projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
tcp: increment sk_drops for dropped rx packets
[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) {
6db8b963
SHY		 835 if ((1 << sk->sk_state) &
836 (TCPF_CLOSE | TCPF_LISTEN)) {
4ed2d765
WB		 837 ret = -EINVAL;
838 break;
839 }
840 sk->sk_tskey = tcp_sk(sk)->snd_una;
841 } else {
842 sk->sk_tskey = 0;
843 }
844 }
b9f40e21 845 sk->sk_tsflags = val;
20d49473
PO		 846 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
847 sock_enable_timestamp(sk,
848 SOCK_TIMESTAMPING_RX_SOFTWARE);
849 else
850 sock_disable_timestamp(sk,
08e29af3 851 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 852 break;
853
e71a4783
SH		 854 case SO_RCVLOWAT:
855 if (val < 0)
856 val = INT_MAX;
857 sk->sk_rcvlowat = val ? : 1;
858 break;
859
860 case SO_RCVTIMEO:
861 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
862 break;
863
864 case SO_SNDTIMEO:
865 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
866 break;
1da177e4 867
e71a4783
SH		 868 case SO_ATTACH_FILTER:
869 ret = -EINVAL;
870 if (optlen == sizeof(struct sock_fprog)) {
871 struct sock_fprog fprog;
1da177e4 872
e71a4783
SH		 873 ret = -EFAULT;
874 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 875 break;
e71a4783
SH		 876
877 ret = sk_attach_filter(&fprog, sk);
878 }
879 break;
880
89aa0758
AS		 881 case SO_ATTACH_BPF:
882 ret = -EINVAL;
883 if (optlen == sizeof(u32)) {
884 u32 ufd;
885
886 ret = -EFAULT;
887 if (copy_from_user(&ufd, optval, sizeof(ufd)))
888 break;
889
890 ret = sk_attach_bpf(ufd, sk);
891 }
892 break;
893
538950a1
CG		 894 case SO_ATTACH_REUSEPORT_CBPF:
895 ret = -EINVAL;
896 if (optlen == sizeof(struct sock_fprog)) {
897 struct sock_fprog fprog;
898
899 ret = -EFAULT;
900 if (copy_from_user(&fprog, optval, sizeof(fprog)))
901 break;
902
903 ret = sk_reuseport_attach_filter(&fprog, sk);
904 }
905 break;
906
907 case SO_ATTACH_REUSEPORT_EBPF:
908 ret = -EINVAL;
909 if (optlen == sizeof(u32)) {
910 u32 ufd;
911
912 ret = -EFAULT;
913 if (copy_from_user(&ufd, optval, sizeof(ufd)))
914 break;
915
916 ret = sk_reuseport_attach_bpf(ufd, sk);
917 }
918 break;
919
e71a4783 920 case SO_DETACH_FILTER:
55b33325 921 ret = sk_detach_filter(sk);
e71a4783 922 break;
1da177e4 923
d59577b6
VB		 924 case SO_LOCK_FILTER:
925 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
926 ret = -EPERM;
927 else
928 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
929 break;
930
e71a4783
SH		 931 case SO_PASSSEC:
932 if (valbool)
933 set_bit(SOCK_PASSSEC, &sock->flags);
934 else
935 clear_bit(SOCK_PASSSEC, &sock->flags);
936 break;
4a19ec58 937 case SO_MARK:
5e1fccc0 938 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 939 ret = -EPERM;
2a91525c 940 else
4a19ec58 941 sk->sk_mark = val;
4a19ec58 942 break;
877ce7c1 943
3b885787 944 case SO_RXQ_OVFL:
8083f0fc 945 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 946 break;
6e3e939f
JB		 947
948 case SO_WIFI_STATUS:
949 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
950 break;
951
ef64a54f
PE		 952 case SO_PEEK_OFF:
953 if (sock->ops->set_peek_off)
12663bfc 954 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 955 else
956 ret = -EOPNOTSUPP;
957 break;
3bdc0eba
BG		 958
959 case SO_NOFCS:
960 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
961 break;
962
7d4c04fc
KJ		 963 case SO_SELECT_ERR_QUEUE:
964 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
965 break;
966
e0d1095a 967#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 968 case SO_BUSY_POLL:
dafcc438
ET		 969 /* allow unprivileged users to decrease the value */
970 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
971 ret = -EPERM;
972 else {
973 if (val < 0)
974 ret = -EINVAL;
975 else
976 sk->sk_ll_usec = val;
977 }
978 break;
979#endif
62748f32
ED		 980
981 case SO_MAX_PACING_RATE:
982 sk->sk_max_pacing_rate = val;
983 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
984 sk->sk_max_pacing_rate);
985 break;
986
70da268b
ED		 987 case SO_INCOMING_CPU:
988 sk->sk_incoming_cpu = val;
989 break;
990
a87cb3e4
TH		 991 case SO_CNX_ADVICE:
992 if (val == 1)
993 dst_negative_advice(sk);
994 break;
e71a4783
SH		 995 default:
996 ret = -ENOPROTOOPT;
997 break;
4ec93edb 998 }
1da177e4
LT		 999 release_sock(sk);
1000 return ret;
1001}
2a91525c 1002EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1003
1004
8f09898b 1005static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1006 struct ucred *ucred)
3f551f94
EB		 1007{
1008 ucred->pid = pid_vnr(pid);
1009 ucred->uid = ucred->gid = -1;
1010 if (cred) {
1011 struct user_namespace *current_ns = current_user_ns();
1012
b2e4f544
EB		 1013 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1014 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1015 }
1016}
1017
1da177e4
LT		 1018int sock_getsockopt(struct socket *sock, int level, int optname,
1019 char __user *optval, int __user *optlen)
1020{
1021 struct sock *sk = sock->sk;
4ec93edb 1022
e71a4783 1023 union {
4ec93edb
YH		 1024 int val;
1025 struct linger ling;
1da177e4
LT		 1026 struct timeval tm;
1027 } v;
4ec93edb 1028
4d0392be 1029 int lv = sizeof(int);
1da177e4 1030 int len;
4ec93edb 1031
e71a4783 1032 if (get_user(len, optlen))
4ec93edb 1033 return -EFAULT;
e71a4783 1034 if (len < 0)
1da177e4 1035 return -EINVAL;
4ec93edb 1036
50fee1de 1037 memset(&v, 0, sizeof(v));
df0bca04 1038
2a91525c 1039 switch (optname) {
e71a4783
SH		 1040 case SO_DEBUG:
1041 v.val = sock_flag(sk, SOCK_DBG);
1042 break;
1043
1044 case SO_DONTROUTE:
1045 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1046 break;
1047
1048 case SO_BROADCAST:
1b23a5df 1049 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1050 break;
1051
1052 case SO_SNDBUF:
1053 v.val = sk->sk_sndbuf;
1054 break;
1055
1056 case SO_RCVBUF:
1057 v.val = sk->sk_rcvbuf;
1058 break;
1059
1060 case SO_REUSEADDR:
1061 v.val = sk->sk_reuse;
1062 break;
1063
055dc21a
TH		 1064 case SO_REUSEPORT:
1065 v.val = sk->sk_reuseport;
1066 break;
1067
e71a4783 1068 case SO_KEEPALIVE:
1b23a5df 1069 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1070 break;
1071
1072 case SO_TYPE:
1073 v.val = sk->sk_type;
1074 break;
1075
49c794e9
JE		 1076 case SO_PROTOCOL:
1077 v.val = sk->sk_protocol;
1078 break;
1079
0d6038ee
JE		 1080 case SO_DOMAIN:
1081 v.val = sk->sk_family;
1082 break;
1083
e71a4783
SH		 1084 case SO_ERROR:
1085 v.val = -sock_error(sk);
2a91525c 1086 if (v.val == 0)
e71a4783
SH		 1087 v.val = xchg(&sk->sk_err_soft, 0);
1088 break;
1089
1090 case SO_OOBINLINE:
1b23a5df 1091 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1092 break;
1093
1094 case SO_NO_CHECK:
28448b80 1095 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1096 break;
1097
1098 case SO_PRIORITY:
1099 v.val = sk->sk_priority;
1100 break;
1101
1102 case SO_LINGER:
1103 lv = sizeof(v.ling);
1b23a5df 1104 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1105 v.ling.l_linger = sk->sk_lingertime / HZ;
1106 break;
1107
1108 case SO_BSDCOMPAT:
1109 sock_warn_obsolete_bsdism("getsockopt");
1110 break;
1111
1112 case SO_TIMESTAMP:
92f37fd2
ED		 1113 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1114 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1115 break;
1116
1117 case SO_TIMESTAMPNS:
1118 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1119 break;
1120
20d49473 1121 case SO_TIMESTAMPING:
b9f40e21 1122 v.val = sk->sk_tsflags;
20d49473
PO		 1123 break;
1124
e71a4783 1125 case SO_RCVTIMEO:
2a91525c 1126 lv = sizeof(struct timeval);
e71a4783
SH		 1127 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1128 v.tm.tv_sec = 0;
1129 v.tm.tv_usec = 0;
1130 } else {
