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 /* Take into consideration the size of the struct sk_buff overhead in the
311 * determination of these values, since that is non-constant across
312 * platforms. This makes socket queueing behavior and performance
313 * not depend upon such differences.
315 #define _SK_MEM_PACKETS 256
316 #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
317 #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
318 #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
320 /* Run time adjustable parameters. */
321 __u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
322 EXPORT_SYMBOL(sysctl_wmem_max);
323 __u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
324 EXPORT_SYMBOL(sysctl_rmem_max);
325 __u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
326 __u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
328 /* Maximal space eaten by iovec or ancillary data plus some space */
329 int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
330 EXPORT_SYMBOL(sysctl_optmem_max);
332 int sysctl_tstamp_allow_data __read_mostly = 1;
334 struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
335 EXPORT_SYMBOL_GPL(memalloc_socks);
338 * sk_set_memalloc - sets %SOCK_MEMALLOC
339 * @sk: socket to set it on
341 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
342 * It's the responsibility of the admin to adjust min_free_kbytes
343 * to meet the requirements
345 void sk_set_memalloc(struct sock *sk)
347 sock_set_flag(sk, SOCK_MEMALLOC);
348 sk->sk_allocation |= __GFP_MEMALLOC;
349 static_key_slow_inc(&memalloc_socks);
351 EXPORT_SYMBOL_GPL(sk_set_memalloc);
353 void sk_clear_memalloc(struct sock *sk)
355 sock_reset_flag(sk, SOCK_MEMALLOC);
356 sk->sk_allocation &= ~__GFP_MEMALLOC;
357 static_key_slow_dec(&memalloc_socks);
360 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
361 * progress of swapping. SOCK_MEMALLOC may be cleared while
362 * it has rmem allocations due to the last swapfile being deactivated
363 * but there is a risk that the socket is unusable due to exceeding
364 * the rmem limits. Reclaim the reserves and obey rmem limits again.
368 EXPORT_SYMBOL_GPL(sk_clear_memalloc);
370 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
373 unsigned int noreclaim_flag;
375 /* these should have been dropped before queueing */
376 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
378 noreclaim_flag = memalloc_noreclaim_save();
379 ret = sk->sk_backlog_rcv(sk, skb);
380 memalloc_noreclaim_restore(noreclaim_flag);
384 EXPORT_SYMBOL(__sk_backlog_rcv);
386 static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
390 if (optlen < sizeof(tv))
392 if (copy_from_user(&tv, optval, sizeof(tv)))
394 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
398 static int warned __read_mostly;
401 if (warned < 10 && net_ratelimit()) {
403 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
404 __func__, current->comm, task_pid_nr(current));
408 *timeo_p = MAX_SCHEDULE_TIMEOUT;
409 if (tv.tv_sec == 0 && tv.tv_usec == 0)
411 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
412 *timeo_p = tv.tv_sec * HZ + DIV_ROUND_UP(tv.tv_usec, USEC_PER_SEC / HZ);
416 static void sock_warn_obsolete_bsdism(const char *name)
419 static char warncomm[TASK_COMM_LEN];
420 if (strcmp(warncomm, current->comm) && warned < 5) {
421 strcpy(warncomm, current->comm);
422 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
428 static bool sock_needs_netstamp(const struct sock *sk)
430 switch (sk->sk_family) {
439 static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
441 if (sk->sk_flags & flags) {
442 sk->sk_flags &= ~flags;
443 if (sock_needs_netstamp(sk) &&
444 !(sk->sk_flags & SK_FLAGS_TIMESTAMP))
445 net_disable_timestamp();
450 int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
453 struct sk_buff_head *list = &sk->sk_receive_queue;
455 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
456 atomic_inc(&sk->sk_drops);
457 trace_sock_rcvqueue_full(sk, skb);
461 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
462 atomic_inc(&sk->sk_drops);
467 skb_set_owner_r(skb, sk);
469 /* we escape from rcu protected region, make sure we dont leak
474 spin_lock_irqsave(&list->lock, flags);
475 sock_skb_set_dropcount(sk, skb);
476 __skb_queue_tail(list, skb);
477 spin_unlock_irqrestore(&list->lock, flags);
479 if (!sock_flag(sk, SOCK_DEAD))
480 sk->sk_data_ready(sk);
483 EXPORT_SYMBOL(__sock_queue_rcv_skb);
485 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
489 err = sk_filter(sk, skb);
493 return __sock_queue_rcv_skb(sk, skb);
495 EXPORT_SYMBOL(sock_queue_rcv_skb);
497 int __sk_receive_skb(struct sock *sk, struct sk_buff *skb,
498 const int nested, unsigned int trim_cap, bool refcounted)
500 int rc = NET_RX_SUCCESS;
502 if (sk_filter_trim_cap(sk, skb, trim_cap))
503 goto discard_and_relse;
507 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
508 atomic_inc(&sk->sk_drops);
509 goto discard_and_relse;
512 bh_lock_sock_nested(sk);
515 if (!sock_owned_by_user(sk)) {
517 * trylock + unlock semantics:
519 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
521 rc = sk_backlog_rcv(sk, skb);
523 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
524 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
526 atomic_inc(&sk->sk_drops);
527 goto discard_and_relse;
539 EXPORT_SYMBOL(__sk_receive_skb);
541 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
543 struct dst_entry *dst = __sk_dst_get(sk);
545 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
546 sk_tx_queue_clear(sk);
547 sk->sk_dst_pending_confirm = 0;
548 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
555 EXPORT_SYMBOL(__sk_dst_check);
557 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
559 struct dst_entry *dst = sk_dst_get(sk);
561 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
569 EXPORT_SYMBOL(sk_dst_check);
571 static int sock_setbindtodevice(struct sock *sk, char __user *optval,
574 int ret = -ENOPROTOOPT;
575 #ifdef CONFIG_NETDEVICES
576 struct net *net = sock_net(sk);
577 char devname[IFNAMSIZ];
582 if (!ns_capable(net->user_ns, CAP_NET_RAW))
589 /* Bind this socket to a particular device like "eth0",
590 * as specified in the passed interface name. If the
591 * name is "" or the option length is zero the socket
594 if (optlen > IFNAMSIZ - 1)
595 optlen = IFNAMSIZ - 1;
596 memset(devname, 0, sizeof(devname));
599 if (copy_from_user(devname, optval, optlen))
603 if (devname[0] != '\0') {
604 struct net_device *dev;
607 dev = dev_get_by_name_rcu(net, devname);
609 index = dev->ifindex;
617 sk->sk_bound_dev_if = index;
629 static int sock_getbindtodevice(struct sock *sk, char __user *optval,
630 int __user *optlen, int len)
632 int ret = -ENOPROTOOPT;
633 #ifdef CONFIG_NETDEVICES
634 struct net *net = sock_net(sk);
635 char devname[IFNAMSIZ];
637 if (sk->sk_bound_dev_if == 0) {
646 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
650 len = strlen(devname) + 1;
653 if (copy_to_user(optval, devname, len))
658 if (put_user(len, optlen))
669 static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
672 sock_set_flag(sk, bit);
674 sock_reset_flag(sk, bit);
677 bool sk_mc_loop(struct sock *sk)
679 if (dev_recursion_level())
683 switch (sk->sk_family) {
685 return inet_sk(sk)->mc_loop;
686 #if IS_ENABLED(CONFIG_IPV6)
688 return inet6_sk(sk)->mc_loop;
694 EXPORT_SYMBOL(sk_mc_loop);
697 * This is meant for all protocols to use and covers goings on
698 * at the socket level. Everything here is generic.
