projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net: add sock_enable_timestamps
[linux-block.git] / net / core / sock.c
CommitLineData
2874c5fd 1// SPDX-License-Identifier: GPL-2.0-or-later
1da177e4
LT		 2/*
3 * INET An implementation of the TCP/IP protocol suite for the LINUX
4 * operating system. INET is implemented using the BSD Socket
5 * interface as the means of communication with the user level.
6 *
7 * Generic socket support routines. Memory allocators, socket lock/release
8 * handler for protocols to use and generic option handler.
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:
1da177e4
LT		 84 */
85
e005d193
JP		 86#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
87
80b14dee 88#include <asm/unaligned.h>
4fc268d2 89#include <linux/capability.h>
1da177e4 90#include <linux/errno.h>
cb820f8e 91#include <linux/errqueue.h>
1da177e4
LT		 92#include <linux/types.h>
93#include <linux/socket.h>
94#include <linux/in.h>
95#include <linux/kernel.h>
1da177e4
LT		 96#include <linux/module.h>
97#include <linux/proc_fs.h>
98#include <linux/seq_file.h>
99#include <linux/sched.h>
f1083048 100#include <linux/sched/mm.h>
1da177e4
LT		 101#include <linux/timer.h>
102#include <linux/string.h>
103#include <linux/sockios.h>
104#include <linux/net.h>
105#include <linux/mm.h>
106#include <linux/slab.h>
107#include <linux/interrupt.h>
108#include <linux/poll.h>
109#include <linux/tcp.h>
110#include <linux/init.h>
a1f8e7f7 111#include <linux/highmem.h>
3f551f94 112#include <linux/user_namespace.h>
c5905afb 113#include <linux/static_key.h>
3969eb38 114#include <linux/memcontrol.h>
8c1ae10d 115#include <linux/prefetch.h>
1da177e4 116
7c0f6ba6 117#include <linux/uaccess.h>
1da177e4
LT		 118
119#include <linux/netdevice.h>
120#include <net/protocol.h>
121#include <linux/skbuff.h>
457c4cbc 122#include <net/net_namespace.h>
2e6599cb 123#include <net/request_sock.h>
1da177e4 124#include <net/sock.h>
20d49473 125#include <linux/net_tstamp.h>
1da177e4
LT		 126#include <net/xfrm.h>
127#include <linux/ipsec.h>
f8451725 128#include <net/cls_cgroup.h>
5bc1421e 129#include <net/netprio_cgroup.h>
eb4cb008 130#include <linux/sock_diag.h>
1da177e4
LT		 131
132#include <linux/filter.h>
538950a1 133#include <net/sock_reuseport.h>
6ac99e8f 134#include <net/bpf_sk_storage.h>
1da177e4 135
3847ce32
SM		 136#include <trace/events/sock.h>
137
1da177e4 138#include <net/tcp.h>
076bb0c8 139#include <net/busy_poll.h>
06021292 140
36b77a52 141static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 142static LIST_HEAD(proto_list);
143
648845ab
TZ		 144static void sock_inuse_add(struct net *net, int val);
145
a3b299da
EB		 146/**
147 * sk_ns_capable - General socket capability test
148 * @sk: Socket to use a capability on or through
149 * @user_ns: The user namespace of the capability to use
150 * @cap: The capability to use
151 *
152 * Test to see if the opener of the socket had when the socket was
153 * created and the current process has the capability @cap in the user
154 * namespace @user_ns.
155 */
156bool sk_ns_capable(const struct sock *sk,
157 struct user_namespace *user_ns, int cap)
158{
159 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
160 ns_capable(user_ns, cap);
161}
162EXPORT_SYMBOL(sk_ns_capable);
163
164/**
165 * sk_capable - Socket global capability test
166 * @sk: Socket to use a capability on or through
e793c0f7 167 * @cap: The global capability to use
a3b299da
EB		 168 *
169 * Test to see if the opener of the socket had when the socket was
170 * created and the current process has the capability @cap in all user
171 * namespaces.
172 */
173bool sk_capable(const struct sock *sk, int cap)
174{
175 return sk_ns_capable(sk, &init_user_ns, cap);
176}
177EXPORT_SYMBOL(sk_capable);
178
179/**
180 * sk_net_capable - Network namespace socket capability test
181 * @sk: Socket to use a capability on or through
182 * @cap: The capability to use
183 *
e793c0f7 184 * Test to see if the opener of the socket had when the socket was created
a3b299da
EB		 185 * and the current process has the capability @cap over the network namespace
186 * the socket is a member of.
187 */
188bool sk_net_capable(const struct sock *sk, int cap)
189{
190 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
191}
192EXPORT_SYMBOL(sk_net_capable);
193
da21f24d
IM		 194/*
195 * Each address family might have different locking rules, so we have
cdfbabfb
DH		 196 * one slock key per address family and separate keys for internal and
197 * userspace sockets.
da21f24d 198 */
a5b5bb9a 199static struct lock_class_key af_family_keys[AF_MAX];
cdfbabfb 200static struct lock_class_key af_family_kern_keys[AF_MAX];
a5b5bb9a 201static struct lock_class_key af_family_slock_keys[AF_MAX];
cdfbabfb 202static struct lock_class_key af_family_kern_slock_keys[AF_MAX];
a5b5bb9a 203
a5b5bb9a
IM		 204/*
205 * Make lock validator output more readable. (we pre-construct these
206 * strings build-time, so that runtime initialization of socket
207 * locks is fast):
208 */
cdfbabfb
DH		 209
210#define _sock_locks(x) \
211 x "AF_UNSPEC", x "AF_UNIX" , x "AF_INET" , \
212 x "AF_AX25" , x "AF_IPX" , x "AF_APPLETALK", \
213 x "AF_NETROM", x "AF_BRIDGE" , x "AF_ATMPVC" , \
214 x "AF_X25" , x "AF_INET6" , x "AF_ROSE" , \
215 x "AF_DECnet", x "AF_NETBEUI" , x "AF_SECURITY" , \
216 x "AF_KEY" , x "AF_NETLINK" , x "AF_PACKET" , \
217 x "AF_ASH" , x "AF_ECONET" , x "AF_ATMSVC" , \
218 x "AF_RDS" , x "AF_SNA" , x "AF_IRDA" , \
219 x "AF_PPPOX" , x "AF_WANPIPE" , x "AF_LLC" , \
220 x "27" , x "28" , x "AF_CAN" , \
221 x "AF_TIPC" , x "AF_BLUETOOTH", x "IUCV" , \
222 x "AF_RXRPC" , x "AF_ISDN" , x "AF_PHONET" , \
223 x "AF_IEEE802154", x "AF_CAIF" , x "AF_ALG" , \
224 x "AF_NFC" , x "AF_VSOCK" , x "AF_KCM" , \
68e8b849
BT		 225 x "AF_QIPCRTR", x "AF_SMC" , x "AF_XDP" , \
226 x "AF_MAX"
cdfbabfb 227
36cbd3dc 228static const char *const af_family_key_strings[AF_MAX+1] = {
cdfbabfb 229 _sock_locks("sk_lock-")
a5b5bb9a 230};
36cbd3dc 231static const char *const af_family_slock_key_strings[AF_MAX+1] = {
cdfbabfb 232 _sock_locks("slock-")
a5b5bb9a 233};
36cbd3dc 234static const char *const af_family_clock_key_strings[AF_MAX+1] = {
cdfbabfb
DH		 235 _sock_locks("clock-")
236};
237
238static const char *const af_family_kern_key_strings[AF_MAX+1] = {
239 _sock_locks("k-sk_lock-")
240};
241static const char *const af_family_kern_slock_key_strings[AF_MAX+1] = {
242 _sock_locks("k-slock-")
243};
244static const char *const af_family_kern_clock_key_strings[AF_MAX+1] = {
245 _sock_locks("k-clock-")
443aef0e 246};
581319c5 247static const char *const af_family_rlock_key_strings[AF_MAX+1] = {
6b431d50 248 _sock_locks("rlock-")
581319c5
PA		 249};
250static const char *const af_family_wlock_key_strings[AF_MAX+1] = {
6b431d50 251 _sock_locks("wlock-")
581319c5
PA		 252};
253static const char *const af_family_elock_key_strings[AF_MAX+1] = {
6b431d50 254 _sock_locks("elock-")
581319c5 255};
da21f24d
IM		 256
257/*
581319c5 258 * sk_callback_lock and sk queues locking rules are per-address-family,
da21f24d
IM		 259 * so split the lock classes by using a per-AF key:
260 */
261static struct lock_class_key af_callback_keys[AF_MAX];
581319c5
PA		 262static struct lock_class_key af_rlock_keys[AF_MAX];
263static struct lock_class_key af_wlock_keys[AF_MAX];
264static struct lock_class_key af_elock_keys[AF_MAX];
cdfbabfb 265static struct lock_class_key af_kern_callback_keys[AF_MAX];
da21f24d 266
1da177e4 267/* Run time adjustable parameters. */
ab32ea5d 268__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 269EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 270__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 271EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 272__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
273__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 274
25985edc 275/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 276int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 277EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 278
b245be1f
WB		 279int sysctl_tstamp_allow_data __read_mostly = 1;
280
a7950ae8
DB		 281DEFINE_STATIC_KEY_FALSE(memalloc_socks_key);
282EXPORT_SYMBOL_GPL(memalloc_socks_key);
c93bdd0e 283
7cb02404
MG		 284/**
285 * sk_set_memalloc - sets %SOCK_MEMALLOC
286 * @sk: socket to set it on
287 *
288 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
289 * It's the responsibility of the admin to adjust min_free_kbytes
290 * to meet the requirements
291 */
292void sk_set_memalloc(struct sock *sk)
293{
294 sock_set_flag(sk, SOCK_MEMALLOC);
295 sk->sk_allocation |= __GFP_MEMALLOC;
a7950ae8 296 static_branch_inc(&memalloc_socks_key);
7cb02404
MG		 297}
298EXPORT_SYMBOL_GPL(sk_set_memalloc);
299
300void sk_clear_memalloc(struct sock *sk)
301{
302 sock_reset_flag(sk, SOCK_MEMALLOC);
303 sk->sk_allocation &= ~__GFP_MEMALLOC;
a7950ae8 304 static_branch_dec(&memalloc_socks_key);
c76562b6
MG		 305
306 /*
307 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
5d753610
MG		 308 * progress of swapping. SOCK_MEMALLOC may be cleared while
309 * it has rmem allocations due to the last swapfile being deactivated
310 * but there is a risk that the socket is unusable due to exceeding
311 * the rmem limits. Reclaim the reserves and obey rmem limits again.
c76562b6 312 */
5d753610 313 sk_mem_reclaim(sk);
7cb02404
MG		 314}
315EXPORT_SYMBOL_GPL(sk_clear_memalloc);
316
b4b9e355
MG		 317int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
318{
319 int ret;
f1083048 320 unsigned int noreclaim_flag;
b4b9e355
MG		 321
322 /* these should have been dropped before queueing */
323 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
324
f1083048 325 noreclaim_flag = memalloc_noreclaim_save();
b4b9e355 326 ret = sk->sk_backlog_rcv(sk, skb);
f1083048 327 memalloc_noreclaim_restore(noreclaim_flag);
b4b9e355
MG		 328
329 return ret;
330}
331EXPORT_SYMBOL(__sk_backlog_rcv);
332
a9beb86a 333static int sock_get_timeout(long timeo, void *optval, bool old_timeval)
fe0c72f3 334{
a9beb86a 335 struct __kernel_sock_timeval tv;
fe0c72f3
AB		 336
337 if (timeo == MAX_SCHEDULE_TIMEOUT) {
338 tv.tv_sec = 0;
339 tv.tv_usec = 0;
340 } else {
341 tv.tv_sec = timeo / HZ;
342 tv.tv_usec = ((timeo % HZ) * USEC_PER_SEC) / HZ;
343 }
344
e6986423 345 if (old_timeval && in_compat_syscall() && !COMPAT_USE_64BIT_TIME) {
fe0c72f3
AB		 346 struct old_timeval32 tv32 = { tv.tv_sec, tv.tv_usec };
347 *(struct old_timeval32 *)optval = tv32;
348 return sizeof(tv32);
349 }
350
a9beb86a
DD		 351 if (old_timeval) {
352 struct __kernel_old_timeval old_tv;
353 old_tv.tv_sec = tv.tv_sec;
354 old_tv.tv_usec = tv.tv_usec;
355 *(struct __kernel_old_timeval *)optval = old_tv;
28e72b26 356 return sizeof(old_tv);
a9beb86a
DD		 357 }
358
28e72b26
VC		 359 *(struct __kernel_sock_timeval *)optval = tv;
360 return sizeof(tv);
fe0c72f3
AB		 361}
362
a9beb86a 363static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen, bool old_timeval)
1da177e4 364{
a9beb86a 365 struct __kernel_sock_timeval tv;
1da177e4 366
e6986423 367 if (old_timeval && in_compat_syscall() && !COMPAT_USE_64BIT_TIME) {
fe0c72f3
AB		 368 struct old_timeval32 tv32;
369
370 if (optlen < sizeof(tv32))
371 return -EINVAL;
372
373 if (copy_from_user(&tv32, optval, sizeof(tv32)))
374 return -EFAULT;
375 tv.tv_sec = tv32.tv_sec;
376 tv.tv_usec = tv32.tv_usec;
a9beb86a
DD		 377 } else if (old_timeval) {
378 struct __kernel_old_timeval old_tv;
379
380 if (optlen < sizeof(old_tv))
381 return -EINVAL;
382 if (copy_from_user(&old_tv, optval, sizeof(old_tv)))
383 return -EFAULT;
384 tv.tv_sec = old_tv.tv_sec;
385 tv.tv_usec = old_tv.tv_usec;
fe0c72f3
AB		 386 } else {
387 if (optlen < sizeof(tv))
388 return -EINVAL;
389 if (copy_from_user(&tv, optval, sizeof(tv)))
390 return -EFAULT;
391 }
ba78073e
VA		 392 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
393 return -EDOM;
1da177e4 394
ba78073e 395 if (tv.tv_sec < 0) {
6f11df83
AM		 396 static int warned __read_mostly;
397
ba78073e 398 *timeo_p = 0;
50aab54f 399 if (warned < 10 && net_ratelimit()) {
ba78073e 400 warned++;
e005d193
JP		 401 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
402 __func__, current->comm, task_pid_nr(current));
50aab54f 403 }
ba78073e
VA		 404 return 0;
405 }
1da177e4
LT		 406 *timeo_p = MAX_SCHEDULE_TIMEOUT;
407 if (tv.tv_sec == 0 && tv.tv_usec == 0)
408 return 0;
a9beb86a
DD		 409 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT / HZ - 1))
410 *timeo_p = tv.tv_sec * HZ + DIV_ROUND_UP((unsigned long)tv.tv_usec, USEC_PER_SEC / HZ);
1da177e4
LT		 411 return 0;
412}
413
414static void sock_warn_obsolete_bsdism(const char *name)
415{
416 static int warned;
417 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 418 if (strcmp(warncomm, current->comm) && warned < 5) {
419 strcpy(warncomm, current->comm);
e005d193
JP		 420 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
421 warncomm, name);
1da177e4
LT		 422 warned++;
423 }
424}
425
080a270f
HFS		 426static bool sock_needs_netstamp(const struct sock *sk)
427{
428 switch (sk->sk_family) {
429 case AF_UNSPEC:
430 case AF_UNIX:
431 return false;
432 default:
433 return true;
434 }
435}
436
08e29af3 437static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 438{
08e29af3
ED		 439 if (sk->sk_flags & flags) {
440 sk->sk_flags &= ~flags;
080a270f
HFS		 441 if (sock_needs_netstamp(sk) &&
442 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 443 net_disable_timestamp();
1da177e4
LT		 444 }
445}
446
447
e6afc8ac 448int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
f0088a50 449{
3b885787
NH		 450 unsigned long flags;
451 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 452
0fd7bac6 453 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 454 atomic_inc(&sk->sk_drops);
3847ce32 455 trace_sock_rcvqueue_full(sk, skb);
766e9037 456 return -ENOMEM;
f0088a50
DV		 457 }
458
c76562b6 459 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 460 atomic_inc(&sk->sk_drops);
461 return -ENOBUFS;
3ab224be
HA		 462 }
463
f0088a50
DV		 464 skb->dev = NULL;
465 skb_set_owner_r(skb, sk);
49ad9599 466
7fee226a
ED		 467 /* we escape from rcu protected region, make sure we dont leak
468 * a norefcounted dst
469 */
470 skb_dst_force(skb);
471
3b885787 472 spin_lock_irqsave(&list->lock, flags);
3bc3b96f 473 sock_skb_set_dropcount(sk, skb);
3b885787
NH		 474 __skb_queue_tail(list, skb);
475 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 476
477 if (!sock_flag(sk, SOCK_DEAD))
676d2369 478 sk->sk_data_ready(sk);
766e9037 479 return 0;
f0088a50 480}
e6afc8ac 481EXPORT_SYMBOL(__sock_queue_rcv_skb);
482
483int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
484{
485 int err;
486
487 err = sk_filter(sk, skb);
488 if (err)
489 return err;
490
491 return __sock_queue_rcv_skb(sk, skb);
492}
f0088a50
DV		 493EXPORT_SYMBOL(sock_queue_rcv_skb);
494
4f0c40d9 495int __sk_receive_skb(struct sock *sk, struct sk_buff *skb,
c3f24cfb 496 const int nested, unsigned int trim_cap, bool refcounted)
f0088a50
DV		 497{
498 int rc = NET_RX_SUCCESS;
499
4f0c40d9 500 if (sk_filter_trim_cap(sk, skb, trim_cap))
f0088a50
DV		 501 goto discard_and_relse;
502
503 skb->dev = NULL;
504
274f482d 505 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 506 atomic_inc(&sk->sk_drops);
507 goto discard_and_relse;
508 }
58a5a7b9
ACM		 509 if (nested)
510 bh_lock_sock_nested(sk);
511 else
512 bh_lock_sock(sk);
a5b5bb9a
IM		 513 if (!sock_owned_by_user(sk)) {
514 /*
515 * trylock + unlock semantics:
516 */
517 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
518
c57943a1 519 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a 520
5facae4f 521 mutex_release(&sk->sk_lock.dep_map, _RET_IP_);
8265792b 522 } else if (sk_add_backlog(sk, skb, READ_ONCE(sk->sk_rcvbuf))) {
8eae939f
ZY		 523 bh_unlock_sock(sk);
524 atomic_inc(&sk->sk_drops);
525 goto discard_and_relse;
526 }
527
f0088a50
DV		 528 bh_unlock_sock(sk);
529out:
c3f24cfb
ED		 530 if (refcounted)
531 sock_put(sk);
f0088a50
DV		 532 return rc;
533discard_and_relse:
534 kfree_skb(skb);
535 goto out;
536}
4f0c40d9 537EXPORT_SYMBOL(__sk_receive_skb);
f0088a50
DV		 538
539struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
540{
b6c6712a 541 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 542
543 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 544 sk_tx_queue_clear(sk);
9b8805a3 545 sk->sk_dst_pending_confirm = 0;
a9b3cd7f 546 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 547 dst_release(dst);
548 return NULL;
549 }
550
551 return dst;
552}
553EXPORT_SYMBOL(__sk_dst_check);
554
555struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
556{
557 struct dst_entry *dst = sk_dst_get(sk);
558
559 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
560 sk_dst_reset(sk);
561 dst_release(dst);
562 return NULL;
563 }
564
565 return dst;
566}
567EXPORT_SYMBOL(sk_dst_check);
568
7594888c 569static int sock_bindtoindex_locked(struct sock *sk, int ifindex)
4878809f
DM		 570{
571 int ret = -ENOPROTOOPT;
572#ifdef CONFIG_NETDEVICES
3b1e0a65 573 struct net *net = sock_net(sk);
4878809f
DM		 574
575 /* Sorry... */
576 ret = -EPERM;
c427bfec 577 if (sk->sk_bound_dev_if && !ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 578 goto out;
579
f5dd3d0c
DH		 580 ret = -EINVAL;
581 if (ifindex < 0)
582 goto out;
583
584 sk->sk_bound_dev_if = ifindex;
585 if (sk->sk_prot->rehash)
586 sk->sk_prot->rehash(sk);
587 sk_dst_reset(sk);
588
589 ret = 0;
590
591out:
592#endif
593
594 return ret;
595}
596
7594888c
CH		 597int sock_bindtoindex(struct sock *sk, int ifindex)
598{
599 int ret;
600
601 lock_sock(sk);
602 ret = sock_bindtoindex_locked(sk, ifindex);
603 release_sock(sk);
604
605 return ret;
606}
607EXPORT_SYMBOL(sock_bindtoindex);
608
f5dd3d0c
DH		 609static int sock_setbindtodevice(struct sock *sk, char __user *optval,
610 int optlen)
611{
612 int ret = -ENOPROTOOPT;
613#ifdef CONFIG_NETDEVICES
614 struct net *net = sock_net(sk);
615 char devname[IFNAMSIZ];
616 int index;
617
4878809f
DM		 618 ret = -EINVAL;
619 if (optlen < 0)
620 goto out;
621
622 /* Bind this socket to a particular device like "eth0",
623 * as specified in the passed interface name. If the
624 * name is "" or the option length is zero the socket
625 * is not bound.
