]> 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
packet: support per-packet fwmark for af_packet sendmsg
[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>
eb4cb008 134#include <linux/sock_diag.h>
1da177e4
LT		 135
136#include <linux/filter.h>
137
3847ce32
SM		 138#include <trace/events/sock.h>
139
1da177e4
LT		 140#ifdef CONFIG_INET
141#include <net/tcp.h>
142#endif
143
076bb0c8 144#include <net/busy_poll.h>
06021292 145
36b77a52 146static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 147static LIST_HEAD(proto_list);
148
a3b299da
EB		 149/**
150 * sk_ns_capable - General socket capability test
151 * @sk: Socket to use a capability on or through
152 * @user_ns: The user namespace of the capability to use
153 * @cap: The capability to use
154 *
155 * Test to see if the opener of the socket had when the socket was
156 * created and the current process has the capability @cap in the user
157 * namespace @user_ns.
158 */
159bool sk_ns_capable(const struct sock *sk,
160 struct user_namespace *user_ns, int cap)
161{
162 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
163 ns_capable(user_ns, cap);
164}
165EXPORT_SYMBOL(sk_ns_capable);
166
167/**
168 * sk_capable - Socket global capability test
169 * @sk: Socket to use a capability on or through
e793c0f7 170 * @cap: The global capability to use
a3b299da
EB		 171 *
172 * Test to see if the opener of the socket had when the socket was
173 * created and the current process has the capability @cap in all user
174 * namespaces.
175 */
176bool sk_capable(const struct sock *sk, int cap)
177{
178 return sk_ns_capable(sk, &init_user_ns, cap);
179}
180EXPORT_SYMBOL(sk_capable);
181
182/**
183 * sk_net_capable - Network namespace socket capability test
184 * @sk: Socket to use a capability on or through
185 * @cap: The capability to use
186 *
e793c0f7 187 * Test to see if the opener of the socket had when the socket was created
a3b299da
EB		 188 * and the current process has the capability @cap over the network namespace
189 * the socket is a member of.
190 */
191bool sk_net_capable(const struct sock *sk, int cap)
192{
193 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
194}
195EXPORT_SYMBOL(sk_net_capable);
196
197
c255a458 198#ifdef CONFIG_MEMCG_KMEM
1d62e436 199int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
d1a4c0b3
GC		 200{
201 struct proto *proto;
202 int ret = 0;
203
36b77a52 204 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 205 list_for_each_entry(proto, &proto_list, node) {
206 if (proto->init_cgroup) {
1d62e436 207 ret = proto->init_cgroup(memcg, ss);
d1a4c0b3
GC		 208 if (ret)
209 goto out;
210 }
211 }
212
36b77a52 213 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 214 return ret;
215out:
216 list_for_each_entry_continue_reverse(proto, &proto_list, node)
217 if (proto->destroy_cgroup)
1d62e436 218 proto->destroy_cgroup(memcg);
36b77a52 219 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 220 return ret;
221}
222
1d62e436 223void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
d1a4c0b3
GC		 224{
225 struct proto *proto;
226
36b77a52 227 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 228 list_for_each_entry_reverse(proto, &proto_list, node)
229 if (proto->destroy_cgroup)
1d62e436 230 proto->destroy_cgroup(memcg);
36b77a52 231 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 232}
233#endif
234
da21f24d
IM		 235/*
236 * Each address family might have different locking rules, so we have
237 * one slock key per address family:
238 */
a5b5bb9a
IM		 239static struct lock_class_key af_family_keys[AF_MAX];
240static struct lock_class_key af_family_slock_keys[AF_MAX];
241
cbda4eaf 242#if defined(CONFIG_MEMCG_KMEM)
c5905afb 243struct static_key memcg_socket_limit_enabled;
e1aab161 244EXPORT_SYMBOL(memcg_socket_limit_enabled);
cbda4eaf 245#endif
e1aab161 246
a5b5bb9a
IM		 247/*
248 * Make lock validator output more readable. (we pre-construct these
249 * strings build-time, so that runtime initialization of socket
250 * locks is fast):
251 */
36cbd3dc 252static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 253 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
254 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
255 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
256 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
257 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
258 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
259 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 260 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 261 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 262 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 263 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 264 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 265 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
456db6a4 266 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 267};
36cbd3dc 268static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 269 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
270 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
271 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
272 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
273 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
274 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
275 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 276 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 277 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 278 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 279 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 280 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 281 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 282 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 283};
36cbd3dc 284static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 285 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
286 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
287 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
288 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
289 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
290 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
291 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 292 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 293 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 294 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 295 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 296 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 297 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 298 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 299};
da21f24d
IM		 300
301/*
302 * sk_callback_lock locking rules are per-address-family,
303 * so split the lock classes by using a per-AF key:
304 */
305static struct lock_class_key af_callback_keys[AF_MAX];
306
1da177e4
LT		 307/* Take into consideration the size of the struct sk_buff overhead in the
308 * determination of these values, since that is non-constant across
309 * platforms. This makes socket queueing behavior and performance
310 * not depend upon such differences.
311 */
312#define _SK_MEM_PACKETS 256
87fb4b7b 313#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 314#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
315#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
316
317/* Run time adjustable parameters. */
ab32ea5d 318__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 319EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 320__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 321EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 322__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
323__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 324
25985edc 325/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 326int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 327EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 328
b245be1f
WB		 329int sysctl_tstamp_allow_data __read_mostly = 1;
330
c93bdd0e
MG		 331struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
332EXPORT_SYMBOL_GPL(memalloc_socks);
333
7cb02404
MG		 334/**
335 * sk_set_memalloc - sets %SOCK_MEMALLOC
336 * @sk: socket to set it on
337 *
338 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
339 * It's the responsibility of the admin to adjust min_free_kbytes
340 * to meet the requirements
341 */
342void sk_set_memalloc(struct sock *sk)
343{
344 sock_set_flag(sk, SOCK_MEMALLOC);
345 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 346 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 347}
348EXPORT_SYMBOL_GPL(sk_set_memalloc);
349
350void sk_clear_memalloc(struct sock *sk)
351{
352 sock_reset_flag(sk, SOCK_MEMALLOC);
353 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 354 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 355
356 /*
357 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
5d753610
MG		 358 * progress of swapping. SOCK_MEMALLOC may be cleared while
359 * it has rmem allocations due to the last swapfile being deactivated
360 * but there is a risk that the socket is unusable due to exceeding
361 * the rmem limits. Reclaim the reserves and obey rmem limits again.
c76562b6 362 */
5d753610 363 sk_mem_reclaim(sk);
7cb02404
MG		 364}
365EXPORT_SYMBOL_GPL(sk_clear_memalloc);
366
b4b9e355
MG		 367int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
368{
369 int ret;
370 unsigned long pflags = current->flags;
371
372 /* these should have been dropped before queueing */
373 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
374
375 current->flags |= PF_MEMALLOC;
376 ret = sk->sk_backlog_rcv(sk, skb);
377 tsk_restore_flags(current, pflags, PF_MEMALLOC);
378
379 return ret;
380}
381EXPORT_SYMBOL(__sk_backlog_rcv);
382
1da177e4
LT		 383static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
384{
385 struct timeval tv;
386
387 if (optlen < sizeof(tv))
388 return -EINVAL;
389 if (copy_from_user(&tv, optval, sizeof(tv)))
390 return -EFAULT;
ba78073e
VA		 391 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
392 return -EDOM;
1da177e4 393
ba78073e 394 if (tv.tv_sec < 0) {
6f11df83
AM		 395 static int warned __read_mostly;
396
ba78073e 397 *timeo_p = 0;
50aab54f 398 if (warned < 10 && net_ratelimit()) {
ba78073e 399 warned++;
e005d193
JP		 400 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
401 __func__, current->comm, task_pid_nr(current));
50aab54f 402 }
ba78073e
VA		 403 return 0;
404 }
1da177e4
LT		 405 *timeo_p = MAX_SCHEDULE_TIMEOUT;
406 if (tv.tv_sec == 0 && tv.tv_usec == 0)
407 return 0;
408 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
409 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
410 return 0;
411}
412
413static void sock_warn_obsolete_bsdism(const char *name)
414{
415 static int warned;
416 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 417 if (strcmp(warncomm, current->comm) && warned < 5) {
418 strcpy(warncomm, current->comm);
e005d193
JP		 419 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
420 warncomm, name);
1da177e4
LT		 421 warned++;
422 }
423}
424
08e29af3
ED		 425#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
426
427static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 428{
08e29af3
ED		 429 if (sk->sk_flags & flags) {
430 sk->sk_flags &= ~flags;
431 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 432 net_disable_timestamp();
1da177e4
LT		 433 }
434}
435
436
f0088a50
DV		 437int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
438{
766e9037 439 int err;
3b885787
NH		 440 unsigned long flags;
441 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 442
0fd7bac6 443 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 444 atomic_inc(&sk->sk_drops);
3847ce32 445 trace_sock_rcvqueue_full(sk, skb);
766e9037 446 return -ENOMEM;
f0088a50
DV		 447 }
448
fda9ef5d 449 err = sk_filter(sk, skb);
f0088a50 450 if (err)
766e9037 451 return err;
f0088a50 452
c76562b6 453 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 454 atomic_inc(&sk->sk_drops);
455 return -ENOBUFS;
3ab224be
HA		 456 }
457
f0088a50
DV		 458 skb->dev = NULL;
459 skb_set_owner_r(skb, sk);
49ad9599 460
7fee226a
ED		 461 /* we escape from rcu protected region, make sure we dont leak
462 * a norefcounted dst
463 */
464 skb_dst_force(skb);
465
3b885787 466 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 467 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 468 __skb_queue_tail(list, skb);
469 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 470
471 if (!sock_flag(sk, SOCK_DEAD))
676d2369 472 sk->sk_data_ready(sk);
766e9037 473 return 0;
f0088a50
DV		 474}
475EXPORT_SYMBOL(sock_queue_rcv_skb);
476
58a5a7b9 477int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 478{
479 int rc = NET_RX_SUCCESS;
480
fda9ef5d 481 if (sk_filter(sk, skb))
f0088a50
DV		 482 goto discard_and_relse;
483
484 skb->dev = NULL;
485
274f482d 486 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 487 atomic_inc(&sk->sk_drops);
488 goto discard_and_relse;
489 }
58a5a7b9
ACM		 490 if (nested)
491 bh_lock_sock_nested(sk);
492 else
493 bh_lock_sock(sk);
a5b5bb9a
IM		 494 if (!sock_owned_by_user(sk)) {
495 /*
496 * trylock + unlock semantics:
497 */
498 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
499
c57943a1 500 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 501
502 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 503 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 504 bh_unlock_sock(sk);
505 atomic_inc(&sk->sk_drops);
506 goto discard_and_relse;
507 }
508
f0088a50
DV		 509 bh_unlock_sock(sk);
510out:
511 sock_put(sk);
512 return rc;
513discard_and_relse:
514 kfree_skb(skb);
515 goto out;
516}
517EXPORT_SYMBOL(sk_receive_skb);
518
519struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
520{
b6c6712a 521 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 522
523 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 524 sk_tx_queue_clear(sk);
a9b3cd7f 525 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 526 dst_release(dst);
527 return NULL;
528 }
529
530 return dst;
531}
532EXPORT_SYMBOL(__sk_dst_check);
533
534struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
535{
536 struct dst_entry *dst = sk_dst_get(sk);
537
538 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
539 sk_dst_reset(sk);
540 dst_release(dst);
541 return NULL;
542 }
543
544 return dst;
545}
546EXPORT_SYMBOL(sk_dst_check);
547
c91f6df2
BH		 548static int sock_setbindtodevice(struct sock *sk, char __user *optval,
549 int optlen)
4878809f
DM		 550{
551 int ret = -ENOPROTOOPT;
552#ifdef CONFIG_NETDEVICES
3b1e0a65 553 struct net *net = sock_net(sk);
4878809f
DM		 554 char devname[IFNAMSIZ];
555 int index;
556
557 /* Sorry... */
558 ret = -EPERM;
5e1fccc0 559 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 560 goto out;
561
562 ret = -EINVAL;
563 if (optlen < 0)
564 goto out;
565
566 /* Bind this socket to a particular device like "eth0",
567 * as specified in the passed interface name. If the
568 * name is "" or the option length is zero the socket
569 * is not bound.
