projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'linux-4.6' of git://github.com/skeggsb/linux into drm-fixes
[deliverable/linux.git] / net / core / sock.c
CommitLineData
1da177e4
LT		 1/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Generic socket support routines. Memory allocators, socket lock/release
7 * handler for protocols to use and generic option handler.
8 *
9 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT		 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
14 *
15 * Fixes:
16 * Alan Cox : Numerous verify_area() problems
17 * Alan Cox : Connecting on a connecting socket
18 * now returns an error for tcp.
19 * Alan Cox : sock->protocol is set correctly.
20 * and is not sometimes left as 0.
21 * Alan Cox : connect handles icmp errors on a
22 * connect properly. Unfortunately there
23 * is a restart syscall nasty there. I
24 * can't match BSD without hacking the C
25 * library. Ideas urgently sought!
26 * Alan Cox : Disallow bind() to addresses that are
27 * not ours - especially broadcast ones!!
28 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost)
29 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets,
30 * instead they leave that for the DESTROY timer.
31 * Alan Cox : Clean up error flag in accept
32 * Alan Cox : TCP ack handling is buggy, the DESTROY timer
33 * was buggy. Put a remove_sock() in the handler
34 * for memory when we hit 0. Also altered the timer
4ec93edb 35 * code. The ACK stuff can wait and needs major
1da177e4
LT		 36 * TCP layer surgery.
37 * Alan Cox : Fixed TCP ack bug, removed remove sock
38 * and fixed timer/inet_bh race.
39 * Alan Cox : Added zapped flag for TCP
40 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code
41 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb
42 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources
43 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing.
44 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so...
45 * Rick Sladkey : Relaxed UDP rules for matching packets.
46 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support
47 * Pauline Middelink : identd support
48 * Alan Cox : Fixed connect() taking signals I think.
49 * Alan Cox : SO_LINGER supported
50 * Alan Cox : Error reporting fixes
51 * Anonymous : inet_create tidied up (sk->reuse setting)
52 * Alan Cox : inet sockets don't set sk->type!
53 * Alan Cox : Split socket option code
54 * Alan Cox : Callbacks
55 * Alan Cox : Nagle flag for Charles & Johannes stuff
56 * Alex : Removed restriction on inet fioctl
57 * Alan Cox : Splitting INET from NET core
58 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt()
59 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code
60 * Alan Cox : Split IP from generic code
61 * Alan Cox : New kfree_skbmem()
62 * Alan Cox : Make SO_DEBUG superuser only.
63 * Alan Cox : Allow anyone to clear SO_DEBUG
64 * (compatibility fix)
65 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput.
66 * Alan Cox : Allocator for a socket is settable.
67 * Alan Cox : SO_ERROR includes soft errors.
68 * Alan Cox : Allow NULL arguments on some SO_ opts
69 * Alan Cox : Generic socket allocation to make hooks
70 * easier (suggested by Craig Metz).
71 * Michael Pall : SO_ERROR returns positive errno again
72 * Steve Whitehouse: Added default destructor to free
73 * protocol private data.
74 * Steve Whitehouse: Added various other default routines
75 * common to several socket families.
76 * Chris Evans : Call suser() check last on F_SETOWN
77 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER.
78 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s()
79 * Andi Kleen : Fix write_space callback
80 * Chris Evans : Security fixes - signedness again
81 * Arnaldo C. Melo : cleanups, use skb_queue_purge
82 *
83 * To Fix:
84 *
85 *
86 * This program is free software; you can redistribute it and/or
87 * modify it under the terms of the GNU General Public License
88 * as published by the Free Software Foundation; either version
89 * 2 of the License, or (at your option) any later version.
90 */
91
e005d193
JP		 92#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
93
4fc268d2 94#include <linux/capability.h>
1da177e4 95#include <linux/errno.h>
cb820f8e 96#include <linux/errqueue.h>
1da177e4
LT		 97#include <linux/types.h>
98#include <linux/socket.h>
99#include <linux/in.h>
100#include <linux/kernel.h>
1da177e4
LT		 101#include <linux/module.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
104#include <linux/sched.h>
105#include <linux/timer.h>
106#include <linux/string.h>
107#include <linux/sockios.h>
108#include <linux/net.h>
109#include <linux/mm.h>
110#include <linux/slab.h>
111#include <linux/interrupt.h>
112#include <linux/poll.h>
113#include <linux/tcp.h>
114#include <linux/init.h>
a1f8e7f7 115#include <linux/highmem.h>
3f551f94 116#include <linux/user_namespace.h>
c5905afb 117#include <linux/static_key.h>
3969eb38 118#include <linux/memcontrol.h>
8c1ae10d 119#include <linux/prefetch.h>
1da177e4
LT		 120
121#include <asm/uaccess.h>
1da177e4
LT		 122
123#include <linux/netdevice.h>
124#include <net/protocol.h>
125#include <linux/skbuff.h>
457c4cbc 126#include <net/net_namespace.h>
2e6599cb 127#include <net/request_sock.h>
1da177e4 128#include <net/sock.h>
20d49473 129#include <linux/net_tstamp.h>
1da177e4
LT		 130#include <net/xfrm.h>
131#include <linux/ipsec.h>
f8451725 132#include <net/cls_cgroup.h>
5bc1421e 133#include <net/netprio_cgroup.h>
eb4cb008 134#include <linux/sock_diag.h>
1da177e4
LT		 135
136#include <linux/filter.h>
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" ,
456db6a4 224 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 225};
36cbd3dc 226static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 227 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
228 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
229 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
230 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
231 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
232 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
233 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 234 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 235 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 236 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 237 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 238 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 239 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 240 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 241};
36cbd3dc 242static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 243 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
244 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
245 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
246 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
247 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
248 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
249 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 250 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 251 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 252 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 253 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 254 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 255 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 256 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 257};
da21f24d
IM		 258
259/*
260 * sk_callback_lock locking rules are per-address-family,
261 * so split the lock classes by using a per-AF key:
262 */
263static struct lock_class_key af_callback_keys[AF_MAX];
264
1da177e4
LT		 265/* Take into consideration the size of the struct sk_buff overhead in the
266 * determination of these values, since that is non-constant across
267 * platforms. This makes socket queueing behavior and performance
268 * not depend upon such differences.
269 */
270#define _SK_MEM_PACKETS 256
87fb4b7b 271#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 272#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
273#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
274
275/* Run time adjustable parameters. */
ab32ea5d 276__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 277EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 278__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 279EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 280__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
281__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 282
25985edc 283/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 284int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 285EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 286
b245be1f
WB		 287int sysctl_tstamp_allow_data __read_mostly = 1;
288
c93bdd0e
MG		 289struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
290EXPORT_SYMBOL_GPL(memalloc_socks);
291
7cb02404
MG		 292/**
293 * sk_set_memalloc - sets %SOCK_MEMALLOC
294 * @sk: socket to set it on
295 *
296 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
297 * It's the responsibility of the admin to adjust min_free_kbytes
298 * to meet the requirements
299 */
300void sk_set_memalloc(struct sock *sk)
301{
302 sock_set_flag(sk, SOCK_MEMALLOC);
303 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 304 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 305}
306EXPORT_SYMBOL_GPL(sk_set_memalloc);
307
308void sk_clear_memalloc(struct sock *sk)
309{
310 sock_reset_flag(sk, SOCK_MEMALLOC);
311 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 312 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 313
314 /*
315 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
5d753610
MG		 316 * progress of swapping. SOCK_MEMALLOC may be cleared while
317 * it has rmem allocations due to the last swapfile being deactivated
318 * but there is a risk that the socket is unusable due to exceeding
319 * the rmem limits. Reclaim the reserves and obey rmem limits again.
c76562b6 320 */
5d753610 321 sk_mem_reclaim(sk);
7cb02404
MG		 322}
323EXPORT_SYMBOL_GPL(sk_clear_memalloc);
324
b4b9e355
MG		 325int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
326{
327 int ret;
328 unsigned long pflags = current->flags;
329
330 /* these should have been dropped before queueing */
331 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
332
333 current->flags |= PF_MEMALLOC;
334 ret = sk->sk_backlog_rcv(sk, skb);
335 tsk_restore_flags(current, pflags, PF_MEMALLOC);
336
337 return ret;
338}
339EXPORT_SYMBOL(__sk_backlog_rcv);
340
1da177e4
LT		 341static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
342{
343 struct timeval tv;
344
345 if (optlen < sizeof(tv))
346 return -EINVAL;
347 if (copy_from_user(&tv, optval, sizeof(tv)))
348 return -EFAULT;
ba78073e
VA		 349 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
350 return -EDOM;
1da177e4 351
ba78073e 352 if (tv.tv_sec < 0) {
6f11df83
AM		 353 static int warned __read_mostly;
354
ba78073e 355 *timeo_p = 0;
50aab54f 356 if (warned < 10 && net_ratelimit()) {
ba78073e 357 warned++;
e005d193
JP		 358 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
359 __func__, current->comm, task_pid_nr(current));
50aab54f 360 }
ba78073e
VA		 361 return 0;
362 }
1da177e4
LT		 363 *timeo_p = MAX_SCHEDULE_TIMEOUT;
364 if (tv.tv_sec == 0 && tv.tv_usec == 0)
365 return 0;
366 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
367 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
368 return 0;
369}
370
371static void sock_warn_obsolete_bsdism(const char *name)
372{
373 static int warned;
374 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 375 if (strcmp(warncomm, current->comm) && warned < 5) {
376 strcpy(warncomm, current->comm);
e005d193
JP		 377 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
378 warncomm, name);
1da177e4
LT		 379 warned++;
380 }
381}
382
080a270f
HFS		 383static bool sock_needs_netstamp(const struct sock *sk)
384{
385 switch (sk->sk_family) {
386 case AF_UNSPEC:
387 case AF_UNIX:
388 return false;
389 default:
390 return true;
391 }
392}
393
08e29af3 394static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 395{
08e29af3
ED		 396 if (sk->sk_flags & flags) {
397 sk->sk_flags &= ~flags;
080a270f
HFS		 398 if (sock_needs_netstamp(sk) &&
399 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 400 net_disable_timestamp();
1da177e4
LT		 401 }
402}
403
404
f0088a50
DV		 405int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
406{
766e9037 407 int err;
3b885787
NH		 408 unsigned long flags;
409 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 410
0fd7bac6 411 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 412 atomic_inc(&sk->sk_drops);
3847ce32 413 trace_sock_rcvqueue_full(sk, skb);
766e9037 414 return -ENOMEM;
f0088a50
DV		 415 }
416
fda9ef5d 417 err = sk_filter(sk, skb);
f0088a50 418 if (err)
766e9037 419 return err;
f0088a50 420
c76562b6 421 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 422 atomic_inc(&sk->sk_drops);
423 return -ENOBUFS;
3ab224be
HA		 424 }
425
f0088a50
DV		 426 skb->dev = NULL;
427 skb_set_owner_r(skb, sk);
49ad9599 428
7fee226a
ED		 429 /* we escape from rcu protected region, make sure we dont leak
430 * a norefcounted dst
431 */
432 skb_dst_force(skb);
433
3b885787 434 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 435 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 436 __skb_queue_tail(list, skb);
437 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 438
439 if (!sock_flag(sk, SOCK_DEAD))
676d2369 440 sk->sk_data_ready(sk);
766e9037 441 return 0;
f0088a50
DV		 442}
443EXPORT_SYMBOL(sock_queue_rcv_skb);
444
58a5a7b9 445int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 446{
447 int rc = NET_RX_SUCCESS;
448
fda9ef5d 449 if (sk_filter(sk, skb))
f0088a50
DV		 450 goto discard_and_relse;
451
452 skb->dev = NULL;
453
274f482d 454 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 455 atomic_inc(&sk->sk_drops);
456 goto discard_and_relse;
457 }
58a5a7b9
ACM		 458 if (nested)
459 bh_lock_sock_nested(sk);
460 else
461 bh_lock_sock(sk);
a5b5bb9a
IM		 462 if (!sock_owned_by_user(sk)) {
463 /*
464 * trylock + unlock semantics:
465 */
466 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
467
c57943a1 468 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 469
470 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 471 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 472 bh_unlock_sock(sk);
473 atomic_inc(&sk->sk_drops);
474 goto discard_and_relse;
475 }
476
f0088a50
DV		 477 bh_unlock_sock(sk);
478out:
479 sock_put(sk);
480 return rc;
481discard_and_relse:
482 kfree_skb(skb);
483 goto out;
484}
485EXPORT_SYMBOL(sk_receive_skb);
486
487struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
488{
b6c6712a 489 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 490
491 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 492 sk_tx_queue_clear(sk);
a9b3cd7f 493 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 494 dst_release(dst);
495 return NULL;
496 }
497
498 return dst;
499}
500EXPORT_SYMBOL(__sk_dst_check);
501
502struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
503{
504 struct dst_entry *dst = sk_dst_get(sk);
505
506 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
507 sk_dst_reset(sk);
508 dst_release(dst);
509 return NULL;
510 }
511
512 return dst;
513}
514EXPORT_SYMBOL(sk_dst_check);
515
c91f6df2
BH		 516static int sock_setbindtodevice(struct sock *sk, char __user *optval,
517 int optlen)
4878809f
DM		 518{
519 int ret = -ENOPROTOOPT;
520#ifdef CONFIG_NETDEVICES
3b1e0a65 521 struct net *net = sock_net(sk);
4878809f
DM		 522 char devname[IFNAMSIZ];
523 int index;
524
525 /* Sorry... */
526 ret = -EPERM;
5e1fccc0 527 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 528 goto out;
529
530 ret = -EINVAL;
531 if (optlen < 0)
532 goto out;
533
534 /* Bind this socket to a particular device like "eth0",
535 * as specified in the passed interface name. If the
536 * name is "" or the option length is zero the socket
537 * is not bound.