1131 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1132 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1133 }
1134 break;
1135
1136 case SO_SNDTIMEO:
2a91525c 1137 lv = sizeof(struct timeval);
e71a4783
SH		 1138 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1139 v.tm.tv_sec = 0;
1140 v.tm.tv_usec = 0;
1141 } else {
1142 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1143 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1144 }
1145 break;
1da177e4 1146
e71a4783
SH		 1147 case SO_RCVLOWAT:
1148 v.val = sk->sk_rcvlowat;
1149 break;
1da177e4 1150
e71a4783 1151 case SO_SNDLOWAT:
2a91525c 1152 v.val = 1;
e71a4783 1153 break;
1da177e4 1154
e71a4783 1155 case SO_PASSCRED:
82981930 1156 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1157 break;
1da177e4 1158
e71a4783 1159 case SO_PEERCRED:
109f6e39
EB		 1160 {
1161 struct ucred peercred;
1162 if (len > sizeof(peercred))
1163 len = sizeof(peercred);
1164 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1165 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1166 return -EFAULT;
1167 goto lenout;
109f6e39 1168 }
1da177e4 1169
e71a4783
SH		 1170 case SO_PEERNAME:
1171 {
1172 char address[128];
1173
1174 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1175 return -ENOTCONN;
1176 if (lv < len)
1177 return -EINVAL;
1178 if (copy_to_user(optval, address, len))
1179 return -EFAULT;
1180 goto lenout;
1181 }
1da177e4 1182
e71a4783
SH		 1183 /* Dubious BSD thing... Probably nobody even uses it, but
1184 * the UNIX standard wants it for whatever reason... -DaveM
1185 */
1186 case SO_ACCEPTCONN:
1187 v.val = sk->sk_state == TCP_LISTEN;
1188 break;
1da177e4 1189
e71a4783 1190 case SO_PASSSEC:
82981930 1191 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1192 break;
877ce7c1 1193
e71a4783
SH		 1194 case SO_PEERSEC:
1195 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1196
4a19ec58
LAT		 1197 case SO_MARK:
1198 v.val = sk->sk_mark;
1199 break;
1200
3b885787 1201 case SO_RXQ_OVFL:
1b23a5df 1202 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1203 break;
1204
6e3e939f 1205 case SO_WIFI_STATUS:
1b23a5df 1206 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1207 break;
1208
ef64a54f
PE		 1209 case SO_PEEK_OFF:
1210 if (!sock->ops->set_peek_off)
1211 return -EOPNOTSUPP;
1212
1213 v.val = sk->sk_peek_off;
1214 break;
bc2f7996 1215 case SO_NOFCS:
1b23a5df 1216 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1217 break;
c91f6df2 1218
f7b86bfe 1219 case SO_BINDTODEVICE:
c91f6df2
BH		 1220 return sock_getbindtodevice(sk, optval, optlen, len);
1221
a8fc9277
PE		 1222 case SO_GET_FILTER:
1223 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1224 if (len < 0)
1225 return len;
1226
1227 goto lenout;
c91f6df2 1228
d59577b6
VB		 1229 case SO_LOCK_FILTER:
1230 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1231 break;
1232
ea02f941
MS		 1233 case SO_BPF_EXTENSIONS:
1234 v.val = bpf_tell_extensions();
1235 break;
1236
7d4c04fc
KJ		 1237 case SO_SELECT_ERR_QUEUE:
1238 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1239 break;
1240
e0d1095a 1241#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1242 case SO_BUSY_POLL:
dafcc438
ET		 1243 v.val = sk->sk_ll_usec;
1244 break;
1245#endif
1246
62748f32
ED		 1247 case SO_MAX_PACING_RATE:
1248 v.val = sk->sk_max_pacing_rate;
1249 break;
1250
2c8c56e1
ED		 1251 case SO_INCOMING_CPU:
1252 v.val = sk->sk_incoming_cpu;
1253 break;
1254
e71a4783 1255 default:
443b5991
YH		 1256 /* We implement the SO_SNDLOWAT etc to not be settable
1257 * (1003.1g 7).
1258 */
e71a4783 1259 return -ENOPROTOOPT;
1da177e4 1260 }
e71a4783 1261
1da177e4
LT		 1262 if (len > lv)
1263 len = lv;
1264 if (copy_to_user(optval, &v, len))
1265 return -EFAULT;
1266lenout:
4ec93edb
YH		 1267 if (put_user(len, optlen))
1268 return -EFAULT;
1269 return 0;
1da177e4
LT		 1270}
1271
a5b5bb9a
IM		 1272/*
1273 * Initialize an sk_lock.
1274 *
1275 * (We also register the sk_lock with the lock validator.)
1276 */
b6f99a21 1277static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1278{
ed07536e
PZ		 1279 sock_lock_init_class_and_name(sk,
1280 af_family_slock_key_strings[sk->sk_family],
1281 af_family_slock_keys + sk->sk_family,
1282 af_family_key_strings[sk->sk_family],
1283 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1284}
1285
4dc6dc71
ED		 1286/*
1287 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1288 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1289 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1290 */
f1a6c4da
PE		 1291static void sock_copy(struct sock *nsk, const struct sock *osk)
1292{
1293#ifdef CONFIG_SECURITY_NETWORK
1294 void *sptr = nsk->sk_security;
1295#endif
68835aba
ED		 1296 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1297
1298 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1299 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1300
f1a6c4da
PE		 1301#ifdef CONFIG_SECURITY_NETWORK
1302 nsk->sk_security = sptr;
1303 security_sk_clone(osk, nsk);
1304#endif
1305}
1306
fcbdf09d
OP		 1307void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1308{
1309 unsigned long nulls1, nulls2;
1310
1311 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1312 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1313 if (nulls1 > nulls2)
1314 swap(nulls1, nulls2);
1315
1316 if (nulls1 != 0)
1317 memset((char *)sk, 0, nulls1);
1318 memset((char *)sk + nulls1 + sizeof(void *), 0,
1319 nulls2 - nulls1 - sizeof(void *));
1320 memset((char *)sk + nulls2 + sizeof(void *), 0,
1321 size - nulls2 - sizeof(void *));
1322}
1323EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1324
2e4afe7b
PE		 1325static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1326 int family)
c308c1b2
PE		 1327{
1328 struct sock *sk;
1329 struct kmem_cache *slab;
1330
1331 slab = prot->slab;
e912b114
ED		 1332 if (slab != NULL) {
1333 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1334 if (!sk)
1335 return sk;
1336 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1337 if (prot->clear_sk)
1338 prot->clear_sk(sk, prot->obj_size);
1339 else
1340 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1341 }
fcbdf09d 1342 } else
c308c1b2
PE		 1343 sk = kmalloc(prot->obj_size, priority);
1344
2e4afe7b 1345 if (sk != NULL) {
a98b65a3
VN		 1346 kmemcheck_annotate_bitfield(sk, flags);
1347
2e4afe7b
PE		 1348 if (security_sk_alloc(sk, family, priority))
1349 goto out_free;
1350
1351 if (!try_module_get(prot->owner))
1352 goto out_free_sec;
e022f0b4 1353 sk_tx_queue_clear(sk);
bd1060a1 1354 cgroup_sk_alloc(&sk->sk_cgrp_data);
2e4afe7b
PE		 1355 }
1356
c308c1b2 1357 return sk;
2e4afe7b
PE		 1358
1359out_free_sec:
1360 security_sk_free(sk);
1361out_free:
1362 if (slab != NULL)
1363 kmem_cache_free(slab, sk);
1364 else
1365 kfree(sk);
1366 return NULL;
c308c1b2
PE		 1367}
1368
1369static void sk_prot_free(struct proto *prot, struct sock *sk)
1370{
1371 struct kmem_cache *slab;
2e4afe7b 1372 struct module *owner;
c308c1b2 1373
2e4afe7b 1374 owner = prot->owner;
c308c1b2 1375 slab = prot->slab;
2e4afe7b 1376
bd1060a1 1377 cgroup_sk_free(&sk->sk_cgrp_data);
2e4afe7b 1378 security_sk_free(sk);
c308c1b2
PE		 1379 if (slab != NULL)
1380 kmem_cache_free(slab, sk);
1381 else
1382 kfree(sk);
2e4afe7b 1383 module_put(owner);
c308c1b2
PE		 1384}
1385
1da177e4
LT		 1386/**
1387 * sk_alloc - All socket objects are allocated here
c4ea43c5 1388 * @net: the applicable net namespace
4dc3b16b
PP		 1389 * @family: protocol family
1390 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1391 * @prot: struct proto associated with this new sock instance
11aa9c28 1392 * @kern: is this to be a kernel socket?