570 */
571 if (optlen > IFNAMSIZ - 1)
572 optlen = IFNAMSIZ - 1;
573 memset(devname, 0, sizeof(devname));
574
575 ret = -EFAULT;
576 if (copy_from_user(devname, optval, optlen))
577 goto out;
578
000ba2e4
DM		 579 index = 0;
580 if (devname[0] != '\0') {
bf8e56bf 581 struct net_device *dev;
4878809f 582
bf8e56bf
ED		 583 rcu_read_lock();
584 dev = dev_get_by_name_rcu(net, devname);
585 if (dev)
586 index = dev->ifindex;
587 rcu_read_unlock();
4878809f
DM		 588 ret = -ENODEV;
589 if (!dev)
590 goto out;
4878809f
DM		 591 }
592
593 lock_sock(sk);
594 sk->sk_bound_dev_if = index;
595 sk_dst_reset(sk);
596 release_sock(sk);
597
598 ret = 0;
599
600out:
601#endif
602
603 return ret;
604}
605
c91f6df2
BH		 606static int sock_getbindtodevice(struct sock *sk, char __user *optval,
607 int __user *optlen, int len)
608{
609 int ret = -ENOPROTOOPT;
610#ifdef CONFIG_NETDEVICES
611 struct net *net = sock_net(sk);
c91f6df2 612 char devname[IFNAMSIZ];
c91f6df2
BH		 613
614 if (sk->sk_bound_dev_if == 0) {
615 len = 0;
616 goto zero;
617 }
618
619 ret = -EINVAL;
620 if (len < IFNAMSIZ)
621 goto out;
622
5dbe7c17
NS		 623 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
624 if (ret)
c91f6df2 625 goto out;
c91f6df2
BH		 626
627 len = strlen(devname) + 1;
628
629 ret = -EFAULT;
630 if (copy_to_user(optval, devname, len))
631 goto out;
632
633zero:
634 ret = -EFAULT;
635 if (put_user(len, optlen))
636 goto out;
637
638 ret = 0;
639
640out:
641#endif
642
643 return ret;
644}
645
c0ef877b
PE		 646static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
647{
648 if (valbool)
649 sock_set_flag(sk, bit);
650 else
651 sock_reset_flag(sk, bit);
652}
653
f60e5990 654bool sk_mc_loop(struct sock *sk)
655{
656 if (dev_recursion_level())
657 return false;
658 if (!sk)
659 return true;
660 switch (sk->sk_family) {
661 case AF_INET:
662 return inet_sk(sk)->mc_loop;
663#if IS_ENABLED(CONFIG_IPV6)
664 case AF_INET6:
665 return inet6_sk(sk)->mc_loop;
666#endif
667 }
668 WARN_ON(1);
669 return true;
670}
671EXPORT_SYMBOL(sk_mc_loop);
672
1da177e4
LT		 673/*
674 * This is meant for all protocols to use and covers goings on
675 * at the socket level. Everything here is generic.
676 */
677
678int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 679 char __user *optval, unsigned int optlen)
1da177e4 680{
2a91525c 681 struct sock *sk = sock->sk;
1da177e4
LT		 682 int val;
683 int valbool;
684 struct linger ling;
685 int ret = 0;
4ec93edb 686
1da177e4
LT		 687 /*
688 * Options without arguments
689 */
690
4878809f 691 if (optname == SO_BINDTODEVICE)
c91f6df2 692 return sock_setbindtodevice(sk, optval, optlen);
4878809f 693
e71a4783
SH		 694 if (optlen < sizeof(int))
695 return -EINVAL;
4ec93edb 696
1da177e4
LT		 697 if (get_user(val, (int __user *)optval))
698 return -EFAULT;
4ec93edb 699
2a91525c 700 valbool = val ? 1 : 0;
1da177e4
LT		 701
702 lock_sock(sk);
703
2a91525c 704 switch (optname) {
e71a4783 705 case SO_DEBUG:
2a91525c 706 if (val && !capable(CAP_NET_ADMIN))
e71a4783 707 ret = -EACCES;
2a91525c 708 else
c0ef877b 709 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 710 break;
711 case SO_REUSEADDR:
4a17fd52 712 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 713 break;
055dc21a
TH		 714 case SO_REUSEPORT:
715 sk->sk_reuseport = valbool;
716 break;
e71a4783 717 case SO_TYPE:
49c794e9 718 case SO_PROTOCOL:
0d6038ee 719 case SO_DOMAIN:
e71a4783
SH		 720 case SO_ERROR:
721 ret = -ENOPROTOOPT;
722 break;
723 case SO_DONTROUTE:
c0ef877b 724 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 725 break;
726 case SO_BROADCAST:
727 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
728 break;
729 case SO_SNDBUF:
730 /* Don't error on this BSD doesn't and if you think
82981930
ED		 731 * about it this is right. Otherwise apps have to
732 * play 'guess the biggest size' games. RCVBUF/SNDBUF
733 * are treated in BSD as hints
734 */
735 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 736set_sndbuf:
e71a4783 737 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 738 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
739 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 740 sk->sk_write_space(sk);
741 break;
1da177e4 742
e71a4783
SH		 743 case SO_SNDBUFFORCE:
744 if (!capable(CAP_NET_ADMIN)) {
745 ret = -EPERM;
746 break;
747 }
748 goto set_sndbuf;
b0573dea 749
e71a4783
SH		 750 case SO_RCVBUF:
751 /* Don't error on this BSD doesn't and if you think
82981930
ED		 752 * about it this is right. Otherwise apps have to
753 * play 'guess the biggest size' games. RCVBUF/SNDBUF
754 * are treated in BSD as hints
755 */
756 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 757set_rcvbuf:
e71a4783
SH		 758 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
759 /*
760 * We double it on the way in to account for
761 * "struct sk_buff" etc. overhead. Applications
762 * assume that the SO_RCVBUF setting they make will
763 * allow that much actual data to be received on that
764 * socket.
765 *
766 * Applications are unaware that "struct sk_buff" and
767 * other overheads allocate from the receive buffer
768 * during socket buffer allocation.
769 *
770 * And after considering the possible alternatives,
771 * returning the value we actually used in getsockopt
772 * is the most desirable behavior.
773 */
82981930 774 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 775 break;
776
777 case SO_RCVBUFFORCE:
778 if (!capable(CAP_NET_ADMIN)) {
779 ret = -EPERM;
1da177e4 780 break;
e71a4783
SH		 781 }
782 goto set_rcvbuf;
1da177e4 783
e71a4783 784 case SO_KEEPALIVE:
1da177e4 785#ifdef CONFIG_INET
3e10986d
ED		 786 if (sk->sk_protocol == IPPROTO_TCP &&
787 sk->sk_type == SOCK_STREAM)
e71a4783 788 tcp_set_keepalive(sk, valbool);
1da177e4 789#endif
e71a4783
SH		 790 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
791 break;
792
793 case SO_OOBINLINE:
794 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
795 break;
796
797 case SO_NO_CHECK:
28448b80 798 sk->sk_no_check_tx = valbool;
e71a4783
SH		 799 break;
800
801 case SO_PRIORITY:
5e1fccc0
EB		 802 if ((val >= 0 && val <= 6) ||
803 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 804 sk->sk_priority = val;
805 else
806 ret = -EPERM;
807 break;
808
809 case SO_LINGER:
810 if (optlen < sizeof(ling)) {
811 ret = -EINVAL; /* 1003.1g */
1da177e4 812 break;
e71a4783 813 }
2a91525c 814 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 815 ret = -EFAULT;
1da177e4 816 break;
e71a4783
SH		 817 }
818 if (!ling.l_onoff)
819 sock_reset_flag(sk, SOCK_LINGER);
820 else {
1da177e4 821#if (BITS_PER_LONG == 32)
e71a4783
SH		 822 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
823 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 824 else
e71a4783
SH		 825#endif
826 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
827 sock_set_flag(sk, SOCK_LINGER);
828 }
829 break;
830
831 case SO_BSDCOMPAT:
832 sock_warn_obsolete_bsdism("setsockopt");
833 break;
834
835 case SO_PASSCRED:
836 if (valbool)
837 set_bit(SOCK_PASSCRED, &sock->flags);
838 else
839 clear_bit(SOCK_PASSCRED, &sock->flags);
840 break;
841
842 case SO_TIMESTAMP:
92f37fd2 843 case SO_TIMESTAMPNS:
e71a4783 844 if (valbool) {
92f37fd2
ED		 845 if (optname == SO_TIMESTAMP)
846 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
847 else
848 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 849 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 850 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 851 } else {
e71a4783 852 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 853 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
854 }
e71a4783
SH		 855 break;
856
20d49473
PO		 857 case SO_TIMESTAMPING:
858 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 859 ret = -EINVAL;
20d49473
PO		 860 break;
861 }
b245be1f 862
09c2d251 863 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765
WB		 864 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
865 if (sk->sk_protocol == IPPROTO_TCP) {
866 if (sk->sk_state != TCP_ESTABLISHED) {
867 ret = -EINVAL;
868 break;
869 }
870 sk->sk_tskey = tcp_sk(sk)->snd_una;
871 } else {
872 sk->sk_tskey = 0;
873 }
874 }
b9f40e21 875 sk->sk_tsflags = val;
20d49473
PO		 876 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
877 sock_enable_timestamp(sk,
878 SOCK_TIMESTAMPING_RX_SOFTWARE);
879 else
880 sock_disable_timestamp(sk,
08e29af3 881 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 882 break;
883
e71a4783
SH		 884 case SO_RCVLOWAT:
885 if (val < 0)
886 val = INT_MAX;
887 sk->sk_rcvlowat = val ? : 1;
888 break;
889
890 case SO_RCVTIMEO:
891 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
892 break;
893
894 case SO_SNDTIMEO:
895 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
896 break;
1da177e4 897
e71a4783
SH		 898 case SO_ATTACH_FILTER:
899 ret = -EINVAL;
900 if (optlen == sizeof(struct sock_fprog)) {
901 struct sock_fprog fprog;
1da177e4 902
e71a4783
SH		 903 ret = -EFAULT;
904 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 905 break;
e71a4783
SH		 906
907 ret = sk_attach_filter(&fprog, sk);
908 }
909 break;
910
89aa0758
