projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
cxgb4: Simplify PCIe Completion Timeout setting
[linux-2.6-block.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>
f1083048 105#include <linux/sched/mm.h>
1da177e4
LT		 106#include <linux/timer.h>
107#include <linux/string.h>
108#include <linux/sockios.h>
109#include <linux/net.h>
110#include <linux/mm.h>
111#include <linux/slab.h>
112#include <linux/interrupt.h>
113#include <linux/poll.h>
114#include <linux/tcp.h>
115#include <linux/init.h>
a1f8e7f7 116#include <linux/highmem.h>
3f551f94 117#include <linux/user_namespace.h>
c5905afb 118#include <linux/static_key.h>
3969eb38 119#include <linux/memcontrol.h>
8c1ae10d 120#include <linux/prefetch.h>
1da177e4 121
7c0f6ba6 122#include <linux/uaccess.h>
1da177e4
LT		 123
124#include <linux/netdevice.h>
125#include <net/protocol.h>
126#include <linux/skbuff.h>
457c4cbc 127#include <net/net_namespace.h>
2e6599cb 128#include <net/request_sock.h>
1da177e4 129#include <net/sock.h>
20d49473 130#include <linux/net_tstamp.h>
1da177e4
LT		 131#include <net/xfrm.h>
132#include <linux/ipsec.h>
f8451725 133#include <net/cls_cgroup.h>
5bc1421e 134#include <net/netprio_cgroup.h>
eb4cb008 135#include <linux/sock_diag.h>
1da177e4
LT		 136
137#include <linux/filter.h>
538950a1 138#include <net/sock_reuseport.h>
1da177e4 139
3847ce32
SM		 140#include <trace/events/sock.h>
141
1da177e4 142#include <net/tcp.h>
076bb0c8 143#include <net/busy_poll.h>
06021292 144
36b77a52 145static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 146static LIST_HEAD(proto_list);
147
a3b299da
EB		 148/**
149 * sk_ns_capable - General socket capability test
150 * @sk: Socket to use a capability on or through
151 * @user_ns: The user namespace of the capability to use
152 * @cap: The capability to use
153 *
154 * Test to see if the opener of the socket had when the socket was
155 * created and the current process has the capability @cap in the user
156 * namespace @user_ns.
157 */
158bool sk_ns_capable(const struct sock *sk,
159 struct user_namespace *user_ns, int cap)
160{
161 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
162 ns_capable(user_ns, cap);
163}
164EXPORT_SYMBOL(sk_ns_capable);
165
166/**
167 * sk_capable - Socket global capability test
168 * @sk: Socket to use a capability on or through
e793c0f7 169 * @cap: The global capability to use
a3b299da
EB		 170 *
171 * Test to see if the opener of the socket had when the socket was
172 * created and the current process has the capability @cap in all user
173 * namespaces.
174 */
175bool sk_capable(const struct sock *sk, int cap)
176{
177 return sk_ns_capable(sk, &init_user_ns, cap);
178}
179EXPORT_SYMBOL(sk_capable);
180
181/**
182 * sk_net_capable - Network namespace socket capability test
183 * @sk: Socket to use a capability on or through
184 * @cap: The capability to use
185 *
e793c0f7 186 * Test to see if the opener of the socket had when the socket was created
a3b299da
EB		 187 * and the current process has the capability @cap over the network namespace
188 * the socket is a member of.
189 */
190bool sk_net_capable(const struct sock *sk, int cap)
191{
192 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
193}
194EXPORT_SYMBOL(sk_net_capable);
195
da21f24d
IM		 196/*
197 * Each address family might have different locking rules, so we have
cdfbabfb
DH		 198 * one slock key per address family and separate keys for internal and
199 * userspace sockets.
da21f24d 200 */
a5b5bb9a 201static struct lock_class_key af_family_keys[AF_MAX];
cdfbabfb 202static struct lock_class_key af_family_kern_keys[AF_MAX];
a5b5bb9a 203static struct lock_class_key af_family_slock_keys[AF_MAX];
cdfbabfb 204static struct lock_class_key af_family_kern_slock_keys[AF_MAX];
a5b5bb9a 205
a5b5bb9a
IM		 206/*
207 * Make lock validator output more readable. (we pre-construct these
208 * strings build-time, so that runtime initialization of socket
209 * locks is fast):
210 */
cdfbabfb
DH		 211
212#define _sock_locks(x) \
213 x "AF_UNSPEC", x "AF_UNIX" , x "AF_INET" , \
214 x "AF_AX25" , x "AF_IPX" , x "AF_APPLETALK", \
215 x "AF_NETROM", x "AF_BRIDGE" , x "AF_ATMPVC" , \
216 x "AF_X25" , x "AF_INET6" , x "AF_ROSE" , \
217 x "AF_DECnet", x "AF_NETBEUI" , x "AF_SECURITY" , \
218 x "AF_KEY" , x "AF_NETLINK" , x "AF_PACKET" , \
219 x "AF_ASH" , x "AF_ECONET" , x "AF_ATMSVC" , \
220 x "AF_RDS" , x "AF_SNA" , x "AF_IRDA" , \
221 x "AF_PPPOX" , x "AF_WANPIPE" , x "AF_LLC" , \
222 x "27" , x "28" , x "AF_CAN" , \
223 x "AF_TIPC" , x "AF_BLUETOOTH", x "IUCV" , \
224 x "AF_RXRPC" , x "AF_ISDN" , x "AF_PHONET" , \
225 x "AF_IEEE802154", x "AF_CAIF" , x "AF_ALG" , \
226 x "AF_NFC" , x "AF_VSOCK" , x "AF_KCM" , \
227 x "AF_QIPCRTR", x "AF_SMC" , x "AF_MAX"
228
36cbd3dc 229static const char *const af_family_key_strings[AF_MAX+1] = {
cdfbabfb 230 _sock_locks("sk_lock-")
a5b5bb9a 231};
36cbd3dc 232static const char *const af_family_slock_key_strings[AF_MAX+1] = {
cdfbabfb 233 _sock_locks("slock-")
a5b5bb9a 234};
36cbd3dc 235static const char *const af_family_clock_key_strings[AF_MAX+1] = {
cdfbabfb
DH		 236 _sock_locks("clock-")
237};
238
239static const char *const af_family_kern_key_strings[AF_MAX+1] = {
240 _sock_locks("k-sk_lock-")
241};
242static const char *const af_family_kern_slock_key_strings[AF_MAX+1] = {
243 _sock_locks("k-slock-")
244};
245static const char *const af_family_kern_clock_key_strings[AF_MAX+1] = {
246 _sock_locks("k-clock-")
443aef0e 247};
581319c5
PA		 248static const char *const af_family_rlock_key_strings[AF_MAX+1] = {
249 "rlock-AF_UNSPEC", "rlock-AF_UNIX" , "rlock-AF_INET" ,
250 "rlock-AF_AX25" , "rlock-AF_IPX" , "rlock-AF_APPLETALK",
251 "rlock-AF_NETROM", "rlock-AF_BRIDGE" , "rlock-AF_ATMPVC" ,
252 "rlock-AF_X25" , "rlock-AF_INET6" , "rlock-AF_ROSE" ,
253 "rlock-AF_DECnet", "rlock-AF_NETBEUI" , "rlock-AF_SECURITY" ,
254 "rlock-AF_KEY" , "rlock-AF_NETLINK" , "rlock-AF_PACKET" ,
255 "rlock-AF_ASH" , "rlock-AF_ECONET" , "rlock-AF_ATMSVC" ,
256 "rlock-AF_RDS" , "rlock-AF_SNA" , "rlock-AF_IRDA" ,
257 "rlock-AF_PPPOX" , "rlock-AF_WANPIPE" , "rlock-AF_LLC" ,
258 "rlock-27" , "rlock-28" , "rlock-AF_CAN" ,
259 "rlock-AF_TIPC" , "rlock-AF_BLUETOOTH", "rlock-AF_IUCV" ,
260 "rlock-AF_RXRPC" , "rlock-AF_ISDN" , "rlock-AF_PHONET" ,
261 "rlock-AF_IEEE802154", "rlock-AF_CAIF" , "rlock-AF_ALG" ,
262 "rlock-AF_NFC" , "rlock-AF_VSOCK" , "rlock-AF_KCM" ,
263 "rlock-AF_QIPCRTR", "rlock-AF_SMC" , "rlock-AF_MAX"
264};
265static const char *const af_family_wlock_key_strings[AF_MAX+1] = {
266 "wlock-AF_UNSPEC", "wlock-AF_UNIX" , "wlock-AF_INET" ,
267 "wlock-AF_AX25" , "wlock-AF_IPX" , "wlock-AF_APPLETALK",
268 "wlock-AF_NETROM", "wlock-AF_BRIDGE" , "wlock-AF_ATMPVC" ,
269 "wlock-AF_X25" , "wlock-AF_INET6" , "wlock-AF_ROSE" ,
270 "wlock-AF_DECnet", "wlock-AF_NETBEUI" , "wlock-AF_SECURITY" ,
271 "wlock-AF_KEY" , "wlock-AF_NETLINK" , "wlock-AF_PACKET" ,
272 "wlock-AF_ASH" , "wlock-AF_ECONET" , "wlock-AF_ATMSVC" ,
273 "wlock-AF_RDS" , "wlock-AF_SNA" , "wlock-AF_IRDA" ,
274 "wlock-AF_PPPOX" , "wlock-AF_WANPIPE" , "wlock-AF_LLC" ,
275 "wlock-27" , "wlock-28" , "wlock-AF_CAN" ,
276 "wlock-AF_TIPC" , "wlock-AF_BLUETOOTH", "wlock-AF_IUCV" ,
277 "wlock-AF_RXRPC" , "wlock-AF_ISDN" , "wlock-AF_PHONET" ,
278 "wlock-AF_IEEE802154", "wlock-AF_CAIF" , "wlock-AF_ALG" ,
279 "wlock-AF_NFC" , "wlock-AF_VSOCK" , "wlock-AF_KCM" ,
280 "wlock-AF_QIPCRTR", "wlock-AF_SMC" , "wlock-AF_MAX"
281};
282static const char *const af_family_elock_key_strings[AF_MAX+1] = {
283 "elock-AF_UNSPEC", "elock-AF_UNIX" , "elock-AF_INET" ,
284 "elock-AF_AX25" , "elock-AF_IPX" , "elock-AF_APPLETALK",
285 "elock-AF_NETROM", "elock-AF_BRIDGE" , "elock-AF_ATMPVC" ,
286 "elock-AF_X25" , "elock-AF_INET6" , "elock-AF_ROSE" ,
287 "elock-AF_DECnet", "elock-AF_NETBEUI" , "elock-AF_SECURITY" ,
288 "elock-AF_KEY" , "elock-AF_NETLINK" , "elock-AF_PACKET" ,
289 "elock-AF_ASH" , "elock-AF_ECONET" , "elock-AF_ATMSVC" ,
290 "elock-AF_RDS" , "elock-AF_SNA" , "elock-AF_IRDA" ,
291 "elock-AF_PPPOX" , "elock-AF_WANPIPE" , "elock-AF_LLC" ,
292 "elock-27" , "elock-28" , "elock-AF_CAN" ,
293 "elock-AF_TIPC" , "elock-AF_BLUETOOTH", "elock-AF_IUCV" ,
294 "elock-AF_RXRPC" , "elock-AF_ISDN" , "elock-AF_PHONET" ,
295 "elock-AF_IEEE802154", "elock-AF_CAIF" , "elock-AF_ALG" ,
296 "elock-AF_NFC" , "elock-AF_VSOCK" , "elock-AF_KCM" ,
297 "elock-AF_QIPCRTR", "elock-AF_SMC" , "elock-AF_MAX"
298};
da21f24d
IM		 299
300/*
581319c5 301 * sk_callback_lock and sk queues locking rules are per-address-family,
da21f24d
IM		 302 * so split the lock classes by using a per-AF key:
303 */
304static struct lock_class_key af_callback_keys[AF_MAX];
581319c5
PA		 305static struct lock_class_key af_rlock_keys[AF_MAX];
306static struct lock_class_key af_wlock_keys[AF_MAX];
307static struct lock_class_key af_elock_keys[AF_MAX];
cdfbabfb 308static struct lock_class_key af_kern_callback_keys[AF_MAX];
da21f24d 309
1da177e4 310/* Run time adjustable parameters. */
ab32ea5d 311__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 312EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 313__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 314EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 315__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
316__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 317
25985edc 318/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 319int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 320EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 321
b245be1f
WB		 322int sysctl_tstamp_allow_data __read_mostly = 1;
323
c93bdd0e
MG		 324struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
325EXPORT_SYMBOL_GPL(memalloc_socks);
326
7cb02404
MG		 327/**
328 * sk_set_memalloc - sets %SOCK_MEMALLOC
329 * @sk: socket to set it on
330 *
331 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
332 * It's the responsibility of the admin to adjust min_free_kbytes
333 * to meet the requirements
334 */
335void sk_set_memalloc(struct sock *sk)
336{
337 sock_set_flag(sk, SOCK_MEMALLOC);
338 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 339 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 340}
341EXPORT_SYMBOL_GPL(sk_set_memalloc);
342
343void sk_clear_memalloc(struct sock *sk)
344{
345 sock_reset_flag(sk, SOCK_MEMALLOC);
346 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 347 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 348
349 /*
350 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
5d753610
MG		 351 * progress of swapping. SOCK_MEMALLOC may be cleared while
352 * it has rmem allocations due to the last swapfile being deactivated
353 * but there is a risk that the socket is unusable due to exceeding
354 * the rmem limits. Reclaim the reserves and obey rmem limits again.