1da177e4 1393 */
1b8d7ae4 1394struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1395 struct proto *prot, int kern)
1da177e4 1396{
c308c1b2 1397 struct sock *sk;
1da177e4 1398
154adbc8 1399 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1400 if (sk) {
154adbc8
PE		 1401 sk->sk_family = family;
1402 /*
1403 * See comment in struct sock definition to understand
1404 * why we need sk_prot_creator -acme
1405 */
1406 sk->sk_prot = sk->sk_prot_creator = prot;
1407 sock_lock_init(sk);
26abe143
EB		 1408 sk->sk_net_refcnt = kern ? 0 : 1;
1409 if (likely(sk->sk_net_refcnt))
1410 get_net(net);
1411 sock_net_set(sk, net);
d66ee058 1412 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1413
2a56a1fe
TH		 1414 sock_update_classid(&sk->sk_cgrp_data);
1415 sock_update_netprioidx(&sk->sk_cgrp_data);
1da177e4 1416 }
a79af59e 1417
2e4afe7b 1418 return sk;
1da177e4 1419}
2a91525c 1420EXPORT_SYMBOL(sk_alloc);
1da177e4 1421
a4298e45
ED		 1422/* Sockets having SOCK_RCU_FREE will call this function after one RCU
1423 * grace period. This is the case for UDP sockets and TCP listeners.
1424 */
1425static void __sk_destruct(struct rcu_head *head)
1da177e4 1426{
a4298e45 1427 struct sock *sk = container_of(head, struct sock, sk_rcu);
1da177e4 1428 struct sk_filter *filter;
1da177e4
LT		 1429
1430 if (sk->sk_destruct)
1431 sk->sk_destruct(sk);
1432
a898def2
PM		 1433 filter = rcu_dereference_check(sk->sk_filter,
1434 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1435 if (filter) {
309dd5fc 1436 sk_filter_uncharge(sk, filter);
a9b3cd7f 1437 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4 1438 }
538950a1
CG		 1439 if (rcu_access_pointer(sk->sk_reuseport_cb))
1440 reuseport_detach_sock(sk);
1da177e4 1441
08e29af3 1442 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1443
1444 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1445 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1446 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1447
109f6e39
EB		 1448 if (sk->sk_peer_cred)
1449 put_cred(sk->sk_peer_cred);
1450 put_pid(sk->sk_peer_pid);
26abe143
EB		 1451 if (likely(sk->sk_net_refcnt))
1452 put_net(sock_net(sk));
c308c1b2 1453 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1454}
2b85a34e 1455
a4298e45
ED		 1456void sk_destruct(struct sock *sk)
1457{
1458 if (sock_flag(sk, SOCK_RCU_FREE))
1459 call_rcu(&sk->sk_rcu, __sk_destruct);
1460 else
1461 __sk_destruct(&sk->sk_rcu);
1462}
1463
eb4cb008
CG		 1464static void __sk_free(struct sock *sk)
1465{
b922622e 1466 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1467 sock_diag_broadcast_destroy(sk);
1468 else
1469 sk_destruct(sk);
1470}
1471
2b85a34e
ED		 1472void sk_free(struct sock *sk)
1473{
1474 /*
25985edc 1475 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1476 * some packets are still in some tx queue.
1477 * If not null, sock_wfree() will call __sk_free(sk) later
1478 */
1479 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1480 __sk_free(sk);
1481}
2a91525c 1482EXPORT_SYMBOL(sk_free);
1da177e4 1483
e56c57d0
ED		 1484/**
1485 * sk_clone_lock - clone a socket, and lock its clone
1486 * @sk: the socket to clone
1487 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1488 *
1489 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1490 */
1491struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1492{
8fd1d178 1493 struct sock *newsk;
278571ba 1494 bool is_charged = true;
87d11ceb 1495
8fd1d178 1496 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1497 if (newsk != NULL) {
1498 struct sk_filter *filter;
1499
892c141e 1500 sock_copy(newsk, sk);
87d11ceb
ACM		 1501
1502 /* SANITY */
8a681736
SV		 1503 if (likely(newsk->sk_net_refcnt))
1504 get_net(sock_net(newsk));
87d11ceb
ACM		 1505 sk_node_init(&newsk->sk_node);
1506 sock_lock_init(newsk);
1507 bh_lock_sock(newsk);
fa438ccf 1508 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1509 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1510
1511 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1512 /*
1513 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1514 */
1515 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1516 atomic_set(&newsk->sk_omem_alloc, 0);
1517 skb_queue_head_init(&newsk->sk_receive_queue);
1518 skb_queue_head_init(&newsk->sk_write_queue);
1519
87d11ceb 1520 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1521 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1522 af_callback_keys + newsk->sk_family,
1523 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1524
1525 newsk->sk_dst_cache = NULL;
1526 newsk->sk_wmem_queued = 0;
1527 newsk->sk_forward_alloc = 0;
1528 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1529 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1530
1531 sock_reset_flag(newsk, SOCK_DONE);
1532 skb_queue_head_init(&newsk->sk_error_queue);
1533
0d7da9dd 1534 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1535 if (filter != NULL)
278571ba
AS		 1536 /* though it's an empty new sock, the charging may fail
1537 * if sysctl_optmem_max was changed between creation of
1538 * original socket and cloning
1539 */
1540 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1541
d188ba86 1542 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
87d11ceb
ACM		 1543 /* It is still raw copy of parent, so invalidate
1544 * destructor and make plain sk_free() */
1545 newsk->sk_destruct = NULL;
b0691c8e 1546 bh_unlock_sock(newsk);
87d11ceb
ACM		 1547 sk_free(newsk);
1548 newsk = NULL;
1549 goto out;
1550 }
fa463497 1551 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
87d11ceb
ACM		 1552
1553 newsk->sk_err = 0;
1554 newsk->sk_priority = 0;
2c8c56e1 1555 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1556 atomic64_set(&newsk->sk_cookie, 0);
4dc6dc71
ED		 1557 /*
1558 * Before updating sk_refcnt, we must commit prior changes to memory
1559 * (Documentation/RCU/rculist_nulls.txt for details)
1560 */
1561 smp_wmb();
87d11ceb
ACM		 1562 atomic_set(&newsk->sk_refcnt, 2);
1563
1564 /*
1565 * Increment the counter in the same struct proto as the master
1566 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1567 * is the same as sk->sk_prot->socks, as this field was copied
1568 * with memcpy).
1569 *
1570 * This _changes_ the previous behaviour, where
1571 * tcp_create_openreq_child always was incrementing the
1572 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1573 * to be taken into account in all callers. -acme
1574 */
1575 sk_refcnt_debug_inc(newsk);
972692e0 1576 sk_set_socket(newsk, NULL);
43815482 1577 newsk->sk_wq = NULL;
87d11ceb 1578
baac50bb 1579 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
3d596f7b 1580 sock_update_memcg(newsk);
f3f511e1 1581
87d11ceb 1582 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1583 sk_sockets_allocated_inc(newsk);
704da560 1584
080a270f
HFS		 1585 if (sock_needs_netstamp(sk) &&
1586 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1587 net_enable_timestamp();
87d11ceb
ACM		 1588 }
1589out:
1590 return newsk;
1591}
e56c57d0 1592EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1593
9958089a
AK		 1594void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1595{
d6a4e26a
ED		 1596 u32 max_segs = 1;
1597
6bd4f355 1598 sk_dst_set(sk, dst);
9958089a
AK		 1599 sk->sk_route_caps = dst->dev->features;
1600 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1601 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1602 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1603 if (sk_can_gso(sk)) {
82cc1a7a 1604 if (dst->header_len) {
9958089a 1605 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1606 } else {
9958089a 1607 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1608 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1609 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1610 }
9958089a 1611 }
d6a4e26a 1612 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1613}
1614EXPORT_SYMBOL_GPL(sk_setup_caps);
1615
1da177e4
LT		 1616/*
1617 * Simple resource managers for sockets.
1618 */
1619
1620
4ec93edb
YH		 1621/*
1622 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1623 */
1624void sock_wfree(struct sk_buff *skb)
1625{
1626 struct sock *sk = skb->sk;
d99927f4 1627 unsigned int len = skb->truesize;
1da177e4 1628
d99927f4
ED		 1629 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1630 /*
1631 * Keep a reference on sk_wmem_alloc, this will be released
1632 * after sk_write_space() call
1633 */
1634 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1635 sk->sk_write_space(sk);
d99927f4
ED		 1636 len = 1;
1637 }
2b85a34e 1638 /*
d99927f4
ED		 1639 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1640 * could not do because of in-flight packets
2b85a34e 1641 */
d99927f4 1642 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1643 __sk_free(sk);
1da177e4 1644}
2a91525c 1645EXPORT_SYMBOL(sock_wfree);
1da177e4 1646
9e17f8a4
ED		 1647void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1648{
1649 skb_orphan(skb);
1650 skb->sk = sk;
1651#ifdef CONFIG_INET
1652 if (unlikely(!sk_fullsock(sk))) {
1653 skb->destructor = sock_edemux;
1654 sock_hold(sk);
1655 return;
1656 }
1657#endif
1658 skb->destructor = sock_wfree;
1659 skb_set_hash_from_sk(skb, sk);
1660 /*
1661 * We used to take a refcount on sk, but following operation
1662 * is enough to guarantee sk_free() wont free this sock until
1663 * all in-flight packets are completed
1664 */
1665 atomic_add(skb->truesize, &sk->sk_wmem_alloc);
1666}
1667EXPORT_SYMBOL(skb_set_owner_w);
1668
f2f872f9
ED		 1669void skb_orphan_partial(struct sk_buff *skb)
1670{
1671 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1672 * so we do not completely orphan skb, but transfert all
1673 * accounted bytes but one, to avoid unexpected reorders.