AS		 911 case SO_ATTACH_BPF:
912 ret = -EINVAL;
913 if (optlen == sizeof(u32)) {
914 u32 ufd;
915
916 ret = -EFAULT;
917 if (copy_from_user(&ufd, optval, sizeof(ufd)))
918 break;
919
920 ret = sk_attach_bpf(ufd, sk);
921 }
922 break;
923
e71a4783 924 case SO_DETACH_FILTER:
55b33325 925 ret = sk_detach_filter(sk);
e71a4783 926 break;
1da177e4 927
d59577b6
VB		 928 case SO_LOCK_FILTER:
929 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
930 ret = -EPERM;
931 else
932 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
933 break;
934
e71a4783
SH		 935 case SO_PASSSEC:
936 if (valbool)
937 set_bit(SOCK_PASSSEC, &sock->flags);
938 else
939 clear_bit(SOCK_PASSSEC, &sock->flags);
940 break;
4a19ec58 941 case SO_MARK:
5e1fccc0 942 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 943 ret = -EPERM;
2a91525c 944 else
4a19ec58 945 sk->sk_mark = val;
4a19ec58 946 break;
877ce7c1 947
3b885787 948 case SO_RXQ_OVFL:
8083f0fc 949 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 950 break;
6e3e939f
JB		 951
952 case SO_WIFI_STATUS:
953 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
954 break;
955
ef64a54f
PE		 956 case SO_PEEK_OFF:
957 if (sock->ops->set_peek_off)
12663bfc 958 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 959 else
960 ret = -EOPNOTSUPP;
961 break;
3bdc0eba
BG		 962
963 case SO_NOFCS:
964 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
965 break;
966
7d4c04fc
KJ		 967 case SO_SELECT_ERR_QUEUE:
968 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
969 break;
970
e0d1095a 971#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 972 case SO_BUSY_POLL:
dafcc438
ET		 973 /* allow unprivileged users to decrease the value */
974 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
975 ret = -EPERM;
976 else {
977 if (val < 0)
978 ret = -EINVAL;
979 else
980 sk->sk_ll_usec = val;
981 }
982 break;
983#endif
62748f32
ED		 984
985 case SO_MAX_PACING_RATE:
986 sk->sk_max_pacing_rate = val;
987 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
988 sk->sk_max_pacing_rate);
989 break;
990
e71a4783
SH		 991 default:
992 ret = -ENOPROTOOPT;
993 break;
4ec93edb 994 }
1da177e4
LT		 995 release_sock(sk);
996 return ret;
997}
2a91525c 998EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 999
1000
8f09898b 1001static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1002 struct ucred *ucred)
3f551f94
EB		 1003{
1004 ucred->pid = pid_vnr(pid);
1005 ucred->uid = ucred->gid = -1;
1006 if (cred) {
1007 struct user_namespace *current_ns = current_user_ns();
1008
b2e4f544
EB		 1009 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1010 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1011 }
1012}
1013
1da177e4
LT		 1014int sock_getsockopt(struct socket *sock, int level, int optname,
1015 char __user *optval, int __user *optlen)
1016{
1017 struct sock *sk = sock->sk;
4ec93edb 1018
e71a4783 1019 union {
4ec93edb
YH		 1020 int val;
1021 struct linger ling;
1da177e4
LT		 1022 struct timeval tm;
1023 } v;
4ec93edb 1024
4d0392be 1025 int lv = sizeof(int);
1da177e4 1026 int len;
4ec93edb 1027
e71a4783 1028 if (get_user(len, optlen))
4ec93edb 1029 return -EFAULT;
e71a4783 1030 if (len < 0)
1da177e4 1031 return -EINVAL;
4ec93edb 1032
50fee1de 1033 memset(&v, 0, sizeof(v));
df0bca04 1034
2a91525c 1035 switch (optname) {
e71a4783
SH		 1036 case SO_DEBUG:
1037 v.val = sock_flag(sk, SOCK_DBG);
1038 break;
1039
1040 case SO_DONTROUTE:
1041 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1042 break;
1043
1044 case SO_BROADCAST:
1b23a5df 1045 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1046 break;
1047
1048 case SO_SNDBUF:
1049 v.val = sk->sk_sndbuf;
1050 break;
1051
1052 case SO_RCVBUF:
1053 v.val = sk->sk_rcvbuf;
1054 break;
1055
1056 case SO_REUSEADDR:
1057 v.val = sk->sk_reuse;
1058 break;
1059
055dc21a
TH		 1060 case SO_REUSEPORT:
1061 v.val = sk->sk_reuseport;
1062 break;
1063
e71a4783 1064 case SO_KEEPALIVE:
1b23a5df 1065 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1066 break;
1067
1068 case SO_TYPE:
1069 v.val = sk->sk_type;
1070 break;
1071
49c794e9
JE		 1072 case SO_PROTOCOL:
1073 v.val = sk->sk_protocol;
1074 break;
1075
0d6038ee
JE		 1076 case SO_DOMAIN:
1077 v.val = sk->sk_family;
1078 break;
1079
e71a4783
SH		 1080 case SO_ERROR:
1081 v.val = -sock_error(sk);
2a91525c 1082 if (v.val == 0)
e71a4783
SH		 1083 v.val = xchg(&sk->sk_err_soft, 0);
1084 break;
1085
1086 case SO_OOBINLINE:
1b23a5df 1087 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1088 break;
1089
1090 case SO_NO_CHECK:
28448b80 1091 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1092 break;
1093
1094 case SO_PRIORITY:
1095 v.val = sk->sk_priority;
1096 break;
1097
1098 case SO_LINGER:
1099 lv = sizeof(v.ling);
1b23a5df 1100 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1101 v.ling.l_linger = sk->sk_lingertime / HZ;
1102 break;
1103
1104 case SO_BSDCOMPAT:
1105 sock_warn_obsolete_bsdism("getsockopt");
1106 break;
1107
1108 case SO_TIMESTAMP:
92f37fd2
ED		 1109 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1110 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1111 break;
1112
1113 case SO_TIMESTAMPNS:
1114 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1115 break;
1116
20d49473 1117 case SO_TIMESTAMPING:
b9f40e21 1118 v.val = sk->sk_tsflags;
20d49473
PO		 1119 break;
1120
e71a4783 1121 case SO_RCVTIMEO:
2a91525c 1122 lv = sizeof(struct timeval);
e71a4783
SH		 1123 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1124 v.tm.tv_sec = 0;
1125 v.tm.tv_usec = 0;
1126 } else {
1127 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1128 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1129 }
1130 break;
1131
1132 case SO_SNDTIMEO:
2a91525c 1133 lv = sizeof(struct timeval);
e71a4783
SH		 1134 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1135 v.tm.tv_sec = 0;
1136 v.tm.tv_usec = 0;
1137 } else {
1138 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1139 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1140 }
1141 break;
1da177e4 1142
e71a4783
SH		 1143 case SO_RCVLOWAT:
1144 v.val = sk->sk_rcvlowat;
1145 break;
1da177e4 1146
e71a4783 1147 case SO_SNDLOWAT:
2a91525c 1148 v.val = 1;
e71a4783 1149 break;
1da177e4 1150
e71a4783 1151 case SO_PASSCRED:
82981930 1152 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1153 break;
1da177e4 1154
e71a4783 1155 case SO_PEERCRED:
109f6e39
EB		 1156 {
1157 struct ucred peercred;
1158 if (len > sizeof(peercred))
1159 len = sizeof(peercred);
1160 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1161 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1162 return -EFAULT;
1163 goto lenout;
109f6e39 1164 }
1da177e4 1165
e71a4783
SH		 1166 case SO_PEERNAME:
1167 {
1168 char address[128];
1169
1170 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1171 return -ENOTCONN;
1172 if (lv < len)
1173 return -EINVAL;
1174 if (copy_to_user(optval, address, len))
1175 return -EFAULT;
1176 goto lenout;
1177 }
1da177e4 1178
e71a4783
SH		 1179 /* Dubious BSD thing... Probably nobody even uses it, but
1180 * the UNIX standard wants it for whatever reason... -DaveM
1181 */
1182 case SO_ACCEPTCONN:
1183 v.val = sk->sk_state == TCP_LISTEN;
1184 break;
1da177e4 1185
e71a4783 1186 case SO_PASSSEC:
82981930 1187 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1188 break;
877ce7c1 1189
e71a4783
SH		 1190 case SO_PEERSEC:
1191 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1192
4a19ec58
LAT		 1193 case SO_MARK:
1194 v.val = sk->sk_mark;
1195 break;
1196
3b885787 1197 case SO_RXQ_OVFL:
1b23a5df 1198 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1199 break;
1200
6e3e939f 1201 case SO_WIFI_STATUS:
1b23a5df 1202 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1203 break;
1204
ef64a54f
PE		 1205 case SO_PEEK_OFF:
1206 if (!sock->ops->set_peek_off)
1207 return -EOPNOTSUPP;
1208
1209 v.val = sk->sk_peek_off;
1210 break;
bc2f7996 1211 case SO_NOFCS:
1b23a5df 1212 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1213 break;
c91f6df2 1214
f7b86bfe 1215 case SO_BINDTODEVICE:
c91f6df2
BH		 1216 return sock_getbindtodevice(sk, optval, optlen, len);
1217
a8fc9277
PE		 1218 case SO_GET_FILTER:
1219 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1220 if (len < 0)
1221 return len;
1222
1223 goto lenout;
c91f6df2 1224
d59577b6
VB		 1225 case SO_LOCK_FILTER:
1226 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1227 break;
1228
ea02f941
MS		 1229 case SO_BPF_EXTENSIONS:
1230 v.val = bpf_tell_extensions();
1231 break;
1232
7d4c04fc
KJ		 1233 case SO_SELECT_ERR_QUEUE:
1234 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1235 break;
1236
e0d1095a 1237#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1238 case SO_BUSY_POLL:
dafcc438
ET		 1239 v.val = sk->sk_ll_usec;
1240 break;
1241#endif
1242
62748f32
ED		 1243 case SO_MAX_PACING_RATE:
1244 v.val = sk->sk_max_pacing_rate;
1245 break;
1246
2c8c56e1
ED		 1247 case SO_INCOMING_CPU:
1248 v.val = sk->sk_incoming_cpu;
1249 break;
1250
e71a4783 1251 default:
443b5991
YH		 1252 /* We implement the SO_SNDLOWAT etc to not be settable
1253 * (1003.1g 7).