c76562b6 355 */
5d753610 356 sk_mem_reclaim(sk);
7cb02404
MG		 357}
358EXPORT_SYMBOL_GPL(sk_clear_memalloc);
359
b4b9e355
MG		 360int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
361{
362 int ret;
f1083048 363 unsigned int noreclaim_flag;
b4b9e355
MG		 364
365 /* these should have been dropped before queueing */
366 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
367
f1083048 368 noreclaim_flag = memalloc_noreclaim_save();
b4b9e355 369 ret = sk->sk_backlog_rcv(sk, skb);
f1083048 370 memalloc_noreclaim_restore(noreclaim_flag);
b4b9e355
MG		 371
372 return ret;
373}
374EXPORT_SYMBOL(__sk_backlog_rcv);
375
1da177e4
LT		 376static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
377{
378 struct timeval tv;
379
380 if (optlen < sizeof(tv))
381 return -EINVAL;
382 if (copy_from_user(&tv, optval, sizeof(tv)))
383 return -EFAULT;
ba78073e
VA		 384 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
385 return -EDOM;
1da177e4 386
ba78073e 387 if (tv.tv_sec < 0) {
6f11df83
AM		 388 static int warned __read_mostly;
389
ba78073e 390 *timeo_p = 0;
50aab54f 391 if (warned < 10 && net_ratelimit()) {
ba78073e 392 warned++;
e005d193
JP		 393 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
394 __func__, current->comm, task_pid_nr(current));
50aab54f 395 }
ba78073e
VA		 396 return 0;
397 }
1da177e4
LT		 398 *timeo_p = MAX_SCHEDULE_TIMEOUT;
399 if (tv.tv_sec == 0 && tv.tv_usec == 0)
400 return 0;
401 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
8ccde4c5 402 *timeo_p = tv.tv_sec * HZ + DIV_ROUND_UP(tv.tv_usec, USEC_PER_SEC / HZ);
1da177e4
LT		 403 return 0;
404}
405
406static void sock_warn_obsolete_bsdism(const char *name)
407{
408 static int warned;
409 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 410 if (strcmp(warncomm, current->comm) && warned < 5) {
411 strcpy(warncomm, current->comm);
e005d193
JP		 412 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
413 warncomm, name);
1da177e4
LT		 414 warned++;
415 }
416}
417
080a270f
HFS		 418static bool sock_needs_netstamp(const struct sock *sk)
419{
420 switch (sk->sk_family) {
421 case AF_UNSPEC:
422 case AF_UNIX:
423 return false;
424 default:
425 return true;
426 }
427}
428
08e29af3 429static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 430{
08e29af3
ED		 431 if (sk->sk_flags & flags) {
432 sk->sk_flags &= ~flags;
080a270f
HFS		 433 if (sock_needs_netstamp(sk) &&
434 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 435 net_disable_timestamp();
1da177e4
LT		 436 }
437}
438
439
e6afc8ac 440int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
f0088a50 441{
3b885787
NH		 442 unsigned long flags;
443 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 444
0fd7bac6 445 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 446 atomic_inc(&sk->sk_drops);
3847ce32 447 trace_sock_rcvqueue_full(sk, skb);
766e9037 448 return -ENOMEM;
f0088a50
DV		 449 }
450
c76562b6 451 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 452 atomic_inc(&sk->sk_drops);
453 return -ENOBUFS;
3ab224be
HA		 454 }
455
f0088a50
DV		 456 skb->dev = NULL;
457 skb_set_owner_r(skb, sk);
49ad9599 458
7fee226a
ED		 459 /* we escape from rcu protected region, make sure we dont leak
460 * a norefcounted dst
461 */
462 skb_dst_force(skb);
463
3b885787 464 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 465 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 466 __skb_queue_tail(list, skb);
467 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 468
469 if (!sock_flag(sk, SOCK_DEAD))
676d2369 470 sk->sk_data_ready(sk);
766e9037 471 return 0;
f0088a50 472}
e6afc8ac 473EXPORT_SYMBOL(__sock_queue_rcv_skb);
474
475int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
476{
477 int err;
478
479 err = sk_filter(sk, skb);
480 if (err)
481 return err;
482
483 return __sock_queue_rcv_skb(sk, skb);
484}
f0088a50
DV		 485EXPORT_SYMBOL(sock_queue_rcv_skb);
486
4f0c40d9 487int __sk_receive_skb(struct sock *sk, struct sk_buff *skb,
c3f24cfb 488 const int nested, unsigned int trim_cap, bool refcounted)
f0088a50
DV		 489{
490 int rc = NET_RX_SUCCESS;
491
4f0c40d9 492 if (sk_filter_trim_cap(sk, skb, trim_cap))
f0088a50
DV		 493 goto discard_and_relse;
494
495 skb->dev = NULL;
496
274f482d 497 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 498 atomic_inc(&sk->sk_drops);
499 goto discard_and_relse;
500 }
58a5a7b9
ACM		 501 if (nested)
502 bh_lock_sock_nested(sk);
503 else
504 bh_lock_sock(sk);
a5b5bb9a
IM		 505 if (!sock_owned_by_user(sk)) {
506 /*
507 * trylock + unlock semantics:
508 */
509 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
510
c57943a1 511 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 512
513 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 514 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 515 bh_unlock_sock(sk);
516 atomic_inc(&sk->sk_drops);
517 goto discard_and_relse;
518 }
519
f0088a50
DV		 520 bh_unlock_sock(sk);
521out:
c3f24cfb
ED		 522 if (refcounted)
523 sock_put(sk);
f0088a50
DV		 524 return rc;
525discard_and_relse:
526 kfree_skb(skb);
527 goto out;
528}
4f0c40d9 529EXPORT_SYMBOL(__sk_receive_skb);
f0088a50
DV		 530
531struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
532{
b6c6712a 533 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 534
535 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 536 sk_tx_queue_clear(sk);
9b8805a3 537 sk->sk_dst_pending_confirm = 0;
a9b3cd7f 538 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 539 dst_release(dst);
540 return NULL;
541 }
542
543 return dst;
544}
545EXPORT_SYMBOL(__sk_dst_check);
546
547struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
548{
549 struct dst_entry *dst = sk_dst_get(sk);
550
551 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
552 sk_dst_reset(sk);
553 dst_release(dst);
554 return NULL;
555 }
556
557 return dst;
558}
559EXPORT_SYMBOL(sk_dst_check);
560
c91f6df2
BH		 561static int sock_setbindtodevice(struct sock *sk, char __user *optval,
562 int optlen)
4878809f
DM		 563{
564 int ret = -ENOPROTOOPT;
565#ifdef CONFIG_NETDEVICES
3b1e0a65 566 struct net *net = sock_net(sk);
4878809f
DM		 567 char devname[IFNAMSIZ];
568 int index;
569
570 /* Sorry... */
571 ret = -EPERM;
5e1fccc0 572 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 573 goto out;
574
575 ret = -EINVAL;
576 if (optlen < 0)
577 goto out;
578
579 /* Bind this socket to a particular device like "eth0",
580 * as specified in the passed interface name. If the
581 * name is "" or the option length is zero the socket
582 * is not bound.
583 */
584 if (optlen > IFNAMSIZ - 1)
585 optlen = IFNAMSIZ - 1;
586 memset(devname, 0, sizeof(devname));
587
588 ret = -EFAULT;
589 if (copy_from_user(devname, optval, optlen))
590 goto out;
591
000ba2e4
DM		 592 index = 0;
593 if (devname[0] != '\0') {
bf8e56bf 594 struct net_device *dev;
4878809f 595
bf8e56bf
ED		 596 rcu_read_lock();
597 dev = dev_get_by_name_rcu(net, devname);
598 if (dev)
599 index = dev->ifindex;
600 rcu_read_unlock();
4878809f
DM		 601 ret = -ENODEV;
602 if (!dev)
603 goto out;
4878809f
DM		 604 }
605
606 lock_sock(sk);
607 sk->sk_bound_dev_if = index;
608 sk_dst_reset(sk);
609 release_sock(sk);
610
611 ret = 0;
612
613out:
614#endif
615
616 return ret;
617}
618
c91f6df2
BH		 619static int sock_getbindtodevice(struct sock *sk, char __user *optval,
620 int __user *optlen, int len)
621{
622 int ret = -ENOPROTOOPT;
623#ifdef CONFIG_NETDEVICES
624 struct net *net = sock_net(sk);
c91f6df2 625 char devname[IFNAMSIZ];
c91f6df2
BH		 626
627 if (sk->sk_bound_dev_if == 0) {
628 len = 0;
629 goto zero;
630 }
631
632 ret = -EINVAL;
633 if (len < IFNAMSIZ)
634 goto out;
635
5dbe7c17
NS		 636 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
637 if (ret)
c91f6df2 638 goto out;
c91f6df2
BH		 639
640 len = strlen(devname) + 1;
641
642 ret = -EFAULT;
643 if (copy_to_user(optval, devname, len))
644 goto out;
645
646zero:
647 ret = -EFAULT;
648 if (put_user(len, optlen))
649 goto out;
650
651 ret = 0;
652
653out:
654#endif
655
656 return ret;
657}
658
c0ef877b
PE		 659static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
660{
661 if (valbool)
662 sock_set_flag(sk, bit);
663 else
664 sock_reset_flag(sk, bit);
665}
666
f60e5990 667bool sk_mc_loop(struct sock *sk)
668{
669 if (dev_recursion_level())
670 return false;
671 if (!sk)
672 return true;
673 switch (sk->sk_family) {
674 case AF_INET:
675 return inet_sk(sk)->mc_loop;
676#if IS_ENABLED(CONFIG_IPV6)
677 case AF_INET6:
678 return inet6_sk(sk)->mc_loop;
679#endif
680 }
681 WARN_ON(1);
682 return true;
683}
684EXPORT_SYMBOL(sk_mc_loop);
685
1da177e4
LT		 686/*
687 * This is meant for all protocols to use and covers goings on
688 * at the socket level. Everything here is generic.
689 */
690
691int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 692 char __user *optval, unsigned int optlen)
1da177e4 693{
2a91525c 694 struct sock *sk = sock->sk;
1da177e4
LT		 695 int val;
696 int valbool;
697 struct linger ling;
698 int ret = 0;
4ec93edb 699
1da177e4
LT		 700 /*
701 * Options without arguments
702 */
703
4878809f 704 if (optname == SO_BINDTODEVICE)
c91f6df2 705 return sock_setbindtodevice(sk, optval, optlen);
4878809f 706
e71a4783
SH		 707 if (optlen < sizeof(int))
708 return -EINVAL;
4ec93edb 709
1da177e4
LT		 710 if (get_user(val, (int __user *)optval))
711 return -EFAULT;
4ec93edb 712
2a91525c 713 valbool = val ? 1 : 0;
1da177e4
LT		 714
715 lock_sock(sk);
716
2a91525c 717 switch (optname) {
e71a4783 718 case SO_DEBUG:
2a91525c 719 if (val && !capable(CAP_NET_ADMIN))
e71a4783 720 ret = -EACCES;
2a91525c 721 else
c0ef877b 722 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 723 break;
724 case SO_REUSEADDR:
4a17fd52 725 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 726 break;
055dc21a
TH		 727 case SO_REUSEPORT:
728 sk->sk_reuseport = valbool;
729 break;
e71a4783 730 case SO_TYPE:
49c794e9 731 case SO_PROTOCOL:
0d6038ee 732 case SO_DOMAIN:
e71a4783
SH		 733 case SO_ERROR:
734 ret = -ENOPROTOOPT;
735 break;
736 case SO_DONTROUTE:
c0ef877b 737 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 738 break;
739 case SO_BROADCAST:
740 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
741 break;
742 case SO_SNDBUF:
743 /* Don't error on this BSD doesn't and if you think
82981930
ED		 744 * about it this is right. Otherwise apps have to
745 * play 'guess the biggest size' games. RCVBUF/SNDBUF
746 * are treated in BSD as hints
747 */
748 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 749set_sndbuf:
e71a4783 750 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
b98b0bc8 751 sk->sk_sndbuf = max_t(int, val * 2, SOCK_MIN_SNDBUF);
82981930 752 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 753 sk->sk_write_space(sk);
754 break;
1da177e4 755
e71a4783
SH		 756 case SO_SNDBUFFORCE:
757 if (!capable(CAP_NET_ADMIN)) {
758 ret = -EPERM;
759 break;
760 }
761 goto set_sndbuf;
b0573dea 762
e71a4783
SH		 763 case SO_RCVBUF:
764 /* Don't error on this BSD doesn't and if you think
82981930
ED		 765 * about it this is right. Otherwise apps have to
766 * play 'guess the biggest size' games. RCVBUF/SNDBUF
767 * are treated in BSD as hints
768 */
769 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 770set_rcvbuf:
e71a4783
SH		 771 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
772 /*
773 * We double it on the way in to account for
774 * "struct sk_buff" etc. overhead. Applications
775 * assume that the SO_RCVBUF setting they make will
776 * allow that much actual data to be received on that
777 * socket.
778 *
779 * Applications are unaware that "struct sk_buff" and
780 * other overheads allocate from the receive buffer
781 * during socket buffer allocation.
782 *
783 * And after considering the possible alternatives,
784 * returning the value we actually used in getsockopt
785 * is the most desirable behavior.