1674 */
1675 if (skb->destructor == sock_wfree
1676#ifdef CONFIG_INET
1677 || skb->destructor == tcp_wfree
1678#endif
1679 ) {
1680 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1681 skb->truesize = 1;
1682 } else {
1683 skb_orphan(skb);
1684 }
1685}
1686EXPORT_SYMBOL(skb_orphan_partial);
1687
4ec93edb
YH		 1688/*
1689 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1690 */
1691void sock_rfree(struct sk_buff *skb)
1692{
1693 struct sock *sk = skb->sk;
d361fd59 1694 unsigned int len = skb->truesize;
1da177e4 1695
d361fd59
ED		 1696 atomic_sub(len, &sk->sk_rmem_alloc);
1697 sk_mem_uncharge(sk, len);
1da177e4 1698}
2a91525c 1699EXPORT_SYMBOL(sock_rfree);
1da177e4 1700
7768eed8
OH		 1701/*
1702 * Buffer destructor for skbs that are not used directly in read or write
1703 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1704 */
62bccb8c
AD		 1705void sock_efree(struct sk_buff *skb)
1706{
1707 sock_put(skb->sk);
1708}
1709EXPORT_SYMBOL(sock_efree);
1710
976d0201 1711kuid_t sock_i_uid(struct sock *sk)
1da177e4 1712{
976d0201 1713 kuid_t uid;
1da177e4 1714
f064af1e 1715 read_lock_bh(&sk->sk_callback_lock);
976d0201 1716 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1717 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1718 return uid;
1719}
2a91525c 1720EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1721
1722unsigned long sock_i_ino(struct sock *sk)
1723{
1724 unsigned long ino;
1725
f064af1e 1726 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1727 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1728 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1729 return ino;
1730}
2a91525c 1731EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1732
1733/*
1734 * Allocate a skb from the socket's send buffer.
1735 */
86a76caf 1736struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1737 gfp_t priority)
1da177e4
LT		 1738{
1739 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1740 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1741 if (skb) {
1742 skb_set_owner_w(skb, sk);
1743 return skb;
1744 }
1745 }
1746 return NULL;
1747}
2a91525c 1748EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1749
4ec93edb 1750/*
1da177e4 1751 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1752 */
dd0fc66f 1753void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1754{
95c96174 1755 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1756 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1757 void *mem;
1758 /* First do the add, to avoid the race if kmalloc
4ec93edb 1759 * might sleep.
1da177e4
LT		 1760 */
1761 atomic_add(size, &sk->sk_omem_alloc);
1762 mem = kmalloc(size, priority);
1763 if (mem)
1764 return mem;
1765 atomic_sub(size, &sk->sk_omem_alloc);
1766 }
1767 return NULL;
1768}
2a91525c 1769EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1770
79e88659
DB		 1771/* Free an option memory block. Note, we actually want the inline
1772 * here as this allows gcc to detect the nullify and fold away the
1773 * condition entirely.
1da177e4 1774 */
79e88659
DB		 1775static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1776 const bool nullify)
1da177e4 1777{
e53da5fb
DM		 1778 if (WARN_ON_ONCE(!mem))
1779 return;
79e88659
DB		 1780 if (nullify)
1781 kzfree(mem);
1782 else
1783 kfree(mem);
1da177e4
LT		 1784 atomic_sub(size, &sk->sk_omem_alloc);
1785}
79e88659
DB		 1786
1787void sock_kfree_s(struct sock *sk, void *mem, int size)
1788{
1789 __sock_kfree_s(sk, mem, size, false);
1790}
2a91525c 1791EXPORT_SYMBOL(sock_kfree_s);
1da177e4 1792
79e88659
DB		 1793void sock_kzfree_s(struct sock *sk, void *mem, int size)
1794{
1795 __sock_kfree_s(sk, mem, size, true);
1796}
1797EXPORT_SYMBOL(sock_kzfree_s);
1798
1da177e4
LT		 1799/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1800 I think, these locks should be removed for datagram sockets.
1801 */
2a91525c 1802static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1803{
1804 DEFINE_WAIT(wait);
1805
9cd3e072 1806 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1da177e4
LT		 1807 for (;;) {
1808 if (!timeo)
1809 break;
1810 if (signal_pending(current))
1811 break;
1812 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1813 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1814 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1815 break;
1816 if (sk->sk_shutdown & SEND_SHUTDOWN)
1817 break;
1818 if (sk->sk_err)
1819 break;
1820 timeo = schedule_timeout(timeo);
1821 }
aa395145 1822 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1823 return timeo;
1824}
1825
1826
1827/*
1828 * Generic send/receive buffer handlers
1829 */
1830
4cc7f68d
HX		 1831struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1832 unsigned long data_len, int noblock,
28d64271 1833 int *errcode, int max_page_order)
1da177e4 1834{
2e4e4410 1835 struct sk_buff *skb;
1da177e4
LT		 1836 long timeo;
1837 int err;
1838
1da177e4 1839 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 1840 for (;;) {
1da177e4
LT		 1841 err = sock_error(sk);
1842 if (err != 0)
1843 goto failure;
1844
1845 err = -EPIPE;
1846 if (sk->sk_shutdown & SEND_SHUTDOWN)
1847 goto failure;
1848
2e4e4410
ED		 1849 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1850 break;
28d64271 1851
9cd3e072 1852 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2e4e4410
ED		 1853 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1854 err = -EAGAIN;
1855 if (!timeo)
1da177e4 1856 goto failure;
2e4e4410
ED		 1857 if (signal_pending(current))
1858 goto interrupted;
1859 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 1860 }
2e4e4410
ED		 1861 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1862 errcode, sk->sk_allocation);
1863 if (skb)
1864 skb_set_owner_w(skb, sk);
1da177e4
LT		 1865 return skb;
1866
1867interrupted:
1868 err = sock_intr_errno(timeo);
1869failure:
1870 *errcode = err;
1871 return NULL;
1872}
4cc7f68d 1873EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1874
4ec93edb 1875struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1876 int noblock, int *errcode)
1877{
28d64271 1878 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1879}
2a91525c 1880EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1881
39771b12
WB		 1882int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1883 struct sockcm_cookie *sockc)
1884{
3dd17e63
SHY		 1885 u32 tsflags;
1886
39771b12
WB		 1887 switch (cmsg->cmsg_type) {
1888 case SO_MARK:
1889 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1890 return -EPERM;
1891 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1892 return -EINVAL;
1893 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
1894 break;
3dd17e63
SHY		 1895 case SO_TIMESTAMPING:
1896 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
1897 return -EINVAL;
1898
1899 tsflags = *(u32 *)CMSG_DATA(cmsg);
1900 if (tsflags & ~SOF_TIMESTAMPING_TX_RECORD_MASK)
1901 return -EINVAL;
1902
1903 sockc->tsflags &= ~SOF_TIMESTAMPING_TX_RECORD_MASK;
1904 sockc->tsflags |= tsflags;
1905 break;
39771b12
WB		 1906 default:
1907 return -EINVAL;
1908 }
1909 return 0;
1910}
1911EXPORT_SYMBOL(__sock_cmsg_send);
1912
f28ea365
EJ		 1913int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1914 struct sockcm_cookie *sockc)
1915{
1916 struct cmsghdr *cmsg;
39771b12 1917 int ret;
f28ea365
EJ		 1918
1919 for_each_cmsghdr(cmsg, msg) {
1920 if (!CMSG_OK(msg, cmsg))
1921 return -EINVAL;
1922 if (cmsg->cmsg_level != SOL_SOCKET)
1923 continue;
39771b12
WB		 1924 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
1925 if (ret)
1926 return ret;
f28ea365
EJ		 1927 }
1928 return 0;
1929}
1930EXPORT_SYMBOL(sock_cmsg_send);
1931
5640f768
ED		 1932/* On 32bit arches, an skb frag is limited to 2^15 */
1933#define SKB_FRAG_PAGE_ORDER get_order(32768)
1934
400dfd3a
ED		 1935/**
1936 * skb_page_frag_refill - check that a page_frag contains enough room
1937 * @sz: minimum size of the fragment we want to get
1938 * @pfrag: pointer to page_frag
82d5e2b8 1939 * @gfp: priority for memory allocation
400dfd3a
ED		 1940 *
1941 * Note: While this allocator tries to use high order pages, there is
1942 * no guarantee that allocations succeed. Therefore, @sz MUST be
1943 * less or equal than PAGE_SIZE.
