]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - net/core/sock.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'efi-urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/efi/efi into...
[mirror_ubuntu-artful-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 120
7c0f6ba6 121#include <linux/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>
538950a1 137#include <net/sock_reuseport.h>
1da177e4 138
3847ce32
SM		 139#include <trace/events/sock.h>
140
1da177e4
LT		 141#ifdef CONFIG_INET
142#include <net/tcp.h>
143#endif
144
076bb0c8 145#include <net/busy_poll.h>
06021292 146
36b77a52 147static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 148static LIST_HEAD(proto_list);
149
a3b299da
EB		 150/**
151 * sk_ns_capable - General socket capability test
152 * @sk: Socket to use a capability on or through
153 * @user_ns: The user namespace of the capability to use
154 * @cap: The capability to use
155 *
156 * Test to see if the opener of the socket had when the socket was
157 * created and the current process has the capability @cap in the user
158 * namespace @user_ns.
159 */
160bool sk_ns_capable(const struct sock *sk,
161 struct user_namespace *user_ns, int cap)
162{
163 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
164 ns_capable(user_ns, cap);
165}
166EXPORT_SYMBOL(sk_ns_capable);
167
168/**
169 * sk_capable - Socket global capability test
170 * @sk: Socket to use a capability on or through
e793c0f7 171 * @cap: The global capability to use
a3b299da
EB		 172 *
173 * Test to see if the opener of the socket had when the socket was
174 * created and the current process has the capability @cap in all user
175 * namespaces.
176 */
177bool sk_capable(const struct sock *sk, int cap)
178{
179 return sk_ns_capable(sk, &init_user_ns, cap);
180}
181EXPORT_SYMBOL(sk_capable);
182
183/**
184 * sk_net_capable - Network namespace socket capability test
185 * @sk: Socket to use a capability on or through
186 * @cap: The capability to use
187 *
e793c0f7 188 * Test to see if the opener of the socket had when the socket was created
a3b299da
EB		 189 * and the current process has the capability @cap over the network namespace
190 * the socket is a member of.
191 */
192bool sk_net_capable(const struct sock *sk, int cap)
193{
194 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
195}
196EXPORT_SYMBOL(sk_net_capable);
197
da21f24d
IM		 198/*
199 * Each address family might have different locking rules, so we have
200 * one slock key per address family:
201 */
a5b5bb9a
IM		 202static struct lock_class_key af_family_keys[AF_MAX];
203static struct lock_class_key af_family_slock_keys[AF_MAX];
204
a5b5bb9a
IM		 205/*
206 * Make lock validator output more readable. (we pre-construct these
207 * strings build-time, so that runtime initialization of socket
208 * locks is fast):
209 */
36cbd3dc 210static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 211 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
212 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
213 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
214 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
215 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
216 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
217 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 218 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 219 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 220 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 221 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 222 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 223 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
0a1a37b6 224 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_KCM" ,
02ac5d14 225 "sk_lock-AF_QIPCRTR", "sk_lock-AF_SMC" , "sk_lock-AF_MAX"
a5b5bb9a 226};
36cbd3dc 227static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 228 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
229 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
230 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
231 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
232 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
233 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
234 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 235 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 236 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 237 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 238 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 239 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 240 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
0a1a37b6 241 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_KCM" ,
02ac5d14 242 "slock-AF_QIPCRTR", "slock-AF_SMC" , "slock-AF_MAX"
a5b5bb9a 243};
36cbd3dc 244static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 245 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
246 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
247 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
248 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
249 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
250 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
251 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 252 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 253 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 254 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 255 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 256 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 257 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
0a1a37b6 258 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_KCM" ,
526735dd 259 "clock-AF_QIPCRTR", "clock-AF_SMC" , "clock-AF_MAX"
443aef0e 260};
da21f24d
IM		 261
262/*
263 * sk_callback_lock locking rules are per-address-family,
264 * so split the lock classes by using a per-AF key:
265 */
266static struct lock_class_key af_callback_keys[AF_MAX];
267
1da177e4
LT		 268/* Take into consideration the size of the struct sk_buff overhead in the
269 * determination of these values, since that is non-constant across
270 * platforms. This makes socket queueing behavior and performance
271 * not depend upon such differences.
272 */
273#define _SK_MEM_PACKETS 256
87fb4b7b 274#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 275#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
276#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
277
278/* Run time adjustable parameters. */
ab32ea5d 279__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 280EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 281__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 282EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 283__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
284__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 285
25985edc 286/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 287int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 288EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 289
b245be1f
WB		 290int sysctl_tstamp_allow_data __read_mostly = 1;
291
c93bdd0e
MG		 292struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
293EXPORT_SYMBOL_GPL(memalloc_socks);
294
7cb02404
MG		 295/**
296 * sk_set_memalloc - sets %SOCK_MEMALLOC
297 * @sk: socket to set it on
298 *
299 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
300 * It's the responsibility of the admin to adjust min_free_kbytes
301 * to meet the requirements
302 */
303void sk_set_memalloc(struct sock *sk)
304{
305 sock_set_flag(sk, SOCK_MEMALLOC);
306 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 307 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 308}
309EXPORT_SYMBOL_GPL(sk_set_memalloc);
310
311void sk_clear_memalloc(struct sock *sk)
312{
313 sock_reset_flag(sk, SOCK_MEMALLOC);
314 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 315 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 316
317 /*
318 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
5d753610
MG		 319 * progress of swapping. SOCK_MEMALLOC may be cleared while
320 * it has rmem allocations due to the last swapfile being deactivated
321 * but there is a risk that the socket is unusable due to exceeding
322 * the rmem limits. Reclaim the reserves and obey rmem limits again.
c76562b6 323 */
5d753610 324 sk_mem_reclaim(sk);
7cb02404
MG		 325}
326EXPORT_SYMBOL_GPL(sk_clear_memalloc);
327
b4b9e355
MG		 328int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
329{
330 int ret;
331 unsigned long pflags = current->flags;
332
333 /* these should have been dropped before queueing */
334 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
335
336 current->flags |= PF_MEMALLOC;
337 ret = sk->sk_backlog_rcv(sk, skb);
338 tsk_restore_flags(current, pflags, PF_MEMALLOC);
339
340 return ret;
341}
342EXPORT_SYMBOL(__sk_backlog_rcv);
343
1da177e4
LT		 344static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
345{
346 struct timeval tv;
347
348 if (optlen < sizeof(tv))
349 return -EINVAL;
350 if (copy_from_user(&tv, optval, sizeof(tv)))
351 return -EFAULT;
ba78073e
VA		 352 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
353 return -EDOM;
1da177e4 354
ba78073e 355 if (tv.tv_sec < 0) {
6f11df83
AM		 356 static int warned __read_mostly;
357
ba78073e 358 *timeo_p = 0;
50aab54f 359 if (warned < 10 && net_ratelimit()) {
ba78073e 360 warned++;
e005d193
JP		 361 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
362 __func__, current->comm, task_pid_nr(current));
50aab54f 363 }
ba78073e
VA		 364 return 0;
365 }
1da177e4
LT		 366 *timeo_p = MAX_SCHEDULE_TIMEOUT;
367 if (tv.tv_sec == 0 && tv.tv_usec == 0)
368 return 0;
369 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
8ccde4c5 370 *timeo_p = tv.tv_sec * HZ + DIV_ROUND_UP(tv.tv_usec, USEC_PER_SEC / HZ);
1da177e4
LT		 371 return 0;
372}
373
374static void sock_warn_obsolete_bsdism(const char *name)
375{
376 static int warned;
377 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 378 if (strcmp(warncomm, current->comm) && warned < 5) {
379 strcpy(warncomm, current->comm);
e005d193
JP		 380 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
381 warncomm, name);
1da177e4
LT		 382 warned++;
383 }
384}
385
080a270f
HFS		 386static bool sock_needs_netstamp(const struct sock *sk)
387{
388 switch (sk->sk_family) {
389 case AF_UNSPEC:
390 case AF_UNIX:
391 return false;
392 default:
393 return true;
394 }
395}
396
08e29af3 397static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 398{
08e29af3
ED		 399 if (sk->sk_flags & flags) {
400 sk->sk_flags &= ~flags;
080a270f
HFS		 401 if (sock_needs_netstamp(sk) &&
402 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 403 net_disable_timestamp();
1da177e4
LT		 404 }
405}
406
407
e6afc8ac 408int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
f0088a50 409{
3b885787
NH		 410 unsigned long flags;
411 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 412
0fd7bac6 413 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 414 atomic_inc(&sk->sk_drops);
3847ce32 415 trace_sock_rcvqueue_full(sk, skb);
766e9037 416 return -ENOMEM;
f0088a50
DV		 417 }
418
c76562b6 419 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 420 atomic_inc(&sk->sk_drops);
421 return -ENOBUFS;
3ab224be
HA		 422 }
423
f0088a50
DV		 424 skb->dev = NULL;
425 skb_set_owner_r(skb, sk);
49ad9599 426
7fee226a
ED		 427 /* we escape from rcu protected region, make sure we dont leak
428 * a norefcounted dst
429 */
430 skb_dst_force(skb);
431
3b885787 432 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 433 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 434 __skb_queue_tail(list, skb);
435 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 436
437 if (!sock_flag(sk, SOCK_DEAD))
676d2369 438 sk->sk_data_ready(sk);
766e9037 439 return 0;
f0088a50 440}
e6afc8ac 441EXPORT_SYMBOL(__sock_queue_rcv_skb);
442
443int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
444{
445 int err;
446
447 err = sk_filter(sk, skb);
448 if (err)
449 return err;
450
451 return __sock_queue_rcv_skb(sk, skb);
452}
f0088a50
DV		 453EXPORT_SYMBOL(sock_queue_rcv_skb);
454
4f0c40d9 455int __sk_receive_skb(struct sock *sk, struct sk_buff *skb,
c3f24cfb 456 const int nested, unsigned int trim_cap, bool refcounted)
f0088a50
DV		 457{
458 int rc = NET_RX_SUCCESS;
459
4f0c40d9 460 if (sk_filter_trim_cap(sk, skb, trim_cap))
f0088a50
DV		 461 goto discard_and_relse;
462
463 skb->dev = NULL;
464
274f482d 465 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 466 atomic_inc(&sk->sk_drops);
467 goto discard_and_relse;
468 }
58a5a7b9
ACM		 469 if (nested)
470 bh_lock_sock_nested(sk);
471 else
472 bh_lock_sock(sk);
a5b5bb9a
IM		 473 if (!sock_owned_by_user(sk)) {
474 /*
475 * trylock + unlock semantics:
476 */
477 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
478
c57943a1 479 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 480
481 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 482 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 483 bh_unlock_sock(sk);
484 atomic_inc(&sk->sk_drops);
485 goto discard_and_relse;
486 }
487
f0088a50
DV		 488 bh_unlock_sock(sk);
489out:
c3f24cfb
ED		 490 if (refcounted)
491 sock_put(sk);
f0088a50
DV		 492 return rc;
493discard_and_relse:
494 kfree_skb(skb);
495 goto out;
496}
4f0c40d9 497EXPORT_SYMBOL(__sk_receive_skb);
f0088a50
DV		 498
499struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
500{
b6c6712a 501 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 502
503 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 504 sk_tx_queue_clear(sk);
9b8805a3 505 sk->sk_dst_pending_confirm = 0;
a9b3cd7f 506 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 507 dst_release(dst);
508 return NULL;
509 }
510
511 return dst;
512}
513EXPORT_SYMBOL(__sk_dst_check);
514
515struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
516{
517 struct dst_entry *dst = sk_dst_get(sk);
518
519 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
520 sk_dst_reset(sk);
521 dst_release(dst);
522 return NULL;
523 }
524
525 return dst;
526}
527EXPORT_SYMBOL(sk_dst_check);
528
c91f6df2
BH		 529static int sock_setbindtodevice(struct sock *sk, char __user *optval,
530 int optlen)
4878809f
DM		 531{
532 int ret = -ENOPROTOOPT;
533#ifdef CONFIG_NETDEVICES
3b1e0a65 534 struct net *net = sock_net(sk);
4878809f
DM		 535 char devname[IFNAMSIZ];
536 int index;
537
538 /* Sorry... */
539 ret = -EPERM;
5e1fccc0 540 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 541 goto out;
542
543 ret = -EINVAL;
544 if (optlen < 0)
545 goto out;
546
547 /* Bind this socket to a particular device like "eth0",
548 * as specified in the passed interface name. If the
549 * name is "" or the option length is zero the socket
550 * is not bound.
