2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * Generic socket support routines. Memory allocators, socket lock/release
7 * handler for protocols to use and generic option handler.
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>
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
35 * code. The ACK stuff can wait and needs major
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
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
86 * This program is free software; you can redistribute it and/or
87 * modify it under the terms of the GNU General Public License
88 * as published by the Free Software Foundation; either version
89 * 2 of the License, or (at your option) any later version.
92 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
94 #include <linux/capability.h>
95 #include <linux/errno.h>
96 #include <linux/errqueue.h>
97 #include <linux/types.h>
98 #include <linux/socket.h>
100 #include <linux/kernel.h>
101 #include <linux/module.h>
102 #include <linux/proc_fs.h>
103 #include <linux/seq_file.h>
104 #include <linux/sched.h>
105 #include <linux/sched/mm.h>
106 #include <linux/timer.h>
107 #include <linux/string.h>
108 #include <linux/sockios.h>
109 #include <linux/net.h>
110 #include <linux/mm.h>
111 #include <linux/slab.h>
112 #include <linux/interrupt.h>
113 #include <linux/poll.h>
114 #include <linux/tcp.h>
115 #include <linux/init.h>
116 #include <linux/highmem.h>
117 #include <linux/user_namespace.h>
118 #include <linux/static_key.h>
119 #include <linux/memcontrol.h>
120 #include <linux/prefetch.h>
122 #include <linux/uaccess.h>
124 #include <linux/netdevice.h>
125 #include <net/protocol.h>
126 #include <linux/skbuff.h>
127 #include <net/net_namespace.h>
128 #include <net/request_sock.h>
129 #include <net/sock.h>
130 #include <linux/net_tstamp.h>
131 #include <net/xfrm.h>
132 #include <linux/ipsec.h>
133 #include <net/cls_cgroup.h>
134 #include <net/netprio_cgroup.h>
135 #include <linux/sock_diag.h>
137 #include <linux/filter.h>
138 #include <net/sock_reuseport.h>
140 #include <trace/events/sock.h>
143 #include <net/busy_poll.h>
145 static DEFINE_MUTEX(proto_list_mutex);
146 static LIST_HEAD(proto_list);
149 * sk_ns_capable - General socket capability test
150 * @sk: Socket to use a capability on or through
151 * @user_ns: The user namespace of the capability to use
152 * @cap: The capability to use
154 * Test to see if the opener of the socket had when the socket was
155 * created and the current process has the capability @cap in the user
156 * namespace @user_ns.
158 bool sk_ns_capable(const struct sock *sk,
159 struct user_namespace *user_ns, int cap)
161 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
162 ns_capable(user_ns, cap);
164 EXPORT_SYMBOL(sk_ns_capable);
167 * sk_capable - Socket global capability test
168 * @sk: Socket to use a capability on or through
169 * @cap: The global capability to use
171 * Test to see if the opener of the socket had when the socket was
172 * created and the current process has the capability @cap in all user
175 bool sk_capable(const struct sock *sk, int cap)
177 return sk_ns_capable(sk, &init_user_ns, cap);
179 EXPORT_SYMBOL(sk_capable);
182 * sk_net_capable - Network namespace socket capability test
183 * @sk: Socket to use a capability on or through
184 * @cap: The capability to use
186 * Test to see if the opener of the socket had when the socket was created
187 * and the current process has the capability @cap over the network namespace
188 * the socket is a member of.
190 bool sk_net_capable(const struct sock *sk, int cap)
192 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
194 EXPORT_SYMBOL(sk_net_capable);
197 * Each address family might have different locking rules, so we have
198 * one slock key per address family and separate keys for internal and
201 static struct lock_class_key af_family_keys[AF_MAX];
202 static struct lock_class_key af_family_kern_keys[AF_MAX];
203 static struct lock_class_key af_family_slock_keys[AF_MAX];
204 static struct lock_class_key af_family_kern_slock_keys[AF_MAX];
207 * Make lock validator output more readable. (we pre-construct these
208 * strings build-time, so that runtime initialization of socket
212 #define _sock_locks(x) \
213 x "AF_UNSPEC", x "AF_UNIX" , x "AF_INET" , \
214 x "AF_AX25" , x "AF_IPX" , x "AF_APPLETALK", \
215 x "AF_NETROM", x "AF_BRIDGE" , x "AF_ATMPVC" , \
216 x "AF_X25" , x "AF_INET6" , x "AF_ROSE" , \
217 x "AF_DECnet", x "AF_NETBEUI" , x "AF_SECURITY" , \
218 x "AF_KEY" , x "AF_NETLINK" , x "AF_PACKET" , \
219 x "AF_ASH" , x "AF_ECONET" , x "AF_ATMSVC" , \
220 x "AF_RDS" , x "AF_SNA" , x "AF_IRDA" , \
221 x "AF_PPPOX" , x "AF_WANPIPE" , x "AF_LLC" , \
222 x "27" , x "28" , x "AF_CAN" , \
223 x "AF_TIPC" , x "AF_BLUETOOTH", x "IUCV" , \
224 x "AF_RXRPC" , x "AF_ISDN" , x "AF_PHONET" , \
225 x "AF_IEEE802154", x "AF_CAIF" , x "AF_ALG" , \
226 x "AF_NFC" , x "AF_VSOCK" , x "AF_KCM" , \
227 x "AF_QIPCRTR", x "AF_SMC" , x "AF_MAX"
229 static const char *const af_family_key_strings[AF_MAX+1] = {
230 _sock_locks("sk_lock-")
232 static const char *const af_family_slock_key_strings[AF_MAX+1] = {
233 _sock_locks("slock-")
235 static const char *const af_family_clock_key_strings[AF_MAX+1] = {
236 _sock_locks("clock-")
239 static const char *const af_family_kern_key_strings[AF_MAX+1] = {
240 _sock_locks("k-sk_lock-")
242 static const char *const af_family_kern_slock_key_strings[AF_MAX+1] = {
243 _sock_locks("k-slock-")
245 static const char *const af_family_kern_clock_key_strings[AF_MAX+1] = {
246 _sock_locks("k-clock-")
248 static const char *const af_family_rlock_key_strings[AF_MAX+1] = {
249 "rlock-AF_UNSPEC", "rlock-AF_UNIX" , "rlock-AF_INET" ,
250 "rlock-AF_AX25" , "rlock-AF_IPX" , "rlock-AF_APPLETALK",
251 "rlock-AF_NETROM", "rlock-AF_BRIDGE" , "rlock-AF_ATMPVC" ,
252 "rlock-AF_X25" , "rlock-AF_INET6" , "rlock-AF_ROSE" ,
253 "rlock-AF_DECnet", "rlock-AF_NETBEUI" , "rlock-AF_SECURITY" ,
254 "rlock-AF_KEY" , "rlock-AF_NETLINK" , "rlock-AF_PACKET" ,
255 "rlock-AF_ASH" , "rlock-AF_ECONET" , "rlock-AF_ATMSVC" ,
256 "rlock-AF_RDS" , "rlock-AF_SNA" , "rlock-AF_IRDA" ,
257 "rlock-AF_PPPOX" , "rlock-AF_WANPIPE" , "rlock-AF_LLC" ,
258 "rlock-27" , "rlock-28" , "rlock-AF_CAN" ,
259 "rlock-AF_TIPC" , "rlock-AF_BLUETOOTH", "rlock-AF_IUCV" ,
260 "rlock-AF_RXRPC" , "rlock-AF_ISDN" , "rlock-AF_PHONET" ,
261 "rlock-AF_IEEE802154", "rlock-AF_CAIF" , "rlock-AF_ALG" ,
262 "rlock-AF_NFC" , "rlock-AF_VSOCK" , "rlock-AF_KCM" ,
263 "rlock-AF_QIPCRTR", "rlock-AF_SMC" , "rlock-AF_MAX"
265 static const char *const af_family_wlock_key_strings[AF_MAX+1] = {
266 "wlock-AF_UNSPEC", "wlock-AF_UNIX" , "wlock-AF_INET" ,
267 "wlock-AF_AX25" , "wlock-AF_IPX" , "wlock-AF_APPLETALK",
268 "wlock-AF_NETROM", "wlock-AF_BRIDGE" , "wlock-AF_ATMPVC" ,
269 "wlock-AF_X25" , "wlock-AF_INET6" , "wlock-AF_ROSE" ,
270 "wlock-AF_DECnet", "wlock-AF_NETBEUI" , "wlock-AF_SECURITY" ,
271 "wlock-AF_KEY" , "wlock-AF_NETLINK" , "wlock-AF_PACKET" ,
272 "wlock-AF_ASH" , "wlock-AF_ECONET" , "wlock-AF_ATMSVC" ,
273 "wlock-AF_RDS" , "wlock-AF_SNA" , "wlock-AF_IRDA" ,
274 "wlock-AF_PPPOX" , "wlock-AF_WANPIPE" , "wlock-AF_LLC" ,
275 "wlock-27" , "wlock-28" , "wlock-AF_CAN" ,
276 "wlock-AF_TIPC" , "wlock-AF_BLUETOOTH", "wlock-AF_IUCV" ,
277 "wlock-AF_RXRPC" , "wlock-AF_ISDN" , "wlock-AF_PHONET" ,
278 "wlock-AF_IEEE802154", "wlock-AF_CAIF" , "wlock-AF_ALG" ,
279 "wlock-AF_NFC" , "wlock-AF_VSOCK" , "wlock-AF_KCM" ,
280 "wlock-AF_QIPCRTR", "wlock-AF_SMC" , "wlock-AF_MAX"
282 static const char *const af_family_elock_key_strings[AF_MAX+1] = {
283 "elock-AF_UNSPEC", "elock-AF_UNIX" , "elock-AF_INET" ,
284 "elock-AF_AX25" , "elock-AF_IPX" , "elock-AF_APPLETALK",
285 "elock-AF_NETROM", "elock-AF_BRIDGE" , "elock-AF_ATMPVC" ,
286 "elock-AF_X25" , "elock-AF_INET6" , "elock-AF_ROSE" ,
287 "elock-AF_DECnet", "elock-AF_NETBEUI" , "elock-AF_SECURITY" ,
288 "elock-AF_KEY" , "elock-AF_NETLINK" , "elock-AF_PACKET" ,
289 "elock-AF_ASH" , "elock-AF_ECONET" , "elock-AF_ATMSVC" ,
290 "elock-AF_RDS" , "elock-AF_SNA" , "elock-AF_IRDA" ,
291 "elock-AF_PPPOX" , "elock-AF_WANPIPE" , "elock-AF_LLC" ,
292 "elock-27" , "elock-28" , "elock-AF_CAN" ,
293 "elock-AF_TIPC" , "elock-AF_BLUETOOTH", "elock-AF_IUCV" ,
294 "elock-AF_RXRPC" , "elock-AF_ISDN" , "elock-AF_PHONET" ,
295 "elock-AF_IEEE802154", "elock-AF_CAIF" , "elock-AF_ALG" ,
296 "elock-AF_NFC" , "elock-AF_VSOCK" , "elock-AF_KCM" ,
297 "elock-AF_QIPCRTR", "elock-AF_SMC" , "elock-AF_MAX"
301 * sk_callback_lock and sk queues locking rules are per-address-family,
302 * so split the lock classes by using a per-AF key:
304 static struct lock_class_key af_callback_keys[AF_MAX];
305 static struct lock_class_key af_rlock_keys[AF_MAX];
306 static struct lock_class_key af_wlock_keys[AF_MAX];
307 static struct lock_class_key af_elock_keys[AF_MAX];