1254 */
e71a4783 1255 return -ENOPROTOOPT;
1da177e4 1256 }
e71a4783 1257
1da177e4
LT		 1258 if (len > lv)
1259 len = lv;
1260 if (copy_to_user(optval, &v, len))
1261 return -EFAULT;
1262lenout:
4ec93edb
YH		 1263 if (put_user(len, optlen))
1264 return -EFAULT;
1265 return 0;
1da177e4
LT		 1266}
1267
a5b5bb9a
IM		 1268/*
1269 * Initialize an sk_lock.
1270 *
1271 * (We also register the sk_lock with the lock validator.)
1272 */
b6f99a21 1273static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1274{
ed07536e
PZ		 1275 sock_lock_init_class_and_name(sk,
1276 af_family_slock_key_strings[sk->sk_family],
1277 af_family_slock_keys + sk->sk_family,
1278 af_family_key_strings[sk->sk_family],
1279 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1280}
1281
4dc6dc71
ED		 1282/*
1283 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1284 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1285 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1286 */
f1a6c4da
PE		 1287static void sock_copy(struct sock *nsk, const struct sock *osk)
1288{
1289#ifdef CONFIG_SECURITY_NETWORK
1290 void *sptr = nsk->sk_security;
1291#endif
68835aba
ED		 1292 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1293
1294 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1295 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1296
f1a6c4da
PE		 1297#ifdef CONFIG_SECURITY_NETWORK
1298 nsk->sk_security = sptr;
1299 security_sk_clone(osk, nsk);
1300#endif
1301}
1302
fcbdf09d
OP		 1303void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1304{
1305 unsigned long nulls1, nulls2;
1306
1307 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1308 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1309 if (nulls1 > nulls2)
1310 swap(nulls1, nulls2);
1311
1312 if (nulls1 != 0)
1313 memset((char *)sk, 0, nulls1);
1314 memset((char *)sk + nulls1 + sizeof(void *), 0,
1315 nulls2 - nulls1 - sizeof(void *));
1316 memset((char *)sk + nulls2 + sizeof(void *), 0,
1317 size - nulls2 - sizeof(void *));
1318}
1319EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1320
2e4afe7b
PE		 1321static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1322 int family)
c308c1b2
PE		 1323{
1324 struct sock *sk;
1325 struct kmem_cache *slab;
1326
1327 slab = prot->slab;
e912b114
ED		 1328 if (slab != NULL) {
1329 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1330 if (!sk)
1331 return sk;
1332 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1333 if (prot->clear_sk)
1334 prot->clear_sk(sk, prot->obj_size);
1335 else
1336 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1337 }
fcbdf09d 1338 } else
c308c1b2
PE		 1339 sk = kmalloc(prot->obj_size, priority);
1340
2e4afe7b 1341 if (sk != NULL) {
a98b65a3
VN		 1342 kmemcheck_annotate_bitfield(sk, flags);
1343
2e4afe7b
PE		 1344 if (security_sk_alloc(sk, family, priority))
1345 goto out_free;
1346
1347 if (!try_module_get(prot->owner))
1348 goto out_free_sec;
e022f0b4 1349 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1350 }
1351
c308c1b2 1352 return sk;
2e4afe7b
PE		 1353
1354out_free_sec:
1355 security_sk_free(sk);
1356out_free:
1357 if (slab != NULL)
1358 kmem_cache_free(slab, sk);
1359 else
1360 kfree(sk);
1361 return NULL;
c308c1b2
PE		 1362}
1363
1364static void sk_prot_free(struct proto *prot, struct sock *sk)
1365{
1366 struct kmem_cache *slab;
2e4afe7b 1367 struct module *owner;
c308c1b2 1368
2e4afe7b 1369 owner = prot->owner;
c308c1b2 1370 slab = prot->slab;
2e4afe7b
PE		 1371
1372 security_sk_free(sk);
c308c1b2
PE		 1373 if (slab != NULL)
1374 kmem_cache_free(slab, sk);
1375 else
1376 kfree(sk);
2e4afe7b 1377 module_put(owner);
c308c1b2
PE		 1378}
1379
86f8515f 1380#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1381void sock_update_netprioidx(struct sock *sk)
5bc1421e 1382{
5bc1421e
NH		 1383 if (in_interrupt())
1384 return;
2b73bc65 1385
6ffd4641 1386 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1387}
1388EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1389#endif
1390
1da177e4
LT		 1391/**
1392 * sk_alloc - All socket objects are allocated here
c4ea43c5 1393 * @net: the applicable net namespace
4dc3b16b
PP		 1394 * @family: protocol family
1395 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1396 * @prot: struct proto associated with this new sock instance
11aa9c28 1397 * @kern: is this to be a kernel socket?
1da177e4 1398 */
1b8d7ae4 1399struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1400 struct proto *prot, int kern)
1da177e4 1401{
c308c1b2 1402 struct sock *sk;
1da177e4 1403
154adbc8 1404 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1405 if (sk) {
154adbc8
PE		 1406 sk->sk_family = family;
1407 /*
1408 * See comment in struct sock definition to understand
1409 * why we need sk_prot_creator -acme
1410 */
1411 sk->sk_prot = sk->sk_prot_creator = prot;
1412 sock_lock_init(sk);
26abe143
EB		 1413 sk->sk_net_refcnt = kern ? 0 : 1;
1414 if (likely(sk->sk_net_refcnt))
1415 get_net(net);
1416 sock_net_set(sk, net);
d66ee058 1417 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1418
211d2f97 1419 sock_update_classid(sk);
6ffd4641 1420 sock_update_netprioidx(sk);
1da177e4 1421 }
a79af59e 1422
2e4afe7b 1423 return sk;
1da177e4 1424}
2a91525c 1425EXPORT_SYMBOL(sk_alloc);
1da177e4 1426
eb4cb008 1427void sk_destruct(struct sock *sk)
1da177e4
LT		 1428{
1429 struct sk_filter *filter;
1da177e4
LT		 1430
1431 if (sk->sk_destruct)
1432 sk->sk_destruct(sk);
1433
a898def2
PM		 1434 filter = rcu_dereference_check(sk->sk_filter,
1435 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1436 if (filter) {
309dd5fc 1437 sk_filter_uncharge(sk, filter);
a9b3cd7f 1438 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1439 }
1440
08e29af3 1441 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1442
1443 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1444 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1445 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1446
109f6e39
EB		 1447 if (sk->sk_peer_cred)
1448 put_cred(sk->sk_peer_cred);
1449 put_pid(sk->sk_peer_pid);
26abe143
EB		 1450 if (likely(sk->sk_net_refcnt))
1451 put_net(sock_net(sk));
c308c1b2 1452 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1453}
2b85a34e 1454
eb4cb008
CG		 1455static void __sk_free(struct sock *sk)
1456{
b922622e 1457 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1458 sock_diag_broadcast_destroy(sk);
1459 else
1460 sk_destruct(sk);
1461}
1462
2b85a34e
ED		 1463void sk_free(struct sock *sk)
1464{
1465 /*
25985edc 1466 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1467 * some packets are still in some tx queue.
1468 * If not null, sock_wfree() will call __sk_free(sk) later
1469 */
1470 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1471 __sk_free(sk);
1472}
2a91525c 1473EXPORT_SYMBOL(sk_free);
1da177e4 1474
475f1b52
SR		 1475static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1476{
1477 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1478 sock_update_memcg(newsk);
1479}
1480
e56c57d0
ED		 1481/**
1482 * sk_clone_lock - clone a socket, and lock its clone
1483 * @sk: the socket to clone
1484 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1485 *
1486 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1487 */
1488struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1489{
8fd1d178 1490 struct sock *newsk;
278571ba 1491 bool is_charged = true;
87d11ceb 1492
8fd1d178 1493 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1494 if (newsk != NULL) {
1495 struct sk_filter *filter;
1496
892c141e 1497 sock_copy(newsk, sk);
87d11ceb
ACM		 1498
1499 /* SANITY */
8a681736
SV		 1500 if (likely(newsk->sk_net_refcnt))
1501 get_net(sock_net(newsk));
87d11ceb
ACM		 1502 sk_node_init(&newsk->sk_node);
1503 sock_lock_init(newsk);
1504 bh_lock_sock(newsk);
fa438ccf 1505 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1506 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1507
1508 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1509 /*
1510 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1511 */
1512 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1513 atomic_set(&newsk->sk_omem_alloc, 0);
1514 skb_queue_head_init(&newsk->sk_receive_queue);
1515 skb_queue_head_init(&newsk->sk_write_queue);
1516
b6c6712a 1517 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1518 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1519 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1520 af_callback_keys + newsk->sk_family,
1521 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1522
1523 newsk->sk_dst_cache = NULL;
1524 newsk->sk_wmem_queued = 0;
1525 newsk->sk_forward_alloc = 0;
1526 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1527 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1528
1529 sock_reset_flag(newsk, SOCK_DONE);
1530 skb_queue_head_init(&newsk->sk_error_queue);
1531
0d7da9dd 1532 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1533 if (filter != NULL)
278571ba
AS		 1534 /* though it's an empty new sock, the charging may fail
1535 * if sysctl_optmem_max was changed between creation of
1536 * original socket and cloning
1537 */
1538 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1539
278571ba 1540 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk))) {
87d11ceb
ACM		 1541 /* It is still raw copy of parent, so invalidate
1542 * destructor and make plain sk_free() */
1543 newsk->sk_destruct = NULL;
b0691c8e 1544 bh_unlock_sock(newsk);
87d11ceb
ACM		 1545 sk_free(newsk);
1546 newsk = NULL;
1547 goto out;
1548 }
1549
1550 newsk->sk_err = 0;
1551 newsk->sk_priority = 0;
2c8c56e1 1552 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1553 atomic64_set(&newsk->sk_cookie, 0);
4dc6dc71
ED		 1554 /*
1555 * Before updating sk_refcnt, we must commit prior changes to memory
1556 * (Documentation/RCU/rculist_nulls.txt for details)
1557 */
1558 smp_wmb();
87d11ceb
ACM		 1559 atomic_set(&newsk->sk_refcnt, 2);
1560
1561 /*
1562 * Increment the counter in the same struct proto as the master
1563 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1564 * is the same as sk->sk_prot->socks, as this field was copied
1565 * with memcpy).