1944 */
d9b2938a 1945bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 1946{
5640f768 1947 if (pfrag->page) {
fe896d18 1948 if (page_ref_count(pfrag->page) == 1) {
5640f768
ED		 1949 pfrag->offset = 0;
1950 return true;
1951 }
400dfd3a 1952 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1953 return true;
1954 put_page(pfrag->page);
1955 }
1956
d9b2938a
ED		 1957 pfrag->offset = 0;
1958 if (SKB_FRAG_PAGE_ORDER) {
d0164adc
MG		 1959 /* Avoid direct reclaim but allow kswapd to wake */
1960 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
1961 __GFP_COMP | __GFP_NOWARN |
1962 __GFP_NORETRY,
d9b2938a 1963 SKB_FRAG_PAGE_ORDER);
5640f768 1964 if (likely(pfrag->page)) {
d9b2938a 1965 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 1966 return true;
1967 }
d9b2938a
ED		 1968 }
1969 pfrag->page = alloc_page(gfp);
1970 if (likely(pfrag->page)) {
1971 pfrag->size = PAGE_SIZE;
1972 return true;
1973 }
400dfd3a
ED		 1974 return false;
1975}
1976EXPORT_SYMBOL(skb_page_frag_refill);
1977
1978bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1979{
1980 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1981 return true;
1982
5640f768
ED		 1983 sk_enter_memory_pressure(sk);
1984 sk_stream_moderate_sndbuf(sk);
1985 return false;
1986}
1987EXPORT_SYMBOL(sk_page_frag_refill);
1988
1da177e4 1989static void __lock_sock(struct sock *sk)
f39234d6
NK		 1990 __releases(&sk->sk_lock.slock)
1991 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1992{
1993 DEFINE_WAIT(wait);
1994
e71a4783 1995 for (;;) {
1da177e4
LT		 1996 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1997 TASK_UNINTERRUPTIBLE);
1998 spin_unlock_bh(&sk->sk_lock.slock);
1999 schedule();
2000 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 2001 if (!sock_owned_by_user(sk))
1da177e4
LT		 2002 break;
2003 }
2004 finish_wait(&sk->sk_lock.wq, &wait);
2005}
2006
2007static void __release_sock(struct sock *sk)
f39234d6
NK		 2008 __releases(&sk->sk_lock.slock)
2009 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 2010{
2011 struct sk_buff *skb = sk->sk_backlog.head;
2012
2013 do {
2014 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
2015 bh_unlock_sock(sk);
2016
2017 do {
2018 struct sk_buff *next = skb->next;
2019
e4cbb02a 2020 prefetch(next);
7fee226a 2021 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 2022 skb->next = NULL;
c57943a1 2023 sk_backlog_rcv(sk, skb);
1da177e4
LT		 2024
2025 /*
2026 * We are in process context here with softirqs
2027 * disabled, use cond_resched_softirq() to preempt.
2028 * This is safe to do because we've taken the backlog
2029 * queue private:
2030 */
2031 cond_resched_softirq();
2032
2033 skb = next;
2034 } while (skb != NULL);
2035
2036 bh_lock_sock(sk);
e71a4783 2037 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 2038
2039 /*
2040 * Doing the zeroing here guarantee we can not loop forever
2041 * while a wild producer attempts to flood us.
2042 */
2043 sk->sk_backlog.len = 0;
1da177e4
LT		 2044}
2045
2046/**
2047 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 2048 * @sk: sock to wait on
2049 * @timeo: for how long
dfbafc99 2050 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 2051 *
2052 * Now socket state including sk->sk_err is changed only under lock,
2053 * hence we may omit checks after joining wait queue.
2054 * We check receive queue before schedule() only as optimization;
2055 * it is very likely that release_sock() added new data.
2056 */
dfbafc99 2057int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4
LT		 2058{
2059 int rc;
2060 DEFINE_WAIT(wait);
2061
aa395145 2062 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
9cd3e072 2063 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
dfbafc99 2064 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);
9cd3e072 2065 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
aa395145 2066 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2067 return rc;
2068}
1da177e4
LT		 2069EXPORT_SYMBOL(sk_wait_data);
2070
3ab224be
HA		 2071/**
2072 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2073 * @sk: socket
2074 * @size: memory size to allocate
2075 * @kind: allocation type
2076 *
2077 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2078 * rmem allocation. This function assumes that protocols which have
2079 * memory_pressure use sk_wmem_queued as write buffer accounting.
2080 */
2081int __sk_mem_schedule(struct sock *sk, int size, int kind)
2082{
2083 struct proto *prot = sk->sk_prot;
2084 int amt = sk_mem_pages(size);
8d987e5c 2085 long allocated;
3ab224be
HA		 2086
2087 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2088
e805605c
JW		 2089 allocated = sk_memory_allocated_add(sk, amt);
2090
baac50bb
JW		 2091 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2092 !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
e805605c 2093 goto suppress_allocation;
3ab224be
HA		 2094
2095 /* Under limit. */
e805605c 2096 if (allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2097 sk_leave_memory_pressure(sk);
3ab224be
HA		 2098 return 1;
2099 }
2100
e805605c
JW		 2101 /* Under pressure. */
2102 if (allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2103 sk_enter_memory_pressure(sk);
3ab224be 2104
e805605c
JW		 2105 /* Over hard limit. */
2106 if (allocated > sk_prot_mem_limits(sk, 2))
3ab224be
HA		 2107 goto suppress_allocation;
2108
2109 /* guarantee minimum buffer size under pressure */
2110 if (kind == SK_MEM_RECV) {
2111 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2112 return 1;
180d8cd9 2113
3ab224be
HA		 2114 } else { /* SK_MEM_SEND */
2115 if (sk->sk_type == SOCK_STREAM) {
2116 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2117 return 1;
2118 } else if (atomic_read(&sk->sk_wmem_alloc) <
2119 prot->sysctl_wmem[0])
2120 return 1;
2121 }
2122
180d8cd9 2123 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2124 int alloc;
2125
180d8cd9 2126 if (!sk_under_memory_pressure(sk))
1748376b 2127 return 1;
180d8cd9
GC		 2128 alloc = sk_sockets_allocated_read_positive(sk);
2129 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2130 sk_mem_pages(sk->sk_wmem_queued +
2131 atomic_read(&sk->sk_rmem_alloc) +
2132 sk->sk_forward_alloc))
2133 return 1;
2134 }
2135
2136suppress_allocation:
2137
2138 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2139 sk_stream_moderate_sndbuf(sk);
2140
2141 /* Fail only if socket is _under_ its sndbuf.
2142 * In this case we cannot block, so that we have to fail.
2143 */
2144 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2145 return 1;
2146 }
2147
3847ce32
SM		 2148 trace_sock_exceed_buf_limit(sk, prot, allocated);
2149
3ab224be
HA		 2150 /* Alas. Undo changes. */
2151 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2152
0e90b31f 2153 sk_memory_allocated_sub(sk, amt);
180d8cd9 2154
baac50bb
JW		 2155 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2156 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
e805605c 2157
3ab224be
HA		 2158 return 0;
2159}
3ab224be
HA		 2160EXPORT_SYMBOL(__sk_mem_schedule);
2161
2162/**
69dba9bb 2163 * __sk_mem_reclaim - reclaim memory_allocated
3ab224be 2164 * @sk: socket
1a24e04e 2165 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
3ab224be 2166 */
1a24e04e 2167void __sk_mem_reclaim(struct sock *sk, int amount)
3ab224be 2168{
1a24e04e
ED		 2169 amount >>= SK_MEM_QUANTUM_SHIFT;
2170 sk_memory_allocated_sub(sk, amount);
2171 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
3ab224be 2172
baac50bb
JW		 2173 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2174 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
e805605c 2175
180d8cd9
GC		 2176 if (sk_under_memory_pressure(sk) &&
2177 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2178 sk_leave_memory_pressure(sk);
3ab224be 2179}
3ab224be
HA		 2180EXPORT_SYMBOL(__sk_mem_reclaim);
2181
2182
1da177e4
LT		 2183/*
2184 * Set of default routines for initialising struct proto_ops when
2185 * the protocol does not support a particular function. In certain
2186 * cases where it makes no sense for a protocol to have a "do nothing"
2187 * function, some default processing is provided.
