 2775 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2776
7e56b5d6 2777 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2778 goto out_free_request_sock_slab;
2779
6d6ee43e 2780 prot->twsk_prot->twsk_slab =
7e56b5d6 2781 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2782 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2783 0,
2784 SLAB_HWCACHE_ALIGN |
2785 prot->slab_flags,
20c2df83 2786 NULL);
6d6ee43e 2787 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2788 goto out_free_timewait_sock_slab_name;
2789 }
1da177e4
LT		 2790 }
2791
36b77a52 2792 mutex_lock(&proto_list_mutex);
1da177e4 2793 list_add(&prot->node, &proto_list);
13ff3d6f 2794 assign_proto_idx(prot);
36b77a52 2795 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2796 return 0;
2797
8feaf0c0 2798out_free_timewait_sock_slab_name:
7e56b5d6 2799 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2800out_free_request_sock_slab:
2801 if (prot->rsk_prot && prot->rsk_prot->slab) {
2802 kmem_cache_destroy(prot->rsk_prot->slab);
2803 prot->rsk_prot->slab = NULL;
2804 }
2e6599cb 2805out_free_request_sock_slab_name:
72150e9b
DC		 2806 if (prot->rsk_prot)
2807 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2808out_free_sock_slab:
2809 kmem_cache_destroy(prot->slab);
2810 prot->slab = NULL;
b733c007
PE		 2811out:
2812 return -ENOBUFS;
1da177e4 2813}
1da177e4
LT		 2814EXPORT_SYMBOL(proto_register);
2815
2816void proto_unregister(struct proto *prot)
2817{
36b77a52 2818 mutex_lock(&proto_list_mutex);
13ff3d6f 2819 release_proto_idx(prot);
0a3f4358 2820 list_del(&prot->node);
36b77a52 2821 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2822
2823 if (prot->slab != NULL) {
2824 kmem_cache_destroy(prot->slab);
2825 prot->slab = NULL;
2826 }
2827
2e6599cb 2828 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2829 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2830 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2831 prot->rsk_prot->slab = NULL;
2832 }
2833
6d6ee43e 2834 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2835 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2836 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2837 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2838 }
1da177e4 2839}
1da177e4
LT		 2840EXPORT_SYMBOL(proto_unregister);
2841
2842#ifdef CONFIG_PROC_FS
1da177e4 2843static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2844 __acquires(proto_list_mutex)
1da177e4 2845{
36b77a52 2846 mutex_lock(&proto_list_mutex);
60f0438a 2847 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2848}
2849
2850static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2851{
60f0438a 2852 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2853}
2854
2855static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2856 __releases(proto_list_mutex)
1da177e4 2857{
36b77a52 2858 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2859}
2860
2861static char proto_method_implemented(const void *method)
2862{
2863 return method == NULL ? 'n' : 'y';
2864}
180d8cd9
GC		 2865static long sock_prot_memory_allocated(struct proto *proto)
2866{
cb75a36c 2867 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2868}
2869
2870static char *sock_prot_memory_pressure(struct proto *proto)
2871{
2872 return proto->memory_pressure != NULL ?
2873 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2874}
1da177e4
LT		 2875
2876static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2877{
180d8cd9 2878
8d987e5c 2879 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2880 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2881 proto->name,
2882 proto->obj_size,
14e943db 2883 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2884 sock_prot_memory_allocated(proto),
2885 sock_prot_memory_pressure(proto),
1da177e4
LT		 2886 proto->max_header,
2887 proto->slab == NULL ? "no" : "yes",
2888 module_name(proto->owner),
2889 proto_method_implemented(proto->close),
2890 proto_method_implemented(proto->connect),
2891 proto_method_implemented(proto->disconnect),
2892 proto_method_implemented(proto->accept),
2893 proto_method_implemented(proto->ioctl),
2894 proto_method_implemented(proto->init),
2895 proto_method_implemented(proto->destroy),
2896 proto_method_implemented(proto->shutdown),
2897 proto_method_implemented(proto->setsockopt),
2898 proto_method_implemented(proto->getsockopt),
2899 proto_method_implemented(proto->sendmsg),
2900 proto_method_implemented(proto->recvmsg),
2901 proto_method_implemented(proto->sendpage),
2902 proto_method_implemented(proto->bind),
2903 proto_method_implemented(proto->backlog_rcv),
2904 proto_method_implemented(proto->hash),
2905 proto_method_implemented(proto->unhash),
2906 proto_method_implemented(proto->get_port),
2907 proto_method_implemented(proto->enter_memory_pressure));
2908}
2909
2910static int proto_seq_show(struct seq_file *seq, void *v)
2911{
60f0438a 2912 if (v == &proto_list)
1da177e4
LT		 2913 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2914 "protocol",
2915 "size",
2916 "sockets",
2917 "memory",
2918 "press",
2919 "maxhdr",
2920 "slab",
2921 "module",
2922 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2923 else
60f0438a 2924 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2925 return 0;
2926}
2927
f690808e 2928static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2929 .start = proto_seq_start,
2930 .next = proto_seq_next,
2931 .stop = proto_seq_stop,
2932 .show = proto_seq_show,
2933};
2934
2935static int proto_seq_open(struct inode *inode, struct file *file)
2936{
14e943db
ED		 2937 return seq_open_net(inode, file, &proto_seq_ops,
2938 sizeof(struct seq_net_private));
1da177e4
LT		 2939}
2940
9a32144e 2941static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2942 .owner = THIS_MODULE,
2943 .open = proto_seq_open,
2944 .read = seq_read,
2945 .llseek = seq_lseek,
14e943db
ED		 2946 .release = seq_release_net,
2947};
2948
2949static __net_init int proto_init_net(struct net *net)
2950{
d4beaa66 2951 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2952 return -ENOMEM;
2953
2954 return 0;
2955}
2956
2957static __net_exit void proto_exit_net(struct net *net)
2958{
ece31ffd 2959 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2960}
2961
2962
2963static __net_initdata struct pernet_operations proto_net_ops = {
2964 .init = proto_init_net,
2965 .exit = proto_exit_net,
1da177e4
LT		 2966};
2967
2968static int __init proto_init(void)
2969{
14e943db 2970 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2971}
2972
2973subsys_initcall(proto_init);
2974
2975#endif /* PROC_FS */
ubuntu-zesty-kernel mirror