551 */
552 if (optlen > IFNAMSIZ - 1)
553 optlen = IFNAMSIZ - 1;
554 memset(devname, 0, sizeof(devname));
555
556 ret = -EFAULT;
557 if (copy_from_user(devname, optval, optlen))
558 goto out;
559
000ba2e4
DM		 560 index = 0;
561 if (devname[0] != '\0') {
bf8e56bf 562 struct net_device *dev;
4878809f 563
bf8e56bf
ED		 564 rcu_read_lock();
565 dev = dev_get_by_name_rcu(net, devname);
566 if (dev)
567 index = dev->ifindex;
568 rcu_read_unlock();
4878809f
DM		 569 ret = -ENODEV;
570 if (!dev)
571 goto out;
4878809f
DM		 572 }
573
574 lock_sock(sk);
575 sk->sk_bound_dev_if = index;
576 sk_dst_reset(sk);
577 release_sock(sk);
578
579 ret = 0;
580
581out:
582#endif
583
584 return ret;
585}
586
c91f6df2
BH		 587static int sock_getbindtodevice(struct sock *sk, char __user *optval,
588 int __user *optlen, int len)
589{
590 int ret = -ENOPROTOOPT;
591#ifdef CONFIG_NETDEVICES
592 struct net *net = sock_net(sk);
c91f6df2 593 char devname[IFNAMSIZ];
c91f6df2
BH		 594
595 if (sk->sk_bound_dev_if == 0) {
596 len = 0;
597 goto zero;
598 }
599
600 ret = -EINVAL;
601 if (len < IFNAMSIZ)
602 goto out;
603
5dbe7c17
NS		 604 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
605 if (ret)
c91f6df2 606 goto out;
c91f6df2
BH		 607
608 len = strlen(devname) + 1;
609
610 ret = -EFAULT;
611 if (copy_to_user(optval, devname, len))
612 goto out;
613
614zero:
615 ret = -EFAULT;
616 if (put_user(len, optlen))
617 goto out;
618
619 ret = 0;
620
621out:
622#endif
623
624 return ret;
625}
626
c0ef877b
PE		 627static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
628{
629 if (valbool)
630 sock_set_flag(sk, bit);
631 else
632 sock_reset_flag(sk, bit);
633}
634
f60e5990 635bool sk_mc_loop(struct sock *sk)
636{
637 if (dev_recursion_level())
638 return false;
639 if (!sk)
640 return true;
641 switch (sk->sk_family) {
642 case AF_INET:
643 return inet_sk(sk)->mc_loop;
644#if IS_ENABLED(CONFIG_IPV6)
645 case AF_INET6:
646 return inet6_sk(sk)->mc_loop;
647#endif
648 }
649 WARN_ON(1);
650 return true;
651}
652EXPORT_SYMBOL(sk_mc_loop);
653
1da177e4
LT		 654/*
655 * This is meant for all protocols to use and covers goings on
656 * at the socket level. Everything here is generic.
657 */
658
659int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 660 char __user *optval, unsigned int optlen)
1da177e4 661{
2a91525c 662 struct sock *sk = sock->sk;
1da177e4
LT		 663 int val;
664 int valbool;
665 struct linger ling;
666 int ret = 0;
4ec93edb 667
1da177e4
LT		 668 /*
669 * Options without arguments
670 */
671
4878809f 672 if (optname == SO_BINDTODEVICE)
c91f6df2 673 return sock_setbindtodevice(sk, optval, optlen);
4878809f 674
e71a4783
SH		 675 if (optlen < sizeof(int))
676 return -EINVAL;
4ec93edb 677
1da177e4
LT		 678 if (get_user(val, (int __user *)optval))
679 return -EFAULT;
4ec93edb 680
2a91525c 681 valbool = val ? 1 : 0;
1da177e4
LT		 682
683 lock_sock(sk);
684
2a91525c 685 switch (optname) {
e71a4783 686 case SO_DEBUG:
2a91525c 687 if (val && !capable(CAP_NET_ADMIN))
e71a4783 688 ret = -EACCES;
2a91525c 689 else
c0ef877b 690 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 691 break;
692 case SO_REUSEADDR:
4a17fd52 693 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 694 break;
055dc21a
TH		 695 case SO_REUSEPORT:
696 sk->sk_reuseport = valbool;
697 break;
e71a4783 698 case SO_TYPE:
49c794e9 699 case SO_PROTOCOL:
0d6038ee 700 case SO_DOMAIN:
e71a4783
SH		 701 case SO_ERROR:
702 ret = -ENOPROTOOPT;
703 break;
704 case SO_DONTROUTE:
c0ef877b 705 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 706 break;
707 case SO_BROADCAST:
708 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
709 break;
710 case SO_SNDBUF:
711 /* Don't error on this BSD doesn't and if you think
82981930
ED		 712 * about it this is right. Otherwise apps have to
713 * play 'guess the biggest size' games. RCVBUF/SNDBUF
714 * are treated in BSD as hints
715 */
716 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 717set_sndbuf:
e71a4783 718 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
b98b0bc8 719 sk->sk_sndbuf = max_t(int, val * 2, SOCK_MIN_SNDBUF);
82981930 720 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 721 sk->sk_write_space(sk);
722 break;
1da177e4 723
e71a4783
SH		 724 case SO_SNDBUFFORCE:
725 if (!capable(CAP_NET_ADMIN)) {
726 ret = -EPERM;
727 break;
728 }
729 goto set_sndbuf;
b0573dea 730
e71a4783
SH		 731 case SO_RCVBUF:
732 /* Don't error on this BSD doesn't and if you think
82981930
ED		 733 * about it this is right. Otherwise apps have to
734 * play 'guess the biggest size' games. RCVBUF/SNDBUF
735 * are treated in BSD as hints
736 */
737 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 738set_rcvbuf:
e71a4783
SH		 739 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
740 /*
741 * We double it on the way in to account for
742 * "struct sk_buff" etc. overhead. Applications
743 * assume that the SO_RCVBUF setting they make will
744 * allow that much actual data to be received on that
745 * socket.
746 *
747 * Applications are unaware that "struct sk_buff" and
748 * other overheads allocate from the receive buffer
749 * during socket buffer allocation.
750 *
751 * And after considering the possible alternatives,
752 * returning the value we actually used in getsockopt
753 * is the most desirable behavior.
754 */
b98b0bc8 755 sk->sk_rcvbuf = max_t(int, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 756 break;
757
758 case SO_RCVBUFFORCE:
759 if (!capable(CAP_NET_ADMIN)) {
760 ret = -EPERM;
1da177e4 761 break;
e71a4783
SH		 762 }
763 goto set_rcvbuf;
1da177e4 764
e71a4783 765 case SO_KEEPALIVE:
4b9d07a4
UB		 766 if (sk->sk_prot->keepalive)
767 sk->sk_prot->keepalive(sk, valbool);
e71a4783
SH		 768 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
769 break;
770
771 case SO_OOBINLINE:
772 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
773 break;
774
775 case SO_NO_CHECK:
28448b80 776 sk->sk_no_check_tx = valbool;
e71a4783
SH		 777 break;
778
779 case SO_PRIORITY:
5e1fccc0
EB		 780 if ((val >= 0 && val <= 6) ||
781 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 782 sk->sk_priority = val;
783 else
784 ret = -EPERM;
785 break;
786
787 case SO_LINGER:
788 if (optlen < sizeof(ling)) {
789 ret = -EINVAL; /* 1003.1g */
1da177e4 790 break;
e71a4783 791 }
2a91525c 792 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 793 ret = -EFAULT;
1da177e4 794 break;
e71a4783
SH		 795 }
796 if (!ling.l_onoff)
797 sock_reset_flag(sk, SOCK_LINGER);
798 else {
1da177e4 799#if (BITS_PER_LONG == 32)
e71a4783
SH		 800 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
801 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 802 else
e71a4783
SH		 803#endif
804 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
805 sock_set_flag(sk, SOCK_LINGER);
806 }
807 break;
808
809 case SO_BSDCOMPAT:
810 sock_warn_obsolete_bsdism("setsockopt");
811 break;
812
813 case SO_PASSCRED:
814 if (valbool)
815 set_bit(SOCK_PASSCRED, &sock->flags);
816 else
817 clear_bit(SOCK_PASSCRED, &sock->flags);
818 break;
819
820 case SO_TIMESTAMP:
92f37fd2 821 case SO_TIMESTAMPNS:
e71a4783 822 if (valbool) {
92f37fd2
ED		 823 if (optname == SO_TIMESTAMP)
824 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
825 else
826 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 827 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 828 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 829 } else {
e71a4783 830 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 831 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
832 }
e71a4783
SH		 833 break;
834
20d49473
PO		 835 case SO_TIMESTAMPING:
836 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 837 ret = -EINVAL;
20d49473
PO		 838 break;
839 }
b245be1f 840
09c2d251 841 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765 842 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
ac5cc977
WC		 843 if (sk->sk_protocol == IPPROTO_TCP &&
844 sk->sk_type == SOCK_STREAM) {
6db8b963
SHY		 845 if ((1 << sk->sk_state) &
846 (TCPF_CLOSE | TCPF_LISTEN)) {
4ed2d765
WB		 847 ret = -EINVAL;
848 break;
849 }
850 sk->sk_tskey = tcp_sk(sk)->snd_una;
851 } else {
852 sk->sk_tskey = 0;
853 }
854 }
1c885808
FY		 855
856 if (val & SOF_TIMESTAMPING_OPT_STATS &&
857 !(val & SOF_TIMESTAMPING_OPT_TSONLY)) {
858 ret = -EINVAL;
859 break;
860 }
861
b9f40e21 862 sk->sk_tsflags = val;
20d49473
PO		 863 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
864 sock_enable_timestamp(sk,
865 SOCK_TIMESTAMPING_RX_SOFTWARE);
866 else
867 sock_disable_timestamp(sk,
08e29af3 868 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 869 break;
870
e71a4783
SH		 871 case SO_RCVLOWAT:
872 if (val < 0)
873 val = INT_MAX;
874 sk->sk_rcvlowat = val ? : 1;
875 break;
876
877 case SO_RCVTIMEO:
878 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
879 break;
880
881 case SO_SNDTIMEO:
882 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
883 break;
1da177e4 884
e71a4783
SH		 885 case SO_ATTACH_FILTER:
886 ret = -EINVAL;
887 if (optlen == sizeof(struct sock_fprog)) {
888 struct sock_fprog fprog;
1da177e4 889
e71a4783
SH		 890 ret = -EFAULT;
891 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 892 break;
e71a4783
SH		 893
894 ret = sk_attach_filter(&fprog, sk);
895 }
896 break;
897
89aa0758
AS		 898 case SO_ATTACH_BPF:
899 ret = -EINVAL;
900 if (optlen == sizeof(u32)) {
901 u32 ufd;
902
903 ret = -EFAULT;
904 if (copy_from_user(&ufd, optval, sizeof(ufd)))
905 break;
906
907 ret = sk_attach_bpf(ufd, sk);
908 }
909 break;
910
538950a1
CG		 911 case SO_ATTACH_REUSEPORT_CBPF:
912 ret = -EINVAL;
913 if (optlen == sizeof(struct sock_fprog)) {
914 struct sock_fprog fprog;
915
916 ret = -EFAULT;
917 if (copy_from_user(&fprog, optval, sizeof(fprog)))
918 break;
919
920 ret = sk_reuseport_attach_filter(&fprog, sk);
921 }
922 break;