308 static struct lock_class_key af_kern_callback_keys[AF_MAX];
310 /* Run time adjustable parameters. */
311 __u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
312 EXPORT_SYMBOL(sysctl_wmem_max);
313 __u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
314 EXPORT_SYMBOL(sysctl_rmem_max);
315 __u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
316 __u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
318 /* Maximal space eaten by iovec or ancillary data plus some space */
319 int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
320 EXPORT_SYMBOL(sysctl_optmem_max);
322 int sysctl_tstamp_allow_data __read_mostly = 1;
324 struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
325 EXPORT_SYMBOL_GPL(memalloc_socks);
328 * sk_set_memalloc - sets %SOCK_MEMALLOC
329 * @sk: socket to set it on
331 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
332 * It's the responsibility of the admin to adjust min_free_kbytes
333 * to meet the requirements
335 void sk_set_memalloc(struct sock *sk)
337 sock_set_flag(sk, SOCK_MEMALLOC);
338 sk->sk_allocation |= __GFP_MEMALLOC;
339 static_key_slow_inc(&memalloc_socks);
341 EXPORT_SYMBOL_GPL(sk_set_memalloc);
343 void sk_clear_memalloc(struct sock *sk)
345 sock_reset_flag(sk, SOCK_MEMALLOC);
346 sk->sk_allocation &= ~__GFP_MEMALLOC;
347 static_key_slow_dec(&memalloc_socks);
350 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
351 * progress of swapping. SOCK_MEMALLOC may be cleared while
352 * it has rmem allocations due to the last swapfile being deactivated
353 * but there is a risk that the socket is unusable due to exceeding
354 * the rmem limits. Reclaim the reserves and obey rmem limits again.
358 EXPORT_SYMBOL_GPL(sk_clear_memalloc);
360 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
363 unsigned int noreclaim_flag;
365 /* these should have been dropped before queueing */
366 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
368 noreclaim_flag = memalloc_noreclaim_save();
369 ret = sk->sk_backlog_rcv(sk, skb);
370 memalloc_noreclaim_restore(noreclaim_flag);
374 EXPORT_SYMBOL(__sk_backlog_rcv);
376 static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
380 if (optlen < sizeof(tv))
382 if (copy_from_user(&tv, optval, sizeof(tv)))
384 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
388 static int warned __read_mostly;
391 if (warned < 10 && net_ratelimit()) {
393 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
394 __func__, current->comm, task_pid_nr(current));
398 *timeo_p = MAX_SCHEDULE_TIMEOUT;
399 if (tv.tv_sec == 0 && tv.tv_usec == 0)
401 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
402 *timeo_p = tv.tv_sec * HZ + DIV_ROUND_UP(tv.tv_usec, USEC_PER_SEC / HZ);
406 static void sock_warn_obsolete_bsdism(const char *name)
409 static char warncomm[TASK_COMM_LEN];
410 if (strcmp(warncomm, current->comm) && warned < 5) {
411 strcpy(warncomm, current->comm);
412 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
418 static bool sock_needs_netstamp(const struct sock *sk)
420 switch (sk->sk_family) {
429 static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
431 if (sk->sk_flags & flags) {
432 sk->sk_flags &= ~flags;
433 if (sock_needs_netstamp(sk) &&
434 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
435 net_disable_timestamp();
440 int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
443 struct sk_buff_head *list = &sk->sk_receive_queue;
445 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
446 atomic_inc(&sk->sk_drops);
447 trace_sock_rcvqueue_full(sk, skb);
451 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
452 atomic_inc(&sk->sk_drops);
457 skb_set_owner_r(skb, sk);
459 /* we escape from rcu protected region, make sure we dont leak
464 spin_lock_irqsave(&list->lock, flags);
465 sock_skb_set_dropcount(sk, skb);
466 __skb_queue_tail(list, skb);
467 spin_unlock_irqrestore(&list->lock, flags);
469 if (!sock_flag(sk, SOCK_DEAD))
470 sk->sk_data_ready(sk);
473 EXPORT_SYMBOL(__sock_queue_rcv_skb);
475 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
479 err = sk_filter(sk, skb);
483 return __sock_queue_rcv_skb(sk, skb);
485 EXPORT_SYMBOL(sock_queue_rcv_skb);
487 int __sk_receive_skb(struct sock *sk, struct sk_buff *skb,
488 const int nested, unsigned int trim_cap, bool refcounted)
490 int rc = NET_RX_SUCCESS;
492 if (sk_filter_trim_cap(sk, skb, trim_cap))
493 goto discard_and_relse;
497 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
498 atomic_inc(&sk->sk_drops);
499 goto discard_and_relse;
502 bh_lock_sock_nested(sk);
505 if (!sock_owned_by_user(sk)) {
507 * trylock + unlock semantics:
509 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
511 rc = sk_backlog_rcv(sk, skb);
513 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
514 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
516 atomic_inc(&sk->sk_drops);
517 goto discard_and_relse;
529 EXPORT_SYMBOL(__sk_receive_skb);
531 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
533 struct dst_entry *dst = __sk_dst_get(sk);
535 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
536 sk_tx_queue_clear(sk);
537 sk->sk_dst_pending_confirm = 0;
538 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
545 EXPORT_SYMBOL(__sk_dst_check);
547 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
549 struct dst_entry *dst = sk_dst_get(sk);
551 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
559 EXPORT_SYMBOL(sk_dst_check);
561 static int sock_setbindtodevice(struct sock *sk, char __user *optval,
564 int ret = -ENOPROTOOPT;
565 #ifdef CONFIG_NETDEVICES
566 struct net *net = sock_net(sk);
567 char devname[IFNAMSIZ];
572 if (!ns_capable(net->user_ns, CAP_NET_RAW))
579 /* Bind this socket to a particular device like "eth0",
580 * as specified in the passed interface name. If the
581 * name is "" or the option length is zero the socket
584 if (optlen > IFNAMSIZ - 1)
585 optlen = IFNAMSIZ - 1;
586 memset(devname, 0, sizeof(devname));
589 if (copy_from_user(devname, optval, optlen))
593 if (devname[0] != '\0') {
594 struct net_device *dev;
597 dev = dev_get_by_name_rcu(net, devname);
599 index = dev->ifindex;
607 sk->sk_bound_dev_if = index;
619 static int sock_getbindtodevice(struct sock *sk, char __user *optval,
620 int __user *optlen, int len)
622 int ret = -ENOPROTOOPT;
623 #ifdef CONFIG_NETDEVICES
624 struct net *net = sock_net(sk);
625 char devname[IFNAMSIZ];
627 if (sk->sk_bound_dev_if == 0) {
636 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
640 len = strlen(devname) + 1;
643 if (copy_to_user(optval, devname, len))
648 if (put_user(len, optlen))
659 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
662 sock_set_flag(sk, bit);
664 sock_reset_flag(sk, bit);
667 bool sk_mc_loop(struct sock *sk)
669 if (dev_recursion_level())
673 switch (sk->sk_family) {
675 return inet_sk(sk)->mc_loop;
676 #if IS_ENABLED(CONFIG_IPV6)
678 return inet6_sk(sk)->mc_loop;
684 EXPORT_SYMBOL(sk_mc_loop);
687 * This is meant for all protocols to use and covers goings on
688 * at the socket level. Everything here is generic.
691 int sock_setsockopt(struct socket *sock, int level, int optname,
692 char __user *optval, unsigned int optlen)
694 struct sock *sk = sock->sk;
701 * Options without arguments
704 if (optname == SO_BINDTODEVICE)
705 return sock_setbindtodevice(sk, optval, optlen);
707 if (optlen < sizeof(int))
710 if (get_user(val, (int __user *)optval))
713 valbool = val ? 1 : 0;
719 if (val && !capable(CAP_NET_ADMIN))
722 sock_valbool_flag(sk, SOCK_DBG, valbool);
725 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
728 sk->sk_reuseport = valbool;
737 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
741 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
744 /* Don't error on this BSD doesn't and if you think
745 * about it this is right. Otherwise apps have to
746 * play 'guess the biggest size' games. RCVBUF/SNDBUF
747 * are treated in BSD as hints
749 val = min_t(u32, val, sysctl_wmem_max);
751 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
752 sk->sk_sndbuf = max_t(int, val * 2, SOCK_MIN_SNDBUF);
753 /* Wake up sending tasks if we upped the value. */
754 sk->sk_write_space(sk);
758 if (!capable(CAP_NET_ADMIN)) {
765 /* Don't error on this BSD doesn't and if you think
766 * about it this is right. Otherwise apps have to
767 * play 'guess the biggest size' games. RCVBUF/SNDBUF
768 * are treated in BSD as hints
770 val = min_t(u32, val, sysctl_rmem_max);
772 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
774 * We double it on the way in to account for
775 * "struct sk_buff" etc. overhead. Applications
776 * assume that the SO_RCVBUF setting they make will
777 * allow that much actual data to be received on that
780 * Applications are unaware that "struct sk_buff" and
781 * other overheads allocate from the receive buffer
782 * during socket buffer allocation.