538 */
539 if (optlen > IFNAMSIZ - 1)
540 optlen = IFNAMSIZ - 1;
541 memset(devname, 0, sizeof(devname));
542
543 ret = -EFAULT;
544 if (copy_from_user(devname, optval, optlen))
545 goto out;
546
000ba2e4
DM		 547 index = 0;
548 if (devname[0] != '\0') {
bf8e56bf 549 struct net_device *dev;
4878809f 550
bf8e56bf
ED		 551 rcu_read_lock();
552 dev = dev_get_by_name_rcu(net, devname);
553 if (dev)
554 index = dev->ifindex;
555 rcu_read_unlock();
4878809f
DM		 556 ret = -ENODEV;
557 if (!dev)
558 goto out;
4878809f
DM		 559 }
560
561 lock_sock(sk);
562 sk->sk_bound_dev_if = index;
563 sk_dst_reset(sk);
564 release_sock(sk);
565
566 ret = 0;
567
568out:
569#endif
570
571 return ret;
572}
573
c91f6df2
BH		 574static int sock_getbindtodevice(struct sock *sk, char __user *optval,
575 int __user *optlen, int len)
576{
577 int ret = -ENOPROTOOPT;
578#ifdef CONFIG_NETDEVICES
579 struct net *net = sock_net(sk);
c91f6df2 580 char devname[IFNAMSIZ];
c91f6df2
BH		 581
582 if (sk->sk_bound_dev_if == 0) {
583 len = 0;
584 goto zero;
585 }
586
587 ret = -EINVAL;
588 if (len < IFNAMSIZ)
589 goto out;
590
5dbe7c17
NS		 591 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
592 if (ret)
c91f6df2 593 goto out;
c91f6df2
BH		 594
595 len = strlen(devname) + 1;
596
597 ret = -EFAULT;
598 if (copy_to_user(optval, devname, len))
599 goto out;
600
601zero:
602 ret = -EFAULT;
603 if (put_user(len, optlen))
604 goto out;
605
606 ret = 0;
607
608out:
609#endif
610
611 return ret;
612}
613
c0ef877b
PE		 614static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
615{
616 if (valbool)
617 sock_set_flag(sk, bit);
618 else
619 sock_reset_flag(sk, bit);
620}
621
f60e5990 622bool sk_mc_loop(struct sock *sk)
623{
624 if (dev_recursion_level())
625 return false;
626 if (!sk)
627 return true;
628 switch (sk->sk_family) {
629 case AF_INET:
630 return inet_sk(sk)->mc_loop;
631#if IS_ENABLED(CONFIG_IPV6)
632 case AF_INET6:
633 return inet6_sk(sk)->mc_loop;
634#endif
635 }
636 WARN_ON(1);
637 return true;
638}
639EXPORT_SYMBOL(sk_mc_loop);
640
1da177e4
LT		 641/*
642 * This is meant for all protocols to use and covers goings on
643 * at the socket level. Everything here is generic.
644 */
645
646int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 647 char __user *optval, unsigned int optlen)
1da177e4 648{
2a91525c 649 struct sock *sk = sock->sk;
1da177e4
LT		 650 int val;
651 int valbool;
652 struct linger ling;
653 int ret = 0;
4ec93edb 654
1da177e4
LT		 655 /*
656 * Options without arguments
657 */
658
4878809f 659 if (optname == SO_BINDTODEVICE)
c91f6df2 660 return sock_setbindtodevice(sk, optval, optlen);
4878809f 661
e71a4783
SH		 662 if (optlen < sizeof(int))
663 return -EINVAL;
4ec93edb 664
1da177e4
LT		 665 if (get_user(val, (int __user *)optval))
666 return -EFAULT;
4ec93edb 667
2a91525c 668 valbool = val ? 1 : 0;
1da177e4
LT		 669
670 lock_sock(sk);
671
2a91525c 672 switch (optname) {
e71a4783 673 case SO_DEBUG:
2a91525c 674 if (val && !capable(CAP_NET_ADMIN))
e71a4783 675 ret = -EACCES;
2a91525c 676 else
c0ef877b 677 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 678 break;
679 case SO_REUSEADDR:
4a17fd52 680 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 681 break;
055dc21a
TH		 682 case SO_REUSEPORT:
683 sk->sk_reuseport = valbool;
684 break;
e71a4783 685 case SO_TYPE:
49c794e9 686 case SO_PROTOCOL:
0d6038ee 687 case SO_DOMAIN:
e71a4783
SH		 688 case SO_ERROR:
689 ret = -ENOPROTOOPT;
690 break;
691 case SO_DONTROUTE:
c0ef877b 692 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 693 break;
694 case SO_BROADCAST:
695 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
696 break;
697 case SO_SNDBUF:
698 /* Don't error on this BSD doesn't and if you think
82981930
ED		 699 * about it this is right. Otherwise apps have to
700 * play 'guess the biggest size' games. RCVBUF/SNDBUF
701 * are treated in BSD as hints
702 */
703 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 704set_sndbuf:
e71a4783 705 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 706 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
707 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 708 sk->sk_write_space(sk);
709 break;
1da177e4 710
e71a4783
SH		 711 case SO_SNDBUFFORCE:
712 if (!capable(CAP_NET_ADMIN)) {
713 ret = -EPERM;
714 break;
715 }
716 goto set_sndbuf;
b0573dea 717
e71a4783
SH		 718 case SO_RCVBUF:
719 /* Don't error on this BSD doesn't and if you think
82981930
ED		 720 * about it this is right. Otherwise apps have to
721 * play 'guess the biggest size' games. RCVBUF/SNDBUF
722 * are treated in BSD as hints
723 */
724 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 725set_rcvbuf:
e71a4783
SH		 726 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
727 /*
728 * We double it on the way in to account for
729 * "struct sk_buff" etc. overhead. Applications
730 * assume that the SO_RCVBUF setting they make will
731 * allow that much actual data to be received on that
732 * socket.
733 *
734 * Applications are unaware that "struct sk_buff" and
735 * other overheads allocate from the receive buffer
736 * during socket buffer allocation.
737 *
738 * And after considering the possible alternatives,
739 * returning the value we actually used in getsockopt
740 * is the most desirable behavior.