626 */
627 if (optlen > IFNAMSIZ - 1)
628 optlen = IFNAMSIZ - 1;
629 memset(devname, 0, sizeof(devname));
630
631 ret = -EFAULT;
632 if (copy_from_user(devname, optval, optlen))
633 goto out;
634
000ba2e4
DM		 635 index = 0;
636 if (devname[0] != '\0') {
bf8e56bf 637 struct net_device *dev;
4878809f 638
bf8e56bf
ED		 639 rcu_read_lock();
640 dev = dev_get_by_name_rcu(net, devname);
641 if (dev)
642 index = dev->ifindex;
643 rcu_read_unlock();
4878809f
DM		 644 ret = -ENODEV;
645 if (!dev)
646 goto out;
4878809f
DM		 647 }
648
7594888c 649 return sock_bindtoindex(sk, index);
4878809f
DM		 650out:
651#endif
652
653 return ret;
654}
655
c91f6df2
BH		 656static int sock_getbindtodevice(struct sock *sk, char __user *optval,
657 int __user *optlen, int len)
658{
659 int ret = -ENOPROTOOPT;
660#ifdef CONFIG_NETDEVICES
661 struct net *net = sock_net(sk);
c91f6df2 662 char devname[IFNAMSIZ];
c91f6df2
BH		 663
664 if (sk->sk_bound_dev_if == 0) {
665 len = 0;
666 goto zero;
667 }
668
669 ret = -EINVAL;
670 if (len < IFNAMSIZ)
671 goto out;
672
5dbe7c17
NS		 673 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
674 if (ret)
c91f6df2 675 goto out;
c91f6df2
BH		 676
677 len = strlen(devname) + 1;
678
679 ret = -EFAULT;
680 if (copy_to_user(optval, devname, len))
681 goto out;
682
683zero:
684 ret = -EFAULT;
685 if (put_user(len, optlen))
686 goto out;
687
688 ret = 0;
689
690out:
691#endif
692
693 return ret;
694}
695
193d357d
AD		 696static inline void sock_valbool_flag(struct sock *sk, enum sock_flags bit,
697 int valbool)
c0ef877b
PE		 698{
699 if (valbool)
700 sock_set_flag(sk, bit);
701 else
702 sock_reset_flag(sk, bit);
703}
704
f60e5990 705bool sk_mc_loop(struct sock *sk)
706{
707 if (dev_recursion_level())
708 return false;
709 if (!sk)
710 return true;
711 switch (sk->sk_family) {
712 case AF_INET:
713 return inet_sk(sk)->mc_loop;
714#if IS_ENABLED(CONFIG_IPV6)
715 case AF_INET6:
716 return inet6_sk(sk)->mc_loop;
717#endif
718 }
719 WARN_ON(1);
720 return true;
721}
722EXPORT_SYMBOL(sk_mc_loop);
723
b58f0e8f
CH		 724void sock_set_reuseaddr(struct sock *sk)
725{
726 lock_sock(sk);
727 sk->sk_reuse = SK_CAN_REUSE;
728 release_sock(sk);
729}
730EXPORT_SYMBOL(sock_set_reuseaddr);
731
c433594c
CH		 732void sock_no_linger(struct sock *sk)
733{
734 lock_sock(sk);
735 sk->sk_lingertime = 0;
736 sock_set_flag(sk, SOCK_LINGER);
737 release_sock(sk);
738}
739EXPORT_SYMBOL(sock_no_linger);
740
6e434967
CH		 741void sock_set_priority(struct sock *sk, u32 priority)
742{
743 lock_sock(sk);
744 sk->sk_priority = priority;
745 release_sock(sk);
746}
747EXPORT_SYMBOL(sock_set_priority);
748
76ee0785
CH		 749void sock_set_sndtimeo(struct sock *sk, s64 secs)
750{
751 lock_sock(sk);
752 if (secs && secs < MAX_SCHEDULE_TIMEOUT / HZ - 1)
753 sk->sk_sndtimeo = secs * HZ;
754 else
755 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
756 release_sock(sk);
757}
758EXPORT_SYMBOL(sock_set_sndtimeo);
759
783da70e
CH		 760static void __sock_set_timestamps(struct sock *sk, bool val, bool new, bool ns)
761{
762 if (val) {
763 sock_valbool_flag(sk, SOCK_TSTAMP_NEW, new);
764 sock_valbool_flag(sk, SOCK_RCVTSTAMPNS, ns);
765 sock_set_flag(sk, SOCK_RCVTSTAMP);
766 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
767 } else {
768 sock_reset_flag(sk, SOCK_RCVTSTAMP);
769 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
770 sock_reset_flag(sk, SOCK_TSTAMP_NEW);
771 }
772}
773
774void sock_enable_timestamps(struct sock *sk)
775{
776 lock_sock(sk);
777 __sock_set_timestamps(sk, true, false, true);
778 release_sock(sk);
779}
780EXPORT_SYMBOL(sock_enable_timestamps);
781
1da177e4
LT		 782/*
783 * This is meant for all protocols to use and covers goings on
784 * at the socket level. Everything here is generic.
785 */
786
787int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 788 char __user *optval, unsigned int optlen)
1da177e4 789{
80b14dee 790 struct sock_txtime sk_txtime;
2a91525c 791 struct sock *sk = sock->sk;
1da177e4
LT		 792 int val;
793 int valbool;
794 struct linger ling;
795 int ret = 0;
4ec93edb 796
1da177e4
LT		 797 /*
798 * Options without arguments
799 */
800
4878809f 801 if (optname == SO_BINDTODEVICE)
c91f6df2 802 return sock_setbindtodevice(sk, optval, optlen);
4878809f 803
e71a4783
SH		 804 if (optlen < sizeof(int))
805 return -EINVAL;
4ec93edb 806
1da177e4
LT		 807 if (get_user(val, (int __user *)optval))
808 return -EFAULT;
4ec93edb 809
2a91525c 810 valbool = val ? 1 : 0;
1da177e4
LT		 811
812 lock_sock(sk);
813
2a91525c 814 switch (optname) {
e71a4783 815 case SO_DEBUG:
2a91525c 816 if (val && !capable(CAP_NET_ADMIN))
e71a4783 817 ret = -EACCES;
2a91525c 818 else
c0ef877b 819 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 820 break;
821 case SO_REUSEADDR:
cdb8744d 822 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 823 break;
055dc21a
TH		 824 case SO_REUSEPORT:
825 sk->sk_reuseport = valbool;
826 break;
e71a4783 827 case SO_TYPE:
49c794e9 828 case SO_PROTOCOL:
0d6038ee 829 case SO_DOMAIN:
e71a4783
SH		 830 case SO_ERROR:
831 ret = -ENOPROTOOPT;
832 break;
833 case SO_DONTROUTE:
c0ef877b 834 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
0fbe82e6 835 sk_dst_reset(sk);
e71a4783
SH		 836 break;
837 case SO_BROADCAST:
838 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
839 break;
840 case SO_SNDBUF:
841 /* Don't error on this BSD doesn't and if you think
82981930
ED		 842 * about it this is right. Otherwise apps have to
843 * play 'guess the biggest size' games. RCVBUF/SNDBUF
844 * are treated in BSD as hints
845 */
846 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 847set_sndbuf:
4057765f
GN		 848 /* Ensure val * 2 fits into an int, to prevent max_t()
849 * from treating it as a negative value.
850 */
851 val = min_t(int, val, INT_MAX / 2);
e71a4783 852 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
e292f05e
ED		 853 WRITE_ONCE(sk->sk_sndbuf,
854 max_t(int, val * 2, SOCK_MIN_SNDBUF));
82981930 855 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 856 sk->sk_write_space(sk);
857 break;
1da177e4 858
e71a4783
SH		 859 case SO_SNDBUFFORCE:
860 if (!capable(CAP_NET_ADMIN)) {
861 ret = -EPERM;
862 break;
863 }
4057765f
GN		 864
865 /* No negative values (to prevent underflow, as val will be
866 * multiplied by 2).
867 */
868 if (val < 0)
869 val = 0;
e71a4783 870 goto set_sndbuf;
b0573dea 871
e71a4783
SH		 872 case SO_RCVBUF:
873 /* Don't error on this BSD doesn't and if you think
82981930
ED		 874 * about it this is right. Otherwise apps have to
875 * play 'guess the biggest size' games. RCVBUF/SNDBUF
876 * are treated in BSD as hints
877 */
878 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 879set_rcvbuf:
4057765f
GN		 880 /* Ensure val * 2 fits into an int, to prevent max_t()
881 * from treating it as a negative value.
882 */
883 val = min_t(int, val, INT_MAX / 2);
e71a4783
SH		 884 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
885 /*
886 * We double it on the way in to account for
887 * "struct sk_buff" etc. overhead. Applications
888 * assume that the SO_RCVBUF setting they make will
889 * allow that much actual data to be received on that
890 * socket.
891 *
892 * Applications are unaware that "struct sk_buff" and
893 * other overheads allocate from the receive buffer
894 * during socket buffer allocation.
895 *
896 * And after considering the possible alternatives,
897 * returning the value we actually used in getsockopt
898 * is the most desirable behavior.
899 */
ebb3b78d
ED		 900 WRITE_ONCE(sk->sk_rcvbuf,
901 max_t(int, val * 2, SOCK_MIN_RCVBUF));
e71a4783
SH		 902 break;
903
904 case SO_RCVBUFFORCE:
905 if (!capable(CAP_NET_ADMIN)) {
906 ret = -EPERM;
1da177e4 907 break;
e71a4783 908 }
4057765f
GN		 909
910 /* No negative values (to prevent underflow, as val will be
911 * multiplied by 2).