923
924 case SO_ATTACH_REUSEPORT_EBPF:
925 ret = -EINVAL;
926 if (optlen == sizeof(u32)) {
927 u32 ufd;
928
929 ret = -EFAULT;
930 if (copy_from_user(&ufd, optval, sizeof(ufd)))
931 break;
932
933 ret = sk_reuseport_attach_bpf(ufd, sk);
934 }
935 break;
936
e71a4783 937 case SO_DETACH_FILTER:
55b33325 938 ret = sk_detach_filter(sk);
e71a4783 939 break;
1da177e4 940
d59577b6
VB		 941 case SO_LOCK_FILTER:
942 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
943 ret = -EPERM;
944 else
945 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
946 break;
947
e71a4783
SH		 948 case SO_PASSSEC:
949 if (valbool)
950 set_bit(SOCK_PASSSEC, &sock->flags);
951 else
952 clear_bit(SOCK_PASSSEC, &sock->flags);
953 break;
4a19ec58 954 case SO_MARK:
5e1fccc0 955 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 956 ret = -EPERM;
2a91525c 957 else
4a19ec58 958 sk->sk_mark = val;
4a19ec58 959 break;
877ce7c1 960
3b885787 961 case SO_RXQ_OVFL:
8083f0fc 962 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 963 break;
6e3e939f
JB		 964
965 case SO_WIFI_STATUS:
966 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
967 break;
968
ef64a54f
PE		 969 case SO_PEEK_OFF:
970 if (sock->ops->set_peek_off)
12663bfc 971 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 972 else
973 ret = -EOPNOTSUPP;
974 break;
3bdc0eba
BG		 975
976 case SO_NOFCS:
977 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
978 break;
979
7d4c04fc
KJ		 980 case SO_SELECT_ERR_QUEUE:
981 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
982 break;
983
e0d1095a 984#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 985 case SO_BUSY_POLL:
dafcc438
ET		 986 /* allow unprivileged users to decrease the value */
987 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
988 ret = -EPERM;
989 else {
990 if (val < 0)
991 ret = -EINVAL;
992 else
993 sk->sk_ll_usec = val;
994 }
995 break;
996#endif
62748f32
ED		 997
998 case SO_MAX_PACING_RATE:
999 sk->sk_max_pacing_rate = val;
1000 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
1001 sk->sk_max_pacing_rate);
1002 break;
1003
70da268b
ED		 1004 case SO_INCOMING_CPU:
1005 sk->sk_incoming_cpu = val;
1006 break;
1007
a87cb3e4
TH		 1008 case SO_CNX_ADVICE:
1009 if (val == 1)
1010 dst_negative_advice(sk);
1011 break;
e71a4783
SH		 1012 default:
1013 ret = -ENOPROTOOPT;
1014 break;
4ec93edb 1015 }
1da177e4
LT		 1016 release_sock(sk);
1017 return ret;
1018}
2a91525c 1019EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1020
1021
8f09898b 1022static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1023 struct ucred *ucred)
3f551f94
EB		 1024{
1025 ucred->pid = pid_vnr(pid);
1026 ucred->uid = ucred->gid = -1;
1027 if (cred) {
1028 struct user_namespace *current_ns = current_user_ns();
1029
b2e4f544
EB		 1030 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1031 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1032 }
1033}
1034
1da177e4
LT		 1035int sock_getsockopt(struct socket *sock, int level, int optname,
1036 char __user *optval, int __user *optlen)
1037{
1038 struct sock *sk = sock->sk;
4ec93edb 1039
e71a4783 1040 union {
4ec93edb
YH		 1041 int val;
1042 struct linger ling;
1da177e4
LT		 1043 struct timeval tm;
1044 } v;
4ec93edb 1045
4d0392be 1046 int lv = sizeof(int);
1da177e4 1047 int len;
4ec93edb 1048
e71a4783 1049 if (get_user(len, optlen))
4ec93edb 1050 return -EFAULT;
e71a4783 1051 if (len < 0)
1da177e4 1052 return -EINVAL;
4ec93edb 1053
50fee1de 1054 memset(&v, 0, sizeof(v));
df0bca04 1055
2a91525c 1056 switch (optname) {
e71a4783
SH		 1057 case SO_DEBUG:
1058 v.val = sock_flag(sk, SOCK_DBG);
1059 break;
1060
1061 case SO_DONTROUTE:
1062 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1063 break;
1064
1065 case SO_BROADCAST:
1b23a5df 1066 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1067 break;
1068
1069 case SO_SNDBUF:
1070 v.val = sk->sk_sndbuf;
1071 break;
1072
1073 case SO_RCVBUF:
1074 v.val = sk->sk_rcvbuf;
1075 break;
1076
1077 case SO_REUSEADDR:
1078 v.val = sk->sk_reuse;
1079 break;
1080
055dc21a
TH		 1081 case SO_REUSEPORT:
1082 v.val = sk->sk_reuseport;
1083 break;
1084
e71a4783 1085 case SO_KEEPALIVE:
1b23a5df 1086 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1087 break;
1088
1089 case SO_TYPE:
1090 v.val = sk->sk_type;
1091 break;
1092
49c794e9
JE		 1093 case SO_PROTOCOL:
1094 v.val = sk->sk_protocol;
1095 break;
1096
0d6038ee
JE		 1097 case SO_DOMAIN:
1098 v.val = sk->sk_family;
1099 break;
1100
e71a4783
SH		 1101 case SO_ERROR:
1102 v.val = -sock_error(sk);
2a91525c 1103 if (v.val == 0)
e71a4783
SH		 1104 v.val = xchg(&sk->sk_err_soft, 0);
1105 break;
1106
1107 case SO_OOBINLINE:
1b23a5df 1108 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1109 break;
1110
1111 case SO_NO_CHECK:
28448b80 1112 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1113 break;
1114
1115 case SO_PRIORITY:
1116 v.val = sk->sk_priority;
1117 break;
1118
1119 case SO_LINGER:
1120 lv = sizeof(v.ling);
1b23a5df 1121 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1122 v.ling.l_linger = sk->sk_lingertime / HZ;
1123 break;
1124
1125 case SO_BSDCOMPAT:
1126 sock_warn_obsolete_bsdism("getsockopt");
1127 break;
1128
1129 case SO_TIMESTAMP:
92f37fd2
ED		 1130 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1131 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1132 break;
1133
1134 case SO_TIMESTAMPNS:
1135 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1136 break;
1137
20d49473 1138 case SO_TIMESTAMPING:
b9f40e21 1139 v.val = sk->sk_tsflags;
20d49473
PO		 1140 break;
1141
e71a4783 1142 case SO_RCVTIMEO:
2a91525c 1143 lv = sizeof(struct timeval);
e71a4783
SH		 1144 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1145 v.tm.tv_sec = 0;
1146 v.tm.tv_usec = 0;
1147 } else {
1148 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
8ccde4c5 1149 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * USEC_PER_SEC) / HZ;
e71a4783
SH		 1150 }
1151 break;
1152
1153 case SO_SNDTIMEO:
2a91525c 1154 lv = sizeof(struct timeval);
e71a4783
SH		 1155 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1156 v.tm.tv_sec = 0;
1157 v.tm.tv_usec = 0;
1158 } else {
1159 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
8ccde4c5 1160 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * USEC_PER_SEC) / HZ;
e71a4783
SH		 1161 }
1162 break;
1da177e4 1163
e71a4783
SH		 1164 case SO_RCVLOWAT:
1165 v.val = sk->sk_rcvlowat;
1166 break;
1da177e4 1167
e71a4783 1168 case SO_SNDLOWAT:
2a91525c 1169 v.val = 1;
e71a4783 1170 break;
1da177e4 1171
e71a4783 1172 case SO_PASSCRED:
82981930 1173 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1174 break;
1da177e4 1175
e71a4783 1176 case SO_PEERCRED:
109f6e39
EB		 1177 {
1178 struct ucred peercred;
1179 if (len > sizeof(peercred))
1180 len = sizeof(peercred);
1181 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1182 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1183 return -EFAULT;
1184 goto lenout;
109f6e39 1185 }
1da177e4 1186
e71a4783
SH		 1187 case SO_PEERNAME:
1188 {
1189 char address[128];
1190
1191 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1192 return -ENOTCONN;
1193 if (lv < len)
1194 return -EINVAL;
1195 if (copy_to_user(optval, address, len))
1196 return -EFAULT;
1197 goto lenout;
1198 }
1da177e4 1199
e71a4783
SH		 1200 /* Dubious BSD thing... Probably nobody even uses it, but
1201 * the UNIX standard wants it for whatever reason... -DaveM
1202 */
1203 case SO_ACCEPTCONN:
1204 v.val = sk->sk_state == TCP_LISTEN;
1205 break;
1da177e4 1206
e71a4783 1207 case SO_PASSSEC:
82981930 1208 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1209 break;
877ce7c1 1210
e71a4783
SH		 1211 case SO_PEERSEC:
1212 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1213
4a19ec58
LAT		 1214 case SO_MARK:
1215 v.val = sk->sk_mark;
1216 break;
1217
3b885787 1218 case SO_RXQ_OVFL:
1b23a5df 1219 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1220 break;
1221
6e3e939f 1222 case SO_WIFI_STATUS:
1b23a5df 1223 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1224 break;
1225
ef64a54f
PE		 1226 case SO_PEEK_OFF:
1227 if (!sock->ops->set_peek_off)
1228 return -EOPNOTSUPP;
1229
1230 v.val = sk->sk_peek_off;
1231 break;
bc2f7996 1232 case SO_NOFCS:
1b23a5df 1233 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1234 break;
c91f6df2 1235
f7b86bfe 1236 case SO_BINDTODEVICE:
c91f6df2
BH		 1237 return sock_getbindtodevice(sk, optval, optlen, len);
1238
a8fc9277
PE		 1239 case SO_GET_FILTER:
1240 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1241 if (len < 0)
1242 return len;
1243
1244 goto lenout;
c91f6df2 1245
d59577b6
VB		 1246 case SO_LOCK_FILTER:
1247 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1248 break;
1249
ea02f941
MS		 1250 case SO_BPF_EXTENSIONS:
1251 v.val = bpf_tell_extensions();
1252 break;
1253
7d4c04fc
KJ		 1254 case SO_SELECT_ERR_QUEUE:
1255 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1256 break;
1257
e0d1095a 1258#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1259 case SO_BUSY_POLL:
dafcc438
ET		 1260 v.val = sk->sk_ll_usec;
1261 break;
1262#endif
1263
62748f32
ED		 1264 case SO_MAX_PACING_RATE:
1265 v.val = sk->sk_max_pacing_rate;
1266 break;
1267
2c8c56e1
ED		 1268 case SO_INCOMING_CPU:
1269 v.val = sk->sk_incoming_cpu;
1270 break;
1271
e71a4783 1272 default:
443b5991
YH		 1273 /* We implement the SO_SNDLOWAT etc to not be settable
1274 * (1003.1g 7).
1275 */
e71a4783 1276 return -ENOPROTOOPT;
1da177e4 1277 }
e71a4783 1278
1da177e4
LT		 1279 if (len > lv)
1280 len = lv;
1281 if (copy_to_user(optval, &v, len))
1282 return -EFAULT;
1283lenout:
4ec93edb
YH		 1284 if (put_user(len, optlen))
1285 return -EFAULT;
1286 return 0;
1da177e4
LT		 1287}
1288
a5b5bb9a
IM		 1289/*
1290 * Initialize an sk_lock.
1291 *
1292 * (We also register the sk_lock with the lock validator.)