741 */
82981930 742 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 743 break;
744
745 case SO_RCVBUFFORCE:
746 if (!capable(CAP_NET_ADMIN)) {
747 ret = -EPERM;
1da177e4 748 break;
e71a4783
SH		 749 }
750 goto set_rcvbuf;
1da177e4 751
e71a4783 752 case SO_KEEPALIVE:
1da177e4 753#ifdef CONFIG_INET
3e10986d
ED		 754 if (sk->sk_protocol == IPPROTO_TCP &&
755 sk->sk_type == SOCK_STREAM)
e71a4783 756 tcp_set_keepalive(sk, valbool);
1da177e4 757#endif
e71a4783
SH		 758 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
759 break;
760
761 case SO_OOBINLINE:
762 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
763 break;
764
765 case SO_NO_CHECK:
28448b80 766 sk->sk_no_check_tx = valbool;
e71a4783
SH		 767 break;
768
769 case SO_PRIORITY:
5e1fccc0
EB		 770 if ((val >= 0 && val <= 6) ||
771 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 772 sk->sk_priority = val;
773 else
774 ret = -EPERM;
775 break;
776
777 case SO_LINGER:
778 if (optlen < sizeof(ling)) {
779 ret = -EINVAL; /* 1003.1g */
1da177e4 780 break;
e71a4783 781 }
2a91525c 782 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 783 ret = -EFAULT;
1da177e4 784 break;
e71a4783
SH		 785 }
786 if (!ling.l_onoff)
787 sock_reset_flag(sk, SOCK_LINGER);
788 else {
1da177e4 789#if (BITS_PER_LONG == 32)
e71a4783
SH		 790 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
791 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 792 else
e71a4783
SH		 793#endif
794 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
795 sock_set_flag(sk, SOCK_LINGER);
796 }
797 break;
798
799 case SO_BSDCOMPAT:
800 sock_warn_obsolete_bsdism("setsockopt");
801 break;
802
803 case SO_PASSCRED:
804 if (valbool)
805 set_bit(SOCK_PASSCRED, &sock->flags);
806 else
807 clear_bit(SOCK_PASSCRED, &sock->flags);
808 break;
809
810 case SO_TIMESTAMP:
92f37fd2 811 case SO_TIMESTAMPNS:
e71a4783 812 if (valbool) {
92f37fd2
ED		 813 if (optname == SO_TIMESTAMP)
814 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
815 else
816 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 817 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 818 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 819 } else {
e71a4783 820 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 821 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
822 }
e71a4783
SH		 823 break;
824
20d49473
PO		 825 case SO_TIMESTAMPING:
826 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 827 ret = -EINVAL;
20d49473
PO		 828 break;
829 }
b245be1f 830
09c2d251 831 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765 832 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
ac5cc977
WC		 833 if (sk->sk_protocol == IPPROTO_TCP &&
834 sk->sk_type == SOCK_STREAM) {
4ed2d765
WB		 835 if (sk->sk_state != TCP_ESTABLISHED) {
836 ret = -EINVAL;
837 break;
838 }
839 sk->sk_tskey = tcp_sk(sk)->snd_una;
840 } else {
841 sk->sk_tskey = 0;
842 }
843 }
b9f40e21 844 sk->sk_tsflags = val;
20d49473
PO		 845 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
846 sock_enable_timestamp(sk,
847 SOCK_TIMESTAMPING_RX_SOFTWARE);
848 else
849 sock_disable_timestamp(sk,
08e29af3 850 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 851 break;
852
e71a4783
SH		 853 case SO_RCVLOWAT:
854 if (val < 0)
855 val = INT_MAX;
856 sk->sk_rcvlowat = val ? : 1;
857 break;
858
859 case SO_RCVTIMEO:
860 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
861 break;
862
863 case SO_SNDTIMEO:
864 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
865 break;
1da177e4 866
e71a4783
SH		 867 case SO_ATTACH_FILTER:
868 ret = -EINVAL;
869 if (optlen == sizeof(struct sock_fprog)) {
870 struct sock_fprog fprog;
1da177e4 871
e71a4783
SH		 872 ret = -EFAULT;
873 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 874 break;
e71a4783
SH		 875
876 ret = sk_attach_filter(&fprog, sk);
877 }
878 break;
879
89aa0758
AS		 880 case SO_ATTACH_BPF:
881 ret = -EINVAL;
882 if (optlen == sizeof(u32)) {
883 u32 ufd;
884
885 ret = -EFAULT;
886 if (copy_from_user(&ufd, optval, sizeof(ufd)))
887 break;
888
889 ret = sk_attach_bpf(ufd, sk);
890 }
891 break;
892
538950a1
CG		 893 case SO_ATTACH_REUSEPORT_CBPF:
894 ret = -EINVAL;
895 if (optlen == sizeof(struct sock_fprog)) {
896 struct sock_fprog fprog;
897
898 ret = -EFAULT;
899 if (copy_from_user(&fprog, optval, sizeof(fprog)))
900 break;
901
902 ret = sk_reuseport_attach_filter(&fprog, sk);
903 }
904 break;
905
906 case SO_ATTACH_REUSEPORT_EBPF:
907 ret = -EINVAL;
908 if (optlen == sizeof(u32)) {
909 u32 ufd;
910
911 ret = -EFAULT;
912 if (copy_from_user(&ufd, optval, sizeof(ufd)))
913 break;
914
915 ret = sk_reuseport_attach_bpf(ufd, sk);
916 }
917 break;
918
e71a4783 919 case SO_DETACH_FILTER:
55b33325 920 ret = sk_detach_filter(sk);
e71a4783 921 break;
1da177e4 922
d59577b6
VB		 923 case SO_LOCK_FILTER:
924 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
925 ret = -EPERM;
926 else
927 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
928 break;
929
e71a4783
SH		 930 case SO_PASSSEC:
931 if (valbool)
932 set_bit(SOCK_PASSSEC, &sock->flags);
933 else
934 clear_bit(SOCK_PASSSEC, &sock->flags);
935 break;
4a19ec58 936 case SO_MARK:
5e1fccc0 937 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 938 ret = -EPERM;
2a91525c 939 else
4a19ec58 940 sk->sk_mark = val;
4a19ec58 941 break;
877ce7c1 942
3b885787 943 case SO_RXQ_OVFL:
8083f0fc 944 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 945 break;
6e3e939f
JB		 946
947 case SO_WIFI_STATUS:
948 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
949 break;
950
ef64a54f
PE		 951 case SO_PEEK_OFF:
952 if (sock->ops->set_peek_off)
12663bfc 953 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 954 else
955 ret = -EOPNOTSUPP;
956 break;
3bdc0eba
BG		 957
958 case SO_NOFCS:
959 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
960 break;
961
7d4c04fc
KJ		 962 case SO_SELECT_ERR_QUEUE:
963 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
964 break;
965
e0d1095a 966#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 967 case SO_BUSY_POLL:
dafcc438
ET		 968 /* allow unprivileged users to decrease the value */
969 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
970 ret = -EPERM;
971 else {
972 if (val < 0)
973 ret = -EINVAL;
974 else
975 sk->sk_ll_usec = val;
976 }
977 break;
978#endif
62748f32
ED		 979
980 case SO_MAX_PACING_RATE:
981 sk->sk_max_pacing_rate = val;
982 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
983 sk->sk_max_pacing_rate);
984 break;
985
70da268b
ED		 986 case SO_INCOMING_CPU:
987 sk->sk_incoming_cpu = val;
988 break;
989
a87cb3e4
TH		 990 case SO_CNX_ADVICE:
991 if (val == 1)
992 dst_negative_advice(sk);
993 break;
e71a4783
SH		 994 default:
995 ret = -ENOPROTOOPT;
996 break;
4ec93edb 997 }
1da177e4
LT		 998 release_sock(sk);
999 return ret;
1000}
2a91525c 1001EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1002
1003
8f09898b 1004static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1005 struct ucred *ucred)
3f551f94
EB		 1006{
1007 ucred->pid = pid_vnr(pid);
1008 ucred->uid = ucred->gid = -1;
1009 if (cred) {
1010 struct user_namespace *current_ns = current_user_ns();
1011
b2e4f544
EB		 1012 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1013 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1014 }
1015}
1016
1da177e4
LT		 1017int sock_getsockopt(struct socket *sock, int level, int optname,
1018 char __user *optval, int __user *optlen)
1019{
1020 struct sock *sk = sock->sk;
4ec93edb 1021
e71a4783 1022 union {
4ec93edb
YH		 1023 int val;
1024 struct linger ling;
1da177e4
LT		 1025 struct timeval tm;
1026 } v;
4ec93edb 1027
4d0392be 1028 int lv = sizeof(int);
1da177e4 1029 int len;
4ec93edb 1030
e71a4783 1031 if (get_user(len, optlen))
4ec93edb 1032 return -EFAULT;
e71a4783 1033 if (len < 0)
1da177e4 1034 return -EINVAL;
4ec93edb 1035
50fee1de 1036 memset(&v, 0, sizeof(v));
df0bca04 1037
2a91525c 1038 switch (optname) {
e71a4783
SH		 1039 case SO_DEBUG:
1040 v.val = sock_flag(sk, SOCK_DBG);
1041 break;
1042
1043 case SO_DONTROUTE:
1044 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1045 break;
1046
1047 case SO_BROADCAST:
1b23a5df 1048 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1049 break;
1050
1051 case SO_SNDBUF:
1052 v.val = sk->sk_sndbuf;
1053 break;
1054
1055 case SO_RCVBUF:
1056 v.val = sk->sk_rcvbuf;
1057 break;
1058
1059 case SO_REUSEADDR:
1060 v.val = sk->sk_reuse;
1061 break;
1062
055dc21a
TH		 1063 case SO_REUSEPORT:
1064 v.val = sk->sk_reuseport;
1065 break;
1066
e71a4783 1067 case SO_KEEPALIVE:
1b23a5df 1068 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1069 break;
1070
1071 case SO_TYPE:
1072 v.val = sk->sk_type;
1073 break;
1074
49c794e9
JE		 1075 case SO_PROTOCOL:
1076 v.val = sk->sk_protocol;
1077 break;
1078
0d6038ee
JE		 1079 case SO_DOMAIN:
1080 v.val = sk->sk_family;
1081 break;
1082
e71a4783
SH		 1083 case SO_ERROR:
1084 v.val = -sock_error(sk);
2a91525c 1085 if (v.val == 0)
e71a4783
SH		 1086 v.val = xchg(&sk->sk_err_soft, 0);
1087 break;
1088
1089 case SO_OOBINLINE:
1b23a5df 1090 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1091 break;
1092
1093 case SO_NO_CHECK:
28448b80 1094 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1095 break;
1096
1097 case SO_PRIORITY:
1098 v.val = sk->sk_priority;
1099 break;
1100
1101 case SO_LINGER:
1102 lv = sizeof(v.ling);
1b23a5df 1103 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1104 v.ling.l_linger = sk->sk_lingertime / HZ;
1105 break;
1106
1107 case SO_BSDCOMPAT:
1108 sock_warn_obsolete_bsdism("getsockopt");
1109 break;
1110
1111 case SO_TIMESTAMP:
92f37fd2
ED		 1112 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1113 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1114 break;
1115
1116 case SO_TIMESTAMPNS:
1117 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1118 break;
1119
20d49473 1120 case SO_TIMESTAMPING:
b9f40e21 1121 v.val = sk->sk_tsflags;
20d49473
PO		 1122 break;
1123
e71a4783 1124 case SO_RCVTIMEO:
2a91525c 1125 lv = sizeof(struct timeval);
e71a4783
SH		 1126 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1127 v.tm.tv_sec = 0;
1128 v.tm.tv_usec = 0;
1129 } else {
1130 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1131 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1132 }
1133 break;
1134
1135 case SO_SNDTIMEO:
2a91525c 1136 lv = sizeof(struct timeval);
e71a4783
SH		 1137 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1138 v.tm.tv_sec = 0;
1139 v.tm.tv_usec = 0;
1140 } else {
1141 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1142 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1143 }
1144 break;
1da177e4 1145
e71a4783
SH		 1146 case SO_RCVLOWAT:
1147 v.val = sk->sk_rcvlowat;
1148 break;
1da177e4 1149
e71a4783 1150 case SO_SNDLOWAT:
2a91525c 1151 v.val = 1;
e71a4783 1152 break;
1da177e4 1153
e71a4783 1154 case SO_PASSCRED:
82981930 1155 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1156 break;
1da177e4 1157
e71a4783 1158 case SO_PEERCRED:
109f6e39
EB		 1159 {
1160 struct ucred peercred;
1161 if (len > sizeof(peercred))
1162 len = sizeof(peercred);
1163 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1164 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1165 return -EFAULT;
1166 goto lenout;
109f6e39 1167 }
1da177e4 1168
e71a4783
SH		 1169 case SO_PEERNAME:
1170 {
1171 char address[128];
1172
1173 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1174 return -ENOTCONN;
1175 if (lv < len)
1176 return -EINVAL;
1177 if (copy_to_user(optval, address, len))
1178 return -EFAULT;
1179 goto lenout;
1180 }
1da177e4 1181
e71a4783
SH		 1182 /* Dubious BSD thing... Probably nobody even uses it, but
1183 * the UNIX standard wants it for whatever reason... -DaveM
1184 */
1185 case SO_ACCEPTCONN:
1186 v.val = sk->sk_state == TCP_LISTEN;
1187 break;
1da177e4 1188
e71a4783 1189 case SO_PASSSEC:
82981930 1190 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1191 break;
877ce7c1 1192
e71a4783
SH		 1193 case SO_PEERSEC:
1194 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1195
4a19ec58
LAT		 1196 case SO_MARK:
1197 v.val = sk->sk_mark;
1198 break;
1199
3b885787 1200 case SO_RXQ_OVFL:
1b23a5df 1201 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1202 break;
1203
6e3e939f 1204 case SO_WIFI_STATUS:
1b23a5df 1205 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1206 break;
1207
ef64a54f
PE		 1208 case SO_PEEK_OFF:
1209 if (!sock->ops->set_peek_off)
1210 return -EOPNOTSUPP;
1211
1212 v.val = sk->sk_peek_off;
1213 break;
bc2f7996 1214 case SO_NOFCS:
1b23a5df 1215 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1216 break;
c91f6df2 1217
f7b86bfe 1218 case SO_BINDTODEVICE:
c91f6df2
BH		 1219 return sock_getbindtodevice(sk, optval, optlen, len);
1220
a8fc9277
PE		 1221 case SO_GET_FILTER:
1222 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1223 if (len < 0)
1224 return len;
1225
1226 goto lenout;
c91f6df2 1227
d59577b6
VB		 1228 case SO_LOCK_FILTER:
1229 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1230 break;
1231
ea02f941
MS		 1232 case SO_BPF_EXTENSIONS:
1233 v.val = bpf_tell_extensions();
1234 break;
1235
7d4c04fc
KJ		 1236 case SO_SELECT_ERR_QUEUE:
1237 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1238 break;
1239
e0d1095a 1240#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1241 case SO_BUSY_POLL:
dafcc438
ET		 1242 v.val = sk->sk_ll_usec;
1243 break;
1244#endif
1245
62748f32
ED		 1246 case SO_MAX_PACING_RATE:
1247 v.val = sk->sk_max_pacing_rate;
1248 break;
1249
2c8c56e1
ED		 1250 case SO_INCOMING_CPU:
1251 v.val = sk->sk_incoming_cpu;
1252 break;
1253
e71a4783 1254 default:
443b5991
YH		 1255 /* We implement the SO_SNDLOWAT etc to not be settable
1256 * (1003.1g 7).