786 */
b98b0bc8 787 sk->sk_rcvbuf = max_t(int, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 788 break;
789
790 case SO_RCVBUFFORCE:
791 if (!capable(CAP_NET_ADMIN)) {
792 ret = -EPERM;
1da177e4 793 break;
e71a4783
SH		 794 }
795 goto set_rcvbuf;
1da177e4 796
e71a4783 797 case SO_KEEPALIVE:
4b9d07a4
UB		 798 if (sk->sk_prot->keepalive)
799 sk->sk_prot->keepalive(sk, valbool);
e71a4783
SH		 800 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
801 break;
802
803 case SO_OOBINLINE:
804 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
805 break;
806
807 case SO_NO_CHECK:
28448b80 808 sk->sk_no_check_tx = valbool;
e71a4783
SH		 809 break;
810
811 case SO_PRIORITY:
5e1fccc0
EB		 812 if ((val >= 0 && val <= 6) ||
813 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 814 sk->sk_priority = val;
815 else
816 ret = -EPERM;
817 break;
818
819 case SO_LINGER:
820 if (optlen < sizeof(ling)) {
821 ret = -EINVAL; /* 1003.1g */
1da177e4 822 break;
e71a4783 823 }
2a91525c 824 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 825 ret = -EFAULT;
1da177e4 826 break;
e71a4783
SH		 827 }
828 if (!ling.l_onoff)
829 sock_reset_flag(sk, SOCK_LINGER);
830 else {
1da177e4 831#if (BITS_PER_LONG == 32)
e71a4783
SH		 832 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
833 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 834 else
e71a4783
SH		 835#endif
836 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
837 sock_set_flag(sk, SOCK_LINGER);
838 }
839 break;
840
841 case SO_BSDCOMPAT:
842 sock_warn_obsolete_bsdism("setsockopt");
843 break;
844
845 case SO_PASSCRED:
846 if (valbool)
847 set_bit(SOCK_PASSCRED, &sock->flags);
848 else
849 clear_bit(SOCK_PASSCRED, &sock->flags);
850 break;
851
852 case SO_TIMESTAMP:
92f37fd2 853 case SO_TIMESTAMPNS:
e71a4783 854 if (valbool) {
92f37fd2
ED		 855 if (optname == SO_TIMESTAMP)
856 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
857 else
858 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 859 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 860 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 861 } else {
e71a4783 862 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 863 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
864 }
e71a4783
SH		 865 break;
866
20d49473
PO		 867 case SO_TIMESTAMPING:
868 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 869 ret = -EINVAL;
20d49473
PO		 870 break;
871 }
b245be1f 872
09c2d251 873 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765 874 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
ac5cc977
WC		 875 if (sk->sk_protocol == IPPROTO_TCP &&
876 sk->sk_type == SOCK_STREAM) {
6db8b963
SHY		 877 if ((1 << sk->sk_state) &
878 (TCPF_CLOSE | TCPF_LISTEN)) {
4ed2d765
WB		 879 ret = -EINVAL;
880 break;
881 }
882 sk->sk_tskey = tcp_sk(sk)->snd_una;
883 } else {
884 sk->sk_tskey = 0;
885 }
886 }
1c885808
FY		 887
888 if (val & SOF_TIMESTAMPING_OPT_STATS &&
889 !(val & SOF_TIMESTAMPING_OPT_TSONLY)) {
890 ret = -EINVAL;
891 break;
892 }
893
b9f40e21 894 sk->sk_tsflags = val;
20d49473
PO		 895 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
896 sock_enable_timestamp(sk,
897 SOCK_TIMESTAMPING_RX_SOFTWARE);
898 else
899 sock_disable_timestamp(sk,
08e29af3 900 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 901 break;
902
e71a4783
SH		 903 case SO_RCVLOWAT:
904 if (val < 0)
905 val = INT_MAX;
906 sk->sk_rcvlowat = val ? : 1;
907 break;
908
909 case SO_RCVTIMEO:
910 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
911 break;
912
913 case SO_SNDTIMEO:
914 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
915 break;
1da177e4 916
e71a4783
SH		 917 case SO_ATTACH_FILTER:
918 ret = -EINVAL;
919 if (optlen == sizeof(struct sock_fprog)) {
920 struct sock_fprog fprog;
1da177e4 921
e71a4783
SH		 922 ret = -EFAULT;
923 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 924 break;
e71a4783
SH		 925
926 ret = sk_attach_filter(&fprog, sk);
927 }
928 break;
929
89aa0758
AS		 930 case SO_ATTACH_BPF:
931 ret = -EINVAL;
932 if (optlen == sizeof(u32)) {
933 u32 ufd;
934
935 ret = -EFAULT;
936 if (copy_from_user(&ufd, optval, sizeof(ufd)))
937 break;
938
939 ret = sk_attach_bpf(ufd, sk);
940 }
941 break;
942
538950a1
CG		 943 case SO_ATTACH_REUSEPORT_CBPF:
944 ret = -EINVAL;
945 if (optlen == sizeof(struct sock_fprog)) {
946 struct sock_fprog fprog;
947
948 ret = -EFAULT;
949 if (copy_from_user(&fprog, optval, sizeof(fprog)))
950 break;
951
952 ret = sk_reuseport_attach_filter(&fprog, sk);
953 }
954 break;
955
956 case SO_ATTACH_REUSEPORT_EBPF:
957 ret = -EINVAL;
958 if (optlen == sizeof(u32)) {
959 u32 ufd;
960
961 ret = -EFAULT;
962 if (copy_from_user(&ufd, optval, sizeof(ufd)))
963 break;
964
965 ret = sk_reuseport_attach_bpf(ufd, sk);
966 }
967 break;
968
e71a4783 969 case SO_DETACH_FILTER:
55b33325 970 ret = sk_detach_filter(sk);
e71a4783 971 break;
1da177e4 972
d59577b6
VB		 973 case SO_LOCK_FILTER:
974 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
975 ret = -EPERM;
976 else
977 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
978 break;
979
e71a4783
SH		 980 case SO_PASSSEC:
981 if (valbool)
982 set_bit(SOCK_PASSSEC, &sock->flags);
983 else
984 clear_bit(SOCK_PASSSEC, &sock->flags);
985 break;
4a19ec58 986 case SO_MARK:
5e1fccc0 987 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 988 ret = -EPERM;
2a91525c 989 else
4a19ec58 990 sk->sk_mark = val;
4a19ec58 991 break;
877ce7c1 992
3b885787 993 case SO_RXQ_OVFL:
8083f0fc 994 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 995 break;
6e3e939f
JB		 996
997 case SO_WIFI_STATUS:
998 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
999 break;
1000
ef64a54f
PE		 1001 case SO_PEEK_OFF:
1002 if (sock->ops->set_peek_off)
12663bfc 1003 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 1004 else
1005 ret = -EOPNOTSUPP;
1006 break;
3bdc0eba
BG		 1007
1008 case SO_NOFCS:
1009 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
1010 break;
1011
7d4c04fc
KJ		 1012 case SO_SELECT_ERR_QUEUE:
1013 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
1014 break;
1015
e0d1095a 1016#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1017 case SO_BUSY_POLL:
dafcc438
ET		 1018 /* allow unprivileged users to decrease the value */
1019 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
1020 ret = -EPERM;
1021 else {
1022 if (val < 0)
1023 ret = -EINVAL;
1024 else
1025 sk->sk_ll_usec = val;
1026 }
1027 break;
1028#endif
62748f32
ED		 1029
1030 case SO_MAX_PACING_RATE:
218af599
ED		 1031 if (val != ~0U)
1032 cmpxchg(&sk->sk_pacing_status,
1033 SK_PACING_NONE,
1034 SK_PACING_NEEDED);
62748f32
ED		 1035 sk->sk_max_pacing_rate = val;
1036 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
1037 sk->sk_max_pacing_rate);
1038 break;
1039
70da268b
ED		 1040 case SO_INCOMING_CPU:
1041 sk->sk_incoming_cpu = val;
1042 break;
1043
a87cb3e4
TH		 1044 case SO_CNX_ADVICE:
1045 if (val == 1)
1046 dst_negative_advice(sk);
1047 break;
76851d12
WB		 1048
1049 case SO_ZEROCOPY:
1050 if (sk->sk_family != PF_INET && sk->sk_family != PF_INET6)
1051 ret = -ENOTSUPP;
1052 else if (sk->sk_protocol != IPPROTO_TCP)
1053 ret = -ENOTSUPP;
1054 else if (sk->sk_state != TCP_CLOSE)
1055 ret = -EBUSY;
1056 else if (val < 0 || val > 1)
1057 ret = -EINVAL;
1058 else
1059 sock_valbool_flag(sk, SOCK_ZEROCOPY, valbool);
1060 break;
1061
e71a4783
SH		 1062 default:
1063 ret = -ENOPROTOOPT;
1064 break;
4ec93edb 1065 }
1da177e4
LT		 1066 release_sock(sk);
1067 return ret;
1068}
2a91525c 1069EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1070
1071
8f09898b 1072static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1073 struct ucred *ucred)
3f551f94
EB		 1074{
1075 ucred->pid = pid_vnr(pid);
1076 ucred->uid = ucred->gid = -1;
1077 if (cred) {
1078 struct user_namespace *current_ns = current_user_ns();
1079
b2e4f544
EB		 1080 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1081 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1082 }
1083}
1084
28b5ba2a
DH		 1085static int groups_to_user(gid_t __user *dst, const struct group_info *src)
1086{
1087 struct user_namespace *user_ns = current_user_ns();
1088 int i;
1089
1090 for (i = 0; i < src->ngroups; i++)
1091 if (put_user(from_kgid_munged(user_ns, src->gid[i]), dst + i))
1092 return -EFAULT;
1093
1094 return 0;
1095}
1096
1da177e4
LT		 1097int sock_getsockopt(struct socket *sock, int level, int optname,
1098 char __user *optval, int __user *optlen)
1099{
1100 struct sock *sk = sock->sk;
4ec93edb 1101
e71a4783 1102 union {
4ec93edb 1103 int val;
5daab9db 1104 u64 val64;
4ec93edb 1105 struct linger ling;
1da177e4
LT		 1106 struct timeval tm;
1107 } v;
4ec93edb 1108
4d0392be 1109 int lv = sizeof(int);
1da177e4 1110 int len;
4ec93edb 1111
e71a4783 1112 if (get_user(len, optlen))
4ec93edb 1113 return -EFAULT;
e71a4783 1114 if (len < 0)
1da177e4 1115 return -EINVAL;
4ec93edb 1116
50fee1de 1117 memset(&v, 0, sizeof(v));
df0bca04 1118
2a91525c 1119 switch (optname) {
e71a4783
SH		 1120 case SO_DEBUG:
1121 v.val = sock_flag(sk, SOCK_DBG);
1122 break;
1123
1124 case SO_DONTROUTE:
1125 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1126 break;
1127
1128 case SO_BROADCAST:
1b23a5df 1129 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1130 break;
1131
1132 case SO_SNDBUF:
1133 v.val = sk->sk_sndbuf;
1134 break;
1135
1136 case SO_RCVBUF:
1137 v.val = sk->sk_rcvbuf;
1138 break;
1139
1140 case SO_REUSEADDR:
1141 v.val = sk->sk_reuse;
1142 break;
1143
055dc21a
TH		 1144 case SO_REUSEPORT:
1145 v.val = sk->sk_reuseport;
1146 break;
1147
e71a4783 1148 case SO_KEEPALIVE:
1b23a5df 1149 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1150 break;
1151
1152 case SO_TYPE:
1153 v.val = sk->sk_type;
1154 break;
1155
49c794e9
JE		 1156 case SO_PROTOCOL:
1157 v.val = sk->sk_protocol;
1158 break;
1159
0d6038ee
JE		 1160 case SO_DOMAIN:
1161 v.val = sk->sk_family;
1162 break;
1163
e71a4783
SH		 1164 case SO_ERROR:
1165 v.val = -sock_error(sk);
2a91525c 1166 if (v.val == 0)
e71a4783
SH		 1167 v.val = xchg(&sk->sk_err_soft, 0);
1168 break;
1169
1170 case SO_OOBINLINE:
1b23a5df 1171 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1172 break;
1173
1174 case SO_NO_CHECK:
28448b80 1175 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1176 break;
1177
1178 case SO_PRIORITY:
1179 v.val = sk->sk_priority;
1180 break;
1181
1182 case SO_LINGER:
1183 lv = sizeof(v.ling);
1b23a5df 1184 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1185 v.ling.l_linger = sk->sk_lingertime / HZ;
1186 break;
1187
1188 case SO_BSDCOMPAT:
1189 sock_warn_obsolete_bsdism("getsockopt");
1190 break;
1191
1192 case SO_TIMESTAMP:
92f37fd2
ED		 1193 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1194 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1195 break;
1196
1197 case SO_TIMESTAMPNS:
1198 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1199 break;
1200
20d49473 1201 case SO_TIMESTAMPING:
b9f40e21 1202 v.val = sk->sk_tsflags;
20d49473
PO		 1203 break;
1204
e71a4783 1205 case SO_RCVTIMEO:
2a91525c 1206 lv = sizeof(struct timeval);
e71a4783
SH		 1207 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1208 v.tm.tv_sec = 0;
1209 v.tm.tv_usec = 0;
1210 } else {
1211 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
8ccde4c5 1212 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * USEC_PER_SEC) / HZ;
e71a4783
SH		 1213 }
1214 break;
1215
1216 case SO_SNDTIMEO:
2a91525c 1217 lv = sizeof(struct timeval);
e71a4783
SH		 1218 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1219 v.tm.tv_sec = 0;
1220 v.tm.tv_usec = 0;
1221 } else {
1222 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
8ccde4c5 1223 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * USEC_PER_SEC) / HZ;
e71a4783
SH		 1224 }
1225 break;
1da177e4 1226
e71a4783
SH		 1227 case SO_RCVLOWAT:
1228 v.val = sk->sk_rcvlowat;
1229 break;
1da177e4 1230
e71a4783 1231 case SO_SNDLOWAT:
2a91525c 1232 v.val = 1;
e71a4783 1233 break;
1da177e4 1234
e71a4783 1235 case SO_PASSCRED:
82981930 1236 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1237 break;
1da177e4 1238
e71a4783 1239 case SO_PEERCRED:
109f6e39
EB		 1240 {
1241 struct ucred peercred;
1242 if (len > sizeof(peercred))
1243 len = sizeof(peercred);
1244 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1245 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1246 return -EFAULT;
1247 goto lenout;
109f6e39 1248 }
1da177e4 1249
28b5ba2a
DH		 1250 case SO_PEERGROUPS:
1251 {
1252 int ret, n;
1253
1254 if (!sk->sk_peer_cred)
1255 return -ENODATA;
1256
1257 n = sk->sk_peer_cred->group_info->ngroups;
1258 if (len < n * sizeof(gid_t)) {
1259 len = n * sizeof(gid_t);
1260 return put_user(len, optlen) ? -EFAULT : -ERANGE;
1261 }
1262 len = n * sizeof(gid_t);
1263
1264 ret = groups_to_user((gid_t __user *)optval,
1265 sk->sk_peer_cred->group_info);
1266 if (ret)
1267 return ret;
1268 goto lenout;
1269 }
1270
e71a4783
SH		 1271 case SO_PEERNAME:
1272 {
1273 char address[128];
1274
1275 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1276 return -ENOTCONN;
1277 if (lv < len)
1278 return -EINVAL;
1279 if (copy_to_user(optval, address, len))
1280 return -EFAULT;
1281 goto lenout;
1282 }
1da177e4 1283
e71a4783
SH		 1284 /* Dubious BSD thing... Probably nobody even uses it, but
1285 * the UNIX standard wants it for whatever reason... -DaveM
1286 */
1287 case SO_ACCEPTCONN:
1288 v.val = sk->sk_state == TCP_LISTEN;
1289 break;
1da177e4 1290
e71a4783 1291 case SO_PASSSEC:
82981930 1292 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1293 break;
877ce7c1 1294
e71a4783
SH		 1295 case SO_PEERSEC:
1296 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1297
4a19ec58
LAT		 1298 case SO_MARK:
1299 v.val = sk->sk_mark;
1300 break;
1301
3b885787 1302 case SO_RXQ_OVFL:
1b23a5df 1303 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1304 break;
1305
6e3e939f 1306 case SO_WIFI_STATUS:
1b23a5df 1307 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1308 break;
1309
ef64a54f
PE		 1310 case SO_PEEK_OFF:
1311 if (!sock->ops->set_peek_off)
1312 return -EOPNOTSUPP;
1313
1314 v.val = sk->sk_peek_off;
1315 break;
bc2f7996 1316 case SO_NOFCS:
1b23a5df 1317 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1318 break;
c91f6df2 1319
f7b86bfe 1320 case SO_BINDTODEVICE:
c91f6df2
BH		 1321 return sock_getbindtodevice(sk, optval, optlen, len);
1322
a8fc9277
PE		 1323 case SO_GET_FILTER:
1324 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1325 if (len < 0)
1326 return len;
1327
1328 goto lenout;
c91f6df2 1329
d59577b6
VB		 1330 case SO_LOCK_FILTER:
1331 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1332 break;
1333
ea02f941
MS		 1334 case SO_BPF_EXTENSIONS:
1335 v.val = bpf_tell_extensions();
1336 break;
1337
7d4c04fc
KJ		 1338 case SO_SELECT_ERR_QUEUE:
1339 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1340 break;
1341
e0d1095a 1342#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1343 case SO_BUSY_POLL:
dafcc438
ET		 1344 v.val = sk->sk_ll_usec;
1345 break;
1346#endif
1347
62748f32
ED		 1348 case SO_MAX_PACING_RATE:
1349 v.val = sk->sk_max_pacing_rate;
1350 break;
1351
2c8c56e1
ED		 1352 case SO_INCOMING_CPU:
1353 v.val = sk->sk_incoming_cpu;
1354 break;
1355
a2d133b1
JH		 1356 case SO_MEMINFO:
1357 {
1358 u32 meminfo[SK_MEMINFO_VARS];
1359
1360 if (get_user(len, optlen))
1361 return -EFAULT;
1362
1363 sk_get_meminfo(sk, meminfo);
1364
1365 len = min_t(unsigned int, len, sizeof(meminfo));
1366 if (copy_to_user(optval, &meminfo, len))
1367 return -EFAULT;
1368
1369 goto lenout;
1370 }
6d433902
SS		 1371
1372#ifdef CONFIG_NET_RX_BUSY_POLL
1373 case SO_INCOMING_NAPI_ID:
1374 v.val = READ_ONCE(sk->sk_napi_id);
1375
1376 /* aggregate non-NAPI IDs down to 0 */
1377 if (v.val < MIN_NAPI_ID)
1378 v.val = 0;
1379
1380 break;
1381#endif
1382
5daab9db
CF		 1383 case SO_COOKIE:
1384 lv = sizeof(u64);
1385 if (len < lv)
1386 return -EINVAL;
1387 v.val64 = sock_gen_cookie(sk);
1388 break;
1389
76851d12
WB		 1390 case SO_ZEROCOPY:
1391 v.val = sock_flag(sk, SOCK_ZEROCOPY);
1392 break;
1393
e71a4783 1394 default:
443b5991
YH		 1395 /* We implement the SO_SNDLOWAT etc to not be settable
1396 * (1003.1g 7).