912 */
913 if (val < 0)
914 val = 0;
e71a4783 915 goto set_rcvbuf;
1da177e4 916
e71a4783 917 case SO_KEEPALIVE:
4b9d07a4
UB		 918 if (sk->sk_prot->keepalive)
919 sk->sk_prot->keepalive(sk, valbool);
e71a4783
SH		 920 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
921 break;
922
923 case SO_OOBINLINE:
924 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
925 break;
926
927 case SO_NO_CHECK:
28448b80 928 sk->sk_no_check_tx = valbool;
e71a4783
SH		 929 break;
930
931 case SO_PRIORITY:
5e1fccc0
EB		 932 if ((val >= 0 && val <= 6) ||
933 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 934 sk->sk_priority = val;
935 else
936 ret = -EPERM;
937 break;
938
939 case SO_LINGER:
940 if (optlen < sizeof(ling)) {
941 ret = -EINVAL; /* 1003.1g */
1da177e4 942 break;
e71a4783 943 }
2a91525c 944 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 945 ret = -EFAULT;
1da177e4 946 break;
e71a4783
SH		 947 }
948 if (!ling.l_onoff)
949 sock_reset_flag(sk, SOCK_LINGER);
950 else {
1da177e4 951#if (BITS_PER_LONG == 32)
e71a4783
SH		 952 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
953 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 954 else
e71a4783
SH		 955#endif
956 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
957 sock_set_flag(sk, SOCK_LINGER);
958 }
959 break;
960
961 case SO_BSDCOMPAT:
962 sock_warn_obsolete_bsdism("setsockopt");
963 break;
964
965 case SO_PASSCRED:
966 if (valbool)
967 set_bit(SOCK_PASSCRED, &sock->flags);
968 else
969 clear_bit(SOCK_PASSCRED, &sock->flags);
970 break;
971
7f1bc6e9 972 case SO_TIMESTAMP_OLD:
783da70e
CH		 973 __sock_set_timestamps(sk, valbool, false, false);
974 break;
887feae3 975 case SO_TIMESTAMP_NEW:
783da70e
CH		 976 __sock_set_timestamps(sk, valbool, true, false);
977 break;
7f1bc6e9 978 case SO_TIMESTAMPNS_OLD:
783da70e
CH		 979 __sock_set_timestamps(sk, valbool, false, true);
980 break;
887feae3 981 case SO_TIMESTAMPNS_NEW:
783da70e 982 __sock_set_timestamps(sk, valbool, true, true);
e71a4783 983 break;
9718475e
DD		 984 case SO_TIMESTAMPING_NEW:
985 sock_set_flag(sk, SOCK_TSTAMP_NEW);
ff7653f9 986 /* fall through */
7f1bc6e9 987 case SO_TIMESTAMPING_OLD:
20d49473 988 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 989 ret = -EINVAL;
20d49473
PO		 990 break;
991 }
b245be1f 992
09c2d251 993 if (val & SOF_TIMESTAMPING_OPT_ID &&
4ed2d765 994 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
ac5cc977
WC		 995 if (sk->sk_protocol == IPPROTO_TCP &&
996 sk->sk_type == SOCK_STREAM) {
6db8b963
SHY		 997 if ((1 << sk->sk_state) &
998 (TCPF_CLOSE | TCPF_LISTEN)) {
4ed2d765
WB		 999 ret = -EINVAL;
1000 break;
1001 }
1002 sk->sk_tskey = tcp_sk(sk)->snd_una;
1003 } else {
1004 sk->sk_tskey = 0;
1005 }
1006 }
1c885808
FY		 1007
1008 if (val & SOF_TIMESTAMPING_OPT_STATS &&
1009 !(val & SOF_TIMESTAMPING_OPT_TSONLY)) {
1010 ret = -EINVAL;
1011 break;
1012 }
1013
b9f40e21 1014 sk->sk_tsflags = val;
20d49473
PO		 1015 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
1016 sock_enable_timestamp(sk,
1017 SOCK_TIMESTAMPING_RX_SOFTWARE);
9718475e
DD		 1018 else {
1019 if (optname == SO_TIMESTAMPING_NEW)
1020 sock_reset_flag(sk, SOCK_TSTAMP_NEW);
1021
20d49473 1022 sock_disable_timestamp(sk,
08e29af3 1023 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
9718475e 1024 }
20d49473
PO		 1025 break;
1026
e71a4783
SH		 1027 case SO_RCVLOWAT:
1028 if (val < 0)
1029 val = INT_MAX;
d1361840
ED		 1030 if (sock->ops->set_rcvlowat)
1031 ret = sock->ops->set_rcvlowat(sk, val);
1032 else
eac66402 1033 WRITE_ONCE(sk->sk_rcvlowat, val ? : 1);
e71a4783
SH		 1034 break;
1035
45bdc661 1036 case SO_RCVTIMEO_OLD:
a9beb86a
DD		 1037 case SO_RCVTIMEO_NEW:
1038 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen, optname == SO_RCVTIMEO_OLD);
e71a4783
SH		 1039 break;
1040
45bdc661 1041 case SO_SNDTIMEO_OLD:
a9beb86a
DD		 1042 case SO_SNDTIMEO_NEW:
1043 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen, optname == SO_SNDTIMEO_OLD);
e71a4783 1044 break;
1da177e4 1045
e71a4783
SH		 1046 case SO_ATTACH_FILTER:
1047 ret = -EINVAL;
1048 if (optlen == sizeof(struct sock_fprog)) {
1049 struct sock_fprog fprog;
1da177e4 1050
e71a4783
SH		 1051 ret = -EFAULT;
1052 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 1053 break;
e71a4783
SH		 1054
1055 ret = sk_attach_filter(&fprog, sk);
1056 }
1057 break;
1058
89aa0758
AS		 1059 case SO_ATTACH_BPF:
1060 ret = -EINVAL;
1061 if (optlen == sizeof(u32)) {
1062 u32 ufd;
1063
1064 ret = -EFAULT;
1065 if (copy_from_user(&ufd, optval, sizeof(ufd)))
1066 break;
1067
1068 ret = sk_attach_bpf(ufd, sk);
1069 }
1070 break;
1071
538950a1
CG		 1072 case SO_ATTACH_REUSEPORT_CBPF:
1073 ret = -EINVAL;
1074 if (optlen == sizeof(struct sock_fprog)) {
1075 struct sock_fprog fprog;
1076
1077 ret = -EFAULT;
1078 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1079 break;
1080
1081 ret = sk_reuseport_attach_filter(&fprog, sk);
1082 }
1083 break;
1084
1085 case SO_ATTACH_REUSEPORT_EBPF:
1086 ret = -EINVAL;
1087 if (optlen == sizeof(u32)) {
1088 u32 ufd;
1089
1090 ret = -EFAULT;
1091 if (copy_from_user(&ufd, optval, sizeof(ufd)))
1092 break;
1093
1094 ret = sk_reuseport_attach_bpf(ufd, sk);
1095 }
1096 break;
1097
99f3a064
MKL		 1098 case SO_DETACH_REUSEPORT_BPF:
1099 ret = reuseport_detach_prog(sk);
1100 break;
1101
e71a4783 1102 case SO_DETACH_FILTER:
55b33325 1103 ret = sk_detach_filter(sk);
e71a4783 1104 break;
1da177e4 1105
d59577b6
VB		 1106 case SO_LOCK_FILTER:
1107 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
1108 ret = -EPERM;
1109 else
1110 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
1111 break;
1112
e71a4783
SH		 1113 case SO_PASSSEC:
1114 if (valbool)
1115 set_bit(SOCK_PASSSEC, &sock->flags);
1116 else
1117 clear_bit(SOCK_PASSSEC, &sock->flags);
1118 break;
4a19ec58 1119 case SO_MARK:
50254256 1120 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
4a19ec58 1121 ret = -EPERM;
50254256 1122 } else if (val != sk->sk_mark) {
4a19ec58 1123 sk->sk_mark = val;
50254256
DB		 1124 sk_dst_reset(sk);
1125 }
4a19ec58 1126 break;
877ce7c1 1127
3b885787 1128 case SO_RXQ_OVFL:
8083f0fc 1129 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 1130 break;
6e3e939f
JB		 1131
1132 case SO_WIFI_STATUS:
1133 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
1134 break;
1135
ef64a54f
PE		 1136 case SO_PEEK_OFF:
1137 if (sock->ops->set_peek_off)
12663bfc 1138 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 1139 else
1140 ret = -EOPNOTSUPP;
1141 break;
3bdc0eba
BG		 1142
1143 case SO_NOFCS:
1144 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
1145 break;
1146
7d4c04fc
KJ		 1147 case SO_SELECT_ERR_QUEUE:
1148 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
1149 break;
1150
e0d1095a 1151#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1152 case SO_BUSY_POLL:
dafcc438
ET		 1153 /* allow unprivileged users to decrease the value */
1154 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
1155 ret = -EPERM;
1156 else {
1157 if (val < 0)
1158 ret = -EINVAL;
1159 else
1160 sk->sk_ll_usec = val;
1161 }
1162 break;
1163#endif
62748f32
ED		 1164
1165 case SO_MAX_PACING_RATE:
6bdef102
ED		 1166 {
1167 unsigned long ulval = (val == ~0U) ? ~0UL : val;
1168
1169 if (sizeof(ulval) != sizeof(val) &&
1170 optlen >= sizeof(ulval) &&
1171 get_user(ulval, (unsigned long __user *)optval)) {
1172 ret = -EFAULT;
1173 break;
1174 }
1175 if (ulval != ~0UL)
218af599
ED		 1176 cmpxchg(&sk->sk_pacing_status,
1177 SK_PACING_NONE,
1178 SK_PACING_NEEDED);
6bdef102
ED		 1179 sk->sk_max_pacing_rate = ulval;
1180 sk->sk_pacing_rate = min(sk->sk_pacing_rate, ulval);
62748f32 1181 break;
6bdef102 1182 }
70da268b 1183 case SO_INCOMING_CPU:
7170a977 1184 WRITE_ONCE(sk->sk_incoming_cpu, val);
70da268b
ED		 1185 break;
1186
a87cb3e4
TH		 1187 case SO_CNX_ADVICE:
1188 if (val == 1)
1189 dst_negative_advice(sk);
1190 break;
76851d12
WB		 1191
1192 case SO_ZEROCOPY:
28190752 1193 if (sk->sk_family == PF_INET || sk->sk_family == PF_INET6) {
b5947e5d
WB		 1194 if (!((sk->sk_type == SOCK_STREAM &&
1195 sk->sk_protocol == IPPROTO_TCP) ||
1196 (sk->sk_type == SOCK_DGRAM &&
1197 sk->sk_protocol == IPPROTO_UDP)))
28190752 1198 ret = -ENOTSUPP;
28190752 1199 } else if (sk->sk_family != PF_RDS) {
76851d12 1200 ret = -ENOTSUPP;
28190752
SV		 1201 }
1202 if (!ret) {
1203 if (val < 0 || val > 1)
1204 ret = -EINVAL;
1205 else
1206 sock_valbool_flag(sk, SOCK_ZEROCOPY, valbool);
28190752 1207 }
334e6413
JSP		 1208 break;
1209
80b14dee 1210 case SO_TXTIME:
790709f2 1211 if (optlen != sizeof(struct sock_txtime)) {
80b14dee 1212 ret = -EINVAL;
790709f2 1213 break;
80b14dee
RC		 1214 } else if (copy_from_user(&sk_txtime, optval,
1215 sizeof(struct sock_txtime))) {
1216 ret = -EFAULT;
790709f2 1217 break;
80b14dee
RC		 1218 } else if (sk_txtime.flags & ~SOF_TXTIME_FLAGS_MASK) {
1219 ret = -EINVAL;
790709f2
ED		 1220 break;
1221 }
1222 /* CLOCK_MONOTONIC is only used by sch_fq, and this packet
1223 * scheduler has enough safe guards.
1224 */
1225 if (sk_txtime.clockid != CLOCK_MONOTONIC &&
1226 !ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN)) {
1227 ret = -EPERM;
1228 break;
80b14dee 1229 }
790709f2
ED		 1230 sock_valbool_flag(sk, SOCK_TXTIME, true);
1231 sk->sk_clockid = sk_txtime.clockid;
1232 sk->sk_txtime_deadline_mode =
1233 !!(sk_txtime.flags & SOF_TXTIME_DEADLINE_MODE);
1234 sk->sk_txtime_report_errors =
1235 !!(sk_txtime.flags & SOF_TXTIME_REPORT_ERRORS);
80b14dee
RC		 1236 break;
1237
f5dd3d0c 1238 case SO_BINDTOIFINDEX:
7594888c 1239 ret = sock_bindtoindex_locked(sk, val);
f5dd3d0c
DH		 1240 break;
1241
e71a4783
SH		 1242 default:
1243 ret = -ENOPROTOOPT;
1244 break;
4ec93edb 1245 }
1da177e4
LT		 1246 release_sock(sk);
1247 return ret;
1248}
2a91525c 1249EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 1250
1251
8f09898b 1252static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1253 struct ucred *ucred)
3f551f94
EB		 1254{
1255 ucred->pid = pid_vnr(pid);
1256 ucred->uid = ucred->gid = -1;
1257 if (cred) {
1258 struct user_namespace *current_ns = current_user_ns();
1259
b2e4f544
EB		 1260 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1261 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 1262 }
1263}
1264
28b5ba2a
DH		 1265static int groups_to_user(gid_t __user *dst, const struct group_info *src)
1266{
1267 struct user_namespace *user_ns = current_user_ns();
1268 int i;
1269
1270 for (i = 0; i < src->ngroups; i++)
1271 if (put_user(from_kgid_munged(user_ns, src->gid[i]), dst + i))
1272 return -EFAULT;
1273
1274 return 0;
1275}
1276
1da177e4
LT		 1277int sock_getsockopt(struct socket *sock, int level, int optname,
1278 char __user *optval, int __user *optlen)
1279{
1280 struct sock *sk = sock->sk;
4ec93edb 1281
e71a4783 1282 union {
4ec93edb 1283 int val;
5daab9db 1284 u64 val64;
677f136c 1285 unsigned long ulval;
4ec93edb 1286 struct linger ling;
fe0c72f3
AB		 1287 struct old_timeval32 tm32;
1288 struct __kernel_old_timeval tm;
a9beb86a 1289 struct __kernel_sock_timeval stm;
80b14dee 1290 struct sock_txtime txtime;
1da177e4 1291 } v;
4ec93edb 1292
4d0392be 1293 int lv = sizeof(int);
1da177e4 1294 int len;
4ec93edb 1295
e71a4783 1296 if (get_user(len, optlen))
4ec93edb 1297 return -EFAULT;
e71a4783 1298 if (len < 0)
1da177e4 1299 return -EINVAL;
4ec93edb 1300
50fee1de 1301 memset(&v, 0, sizeof(v));
df0bca04 1302
2a91525c 1303 switch (optname) {
e71a4783
SH		 1304 case SO_DEBUG:
1305 v.val = sock_flag(sk, SOCK_DBG);
1306 break;
1307
1308 case SO_DONTROUTE:
1309 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1310 break;
1311
1312 case SO_BROADCAST:
1b23a5df 1313 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1314 break;
1315
1316 case SO_SNDBUF:
1317 v.val = sk->sk_sndbuf;
1318 break;
1319
1320 case SO_RCVBUF:
1321 v.val = sk->sk_rcvbuf;
1322 break;
1323
1324 case SO_REUSEADDR:
1325 v.val = sk->sk_reuse;
1326 break;
1327
055dc21a
TH		 1328 case SO_REUSEPORT:
1329 v.val = sk->sk_reuseport;
1330 break;
1331
e71a4783 1332 case SO_KEEPALIVE:
1b23a5df 1333 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1334 break;
1335
1336 case SO_TYPE:
1337 v.val = sk->sk_type;
1338 break;
1339
49c794e9
JE		 1340 case SO_PROTOCOL:
1341 v.val = sk->sk_protocol;
1342 break;
1343
0d6038ee
JE		 1344 case SO_DOMAIN:
1345 v.val = sk->sk_family;
1346 break;
1347
e71a4783
SH		 1348 case SO_ERROR:
1349 v.val = -sock_error(sk);
2a91525c 1350 if (v.val == 0)
e71a4783
SH		 1351 v.val = xchg(&sk->sk_err_soft, 0);
1352 break;
1353
1354 case SO_OOBINLINE:
1b23a5df 1355 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1356 break;
1357
1358 case SO_NO_CHECK:
28448b80 1359 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1360 break;
1361
1362 case SO_PRIORITY:
1363 v.val = sk->sk_priority;
1364 break;
1365
1366 case SO_LINGER:
1367 lv = sizeof(v.ling);
1b23a5df 1368 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1369 v.ling.l_linger = sk->sk_lingertime / HZ;
1370 break;
1371
1372 case SO_BSDCOMPAT:
1373 sock_warn_obsolete_bsdism("getsockopt");
1374 break;
1375
7f1bc6e9 1376 case SO_TIMESTAMP_OLD:
92f37fd2 1377 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
887feae3 1378 !sock_flag(sk, SOCK_TSTAMP_NEW) &&
92f37fd2
ED		 1379 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1380 break;
1381
7f1bc6e9 1382 case SO_TIMESTAMPNS_OLD:
887feae3
DD		 1383 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS) && !sock_flag(sk, SOCK_TSTAMP_NEW);
1384 break;
1385
1386 case SO_TIMESTAMP_NEW:
1387 v.val = sock_flag(sk, SOCK_RCVTSTAMP) && sock_flag(sk, SOCK_TSTAMP_NEW);
1388 break;
1389
1390 case SO_TIMESTAMPNS_NEW:
1391 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS) && sock_flag(sk, SOCK_TSTAMP_NEW);
e71a4783
SH		 1392 break;
1393
7f1bc6e9 1394 case SO_TIMESTAMPING_OLD:
b9f40e21 1395 v.val = sk->sk_tsflags;
20d49473
PO		 1396 break;
1397
a9beb86a
DD		 1398 case SO_RCVTIMEO_OLD:
1399 case SO_RCVTIMEO_NEW:
1400 lv = sock_get_timeout(sk->sk_rcvtimeo, &v, SO_RCVTIMEO_OLD == optname);
e71a4783
SH		 1401 break;
1402
a9beb86a
DD		 1403 case SO_SNDTIMEO_OLD:
1404 case SO_SNDTIMEO_NEW:
1405 lv = sock_get_timeout(sk->sk_sndtimeo, &v, SO_SNDTIMEO_OLD == optname);
e71a4783 1406 break;
1da177e4 1407
e71a4783
SH		 1408 case SO_RCVLOWAT:
1409 v.val = sk->sk_rcvlowat;
1410 break;
1da177e4 1411
e71a4783 1412 case SO_SNDLOWAT:
2a91525c 1413 v.val = 1;
e71a4783 1414 break;
1da177e4 1415
e71a4783 1416 case SO_PASSCRED:
82981930 1417 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1418 break;
1da177e4 1419
e71a4783 1420 case SO_PEERCRED:
109f6e39
EB		 1421 {
1422 struct ucred peercred;
1423 if (len > sizeof(peercred))
1424 len = sizeof(peercred);
1425 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1426 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1427 return -EFAULT;
1428 goto lenout;
109f6e39 1429 }
1da177e4 1430
28b5ba2a
DH		 1431 case SO_PEERGROUPS:
1432 {
1433 int ret, n;
1434
1435 if (!sk->sk_peer_cred)
1436 return -ENODATA;
1437
1438 n = sk->sk_peer_cred->group_info->ngroups;
1439 if (len < n * sizeof(gid_t)) {
1440 len = n * sizeof(gid_t);
1441 return put_user(len, optlen) ? -EFAULT : -ERANGE;
1442 }
1443 len = n * sizeof(gid_t);
1444
1445 ret = groups_to_user((gid_t __user *)optval,
1446 sk->sk_peer_cred->group_info);
1447 if (ret)
1448 return ret;
1449 goto lenout;
1450 }
1451
e71a4783
SH		 1452 case SO_PEERNAME:
1453 {
1454 char address[128];
1455
9b2c45d4
DV		 1456 lv = sock->ops->getname(sock, (struct sockaddr *)address, 2);
1457 if (lv < 0)
e71a4783
SH		 1458 return -ENOTCONN;
1459 if (lv < len)
1460 return -EINVAL;
1461 if (copy_to_user(optval, address, len))
1462 return -EFAULT;
1463 goto lenout;
1464 }
1da177e4 1465
e71a4783
SH		 1466 /* Dubious BSD thing... Probably nobody even uses it, but
1467 * the UNIX standard wants it for whatever reason... -DaveM
1468 */
1469 case SO_ACCEPTCONN:
1470 v.val = sk->sk_state == TCP_LISTEN;
1471 break;
1da177e4 1472
e71a4783 1473 case SO_PASSSEC:
82981930 1474 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1475 break;
877ce7c1 1476
e71a4783
SH		 1477 case SO_PEERSEC:
1478 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1479
4a19ec58
LAT		 1480 case SO_MARK:
1481 v.val = sk->sk_mark;
1482 break;
1483
3b885787 1484 case SO_RXQ_OVFL:
1b23a5df 1485 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1486 break;
1487
6e3e939f 1488 case SO_WIFI_STATUS:
1b23a5df 1489 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1490 break;
1491
ef64a54f
PE		 1492 case SO_PEEK_OFF:
1493 if (!sock->ops->set_peek_off)
1494 return -EOPNOTSUPP;
1495
1496 v.val = sk->sk_peek_off;
1497 break;
bc2f7996 1498 case SO_NOFCS:
1b23a5df 1499 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1500 break;
c91f6df2 1501
f7b86bfe 1502 case SO_BINDTODEVICE:
c91f6df2
BH		 1503 return sock_getbindtodevice(sk, optval, optlen, len);
1504
a8fc9277
PE		 1505 case SO_GET_FILTER:
1506 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1507 if (len < 0)
1508 return len;
1509
1510 goto lenout;
c91f6df2 1511
d59577b6
VB		 1512 case SO_LOCK_FILTER:
1513 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1514 break;
1515
ea02f941
MS		 1516 case SO_BPF_EXTENSIONS:
1517 v.val = bpf_tell_extensions();
1518 break;
1519
7d4c04fc
KJ		 1520 case SO_SELECT_ERR_QUEUE:
1521 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1522 break;
1523
e0d1095a 1524#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1525 case SO_BUSY_POLL:
dafcc438
ET		 1526 v.val = sk->sk_ll_usec;
1527 break;
1528#endif
1529
62748f32 1530 case SO_MAX_PACING_RATE:
677f136c
ED		 1531 if (sizeof(v.ulval) != sizeof(v.val) && len >= sizeof(v.ulval)) {
1532 lv = sizeof(v.ulval);
1533 v.ulval = sk->sk_max_pacing_rate;
1534 } else {
1535 /* 32bit version */
1536 v.val = min_t(unsigned long, sk->sk_max_pacing_rate, ~0U);
1537 }
62748f32
ED		 1538 break;
1539
2c8c56e1 1540 case SO_INCOMING_CPU:
7170a977 1541 v.val = READ_ONCE(sk->sk_incoming_cpu);
2c8c56e1
ED		 1542 break;
1543
a2d133b1
JH		 1544 case SO_MEMINFO:
1545 {
1546 u32 meminfo[SK_MEMINFO_VARS];
1547
a2d133b1
JH		 1548 sk_get_meminfo(sk, meminfo);
1549
1550 len = min_t(unsigned int, len, sizeof(meminfo));
1551 if (copy_to_user(optval, &meminfo, len))
1552 return -EFAULT;
1553
1554 goto lenout;
1555 }
6d433902
SS		 1556
1557#ifdef CONFIG_NET_RX_BUSY_POLL
1558 case SO_INCOMING_NAPI_ID:
1559 v.val = READ_ONCE(sk->sk_napi_id);
1560
1561 /* aggregate non-NAPI IDs down to 0 */
1562 if (v.val < MIN_NAPI_ID)
1563 v.val = 0;
1564
1565 break;
1566#endif
1567
5daab9db
CF		 1568 case SO_COOKIE:
1569 lv = sizeof(u64);
1570 if (len < lv)
1571 return -EINVAL;
1572 v.val64 = sock_gen_cookie(sk);
1573 break;
1574
76851d12
WB		 1575 case SO_ZEROCOPY:
1576 v.val = sock_flag(sk, SOCK_ZEROCOPY);
1577 break;
1578
80b14dee
RC		 1579 case SO_TXTIME:
1580 lv = sizeof(v.txtime);
1581 v.txtime.clockid = sk->sk_clockid;
1582 v.txtime.flags |= sk->sk_txtime_deadline_mode ?
