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.
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);
1060 if (sk->sk_family != PF_INET && sk->sk_family != PF_INET6)
1062 else if (sk->sk_protocol != IPPROTO_TCP)
1064 else if (sk->sk_state != TCP_CLOSE)
1066 else if (val < 0 || val > 1)
1069 sock_valbool_flag(sk, SOCK_ZEROCOPY, valbool);
1079 EXPORT_SYMBOL(sock_setsockopt);
1082 static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1083 struct ucred *ucred)
1085 ucred->pid = pid_vnr(pid);
1086 ucred->uid = ucred->gid = -1;
1088 struct user_namespace *current_ns = current_user_ns();
1090 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1091 ucred->gid = from_kgid_munged(current_ns, cred->egid);
1095 static int groups_to_user(gid_t __user *dst, const struct group_info *src)
1097 struct user_namespace *user_ns = current_user_ns();
1100 for (i = 0; i < src->ngroups; i++)
1101 if (put_user(from_kgid_munged(user_ns, src->gid[i]), dst + i))
1107 int sock_getsockopt(struct socket *sock, int level, int optname,
1108 char __user *optval, int __user *optlen)
1110 struct sock *sk = sock->sk;
1119 int lv = sizeof(int);
1122 if (get_user(len, optlen))
1127 memset(&v, 0, sizeof(v));
1131 v.val = sock_flag(sk, SOCK_DBG);
1135 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1139 v.val = sock_flag(sk, SOCK_BROADCAST);
1143 v.val = sk->sk_sndbuf;
1147 v.val = sk->sk_rcvbuf;
1151 v.val = sk->sk_reuse;
1155 v.val = sk->sk_reuseport;
1159 v.val = sock_flag(sk, SOCK_KEEPOPEN);
1163 v.val = sk->sk_type;
1167 v.val = sk->sk_protocol;
1171 v.val = sk->sk_family;
1175 v.val = -sock_error(sk);
1177 v.val = xchg(&sk->sk_err_soft, 0);
1181 v.val = sock_flag(sk, SOCK_URGINLINE);
1185 v.val = sk->sk_no_check_tx;
1189 v.val = sk->sk_priority;
1193 lv = sizeof(v.ling);
1194 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
1195 v.ling.l_linger = sk->sk_lingertime / HZ;
1199 sock_warn_obsolete_bsdism("getsockopt");
1203 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1204 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1207 case SO_TIMESTAMPNS:
1208 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
1211 case SO_TIMESTAMPING:
1212 v.val = sk->sk_tsflags;
1216 lv = sizeof(struct timeval);
1217 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1221 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1222 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * USEC_PER_SEC) / HZ;
1227 lv = sizeof(struct timeval);
1228 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1232 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1233 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * USEC_PER_SEC) / HZ;
1238 v.val = sk->sk_rcvlowat;
1246 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
1251 struct ucred peercred;
1252 if (len > sizeof(peercred))
1253 len = sizeof(peercred);
1254 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1255 if (copy_to_user(optval, &peercred, len))
1264 if (!sk->sk_peer_cred)
1267 n = sk->sk_peer_cred->group_info->ngroups;
1268 if (len < n * sizeof(gid_t)) {
1269 len = n * sizeof(gid_t);
1270 return put_user(len, optlen) ? -EFAULT : -ERANGE;
1272 len = n * sizeof(gid_t);
1274 ret = groups_to_user((gid_t __user *)optval,
1275 sk->sk_peer_cred->group_info);
1285 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1289 if (copy_to_user(optval, address, len))
1294 /* Dubious BSD thing... Probably nobody even uses it, but
1295 * the UNIX standard wants it for whatever reason... -DaveM
1298 v.val = sk->sk_state == TCP_LISTEN;
1302 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
1306 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1309 v.val = sk->sk_mark;
1313 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
1316 case SO_WIFI_STATUS:
1317 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
1321 if (!sock->ops->set_peek_off)
1324 v.val = sk->sk_peek_off;
1327 v.val = sock_flag(sk, SOCK_NOFCS);
1330 case SO_BINDTODEVICE:
1331 return sock_getbindtodevice(sk, optval, optlen, len);
1334 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1340 case SO_LOCK_FILTER:
1341 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1344 case SO_BPF_EXTENSIONS:
1345 v.val = bpf_tell_extensions();
1348 case SO_SELECT_ERR_QUEUE:
1349 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1352 #ifdef CONFIG_NET_RX_BUSY_POLL
1354 v.val = sk->sk_ll_usec;
1358 case SO_MAX_PACING_RATE:
1359 v.val = sk->sk_max_pacing_rate;
1362 case SO_INCOMING_CPU:
1363 v.val = sk->sk_incoming_cpu;
1368 u32 meminfo[SK_MEMINFO_VARS];
1370 if (get_user(len, optlen))
1373 sk_get_meminfo(sk, meminfo);
1375 len = min_t(unsigned int, len, sizeof(meminfo));
1376 if (copy_to_user(optval, &meminfo, len))
1382 #ifdef CONFIG_NET_RX_BUSY_POLL
1383 case SO_INCOMING_NAPI_ID:
1384 v.val = READ_ONCE(sk->sk_napi_id);
1386 /* aggregate non-NAPI IDs down to 0 */
1387 if (v.val < MIN_NAPI_ID)
1397 v.val64 = sock_gen_cookie(sk);
1401 v.val = sock_flag(sk, SOCK_ZEROCOPY);
1405 /* We implement the SO_SNDLOWAT etc to not be settable
1408 return -ENOPROTOOPT;
1413 if (copy_to_user(optval, &v, len))
1416 if (put_user(len, optlen))
1422 * Initialize an sk_lock.
1424 * (We also register the sk_lock with the lock validator.)
1426 static inline void sock_lock_init(struct sock *sk)
1428 if (sk->sk_kern_sock)
1429 sock_lock_init_class_and_name(
1431 af_family_kern_slock_key_strings[sk->sk_family],
1432 af_family_kern_slock_keys + sk->sk_family,
1433 af_family_kern_key_strings[sk->sk_family],
1434 af_family_kern_keys + sk->sk_family);
1436 sock_lock_init_class_and_name(
1438 af_family_slock_key_strings[sk->sk_family],
1439 af_family_slock_keys + sk->sk_family,
1440 af_family_key_strings[sk->sk_family],
1441 af_family_keys + sk->sk_family);
1445 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1446 * even temporarly, because of RCU lookups. sk_node should also be left as is.