1257 */
e71a4783 1258 return -ENOPROTOOPT;
1da177e4 1259 }
e71a4783 1260
1da177e4
LT		 1261 if (len > lv)
1262 len = lv;
1263 if (copy_to_user(optval, &v, len))
1264 return -EFAULT;
1265lenout:
4ec93edb
YH		 1266 if (put_user(len, optlen))
1267 return -EFAULT;
1268 return 0;
1da177e4
LT		 1269}
1270
a5b5bb9a
IM		 1271/*
1272 * Initialize an sk_lock.
1273 *
1274 * (We also register the sk_lock with the lock validator.)
1275 */
b6f99a21 1276static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1277{
ed07536e
PZ		 1278 sock_lock_init_class_and_name(sk,
1279 af_family_slock_key_strings[sk->sk_family],
1280 af_family_slock_keys + sk->sk_family,
1281 af_family_key_strings[sk->sk_family],
1282 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1283}
1284
4dc6dc71
ED		 1285/*
1286 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1287 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1288 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1289 */
f1a6c4da
PE		 1290static void sock_copy(struct sock *nsk, const struct sock *osk)
1291{
1292#ifdef CONFIG_SECURITY_NETWORK
1293 void *sptr = nsk->sk_security;
1294#endif
68835aba
ED		 1295 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1296
1297 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1298 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1299
f1a6c4da
PE		 1300#ifdef CONFIG_SECURITY_NETWORK
1301 nsk->sk_security = sptr;
1302 security_sk_clone(osk, nsk);
1303#endif
1304}
1305
fcbdf09d
OP		 1306void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1307{
1308 unsigned long nulls1, nulls2;
1309
1310 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1311 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1312 if (nulls1 > nulls2)
1313 swap(nulls1, nulls2);
1314
1315 if (nulls1 != 0)
1316 memset((char *)sk, 0, nulls1);
1317 memset((char *)sk + nulls1 + sizeof(void *), 0,
1318 nulls2 - nulls1 - sizeof(void *));
1319 memset((char *)sk + nulls2 + sizeof(void *), 0,
1320 size - nulls2 - sizeof(void *));
1321}
1322EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
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;
1335 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1336 if (prot->clear_sk)
1337 prot->clear_sk(sk, prot->obj_size);
1338 else
1339 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1340 }
fcbdf09d 1341 } else
c308c1b2
PE		 1342 sk = kmalloc(prot->obj_size, priority);
1343
2e4afe7b 1344 if (sk != NULL) {
a98b65a3
VN		 1345 kmemcheck_annotate_bitfield(sk, flags);
1346
2e4afe7b
PE		 1347 if (security_sk_alloc(sk, family, priority))
1348 goto out_free;
1349
1350 if (!try_module_get(prot->owner))
1351 goto out_free_sec;
e022f0b4 1352 sk_tx_queue_clear(sk);
bd1060a1 1353 cgroup_sk_alloc(&sk->sk_cgrp_data);
2e4afe7b
PE		 1354 }
1355
c308c1b2 1356 return sk;
2e4afe7b
PE		 1357
1358out_free_sec:
1359 security_sk_free(sk);
1360out_free:
1361 if (slab != NULL)
1362 kmem_cache_free(slab, sk);
1363 else
1364 kfree(sk);
1365 return NULL;
c308c1b2
PE		 1366}
1367
1368static void sk_prot_free(struct proto *prot, struct sock *sk)
1369{
1370 struct kmem_cache *slab;
2e4afe7b 1371 struct module *owner;
c308c1b2 1372
2e4afe7b 1373 owner = prot->owner;
c308c1b2 1374 slab = prot->slab;
2e4afe7b 1375
bd1060a1 1376 cgroup_sk_free(&sk->sk_cgrp_data);
2e4afe7b 1377 security_sk_free(sk);
c308c1b2
PE		 1378 if (slab != NULL)
1379 kmem_cache_free(slab, sk);
1380 else
1381 kfree(sk);
2e4afe7b 1382 module_put(owner);
c308c1b2
PE		 1383}
1384
1da177e4
LT		 1385/**
1386 * sk_alloc - All socket objects are allocated here
c4ea43c5 1387 * @net: the applicable net namespace
4dc3b16b
PP		 1388 * @family: protocol family
1389 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1390 * @prot: struct proto associated with this new sock instance
11aa9c28 1391 * @kern: is this to be a kernel socket?
1da177e4 1392 */
1b8d7ae4 1393struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1394 struct proto *prot, int kern)
1da177e4 1395{
c308c1b2 1396 struct sock *sk;
1da177e4 1397
154adbc8 1398 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1399 if (sk) {
154adbc8
PE		 1400 sk->sk_family = family;
1401 /*
1402 * See comment in struct sock definition to understand
1403 * why we need sk_prot_creator -acme
1404 */
1405 sk->sk_prot = sk->sk_prot_creator = prot;
1406 sock_lock_init(sk);
26abe143
EB		 1407 sk->sk_net_refcnt = kern ? 0 : 1;
1408 if (likely(sk->sk_net_refcnt))
1409 get_net(net);
1410 sock_net_set(sk, net);
d66ee058 1411 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1412
2a56a1fe
TH		 1413 sock_update_classid(&sk->sk_cgrp_data);
1414 sock_update_netprioidx(&sk->sk_cgrp_data);
1da177e4 1415 }
a79af59e 1416
2e4afe7b 1417 return sk;
1da177e4 1418}
2a91525c 1419EXPORT_SYMBOL(sk_alloc);
1da177e4 1420
eb4cb008 1421void sk_destruct(struct sock *sk)
1da177e4
LT		 1422{
1423 struct sk_filter *filter;
1da177e4
LT		 1424
1425 if (sk->sk_destruct)
1426 sk->sk_destruct(sk);
1427
a898def2
PM		 1428 filter = rcu_dereference_check(sk->sk_filter,
1429 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1430 if (filter) {
309dd5fc 1431 sk_filter_uncharge(sk, filter);
a9b3cd7f 1432 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4 1433 }
538950a1
CG		 1434 if (rcu_access_pointer(sk->sk_reuseport_cb))
1435 reuseport_detach_sock(sk);
1da177e4 1436
08e29af3 1437 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1438
1439 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1440 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1441 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1442
109f6e39
EB		 1443 if (sk->sk_peer_cred)
1444 put_cred(sk->sk_peer_cred);
1445 put_pid(sk->sk_peer_pid);
26abe143
EB		 1446 if (likely(sk->sk_net_refcnt))
1447 put_net(sock_net(sk));
c308c1b2 1448 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1449}
2b85a34e 1450
eb4cb008
CG		 1451static void __sk_free(struct sock *sk)
1452{
b922622e 1453 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1454 sock_diag_broadcast_destroy(sk);
1455 else
1456 sk_destruct(sk);
1457}
1458
2b85a34e
ED		 1459void sk_free(struct sock *sk)
1460{
1461 /*
25985edc 1462 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1463 * some packets are still in some tx queue.
1464 * If not null, sock_wfree() will call __sk_free(sk) later
1465 */
1466 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1467 __sk_free(sk);
1468}
2a91525c 1469EXPORT_SYMBOL(sk_free);
1da177e4 1470
e56c57d0
ED		 1471/**
1472 * sk_clone_lock - clone a socket, and lock its clone
1473 * @sk: the socket to clone
1474 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1475 *
1476 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1477 */
1478struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1479{
8fd1d178 1480 struct sock *newsk;
278571ba 1481 bool is_charged = true;
87d11ceb 1482
8fd1d178 1483 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1484 if (newsk != NULL) {
1485 struct sk_filter *filter;
1486
892c141e 1487 sock_copy(newsk, sk);
87d11ceb
ACM		 1488
1489 /* SANITY */
8a681736
SV		 1490 if (likely(newsk->sk_net_refcnt))
1491 get_net(sock_net(newsk));
87d11ceb
ACM		 1492 sk_node_init(&newsk->sk_node);
1493 sock_lock_init(newsk);
1494 bh_lock_sock(newsk);
fa438ccf 1495 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1496 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1497
1498 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1499 /*
1500 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1501 */
1502 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1503 atomic_set(&newsk->sk_omem_alloc, 0);
1504 skb_queue_head_init(&newsk->sk_receive_queue);
1505 skb_queue_head_init(&newsk->sk_write_queue);
1506
87d11ceb 1507 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1508 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1509 af_callback_keys + newsk->sk_family,
1510 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1511
1512 newsk->sk_dst_cache = NULL;
1513 newsk->sk_wmem_queued = 0;
1514 newsk->sk_forward_alloc = 0;
1515 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1516 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1517
1518 sock_reset_flag(newsk, SOCK_DONE);
1519 skb_queue_head_init(&newsk->sk_error_queue);
1520
0d7da9dd 1521 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1522 if (filter != NULL)
278571ba
AS		 1523 /* though it's an empty new sock, the charging may fail
1524 * if sysctl_optmem_max was changed between creation of
1525 * original socket and cloning
1526 */
1527 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1528
d188ba86 1529 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
87d11ceb
ACM		 1530 /* It is still raw copy of parent, so invalidate
1531 * destructor and make plain sk_free() */
1532 newsk->sk_destruct = NULL;
b0691c8e 1533 bh_unlock_sock(newsk);
87d11ceb
ACM		 1534 sk_free(newsk);
1535 newsk = NULL;
1536 goto out;
1537 }
fa463497 1538 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
87d11ceb
ACM		 1539
1540 newsk->sk_err = 0;
1541 newsk->sk_priority = 0;
2c8c56e1 1542 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1543 atomic64_set(&newsk->sk_cookie, 0);
4dc6dc71
ED		 1544 /*
1545 * Before updating sk_refcnt, we must commit prior changes to memory
1546 * (Documentation/RCU/rculist_nulls.txt for details)
1547 */
1548 smp_wmb();
87d11ceb
ACM		 1549 atomic_set(&newsk->sk_refcnt, 2);
1550
1551 /*
1552 * Increment the counter in the same struct proto as the master
1553 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1554 * is the same as sk->sk_prot->socks, as this field was copied
1555 * with memcpy).