784 * And after considering the possible alternatives,
785 * returning the value we actually used in getsockopt
786 * is the most desirable behavior.
788 sk->sk_rcvbuf = max_t(int, val * 2, SOCK_MIN_RCVBUF);
792 if (!capable(CAP_NET_ADMIN)) {
799 if (sk->sk_prot->keepalive)
800 sk->sk_prot->keepalive(sk, valbool);
801 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
805 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
809 sk->sk_no_check_tx = valbool;
813 if ((val >= 0 && val <= 6) ||
814 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
815 sk->sk_priority = val;
821 if (optlen < sizeof(ling)) {
822 ret = -EINVAL; /* 1003.1g */
825 if (copy_from_user(&ling, optval, sizeof(ling))) {
830 sock_reset_flag(sk, SOCK_LINGER);
832 #if (BITS_PER_LONG == 32)
833 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
834 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
837 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
838 sock_set_flag(sk, SOCK_LINGER);
843 sock_warn_obsolete_bsdism("setsockopt");
848 set_bit(SOCK_PASSCRED, &sock->flags);
850 clear_bit(SOCK_PASSCRED, &sock->flags);
856 if (optname == SO_TIMESTAMP)
857 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
859 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
860 sock_set_flag(sk, SOCK_RCVTSTAMP);
861 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
863 sock_reset_flag(sk, SOCK_RCVTSTAMP);
864 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
868 case SO_TIMESTAMPING:
869 if (val & ~SOF_TIMESTAMPING_MASK) {
874 if (val & SOF_TIMESTAMPING_OPT_ID &&
875 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
876 if (sk->sk_protocol == IPPROTO_TCP &&
877 sk->sk_type == SOCK_STREAM) {
878 if ((1 << sk->sk_state) &
879 (TCPF_CLOSE | TCPF_LISTEN)) {
883 sk->sk_tskey = tcp_sk(sk)->snd_una;
889 if (val & SOF_TIMESTAMPING_OPT_STATS &&
890 !(val & SOF_TIMESTAMPING_OPT_TSONLY)) {
895 sk->sk_tsflags = val;
896 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
897 sock_enable_timestamp(sk,
898 SOCK_TIMESTAMPING_RX_SOFTWARE);
900 sock_disable_timestamp(sk,
901 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
907 sk->sk_rcvlowat = val ? : 1;
911 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
915 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
918 case SO_ATTACH_FILTER:
920 if (optlen == sizeof(struct sock_fprog)) {
921 struct sock_fprog fprog;
924 if (copy_from_user(&fprog, optval, sizeof(fprog)))
927 ret = sk_attach_filter(&fprog, sk);
933 if (optlen == sizeof(u32)) {
937 if (copy_from_user(&ufd, optval, sizeof(ufd)))
940 ret = sk_attach_bpf(ufd, sk);
944 case SO_ATTACH_REUSEPORT_CBPF:
946 if (optlen == sizeof(struct sock_fprog)) {
947 struct sock_fprog fprog;
950 if (copy_from_user(&fprog, optval, sizeof(fprog)))
953 ret = sk_reuseport_attach_filter(&fprog, sk);
957 case SO_ATTACH_REUSEPORT_EBPF:
959 if (optlen == sizeof(u32)) {
963 if (copy_from_user(&ufd, optval, sizeof(ufd)))
966 ret = sk_reuseport_attach_bpf(ufd, sk);
970 case SO_DETACH_FILTER:
971 ret = sk_detach_filter(sk);
975 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
978 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
983 set_bit(SOCK_PASSSEC, &sock->flags);
985 clear_bit(SOCK_PASSSEC, &sock->flags);
988 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
995 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
999 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
1003 if (sock->ops->set_peek_off)
1004 ret = sock->ops->set_peek_off(sk, val);
1010 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
1013 case SO_SELECT_ERR_QUEUE:
1014 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
1017 #ifdef CONFIG_NET_RX_BUSY_POLL
1019 /* allow unprivileged users to decrease the value */
1020 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
1026 WRITE_ONCE(sk->sk_ll_usec, val);
1031 case SO_MAX_PACING_RATE:
1033 cmpxchg(&sk->sk_pacing_status,
1036 sk->sk_max_pacing_rate = val;
1037 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
1038 sk->sk_max_pacing_rate);
1041 case SO_INCOMING_CPU:
1042 WRITE_ONCE(sk->sk_incoming_cpu, val);
1047 dst_negative_advice(sk);
1051 if (sk->sk_family != PF_INET && sk->sk_family != PF_INET6)
1053 else if (sk->sk_protocol != IPPROTO_TCP)
1055 else if (sk->sk_state != TCP_CLOSE)
1057 else if (val < 0 || val > 1)
1060 sock_valbool_flag(sk, SOCK_ZEROCOPY, valbool);
1070 EXPORT_SYMBOL(sock_setsockopt);
1072 static const struct cred *sk_get_peer_cred(struct sock *sk)
1074 const struct cred *cred;
1076 spin_lock(&sk->sk_peer_lock);
1077 cred = get_cred(sk->sk_peer_cred);
1078 spin_unlock(&sk->sk_peer_lock);
1083 static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1084 struct ucred *ucred)
1086 ucred->pid = pid_vnr(pid);
1087 ucred->uid = ucred->gid = -1;
1089 struct user_namespace *current_ns = current_user_ns();
1091 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1092 ucred->gid = from_kgid_munged(current_ns, cred->egid);
1096 static int groups_to_user(gid_t __user *dst, const struct group_info *src)
1098 struct user_namespace *user_ns = current_user_ns();
1101 for (i = 0; i < src->ngroups; i++)
1102 if (put_user(from_kgid_munged(user_ns, src->gid[i]), dst + i))
1108 int sock_getsockopt(struct socket *sock, int level, int optname,
1109 char __user *optval, int __user *optlen)
1111 struct sock *sk = sock->sk;
1120 int lv = sizeof(int);
1123 if (get_user(len, optlen))
1128 memset(&v, 0, sizeof(v));
1132 v.val = sock_flag(sk, SOCK_DBG);
1136 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1140 v.val = sock_flag(sk, SOCK_BROADCAST);
1144 v.val = sk->sk_sndbuf;
1148 v.val = sk->sk_rcvbuf;
1152 v.val = sk->sk_reuse;
1156 v.val = sk->sk_reuseport;
1160 v.val = sock_flag(sk, SOCK_KEEPOPEN);
1164 v.val = sk->sk_type;
1168 v.val = sk->sk_protocol;
1172 v.val = sk->sk_family;
1176 v.val = -sock_error(sk);
1178 v.val = xchg(&sk->sk_err_soft, 0);
1182 v.val = sock_flag(sk, SOCK_URGINLINE);
1186 v.val = sk->sk_no_check_tx;
1190 v.val = sk->sk_priority;
1194 lv = sizeof(v.ling);
1195 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
1196 v.ling.l_linger = sk->sk_lingertime / HZ;
1200 sock_warn_obsolete_bsdism("getsockopt");
1204 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1205 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1208 case SO_TIMESTAMPNS:
1209 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
1212 case SO_TIMESTAMPING:
1213 v.val = sk->sk_tsflags;
1217 lv = sizeof(struct timeval);
1218 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1222 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1223 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * USEC_PER_SEC) / HZ;
1228 lv = sizeof(struct timeval);
1229 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1233 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1234 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * USEC_PER_SEC) / HZ;
1239 v.val = sk->sk_rcvlowat;
1247 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
1252 struct ucred peercred;
1253 if (len > sizeof(peercred))
1254 len = sizeof(peercred);
1256 spin_lock(&sk->sk_peer_lock);
1257 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1258 spin_unlock(&sk->sk_peer_lock);
1260 if (copy_to_user(optval, &peercred, len))
1267 const struct cred *cred;
1270 cred = sk_get_peer_cred(sk);
1274 n = cred->group_info->ngroups;
1275 if (len < n * sizeof(gid_t)) {
1276 len = n * sizeof(gid_t);
1278 return put_user(len, optlen) ? -EFAULT : -ERANGE;
1280 len = n * sizeof(gid_t);
1282 ret = groups_to_user((gid_t __user *)optval, cred->group_info);
1293 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1297 if (copy_to_user(optval, address, len))
1302 /* Dubious BSD thing... Probably nobody even uses it, but
1303 * the UNIX standard wants it for whatever reason... -DaveM
1306 v.val = sk->sk_state == TCP_LISTEN;
1310 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
1314 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1317 v.val = sk->sk_mark;
1321 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
1324 case SO_WIFI_STATUS:
1325 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
1329 if (!sock->ops->set_peek_off)
1332 v.val = sk->sk_peek_off;
1335 v.val = sock_flag(sk, SOCK_NOFCS);
1338 case SO_BINDTODEVICE:
1339 return sock_getbindtodevice(sk, optval, optlen, len);
1342 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1348 case SO_LOCK_FILTER:
1349 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1352 case SO_BPF_EXTENSIONS:
1353 v.val = bpf_tell_extensions();
1356 case SO_SELECT_ERR_QUEUE:
1357 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1360 #ifdef CONFIG_NET_RX_BUSY_POLL
1362 v.val = sk->sk_ll_usec;
1366 case SO_MAX_PACING_RATE:
1367 v.val = sk->sk_max_pacing_rate;
1370 case SO_INCOMING_CPU:
1371 v.val = READ_ONCE(sk->sk_incoming_cpu);
1376 u32 meminfo[SK_MEMINFO_VARS];
1378 sk_get_meminfo(sk, meminfo);
1380 len = min_t(unsigned int, len, sizeof(meminfo));
1381 if (copy_to_user(optval, &meminfo, len))
1387 #ifdef CONFIG_NET_RX_BUSY_POLL
1388 case SO_INCOMING_NAPI_ID:
1389 v.val = READ_ONCE(sk->sk_napi_id);
1391 /* aggregate non-NAPI IDs down to 0 */
1392 if (v.val < MIN_NAPI_ID)
1402 v.val64 = sock_gen_cookie(sk);
1406 v.val = sock_flag(sk, SOCK_ZEROCOPY);
1410 /* We implement the SO_SNDLOWAT etc to not be settable
1413 return -ENOPROTOOPT;
1418 if (copy_to_user(optval, &v, len))
1421 if (put_user(len, optlen))
1427 * Initialize an sk_lock.