1447 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
1449 static void sock_copy(struct sock *nsk, const struct sock *osk)
1451 #ifdef CONFIG_SECURITY_NETWORK
1452 void *sptr = nsk->sk_security;
1454 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1456 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1457 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1459 #ifdef CONFIG_SECURITY_NETWORK
1460 nsk->sk_security = sptr;
1461 security_sk_clone(osk, nsk);
1465 static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1469 struct kmem_cache *slab;
1473 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1476 if (priority & __GFP_ZERO)
1477 sk_prot_clear_nulls(sk, prot->obj_size);
1479 sk = kmalloc(prot->obj_size, priority);
1482 kmemcheck_annotate_bitfield(sk, flags);
1484 if (security_sk_alloc(sk, family, priority))
1487 if (!try_module_get(prot->owner))
1489 sk_tx_queue_clear(sk);
1495 security_sk_free(sk);
1498 kmem_cache_free(slab, sk);
1504 static void sk_prot_free(struct proto *prot, struct sock *sk)
1506 struct kmem_cache *slab;
1507 struct module *owner;
1509 owner = prot->owner;
1512 cgroup_sk_free(&sk->sk_cgrp_data);
1513 mem_cgroup_sk_free(sk);
1514 security_sk_free(sk);
1516 kmem_cache_free(slab, sk);
1523 * sk_alloc - All socket objects are allocated here
1524 * @net: the applicable net namespace
1525 * @family: protocol family
1526 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1527 * @prot: struct proto associated with this new sock instance
1528 * @kern: is this to be a kernel socket?
1530 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1531 struct proto *prot, int kern)
1535 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1537 sk->sk_family = family;
1539 * See comment in struct sock definition to understand
1540 * why we need sk_prot_creator -acme
1542 sk->sk_prot = sk->sk_prot_creator = prot;
1543 sk->sk_kern_sock = kern;
1545 sk->sk_net_refcnt = kern ? 0 : 1;
1546 if (likely(sk->sk_net_refcnt))
1548 sock_net_set(sk, net);
1549 refcount_set(&sk->sk_wmem_alloc, 1);
1551 mem_cgroup_sk_alloc(sk);
1552 cgroup_sk_alloc(&sk->sk_cgrp_data);
1553 sock_update_classid(&sk->sk_cgrp_data);
1554 sock_update_netprioidx(&sk->sk_cgrp_data);
1559 EXPORT_SYMBOL(sk_alloc);
1561 /* Sockets having SOCK_RCU_FREE will call this function after one RCU
1562 * grace period. This is the case for UDP sockets and TCP listeners.
1564 static void __sk_destruct(struct rcu_head *head)
1566 struct sock *sk = container_of(head, struct sock, sk_rcu);
1567 struct sk_filter *filter;
1569 if (sk->sk_destruct)
1570 sk->sk_destruct(sk);
1572 filter = rcu_dereference_check(sk->sk_filter,
1573 refcount_read(&sk->sk_wmem_alloc) == 0);
1575 sk_filter_uncharge(sk, filter);
1576 RCU_INIT_POINTER(sk->sk_filter, NULL);
1578 if (rcu_access_pointer(sk->sk_reuseport_cb))
1579 reuseport_detach_sock(sk);
1581 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1583 if (atomic_read(&sk->sk_omem_alloc))
1584 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1585 __func__, atomic_read(&sk->sk_omem_alloc));
1587 if (sk->sk_frag.page) {
1588 put_page(sk->sk_frag.page);
1589 sk->sk_frag.page = NULL;
1592 if (sk->sk_peer_cred)
1593 put_cred(sk->sk_peer_cred);
1594 put_pid(sk->sk_peer_pid);
1595 if (likely(sk->sk_net_refcnt))
1596 put_net(sock_net(sk));
1597 sk_prot_free(sk->sk_prot_creator, sk);
1600 void sk_destruct(struct sock *sk)
1602 if (sock_flag(sk, SOCK_RCU_FREE))
1603 call_rcu(&sk->sk_rcu, __sk_destruct);
1605 __sk_destruct(&sk->sk_rcu);
1608 static void __sk_free(struct sock *sk)
1610 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
1611 sock_diag_broadcast_destroy(sk);
1616 void sk_free(struct sock *sk)
1619 * We subtract one from sk_wmem_alloc and can know if
1620 * some packets are still in some tx queue.
1621 * If not null, sock_wfree() will call __sk_free(sk) later
1623 if (refcount_dec_and_test(&sk->sk_wmem_alloc))
1626 EXPORT_SYMBOL(sk_free);
1628 static void sk_init_common(struct sock *sk)
1630 skb_queue_head_init(&sk->sk_receive_queue);
1631 skb_queue_head_init(&sk->sk_write_queue);
1632 skb_queue_head_init(&sk->sk_error_queue);
1634 rwlock_init(&sk->sk_callback_lock);
1635 lockdep_set_class_and_name(&sk->sk_receive_queue.lock,
1636 af_rlock_keys + sk->sk_family,
1637 af_family_rlock_key_strings[sk->sk_family]);
1638 lockdep_set_class_and_name(&sk->sk_write_queue.lock,
1639 af_wlock_keys + sk->sk_family,
1640 af_family_wlock_key_strings[sk->sk_family]);
1641 lockdep_set_class_and_name(&sk->sk_error_queue.lock,
1642 af_elock_keys + sk->sk_family,
1643 af_family_elock_key_strings[sk->sk_family]);
1644 lockdep_set_class_and_name(&sk->sk_callback_lock,
1645 af_callback_keys + sk->sk_family,
1646 af_family_clock_key_strings[sk->sk_family]);
1650 * sk_clone_lock - clone a socket, and lock its clone
1651 * @sk: the socket to clone
1652 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1654 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1656 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
1659 bool is_charged = true;
1661 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
1662 if (newsk != NULL) {
1663 struct sk_filter *filter;
1665 sock_copy(newsk, sk);
1668 if (likely(newsk->sk_net_refcnt))
1669 get_net(sock_net(newsk));
1670 sk_node_init(&newsk->sk_node);
1671 sock_lock_init(newsk);
1672 bh_lock_sock(newsk);
1673 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
1674 newsk->sk_backlog.len = 0;
1676 atomic_set(&newsk->sk_rmem_alloc, 0);
1678 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1680 refcount_set(&newsk->sk_wmem_alloc, 1);
1681 atomic_set(&newsk->sk_omem_alloc, 0);
1682 sk_init_common(newsk);
1684 newsk->sk_dst_cache = NULL;
1685 newsk->sk_dst_pending_confirm = 0;
1686 newsk->sk_wmem_queued = 0;
1687 newsk->sk_forward_alloc = 0;
1688 atomic_set(&newsk->sk_drops, 0);
1689 newsk->sk_send_head = NULL;
1690 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1691 atomic_set(&newsk->sk_zckey, 0);
1693 sock_reset_flag(newsk, SOCK_DONE);
1695 filter = rcu_dereference_protected(newsk->sk_filter, 1);
1697 /* though it's an empty new sock, the charging may fail
1698 * if sysctl_optmem_max was changed between creation of
1699 * original socket and cloning
1701 is_charged = sk_filter_charge(newsk, filter);
1703 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk, sk))) {
1704 /* We need to make sure that we don't uncharge the new
1705 * socket if we couldn't charge it in the first place
1706 * as otherwise we uncharge the parent's filter.
