]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - net/core/sock.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: Pass kern from net_proto_family.create to sk_alloc
[mirror_ubuntu-bionic-kernel.git] / net / core / sock.c
CommitLineData
1da177e4
LT		 1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Generic socket support routines. Memory allocators, socket lock/release
7 * handler for protocols to use and generic option handler.
8 *
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT		 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
14 *
15 * Fixes:
16 * Alan Cox : Numerous verify_area() problems
17 * Alan Cox : Connecting on a connecting socket
18 * now returns an error for tcp.
19 * Alan Cox : sock->protocol is set correctly.
20 * and is not sometimes left as 0.
21 * Alan Cox : connect handles icmp errors on a
22 * connect properly. Unfortunately there
23 * is a restart syscall nasty there. I
24 * can't match BSD without hacking the C
25 * library. Ideas urgently sought!
26 * Alan Cox : Disallow bind() to addresses that are
27 * not ours - especially broadcast ones!!
28 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost)
29 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets,
30 * instead they leave that for the DESTROY timer.
31 * Alan Cox : Clean up error flag in accept
32 * Alan Cox : TCP ack handling is buggy, the DESTROY timer
33 * was buggy. Put a remove_sock() in the handler
34 * for memory when we hit 0. Also altered the timer
4ec93edb 35 * code. The ACK stuff can wait and needs major
1da177e4
LT		 36 * TCP layer surgery.
37 * Alan Cox : Fixed TCP ack bug, removed remove sock
38 * and fixed timer/inet_bh race.
39 * Alan Cox : Added zapped flag for TCP
40 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code
41 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb
42 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources
43 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing.
44 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so...
45 * Rick Sladkey : Relaxed UDP rules for matching packets.
46 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support
47 * Pauline Middelink : identd support
48 * Alan Cox : Fixed connect() taking signals I think.
49 * Alan Cox : SO_LINGER supported
50 * Alan Cox : Error reporting fixes
51 * Anonymous : inet_create tidied up (sk->reuse setting)
52 * Alan Cox : inet sockets don't set sk->type!
53 * Alan Cox : Split socket option code
54 * Alan Cox : Callbacks
55 * Alan Cox : Nagle flag for Charles & Johannes stuff
56 * Alex : Removed restriction on inet fioctl
57 * Alan Cox : Splitting INET from NET core
58 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt()
59 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code
60 * Alan Cox : Split IP from generic code
61 * Alan Cox : New kfree_skbmem()
62 * Alan Cox : Make SO_DEBUG superuser only.
63 * Alan Cox : Allow anyone to clear SO_DEBUG
64 * (compatibility fix)
65 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput.
66 * Alan Cox : Allocator for a socket is settable.
67 * Alan Cox : SO_ERROR includes soft errors.
68 * Alan Cox : Allow NULL arguments on some SO_ opts
69 * Alan Cox : Generic socket allocation to make hooks
70 * easier (suggested by Craig Metz).
71 * Michael Pall : SO_ERROR returns positive errno again
72 * Steve Whitehouse: Added default destructor to free
73 * protocol private data.
74 * Steve Whitehouse: Added various other default routines
75 * common to several socket families.
76 * Chris Evans : Call suser() check last on F_SETOWN
77 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER.
78 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s()
79 * Andi Kleen : Fix write_space callback
80 * Chris Evans : Security fixes - signedness again
81 * Arnaldo C. Melo : cleanups, use skb_queue_purge
82 *
83 * To Fix:
84 *
85 *
86 * This program is free software; you can redistribute it and/or
87 * modify it under the terms of the GNU General Public License
88 * as published by the Free Software Foundation; either version
89 * 2 of the License, or (at your option) any later version.
90 */
91
e005d193
JP		 92#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
93
4fc268d2 94#include <linux/capability.h>
1da177e4 95#include <linux/errno.h>
cb820f8e 96#include <linux/errqueue.h>
1da177e4
LT		 97#include <linux/types.h>
98#include <linux/socket.h>
99#include <linux/in.h>
100#include <linux/kernel.h>
1da177e4
LT		 101#include <linux/module.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
104#include <linux/sched.h>
105#include <linux/timer.h>
106#include <linux/string.h>
107#include <linux/sockios.h>
108#include <linux/net.h>
109#include <linux/mm.h>
110#include <linux/slab.h>
111#include <linux/interrupt.h>
112#include <linux/poll.h>
113#include <linux/tcp.h>
114#include <linux/init.h>
a1f8e7f7 115#include <linux/highmem.h>
3f551f94 116#include <linux/user_namespace.h>
c5905afb 117#include <linux/static_key.h>
3969eb38 118#include <linux/memcontrol.h>
8c1ae10d 119#include <linux/prefetch.h>
1da177e4
LT		 120
121#include <asm/uaccess.h>
1da177e4
LT		 122
123#include <linux/netdevice.h>
124#include <net/protocol.h>
125#include <linux/skbuff.h>
457c4cbc 126#include <net/net_namespace.h>
2e6599cb 127#include <net/request_sock.h>
1da177e4 128#include <net/sock.h>
20d49473 129#include <linux/net_tstamp.h>
1da177e4
LT		 130#include <net/xfrm.h>
131#include <linux/ipsec.h>
f8451725 132#include <net/cls_cgroup.h>
5bc1421e 133#include <net/netprio_cgroup.h>
1da177e4
LT		 134
135#include <linux/filter.h>
136
3847ce32
SM		 137#include <trace/events/sock.h>
138
1da177e4
LT		 139#ifdef CONFIG_INET
140#include <net/tcp.h>
141#endif
142
076bb0c8 143#include <net/busy_poll.h>
06021292 144
36b77a52 145static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 146static LIST_HEAD(proto_list);
147
a3b299da
EB		 148/**
149 * sk_ns_capable - General socket capability test
150 * @sk: Socket to use a capability on or through
151 * @user_ns: The user namespace of the capability to use
152 * @cap: The capability to use
153 *
154 * Test to see if the opener of the socket had when the socket was
155 * created and the current process has the capability @cap in the user
156 * namespace @user_ns.
157 */
158bool sk_ns_capable(const struct sock *sk,
159 struct user_namespace *user_ns, int cap)
160{
161 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
162 ns_capable(user_ns, cap);
163}
164EXPORT_SYMBOL(sk_ns_capable);
165
166/**
167 * sk_capable - Socket global capability test
168 * @sk: Socket to use a capability on or through
e793c0f7 169 * @cap: The global capability to use
a3b299da
EB		 170 *
171 * Test to see if the opener of the socket had when the socket was
172 * created and the current process has the capability @cap in all user
173 * namespaces.
174 */
175bool sk_capable(const struct sock *sk, int cap)
176{
177 return sk_ns_capable(sk, &init_user_ns, cap);
178}
179EXPORT_SYMBOL(sk_capable);
180
181/**
182 * sk_net_capable - Network namespace socket capability test
183 * @sk: Socket to use a capability on or through
184 * @cap: The capability to use
185 *
e793c0f7 186 * Test to see if the opener of the socket had when the socket was created
a3b299da
EB		 187 * and the current process has the capability @cap over the network namespace
188 * the socket is a member of.
189 */
190bool sk_net_capable(const struct sock *sk, int cap)
191{
192 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
193}
194EXPORT_SYMBOL(sk_net_capable);
195
196
c255a458 197#ifdef CONFIG_MEMCG_KMEM
1d62e436 198int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
d1a4c0b3
GC		 199{
200 struct proto *proto;
201 int ret = 0;
202
36b77a52 203 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 204 list_for_each_entry(proto, &proto_list, node) {
205 if (proto->init_cgroup) {
1d62e436 206 ret = proto->init_cgroup(memcg, ss);
d1a4c0b3
GC		 207 if (ret)
208 goto out;
209 }
210 }
211
36b77a52 212 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 213 return ret;
214out:
215 list_for_each_entry_continue_reverse(proto, &proto_list, node)
216 if (proto->destroy_cgroup)
1d62e436 217 proto->destroy_cgroup(memcg);
36b77a52 218 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 219 return ret;
220}
221
1d62e436 222void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
d1a4c0b3
GC		 223{
224 struct proto *proto;
225
36b77a52 226 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 227 list_for_each_entry_reverse(proto, &proto_list, node)
228 if (proto->destroy_cgroup)
1d62e436 229 proto->destroy_cgroup(memcg);
36b77a52 230 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 231}
232#endif
233
da21f24d
IM		 234/*
235 * Each address family might have different locking rules, so we have
236 * one slock key per address family:
237 */
a5b5bb9a
IM		 238static struct lock_class_key af_family_keys[AF_MAX];
239static struct lock_class_key af_family_slock_keys[AF_MAX];
240
cbda4eaf 241#if defined(CONFIG_MEMCG_KMEM)
c5905afb 242struct static_key memcg_socket_limit_enabled;
e1aab161 243EXPORT_SYMBOL(memcg_socket_limit_enabled);
cbda4eaf 244#endif
e1aab161 245
a5b5bb9a
IM		 246/*
247 * Make lock validator output more readable. (we pre-construct these
248 * strings build-time, so that runtime initialization of socket
249 * locks is fast):
250 */
36cbd3dc 251static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 252 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
253 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
254 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
255 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
256 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
257 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
258 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 259 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 260 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 261 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 262 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 263 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 264 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
456db6a4 265 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 266};
36cbd3dc 267static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 268 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
269 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
270 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
271 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
272 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
273 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
274 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 275 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 276 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 277 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 278 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 279 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 280 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 281 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 282};
36cbd3dc 283static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 284 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
285 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
286 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
287 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
288 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
289 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
290 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 291 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 292 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 293 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 294 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 295 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 296 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 297 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 298};
da21f24d
IM		 299
300/*
301 * sk_callback_lock locking rules are per-address-family,
302 * so split the lock classes by using a per-AF key:
303 */
304static struct lock_class_key af_callback_keys[AF_MAX];
305
1da177e4
LT		 306/* Take into consideration the size of the struct sk_buff overhead in the
307 * determination of these values, since that is non-constant across
308 * platforms. This makes socket queueing behavior and performance
309 * not depend upon such differences.