1429 * (We also register the sk_lock with the lock validator.)
1431 static inline void sock_lock_init(struct sock *sk)
1433 if (sk->sk_kern_sock)
1434 sock_lock_init_class_and_name(
1436 af_family_kern_slock_key_strings[sk->sk_family],
1437 af_family_kern_slock_keys + sk->sk_family,
1438 af_family_kern_key_strings[sk->sk_family],
1439 af_family_kern_keys + sk->sk_family);
1441 sock_lock_init_class_and_name(
1443 af_family_slock_key_strings[sk->sk_family],
1444 af_family_slock_keys + sk->sk_family,
1445 af_family_key_strings[sk->sk_family],
1446 af_family_keys + sk->sk_family);
1450 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1451 * even temporarly, because of RCU lookups. sk_node should also be left as is.
1452 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
1454 static void sock_copy(struct sock *nsk, const struct sock *osk)
1456 #ifdef CONFIG_SECURITY_NETWORK
1457 void *sptr = nsk->sk_security;
1459 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1461 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1462 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1464 #ifdef CONFIG_SECURITY_NETWORK
1465 nsk->sk_security = sptr;
1466 security_sk_clone(osk, nsk);
1470 static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1474 struct kmem_cache *slab;
1478 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1481 if (priority & __GFP_ZERO)
1482 sk_prot_clear_nulls(sk, prot->obj_size);
1484 sk = kmalloc(prot->obj_size, priority);
1487 if (security_sk_alloc(sk, family, priority))
1490 if (!try_module_get(prot->owner))
1492 sk_tx_queue_clear(sk);
1498 security_sk_free(sk);
1501 kmem_cache_free(slab, sk);
1507 static void sk_prot_free(struct proto *prot, struct sock *sk)
1509 struct kmem_cache *slab;
1510 struct module *owner;
1512 owner = prot->owner;
1515 cgroup_sk_free(&sk->sk_cgrp_data);
1516 mem_cgroup_sk_free(sk);
1517 security_sk_free(sk);
1519 kmem_cache_free(slab, sk);
1526 * sk_alloc - All socket objects are allocated here
1527 * @net: the applicable net namespace
1528 * @family: protocol family
1529 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1530 * @prot: struct proto associated with this new sock instance
1531 * @kern: is this to be a kernel socket?
1533 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1534 struct proto *prot, int kern)
1538 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1540 sk->sk_family = family;
1542 * See comment in struct sock definition to understand
1543 * why we need sk_prot_creator -acme
1545 sk->sk_prot = sk->sk_prot_creator = prot;
1546 sk->sk_kern_sock = kern;
1548 sk->sk_net_refcnt = kern ? 0 : 1;
1549 if (likely(sk->sk_net_refcnt))
1551 sock_net_set(sk, net);
1552 refcount_set(&sk->sk_wmem_alloc, 1);
1554 mem_cgroup_sk_alloc(sk);
1555 cgroup_sk_alloc(&sk->sk_cgrp_data);
1556 sock_update_classid(&sk->sk_cgrp_data);
1557 sock_update_netprioidx(&sk->sk_cgrp_data);
1558 sk_tx_queue_clear(sk);
1563 EXPORT_SYMBOL(sk_alloc);
1565 /* Sockets having SOCK_RCU_FREE will call this function after one RCU
1566 * grace period. This is the case for UDP sockets and TCP listeners.
1568 static void __sk_destruct(struct rcu_head *head)
1570 struct sock *sk = container_of(head, struct sock, sk_rcu);
1571 struct sk_filter *filter;
1573 if (sk->sk_destruct)
1574 sk->sk_destruct(sk);
1576 filter = rcu_dereference_check(sk->sk_filter,
1577 refcount_read(&sk->sk_wmem_alloc) == 0);
1579 sk_filter_uncharge(sk, filter);
1580 RCU_INIT_POINTER(sk->sk_filter, NULL);
1583 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1585 if (atomic_read(&sk->sk_omem_alloc))
1586 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1587 __func__, atomic_read(&sk->sk_omem_alloc));
1589 if (sk->sk_frag.page) {
1590 put_page(sk->sk_frag.page);
1591 sk->sk_frag.page = NULL;
1594 /* We do not need to acquire sk->sk_peer_lock, we are the last user. */
1595 put_cred(sk->sk_peer_cred);
1596 put_pid(sk->sk_peer_pid);
1598 if (likely(sk->sk_net_refcnt))
1599 put_net(sock_net(sk));
1600 sk_prot_free(sk->sk_prot_creator, sk);
1603 void sk_destruct(struct sock *sk)
1605 bool use_call_rcu = sock_flag(sk, SOCK_RCU_FREE);
1607 if (rcu_access_pointer(sk->sk_reuseport_cb)) {
1608 reuseport_detach_sock(sk);
1609 use_call_rcu = true;
1613 call_rcu(&sk->sk_rcu, __sk_destruct);
1615 __sk_destruct(&sk->sk_rcu);
1618 static void __sk_free(struct sock *sk)
1620 if (unlikely(sk->sk_net_refcnt && sock_diag_has_destroy_listeners(sk)))
1621 sock_diag_broadcast_destroy(sk);
1626 void sk_free(struct sock *sk)
1629 * We subtract one from sk_wmem_alloc and can know if
1630 * some packets are still in some tx queue.
1631 * If not null, sock_wfree() will call __sk_free(sk) later
1633 if (refcount_dec_and_test(&sk->sk_wmem_alloc))
1636 EXPORT_SYMBOL(sk_free);
1638 static void sk_init_common(struct sock *sk)
1640 skb_queue_head_init(&sk->sk_receive_queue);
1641 skb_queue_head_init(&sk->sk_write_queue);
1642 skb_queue_head_init(&sk->sk_error_queue);
1644 rwlock_init(&sk->sk_callback_lock);
1645 lockdep_set_class_and_name(&sk->sk_receive_queue.lock,
1646 af_rlock_keys + sk->sk_family,
1647 af_family_rlock_key_strings[sk->sk_family]);
1648 lockdep_set_class_and_name(&sk->sk_write_queue.lock,
1649 af_wlock_keys + sk->sk_family,
1650 af_family_wlock_key_strings[sk->sk_family]);
1651 lockdep_set_class_and_name(&sk->sk_error_queue.lock,
1652 af_elock_keys + sk->sk_family,
1653 af_family_elock_key_strings[sk->sk_family]);
1654 lockdep_set_class_and_name(&sk->sk_callback_lock,
1655 af_callback_keys + sk->sk_family,
1656 af_family_clock_key_strings[sk->sk_family]);
1660 * sk_clone_lock - clone a socket, and lock its clone
1661 * @sk: the socket to clone
1662 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1664 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1666 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
1669 bool is_charged = true;
1671 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
1672 if (newsk != NULL) {
1673 struct sk_filter *filter;
1675 sock_copy(newsk, sk);
1677 newsk->sk_prot_creator = sk->sk_prot;
1680 if (likely(newsk->sk_net_refcnt))
1681 get_net(sock_net(newsk));
1682 sk_node_init(&newsk->sk_node);
1683 sock_lock_init(newsk);
1684 bh_lock_sock(newsk);
1685 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
1686 newsk->sk_backlog.len = 0;
1688 atomic_set(&newsk->sk_rmem_alloc, 0);
1690 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1692 refcount_set(&newsk->sk_wmem_alloc, 1);
1693 atomic_set(&newsk->sk_omem_alloc, 0);
1694 sk_init_common(newsk);
1696 newsk->sk_dst_cache = NULL;
1697 newsk->sk_dst_pending_confirm = 0;
1698 newsk->sk_wmem_queued = 0;
1699 newsk->sk_forward_alloc = 0;
1700 atomic_set(&newsk->sk_drops, 0);
1701 newsk->sk_send_head = NULL;
1702 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1703 atomic_set(&newsk->sk_zckey, 0);
1705 sock_reset_flag(newsk, SOCK_DONE);
1707 /* sk->sk_memcg will be populated at accept() time */
1708 newsk->sk_memcg = NULL;
1710 cgroup_sk_clone(&newsk->sk_cgrp_data);
1713 filter = rcu_dereference(sk->sk_filter);
1715 /* though it's an empty new sock, the charging may fail
1716 * if sysctl_optmem_max was changed between creation of
1717 * original socket and cloning
1719 is_charged = sk_filter_charge(newsk, filter);
1720 RCU_INIT_POINTER(newsk->sk_filter, filter);
1723 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
1724 /* We need to make sure that we don't uncharge the new
1725 * socket if we couldn't charge it in the first place
1726 * as otherwise we uncharge the parent's filter.