1397 */
e71a4783 1398 return -ENOPROTOOPT;
1da177e4 1399 }
e71a4783 1400
1da177e4
LT		 1401 if (len > lv)
1402 len = lv;
1403 if (copy_to_user(optval, &v, len))
1404 return -EFAULT;
1405lenout:
4ec93edb
YH		 1406 if (put_user(len, optlen))
1407 return -EFAULT;
1408 return 0;
1da177e4
LT		 1409}
1410
a5b5bb9a
IM		 1411/*
1412 * Initialize an sk_lock.
1413 *
1414 * (We also register the sk_lock with the lock validator.)
1415 */
b6f99a21 1416static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1417{
cdfbabfb
DH		 1418 if (sk->sk_kern_sock)
1419 sock_lock_init_class_and_name(
1420 sk,
1421 af_family_kern_slock_key_strings[sk->sk_family],
1422 af_family_kern_slock_keys + sk->sk_family,
1423 af_family_kern_key_strings[sk->sk_family],
1424 af_family_kern_keys + sk->sk_family);
1425 else
1426 sock_lock_init_class_and_name(
1427 sk,
ed07536e
PZ		 1428 af_family_slock_key_strings[sk->sk_family],
1429 af_family_slock_keys + sk->sk_family,
1430 af_family_key_strings[sk->sk_family],
1431 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1432}
1433
4dc6dc71
ED		 1434/*
1435 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1436 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1437 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1438 */
f1a6c4da
PE		 1439static void sock_copy(struct sock *nsk, const struct sock *osk)
1440{
1441#ifdef CONFIG_SECURITY_NETWORK
1442 void *sptr = nsk->sk_security;
1443#endif
68835aba
ED		 1444 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1445
1446 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1447 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1448
f1a6c4da
PE		 1449#ifdef CONFIG_SECURITY_NETWORK
1450 nsk->sk_security = sptr;
1451 security_sk_clone(osk, nsk);
1452#endif
1453}
1454
2e4afe7b
PE		 1455static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1456 int family)
c308c1b2
PE		 1457{
1458 struct sock *sk;
1459 struct kmem_cache *slab;
1460
1461 slab = prot->slab;
e912b114
ED		 1462 if (slab != NULL) {
1463 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1464 if (!sk)
1465 return sk;
ba2489b0
ED		 1466 if (priority & __GFP_ZERO)
1467 sk_prot_clear_nulls(sk, prot->obj_size);
fcbdf09d 1468 } else
c308c1b2
PE		 1469 sk = kmalloc(prot->obj_size, priority);
1470
2e4afe7b
PE		 1471 if (sk != NULL) {
1472 if (security_sk_alloc(sk, family, priority))
1473 goto out_free;
1474
1475 if (!try_module_get(prot->owner))
1476 goto out_free_sec;
e022f0b4 1477 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1478 }
1479
c308c1b2 1480 return sk;
2e4afe7b
PE		 1481
1482out_free_sec:
1483 security_sk_free(sk);
1484out_free:
1485 if (slab != NULL)
1486 kmem_cache_free(slab, sk);
1487 else
1488 kfree(sk);
1489 return NULL;
c308c1b2
PE		 1490}
1491
1492static void sk_prot_free(struct proto *prot, struct sock *sk)
1493{
1494 struct kmem_cache *slab;
2e4afe7b 1495 struct module *owner;
c308c1b2 1496
2e4afe7b 1497 owner = prot->owner;
c308c1b2 1498 slab = prot->slab;
2e4afe7b 1499
bd1060a1 1500 cgroup_sk_free(&sk->sk_cgrp_data);
2d758073 1501 mem_cgroup_sk_free(sk);
2e4afe7b 1502 security_sk_free(sk);
c308c1b2
PE		 1503 if (slab != NULL)
1504 kmem_cache_free(slab, sk);
1505 else
1506 kfree(sk);
2e4afe7b 1507 module_put(owner);
c308c1b2
PE		 1508}
1509
1da177e4
LT		 1510/**
1511 * sk_alloc - All socket objects are allocated here
c4ea43c5 1512 * @net: the applicable net namespace
4dc3b16b
PP		 1513 * @family: protocol family
1514 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1515 * @prot: struct proto associated with this new sock instance
11aa9c28 1516 * @kern: is this to be a kernel socket?
1da177e4 1517 */
1b8d7ae4 1518struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1519 struct proto *prot, int kern)
1da177e4 1520{
c308c1b2 1521 struct sock *sk;
1da177e4 1522
154adbc8 1523 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1524 if (sk) {
154adbc8
PE		 1525 sk->sk_family = family;
1526 /*
1527 * See comment in struct sock definition to understand
1528 * why we need sk_prot_creator -acme
1529 */
1530 sk->sk_prot = sk->sk_prot_creator = prot;
cdfbabfb 1531 sk->sk_kern_sock = kern;
154adbc8 1532 sock_lock_init(sk);
26abe143
EB		 1533 sk->sk_net_refcnt = kern ? 0 : 1;
1534 if (likely(sk->sk_net_refcnt))
1535 get_net(net);
1536 sock_net_set(sk, net);
14afee4b 1537 refcount_set(&sk->sk_wmem_alloc, 1);
f8451725 1538
2d758073 1539 mem_cgroup_sk_alloc(sk);
d979a39d 1540 cgroup_sk_alloc(&sk->sk_cgrp_data);
2a56a1fe
TH		 1541 sock_update_classid(&sk->sk_cgrp_data);
1542 sock_update_netprioidx(&sk->sk_cgrp_data);
1da177e4 1543 }
a79af59e 1544
2e4afe7b 1545 return sk;
1da177e4 1546}
2a91525c 1547EXPORT_SYMBOL(sk_alloc);
1da177e4 1548
a4298e45
ED		 1549/* Sockets having SOCK_RCU_FREE will call this function after one RCU
1550 * grace period. This is the case for UDP sockets and TCP listeners.
1551 */
1552static void __sk_destruct(struct rcu_head *head)
1da177e4 1553{
a4298e45 1554 struct sock *sk = container_of(head, struct sock, sk_rcu);
1da177e4 1555 struct sk_filter *filter;
1da177e4
LT		 1556
1557 if (sk->sk_destruct)
1558 sk->sk_destruct(sk);
1559
a898def2 1560 filter = rcu_dereference_check(sk->sk_filter,
14afee4b 1561 refcount_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1562 if (filter) {
309dd5fc 1563 sk_filter_uncharge(sk, filter);
a9b3cd7f 1564 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4 1565 }
538950a1
CG		 1566 if (rcu_access_pointer(sk->sk_reuseport_cb))
1567 reuseport_detach_sock(sk);
1da177e4 1568
08e29af3 1569 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1570
1571 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1572 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1573 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1574
22a0e18e
ED		 1575 if (sk->sk_frag.page) {
1576 put_page(sk->sk_frag.page);
1577 sk->sk_frag.page = NULL;
1578 }
1579
109f6e39
EB		 1580 if (sk->sk_peer_cred)
1581 put_cred(sk->sk_peer_cred);
1582 put_pid(sk->sk_peer_pid);
26abe143
EB		 1583 if (likely(sk->sk_net_refcnt))
1584 put_net(sock_net(sk));
c308c1b2 1585 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1586}
2b85a34e 1587
a4298e45
ED		 1588void sk_destruct(struct sock *sk)
1589{
1590 if (sock_flag(sk, SOCK_RCU_FREE))
1591 call_rcu(&sk->sk_rcu, __sk_destruct);
1592 else
1593 __sk_destruct(&sk->sk_rcu);
1594}
1595
eb4cb008
CG		 1596static void __sk_free(struct sock *sk)
1597{
b922622e 1598 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
eb4cb008
CG		 1599 sock_diag_broadcast_destroy(sk);
1600 else
1601 sk_destruct(sk);
1602}
1603
2b85a34e
ED		 1604void sk_free(struct sock *sk)
1605{
1606 /*
25985edc 1607 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1608 * some packets are still in some tx queue.
1609 * If not null, sock_wfree() will call __sk_free(sk) later
1610 */
14afee4b 1611 if (refcount_dec_and_test(&sk->sk_wmem_alloc))
2b85a34e
ED		 1612 __sk_free(sk);
1613}
2a91525c 1614EXPORT_SYMBOL(sk_free);
1da177e4 1615
581319c5
PA		 1616static void sk_init_common(struct sock *sk)
1617{
1618 skb_queue_head_init(&sk->sk_receive_queue);
1619 skb_queue_head_init(&sk->sk_write_queue);
1620 skb_queue_head_init(&sk->sk_error_queue);
1621
1622 rwlock_init(&sk->sk_callback_lock);
1623 lockdep_set_class_and_name(&sk->sk_receive_queue.lock,
1624 af_rlock_keys + sk->sk_family,
1625 af_family_rlock_key_strings[sk->sk_family]);
1626 lockdep_set_class_and_name(&sk->sk_write_queue.lock,
1627 af_wlock_keys + sk->sk_family,
1628 af_family_wlock_key_strings[sk->sk_family]);
1629 lockdep_set_class_and_name(&sk->sk_error_queue.lock,
1630 af_elock_keys + sk->sk_family,
1631 af_family_elock_key_strings[sk->sk_family]);
1632 lockdep_set_class_and_name(&sk->sk_callback_lock,
1633 af_callback_keys + sk->sk_family,
1634 af_family_clock_key_strings[sk->sk_family]);
1635}
1636
e56c57d0
ED		 1637/**
1638 * sk_clone_lock - clone a socket, and lock its clone
1639 * @sk: the socket to clone
1640 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1641 *
1642 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1643 */
1644struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1645{
8fd1d178 1646 struct sock *newsk;
278571ba 1647 bool is_charged = true;
87d11ceb 1648
8fd1d178 1649 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1650 if (newsk != NULL) {
1651 struct sk_filter *filter;
1652
892c141e 1653 sock_copy(newsk, sk);
87d11ceb 1654
9d538fa6
CP		 1655 newsk->sk_prot_creator = sk->sk_prot;
1656
87d11ceb 1657 /* SANITY */
8a681736
SV		 1658 if (likely(newsk->sk_net_refcnt))
1659 get_net(sock_net(newsk));
87d11ceb
ACM		 1660 sk_node_init(&newsk->sk_node);
1661 sock_lock_init(newsk);
1662 bh_lock_sock(newsk);
fa438ccf 1663 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1664 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1665
1666 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1667 /*
1668 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1669 */
14afee4b 1670 refcount_set(&newsk->sk_wmem_alloc, 1);
87d11ceb 1671 atomic_set(&newsk->sk_omem_alloc, 0);
581319c5 1672 sk_init_common(newsk);
87d11ceb
ACM		 1673
1674 newsk->sk_dst_cache = NULL;
9b8805a3 1675 newsk->sk_dst_pending_confirm = 0;
87d11ceb
ACM		 1676 newsk->sk_wmem_queued = 0;
1677 newsk->sk_forward_alloc = 0;
9f1c2674
ED		 1678
1679 /* sk->sk_memcg will be populated at accept() time */
1680 newsk->sk_memcg = NULL;
1681
9caad864 1682 atomic_set(&newsk->sk_drops, 0);
87d11ceb 1683 newsk->sk_send_head = NULL;
87d11ceb 1684 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
52267790 1685 atomic_set(&newsk->sk_zckey, 0);
87d11ceb
ACM		 1686
1687 sock_reset_flag(newsk, SOCK_DONE);
c0576e39 1688 cgroup_sk_alloc(&newsk->sk_cgrp_data);
87d11ceb 1689
eefca20e
ED		 1690 rcu_read_lock();
1691 filter = rcu_dereference(sk->sk_filter);
87d11ceb 1692 if (filter != NULL)
278571ba
AS		 1693 /* though it's an empty new sock, the charging may fail
1694 * if sysctl_optmem_max was changed between creation of
1695 * original socket and cloning
1696 */
1697 is_charged = sk_filter_charge(newsk, filter);
eefca20e
ED		 1698 RCU_INIT_POINTER(newsk->sk_filter, filter);
1699 rcu_read_unlock();
87d11ceb 1700
d188ba86 1701 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
a97e50cc
DB		 1702 /* We need to make sure that we don't uncharge the new
1703 * socket if we couldn't charge it in the first place
1704 * as otherwise we uncharge the parent's filter.