1583 SOF_TXTIME_DEADLINE_MODE : 0;
4b15c707
JSP		 1584 v.txtime.flags |= sk->sk_txtime_report_errors ?
1585 SOF_TXTIME_REPORT_ERRORS : 0;
80b14dee
RC		 1586 break;
1587
f5dd3d0c
DH		 1588 case SO_BINDTOIFINDEX:
1589 v.val = sk->sk_bound_dev_if;
1590 break;
1591
e71a4783 1592 default:
443b5991
YH		 1593 /* We implement the SO_SNDLOWAT etc to not be settable
1594 * (1003.1g 7).
1595 */
e71a4783 1596 return -ENOPROTOOPT;
1da177e4 1597 }
e71a4783 1598
1da177e4
LT		 1599 if (len > lv)
1600 len = lv;
1601 if (copy_to_user(optval, &v, len))
1602 return -EFAULT;
1603lenout:
4ec93edb
YH		 1604 if (put_user(len, optlen))
1605 return -EFAULT;
1606 return 0;
1da177e4
LT		 1607}
1608
a5b5bb9a
IM		 1609/*
1610 * Initialize an sk_lock.
1611 *
1612 * (We also register the sk_lock with the lock validator.)
1613 */
b6f99a21 1614static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1615{
cdfbabfb
DH		 1616 if (sk->sk_kern_sock)
1617 sock_lock_init_class_and_name(
1618 sk,
1619 af_family_kern_slock_key_strings[sk->sk_family],
1620 af_family_kern_slock_keys + sk->sk_family,
1621 af_family_kern_key_strings[sk->sk_family],
1622 af_family_kern_keys + sk->sk_family);
1623 else
1624 sock_lock_init_class_and_name(
1625 sk,
ed07536e
PZ		 1626 af_family_slock_key_strings[sk->sk_family],
1627 af_family_slock_keys + sk->sk_family,
1628 af_family_key_strings[sk->sk_family],
1629 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1630}
1631
4dc6dc71
ED		 1632/*
1633 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1634 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1635 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1636 */
f1a6c4da
PE		 1637static void sock_copy(struct sock *nsk, const struct sock *osk)
1638{
b8e202d1 1639 const struct proto *prot = READ_ONCE(osk->sk_prot);
f1a6c4da
PE		 1640#ifdef CONFIG_SECURITY_NETWORK
1641 void *sptr = nsk->sk_security;
1642#endif
68835aba
ED		 1643 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1644
1645 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
b8e202d1 1646 prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
68835aba 1647
f1a6c4da
PE		 1648#ifdef CONFIG_SECURITY_NETWORK
1649 nsk->sk_security = sptr;
1650 security_sk_clone(osk, nsk);
1651#endif
1652}
1653
2e4afe7b
PE		 1654static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1655 int family)
c308c1b2
PE		 1656{
1657 struct sock *sk;
1658 struct kmem_cache *slab;
1659
1660 slab = prot->slab;
e912b114
ED		 1661 if (slab != NULL) {
1662 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1663 if (!sk)
1664 return sk;
6471384a 1665 if (want_init_on_alloc(priority))
ba2489b0 1666 sk_prot_clear_nulls(sk, prot->obj_size);
fcbdf09d 1667 } else
c308c1b2
PE		 1668 sk = kmalloc(prot->obj_size, priority);
1669
2e4afe7b
PE		 1670 if (sk != NULL) {
1671 if (security_sk_alloc(sk, family, priority))
1672 goto out_free;
1673
1674 if (!try_module_get(prot->owner))
1675 goto out_free_sec;
e022f0b4 1676 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1677 }
1678
c308c1b2 1679 return sk;
2e4afe7b
PE		 1680
1681out_free_sec:
1682 security_sk_free(sk);
1683out_free:
1684 if (slab != NULL)
1685 kmem_cache_free(slab, sk);
1686 else
1687 kfree(sk);
1688 return NULL;
c308c1b2
PE		 1689}
1690
1691static void sk_prot_free(struct proto *prot, struct sock *sk)
1692{
1693 struct kmem_cache *slab;
2e4afe7b 1694 struct module *owner;
c308c1b2 1695
2e4afe7b 1696 owner = prot->owner;
c308c1b2 1697 slab = prot->slab;
2e4afe7b 1698
bd1060a1 1699 cgroup_sk_free(&sk->sk_cgrp_data);
2d758073 1700 mem_cgroup_sk_free(sk);
2e4afe7b 1701 security_sk_free(sk);
c308c1b2
PE		 1702 if (slab != NULL)
1703 kmem_cache_free(slab, sk);
1704 else
1705 kfree(sk);
2e4afe7b 1706 module_put(owner);
c308c1b2
PE		 1707}
1708
1da177e4
LT		 1709/**
1710 * sk_alloc - All socket objects are allocated here
c4ea43c5 1711 * @net: the applicable net namespace
4dc3b16b
PP		 1712 * @family: protocol family
1713 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1714 * @prot: struct proto associated with this new sock instance
11aa9c28 1715 * @kern: is this to be a kernel socket?
1da177e4 1716 */
1b8d7ae4 1717struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1718 struct proto *prot, int kern)
1da177e4 1719{
c308c1b2 1720 struct sock *sk;
1da177e4 1721
154adbc8 1722 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1723 if (sk) {
154adbc8
PE		 1724 sk->sk_family = family;
1725 /*
1726 * See comment in struct sock definition to understand
1727 * why we need sk_prot_creator -acme
1728 */
1729 sk->sk_prot = sk->sk_prot_creator = prot;
cdfbabfb 1730 sk->sk_kern_sock = kern;
154adbc8 1731 sock_lock_init(sk);
26abe143 1732 sk->sk_net_refcnt = kern ? 0 : 1;
648845ab 1733 if (likely(sk->sk_net_refcnt)) {
26abe143 1734 get_net(net);
648845ab
TZ		 1735 sock_inuse_add(net, 1);
1736 }
1737
26abe143 1738 sock_net_set(sk, net);
14afee4b 1739 refcount_set(&sk->sk_wmem_alloc, 1);
f8451725 1740
2d758073 1741 mem_cgroup_sk_alloc(sk);
d979a39d 1742 cgroup_sk_alloc(&sk->sk_cgrp_data);
2a56a1fe
TH		 1743 sock_update_classid(&sk->sk_cgrp_data);
1744 sock_update_netprioidx(&sk->sk_cgrp_data);
1da177e4 1745 }
a79af59e 1746
2e4afe7b 1747 return sk;
1da177e4 1748}
2a91525c 1749EXPORT_SYMBOL(sk_alloc);
1da177e4 1750
a4298e45
ED		 1751/* Sockets having SOCK_RCU_FREE will call this function after one RCU
1752 * grace period. This is the case for UDP sockets and TCP listeners.
1753 */
1754static void __sk_destruct(struct rcu_head *head)
1da177e4 1755{
a4298e45 1756 struct sock *sk = container_of(head, struct sock, sk_rcu);
1da177e4 1757 struct sk_filter *filter;
1da177e4
LT		 1758
1759 if (sk->sk_destruct)
1760 sk->sk_destruct(sk);
1761
a898def2 1762 filter = rcu_dereference_check(sk->sk_filter,
14afee4b 1763 refcount_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1764 if (filter) {
309dd5fc 1765 sk_filter_uncharge(sk, filter);
a9b3cd7f 1766 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1767 }
1768
08e29af3 1769 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4 1770
6ac99e8f
MKL		 1771#ifdef CONFIG_BPF_SYSCALL
1772 bpf_sk_storage_free(sk);
1773#endif
1774
1da177e4 1775 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1776 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1777 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1778
22a0e18e
ED		 1779 if (sk->sk_frag.page) {
1780 put_page(sk->sk_frag.page);
1781 sk->sk_frag.page = NULL;
1782 }
1783
109f6e39
EB		 1784 if (sk->sk_peer_cred)
1785 put_cred(sk->sk_peer_cred);
1786 put_pid(sk->sk_peer_pid);
26abe143
EB		 1787 if (likely(sk->sk_net_refcnt))
1788 put_net(sock_net(sk));
c308c1b2 1789 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1790}
2b85a34e 1791
a4298e45
ED		 1792void sk_destruct(struct sock *sk)
1793{
8c7138b3
MKL		 1794 bool use_call_rcu = sock_flag(sk, SOCK_RCU_FREE);
1795
1796 if (rcu_access_pointer(sk->sk_reuseport_cb)) {
1797 reuseport_detach_sock(sk);
1798 use_call_rcu = true;
1799 }
1800
1801 if (use_call_rcu)
a4298e45
ED		 1802 call_rcu(&sk->sk_rcu, __sk_destruct);
1803 else
1804 __sk_destruct(&sk->sk_rcu);
1805}
1806
eb4cb008
CG		 1807static void __sk_free(struct sock *sk)
1808{
648845ab
TZ		 1809 if (likely(sk->sk_net_refcnt))
1810 sock_inuse_add(sock_net(sk), -1);
1811
9709020c 1812 if (unlikely(sk->sk_net_refcnt && sock_diag_has_destroy_listeners(sk)))
eb4cb008
CG		 1813 sock_diag_broadcast_destroy(sk);
1814 else
1815 sk_destruct(sk);
1816}
1817
2b85a34e
ED		 1818void sk_free(struct sock *sk)
1819{
1820 /*
25985edc 1821 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1822 * some packets are still in some tx queue.
1823 * If not null, sock_wfree() will call __sk_free(sk) later
1824 */
14afee4b 1825 if (refcount_dec_and_test(&sk->sk_wmem_alloc))
2b85a34e
ED		 1826 __sk_free(sk);
1827}
2a91525c 1828EXPORT_SYMBOL(sk_free);
1da177e4 1829
581319c5
PA		 1830static void sk_init_common(struct sock *sk)
1831{
1832 skb_queue_head_init(&sk->sk_receive_queue);
1833 skb_queue_head_init(&sk->sk_write_queue);
1834 skb_queue_head_init(&sk->sk_error_queue);
1835
1836 rwlock_init(&sk->sk_callback_lock);
1837 lockdep_set_class_and_name(&sk->sk_receive_queue.lock,
1838 af_rlock_keys + sk->sk_family,
1839 af_family_rlock_key_strings[sk->sk_family]);
1840 lockdep_set_class_and_name(&sk->sk_write_queue.lock,
1841 af_wlock_keys + sk->sk_family,
1842 af_family_wlock_key_strings[sk->sk_family]);
1843 lockdep_set_class_and_name(&sk->sk_error_queue.lock,
1844 af_elock_keys + sk->sk_family,
1845 af_family_elock_key_strings[sk->sk_family]);
1846 lockdep_set_class_and_name(&sk->sk_callback_lock,
1847 af_callback_keys + sk->sk_family,
1848 af_family_clock_key_strings[sk->sk_family]);
1849}
1850
e56c57d0
ED		 1851/**
1852 * sk_clone_lock - clone a socket, and lock its clone
1853 * @sk: the socket to clone
1854 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1855 *
1856 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1857 */
1858struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1859{
b8e202d1 1860 struct proto *prot = READ_ONCE(sk->sk_prot);
8fd1d178 1861 struct sock *newsk;
278571ba 1862 bool is_charged = true;
87d11ceb 1863
b8e202d1 1864 newsk = sk_prot_alloc(prot, priority, sk->sk_family);
87d11ceb
ACM		 1865 if (newsk != NULL) {
1866 struct sk_filter *filter;
1867
892c141e 1868 sock_copy(newsk, sk);
87d11ceb 1869
b8e202d1 1870 newsk->sk_prot_creator = prot;
9d538fa6 1871
87d11ceb 1872 /* SANITY */
8a681736
SV		 1873 if (likely(newsk->sk_net_refcnt))
1874 get_net(sock_net(newsk));
87d11ceb
ACM		 1875 sk_node_init(&newsk->sk_node);
1876 sock_lock_init(newsk);
1877 bh_lock_sock(newsk);
fa438ccf 1878 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1879 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1880
1881 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1882 /*
1883 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1884 */
14afee4b 1885 refcount_set(&newsk->sk_wmem_alloc, 1);
87d11ceb 1886 atomic_set(&newsk->sk_omem_alloc, 0);
581319c5 1887 sk_init_common(newsk);
87d11ceb
ACM		 1888
1889 newsk->sk_dst_cache = NULL;
9b8805a3 1890 newsk->sk_dst_pending_confirm = 0;
87d11ceb
ACM		 1891 newsk->sk_wmem_queued = 0;
1892 newsk->sk_forward_alloc = 0;
9caad864 1893 atomic_set(&newsk->sk_drops, 0);
87d11ceb 1894 newsk->sk_send_head = NULL;
87d11ceb 1895 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
52267790 1896 atomic_set(&newsk->sk_zckey, 0);
87d11ceb
ACM		 1897
1898 sock_reset_flag(newsk, SOCK_DONE);
d752a498
SB		 1899
1900 /* sk->sk_memcg will be populated at accept() time */
1901 newsk->sk_memcg = NULL;
1902
c0576e39 1903 cgroup_sk_alloc(&newsk->sk_cgrp_data);
87d11ceb 1904
eefca20e
ED		 1905 rcu_read_lock();
1906 filter = rcu_dereference(sk->sk_filter);
87d11ceb 1907 if (filter != NULL)
278571ba
AS		 1908 /* though it's an empty new sock, the charging may fail
1909 * if sysctl_optmem_max was changed between creation of
1910 * original socket and cloning
1911 */
1912 is_charged = sk_filter_charge(newsk, filter);
eefca20e
ED		 1913 RCU_INIT_POINTER(newsk->sk_filter, filter);
1914 rcu_read_unlock();
87d11ceb 1915
d188ba86 1916 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
a97e50cc
DB		 1917 /* We need to make sure that we don't uncharge the new
1918 * socket if we couldn't charge it in the first place
1919 * as otherwise we uncharge the parent's filter.