1566 *
1567 * This _changes_ the previous behaviour, where
1568 * tcp_create_openreq_child always was incrementing the
1569 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1570 * to be taken into account in all callers. -acme
1571 */
1572 sk_refcnt_debug_inc(newsk);
972692e0 1573 sk_set_socket(newsk, NULL);
43815482 1574 newsk->sk_wq = NULL;
87d11ceb 1575
f3f511e1
GC		 1576 sk_update_clone(sk, newsk);
1577
87d11ceb 1578 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1579 sk_sockets_allocated_inc(newsk);
704da560 1580
08e29af3 1581 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1582 net_enable_timestamp();
87d11ceb
ACM		 1583 }
1584out:
1585 return newsk;
1586}
e56c57d0 1587EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1588
9958089a
AK		 1589void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1590{
d6a4e26a
ED		 1591 u32 max_segs = 1;
1592
9958089a
AK		 1593 __sk_dst_set(sk, dst);
1594 sk->sk_route_caps = dst->dev->features;
1595 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1596 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1597 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1598 if (sk_can_gso(sk)) {
82cc1a7a 1599 if (dst->header_len) {
9958089a 1600 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1601 } else {
9958089a 1602 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1603 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1604 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1605 }
9958089a 1606 }
d6a4e26a 1607 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1608}
1609EXPORT_SYMBOL_GPL(sk_setup_caps);
1610
1da177e4
LT		 1611/*
1612 * Simple resource managers for sockets.
1613 */
1614
1615
4ec93edb
YH		 1616/*
1617 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1618 */
1619void sock_wfree(struct sk_buff *skb)
1620{
1621 struct sock *sk = skb->sk;
d99927f4 1622 unsigned int len = skb->truesize;
1da177e4 1623
d99927f4
ED		 1624 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1625 /*
1626 * Keep a reference on sk_wmem_alloc, this will be released
1627 * after sk_write_space() call
1628 */
1629 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1630 sk->sk_write_space(sk);
d99927f4
ED		 1631 len = 1;
1632 }
2b85a34e 1633 /*
d99927f4
ED		 1634 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1635 * could not do because of in-flight packets
2b85a34e 1636 */
d99927f4 1637 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1638 __sk_free(sk);
1da177e4 1639}
2a91525c 1640EXPORT_SYMBOL(sock_wfree);
1da177e4 1641
f2f872f9
ED		 1642void skb_orphan_partial(struct sk_buff *skb)
1643{
1644 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1645 * so we do not completely orphan skb, but transfert all
1646 * accounted bytes but one, to avoid unexpected reorders.
1647 */
1648 if (skb->destructor == sock_wfree
1649#ifdef CONFIG_INET
1650 || skb->destructor == tcp_wfree
1651#endif
1652 ) {
1653 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1654 skb->truesize = 1;
1655 } else {
1656 skb_orphan(skb);
1657 }
1658}
1659EXPORT_SYMBOL(skb_orphan_partial);
1660
4ec93edb
YH		 1661/*
1662 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1663 */
1664void sock_rfree(struct sk_buff *skb)
1665{
1666 struct sock *sk = skb->sk;
d361fd59 1667 unsigned int len = skb->truesize;
1da177e4 1668
d361fd59
ED		 1669 atomic_sub(len, &sk->sk_rmem_alloc);
1670 sk_mem_uncharge(sk, len);
1da177e4 1671}
2a91525c 1672EXPORT_SYMBOL(sock_rfree);
1da177e4 1673
7768eed8
OH		 1674/*
1675 * Buffer destructor for skbs that are not used directly in read or write
1676 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1677 */
62bccb8c
AD		 1678void sock_efree(struct sk_buff *skb)
1679{
1680 sock_put(skb->sk);
1681}
1682EXPORT_SYMBOL(sock_efree);
1683
976d0201 1684kuid_t sock_i_uid(struct sock *sk)
1da177e4 1685{
976d0201 1686 kuid_t uid;
1da177e4 1687
f064af1e 1688 read_lock_bh(&sk->sk_callback_lock);
976d0201 1689 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1690 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1691 return uid;
1692}
2a91525c 1693EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1694
1695unsigned long sock_i_ino(struct sock *sk)
1696{
1697 unsigned long ino;
1698
f064af1e 1699 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1700 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1701 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1702 return ino;
1703}
2a91525c 1704EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1705
1706/*
1707 * Allocate a skb from the socket's send buffer.
1708 */
86a76caf 1709struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1710 gfp_t priority)
1da177e4
LT		 1711{
1712 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1713 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1714 if (skb) {
1715 skb_set_owner_w(skb, sk);
1716 return skb;
1717 }
1718 }
1719 return NULL;
1720}
2a91525c 1721EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1722
4ec93edb 1723/*
1da177e4 1724 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1725 */
dd0fc66f 1726void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1727{
95c96174 1728 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1729 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1730 void *mem;
1731 /* First do the add, to avoid the race if kmalloc
4ec93edb 1732 * might sleep.
1da177e4
LT		 1733 */
1734 atomic_add(size, &sk->sk_omem_alloc);
1735 mem = kmalloc(size, priority);
1736 if (mem)
1737 return mem;
1738 atomic_sub(size, &sk->sk_omem_alloc);
1739 }
1740 return NULL;
1741}
2a91525c 1742EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1743
79e88659
DB		 1744/* Free an option memory block. Note, we actually want the inline
1745 * here as this allows gcc to detect the nullify and fold away the
1746 * condition entirely.
1da177e4 1747 */
79e88659
DB		 1748static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1749 const bool nullify)
1da177e4 1750{
e53da5fb
DM		 1751 if (WARN_ON_ONCE(!mem))
1752 return;
79e88659
DB		 1753 if (nullify)
1754 kzfree(mem);
1755 else
1756 kfree(mem);
1da177e4
LT		 1757 atomic_sub(size, &sk->sk_omem_alloc);
1758}
79e88659
DB		 1759
1760void sock_kfree_s(struct sock *sk, void *mem, int size)
1761{
1762 __sock_kfree_s(sk, mem, size, false);
1763}
2a91525c 1764EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1765
79e88659
DB		 1766void sock_kzfree_s(struct sock *sk, void *mem, int size)
1767{
1768 __sock_kfree_s(sk, mem, size, true);
1769}
1770EXPORT_SYMBOL(sock_kzfree_s);
1771
1da177e4
LT		 1772/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1773 I think, these locks should be removed for datagram sockets.
1774 */
2a91525c 1775static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1776{
1777 DEFINE_WAIT(wait);
1778
1779 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1780 for (;;) {
1781 if (!timeo)
1782 break;
1783 if (signal_pending(current))
1784 break;
1785 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1786 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1787 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1788 break;
1789 if (sk->sk_shutdown & SEND_SHUTDOWN)
1790 break;
1791 if (sk->sk_err)
1792 break;
1793 timeo = schedule_timeout(timeo);
1794 }
aa395145 1795 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1796 return timeo;
1797}
1798
1799
1800/*
1801 * Generic send/receive buffer handlers
1802 */
1803
4cc7f68d
HX		 1804struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1805 unsigned long data_len, int noblock,
28d64271 1806 int *errcode, int max_page_order)
1da177e4 1807{
2e4e4410 1808 struct sk_buff *skb;
1da177e4
LT		 1809 long timeo;
1810 int err;
1811
1da177e4 1812 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1813 for (;;) {
1da177e4
LT		 1814 err = sock_error(sk);
1815 if (err != 0)
1816 goto failure;
1817
1818 err = -EPIPE;
1819 if (sk->sk_shutdown & SEND_SHUTDOWN)
1820 goto failure;
1821
2e4e4410
ED		 1822 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1823 break;
28d64271 1824
2e4e4410
ED		 1825 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1826 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1827 err = -EAGAIN;
1828 if (!timeo)
1da177e4 1829 goto failure;
2e4e4410
ED		 1830 if (signal_pending(current))
1831 goto interrupted;
1832 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1833 }
2e4e4410
ED		 1834 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1835 errcode, sk->sk_allocation);
1836 if (skb)
1837 skb_set_owner_w(skb, sk);
1da177e4
LT		 1838 return skb;
1839
1840interrupted:
1841 err = sock_intr_errno(timeo);
1842failure:
1843 *errcode = err;
1844 return NULL;
1845}
4cc7f68d 1846EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1847
4ec93edb 1848struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1849 int noblock, int *errcode)
1850{
28d64271 1851 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1852}
2a91525c 1853EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1854
f28ea365
EJ		 1855int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1856 struct sockcm_cookie *sockc)
1857{
1858 struct cmsghdr *cmsg;
1859
1860 for_each_cmsghdr(cmsg, msg) {
1861 if (!CMSG_OK(msg, cmsg))
1862 return -EINVAL;
1863 if (cmsg->cmsg_level != SOL_SOCKET)
1864 continue;
1865 switch (cmsg->cmsg_type) {
1866 case SO_MARK:
1867 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1868 return -EPERM;
1869 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1870 return -EINVAL;
1871 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
1872 break;
1873 default:
1874 return -EINVAL;
1875 }
1876 }
1877 return 0;
1878}
1879EXPORT_SYMBOL(sock_cmsg_send);
1880
5640f768
ED		 1881/* On 32bit arches, an skb frag is limited to 2^15 */
1882#define SKB_FRAG_PAGE_ORDER get_order(32768)
1883
400dfd3a
ED		 1884/**
1885 * skb_page_frag_refill - check that a page_frag contains enough room
1886 * @sz: minimum size of the fragment we want to get
1887 * @pfrag: pointer to page_frag
82d5e2b8 1888 * @gfp: priority for memory allocation
400dfd3a
ED		 1889 *
1890 * Note: While this allocator tries to use high order pages, there is
1891 * no guarantee that allocations succeed. Therefore, @sz MUST be
1892 * less or equal than PAGE_SIZE.
1893 */
d9b2938a 1894bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1895{
5640f768
ED		 1896 if (pfrag->page) {
1897 if (atomic_read(&pfrag->page->_count) == 1) {
1898 pfrag->offset = 0;
1899 return true;
1900 }
400dfd3a 1901 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1902 return true;
1903 put_page(pfrag->page);
1904 }
1905
d9b2938a
ED		 1906 pfrag->offset = 0;
1907 if (SKB_FRAG_PAGE_ORDER) {
fb05e7a8 1908 pfrag->page = alloc_pages((gfp & ~__GFP_WAIT) | __GFP_COMP |
d9b2938a
ED		 1909 __GFP_NOWARN | __GFP_NORETRY,
1910 SKB_FRAG_PAGE_ORDER);
5640f768 1911 if (likely(pfrag->page)) {
d9b2938a 1912 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1913 return true;
1914 }
d9b2938a
ED		 1915 }
1916 pfrag->page = alloc_page(gfp);
1917 if (likely(pfrag->page)) {
1918 pfrag->size = PAGE_SIZE;
1919 return true;
1920 }
400dfd3a
ED		 1921 return false;
1922}
1923EXPORT_SYMBOL(skb_page_frag_refill);
1924
1925bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1926{
1927 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1928 return true;
1929
5640f768
ED		 1930 sk_enter_memory_pressure(sk);
1931 sk_stream_moderate_sndbuf(sk);
1932 return false;
1933}
1934EXPORT_SYMBOL(sk_page_frag_refill);
1935
1da177e4 1936static void __lock_sock(struct sock *sk)
f39234d6
NK		 1937 __releases(&sk->sk_lock.slock)
1938 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1939{
1940 DEFINE_WAIT(wait);
1941
e71a4783 1942 for (;;) {
1da177e4
LT		 1943 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1944 TASK_UNINTERRUPTIBLE);
1945 spin_unlock_bh(&sk->sk_lock.slock);
1946 schedule();
1947 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1948 if (!sock_owned_by_user(sk))
1da177e4
LT		 1949 break;
1950 }
1951 finish_wait(&sk->sk_lock.wq, &wait);
1952}
1953
1954static void __release_sock(struct sock *sk)
f39234d6
NK		 1955 __releases(&sk->sk_lock.slock)
1956 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1957{
1958 struct sk_buff *skb = sk->sk_backlog.head;
1959
1960 do {
1961 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1962 bh_unlock_sock(sk);
1963
1964 do {
1965 struct sk_buff *next = skb->next;
1966
e4cbb02a 1967 prefetch(next);
7fee226a 1968 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1969 skb->next = NULL;
c57943a1 1970 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1971
1972 /*
1973 * We are in process context here with softirqs
1974 * disabled, use cond_resched_softirq() to preempt.