701 int sock_setsockopt(struct socket *sock, int level, int optname,
702 char __user *optval, unsigned int optlen)
704 struct sock *sk = sock->sk;
711 * Options without arguments
714 if (optname == SO_BINDTODEVICE)
715 return sock_setbindtodevice(sk, optval, optlen);
717 if (optlen < sizeof(int))
720 if (get_user(val, (int __user *)optval))
723 valbool = val ? 1 : 0;
729 if (val && !capable(CAP_NET_ADMIN))
732 sock_valbool_flag(sk, SOCK_DBG, valbool);
735 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
738 sk->sk_reuseport = valbool;
747 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
750 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
753 /* Don't error on this BSD doesn't and if you think
754 * about it this is right. Otherwise apps have to
755 * play 'guess the biggest size' games. RCVBUF/SNDBUF
756 * are treated in BSD as hints
758 val = min_t(u32, val, sysctl_wmem_max);
760 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
761 sk->sk_sndbuf = max_t(int, val * 2, SOCK_MIN_SNDBUF);
762 /* Wake up sending tasks if we upped the value. */
763 sk->sk_write_space(sk);
767 if (!capable(CAP_NET_ADMIN)) {
774 /* Don't error on this BSD doesn't and if you think
775 * about it this is right. Otherwise apps have to
776 * play 'guess the biggest size' games. RCVBUF/SNDBUF
777 * are treated in BSD as hints
779 val = min_t(u32, val, sysctl_rmem_max);
781 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
783 * We double it on the way in to account for
784 * "struct sk_buff" etc. overhead. Applications
785 * assume that the SO_RCVBUF setting they make will
786 * allow that much actual data to be received on that
789 * Applications are unaware that "struct sk_buff" and
790 * other overheads allocate from the receive buffer
791 * during socket buffer allocation.
793 * And after considering the possible alternatives,
794 * returning the value we actually used in getsockopt
795 * is the most desirable behavior.
797 sk->sk_rcvbuf = max_t(int, val * 2, SOCK_MIN_RCVBUF);
801 if (!capable(CAP_NET_ADMIN)) {
808 if (sk->sk_prot->keepalive)
809 sk->sk_prot->keepalive(sk, valbool);
810 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
814 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
818 sk->sk_no_check_tx = valbool;
822 if ((val >= 0 && val <= 6) ||
823 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
824 sk->sk_priority = val;
830 if (optlen < sizeof(ling)) {
831 ret = -EINVAL; /* 1003.1g */
834 if (copy_from_user(&ling, optval, sizeof(ling))) {
839 sock_reset_flag(sk, SOCK_LINGER);
841 #if (BITS_PER_LONG == 32)
842 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
843 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
846 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
847 sock_set_flag(sk, SOCK_LINGER);
852 sock_warn_obsolete_bsdism("setsockopt");
857 set_bit(SOCK_PASSCRED, &sock->flags);
859 clear_bit(SOCK_PASSCRED, &sock->flags);
865 if (optname == SO_TIMESTAMP)
866 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
868 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
869 sock_set_flag(sk, SOCK_RCVTSTAMP);
870 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
872 sock_reset_flag(sk, SOCK_RCVTSTAMP);
873 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
877 case SO_TIMESTAMPING:
878 if (val & ~SOF_TIMESTAMPING_MASK) {
883 if (val & SOF_TIMESTAMPING_OPT_ID &&
884 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
885 if (sk->sk_protocol == IPPROTO_TCP &&
886 sk->sk_type == SOCK_STREAM) {
887 if ((1 << sk->sk_state) &
888 (TCPF_CLOSE | TCPF_LISTEN)) {
892 sk->sk_tskey = tcp_sk(sk)->snd_una;
898 if (val & SOF_TIMESTAMPING_OPT_STATS &&
899 !(val & SOF_TIMESTAMPING_OPT_TSONLY)) {
904 sk->sk_tsflags = val;
905 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
906 sock_enable_timestamp(sk,
907 SOCK_TIMESTAMPING_RX_SOFTWARE);
909 sock_disable_timestamp(sk,
910 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
916 sk->sk_rcvlowat = val ? : 1;
920 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
924 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
927 case SO_ATTACH_FILTER:
929 if (optlen == sizeof(struct sock_fprog)) {
930 struct sock_fprog fprog;
933 if (copy_from_user(&fprog, optval, sizeof(fprog)))
936 ret = sk_attach_filter(&fprog, sk);
942 if (optlen == sizeof(u32)) {
946 if (copy_from_user(&ufd, optval, sizeof(ufd)))
949 ret = sk_attach_bpf(ufd, sk);
953 case SO_ATTACH_REUSEPORT_CBPF:
955 if (optlen == sizeof(struct sock_fprog)) {
956 struct sock_fprog fprog;
959 if (copy_from_user(&fprog, optval, sizeof(fprog)))
962 ret = sk_reuseport_attach_filter(&fprog, sk);
966 case SO_ATTACH_REUSEPORT_EBPF:
968 if (optlen == sizeof(u32)) {
972 if (copy_from_user(&ufd, optval, sizeof(ufd)))
975 ret = sk_reuseport_attach_bpf(ufd, sk);
979 case SO_DETACH_FILTER:
980 ret = sk_detach_filter(sk);
984 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
987 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
992 set_bit(SOCK_PASSSEC, &sock->flags);
994 clear_bit(SOCK_PASSSEC, &sock->flags);
997 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
1004 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
1007 case SO_WIFI_STATUS:
1008 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
1012 if (sock->ops->set_peek_off)
1013 ret = sock->ops->set_peek_off(sk, val);
1019 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
1022 case SO_SELECT_ERR_QUEUE:
1023 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
1026 #ifdef CONFIG_NET_RX_BUSY_POLL
1028 /* allow unprivileged users to decrease the value */
1029 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
1035 sk->sk_ll_usec = val;
1040 case SO_MAX_PACING_RATE:
1042 cmpxchg(&sk->sk_pacing_status,
1045 sk->sk_max_pacing_rate = val;
1046 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
1047 sk->sk_max_pacing_rate);
1050 case SO_INCOMING_CPU:
1051 sk->sk_incoming_cpu = val;
1056 dst_negative_advice(sk);
1065 EXPORT_SYMBOL(sock_setsockopt);
1068 static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1069 struct ucred *ucred)
1071 ucred->pid = pid_vnr(pid);
1072 ucred->uid = ucred->gid = -1;
1074 struct user_namespace *current_ns = current_user_ns();
1076 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1077 ucred->gid = from_kgid_munged(current_ns, cred->egid);
1081 static int groups_to_user(gid_t __user *dst, const struct group_info *src)
1083 struct user_namespace *user_ns = current_user_ns();
1086 for (i = 0; i < src->ngroups; i++)
1087 if (put_user(from_kgid_munged(user_ns, src->gid[i]), dst + i))
1093 int sock_getsockopt(struct socket *sock, int level, int optname,
1094 char __user *optval, int __user *optlen)
1096 struct sock *sk = sock->sk;
1105 int lv = sizeof(int);
1108 if (get_user(len, optlen))
1113 memset(&v, 0, sizeof(v));
1117 v.val = sock_flag(sk, SOCK_DBG);
1121 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1125 v.val = sock_flag(sk, SOCK_BROADCAST);
1129 v.val = sk->sk_sndbuf;
1133 v.val = sk->sk_rcvbuf;
1137 v.val = sk->sk_reuse;
1141 v.val = sk->sk_reuseport;
1145 v.val = sock_flag(sk, SOCK_KEEPOPEN);
1149 v.val = sk->sk_type;
1153 v.val = sk->sk_protocol;
1157 v.val = sk->sk_family;
1161 v.val = -sock_error(sk);
1163 v.val = xchg(&sk->sk_err_soft, 0);
1167 v.val = sock_flag(sk, SOCK_URGINLINE);
1171 v.val = sk->sk_no_check_tx;
1175 v.val = sk->sk_priority;
1179 lv = sizeof(v.ling);
1180 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
1181 v.ling.l_linger = sk->sk_lingertime / HZ;
1185 sock_warn_obsolete_bsdism("getsockopt");
1189 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1190 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1193 case SO_TIMESTAMPNS:
1194 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
1197 case SO_TIMESTAMPING:
1198 v.val = sk->sk_tsflags;
1202 lv = sizeof(struct timeval);
1203 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1207 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1208 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * USEC_PER_SEC) / HZ;
1213 lv = sizeof(struct timeval);
1214 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1218 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1219 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * USEC_PER_SEC) / HZ;
1224 v.val = sk->sk_rcvlowat;
1232 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
1237 struct ucred peercred;
1238 if (len > sizeof(peercred))
1239 len = sizeof(peercred);
1240 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1241 if (copy_to_user(optval, &peercred, len))
1250 if (!sk->sk_peer_cred)
1253 n = sk->sk_peer_cred->group_info->ngroups;
1254 if (len < n * sizeof(gid_t)) {
1255 len = n * sizeof(gid_t);
1256 return put_user(len, optlen) ? -EFAULT : -ERANGE;
1258 len = n * sizeof(gid_t);
1260 ret = groups_to_user((gid_t __user *)optval,
1261 sk->sk_peer_cred->group_info);
1271 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1275 if (copy_to_user(optval, address, len))
1280 /* Dubious BSD thing... Probably nobody even uses it, but
1281 * the UNIX standard wants it for whatever reason... -DaveM
1284 v.val = sk->sk_state == TCP_LISTEN;
1288 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
1292 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1295 v.val = sk->sk_mark;
1299 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
1302 case SO_WIFI_STATUS:
1303 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
1307 if (!sock->ops->set_peek_off)
1310 v.val = sk->sk_peek_off;
1313 v.val = sock_flag(sk, SOCK_NOFCS);
1316 case SO_BINDTODEVICE:
1317 return sock_getbindtodevice(sk, optval, optlen, len);
1320 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1326 case SO_LOCK_FILTER:
1327 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1330 case SO_BPF_EXTENSIONS:
1331 v.val = bpf_tell_extensions();
1334 case SO_SELECT_ERR_QUEUE:
1335 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1338 #ifdef CONFIG_NET_RX_BUSY_POLL
1340 v.val = sk->sk_ll_usec;
1344 case SO_MAX_PACING_RATE:
1345 v.val = sk->sk_max_pacing_rate;
1348 case SO_INCOMING_CPU:
1349 v.val = sk->sk_incoming_cpu;
1354 u32 meminfo[SK_MEMINFO_VARS];
1356 if (get_user(len, optlen))
1359 sk_get_meminfo(sk, meminfo);
1361 len = min_t(unsigned int, len, sizeof(meminfo));
1362 if (copy_to_user(optval, &meminfo, len))
1368 #ifdef CONFIG_NET_RX_BUSY_POLL
1369 case SO_INCOMING_NAPI_ID:
1370 v.val = READ_ONCE(sk->sk_napi_id);
1372 /* aggregate non-NAPI IDs down to 0 */
1373 if (v.val < MIN_NAPI_ID)
1383 v.val64 = sock_gen_cookie(sk);
1387 /* We implement the SO_SNDLOWAT etc to not be settable
1390 return -ENOPROTOOPT;
1395 if (copy_to_user(optval, &v, len))
1398 if (put_user(len, optlen))
1404 * Initialize an sk_lock.