310 */
311#define _SK_MEM_PACKETS 256
87fb4b7b 312#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 313#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
314#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
315
316/* Run time adjustable parameters. */
ab32ea5d 317__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 318EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 319__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 320EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 321__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
322__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 323
25985edc 324/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 325int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 326EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 327
b245be1f
WB		 328int sysctl_tstamp_allow_data __read_mostly = 1;
329
c93bdd0e
MG		 330struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
331EXPORT_SYMBOL_GPL(memalloc_socks);
332
7cb02404
MG		 333/**
334 * sk_set_memalloc - sets %SOCK_MEMALLOC
335 * @sk: socket to set it on
336 *
337 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
338 * It's the responsibility of the admin to adjust min_free_kbytes
339 * to meet the requirements
340 */
341void sk_set_memalloc(struct sock *sk)
342{
343 sock_set_flag(sk, SOCK_MEMALLOC);
344 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 345 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 346}
347EXPORT_SYMBOL_GPL(sk_set_memalloc);
348
349void sk_clear_memalloc(struct sock *sk)
350{
351 sock_reset_flag(sk, SOCK_MEMALLOC);
352 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 353 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 354
355 /*
356 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
357 * progress of swapping. However, if SOCK_MEMALLOC is cleared while
358 * it has rmem allocations there is a risk that the user of the
359 * socket cannot make forward progress due to exceeding the rmem
360 * limits. By rights, sk_clear_memalloc() should only be called
361 * on sockets being torn down but warn and reset the accounting if
362 * that assumption breaks.
363 */
364 if (WARN_ON(sk->sk_forward_alloc))
365 sk_mem_reclaim(sk);
7cb02404
MG		 366}
367EXPORT_SYMBOL_GPL(sk_clear_memalloc);
368
b4b9e355
MG		 369int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
370{
371 int ret;
372 unsigned long pflags = current->flags;
373
374 /* these should have been dropped before queueing */
375 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
376
377 current->flags |= PF_MEMALLOC;
378 ret = sk->sk_backlog_rcv(sk, skb);
379 tsk_restore_flags(current, pflags, PF_MEMALLOC);
380
381 return ret;
382}
383EXPORT_SYMBOL(__sk_backlog_rcv);
384
1da177e4
LT		 385static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
386{
387 struct timeval tv;
388
389 if (optlen < sizeof(tv))
390 return -EINVAL;
391 if (copy_from_user(&tv, optval, sizeof(tv)))
392 return -EFAULT;
ba78073e
VA		 393 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
394 return -EDOM;
1da177e4 395
ba78073e 396 if (tv.tv_sec < 0) {
6f11df83
AM		 397 static int warned __read_mostly;
398
ba78073e 399 *timeo_p = 0;
50aab54f 400 if (warned < 10 && net_ratelimit()) {
ba78073e 401 warned++;
e005d193
JP		 402 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
403 __func__, current->comm, task_pid_nr(current));
50aab54f 404 }
ba78073e
VA		 405 return 0;
406 }
1da177e4
LT		 407 *timeo_p = MAX_SCHEDULE_TIMEOUT;
408 if (tv.tv_sec == 0 && tv.tv_usec == 0)
409 return 0;
410 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
411 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
412 return 0;
413}
414
415static void sock_warn_obsolete_bsdism(const char *name)
416{
417 static int warned;
418 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 419 if (strcmp(warncomm, current->comm) && warned < 5) {
420 strcpy(warncomm, current->comm);
e005d193
JP		 421 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
422 warncomm, name);
1da177e4
LT		 423 warned++;
424 }
425}
426
08e29af3
ED		 427#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
428
429static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 430{
08e29af3
ED		 431 if (sk->sk_flags & flags) {
432 sk->sk_flags &= ~flags;
433 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 434 net_disable_timestamp();
1da177e4
LT		 435 }
436}
437
438
f0088a50
DV		 439int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
440{
766e9037 441 int err;
3b885787
NH		 442 unsigned long flags;
443 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 444
0fd7bac6 445 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 446 atomic_inc(&sk->sk_drops);
3847ce32 447 trace_sock_rcvqueue_full(sk, skb);
766e9037 448 return -ENOMEM;
f0088a50
DV		 449 }
450
fda9ef5d 451 err = sk_filter(sk, skb);
f0088a50 452 if (err)
766e9037 453 return err;
f0088a50 454
c76562b6 455 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 456 atomic_inc(&sk->sk_drops);
457 return -ENOBUFS;
3ab224be
HA		 458 }
459
f0088a50
DV		 460 skb->dev = NULL;
461 skb_set_owner_r(skb, sk);
49ad9599 462
7fee226a
ED		 463 /* we escape from rcu protected region, make sure we dont leak
464 * a norefcounted dst
465 */
466 skb_dst_force(skb);
467
3b885787 468 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 469 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 470 __skb_queue_tail(list, skb);
471 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 472
473 if (!sock_flag(sk, SOCK_DEAD))
676d2369 474 sk->sk_data_ready(sk);
766e9037 475 return 0;
f0088a50
DV		 476}
477EXPORT_SYMBOL(sock_queue_rcv_skb);
478
58a5a7b9 479int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 480{
481 int rc = NET_RX_SUCCESS;
482
fda9ef5d 483 if (sk_filter(sk, skb))
f0088a50
DV		 484 goto discard_and_relse;
485
486 skb->dev = NULL;
487
274f482d 488 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 489 atomic_inc(&sk->sk_drops);
490 goto discard_and_relse;
491 }
58a5a7b9
ACM		 492 if (nested)
493 bh_lock_sock_nested(sk);
494 else
495 bh_lock_sock(sk);
a5b5bb9a
IM		 496 if (!sock_owned_by_user(sk)) {
497 /*
498 * trylock + unlock semantics:
499 */
500 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
501
c57943a1 502 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 503
504 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 505 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 506 bh_unlock_sock(sk);
507 atomic_inc(&sk->sk_drops);
508 goto discard_and_relse;
509 }
510
f0088a50
DV		 511 bh_unlock_sock(sk);
512out:
513 sock_put(sk);
514 return rc;
515discard_and_relse:
516 kfree_skb(skb);
517 goto out;
518}
519EXPORT_SYMBOL(sk_receive_skb);
520
521struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
522{
b6c6712a 523 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 524
525 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 526 sk_tx_queue_clear(sk);
a9b3cd7f 527 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 528 dst_release(dst);
529 return NULL;
530 }
531
532 return dst;
533}
534EXPORT_SYMBOL(__sk_dst_check);
535
536struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
537{
538 struct dst_entry *dst = sk_dst_get(sk);
539
540 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
541 sk_dst_reset(sk);
542 dst_release(dst);
543 return NULL;
544 }
545
546 return dst;
547}
548EXPORT_SYMBOL(sk_dst_check);
549
c91f6df2
BH		 550static int sock_setbindtodevice(struct sock *sk, char __user *optval,
551 int optlen)
4878809f
DM		 552{
553 int ret = -ENOPROTOOPT;
554#ifdef CONFIG_NETDEVICES
3b1e0a65 555 struct net *net = sock_net(sk);
4878809f
DM		 556 char devname[IFNAMSIZ];
557 int index;
558
559 /* Sorry... */
560 ret = -EPERM;
5e1fccc0 561 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 562 goto out;
563
564 ret = -EINVAL;
565 if (optlen < 0)
566 goto out;
567
568 /* Bind this socket to a particular device like "eth0",
569 * as specified in the passed interface name. If the
570 * name is "" or the option length is zero the socket
571 * is not bound.
572 */
573 if (optlen > IFNAMSIZ - 1)
574 optlen = IFNAMSIZ - 1;
575 memset(devname, 0, sizeof(devname));
576
577 ret = -EFAULT;
578 if (copy_from_user(devname, optval, optlen))
579 goto out;
580
000ba2e4
DM		 581 index = 0;
582 if (devname[0] != '\0') {
bf8e56bf 583 struct net_device *dev;
4878809f 584
bf8e56bf
ED		 585 rcu_read_lock();
586 dev = dev_get_by_name_rcu(net, devname);
587 if (dev)
588 index = dev->ifindex;
589 rcu_read_unlock();
4878809f
DM		 590 ret = -ENODEV;
591 if (!dev)
592 goto out;
4878809f
DM		 593 }
594
595 lock_sock(sk);
596 sk->sk_bound_dev_if = index;
597 sk_dst_reset(sk);
598 release_sock(sk);
599
600 ret = 0;
601
602out:
603#endif
604
605 return ret;
606}
607
c91f6df2
BH		 608static int sock_getbindtodevice(struct sock *sk, char __user *optval,
609 int __user *optlen, int len)
610{
611 int ret = -ENOPROTOOPT;
612#ifdef CONFIG_NETDEVICES
613 struct net *net = sock_net(sk);
c91f6df2 614 char devname[IFNAMSIZ];
c91f6df2
BH		 615
616 if (sk->sk_bound_dev_if == 0) {
617 len = 0;
618 goto zero;
619 }
620
621 ret = -EINVAL;
622 if (len < IFNAMSIZ)
623 goto out;
624
5dbe7c17
NS		 625 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
626 if (ret)
c91f6df2 627 goto out;
c91f6df2
BH		 628
629 len = strlen(devname) + 1;
630
631 ret = -EFAULT;
632 if (copy_to_user(optval, devname, len))
633 goto out;
634
635zero:
636 ret = -EFAULT;
637 if (put_user(len, optlen))
638 goto out;
639
640 ret = 0;
641
642out:
643#endif
644
645 return ret;
646}
647
c0ef877b
PE		 648static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
649{
650 if (valbool)
651 sock_set_flag(sk, bit);
652 else
653 sock_reset_flag(sk, bit);
654}
655
f60e5990 656bool sk_mc_loop(struct sock *sk)
657{
658 if (dev_recursion_level())
659 return false;
660 if (!sk)
661 return true;
662 switch (sk->sk_family) {
663 case AF_INET:
664 return inet_sk(sk)->mc_loop;
665#if IS_ENABLED(CONFIG_IPV6)
666 case AF_INET6:
667 return inet6_sk(sk)->mc_loop;
668#endif
669 }
670 WARN_ON(1);
671 return true;
672}
673EXPORT_SYMBOL(sk_mc_loop);
674
1da177e4
LT		 675/*
676 * This is meant for all protocols to use and covers goings on
677 * at the socket level. Everything here is generic.