1293 */
b6f99a21 1294static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1295{
ed07536e
PZ		 1296 sock_lock_init_class_and_name(sk,
1297 af_family_slock_key_strings[sk->sk_family],
1298 af_family_slock_keys + sk->sk_family,
1299 af_family_key_strings[sk->sk_family],
1300 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1301}
1302
4dc6dc71
ED		 1303/*
1304 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1305 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1306 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1307 */
f1a6c4da
PE		 1308static void sock_copy(struct sock *nsk, const struct sock *osk)
1309{
1310#ifdef CONFIG_SECURITY_NETWORK
1311 void *sptr = nsk->sk_security;
1312#endif
68835aba
ED		 1313 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1314
1315 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1316 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1317
f1a6c4da
PE		 1318#ifdef CONFIG_SECURITY_NETWORK
1319 nsk->sk_security = sptr;
1320 security_sk_clone(osk, nsk);
1321#endif
1322}
1323
2e4afe7b
PE		 1324static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1325 int family)
c308c1b2
PE		 1326{
1327 struct sock *sk;
1328 struct kmem_cache *slab;
1329
1330 slab = prot->slab;
e912b114
ED		 1331 if (slab != NULL) {
1332 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1333 if (!sk)
1334 return sk;
ba2489b0
ED		 1335 if (priority & __GFP_ZERO)
1336 sk_prot_clear_nulls(sk, prot->obj_size);
fcbdf09d 1337 } else
c308c1b2
PE		 1338 sk = kmalloc(prot->obj_size, priority);
1339
2e4afe7b 1340 if (sk != NULL) {
a98b65a3
VN		 1341 kmemcheck_annotate_bitfield(sk, flags);
1342
2e4afe7b
PE		 1343 if (security_sk_alloc(sk, family, priority))
1344 goto out_free;
1345
1346 if (!try_module_get(prot->owner))
1347 goto out_free_sec;
e022f0b4 1348 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1349 }
1350
c308c1b2 1351 return sk;
2e4afe7b
PE		 1352
1353out_free_sec:
1354 security_sk_free(sk);
1355out_free:
1356 if (slab != NULL)
1357 kmem_cache_free(slab, sk);
1358 else
1359 kfree(sk);
1360 return NULL;
c308c1b2
PE		 1361}
1362
1363static void sk_prot_free(struct proto *prot, struct sock *sk)
1364{
1365 struct kmem_cache *slab;
2e4afe7b 1366 struct module *owner;
c308c1b2 1367
2e4afe7b 1368 owner = prot->owner;
c308c1b2 1369 slab = prot->slab;
2e4afe7b 1370
bd1060a1 1371 cgroup_sk_free(&sk->sk_cgrp_data);
2d758073 1372 mem_cgroup_sk_free(sk);
2e4afe7b 1373 security_sk_free(sk);
c308c1b2
PE		 1374 if (slab != NULL)
1375 kmem_cache_free(slab, sk);
1376 else
1377 kfree(sk);
2e4afe7b 1378 module_put(owner);
c308c1b2
PE		 1379}
1380
1da177e4
LT		 1381/**
1382 * sk_alloc - All socket objects are allocated here
c4ea43c5 1383 * @net: the applicable net namespace
4dc3b16b
PP		 1384 * @family: protocol family
1385 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1386 * @prot: struct proto associated with this new sock instance
11aa9c28 1387 * @kern: is this to be a kernel socket?
1da177e4 1388 */
1b8d7ae4 1389struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1390 struct proto *prot, int kern)
1da177e4 1391{
c308c1b2 1392 struct sock *sk;
1da177e4 1393
154adbc8 1394 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1395 if (sk) {
154adbc8
PE		 1396 sk->sk_family = family;
1397 /*
1398 * See comment in struct sock definition to understand
1399 * why we need sk_prot_creator -acme
1400 */
1401 sk->sk_prot = sk->sk_prot_creator = prot;
1402 sock_lock_init(sk);
26abe143
EB		 1403 sk->sk_net_refcnt = kern ? 0 : 1;
1404 if (likely(sk->sk_net_refcnt))
1405 get_net(net);
1406 sock_net_set(sk, net);
d66ee058 1407 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1408
2d758073 1409 mem_cgroup_sk_alloc(sk);
d979a39d 1410 cgroup_sk_alloc(&sk->sk_cgrp_data);
2a56a1fe
TH		 1411 sock_update_classid(&sk->sk_cgrp_data);
1412 sock_update_netprioidx(&sk->sk_cgrp_data);
1da177e4 1413 }
a79af59e 1414
2e4afe7b 1415 return sk;
1da177e4 1416}
2a91525c 1417EXPORT_SYMBOL(sk_alloc);
1da177e4 1418
a4298e45
ED		 1419/* Sockets having SOCK_RCU_FREE will call this function after one RCU
1420 * grace period. This is the case for UDP sockets and TCP listeners.
1421 */
1422static void __sk_destruct(struct rcu_head *head)
1da177e4 1423{
a4298e45 1424 struct sock *sk = container_of(head, struct sock, sk_rcu);
1da177e4 1425 struct sk_filter *filter;
1da177e4
LT		 1426
1427 if (sk->sk_destruct)
1428 sk->sk_destruct(sk);
1429
a898def2
PM		 1430 filter = rcu_dereference_check(sk->sk_filter,
1431 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1432 if (filter) {
309dd5fc 1433 sk_filter_uncharge(sk, filter);
a9b3cd7f 1434 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4 1435 }
538950a1
CG		 1436 if (rcu_access_pointer(sk->sk_reuseport_cb))
1437 reuseport_detach_sock(sk);
1da177e4 1438
08e29af3 1439 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1440
1441 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1442 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1443 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1444
109f6e39
EB		 1445 if (sk->sk_peer_cred)
1446 put_cred(sk->sk_peer_cred);
1447 put_pid(sk->sk_peer_pid);
26abe143
EB		 1448 if (likely(sk->sk_net_refcnt))
1449 put_net(sock_net(sk));
c308c1b2 1450 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1451}
2b85a34e 1452
a4298e45
ED		 1453void sk_destruct(struct sock *sk)
1454{
1455 if (sock_flag(sk, SOCK_RCU_FREE))
1456 call_rcu(&sk->sk_rcu, __sk_destruct);
1457 else
1458 __sk_destruct(&sk->sk_rcu);
1459}
1460
eb4cb008
CG		 1461static void __sk_free(struct sock *sk)
1462{
b922622e 1463 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1464 sock_diag_broadcast_destroy(sk);
1465 else
1466 sk_destruct(sk);
1467}
1468
2b85a34e
ED		 1469void sk_free(struct sock *sk)
1470{
1471 /*
25985edc 1472 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1473 * some packets are still in some tx queue.
1474 * If not null, sock_wfree() will call __sk_free(sk) later
1475 */
1476 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1477 __sk_free(sk);
1478}
2a91525c 1479EXPORT_SYMBOL(sk_free);
1da177e4 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
87d11ceb 1517 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1518 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1519 af_callback_keys + newsk->sk_family,
1520 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1521
1522 newsk->sk_dst_cache = NULL;
9b8805a3 1523 newsk->sk_dst_pending_confirm = 0;
87d11ceb
ACM		 1524 newsk->sk_wmem_queued = 0;
1525 newsk->sk_forward_alloc = 0;
9caad864 1526 atomic_set(&newsk->sk_drops, 0);
87d11ceb 1527 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1528 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1529
1530 sock_reset_flag(newsk, SOCK_DONE);
1531 skb_queue_head_init(&newsk->sk_error_queue);
1532
0d7da9dd 1533 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1534 if (filter != NULL)
278571ba
AS		 1535 /* though it's an empty new sock, the charging may fail
1536 * if sysctl_optmem_max was changed between creation of
1537 * original socket and cloning
1538 */
1539 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1540
d188ba86 1541 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
94352d45 1542 sk_free_unlock_clone(newsk);
87d11ceb
ACM		 1543 newsk = NULL;
1544 goto out;
1545 }
fa463497 1546 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
87d11ceb
ACM		 1547
1548 newsk->sk_err = 0;
e551c32d 1549 newsk->sk_err_soft = 0;
87d11ceb 1550 newsk->sk_priority = 0;
2c8c56e1 1551 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1552 atomic64_set(&newsk->sk_cookie, 0);
d979a39d 1553
2d758073 1554 mem_cgroup_sk_alloc(newsk);
d979a39d
JW		 1555 cgroup_sk_alloc(&newsk->sk_cgrp_data);
1556
4dc6dc71
ED		 1557 /*
1558 * Before updating sk_refcnt, we must commit prior changes to memory
1559 * (Documentation/RCU/rculist_nulls.txt for details)
1560 */
1561 smp_wmb();
87d11ceb
ACM		 1562 atomic_set(&newsk->sk_refcnt, 2);
1563
1564 /*
1565 * Increment the counter in the same struct proto as the master
1566 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1567 * is the same as sk->sk_prot->socks, as this field was copied
1568 * with memcpy).
1569 *
1570 * This _changes_ the previous behaviour, where
1571 * tcp_create_openreq_child always was incrementing the
1572 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1573 * to be taken into account in all callers. -acme
1574 */
1575 sk_refcnt_debug_inc(newsk);
972692e0 1576 sk_set_socket(newsk, NULL);
43815482 1577 newsk->sk_wq = NULL;
87d11ceb
ACM		 1578
1579 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1580 sk_sockets_allocated_inc(newsk);
704da560 1581
080a270f
HFS		 1582 if (sock_needs_netstamp(sk) &&
1583 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1584 net_enable_timestamp();
87d11ceb
ACM		 1585 }
1586out:
1587 return newsk;
1588}
e56c57d0 1589EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1590
94352d45
ACM		 1591void sk_free_unlock_clone(struct sock *sk)
1592{
1593 /* It is still raw copy of parent, so invalidate
1594 * destructor and make plain sk_free() */
1595 sk->sk_destruct = NULL;
1596 bh_unlock_sock(sk);
1597 sk_free(sk);
1598}
1599EXPORT_SYMBOL_GPL(sk_free_unlock_clone);
1600
9958089a
AK		 1601void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1602{
d6a4e26a
ED		 1603 u32 max_segs = 1;
1604
6bd4f355 1605 sk_dst_set(sk, dst);
9958089a
AK		 1606 sk->sk_route_caps = dst->dev->features;
1607 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1608 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1609 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1610 if (sk_can_gso(sk)) {
82cc1a7a 1611 if (dst->header_len) {
9958089a 1612 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1613 } else {
9958089a 1614 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1615 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1616 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1617 }
9958089a 1618 }
d6a4e26a 1619 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1620}
1621EXPORT_SYMBOL_GPL(sk_setup_caps);
1622
1da177e4
LT		 1623/*
1624 * Simple resource managers for sockets.
1625 */
1626
1627
4ec93edb
YH		 1628/*
1629 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1630 */
1631void sock_wfree(struct sk_buff *skb)
1632{
1633 struct sock *sk = skb->sk;
d99927f4 1634 unsigned int len = skb->truesize;
1da177e4 1635
d99927f4
ED		 1636 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1637 /*
1638 * Keep a reference on sk_wmem_alloc, this will be released
1639 * after sk_write_space() call
1640 */
1641 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1642 sk->sk_write_space(sk);
d99927f4
ED		 1643 len = 1;
1644 }
2b85a34e 1645 /*
d99927f4
ED		 1646 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1647 * could not do because of in-flight packets
2b85a34e 1648 */
d99927f4 1649 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1650 __sk_free(sk);
1da177e4 1651}
2a91525c 1652EXPORT_SYMBOL(sock_wfree);
1da177e4 1653
1d2077ac
ED		 1654/* This variant of sock_wfree() is used by TCP,
1655 * since it sets SOCK_USE_WRITE_QUEUE.