1705 */
1706 if (!is_charged)
1707 RCU_INIT_POINTER(newsk->sk_filter, NULL);
94352d45 1708 sk_free_unlock_clone(newsk);
87d11ceb
ACM		 1709 newsk = NULL;
1710 goto out;
1711 }
fa463497 1712 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
87d11ceb
ACM		 1713
1714 newsk->sk_err = 0;
e551c32d 1715 newsk->sk_err_soft = 0;
87d11ceb 1716 newsk->sk_priority = 0;
2c8c56e1 1717 newsk->sk_incoming_cpu = raw_smp_processor_id();
33cf7c90 1718 atomic64_set(&newsk->sk_cookie, 0);
d979a39d 1719
4dc6dc71
ED		 1720 /*
1721 * Before updating sk_refcnt, we must commit prior changes to memory
1722 * (Documentation/RCU/rculist_nulls.txt for details)
1723 */
1724 smp_wmb();
41c6d650 1725 refcount_set(&newsk->sk_refcnt, 2);
87d11ceb
ACM		 1726
1727 /*
1728 * Increment the counter in the same struct proto as the master
1729 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1730 * is the same as sk->sk_prot->socks, as this field was copied
1731 * with memcpy).
1732 *
1733 * This _changes_ the previous behaviour, where
1734 * tcp_create_openreq_child always was incrementing the
1735 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1736 * to be taken into account in all callers. -acme
1737 */
1738 sk_refcnt_debug_inc(newsk);
972692e0 1739 sk_set_socket(newsk, NULL);
43815482 1740 newsk->sk_wq = NULL;
87d11ceb
ACM		 1741
1742 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1743 sk_sockets_allocated_inc(newsk);
704da560 1744
080a270f
HFS		 1745 if (sock_needs_netstamp(sk) &&
1746 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1747 net_enable_timestamp();
87d11ceb
ACM		 1748 }
1749out:
1750 return newsk;
1751}
e56c57d0 1752EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1753
94352d45
ACM		 1754void sk_free_unlock_clone(struct sock *sk)
1755{
1756 /* It is still raw copy of parent, so invalidate
1757 * destructor and make plain sk_free() */
1758 sk->sk_destruct = NULL;
1759 bh_unlock_sock(sk);
1760 sk_free(sk);
1761}
1762EXPORT_SYMBOL_GPL(sk_free_unlock_clone);
1763
9958089a
AK		 1764void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1765{
d6a4e26a
ED		 1766 u32 max_segs = 1;
1767
6bd4f355 1768 sk_dst_set(sk, dst);
9958089a
AK		 1769 sk->sk_route_caps = dst->dev->features;
1770 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1771 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1772 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1773 if (sk_can_gso(sk)) {
f70f250a 1774 if (dst->header_len && !xfrm_dst_offload_ok(dst)) {
9958089a 1775 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1776 } else {
9958089a 1777 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1778 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 1779 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 1780 }
9958089a 1781 }
d6a4e26a 1782 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 1783}
1784EXPORT_SYMBOL_GPL(sk_setup_caps);
1785
1da177e4
LT		 1786/*
1787 * Simple resource managers for sockets.
1788 */
1789
1790
4ec93edb
YH		 1791/*
1792 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1793 */
1794void sock_wfree(struct sk_buff *skb)
1795{
1796 struct sock *sk = skb->sk;
d99927f4 1797 unsigned int len = skb->truesize;
1da177e4 1798
d99927f4
ED		 1799 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1800 /*
1801 * Keep a reference on sk_wmem_alloc, this will be released
1802 * after sk_write_space() call
1803 */
14afee4b 1804 WARN_ON(refcount_sub_and_test(len - 1, &sk->sk_wmem_alloc));
1da177e4 1805 sk->sk_write_space(sk);
d99927f4
ED		 1806 len = 1;
1807 }
2b85a34e 1808 /*
d99927f4
ED		 1809 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1810 * could not do because of in-flight packets
2b85a34e 1811 */
14afee4b 1812 if (refcount_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1813 __sk_free(sk);
1da177e4 1814}
2a91525c 1815EXPORT_SYMBOL(sock_wfree);
1da177e4 1816
1d2077ac
ED		 1817/* This variant of sock_wfree() is used by TCP,
1818 * since it sets SOCK_USE_WRITE_QUEUE.
1819 */
1820void __sock_wfree(struct sk_buff *skb)
1821{
1822 struct sock *sk = skb->sk;
1823
14afee4b 1824 if (refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc))
1d2077ac
ED		 1825 __sk_free(sk);
1826}
1827
9e17f8a4
ED		 1828void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1829{
1830 skb_orphan(skb);
1831 skb->sk = sk;
1832#ifdef CONFIG_INET
1833 if (unlikely(!sk_fullsock(sk))) {
1834 skb->destructor = sock_edemux;
1835 sock_hold(sk);
1836 return;
1837 }
1838#endif
1839 skb->destructor = sock_wfree;
1840 skb_set_hash_from_sk(skb, sk);
1841 /*
1842 * We used to take a refcount on sk, but following operation
1843 * is enough to guarantee sk_free() wont free this sock until
1844 * all in-flight packets are completed
1845 */
14afee4b 1846 refcount_add(skb->truesize, &sk->sk_wmem_alloc);
9e17f8a4
ED		 1847}
1848EXPORT_SYMBOL(skb_set_owner_w);
1849
1d2077ac
ED		 1850/* This helper is used by netem, as it can hold packets in its
1851 * delay queue. We want to allow the owner socket to send more
1852 * packets, as if they were already TX completed by a typical driver.
1853 * But we also want to keep skb->sk set because some packet schedulers
f6ba8d33 1854 * rely on it (sch_fq for example).
1d2077ac 1855 */
f2f872f9
ED		 1856void skb_orphan_partial(struct sk_buff *skb)
1857{
f6ba8d33 1858 if (skb_is_tcp_pure_ack(skb))
1d2077ac
ED		 1859 return;
1860
f2f872f9
ED		 1861 if (skb->destructor == sock_wfree
1862#ifdef CONFIG_INET
1863 || skb->destructor == tcp_wfree
1864#endif
1865 ) {
f6ba8d33
ED		 1866 struct sock *sk = skb->sk;
1867
41c6d650 1868 if (refcount_inc_not_zero(&sk->sk_refcnt)) {
14afee4b 1869 WARN_ON(refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc));
f6ba8d33
ED		 1870 skb->destructor = sock_efree;
1871 }
f2f872f9
ED		 1872 } else {
1873 skb_orphan(skb);
1874 }
1875}
1876EXPORT_SYMBOL(skb_orphan_partial);
1877
4ec93edb
YH		 1878/*
1879 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1880 */
1881void sock_rfree(struct sk_buff *skb)
1882{
1883 struct sock *sk = skb->sk;
d361fd59 1884 unsigned int len = skb->truesize;
1da177e4 1885
d361fd59
ED		 1886 atomic_sub(len, &sk->sk_rmem_alloc);
1887 sk_mem_uncharge(sk, len);
1da177e4 1888}
2a91525c 1889EXPORT_SYMBOL(sock_rfree);
1da177e4 1890
7768eed8
OH		 1891/*
1892 * Buffer destructor for skbs that are not used directly in read or write
1893 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1894 */
62bccb8c
AD		 1895void sock_efree(struct sk_buff *skb)
1896{
1897 sock_put(skb->sk);
1898}
1899EXPORT_SYMBOL(sock_efree);
1900
976d0201 1901kuid_t sock_i_uid(struct sock *sk)
1da177e4 1902{
976d0201 1903 kuid_t uid;
1da177e4 1904
f064af1e 1905 read_lock_bh(&sk->sk_callback_lock);
976d0201 1906 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1907 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1908 return uid;
1909}
2a91525c 1910EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1911
1912unsigned long sock_i_ino(struct sock *sk)
1913{
1914 unsigned long ino;
1915
f064af1e 1916 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1917 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1918 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1919 return ino;
1920}
2a91525c 1921EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1922
1923/*
1924 * Allocate a skb from the socket's send buffer.
1925 */
86a76caf 1926struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1927 gfp_t priority)
1da177e4 1928{
14afee4b 1929 if (force || refcount_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1930 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1931 if (skb) {
1932 skb_set_owner_w(skb, sk);
1933 return skb;
1934 }
1935 }
1936 return NULL;
1937}
2a91525c 1938EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1939
98ba0bd5
WB		 1940static void sock_ofree(struct sk_buff *skb)
1941{
1942 struct sock *sk = skb->sk;
1943
1944 atomic_sub(skb->truesize, &sk->sk_omem_alloc);
1945}
1946
1947struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1948 gfp_t priority)
1949{
1950 struct sk_buff *skb;
1951
1952 /* small safe race: SKB_TRUESIZE may differ from final skb->truesize */
1953 if (atomic_read(&sk->sk_omem_alloc) + SKB_TRUESIZE(size) >
1954 sysctl_optmem_max)
1955 return NULL;
1956
1957 skb = alloc_skb(size, priority);
1958 if (!skb)
1959 return NULL;
1960
1961 atomic_add(skb->truesize, &sk->sk_omem_alloc);
1962 skb->sk = sk;
1963 skb->destructor = sock_ofree;
1964 return skb;
1965}
1966
4ec93edb 1967/*
1da177e4 1968 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1969 */
dd0fc66f 1970void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1971{
95c96174 1972 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1973 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1974 void *mem;
1975 /* First do the add, to avoid the race if kmalloc
4ec93edb 1976 * might sleep.
1da177e4
LT		 1977 */
1978 atomic_add(size, &sk->sk_omem_alloc);
1979 mem = kmalloc(size, priority);
1980 if (mem)
1981 return mem;
1982 atomic_sub(size, &sk->sk_omem_alloc);
1983 }
1984 return NULL;
1985}
2a91525c 1986EXPORT_SYMBOL(sock_kmalloc);
1da177e4 1987
79e88659
DB		 1988/* Free an option memory block. Note, we actually want the inline
1989 * here as this allows gcc to detect the nullify and fold away the
1990 * condition entirely.
1da177e4 1991 */
79e88659
DB		 1992static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1993 const bool nullify)
1da177e4 1994{
e53da5fb
DM		 1995 if (WARN_ON_ONCE(!mem))
1996 return;
79e88659
DB		 1997 if (nullify)
1998 kzfree(mem);
1999 else
2000 kfree(mem);
1da177e4
LT		 2001 atomic_sub(size, &sk->sk_omem_alloc);
2002}
79e88659
DB		 2003
2004void sock_kfree_s(struct sock *sk, void *mem, int size)
2005{
2006 __sock_kfree_s(sk, mem, size, false);
2007}
2a91525c 2008EXPORT_SYMBOL(sock_kfree_s);
1da177e4 2009
79e88659
DB		 2010void sock_kzfree_s(struct sock *sk, void *mem, int size)
2011{
2012 __sock_kfree_s(sk, mem, size, true);
2013}
2014EXPORT_SYMBOL(sock_kzfree_s);