678 */
679
680int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 681 char __user *optval, unsigned int optlen)
1da177e4 682{
2a91525c 683 struct sock *sk = sock->sk;
1da177e4
LT		 684 int val;
685 int valbool;
686 struct linger ling;
687 int ret = 0;
4ec93edb 688
1da177e4
LT		 689 /*
690 * Options without arguments
691 */
692
4878809f 693 if (optname == SO_BINDTODEVICE)
c91f6df2 694 return sock_setbindtodevice(sk, optval, optlen);
4878809f 695
e71a4783
SH		 696 if (optlen < sizeof(int))
697 return -EINVAL;
4ec93edb 698
1da177e4
LT		 699 if (get_user(val, (int __user *)optval))
700 return -EFAULT;
4ec93edb 701
2a91525c 702 valbool = val ? 1 : 0;
1da177e4
LT		 703
704 lock_sock(sk);
705
2a91525c 706 switch (optname) {
e71a4783 707 case SO_DEBUG:
2a91525c 708 if (val && !capable(CAP_NET_ADMIN))
e71a4783 709 ret = -EACCES;
2a91525c 710 else
c0ef877b 711 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 712 break;
713 case SO_REUSEADDR:
4a17fd52 714 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 715 break;
055dc21a
TH		 716 case SO_REUSEPORT:
717 sk->sk_reuseport = valbool;
718 break;
e71a4783 719 case SO_TYPE:
49c794e9 720 case SO_PROTOCOL:
0d6038ee 721 case SO_DOMAIN:
e71a4783
SH		 722 case SO_ERROR:
723 ret = -ENOPROTOOPT;
724 break;
725 case SO_DONTROUTE:
c0ef877b 726 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 727 break;
728 case SO_BROADCAST:
729 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
730 break;
731 case SO_SNDBUF:
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_wmem_max);
b0573dea 738set_sndbuf:
e71a4783 739 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 740 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
741 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 742 sk->sk_write_space(sk);
743 break;
1da177e4 744
e71a4783
SH		 745 case SO_SNDBUFFORCE:
746 if (!capable(CAP_NET_ADMIN)) {
747 ret = -EPERM;
748 break;
749 }
750 goto set_sndbuf;
b0573dea 751
e71a4783
SH		 752 case SO_RCVBUF:
753 /* Don't error on this BSD doesn't and if you think
82981930
ED		 754 * about it this is right. Otherwise apps have to
755 * play 'guess the biggest size' games. RCVBUF/SNDBUF
756 * are treated in BSD as hints
757 */
758 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 759set_rcvbuf:
e71a4783
SH		 760 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
761 /*
762 * We double it on the way in to account for
763 * "struct sk_buff" etc. overhead. Applications
764 * assume that the SO_RCVBUF setting they make will
765 * allow that much actual data to be received on that
766 * socket.
767 *
768 * Applications are unaware that "struct sk_buff" and
769 * other overheads allocate from the receive buffer
770 * during socket buffer allocation.
771 *
772 * And after considering the possible alternatives,
773 * returning the value we actually used in getsockopt
774 * is the most desirable behavior.
775 */
82981930 776 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 777 break;
778
779 case SO_RCVBUFFORCE:
780 if (!capable(CAP_NET_ADMIN)) {
781 ret = -EPERM;
1da177e4 782 break;
e71a4783
SH		 783 }
784 goto set_rcvbuf;
1da177e4 785
e71a4783 786 case SO_KEEPALIVE:
1da177e4 787#ifdef CONFIG_INET
3e10986d
ED		 788 if (sk->sk_protocol == IPPROTO_TCP &&
789 sk->sk_type == SOCK_STREAM)
e71a4783 790 tcp_set_keepalive(sk, valbool);
1da177e4 791#endif
e71a4783
SH		 792 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
793 break;
794
795 case SO_OOBINLINE:
796 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
797 break;
798
799 case SO_NO_CHECK:
28448b80 800 sk->sk_no_check_tx = valbool;
e71a4783
SH		 801 break;
802
803 case SO_PRIORITY:
5e1fccc0
EB		 804 if ((val >= 0 && val <= 6) ||
805 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 806 sk->sk_priority = val;
807 else
808 ret = -EPERM;
809 break;
810
811 case SO_LINGER:
812 if (optlen < sizeof(ling)) {
813 ret = -EINVAL; /* 1003.1g */
1da177e4 814 break;
e71a4783 815 }
2a91525c 816 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 817 ret = -EFAULT;
1da177e4 818 break;
e71a4783
SH		 819 }
820 if (!ling.l_onoff)
821 sock_reset_flag(sk, SOCK_LINGER);
822 else {
1da177e4 823#if (BITS_PER_LONG == 32)
e71a4783
SH		 824 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
825 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 826 else
e71a4783
SH		 827#endif
828 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
829 sock_set_flag(sk, SOCK_LINGER);
830 }
831 break;
832
833 case SO_BSDCOMPAT:
834 sock_warn_obsolete_bsdism("setsockopt");
835 break;
836
837 case SO_PASSCRED:
838 if (valbool)
839 set_bit(SOCK_PASSCRED, &sock->flags);
840 else
841 clear_bit(SOCK_PASSCRED, &sock->flags);
842 break;
843
844 case SO_TIMESTAMP:
92f37fd2 845 case SO_TIMESTAMPNS:
e71a4783 846 if (valbool) {
92f37fd2
ED		 847 if (optname == SO_TIMESTAMP)
848 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
849 else
850 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 851 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 852 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 853 } else {
e71a4783 854 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 855 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
856 }
e71a4783
SH		 857 break;
858
20d49473
PO		 859 case SO_TIMESTAMPING:
860 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 861 ret = -EINVAL;
20d49473
PO		 862 break;
863 }
b245be1f 864
09c2d251 865 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765
WB		 866 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
867 if (sk->sk_protocol == IPPROTO_TCP) {
868 if (sk->sk_state != TCP_ESTABLISHED) {
869 ret = -EINVAL;
870 break;
871 }
872 sk->sk_tskey = tcp_sk(sk)->snd_una;
873 } else {
874 sk->sk_tskey = 0;
875 }
876 }
b9f40e21 877 sk->sk_tsflags = val;
20d49473
PO		 878 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
879 sock_enable_timestamp(sk,
880 SOCK_TIMESTAMPING_RX_SOFTWARE);
881 else
882 sock_disable_timestamp(sk,
08e29af3 883 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 884 break;
885
e71a4783
SH		 886 case SO_RCVLOWAT:
887 if (val < 0)
888 val = INT_MAX;
889 sk->sk_rcvlowat = val ? : 1;
890 break;
891
892 case SO_RCVTIMEO:
893 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
894 break;
895
896 case SO_SNDTIMEO:
897 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
898 break;
1da177e4 899
e71a4783
SH		 900 case SO_ATTACH_FILTER:
901 ret = -EINVAL;
902 if (optlen == sizeof(struct sock_fprog)) {
903 struct sock_fprog fprog;
1da177e4 904
e71a4783
SH		 905 ret = -EFAULT;
906 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 907 break;
e71a4783
SH		 908
909 ret = sk_attach_filter(&fprog, sk);
910 }
911 break;
912
89aa0758
AS		 913 case SO_ATTACH_BPF:
914 ret = -EINVAL;
915 if (optlen == sizeof(u32)) {
916 u32 ufd;
917
918 ret = -EFAULT;
919 if (copy_from_user(&ufd, optval, sizeof(ufd)))
920 break;
921
922 ret = sk_attach_bpf(ufd, sk);
923 }
924 break;
925
e71a4783 926 case SO_DETACH_FILTER:
55b33325 927 ret = sk_detach_filter(sk);
e71a4783 928 break;
1da177e4 929
d59577b6
VB		 930 case SO_LOCK_FILTER:
931 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
932 ret = -EPERM;
933 else
934 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
935 break;
936
e71a4783
SH		 937 case SO_PASSSEC:
938 if (valbool)
939 set_bit(SOCK_PASSSEC, &sock->flags);
940 else
941 clear_bit(SOCK_PASSSEC, &sock->flags);
942 break;
4a19ec58 943 case SO_MARK:
5e1fccc0 944 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 945 ret = -EPERM;
2a91525c 946 else
4a19ec58 947 sk->sk_mark = val;
4a19ec58 948 break;
877ce7c1 949
3b885787 950 case SO_RXQ_OVFL:
8083f0fc 951 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 952 break;
6e3e939f
JB		 953
954 case SO_WIFI_STATUS:
955 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
956 break;
957
ef64a54f
PE		 958 case SO_PEEK_OFF:
959 if (sock->ops->set_peek_off)
12663bfc 960 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 961 else
962 ret = -EOPNOTSUPP;
963 break;
3bdc0eba
BG		 964
965 case SO_NOFCS:
966 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
967 break;
968
7d4c04fc
KJ		 969 case SO_SELECT_ERR_QUEUE:
970 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
971 break;
972
e0d1095a 973#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 974 case SO_BUSY_POLL:
dafcc438
ET		 975 /* allow unprivileged users to decrease the value */
976 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
977 ret = -EPERM;
978 else {
979 if (val < 0)
980 ret = -EINVAL;
981 else
982 sk->sk_ll_usec = val;
983 }
984 break;
985#endif
62748f32
ED		 986
987 case SO_MAX_PACING_RATE:
988 sk->sk_max_pacing_rate = val;
989 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
990 sk->sk_max_pacing_rate);
991 break;
992
e71a4783
SH		 993 default:
994 ret = -ENOPROTOOPT;
995 break;
4ec93edb 996 }
1da177e4
LT		 997 release_sock(sk);
998 return ret;
999}
2a91525c 1000EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1001
1002
8f09898b 1003static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1004 struct ucred *ucred)
3f551f94
EB		 1005{
1006 ucred->pid = pid_vnr(pid);
1007 ucred->uid = ucred->gid = -1;
1008 if (cred) {
1009 struct user_namespace *current_ns = current_user_ns();
1010
b2e4f544
EB		 1011 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1012 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1013 }
1014}
1015
1da177e4
LT		 1016int sock_getsockopt(struct socket *sock, int level, int optname,
1017 char __user *optval, int __user *optlen)
1018{
1019 struct sock *sk = sock->sk;
4ec93edb 1020
e71a4783 1021 union {
4ec93edb
YH		 1022 int val;
1023 struct linger ling;
1da177e4
LT		 1024 struct timeval tm;
1025 } v;
4ec93edb 1026
4d0392be 1027 int lv = sizeof(int);
1da177e4 1028 int len;
4ec93edb 1029
e71a4783 1030 if (get_user(len, optlen))
4ec93edb 1031 return -EFAULT;
e71a4783 1032 if (len < 0)
1da177e4 1033 return -EINVAL;
4ec93edb 1034
50fee1de 1035 memset(&v, 0, sizeof(v));
df0bca04 1036
2a91525c 1037 switch (optname) {
e71a4783
SH		 1038 case SO_DEBUG:
1039 v.val = sock_flag(sk, SOCK_DBG);