1975 * This is safe to do because we've taken the backlog
1976 * queue private:
1977 */
1978 cond_resched_softirq();
1979
1980 skb = next;
1981 } while (skb != NULL);
1982
1983 bh_lock_sock(sk);
e71a4783 1984 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1985
1986 /*
1987 * Doing the zeroing here guarantee we can not loop forever
1988 * while a wild producer attempts to flood us.
1989 */
1990 sk->sk_backlog.len = 0;
1da177e4
LT		 1991}
1992
1993/**
1994 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1995 * @sk: sock to wait on
1996 * @timeo: for how long
dfbafc99 1997 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 1998 *
1999 * Now socket state including sk->sk_err is changed only under lock,
2000 * hence we may omit checks after joining wait queue.
2001 * We check receive queue before schedule() only as optimization;
2002 * it is very likely that release_sock() added new data.
2003 */
dfbafc99 2004int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4
LT		 2005{
2006 int rc;
2007 DEFINE_WAIT(wait);
2008
aa395145 2009 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4 2010 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
dfbafc99 2011 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);
1da177e4 2012 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 2013 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2014 return rc;
2015}
1da177e4
LT		 2016EXPORT_SYMBOL(sk_wait_data);
2017
3ab224be
HA		 2018/**
2019 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2020 * @sk: socket
2021 * @size: memory size to allocate
2022 * @kind: allocation type
2023 *
2024 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2025 * rmem allocation. This function assumes that protocols which have
2026 * memory_pressure use sk_wmem_queued as write buffer accounting.
2027 */
2028int __sk_mem_schedule(struct sock *sk, int size, int kind)
2029{
2030 struct proto *prot = sk->sk_prot;
2031 int amt = sk_mem_pages(size);
8d987e5c 2032 long allocated;
e1aab161 2033 int parent_status = UNDER_LIMIT;
3ab224be
HA		 2034
2035 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2036
e1aab161 2037 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 2038
2039 /* Under limit. */
e1aab161
GC		 2040 if (parent_status == UNDER_LIMIT &&
2041 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2042 sk_leave_memory_pressure(sk);
3ab224be
HA		 2043 return 1;
2044 }
2045
e1aab161
GC		 2046 /* Under pressure. (we or our parents) */
2047 if ((parent_status > SOFT_LIMIT) ||
2048 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2049 sk_enter_memory_pressure(sk);
3ab224be 2050
e1aab161
GC		 2051 /* Over hard limit (we or our parents) */
2052 if ((parent_status == OVER_LIMIT) ||
2053 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 2054 goto suppress_allocation;
2055
2056 /* guarantee minimum buffer size under pressure */
2057 if (kind == SK_MEM_RECV) {
2058 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2059 return 1;
180d8cd9 2060
3ab224be
HA		 2061 } else { /* SK_MEM_SEND */
2062 if (sk->sk_type == SOCK_STREAM) {
2063 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2064 return 1;
2065 } else if (atomic_read(&sk->sk_wmem_alloc) <
2066 prot->sysctl_wmem[0])
2067 return 1;
2068 }
2069
180d8cd9 2070 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2071 int alloc;
2072
180d8cd9 2073 if (!sk_under_memory_pressure(sk))
1748376b 2074 return 1;
180d8cd9
GC		 2075 alloc = sk_sockets_allocated_read_positive(sk);
2076 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2077 sk_mem_pages(sk->sk_wmem_queued +
2078 atomic_read(&sk->sk_rmem_alloc) +
2079 sk->sk_forward_alloc))
2080 return 1;
2081 }
2082
2083suppress_allocation:
2084
2085 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2086 sk_stream_moderate_sndbuf(sk);
2087
2088 /* Fail only if socket is _under_ its sndbuf.
2089 * In this case we cannot block, so that we have to fail.
2090 */
2091 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2092 return 1;
2093 }
2094
3847ce32
SM		 2095 trace_sock_exceed_buf_limit(sk, prot, allocated);
2096
3ab224be
HA		 2097 /* Alas. Undo changes. */
2098 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2099
0e90b31f 2100 sk_memory_allocated_sub(sk, amt);
180d8cd9 2101
3ab224be
HA		 2102 return 0;
2103}
3ab224be
HA		 2104EXPORT_SYMBOL(__sk_mem_schedule);
2105
2106/**
69dba9bb 2107 * __sk_mem_reclaim - reclaim memory_allocated
3ab224be 2108 * @sk: socket
1a24e04e 2109 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
3ab224be 2110 */
1a24e04e 2111void __sk_mem_reclaim(struct sock *sk, int amount)
3ab224be 2112{
1a24e04e
ED		 2113 amount >>= SK_MEM_QUANTUM_SHIFT;
2114 sk_memory_allocated_sub(sk, amount);
2115 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
3ab224be 2116
180d8cd9
GC		 2117 if (sk_under_memory_pressure(sk) &&
2118 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2119 sk_leave_memory_pressure(sk);
3ab224be 2120}
3ab224be
HA		 2121EXPORT_SYMBOL(__sk_mem_reclaim);
2122
2123
1da177e4
LT		 2124/*
2125 * Set of default routines for initialising struct proto_ops when
2126 * the protocol does not support a particular function. In certain
2127 * cases where it makes no sense for a protocol to have a "do nothing"
2128 * function, some default processing is provided.
2129 */
2130
2131int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2132{
2133 return -EOPNOTSUPP;
2134}
2a91525c 2135EXPORT_SYMBOL(sock_no_bind);
1da177e4 2136
4ec93edb 2137int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2138 int len, int flags)
2139{
2140 return -EOPNOTSUPP;
2141}
2a91525c 2142EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2143
2144int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2145{
2146 return -EOPNOTSUPP;
2147}
2a91525c 2148EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2149
2150int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2151{
2152 return -EOPNOTSUPP;
2153}
2a91525c 2154EXPORT_SYMBOL(sock_no_accept);
1da177e4 2155
4ec93edb 2156int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2157 int *len, int peer)
2158{
2159 return -EOPNOTSUPP;
2160}
2a91525c 2161EXPORT_SYMBOL(sock_no_getname);
1da177e4 2162
2a91525c 2163unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2164{
2165 return 0;
2166}
2a91525c 2167EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2168
2169int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2170{
2171 return -EOPNOTSUPP;
2172}
2a91525c 2173EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2174
2175int sock_no_listen(struct socket *sock, int backlog)
2176{
2177 return -EOPNOTSUPP;
2178}
2a91525c 2179EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2180
2181int sock_no_shutdown(struct socket *sock, int how)
2182{
2183 return -EOPNOTSUPP;
2184}
2a91525c 2185EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2186
2187int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2188 char __user *optval, unsigned int optlen)
1da177e4
LT		 2189{
2190 return -EOPNOTSUPP;
2191}
2a91525c 2192EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2193
2194int sock_no_getsockopt(struct socket *sock, int level, int optname,
2195 char __user *optval, int __user *optlen)
2196{
2197 return -EOPNOTSUPP;
2198}
2a91525c 2199EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2200
1b784140 2201int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2202{
2203 return -EOPNOTSUPP;
2204}
2a91525c 2205EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2206
1b784140
YX		 2207int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2208 int flags)
1da177e4
LT		 2209{
2210 return -EOPNOTSUPP;
2211}
2a91525c 2212EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2213
2214int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2215{
2216 /* Mirror missing mmap method error code */
2217 return -ENODEV;
2218}
2a91525c 2219EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2220
2221ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2222{
2223 ssize_t res;
2224 struct msghdr msg = {.msg_flags = flags};
2225 struct kvec iov;
2226 char *kaddr = kmap(page);
2227 iov.iov_base = kaddr + offset;
2228 iov.iov_len = size;
2229 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2230 kunmap(page);
2231 return res;
2232}
2a91525c 2233EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2234
2235/*
2236 * Default Socket Callbacks
2237 */
2238
2239static void sock_def_wakeup(struct sock *sk)
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))
2246 wake_up_interruptible_all(&wq->wait);
2247 rcu_read_unlock();
1da177e4
LT		 2248}
2249
2250static void sock_def_error_report(struct sock *sk)
2251{
43815482
ED		 2252 struct socket_wq *wq;
2253
2254 rcu_read_lock();
2255 wq = rcu_dereference(sk->sk_wq);
2256 if (wq_has_sleeper(wq))
2257 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2258 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2259 rcu_read_unlock();
1da177e4
LT		 2260}
2261
676d2369 2262static void sock_def_readable(struct sock *sk)
1da177e4 2263{
43815482
ED		 2264 struct socket_wq *wq;
2265
2266 rcu_read_lock();
2267 wq = rcu_dereference(sk->sk_wq);
2268 if (wq_has_sleeper(wq))
2c6607c6 2269 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2270 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2271 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2272 rcu_read_unlock();
1da177e4
LT		 2273}
2274
2275static void sock_def_write_space(struct sock *sk)
2276{
43815482
ED		 2277 struct socket_wq *wq;
2278
2279 rcu_read_lock();
1da177e4
LT		 2280
2281 /* Do not wake up a writer until he can make "significant"
2282 * progress. --DaveM
2283 */
e71a4783 2284 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2285 wq = rcu_dereference(sk->sk_wq);
2286 if (wq_has_sleeper(wq))
2287 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2288 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2289
2290 /* Should agree with poll, otherwise some programs break */
2291 if (sock_writeable(sk))
8d8ad9d7 2292 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2293 }
2294
43815482 2295 rcu_read_unlock();
1da177e4
LT		 2296}
2297
2298static void sock_def_destruct(struct sock *sk)
2299{
1da177e4
LT		 2300}
2301
2302void sk_send_sigurg(struct sock *sk)
2303{
2304 if (sk->sk_socket && sk->sk_socket->file)
2305 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2306 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2307}
2a91525c 2308EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2309
2310void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2311 unsigned long expires)