1406 * (We also register the sk_lock with the lock validator.)
1408 static inline void sock_lock_init(struct sock *sk)
1410 if (sk->sk_kern_sock)
1411 sock_lock_init_class_and_name(
1413 af_family_kern_slock_key_strings[sk->sk_family],
1414 af_family_kern_slock_keys + sk->sk_family,
1415 af_family_kern_key_strings[sk->sk_family],
1416 af_family_kern_keys + sk->sk_family);
1418 sock_lock_init_class_and_name(
1420 af_family_slock_key_strings[sk->sk_family],
1421 af_family_slock_keys + sk->sk_family,
1422 af_family_key_strings[sk->sk_family],
1423 af_family_keys + sk->sk_family);
1427 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1428 * even temporarly, because of RCU lookups. sk_node should also be left as is.
1429 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
1431 static void sock_copy(struct sock *nsk, const struct sock *osk)
1433 #ifdef CONFIG_SECURITY_NETWORK
1434 void *sptr = nsk->sk_security;
1436 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1438 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1439 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1441 #ifdef CONFIG_SECURITY_NETWORK
1442 nsk->sk_security = sptr;
1443 security_sk_clone(osk, nsk);
1447 static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1451 struct kmem_cache *slab;
1455 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1458 if (priority & __GFP_ZERO)
1459 sk_prot_clear_nulls(sk, prot->obj_size);
1461 sk = kmalloc(prot->obj_size, priority);
1464 kmemcheck_annotate_bitfield(sk, flags);
1466 if (security_sk_alloc(sk, family, priority))
1469 if (!try_module_get(prot->owner))
1471 sk_tx_queue_clear(sk);
1477 security_sk_free(sk);
1480 kmem_cache_free(slab, sk);
1486 static void sk_prot_free(struct proto *prot, struct sock *sk)
1488 struct kmem_cache *slab;
1489 struct module *owner;
1491 owner = prot->owner;
1494 cgroup_sk_free(&sk->sk_cgrp_data);
1495 mem_cgroup_sk_free(sk);
1496 security_sk_free(sk);
1498 kmem_cache_free(slab, sk);
1505 * sk_alloc - All socket objects are allocated here
1506 * @net: the applicable net namespace
1507 * @family: protocol family
1508 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1509 * @prot: struct proto associated with this new sock instance
1510 * @kern: is this to be a kernel socket?
1512 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1513 struct proto *prot, int kern)
1517 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1519 sk->sk_family = family;
1521 * See comment in struct sock definition to understand
1522 * why we need sk_prot_creator -acme
1524 sk->sk_prot = sk->sk_prot_creator = prot;
1525 sk->sk_kern_sock = kern;
1527 sk->sk_net_refcnt = kern ? 0 : 1;
1528 if (likely(sk->sk_net_refcnt))
1530 sock_net_set(sk, net);
1531 refcount_set(&sk->sk_wmem_alloc, 1);
1533 mem_cgroup_sk_alloc(sk);
1534 cgroup_sk_alloc(&sk->sk_cgrp_data);
1535 sock_update_classid(&sk->sk_cgrp_data);
1536 sock_update_netprioidx(&sk->sk_cgrp_data);
1541 EXPORT_SYMBOL(sk_alloc);
1543 /* Sockets having SOCK_RCU_FREE will call this function after one RCU
1544 * grace period. This is the case for UDP sockets and TCP listeners.
1546 static void __sk_destruct(struct rcu_head *head)
1548 struct sock *sk = container_of(head, struct sock, sk_rcu);
1549 struct sk_filter *filter;
1551 if (sk->sk_destruct)
1552 sk->sk_destruct(sk);
1554 filter = rcu_dereference_check(sk->sk_filter,
1555 refcount_read(&sk->sk_wmem_alloc) == 0);
1557 sk_filter_uncharge(sk, filter);
1558 RCU_INIT_POINTER(sk->sk_filter, NULL);
1560 if (rcu_access_pointer(sk->sk_reuseport_cb))
1561 reuseport_detach_sock(sk);
1563 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1565 if (atomic_read(&sk->sk_omem_alloc))
1566 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1567 __func__, atomic_read(&sk->sk_omem_alloc));
1569 if (sk->sk_frag.page) {
1570 put_page(sk->sk_frag.page);
1571 sk->sk_frag.page = NULL;
1574 if (sk->sk_peer_cred)
1575 put_cred(sk->sk_peer_cred);
1576 put_pid(sk->sk_peer_pid);
1577 if (likely(sk->sk_net_refcnt))
1578 put_net(sock_net(sk));
1579 sk_prot_free(sk->sk_prot_creator, sk);
1582 void sk_destruct(struct sock *sk)
1584 if (sock_flag(sk, SOCK_RCU_FREE))
1585 call_rcu(&sk->sk_rcu, __sk_destruct);
1587 __sk_destruct(&sk->sk_rcu);
1590 static void __sk_free(struct sock *sk)
1592 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
1593 sock_diag_broadcast_destroy(sk);
1598 void sk_free(struct sock *sk)
1601 * We subtract one from sk_wmem_alloc and can know if
1602 * some packets are still in some tx queue.