1556 *
1557 * This _changes_ the previous behaviour, where
1558 * tcp_create_openreq_child always was incrementing the
1559 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1560 * to be taken into account in all callers. -acme
1561 */
1562 sk_refcnt_debug_inc(newsk);
972692e0 1563 sk_set_socket(newsk, NULL);
43815482 1564 newsk->sk_wq = NULL;
87d11ceb 1565
baac50bb 1566 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
3d596f7b 1567 sock_update_memcg(newsk);
f3f511e1 1568
87d11ceb 1569 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1570 sk_sockets_allocated_inc(newsk);
704da560 1571
080a270f
HFS		 1572 if (sock_needs_netstamp(sk) &&
1573 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1574 net_enable_timestamp();
87d11ceb
ACM		 1575 }
1576out:
1577 return newsk;
1578}
e56c57d0 1579EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1580
9958089a
AK		 1581void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1582{
d6a4e26a
ED		 1583 u32 max_segs = 1;
1584
6bd4f355 1585 sk_dst_set(sk, dst);
9958089a
AK		 1586 sk->sk_route_caps = dst->dev->features;
1587 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1588 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1589 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1590 if (sk_can_gso(sk)) {
82cc1a7a 1591 if (dst->header_len) {
9958089a 1592 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1593 } else {
9958089a 1594 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1595 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1596 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1597 }
9958089a 1598 }
d6a4e26a 1599 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1600}
1601EXPORT_SYMBOL_GPL(sk_setup_caps);
1602
1da177e4
LT		 1603/*
1604 * Simple resource managers for sockets.
1605 */
1606
1607
4ec93edb
YH		 1608/*
1609 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1610 */
1611void sock_wfree(struct sk_buff *skb)
1612{
1613 struct sock *sk = skb->sk;
d99927f4 1614 unsigned int len = skb->truesize;
1da177e4 1615
d99927f4
ED		 1616 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1617 /*
1618 * Keep a reference on sk_wmem_alloc, this will be released
1619 * after sk_write_space() call
1620 */
1621 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1622 sk->sk_write_space(sk);
d99927f4
ED		 1623 len = 1;
1624 }
2b85a34e 1625 /*
d99927f4
ED		 1626 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1627 * could not do because of in-flight packets
2b85a34e 1628 */
d99927f4 1629 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1630 __sk_free(sk);
1da177e4 1631}
2a91525c 1632EXPORT_SYMBOL(sock_wfree);
1da177e4 1633
9e17f8a4
ED		 1634void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1635{
1636 skb_orphan(skb);
1637 skb->sk = sk;
1638#ifdef CONFIG_INET
1639 if (unlikely(!sk_fullsock(sk))) {
1640 skb->destructor = sock_edemux;
1641 sock_hold(sk);
1642 return;
1643 }
1644#endif
1645 skb->destructor = sock_wfree;
1646 skb_set_hash_from_sk(skb, sk);
1647 /*
1648 * We used to take a refcount on sk, but following operation
1649 * is enough to guarantee sk_free() wont free this sock until
1650 * all in-flight packets are completed
1651 */
1652 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1653}
1654EXPORT_SYMBOL(skb_set_owner_w);
1655
f2f872f9
ED		 1656void skb_orphan_partial(struct sk_buff *skb)
1657{
1658 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1659 * so we do not completely orphan skb, but transfert all
1660 * accounted bytes but one, to avoid unexpected reorders.
1661 */
1662 if (skb->destructor == sock_wfree
1663#ifdef CONFIG_INET
1664 || skb->destructor == tcp_wfree
1665#endif
1666 ) {
1667 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1668 skb->truesize = 1;
1669 } else {
1670 skb_orphan(skb);
1671 }
1672}
1673EXPORT_SYMBOL(skb_orphan_partial);
1674
4ec93edb
YH		 1675/*
1676 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1677 */
1678void sock_rfree(struct sk_buff *skb)
1679{
1680 struct sock *sk = skb->sk;
d361fd59 1681 unsigned int len = skb->truesize;
1da177e4 1682
d361fd59
ED		 1683 atomic_sub(len, &sk->sk_rmem_alloc);
1684 sk_mem_uncharge(sk, len);
1da177e4 1685}
2a91525c 1686EXPORT_SYMBOL(sock_rfree);
1da177e4 1687
7768eed8
OH		 1688/*
1689 * Buffer destructor for skbs that are not used directly in read or write
1690 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1691 */
62bccb8c
AD		 1692void sock_efree(struct sk_buff *skb)
1693{
1694 sock_put(skb->sk);
1695}
1696EXPORT_SYMBOL(sock_efree);
1697
976d0201 1698kuid_t sock_i_uid(struct sock *sk)
1da177e4 1699{
976d0201 1700 kuid_t uid;
1da177e4 1701
f064af1e 1702 read_lock_bh(&sk->sk_callback_lock);
976d0201 1703 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1704 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1705 return uid;
1706}
2a91525c 1707EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1708
1709unsigned long sock_i_ino(struct sock *sk)
1710{
1711 unsigned long ino;
1712
f064af1e 1713 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1714 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1715 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1716 return ino;
1717}
2a91525c 1718EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1719
1720/*
1721 * Allocate a skb from the socket's send buffer.
1722 */
86a76caf 1723struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1724 gfp_t priority)
1da177e4
LT		 1725{
1726 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1727 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1728 if (skb) {
1729 skb_set_owner_w(skb, sk);
1730 return skb;
1731 }
1732 }
1733 return NULL;
1734}
2a91525c 1735EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1736
4ec93edb 1737/*
1da177e4 1738 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1739 */
dd0fc66f 1740void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1741{
95c96174 1742 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1743 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1744 void *mem;
1745 /* First do the add, to avoid the race if kmalloc
4ec93edb 1746 * might sleep.
1da177e4
LT		 1747 */
1748 atomic_add(size, &sk->sk_omem_alloc);
1749 mem = kmalloc(size, priority);
1750 if (mem)
1751 return mem;
1752 atomic_sub(size, &sk->sk_omem_alloc);
1753 }
1754 return NULL;
1755}
2a91525c 1756EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1757
79e88659
DB		 1758/* Free an option memory block. Note, we actually want the inline
1759 * here as this allows gcc to detect the nullify and fold away the
1760 * condition entirely.
1da177e4 1761 */
79e88659
DB		 1762static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1763 const bool nullify)
1da177e4 1764{
e53da5fb
DM		 1765 if (WARN_ON_ONCE(!mem))
1766 return;
79e88659
DB		 1767 if (nullify)
1768 kzfree(mem);
1769 else
1770 kfree(mem);
1da177e4
LT		 1771 atomic_sub(size, &sk->sk_omem_alloc);
1772}
79e88659
DB		 1773
1774void sock_kfree_s(struct sock *sk, void *mem, int size)
1775{
1776 __sock_kfree_s(sk, mem, size, false);
1777}
2a91525c 1778EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1779
79e88659
DB		 1780void sock_kzfree_s(struct sock *sk, void *mem, int size)
1781{
1782 __sock_kfree_s(sk, mem, size, true);
1783}
1784EXPORT_SYMBOL(sock_kzfree_s);
1785
1da177e4
LT		 1786/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1787 I think, these locks should be removed for datagram sockets.
1788 */
2a91525c 1789static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1790{
1791 DEFINE_WAIT(wait);
1792
9cd3e072 1793 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1da177e4
LT		 1794 for (;;) {
1795 if (!timeo)
1796 break;
1797 if (signal_pending(current))
1798 break;
1799 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1800 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1801 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1802 break;
1803 if (sk->sk_shutdown & SEND_SHUTDOWN)
1804 break;
1805 if (sk->sk_err)
1806 break;
1807 timeo = schedule_timeout(timeo);
1808 }
aa395145 1809 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1810 return timeo;
1811}
1812
1813
1814/*
1815 * Generic send/receive buffer handlers
1816 */
1817
4cc7f68d
HX		 1818struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1819 unsigned long data_len, int noblock,
28d64271 1820 int *errcode, int max_page_order)
1da177e4 1821{
2e4e4410 1822 struct sk_buff *skb;
1da177e4
LT		 1823 long timeo;
1824 int err;
1825
1da177e4 1826 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1827 for (;;) {
1da177e4
LT		 1828 err = sock_error(sk);
1829 if (err != 0)
1830 goto failure;
1831
1832 err = -EPIPE;
1833 if (sk->sk_shutdown & SEND_SHUTDOWN)
1834 goto failure;
1835
2e4e4410
ED		 1836 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1837 break;
28d64271 1838
9cd3e072 1839 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2e4e4410
ED		 1840 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1841 err = -EAGAIN;
1842 if (!timeo)
1da177e4 1843 goto failure;
2e4e4410
ED		 1844 if (signal_pending(current))
1845 goto interrupted;
1846 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1847 }
2e4e4410
ED		 1848 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1849 errcode, sk->sk_allocation);
1850 if (skb)
1851 skb_set_owner_w(skb, sk);
1da177e4
LT		 1852 return skb;
1853
1854interrupted:
1855 err = sock_intr_errno(timeo);
1856failure:
1857 *errcode = err;
1858 return NULL;
1859}
4cc7f68d 1860EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1861
4ec93edb 1862struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1863 int noblock, int *errcode)
1864{
28d64271 1865 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1866}
2a91525c 1867EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1868
f28ea365
EJ		 1869int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1870 struct sockcm_cookie *sockc)
1871{
1872 struct cmsghdr *cmsg;
1873
1874 for_each_cmsghdr(cmsg, msg) {
1875 if (!CMSG_OK(msg, cmsg))
1876 return -EINVAL;
1877 if (cmsg->cmsg_level != SOL_SOCKET)
1878 continue;
1879 switch (cmsg->cmsg_type) {
1880 case SO_MARK:
1881 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1882 return -EPERM;
1883 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1884 return -EINVAL;
1885 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
1886 break;
1887 default:
1888 return -EINVAL;
1889 }
1890 }
1891 return 0;
1892}
1893EXPORT_SYMBOL(sock_cmsg_send);
1894
5640f768
ED		 1895/* On 32bit arches, an skb frag is limited to 2^15 */
1896#define SKB_FRAG_PAGE_ORDER get_order(32768)
1897
400dfd3a
ED		 1898/**
1899 * skb_page_frag_refill - check that a page_frag contains enough room
1900 * @sz: minimum size of the fragment we want to get
1901 * @pfrag: pointer to page_frag
82d5e2b8 1902 * @gfp: priority for memory allocation
400dfd3a
ED		 1903 *
1904 * Note: While this allocator tries to use high order pages, there is
1905 * no guarantee that allocations succeed. Therefore, @sz MUST be
1906 * less or equal than PAGE_SIZE.