1709 RCU_INIT_POINTER(newsk->sk_filter, NULL);
1710 sk_free_unlock_clone(newsk);
1714 RCU_INIT_POINTER(newsk->sk_reuseport_cb, NULL);
1717 newsk->sk_err_soft = 0;
1718 newsk->sk_priority = 0;
1719 newsk->sk_incoming_cpu = raw_smp_processor_id();
1720 atomic64_set(&newsk->sk_cookie, 0);
1722 mem_cgroup_sk_alloc(newsk);
1723 cgroup_sk_alloc(&newsk->sk_cgrp_data);
1726 * Before updating sk_refcnt, we must commit prior changes to memory
1727 * (Documentation/RCU/rculist_nulls.txt for details)
1730 refcount_set(&newsk->sk_refcnt, 2);
1733 * Increment the counter in the same struct proto as the master
1734 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1735 * is the same as sk->sk_prot->socks, as this field was copied
1738 * This _changes_ the previous behaviour, where
1739 * tcp_create_openreq_child always was incrementing the
1740 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1741 * to be taken into account in all callers. -acme
1743 sk_refcnt_debug_inc(newsk);
1744 sk_set_socket(newsk, NULL);
1745 newsk->sk_wq = NULL;
1747 if (newsk->sk_prot->sockets_allocated)
1748 sk_sockets_allocated_inc(newsk);
1750 if (sock_needs_netstamp(sk) &&
1751 newsk->sk_flags & SK_FLAGS_TIMESTAMP)
1752 net_enable_timestamp();
1757 EXPORT_SYMBOL_GPL(sk_clone_lock);
1759 void sk_free_unlock_clone(struct sock *sk)
1761 /* It is still raw copy of parent, so invalidate
1762 * destructor and make plain sk_free() */
1763 sk->sk_destruct = NULL;
1767 EXPORT_SYMBOL_GPL(sk_free_unlock_clone);
1769 void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1773 sk_dst_set(sk, dst);
1774 sk->sk_route_caps = dst->dev->features;
1775 if (sk->sk_route_caps & NETIF_F_GSO)
1776 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
1777 sk->sk_route_caps &= ~sk->sk_route_nocaps;
1778 if (sk_can_gso(sk)) {
1779 if (dst->header_len) {
1780 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
1782 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
1783 sk->sk_gso_max_size = dst->dev->gso_max_size;
1784 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
1787 sk->sk_gso_max_segs = max_segs;
1789 EXPORT_SYMBOL_GPL(sk_setup_caps);
1792 * Simple resource managers for sockets.
1797 * Write buffer destructor automatically called from kfree_skb.
1799 void sock_wfree(struct sk_buff *skb)
1801 struct sock *sk = skb->sk;
1802 unsigned int len = skb->truesize;
1804 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1806 * Keep a reference on sk_wmem_alloc, this will be released
1807 * after sk_write_space() call
1809 WARN_ON(refcount_sub_and_test(len - 1, &sk->sk_wmem_alloc));
1810 sk->sk_write_space(sk);
1814 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1815 * could not do because of in-flight packets
1817 if (refcount_sub_and_test(len, &sk->sk_wmem_alloc))
1820 EXPORT_SYMBOL(sock_wfree);
1822 /* This variant of sock_wfree() is used by TCP,
1823 * since it sets SOCK_USE_WRITE_QUEUE.
1825 void __sock_wfree(struct sk_buff *skb)
1827 struct sock *sk = skb->sk;
1829 if (refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc))
1833 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
1838 if (unlikely(!sk_fullsock(sk))) {
1839 skb->destructor = sock_edemux;
1844 skb->destructor = sock_wfree;
1845 skb_set_hash_from_sk(skb, sk);
1847 * We used to take a refcount on sk, but following operation
1848 * is enough to guarantee sk_free() wont free this sock until
1849 * all in-flight packets are completed
1851 refcount_add(skb->truesize, &sk->sk_wmem_alloc);
1853 EXPORT_SYMBOL(skb_set_owner_w);
1855 /* This helper is used by netem, as it can hold packets in its
1856 * delay queue. We want to allow the owner socket to send more
1857 * packets, as if they were already TX completed by a typical driver.
1858 * But we also want to keep skb->sk set because some packet schedulers
1859 * rely on it (sch_fq for example).
1861 void skb_orphan_partial(struct sk_buff *skb)
1863 if (skb_is_tcp_pure_ack(skb))
1866 if (skb->destructor == sock_wfree
1868 || skb->destructor == tcp_wfree
1871 struct sock *sk = skb->sk;
1873 if (refcount_inc_not_zero(&sk->sk_refcnt)) {
1874 WARN_ON(refcount_sub_and_test(skb->truesize, &sk->sk_wmem_alloc));
1875 skb->destructor = sock_efree;
1881 EXPORT_SYMBOL(skb_orphan_partial);
1884 * Read buffer destructor automatically called from kfree_skb.
1886 void sock_rfree(struct sk_buff *skb)
1888 struct sock *sk = skb->sk;
1889 unsigned int len = skb->truesize;
1891 atomic_sub(len, &sk->sk_rmem_alloc);
1892 sk_mem_uncharge(sk, len);
1894 EXPORT_SYMBOL(sock_rfree);
1897 * Buffer destructor for skbs that are not used directly in read or write
1898 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1900 void sock_efree(struct sk_buff *skb)
1904 EXPORT_SYMBOL(sock_efree);
1906 kuid_t sock_i_uid(struct sock *sk)
1910 read_lock_bh(&sk->sk_callback_lock);
1911 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
1912 read_unlock_bh(&sk->sk_callback_lock);
1915 EXPORT_SYMBOL(sock_i_uid);
1917 unsigned long sock_i_ino(struct sock *sk)
1921 read_lock_bh(&sk->sk_callback_lock);
1922 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
1923 read_unlock_bh(&sk->sk_callback_lock);
1926 EXPORT_SYMBOL(sock_i_ino);
1929 * Allocate a skb from the socket's send buffer.