1920 */
1921 if (!is_charged)
1922 RCU_INIT_POINTER(newsk->sk_filter, NULL);
94352d45 1923 sk_free_unlock_clone(newsk);
87d11ceb
ACM		 1924 newsk = NULL;
1925 goto out;
1926 }
fa463497 1927 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
8f51dfc7
SF		 1928
1929 if (bpf_sk_storage_clone(sk, newsk)) {
1930 sk_free_unlock_clone(newsk);
1931 newsk = NULL;
1932 goto out;
1933 }
87d11ceb 1934
f1ff5ce2
JS		 1935 /* Clear sk_user_data if parent had the pointer tagged
1936 * as not suitable for copying when cloning.
1937 */
1938 if (sk_user_data_is_nocopy(newsk))
7a1ca972 1939 newsk->sk_user_data = NULL;
f1ff5ce2 1940
87d11ceb 1941 newsk->sk_err = 0;
e551c32d 1942 newsk->sk_err_soft = 0;
87d11ceb 1943 newsk->sk_priority = 0;
2c8c56e1 1944 newsk->sk_incoming_cpu = raw_smp_processor_id();
648845ab
TZ		 1945 if (likely(newsk->sk_net_refcnt))
1946 sock_inuse_add(sock_net(newsk), 1);
d979a39d 1947
4dc6dc71
ED		 1948 /*
1949 * Before updating sk_refcnt, we must commit prior changes to memory
1950 * (Documentation/RCU/rculist_nulls.txt for details)
1951 */
1952 smp_wmb();
41c6d650 1953 refcount_set(&newsk->sk_refcnt, 2);
87d11ceb
ACM		 1954
1955 /*
1956 * Increment the counter in the same struct proto as the master
1957 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1958 * is the same as sk->sk_prot->socks, as this field was copied
1959 * with memcpy).
1960 *
1961 * This _changes_ the previous behaviour, where
1962 * tcp_create_openreq_child always was incrementing the
1963 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1964 * to be taken into account in all callers. -acme
1965 */
1966 sk_refcnt_debug_inc(newsk);
972692e0 1967 sk_set_socket(newsk, NULL);
c2f26e8f 1968 RCU_INIT_POINTER(newsk->sk_wq, NULL);
87d11ceb
ACM		 1969
1970 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1971 sk_sockets_allocated_inc(newsk);
704da560 1972
080a270f
HFS		 1973 if (sock_needs_netstamp(sk) &&
1974 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1975 net_enable_timestamp();
87d11ceb
ACM		 1976 }
1977out:
1978 return newsk;
1979}
e56c57d0 1980EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1981
94352d45
ACM		 1982void sk_free_unlock_clone(struct sock *sk)
1983{
1984 /* It is still raw copy of parent, so invalidate
1985 * destructor and make plain sk_free() */
1986 sk->sk_destruct = NULL;
1987 bh_unlock_sock(sk);
1988 sk_free(sk);
1989}
1990EXPORT_SYMBOL_GPL(sk_free_unlock_clone);
1991
9958089a
AK		 1992void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1993{
d6a4e26a
ED		 1994 u32 max_segs = 1;
1995
6bd4f355 1996 sk_dst_set(sk, dst);
0a6b2a1d 1997 sk->sk_route_caps = dst->dev->features | sk->sk_route_forced_caps;
9958089a 1998 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1999 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 2000 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 2001 if (sk_can_gso(sk)) {
f70f250a 2002 if (dst->header_len && !xfrm_dst_offload_ok(dst)) {
9958089a 2003 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 2004 } else {
9958089a 2005 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 2006 sk->sk_gso_max_size = dst->dev->gso_max_size;
d6a4e26a 2007 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
82cc1a7a 2008 }
9958089a 2009 }
d6a4e26a 2010 sk->sk_gso_max_segs = max_segs;
9958089a
AK		 2011}
2012EXPORT_SYMBOL_GPL(sk_setup_caps);
2013
1da177e4
LT		 2014/*
2015 * Simple resource managers for sockets.
2016 */
2017
2018
4ec93edb
YH		 2019/*
2020 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 2021 */
2022void sock_wfree(struct sk_buff *skb)
2023{
2024 struct sock *sk = skb->sk;
d99927f4 2025 unsigned int len = skb->truesize;
1da177e4 2026
d99927f4
ED		 2027 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
2028 /*
2029 * Keep a reference on sk_wmem_alloc, this will be released
2030 * after sk_write_space() call
2031 */
14afee4b 2032 WARN_ON(refcount_sub_and_test(len - 1, &sk->sk_wmem_alloc));
1da177e4 2033 sk->sk_write_space(sk);
d99927f4
ED		 2034 len = 1;
2035 }
2b85a34e 2036 /*
d99927f4
ED		 2037 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
2038 * could not do because of in-flight packets
2b85a34e 2039 */
14afee4b 2040 if (refcount_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 2041 __sk_free(sk);
1da177e4 2042}
2a91525c 2043EXPORT_SYMBOL(sock_wfree);
1da177e4 2044
1d2077ac
ED		 2045/* This variant of sock_wfree() is used by TCP,
2046 * since it sets SOCK_USE_WRITE_QUEUE.
2047 */
2048void __sock_wfree(struct sk_buff *skb)
2049{
2050 struct sock *sk = skb->sk;
2051
14afee4b 2052 if (refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc))
1d2077ac
ED		 2053 __sk_free(sk);
2054}
2055
9e17f8a4
ED		 2056void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
2057{
2058 skb_orphan(skb);
2059 skb->sk = sk;
2060#ifdef CONFIG_INET
2061 if (unlikely(!sk_fullsock(sk))) {
2062 skb->destructor = sock_edemux;
2063 sock_hold(sk);
2064 return;
2065 }
2066#endif
2067 skb->destructor = sock_wfree;
2068 skb_set_hash_from_sk(skb, sk);
2069 /*
2070 * We used to take a refcount on sk, but following operation
2071 * is enough to guarantee sk_free() wont free this sock until
2072 * all in-flight packets are completed
2073 */
14afee4b 2074 refcount_add(skb->truesize, &sk->sk_wmem_alloc);
9e17f8a4
ED		 2075}
2076EXPORT_SYMBOL(skb_set_owner_w);
2077
41477662
JK		 2078static bool can_skb_orphan_partial(const struct sk_buff *skb)
2079{
2080#ifdef CONFIG_TLS_DEVICE
2081 /* Drivers depend on in-order delivery for crypto offload,
2082 * partial orphan breaks out-of-order-OK logic.
2083 */
2084 if (skb->decrypted)
2085 return false;
2086#endif
2087 return (skb->destructor == sock_wfree ||
2088 (IS_ENABLED(CONFIG_INET) && skb->destructor == tcp_wfree));
2089}
2090
1d2077ac
ED		 2091/* This helper is used by netem, as it can hold packets in its
2092 * delay queue. We want to allow the owner socket to send more
2093 * packets, as if they were already TX completed by a typical driver.
2094 * But we also want to keep skb->sk set because some packet schedulers
f6ba8d33 2095 * rely on it (sch_fq for example).
1d2077ac 2096 */
f2f872f9
ED		 2097void skb_orphan_partial(struct sk_buff *skb)
2098{
f6ba8d33 2099 if (skb_is_tcp_pure_ack(skb))
1d2077ac
ED		 2100 return;
2101
41477662 2102 if (can_skb_orphan_partial(skb)) {
f6ba8d33
ED		 2103 struct sock *sk = skb->sk;
2104
41c6d650 2105 if (refcount_inc_not_zero(&sk->sk_refcnt)) {
14afee4b 2106 WARN_ON(refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc));
f6ba8d33
ED		 2107 skb->destructor = sock_efree;
2108 }
f2f872f9
ED		 2109 } else {
2110 skb_orphan(skb);
2111 }
2112}
2113EXPORT_SYMBOL(skb_orphan_partial);
2114
4ec93edb
YH		 2115/*
2116 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 2117 */
2118void sock_rfree(struct sk_buff *skb)
2119{
2120 struct sock *sk = skb->sk;
d361fd59 2121 unsigned int len = skb->truesize;
1da177e4 2122
d361fd59
ED		 2123 atomic_sub(len, &sk->sk_rmem_alloc);
2124 sk_mem_uncharge(sk, len);
1da177e4 2125}
2a91525c 2126EXPORT_SYMBOL(sock_rfree);
1da177e4 2127
7768eed8
OH		 2128/*
2129 * Buffer destructor for skbs that are not used directly in read or write
2130 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
2131 */
62bccb8c
AD		 2132void sock_efree(struct sk_buff *skb)
2133{
2134 sock_put(skb->sk);
2135}
2136EXPORT_SYMBOL(sock_efree);
2137
cf7fbe66
JS		 2138/* Buffer destructor for prefetch/receive path where reference count may
2139 * not be held, e.g. for listen sockets.
2140 */
2141#ifdef CONFIG_INET
2142void sock_pfree(struct sk_buff *skb)
2143{
7ae215d2
JS		 2144 if (sk_is_refcounted(skb->sk))
2145 sock_gen_put(skb->sk);
cf7fbe66
JS		 2146}
2147EXPORT_SYMBOL(sock_pfree);
2148#endif /* CONFIG_INET */
2149
976d0201 2150kuid_t sock_i_uid(struct sock *sk)
1da177e4 2151{
976d0201 2152 kuid_t uid;
1da177e4 2153
f064af1e 2154 read_lock_bh(&sk->sk_callback_lock);
976d0201 2155 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 2156 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 2157 return uid;
2158}
2a91525c 2159EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 2160
2161unsigned long sock_i_ino(struct sock *sk)
2162{
2163 unsigned long ino;
2164
f064af1e 2165 read_lock_bh(&sk->sk_callback_lock);
1da177e4 2166 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 2167 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 2168 return ino;
2169}
2a91525c 2170EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 2171
2172/*
2173 * Allocate a skb from the socket's send buffer.
2174 */
86a76caf 2175struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 2176 gfp_t priority)
1da177e4 2177{
e292f05e
ED		 2178 if (force ||
2179 refcount_read(&sk->sk_wmem_alloc) < READ_ONCE(sk->sk_sndbuf)) {
2a91525c 2180 struct sk_buff *skb = alloc_skb(size, priority);
e292f05e 2181
1da177e4
LT		 2182 if (skb) {
2183 skb_set_owner_w(skb, sk);
2184 return skb;
2185 }
2186 }
2187 return NULL;
2188}
2a91525c 2189EXPORT_SYMBOL(sock_wmalloc);
1da177e4 2190
98ba0bd5
WB		 2191static void sock_ofree(struct sk_buff *skb)
2192{
2193 struct sock *sk = skb->sk;
2194
2195 atomic_sub(skb->truesize, &sk->sk_omem_alloc);
2196}
2197
2198struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
2199 gfp_t priority)
2200{
2201 struct sk_buff *skb;
2202
2203 /* small safe race: SKB_TRUESIZE may differ from final skb->truesize */
2204 if (atomic_read(&sk->sk_omem_alloc) + SKB_TRUESIZE(size) >
2205 sysctl_optmem_max)
2206 return NULL;
2207
2208 skb = alloc_skb(size, priority);
2209 if (!skb)
2210 return NULL;
2211
2212 atomic_add(skb->truesize, &sk->sk_omem_alloc);
2213 skb->sk = sk;
2214 skb->destructor = sock_ofree;
2215 return skb;
2216}
2217
4ec93edb 2218/*
1da177e4 2219 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 2220 */
dd0fc66f 2221void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 2222{
95c96174 2223 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 2224 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
2225 void *mem;
2226 /* First do the add, to avoid the race if kmalloc
4ec93edb 2227 * might sleep.
1da177e4
LT		 2228 */
2229 atomic_add(size, &sk->sk_omem_alloc);
2230 mem = kmalloc(size, priority);
2231 if (mem)
2232 return mem;
2233 atomic_sub(size, &sk->sk_omem_alloc);
2234 }
2235 return NULL;
2236}
2a91525c 2237EXPORT_SYMBOL(sock_kmalloc);
1da177e4 2238
79e88659
DB		 2239/* Free an option memory block. Note, we actually want the inline
2240 * here as this allows gcc to detect the nullify and fold away the
2241 * condition entirely.
1da177e4 2242 */
79e88659
DB		 2243static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
2244 const bool nullify)
1da177e4 2245{
e53da5fb
DM		 2246 if (WARN_ON_ONCE(!mem))
2247 return;
79e88659
DB		 2248 if (nullify)
2249 kzfree(mem);
2250 else
2251 kfree(mem);
1da177e4
LT		 2252 atomic_sub(size, &sk->sk_omem_alloc);
2253}
79e88659
DB		 2254
2255void sock_kfree_s(struct sock *sk, void *mem, int size)
2256{
2257 __sock_kfree_s(sk, mem, size, false);
2258}
2a91525c 2259EXPORT_SYMBOL(sock_kfree_s);
1da177e4 2260
79e88659
DB		 2261void sock_kzfree_s(struct sock *sk, void *mem, int size)
2262{
2263 __sock_kfree_s(sk, mem, size, true);
2264}
2265EXPORT_SYMBOL(sock_kzfree_s);