1656 */
1657void __sock_wfree(struct sk_buff *skb)
1658{
1659 struct sock *sk = skb->sk;
1660
1661 if (atomic_sub_and_test(skb->truesize, &sk->sk_wmem_alloc))
1662 __sk_free(sk);
1663}
1664
9e17f8a4
ED		 1665void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1666{
1667 skb_orphan(skb);
1668 skb->sk = sk;
1669#ifdef CONFIG_INET
1670 if (unlikely(!sk_fullsock(sk))) {
1671 skb->destructor = sock_edemux;
1672 sock_hold(sk);
1673 return;
1674 }
1675#endif
1676 skb->destructor = sock_wfree;
1677 skb_set_hash_from_sk(skb, sk);
1678 /*
1679 * We used to take a refcount on sk, but following operation
1680 * is enough to guarantee sk_free() wont free this sock until
1681 * all in-flight packets are completed
1682 */
1683 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1684}
1685EXPORT_SYMBOL(skb_set_owner_w);
1686
1d2077ac
ED		 1687/* This helper is used by netem, as it can hold packets in its
1688 * delay queue. We want to allow the owner socket to send more
1689 * packets, as if they were already TX completed by a typical driver.
1690 * But we also want to keep skb->sk set because some packet schedulers
1691 * rely on it (sch_fq for example). So we set skb->truesize to a small
1692 * amount (1) and decrease sk_wmem_alloc accordingly.
1693 */
f2f872f9
ED		 1694void skb_orphan_partial(struct sk_buff *skb)
1695{
1d2077ac
ED		 1696 /* If this skb is a TCP pure ACK or already went here,
1697 * we have nothing to do. 2 is already a very small truesize.
1698 */
1699 if (skb->truesize <= 2)
1700 return;
1701
f2f872f9
ED		 1702 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1703 * so we do not completely orphan skb, but transfert all
1704 * accounted bytes but one, to avoid unexpected reorders.
1705 */
1706 if (skb->destructor == sock_wfree
1707#ifdef CONFIG_INET
1708 || skb->destructor == tcp_wfree
1709#endif
1710 ) {
1711 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1712 skb->truesize = 1;
1713 } else {
1714 skb_orphan(skb);
1715 }
1716}
1717EXPORT_SYMBOL(skb_orphan_partial);
1718
4ec93edb
YH		 1719/*
1720 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1721 */
1722void sock_rfree(struct sk_buff *skb)
1723{
1724 struct sock *sk = skb->sk;
d361fd59 1725 unsigned int len = skb->truesize;
1da177e4 1726
d361fd59
ED		 1727 atomic_sub(len, &sk->sk_rmem_alloc);
1728 sk_mem_uncharge(sk, len);
1da177e4 1729}
2a91525c 1730EXPORT_SYMBOL(sock_rfree);
1da177e4 1731
7768eed8
OH		 1732/*
1733 * Buffer destructor for skbs that are not used directly in read or write
1734 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1735 */
62bccb8c
AD		 1736void sock_efree(struct sk_buff *skb)
1737{
1738 sock_put(skb->sk);
1739}
1740EXPORT_SYMBOL(sock_efree);
1741
976d0201 1742kuid_t sock_i_uid(struct sock *sk)
1da177e4 1743{
976d0201 1744 kuid_t uid;
1da177e4 1745
f064af1e 1746 read_lock_bh(&sk->sk_callback_lock);
976d0201 1747 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1748 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1749 return uid;
1750}
2a91525c 1751EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1752
1753unsigned long sock_i_ino(struct sock *sk)
1754{
1755 unsigned long ino;
1756
f064af1e 1757 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1758 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1759 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1760 return ino;
1761}
2a91525c 1762EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1763
1764/*
1765 * Allocate a skb from the socket's send buffer.
1766 */
86a76caf 1767struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1768 gfp_t priority)
1da177e4
LT		 1769{
1770 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1771 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1772 if (skb) {
1773 skb_set_owner_w(skb, sk);
1774 return skb;
1775 }
1776 }
1777 return NULL;
1778}
2a91525c 1779EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1780
4ec93edb 1781/*
1da177e4 1782 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1783 */
dd0fc66f 1784void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1785{
95c96174 1786 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1787 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1788 void *mem;
1789 /* First do the add, to avoid the race if kmalloc
4ec93edb 1790 * might sleep.
1da177e4
LT		 1791 */
1792 atomic_add(size, &sk->sk_omem_alloc);
1793 mem = kmalloc(size, priority);
1794 if (mem)
1795 return mem;
1796 atomic_sub(size, &sk->sk_omem_alloc);
1797 }
1798 return NULL;
1799}
2a91525c 1800EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1801
79e88659
DB		 1802/* Free an option memory block. Note, we actually want the inline
1803 * here as this allows gcc to detect the nullify and fold away the
1804 * condition entirely.
1da177e4 1805 */
79e88659
DB		 1806static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1807 const bool nullify)
1da177e4 1808{
e53da5fb
DM		 1809 if (WARN_ON_ONCE(!mem))
1810 return;
79e88659
DB		 1811 if (nullify)
1812 kzfree(mem);
1813 else
1814 kfree(mem);
1da177e4
LT		 1815 atomic_sub(size, &sk->sk_omem_alloc);
1816}
79e88659
DB		 1817
1818void sock_kfree_s(struct sock *sk, void *mem, int size)
1819{
1820 __sock_kfree_s(sk, mem, size, false);
1821}
2a91525c 1822EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1823
79e88659
DB		 1824void sock_kzfree_s(struct sock *sk, void *mem, int size)
1825{
1826 __sock_kfree_s(sk, mem, size, true);
1827}
1828EXPORT_SYMBOL(sock_kzfree_s);
1829
1da177e4
LT		 1830/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1831 I think, these locks should be removed for datagram sockets.
1832 */
2a91525c 1833static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1834{
1835 DEFINE_WAIT(wait);
1836
9cd3e072 1837 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1da177e4
LT		 1838 for (;;) {
1839 if (!timeo)
1840 break;
1841 if (signal_pending(current))
1842 break;
1843 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1844 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1845 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1846 break;
1847 if (sk->sk_shutdown & SEND_SHUTDOWN)
1848 break;
1849 if (sk->sk_err)
1850 break;
1851 timeo = schedule_timeout(timeo);
1852 }
aa395145 1853 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1854 return timeo;
1855}
1856
1857
1858/*
1859 * Generic send/receive buffer handlers
1860 */
1861
4cc7f68d
HX		 1862struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1863 unsigned long data_len, int noblock,
28d64271 1864 int *errcode, int max_page_order)
1da177e4 1865{
2e4e4410 1866 struct sk_buff *skb;
1da177e4
LT		 1867 long timeo;
1868 int err;
1869
1da177e4 1870 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1871 for (;;) {
1da177e4
LT		 1872 err = sock_error(sk);
1873 if (err != 0)
1874 goto failure;
1875
1876 err = -EPIPE;
1877 if (sk->sk_shutdown & SEND_SHUTDOWN)
1878 goto failure;
1879
2e4e4410
ED		 1880 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1881 break;
28d64271 1882
9cd3e072 1883 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2e4e4410
ED		 1884 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1885 err = -EAGAIN;
1886 if (!timeo)
1da177e4 1887 goto failure;
2e4e4410
ED		 1888 if (signal_pending(current))
1889 goto interrupted;
1890 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1891 }
2e4e4410
ED		 1892 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1893 errcode, sk->sk_allocation);
1894 if (skb)
1895 skb_set_owner_w(skb, sk);
1da177e4
LT		 1896 return skb;
1897
1898interrupted:
1899 err = sock_intr_errno(timeo);
1900failure:
1901 *errcode = err;
1902 return NULL;
1903}
4cc7f68d 1904EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1905
4ec93edb 1906struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1907 int noblock, int *errcode)
1908{
28d64271 1909 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1910}
2a91525c 1911EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1912
39771b12
WB		 1913int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1914 struct sockcm_cookie *sockc)
1915{
3dd17e63
SHY		 1916 u32 tsflags;
1917
39771b12
WB		 1918 switch (cmsg->cmsg_type) {
1919 case SO_MARK:
1920 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1921 return -EPERM;
1922 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1923 return -EINVAL;
1924 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
1925 break;
3dd17e63
SHY		 1926 case SO_TIMESTAMPING:
1927 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1928 return -EINVAL;
1929
1930 tsflags = *(u32 *)CMSG_DATA(cmsg);
1931 if (tsflags & ~SOF_TIMESTAMPING_TX_RECORD_MASK)
1932 return -EINVAL;
1933
1934 sockc->tsflags &= ~SOF_TIMESTAMPING_TX_RECORD_MASK;
1935 sockc->tsflags |= tsflags;
1936 break;
779f1ede
SHY		 1937 /* SCM_RIGHTS and SCM_CREDENTIALS are semantically in SOL_UNIX. */
1938 case SCM_RIGHTS:
1939 case SCM_CREDENTIALS:
1940 break;
39771b12
WB		 1941 default:
1942 return -EINVAL;
1943 }
1944 return 0;
1945}
1946EXPORT_SYMBOL(__sock_cmsg_send);
1947
f28ea365
EJ		 1948int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1949 struct sockcm_cookie *sockc)
1950{
1951 struct cmsghdr *cmsg;
39771b12 1952 int ret;
f28ea365
EJ		 1953
1954 for_each_cmsghdr(cmsg, msg) {
1955 if (!CMSG_OK(msg, cmsg))
1956 return -EINVAL;
1957 if (cmsg->cmsg_level != SOL_SOCKET)
1958 continue;
39771b12
WB		 1959 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
1960 if (ret)
1961 return ret;
f28ea365
EJ		 1962 }
1963 return 0;
1964}
1965EXPORT_SYMBOL(sock_cmsg_send);
1966
5640f768
ED		 1967/* On 32bit arches, an skb frag is limited to 2^15 */
1968#define SKB_FRAG_PAGE_ORDER get_order(32768)
1969
400dfd3a
ED		 1970/**
1971 * skb_page_frag_refill - check that a page_frag contains enough room
1972 * @sz: minimum size of the fragment we want to get
1973 * @pfrag: pointer to page_frag
82d5e2b8 1974 * @gfp: priority for memory allocation
400dfd3a
ED		 1975 *
1976 * Note: While this allocator tries to use high order pages, there is
1977 * no guarantee that allocations succeed. Therefore, @sz MUST be
1978 * less or equal than PAGE_SIZE.
1979 */
d9b2938a 1980bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1981{
5640f768 1982 if (pfrag->page) {
fe896d18 1983 if (page_ref_count(pfrag->page) == 1) {
5640f768
ED		 1984 pfrag->offset = 0;
1985 return true;
1986 }
400dfd3a 1987 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1988 return true;
1989 put_page(pfrag->page);
1990 }
1991
d9b2938a
ED		 1992 pfrag->offset = 0;
1993 if (SKB_FRAG_PAGE_ORDER) {
d0164adc
MG		 1994 /* Avoid direct reclaim but allow kswapd to wake */
1995 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
1996 __GFP_COMP | __GFP_NOWARN |
1997 __GFP_NORETRY,
d9b2938a 1998 SKB_FRAG_PAGE_ORDER);
5640f768 1999 if (likely(pfrag->page)) {
d9b2938a 2000 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 2001 return true;
2002 }
d9b2938a
ED		 2003 }
2004 pfrag->page = alloc_page(gfp);
2005 if (likely(pfrag->page)) {
2006 pfrag->size = PAGE_SIZE;
2007 return true;
2008 }
400dfd3a
ED		 2009 return false;
2010}
2011EXPORT_SYMBOL(skb_page_frag_refill);
2012
2013bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
2014{
2015 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
2016 return true;
2017
5640f768
ED		 2018 sk_enter_memory_pressure(sk);
2019 sk_stream_moderate_sndbuf(sk);
2020 return false;
2021}
2022EXPORT_SYMBOL(sk_page_frag_refill);
2023
1da177e4 2024static void __lock_sock(struct sock *sk)
f39234d6
NK		 2025 __releases(&sk->sk_lock.slock)
2026 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 2027{
2028 DEFINE_WAIT(wait);
2029
e71a4783 2030 for (;;) {
1da177e4
LT		 2031 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
2032 TASK_UNINTERRUPTIBLE);
2033 spin_unlock_bh(&sk->sk_lock.slock);
2034 schedule();
2035 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 2036 if (!sock_owned_by_user(sk))
1da177e4
LT		 2037 break;
2038 }
2039 finish_wait(&sk->sk_lock.wq, &wait);
2040}
2041
2042static void __release_sock(struct sock *sk)
f39234d6
NK		 2043 __releases(&sk->sk_lock.slock)
2044 __acquires(&sk->sk_lock.slock)
1da177e4 2045{
5413d1ba 2046 struct sk_buff *skb, *next;
1da177e4 2047
5413d1ba 2048 while ((skb = sk->sk_backlog.head) != NULL) {
1da177e4 2049 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1da177e4 2050
5413d1ba 2051 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4 2052
5413d1ba
ED		 2053 do {
2054 next = skb->next;
e4cbb02a 2055 prefetch(next);
7fee226a 2056 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 2057 skb->next = NULL;
c57943a1 2058 sk_backlog_rcv(sk, skb);
1da177e4 2059
5413d1ba 2060 cond_resched();
1da177e4
LT		 2061
2062 skb = next;
2063 } while (skb != NULL);
2064
5413d1ba
ED		 2065 spin_lock_bh(&sk->sk_lock.slock);
2066 }
8eae939f
ZY		 2067
2068 /*
2069 * Doing the zeroing here guarantee we can not loop forever
2070 * while a wild producer attempts to flood us.