1603 * If not null, sock_wfree() will call __sk_free(sk) later
1605 if (refcount_dec_and_test(&sk->sk_wmem_alloc))
1608 EXPORT_SYMBOL(sk_free);
1610 static void sk_init_common(struct sock *sk)
1612 skb_queue_head_init(&sk->sk_receive_queue);
1613 skb_queue_head_init(&sk->sk_write_queue);
1614 skb_queue_head_init(&sk->sk_error_queue);
1616 rwlock_init(&sk->sk_callback_lock);
1617 lockdep_set_class_and_name(&sk->sk_receive_queue.lock,
1618 af_rlock_keys + sk->sk_family,
1619 af_family_rlock_key_strings[sk->sk_family]);
1620 lockdep_set_class_and_name(&sk->sk_write_queue.lock,
1621 af_wlock_keys + sk->sk_family,
1622 af_family_wlock_key_strings[sk->sk_family]);
1623 lockdep_set_class_and_name(&sk->sk_error_queue.lock,
1624 af_elock_keys + sk->sk_family,
1625 af_family_elock_key_strings[sk->sk_family]);
1626 lockdep_set_class_and_name(&sk->sk_callback_lock,
1627 af_callback_keys + sk->sk_family,
1628 af_family_clock_key_strings[sk->sk_family]);
1632 * sk_clone_lock - clone a socket, and lock its clone
1633 * @sk: the socket to clone
1634 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1636 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1638 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
1641 bool is_charged = true;
1643 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
1644 if (newsk != NULL) {
1645 struct sk_filter *filter;
1647 sock_copy(newsk, sk);
1650 if (likely(newsk->sk_net_refcnt))
1651 get_net(sock_net(newsk));
1652 sk_node_init(&newsk->sk_node);
1653 sock_lock_init(newsk);
1654 bh_lock_sock(newsk);
1655 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
1656 newsk->sk_backlog.len = 0;
1658 atomic_set(&newsk->sk_rmem_alloc, 0);
1660 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1662 refcount_set(&newsk->sk_wmem_alloc, 1);
1663 atomic_set(&newsk->sk_omem_alloc, 0);
1664 sk_init_common(newsk);
1666 newsk->sk_dst_cache = NULL;
1667 newsk->sk_dst_pending_confirm = 0;
1668 newsk->sk_wmem_queued = 0;
1669 newsk->sk_forward_alloc = 0;
1670 atomic_set(&newsk->sk_drops, 0);
1671 newsk->sk_send_head = NULL;
1672 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1674 sock_reset_flag(newsk, SOCK_DONE);
1676 filter = rcu_dereference_protected(newsk->sk_filter, 1);
1678 /* though it's an empty new sock, the charging may fail
1679 * if sysctl_optmem_max was changed between creation of
1680 * original socket and cloning
1682 is_charged = sk_filter_charge(newsk, filter);
1684 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
1685 /* We need to make sure that we don't uncharge the new
1686 * socket if we couldn't charge it in the first place
1687 * as otherwise we uncharge the parent's filter.
1690 RCU_INIT_POINTER(newsk->sk_filter, NULL);
1691 sk_free_unlock_clone(newsk);
1695 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
1698 newsk->sk_err_soft = 0;
1699 newsk->sk_priority = 0;
1700 newsk->sk_incoming_cpu = raw_smp_processor_id();
1701 atomic64_set(&newsk->sk_cookie, 0);
1703 mem_cgroup_sk_alloc(newsk);
1704 cgroup_sk_alloc(&newsk->sk_cgrp_data);
1707 * Before updating sk_refcnt, we must commit prior changes to memory
1708 * (Documentation/RCU/rculist_nulls.txt for details)
1711 refcount_set(&newsk->sk_refcnt, 2);
1714 * Increment the counter in the same struct proto as the master
1715 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1716 * is the same as sk->sk_prot->socks, as this field was copied
1719 * This _changes_ the previous behaviour, where
1720 * tcp_create_openreq_child always was incrementing the
1721 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1722 * to be taken into account in all callers. -acme
1724 sk_refcnt_debug_inc(newsk);
1725 sk_set_socket(newsk, NULL);
1726 newsk->sk_wq = NULL;
1728 if (newsk->sk_prot->sockets_allocated)
1729 sk_sockets_allocated_inc(newsk);
1731 if (sock_needs_netstamp(sk) &&
1732 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
1733 net_enable_timestamp();
1738 EXPORT_SYMBOL_GPL(sk_clone_lock);
1740 void sk_free_unlock_clone(struct sock *sk)
1742 /* It is still raw copy of parent, so invalidate
1743 * destructor and make plain sk_free() */
1744 sk->sk_destruct = NULL;
1748 EXPORT_SYMBOL_GPL(sk_free_unlock_clone);
1750 void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1754 sk_dst_set(sk, dst);
1755 sk->sk_route_caps = dst->dev->features;
1756 if (sk->sk_route_caps & NETIF_F_GSO)
1757 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
1758 sk->sk_route_caps &= ~sk->sk_route_nocaps;
1759 if (sk_can_gso(sk)) {
1760 if (dst->header_len) {
1761 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
1763 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
1764 sk->sk_gso_max_size = dst->dev->gso_max_size;
1765 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
1768 sk->sk_gso_max_segs = max_segs;
1770 EXPORT_SYMBOL_GPL(sk_setup_caps);
1773 * Simple resource managers for sockets.
1778 * Write buffer destructor automatically called from kfree_skb.
1780 void sock_wfree(struct sk_buff *skb)
1782 struct sock *sk = skb->sk;
1783 unsigned int len = skb->truesize;
1785 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1787 * Keep a reference on sk_wmem_alloc, this will be released
1788 * after sk_write_space() call
1790 WARN_ON(refcount_sub_and_test(len - 1, &sk->sk_wmem_alloc));
1791 sk->sk_write_space(sk);
1795 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1796 * could not do because of in-flight packets
1798 if (refcount_sub_and_test(len, &sk->sk_wmem_alloc))
1801 EXPORT_SYMBOL(sock_wfree);
1803 /* This variant of sock_wfree() is used by TCP,
1804 * since it sets SOCK_USE_WRITE_QUEUE.
1806 void __sock_wfree(struct sk_buff *skb)
1808 struct sock *sk = skb->sk;
1810 if (refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc))
1814 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1819 if (unlikely(!sk_fullsock(sk))) {
1820 skb->destructor = sock_edemux;
1825 skb->destructor = sock_wfree;
1826 skb_set_hash_from_sk(skb, sk);
1828 * We used to take a refcount on sk, but following operation
1829 * is enough to guarantee sk_free() wont free this sock until
1830 * all in-flight packets are completed
1832 refcount_add(skb->truesize, &sk->sk_wmem_alloc);
1834 EXPORT_SYMBOL(skb_set_owner_w);
1836 /* This helper is used by netem, as it can hold packets in its
1837 * delay queue. We want to allow the owner socket to send more
1838 * packets, as if they were already TX completed by a typical driver.
1839 * But we also want to keep skb->sk set because some packet schedulers
1840 * rely on it (sch_fq for example).
1842 void skb_orphan_partial(struct sk_buff *skb)
1844 if (skb_is_tcp_pure_ack(skb))
1847 if (skb->destructor == sock_wfree
1849 || skb->destructor == tcp_wfree
1852 struct sock *sk = skb->sk;
1854 if (refcount_inc_not_zero(&sk->sk_refcnt)) {
1855 WARN_ON(refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc));
1856 skb->destructor = sock_efree;
1862 EXPORT_SYMBOL(skb_orphan_partial);
1865 * Read buffer destructor automatically called from kfree_skb.
1867 void sock_rfree(struct sk_buff *skb)
1869 struct sock *sk = skb->sk;
1870 unsigned int len = skb->truesize;
1872 atomic_sub(len, &sk->sk_rmem_alloc);
1873 sk_mem_uncharge(sk, len);
1875 EXPORT_SYMBOL(sock_rfree);
1878 * Buffer destructor for skbs that are not used directly in read or write
1879 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1881 void sock_efree(struct sk_buff *skb)
1885 EXPORT_SYMBOL(sock_efree);
1887 kuid_t sock_i_uid(struct sock *sk)
1891 read_lock_bh(&sk->sk_callback_lock);
1892 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
1893 read_unlock_bh(&sk->sk_callback_lock);
1896 EXPORT_SYMBOL(sock_i_uid);
1898 unsigned long sock_i_ino(struct sock *sk)
1902 read_lock_bh(&sk->sk_callback_lock);
1903 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
1904 read_unlock_bh(&sk->sk_callback_lock);
1907 EXPORT_SYMBOL(sock_i_ino);
1910 * Allocate a skb from the socket's send buffer.
1912 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1915 if (force || refcount_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
1916 struct sk_buff *skb = alloc_skb(size, priority);
1918 skb_set_owner_w(skb, sk);
1924 EXPORT_SYMBOL(sock_wmalloc);
1927 * Allocate a memory block from the socket's option memory buffer.
1929 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1931 if ((unsigned int)size <= sysctl_optmem_max &&
1932 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1934 /* First do the add, to avoid the race if kmalloc
1937 atomic_add(size, &sk->sk_omem_alloc);
1938 mem = kmalloc(size, priority);
1941 atomic_sub(size, &sk->sk_omem_alloc);
1945 EXPORT_SYMBOL(sock_kmalloc);
1947 /* Free an option memory block. Note, we actually want the inline
1948 * here as this allows gcc to detect the nullify and fold away the
1949 * condition entirely.