2266
1da177e4
LT		 2267/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
2268 I think, these locks should be removed for datagram sockets.
2269 */
2a91525c 2270static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 2271{
2272 DEFINE_WAIT(wait);
2273
9cd3e072 2274 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1da177e4
LT		 2275 for (;;) {
2276 if (!timeo)
2277 break;
2278 if (signal_pending(current))
2279 break;
2280 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 2281 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
e292f05e 2282 if (refcount_read(&sk->sk_wmem_alloc) < READ_ONCE(sk->sk_sndbuf))
1da177e4
LT		 2283 break;
2284 if (sk->sk_shutdown & SEND_SHUTDOWN)
2285 break;
2286 if (sk->sk_err)
2287 break;
2288 timeo = schedule_timeout(timeo);
2289 }
aa395145 2290 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2291 return timeo;
2292}
2293
2294
2295/*
2296 * Generic send/receive buffer handlers
2297 */
2298
4cc7f68d
HX		 2299struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
2300 unsigned long data_len, int noblock,
28d64271 2301 int *errcode, int max_page_order)
1da177e4 2302{
2e4e4410 2303 struct sk_buff *skb;
1da177e4
LT		 2304 long timeo;
2305 int err;
2306
1da177e4 2307 timeo = sock_sndtimeo(sk, noblock);
2e4e4410 2308 for (;;) {
1da177e4
LT		 2309 err = sock_error(sk);
2310 if (err != 0)
2311 goto failure;
2312
2313 err = -EPIPE;
2314 if (sk->sk_shutdown & SEND_SHUTDOWN)
2315 goto failure;
2316
e292f05e 2317 if (sk_wmem_alloc_get(sk) < READ_ONCE(sk->sk_sndbuf))
2e4e4410 2318 break;
28d64271 2319
9cd3e072 2320 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2e4e4410
ED		 2321 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2322 err = -EAGAIN;
2323 if (!timeo)
1da177e4 2324 goto failure;
2e4e4410
ED		 2325 if (signal_pending(current))
2326 goto interrupted;
2327 timeo = sock_wait_for_wmem(sk, timeo);
1da177e4 2328 }
2e4e4410
ED		 2329 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
2330 errcode, sk->sk_allocation);
2331 if (skb)
2332 skb_set_owner_w(skb, sk);
1da177e4
LT		 2333 return skb;
2334
2335interrupted:
2336 err = sock_intr_errno(timeo);
2337failure:
2338 *errcode = err;
2339 return NULL;
2340}
4cc7f68d 2341EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 2342
4ec93edb 2343struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 2344 int noblock, int *errcode)
2345{
28d64271 2346 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 2347}
2a91525c 2348EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 2349
39771b12
WB		 2350int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
2351 struct sockcm_cookie *sockc)
2352{
3dd17e63
SHY		 2353 u32 tsflags;
2354
39771b12
WB		 2355 switch (cmsg->cmsg_type) {
2356 case SO_MARK:
2357 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2358 return -EPERM;
2359 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2360 return -EINVAL;
2361 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
2362 break;
7f1bc6e9 2363 case SO_TIMESTAMPING_OLD:
3dd17e63
SHY		 2364 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2365 return -EINVAL;
2366
2367 tsflags = *(u32 *)CMSG_DATA(cmsg);
2368 if (tsflags & ~SOF_TIMESTAMPING_TX_RECORD_MASK)
2369 return -EINVAL;
2370
2371 sockc->tsflags &= ~SOF_TIMESTAMPING_TX_RECORD_MASK;
2372 sockc->tsflags |= tsflags;
2373 break;
80b14dee
RC		 2374 case SCM_TXTIME:
2375 if (!sock_flag(sk, SOCK_TXTIME))
2376 return -EINVAL;
2377 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u64)))
2378 return -EINVAL;
2379 sockc->transmit_time = get_unaligned((u64 *)CMSG_DATA(cmsg));
2380 break;
779f1ede
SHY		 2381 /* SCM_RIGHTS and SCM_CREDENTIALS are semantically in SOL_UNIX. */
2382 case SCM_RIGHTS:
2383 case SCM_CREDENTIALS:
2384 break;
39771b12
WB		 2385 default:
2386 return -EINVAL;
2387 }
2388 return 0;
2389}
2390EXPORT_SYMBOL(__sock_cmsg_send);
2391
f28ea365
EJ		 2392int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
2393 struct sockcm_cookie *sockc)
2394{
2395 struct cmsghdr *cmsg;
39771b12 2396 int ret;
f28ea365
EJ		 2397
2398 for_each_cmsghdr(cmsg, msg) {
2399 if (!CMSG_OK(msg, cmsg))
2400 return -EINVAL;
2401 if (cmsg->cmsg_level != SOL_SOCKET)
2402 continue;
39771b12
WB		 2403 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
2404 if (ret)
2405 return ret;
f28ea365
EJ		 2406 }
2407 return 0;
2408}
2409EXPORT_SYMBOL(sock_cmsg_send);
2410
06044751
ED		 2411static void sk_enter_memory_pressure(struct sock *sk)
2412{
2413 if (!sk->sk_prot->enter_memory_pressure)
2414 return;
2415
2416 sk->sk_prot->enter_memory_pressure(sk);
2417}
2418
2419static void sk_leave_memory_pressure(struct sock *sk)
2420{
2421 if (sk->sk_prot->leave_memory_pressure) {
2422 sk->sk_prot->leave_memory_pressure(sk);
2423 } else {
2424 unsigned long *memory_pressure = sk->sk_prot->memory_pressure;
2425
503978ac
ED		 2426 if (memory_pressure && READ_ONCE(*memory_pressure))
2427 WRITE_ONCE(*memory_pressure, 0);
06044751
ED		 2428 }
2429}
2430
5640f768 2431#define SKB_FRAG_PAGE_ORDER get_order(32768)
ce27ec60 2432DEFINE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key);
5640f768 2433
400dfd3a
ED		 2434/**
2435 * skb_page_frag_refill - check that a page_frag contains enough room
2436 * @sz: minimum size of the fragment we want to get
2437 * @pfrag: pointer to page_frag
82d5e2b8 2438 * @gfp: priority for memory allocation
400dfd3a
ED		 2439 *
2440 * Note: While this allocator tries to use high order pages, there is
2441 * no guarantee that allocations succeed. Therefore, @sz MUST be
2442 * less or equal than PAGE_SIZE.
2443 */
d9b2938a 2444bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
5640f768 2445{
5640f768 2446 if (pfrag->page) {
fe896d18 2447 if (page_ref_count(pfrag->page) == 1) {
5640f768
ED		 2448 pfrag->offset = 0;
2449 return true;
2450 }
400dfd3a 2451 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 2452 return true;
2453 put_page(pfrag->page);
2454 }
2455
d9b2938a 2456 pfrag->offset = 0;
ce27ec60
ED		 2457 if (SKB_FRAG_PAGE_ORDER &&
2458 !static_branch_unlikely(&net_high_order_alloc_disable_key)) {
d0164adc
MG		 2459 /* Avoid direct reclaim but allow kswapd to wake */
2460 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
2461 __GFP_COMP | __GFP_NOWARN |
2462 __GFP_NORETRY,
d9b2938a 2463 SKB_FRAG_PAGE_ORDER);
5640f768 2464 if (likely(pfrag->page)) {
d9b2938a 2465 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
5640f768
ED		 2466 return true;
2467 }
d9b2938a
ED		 2468 }
2469 pfrag->page = alloc_page(gfp);
2470 if (likely(pfrag->page)) {
2471 pfrag->size = PAGE_SIZE;
2472 return true;
2473 }
400dfd3a
ED		 2474 return false;
2475}
2476EXPORT_SYMBOL(skb_page_frag_refill);
2477
2478bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
2479{
2480 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
2481 return true;
2482
5640f768
ED		 2483 sk_enter_memory_pressure(sk);
2484 sk_stream_moderate_sndbuf(sk);
2485 return false;
2486}
2487EXPORT_SYMBOL(sk_page_frag_refill);
2488
1da177e4 2489static void __lock_sock(struct sock *sk)
f39234d6
NK		 2490 __releases(&sk->sk_lock.slock)
2491 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 2492{
2493 DEFINE_WAIT(wait);
2494
e71a4783 2495 for (;;) {
1da177e4
LT		 2496 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
2497 TASK_UNINTERRUPTIBLE);
2498 spin_unlock_bh(&sk->sk_lock.slock);
2499 schedule();
2500 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 2501 if (!sock_owned_by_user(sk))
1da177e4
LT		 2502 break;
2503 }
2504 finish_wait(&sk->sk_lock.wq, &wait);
2505}
2506
8873c064 2507void __release_sock(struct sock *sk)
f39234d6
NK		 2508 __releases(&sk->sk_lock.slock)
2509 __acquires(&sk->sk_lock.slock)
1da177e4 2510{
5413d1ba 2511 struct sk_buff *skb, *next;
1da177e4 2512
5413d1ba 2513 while ((skb = sk->sk_backlog.head) != NULL) {
1da177e4 2514 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1da177e4 2515
5413d1ba 2516 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4 2517
5413d1ba
ED		 2518 do {
2519 next = skb->next;
e4cbb02a 2520 prefetch(next);
7fee226a 2521 WARN_ON_ONCE(skb_dst_is_noref(skb));
a8305bff 2522 skb_mark_not_on_list(skb);
c57943a1 2523 sk_backlog_rcv(sk, skb);
1da177e4 2524
5413d1ba 2525 cond_resched();
1da177e4
LT		 2526
2527 skb = next;
2528 } while (skb != NULL);
2529
5413d1ba
ED		 2530 spin_lock_bh(&sk->sk_lock.slock);
2531 }
8eae939f
ZY		 2532
2533 /*
2534 * Doing the zeroing here guarantee we can not loop forever
2535 * while a wild producer attempts to flood us.
2536 */
2537 sk->sk_backlog.len = 0;
1da177e4
LT		 2538}
2539
d41a69f1
ED		 2540void __sk_flush_backlog(struct sock *sk)
2541{
2542 spin_lock_bh(&sk->sk_lock.slock);
2543 __release_sock(sk);
2544 spin_unlock_bh(&sk->sk_lock.slock);
2545}
2546
1da177e4
LT		 2547/**
2548 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 2549 * @sk: sock to wait on
2550 * @timeo: for how long
dfbafc99 2551 * @skb: last skb seen on sk_receive_queue
1da177e4
LT		 2552 *
2553 * Now socket state including sk->sk_err is changed only under lock,
2554 * hence we may omit checks after joining wait queue.
2555 * We check receive queue before schedule() only as optimization;
2556 * it is very likely that release_sock() added new data.
2557 */
dfbafc99 2558int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
1da177e4 2559{
d9dc8b0f 2560 DEFINE_WAIT_FUNC(wait, woken_wake_function);
1da177e4 2561 int rc;
1da177e4 2562
d9dc8b0f 2563 add_wait_queue(sk_sleep(sk), &wait);
9cd3e072 2564 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
d9dc8b0f 2565 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb, &wait);
9cd3e072 2566 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
d9dc8b0f 2567 remove_wait_queue(sk_sleep(sk), &wait);
1da177e4
LT		 2568 return rc;
2569}
1da177e4
LT		 2570EXPORT_SYMBOL(sk_wait_data);
2571
3ab224be 2572/**
f8c3bf00 2573 * __sk_mem_raise_allocated - increase memory_allocated
3ab224be
HA		 2574 * @sk: socket
2575 * @size: memory size to allocate
f8c3bf00 2576 * @amt: pages to allocate
3ab224be
HA		 2577 * @kind: allocation type
2578 *
f8c3bf00 2579 * Similar to __sk_mem_schedule(), but does not update sk_forward_alloc
3ab224be 2580 */
f8c3bf00 2581int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
3ab224be
HA		 2582{
2583 struct proto *prot = sk->sk_prot;
f8c3bf00 2584 long allocated = sk_memory_allocated_add(sk, amt);
d6f19938 2585 bool charged = true;
e805605c 2586
baac50bb 2587 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
d6f19938 2588 !(charged = mem_cgroup_charge_skmem(sk->sk_memcg, amt)))
e805605c 2589 goto suppress_allocation;
3ab224be
HA		 2590
2591 /* Under limit. */
e805605c 2592 if (allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2593 sk_leave_memory_pressure(sk);
3ab224be
HA		 2594 return 1;
2595 }
2596
e805605c
JW		 2597 /* Under pressure. */
2598 if (allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2599 sk_enter_memory_pressure(sk);
3ab224be 2600
e805605c
JW		 2601 /* Over hard limit. */
2602 if (allocated > sk_prot_mem_limits(sk, 2))
3ab224be
HA		 2603 goto suppress_allocation;
2604
2605 /* guarantee minimum buffer size under pressure */
2606 if (kind == SK_MEM_RECV) {
a3dcaf17 2607 if (atomic_read(&sk->sk_rmem_alloc) < sk_get_rmem0(sk, prot))
3ab224be 2608 return 1;
180d8cd9 2609
3ab224be 2610 } else { /* SK_MEM_SEND */
a3dcaf17
ED		 2611 int wmem0 = sk_get_wmem0(sk, prot);
2612
3ab224be 2613 if (sk->sk_type == SOCK_STREAM) {
a3dcaf17 2614 if (sk->sk_wmem_queued < wmem0)
3ab224be 2615 return 1;
a3dcaf17 2616 } else if (refcount_read(&sk->sk_wmem_alloc) < wmem0) {
3ab224be 2617 return 1;
a3dcaf17 2618 }
3ab224be
HA		 2619 }
2620
180d8cd9 2621 if (sk_has_memory_pressure(sk)) {
5bf325a5 2622 u64 alloc;
1748376b 2623
180d8cd9 2624 if (!sk_under_memory_pressure(sk))
1748376b 2625 return 1;
180d8cd9
GC		 2626 alloc = sk_sockets_allocated_read_positive(sk);
2627 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2628 sk_mem_pages(sk->sk_wmem_queued +
2629 atomic_read(&sk->sk_rmem_alloc) +
2630 sk->sk_forward_alloc))
2631 return 1;
2632 }
2633
2634suppress_allocation:
2635
2636 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2637 sk_stream_moderate_sndbuf(sk);
2638
2639 /* Fail only if socket is _under_ its sndbuf.
2640 * In this case we cannot block, so that we have to fail.
2641 */
2642 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2643 return 1;
2644 }
2645
d6f19938
YS		 2646 if (kind == SK_MEM_SEND || (kind == SK_MEM_RECV && charged))
2647 trace_sock_exceed_buf_limit(sk, prot, allocated, kind);
3847ce32 2648
0e90b31f 2649 sk_memory_allocated_sub(sk, amt);
180d8cd9 2650
baac50bb
JW		 2651 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2652 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
e805605c 2653
3ab224be
HA		 2654 return 0;
2655}
f8c3bf00
PA		 2656EXPORT_SYMBOL(__sk_mem_raise_allocated);
2657
2658/**
2659 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2660 * @sk: socket
2661 * @size: memory size to allocate
2662 * @kind: allocation type
2663 *
2664 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2665 * rmem allocation. This function assumes that protocols which have
2666 * memory_pressure use sk_wmem_queued as write buffer accounting.
2667 */
2668int __sk_mem_schedule(struct sock *sk, int size, int kind)
2669{
2670 int ret, amt = sk_mem_pages(size);
2671
2672 sk->sk_forward_alloc += amt << SK_MEM_QUANTUM_SHIFT;
2673 ret = __sk_mem_raise_allocated(sk, size, amt, kind);
2674 if (!ret)
2675 sk->sk_forward_alloc -= amt << SK_MEM_QUANTUM_SHIFT;
2676 return ret;
2677}
3ab224be
HA		 2678EXPORT_SYMBOL(__sk_mem_schedule);
2679
2680/**
f8c3bf00 2681 * __sk_mem_reduce_allocated - reclaim memory_allocated
3ab224be 2682 * @sk: socket
f8c3bf00
PA		 2683 * @amount: number of quanta
2684 *
2685 * Similar to __sk_mem_reclaim(), but does not update sk_forward_alloc
3ab224be 2686 */
f8c3bf00 2687void __sk_mem_reduce_allocated(struct sock *sk, int amount)
3ab224be 2688{
1a24e04e 2689 sk_memory_allocated_sub(sk, amount);
3ab224be 2690
baac50bb
JW		 2691 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2692 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
e805605c 2693
180d8cd9
GC		 2694 if (sk_under_memory_pressure(sk) &&
2695 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2696 sk_leave_memory_pressure(sk);
3ab224be 2697}
f8c3bf00
PA		 2698EXPORT_SYMBOL(__sk_mem_reduce_allocated);
2699
2700/**
2701 * __sk_mem_reclaim - reclaim sk_forward_alloc and memory_allocated
2702 * @sk: socket
2703 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
2704 */
2705void __sk_mem_reclaim(struct sock *sk, int amount)
2706{
2707 amount >>= SK_MEM_QUANTUM_SHIFT;
2708 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
2709 __sk_mem_reduce_allocated(sk, amount);
2710}
3ab224be
HA		 2711EXPORT_SYMBOL(__sk_mem_reclaim);
2712
627d2d6b 2713int sk_set_peek_off(struct sock *sk, int val)
2714{
627d2d6b 2715 sk->sk_peek_off = val;
2716 return 0;
2717}
2718EXPORT_SYMBOL_GPL(sk_set_peek_off);
3ab224be 2719
1da177e4
LT		 2720/*
2721 * Set of default routines for initialising struct proto_ops when
2722 * the protocol does not support a particular function. In certain
2723 * cases where it makes no sense for a protocol to have a "do nothing"
2724 * function, some default processing is provided.