1931 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1934 if (force || refcount_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
1935 struct sk_buff *skb = alloc_skb(size, priority);
1937 skb_set_owner_w(skb, sk);
1943 EXPORT_SYMBOL(sock_wmalloc);
1945 static void sock_ofree(struct sk_buff *skb)
1947 struct sock *sk = skb->sk;
1949 atomic_sub(skb->truesize, &sk->sk_omem_alloc);
1952 struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1955 struct sk_buff *skb;
1957 /* small safe race: SKB_TRUESIZE may differ from final skb->truesize */
1958 if (atomic_read(&sk->sk_omem_alloc) + SKB_TRUESIZE(size) >
1962 skb = alloc_skb(size, priority);
1966 atomic_add(skb->truesize, &sk->sk_omem_alloc);
1968 skb->destructor = sock_ofree;
1973 * Allocate a memory block from the socket's option memory buffer.
1975 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1977 if ((unsigned int)size <= sysctl_optmem_max &&
1978 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1980 /* First do the add, to avoid the race if kmalloc
1983 atomic_add(size, &sk->sk_omem_alloc);
1984 mem = kmalloc(size, priority);
1987 atomic_sub(size, &sk->sk_omem_alloc);
1991 EXPORT_SYMBOL(sock_kmalloc);
1993 /* Free an option memory block. Note, we actually want the inline
1994 * here as this allows gcc to detect the nullify and fold away the
1995 * condition entirely.
1997 static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
2000 if (WARN_ON_ONCE(!mem))
2006 atomic_sub(size, &sk->sk_omem_alloc);
2009 void sock_kfree_s(struct sock *sk, void *mem, int size)
2011 __sock_kfree_s(sk, mem, size, false);
2013 EXPORT_SYMBOL(sock_kfree_s);
2015 void sock_kzfree_s(struct sock *sk, void *mem, int size)
2017 __sock_kfree_s(sk, mem, size, true);
2019 EXPORT_SYMBOL(sock_kzfree_s);
2021 /* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
2022 I think, these locks should be removed for datagram sockets.
2024 static long sock_wait_for_wmem(struct sock *sk, long timeo)
2028 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2032 if (signal_pending(current))
2034 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2035 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
2036 if (refcount_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
2038 if (sk->sk_shutdown & SEND_SHUTDOWN)
2042 timeo = schedule_timeout(timeo);
2044 finish_wait(sk_sleep(sk), &wait);
2050 * Generic send/receive buffer handlers
2053 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
2054 unsigned long data_len, int noblock,
2055 int *errcode, int max_page_order)
2057 struct sk_buff *skb;
2061 timeo = sock_sndtimeo(sk, noblock);
2063 err = sock_error(sk);
2068 if (sk->sk_shutdown & SEND_SHUTDOWN)
2071 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
2074 sk_set_bit(SOCKWQ_ASYNC_NOSPACE, sk);
2075 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
2079 if (signal_pending(current))
2081 timeo = sock_wait_for_wmem(sk, timeo);
2083 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
2084 errcode, sk->sk_allocation);
2086 skb_set_owner_w(skb, sk);
2090 err = sock_intr_errno(timeo);
2095 EXPORT_SYMBOL(sock_alloc_send_pskb);
2097 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
2098 int noblock, int *errcode)
2100 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
2102 EXPORT_SYMBOL(sock_alloc_send_skb);
2104 int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
2105 struct sockcm_cookie *sockc)
2109 switch (cmsg->cmsg_type) {
2111 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
2113 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2115 sockc->mark = *(u32 *)CMSG_DATA(cmsg);
2117 case SO_TIMESTAMPING:
2118 if (cmsg->cmsg_len != CMSG_LEN(sizeof(u32)))
2121 tsflags = *(u32 *)CMSG_DATA(cmsg);
2122 if (tsflags & ~SOF_TIMESTAMPING_TX_RECORD_MASK)
2125 sockc->tsflags &= ~SOF_TIMESTAMPING_TX_RECORD_MASK;
2126 sockc->tsflags |= tsflags;
2128 /* SCM_RIGHTS and SCM_CREDENTIALS are semantically in SOL_UNIX. */
2130 case SCM_CREDENTIALS:
2137 EXPORT_SYMBOL(__sock_cmsg_send);
2139 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
2140 struct sockcm_cookie *sockc)
2142 struct cmsghdr *cmsg;
2145 for_each_cmsghdr(cmsg, msg) {
2146 if (!CMSG_OK(msg, cmsg))
2148 if (cmsg->cmsg_level != SOL_SOCKET)
2150 ret = __sock_cmsg_send(sk, msg, cmsg, sockc);
2156 EXPORT_SYMBOL(sock_cmsg_send);
2158 static void sk_enter_memory_pressure(struct sock *sk)
2160 if (!sk->sk_prot->enter_memory_pressure)
2163 sk->sk_prot->enter_memory_pressure(sk);
2166 static void sk_leave_memory_pressure(struct sock *sk)
2168 if (sk->sk_prot->leave_memory_pressure) {
2169 sk->sk_prot->leave_memory_pressure(sk);
2171 unsigned long *memory_pressure = sk->sk_prot->memory_pressure;
2173 if (memory_pressure && *memory_pressure)
2174 *memory_pressure = 0;
2178 /* On 32bit arches, an skb frag is limited to 2^15 */
2179 #define SKB_FRAG_PAGE_ORDER get_order(32768)
2182 * skb_page_frag_refill - check that a page_frag contains enough room
2183 * @sz: minimum size of the fragment we want to get
2184 * @pfrag: pointer to page_frag
2185 * @gfp: priority for memory allocation
2187 * Note: While this allocator tries to use high order pages, there is
2188 * no guarantee that allocations succeed. Therefore, @sz MUST be
2189 * less or equal than PAGE_SIZE.