2188 */
2189
2190int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2191{
2192 return -EOPNOTSUPP;
2193}
2a91525c 2194EXPORT_SYMBOL(sock_no_bind);
1da177e4 2195
4ec93edb 2196int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2197 int len, int flags)
2198{
2199 return -EOPNOTSUPP;
2200}
2a91525c 2201EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2202
2203int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2204{
2205 return -EOPNOTSUPP;
2206}
2a91525c 2207EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2208
2209int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2210{
2211 return -EOPNOTSUPP;
2212}
2a91525c 2213EXPORT_SYMBOL(sock_no_accept);
1da177e4 2214
4ec93edb 2215int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2216 int *len, int peer)
2217{
2218 return -EOPNOTSUPP;
2219}
2a91525c 2220EXPORT_SYMBOL(sock_no_getname);
1da177e4 2221
2a91525c 2222unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2223{
2224 return 0;
2225}
2a91525c 2226EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2227
2228int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2229{
2230 return -EOPNOTSUPP;
2231}
2a91525c 2232EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2233
2234int sock_no_listen(struct socket *sock, int backlog)
2235{
2236 return -EOPNOTSUPP;
2237}
2a91525c 2238EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2239
2240int sock_no_shutdown(struct socket *sock, int how)
2241{
2242 return -EOPNOTSUPP;
2243}
2a91525c 2244EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2245
2246int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2247 char __user *optval, unsigned int optlen)
1da177e4
LT		 2248{
2249 return -EOPNOTSUPP;
2250}
2a91525c 2251EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2252
2253int sock_no_getsockopt(struct socket *sock, int level, int optname,
2254 char __user *optval, int __user *optlen)
2255{
2256 return -EOPNOTSUPP;
2257}
2a91525c 2258EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2259
1b784140 2260int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2261{
2262 return -EOPNOTSUPP;
2263}
2a91525c 2264EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2265
1b784140
YX		 2266int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2267 int flags)
1da177e4
LT		 2268{
2269 return -EOPNOTSUPP;
2270}
2a91525c 2271EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2272
2273int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2274{
2275 /* Mirror missing mmap method error code */
2276 return -ENODEV;
2277}
2a91525c 2278EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2279
2280ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2281{
2282 ssize_t res;
2283 struct msghdr msg = {.msg_flags = flags};
2284 struct kvec iov;
2285 char *kaddr = kmap(page);
2286 iov.iov_base = kaddr + offset;
2287 iov.iov_len = size;
2288 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2289 kunmap(page);
2290 return res;
2291}
2a91525c 2292EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2293
2294/*
2295 * Default Socket Callbacks
2296 */
2297
2298static void sock_def_wakeup(struct sock *sk)
2299{
43815482
ED		 2300 struct socket_wq *wq;
2301
2302 rcu_read_lock();
2303 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2304 if (skwq_has_sleeper(wq))
43815482
ED		 2305 wake_up_interruptible_all(&wq->wait);
2306 rcu_read_unlock();
1da177e4
LT		 2307}
2308
2309static void sock_def_error_report(struct sock *sk)
2310{
43815482
ED		 2311 struct socket_wq *wq;
2312
2313 rcu_read_lock();
2314 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2315 if (skwq_has_sleeper(wq))
43815482 2316 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2317 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2318 rcu_read_unlock();
1da177e4
LT		 2319}
2320
676d2369 2321static void sock_def_readable(struct sock *sk)
1da177e4 2322{
43815482
ED		 2323 struct socket_wq *wq;
2324
2325 rcu_read_lock();
2326 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2327 if (skwq_has_sleeper(wq))
2c6607c6 2328 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2329 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2330 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2331 rcu_read_unlock();
1da177e4
LT		 2332}
2333
2334static void sock_def_write_space(struct sock *sk)
2335{
43815482
ED		 2336 struct socket_wq *wq;
2337
2338 rcu_read_lock();
1da177e4
LT		 2339
2340 /* Do not wake up a writer until he can make "significant"
2341 * progress. --DaveM
2342 */
e71a4783 2343 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482 2344 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2345 if (skwq_has_sleeper(wq))
43815482 2346 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2347 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2348
2349 /* Should agree with poll, otherwise some programs break */
2350 if (sock_writeable(sk))
8d8ad9d7 2351 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2352 }
2353
43815482 2354 rcu_read_unlock();
1da177e4
LT		 2355}
2356
2357static void sock_def_destruct(struct sock *sk)
2358{
1da177e4
LT		 2359}
2360
2361void sk_send_sigurg(struct sock *sk)
2362{
2363 if (sk->sk_socket && sk->sk_socket->file)
2364 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2365 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2366}
2a91525c 2367EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2368
2369void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2370 unsigned long expires)
2371{
2372 if (!mod_timer(timer, expires))
2373 sock_hold(sk);
2374}
1da177e4
LT		 2375EXPORT_SYMBOL(sk_reset_timer);
2376
2377void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2378{
25cc4ae9 2379 if (del_timer(timer))
1da177e4
LT		 2380 __sock_put(sk);
2381}
1da177e4
LT		 2382EXPORT_SYMBOL(sk_stop_timer);
2383
2384void sock_init_data(struct socket *sock, struct sock *sk)
2385{
2386 skb_queue_head_init(&sk->sk_receive_queue);
2387 skb_queue_head_init(&sk->sk_write_queue);
2388 skb_queue_head_init(&sk->sk_error_queue);
2389
2390 sk->sk_send_head = NULL;
2391
2392 init_timer(&sk->sk_timer);
4ec93edb 2393
1da177e4
LT		 2394 sk->sk_allocation = GFP_KERNEL;
2395 sk->sk_rcvbuf = sysctl_rmem_default;
2396 sk->sk_sndbuf = sysctl_wmem_default;
2397 sk->sk_state = TCP_CLOSE;
972692e0 2398 sk_set_socket(sk, sock);
1da177e4
LT		 2399
2400 sock_set_flag(sk, SOCK_ZAPPED);
2401
e71a4783 2402 if (sock) {
1da177e4 2403 sk->sk_type = sock->type;
43815482 2404 sk->sk_wq = sock->wq;
1da177e4
LT		 2405 sock->sk = sk;
2406 } else
43815482 2407 sk->sk_wq = NULL;
1da177e4 2408
1da177e4 2409 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2410 lockdep_set_class_and_name(&sk->sk_callback_lock,
2411 af_callback_keys + sk->sk_family,
2412 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2413
2414 sk->sk_state_change = sock_def_wakeup;
2415 sk->sk_data_ready = sock_def_readable;
2416 sk->sk_write_space = sock_def_write_space;
2417 sk->sk_error_report = sock_def_error_report;
2418 sk->sk_destruct = sock_def_destruct;
2419
5640f768
ED		 2420 sk->sk_frag.page = NULL;
2421 sk->sk_frag.offset = 0;
ef64a54f 2422 sk->sk_peek_off = -1;
1da177e4 2423
109f6e39
EB		 2424 sk->sk_peer_pid = NULL;
2425 sk->sk_peer_cred = NULL;
1da177e4
LT		 2426 sk->sk_write_pending = 0;
2427 sk->sk_rcvlowat = 1;
2428 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2429 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2430
f37f0afb 2431 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2432
e0d1095a 2433#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2434 sk->sk_napi_id = 0;
64b0dc51 2435 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2436#endif
2437
62748f32 2438 sk->sk_max_pacing_rate = ~0U;
7eec4174 2439 sk->sk_pacing_rate = ~0U;
70da268b 2440 sk->sk_incoming_cpu = -1;
4dc6dc71
ED		 2441 /*
2442 * Before updating sk_refcnt, we must commit prior changes to memory
2443 * (Documentation/RCU/rculist_nulls.txt for details)
2444 */
2445 smp_wmb();
1da177e4 2446 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2447 atomic_set(&sk->sk_drops, 0);
1da177e4 2448}
2a91525c 2449EXPORT_SYMBOL(sock_init_data);
1da177e4 2450
b5606c2d 2451void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2452{
2453 might_sleep();
a5b5bb9a 2454 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2455 if (sk->sk_lock.owned)
1da177e4 2456 __lock_sock(sk);
d2e9117c 2457 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2458 spin_unlock(&sk->sk_lock.slock);
2459 /*
2460 * The sk_lock has mutex_lock() semantics here:
2461 */
fcc70d5f 2462 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2463 local_bh_enable();
1da177e4 2464}
fcc70d5f 2465EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2466
b5606c2d 2467void release_sock(struct sock *sk)
1da177e4 2468{
a5b5bb9a
IM		 2469 /*
2470 * The sk_lock has mutex_unlock() semantics:
2471 */
2472 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2473
2474 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2475 if (sk->sk_backlog.tail)
2476 __release_sock(sk);
46d3ceab 2477
c3f9b018
ED		 2478 /* Warning : release_cb() might need to release sk ownership,
2479 * ie call sock_release_ownership(sk) before us.