1907 */
d9b2938a 1908bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1909{
5640f768 1910 if (pfrag->page) {
fe896d18 1911 if (page_ref_count(pfrag->page) == 1) {
5640f768
ED		 1912 pfrag->offset = 0;
1913 return true;
1914 }
400dfd3a 1915 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1916 return true;
1917 put_page(pfrag->page);
1918 }
1919
d9b2938a
ED		 1920 pfrag->offset = 0;
1921 if (SKB_FRAG_PAGE_ORDER) {
d0164adc
MG		 1922 /* Avoid direct reclaim but allow kswapd to wake */
1923 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
1924 __GFP_COMP | __GFP_NOWARN |
1925 __GFP_NORETRY,
d9b2938a 1926 SKB_FRAG_PAGE_ORDER);
5640f768 1927 if (likely(pfrag->page)) {
d9b2938a 1928 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1929 return true;
1930 }
d9b2938a
ED		 1931 }
1932 pfrag->page = alloc_page(gfp);
1933 if (likely(pfrag->page)) {
1934 pfrag->size = PAGE_SIZE;
1935 return true;
1936 }
400dfd3a
ED		 1937 return false;
1938}
1939EXPORT_SYMBOL(skb_page_frag_refill);
1940
1941bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1942{
1943 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1944 return true;
1945
5640f768
ED		 1946 sk_enter_memory_pressure(sk);
1947 sk_stream_moderate_sndbuf(sk);
1948 return false;
1949}
1950EXPORT_SYMBOL(sk_page_frag_refill);
1951
1da177e4 1952static void __lock_sock(struct sock *sk)
f39234d6
NK		 1953 __releases(&sk->sk_lock.slock)
1954 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1955{
1956 DEFINE_WAIT(wait);
1957
e71a4783 1958 for (;;) {
1da177e4
LT		 1959 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1960 TASK_UNINTERRUPTIBLE);
1961 spin_unlock_bh(&sk->sk_lock.slock);
1962 schedule();
1963 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1964 if (!sock_owned_by_user(sk))
1da177e4
LT		 1965 break;
1966 }
1967 finish_wait(&sk->sk_lock.wq, &wait);
1968}
1969
1970static void __release_sock(struct sock *sk)
f39234d6
NK		 1971 __releases(&sk->sk_lock.slock)
1972 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1973{
1974 struct sk_buff *skb = sk->sk_backlog.head;
1975
1976 do {
1977 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1978 bh_unlock_sock(sk);
1979
1980 do {
1981 struct sk_buff *next = skb->next;
1982
e4cbb02a 1983 prefetch(next);
7fee226a 1984 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1985 skb->next = NULL;
c57943a1 1986 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1987
1988 /*
1989 * We are in process context here with softirqs
1990 * disabled, use cond_resched_softirq() to preempt.
1991 * This is safe to do because we've taken the backlog
1992 * queue private:
1993 */
1994 cond_resched_softirq();
1995
1996 skb = next;
1997 } while (skb != NULL);
1998
1999 bh_lock_sock(sk);
e71a4783 2000 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 2001
2002 /*
2003 * Doing the zeroing here guarantee we can not loop forever
2004 * while a wild producer attempts to flood us.
2005 */
2006 sk->sk_backlog.len = 0;
1da177e4
LT		 2007}
2008
2009/**
2010 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 2011 * @sk: sock to wait on
2012 * @timeo: for how long
dfbafc99 2013 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 2014 *
2015 * Now socket state including sk->sk_err is changed only under lock,
2016 * hence we may omit checks after joining wait queue.
2017 * We check receive queue before schedule() only as optimization;
2018 * it is very likely that release_sock() added new data.
2019 */
dfbafc99 2020int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4
LT		 2021{
2022 int rc;
2023 DEFINE_WAIT(wait);
2024
aa395145 2025 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
9cd3e072 2026 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
dfbafc99 2027 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);
9cd3e072 2028 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
aa395145 2029 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2030 return rc;
2031}
1da177e4
LT		 2032EXPORT_SYMBOL(sk_wait_data);
2033
3ab224be
HA		 2034/**
2035 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2036 * @sk: socket
2037 * @size: memory size to allocate
2038 * @kind: allocation type
2039 *
2040 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2041 * rmem allocation. This function assumes that protocols which have
2042 * memory_pressure use sk_wmem_queued as write buffer accounting.
2043 */
2044int __sk_mem_schedule(struct sock *sk, int size, int kind)
2045{
2046 struct proto *prot = sk->sk_prot;
2047 int amt = sk_mem_pages(size);
8d987e5c 2048 long allocated;
3ab224be
HA		 2049
2050 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2051
e805605c
JW		 2052 allocated = sk_memory_allocated_add(sk, amt);
2053
baac50bb
JW		 2054 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2055 !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
e805605c 2056 goto suppress_allocation;
3ab224be
HA		 2057
2058 /* Under limit. */
e805605c 2059 if (allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2060 sk_leave_memory_pressure(sk);
3ab224be
HA		 2061 return 1;
2062 }
2063
e805605c
JW		 2064 /* Under pressure. */
2065 if (allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2066 sk_enter_memory_pressure(sk);
3ab224be 2067
e805605c
JW		 2068 /* Over hard limit. */
2069 if (allocated > sk_prot_mem_limits(sk, 2))
3ab224be
HA		 2070 goto suppress_allocation;
2071
2072 /* guarantee minimum buffer size under pressure */
2073 if (kind == SK_MEM_RECV) {
2074 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2075 return 1;
180d8cd9 2076
3ab224be
HA		 2077 } else { /* SK_MEM_SEND */
2078 if (sk->sk_type == SOCK_STREAM) {
2079 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2080 return 1;
2081 } else if (atomic_read(&sk->sk_wmem_alloc) <
2082 prot->sysctl_wmem[0])
2083 return 1;
2084 }
2085
180d8cd9 2086 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2087 int alloc;
2088
180d8cd9 2089 if (!sk_under_memory_pressure(sk))
1748376b 2090 return 1;
180d8cd9
GC		 2091 alloc = sk_sockets_allocated_read_positive(sk);
2092 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2093 sk_mem_pages(sk->sk_wmem_queued +
2094 atomic_read(&sk->sk_rmem_alloc) +
2095 sk->sk_forward_alloc))
2096 return 1;
2097 }
2098
2099suppress_allocation:
2100
2101 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2102 sk_stream_moderate_sndbuf(sk);
2103
2104 /* Fail only if socket is _under_ its sndbuf.
2105 * In this case we cannot block, so that we have to fail.
2106 */
2107 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2108 return 1;
2109 }
2110
3847ce32
SM		 2111 trace_sock_exceed_buf_limit(sk, prot, allocated);
2112
3ab224be
HA		 2113 /* Alas. Undo changes. */
2114 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2115
0e90b31f 2116 sk_memory_allocated_sub(sk, amt);
180d8cd9 2117
baac50bb
JW		 2118 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2119 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
e805605c 2120
3ab224be
HA		 2121 return 0;
2122}
3ab224be
HA		 2123EXPORT_SYMBOL(__sk_mem_schedule);
2124
2125/**
69dba9bb 2126 * __sk_mem_reclaim - reclaim memory_allocated
3ab224be 2127 * @sk: socket
1a24e04e 2128 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
3ab224be 2129 */
1a24e04e 2130void __sk_mem_reclaim(struct sock *sk, int amount)
3ab224be 2131{
1a24e04e
ED		 2132 amount >>= SK_MEM_QUANTUM_SHIFT;
2133 sk_memory_allocated_sub(sk, amount);
2134 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
3ab224be 2135
baac50bb
JW		 2136 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2137 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
e805605c 2138
180d8cd9
GC		 2139 if (sk_under_memory_pressure(sk) &&
2140 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2141 sk_leave_memory_pressure(sk);
3ab224be 2142}
3ab224be
HA		 2143EXPORT_SYMBOL(__sk_mem_reclaim);
2144
2145
1da177e4
LT		 2146/*
2147 * Set of default routines for initialising struct proto_ops when
2148 * the protocol does not support a particular function. In certain
2149 * cases where it makes no sense for a protocol to have a "do nothing"
2150 * function, some default processing is provided.
2151 */
2152
2153int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2154{
2155 return -EOPNOTSUPP;
2156}
2a91525c 2157EXPORT_SYMBOL(sock_no_bind);
1da177e4 2158
4ec93edb 2159int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2160 int len, int flags)
2161{
2162 return -EOPNOTSUPP;
2163}
2a91525c 2164EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2165
2166int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2167{
2168 return -EOPNOTSUPP;
2169}
2a91525c 2170EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2171
2172int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2173{
2174 return -EOPNOTSUPP;
2175}
2a91525c 2176EXPORT_SYMBOL(sock_no_accept);
1da177e4 2177
4ec93edb 2178int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2179 int *len, int peer)
2180{
2181 return -EOPNOTSUPP;
2182}
2a91525c 2183EXPORT_SYMBOL(sock_no_getname);
1da177e4 2184
2a91525c 2185unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2186{
2187 return 0;
2188}
2a91525c 2189EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2190
2191int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2192{
2193 return -EOPNOTSUPP;
2194}
2a91525c 2195EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2196
2197int sock_no_listen(struct socket *sock, int backlog)
2198{
2199 return -EOPNOTSUPP;
2200}
2a91525c 2201EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2202
2203int sock_no_shutdown(struct socket *sock, int how)
2204{
2205 return -EOPNOTSUPP;
2206}
2a91525c 2207EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2208
2209int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2210 char __user *optval, unsigned int optlen)
1da177e4
LT		 2211{
2212 return -EOPNOTSUPP;
2213}
2a91525c 2214EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2215
2216int sock_no_getsockopt(struct socket *sock, int level, int optname,
2217 char __user *optval, int __user *optlen)
2218{
2219 return -EOPNOTSUPP;
2220}
2a91525c 2221EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2222
1b784140 2223int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2224{
2225 return -EOPNOTSUPP;
2226}
2a91525c 2227EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2228
1b784140
YX		 2229int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2230 int flags)
1da177e4
LT		 2231{
2232 return -EOPNOTSUPP;
2233}
2a91525c 2234EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2235
2236int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2237{
2238 /* Mirror missing mmap method error code */
2239 return -ENODEV;
2240}
2a91525c 2241EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2242
2243ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2244{
2245 ssize_t res;
2246 struct msghdr msg = {.msg_flags = flags};
2247 struct kvec iov;
2248 char *kaddr = kmap(page);
2249 iov.iov_base = kaddr + offset;
2250 iov.iov_len = size;
2251 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2252 kunmap(page);
2253 return res;
2254}
2a91525c 2255EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2256
2257/*
2258 * Default Socket Callbacks
2259 */
2260
2261static void sock_def_wakeup(struct sock *sk)
2262{
43815482
ED		 2263 struct socket_wq *wq;
2264
2265 rcu_read_lock();
2266 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2267 if (skwq_has_sleeper(wq))
43815482
ED		 2268 wake_up_interruptible_all(&wq->wait);
2269 rcu_read_unlock();
1da177e4
LT		 2270}
2271
2272static void sock_def_error_report(struct sock *sk)
2273{
43815482
ED		 2274 struct socket_wq *wq;
2275
2276 rcu_read_lock();
2277 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2278 if (skwq_has_sleeper(wq))
43815482 2279 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2280 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2281 rcu_read_unlock();