2191 bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
2194 if (page_ref_count(pfrag->page) == 1) {
2198 if (pfrag->offset + sz <= pfrag->size)
2200 put_page(pfrag->page);
2204 if (SKB_FRAG_PAGE_ORDER) {
2205 /* Avoid direct reclaim but allow kswapd to wake */
2206 pfrag->page = alloc_pages((gfp & ~__GFP_DIRECT_RECLAIM) |
2207 __GFP_COMP | __GFP_NOWARN |
2209 SKB_FRAG_PAGE_ORDER);
2210 if (likely(pfrag->page)) {
2211 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
2215 pfrag->page = alloc_page(gfp);
2216 if (likely(pfrag->page)) {
2217 pfrag->size = PAGE_SIZE;
2222 EXPORT_SYMBOL(skb_page_frag_refill);
2224 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
2226 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
2229 sk_enter_memory_pressure(sk);
2230 sk_stream_moderate_sndbuf(sk);
2233 EXPORT_SYMBOL(sk_page_frag_refill);
2235 static void __lock_sock(struct sock *sk)
2236 __releases(&sk->sk_lock.slock)
2237 __acquires(&sk->sk_lock.slock)
2242 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
2243 TASK_UNINTERRUPTIBLE);
2244 spin_unlock_bh(&sk->sk_lock.slock);
2246 spin_lock_bh(&sk->sk_lock.slock);
2247 if (!sock_owned_by_user(sk))
2250 finish_wait(&sk->sk_lock.wq, &wait);
2253 static void __release_sock(struct sock *sk)
2254 __releases(&sk->sk_lock.slock)
2255 __acquires(&sk->sk_lock.slock)
2257 struct sk_buff *skb, *next;
2259 while ((skb = sk->sk_backlog.head) != NULL) {
2260 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
2262 spin_unlock_bh(&sk->sk_lock.slock);
2267 WARN_ON_ONCE(skb_dst_is_noref(skb));
2269 sk_backlog_rcv(sk, skb);
2274 } while (skb != NULL);
2276 spin_lock_bh(&sk->sk_lock.slock);
2280 * Doing the zeroing here guarantee we can not loop forever
2281 * while a wild producer attempts to flood us.
2283 sk->sk_backlog.len = 0;
2286 void __sk_flush_backlog(struct sock *sk)
2288 spin_lock_bh(&sk->sk_lock.slock);
2290 spin_unlock_bh(&sk->sk_lock.slock);
2294 * sk_wait_data - wait for data to arrive at sk_receive_queue
2295 * @sk: sock to wait on
2296 * @timeo: for how long
2297 * @skb: last skb seen on sk_receive_queue
2299 * Now socket state including sk->sk_err is changed only under lock,
2300 * hence we may omit checks after joining wait queue.
2301 * We check receive queue before schedule() only as optimization;
2302 * it is very likely that release_sock() added new data.
2304 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
2306 DEFINE_WAIT_FUNC(wait, woken_wake_function);
2309 add_wait_queue(sk_sleep(sk), &wait);
2310 sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk);
2311 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb, &wait);
2312 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk);
2313 remove_wait_queue(sk_sleep(sk), &wait);
2316 EXPORT_SYMBOL(sk_wait_data);
2319 * __sk_mem_raise_allocated - increase memory_allocated
2321 * @size: memory size to allocate
2322 * @amt: pages to allocate
2323 * @kind: allocation type
2325 * Similar to __sk_mem_schedule(), but does not update sk_forward_alloc
2327 int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind)
2329 struct proto *prot = sk->sk_prot;
2330 long allocated = sk_memory_allocated_add(sk, amt);
2332 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
2333 !mem_cgroup_charge_skmem(sk->sk_memcg, amt))
2334 goto suppress_allocation;
2337 if (allocated <= sk_prot_mem_limits(sk, 0)) {
2338 sk_leave_memory_pressure(sk);
2342 /* Under pressure. */
2343 if (allocated > sk_prot_mem_limits(sk, 1))
2344 sk_enter_memory_pressure(sk);
2346 /* Over hard limit. */
2347 if (allocated > sk_prot_mem_limits(sk, 2))
2348 goto suppress_allocation;
2350 /* guarantee minimum buffer size under pressure */
2351 if (kind == SK_MEM_RECV) {
2352 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2355 } else { /* SK_MEM_SEND */
2356 if (sk->sk_type == SOCK_STREAM) {
2357 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2359 } else if (refcount_read(&sk->sk_wmem_alloc) <
2360 prot->sysctl_wmem[0])
2364 if (sk_has_memory_pressure(sk)) {
2367 if (!sk_under_memory_pressure(sk))
2369 alloc = sk_sockets_allocated_read_positive(sk);
2370 if (sk_prot_mem_limits(sk, 2) > alloc *
2371 sk_mem_pages(sk->sk_wmem_queued +
2372 atomic_read(&sk->sk_rmem_alloc) +
2373 sk->sk_forward_alloc))
2377 suppress_allocation:
2379 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2380 sk_stream_moderate_sndbuf(sk);
2382 /* Fail only if socket is _under_ its sndbuf.
2383 * In this case we cannot block, so that we have to fail.
2385 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2389 trace_sock_exceed_buf_limit(sk, prot, allocated);
2391 sk_memory_allocated_sub(sk, amt);
2393 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2394 mem_cgroup_uncharge_skmem(sk->sk_memcg, amt);
2398 EXPORT_SYMBOL(__sk_mem_raise_allocated);
2401 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2403 * @size: memory size to allocate
2404 * @kind: allocation type
2406 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2407 * rmem allocation. This function assumes that protocols which have
2408 * memory_pressure use sk_wmem_queued as write buffer accounting.
2410 int __sk_mem_schedule(struct sock *sk, int size, int kind)
2412 int ret, amt = sk_mem_pages(size);
2414 sk->sk_forward_alloc += amt << SK_MEM_QUANTUM_SHIFT;
2415 ret = __sk_mem_raise_allocated(sk, size, amt, kind);
2417 sk->sk_forward_alloc -= amt << SK_MEM_QUANTUM_SHIFT;
2420 EXPORT_SYMBOL(__sk_mem_schedule);
2423 * __sk_mem_reduce_allocated - reclaim memory_allocated
2425 * @amount: number of quanta
2427 * Similar to __sk_mem_reclaim(), but does not update sk_forward_alloc
2429 void __sk_mem_reduce_allocated(struct sock *sk, int amount)
2431 sk_memory_allocated_sub(sk, amount);
2433 if (mem_cgroup_sockets_enabled && sk->sk_memcg)
2434 mem_cgroup_uncharge_skmem(sk->sk_memcg, amount);
2436 if (sk_under_memory_pressure(sk) &&
2437 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2438 sk_leave_memory_pressure(sk);
2440 EXPORT_SYMBOL(__sk_mem_reduce_allocated);
2443 * __sk_mem_reclaim - reclaim sk_forward_alloc and memory_allocated
2445 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
2447 void __sk_mem_reclaim(struct sock *sk, int amount)
2449 amount >>= SK_MEM_QUANTUM_SHIFT;
2450 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
2451 __sk_mem_reduce_allocated(sk, amount);
2453 EXPORT_SYMBOL(__sk_mem_reclaim);
2455 int sk_set_peek_off(struct sock *sk, int val)
2460 sk->sk_peek_off = val;
2463 EXPORT_SYMBOL_GPL(sk_set_peek_off);
2466 * Set of default routines for initialising struct proto_ops when
2467 * the protocol does not support a particular function. In certain
2468 * cases where it makes no sense for a protocol to have a "do nothing"
2469 * function, some default processing is provided.