1951 static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1954 if (WARN_ON_ONCE(!mem))
1960 atomic_sub(size, &sk->sk_omem_alloc);
1963 void sock_kfree_s(struct sock *sk, void *mem, int size)
1965 __sock_kfree_s(sk, mem, size, false);
1967 EXPORT_SYMBOL(sock_kfree_s);
1969 void sock_kzfree_s(struct sock *sk, void *mem, int size)
1971 __sock_kfree_s(sk, mem, size, true);
1973 EXPORT_SYMBOL(sock_kzfree_s);
1975 /* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1976 I think, these locks should be removed for datagram sockets.
1978 static long sock_wait_for_wmem(struct sock *sk, long timeo)
1982 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
1986 if (signal_pending(current))
1988 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1989 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1990 if (refcount_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1992 if (sk->sk_shutdown & SEND_SHUTDOWN)
1996 timeo = schedule_timeout(timeo);
1998 finish_wait(sk_sleep(sk), &wait);
2004 * Generic send/receive buffer handlers
2007 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
2008 unsigned long data_len, int noblock,
2009 int *errcode, int max_page_order)
2011 struct sk_buff *skb;
2015 timeo = sock_sndtimeo(sk, noblock);
2017 err = sock_error(sk);
2022 if (sk->sk_shutdown & SEND_SHUTDOWN)
2025 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
2028 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2029 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2033 if (signal_pending(current))
2035 timeo = sock_wait_for_wmem(sk, timeo);
2037 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
2038 errcode, sk->sk_allocation);
2040 skb_set_owner_w(skb, sk);
2044 err = sock_intr_errno(timeo);
2049 EXPORT_SYMBOL(sock_alloc_send_pskb);
2051 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
2052 int noblock, int *errcode)
2054 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
2056 EXPORT_SYMBOL(sock_alloc_send_skb);
2058 int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
2059 struct sockcm_cookie *sockc)
2063 switch (cmsg->cmsg_type) {
2065 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2067 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2069 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
2071 case SO_TIMESTAMPING:
2072 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2075 tsflags = *(u32 *)CMSG_DATA(cmsg);
2076 if (tsflags & ~SOF_TIMESTAMPING_TX_RECORD_MASK)
2079 sockc->tsflags &= ~SOF_TIMESTAMPING_TX_RECORD_MASK;
2080 sockc->tsflags |= tsflags;
2082 /* SCM_RIGHTS and SCM_CREDENTIALS are semantically in SOL_UNIX. */
2084 case SCM_CREDENTIALS:
2091 EXPORT_SYMBOL(__sock_cmsg_send);
2093 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
2094 struct sockcm_cookie *sockc)
2096 struct cmsghdr *cmsg;
2099 for_each_cmsghdr(cmsg, msg) {
2100 if (!CMSG_OK(msg, cmsg))
2102 if (cmsg->cmsg_level != SOL_SOCKET)
2104 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
2110 EXPORT_SYMBOL(sock_cmsg_send);
2112 static void sk_enter_memory_pressure(struct sock *sk)
2114 if (!sk->sk_prot->enter_memory_pressure)
2117 sk->sk_prot->enter_memory_pressure(sk);
2120 static void sk_leave_memory_pressure(struct sock *sk)
2122 if (sk->sk_prot->leave_memory_pressure) {
2123 sk->sk_prot->leave_memory_pressure(sk);
2125 unsigned long *memory_pressure = sk->sk_prot->memory_pressure;
2127 if (memory_pressure && *memory_pressure)
2128 *memory_pressure = 0;
2132 /* On 32bit arches, an skb frag is limited to 2^15 */
2133 #define SKB_FRAG_PAGE_ORDER get_order(32768)
2136 * skb_page_frag_refill - check that a page_frag contains enough room
2137 * @sz: minimum size of the fragment we want to get
2138 * @pfrag: pointer to page_frag
2139 * @gfp: priority for memory allocation
2141 * Note: While this allocator tries to use high order pages, there is
2142 * no guarantee that allocations succeed. Therefore, @sz MUST be
2143 * less or equal than PAGE_SIZE.
2145 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
2148 if (page_ref_count(pfrag->page) == 1) {
2152 if (pfrag->offset + sz <= pfrag->size)
2154 put_page(pfrag->page);
2158 if (SKB_FRAG_PAGE_ORDER) {
2159 /* Avoid direct reclaim but allow kswapd to wake */
2160 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
2161 __GFP_COMP | __GFP_NOWARN |
2163 SKB_FRAG_PAGE_ORDER);
2164 if (likely(pfrag->page)) {
2165 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
2169 pfrag->page = alloc_page(gfp);
2170 if (likely(pfrag->page)) {
2171 pfrag->size = PAGE_SIZE;
2176 EXPORT_SYMBOL(skb_page_frag_refill);
2178 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
2180 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
2183 sk_enter_memory_pressure(sk);
2184 sk_stream_moderate_sndbuf(sk);
2187 EXPORT_SYMBOL(sk_page_frag_refill);
2189 static void __lock_sock(struct sock *sk)
2190 __releases(&sk->sk_lock.slock)
2191 __acquires(&sk->sk_lock.slock)
2196 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
2197 TASK_UNINTERRUPTIBLE);
2198 spin_unlock_bh(&sk->sk_lock.slock);
2200 spin_lock_bh(&sk->sk_lock.slock);
2201 if (!sock_owned_by_user(sk))
2204 finish_wait(&sk->sk_lock.wq, &wait);
2207 static void __release_sock(struct sock *sk)
2208 __releases(&sk->sk_lock.slock)
2209 __acquires(&sk->sk_lock.slock)
2211 struct sk_buff *skb, *next;
2213 while ((skb = sk->sk_backlog.head) != NULL) {
2214 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
2216 spin_unlock_bh(&sk->sk_lock.slock);
2221 WARN_ON_ONCE(skb_dst_is_noref(skb));
2223 sk_backlog_rcv(sk, skb);
2228 } while (skb != NULL);
2230 spin_lock_bh(&sk->sk_lock.slock);
2234 * Doing the zeroing here guarantee we can not loop forever
2235 * while a wild producer attempts to flood us.
2237 sk->sk_backlog.len = 0;
2240 void __sk_flush_backlog(struct sock *sk)
2242 spin_lock_bh(&sk->sk_lock.slock);
2244 spin_unlock_bh(&sk->sk_lock.slock);
2248 * sk_wait_data - wait for data to arrive at sk_receive_queue
2249 * @sk: sock to wait on
2250 * @timeo: for how long
2251 * @skb: last skb seen on sk_receive_queue
2253 * Now socket state including sk->sk_err is changed only under lock,
2254 * hence we may omit checks after joining wait queue.
2255 * We check receive queue before schedule() only as optimization;
2256 * it is very likely that release_sock() added new data.
2258 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
2260 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2263 add_wait_queue(sk_sleep(sk), &wait);
2264 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
2265 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb, &wait);
2266 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
2267 remove_wait_queue(sk_sleep(sk), &wait);
2270 EXPORT_SYMBOL(sk_wait_data);
2273 * __sk_mem_raise_allocated - increase memory_allocated
2275 * @size: memory size to allocate
2276 * @amt: pages to allocate
2277 * @kind: allocation type
2279 * Similar to __sk_mem_schedule(), but does not update sk_forward_alloc
2281 int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
2283 struct proto *prot = sk->sk_prot;
2284 long allocated = sk_memory_allocated_add(sk, amt);
2286 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2287 !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
2288 goto suppress_allocation;
2291 if (allocated <= sk_prot_mem_limits(sk, 0)) {
2292 sk_leave_memory_pressure(sk);
2296 /* Under pressure. */
2297 if (allocated > sk_prot_mem_limits(sk, 1))
2298 sk_enter_memory_pressure(sk);
2300 /* Over hard limit. */
2301 if (allocated > sk_prot_mem_limits(sk, 2))
2302 goto suppress_allocation;
2304 /* guarantee minimum buffer size under pressure */
2305 if (kind == SK_MEM_RECV) {
2306 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2309 } else { /* SK_MEM_SEND */
2310 if (sk->sk_type == SOCK_STREAM) {
2311 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2313 } else if (refcount_read(&sk->sk_wmem_alloc) <
2314 prot->sysctl_wmem[0])
2318 if (sk_has_memory_pressure(sk)) {
2321 if (!sk_under_memory_pressure(sk))
2323 alloc = sk_sockets_allocated_read_positive(sk);
2324 if (sk_prot_mem_limits(sk, 2) > alloc *
2325 sk_mem_pages(sk->sk_wmem_queued +
2326 atomic_read(&sk->sk_rmem_alloc) +
2327 sk->sk_forward_alloc))
2331 suppress_allocation:
2333 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2334 sk_stream_moderate_sndbuf(sk);
2336 /* Fail only if socket is _under_ its sndbuf.