2312{
2313 if (!mod_timer(timer, expires))
2314 sock_hold(sk);
2315}
1da177e4
LT		 2316EXPORT_SYMBOL(sk_reset_timer);
2317
2318void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2319{
25cc4ae9 2320 if (del_timer(timer))
1da177e4
LT		 2321 __sock_put(sk);
2322}
1da177e4
LT		 2323EXPORT_SYMBOL(sk_stop_timer);
2324
2325void sock_init_data(struct socket *sock, struct sock *sk)
2326{
2327 skb_queue_head_init(&sk->sk_receive_queue);
2328 skb_queue_head_init(&sk->sk_write_queue);
2329 skb_queue_head_init(&sk->sk_error_queue);
2330
2331 sk->sk_send_head = NULL;
2332
2333 init_timer(&sk->sk_timer);
4ec93edb 2334
1da177e4
LT		 2335 sk->sk_allocation = GFP_KERNEL;
2336 sk->sk_rcvbuf = sysctl_rmem_default;
2337 sk->sk_sndbuf = sysctl_wmem_default;
2338 sk->sk_state = TCP_CLOSE;
972692e0 2339 sk_set_socket(sk, sock);
1da177e4
LT		 2340
2341 sock_set_flag(sk, SOCK_ZAPPED);
2342
e71a4783 2343 if (sock) {
1da177e4 2344 sk->sk_type = sock->type;
43815482 2345 sk->sk_wq = sock->wq;
1da177e4
LT		 2346 sock->sk = sk;
2347 } else
43815482 2348 sk->sk_wq = NULL;
1da177e4 2349
b6c6712a 2350 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2351 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2352 lockdep_set_class_and_name(&sk->sk_callback_lock,
2353 af_callback_keys + sk->sk_family,
2354 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2355
2356 sk->sk_state_change = sock_def_wakeup;
2357 sk->sk_data_ready = sock_def_readable;
2358 sk->sk_write_space = sock_def_write_space;
2359 sk->sk_error_report = sock_def_error_report;
2360 sk->sk_destruct = sock_def_destruct;
2361
5640f768
ED		 2362 sk->sk_frag.page = NULL;
2363 sk->sk_frag.offset = 0;
ef64a54f 2364 sk->sk_peek_off = -1;
1da177e4 2365
109f6e39
EB		 2366 sk->sk_peer_pid = NULL;
2367 sk->sk_peer_cred = NULL;
1da177e4
LT		 2368 sk->sk_write_pending = 0;
2369 sk->sk_rcvlowat = 1;
2370 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2371 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2372
f37f0afb 2373 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2374
e0d1095a 2375#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2376 sk->sk_napi_id = 0;
64b0dc51 2377 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2378#endif
2379
62748f32 2380 sk->sk_max_pacing_rate = ~0U;
7eec4174 2381 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2382 /*
2383 * Before updating sk_refcnt, we must commit prior changes to memory
2384 * (Documentation/RCU/rculist_nulls.txt for details)
2385 */
2386 smp_wmb();
1da177e4 2387 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2388 atomic_set(&sk->sk_drops, 0);
1da177e4 2389}
2a91525c 2390EXPORT_SYMBOL(sock_init_data);
1da177e4 2391
b5606c2d 2392void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2393{
2394 might_sleep();
a5b5bb9a 2395 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2396 if (sk->sk_lock.owned)
1da177e4 2397 __lock_sock(sk);
d2e9117c 2398 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2399 spin_unlock(&sk->sk_lock.slock);
2400 /*
2401 * The sk_lock has mutex_lock() semantics here:
2402 */
fcc70d5f 2403 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2404 local_bh_enable();
1da177e4 2405}
fcc70d5f 2406EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2407
b5606c2d 2408void release_sock(struct sock *sk)
1da177e4 2409{
a5b5bb9a
IM		 2410 /*
2411 * The sk_lock has mutex_unlock() semantics:
2412 */
2413 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2414
2415 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2416 if (sk->sk_backlog.tail)
2417 __release_sock(sk);
46d3ceab 2418
c3f9b018
ED		 2419 /* Warning : release_cb() might need to release sk ownership,
2420 * ie call sock_release_ownership(sk) before us.
2421 */
46d3ceab
ED		 2422 if (sk->sk_prot->release_cb)
2423 sk->sk_prot->release_cb(sk);
2424
c3f9b018 2425 sock_release_ownership(sk);
a5b5bb9a
IM		 2426 if (waitqueue_active(&sk->sk_lock.wq))
2427 wake_up(&sk->sk_lock.wq);
2428 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2429}
2430EXPORT_SYMBOL(release_sock);
2431
8a74ad60
ED		 2432/**
2433 * lock_sock_fast - fast version of lock_sock
2434 * @sk: socket
2435 *
2436 * This version should be used for very small section, where process wont block
2437 * return false if fast path is taken
2438 * sk_lock.slock locked, owned = 0, BH disabled
2439 * return true if slow path is taken
2440 * sk_lock.slock unlocked, owned = 1, BH enabled
2441 */
2442bool lock_sock_fast(struct sock *sk)
2443{
2444 might_sleep();
2445 spin_lock_bh(&sk->sk_lock.slock);
2446
2447 if (!sk->sk_lock.owned)
2448 /*
2449 * Note : We must disable BH
2450 */
2451 return false;
2452
2453 __lock_sock(sk);
2454 sk->sk_lock.owned = 1;
2455 spin_unlock(&sk->sk_lock.slock);
2456 /*
2457 * The sk_lock has mutex_lock() semantics here:
2458 */
2459 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2460 local_bh_enable();
2461 return true;
2462}
2463EXPORT_SYMBOL(lock_sock_fast);
2464
1da177e4 2465int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2466{
b7aa0bf7 2467 struct timeval tv;
1da177e4 2468 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2469 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2470 tv = ktime_to_timeval(sk->sk_stamp);
2471 if (tv.tv_sec == -1)
1da177e4 2472 return -ENOENT;
b7aa0bf7
ED		 2473 if (tv.tv_sec == 0) {
2474 sk->sk_stamp = ktime_get_real();
2475 tv = ktime_to_timeval(sk->sk_stamp);
2476 }
2477 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2478}
1da177e4
LT		 2479EXPORT_SYMBOL(sock_get_timestamp);
2480
ae40eb1e
ED		 2481int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2482{
2483 struct timespec ts;
2484 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2485 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2486 ts = ktime_to_timespec(sk->sk_stamp);
2487 if (ts.tv_sec == -1)
2488 return -ENOENT;
2489 if (ts.tv_sec == 0) {
2490 sk->sk_stamp = ktime_get_real();
2491 ts = ktime_to_timespec(sk->sk_stamp);
2492 }
2493 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2494}
2495EXPORT_SYMBOL(sock_get_timestampns);
2496
20d49473 2497void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2498{
20d49473 2499 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2500 unsigned long previous_flags = sk->sk_flags;
2501
20d49473
PO		 2502 sock_set_flag(sk, flag);
2503 /*
2504 * we just set one of the two flags which require net
2505 * time stamping, but time stamping might have been on
2506 * already because of the other one
2507 */
08e29af3 2508 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2509 net_enable_timestamp();
1da177e4
LT		 2510 }
2511}
1da177e4 2512
cb820f8e
RC		 2513int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2514 int level, int type)
2515{
2516 struct sock_exterr_skb *serr;
364a9e93 2517 struct sk_buff *skb;
cb820f8e
RC		 2518 int copied, err;
2519
2520 err = -EAGAIN;
364a9e93 2521 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2522 if (skb == NULL)
2523 goto out;
2524
2525 copied = skb->len;
2526 if (copied > len) {
2527 msg->msg_flags |= MSG_TRUNC;
2528 copied = len;
2529 }
51f3d02b 2530 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2531 if (err)
2532 goto out_free_skb;
2533
2534 sock_recv_timestamp(msg, sk, skb);
2535
2536 serr = SKB_EXT_ERR(skb);
2537 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2538
2539 msg->msg_flags |= MSG_ERRQUEUE;
2540 err = copied;
2541
cb820f8e
RC		 2542out_free_skb:
2543 kfree_skb(skb);
2544out:
2545 return err;
2546}
2547EXPORT_SYMBOL(sock_recv_errqueue);
2548
1da177e4
LT		 2549/*
2550 * Get a socket option on an socket.
2551 *
2552 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2553 * asynchronous errors should be reported by getsockopt. We assume
2554 * this means if you specify SO_ERROR (otherwise whats the point of it).
2555 */
2556int sock_common_getsockopt(struct socket *sock, int level, int optname,
2557 char __user *optval, int __user *optlen)
2558{
2559 struct sock *sk = sock->sk;
2560
2561 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2562}
1da177e4
LT		 2563EXPORT_SYMBOL(sock_common_getsockopt);
2564
3fdadf7d 2565#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2566int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2567 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2568{
2569 struct sock *sk = sock->sk;
2570
1e51f951 2571 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2572 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2573 optval, optlen);
3fdadf7d
DM		 2574 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2575}
2576EXPORT_SYMBOL(compat_sock_common_getsockopt);
2577#endif
2578
1b784140
YX		 2579int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2580 int flags)
1da177e4
LT		 2581{
2582 struct sock *sk = sock->sk;
2583 int addr_len = 0;
2584 int err;
2585
1b784140 2586 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2587 flags & ~MSG_DONTWAIT, &addr_len);
2588 if (err >= 0)
2589 msg->msg_namelen = addr_len;
2590 return err;
2591}
1da177e4
LT		 2592EXPORT_SYMBOL(sock_common_recvmsg);
2593
2594/*
2595 * Set socket options on an inet socket.
2596 */
2597int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2598 char __user *optval, unsigned int optlen)
1da177e4
LT		 2599{
2600 struct sock *sk = sock->sk;
2601
2602 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2603}
1da177e4
LT		 2604EXPORT_SYMBOL(sock_common_setsockopt);
2605
3fdadf7d 2606#ifdef CONFIG_COMPAT
543d9cfe 2607int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2608 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2609{
2610 struct sock *sk = sock->sk;
2611
543d9cfe
ACM		 2612 if (sk->sk_prot->compat_setsockopt != NULL)
2613 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2614 optval, optlen);
3fdadf7d
DM		 2615 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2616}
2617EXPORT_SYMBOL(compat_sock_common_setsockopt);
2618#endif
2619
1da177e4
LT		 2620void sk_common_release(struct sock *sk)
2621{
2622 if (sk->sk_prot->destroy)
2623 sk->sk_prot->destroy(sk);
2624
2625 /*
2626 * Observation: when sock_common_release is called, processes have
2627 * no access to socket. But net still has.
2628 * Step one, detach it from networking:
2629 *
2630 * A. Remove from hash tables.
2631 */
2632
2633 sk->sk_prot->unhash(sk);
2634
2635 /*
2636 * In this point socket cannot receive new packets, but it is possible
2637 * that some packets are in flight because some CPU runs receiver and
2638 * did hash table lookup before we unhashed socket. They will achieve
2639 * receive queue and will be purged by socket destructor.