1da177e4
LT		 2282}
2283
676d2369 2284static void sock_def_readable(struct sock *sk)
1da177e4 2285{
43815482
ED		 2286 struct socket_wq *wq;
2287
2288 rcu_read_lock();
2289 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2290 if (skwq_has_sleeper(wq))
2c6607c6 2291 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2292 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2293 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2294 rcu_read_unlock();
1da177e4
LT		 2295}
2296
2297static void sock_def_write_space(struct sock *sk)
2298{
43815482
ED		 2299 struct socket_wq *wq;
2300
2301 rcu_read_lock();
1da177e4
LT		 2302
2303 /* Do not wake up a writer until he can make "significant"
2304 * progress. --DaveM
2305 */
e71a4783 2306 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482 2307 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2308 if (skwq_has_sleeper(wq))
43815482 2309 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2310 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2311
2312 /* Should agree with poll, otherwise some programs break */
2313 if (sock_writeable(sk))
8d8ad9d7 2314 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2315 }
2316
43815482 2317 rcu_read_unlock();
1da177e4
LT		 2318}
2319
2320static void sock_def_destruct(struct sock *sk)
2321{
1da177e4
LT		 2322}
2323
2324void sk_send_sigurg(struct sock *sk)
2325{
2326 if (sk->sk_socket && sk->sk_socket->file)
2327 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2328 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2329}
2a91525c 2330EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2331
2332void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2333 unsigned long expires)
2334{
2335 if (!mod_timer(timer, expires))
2336 sock_hold(sk);
2337}
1da177e4
LT		 2338EXPORT_SYMBOL(sk_reset_timer);
2339
2340void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2341{
25cc4ae9 2342 if (del_timer(timer))
1da177e4
LT		 2343 __sock_put(sk);
2344}
1da177e4
LT		 2345EXPORT_SYMBOL(sk_stop_timer);
2346
2347void sock_init_data(struct socket *sock, struct sock *sk)
2348{
2349 skb_queue_head_init(&sk->sk_receive_queue);
2350 skb_queue_head_init(&sk->sk_write_queue);
2351 skb_queue_head_init(&sk->sk_error_queue);
2352
2353 sk->sk_send_head = NULL;
2354
2355 init_timer(&sk->sk_timer);
4ec93edb 2356
1da177e4
LT		 2357 sk->sk_allocation = GFP_KERNEL;
2358 sk->sk_rcvbuf = sysctl_rmem_default;
2359 sk->sk_sndbuf = sysctl_wmem_default;
2360 sk->sk_state = TCP_CLOSE;
972692e0 2361 sk_set_socket(sk, sock);
1da177e4
LT		 2362
2363 sock_set_flag(sk, SOCK_ZAPPED);
2364
e71a4783 2365 if (sock) {
1da177e4 2366 sk->sk_type = sock->type;
43815482 2367 sk->sk_wq = sock->wq;
1da177e4
LT		 2368 sock->sk = sk;
2369 } else
43815482 2370 sk->sk_wq = NULL;
1da177e4 2371
1da177e4 2372 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2373 lockdep_set_class_and_name(&sk->sk_callback_lock,
2374 af_callback_keys + sk->sk_family,
2375 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2376
2377 sk->sk_state_change = sock_def_wakeup;
2378 sk->sk_data_ready = sock_def_readable;
2379 sk->sk_write_space = sock_def_write_space;
2380 sk->sk_error_report = sock_def_error_report;
2381 sk->sk_destruct = sock_def_destruct;
2382
5640f768
ED		 2383 sk->sk_frag.page = NULL;
2384 sk->sk_frag.offset = 0;
ef64a54f 2385 sk->sk_peek_off = -1;
1da177e4 2386
109f6e39
EB		 2387 sk->sk_peer_pid = NULL;
2388 sk->sk_peer_cred = NULL;
1da177e4
LT		 2389 sk->sk_write_pending = 0;
2390 sk->sk_rcvlowat = 1;
2391 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2392 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2393
f37f0afb 2394 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2395
e0d1095a 2396#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2397 sk->sk_napi_id = 0;
64b0dc51 2398 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2399#endif
2400
62748f32 2401 sk->sk_max_pacing_rate = ~0U;
7eec4174 2402 sk->sk_pacing_rate = ~0U;
70da268b 2403 sk->sk_incoming_cpu = -1;
4dc6dc71
ED		 2404 /*
2405 * Before updating sk_refcnt, we must commit prior changes to memory
2406 * (Documentation/RCU/rculist_nulls.txt for details)
2407 */
2408 smp_wmb();
1da177e4 2409 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2410 atomic_set(&sk->sk_drops, 0);
1da177e4 2411}
2a91525c 2412EXPORT_SYMBOL(sock_init_data);
1da177e4 2413
b5606c2d 2414void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2415{
2416 might_sleep();
a5b5bb9a 2417 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2418 if (sk->sk_lock.owned)
1da177e4 2419 __lock_sock(sk);
d2e9117c 2420 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2421 spin_unlock(&sk->sk_lock.slock);
2422 /*
2423 * The sk_lock has mutex_lock() semantics here:
2424 */
fcc70d5f 2425 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2426 local_bh_enable();
1da177e4 2427}
fcc70d5f 2428EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2429
b5606c2d 2430void release_sock(struct sock *sk)
1da177e4 2431{
a5b5bb9a
IM		 2432 /*
2433 * The sk_lock has mutex_unlock() semantics:
2434 */
2435 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2436
2437 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2438 if (sk->sk_backlog.tail)
2439 __release_sock(sk);
46d3ceab 2440
c3f9b018
ED		 2441 /* Warning : release_cb() might need to release sk ownership,
2442 * ie call sock_release_ownership(sk) before us.
2443 */
46d3ceab
ED		 2444 if (sk->sk_prot->release_cb)
2445 sk->sk_prot->release_cb(sk);
2446
c3f9b018 2447 sock_release_ownership(sk);
a5b5bb9a
IM		 2448 if (waitqueue_active(&sk->sk_lock.wq))
2449 wake_up(&sk->sk_lock.wq);
2450 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2451}
2452EXPORT_SYMBOL(release_sock);
2453
8a74ad60
ED		 2454/**
2455 * lock_sock_fast - fast version of lock_sock
2456 * @sk: socket
2457 *
2458 * This version should be used for very small section, where process wont block
2459 * return false if fast path is taken
2460 * sk_lock.slock locked, owned = 0, BH disabled
2461 * return true if slow path is taken
2462 * sk_lock.slock unlocked, owned = 1, BH enabled
2463 */
2464bool lock_sock_fast(struct sock *sk)
2465{
2466 might_sleep();
2467 spin_lock_bh(&sk->sk_lock.slock);
2468
2469 if (!sk->sk_lock.owned)
2470 /*
2471 * Note : We must disable BH
2472 */
2473 return false;
2474
2475 __lock_sock(sk);
2476 sk->sk_lock.owned = 1;
2477 spin_unlock(&sk->sk_lock.slock);
2478 /*
2479 * The sk_lock has mutex_lock() semantics here:
2480 */
2481 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2482 local_bh_enable();
2483 return true;
2484}
2485EXPORT_SYMBOL(lock_sock_fast);
2486
1da177e4 2487int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2488{
b7aa0bf7 2489 struct timeval tv;
1da177e4 2490 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2491 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2492 tv = ktime_to_timeval(sk->sk_stamp);
2493 if (tv.tv_sec == -1)
1da177e4 2494 return -ENOENT;
b7aa0bf7
ED		 2495 if (tv.tv_sec == 0) {
2496 sk->sk_stamp = ktime_get_real();
2497 tv = ktime_to_timeval(sk->sk_stamp);
2498 }
2499 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2500}
1da177e4
LT		 2501EXPORT_SYMBOL(sock_get_timestamp);
2502
ae40eb1e
ED		 2503int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2504{
2505 struct timespec ts;
2506 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2507 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2508 ts = ktime_to_timespec(sk->sk_stamp);
2509 if (ts.tv_sec == -1)
2510 return -ENOENT;
2511 if (ts.tv_sec == 0) {
2512 sk->sk_stamp = ktime_get_real();
2513 ts = ktime_to_timespec(sk->sk_stamp);
2514 }
2515 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2516}
2517EXPORT_SYMBOL(sock_get_timestampns);
2518
20d49473 2519void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2520{
20d49473 2521 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2522 unsigned long previous_flags = sk->sk_flags;
2523
20d49473
PO		 2524 sock_set_flag(sk, flag);
2525 /*
2526 * we just set one of the two flags which require net
2527 * time stamping, but time stamping might have been on
2528 * already because of the other one
2529 */
080a270f
HFS		 2530 if (sock_needs_netstamp(sk) &&
2531 !(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2532 net_enable_timestamp();
1da177e4
LT		 2533 }
2534}
1da177e4 2535
cb820f8e
RC		 2536int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2537 int level, int type)
2538{
2539 struct sock_exterr_skb *serr;
364a9e93 2540 struct sk_buff *skb;
cb820f8e
RC		 2541 int copied, err;
2542
2543 err = -EAGAIN;
364a9e93 2544 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2545 if (skb == NULL)
2546 goto out;
2547
2548 copied = skb->len;
2549 if (copied > len) {
2550 msg->msg_flags |= MSG_TRUNC;
2551 copied = len;
2552 }
51f3d02b 2553 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2554 if (err)
2555 goto out_free_skb;
2556
2557 sock_recv_timestamp(msg, sk, skb);
2558
2559 serr = SKB_EXT_ERR(skb);
2560 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2561
2562 msg->msg_flags |= MSG_ERRQUEUE;
2563 err = copied;
2564
cb820f8e
RC		 2565out_free_skb:
2566 kfree_skb(skb);
2567out:
2568 return err;
2569}
2570EXPORT_SYMBOL(sock_recv_errqueue);
2571
1da177e4
LT		 2572/*
2573 * Get a socket option on an socket.
2574 *
2575 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2576 * asynchronous errors should be reported by getsockopt. We assume
2577 * this means if you specify SO_ERROR (otherwise whats the point of it).
2578 */
2579int sock_common_getsockopt(struct socket *sock, int level, int optname,
2580 char __user *optval, int __user *optlen)
2581{
2582 struct sock *sk = sock->sk;
2583
2584 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2585}
1da177e4
LT		 2586EXPORT_SYMBOL(sock_common_getsockopt);
2587
3fdadf7d 2588#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2589int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2590 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2591{
2592 struct sock *sk = sock->sk;
2593
1e51f951 2594 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2595 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2596 optval, optlen);
3fdadf7d
DM		 2597 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2598}
2599EXPORT_SYMBOL(compat_sock_common_getsockopt);
2600#endif
2601
1b784140
YX		 2602int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2603 int flags)
1da177e4
LT		 2604{
2605 struct sock *sk = sock->sk;
2606 int addr_len = 0;
2607 int err;
2608
1b784140 2609 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2610 flags & ~MSG_DONTWAIT, &addr_len);
2611 if (err >= 0)
2612 msg->msg_namelen = addr_len;
2613 return err;
2614}
1da177e4
LT		 2615EXPORT_SYMBOL(sock_common_recvmsg);
2616
2617/*
2618 * Set socket options on an inet socket.
2619 */
2620int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2621 char __user *optval, unsigned int optlen)
1da177e4
LT		 2622{
2623 struct sock *sk = sock->sk;
2624
2625 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2626}
1da177e4
LT		 2627EXPORT_SYMBOL(sock_common_setsockopt);
2628
3fdadf7d 2629#ifdef CONFIG_COMPAT
543d9cfe 2630int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2631 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2632{
2633 struct sock *sk = sock->sk;
2634
543d9cfe
ACM		 2635 if (sk->sk_prot->compat_setsockopt != NULL)
2636 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2637 optval, optlen);
3fdadf7d
DM		 2638 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2639}
2640EXPORT_SYMBOL(compat_sock_common_setsockopt);
2641#endif
2642
1da177e4
LT		 2643void sk_common_release(struct sock *sk)
2644{
2645 if (sk->sk_prot->destroy)
2646 sk->sk_prot->destroy(sk);
2647
2648 /*
2649 * Observation: when sock_common_release is called, processes have
2650 * no access to socket. But net still has.