2071 */
2072 sk->sk_backlog.len = 0;
1da177e4
LT		 2073}
2074
d41a69f1
ED		 2075void __sk_flush_backlog(struct sock *sk)
2076{
2077 spin_lock_bh(&sk->sk_lock.slock);
2078 __release_sock(sk);
2079 spin_unlock_bh(&sk->sk_lock.slock);
2080}
2081
1da177e4
LT		 2082/**
2083 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 2084 * @sk: sock to wait on
2085 * @timeo: for how long
dfbafc99 2086 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 2087 *
2088 * Now socket state including sk->sk_err is changed only under lock,
2089 * hence we may omit checks after joining wait queue.
2090 * We check receive queue before schedule() only as optimization;
2091 * it is very likely that release_sock() added new data.
2092 */
dfbafc99 2093int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4 2094{
d9dc8b0f 2095 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1da177e4 2096 int rc;
1da177e4 2097
d9dc8b0f 2098 add_wait_queue(sk_sleep(sk), &wait);
9cd3e072 2099 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
d9dc8b0f 2100 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb, &wait);
9cd3e072 2101 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
d9dc8b0f 2102 remove_wait_queue(sk_sleep(sk), &wait);
1da177e4
LT		 2103 return rc;
2104}
1da177e4
LT		 2105EXPORT_SYMBOL(sk_wait_data);
2106
3ab224be 2107/**
f8c3bf00 2108 * __sk_mem_raise_allocated - increase memory_allocated
3ab224be
HA		 2109 * @sk: socket
2110 * @size: memory size to allocate
f8c3bf00 2111 * @amt: pages to allocate
3ab224be
HA		 2112 * @kind: allocation type
2113 *
f8c3bf00 2114 * Similar to __sk_mem_schedule(), but does not update sk_forward_alloc
3ab224be 2115 */
f8c3bf00 2116int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
3ab224be
HA		 2117{
2118 struct proto *prot = sk->sk_prot;
f8c3bf00 2119 long allocated = sk_memory_allocated_add(sk, amt);
e805605c 2120
baac50bb
JW		 2121 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2122 !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
e805605c 2123 goto suppress_allocation;
3ab224be
HA		 2124
2125 /* Under limit. */
e805605c 2126 if (allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2127 sk_leave_memory_pressure(sk);
3ab224be
HA		 2128 return 1;
2129 }
2130
e805605c
JW		 2131 /* Under pressure. */
2132 if (allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2133 sk_enter_memory_pressure(sk);
3ab224be 2134
e805605c
JW		 2135 /* Over hard limit. */
2136 if (allocated > sk_prot_mem_limits(sk, 2))
3ab224be
HA		 2137 goto suppress_allocation;
2138
2139 /* guarantee minimum buffer size under pressure */
2140 if (kind == SK_MEM_RECV) {
2141 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2142 return 1;
180d8cd9 2143
3ab224be
HA		 2144 } else { /* SK_MEM_SEND */
2145 if (sk->sk_type == SOCK_STREAM) {
2146 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2147 return 1;
2148 } else if (atomic_read(&sk->sk_wmem_alloc) <
2149 prot->sysctl_wmem[0])
2150 return 1;
2151 }
2152
180d8cd9 2153 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2154 int alloc;
2155
180d8cd9 2156 if (!sk_under_memory_pressure(sk))
1748376b 2157 return 1;
180d8cd9
GC		 2158 alloc = sk_sockets_allocated_read_positive(sk);
2159 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2160 sk_mem_pages(sk->sk_wmem_queued +
2161 atomic_read(&sk->sk_rmem_alloc) +
2162 sk->sk_forward_alloc))
2163 return 1;
2164 }
2165
2166suppress_allocation:
2167
2168 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2169 sk_stream_moderate_sndbuf(sk);
2170
2171 /* Fail only if socket is _under_ its sndbuf.
2172 * In this case we cannot block, so that we have to fail.
2173 */
2174 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2175 return 1;
2176 }
2177
3847ce32
SM		 2178 trace_sock_exceed_buf_limit(sk, prot, allocated);
2179
0e90b31f 2180 sk_memory_allocated_sub(sk, amt);
180d8cd9 2181
baac50bb
JW		 2182 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2183 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
e805605c 2184
3ab224be
HA		 2185 return 0;
2186}
f8c3bf00
PA		 2187EXPORT_SYMBOL(__sk_mem_raise_allocated);
2188
2189/**
2190 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2191 * @sk: socket
2192 * @size: memory size to allocate
2193 * @kind: allocation type
2194 *
2195 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2196 * rmem allocation. This function assumes that protocols which have
2197 * memory_pressure use sk_wmem_queued as write buffer accounting.
2198 */
2199int __sk_mem_schedule(struct sock *sk, int size, int kind)
2200{
2201 int ret, amt = sk_mem_pages(size);
2202
2203 sk->sk_forward_alloc += amt << SK_MEM_QUANTUM_SHIFT;
2204 ret = __sk_mem_raise_allocated(sk, size, amt, kind);
2205 if (!ret)
2206 sk->sk_forward_alloc -= amt << SK_MEM_QUANTUM_SHIFT;
2207 return ret;
2208}
3ab224be
HA		 2209EXPORT_SYMBOL(__sk_mem_schedule);
2210
2211/**
f8c3bf00 2212 * __sk_mem_reduce_allocated - reclaim memory_allocated
3ab224be 2213 * @sk: socket
f8c3bf00
PA		 2214 * @amount: number of quanta
2215 *
2216 * Similar to __sk_mem_reclaim(), but does not update sk_forward_alloc
3ab224be 2217 */
f8c3bf00 2218void __sk_mem_reduce_allocated(struct sock *sk, int amount)
3ab224be 2219{
1a24e04e 2220 sk_memory_allocated_sub(sk, amount);
3ab224be 2221
baac50bb
JW		 2222 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2223 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
e805605c 2224
180d8cd9
GC		 2225 if (sk_under_memory_pressure(sk) &&
2226 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2227 sk_leave_memory_pressure(sk);
3ab224be 2228}
f8c3bf00
PA		 2229EXPORT_SYMBOL(__sk_mem_reduce_allocated);
2230
2231/**
2232 * __sk_mem_reclaim - reclaim sk_forward_alloc and memory_allocated
2233 * @sk: socket
2234 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
2235 */
2236void __sk_mem_reclaim(struct sock *sk, int amount)
2237{
2238 amount >>= SK_MEM_QUANTUM_SHIFT;
2239 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
2240 __sk_mem_reduce_allocated(sk, amount);
2241}
3ab224be
HA		 2242EXPORT_SYMBOL(__sk_mem_reclaim);
2243
627d2d6b 2244int sk_set_peek_off(struct sock *sk, int val)
2245{
2246 if (val < 0)
2247 return -EINVAL;
2248
2249 sk->sk_peek_off = val;
2250 return 0;
2251}
2252EXPORT_SYMBOL_GPL(sk_set_peek_off);
3ab224be 2253
1da177e4
LT		 2254/*
2255 * Set of default routines for initialising struct proto_ops when
2256 * the protocol does not support a particular function. In certain
2257 * cases where it makes no sense for a protocol to have a "do nothing"
2258 * function, some default processing is provided.