2725 */
2726
2727int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2728{
2729 return -EOPNOTSUPP;
2730}
2a91525c 2731EXPORT_SYMBOL(sock_no_bind);
1da177e4 2732
4ec93edb 2733int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2734 int len, int flags)
2735{
2736 return -EOPNOTSUPP;
2737}
2a91525c 2738EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2739
2740int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2741{
2742 return -EOPNOTSUPP;
2743}
2a91525c 2744EXPORT_SYMBOL(sock_no_socketpair);
1da177e4 2745
cdfbabfb
DH		 2746int sock_no_accept(struct socket *sock, struct socket *newsock, int flags,
2747 bool kern)
1da177e4
LT		 2748{
2749 return -EOPNOTSUPP;
2750}
2a91525c 2751EXPORT_SYMBOL(sock_no_accept);
1da177e4 2752
4ec93edb 2753int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
9b2c45d4 2754 int peer)
1da177e4
LT		 2755{
2756 return -EOPNOTSUPP;
2757}
2a91525c 2758EXPORT_SYMBOL(sock_no_getname);
1da177e4 2759
1da177e4
LT		 2760int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2761{
2762 return -EOPNOTSUPP;
2763}
2a91525c 2764EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2765
2766int sock_no_listen(struct socket *sock, int backlog)
2767{
2768 return -EOPNOTSUPP;
2769}
2a91525c 2770EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2771
2772int sock_no_shutdown(struct socket *sock, int how)
2773{
2774 return -EOPNOTSUPP;
2775}
2a91525c 2776EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2777
2778int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2779 char __user *optval, unsigned int optlen)
1da177e4
LT		 2780{
2781 return -EOPNOTSUPP;
2782}
2a91525c 2783EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2784
2785int sock_no_getsockopt(struct socket *sock, int level, int optname,
2786 char __user *optval, int __user *optlen)
2787{
2788 return -EOPNOTSUPP;
2789}
2a91525c 2790EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4 2791
1b784140 2792int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
1da177e4
LT		 2793{
2794 return -EOPNOTSUPP;
2795}
2a91525c 2796EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4 2797
306b13eb
TH		 2798int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *m, size_t len)
2799{
2800 return -EOPNOTSUPP;
2801}
2802EXPORT_SYMBOL(sock_no_sendmsg_locked);
2803
1b784140
YX		 2804int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2805 int flags)
1da177e4
LT		 2806{
2807 return -EOPNOTSUPP;
2808}
2a91525c 2809EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2810
2811int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2812{
2813 /* Mirror missing mmap method error code */
2814 return -ENODEV;
2815}
2a91525c 2816EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2817
2818ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2819{
2820 ssize_t res;
2821 struct msghdr msg = {.msg_flags = flags};
2822 struct kvec iov;
2823 char *kaddr = kmap(page);
2824 iov.iov_base = kaddr + offset;
2825 iov.iov_len = size;
2826 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2827 kunmap(page);
2828 return res;
2829}
2a91525c 2830EXPORT_SYMBOL(sock_no_sendpage);
1da177e4 2831
306b13eb
TH		 2832ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
2833 int offset, size_t size, int flags)
2834{
2835 ssize_t res;
2836 struct msghdr msg = {.msg_flags = flags};
2837 struct kvec iov;
2838 char *kaddr = kmap(page);
2839
2840 iov.iov_base = kaddr + offset;
2841 iov.iov_len = size;
2842 res = kernel_sendmsg_locked(sk, &msg, &iov, 1, size);
2843 kunmap(page);
2844 return res;
2845}
2846EXPORT_SYMBOL(sock_no_sendpage_locked);
2847
1da177e4
LT		 2848/*
2849 * Default Socket Callbacks
2850 */
2851
2852static void sock_def_wakeup(struct sock *sk)
2853{
43815482
ED		 2854 struct socket_wq *wq;
2855
2856 rcu_read_lock();
2857 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2858 if (skwq_has_sleeper(wq))
43815482
ED		 2859 wake_up_interruptible_all(&wq->wait);
2860 rcu_read_unlock();
1da177e4
LT		 2861}
2862
2863static void sock_def_error_report(struct sock *sk)
2864{
43815482
ED		 2865 struct socket_wq *wq;
2866
2867 rcu_read_lock();
2868 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2869 if (skwq_has_sleeper(wq))
a9a08845 2870 wake_up_interruptible_poll(&wq->wait, EPOLLERR);
8d8ad9d7 2871 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2872 rcu_read_unlock();
1da177e4
LT		 2873}
2874
43a825af 2875void sock_def_readable(struct sock *sk)
1da177e4 2876{
43815482
ED		 2877 struct socket_wq *wq;
2878
2879 rcu_read_lock();
2880 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2881 if (skwq_has_sleeper(wq))
a9a08845
LT		 2882 wake_up_interruptible_sync_poll(&wq->wait, EPOLLIN | EPOLLPRI |
2883 EPOLLRDNORM | EPOLLRDBAND);
8d8ad9d7 2884 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2885 rcu_read_unlock();
1da177e4
LT		 2886}
2887
2888static void sock_def_write_space(struct sock *sk)
2889{
43815482
ED		 2890 struct socket_wq *wq;
2891
2892 rcu_read_lock();
1da177e4
LT		 2893
2894 /* Do not wake up a writer until he can make "significant"
2895 * progress. --DaveM
2896 */
e292f05e 2897 if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= READ_ONCE(sk->sk_sndbuf)) {
43815482 2898 wq = rcu_dereference(sk->sk_wq);
1ce0bf50 2899 if (skwq_has_sleeper(wq))
a9a08845
LT		 2900 wake_up_interruptible_sync_poll(&wq->wait, EPOLLOUT |
2901 EPOLLWRNORM | EPOLLWRBAND);
1da177e4
LT		 2902
2903 /* Should agree with poll, otherwise some programs break */
2904 if (sock_writeable(sk))
8d8ad9d7 2905 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2906 }
2907
43815482 2908 rcu_read_unlock();
1da177e4
LT		 2909}
2910
2911static void sock_def_destruct(struct sock *sk)
2912{
1da177e4
LT		 2913}
2914
2915void sk_send_sigurg(struct sock *sk)
2916{
2917 if (sk->sk_socket && sk->sk_socket->file)
2918 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2919 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2920}
2a91525c 2921EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2922
2923void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2924 unsigned long expires)
2925{
2926 if (!mod_timer(timer, expires))
2927 sock_hold(sk);
2928}
1da177e4
LT		 2929EXPORT_SYMBOL(sk_reset_timer);
2930
2931void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2932{
25cc4ae9 2933 if (del_timer(timer))
1da177e4
LT		 2934 __sock_put(sk);
2935}
1da177e4
LT		 2936EXPORT_SYMBOL(sk_stop_timer);
2937
2938void sock_init_data(struct socket *sock, struct sock *sk)
2939{
581319c5 2940 sk_init_common(sk);
1da177e4
LT		 2941 sk->sk_send_head = NULL;
2942
99767f27 2943 timer_setup(&sk->sk_timer, NULL, 0);
4ec93edb 2944
1da177e4
LT		 2945 sk->sk_allocation = GFP_KERNEL;
2946 sk->sk_rcvbuf = sysctl_rmem_default;
2947 sk->sk_sndbuf = sysctl_wmem_default;
2948 sk->sk_state = TCP_CLOSE;
972692e0 2949 sk_set_socket(sk, sock);
1da177e4
LT		 2950
2951 sock_set_flag(sk, SOCK_ZAPPED);
2952
e71a4783 2953 if (sock) {
1da177e4 2954 sk->sk_type = sock->type;
333f7909 2955 RCU_INIT_POINTER(sk->sk_wq, &sock->wq);
1da177e4 2956 sock->sk = sk;
86741ec2
LC		 2957 sk->sk_uid = SOCK_INODE(sock)->i_uid;
2958 } else {
c2f26e8f 2959 RCU_INIT_POINTER(sk->sk_wq, NULL);
86741ec2
LC		 2960 sk->sk_uid = make_kuid(sock_net(sk)->user_ns, 0);
2961 }
1da177e4 2962
1da177e4 2963 rwlock_init(&sk->sk_callback_lock);
cdfbabfb
DH		 2964 if (sk->sk_kern_sock)
2965 lockdep_set_class_and_name(
2966 &sk->sk_callback_lock,
2967 af_kern_callback_keys + sk->sk_family,
2968 af_family_kern_clock_key_strings[sk->sk_family]);
2969 else
2970 lockdep_set_class_and_name(
2971 &sk->sk_callback_lock,
443aef0e
PZ		 2972 af_callback_keys + sk->sk_family,
2973 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2974
2975 sk->sk_state_change = sock_def_wakeup;
2976 sk->sk_data_ready = sock_def_readable;
2977 sk->sk_write_space = sock_def_write_space;
2978 sk->sk_error_report = sock_def_error_report;
2979 sk->sk_destruct = sock_def_destruct;
2980
5640f768
ED		 2981 sk->sk_frag.page = NULL;
2982 sk->sk_frag.offset = 0;
ef64a54f 2983 sk->sk_peek_off = -1;
1da177e4 2984
109f6e39
EB		 2985 sk->sk_peer_pid = NULL;
2986 sk->sk_peer_cred = NULL;
1da177e4
LT		 2987 sk->sk_write_pending = 0;
2988 sk->sk_rcvlowat = 1;
2989 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2990 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2991
6c7c98ba 2992 sk->sk_stamp = SK_DEFAULT_STAMP;
3a0ed3e9
DD		 2993#if BITS_PER_LONG==32
2994 seqlock_init(&sk->sk_stamp_seq);
2995#endif
52267790 2996 atomic_set(&sk->sk_zckey, 0);
1da177e4 2997
e0d1095a 2998#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2999 sk->sk_napi_id = 0;
64b0dc51 3000 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 3001#endif
3002
76a9ebe8
ED		 3003 sk->sk_max_pacing_rate = ~0UL;
3004 sk->sk_pacing_rate = ~0UL;
7c68fa2b 3005 WRITE_ONCE(sk->sk_pacing_shift, 10);
70da268b 3006 sk->sk_incoming_cpu = -1;
c6345ce7
AN		 3007
3008 sk_rx_queue_clear(sk);
4dc6dc71
ED		 3009 /*
3010 * Before updating sk_refcnt, we must commit prior changes to memory
3011 * (Documentation/RCU/rculist_nulls.txt for details)
3012 */
3013 smp_wmb();
41c6d650 3014 refcount_set(&sk->sk_refcnt, 1);
33c732c3 3015 atomic_set(&sk->sk_drops, 0);
1da177e4 3016}
2a91525c 3017EXPORT_SYMBOL(sock_init_data);
1da177e4 3018
b5606c2d 3019void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 3020{
3021 might_sleep();
a5b5bb9a 3022 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 3023 if (sk->sk_lock.owned)
1da177e4 3024 __lock_sock(sk);
d2e9117c 3025 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 3026 spin_unlock(&sk->sk_lock.slock);
3027 /*
3028 * The sk_lock has mutex_lock() semantics here:
3029 */
fcc70d5f 3030 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 3031 local_bh_enable();
1da177e4 3032}
fcc70d5f 3033EXPORT_SYMBOL(lock_sock_nested);
1da177e4 3034
b5606c2d 3035void release_sock(struct sock *sk)
1da177e4 3036{
a5b5bb9a 3037 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 3038 if (sk->sk_backlog.tail)
3039 __release_sock(sk);
46d3ceab 3040
c3f9b018
ED		 3041 /* Warning : release_cb() might need to release sk ownership,
3042 * ie call sock_release_ownership(sk) before us.
3043 */
46d3ceab
ED		 3044 if (sk->sk_prot->release_cb)
3045 sk->sk_prot->release_cb(sk);
3046
c3f9b018 3047 sock_release_ownership(sk);
a5b5bb9a
IM		 3048 if (waitqueue_active(&sk->sk_lock.wq))
3049 wake_up(&sk->sk_lock.wq);
3050 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 3051}
3052EXPORT_SYMBOL(release_sock);
3053
8a74ad60
ED		 3054/**
3055 * lock_sock_fast - fast version of lock_sock
3056 * @sk: socket
3057 *
3058 * This version should be used for very small section, where process wont block
d651983d
MCC		 3059 * return false if fast path is taken:
3060 *
8a74ad60 3061 * sk_lock.slock locked, owned = 0, BH disabled
d651983d
MCC		 3062 *
3063 * return true if slow path is taken:
3064 *
8a74ad60
ED		 3065 * sk_lock.slock unlocked, owned = 1, BH enabled
3066 */
3067bool lock_sock_fast(struct sock *sk)
3068{
3069 might_sleep();
3070 spin_lock_bh(&sk->sk_lock.slock);
3071
3072 if (!sk->sk_lock.owned)
3073 /*
3074 * Note : We must disable BH
3075 */
3076 return false;
3077
3078 __lock_sock(sk);
3079 sk->sk_lock.owned = 1;
3080 spin_unlock(&sk->sk_lock.slock);
3081 /*
3082 * The sk_lock has mutex_lock() semantics here:
3083 */
3084 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
3085 local_bh_enable();
3086 return true;
3087}
3088EXPORT_SYMBOL(lock_sock_fast);
3089
c7cbdbf2
AB		 3090int sock_gettstamp(struct socket *sock, void __user *userstamp,
3091 bool timeval, bool time32)
4ec93edb 3092{
c7cbdbf2
AB		 3093 struct sock *sk = sock->sk;
3094 struct timespec64 ts;
9dae3497
YS		 3095
3096 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
c7cbdbf2
AB		 3097 ts = ktime_to_timespec64(sock_read_timestamp(sk));
3098 if (ts.tv_sec == -1)
1da177e4 3099 return -ENOENT;
c7cbdbf2 3100 if (ts.tv_sec == 0) {
3a0ed3e9 3101 ktime_t kt = ktime_get_real();
f95f96a4 3102 sock_write_timestamp(sk, kt);
c7cbdbf2 3103 ts = ktime_to_timespec64(kt);
b7aa0bf7 3104 }
1da177e4 3105
c7cbdbf2
AB		 3106 if (timeval)
3107 ts.tv_nsec /= 1000;
9dae3497 3108
c7cbdbf2
AB		 3109#ifdef CONFIG_COMPAT_32BIT_TIME
3110 if (time32)
3111 return put_old_timespec32(&ts, userstamp);
3112#endif
3113#ifdef CONFIG_SPARC64
3114 /* beware of padding in sparc64 timeval */
3115 if (timeval && !in_compat_syscall()) {
3116 struct __kernel_old_timeval __user tv = {
c98f4822
SR		 3117 .tv_sec = ts.tv_sec,
3118 .tv_usec = ts.tv_nsec,
c7cbdbf2 3119 };
c98f4822 3120 if (copy_to_user(userstamp, &tv, sizeof(tv)))
c7cbdbf2
AB		 3121 return -EFAULT;
3122 return 0;
ae40eb1e 3123 }
c7cbdbf2
AB		 3124#endif
3125 return put_timespec64(&ts, userstamp);
ae40eb1e 3126}
c7cbdbf2 3127EXPORT_SYMBOL(sock_gettstamp);
ae40eb1e 3128
193d357d 3129void sock_enable_timestamp(struct sock *sk, enum sock_flags flag)
4ec93edb 3130{
20d49473 3131 if (!sock_flag(sk, flag)) {
08e29af3
ED		 3132 unsigned long previous_flags = sk->sk_flags;
3133
20d49473
PO		 3134 sock_set_flag(sk, flag);
3135 /*
3136 * we just set one of the two flags which require net
3137 * time stamping, but time stamping might have been on
3138 * already because of the other one
3139 */
080a270f
HFS		 3140 if (sock_needs_netstamp(sk) &&
3141 !(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 3142 net_enable_timestamp();
1da177e4
LT		 3143 }
3144}
1da177e4 3145
cb820f8e
RC		 3146int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
3147 int level, int type)
3148{
3149 struct sock_exterr_skb *serr;
364a9e93 3150 struct sk_buff *skb;
cb820f8e
RC		 3151 int copied, err;
3152
3153 err = -EAGAIN;
364a9e93 3154 skb = sock_dequeue_err_skb(sk);
cb820f8e
RC		 3155 if (skb == NULL)
3156 goto out;
3157
3158 copied = skb->len;
3159 if (copied > len) {
3160 msg->msg_flags |= MSG_TRUNC;
3161 copied = len;
3162 }
51f3d02b 3163 err = skb_copy_datagram_msg(skb, 0, msg, copied);
cb820f8e
RC		 3164 if (err)
3165 goto out_free_skb;
3166
3167 sock_recv_timestamp(msg, sk, skb);
3168
3169 serr = SKB_EXT_ERR(skb);
3170 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
3171
3172 msg->msg_flags |= MSG_ERRQUEUE;
3173 err = copied;
3174
cb820f8e
RC		 3175out_free_skb:
3176 kfree_skb(skb);
3177out:
3178 return err;
3179}
3180EXPORT_SYMBOL(sock_recv_errqueue);
3181
1da177e4
LT		 3182/*
3183 * Get a socket option on an socket.
3184 *
3185 * FIX: POSIX 1003.1g is very ambiguous here. It states that
3186 * asynchronous errors should be reported by getsockopt. We assume
3187 * this means if you specify SO_ERROR (otherwise whats the point of it).
3188 */
3189int sock_common_getsockopt(struct socket *sock, int level, int optname,
3190 char __user *optval, int __user *optlen)
3191{
3192 struct sock *sk = sock->sk;
3193
3194 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
3195}
1da177e4
LT		 3196EXPORT_SYMBOL(sock_common_getsockopt);
3197
3fdadf7d 3198#ifdef CONFIG_COMPAT
543d9cfe
ACM		 3199int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
3200 char __user *optval, int __user *optlen)
3fdadf7d
DM		 3201{
3202 struct sock *sk = sock->sk;
3203
1e51f951 3204 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 3205 return sk->sk_prot->compat_getsockopt(sk, level, optname,
3206 optval, optlen);
3fdadf7d
DM		 3207 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
3208}
3209EXPORT_SYMBOL(compat_sock_common_getsockopt);
3210#endif
3211
1b784140
YX		 3212int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
3213 int flags)
1da177e4
LT		 3214{
3215 struct sock *sk = sock->sk;
3216 int addr_len = 0;
3217 int err;
3218
1b784140 3219 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
1da177e4
LT		 3220 flags & ~MSG_DONTWAIT, &addr_len);
3221 if (err >= 0)
3222 msg->msg_namelen = addr_len;
3223 return err;
3224}
1da177e4
LT		 3225EXPORT_SYMBOL(sock_common_recvmsg);
3226
3227/*
3228 * Set socket options on an inet socket.