1040 break;
1041
1042 case SO_DONTROUTE:
1043 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1044 break;
1045
1046 case SO_BROADCAST:
1b23a5df 1047 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1048 break;
1049
1050 case SO_SNDBUF:
1051 v.val = sk->sk_sndbuf;
1052 break;
1053
1054 case SO_RCVBUF:
1055 v.val = sk->sk_rcvbuf;
1056 break;
1057
1058 case SO_REUSEADDR:
1059 v.val = sk->sk_reuse;
1060 break;
1061
055dc21a
TH		 1062 case SO_REUSEPORT:
1063 v.val = sk->sk_reuseport;
1064 break;
1065
e71a4783 1066 case SO_KEEPALIVE:
1b23a5df 1067 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1068 break;
1069
1070 case SO_TYPE:
1071 v.val = sk->sk_type;
1072 break;
1073
49c794e9
JE		 1074 case SO_PROTOCOL:
1075 v.val = sk->sk_protocol;
1076 break;
1077
0d6038ee
JE		 1078 case SO_DOMAIN:
1079 v.val = sk->sk_family;
1080 break;
1081
e71a4783
SH		 1082 case SO_ERROR:
1083 v.val = -sock_error(sk);
2a91525c 1084 if (v.val == 0)
e71a4783
SH		 1085 v.val = xchg(&sk->sk_err_soft, 0);
1086 break;
1087
1088 case SO_OOBINLINE:
1b23a5df 1089 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1090 break;
1091
1092 case SO_NO_CHECK:
28448b80 1093 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1094 break;
1095
1096 case SO_PRIORITY:
1097 v.val = sk->sk_priority;
1098 break;
1099
1100 case SO_LINGER:
1101 lv = sizeof(v.ling);
1b23a5df 1102 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1103 v.ling.l_linger = sk->sk_lingertime / HZ;
1104 break;
1105
1106 case SO_BSDCOMPAT:
1107 sock_warn_obsolete_bsdism("getsockopt");
1108 break;
1109
1110 case SO_TIMESTAMP:
92f37fd2
ED		 1111 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1112 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1113 break;
1114
1115 case SO_TIMESTAMPNS:
1116 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1117 break;
1118
20d49473 1119 case SO_TIMESTAMPING:
b9f40e21 1120 v.val = sk->sk_tsflags;
20d49473
PO		 1121 break;
1122
e71a4783 1123 case SO_RCVTIMEO:
2a91525c 1124 lv = sizeof(struct timeval);
e71a4783
SH		 1125 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1126 v.tm.tv_sec = 0;
1127 v.tm.tv_usec = 0;
1128 } else {
1129 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1130 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1131 }
1132 break;
1133
1134 case SO_SNDTIMEO:
2a91525c 1135 lv = sizeof(struct timeval);
e71a4783
SH		 1136 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1137 v.tm.tv_sec = 0;
1138 v.tm.tv_usec = 0;
1139 } else {
1140 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1141 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1142 }
1143 break;
1da177e4 1144
e71a4783
SH		 1145 case SO_RCVLOWAT:
1146 v.val = sk->sk_rcvlowat;
1147 break;
1da177e4 1148
e71a4783 1149 case SO_SNDLOWAT:
2a91525c 1150 v.val = 1;
e71a4783 1151 break;
1da177e4 1152
e71a4783 1153 case SO_PASSCRED:
82981930 1154 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1155 break;
1da177e4 1156
e71a4783 1157 case SO_PEERCRED:
109f6e39
EB		 1158 {
1159 struct ucred peercred;
1160 if (len > sizeof(peercred))
1161 len = sizeof(peercred);
1162 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1163 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1164 return -EFAULT;
1165 goto lenout;
109f6e39 1166 }
1da177e4 1167
e71a4783
SH		 1168 case SO_PEERNAME:
1169 {
1170 char address[128];
1171
1172 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1173 return -ENOTCONN;
1174 if (lv < len)
1175 return -EINVAL;
1176 if (copy_to_user(optval, address, len))
1177 return -EFAULT;
1178 goto lenout;
1179 }
1da177e4 1180
e71a4783
SH		 1181 /* Dubious BSD thing... Probably nobody even uses it, but
1182 * the UNIX standard wants it for whatever reason... -DaveM
1183 */
1184 case SO_ACCEPTCONN:
1185 v.val = sk->sk_state == TCP_LISTEN;
1186 break;
1da177e4 1187
e71a4783 1188 case SO_PASSSEC:
82981930 1189 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1190 break;
877ce7c1 1191
e71a4783
SH		 1192 case SO_PEERSEC:
1193 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1194
4a19ec58
LAT		 1195 case SO_MARK:
1196 v.val = sk->sk_mark;
1197 break;
1198
3b885787 1199 case SO_RXQ_OVFL:
1b23a5df 1200 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1201 break;
1202
6e3e939f 1203 case SO_WIFI_STATUS:
1b23a5df 1204 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1205 break;
1206
ef64a54f
PE		 1207 case SO_PEEK_OFF:
1208 if (!sock->ops->set_peek_off)
1209 return -EOPNOTSUPP;
1210
1211 v.val = sk->sk_peek_off;
1212 break;
bc2f7996 1213 case SO_NOFCS:
1b23a5df 1214 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1215 break;
c91f6df2 1216
f7b86bfe 1217 case SO_BINDTODEVICE:
c91f6df2
BH		 1218 return sock_getbindtodevice(sk, optval, optlen, len);
1219
a8fc9277
PE		 1220 case SO_GET_FILTER:
1221 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1222 if (len < 0)
1223 return len;
1224
1225 goto lenout;
c91f6df2 1226
d59577b6
VB		 1227 case SO_LOCK_FILTER:
1228 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1229 break;
1230
ea02f941
MS		 1231 case SO_BPF_EXTENSIONS:
1232 v.val = bpf_tell_extensions();
1233 break;
1234
7d4c04fc
KJ		 1235 case SO_SELECT_ERR_QUEUE:
1236 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1237 break;
1238
e0d1095a 1239#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1240 case SO_BUSY_POLL:
dafcc438
ET		 1241 v.val = sk->sk_ll_usec;
1242 break;
1243#endif
1244
62748f32
ED		 1245 case SO_MAX_PACING_RATE:
1246 v.val = sk->sk_max_pacing_rate;
1247 break;
1248
2c8c56e1
ED		 1249 case SO_INCOMING_CPU:
1250 v.val = sk->sk_incoming_cpu;
1251 break;
1252
e71a4783 1253 default:
443b5991
YH		 1254 /* We implement the SO_SNDLOWAT etc to not be settable
1255 * (1003.1g 7).
1256 */
e71a4783 1257 return -ENOPROTOOPT;
1da177e4 1258 }
e71a4783 1259
1da177e4
LT		 1260 if (len > lv)
1261 len = lv;
1262 if (copy_to_user(optval, &v, len))
1263 return -EFAULT;
1264lenout:
4ec93edb
YH		 1265 if (put_user(len, optlen))
1266 return -EFAULT;
1267 return 0;
1da177e4
LT		 1268}
1269
a5b5bb9a
IM		 1270/*
1271 * Initialize an sk_lock.
1272 *
1273 * (We also register the sk_lock with the lock validator.)
1274 */
b6f99a21 1275static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1276{
ed07536e
PZ		 1277 sock_lock_init_class_and_name(sk,
1278 af_family_slock_key_strings[sk->sk_family],
1279 af_family_slock_keys + sk->sk_family,
1280 af_family_key_strings[sk->sk_family],
1281 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1282}
1283
4dc6dc71
ED		 1284/*
1285 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1286 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1287 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1288 */
f1a6c4da
PE		 1289static void sock_copy(struct sock *nsk, const struct sock *osk)
1290{
1291#ifdef CONFIG_SECURITY_NETWORK
1292 void *sptr = nsk->sk_security;
1293#endif
68835aba
ED		 1294 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1295
1296 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1297 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1298
f1a6c4da
PE		 1299#ifdef CONFIG_SECURITY_NETWORK
1300 nsk->sk_security = sptr;
1301 security_sk_clone(osk, nsk);
1302#endif
1303}
1304
fcbdf09d
OP		 1305void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1306{
1307 unsigned long nulls1, nulls2;
1308
1309 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1310 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1311 if (nulls1 > nulls2)
1312 swap(nulls1, nulls2);
1313
1314 if (nulls1 != 0)
1315 memset((char *)sk, 0, nulls1);
1316 memset((char *)sk + nulls1 + sizeof(void *), 0,
1317 nulls2 - nulls1 - sizeof(void *));
1318 memset((char *)sk + nulls2 + sizeof(void *), 0,
1319 size - nulls2 - sizeof(void *));
1320}
1321EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1322
2e4afe7b
PE		 1323static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1324 int family)
c308c1b2
PE		 1325{
1326 struct sock *sk;
1327 struct kmem_cache *slab;
1328
1329 slab = prot->slab;
e912b114
ED		 1330 if (slab != NULL) {
1331 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1332 if (!sk)
1333 return sk;
1334 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1335 if (prot->clear_sk)
1336 prot->clear_sk(sk, prot->obj_size);
1337 else
1338 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1339 }
fcbdf09d 1340 } else
c308c1b2
PE		 1341 sk = kmalloc(prot->obj_size, priority);
1342
2e4afe7b 1343 if (sk != NULL) {
a98b65a3
VN		 1344 kmemcheck_annotate_bitfield(sk, flags);
1345
2e4afe7b
PE		 1346 if (security_sk_alloc(sk, family, priority))
1347 goto out_free;
1348
1349 if (!try_module_get(prot->owner))
1350 goto out_free_sec;
e022f0b4 1351 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1352 }
1353
c308c1b2 1354 return sk;
2e4afe7b
PE		 1355
1356out_free_sec:
1357 security_sk_free(sk);
1358out_free:
1359 if (slab != NULL)
1360 kmem_cache_free(slab, sk);
1361 else
1362 kfree(sk);
1363 return NULL;
c308c1b2
PE		 1364}
1365
1366static void sk_prot_free(struct proto *prot, struct sock *sk)
1367{
1368 struct kmem_cache *slab;
2e4afe7b 1369 struct module *owner;
c308c1b2 1370
2e4afe7b 1371 owner = prot->owner;
c308c1b2 1372 slab = prot->slab;
2e4afe7b
PE		 1373
1374 security_sk_free(sk);
c308c1b2
PE		 1375 if (slab != NULL)
1376 kmem_cache_free(slab, sk);
1377 else
1378 kfree(sk);
2e4afe7b 1379 module_put(owner);
c308c1b2
PE		 1380}
1381
86f8515f 1382#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1383void sock_update_netprioidx(struct sock *sk)
5bc1421e 1384{
5bc1421e
NH		 1385 if (in_interrupt())
1386 return;
2b73bc65 1387
6ffd4641 1388 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1389}
1390EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1391#endif
1392
1da177e4
LT		 1393/**
1394 * sk_alloc - All socket objects are allocated here
c4ea43c5 1395 * @net: the applicable net namespace
4dc3b16b
PP		 1396 * @family: protocol family
1397 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1398 * @prot: struct proto associated with this new sock instance
11aa9c28 1399 * @kern: is this to be a kernel socket?