1729 RCU_INIT_POINTER(newsk->sk_filter, NULL);
1730 sk_free_unlock_clone(newsk);
1734 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
1737 newsk->sk_err_soft = 0;
1738 newsk->sk_priority = 0;
1739 newsk->sk_incoming_cpu = raw_smp_processor_id();
1740 atomic64_set(&newsk->sk_cookie, 0);
1743 * Before updating sk_refcnt, we must commit prior changes to memory
1744 * (Documentation/RCU/rculist_nulls.txt for details)
1747 refcount_set(&newsk->sk_refcnt, 2);
1750 * Increment the counter in the same struct proto as the master
1751 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1752 * is the same as sk->sk_prot->socks, as this field was copied
1755 * This _changes_ the previous behaviour, where
1756 * tcp_create_openreq_child always was incrementing the
1757 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1758 * to be taken into account in all callers. -acme
1760 sk_refcnt_debug_inc(newsk);
1761 sk_set_socket(newsk, NULL);
1762 sk_tx_queue_clear(newsk);
1763 newsk->sk_wq = NULL;
1765 if (newsk->sk_prot->sockets_allocated)
1766 sk_sockets_allocated_inc(newsk);
1768 if (sock_needs_netstamp(sk) &&
1769 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
1770 net_enable_timestamp();
1775 EXPORT_SYMBOL_GPL(sk_clone_lock);
1777 void sk_free_unlock_clone(struct sock *sk)
1779 /* It is still raw copy of parent, so invalidate
1780 * destructor and make plain sk_free() */
1781 sk->sk_destruct = NULL;
1785 EXPORT_SYMBOL_GPL(sk_free_unlock_clone);
1787 void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1791 sk_dst_set(sk, dst);
1792 sk->sk_route_caps = dst->dev->features;
1793 if (sk->sk_route_caps & NETIF_F_GSO)
1794 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
1795 sk->sk_route_caps &= ~sk->sk_route_nocaps;
1796 if (sk_can_gso(sk)) {
1797 if (dst->header_len && !xfrm_dst_offload_ok(dst)) {
1798 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
1800 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
1801 sk->sk_gso_max_size = dst->dev->gso_max_size;
1802 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
1805 sk->sk_gso_max_segs = max_segs;
1807 EXPORT_SYMBOL_GPL(sk_setup_caps);
1810 * Simple resource managers for sockets.
1815 * Write buffer destructor automatically called from kfree_skb.
1817 void sock_wfree(struct sk_buff *skb)
1819 struct sock *sk = skb->sk;
1820 unsigned int len = skb->truesize;
1822 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1824 * Keep a reference on sk_wmem_alloc, this will be released
1825 * after sk_write_space() call
1827 WARN_ON(refcount_sub_and_test(len - 1, &sk->sk_wmem_alloc));
1828 sk->sk_write_space(sk);
1832 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1833 * could not do because of in-flight packets
1835 if (refcount_sub_and_test(len, &sk->sk_wmem_alloc))
1838 EXPORT_SYMBOL(sock_wfree);
1840 /* This variant of sock_wfree() is used by TCP,
1841 * since it sets SOCK_USE_WRITE_QUEUE.
1843 void __sock_wfree(struct sk_buff *skb)
1845 struct sock *sk = skb->sk;
1847 if (refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc))
1851 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1856 if (unlikely(!sk_fullsock(sk))) {
1857 skb->destructor = sock_edemux;
1862 skb->destructor = sock_wfree;
1863 skb_set_hash_from_sk(skb, sk);
1865 * We used to take a refcount on sk, but following operation
1866 * is enough to guarantee sk_free() wont free this sock until
1867 * all in-flight packets are completed
1869 refcount_add(skb->truesize, &sk->sk_wmem_alloc);
1871 EXPORT_SYMBOL(skb_set_owner_w);
1873 /* This helper is used by netem, as it can hold packets in its
1874 * delay queue. We want to allow the owner socket to send more
1875 * packets, as if they were already TX completed by a typical driver.
1876 * But we also want to keep skb->sk set because some packet schedulers
1877 * rely on it (sch_fq for example).
1879 void skb_orphan_partial(struct sk_buff *skb)
1881 if (skb_is_tcp_pure_ack(skb))
1884 if (skb->destructor == sock_wfree
1886 || skb->destructor == tcp_wfree
1889 struct sock *sk = skb->sk;
1891 if (refcount_inc_not_zero(&sk->sk_refcnt)) {
1892 WARN_ON(refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc));
1893 skb->destructor = sock_efree;
1899 EXPORT_SYMBOL(skb_orphan_partial);
1902 * Read buffer destructor automatically called from kfree_skb.
1904 void sock_rfree(struct sk_buff *skb)
1906 struct sock *sk = skb->sk;
1907 unsigned int len = skb->truesize;
1909 atomic_sub(len, &sk->sk_rmem_alloc);
1910 sk_mem_uncharge(sk, len);
1912 EXPORT_SYMBOL(sock_rfree);
1915 * Buffer destructor for skbs that are not used directly in read or write
1916 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1918 void sock_efree(struct sk_buff *skb)
1922 EXPORT_SYMBOL(sock_efree);
1924 kuid_t sock_i_uid(struct sock *sk)
1928 read_lock_bh(&sk->sk_callback_lock);
1929 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
1930 read_unlock_bh(&sk->sk_callback_lock);
1933 EXPORT_SYMBOL(sock_i_uid);
1935 unsigned long sock_i_ino(struct sock *sk)
1939 read_lock_bh(&sk->sk_callback_lock);
1940 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
1941 read_unlock_bh(&sk->sk_callback_lock);
1944 EXPORT_SYMBOL(sock_i_ino);
1947 * Allocate a skb from the socket's send buffer.
1949 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1952 if (force || refcount_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
1953 struct sk_buff *skb = alloc_skb(size, priority);
1955 skb_set_owner_w(skb, sk);
1961 EXPORT_SYMBOL(sock_wmalloc);
1963 static void sock_ofree(struct sk_buff *skb)
1965 struct sock *sk = skb->sk;
1967 atomic_sub(skb->truesize, &sk->sk_omem_alloc);
1970 struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1973 struct sk_buff *skb;
1975 /* small safe race: SKB_TRUESIZE may differ from final skb->truesize */
1976 if (atomic_read(&sk->sk_omem_alloc) + SKB_TRUESIZE(size) >
1980 skb = alloc_skb(size, priority);
1984 atomic_add(skb->truesize, &sk->sk_omem_alloc);
1986 skb->destructor = sock_ofree;
1991 * Allocate a memory block from the socket's option memory buffer.
1993 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1995 if ((unsigned int)size <= sysctl_optmem_max &&
1996 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1998 /* First do the add, to avoid the race if kmalloc
2001 atomic_add(size, &sk->sk_omem_alloc);
2002 mem = kmalloc(size, priority);
2005 atomic_sub(size, &sk->sk_omem_alloc);
2009 EXPORT_SYMBOL(sock_kmalloc);
2011 /* Free an option memory block. Note, we actually want the inline
2012 * here as this allows gcc to detect the nullify and fold away the
2013 * condition entirely.
2015 static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
2018 if (WARN_ON_ONCE(!mem))
2024 atomic_sub(size, &sk->sk_omem_alloc);
2027 void sock_kfree_s(struct sock *sk, void *mem, int size)
2029 __sock_kfree_s(sk, mem, size, false);
2031 EXPORT_SYMBOL(sock_kfree_s);
2033 void sock_kzfree_s(struct sock *sk, void *mem, int size)
2035 __sock_kfree_s(sk, mem, size, true);
2037 EXPORT_SYMBOL(sock_kzfree_s);
2039 /* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
2040 I think, these locks should be removed for datagram sockets.
2042 static long sock_wait_for_wmem(struct sock *sk, long timeo)
2046 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2050 if (signal_pending(current))
2052 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2053 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
2054 if (refcount_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
2056 if (sk->sk_shutdown & SEND_SHUTDOWN)
2060 timeo = schedule_timeout(timeo);
2062 finish_wait(sk_sleep(sk), &wait);
2068 * Generic send/receive buffer handlers
2071 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
2072 unsigned long data_len, int noblock,
2073 int *errcode, int max_page_order)
2075 struct sk_buff *skb;
2079 timeo = sock_sndtimeo(sk, noblock);
2081 err = sock_error(sk);
2086 if (sk->sk_shutdown & SEND_SHUTDOWN)
2089 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
2092 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2093 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2097 if (signal_pending(current))
2099 timeo = sock_wait_for_wmem(sk, timeo);
2101 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
2102 errcode, sk->sk_allocation);
2104 skb_set_owner_w(skb, sk);
2108 err = sock_intr_errno(timeo);
2113 EXPORT_SYMBOL(sock_alloc_send_pskb);
2115 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
2116 int noblock, int *errcode)
2118 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
2120 EXPORT_SYMBOL(sock_alloc_send_skb);
2122 int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
2123 struct sockcm_cookie *sockc)
2127 switch (cmsg->cmsg_type) {
2129 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2131 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2133 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
2135 case SO_TIMESTAMPING:
2136 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2139 tsflags = *(u32 *)CMSG_DATA(cmsg);
2140 if (tsflags & ~SOF_TIMESTAMPING_TX_RECORD_MASK)
2143 sockc->tsflags &= ~SOF_TIMESTAMPING_TX_RECORD_MASK;
2144 sockc->tsflags |= tsflags;
2146 /* SCM_RIGHTS and SCM_CREDENTIALS are semantically in SOL_UNIX. */
2148 case SCM_CREDENTIALS:
2155 EXPORT_SYMBOL(__sock_cmsg_send);
2157 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
2158 struct sockcm_cookie *sockc)
2160 struct cmsghdr *cmsg;
2163 for_each_cmsghdr(cmsg, msg) {
2164 if (!CMSG_OK(msg, cmsg))
2166 if (cmsg->cmsg_level != SOL_SOCKET)
2168 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
2174 EXPORT_SYMBOL(sock_cmsg_send);
2176 static void sk_enter_memory_pressure(struct sock *sk)
2178 if (!sk->sk_prot->enter_memory_pressure)
2181 sk->sk_prot->enter_memory_pressure(sk);
2184 static void sk_leave_memory_pressure(struct sock *sk)
2186 if (sk->sk_prot->leave_memory_pressure) {
2187 sk->sk_prot->leave_memory_pressure(sk);
2189 unsigned long *memory_pressure = sk->sk_prot->memory_pressure;
2191 if (memory_pressure && READ_ONCE(*memory_pressure))
2192 WRITE_ONCE(*memory_pressure, 0);
2196 /* On 32bit arches, an skb frag is limited to 2^15 */
2197 #define SKB_FRAG_PAGE_ORDER get_order(32768)
2200 * skb_page_frag_refill - check that a page_frag contains enough room
2201 * @sz: minimum size of the fragment we want to get
2202 * @pfrag: pointer to page_frag
2203 * @gfp: priority for memory allocation
2205 * Note: While this allocator tries to use high order pages, there is
2206 * no guarantee that allocations succeed. Therefore, @sz MUST be
2207 * less or equal than PAGE_SIZE.