2015
1da177e4
LT		 2016/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
2017 I think, these locks should be removed for datagram sockets.
2018 */
2a91525c 2019static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 2020{
2021 DEFINE_WAIT(wait);
2022
9cd3e072 2023 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1da177e4
LT		 2024 for (;;) {
2025 if (!timeo)
2026 break;
2027 if (signal_pending(current))
2028 break;
2029 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 2030 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
14afee4b 2031 if (refcount_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1da177e4
LT		 2032 break;
2033 if (sk->sk_shutdown & SEND_SHUTDOWN)
2034 break;
2035 if (sk->sk_err)
2036 break;
2037 timeo = schedule_timeout(timeo);
2038 }
aa395145 2039 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2040 return timeo;
2041}
2042
2043
2044/*
2045 * Generic send/receive buffer handlers
2046 */
2047
4cc7f68d
HX		 2048struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
2049 unsigned long data_len, int noblock,
28d64271 2050 int *errcode, int max_page_order)
1da177e4 2051{
2e4e4410 2052 struct sk_buff *skb;
1da177e4
LT		 2053 long timeo;
2054 int err;
2055
1da177e4 2056 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 2057 for (;;) {
1da177e4
LT		 2058 err = sock_error(sk);
2059 if (err != 0)
2060 goto failure;
2061
2062 err = -EPIPE;
2063 if (sk->sk_shutdown & SEND_SHUTDOWN)
2064 goto failure;
2065
2e4e4410
ED		 2066 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
2067 break;
28d64271 2068
9cd3e072 2069 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2e4e4410
ED		 2070 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2071 err = -EAGAIN;
2072 if (!timeo)
1da177e4 2073 goto failure;
2e4e4410
ED		 2074 if (signal_pending(current))
2075 goto interrupted;
2076 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 2077 }
2e4e4410
ED		 2078 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
2079 errcode, sk->sk_allocation);
2080 if (skb)
2081 skb_set_owner_w(skb, sk);
1da177e4
LT		 2082 return skb;
2083
2084interrupted:
2085 err = sock_intr_errno(timeo);
2086failure:
2087 *errcode = err;
2088 return NULL;
2089}
4cc7f68d 2090EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 2091
4ec93edb 2092struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 2093 int noblock, int *errcode)
2094{
28d64271 2095 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 2096}
2a91525c 2097EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 2098
39771b12
WB		 2099int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
2100 struct sockcm_cookie *sockc)
2101{
3dd17e63
SHY		 2102 u32 tsflags;
2103
39771b12
WB		 2104 switch (cmsg->cmsg_type) {
2105 case SO_MARK:
2106 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2107 return -EPERM;
2108 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2109 return -EINVAL;
2110 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
2111 break;
3dd17e63
SHY		 2112 case SO_TIMESTAMPING:
2113 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2114 return -EINVAL;
2115
2116 tsflags = *(u32 *)CMSG_DATA(cmsg);
2117 if (tsflags & ~SOF_TIMESTAMPING_TX_RECORD_MASK)
2118 return -EINVAL;
2119
2120 sockc->tsflags &= ~SOF_TIMESTAMPING_TX_RECORD_MASK;
2121 sockc->tsflags |= tsflags;
2122 break;
779f1ede
SHY		 2123 /* SCM_RIGHTS and SCM_CREDENTIALS are semantically in SOL_UNIX. */
2124 case SCM_RIGHTS:
2125 case SCM_CREDENTIALS:
2126 break;
39771b12
WB		 2127 default:
2128 return -EINVAL;
2129 }
2130 return 0;
2131}
2132EXPORT_SYMBOL(__sock_cmsg_send);
2133
f28ea365
EJ		 2134int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
2135 struct sockcm_cookie *sockc)
2136{
2137 struct cmsghdr *cmsg;
39771b12 2138 int ret;
f28ea365
EJ		 2139
2140 for_each_cmsghdr(cmsg, msg) {
2141 if (!CMSG_OK(msg, cmsg))
2142 return -EINVAL;
2143 if (cmsg->cmsg_level != SOL_SOCKET)
2144 continue;
39771b12
WB		 2145 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
2146 if (ret)
2147 return ret;
f28ea365
EJ		 2148 }
2149 return 0;
2150}
2151EXPORT_SYMBOL(sock_cmsg_send);
2152
06044751
ED		 2153static void sk_enter_memory_pressure(struct sock *sk)
2154{
2155 if (!sk->sk_prot->enter_memory_pressure)
2156 return;
2157
2158 sk->sk_prot->enter_memory_pressure(sk);
2159}
2160
2161static void sk_leave_memory_pressure(struct sock *sk)
2162{
2163 if (sk->sk_prot->leave_memory_pressure) {
2164 sk->sk_prot->leave_memory_pressure(sk);
2165 } else {
2166 unsigned long *memory_pressure = sk->sk_prot->memory_pressure;
2167
2168 if (memory_pressure && *memory_pressure)
2169 *memory_pressure = 0;
2170 }
2171}
2172
5640f768
ED		 2173/* On 32bit arches, an skb frag is limited to 2^15 */
2174#define SKB_FRAG_PAGE_ORDER get_order(32768)
2175
400dfd3a
ED		 2176/**
2177 * skb_page_frag_refill - check that a page_frag contains enough room
2178 * @sz: minimum size of the fragment we want to get
2179 * @pfrag: pointer to page_frag
82d5e2b8 2180 * @gfp: priority for memory allocation
400dfd3a
ED		 2181 *
2182 * Note: While this allocator tries to use high order pages, there is
2183 * no guarantee that allocations succeed. Therefore, @sz MUST be
2184 * less or equal than PAGE_SIZE.
2185 */
d9b2938a 2186bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 2187{
5640f768 2188 if (pfrag->page) {
fe896d18 2189 if (page_ref_count(pfrag->page) == 1) {
5640f768
ED		 2190 pfrag->offset = 0;
2191 return true;
2192 }
400dfd3a 2193 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 2194 return true;
2195 put_page(pfrag->page);
2196 }
2197
d9b2938a
ED		 2198 pfrag->offset = 0;
2199 if (SKB_FRAG_PAGE_ORDER) {
d0164adc
MG		 2200 /* Avoid direct reclaim but allow kswapd to wake */
2201 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
2202 __GFP_COMP | __GFP_NOWARN |
2203 __GFP_NORETRY,
d9b2938a 2204 SKB_FRAG_PAGE_ORDER);
5640f768 2205 if (likely(pfrag->page)) {
d9b2938a 2206 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 2207 return true;
2208 }
d9b2938a
ED		 2209 }
2210 pfrag->page = alloc_page(gfp);
2211 if (likely(pfrag->page)) {
2212 pfrag->size = PAGE_SIZE;
2213 return true;
2214 }
400dfd3a
ED		 2215 return false;
2216}
2217EXPORT_SYMBOL(skb_page_frag_refill);
2218
2219bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
2220{
2221 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
2222 return true;
2223
5640f768
ED		 2224 sk_enter_memory_pressure(sk);
2225 sk_stream_moderate_sndbuf(sk);
2226 return false;
2227}
2228EXPORT_SYMBOL(sk_page_frag_refill);
2229
1da177e4 2230static void __lock_sock(struct sock *sk)
f39234d6
NK		 2231 __releases(&sk->sk_lock.slock)
2232 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 2233{
2234 DEFINE_WAIT(wait);
2235
e71a4783 2236 for (;;) {
1da177e4
LT		 2237 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
2238 TASK_UNINTERRUPTIBLE);
2239 spin_unlock_bh(&sk->sk_lock.slock);
2240 schedule();
2241 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 2242 if (!sock_owned_by_user(sk))
1da177e4
LT		 2243 break;
2244 }
2245 finish_wait(&sk->sk_lock.wq, &wait);
2246}
2247
2248static void __release_sock(struct sock *sk)
f39234d6
NK		 2249 __releases(&sk->sk_lock.slock)
2250 __acquires(&sk->sk_lock.slock)
1da177e4 2251{
5413d1ba 2252 struct sk_buff *skb, *next;
1da177e4 2253
5413d1ba 2254 while ((skb = sk->sk_backlog.head) != NULL) {
1da177e4 2255 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1da177e4 2256
5413d1ba 2257 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4 2258
5413d1ba
ED		 2259 do {
2260 next = skb->next;
e4cbb02a 2261 prefetch(next);
7fee226a 2262 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 2263 skb->next = NULL;
c57943a1 2264 sk_backlog_rcv(sk, skb);
1da177e4 2265
5413d1ba 2266 cond_resched();
1da177e4
LT		 2267
2268 skb = next;
2269 } while (skb != NULL);
2270
5413d1ba
ED		 2271 spin_lock_bh(&sk->sk_lock.slock);
2272 }
8eae939f
ZY		 2273
2274 /*
2275 * Doing the zeroing here guarantee we can not loop forever
2276 * while a wild producer attempts to flood us.
2277 */
2278 sk->sk_backlog.len = 0;
1da177e4
LT		 2279}
2280
d41a69f1
ED		 2281void __sk_flush_backlog(struct sock *sk)
2282{
2283 spin_lock_bh(&sk->sk_lock.slock);
2284 __release_sock(sk);
2285 spin_unlock_bh(&sk->sk_lock.slock);
2286}
2287
1da177e4
LT		 2288/**
2289 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 2290 * @sk: sock to wait on
2291 * @timeo: for how long
dfbafc99 2292 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 2293 *
2294 * Now socket state including sk->sk_err is changed only under lock,
2295 * hence we may omit checks after joining wait queue.
2296 * We check receive queue before schedule() only as optimization;
2297 * it is very likely that release_sock() added new data.
2298 */
dfbafc99 2299int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4 2300{
d9dc8b0f 2301 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1da177e4 2302 int rc;
1da177e4 2303
d9dc8b0f 2304 add_wait_queue(sk_sleep(sk), &wait);
9cd3e072 2305 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
d9dc8b0f 2306 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb, &wait);
9cd3e072 2307 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
d9dc8b0f 2308 remove_wait_queue(sk_sleep(sk), &wait);
1da177e4
LT		 2309 return rc;
2310}
1da177e4
LT		 2311EXPORT_SYMBOL(sk_wait_data);
2312
3ab224be 2313/**
f8c3bf00 2314 * __sk_mem_raise_allocated - increase memory_allocated
3ab224be
HA		 2315 * @sk: socket
2316 * @size: memory size to allocate
f8c3bf00 2317 * @amt: pages to allocate
3ab224be
HA		 2318 * @kind: allocation type
2319 *
f8c3bf00 2320 * Similar to __sk_mem_schedule(), but does not update sk_forward_alloc
3ab224be 2321 */
f8c3bf00 2322int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
3ab224be
HA		 2323{
2324 struct proto *prot = sk->sk_prot;
f8c3bf00 2325 long allocated = sk_memory_allocated_add(sk, amt);
e805605c 2326
baac50bb
JW		 2327 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2328 !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
e805605c 2329 goto suppress_allocation;
3ab224be
HA		 2330
2331 /* Under limit. */
e805605c 2332 if (allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2333 sk_leave_memory_pressure(sk);
3ab224be
HA		 2334 return 1;
2335 }
2336
e805605c
JW		 2337 /* Under pressure. */
2338 if (allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2339 sk_enter_memory_pressure(sk);
3ab224be 2340
e805605c
JW		 2341 /* Over hard limit. */
2342 if (allocated > sk_prot_mem_limits(sk, 2))
3ab224be
HA		 2343 goto suppress_allocation;
2344
2345 /* guarantee minimum buffer size under pressure */
2346 if (kind == SK_MEM_RECV) {
a3dcaf17 2347 if (atomic_read(&sk->sk_rmem_alloc) < sk_get_rmem0(sk, prot))
3ab224be 2348 return 1;
180d8cd9 2349
3ab224be 2350 } else { /* SK_MEM_SEND */
a3dcaf17
ED		 2351 int wmem0 = sk_get_wmem0(sk, prot);
2352
3ab224be 2353 if (sk->sk_type == SOCK_STREAM) {
a3dcaf17 2354 if (sk->sk_wmem_queued < wmem0)
3ab224be 2355 return 1;
a3dcaf17 2356 } else if (refcount_read(&sk->sk_wmem_alloc) < wmem0) {
3ab224be 2357 return 1;
a3dcaf17 2358 }
3ab224be
HA		 2359 }
2360
180d8cd9 2361 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2362 int alloc;
2363
180d8cd9 2364 if (!sk_under_memory_pressure(sk))
1748376b 2365 return 1;
180d8cd9
GC		 2366 alloc = sk_sockets_allocated_read_positive(sk);
2367 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2368 sk_mem_pages(sk->sk_wmem_queued +
2369 atomic_read(&sk->sk_rmem_alloc) +
2370 sk->sk_forward_alloc))
2371 return 1;
2372 }
2373
2374suppress_allocation:
2375
2376 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2377 sk_stream_moderate_sndbuf(sk);
2378
2379 /* Fail only if socket is _under_ its sndbuf.
2380 * In this case we cannot block, so that we have to fail.
2381 */
2382 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2383 return 1;
2384 }
2385
3847ce32
SM		 2386 trace_sock_exceed_buf_limit(sk, prot, allocated);
2387
0e90b31f 2388 sk_memory_allocated_sub(sk, amt);
180d8cd9 2389
baac50bb
JW		 2390 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2391 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
e805605c 2392
3ab224be
HA		 2393 return 0;
2394}
f8c3bf00
PA		 2395EXPORT_SYMBOL(__sk_mem_raise_allocated);
2396
2397/**
2398 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2399 * @sk: socket
2400 * @size: memory size to allocate
2401 * @kind: allocation type
2402 *
2403 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2404 * rmem allocation. This function assumes that protocols which have
2405 * memory_pressure use sk_wmem_queued as write buffer accounting.
2406 */
2407int __sk_mem_schedule(struct sock *sk, int size, int kind)
2408{
2409 int ret, amt = sk_mem_pages(size);
2410
2411 sk->sk_forward_alloc += amt << SK_MEM_QUANTUM_SHIFT;
2412 ret = __sk_mem_raise_allocated(sk, size, amt, kind);
2413 if (!ret)
2414 sk->sk_forward_alloc -= amt << SK_MEM_QUANTUM_SHIFT;
2415 return ret;
2416}
3ab224be
HA		 2417EXPORT_SYMBOL(__sk_mem_schedule);
2418
2419/**
f8c3bf00 2420 * __sk_mem_reduce_allocated - reclaim memory_allocated
3ab224be 2421 * @sk: socket
f8c3bf00
PA		 2422 * @amount: number of quanta
2423 *
2424 * Similar to __sk_mem_reclaim(), but does not update sk_forward_alloc
3ab224be 2425 */
f8c3bf00 2426void __sk_mem_reduce_allocated(struct sock *sk, int amount)
3ab224be 2427{
1a24e04e 2428 sk_memory_allocated_sub(sk, amount);
3ab224be 2429
baac50bb
JW		 2430 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2431 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
e805605c 2432
180d8cd9
GC		 2433 if (sk_under_memory_pressure(sk) &&
2434 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2435 sk_leave_memory_pressure(sk);
3ab224be 2436}
f8c3bf00
PA		 2437EXPORT_SYMBOL(__sk_mem_reduce_allocated);
2438
2439/**
2440 * __sk_mem_reclaim - reclaim sk_forward_alloc and memory_allocated
2441 * @sk: socket
2442 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
2443 */
2444void __sk_mem_reclaim(struct sock *sk, int amount)
2445{
2446 amount >>= SK_MEM_QUANTUM_SHIFT;
2447 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
2448 __sk_mem_reduce_allocated(sk, amount);
2449}
3ab224be
HA		 2450EXPORT_SYMBOL(__sk_mem_reclaim);
2451
627d2d6b 2452int sk_set_peek_off(struct sock *sk, int val)
2453{
627d2d6b 2454 sk->sk_peek_off = val;
2455 return 0;
2456}
2457EXPORT_SYMBOL_GPL(sk_set_peek_off);
3ab224be 2458
1da177e4
LT		 2459/*
2460 * Set of default routines for initialising struct proto_ops when
2461 * the protocol does not support a particular function. In certain
2462 * cases where it makes no sense for a protocol to have a "do nothing"
2463 * function, some default processing is provided.