2640 *
2641 * Also we still have packets pending on receive queue and probably,
2642 * our own packets waiting in device queues. sock_destroy will drain
2643 * receive queue, but transmitted packets will delay socket destruction
2644 * until the last reference will be released.
2645 */
2646
2647 sock_orphan(sk);
2648
2649 xfrm_sk_free_policy(sk);
2650
e6848976 2651 sk_refcnt_debug_release(sk);
5640f768
ED		 2652
2653 if (sk->sk_frag.page) {
2654 put_page(sk->sk_frag.page);
2655 sk->sk_frag.page = NULL;
2656 }
2657
1da177e4
LT		 2658 sock_put(sk);
2659}
1da177e4
LT		 2660EXPORT_SYMBOL(sk_common_release);
2661
13ff3d6f
PE		 2662#ifdef CONFIG_PROC_FS
2663#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2664struct prot_inuse {
2665 int val[PROTO_INUSE_NR];
2666};
13ff3d6f
PE		 2667
2668static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2669
2670#ifdef CONFIG_NET_NS
2671void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2672{
d6d9ca0f 2673 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2674}
2675EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2676
2677int sock_prot_inuse_get(struct net *net, struct proto *prot)
2678{
2679 int cpu, idx = prot->inuse_idx;
2680 int res = 0;
2681
2682 for_each_possible_cpu(cpu)
2683 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2684
2685 return res >= 0 ? res : 0;
2686}
2687EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2688
2c8c1e72 2689static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2690{
2691 net->core.inuse = alloc_percpu(struct prot_inuse);
2692 return net->core.inuse ? 0 : -ENOMEM;
2693}
2694
2c8c1e72 2695static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2696{
2697 free_percpu(net->core.inuse);
2698}
2699
2700static struct pernet_operations net_inuse_ops = {
2701 .init = sock_inuse_init_net,
2702 .exit = sock_inuse_exit_net,
2703};
2704
2705static __init int net_inuse_init(void)
2706{
2707 if (register_pernet_subsys(&net_inuse_ops))
2708 panic("Cannot initialize net inuse counters");
2709
2710 return 0;
2711}
2712
2713core_initcall(net_inuse_init);
2714#else
1338d466
PE		 2715static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2716
c29a0bc4 2717void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2718{
d6d9ca0f 2719 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2720}
2721EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2722
c29a0bc4 2723int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2724{
2725 int cpu, idx = prot->inuse_idx;
2726 int res = 0;
2727
2728 for_each_possible_cpu(cpu)
2729 res += per_cpu(prot_inuse, cpu).val[idx];
2730
2731 return res >= 0 ? res : 0;
2732}
2733EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2734#endif
13ff3d6f
PE		 2735
2736static void assign_proto_idx(struct proto *prot)
2737{
2738 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2739
2740 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2741 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2742 return;
2743 }
2744
2745 set_bit(prot->inuse_idx, proto_inuse_idx);
2746}
2747
2748static void release_proto_idx(struct proto *prot)
2749{
2750 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2751 clear_bit(prot->inuse_idx, proto_inuse_idx);
2752}
2753#else
2754static inline void assign_proto_idx(struct proto *prot)
2755{
2756}
2757
2758static inline void release_proto_idx(struct proto *prot)
2759{
2760}
2761#endif
2762
0159dfd3
ED		 2763static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2764{
2765 if (!rsk_prot)
2766 return;
2767 kfree(rsk_prot->slab_name);
2768 rsk_prot->slab_name = NULL;
adf78eda
JL		 2769 kmem_cache_destroy(rsk_prot->slab);
2770 rsk_prot->slab = NULL;
0159dfd3
ED		 2771}
2772
2773static int req_prot_init(const struct proto *prot)
2774{
2775 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2776
2777 if (!rsk_prot)
2778 return 0;
2779
2780 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2781 prot->name);
2782 if (!rsk_prot->slab_name)
2783 return -ENOMEM;
2784
2785 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2786 rsk_prot->obj_size, 0,
e96f78ab 2787 prot->slab_flags, NULL);
0159dfd3
ED		 2788
2789 if (!rsk_prot->slab) {
2790 pr_crit("%s: Can't create request sock SLAB cache!\n",
2791 prot->name);
2792 return -ENOMEM;
2793 }
2794 return 0;
2795}
2796
b733c007
PE		 2797int proto_register(struct proto *prot, int alloc_slab)
2798{
1da177e4
LT		 2799 if (alloc_slab) {
2800 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2801 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2802 NULL);
1da177e4
LT		 2803
2804 if (prot->slab == NULL) {
e005d193
JP		 2805 pr_crit("%s: Can't create sock SLAB cache!\n",
2806 prot->name);
60e7663d 2807 goto out;
1da177e4 2808 }
2e6599cb 2809
0159dfd3
ED		 2810 if (req_prot_init(prot))
2811 goto out_free_request_sock_slab;
8feaf0c0 2812
6d6ee43e 2813 if (prot->twsk_prot != NULL) {
faf23422 2814 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2815
7e56b5d6 2816 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2817 goto out_free_request_sock_slab;
2818
6d6ee43e 2819 prot->twsk_prot->twsk_slab =
7e56b5d6 2820 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2821 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2822 0,
52db70dc 2823 prot->slab_flags,
20c2df83 2824 NULL);
6d6ee43e 2825 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2826 goto out_free_timewait_sock_slab_name;
2827 }
1da177e4
LT		 2828 }
2829
36b77a52 2830 mutex_lock(&proto_list_mutex);
1da177e4 2831 list_add(&prot->node, &proto_list);
13ff3d6f 2832 assign_proto_idx(prot);
36b77a52 2833 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2834 return 0;
2835
8feaf0c0 2836out_free_timewait_sock_slab_name:
7e56b5d6 2837 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2838out_free_request_sock_slab:
0159dfd3
ED		 2839 req_prot_cleanup(prot->rsk_prot);
2840
2e6599cb
ACM		 2841 kmem_cache_destroy(prot->slab);
2842 prot->slab = NULL;
b733c007
PE		 2843out:
2844 return -ENOBUFS;
1da177e4 2845}
1da177e4
LT		 2846EXPORT_SYMBOL(proto_register);
2847
2848void proto_unregister(struct proto *prot)
2849{
36b77a52 2850 mutex_lock(&proto_list_mutex);
13ff3d6f 2851 release_proto_idx(prot);
0a3f4358 2852 list_del(&prot->node);
36b77a52 2853 mutex_unlock(&proto_list_mutex);
1da177e4 2854
adf78eda
JL		 2855 kmem_cache_destroy(prot->slab);
2856 prot->slab = NULL;
1da177e4 2857
0159dfd3 2858 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2859
6d6ee43e 2860 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2861 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2862 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2863 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2864 }
1da177e4 2865}
1da177e4
LT		 2866EXPORT_SYMBOL(proto_unregister);
2867
2868#ifdef CONFIG_PROC_FS
1da177e4 2869static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2870 __acquires(proto_list_mutex)
1da177e4 2871{
36b77a52 2872 mutex_lock(&proto_list_mutex);
60f0438a 2873 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2874}
2875
2876static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2877{
60f0438a 2878 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2879}
2880
2881static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2882 __releases(proto_list_mutex)
1da177e4 2883{
36b77a52 2884 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2885}
2886
2887static char proto_method_implemented(const void *method)
2888{
2889 return method == NULL ? 'n' : 'y';
2890}
180d8cd9
GC		 2891static long sock_prot_memory_allocated(struct proto *proto)
2892{
cb75a36c 2893 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2894}
2895
2896static char *sock_prot_memory_pressure(struct proto *proto)
2897{
2898 return proto->memory_pressure != NULL ?
2899 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2900}
1da177e4
LT		 2901
2902static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2903{
180d8cd9 2904
8d987e5c 2905 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2906 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2907 proto->name,
2908 proto->obj_size,
14e943db 2909 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2910 sock_prot_memory_allocated(proto),
2911 sock_prot_memory_pressure(proto),
1da177e4
LT		 2912 proto->max_header,
2913 proto->slab == NULL ? "no" : "yes",
2914 module_name(proto->owner),
2915 proto_method_implemented(proto->close),
2916 proto_method_implemented(proto->connect),
2917 proto_method_implemented(proto->disconnect),
2918 proto_method_implemented(proto->accept),
2919 proto_method_implemented(proto->ioctl),
2920 proto_method_implemented(proto->init),
2921 proto_method_implemented(proto->destroy),
2922 proto_method_implemented(proto->shutdown),
2923 proto_method_implemented(proto->setsockopt),
2924 proto_method_implemented(proto->getsockopt),
2925 proto_method_implemented(proto->sendmsg),
2926 proto_method_implemented(proto->recvmsg),
2927 proto_method_implemented(proto->sendpage),
2928 proto_method_implemented(proto->bind),
2929 proto_method_implemented(proto->backlog_rcv),
2930 proto_method_implemented(proto->hash),
2931 proto_method_implemented(proto->unhash),
2932 proto_method_implemented(proto->get_port),
2933 proto_method_implemented(proto->enter_memory_pressure));
2934}
2935
2936static int proto_seq_show(struct seq_file *seq, void *v)
2937{
60f0438a 2938 if (v == &proto_list)
1da177e4
LT		 2939 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2940 "protocol",
2941 "size",
2942 "sockets",
2943 "memory",
2944 "press",
2945 "maxhdr",
2946 "slab",
2947 "module",
2948 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2949 else
60f0438a 2950 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2951 return 0;
2952}
2953
f690808e 2954static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2955 .start = proto_seq_start,
2956 .next = proto_seq_next,
2957 .stop = proto_seq_stop,
2958 .show = proto_seq_show,
2959};
2960
2961static int proto_seq_open(struct inode *inode, struct file *file)
2962{
14e943db
ED		 2963 return seq_open_net(inode, file, &proto_seq_ops,
2964 sizeof(struct seq_net_private));
1da177e4
LT		 2965}
2966
9a32144e 2967static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2968 .owner = THIS_MODULE,
2969 .open = proto_seq_open,
2970 .read = seq_read,
2971 .llseek = seq_lseek,
14e943db
ED		 2972 .release = seq_release_net,
2973};
2974
2975static __net_init int proto_init_net(struct net *net)
2976{
d4beaa66 2977 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2978 return -ENOMEM;
2979
2980 return 0;
2981}
2982
2983static __net_exit void proto_exit_net(struct net *net)
2984{
ece31ffd 2985 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2986}
2987
2988
2989static __net_initdata struct pernet_operations proto_net_ops = {
2990 .init = proto_init_net,
2991 .exit = proto_exit_net,
1da177e4
LT		 2992};
2993
2994static int __init proto_init(void)
2995{
14e943db 2996 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2997}
2998
2999subsys_initcall(proto_init);
3000
3001#endif /* PROC_FS */
ubuntu-bionic-kernel mirror