1da177e4 1400 */
1b8d7ae4 1401struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1402 struct proto *prot, int kern)
1da177e4 1403{
c308c1b2 1404 struct sock *sk;
1da177e4 1405
154adbc8 1406 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1407 if (sk) {
154adbc8
PE		 1408 sk->sk_family = family;
1409 /*
1410 * See comment in struct sock definition to understand
1411 * why we need sk_prot_creator -acme
1412 */
1413 sk->sk_prot = sk->sk_prot_creator = prot;
1414 sock_lock_init(sk);
3b1e0a65 1415 sock_net_set(sk, get_net(net));
d66ee058 1416 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1417
211d2f97 1418 sock_update_classid(sk);
6ffd4641 1419 sock_update_netprioidx(sk);
1da177e4 1420 }
a79af59e 1421
2e4afe7b 1422 return sk;
1da177e4 1423}
2a91525c 1424EXPORT_SYMBOL(sk_alloc);
1da177e4 1425
2b85a34e 1426static void __sk_free(struct sock *sk)
1da177e4
LT		 1427{
1428 struct sk_filter *filter;
1da177e4
LT		 1429
1430 if (sk->sk_destruct)
1431 sk->sk_destruct(sk);
1432
a898def2
PM		 1433 filter = rcu_dereference_check(sk->sk_filter,
1434 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1435 if (filter) {
309dd5fc 1436 sk_filter_uncharge(sk, filter);
a9b3cd7f 1437 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1438 }
1439
08e29af3 1440 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1441
1442 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1443 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1444 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1445
109f6e39
EB		 1446 if (sk->sk_peer_cred)
1447 put_cred(sk->sk_peer_cred);
1448 put_pid(sk->sk_peer_pid);
3b1e0a65 1449 put_net(sock_net(sk));
c308c1b2 1450 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1451}
2b85a34e
ED		 1452
1453void sk_free(struct sock *sk)
1454{
1455 /*
25985edc 1456 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1457 * some packets are still in some tx queue.
1458 * If not null, sock_wfree() will call __sk_free(sk) later
1459 */
1460 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1461 __sk_free(sk);
1462}
2a91525c 1463EXPORT_SYMBOL(sk_free);
1da177e4 1464
edf02087 1465/*
25985edc
LDM		 1466 * Last sock_put should drop reference to sk->sk_net. It has already
1467 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1468 * is not an option.
25985edc 1469 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1470 * destroy it in the context of init_net.
1471 */
1472void sk_release_kernel(struct sock *sk)
1473{
1474 if (sk == NULL || sk->sk_socket == NULL)
1475 return;
1476
1477 sock_hold(sk);
3b1e0a65 1478 sock_net_set(sk, get_net(&init_net));
c243d7e2 1479 sock_release(sk->sk_socket);
edf02087
DL		 1480 sock_put(sk);
1481}
45af1754 1482EXPORT_SYMBOL(sk_release_kernel);
edf02087 1483
475f1b52
SR		 1484static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1485{
1486 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1487 sock_update_memcg(newsk);
1488}
1489
e56c57d0
ED		 1490/**
1491 * sk_clone_lock - clone a socket, and lock its clone
1492 * @sk: the socket to clone
1493 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1494 *
1495 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1496 */
1497struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1498{
8fd1d178 1499 struct sock *newsk;
278571ba 1500 bool is_charged = true;
87d11ceb 1501
8fd1d178 1502 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1503 if (newsk != NULL) {
1504 struct sk_filter *filter;
1505
892c141e 1506 sock_copy(newsk, sk);
87d11ceb
ACM		 1507
1508 /* SANITY */
3b1e0a65 1509 get_net(sock_net(newsk));
87d11ceb
ACM		 1510 sk_node_init(&newsk->sk_node);
1511 sock_lock_init(newsk);
1512 bh_lock_sock(newsk);
fa438ccf 1513 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1514 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1515
1516 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1517 /*
1518 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1519 */
1520 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1521 atomic_set(&newsk->sk_omem_alloc, 0);
1522 skb_queue_head_init(&newsk->sk_receive_queue);
1523 skb_queue_head_init(&newsk->sk_write_queue);
1524
b6c6712a 1525 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1526 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1527 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1528 af_callback_keys + newsk->sk_family,
1529 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1530
1531 newsk->sk_dst_cache = NULL;
1532 newsk->sk_wmem_queued = 0;
1533 newsk->sk_forward_alloc = 0;
1534 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1535 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1536
1537 sock_reset_flag(newsk, SOCK_DONE);
1538 skb_queue_head_init(&newsk->sk_error_queue);
1539
0d7da9dd 1540 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1541 if (filter != NULL)
278571ba
AS		 1542 /* though it's an empty new sock, the charging may fail
1543 * if sysctl_optmem_max was changed between creation of
1544 * original socket and cloning
1545 */
1546 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1547
278571ba 1548 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk))) {
87d11ceb
ACM		 1549 /* It is still raw copy of parent, so invalidate
1550 * destructor and make plain sk_free() */
1551 newsk->sk_destruct = NULL;
b0691c8e 1552 bh_unlock_sock(newsk);
87d11ceb
ACM		 1553 sk_free(newsk);
1554 newsk = NULL;
1555 goto out;
1556 }
1557
1558 newsk->sk_err = 0;
1559 newsk->sk_priority = 0;
2c8c56e1 1560 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1561 atomic64_set(&newsk->sk_cookie, 0);
4dc6dc71
ED		 1562 /*
1563 * Before updating sk_refcnt, we must commit prior changes to memory
1564 * (Documentation/RCU/rculist_nulls.txt for details)
1565 */
1566 smp_wmb();
87d11ceb
ACM		 1567 atomic_set(&newsk->sk_refcnt, 2);
1568
1569 /*
1570 * Increment the counter in the same struct proto as the master
1571 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1572 * is the same as sk->sk_prot->socks, as this field was copied
1573 * with memcpy).
1574 *
1575 * This _changes_ the previous behaviour, where
1576 * tcp_create_openreq_child always was incrementing the
1577 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1578 * to be taken into account in all callers. -acme
1579 */
1580 sk_refcnt_debug_inc(newsk);
972692e0 1581 sk_set_socket(newsk, NULL);
43815482 1582 newsk->sk_wq = NULL;
87d11ceb 1583
f3f511e1
GC		 1584 sk_update_clone(sk, newsk);
1585
87d11ceb 1586 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1587 sk_sockets_allocated_inc(newsk);
704da560 1588
08e29af3 1589 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1590 net_enable_timestamp();
87d11ceb
ACM		 1591 }
1592out:
1593 return newsk;
1594}
e56c57d0 1595EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1596
9958089a
AK		 1597void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1598{
1599 __sk_dst_set(sk, dst);
1600 sk->sk_route_caps = dst->dev->features;
1601 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1602 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1603 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1604 if (sk_can_gso(sk)) {
82cc1a7a 1605 if (dst->header_len) {
9958089a 1606 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1607 } else {
9958089a 1608 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1609 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1610 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1611 }
9958089a
AK		 1612 }
1613}
1614EXPORT_SYMBOL_GPL(sk_setup_caps);
1615
1da177e4
LT		 1616/*
1617 * Simple resource managers for sockets.
1618 */
1619
1620
4ec93edb
YH		 1621/*
1622 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1623 */
1624void sock_wfree(struct sk_buff *skb)
1625{
1626 struct sock *sk = skb->sk;
d99927f4 1627 unsigned int len = skb->truesize;
1da177e4 1628
d99927f4
ED		 1629 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1630 /*
1631 * Keep a reference on sk_wmem_alloc, this will be released
1632 * after sk_write_space() call
1633 */
1634 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1635 sk->sk_write_space(sk);
d99927f4
ED		 1636 len = 1;
1637 }
2b85a34e 1638 /*
d99927f4
ED		 1639 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1640 * could not do because of in-flight packets
2b85a34e 1641 */
d99927f4 1642 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1643 __sk_free(sk);
1da177e4 1644}
2a91525c 1645EXPORT_SYMBOL(sock_wfree);
1da177e4 1646
f2f872f9
ED		 1647void skb_orphan_partial(struct sk_buff *skb)
1648{
1649 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1650 * so we do not completely orphan skb, but transfert all
1651 * accounted bytes but one, to avoid unexpected reorders.