2259 */
2260
2261int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2262{
2263 return -EOPNOTSUPP;
2264}
2a91525c 2265EXPORT_SYMBOL(sock_no_bind);
1da177e4 2266
4ec93edb 2267int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2268 int len, int flags)
2269{
2270 return -EOPNOTSUPP;
2271}
2a91525c 2272EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2273
2274int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2275{
2276 return -EOPNOTSUPP;
2277}
2a91525c 2278EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2279
2280int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2281{
2282 return -EOPNOTSUPP;
2283}
2a91525c 2284EXPORT_SYMBOL(sock_no_accept);
1da177e4 2285
4ec93edb 2286int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2287 int *len, int peer)
2288{
2289 return -EOPNOTSUPP;
2290}
2a91525c 2291EXPORT_SYMBOL(sock_no_getname);
1da177e4 2292
2a91525c 2293unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2294{
2295 return 0;
2296}
2a91525c 2297EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2298
2299int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2300{
2301 return -EOPNOTSUPP;
2302}
2a91525c 2303EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2304
2305int sock_no_listen(struct socket *sock, int backlog)
2306{
2307 return -EOPNOTSUPP;
2308}
2a91525c 2309EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2310
2311int sock_no_shutdown(struct socket *sock, int how)
2312{
2313 return -EOPNOTSUPP;
2314}
2a91525c 2315EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2316
2317int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2318 char __user *optval, unsigned int optlen)
1da177e4
LT		 2319{
2320 return -EOPNOTSUPP;
2321}
2a91525c 2322EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2323
2324int sock_no_getsockopt(struct socket *sock, int level, int optname,
2325 char __user *optval, int __user *optlen)
2326{
2327 return -EOPNOTSUPP;
2328}
2a91525c 2329EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2330
1b784140 2331int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2332{
2333 return -EOPNOTSUPP;
2334}
2a91525c 2335EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2336
1b784140
YX		 2337int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2338 int flags)
1da177e4
LT		 2339{
2340 return -EOPNOTSUPP;
2341}
2a91525c 2342EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2343
2344int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2345{
2346 /* Mirror missing mmap method error code */
2347 return -ENODEV;
2348}
2a91525c 2349EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2350
2351ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2352{
2353 ssize_t res;
2354 struct msghdr msg = {.msg_flags = flags};
2355 struct kvec iov;
2356 char *kaddr = kmap(page);
2357 iov.iov_base = kaddr + offset;
2358 iov.iov_len = size;
2359 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2360 kunmap(page);
2361 return res;
2362}
2a91525c 2363EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2364
2365/*
2366 * Default Socket Callbacks
2367 */
2368
2369static void sock_def_wakeup(struct sock *sk)
2370{
43815482
ED		 2371 struct socket_wq *wq;
2372
2373 rcu_read_lock();
2374 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2375 if (skwq_has_sleeper(wq))
43815482
ED		 2376 wake_up_interruptible_all(&wq->wait);
2377 rcu_read_unlock();
1da177e4
LT		 2378}
2379
2380static void sock_def_error_report(struct sock *sk)
2381{
43815482
ED		 2382 struct socket_wq *wq;
2383
2384 rcu_read_lock();
2385 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2386 if (skwq_has_sleeper(wq))
43815482 2387 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2388 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2389 rcu_read_unlock();
1da177e4
LT		 2390}
2391
676d2369 2392static void sock_def_readable(struct sock *sk)
1da177e4 2393{
43815482
ED		 2394 struct socket_wq *wq;
2395
2396 rcu_read_lock();
2397 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2398 if (skwq_has_sleeper(wq))
2c6607c6 2399 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2400 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2401 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2402 rcu_read_unlock();
1da177e4
LT		 2403}
2404
2405static void sock_def_write_space(struct sock *sk)
2406{
43815482
ED		 2407 struct socket_wq *wq;
2408
2409 rcu_read_lock();
1da177e4
LT		 2410
2411 /* Do not wake up a writer until he can make "significant"
2412 * progress. --DaveM
2413 */
e71a4783 2414 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482 2415 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2416 if (skwq_has_sleeper(wq))
43815482 2417 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2418 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2419
2420 /* Should agree with poll, otherwise some programs break */
2421 if (sock_writeable(sk))
8d8ad9d7 2422 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2423 }
2424
43815482 2425 rcu_read_unlock();
1da177e4
LT		 2426}
2427
2428static void sock_def_destruct(struct sock *sk)
2429{
1da177e4
LT		 2430}
2431
2432void sk_send_sigurg(struct sock *sk)
2433{
2434 if (sk->sk_socket && sk->sk_socket->file)
2435 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2436 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2437}
2a91525c 2438EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2439
2440void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2441 unsigned long expires)
2442{
2443 if (!mod_timer(timer, expires))
2444 sock_hold(sk);
2445}
1da177e4
LT		 2446EXPORT_SYMBOL(sk_reset_timer);
2447
2448void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2449{
25cc4ae9 2450 if (del_timer(timer))
1da177e4
LT		 2451 __sock_put(sk);
2452}
1da177e4
LT		 2453EXPORT_SYMBOL(sk_stop_timer);
2454
2455void sock_init_data(struct socket *sock, struct sock *sk)
2456{
2457 skb_queue_head_init(&sk->sk_receive_queue);
2458 skb_queue_head_init(&sk->sk_write_queue);
2459 skb_queue_head_init(&sk->sk_error_queue);
2460
2461 sk->sk_send_head = NULL;
2462
2463 init_timer(&sk->sk_timer);
4ec93edb 2464
1da177e4
LT		 2465 sk->sk_allocation = GFP_KERNEL;
2466 sk->sk_rcvbuf = sysctl_rmem_default;
2467 sk->sk_sndbuf = sysctl_wmem_default;
2468 sk->sk_state = TCP_CLOSE;
972692e0 2469 sk_set_socket(sk, sock);
1da177e4
LT		 2470
2471 sock_set_flag(sk, SOCK_ZAPPED);
2472
e71a4783 2473 if (sock) {
1da177e4 2474 sk->sk_type = sock->type;
43815482 2475 sk->sk_wq = sock->wq;
1da177e4 2476 sock->sk = sk;
86741ec2
LC		 2477 sk->sk_uid = SOCK_INODE(sock)->i_uid;
2478 } else {
43815482 2479 sk->sk_wq = NULL;
86741ec2
LC		 2480 sk->sk_uid = make_kuid(sock_net(sk)->user_ns, 0);
2481 }
1da177e4 2482
1da177e4 2483 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2484 lockdep_set_class_and_name(&sk->sk_callback_lock,
2485 af_callback_keys + sk->sk_family,
2486 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2487
2488 sk->sk_state_change = sock_def_wakeup;
2489 sk->sk_data_ready = sock_def_readable;
2490 sk->sk_write_space = sock_def_write_space;
2491 sk->sk_error_report = sock_def_error_report;
2492 sk->sk_destruct = sock_def_destruct;
2493
5640f768
ED		 2494 sk->sk_frag.page = NULL;
2495 sk->sk_frag.offset = 0;
ef64a54f 2496 sk->sk_peek_off = -1;
1da177e4 2497
109f6e39
EB		 2498 sk->sk_peer_pid = NULL;
2499 sk->sk_peer_cred = NULL;
1da177e4
LT		 2500 sk->sk_write_pending = 0;
2501 sk->sk_rcvlowat = 1;
2502 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2503 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2504
f37f0afb 2505 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2506
e0d1095a 2507#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2508 sk->sk_napi_id = 0;
64b0dc51 2509 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2510#endif
2511
62748f32 2512 sk->sk_max_pacing_rate = ~0U;
7eec4174 2513 sk->sk_pacing_rate = ~0U;
70da268b 2514 sk->sk_incoming_cpu = -1;
4dc6dc71
ED		 2515 /*
2516 * Before updating sk_refcnt, we must commit prior changes to memory
2517 * (Documentation/RCU/rculist_nulls.txt for details)
2518 */
2519 smp_wmb();
1da177e4 2520 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2521 atomic_set(&sk->sk_drops, 0);
1da177e4 2522}
2a91525c 2523EXPORT_SYMBOL(sock_init_data);
1da177e4 2524
b5606c2d 2525void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2526{
2527 might_sleep();
a5b5bb9a 2528 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2529 if (sk->sk_lock.owned)
1da177e4 2530 __lock_sock(sk);
d2e9117c 2531 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2532 spin_unlock(&sk->sk_lock.slock);
2533 /*
2534 * The sk_lock has mutex_lock() semantics here:
2535 */
fcc70d5f 2536 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2537 local_bh_enable();
1da177e4 2538}
fcc70d5f 2539EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2540
b5606c2d 2541void release_sock(struct sock *sk)
1da177e4 2542{
a5b5bb9a 2543 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2544 if (sk->sk_backlog.tail)
2545 __release_sock(sk);
46d3ceab 2546
c3f9b018
ED		 2547 /* Warning : release_cb() might need to release sk ownership,
2548 * ie call sock_release_ownership(sk) before us.
2549 */
46d3ceab
ED		 2550 if (sk->sk_prot->release_cb)
2551 sk->sk_prot->release_cb(sk);
2552
c3f9b018 2553 sock_release_ownership(sk);
a5b5bb9a
IM		 2554 if (waitqueue_active(&sk->sk_lock.wq))
2555 wake_up(&sk->sk_lock.wq);
2556 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2557}
2558EXPORT_SYMBOL(release_sock);
2559
8a74ad60
ED		 2560/**
2561 * lock_sock_fast - fast version of lock_sock
2562 * @sk: socket
2563 *
2564 * This version should be used for very small section, where process wont block
2565 * return false if fast path is taken
2566 * sk_lock.slock locked, owned = 0, BH disabled
2567 * return true if slow path is taken
2568 * sk_lock.slock unlocked, owned = 1, BH enabled
2569 */
2570bool lock_sock_fast(struct sock *sk)
2571{
2572 might_sleep();
2573 spin_lock_bh(&sk->sk_lock.slock);
2574
2575 if (!sk->sk_lock.owned)
2576 /*
2577 * Note : We must disable BH
2578 */
2579 return false;
2580
2581 __lock_sock(sk);
2582 sk->sk_lock.owned = 1;
2583 spin_unlock(&sk->sk_lock.slock);
2584 /*
2585 * The sk_lock has mutex_lock() semantics here:
2586 */
2587 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2588 local_bh_enable();
2589 return true;
2590}
2591EXPORT_SYMBOL(lock_sock_fast);
2592
1da177e4 2593int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2594{
b7aa0bf7 2595 struct timeval tv;
1da177e4 2596 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2597 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2598 tv = ktime_to_timeval(sk->sk_stamp);
2599 if (tv.tv_sec == -1)
1da177e4 2600 return -ENOENT;
b7aa0bf7
ED		 2601 if (tv.tv_sec == 0) {
2602 sk->sk_stamp = ktime_get_real();
2603 tv = ktime_to_timeval(sk->sk_stamp);
2604 }
2605 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2606}
1da177e4
LT		 2607EXPORT_SYMBOL(sock_get_timestamp);
2608
ae40eb1e
ED		 2609int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2610{
2611 struct timespec ts;
2612 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2613 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2614 ts = ktime_to_timespec(sk->sk_stamp);
2615 if (ts.tv_sec == -1)
2616 return -ENOENT;
2617 if (ts.tv_sec == 0) {
2618 sk->sk_stamp = ktime_get_real();
2619 ts = ktime_to_timespec(sk->sk_stamp);
2620 }
2621 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2622}
2623EXPORT_SYMBOL(sock_get_timestampns);
2624
20d49473 2625void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2626{
20d49473 2627 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2628 unsigned long previous_flags = sk->sk_flags;
2629
20d49473
PO		 2630 sock_set_flag(sk, flag);
2631 /*
2632 * we just set one of the two flags which require net
2633 * time stamping, but time stamping might have been on
2634 * already because of the other one
2635 */
080a270f
HFS		 2636 if (sock_needs_netstamp(sk) &&
2637 !(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2638 net_enable_timestamp();
1da177e4
LT		 2639 }
2640}
1da177e4 2641
cb820f8e
RC		 2642int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2643 int level, int type)
2644{
2645 struct sock_exterr_skb *serr;
364a9e93 2646 struct sk_buff *skb;
cb820f8e
RC		 2647 int copied, err;
2648
2649 err = -EAGAIN;
364a9e93 2650 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2651 if (skb == NULL)
2652 goto out;
2653
2654 copied = skb->len;
2655 if (copied > len) {
2656 msg->msg_flags |= MSG_TRUNC;
2657 copied = len;
2658 }
51f3d02b 2659 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2660 if (err)
2661 goto out_free_skb;
2662
2663 sock_recv_timestamp(msg, sk, skb);
2664
2665 serr = SKB_EXT_ERR(skb);
2666 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2667
2668 msg->msg_flags |= MSG_ERRQUEUE;
2669 err = copied;
2670
cb820f8e
RC		 2671out_free_skb:
2672 kfree_skb(skb);
2673out:
2674 return err;
2675}
2676EXPORT_SYMBOL(sock_recv_errqueue);
2677
1da177e4
LT		 2678/*
2679 * Get a socket option on an socket.
2680 *
2681 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2682 * asynchronous errors should be reported by getsockopt. We assume
2683 * this means if you specify SO_ERROR (otherwise whats the point of it).
2684 */
2685int sock_common_getsockopt(struct socket *sock, int level, int optname,
2686 char __user *optval, int __user *optlen)
2687{
2688 struct sock *sk = sock->sk;
2689
2690 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2691}
1da177e4
LT		 2692EXPORT_SYMBOL(sock_common_getsockopt);
2693
3fdadf7d 2694#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2695int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2696 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2697{
2698 struct sock *sk = sock->sk;
2699
1e51f951 2700 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2701 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2702 optval, optlen);
3fdadf7d
DM		 2703 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2704}
2705EXPORT_SYMBOL(compat_sock_common_getsockopt);
2706#endif
2707
1b784140
YX		 2708int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2709 int flags)
1da177e4
LT		 2710{
2711 struct sock *sk = sock->sk;
2712 int addr_len = 0;
2713 int err;
2714
1b784140 2715 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2716 flags & ~MSG_DONTWAIT, &addr_len);
2717 if (err >= 0)
2718 msg->msg_namelen = addr_len;
2719 return err;
2720}
1da177e4
LT		 2721EXPORT_SYMBOL(sock_common_recvmsg);
2722
2723/*
2724 * Set socket options on an inet socket.