2337 * In this case we cannot block, so that we have to fail.
2339 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2343 trace_sock_exceed_buf_limit(sk, prot, allocated);
2345 sk_memory_allocated_sub(sk, amt);
2347 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2348 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
2352 EXPORT_SYMBOL(__sk_mem_raise_allocated);
2355 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2357 * @size: memory size to allocate
2358 * @kind: allocation type
2360 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2361 * rmem allocation. This function assumes that protocols which have
2362 * memory_pressure use sk_wmem_queued as write buffer accounting.
2364 int __sk_mem_schedule(struct sock *sk, int size, int kind)
2366 int ret, amt = sk_mem_pages(size);
2368 sk->sk_forward_alloc += amt << SK_MEM_QUANTUM_SHIFT;
2369 ret = __sk_mem_raise_allocated(sk, size, amt, kind);
2371 sk->sk_forward_alloc -= amt << SK_MEM_QUANTUM_SHIFT;
2374 EXPORT_SYMBOL(__sk_mem_schedule);
2377 * __sk_mem_reduce_allocated - reclaim memory_allocated
2379 * @amount: number of quanta
2381 * Similar to __sk_mem_reclaim(), but does not update sk_forward_alloc
2383 void __sk_mem_reduce_allocated(struct sock *sk, int amount)
2385 sk_memory_allocated_sub(sk, amount);
2387 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2388 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
2390 if (sk_under_memory_pressure(sk) &&
2391 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2392 sk_leave_memory_pressure(sk);
2394 EXPORT_SYMBOL(__sk_mem_reduce_allocated);
2397 * __sk_mem_reclaim - reclaim sk_forward_alloc and memory_allocated
2399 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
2401 void __sk_mem_reclaim(struct sock *sk, int amount)
2403 amount >>= SK_MEM_QUANTUM_SHIFT;
2404 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
2405 __sk_mem_reduce_allocated(sk, amount);
2407 EXPORT_SYMBOL(__sk_mem_reclaim);
2409 int sk_set_peek_off(struct sock *sk, int val)
2414 sk->sk_peek_off = val;
2417 EXPORT_SYMBOL_GPL(sk_set_peek_off);
2420 * Set of default routines for initialising struct proto_ops when
2421 * the protocol does not support a particular function. In certain
2422 * cases where it makes no sense for a protocol to have a "do nothing"
2423 * function, some default processing is provided.
2426 int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2430 EXPORT_SYMBOL(sock_no_bind);
2432 int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
2437 EXPORT_SYMBOL(sock_no_connect);
2439 int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2443 EXPORT_SYMBOL(sock_no_socketpair);
2445 int sock_no_accept(struct socket *sock, struct socket *newsock, int flags,
2450 EXPORT_SYMBOL(sock_no_accept);
2452 int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
2457 EXPORT_SYMBOL(sock_no_getname);
2459 unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
2463 EXPORT_SYMBOL(sock_no_poll);
2465 int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2469 EXPORT_SYMBOL(sock_no_ioctl);
2471 int sock_no_listen(struct socket *sock, int backlog)
2475 EXPORT_SYMBOL(sock_no_listen);
2477 int sock_no_shutdown(struct socket *sock, int how)
2481 EXPORT_SYMBOL(sock_no_shutdown);
2483 int sock_no_setsockopt(struct socket *sock, int level, int optname,
2484 char __user *optval, unsigned int optlen)
2488 EXPORT_SYMBOL(sock_no_setsockopt);
2490 int sock_no_getsockopt(struct socket *sock, int level, int optname,
2491 char __user *optval, int __user *optlen)
2495 EXPORT_SYMBOL(sock_no_getsockopt);
2497 int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
2501 EXPORT_SYMBOL(sock_no_sendmsg);
2503 int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2508 EXPORT_SYMBOL(sock_no_recvmsg);
2510 int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2512 /* Mirror missing mmap method error code */
2515 EXPORT_SYMBOL(sock_no_mmap);
2517 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2520 struct msghdr msg = {.msg_flags = flags};
2522 char *kaddr = kmap(page);
2523 iov.iov_base = kaddr + offset;
2525 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2529 EXPORT_SYMBOL(sock_no_sendpage);
2532 * Default Socket Callbacks
2535 static void sock_def_wakeup(struct sock *sk)
2537 struct socket_wq *wq;
2540 wq = rcu_dereference(sk->sk_wq);
2541 if (skwq_has_sleeper(wq))
2542 wake_up_interruptible_all(&wq->wait);
2546 static void sock_def_error_report(struct sock *sk)
2548 struct socket_wq *wq;
2551 wq = rcu_dereference(sk->sk_wq);
2552 if (skwq_has_sleeper(wq))
2553 wake_up_interruptible_poll(&wq->wait, POLLERR);
2554 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
2558 static void sock_def_readable(struct sock *sk)
2560 struct socket_wq *wq;
2563 wq = rcu_dereference(sk->sk_wq);
2564 if (skwq_has_sleeper(wq))
2565 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
2566 POLLRDNORM | POLLRDBAND);
2567 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
2571 static void sock_def_write_space(struct sock *sk)
2573 struct socket_wq *wq;
2577 /* Do not wake up a writer until he can make "significant"
2580 if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
2581 wq = rcu_dereference(sk->sk_wq);
2582 if (skwq_has_sleeper(wq))
2583 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
2584 POLLWRNORM | POLLWRBAND);
2586 /* Should agree with poll, otherwise some programs break */
2587 if (sock_writeable(sk))
2588 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
2594 static void sock_def_destruct(struct sock *sk)
2598 void sk_send_sigurg(struct sock *sk)
2600 if (sk->sk_socket && sk->sk_socket->file)
2601 if (send_sigurg(&sk->sk_socket->file->f_owner))
2602 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
2604 EXPORT_SYMBOL(sk_send_sigurg);
2606 void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2607 unsigned long expires)
2609 if (!mod_timer(timer, expires))
2612 EXPORT_SYMBOL(sk_reset_timer);
2614 void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2616 if (del_timer(timer))
2619 EXPORT_SYMBOL(sk_stop_timer);
2621 void sock_init_data(struct socket *sock, struct sock *sk)
2624 sk->sk_send_head = NULL;
2626 init_timer(&sk->sk_timer);
2628 sk->sk_allocation = GFP_KERNEL;
2629 sk->sk_rcvbuf = sysctl_rmem_default;
2630 sk->sk_sndbuf = sysctl_wmem_default;
2631 sk->sk_state = TCP_CLOSE;
2632 sk_set_socket(sk, sock);
2634 sock_set_flag(sk, SOCK_ZAPPED);
2637 sk->sk_type = sock->type;
2638 sk->sk_wq = sock->wq;
2640 sk->sk_uid = SOCK_INODE(sock)->i_uid;
2643 sk->sk_uid = make_kuid(sock_net(sk)->user_ns, 0);
2646 rwlock_init(&sk->sk_callback_lock);
2647 if (sk->sk_kern_sock)
2648 lockdep_set_class_and_name(
2649 &sk->sk_callback_lock,
2650 af_kern_callback_keys + sk->sk_family,
2651 af_family_kern_clock_key_strings[sk->sk_family]);
2653 lockdep_set_class_and_name(
2654 &sk->sk_callback_lock,
2655 af_callback_keys + sk->sk_family,
2656 af_family_clock_key_strings[sk->sk_family]);
2658 sk->sk_state_change = sock_def_wakeup;
2659 sk->sk_data_ready = sock_def_readable;
2660 sk->sk_write_space = sock_def_write_space;
2661 sk->sk_error_report = sock_def_error_report;
2662 sk->sk_destruct = sock_def_destruct;
2664 sk->sk_frag.page = NULL;
2665 sk->sk_frag.offset = 0;
2666 sk->sk_peek_off = -1;
2668 sk->sk_peer_pid = NULL;
2669 sk->sk_peer_cred = NULL;
2670 sk->sk_write_pending = 0;
2671 sk->sk_rcvlowat = 1;
2672 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2673 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2675 sk->sk_stamp = SK_DEFAULT_STAMP;
2677 #ifdef CONFIG_NET_RX_BUSY_POLL
2679 sk->sk_ll_usec = sysctl_net_busy_read;
2682 sk->sk_max_pacing_rate = ~0U;
2683 sk->sk_pacing_rate = ~0U;
2684 sk->sk_incoming_cpu = -1;
2686 * Before updating sk_refcnt, we must commit prior changes to memory
2687 * (Documentation/RCU/rculist_nulls.txt for details)
2690 refcount_set(&sk->sk_refcnt, 1);
2691 atomic_set(&sk->sk_drops, 0);
2693 EXPORT_SYMBOL(sock_init_data);
2695 void lock_sock_nested(struct sock *sk, int subclass)
2698 spin_lock_bh(&sk->sk_lock.slock);
2699 if (sk->sk_lock.owned)
2701 sk->sk_lock.owned = 1;
2702 spin_unlock(&sk->sk_lock.slock);
2704 * The sk_lock has mutex_lock() semantics here:
2706 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
2709 EXPORT_SYMBOL(lock_sock_nested);
2711 void release_sock(struct sock *sk)
2713 spin_lock_bh(&sk->sk_lock.slock);
2714 if (sk->sk_backlog.tail)
2717 /* Warning : release_cb() might need to release sk ownership,
2718 * ie call sock_release_ownership(sk) before us.