1652 */
1653 if (skb->destructor == sock_wfree
1654#ifdef CONFIG_INET
1655 || skb->destructor == tcp_wfree
1656#endif
1657 ) {
1658 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1659 skb->truesize = 1;
1660 } else {
1661 skb_orphan(skb);
1662 }
1663}
1664EXPORT_SYMBOL(skb_orphan_partial);
1665
4ec93edb
YH		 1666/*
1667 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1668 */
1669void sock_rfree(struct sk_buff *skb)
1670{
1671 struct sock *sk = skb->sk;
d361fd59 1672 unsigned int len = skb->truesize;
1da177e4 1673
d361fd59
ED		 1674 atomic_sub(len, &sk->sk_rmem_alloc);
1675 sk_mem_uncharge(sk, len);
1da177e4 1676}
2a91525c 1677EXPORT_SYMBOL(sock_rfree);
1da177e4 1678
7768eed8
OH		 1679/*
1680 * Buffer destructor for skbs that are not used directly in read or write
1681 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1682 */
62bccb8c
AD		 1683void sock_efree(struct sk_buff *skb)
1684{
1685 sock_put(skb->sk);
1686}
1687EXPORT_SYMBOL(sock_efree);
1688
976d0201 1689kuid_t sock_i_uid(struct sock *sk)
1da177e4 1690{
976d0201 1691 kuid_t uid;
1da177e4 1692
f064af1e 1693 read_lock_bh(&sk->sk_callback_lock);
976d0201 1694 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1695 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1696 return uid;
1697}
2a91525c 1698EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1699
1700unsigned long sock_i_ino(struct sock *sk)
1701{
1702 unsigned long ino;
1703
f064af1e 1704 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1705 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1706 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1707 return ino;
1708}
2a91525c 1709EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1710
1711/*
1712 * Allocate a skb from the socket's send buffer.
1713 */
86a76caf 1714struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1715 gfp_t priority)
1da177e4
LT		 1716{
1717 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1718 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1719 if (skb) {
1720 skb_set_owner_w(skb, sk);
1721 return skb;
1722 }
1723 }
1724 return NULL;
1725}
2a91525c 1726EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1727
4ec93edb 1728/*
1da177e4 1729 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1730 */
dd0fc66f 1731void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1732{
95c96174 1733 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1734 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1735 void *mem;
1736 /* First do the add, to avoid the race if kmalloc
4ec93edb 1737 * might sleep.
1da177e4
LT		 1738 */
1739 atomic_add(size, &sk->sk_omem_alloc);
1740 mem = kmalloc(size, priority);
1741 if (mem)
1742 return mem;
1743 atomic_sub(size, &sk->sk_omem_alloc);
1744 }
1745 return NULL;
1746}
2a91525c 1747EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1748
79e88659
DB		 1749/* Free an option memory block. Note, we actually want the inline
1750 * here as this allows gcc to detect the nullify and fold away the
1751 * condition entirely.
1da177e4 1752 */
79e88659
DB		 1753static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1754 const bool nullify)
1da177e4 1755{
e53da5fb
DM		 1756 if (WARN_ON_ONCE(!mem))
1757 return;
79e88659
DB		 1758 if (nullify)
1759 kzfree(mem);
1760 else
1761 kfree(mem);
1da177e4
LT		 1762 atomic_sub(size, &sk->sk_omem_alloc);
1763}
79e88659
DB		 1764
1765void sock_kfree_s(struct sock *sk, void *mem, int size)
1766{
1767 __sock_kfree_s(sk, mem, size, false);
1768}
2a91525c 1769EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1770
79e88659
DB		 1771void sock_kzfree_s(struct sock *sk, void *mem, int size)
1772{
1773 __sock_kfree_s(sk, mem, size, true);
1774}
1775EXPORT_SYMBOL(sock_kzfree_s);
1776
1da177e4
LT		 1777/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1778 I think, these locks should be removed for datagram sockets.
1779 */
2a91525c 1780static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1781{
1782 DEFINE_WAIT(wait);
1783
1784 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1785 for (;;) {
1786 if (!timeo)
1787 break;
1788 if (signal_pending(current))
1789 break;
1790 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1791 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1792 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1793 break;
1794 if (sk->sk_shutdown & SEND_SHUTDOWN)
1795 break;
1796 if (sk->sk_err)
1797 break;
1798 timeo = schedule_timeout(timeo);
1799 }
aa395145 1800 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1801 return timeo;
1802}
1803
1804
1805/*
1806 * Generic send/receive buffer handlers
1807 */
1808
4cc7f68d
HX		 1809struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1810 unsigned long data_len, int noblock,
28d64271 1811 int *errcode, int max_page_order)
1da177e4 1812{
2e4e4410 1813 struct sk_buff *skb;
1da177e4
LT		 1814 long timeo;
1815 int err;
1816
1da177e4 1817 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1818 for (;;) {
1da177e4
LT		 1819 err = sock_error(sk);
1820 if (err != 0)
1821 goto failure;
1822
1823 err = -EPIPE;
1824 if (sk->sk_shutdown & SEND_SHUTDOWN)
1825 goto failure;
1826
2e4e4410
ED		 1827 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1828 break;
28d64271 1829
2e4e4410
ED		 1830 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1831 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1832 err = -EAGAIN;
1833 if (!timeo)
1da177e4 1834 goto failure;
2e4e4410
ED		 1835 if (signal_pending(current))
1836 goto interrupted;
1837 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1838 }
2e4e4410
ED		 1839 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1840 errcode, sk->sk_allocation);
1841 if (skb)
1842 skb_set_owner_w(skb, sk);
1da177e4
LT		 1843 return skb;
1844
1845interrupted:
1846 err = sock_intr_errno(timeo);
1847failure:
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
82d5e2b8 1867 * @gfp: priority for memory allocation
400dfd3a
ED		 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 2177
1b784140 2178int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2179{
2180 return -EOPNOTSUPP;
2181}
2a91525c 2182EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2183
1b784140
YX		 2184int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2185 int flags)
1da177e4
LT		 2186{
2187 return -EOPNOTSUPP;
2188}
2a91525c 2189EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2190
2191int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2192{
2193 /* Mirror missing mmap method error code */
2194 return -ENODEV;
2195}
2a91525c 2196EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2197
2198ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2199{
2200 ssize_t res;
2201 struct msghdr msg = {.msg_flags = flags};
2202 struct kvec iov;
2203 char *kaddr = kmap(page);
2204 iov.iov_base = kaddr + offset;
2205 iov.iov_len = size;
2206 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2207 kunmap(page);
2208 return res;
2209}
2a91525c 2210EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2211
2212/*
2213 * Default Socket Callbacks
2214 */
2215
2216static void sock_def_wakeup(struct sock *sk)
2217{
43815482
ED		 2218 struct socket_wq *wq;
2219
2220 rcu_read_lock();
2221 wq = rcu_dereference(sk->sk_wq);
2222 if (wq_has_sleeper(wq))
2223 wake_up_interruptible_all(&wq->wait);
2224 rcu_read_unlock();
1da177e4
LT		 2225}
2226
2227static void sock_def_error_report(struct sock *sk)
2228{
43815482
ED		 2229 struct socket_wq *wq;
2230
2231 rcu_read_lock();
2232 wq = rcu_dereference(sk->sk_wq);
2233 if (wq_has_sleeper(wq))
2234 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2235 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2236 rcu_read_unlock();
1da177e4
LT		 2237}
2238
676d2369 2239static void sock_def_readable(struct sock *sk)
1da177e4 2240{
43815482
ED		 2241 struct socket_wq *wq;
2242
2243 rcu_read_lock();
2244 wq = rcu_dereference(sk->sk_wq);
2245 if (wq_has_sleeper(wq))
2c6607c6 2246 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2247 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2248 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2249 rcu_read_unlock();
1da177e4
LT		 2250}
2251
2252static void sock_def_write_space(struct sock *sk)
2253{
43815482
ED		 2254 struct socket_wq *wq;
2255
2256 rcu_read_lock();
1da177e4
LT		 2257
2258 /* Do not wake up a writer until he can make "significant"
2259 * progress. --DaveM
2260 */
e71a4783 2261 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2262 wq = rcu_dereference(sk->sk_wq);
2263 if (wq_has_sleeper(wq))
2264 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2265 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2266
2267 /* Should agree with poll, otherwise some programs break */
2268 if (sock_writeable(sk))
8d8ad9d7 2269 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2270 }
2271
43815482 2272 rcu_read_unlock();
1da177e4
LT		 2273}
2274
2275static void sock_def_destruct(struct sock *sk)
2276{
a51482bd 2277 kfree(sk->sk_protinfo);
1da177e4
LT		 2278}
2279
2280void sk_send_sigurg(struct sock *sk)
2281{
2282 if (sk->sk_socket && sk->sk_socket->file)
2283 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2284 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2285}
2a91525c 2286EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2287
2288void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2289 unsigned long expires)
2290{
2291 if (!mod_timer(timer, expires))
2292 sock_hold(sk);
2293}
1da177e4
LT		 2294EXPORT_SYMBOL(sk_reset_timer);
2295
2296void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2297{
25cc4ae9 2298 if (del_timer(timer))
1da177e4
LT		 2299 __sock_put(sk);
2300}
1da177e4
LT		 2301EXPORT_SYMBOL(sk_stop_timer);
2302
2303void sock_init_data(struct socket *sock, struct sock *sk)
2304{
2305 skb_queue_head_init(&sk->sk_receive_queue);
2306 skb_queue_head_init(&sk->sk_write_queue);
2307 skb_queue_head_init(&sk->sk_error_queue);
2308
2309 sk->sk_send_head = NULL;
2310
2311 init_timer(&sk->sk_timer);
4ec93edb 2312
1da177e4
LT		 2313 sk->sk_allocation = GFP_KERNEL;
2314 sk->sk_rcvbuf = sysctl_rmem_default;
2315 sk->sk_sndbuf = sysctl_wmem_default;
2316 sk->sk_state = TCP_CLOSE;
972692e0 2317 sk_set_socket(sk, sock);
1da177e4
LT		 2318
2319 sock_set_flag(sk, SOCK_ZAPPED);
2320
e71a4783 2321 if (sock) {
1da177e4 2322 sk->sk_type = sock->type;
43815482 2323 sk->sk_wq = sock->wq;
1da177e4
LT		 2324 sock->sk = sk;
2325 } else
43815482 2326 sk->sk_wq = NULL;
1da177e4 2327
b6c6712a 2328 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2329 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2330 lockdep_set_class_and_name(&sk->sk_callback_lock,
2331 af_callback_keys + sk->sk_family,
2332 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2333
2334 sk->sk_state_change = sock_def_wakeup;
2335 sk->sk_data_ready = sock_def_readable;
2336 sk->sk_write_space = sock_def_write_space;
2337 sk->sk_error_report = sock_def_error_report;
2338 sk->sk_destruct = sock_def_destruct;
2339
5640f768
ED		 2340 sk->sk_frag.page = NULL;
2341 sk->sk_frag.offset = 0;
ef64a54f 2342 sk->sk_peek_off = -1;
1da177e4 2343
109f6e39
EB		 2344 sk->sk_peer_pid = NULL;
2345 sk->sk_peer_cred = NULL;
1da177e4
LT		 2346 sk->sk_write_pending = 0;
2347 sk->sk_rcvlowat = 1;
2348 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2349 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2350
f37f0afb 2351 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2352
e0d1095a 2353#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2354 sk->sk_napi_id = 0;
64b0dc51 2355 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2356#endif
2357
62748f32 2358 sk->sk_max_pacing_rate = ~0U;
7eec4174 2359 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2360 /*
2361 * Before updating sk_refcnt, we must commit prior changes to memory
2362 * (Documentation/RCU/rculist_nulls.txt for details)
2363 */
2364 smp_wmb();
1da177e4 2365 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2366 atomic_set(&sk->sk_drops, 0);
1da177e4 2367}
2a91525c 2368EXPORT_SYMBOL(sock_init_data);
1da177e4 2369
b5606c2d 2370void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2371{
2372 might_sleep();
a5b5bb9a 2373 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2374 if (sk->sk_lock.owned)
1da177e4 2375 __lock_sock(sk);
d2e9117c 2376 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2377 spin_unlock(&sk->sk_lock.slock);
2378 /*
2379 * The sk_lock has mutex_lock() semantics here:
2380 */
fcc70d5f 2381 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2382 local_bh_enable();
1da177e4 2383}
fcc70d5f 2384EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2385
b5606c2d 2386void release_sock(struct sock *sk)
1da177e4 2387{
a5b5bb9a
IM		 2388 /*
2389 * The sk_lock has mutex_unlock() semantics:
2390 */
2391 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2392
2393 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2394 if (sk->sk_backlog.tail)
2395 __release_sock(sk);
46d3ceab 2396
c3f9b018
ED		 2397 /* Warning : release_cb() might need to release sk ownership,
2398 * ie call sock_release_ownership(sk) before us.