2209 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
2212 if (page_ref_count(pfrag->page) == 1) {
2216 if (pfrag->offset + sz <= pfrag->size)
2218 put_page(pfrag->page);
2222 if (SKB_FRAG_PAGE_ORDER) {
2223 /* Avoid direct reclaim but allow kswapd to wake */
2224 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
2225 __GFP_COMP | __GFP_NOWARN |
2227 SKB_FRAG_PAGE_ORDER);
2228 if (likely(pfrag->page)) {
2229 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
2233 pfrag->page = alloc_page(gfp);
2234 if (likely(pfrag->page)) {
2235 pfrag->size = PAGE_SIZE;
2240 EXPORT_SYMBOL(skb_page_frag_refill);
2242 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
2244 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
2247 sk_enter_memory_pressure(sk);
2248 sk_stream_moderate_sndbuf(sk);
2251 EXPORT_SYMBOL(sk_page_frag_refill);
2253 static void __lock_sock(struct sock *sk)
2254 __releases(&sk->sk_lock.slock)
2255 __acquires(&sk->sk_lock.slock)
2260 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
2261 TASK_UNINTERRUPTIBLE);
2262 spin_unlock_bh(&sk->sk_lock.slock);
2264 spin_lock_bh(&sk->sk_lock.slock);
2265 if (!sock_owned_by_user(sk))
2268 finish_wait(&sk->sk_lock.wq, &wait);
2271 void __release_sock(struct sock *sk)
2272 __releases(&sk->sk_lock.slock)
2273 __acquires(&sk->sk_lock.slock)
2275 struct sk_buff *skb, *next;
2277 while ((skb = sk->sk_backlog.head) != NULL) {
2278 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
2280 spin_unlock_bh(&sk->sk_lock.slock);
2285 WARN_ON_ONCE(skb_dst_is_noref(skb));
2287 sk_backlog_rcv(sk, skb);
2292 } while (skb != NULL);
2294 spin_lock_bh(&sk->sk_lock.slock);
2298 * Doing the zeroing here guarantee we can not loop forever
2299 * while a wild producer attempts to flood us.
2301 sk->sk_backlog.len = 0;
2304 void __sk_flush_backlog(struct sock *sk)
2306 spin_lock_bh(&sk->sk_lock.slock);
2308 spin_unlock_bh(&sk->sk_lock.slock);
2312 * sk_wait_data - wait for data to arrive at sk_receive_queue
2313 * @sk: sock to wait on
2314 * @timeo: for how long
2315 * @skb: last skb seen on sk_receive_queue
2317 * Now socket state including sk->sk_err is changed only under lock,
2318 * hence we may omit checks after joining wait queue.
2319 * We check receive queue before schedule() only as optimization;
2320 * it is very likely that release_sock() added new data.
2322 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
2324 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2327 add_wait_queue(sk_sleep(sk), &wait);
2328 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
2329 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb, &wait);
2330 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
2331 remove_wait_queue(sk_sleep(sk), &wait);
2334 EXPORT_SYMBOL(sk_wait_data);
2337 * __sk_mem_raise_allocated - increase memory_allocated
2339 * @size: memory size to allocate
2340 * @amt: pages to allocate
2341 * @kind: allocation type
2343 * Similar to __sk_mem_schedule(), but does not update sk_forward_alloc
2345 int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
2347 struct proto *prot = sk->sk_prot;
2348 long allocated = sk_memory_allocated_add(sk, amt);
2350 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2351 !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
2352 goto suppress_allocation;
2355 if (allocated <= sk_prot_mem_limits(sk, 0)) {
2356 sk_leave_memory_pressure(sk);
2360 /* Under pressure. */
2361 if (allocated > sk_prot_mem_limits(sk, 1))
2362 sk_enter_memory_pressure(sk);
2364 /* Over hard limit. */
2365 if (allocated > sk_prot_mem_limits(sk, 2))
2366 goto suppress_allocation;
2368 /* guarantee minimum buffer size under pressure */
2369 if (kind == SK_MEM_RECV) {
2370 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2373 } else { /* SK_MEM_SEND */
2374 if (sk->sk_type == SOCK_STREAM) {
2375 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2377 } else if (refcount_read(&sk->sk_wmem_alloc) <
2378 prot->sysctl_wmem[0])
2382 if (sk_has_memory_pressure(sk)) {
2385 if (!sk_under_memory_pressure(sk))
2387 alloc = sk_sockets_allocated_read_positive(sk);
2388 if (sk_prot_mem_limits(sk, 2) > alloc *
2389 sk_mem_pages(sk->sk_wmem_queued +
2390 atomic_read(&sk->sk_rmem_alloc) +
2391 sk->sk_forward_alloc))
2395 suppress_allocation:
2397 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2398 sk_stream_moderate_sndbuf(sk);
2400 /* Fail only if socket is _under_ its sndbuf.
2401 * In this case we cannot block, so that we have to fail.
2403 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2407 trace_sock_exceed_buf_limit(sk, prot, allocated);
2409 sk_memory_allocated_sub(sk, amt);
2411 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2412 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
2416 EXPORT_SYMBOL(__sk_mem_raise_allocated);
2419 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2421 * @size: memory size to allocate
2422 * @kind: allocation type
2424 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2425 * rmem allocation. This function assumes that protocols which have
2426 * memory_pressure use sk_wmem_queued as write buffer accounting.
2428 int __sk_mem_schedule(struct sock *sk, int size, int kind)
2430 int ret, amt = sk_mem_pages(size);
2432 sk->sk_forward_alloc += amt << SK_MEM_QUANTUM_SHIFT;
2433 ret = __sk_mem_raise_allocated(sk, size, amt, kind);
2435 sk->sk_forward_alloc -= amt << SK_MEM_QUANTUM_SHIFT;
2438 EXPORT_SYMBOL(__sk_mem_schedule);
2441 * __sk_mem_reduce_allocated - reclaim memory_allocated
2443 * @amount: number of quanta
2445 * Similar to __sk_mem_reclaim(), but does not update sk_forward_alloc
2447 void __sk_mem_reduce_allocated(struct sock *sk, int amount)
2449 sk_memory_allocated_sub(sk, amount);
2451 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2452 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
2454 if (sk_under_memory_pressure(sk) &&
2455 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2456 sk_leave_memory_pressure(sk);
2458 EXPORT_SYMBOL(__sk_mem_reduce_allocated);
2461 * __sk_mem_reclaim - reclaim sk_forward_alloc and memory_allocated
2463 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
2465 void __sk_mem_reclaim(struct sock *sk, int amount)
2467 amount >>= SK_MEM_QUANTUM_SHIFT;
2468 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
2469 __sk_mem_reduce_allocated(sk, amount);
2471 EXPORT_SYMBOL(__sk_mem_reclaim);
2473 int sk_set_peek_off(struct sock *sk, int val)
2475 sk->sk_peek_off = val;
2478 EXPORT_SYMBOL_GPL(sk_set_peek_off);
2481 * Set of default routines for initialising struct proto_ops when
2482 * the protocol does not support a particular function. In certain
2483 * cases where it makes no sense for a protocol to have a "do nothing"
2484 * function, some default processing is provided.
2487 int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2491 EXPORT_SYMBOL(sock_no_bind);
2493 int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
2498 EXPORT_SYMBOL(sock_no_connect);
2500 int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2504 EXPORT_SYMBOL(sock_no_socketpair);
2506 int sock_no_accept(struct socket *sock, struct socket *newsock, int flags,
2511 EXPORT_SYMBOL(sock_no_accept);
2513 int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
2518 EXPORT_SYMBOL(sock_no_getname);
2520 unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
2524 EXPORT_SYMBOL(sock_no_poll);
2526 int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2530 EXPORT_SYMBOL(sock_no_ioctl);
2532 int sock_no_listen(struct socket *sock, int backlog)
2536 EXPORT_SYMBOL(sock_no_listen);
2538 int sock_no_shutdown(struct socket *sock, int how)
2542 EXPORT_SYMBOL(sock_no_shutdown);
2544 int sock_no_setsockopt(struct socket *sock, int level, int optname,
2545 char __user *optval, unsigned int optlen)
2549 EXPORT_SYMBOL(sock_no_setsockopt);
2551 int sock_no_getsockopt(struct socket *sock, int level, int optname,
2552 char __user *optval, int __user *optlen)
2556 EXPORT_SYMBOL(sock_no_getsockopt);
2558 int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
2562 EXPORT_SYMBOL(sock_no_sendmsg);
2564 int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *m, size_t len)
2568 EXPORT_SYMBOL(sock_no_sendmsg_locked);
2570 int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2575 EXPORT_SYMBOL(sock_no_recvmsg);
2577 int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2579 /* Mirror missing mmap method error code */
2582 EXPORT_SYMBOL(sock_no_mmap);
2585 * When a file is received (via SCM_RIGHTS, etc), we must bump the
2586 * various sock-based usage counts.
2588 void __receive_sock(struct file *file)
2590 struct socket *sock;
2594 * The resulting value of "error" is ignored here since we only
2595 * need to take action when the file is a socket and testing
2596 * "sock" for NULL is sufficient.