2480 */
46d3ceab
ED		 2481 if (sk->sk_prot->release_cb)
2482 sk->sk_prot->release_cb(sk);
2483
c3f9b018 2484 sock_release_ownership(sk);
a5b5bb9a
IM		 2485 if (waitqueue_active(&sk->sk_lock.wq))
2486 wake_up(&sk->sk_lock.wq);
2487 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2488}
2489EXPORT_SYMBOL(release_sock);
2490
8a74ad60
ED		 2491/**
2492 * lock_sock_fast - fast version of lock_sock
2493 * @sk: socket
2494 *
2495 * This version should be used for very small section, where process wont block
2496 * return false if fast path is taken
2497 * sk_lock.slock locked, owned = 0, BH disabled
2498 * return true if slow path is taken
2499 * sk_lock.slock unlocked, owned = 1, BH enabled
2500 */
2501bool lock_sock_fast(struct sock *sk)
2502{
2503 might_sleep();
2504 spin_lock_bh(&sk->sk_lock.slock);
2505
2506 if (!sk->sk_lock.owned)
2507 /*
2508 * Note : We must disable BH
2509 */
2510 return false;
2511
2512 __lock_sock(sk);
2513 sk->sk_lock.owned = 1;
2514 spin_unlock(&sk->sk_lock.slock);
2515 /*
2516 * The sk_lock has mutex_lock() semantics here:
2517 */
2518 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2519 local_bh_enable();
2520 return true;
2521}
2522EXPORT_SYMBOL(lock_sock_fast);
2523
1da177e4 2524int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2525{
b7aa0bf7 2526 struct timeval tv;
1da177e4 2527 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2528 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2529 tv = ktime_to_timeval(sk->sk_stamp);
2530 if (tv.tv_sec == -1)
1da177e4 2531 return -ENOENT;
b7aa0bf7
ED		 2532 if (tv.tv_sec == 0) {
2533 sk->sk_stamp = ktime_get_real();
2534 tv = ktime_to_timeval(sk->sk_stamp);
2535 }
2536 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2537}
1da177e4
LT		 2538EXPORT_SYMBOL(sock_get_timestamp);
2539
ae40eb1e
ED		 2540int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2541{
2542 struct timespec ts;
2543 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2544 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2545 ts = ktime_to_timespec(sk->sk_stamp);
2546 if (ts.tv_sec == -1)
2547 return -ENOENT;
2548 if (ts.tv_sec == 0) {
2549 sk->sk_stamp = ktime_get_real();
2550 ts = ktime_to_timespec(sk->sk_stamp);
2551 }
2552 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2553}
2554EXPORT_SYMBOL(sock_get_timestampns);
2555
20d49473 2556void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2557{
20d49473 2558 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2559 unsigned long previous_flags = sk->sk_flags;
2560
20d49473
PO		 2561 sock_set_flag(sk, flag);
2562 /*
2563 * we just set one of the two flags which require net
2564 * time stamping, but time stamping might have been on
2565 * already because of the other one
2566 */
080a270f
HFS		 2567 if (sock_needs_netstamp(sk) &&
2568 !(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2569 net_enable_timestamp();
1da177e4
LT		 2570 }
2571}
1da177e4 2572
cb820f8e
RC		 2573int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2574 int level, int type)
2575{
2576 struct sock_exterr_skb *serr;
364a9e93 2577 struct sk_buff *skb;
cb820f8e
RC		 2578 int copied, err;
2579
2580 err = -EAGAIN;
364a9e93 2581 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2582 if (skb == NULL)
2583 goto out;
2584
2585 copied = skb->len;
2586 if (copied > len) {
2587 msg->msg_flags |= MSG_TRUNC;
2588 copied = len;
2589 }
51f3d02b 2590 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2591 if (err)
2592 goto out_free_skb;
2593
2594 sock_recv_timestamp(msg, sk, skb);
2595
2596 serr = SKB_EXT_ERR(skb);
2597 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2598
2599 msg->msg_flags |= MSG_ERRQUEUE;
2600 err = copied;
2601
cb820f8e
RC		 2602out_free_skb:
2603 kfree_skb(skb);
2604out:
2605 return err;
2606}
2607EXPORT_SYMBOL(sock_recv_errqueue);
2608
1da177e4
LT		 2609/*
2610 * Get a socket option on an socket.
2611 *
2612 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2613 * asynchronous errors should be reported by getsockopt. We assume
2614 * this means if you specify SO_ERROR (otherwise whats the point of it).
2615 */
2616int sock_common_getsockopt(struct socket *sock, int level, int optname,
2617 char __user *optval, int __user *optlen)
2618{
2619 struct sock *sk = sock->sk;
2620
2621 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2622}
1da177e4
LT		 2623EXPORT_SYMBOL(sock_common_getsockopt);
2624
3fdadf7d 2625#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2626int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2627 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2628{
2629 struct sock *sk = sock->sk;
2630
1e51f951 2631 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2632 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2633 optval, optlen);
3fdadf7d
DM		 2634 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2635}
2636EXPORT_SYMBOL(compat_sock_common_getsockopt);
2637#endif
2638
1b784140
YX		 2639int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2640 int flags)
1da177e4
LT		 2641{
2642 struct sock *sk = sock->sk;
2643 int addr_len = 0;
2644 int err;
2645
1b784140 2646 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2647 flags & ~MSG_DONTWAIT, &addr_len);
2648 if (err >= 0)
2649 msg->msg_namelen = addr_len;
2650 return err;
2651}
1da177e4
LT		 2652EXPORT_SYMBOL(sock_common_recvmsg);
2653
2654/*
2655 * Set socket options on an inet socket.
2656 */
2657int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2658 char __user *optval, unsigned int optlen)
1da177e4
LT		 2659{
2660 struct sock *sk = sock->sk;
2661
2662 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2663}
1da177e4
LT		 2664EXPORT_SYMBOL(sock_common_setsockopt);
2665
3fdadf7d 2666#ifdef CONFIG_COMPAT
543d9cfe 2667int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2668 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2669{
2670 struct sock *sk = sock->sk;
2671
543d9cfe
ACM		 2672 if (sk->sk_prot->compat_setsockopt != NULL)
2673 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2674 optval, optlen);
3fdadf7d
DM		 2675 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2676}
2677EXPORT_SYMBOL(compat_sock_common_setsockopt);
2678#endif
2679
1da177e4
LT		 2680void sk_common_release(struct sock *sk)
2681{
2682 if (sk->sk_prot->destroy)
2683 sk->sk_prot->destroy(sk);
2684
2685 /*
2686 * Observation: when sock_common_release is called, processes have
2687 * no access to socket. But net still has.
2688 * Step one, detach it from networking:
2689 *
2690 * A. Remove from hash tables.
2691 */
2692
2693 sk->sk_prot->unhash(sk);
2694
2695 /*
2696 * In this point socket cannot receive new packets, but it is possible
2697 * that some packets are in flight because some CPU runs receiver and
2698 * did hash table lookup before we unhashed socket. They will achieve
2699 * receive queue and will be purged by socket destructor.
2700 *
2701 * Also we still have packets pending on receive queue and probably,
2702 * our own packets waiting in device queues. sock_destroy will drain
2703 * receive queue, but transmitted packets will delay socket destruction
2704 * until the last reference will be released.