2399 */
46d3ceab
ED		 2400 if (sk->sk_prot->release_cb)
2401 sk->sk_prot->release_cb(sk);
2402
c3f9b018 2403 sock_release_ownership(sk);
a5b5bb9a
IM		 2404 if (waitqueue_active(&sk->sk_lock.wq))
2405 wake_up(&sk->sk_lock.wq);
2406 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2407}
2408EXPORT_SYMBOL(release_sock);
2409
8a74ad60
ED		 2410/**
2411 * lock_sock_fast - fast version of lock_sock
2412 * @sk: socket
2413 *
2414 * This version should be used for very small section, where process wont block
2415 * return false if fast path is taken
2416 * sk_lock.slock locked, owned = 0, BH disabled
2417 * return true if slow path is taken
2418 * sk_lock.slock unlocked, owned = 1, BH enabled
2419 */
2420bool lock_sock_fast(struct sock *sk)
2421{
2422 might_sleep();
2423 spin_lock_bh(&sk->sk_lock.slock);
2424
2425 if (!sk->sk_lock.owned)
2426 /*
2427 * Note : We must disable BH
2428 */
2429 return false;
2430
2431 __lock_sock(sk);
2432 sk->sk_lock.owned = 1;
2433 spin_unlock(&sk->sk_lock.slock);
2434 /*
2435 * The sk_lock has mutex_lock() semantics here:
2436 */
2437 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2438 local_bh_enable();
2439 return true;
2440}
2441EXPORT_SYMBOL(lock_sock_fast);
2442
1da177e4 2443int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2444{
b7aa0bf7 2445 struct timeval tv;
1da177e4 2446 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2447 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2448 tv = ktime_to_timeval(sk->sk_stamp);
2449 if (tv.tv_sec == -1)
1da177e4 2450 return -ENOENT;
b7aa0bf7
ED		 2451 if (tv.tv_sec == 0) {
2452 sk->sk_stamp = ktime_get_real();
2453 tv = ktime_to_timeval(sk->sk_stamp);
2454 }
2455 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2456}
1da177e4
LT		 2457EXPORT_SYMBOL(sock_get_timestamp);
2458
ae40eb1e
ED		 2459int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2460{
2461 struct timespec ts;
2462 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2463 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2464 ts = ktime_to_timespec(sk->sk_stamp);
2465 if (ts.tv_sec == -1)
2466 return -ENOENT;
2467 if (ts.tv_sec == 0) {
2468 sk->sk_stamp = ktime_get_real();
2469 ts = ktime_to_timespec(sk->sk_stamp);
2470 }
2471 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2472}
2473EXPORT_SYMBOL(sock_get_timestampns);
2474
20d49473 2475void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2476{
20d49473 2477 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2478 unsigned long previous_flags = sk->sk_flags;
2479
20d49473
PO		 2480 sock_set_flag(sk, flag);
2481 /*
2482 * we just set one of the two flags which require net
2483 * time stamping, but time stamping might have been on
2484 * already because of the other one
2485 */
08e29af3 2486 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2487 net_enable_timestamp();
1da177e4
LT		 2488 }
2489}
1da177e4 2490
cb820f8e
RC		 2491int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2492 int level, int type)
2493{
2494 struct sock_exterr_skb *serr;
364a9e93 2495 struct sk_buff *skb;
cb820f8e
RC		 2496 int copied, err;
2497
2498 err = -EAGAIN;
364a9e93 2499 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2500 if (skb == NULL)
2501 goto out;
2502
2503 copied = skb->len;
2504 if (copied > len) {
2505 msg->msg_flags |= MSG_TRUNC;
2506 copied = len;
2507 }
51f3d02b 2508 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2509 if (err)
2510 goto out_free_skb;
2511
2512 sock_recv_timestamp(msg, sk, skb);
2513
2514 serr = SKB_EXT_ERR(skb);
2515 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2516
2517 msg->msg_flags |= MSG_ERRQUEUE;
2518 err = copied;
2519
cb820f8e
RC		 2520out_free_skb:
2521 kfree_skb(skb);
2522out:
2523 return err;
2524}
2525EXPORT_SYMBOL(sock_recv_errqueue);
2526
1da177e4
LT		 2527/*
2528 * Get a socket option on an socket.
2529 *
2530 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2531 * asynchronous errors should be reported by getsockopt. We assume
2532 * this means if you specify SO_ERROR (otherwise whats the point of it).
2533 */
2534int sock_common_getsockopt(struct socket *sock, int level, int optname,
2535 char __user *optval, int __user *optlen)
2536{
2537 struct sock *sk = sock->sk;
2538
2539 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2540}
1da177e4
LT		 2541EXPORT_SYMBOL(sock_common_getsockopt);
2542
3fdadf7d 2543#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2544int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2545 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2546{
2547 struct sock *sk = sock->sk;
2548
1e51f951 2549 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2550 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2551 optval, optlen);
3fdadf7d
DM		 2552 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2553}
2554EXPORT_SYMBOL(compat_sock_common_getsockopt);
2555#endif
2556
1b784140
YX		 2557int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2558 int flags)
1da177e4
LT		 2559{
2560 struct sock *sk = sock->sk;
2561 int addr_len = 0;
2562 int err;
2563
1b784140 2564 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2565 flags & ~MSG_DONTWAIT, &addr_len);
2566 if (err >= 0)
2567 msg->msg_namelen = addr_len;
2568 return err;
2569}
1da177e4
LT		 2570EXPORT_SYMBOL(sock_common_recvmsg);
2571
2572/*
2573 * Set socket options on an inet socket.
2574 */
2575int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2576 char __user *optval, unsigned int optlen)
1da177e4
LT		 2577{
2578 struct sock *sk = sock->sk;
2579
2580 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2581}
1da177e4
LT		 2582EXPORT_SYMBOL(sock_common_setsockopt);
2583
3fdadf7d 2584#ifdef CONFIG_COMPAT
543d9cfe 2585int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2586 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2587{
2588 struct sock *sk = sock->sk;
2589
543d9cfe
ACM		 2590 if (sk->sk_prot->compat_setsockopt != NULL)
2591 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2592 optval, optlen);
3fdadf7d
DM		 2593 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2594}
2595EXPORT_SYMBOL(compat_sock_common_setsockopt);
2596#endif
2597
1da177e4
LT		 2598void sk_common_release(struct sock *sk)
2599{
2600 if (sk->sk_prot->destroy)
2601 sk->sk_prot->destroy(sk);
2602
2603 /*
2604 * Observation: when sock_common_release is called, processes have
2605 * no access to socket. But net still has.
2606 * Step one, detach it from networking:
2607 *
2608 * A. Remove from hash tables.
2609 */
2610
2611 sk->sk_prot->unhash(sk);
2612
2613 /*
2614 * In this point socket cannot receive new packets, but it is possible
2615 * that some packets are in flight because some CPU runs receiver and
2616 * did hash table lookup before we unhashed socket. They will achieve
2617 * receive queue and will be purged by socket destructor.
2618 *
2619 * Also we still have packets pending on receive queue and probably,
2620 * our own packets waiting in device queues. sock_destroy will drain
2621 * receive queue, but transmitted packets will delay socket destruction
2622 * until the last reference will be released.