2464 */
2465
2466int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2467{
2468 return -EOPNOTSUPP;
2469}
2a91525c 2470EXPORT_SYMBOL(sock_no_bind);
1da177e4 2471
4ec93edb 2472int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2473 int len, int flags)
2474{
2475 return -EOPNOTSUPP;
2476}
2a91525c 2477EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2478
2479int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2480{
2481 return -EOPNOTSUPP;
2482}
2a91525c 2483EXPORT_SYMBOL(sock_no_socketpair);
1da177e4 2484
cdfbabfb
DH		 2485int sock_no_accept(struct socket *sock, struct socket *newsock, int flags,
2486 bool kern)
1da177e4
LT		 2487{
2488 return -EOPNOTSUPP;
2489}
2a91525c 2490EXPORT_SYMBOL(sock_no_accept);
1da177e4 2491
4ec93edb 2492int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2493 int *len, int peer)
2494{
2495 return -EOPNOTSUPP;
2496}
2a91525c 2497EXPORT_SYMBOL(sock_no_getname);
1da177e4 2498
2a91525c 2499unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2500{
2501 return 0;
2502}
2a91525c 2503EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2504
2505int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2506{
2507 return -EOPNOTSUPP;
2508}
2a91525c 2509EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2510
2511int sock_no_listen(struct socket *sock, int backlog)
2512{
2513 return -EOPNOTSUPP;
2514}
2a91525c 2515EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2516
2517int sock_no_shutdown(struct socket *sock, int how)
2518{
2519 return -EOPNOTSUPP;
2520}
2a91525c 2521EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2522
2523int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2524 char __user *optval, unsigned int optlen)
1da177e4
LT		 2525{
2526 return -EOPNOTSUPP;
2527}
2a91525c 2528EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2529
2530int sock_no_getsockopt(struct socket *sock, int level, int optname,
2531 char __user *optval, int __user *optlen)
2532{
2533 return -EOPNOTSUPP;
2534}
2a91525c 2535EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2536
1b784140 2537int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2538{
2539 return -EOPNOTSUPP;
2540}
2a91525c 2541EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2542
306b13eb
TH		 2543int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *m, size_t len)
2544{
2545 return -EOPNOTSUPP;
2546}
2547EXPORT_SYMBOL(sock_no_sendmsg_locked);
2548
1b784140
YX		 2549int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2550 int flags)
1da177e4
LT		 2551{
2552 return -EOPNOTSUPP;
2553}
2a91525c 2554EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2555
2556int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2557{
2558 /* Mirror missing mmap method error code */
2559 return -ENODEV;
2560}
2a91525c 2561EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2562
2563ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2564{
2565 ssize_t res;
2566 struct msghdr msg = {.msg_flags = flags};
2567 struct kvec iov;
2568 char *kaddr = kmap(page);
2569 iov.iov_base = kaddr + offset;
2570 iov.iov_len = size;
2571 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2572 kunmap(page);
2573 return res;
2574}
2a91525c 2575EXPORT_SYMBOL(sock_no_sendpage);
1da177e4 2576
306b13eb
TH		 2577ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
2578 int offset, size_t size, int flags)
2579{
2580 ssize_t res;
2581 struct msghdr msg = {.msg_flags = flags};
2582 struct kvec iov;
2583 char *kaddr = kmap(page);
2584
2585 iov.iov_base = kaddr + offset;
2586 iov.iov_len = size;
2587 res = kernel_sendmsg_locked(sk, &msg, &iov, 1, size);
2588 kunmap(page);
2589 return res;
2590}
2591EXPORT_SYMBOL(sock_no_sendpage_locked);
2592
1da177e4
LT		 2593/*
2594 * Default Socket Callbacks
2595 */
2596
2597static void sock_def_wakeup(struct sock *sk)
2598{
43815482
ED		 2599 struct socket_wq *wq;
2600
2601 rcu_read_lock();
2602 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2603 if (skwq_has_sleeper(wq))
43815482
ED		 2604 wake_up_interruptible_all(&wq->wait);
2605 rcu_read_unlock();
1da177e4
LT		 2606}
2607
2608static void sock_def_error_report(struct sock *sk)
2609{
43815482
ED		 2610 struct socket_wq *wq;
2611
2612 rcu_read_lock();
2613 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2614 if (skwq_has_sleeper(wq))
43815482 2615 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2616 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2617 rcu_read_unlock();
1da177e4
LT		 2618}
2619
676d2369 2620static void sock_def_readable(struct sock *sk)
1da177e4 2621{
43815482
ED		 2622 struct socket_wq *wq;
2623
2624 rcu_read_lock();
2625 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2626 if (skwq_has_sleeper(wq))
2c6607c6 2627 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2628 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2629 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2630 rcu_read_unlock();
1da177e4
LT		 2631}
2632
2633static void sock_def_write_space(struct sock *sk)
2634{
43815482
ED		 2635 struct socket_wq *wq;
2636
2637 rcu_read_lock();
1da177e4
LT		 2638
2639 /* Do not wake up a writer until he can make "significant"
2640 * progress. --DaveM
2641 */
14afee4b 2642 if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482 2643 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2644 if (skwq_has_sleeper(wq))
43815482 2645 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2646 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2647
2648 /* Should agree with poll, otherwise some programs break */
2649 if (sock_writeable(sk))
8d8ad9d7 2650 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2651 }
2652
43815482 2653 rcu_read_unlock();
1da177e4
LT		 2654}
2655
2656static void sock_def_destruct(struct sock *sk)
2657{
1da177e4
LT		 2658}
2659
2660void sk_send_sigurg(struct sock *sk)
2661{
2662 if (sk->sk_socket && sk->sk_socket->file)
2663 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2664 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2665}
2a91525c 2666EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2667
2668void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2669 unsigned long expires)
2670{
2671 if (!mod_timer(timer, expires))
2672 sock_hold(sk);
2673}
1da177e4
LT		 2674EXPORT_SYMBOL(sk_reset_timer);
2675
2676void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2677{
25cc4ae9 2678 if (del_timer(timer))
1da177e4
LT		 2679 __sock_put(sk);
2680}
1da177e4
LT		 2681EXPORT_SYMBOL(sk_stop_timer);
2682
2683void sock_init_data(struct socket *sock, struct sock *sk)
2684{
581319c5 2685 sk_init_common(sk);
1da177e4
LT		 2686 sk->sk_send_head = NULL;
2687
99767f27 2688 timer_setup(&sk->sk_timer, NULL, 0);
4ec93edb 2689
1da177e4
LT		 2690 sk->sk_allocation = GFP_KERNEL;
2691 sk->sk_rcvbuf = sysctl_rmem_default;
2692 sk->sk_sndbuf = sysctl_wmem_default;
2693 sk->sk_state = TCP_CLOSE;
972692e0 2694 sk_set_socket(sk, sock);
1da177e4
LT		 2695
2696 sock_set_flag(sk, SOCK_ZAPPED);
2697
e71a4783 2698 if (sock) {
1da177e4 2699 sk->sk_type = sock->type;
43815482 2700 sk->sk_wq = sock->wq;
1da177e4 2701 sock->sk = sk;
86741ec2
LC		 2702 sk->sk_uid = SOCK_INODE(sock)->i_uid;
2703 } else {
43815482 2704 sk->sk_wq = NULL;
86741ec2
LC		 2705 sk->sk_uid = make_kuid(sock_net(sk)->user_ns, 0);
2706 }
1da177e4 2707
1da177e4 2708 rwlock_init(&sk->sk_callback_lock);
cdfbabfb
DH		 2709 if (sk->sk_kern_sock)
2710 lockdep_set_class_and_name(
2711 &sk->sk_callback_lock,
2712 af_kern_callback_keys + sk->sk_family,
2713 af_family_kern_clock_key_strings[sk->sk_family]);
2714 else
2715 lockdep_set_class_and_name(
2716 &sk->sk_callback_lock,
443aef0e
PZ		 2717 af_callback_keys + sk->sk_family,
2718 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2719
2720 sk->sk_state_change = sock_def_wakeup;
2721 sk->sk_data_ready = sock_def_readable;
2722 sk->sk_write_space = sock_def_write_space;
2723 sk->sk_error_report = sock_def_error_report;
2724 sk->sk_destruct = sock_def_destruct;
2725
5640f768
ED		 2726 sk->sk_frag.page = NULL;
2727 sk->sk_frag.offset = 0;
ef64a54f 2728 sk->sk_peek_off = -1;
1da177e4 2729
109f6e39
EB		 2730 sk->sk_peer_pid = NULL;
2731 sk->sk_peer_cred = NULL;
1da177e4
LT		 2732 sk->sk_write_pending = 0;
2733 sk->sk_rcvlowat = 1;
2734 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2735 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2736
6c7c98ba 2737 sk->sk_stamp = SK_DEFAULT_STAMP;
52267790 2738 atomic_set(&sk->sk_zckey, 0);
1da177e4 2739
e0d1095a 2740#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2741 sk->sk_napi_id = 0;
64b0dc51 2742 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2743#endif
2744
62748f32 2745 sk->sk_max_pacing_rate = ~0U;
7eec4174 2746 sk->sk_pacing_rate = ~0U;
3a9b76fd 2747 sk->sk_pacing_shift = 10;
70da268b 2748 sk->sk_incoming_cpu = -1;
4dc6dc71
ED		 2749 /*
2750 * Before updating sk_refcnt, we must commit prior changes to memory
2751 * (Documentation/RCU/rculist_nulls.txt for details)
2752 */
2753 smp_wmb();
41c6d650 2754 refcount_set(&sk->sk_refcnt, 1);
33c732c3 2755 atomic_set(&sk->sk_drops, 0);
1da177e4 2756}
2a91525c 2757EXPORT_SYMBOL(sock_init_data);
1da177e4 2758
b5606c2d 2759void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2760{
2761 might_sleep();
a5b5bb9a 2762 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2763 if (sk->sk_lock.owned)
1da177e4 2764 __lock_sock(sk);
d2e9117c 2765 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2766 spin_unlock(&sk->sk_lock.slock);
2767 /*
2768 * The sk_lock has mutex_lock() semantics here:
2769 */
fcc70d5f 2770 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2771 local_bh_enable();
1da177e4 2772}
fcc70d5f 2773EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2774
b5606c2d 2775void release_sock(struct sock *sk)
1da177e4 2776{
a5b5bb9a 2777 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2778 if (sk->sk_backlog.tail)
2779 __release_sock(sk);
46d3ceab 2780
c3f9b018
ED		 2781 /* Warning : release_cb() might need to release sk ownership,
2782 * ie call sock_release_ownership(sk) before us.
2783 */
46d3ceab
ED		 2784 if (sk->sk_prot->release_cb)
2785 sk->sk_prot->release_cb(sk);
2786
c3f9b018 2787 sock_release_ownership(sk);
a5b5bb9a
IM		 2788 if (waitqueue_active(&sk->sk_lock.wq))
2789 wake_up(&sk->sk_lock.wq);
2790 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2791}
2792EXPORT_SYMBOL(release_sock);
2793
8a74ad60
ED		 2794/**
2795 * lock_sock_fast - fast version of lock_sock
2796 * @sk: socket
2797 *
2798 * This version should be used for very small section, where process wont block
d651983d
MCC		 2799 * return false if fast path is taken:
2800 *
8a74ad60 2801 * sk_lock.slock locked, owned = 0, BH disabled
d651983d
MCC		 2802 *
2803 * return true if slow path is taken:
2804 *
8a74ad60
ED		 2805 * sk_lock.slock unlocked, owned = 1, BH enabled
2806 */
2807bool lock_sock_fast(struct sock *sk)
2808{
2809 might_sleep();
2810 spin_lock_bh(&sk->sk_lock.slock);
2811
2812 if (!sk->sk_lock.owned)
2813 /*
2814 * Note : We must disable BH
2815 */
2816 return false;
2817
2818 __lock_sock(sk);
2819 sk->sk_lock.owned = 1;
2820 spin_unlock(&sk->sk_lock.slock);
2821 /*
2822 * The sk_lock has mutex_lock() semantics here:
2823 */
2824 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2825 local_bh_enable();
2826 return true;
2827}
2828EXPORT_SYMBOL(lock_sock_fast);
2829
1da177e4 2830int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2831{
b7aa0bf7 2832 struct timeval tv;
1da177e4 2833 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2834 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2835 tv = ktime_to_timeval(sk->sk_stamp);
2836 if (tv.tv_sec == -1)
1da177e4 2837 return -ENOENT;
b7aa0bf7
ED		 2838 if (tv.tv_sec == 0) {
2839 sk->sk_stamp = ktime_get_real();
2840 tv = ktime_to_timeval(sk->sk_stamp);
2841 }
2842 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2843}
1da177e4
LT		 2844EXPORT_SYMBOL(sock_get_timestamp);
2845
ae40eb1e
ED		 2846int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2847{
2848 struct timespec ts;
2849 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2850 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2851 ts = ktime_to_timespec(sk->sk_stamp);
2852 if (ts.tv_sec == -1)
2853 return -ENOENT;
2854 if (ts.tv_sec == 0) {
2855 sk->sk_stamp = ktime_get_real();
2856 ts = ktime_to_timespec(sk->sk_stamp);
2857 }
2858 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2859}
2860EXPORT_SYMBOL(sock_get_timestampns);
2861
20d49473 2862void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2863{
20d49473 2864 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2865 unsigned long previous_flags = sk->sk_flags;
2866
20d49473
PO		 2867 sock_set_flag(sk, flag);
2868 /*
2869 * we just set one of the two flags which require net
2870 * time stamping, but time stamping might have been on
2871 * already because of the other one
2872 */
080a270f
HFS		 2873 if (sock_needs_netstamp(sk) &&
2874 !(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2875 net_enable_timestamp();
1da177e4
LT		 2876 }
2877}
1da177e4 2878
cb820f8e
RC		 2879int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2880 int level, int type)
2881{
2882 struct sock_exterr_skb *serr;
364a9e93 2883 struct sk_buff *skb;
cb820f8e
RC		 2884 int copied, err;
2885
2886 err = -EAGAIN;
364a9e93 2887 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 2888 if (skb == NULL)
2889 goto out;
2890
2891 copied = skb->len;
2892 if (copied > len) {
2893 msg->msg_flags |= MSG_TRUNC;
2894 copied = len;
2895 }
51f3d02b 2896 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 2897 if (err)
2898 goto out_free_skb;
2899
2900 sock_recv_timestamp(msg, sk, skb);
2901
2902 serr = SKB_EXT_ERR(skb);
2903 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2904
2905 msg->msg_flags |= MSG_ERRQUEUE;
2906 err = copied;
2907
cb820f8e
RC		 2908out_free_skb:
2909 kfree_skb(skb);
2910out:
2911 return err;
2912}
2913EXPORT_SYMBOL(sock_recv_errqueue);
2914
1da177e4
LT		 2915/*
2916 * Get a socket option on an socket.
2917 *
2918 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2919 * asynchronous errors should be reported by getsockopt. We assume
2920 * this means if you specify SO_ERROR (otherwise whats the point of it).