2720 if (sk->sk_prot->release_cb)
2721 sk->sk_prot->release_cb(sk);
2723 sock_release_ownership(sk);
2724 if (waitqueue_active(&sk->sk_lock.wq))
2725 wake_up(&sk->sk_lock.wq);
2726 spin_unlock_bh(&sk->sk_lock.slock);
2728 EXPORT_SYMBOL(release_sock);
2731 * lock_sock_fast - fast version of lock_sock
2734 * This version should be used for very small section, where process wont block
2735 * return false if fast path is taken:
2737 * sk_lock.slock locked, owned = 0, BH disabled
2739 * return true if slow path is taken:
2741 * sk_lock.slock unlocked, owned = 1, BH enabled
2743 bool lock_sock_fast(struct sock *sk)
2746 spin_lock_bh(&sk->sk_lock.slock);
2748 if (!sk->sk_lock.owned)
2750 * Note : We must disable BH
2755 sk->sk_lock.owned = 1;
2756 spin_unlock(&sk->sk_lock.slock);
2758 * The sk_lock has mutex_lock() semantics here:
2760 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2764 EXPORT_SYMBOL(lock_sock_fast);
2766 int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
2769 if (!sock_flag(sk, SOCK_TIMESTAMP))
2770 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
2771 tv = ktime_to_timeval(sk->sk_stamp);
2772 if (tv.tv_sec == -1)
2774 if (tv.tv_sec == 0) {
2775 sk->sk_stamp = ktime_get_real();
2776 tv = ktime_to_timeval(sk->sk_stamp);
2778 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
2780 EXPORT_SYMBOL(sock_get_timestamp);
2782 int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2785 if (!sock_flag(sk, SOCK_TIMESTAMP))
2786 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
2787 ts = ktime_to_timespec(sk->sk_stamp);
2788 if (ts.tv_sec == -1)
2790 if (ts.tv_sec == 0) {
2791 sk->sk_stamp = ktime_get_real();
2792 ts = ktime_to_timespec(sk->sk_stamp);
2794 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2796 EXPORT_SYMBOL(sock_get_timestampns);
2798 void sock_enable_timestamp(struct sock *sk, int flag)
2800 if (!sock_flag(sk, flag)) {
2801 unsigned long previous_flags = sk->sk_flags;
2803 sock_set_flag(sk, flag);
2805 * we just set one of the two flags which require net
2806 * time stamping, but time stamping might have been on
2807 * already because of the other one
2809 if (sock_needs_netstamp(sk) &&
2810 !(previous_flags & SK_FLAGS_TIMESTAMP))
2811 net_enable_timestamp();
2815 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2816 int level, int type)
2818 struct sock_exterr_skb *serr;
2819 struct sk_buff *skb;
2823 skb = sock_dequeue_err_skb(sk);
2829 msg->msg_flags |= MSG_TRUNC;
2832 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2836 sock_recv_timestamp(msg, sk, skb);
2838 serr = SKB_EXT_ERR(skb);
2839 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2841 msg->msg_flags |= MSG_ERRQUEUE;
2849 EXPORT_SYMBOL(sock_recv_errqueue);
2852 * Get a socket option on an socket.
2854 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2855 * asynchronous errors should be reported by getsockopt. We assume
2856 * this means if you specify SO_ERROR (otherwise whats the point of it).
2858 int sock_common_getsockopt(struct socket *sock, int level, int optname,
2859 char __user *optval, int __user *optlen)
2861 struct sock *sk = sock->sk;
2863 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2865 EXPORT_SYMBOL(sock_common_getsockopt);
2867 #ifdef CONFIG_COMPAT
2868 int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2869 char __user *optval, int __user *optlen)
2871 struct sock *sk = sock->sk;
2873 if (sk->sk_prot->compat_getsockopt != NULL)
2874 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2876 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2878 EXPORT_SYMBOL(compat_sock_common_getsockopt);
2881 int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2884 struct sock *sk = sock->sk;
2888 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
2889 flags & ~MSG_DONTWAIT, &addr_len);
2891 msg->msg_namelen = addr_len;
2894 EXPORT_SYMBOL(sock_common_recvmsg);
2897 * Set socket options on an inet socket.
2899 int sock_common_setsockopt(struct socket *sock, int level, int optname,
2900 char __user *optval, unsigned int optlen)
2902 struct sock *sk = sock->sk;
2904 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2906 EXPORT_SYMBOL(sock_common_setsockopt);
2908 #ifdef CONFIG_COMPAT
2909 int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
2910 char __user *optval, unsigned int optlen)
2912 struct sock *sk = sock->sk;
2914 if (sk->sk_prot->compat_setsockopt != NULL)
2915 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2917 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2919 EXPORT_SYMBOL(compat_sock_common_setsockopt);
2922 void sk_common_release(struct sock *sk)
2924 if (sk->sk_prot->destroy)
2925 sk->sk_prot->destroy(sk);
2928 * Observation: when sock_common_release is called, processes have
2929 * no access to socket. But net still has.
2930 * Step one, detach it from networking:
2932 * A. Remove from hash tables.
2935 sk->sk_prot->unhash(sk);
2938 * In this point socket cannot receive new packets, but it is possible
2939 * that some packets are in flight because some CPU runs receiver and
2940 * did hash table lookup before we unhashed socket. They will achieve
2941 * receive queue and will be purged by socket destructor.
2943 * Also we still have packets pending on receive queue and probably,
2944 * our own packets waiting in device queues. sock_destroy will drain
2945 * receive queue, but transmitted packets will delay socket destruction
2946 * until the last reference will be released.