2598 sock = sock_from_file(file, &error);
2600 sock_update_netprioidx(&sock->sk->sk_cgrp_data);
2601 sock_update_classid(&sock->sk->sk_cgrp_data);
2605 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2608 struct msghdr msg = {.msg_flags = flags};
2610 char *kaddr = kmap(page);
2611 iov.iov_base = kaddr + offset;
2613 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2617 EXPORT_SYMBOL(sock_no_sendpage);
2619 ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
2620 int offset, size_t size, int flags)
2623 struct msghdr msg = {.msg_flags = flags};
2625 char *kaddr = kmap(page);
2627 iov.iov_base = kaddr + offset;
2629 res = kernel_sendmsg_locked(sk, &msg, &iov, 1, size);
2633 EXPORT_SYMBOL(sock_no_sendpage_locked);
2636 * Default Socket Callbacks
2639 static void sock_def_wakeup(struct sock *sk)
2641 struct socket_wq *wq;
2644 wq = rcu_dereference(sk->sk_wq);
2645 if (skwq_has_sleeper(wq))
2646 wake_up_interruptible_all(&wq->wait);
2650 static void sock_def_error_report(struct sock *sk)
2652 struct socket_wq *wq;
2655 wq = rcu_dereference(sk->sk_wq);
2656 if (skwq_has_sleeper(wq))
2657 wake_up_interruptible_poll(&wq->wait, POLLERR);
2658 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
2662 static void sock_def_readable(struct sock *sk)
2664 struct socket_wq *wq;
2667 wq = rcu_dereference(sk->sk_wq);
2668 if (skwq_has_sleeper(wq))
2669 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
2670 POLLRDNORM | POLLRDBAND);
2671 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
2675 static void sock_def_write_space(struct sock *sk)
2677 struct socket_wq *wq;
2681 /* Do not wake up a writer until he can make "significant"
2684 if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
2685 wq = rcu_dereference(sk->sk_wq);
2686 if (skwq_has_sleeper(wq))
2687 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
2688 POLLWRNORM | POLLWRBAND);
2690 /* Should agree with poll, otherwise some programs break */
2691 if (sock_writeable(sk))
2692 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
2698 static void sock_def_destruct(struct sock *sk)
2702 void sk_send_sigurg(struct sock *sk)
2704 if (sk->sk_socket && sk->sk_socket->file)
2705 if (send_sigurg(&sk->sk_socket->file->f_owner))
2706 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
2708 EXPORT_SYMBOL(sk_send_sigurg);
2710 void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2711 unsigned long expires)
2713 if (!mod_timer(timer, expires))
2716 EXPORT_SYMBOL(sk_reset_timer);
2718 void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2720 if (del_timer(timer))
2723 EXPORT_SYMBOL(sk_stop_timer);
2725 void sock_init_data(struct socket *sock, struct sock *sk)
2728 sk->sk_send_head = NULL;
2730 init_timer(&sk->sk_timer);
2732 sk->sk_allocation = GFP_KERNEL;
2733 sk->sk_rcvbuf = sysctl_rmem_default;
2734 sk->sk_sndbuf = sysctl_wmem_default;
2735 sk->sk_state = TCP_CLOSE;
2736 sk_set_socket(sk, sock);
2738 sock_set_flag(sk, SOCK_ZAPPED);
2741 sk->sk_type = sock->type;
2742 sk->sk_wq = sock->wq;
2744 sk->sk_uid = SOCK_INODE(sock)->i_uid;
2747 sk->sk_uid = make_kuid(sock_net(sk)->user_ns, 0);
2750 rwlock_init(&sk->sk_callback_lock);
2751 if (sk->sk_kern_sock)
2752 lockdep_set_class_and_name(
2753 &sk->sk_callback_lock,
2754 af_kern_callback_keys + sk->sk_family,
2755 af_family_kern_clock_key_strings[sk->sk_family]);
2757 lockdep_set_class_and_name(
2758 &sk->sk_callback_lock,
2759 af_callback_keys + sk->sk_family,
2760 af_family_clock_key_strings[sk->sk_family]);
2762 sk->sk_state_change = sock_def_wakeup;
2763 sk->sk_data_ready = sock_def_readable;
2764 sk->sk_write_space = sock_def_write_space;
2765 sk->sk_error_report = sock_def_error_report;
2766 sk->sk_destruct = sock_def_destruct;
2768 sk->sk_frag.page = NULL;
2769 sk->sk_frag.offset = 0;
2770 sk->sk_peek_off = -1;
2772 sk->sk_peer_pid = NULL;
2773 sk->sk_peer_cred = NULL;
2774 spin_lock_init(&sk->sk_peer_lock);
2776 sk->sk_write_pending = 0;
2777 sk->sk_rcvlowat = 1;
2778 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2779 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2781 sk->sk_stamp = SK_DEFAULT_STAMP;
2782 #if BITS_PER_LONG==32
2783 seqlock_init(&sk->sk_stamp_seq);
2785 atomic_set(&sk->sk_zckey, 0);
2787 #ifdef CONFIG_NET_RX_BUSY_POLL
2789 sk->sk_ll_usec = sysctl_net_busy_read;
2792 sk->sk_max_pacing_rate = ~0U;
2793 sk->sk_pacing_rate = ~0U;
2794 sk->sk_incoming_cpu = -1;
2796 * Before updating sk_refcnt, we must commit prior changes to memory
2797 * (Documentation/RCU/rculist_nulls.txt for details)
2800 refcount_set(&sk->sk_refcnt, 1);
2801 atomic_set(&sk->sk_drops, 0);
2803 EXPORT_SYMBOL(sock_init_data);
2805 void lock_sock_nested(struct sock *sk, int subclass)
2808 spin_lock_bh(&sk->sk_lock.slock);
2809 if (sk->sk_lock.owned)
2811 sk->sk_lock.owned = 1;
2812 spin_unlock(&sk->sk_lock.slock);
2814 * The sk_lock has mutex_lock() semantics here:
2816 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
2819 EXPORT_SYMBOL(lock_sock_nested);
2821 void release_sock(struct sock *sk)
2823 spin_lock_bh(&sk->sk_lock.slock);
2824 if (sk->sk_backlog.tail)
2827 /* Warning : release_cb() might need to release sk ownership,
2828 * ie call sock_release_ownership(sk) before us.
2830 if (sk->sk_prot->release_cb)
2831 sk->sk_prot->release_cb(sk);
2833 sock_release_ownership(sk);
2834 if (waitqueue_active(&sk->sk_lock.wq))
2835 wake_up(&sk->sk_lock.wq);
2836 spin_unlock_bh(&sk->sk_lock.slock);
2838 EXPORT_SYMBOL(release_sock);
2841 * lock_sock_fast - fast version of lock_sock
2844 * This version should be used for very small section, where process wont block
2845 * return false if fast path is taken:
2847 * sk_lock.slock locked, owned = 0, BH disabled
2849 * return true if slow path is taken:
2851 * sk_lock.slock unlocked, owned = 1, BH enabled
2853 bool lock_sock_fast(struct sock *sk)
2856 spin_lock_bh(&sk->sk_lock.slock);
2858 if (!sk->sk_lock.owned)
2860 * Note : We must disable BH
2865 sk->sk_lock.owned = 1;
2866 spin_unlock(&sk->sk_lock.slock);
2868 * The sk_lock has mutex_lock() semantics here:
2870 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2874 EXPORT_SYMBOL(lock_sock_fast);
2876 int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
2879 if (!sock_flag(sk, SOCK_TIMESTAMP))
2880 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
2881 tv = ktime_to_timeval(sk->sk_stamp);
2882 if (tv.tv_sec == -1)
2884 if (tv.tv_sec == 0) {
2885 sk->sk_stamp = ktime_get_real();
2886 tv = ktime_to_timeval(sk->sk_stamp);
2888 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
2890 EXPORT_SYMBOL(sock_get_timestamp);
2892 int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2895 if (!sock_flag(sk, SOCK_TIMESTAMP))
2896 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
2897 ts = ktime_to_timespec(sk->sk_stamp);
2898 if (ts.tv_sec == -1)
2900 if (ts.tv_sec == 0) {
2901 sk->sk_stamp = ktime_get_real();
2902 ts = ktime_to_timespec(sk->sk_stamp);
2904 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2906 EXPORT_SYMBOL(sock_get_timestampns);
2908 void sock_enable_timestamp(struct sock *sk, int flag)
2910 if (!sock_flag(sk, flag)) {
2911 unsigned long previous_flags = sk->sk_flags;
2913 sock_set_flag(sk, flag);
2915 * we just set one of the two flags which require net
2916 * time stamping, but time stamping might have been on
2917 * already because of the other one
2919 if (sock_needs_netstamp(sk) &&
2920 !(previous_flags & SK_FLAGS_TIMESTAMP))
2921 net_enable_timestamp();
2925 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2926 int level, int type)
2928 struct sock_exterr_skb *serr;
2929 struct sk_buff *skb;
2933 skb = sock_dequeue_err_skb(sk);
2939 msg->msg_flags |= MSG_TRUNC;
2942 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2946 sock_recv_timestamp(msg, sk, skb);
2948 serr = SKB_EXT_ERR(skb);
2949 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2951 msg->msg_flags |= MSG_ERRQUEUE;
2959 EXPORT_SYMBOL(sock_recv_errqueue);
2962 * Get a socket option on an socket.
2964 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2965 * asynchronous errors should be reported by getsockopt. We assume
2966 * this means if you specify SO_ERROR (otherwise whats the point of it).
2968 int sock_common_getsockopt(struct socket *sock, int level, int optname,
2969 char __user *optval, int __user *optlen)
2971 struct sock *sk = sock->sk;
2973 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2975 EXPORT_SYMBOL(sock_common_getsockopt);
2977 #ifdef CONFIG_COMPAT
2978 int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2979 char __user *optval, int __user *optlen)
2981 struct sock *sk = sock->sk;
2983 if (sk->sk_prot->compat_getsockopt != NULL)
2984 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2986 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2988 EXPORT_SYMBOL(compat_sock_common_getsockopt);
2991 int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2994 struct sock *sk = sock->sk;
2998 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
2999 flags & ~MSG_DONTWAIT, &addr_len);
3001 msg->msg_namelen = addr_len;
3004 EXPORT_SYMBOL(sock_common_recvmsg);
3007 * Set socket options on an inet socket.
3009 int sock_common_setsockopt(struct socket *sock, int level, int optname,
3010 char __user *optval, unsigned int optlen)
3012 struct sock *sk = sock->sk;
3014 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
3016 EXPORT_SYMBOL(sock_common_setsockopt);
3018 #ifdef CONFIG_COMPAT
3019 int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
3020 char __user *optval, unsigned int optlen)
3022 struct sock *sk = sock->sk;
3024 if (sk->sk_prot->compat_setsockopt != NULL)
3025 return sk->sk_prot->compat_setsockopt(sk, level, optname,
3027 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
3029 EXPORT_SYMBOL(compat_sock_common_setsockopt);
3032 void sk_common_release(struct sock *sk)
3034 if (sk->sk_prot->destroy)
3035 sk->sk_prot->destroy(sk);
3038 * Observation: when sock_common_release is called, processes have
3039 * no access to socket. But net still has.