2651 * Step one, detach it from networking:
2652 *
2653 * A. Remove from hash tables.
2654 */
2655
2656 sk->sk_prot->unhash(sk);
2657
2658 /*
2659 * In this point socket cannot receive new packets, but it is possible
2660 * that some packets are in flight because some CPU runs receiver and
2661 * did hash table lookup before we unhashed socket. They will achieve
2662 * receive queue and will be purged by socket destructor.
2663 *
2664 * Also we still have packets pending on receive queue and probably,
2665 * our own packets waiting in device queues. sock_destroy will drain
2666 * receive queue, but transmitted packets will delay socket destruction
2667 * until the last reference will be released.
2668 */
2669
2670 sock_orphan(sk);
2671
2672 xfrm_sk_free_policy(sk);
2673
e6848976 2674 sk_refcnt_debug_release(sk);
5640f768
ED		 2675
2676 if (sk->sk_frag.page) {
2677 put_page(sk->sk_frag.page);
2678 sk->sk_frag.page = NULL;
2679 }
2680
1da177e4
LT		 2681 sock_put(sk);
2682}
1da177e4
LT		 2683EXPORT_SYMBOL(sk_common_release);
2684
13ff3d6f
PE		 2685#ifdef CONFIG_PROC_FS
2686#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2687struct prot_inuse {
2688 int val[PROTO_INUSE_NR];
2689};
13ff3d6f
PE		 2690
2691static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2692
2693#ifdef CONFIG_NET_NS
2694void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2695{
d6d9ca0f 2696 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2697}
2698EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2699
2700int sock_prot_inuse_get(struct net *net, struct proto *prot)
2701{
2702 int cpu, idx = prot->inuse_idx;
2703 int res = 0;
2704
2705 for_each_possible_cpu(cpu)
2706 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2707
2708 return res >= 0 ? res : 0;
2709}
2710EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2711
2c8c1e72 2712static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2713{
2714 net->core.inuse = alloc_percpu(struct prot_inuse);
2715 return net->core.inuse ? 0 : -ENOMEM;
2716}
2717
2c8c1e72 2718static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2719{
2720 free_percpu(net->core.inuse);
2721}
2722
2723static struct pernet_operations net_inuse_ops = {
2724 .init = sock_inuse_init_net,
2725 .exit = sock_inuse_exit_net,
2726};
2727
2728static __init int net_inuse_init(void)
2729{
2730 if (register_pernet_subsys(&net_inuse_ops))
2731 panic("Cannot initialize net inuse counters");
2732
2733 return 0;
2734}
2735
2736core_initcall(net_inuse_init);
2737#else
1338d466
PE		 2738static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2739
c29a0bc4 2740void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2741{
d6d9ca0f 2742 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2743}
2744EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2745
c29a0bc4 2746int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2747{
2748 int cpu, idx = prot->inuse_idx;
2749 int res = 0;
2750
2751 for_each_possible_cpu(cpu)
2752 res += per_cpu(prot_inuse, cpu).val[idx];
2753
2754 return res >= 0 ? res : 0;
2755}
2756EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2757#endif
13ff3d6f
PE		 2758
2759static void assign_proto_idx(struct proto *prot)
2760{
2761 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2762
2763 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2764 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2765 return;
2766 }
2767
2768 set_bit(prot->inuse_idx, proto_inuse_idx);
2769}
2770
2771static void release_proto_idx(struct proto *prot)
2772{
2773 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2774 clear_bit(prot->inuse_idx, proto_inuse_idx);
2775}
2776#else
2777static inline void assign_proto_idx(struct proto *prot)
2778{
2779}
2780
2781static inline void release_proto_idx(struct proto *prot)
2782{
2783}
2784#endif
2785
0159dfd3
ED		 2786static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2787{
2788 if (!rsk_prot)
2789 return;
2790 kfree(rsk_prot->slab_name);
2791 rsk_prot->slab_name = NULL;
adf78eda
JL		 2792 kmem_cache_destroy(rsk_prot->slab);
2793 rsk_prot->slab = NULL;
0159dfd3
ED		 2794}
2795
2796static int req_prot_init(const struct proto *prot)
2797{
2798 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2799
2800 if (!rsk_prot)
2801 return 0;
2802
2803 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2804 prot->name);
2805 if (!rsk_prot->slab_name)
2806 return -ENOMEM;
2807
2808 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2809 rsk_prot->obj_size, 0,
e96f78ab 2810 prot->slab_flags, NULL);
0159dfd3
ED		 2811
2812 if (!rsk_prot->slab) {
2813 pr_crit("%s: Can't create request sock SLAB cache!\n",
2814 prot->name);
2815 return -ENOMEM;
2816 }
2817 return 0;
2818}
2819
b733c007
PE		 2820int proto_register(struct proto *prot, int alloc_slab)
2821{
1da177e4
LT		 2822 if (alloc_slab) {
2823 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2824 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2825 NULL);
1da177e4
LT		 2826
2827 if (prot->slab == NULL) {
e005d193
JP		 2828 pr_crit("%s: Can't create sock SLAB cache!\n",
2829 prot->name);
60e7663d 2830 goto out;
1da177e4 2831 }
2e6599cb 2832
0159dfd3
ED		 2833 if (req_prot_init(prot))
2834 goto out_free_request_sock_slab;
8feaf0c0 2835
6d6ee43e 2836 if (prot->twsk_prot != NULL) {
faf23422 2837 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2838
7e56b5d6 2839 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2840 goto out_free_request_sock_slab;
2841
6d6ee43e 2842 prot->twsk_prot->twsk_slab =
7e56b5d6 2843 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2844 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2845 0,
52db70dc 2846 prot->slab_flags,
20c2df83 2847 NULL);
6d6ee43e 2848 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2849 goto out_free_timewait_sock_slab_name;
2850 }
1da177e4
LT		 2851 }
2852
36b77a52 2853 mutex_lock(&proto_list_mutex);
1da177e4 2854 list_add(&prot->node, &proto_list);
13ff3d6f 2855 assign_proto_idx(prot);
36b77a52 2856 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2857 return 0;
2858
8feaf0c0 2859out_free_timewait_sock_slab_name:
7e56b5d6 2860 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2861out_free_request_sock_slab:
0159dfd3
ED		 2862 req_prot_cleanup(prot->rsk_prot);
2863
2e6599cb
ACM		 2864 kmem_cache_destroy(prot->slab);
2865 prot->slab = NULL;
b733c007
PE		 2866out:
2867 return -ENOBUFS;
1da177e4 2868}
1da177e4
LT		 2869EXPORT_SYMBOL(proto_register);
2870
2871void proto_unregister(struct proto *prot)
2872{
36b77a52 2873 mutex_lock(&proto_list_mutex);
13ff3d6f 2874 release_proto_idx(prot);
0a3f4358 2875 list_del(&prot->node);
36b77a52 2876 mutex_unlock(&proto_list_mutex);
1da177e4 2877
adf78eda
JL		 2878 kmem_cache_destroy(prot->slab);
2879 prot->slab = NULL;
1da177e4 2880
0159dfd3 2881 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2882
6d6ee43e 2883 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2884 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2885 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2886 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2887 }
1da177e4 2888}
1da177e4
LT		 2889EXPORT_SYMBOL(proto_unregister);
2890
2891#ifdef CONFIG_PROC_FS
1da177e4 2892static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2893 __acquires(proto_list_mutex)
1da177e4 2894{
36b77a52 2895 mutex_lock(&proto_list_mutex);
60f0438a 2896 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2897}
2898
2899static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2900{
60f0438a 2901 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2902}
2903
2904static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2905 __releases(proto_list_mutex)
1da177e4 2906{
36b77a52 2907 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2908}
2909
2910static char proto_method_implemented(const void *method)
2911{
2912 return method == NULL ? 'n' : 'y';
2913}
180d8cd9
GC		 2914static long sock_prot_memory_allocated(struct proto *proto)
2915{
cb75a36c 2916 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2917}
2918
2919static char *sock_prot_memory_pressure(struct proto *proto)
2920{
2921 return proto->memory_pressure != NULL ?
2922 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2923}
1da177e4
LT		 2924
2925static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2926{
180d8cd9 2927
8d987e5c 2928 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2929 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2930 proto->name,
2931 proto->obj_size,
14e943db 2932 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2933 sock_prot_memory_allocated(proto),
2934 sock_prot_memory_pressure(proto),
1da177e4
LT		 2935 proto->max_header,
2936 proto->slab == NULL ? "no" : "yes",
2937 module_name(proto->owner),
2938 proto_method_implemented(proto->close),
2939 proto_method_implemented(proto->connect),
2940 proto_method_implemented(proto->disconnect),
2941 proto_method_implemented(proto->accept),
2942 proto_method_implemented(proto->ioctl),
2943 proto_method_implemented(proto->init),
2944 proto_method_implemented(proto->destroy),
2945 proto_method_implemented(proto->shutdown),
2946 proto_method_implemented(proto->setsockopt),
2947 proto_method_implemented(proto->getsockopt),
2948 proto_method_implemented(proto->sendmsg),
2949 proto_method_implemented(proto->recvmsg),
2950 proto_method_implemented(proto->sendpage),
2951 proto_method_implemented(proto->bind),
2952 proto_method_implemented(proto->backlog_rcv),
2953 proto_method_implemented(proto->hash),
2954 proto_method_implemented(proto->unhash),
2955 proto_method_implemented(proto->get_port),
2956 proto_method_implemented(proto->enter_memory_pressure));
2957}
2958
2959static int proto_seq_show(struct seq_file *seq, void *v)
2960{
60f0438a 2961 if (v == &proto_list)
1da177e4
LT		 2962 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2963 "protocol",
2964 "size",
2965 "sockets",
2966 "memory",
2967 "press",
2968 "maxhdr",
2969 "slab",
2970 "module",
2971 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2972 else
60f0438a 2973 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2974 return 0;
2975}
2976
f690808e 2977static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2978 .start = proto_seq_start,
2979 .next = proto_seq_next,
2980 .stop = proto_seq_stop,
2981 .show = proto_seq_show,
2982};
2983
2984static int proto_seq_open(struct inode *inode, struct file *file)
2985{
14e943db
ED		 2986 return seq_open_net(inode, file, &proto_seq_ops,
2987 sizeof(struct seq_net_private));
1da177e4
LT		 2988}
2989
9a32144e 2990static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2991 .owner = THIS_MODULE,
2992 .open = proto_seq_open,
2993 .read = seq_read,
2994 .llseek = seq_lseek,
14e943db
ED		 2995 .release = seq_release_net,
2996};
2997
2998static __net_init int proto_init_net(struct net *net)
2999{
d4beaa66 3000 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 3001 return -ENOMEM;
3002
3003 return 0;
3004}
3005
3006static __net_exit void proto_exit_net(struct net *net)
3007{
ece31ffd 3008 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 3009}
3010
3011
3012static __net_initdata struct pernet_operations proto_net_ops = {
3013 .init = proto_init_net,
3014 .exit = proto_exit_net,
1da177e4
LT		 3015};
3016
3017static int __init proto_init(void)
3018{
14e943db 3019 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 3020}
3021
3022subsys_initcall(proto_init);
3023
3024#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 1.265303 seconds and 5 git commands to generate.