2951 xfrm_sk_free_policy(sk);
2953 sk_refcnt_debug_release(sk);
2957 EXPORT_SYMBOL(sk_common_release);
2959 void sk_get_meminfo(const struct sock *sk, u32 *mem)
2961 memset(mem, 0, sizeof(*mem) * SK_MEMINFO_VARS);
2963 mem[SK_MEMINFO_RMEM_ALLOC] = sk_rmem_alloc_get(sk);
2964 mem[SK_MEMINFO_RCVBUF] = sk->sk_rcvbuf;
2965 mem[SK_MEMINFO_WMEM_ALLOC] = sk_wmem_alloc_get(sk);
2966 mem[SK_MEMINFO_SNDBUF] = sk->sk_sndbuf;
2967 mem[SK_MEMINFO_FWD_ALLOC] = sk->sk_forward_alloc;
2968 mem[SK_MEMINFO_WMEM_QUEUED] = sk->sk_wmem_queued;
2969 mem[SK_MEMINFO_OPTMEM] = atomic_read(&sk->sk_omem_alloc);
2970 mem[SK_MEMINFO_BACKLOG] = sk->sk_backlog.len;
2971 mem[SK_MEMINFO_DROPS] = atomic_read(&sk->sk_drops);
2974 #ifdef CONFIG_PROC_FS
2975 #define PROTO_INUSE_NR 64 /* should be enough for the first time */
2977 int val[PROTO_INUSE_NR];
2980 static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
2982 #ifdef CONFIG_NET_NS
2983 void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2985 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
2987 EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2989 int sock_prot_inuse_get(struct net *net, struct proto *prot)
2991 int cpu, idx = prot->inuse_idx;
2994 for_each_possible_cpu(cpu)
2995 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2997 return res >= 0 ? res : 0;
2999 EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
3001 static int __net_init sock_inuse_init_net(struct net *net)
3003 net->core.inuse = alloc_percpu(struct prot_inuse);
3004 return net->core.inuse ? 0 : -ENOMEM;
3007 static void __net_exit sock_inuse_exit_net(struct net *net)
3009 free_percpu(net->core.inuse);
3012 static struct pernet_operations net_inuse_ops = {
3013 .init = sock_inuse_init_net,
3014 .exit = sock_inuse_exit_net,
3017 static __init int net_inuse_init(void)
3019 if (register_pernet_subsys(&net_inuse_ops))
3020 panic("Cannot initialize net inuse counters");
3025 core_initcall(net_inuse_init);
3027 static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
3029 void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
3031 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
3033 EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
3035 int sock_prot_inuse_get(struct net *net, struct proto *prot)
3037 int cpu, idx = prot->inuse_idx;
3040 for_each_possible_cpu(cpu)
3041 res += per_cpu(prot_inuse, cpu).val[idx];
3043 return res >= 0 ? res : 0;
3045 EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
3048 static void assign_proto_idx(struct proto *prot)
3050 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
3052 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
3053 pr_err("PROTO_INUSE_NR exhausted\n");
3057 set_bit(prot->inuse_idx, proto_inuse_idx);
3060 static void release_proto_idx(struct proto *prot)
3062 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
3063 clear_bit(prot->inuse_idx, proto_inuse_idx);
3066 static inline void assign_proto_idx(struct proto *prot)
3070 static inline void release_proto_idx(struct proto *prot)
3075 static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
3079 kfree(rsk_prot->slab_name);
3080 rsk_prot->slab_name = NULL;
3081 kmem_cache_destroy(rsk_prot->slab);
3082 rsk_prot->slab = NULL;
3085 static int req_prot_init(const struct proto *prot)
3087 struct request_sock_ops *rsk_prot = prot->rsk_prot;
3092 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
3094 if (!rsk_prot->slab_name)
3097 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
3098 rsk_prot->obj_size, 0,
3099 prot->slab_flags, NULL);
3101 if (!rsk_prot->slab) {
3102 pr_crit("%s: Can't create request sock SLAB cache!\n",
3109 int proto_register(struct proto *prot, int alloc_slab)
3112 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
3113 SLAB_HWCACHE_ALIGN | prot->slab_flags,
3116 if (prot->slab == NULL) {
3117 pr_crit("%s: Can't create sock SLAB cache!\n",
3122 if (req_prot_init(prot))
3123 goto out_free_request_sock_slab;
3125 if (prot->twsk_prot != NULL) {
3126 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
3128 if (prot->twsk_prot->twsk_slab_name == NULL)
3129 goto out_free_request_sock_slab;
3131 prot->twsk_prot->twsk_slab =
3132 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
3133 prot->twsk_prot->twsk_obj_size,
3137 if (prot->twsk_prot->twsk_slab == NULL)
3138 goto out_free_timewait_sock_slab_name;
3142 mutex_lock(&proto_list_mutex);
3143 list_add(&prot->node, &proto_list);
3144 assign_proto_idx(prot);
3145 mutex_unlock(&proto_list_mutex);
3148 out_free_timewait_sock_slab_name:
3149 kfree(prot->twsk_prot->twsk_slab_name);
3150 out_free_request_sock_slab:
3151 req_prot_cleanup(prot->rsk_prot);
3153 kmem_cache_destroy(prot->slab);
3158 EXPORT_SYMBOL(proto_register);
3160 void proto_unregister(struct proto *prot)
3162 mutex_lock(&proto_list_mutex);
3163 release_proto_idx(prot);
3164 list_del(&prot->node);
3165 mutex_unlock(&proto_list_mutex);
3167 kmem_cache_destroy(prot->slab);
3170 req_prot_cleanup(prot->rsk_prot);
3172 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
3173 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
3174 kfree(prot->twsk_prot->twsk_slab_name);
3175 prot->twsk_prot->twsk_slab = NULL;
3178 EXPORT_SYMBOL(proto_unregister);
3180 #ifdef CONFIG_PROC_FS
3181 static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
3182 __acquires(proto_list_mutex)
3184 mutex_lock(&proto_list_mutex);
3185 return seq_list_start_head(&proto_list, *pos);
3188 static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3190 return seq_list_next(v, &proto_list, pos);
3193 static void proto_seq_stop(struct seq_file *seq, void *v)
3194 __releases(proto_list_mutex)
3196 mutex_unlock(&proto_list_mutex);
3199 static char proto_method_implemented(const void *method)
3201 return method == NULL ? 'n' : 'y';
3203 static long sock_prot_memory_allocated(struct proto *proto)
3205 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
3208 static char *sock_prot_memory_pressure(struct proto *proto)
3210 return proto->memory_pressure != NULL ?
3211 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
3214 static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
3217 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
3218 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
3221 sock_prot_inuse_get(seq_file_net(seq), proto),
3222 sock_prot_memory_allocated(proto),
3223 sock_prot_memory_pressure(proto),
3225 proto->slab == NULL ? "no" : "yes",
3226 module_name(proto->owner),
3227 proto_method_implemented(proto->close),
3228 proto_method_implemented(proto->connect),
3229 proto_method_implemented(proto->disconnect),
3230 proto_method_implemented(proto->accept),
3231 proto_method_implemented(proto->ioctl),
3232 proto_method_implemented(proto->init),
3233 proto_method_implemented(proto->destroy),
3234 proto_method_implemented(proto->shutdown),
3235 proto_method_implemented(proto->setsockopt),
3236 proto_method_implemented(proto->getsockopt),
3237 proto_method_implemented(proto->sendmsg),
3238 proto_method_implemented(proto->recvmsg),
3239 proto_method_implemented(proto->sendpage),
3240 proto_method_implemented(proto->bind),
3241 proto_method_implemented(proto->backlog_rcv),
3242 proto_method_implemented(proto->hash),
3243 proto_method_implemented(proto->unhash),
3244 proto_method_implemented(proto->get_port),
3245 proto_method_implemented(proto->enter_memory_pressure));
3248 static int proto_seq_show(struct seq_file *seq, void *v)
3250 if (v == &proto_list)
3251 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3260 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3262 proto_seq_printf(seq, list_entry(v, struct proto, node));
3266 static const struct seq_operations proto_seq_ops = {
3267 .start = proto_seq_start,
3268 .next = proto_seq_next,
3269 .stop = proto_seq_stop,
3270 .show = proto_seq_show,
3273 static int proto_seq_open(struct inode *inode, struct file *file)
3275 return seq_open_net(inode, file, &proto_seq_ops,
3276 sizeof(struct seq_net_private));
3279 static const struct file_operations proto_seq_fops = {
3280 .owner = THIS_MODULE,
3281 .open = proto_seq_open,
3283 .llseek = seq_lseek,
3284 .release = seq_release_net,
3287 static __net_init int proto_init_net(struct net *net)
3289 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
3295 static __net_exit void proto_exit_net(struct net *net)
3297 remove_proc_entry("protocols", net->proc_net);
3301 static __net_initdata struct pernet_operations proto_net_ops = {
3302 .init = proto_init_net,
3303 .exit = proto_exit_net,
3306 static int __init proto_init(void)
3308 return register_pernet_subsys(&proto_net_ops);
3311 subsys_initcall(proto_init);
3313 #endif /* PROC_FS */
3315 #ifdef CONFIG_NET_RX_BUSY_POLL
3316 bool sk_busy_loop_end(void *p, unsigned long start_time)
3318 struct sock *sk = p;
3320 return !skb_queue_empty(&sk->sk_receive_queue) ||
3321 sk_busy_loop_timeout(sk, start_time);
3323 EXPORT_SYMBOL(sk_busy_loop_end);
3324 #endif /* CONFIG_NET_RX_BUSY_POLL */