2472 int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2476 EXPORT_SYMBOL(sock_no_bind);
2478 int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
2483 EXPORT_SYMBOL(sock_no_connect);
2485 int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2489 EXPORT_SYMBOL(sock_no_socketpair);
2491 int sock_no_accept(struct socket *sock, struct socket *newsock, int flags,
2496 EXPORT_SYMBOL(sock_no_accept);
2498 int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
2503 EXPORT_SYMBOL(sock_no_getname);
2505 unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
2509 EXPORT_SYMBOL(sock_no_poll);
2511 int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2515 EXPORT_SYMBOL(sock_no_ioctl);
2517 int sock_no_listen(struct socket *sock, int backlog)
2521 EXPORT_SYMBOL(sock_no_listen);
2523 int sock_no_shutdown(struct socket *sock, int how)
2527 EXPORT_SYMBOL(sock_no_shutdown);
2529 int sock_no_setsockopt(struct socket *sock, int level, int optname,
2530 char __user *optval, unsigned int optlen)
2534 EXPORT_SYMBOL(sock_no_setsockopt);
2536 int sock_no_getsockopt(struct socket *sock, int level, int optname,
2537 char __user *optval, int __user *optlen)
2541 EXPORT_SYMBOL(sock_no_getsockopt);
2543 int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
2547 EXPORT_SYMBOL(sock_no_sendmsg);
2549 int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *m, size_t len)
2553 EXPORT_SYMBOL(sock_no_sendmsg_locked);
2555 int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2560 EXPORT_SYMBOL(sock_no_recvmsg);
2562 int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2564 /* Mirror missing mmap method error code */
2567 EXPORT_SYMBOL(sock_no_mmap);
2569 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2572 struct msghdr msg = {.msg_flags = flags};
2574 char *kaddr = kmap(page);
2575 iov.iov_base = kaddr + offset;
2577 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2581 EXPORT_SYMBOL(sock_no_sendpage);
2583 ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
2584 int offset, size_t size, int flags)
2587 struct msghdr msg = {.msg_flags = flags};
2589 char *kaddr = kmap(page);
2591 iov.iov_base = kaddr + offset;
2593 res = kernel_sendmsg_locked(sk, &msg, &iov, 1, size);
2597 EXPORT_SYMBOL(sock_no_sendpage_locked);
2600 * Default Socket Callbacks
2603 static void sock_def_wakeup(struct sock *sk)
2605 struct socket_wq *wq;
2608 wq = rcu_dereference(sk->sk_wq);
2609 if (skwq_has_sleeper(wq))
2610 wake_up_interruptible_all(&wq->wait);
2614 static void sock_def_error_report(struct sock *sk)
2616 struct socket_wq *wq;
2619 wq = rcu_dereference(sk->sk_wq);
2620 if (skwq_has_sleeper(wq))
2621 wake_up_interruptible_poll(&wq->wait, POLLERR);
2622 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
2626 static void sock_def_readable(struct sock *sk)
2628 struct socket_wq *wq;
2631 wq = rcu_dereference(sk->sk_wq);
2632 if (skwq_has_sleeper(wq))
2633 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
2634 POLLRDNORM | POLLRDBAND);
2635 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
2639 static void sock_def_write_space(struct sock *sk)
2641 struct socket_wq *wq;
2645 /* Do not wake up a writer until he can make "significant"
2648 if ((refcount_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
2649 wq = rcu_dereference(sk->sk_wq);
2650 if (skwq_has_sleeper(wq))
2651 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
2652 POLLWRNORM | POLLWRBAND);
2654 /* Should agree with poll, otherwise some programs break */
2655 if (sock_writeable(sk))
2656 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
2662 static void sock_def_destruct(struct sock *sk)
2666 void sk_send_sigurg(struct sock *sk)
2668 if (sk->sk_socket && sk->sk_socket->file)
2669 if (send_sigurg(&sk->sk_socket->file->f_owner))
2670 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
2672 EXPORT_SYMBOL(sk_send_sigurg);
2674 void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2675 unsigned long expires)
2677 if (!mod_timer(timer, expires))
2680 EXPORT_SYMBOL(sk_reset_timer);
2682 void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2684 if (del_timer(timer))
2687 EXPORT_SYMBOL(sk_stop_timer);
2689 void sock_init_data(struct socket *sock, struct sock *sk)
2692 sk->sk_send_head = NULL;
2694 init_timer(&sk->sk_timer);
2696 sk->sk_allocation = GFP_KERNEL;
2697 sk->sk_rcvbuf = sysctl_rmem_default;
2698 sk->sk_sndbuf = sysctl_wmem_default;
2699 sk->sk_state = TCP_CLOSE;
2700 sk_set_socket(sk, sock);
2702 sock_set_flag(sk, SOCK_ZAPPED);
2705 sk->sk_type = sock->type;
2706 sk->sk_wq = sock->wq;
2708 sk->sk_uid = SOCK_INODE(sock)->i_uid;
2711 sk->sk_uid = make_kuid(sock_net(sk)->user_ns, 0);
2714 rwlock_init(&sk->sk_callback_lock);
2715 if (sk->sk_kern_sock)
2716 lockdep_set_class_and_name(
2717 &sk->sk_callback_lock,
2718 af_kern_callback_keys + sk->sk_family,
2719 af_family_kern_clock_key_strings[sk->sk_family]);
2721 lockdep_set_class_and_name(
2722 &sk->sk_callback_lock,
2723 af_callback_keys + sk->sk_family,
2724 af_family_clock_key_strings[sk->sk_family]);
2726 sk->sk_state_change = sock_def_wakeup;
2727 sk->sk_data_ready = sock_def_readable;
2728 sk->sk_write_space = sock_def_write_space;
2729 sk->sk_error_report = sock_def_error_report;
2730 sk->sk_destruct = sock_def_destruct;
2732 sk->sk_frag.page = NULL;