2921 */
2922int sock_common_getsockopt(struct socket *sock, int level, int optname,
2923 char __user *optval, int __user *optlen)
2924{
2925 struct sock *sk = sock->sk;
2926
2927 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2928}
1da177e4
LT		 2929EXPORT_SYMBOL(sock_common_getsockopt);
2930
3fdadf7d 2931#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2932int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2933 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2934{
2935 struct sock *sk = sock->sk;
2936
1e51f951 2937 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2938 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2939 optval, optlen);
3fdadf7d
DM		 2940 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2941}
2942EXPORT_SYMBOL(compat_sock_common_getsockopt);
2943#endif
2944
1b784140
YX		 2945int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2946 int flags)
1da177e4
LT		 2947{
2948 struct sock *sk = sock->sk;
2949 int addr_len = 0;
2950 int err;
2951
1b784140 2952 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 2953 flags & ~MSG_DONTWAIT, &addr_len);
2954 if (err >= 0)
2955 msg->msg_namelen = addr_len;
2956 return err;
2957}
1da177e4
LT		 2958EXPORT_SYMBOL(sock_common_recvmsg);
2959
2960/*
2961 * Set socket options on an inet socket.
2962 */
2963int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2964 char __user *optval, unsigned int optlen)
1da177e4
LT		 2965{
2966 struct sock *sk = sock->sk;
2967
2968 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2969}
1da177e4
LT		 2970EXPORT_SYMBOL(sock_common_setsockopt);
2971
3fdadf7d 2972#ifdef CONFIG_COMPAT
543d9cfe 2973int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2974 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2975{
2976 struct sock *sk = sock->sk;
2977
543d9cfe
ACM		 2978 if (sk->sk_prot->compat_setsockopt != NULL)
2979 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2980 optval, optlen);
3fdadf7d
DM		 2981 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2982}
2983EXPORT_SYMBOL(compat_sock_common_setsockopt);
2984#endif
2985
1da177e4
LT		 2986void sk_common_release(struct sock *sk)
2987{
2988 if (sk->sk_prot->destroy)
2989 sk->sk_prot->destroy(sk);
2990
2991 /*
2992 * Observation: when sock_common_release is called, processes have
2993 * no access to socket. But net still has.
2994 * Step one, detach it from networking:
2995 *
2996 * A. Remove from hash tables.
2997 */
2998
2999 sk->sk_prot->unhash(sk);
3000
3001 /*
3002 * In this point socket cannot receive new packets, but it is possible
3003 * that some packets are in flight because some CPU runs receiver and
3004 * did hash table lookup before we unhashed socket. They will achieve
3005 * receive queue and will be purged by socket destructor.
3006 *
3007 * Also we still have packets pending on receive queue and probably,
3008 * our own packets waiting in device queues. sock_destroy will drain
3009 * receive queue, but transmitted packets will delay socket destruction
3010 * until the last reference will be released.
3011 */
3012
3013 sock_orphan(sk);
3014
3015 xfrm_sk_free_policy(sk);
3016
e6848976 3017 sk_refcnt_debug_release(sk);
5640f768 3018
1da177e4
LT		 3019 sock_put(sk);
3020}
1da177e4
LT		 3021EXPORT_SYMBOL(sk_common_release);
3022
a2d133b1
JH		 3023void sk_get_meminfo(const struct sock *sk, u32 *mem)
3024{
3025 memset(mem, 0, sizeof(*mem) * SK_MEMINFO_VARS);
3026
3027 mem[SK_MEMINFO_RMEM_ALLOC] = sk_rmem_alloc_get(sk);
3028 mem[SK_MEMINFO_RCVBUF] = sk->sk_rcvbuf;
3029 mem[SK_MEMINFO_WMEM_ALLOC] = sk_wmem_alloc_get(sk);
3030 mem[SK_MEMINFO_SNDBUF] = sk->sk_sndbuf;
3031 mem[SK_MEMINFO_FWD_ALLOC] = sk->sk_forward_alloc;
3032 mem[SK_MEMINFO_WMEM_QUEUED] = sk->sk_wmem_queued;
3033 mem[SK_MEMINFO_OPTMEM] = atomic_read(&sk->sk_omem_alloc);
3034 mem[SK_MEMINFO_BACKLOG] = sk->sk_backlog.len;
3035 mem[SK_MEMINFO_DROPS] = atomic_read(&sk->sk_drops);
3036}
3037
13ff3d6f
PE		 3038#ifdef CONFIG_PROC_FS
3039#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 3040struct prot_inuse {
3041 int val[PROTO_INUSE_NR];
3042};
13ff3d6f
PE		 3043
3044static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159 3045
70ee1159
PE		 3046void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
3047{
d6d9ca0f 3048 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 3049}
3050EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
3051
3052int sock_prot_inuse_get(struct net *net, struct proto *prot)
3053{
3054 int cpu, idx = prot->inuse_idx;
3055 int res = 0;
3056
3057 for_each_possible_cpu(cpu)
3058 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
3059
3060 return res >= 0 ? res : 0;
3061}
3062EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
3063
2c8c1e72 3064static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 3065{
3066 net->core.inuse = alloc_percpu(struct prot_inuse);
3067 return net->core.inuse ? 0 : -ENOMEM;
3068}
3069
2c8c1e72 3070static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 3071{
3072 free_percpu(net->core.inuse);
3073}
3074
3075static struct pernet_operations net_inuse_ops = {
3076 .init = sock_inuse_init_net,
3077 .exit = sock_inuse_exit_net,
3078};
3079
3080static __init int net_inuse_init(void)
3081{
3082 if (register_pernet_subsys(&net_inuse_ops))
3083 panic("Cannot initialize net inuse counters");
3084
3085 return 0;
3086}
3087
3088core_initcall(net_inuse_init);
13ff3d6f
PE		 3089
3090static void assign_proto_idx(struct proto *prot)
3091{
3092 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
3093
3094 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 3095 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 3096 return;
3097 }
3098
3099 set_bit(prot->inuse_idx, proto_inuse_idx);
3100}
3101
3102static void release_proto_idx(struct proto *prot)
3103{
3104 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
3105 clear_bit(prot->inuse_idx, proto_inuse_idx);
3106}
3107#else
3108static inline void assign_proto_idx(struct proto *prot)
3109{
3110}
3111
3112static inline void release_proto_idx(struct proto *prot)
3113{
3114}
3115#endif
3116
0159dfd3
ED		 3117static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
3118{
3119 if (!rsk_prot)
3120 return;
3121 kfree(rsk_prot->slab_name);
3122 rsk_prot->slab_name = NULL;
adf78eda
JL		 3123 kmem_cache_destroy(rsk_prot->slab);
3124 rsk_prot->slab = NULL;
0159dfd3
ED		 3125}
3126
3127static int req_prot_init(const struct proto *prot)
3128{
3129 struct request_sock_ops *rsk_prot = prot->rsk_prot;
3130
3131 if (!rsk_prot)
3132 return 0;
3133
3134 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
3135 prot->name);
3136 if (!rsk_prot->slab_name)
3137 return -ENOMEM;
3138
3139 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
3140 rsk_prot->obj_size, 0,
e96f78ab 3141 prot->slab_flags, NULL);
0159dfd3
ED		 3142
3143 if (!rsk_prot->slab) {
3144 pr_crit("%s: Can't create request sock SLAB cache!\n",
3145 prot->name);
3146 return -ENOMEM;
3147 }
3148 return 0;
3149}
3150
b733c007
PE		 3151int proto_register(struct proto *prot, int alloc_slab)
3152{
1da177e4
LT		 3153 if (alloc_slab) {
3154 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 3155 SLAB_HWCACHE_ALIGN | prot->slab_flags,
3156 NULL);
1da177e4
LT		 3157
3158 if (prot->slab == NULL) {
e005d193
JP		 3159 pr_crit("%s: Can't create sock SLAB cache!\n",
3160 prot->name);
60e7663d 3161 goto out;
1da177e4 3162 }
2e6599cb 3163
0159dfd3
ED		 3164 if (req_prot_init(prot))
3165 goto out_free_request_sock_slab;
8feaf0c0 3166
6d6ee43e 3167 if (prot->twsk_prot != NULL) {
faf23422 3168 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 3169
7e56b5d6 3170 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 3171 goto out_free_request_sock_slab;
3172
6d6ee43e 3173 prot->twsk_prot->twsk_slab =
7e56b5d6 3174 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 3175 prot->twsk_prot->twsk_obj_size,
3ab5aee7 3176 0,
52db70dc 3177 prot->slab_flags,
20c2df83 3178 NULL);
6d6ee43e 3179 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 3180 goto out_free_timewait_sock_slab_name;
3181 }
1da177e4
LT		 3182 }
3183
36b77a52 3184 mutex_lock(&proto_list_mutex);
1da177e4 3185 list_add(&prot->node, &proto_list);
13ff3d6f 3186 assign_proto_idx(prot);
36b77a52 3187 mutex_unlock(&proto_list_mutex);
b733c007
PE		 3188 return 0;
3189
8feaf0c0 3190out_free_timewait_sock_slab_name:
7e56b5d6 3191 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 3192out_free_request_sock_slab:
0159dfd3
ED		 3193 req_prot_cleanup(prot->rsk_prot);
3194
2e6599cb
ACM		 3195 kmem_cache_destroy(prot->slab);
3196 prot->slab = NULL;
b733c007
PE		 3197out:
3198 return -ENOBUFS;
1da177e4 3199}
1da177e4
LT		 3200EXPORT_SYMBOL(proto_register);
3201
3202void proto_unregister(struct proto *prot)
3203{
36b77a52 3204 mutex_lock(&proto_list_mutex);
13ff3d6f 3205 release_proto_idx(prot);
0a3f4358 3206 list_del(&prot->node);
36b77a52 3207 mutex_unlock(&proto_list_mutex);
1da177e4 3208
adf78eda
JL		 3209 kmem_cache_destroy(prot->slab);
3210 prot->slab = NULL;
1da177e4 3211
0159dfd3 3212 req_prot_cleanup(prot->rsk_prot);
2e6599cb 3213
6d6ee43e 3214 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 3215 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 3216 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 3217 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 3218 }
1da177e4 3219}
1da177e4
LT		 3220EXPORT_SYMBOL(proto_unregister);
3221
3222#ifdef CONFIG_PROC_FS
1da177e4 3223static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 3224 __acquires(proto_list_mutex)
1da177e4 3225{
36b77a52 3226 mutex_lock(&proto_list_mutex);
60f0438a 3227 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 3228}
3229
3230static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3231{
60f0438a 3232 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 3233}
3234
3235static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 3236 __releases(proto_list_mutex)
1da177e4 3237{
36b77a52 3238 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 3239}
3240
3241static char proto_method_implemented(const void *method)
3242{
3243 return method == NULL ? 'n' : 'y';
3244}
180d8cd9
GC		 3245static long sock_prot_memory_allocated(struct proto *proto)
3246{
cb75a36c 3247 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 3248}
3249
3250static char *sock_prot_memory_pressure(struct proto *proto)
3251{
3252 return proto->memory_pressure != NULL ?
3253 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
3254}
1da177e4
LT		 3255
3256static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
3257{
180d8cd9 3258
8d987e5c 3259 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 3260 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
3261 proto->name,
3262 proto->obj_size,
14e943db 3263 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 3264 sock_prot_memory_allocated(proto),
3265 sock_prot_memory_pressure(proto),
1da177e4
LT		 3266 proto->max_header,
3267 proto->slab == NULL ? "no" : "yes",
3268 module_name(proto->owner),
3269 proto_method_implemented(proto->close),
3270 proto_method_implemented(proto->connect),
3271 proto_method_implemented(proto->disconnect),
3272 proto_method_implemented(proto->accept),
3273 proto_method_implemented(proto->ioctl),
3274 proto_method_implemented(proto->init),
3275 proto_method_implemented(proto->destroy),
3276 proto_method_implemented(proto->shutdown),
3277 proto_method_implemented(proto->setsockopt),
3278 proto_method_implemented(proto->getsockopt),
3279 proto_method_implemented(proto->sendmsg),
3280 proto_method_implemented(proto->recvmsg),
3281 proto_method_implemented(proto->sendpage),
3282 proto_method_implemented(proto->bind),
3283 proto_method_implemented(proto->backlog_rcv),
3284 proto_method_implemented(proto->hash),
3285 proto_method_implemented(proto->unhash),
3286 proto_method_implemented(proto->get_port),
3287 proto_method_implemented(proto->enter_memory_pressure));
3288}
3289
3290static int proto_seq_show(struct seq_file *seq, void *v)
3291{
60f0438a 3292 if (v == &proto_list)
1da177e4
LT		 3293 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3294 "protocol",
3295 "size",
3296 "sockets",
3297 "memory",
3298 "press",
3299 "maxhdr",
3300 "slab",
3301 "module",
3302 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3303 else
60f0438a 3304 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 3305 return 0;
3306}
3307
f690808e 3308static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 3309 .start = proto_seq_start,
3310 .next = proto_seq_next,
3311 .stop = proto_seq_stop,
3312 .show = proto_seq_show,
3313};
3314
3315static int proto_seq_open(struct inode *inode, struct file *file)
3316{
14e943db
ED		 3317 return seq_open_net(inode, file, &proto_seq_ops,
3318 sizeof(struct seq_net_private));
1da177e4
LT		 3319}
3320
9a32144e 3321static const struct file_operations proto_seq_fops = {
1da177e4
LT		 3322 .owner = THIS_MODULE,
3323 .open = proto_seq_open,
3324 .read = seq_read,
3325 .llseek = seq_lseek,
14e943db
ED		 3326 .release = seq_release_net,
3327};
3328
3329static __net_init int proto_init_net(struct net *net)
3330{
d4beaa66 3331 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 3332 return -ENOMEM;
3333
3334 return 0;
3335}
3336
3337static __net_exit void proto_exit_net(struct net *net)
3338{
ece31ffd 3339 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 3340}
3341
3342
3343static __net_initdata struct pernet_operations proto_net_ops = {
3344 .init = proto_init_net,
3345 .exit = proto_exit_net,
1da177e4
LT		 3346};
3347
3348static int __init proto_init(void)
3349{
14e943db 3350 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 3351}
3352
3353subsys_initcall(proto_init);
3354
3355#endif /* PROC_FS */
7db6b048
SS		 3356
3357#ifdef CONFIG_NET_RX_BUSY_POLL
3358bool sk_busy_loop_end(void *p, unsigned long start_time)
3359{
3360 struct sock *sk = p;
3361
3362 return !skb_queue_empty(&sk->sk_receive_queue) ||
3363 sk_busy_loop_timeout(sk, start_time);
3364}
3365EXPORT_SYMBOL(sk_busy_loop_end);
3366#endif /* CONFIG_NET_RX_BUSY_POLL */
Linux 6.x block layer and io_uring tree(s)