2705 */
2706
2707 sock_orphan(sk);
2708
2709 xfrm_sk_free_policy(sk);
2710
e6848976 2711 sk_refcnt_debug_release(sk);
5640f768
ED		 2712
2713 if (sk->sk_frag.page) {
2714 put_page(sk->sk_frag.page);
2715 sk->sk_frag.page = NULL;
2716 }
2717
1da177e4
LT		 2718 sock_put(sk);
2719}
1da177e4
LT		 2720EXPORT_SYMBOL(sk_common_release);
2721
13ff3d6f
PE		 2722#ifdef CONFIG_PROC_FS
2723#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2724struct prot_inuse {
2725 int val[PROTO_INUSE_NR];
2726};
13ff3d6f
PE		 2727
2728static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2729
2730#ifdef CONFIG_NET_NS
2731void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2732{
d6d9ca0f 2733 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2734}
2735EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2736
2737int sock_prot_inuse_get(struct net *net, struct proto *prot)
2738{
2739 int cpu, idx = prot->inuse_idx;
2740 int res = 0;
2741
2742 for_each_possible_cpu(cpu)
2743 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2744
2745 return res >= 0 ? res : 0;
2746}
2747EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2748
2c8c1e72 2749static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2750{
2751 net->core.inuse = alloc_percpu(struct prot_inuse);
2752 return net->core.inuse ? 0 : -ENOMEM;
2753}
2754
2c8c1e72 2755static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2756{
2757 free_percpu(net->core.inuse);
2758}
2759
2760static struct pernet_operations net_inuse_ops = {
2761 .init = sock_inuse_init_net,
2762 .exit = sock_inuse_exit_net,
2763};
2764
2765static __init int net_inuse_init(void)
2766{
2767 if (register_pernet_subsys(&net_inuse_ops))
2768 panic("Cannot initialize net inuse counters");
2769
2770 return 0;
2771}
2772
2773core_initcall(net_inuse_init);
2774#else
1338d466
PE		 2775static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2776
c29a0bc4 2777void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2778{
d6d9ca0f 2779 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2780}
2781EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2782
c29a0bc4 2783int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2784{
2785 int cpu, idx = prot->inuse_idx;
2786 int res = 0;
2787
2788 for_each_possible_cpu(cpu)
2789 res += per_cpu(prot_inuse, cpu).val[idx];
2790
2791 return res >= 0 ? res : 0;
2792}
2793EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2794#endif
13ff3d6f
PE		 2795
2796static void assign_proto_idx(struct proto *prot)
2797{
2798 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2799
2800 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2801 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2802 return;
2803 }
2804
2805 set_bit(prot->inuse_idx, proto_inuse_idx);
2806}
2807
2808static void release_proto_idx(struct proto *prot)
2809{
2810 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2811 clear_bit(prot->inuse_idx, proto_inuse_idx);
2812}
2813#else
2814static inline void assign_proto_idx(struct proto *prot)
2815{
2816}
2817
2818static inline void release_proto_idx(struct proto *prot)
2819{
2820}
2821#endif
2822
0159dfd3
ED		 2823static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2824{
2825 if (!rsk_prot)
2826 return;
2827 kfree(rsk_prot->slab_name);
2828 rsk_prot->slab_name = NULL;
adf78eda
JL		 2829 kmem_cache_destroy(rsk_prot->slab);
2830 rsk_prot->slab = NULL;
0159dfd3
ED		 2831}
2832
2833static int req_prot_init(const struct proto *prot)
2834{
2835 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2836
2837 if (!rsk_prot)
2838 return 0;
2839
2840 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2841 prot->name);
2842 if (!rsk_prot->slab_name)
2843 return -ENOMEM;
2844
2845 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2846 rsk_prot->obj_size, 0,
e96f78ab 2847 prot->slab_flags, NULL);
0159dfd3
ED		 2848
2849 if (!rsk_prot->slab) {
2850 pr_crit("%s: Can't create request sock SLAB cache!\n",
2851 prot->name);
2852 return -ENOMEM;
2853 }
2854 return 0;
2855}
2856
b733c007
PE		 2857int proto_register(struct proto *prot, int alloc_slab)
2858{
1da177e4
LT		 2859 if (alloc_slab) {
2860 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2861 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2862 NULL);
1da177e4
LT		 2863
2864 if (prot->slab == NULL) {
e005d193
JP		 2865 pr_crit("%s: Can't create sock SLAB cache!\n",
2866 prot->name);
60e7663d 2867 goto out;
1da177e4 2868 }
2e6599cb 2869
0159dfd3
ED		 2870 if (req_prot_init(prot))
2871 goto out_free_request_sock_slab;
8feaf0c0 2872
6d6ee43e 2873 if (prot->twsk_prot != NULL) {
faf23422 2874 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2875
7e56b5d6 2876 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2877 goto out_free_request_sock_slab;
2878
6d6ee43e 2879 prot->twsk_prot->twsk_slab =
7e56b5d6 2880 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2881 prot->twsk_prot->twsk_obj_size,
3ab5aee7 2882 0,
52db70dc 2883 prot->slab_flags,
20c2df83 2884 NULL);
6d6ee43e 2885 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2886 goto out_free_timewait_sock_slab_name;
2887 }
1da177e4
LT		 2888 }
2889
36b77a52 2890 mutex_lock(&proto_list_mutex);
1da177e4 2891 list_add(&prot->node, &proto_list);
13ff3d6f 2892 assign_proto_idx(prot);
36b77a52 2893 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2894 return 0;
2895
8feaf0c0 2896out_free_timewait_sock_slab_name:
7e56b5d6 2897 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 2898out_free_request_sock_slab:
0159dfd3
ED		 2899 req_prot_cleanup(prot->rsk_prot);
2900
2e6599cb
ACM		 2901 kmem_cache_destroy(prot->slab);
2902 prot->slab = NULL;
b733c007
PE		 2903out:
2904 return -ENOBUFS;
1da177e4 2905}
1da177e4
LT		 2906EXPORT_SYMBOL(proto_register);
2907
2908void proto_unregister(struct proto *prot)
2909{
36b77a52 2910 mutex_lock(&proto_list_mutex);
13ff3d6f 2911 release_proto_idx(prot);
0a3f4358 2912 list_del(&prot->node);
36b77a52 2913 mutex_unlock(&proto_list_mutex);
1da177e4 2914
adf78eda
JL		 2915 kmem_cache_destroy(prot->slab);
2916 prot->slab = NULL;
1da177e4 2917
0159dfd3 2918 req_prot_cleanup(prot->rsk_prot);
2e6599cb 2919
6d6ee43e 2920 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2921 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2922 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2923 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2924 }
1da177e4 2925}
1da177e4
LT		 2926EXPORT_SYMBOL(proto_unregister);
2927
2928#ifdef CONFIG_PROC_FS
1da177e4 2929static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2930 __acquires(proto_list_mutex)
1da177e4 2931{
36b77a52 2932 mutex_lock(&proto_list_mutex);
60f0438a 2933 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2934}
2935
2936static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2937{
60f0438a 2938 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2939}
2940
2941static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2942 __releases(proto_list_mutex)
1da177e4 2943{
36b77a52 2944 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2945}
2946
2947static char proto_method_implemented(const void *method)
2948{
2949 return method == NULL ? 'n' : 'y';
2950}
180d8cd9
GC		 2951static long sock_prot_memory_allocated(struct proto *proto)
2952{
cb75a36c 2953 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2954}
2955
2956static char *sock_prot_memory_pressure(struct proto *proto)
2957{
2958 return proto->memory_pressure != NULL ?
2959 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2960}
1da177e4
LT		 2961
2962static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2963{
180d8cd9 2964
8d987e5c 2965 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2966 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2967 proto->name,
2968 proto->obj_size,
14e943db 2969 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2970 sock_prot_memory_allocated(proto),
2971 sock_prot_memory_pressure(proto),
1da177e4
LT		 2972 proto->max_header,
2973 proto->slab == NULL ? "no" : "yes",
2974 module_name(proto->owner),
2975 proto_method_implemented(proto->close),
2976 proto_method_implemented(proto->connect),
2977 proto_method_implemented(proto->disconnect),
2978 proto_method_implemented(proto->accept),
2979 proto_method_implemented(proto->ioctl),
2980 proto_method_implemented(proto->init),
2981 proto_method_implemented(proto->destroy),
2982 proto_method_implemented(proto->shutdown),
2983 proto_method_implemented(proto->setsockopt),
2984 proto_method_implemented(proto->getsockopt),
2985 proto_method_implemented(proto->sendmsg),
2986 proto_method_implemented(proto->recvmsg),
2987 proto_method_implemented(proto->sendpage),
2988 proto_method_implemented(proto->bind),
2989 proto_method_implemented(proto->backlog_rcv),
2990 proto_method_implemented(proto->hash),
2991 proto_method_implemented(proto->unhash),
2992 proto_method_implemented(proto->get_port),
2993 proto_method_implemented(proto->enter_memory_pressure));
2994}
2995
2996static int proto_seq_show(struct seq_file *seq, void *v)
2997{
60f0438a 2998 if (v == &proto_list)
1da177e4
LT		 2999 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3000 "protocol",
3001 "size",
3002 "sockets",
3003 "memory",
3004 "press",
3005 "maxhdr",
3006 "slab",
3007 "module",
3008 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3009 else
60f0438a 3010 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 3011 return 0;
3012}
3013
f690808e 3014static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 3015 .start = proto_seq_start,
3016 .next = proto_seq_next,
3017 .stop = proto_seq_stop,
3018 .show = proto_seq_show,
3019};
3020
3021static int proto_seq_open(struct inode *inode, struct file *file)
3022{
14e943db
ED		 3023 return seq_open_net(inode, file, &proto_seq_ops,
3024 sizeof(struct seq_net_private));
1da177e4
LT		 3025}
3026
9a32144e 3027static const struct file_operations proto_seq_fops = {
1da177e4
LT		 3028 .owner = THIS_MODULE,
3029 .open = proto_seq_open,
3030 .read = seq_read,
3031 .llseek = seq_lseek,
14e943db
ED		 3032 .release = seq_release_net,
3033};
3034
3035static __net_init int proto_init_net(struct net *net)
3036{
d4beaa66 3037 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 3038 return -ENOMEM;
3039
3040 return 0;
3041}
3042
3043static __net_exit void proto_exit_net(struct net *net)
3044{
ece31ffd 3045 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 3046}
3047
3048
3049static __net_initdata struct pernet_operations proto_net_ops = {
3050 .init = proto_init_net,
3051 .exit = proto_exit_net,
1da177e4
LT		 3052};
3053
3054static int __init proto_init(void)
3055{
14e943db 3056 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 3057}
3058
3059subsys_initcall(proto_init);
3060
3061#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 1.112243 seconds and 5 git commands to generate.