3229 */
3230int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 3231 char __user *optval, unsigned int optlen)
1da177e4
LT		 3232{
3233 struct sock *sk = sock->sk;
3234
3235 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
3236}
1da177e4
LT		 3237EXPORT_SYMBOL(sock_common_setsockopt);
3238
3fdadf7d 3239#ifdef CONFIG_COMPAT
543d9cfe 3240int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 3241 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 3242{
3243 struct sock *sk = sock->sk;
3244
543d9cfe
ACM		 3245 if (sk->sk_prot->compat_setsockopt != NULL)
3246 return sk->sk_prot->compat_setsockopt(sk, level, optname,
3247 optval, optlen);
3fdadf7d
DM		 3248 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
3249}
3250EXPORT_SYMBOL(compat_sock_common_setsockopt);
3251#endif
3252
1da177e4
LT		 3253void sk_common_release(struct sock *sk)
3254{
3255 if (sk->sk_prot->destroy)
3256 sk->sk_prot->destroy(sk);
3257
3258 /*
3259 * Observation: when sock_common_release is called, processes have
3260 * no access to socket. But net still has.
3261 * Step one, detach it from networking:
3262 *
3263 * A. Remove from hash tables.
3264 */
3265
3266 sk->sk_prot->unhash(sk);
3267
3268 /*
3269 * In this point socket cannot receive new packets, but it is possible
3270 * that some packets are in flight because some CPU runs receiver and
3271 * did hash table lookup before we unhashed socket. They will achieve
3272 * receive queue and will be purged by socket destructor.
3273 *
3274 * Also we still have packets pending on receive queue and probably,
3275 * our own packets waiting in device queues. sock_destroy will drain
3276 * receive queue, but transmitted packets will delay socket destruction
3277 * until the last reference will be released.
3278 */
3279
3280 sock_orphan(sk);
3281
3282 xfrm_sk_free_policy(sk);
3283
e6848976 3284 sk_refcnt_debug_release(sk);
5640f768 3285
1da177e4
LT		 3286 sock_put(sk);
3287}
1da177e4
LT		 3288EXPORT_SYMBOL(sk_common_release);
3289
a2d133b1
JH		 3290void sk_get_meminfo(const struct sock *sk, u32 *mem)
3291{
3292 memset(mem, 0, sizeof(*mem) * SK_MEMINFO_VARS);
3293
3294 mem[SK_MEMINFO_RMEM_ALLOC] = sk_rmem_alloc_get(sk);
ebb3b78d 3295 mem[SK_MEMINFO_RCVBUF] = READ_ONCE(sk->sk_rcvbuf);
a2d133b1 3296 mem[SK_MEMINFO_WMEM_ALLOC] = sk_wmem_alloc_get(sk);
e292f05e 3297 mem[SK_MEMINFO_SNDBUF] = READ_ONCE(sk->sk_sndbuf);
a2d133b1 3298 mem[SK_MEMINFO_FWD_ALLOC] = sk->sk_forward_alloc;
ab4e846a 3299 mem[SK_MEMINFO_WMEM_QUEUED] = READ_ONCE(sk->sk_wmem_queued);
a2d133b1 3300 mem[SK_MEMINFO_OPTMEM] = atomic_read(&sk->sk_omem_alloc);
70c26558 3301 mem[SK_MEMINFO_BACKLOG] = READ_ONCE(sk->sk_backlog.len);
a2d133b1
JH		 3302 mem[SK_MEMINFO_DROPS] = atomic_read(&sk->sk_drops);
3303}
3304
13ff3d6f
PE		 3305#ifdef CONFIG_PROC_FS
3306#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 3307struct prot_inuse {
3308 int val[PROTO_INUSE_NR];
3309};
13ff3d6f
PE		 3310
3311static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159 3312
70ee1159
PE		 3313void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
3314{
08fc7f81 3315 __this_cpu_add(net->core.prot_inuse->val[prot->inuse_idx], val);
70ee1159
PE		 3316}
3317EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
3318
3319int sock_prot_inuse_get(struct net *net, struct proto *prot)
3320{
3321 int cpu, idx = prot->inuse_idx;
3322 int res = 0;
3323
3324 for_each_possible_cpu(cpu)
08fc7f81 3325 res += per_cpu_ptr(net->core.prot_inuse, cpu)->val[idx];
70ee1159
PE		 3326
3327 return res >= 0 ? res : 0;
3328}
3329EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
3330
648845ab
TZ		 3331static void sock_inuse_add(struct net *net, int val)
3332{
3333 this_cpu_add(*net->core.sock_inuse, val);
3334}
3335
3336int sock_inuse_get(struct net *net)
3337{
3338 int cpu, res = 0;
3339
3340 for_each_possible_cpu(cpu)
3341 res += *per_cpu_ptr(net->core.sock_inuse, cpu);
3342
3343 return res;
3344}
3345
3346EXPORT_SYMBOL_GPL(sock_inuse_get);
3347
2c8c1e72 3348static int __net_init sock_inuse_init_net(struct net *net)
70ee1159 3349{
08fc7f81 3350 net->core.prot_inuse = alloc_percpu(struct prot_inuse);
648845ab
TZ		 3351 if (net->core.prot_inuse == NULL)
3352 return -ENOMEM;
3353
3354 net->core.sock_inuse = alloc_percpu(int);
3355 if (net->core.sock_inuse == NULL)
3356 goto out;
3357
3358 return 0;
3359
3360out:
3361 free_percpu(net->core.prot_inuse);
3362 return -ENOMEM;
70ee1159
PE		 3363}
3364
2c8c1e72 3365static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159 3366{
08fc7f81 3367 free_percpu(net->core.prot_inuse);
648845ab 3368 free_percpu(net->core.sock_inuse);
70ee1159
PE		 3369}
3370
3371static struct pernet_operations net_inuse_ops = {
3372 .init = sock_inuse_init_net,
3373 .exit = sock_inuse_exit_net,
3374};
3375
3376static __init int net_inuse_init(void)
3377{
3378 if (register_pernet_subsys(&net_inuse_ops))
3379 panic("Cannot initialize net inuse counters");
3380
3381 return 0;
3382}
3383
3384core_initcall(net_inuse_init);
13ff3d6f 3385
b45ce321 3386static int assign_proto_idx(struct proto *prot)
13ff3d6f
PE		 3387{
3388 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
3389
3390 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 3391 pr_err("PROTO_INUSE_NR exhausted\n");
b45ce321 3392 return -ENOSPC;
13ff3d6f
PE		 3393 }
3394
3395 set_bit(prot->inuse_idx, proto_inuse_idx);
b45ce321 3396 return 0;
13ff3d6f
PE		 3397}
3398
3399static void release_proto_idx(struct proto *prot)
3400{
3401 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
3402 clear_bit(prot->inuse_idx, proto_inuse_idx);
3403}
3404#else
b45ce321 3405static inline int assign_proto_idx(struct proto *prot)
13ff3d6f 3406{
b45ce321 3407 return 0;
13ff3d6f
PE		 3408}
3409
3410static inline void release_proto_idx(struct proto *prot)
3411{
3412}
648845ab
TZ		 3413
3414static void sock_inuse_add(struct net *net, int val)
3415{
3416}
13ff3d6f
PE		 3417#endif
3418
0159dfd3
ED		 3419static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
3420{
3421 if (!rsk_prot)
3422 return;
3423 kfree(rsk_prot->slab_name);
3424 rsk_prot->slab_name = NULL;
adf78eda
JL		 3425 kmem_cache_destroy(rsk_prot->slab);
3426 rsk_prot->slab = NULL;
0159dfd3
ED		 3427}
3428
3429static int req_prot_init(const struct proto *prot)
3430{
3431 struct request_sock_ops *rsk_prot = prot->rsk_prot;
3432
3433 if (!rsk_prot)
3434 return 0;
3435
3436 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
3437 prot->name);
3438 if (!rsk_prot->slab_name)
3439 return -ENOMEM;
3440
3441 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
3442 rsk_prot->obj_size, 0,
e699e2c6
SB		 3443 SLAB_ACCOUNT | prot->slab_flags,
3444 NULL);
0159dfd3
ED		 3445
3446 if (!rsk_prot->slab) {
3447 pr_crit("%s: Can't create request sock SLAB cache!\n",
3448 prot->name);
3449 return -ENOMEM;
3450 }
3451 return 0;
3452}
3453
b733c007
PE		 3454int proto_register(struct proto *prot, int alloc_slab)
3455{
b45ce321 3456 int ret = -ENOBUFS;
3457
1da177e4 3458 if (alloc_slab) {
30c2c9f1
DW		 3459 prot->slab = kmem_cache_create_usercopy(prot->name,
3460 prot->obj_size, 0,
e699e2c6
SB		 3461 SLAB_HWCACHE_ALIGN | SLAB_ACCOUNT |
3462 prot->slab_flags,
289a4860 3463 prot->useroffset, prot->usersize,
271b72c7 3464 NULL);
1da177e4
LT		 3465
3466 if (prot->slab == NULL) {
e005d193
JP		 3467 pr_crit("%s: Can't create sock SLAB cache!\n",
3468 prot->name);
60e7663d 3469 goto out;
1da177e4 3470 }
2e6599cb 3471
0159dfd3
ED		 3472 if (req_prot_init(prot))
3473 goto out_free_request_sock_slab;
8feaf0c0 3474
6d6ee43e 3475 if (prot->twsk_prot != NULL) {
faf23422 3476 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 3477
7e56b5d6 3478 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 3479 goto out_free_request_sock_slab;
3480
6d6ee43e 3481 prot->twsk_prot->twsk_slab =
7e56b5d6 3482 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 3483 prot->twsk_prot->twsk_obj_size,
3ab5aee7 3484 0,
e699e2c6 3485 SLAB_ACCOUNT |
52db70dc 3486 prot->slab_flags,
20c2df83 3487 NULL);
6d6ee43e 3488 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 3489 goto out_free_timewait_sock_slab_name;
3490 }
1da177e4
LT		 3491 }
3492
36b77a52 3493 mutex_lock(&proto_list_mutex);
b45ce321 3494 ret = assign_proto_idx(prot);
3495 if (ret) {
3496 mutex_unlock(&proto_list_mutex);
3497 goto out_free_timewait_sock_slab_name;
3498 }
1da177e4 3499 list_add(&prot->node, &proto_list);
36b77a52 3500 mutex_unlock(&proto_list_mutex);
b45ce321 3501 return ret;
b733c007 3502
8feaf0c0 3503out_free_timewait_sock_slab_name:
b45ce321 3504 if (alloc_slab && prot->twsk_prot)
3505 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0 3506out_free_request_sock_slab:
b45ce321 3507 if (alloc_slab) {
3508 req_prot_cleanup(prot->rsk_prot);
0159dfd3 3509
b45ce321 3510 kmem_cache_destroy(prot->slab);
3511 prot->slab = NULL;
3512 }
b733c007 3513out:
b45ce321 3514 return ret;
1da177e4 3515}
1da177e4
LT		 3516EXPORT_SYMBOL(proto_register);
3517
3518void proto_unregister(struct proto *prot)
3519{
36b77a52 3520 mutex_lock(&proto_list_mutex);
13ff3d6f 3521 release_proto_idx(prot);
0a3f4358 3522 list_del(&prot->node);
36b77a52 3523 mutex_unlock(&proto_list_mutex);
1da177e4 3524
adf78eda
JL		 3525 kmem_cache_destroy(prot->slab);
3526 prot->slab = NULL;
1da177e4 3527
0159dfd3 3528 req_prot_cleanup(prot->rsk_prot);
2e6599cb 3529
6d6ee43e 3530 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 3531 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 3532 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 3533 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 3534 }
1da177e4 3535}
1da177e4
LT		 3536EXPORT_SYMBOL(proto_unregister);
3537
bf2ae2e4
XL		 3538int sock_load_diag_module(int family, int protocol)
3539{
3540 if (!protocol) {
3541 if (!sock_is_registered(family))
3542 return -ENOENT;
3543
3544 return request_module("net-pf-%d-proto-%d-type-%d", PF_NETLINK,
3545 NETLINK_SOCK_DIAG, family);
3546 }
3547
3548#ifdef CONFIG_INET
3549 if (family == AF_INET &&
c34c1287 3550 protocol != IPPROTO_RAW &&
bf2ae2e4
XL		 3551 !rcu_access_pointer(inet_protos[protocol]))
3552 return -ENOENT;
3553#endif
3554
3555 return request_module("net-pf-%d-proto-%d-type-%d-%d", PF_NETLINK,
3556 NETLINK_SOCK_DIAG, family, protocol);
3557}
3558EXPORT_SYMBOL(sock_load_diag_module);
3559
1da177e4 3560#ifdef CONFIG_PROC_FS
1da177e4 3561static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 3562 __acquires(proto_list_mutex)
1da177e4 3563{
36b77a52 3564 mutex_lock(&proto_list_mutex);
60f0438a 3565 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 3566}
3567
3568static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3569{
60f0438a 3570 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 3571}
3572
3573static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 3574 __releases(proto_list_mutex)
1da177e4 3575{
36b77a52 3576 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 3577}
3578
3579static char proto_method_implemented(const void *method)
3580{
3581 return method == NULL ? 'n' : 'y';
3582}
180d8cd9
GC		 3583static long sock_prot_memory_allocated(struct proto *proto)
3584{
cb75a36c 3585 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 3586}
3587
7a512eb8 3588static const char *sock_prot_memory_pressure(struct proto *proto)
180d8cd9
GC		 3589{
3590 return proto->memory_pressure != NULL ?
3591 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
3592}
1da177e4
LT		 3593
3594static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
3595{
180d8cd9 3596
8d987e5c 3597 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 3598 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
3599 proto->name,
3600 proto->obj_size,
14e943db 3601 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 3602 sock_prot_memory_allocated(proto),
3603 sock_prot_memory_pressure(proto),
1da177e4
LT		 3604 proto->max_header,
3605 proto->slab == NULL ? "no" : "yes",
3606 module_name(proto->owner),
3607 proto_method_implemented(proto->close),
3608 proto_method_implemented(proto->connect),
3609 proto_method_implemented(proto->disconnect),
3610 proto_method_implemented(proto->accept),
3611 proto_method_implemented(proto->ioctl),
3612 proto_method_implemented(proto->init),
3613 proto_method_implemented(proto->destroy),
3614 proto_method_implemented(proto->shutdown),
3615 proto_method_implemented(proto->setsockopt),
3616 proto_method_implemented(proto->getsockopt),
3617 proto_method_implemented(proto->sendmsg),
3618 proto_method_implemented(proto->recvmsg),
3619 proto_method_implemented(proto->sendpage),
3620 proto_method_implemented(proto->bind),
3621 proto_method_implemented(proto->backlog_rcv),
3622 proto_method_implemented(proto->hash),
3623 proto_method_implemented(proto->unhash),
3624 proto_method_implemented(proto->get_port),
3625 proto_method_implemented(proto->enter_memory_pressure));
3626}
3627
3628static int proto_seq_show(struct seq_file *seq, void *v)
3629{
60f0438a 3630 if (v == &proto_list)
1da177e4
LT		 3631 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3632 "protocol",
3633 "size",
3634 "sockets",
3635 "memory",
3636 "press",
3637 "maxhdr",
3638 "slab",
3639 "module",
3640 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3641 else
60f0438a 3642 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 3643 return 0;
3644}
3645
f690808e 3646static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 3647 .start = proto_seq_start,
3648 .next = proto_seq_next,
3649 .stop = proto_seq_stop,
3650 .show = proto_seq_show,
3651};
3652
14e943db
ED		 3653static __net_init int proto_init_net(struct net *net)
3654{
c3506372
CH		 3655 if (!proc_create_net("protocols", 0444, net->proc_net, &proto_seq_ops,
3656 sizeof(struct seq_net_private)))
14e943db
ED		 3657 return -ENOMEM;
3658
3659 return 0;
3660}
3661
3662static __net_exit void proto_exit_net(struct net *net)
3663{
ece31ffd 3664 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 3665}
3666
3667
3668static __net_initdata struct pernet_operations proto_net_ops = {
3669 .init = proto_init_net,
3670 .exit = proto_exit_net,
1da177e4
LT		 3671};
3672
3673static int __init proto_init(void)
3674{
14e943db 3675 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 3676}
3677
3678subsys_initcall(proto_init);
3679
3680#endif /* PROC_FS */
7db6b048
SS		 3681
3682#ifdef CONFIG_NET_RX_BUSY_POLL
3683bool sk_busy_loop_end(void *p, unsigned long start_time)
3684{
3685 struct sock *sk = p;
3686
3f926af3 3687 return !skb_queue_empty_lockless(&sk->sk_receive_queue) ||
7db6b048
SS		 3688 sk_busy_loop_timeout(sk, start_time);
3689}
3690EXPORT_SYMBOL(sk_busy_loop_end);
3691#endif /* CONFIG_NET_RX_BUSY_POLL */
Linux 6.x block layer and io_uring tree(s)