2733 sk->sk_frag.offset = 0;
2734 sk->sk_peek_off = -1;
2736 sk->sk_peer_pid = NULL;
2737 sk->sk_peer_cred = NULL;
2738 sk->sk_write_pending = 0;
2739 sk->sk_rcvlowat = 1;
2740 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2741 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2743 sk->sk_stamp = SK_DEFAULT_STAMP;
2744 atomic_set(&sk->sk_zckey, 0);
2746 #ifdef CONFIG_NET_RX_BUSY_POLL
2748 sk->sk_ll_usec = sysctl_net_busy_read;
2751 sk->sk_max_pacing_rate = ~0U;
2752 sk->sk_pacing_rate = ~0U;
2753 sk->sk_incoming_cpu = -1;
2755 * Before updating sk_refcnt, we must commit prior changes to memory
2756 * (Documentation/RCU/rculist_nulls.txt for details)
2759 refcount_set(&sk->sk_refcnt, 1);
2760 atomic_set(&sk->sk_drops, 0);
2762 EXPORT_SYMBOL(sock_init_data);
2764 void lock_sock_nested(struct sock *sk, int subclass)
2767 spin_lock_bh(&sk->sk_lock.slock);
2768 if (sk->sk_lock.owned)
2770 sk->sk_lock.owned = 1;
2771 spin_unlock(&sk->sk_lock.slock);
2773 * The sk_lock has mutex_lock() semantics here:
2775 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
2778 EXPORT_SYMBOL(lock_sock_nested);
2780 void release_sock(struct sock *sk)
2782 spin_lock_bh(&sk->sk_lock.slock);
2783 if (sk->sk_backlog.tail)
2786 /* Warning : release_cb() might need to release sk ownership,
2787 * ie call sock_release_ownership(sk) before us.
2789 if (sk->sk_prot->release_cb)
2790 sk->sk_prot->release_cb(sk);
2792 sock_release_ownership(sk);
2793 if (waitqueue_active(&sk->sk_lock.wq))
2794 wake_up(&sk->sk_lock.wq);
2795 spin_unlock_bh(&sk->sk_lock.slock);
2797 EXPORT_SYMBOL(release_sock);
2800 * lock_sock_fast - fast version of lock_sock
2803 * This version should be used for very small section, where process wont block
2804 * return false if fast path is taken:
2806 * sk_lock.slock locked, owned = 0, BH disabled
2808 * return true if slow path is taken:
2810 * sk_lock.slock unlocked, owned = 1, BH enabled
2812 bool lock_sock_fast(struct sock *sk)
2815 spin_lock_bh(&sk->sk_lock.slock);
2817 if (!sk->sk_lock.owned)
2819 * Note : We must disable BH
2824 sk->sk_lock.owned = 1;
2825 spin_unlock(&sk->sk_lock.slock);
2827 * The sk_lock has mutex_lock() semantics here:
2829 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2833 EXPORT_SYMBOL(lock_sock_fast);
2835 int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
2838 if (!sock_flag(sk, SOCK_TIMESTAMP))
2839 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
2840 tv = ktime_to_timeval(sk->sk_stamp);
2841 if (tv.tv_sec == -1)
2843 if (tv.tv_sec == 0) {
2844 sk->sk_stamp = ktime_get_real();
2845 tv = ktime_to_timeval(sk->sk_stamp);
2847 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
2849 EXPORT_SYMBOL(sock_get_timestamp);
2851 int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2854 if (!sock_flag(sk, SOCK_TIMESTAMP))
2855 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
2856 ts = ktime_to_timespec(sk->sk_stamp);
2857 if (ts.tv_sec == -1)
2859 if (ts.tv_sec == 0) {
2860 sk->sk_stamp = ktime_get_real();
2861 ts = ktime_to_timespec(sk->sk_stamp);
2863 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2865 EXPORT_SYMBOL(sock_get_timestampns);
2867 void sock_enable_timestamp(struct sock *sk, int flag)
2869 if (!sock_flag(sk, flag)) {
2870 unsigned long previous_flags = sk->sk_flags;
2872 sock_set_flag(sk, flag);
2874 * we just set one of the two flags which require net
2875 * time stamping, but time stamping might have been on
2876 * already because of the other one
2878 if (sock_needs_netstamp(sk) &&
2879 !(previous_flags & SK_FLAGS_TIMESTAMP))
2880 net_enable_timestamp();
2884 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2885 int level, int type)
2887 struct sock_exterr_skb *serr;
2888 struct sk_buff *skb;
2892 skb = sock_dequeue_err_skb(sk);
2898 msg->msg_flags |= MSG_TRUNC;
2901 err = skb_copy_datagram_msg(skb, 0, msg, copied);
2905 sock_recv_timestamp(msg, sk, skb);
2907 serr = SKB_EXT_ERR(skb);
2908 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2910 msg->msg_flags |= MSG_ERRQUEUE;
2918 EXPORT_SYMBOL(sock_recv_errqueue);
2921 * Get a socket option on an socket.
2923 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2924 * asynchronous errors should be reported by getsockopt. We assume
2925 * this means if you specify SO_ERROR (otherwise whats the point of it).
2927 int sock_common_getsockopt(struct socket *sock, int level, int optname,
2928 char __user *optval, int __user *optlen)
2930 struct sock *sk = sock->sk;
2932 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2934 EXPORT_SYMBOL(sock_common_getsockopt);
2936 #ifdef CONFIG_COMPAT
2937 int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2938 char __user *optval, int __user *optlen)
2940 struct sock *sk = sock->sk;
2942 if (sk->sk_prot->compat_getsockopt != NULL)
2943 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2945 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2947 EXPORT_SYMBOL(compat_sock_common_getsockopt);
2950 int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2953 struct sock *sk = sock->sk;
2957 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
2958 flags & ~MSG_DONTWAIT, &addr_len);
2960 msg->msg_namelen = addr_len;
2963 EXPORT_SYMBOL(sock_common_recvmsg);
2966 * Set socket options on an inet socket.