2623 */
2624
2625 sock_orphan(sk);
2626
2627 xfrm_sk_free_policy(sk);
2628
e6848976 2629 sk_refcnt_debug_release(sk);
5640f768
ED		 2630
2631 if (sk->sk_frag.page) {
2632 put_page(sk->sk_frag.page);
2633 sk->sk_frag.page = NULL;
2634 }
2635
1da177e4
LT		 2636 sock_put(sk);
2637}
1da177e4
LT		 2638EXPORT_SYMBOL(sk_common_release);
2639
13ff3d6f
PE		 2640#ifdef CONFIG_PROC_FS
2641#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2642struct prot_inuse {
2643 int val[PROTO_INUSE_NR];
2644};
13ff3d6f
PE		 2645
2646static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2647
2648#ifdef CONFIG_NET_NS
2649void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2650{
d6d9ca0f 2651 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2652}
2653EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2654
2655int sock_prot_inuse_get(struct net *net, struct proto *prot)
2656{
2657 int cpu, idx = prot->inuse_idx;
2658 int res = 0;
2659
2660 for_each_possible_cpu(cpu)
2661 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2662
2663 return res >= 0 ? res : 0;
2664}
2665EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2666
2c8c1e72 2667static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2668{
2669 net->core.inuse = alloc_percpu(struct prot_inuse);
2670 return net->core.inuse ? 0 : -ENOMEM;
2671}
2672
2c8c1e72 2673static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2674{
2675 free_percpu(net->core.inuse);
2676}
2677
2678static struct pernet_operations net_inuse_ops = {
2679 .init = sock_inuse_init_net,
2680 .exit = sock_inuse_exit_net,
2681};
2682
2683static __init int net_inuse_init(void)
2684{
2685 if (register_pernet_subsys(&net_inuse_ops))
2686 panic("Cannot initialize net inuse counters");
2687
2688 return 0;
2689}
2690
2691core_initcall(net_inuse_init);
2692#else
1338d466
PE		 2693static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2694
c29a0bc4 2695void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2696{
d6d9ca0f 2697 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2698}
2699EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2700
c29a0bc4 2701int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2702{
2703 int cpu, idx = prot->inuse_idx;
2704 int res = 0;
2705
2706 for_each_possible_cpu(cpu)
2707 res += per_cpu(prot_inuse, cpu).val[idx];
2708
2709 return res >= 0 ? res : 0;
2710}
2711EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2712#endif
13ff3d6f
PE		 2713
2714static void assign_proto_idx(struct proto *prot)
2715{
2716 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2717
2718 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2719 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2720 return;
2721 }
2722
2723 set_bit(prot->inuse_idx, proto_inuse_idx);
2724}
2725
2726static void release_proto_idx(struct proto *prot)
2727{
2728 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2729 clear_bit(prot->inuse_idx, proto_inuse_idx);
2730}
2731#else
2732static inline void assign_proto_idx(struct proto *prot)
2733{
2734}
2735
2736static inline void release_proto_idx(struct proto *prot)
2737{
2738}
2739#endif
2740
0159dfd3
ED		 2741static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2742{
2743 if (!rsk_prot)
2744 return;
2745 kfree(rsk_prot->slab_name);
2746 rsk_prot->slab_name = NULL;
2747 if (rsk_prot->slab) {
2748 kmem_cache_destroy(rsk_prot->slab);
2749 rsk_prot->slab = NULL;
2750 }
2751}
2752
2753static int req_prot_init(const struct proto *prot)
2754{
2755 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2756
2757 if (!rsk_prot)
2758 return 0;
2759
2760 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2761 prot->name);
2762 if (!rsk_prot->slab_name)
2763 return -ENOMEM;
2764
2765 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2766 rsk_prot->obj_size, 0,
fa76ce73 2767 0, NULL);
0159dfd3
ED		 2768
2769 if (!rsk_prot->slab) {
2770 pr_crit("%s: Can't create request sock SLAB cache!\n",
2771 prot->name);
2772 return -ENOMEM;
2773 }
2774 return 0;
2775}
2776
b733c007
PE		 2777int proto_register(struct proto *prot, int alloc_slab)
2778{
1da177e4
LT		 2779 if (alloc_slab) {
2780 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2781 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2782 NULL);
1da177e4
LT		 2783
2784 if (prot->slab == NULL) {
e005d193
JP		 2785 pr_crit("%s: Can't create sock SLAB cache!\n",
2786 prot->name);
60e7663d 2787 goto out;
1da177e4 2788 }
2e6599cb 2789
0159dfd3
ED		 2790 if (req_prot_init(prot))
2791 goto out_free_request_sock_slab;
8feaf0c0 2792
6d6ee43e 2793 if (prot->twsk_prot != NULL) {
faf23422 2794 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2795
7e56b5d6 2796 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2797 goto out_free_request_sock_slab;
2798
6d6ee43e 2799 prot->twsk_prot->twsk_slab =
7e56b5d6 2800 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2801 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2802 0,
52db70dc 2803 prot->slab_flags,
20c2df83 2804 NULL);
6d6ee43e 2805 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2806 goto out_free_timewait_sock_slab_name;
2807 }
1da177e4
LT		 2808 }
2809
36b77a52 2810 mutex_lock(&proto_list_mutex);
1da177e4 2811 list_add(&prot->node, &proto_list);
13ff3d6f 2812 assign_proto_idx(prot);
36b77a52 2813 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2814 return 0;
2815
8feaf0c0 2816out_free_timewait_sock_slab_name:
7e56b5d6 2817 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2818out_free_request_sock_slab:
0159dfd3
ED		 2819 req_prot_cleanup(prot->rsk_prot);
2820
2e6599cb
ACM		 2821 kmem_cache_destroy(prot->slab);
2822 prot->slab = NULL;
b733c007
PE		 2823out:
2824 return -ENOBUFS;
1da177e4 2825}
1da177e4
LT		 2826EXPORT_SYMBOL(proto_register);
2827
2828void proto_unregister(struct proto *prot)
2829{
36b77a52 2830 mutex_lock(&proto_list_mutex);
13ff3d6f 2831 release_proto_idx(prot);
0a3f4358 2832 list_del(&prot->node);
36b77a52 2833 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2834
2835 if (prot->slab != NULL) {
2836 kmem_cache_destroy(prot->slab);
2837 prot->slab = NULL;
2838 }
2839
0159dfd3 2840 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2841
6d6ee43e 2842 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2843 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2844 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2845 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2846 }
1da177e4 2847}
1da177e4
LT		 2848EXPORT_SYMBOL(proto_unregister);
2849
2850#ifdef CONFIG_PROC_FS
1da177e4 2851static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2852 __acquires(proto_list_mutex)
1da177e4 2853{
36b77a52 2854 mutex_lock(&proto_list_mutex);
60f0438a 2855 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2856}
2857
2858static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2859{
60f0438a 2860 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2861}
2862
2863static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2864 __releases(proto_list_mutex)
1da177e4 2865{
36b77a52 2866 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2867}
2868
2869static char proto_method_implemented(const void *method)
2870{
2871 return method == NULL ? 'n' : 'y';
2872}
180d8cd9
GC		 2873static long sock_prot_memory_allocated(struct proto *proto)
2874{
cb75a36c 2875 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2876}
2877
2878static char *sock_prot_memory_pressure(struct proto *proto)
2879{
2880 return proto->memory_pressure != NULL ?
2881 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2882}
1da177e4
LT		 2883
2884static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2885{
180d8cd9 2886
8d987e5c 2887 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2888 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2889 proto->name,
2890 proto->obj_size,
14e943db 2891 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2892 sock_prot_memory_allocated(proto),
2893 sock_prot_memory_pressure(proto),
1da177e4
LT		 2894 proto->max_header,
2895 proto->slab == NULL ? "no" : "yes",
2896 module_name(proto->owner),
2897 proto_method_implemented(proto->close),
2898 proto_method_implemented(proto->connect),
2899 proto_method_implemented(proto->disconnect),
2900 proto_method_implemented(proto->accept),
2901 proto_method_implemented(proto->ioctl),
2902 proto_method_implemented(proto->init),
2903 proto_method_implemented(proto->destroy),
2904 proto_method_implemented(proto->shutdown),
2905 proto_method_implemented(proto->setsockopt),
2906 proto_method_implemented(proto->getsockopt),
2907 proto_method_implemented(proto->sendmsg),
2908 proto_method_implemented(proto->recvmsg),
2909 proto_method_implemented(proto->sendpage),
2910 proto_method_implemented(proto->bind),
2911 proto_method_implemented(proto->backlog_rcv),
2912 proto_method_implemented(proto->hash),
2913 proto_method_implemented(proto->unhash),
2914 proto_method_implemented(proto->get_port),
2915 proto_method_implemented(proto->enter_memory_pressure));
2916}
2917
2918static int proto_seq_show(struct seq_file *seq, void *v)
2919{
60f0438a 2920 if (v == &proto_list)
1da177e4
LT		 2921 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2922 "protocol",
2923 "size",
2924 "sockets",
2925 "memory",
2926 "press",
2927 "maxhdr",
2928 "slab",
2929 "module",
2930 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2931 else
60f0438a 2932 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2933 return 0;
2934}
2935
f690808e 2936static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2937 .start = proto_seq_start,
2938 .next = proto_seq_next,
2939 .stop = proto_seq_stop,
2940 .show = proto_seq_show,
2941};
2942
2943static int proto_seq_open(struct inode *inode, struct file *file)
2944{
14e943db
ED		 2945 return seq_open_net(inode, file, &proto_seq_ops,
2946 sizeof(struct seq_net_private));
1da177e4
LT		 2947}
2948
9a32144e 2949static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2950 .owner = THIS_MODULE,
2951 .open = proto_seq_open,
2952 .read = seq_read,
2953 .llseek = seq_lseek,
14e943db
ED		 2954 .release = seq_release_net,
2955};
2956
2957static __net_init int proto_init_net(struct net *net)
2958{
d4beaa66 2959 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2960 return -ENOMEM;
2961
2962 return 0;
2963}
2964
2965static __net_exit void proto_exit_net(struct net *net)
2966{
ece31ffd 2967 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2968}
2969
2970
2971static __net_initdata struct pernet_operations proto_net_ops = {
2972 .init = proto_init_net,
2973 .exit = proto_exit_net,
1da177e4
LT		 2974};
2975
2976static int __init proto_init(void)
2977{
14e943db 2978 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2979}
2980
2981subsys_initcall(proto_init);
2982
2983#endif /* PROC_FS */
ubuntu-bionic-kernel mirror