2725 */
2726int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2727 char __user *optval, unsigned int optlen)
1da177e4
LT		 2728{
2729 struct sock *sk = sock->sk;
2730
2731 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2732}
1da177e4
LT		 2733EXPORT_SYMBOL(sock_common_setsockopt);
2734
3fdadf7d 2735#ifdef CONFIG_COMPAT
543d9cfe 2736int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2737 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2738{
2739 struct sock *sk = sock->sk;
2740
543d9cfe
ACM		 2741 if (sk->sk_prot->compat_setsockopt != NULL)
2742 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2743 optval, optlen);
3fdadf7d
DM		 2744 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2745}
2746EXPORT_SYMBOL(compat_sock_common_setsockopt);
2747#endif
2748
1da177e4
LT		 2749void sk_common_release(struct sock *sk)
2750{
2751 if (sk->sk_prot->destroy)
2752 sk->sk_prot->destroy(sk);
2753
2754 /*
2755 * Observation: when sock_common_release is called, processes have
2756 * no access to socket. But net still has.
2757 * Step one, detach it from networking:
2758 *
2759 * A. Remove from hash tables.
2760 */
2761
2762 sk->sk_prot->unhash(sk);
2763
2764 /*
2765 * In this point socket cannot receive new packets, but it is possible
2766 * that some packets are in flight because some CPU runs receiver and
2767 * did hash table lookup before we unhashed socket. They will achieve
2768 * receive queue and will be purged by socket destructor.
2769 *
2770 * Also we still have packets pending on receive queue and probably,
2771 * our own packets waiting in device queues. sock_destroy will drain
2772 * receive queue, but transmitted packets will delay socket destruction
2773 * until the last reference will be released.
2774 */
2775
2776 sock_orphan(sk);
2777
2778 xfrm_sk_free_policy(sk);
2779
e6848976 2780 sk_refcnt_debug_release(sk);
5640f768
ED		 2781
2782 if (sk->sk_frag.page) {
2783 put_page(sk->sk_frag.page);
2784 sk->sk_frag.page = NULL;
2785 }
2786
1da177e4
LT		 2787 sock_put(sk);
2788}
1da177e4
LT		 2789EXPORT_SYMBOL(sk_common_release);
2790
13ff3d6f
PE		 2791#ifdef CONFIG_PROC_FS
2792#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2793struct prot_inuse {
2794 int val[PROTO_INUSE_NR];
2795};
13ff3d6f
PE		 2796
2797static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2798
2799#ifdef CONFIG_NET_NS
2800void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2801{
d6d9ca0f 2802 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2803}
2804EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2805
2806int sock_prot_inuse_get(struct net *net, struct proto *prot)
2807{
2808 int cpu, idx = prot->inuse_idx;
2809 int res = 0;
2810
2811 for_each_possible_cpu(cpu)
2812 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2813
2814 return res >= 0 ? res : 0;
2815}
2816EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2817
2c8c1e72 2818static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2819{
2820 net->core.inuse = alloc_percpu(struct prot_inuse);
2821 return net->core.inuse ? 0 : -ENOMEM;
2822}
2823
2c8c1e72 2824static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2825{
2826 free_percpu(net->core.inuse);
2827}
2828
2829static struct pernet_operations net_inuse_ops = {
2830 .init = sock_inuse_init_net,
2831 .exit = sock_inuse_exit_net,
2832};
2833
2834static __init int net_inuse_init(void)
2835{
2836 if (register_pernet_subsys(&net_inuse_ops))
2837 panic("Cannot initialize net inuse counters");
2838
2839 return 0;
2840}
2841
2842core_initcall(net_inuse_init);
2843#else
1338d466
PE		 2844static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2845
c29a0bc4 2846void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2847{
d6d9ca0f 2848 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2849}
2850EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2851
c29a0bc4 2852int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2853{
2854 int cpu, idx = prot->inuse_idx;
2855 int res = 0;
2856
2857 for_each_possible_cpu(cpu)
2858 res += per_cpu(prot_inuse, cpu).val[idx];
2859
2860 return res >= 0 ? res : 0;
2861}
2862EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2863#endif
13ff3d6f
PE		 2864
2865static void assign_proto_idx(struct proto *prot)
2866{
2867 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2868
2869 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2870 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2871 return;
2872 }
2873
2874 set_bit(prot->inuse_idx, proto_inuse_idx);
2875}
2876
2877static void release_proto_idx(struct proto *prot)
2878{
2879 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2880 clear_bit(prot->inuse_idx, proto_inuse_idx);
2881}
2882#else
2883static inline void assign_proto_idx(struct proto *prot)
2884{
2885}
2886
2887static inline void release_proto_idx(struct proto *prot)
2888{
2889}
2890#endif
2891
0159dfd3
ED		 2892static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2893{
2894 if (!rsk_prot)
2895 return;
2896 kfree(rsk_prot->slab_name);
2897 rsk_prot->slab_name = NULL;
adf78eda
JL		 2898 kmem_cache_destroy(rsk_prot->slab);
2899 rsk_prot->slab = NULL;
0159dfd3
ED		 2900}
2901
2902static int req_prot_init(const struct proto *prot)
2903{
2904 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2905
2906 if (!rsk_prot)
2907 return 0;
2908
2909 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2910 prot->name);
2911 if (!rsk_prot->slab_name)
2912 return -ENOMEM;
2913
2914 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2915 rsk_prot->obj_size, 0,
e96f78ab 2916 prot->slab_flags, NULL);
0159dfd3
ED		 2917
2918 if (!rsk_prot->slab) {
2919 pr_crit("%s: Can't create request sock SLAB cache!\n",
2920 prot->name);
2921 return -ENOMEM;
2922 }
2923 return 0;
2924}
2925
b733c007
PE		 2926int proto_register(struct proto *prot, int alloc_slab)
2927{
1da177e4
LT		 2928 if (alloc_slab) {
2929 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2930 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2931 NULL);
1da177e4
LT		 2932
2933 if (prot->slab == NULL) {
e005d193
JP		 2934 pr_crit("%s: Can't create sock SLAB cache!\n",
2935 prot->name);
60e7663d 2936 goto out;
1da177e4 2937 }
2e6599cb 2938
0159dfd3
ED		 2939 if (req_prot_init(prot))
2940 goto out_free_request_sock_slab;
8feaf0c0 2941
6d6ee43e 2942 if (prot->twsk_prot != NULL) {
faf23422 2943 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2944
7e56b5d6 2945 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2946 goto out_free_request_sock_slab;
2947
6d6ee43e 2948 prot->twsk_prot->twsk_slab =
7e56b5d6 2949 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2950 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2951 0,
52db70dc 2952 prot->slab_flags,
20c2df83 2953 NULL);
6d6ee43e 2954 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2955 goto out_free_timewait_sock_slab_name;
2956 }
1da177e4
LT		 2957 }
2958
36b77a52 2959 mutex_lock(&proto_list_mutex);
1da177e4 2960 list_add(&prot->node, &proto_list);
13ff3d6f 2961 assign_proto_idx(prot);
36b77a52 2962 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2963 return 0;
2964
8feaf0c0 2965out_free_timewait_sock_slab_name:
7e56b5d6 2966 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2967out_free_request_sock_slab:
0159dfd3
ED		 2968 req_prot_cleanup(prot->rsk_prot);
2969
2e6599cb
ACM		 2970 kmem_cache_destroy(prot->slab);
2971 prot->slab = NULL;
b733c007
PE		 2972out:
2973 return -ENOBUFS;
1da177e4 2974}
1da177e4
LT		 2975EXPORT_SYMBOL(proto_register);
2976
2977void proto_unregister(struct proto *prot)
2978{
36b77a52 2979 mutex_lock(&proto_list_mutex);
13ff3d6f 2980 release_proto_idx(prot);
0a3f4358 2981 list_del(&prot->node);
36b77a52 2982 mutex_unlock(&proto_list_mutex);
1da177e4 2983
adf78eda
JL		 2984 kmem_cache_destroy(prot->slab);
2985 prot->slab = NULL;
1da177e4 2986
0159dfd3 2987 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2988
6d6ee43e 2989 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2990 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2991 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2992 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2993 }
1da177e4 2994}
1da177e4
LT		 2995EXPORT_SYMBOL(proto_unregister);
2996
2997#ifdef CONFIG_PROC_FS
1da177e4 2998static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2999 __acquires(proto_list_mutex)
1da177e4 3000{
36b77a52 3001 mutex_lock(&proto_list_mutex);
60f0438a 3002 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 3003}
3004
3005static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3006{
60f0438a 3007 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 3008}
3009
3010static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 3011 __releases(proto_list_mutex)
1da177e4 3012{
36b77a52 3013 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 3014}
3015
3016static char proto_method_implemented(const void *method)
3017{
3018 return method == NULL ? 'n' : 'y';
3019}
180d8cd9
GC		 3020static long sock_prot_memory_allocated(struct proto *proto)
3021{
cb75a36c 3022 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 3023}
3024
3025static char *sock_prot_memory_pressure(struct proto *proto)
3026{
3027 return proto->memory_pressure != NULL ?
3028 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
3029}
1da177e4
LT		 3030
3031static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
3032{
180d8cd9 3033
8d987e5c 3034 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 3035 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
3036 proto->name,
3037 proto->obj_size,
14e943db 3038 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 3039 sock_prot_memory_allocated(proto),
3040 sock_prot_memory_pressure(proto),
1da177e4
LT		 3041 proto->max_header,
3042 proto->slab == NULL ? "no" : "yes",
3043 module_name(proto->owner),
3044 proto_method_implemented(proto->close),
3045 proto_method_implemented(proto->connect),
3046 proto_method_implemented(proto->disconnect),
3047 proto_method_implemented(proto->accept),
3048 proto_method_implemented(proto->ioctl),
3049 proto_method_implemented(proto->init),
3050 proto_method_implemented(proto->destroy),
3051 proto_method_implemented(proto->shutdown),
3052 proto_method_implemented(proto->setsockopt),
3053 proto_method_implemented(proto->getsockopt),
3054 proto_method_implemented(proto->sendmsg),
3055 proto_method_implemented(proto->recvmsg),
3056 proto_method_implemented(proto->sendpage),
3057 proto_method_implemented(proto->bind),
3058 proto_method_implemented(proto->backlog_rcv),
3059 proto_method_implemented(proto->hash),
3060 proto_method_implemented(proto->unhash),
3061 proto_method_implemented(proto->get_port),
3062 proto_method_implemented(proto->enter_memory_pressure));
3063}
3064
3065static int proto_seq_show(struct seq_file *seq, void *v)
3066{
60f0438a 3067 if (v == &proto_list)
1da177e4
LT		 3068 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3069 "protocol",
3070 "size",
3071 "sockets",
3072 "memory",
3073 "press",
3074 "maxhdr",
3075 "slab",
3076 "module",
3077 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3078 else
60f0438a 3079 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 3080 return 0;
3081}
3082
f690808e 3083static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 3084 .start = proto_seq_start,
3085 .next = proto_seq_next,
3086 .stop = proto_seq_stop,
3087 .show = proto_seq_show,
3088};
3089
3090static int proto_seq_open(struct inode *inode, struct file *file)
3091{
14e943db
ED		 3092 return seq_open_net(inode, file, &proto_seq_ops,
3093 sizeof(struct seq_net_private));
1da177e4
LT		 3094}
3095
9a32144e 3096static const struct file_operations proto_seq_fops = {
1da177e4
LT		 3097 .owner = THIS_MODULE,
3098 .open = proto_seq_open,
3099 .read = seq_read,
3100 .llseek = seq_lseek,
14e943db
ED		 3101 .release = seq_release_net,
3102};
3103
3104static __net_init int proto_init_net(struct net *net)
3105{
d4beaa66 3106 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 3107 return -ENOMEM;
3108
3109 return 0;
3110}
3111
3112static __net_exit void proto_exit_net(struct net *net)
3113{
ece31ffd 3114 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 3115}
3116
3117
3118static __net_initdata struct pernet_operations proto_net_ops = {
3119 .init = proto_init_net,
3120 .exit = proto_exit_net,
1da177e4
LT		 3121};
3122
3123static int __init proto_init(void)
3124{
14e943db 3125 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 3126}
3127
3128subsys_initcall(proto_init);
3129
3130#endif /* PROC_FS */
Ubuntu Artful kernel mirror