3040 * Step one, detach it from networking:
3042 * A. Remove from hash tables.
3045 sk->sk_prot->unhash(sk);
3048 * In this point socket cannot receive new packets, but it is possible
3049 * that some packets are in flight because some CPU runs receiver and
3050 * did hash table lookup before we unhashed socket. They will achieve
3051 * receive queue and will be purged by socket destructor.
3053 * Also we still have packets pending on receive queue and probably,
3054 * our own packets waiting in device queues. sock_destroy will drain
3055 * receive queue, but transmitted packets will delay socket destruction
3056 * until the last reference will be released.
3061 xfrm_sk_free_policy(sk);
3063 sk_refcnt_debug_release(sk);
3067 EXPORT_SYMBOL(sk_common_release);
3069 void sk_get_meminfo(const struct sock *sk, u32 *mem)
3071 memset(mem, 0, sizeof(*mem) * SK_MEMINFO_VARS);
3073 mem[SK_MEMINFO_RMEM_ALLOC] = sk_rmem_alloc_get(sk);
3074 mem[SK_MEMINFO_RCVBUF] = sk->sk_rcvbuf;
3075 mem[SK_MEMINFO_WMEM_ALLOC] = sk_wmem_alloc_get(sk);
3076 mem[SK_MEMINFO_SNDBUF] = sk->sk_sndbuf;
3077 mem[SK_MEMINFO_FWD_ALLOC] = sk->sk_forward_alloc;
3078 mem[SK_MEMINFO_WMEM_QUEUED] = sk->sk_wmem_queued;
3079 mem[SK_MEMINFO_OPTMEM] = atomic_read(&sk->sk_omem_alloc);
3080 mem[SK_MEMINFO_BACKLOG] = sk->sk_backlog.len;
3081 mem[SK_MEMINFO_DROPS] = atomic_read(&sk->sk_drops);
3084 #ifdef CONFIG_PROC_FS
3085 #define PROTO_INUSE_NR 64 /* should be enough for the first time */
3087 int val[PROTO_INUSE_NR];
3090 static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
3092 #ifdef CONFIG_NET_NS
3093 void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
3095 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
3097 EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
3099 int sock_prot_inuse_get(struct net *net, struct proto *prot)
3101 int cpu, idx = prot->inuse_idx;
3104 for_each_possible_cpu(cpu)
3105 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
3107 return res >= 0 ? res : 0;
3109 EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
3111 static int __net_init sock_inuse_init_net(struct net *net)
3113 net->core.inuse = alloc_percpu(struct prot_inuse);
3114 return net->core.inuse ? 0 : -ENOMEM;
3117 static void __net_exit sock_inuse_exit_net(struct net *net)
3119 free_percpu(net->core.inuse);
3122 static struct pernet_operations net_inuse_ops = {
3123 .init = sock_inuse_init_net,
3124 .exit = sock_inuse_exit_net,
3127 static __init int net_inuse_init(void)
3129 if (register_pernet_subsys(&net_inuse_ops))
3130 panic("Cannot initialize net inuse counters");
3135 core_initcall(net_inuse_init);
3137 static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
3139 void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
3141 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
3143 EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
3145 int sock_prot_inuse_get(struct net *net, struct proto *prot)
3147 int cpu, idx = prot->inuse_idx;
3150 for_each_possible_cpu(cpu)
3151 res += per_cpu(prot_inuse, cpu).val[idx];
3153 return res >= 0 ? res : 0;
3155 EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
3158 static void assign_proto_idx(struct proto *prot)
3160 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
3162 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
3163 pr_err("PROTO_INUSE_NR exhausted\n");
3167 set_bit(prot->inuse_idx, proto_inuse_idx);
3170 static void release_proto_idx(struct proto *prot)
3172 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
3173 clear_bit(prot->inuse_idx, proto_inuse_idx);
3176 static inline void assign_proto_idx(struct proto *prot)
3180 static inline void release_proto_idx(struct proto *prot)
3185 static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
3189 kfree(rsk_prot->slab_name);
3190 rsk_prot->slab_name = NULL;
3191 kmem_cache_destroy(rsk_prot->slab);
3192 rsk_prot->slab = NULL;
3195 static int req_prot_init(const struct proto *prot)
3197 struct request_sock_ops *rsk_prot = prot->rsk_prot;
3202 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
3204 if (!rsk_prot->slab_name)
3207 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
3208 rsk_prot->obj_size, 0,
3209 prot->slab_flags, NULL);
3211 if (!rsk_prot->slab) {
3212 pr_crit("%s: Can't create request sock SLAB cache!\n",
3219 int proto_register(struct proto *prot, int alloc_slab)
3222 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
3223 SLAB_HWCACHE_ALIGN | prot->slab_flags,
3226 if (prot->slab == NULL) {
3227 pr_crit("%s: Can't create sock SLAB cache!\n",
3232 if (req_prot_init(prot))
3233 goto out_free_request_sock_slab;
3235 if (prot->twsk_prot != NULL) {
3236 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
3238 if (prot->twsk_prot->twsk_slab_name == NULL)
3239 goto out_free_request_sock_slab;
3241 prot->twsk_prot->twsk_slab =
3242 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
3243 prot->twsk_prot->twsk_obj_size,
3247 if (prot->twsk_prot->twsk_slab == NULL)
3248 goto out_free_timewait_sock_slab_name;
3252 mutex_lock(&proto_list_mutex);
3253 list_add(&prot->node, &proto_list);
3254 assign_proto_idx(prot);
3255 mutex_unlock(&proto_list_mutex);
3258 out_free_timewait_sock_slab_name:
3259 kfree(prot->twsk_prot->twsk_slab_name);
3260 out_free_request_sock_slab:
3261 req_prot_cleanup(prot->rsk_prot);
3263 kmem_cache_destroy(prot->slab);
3268 EXPORT_SYMBOL(proto_register);
3270 void proto_unregister(struct proto *prot)
3272 mutex_lock(&proto_list_mutex);
3273 release_proto_idx(prot);
3274 list_del(&prot->node);
3275 mutex_unlock(&proto_list_mutex);
3277 kmem_cache_destroy(prot->slab);
3280 req_prot_cleanup(prot->rsk_prot);
3282 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
3283 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
3284 kfree(prot->twsk_prot->twsk_slab_name);
3285 prot->twsk_prot->twsk_slab = NULL;
3288 EXPORT_SYMBOL(proto_unregister);
3290 #ifdef CONFIG_PROC_FS
3291 static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
3292 __acquires(proto_list_mutex)
3294 mutex_lock(&proto_list_mutex);
3295 return seq_list_start_head(&proto_list, *pos);
3298 static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3300 return seq_list_next(v, &proto_list, pos);
3303 static void proto_seq_stop(struct seq_file *seq, void *v)
3304 __releases(proto_list_mutex)
3306 mutex_unlock(&proto_list_mutex);
3309 static char proto_method_implemented(const void *method)
3311 return method == NULL ? 'n' : 'y';
3313 static long sock_prot_memory_allocated(struct proto *proto)
3315 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
3318 static char *sock_prot_memory_pressure(struct proto *proto)
3320 return proto->memory_pressure != NULL ?
3321 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
3324 static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
3327 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
3328 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
3331 sock_prot_inuse_get(seq_file_net(seq), proto),
3332 sock_prot_memory_allocated(proto),
3333 sock_prot_memory_pressure(proto),
3335 proto->slab == NULL ? "no" : "yes",
3336 module_name(proto->owner),
3337 proto_method_implemented(proto->close),
3338 proto_method_implemented(proto->connect),
3339 proto_method_implemented(proto->disconnect),
3340 proto_method_implemented(proto->accept),
3341 proto_method_implemented(proto->ioctl),
3342 proto_method_implemented(proto->init),
3343 proto_method_implemented(proto->destroy),
3344 proto_method_implemented(proto->shutdown),
3345 proto_method_implemented(proto->setsockopt),
3346 proto_method_implemented(proto->getsockopt),
3347 proto_method_implemented(proto->sendmsg),
3348 proto_method_implemented(proto->recvmsg),
3349 proto_method_implemented(proto->sendpage),
3350 proto_method_implemented(proto->bind),
3351 proto_method_implemented(proto->backlog_rcv),
3352 proto_method_implemented(proto->hash),
3353 proto_method_implemented(proto->unhash),
3354 proto_method_implemented(proto->get_port),
3355 proto_method_implemented(proto->enter_memory_pressure));
3358 static int proto_seq_show(struct seq_file *seq, void *v)
3360 if (v == &proto_list)
3361 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3370 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3372 proto_seq_printf(seq, list_entry(v, struct proto, node));
3376 static const struct seq_operations proto_seq_ops = {
3377 .start = proto_seq_start,
3378 .next = proto_seq_next,
3379 .stop = proto_seq_stop,
3380 .show = proto_seq_show,
3383 static int proto_seq_open(struct inode *inode, struct file *file)
3385 return seq_open_net(inode, file, &proto_seq_ops,
3386 sizeof(struct seq_net_private));
3389 static const struct file_operations proto_seq_fops = {
3390 .owner = THIS_MODULE,
3391 .open = proto_seq_open,
3393 .llseek = seq_lseek,
3394 .release = seq_release_net,
3397 static __net_init int proto_init_net(struct net *net)
3399 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
3405 static __net_exit void proto_exit_net(struct net *net)
3407 remove_proc_entry("protocols", net->proc_net);
3411 static __net_initdata struct pernet_operations proto_net_ops = {
3412 .init = proto_init_net,
3413 .exit = proto_exit_net,
3416 static int __init proto_init(void)
3418 return register_pernet_subsys(&proto_net_ops);
3421 subsys_initcall(proto_init);
3423 #endif /* PROC_FS */
3425 #ifdef CONFIG_NET_RX_BUSY_POLL
3426 bool sk_busy_loop_end(void *p, unsigned long start_time)
3428 struct sock *sk = p;
3430 return !skb_queue_empty_lockless(&sk->sk_receive_queue) ||
3431 sk_busy_loop_timeout(sk, start_time);
3433 EXPORT_SYMBOL(sk_busy_loop_end);
3434 #endif /* CONFIG_NET_RX_BUSY_POLL */