2968 int sock_common_setsockopt(struct socket *sock, int level, int optname,
2969 char __user *optval, unsigned int optlen)
2971 struct sock *sk = sock->sk;
2973 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2975 EXPORT_SYMBOL(sock_common_setsockopt);
2977 #ifdef CONFIG_COMPAT
2978 int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
2979 char __user *optval, unsigned int optlen)
2981 struct sock *sk = sock->sk;
2983 if (sk->sk_prot->compat_setsockopt != NULL)
2984 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2986 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2988 EXPORT_SYMBOL(compat_sock_common_setsockopt);
2991 void sk_common_release(struct sock *sk)
2993 if (sk->sk_prot->destroy)
2994 sk->sk_prot->destroy(sk);
2997 * Observation: when sock_common_release is called, processes have
2998 * no access to socket. But net still has.
2999 * Step one, detach it from networking:
3001 * A. Remove from hash tables.
3004 sk->sk_prot->unhash(sk);
3007 * In this point socket cannot receive new packets, but it is possible
3008 * that some packets are in flight because some CPU runs receiver and
3009 * did hash table lookup before we unhashed socket. They will achieve
3010 * receive queue and will be purged by socket destructor.
3012 * Also we still have packets pending on receive queue and probably,
3013 * our own packets waiting in device queues. sock_destroy will drain
3014 * receive queue, but transmitted packets will delay socket destruction
3015 * until the last reference will be released.
3020 xfrm_sk_free_policy(sk);
3022 sk_refcnt_debug_release(sk);
3026 EXPORT_SYMBOL(sk_common_release);
3028 void sk_get_meminfo(const struct sock *sk, u32 *mem)
3030 memset(mem, 0, sizeof(*mem) * SK_MEMINFO_VARS);
3032 mem[SK_MEMINFO_RMEM_ALLOC] = sk_rmem_alloc_get(sk);
3033 mem[SK_MEMINFO_RCVBUF] = sk->sk_rcvbuf;
3034 mem[SK_MEMINFO_WMEM_ALLOC] = sk_wmem_alloc_get(sk);
3035 mem[SK_MEMINFO_SNDBUF] = sk->sk_sndbuf;
3036 mem[SK_MEMINFO_FWD_ALLOC] = sk->sk_forward_alloc;
3037 mem[SK_MEMINFO_WMEM_QUEUED] = sk->sk_wmem_queued;
3038 mem[SK_MEMINFO_OPTMEM] = atomic_read(&sk->sk_omem_alloc);
3039 mem[SK_MEMINFO_BACKLOG] = sk->sk_backlog.len;
3040 mem[SK_MEMINFO_DROPS] = atomic_read(&sk->sk_drops);
3043 #ifdef CONFIG_PROC_FS
3044 #define PROTO_INUSE_NR 64 /* should be enough for the first time */
3046 int val[PROTO_INUSE_NR];
3049 static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
3051 #ifdef CONFIG_NET_NS
3052 void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
3054 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
3056 EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
3058 int sock_prot_inuse_get(struct net *net, struct proto *prot)
3060 int cpu, idx = prot->inuse_idx;
3063 for_each_possible_cpu(cpu)
3064 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
3066 return res >= 0 ? res : 0;
3068 EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
3070 static int __net_init sock_inuse_init_net(struct net *net)
3072 net->core.inuse = alloc_percpu(struct prot_inuse);
3073 return net->core.inuse ? 0 : -ENOMEM;
3076 static void __net_exit sock_inuse_exit_net(struct net *net)
3078 free_percpu(net->core.inuse);
3081 static struct pernet_operations net_inuse_ops = {
3082 .init = sock_inuse_init_net,
3083 .exit = sock_inuse_exit_net,
3086 static __init int net_inuse_init(void)
3088 if (register_pernet_subsys(&net_inuse_ops))
3089 panic("Cannot initialize net inuse counters");
3094 core_initcall(net_inuse_init);
3096 static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
3098 void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
3100 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
3102 EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
3104 int sock_prot_inuse_get(struct net *net, struct proto *prot)
3106 int cpu, idx = prot->inuse_idx;
3109 for_each_possible_cpu(cpu)
3110 res += per_cpu(prot_inuse, cpu).val[idx];
3112 return res >= 0 ? res : 0;
3114 EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
3117 static void assign_proto_idx(struct proto *prot)
3119 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
3121 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
3122 pr_err("PROTO_INUSE_NR exhausted\n");
3126 set_bit(prot->inuse_idx, proto_inuse_idx);
3129 static void release_proto_idx(struct proto *prot)
3131 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
3132 clear_bit(prot->inuse_idx, proto_inuse_idx);
3135 static inline void assign_proto_idx(struct proto *prot)
3139 static inline void release_proto_idx(struct proto *prot)
3144 static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
3148 kfree(rsk_prot->slab_name);
3149 rsk_prot->slab_name = NULL;
3150 kmem_cache_destroy(rsk_prot->slab);
3151 rsk_prot->slab = NULL;
3154 static int req_prot_init(const struct proto *prot)
3156 struct request_sock_ops *rsk_prot = prot->rsk_prot;
3161 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
3163 if (!rsk_prot->slab_name)
3166 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
3167 rsk_prot->obj_size, 0,
3168 prot->slab_flags, NULL);
3170 if (!rsk_prot->slab) {
3171 pr_crit("%s: Can't create request sock SLAB cache!\n",
3178 int proto_register(struct proto *prot, int alloc_slab)
3181 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
3182 SLAB_HWCACHE_ALIGN | prot->slab_flags,
3185 if (prot->slab == NULL) {
3186 pr_crit("%s: Can't create sock SLAB cache!\n",
3191 if (req_prot_init(prot))
3192 goto out_free_request_sock_slab;
3194 if (prot->twsk_prot != NULL) {
3195 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
3197 if (prot->twsk_prot->twsk_slab_name == NULL)
3198 goto out_free_request_sock_slab;
3200 prot->twsk_prot->twsk_slab =
3201 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
3202 prot->twsk_prot->twsk_obj_size,
3206 if (prot->twsk_prot->twsk_slab == NULL)
3207 goto out_free_timewait_sock_slab_name;
3211 mutex_lock(&proto_list_mutex);
3212 list_add(&prot->node, &proto_list);
3213 assign_proto_idx(prot);
3214 mutex_unlock(&proto_list_mutex);
3217 out_free_timewait_sock_slab_name:
3218 kfree(prot->twsk_prot->twsk_slab_name);
3219 out_free_request_sock_slab:
3220 req_prot_cleanup(prot->rsk_prot);
3222 kmem_cache_destroy(prot->slab);
3227 EXPORT_SYMBOL(proto_register);
3229 void proto_unregister(struct proto *prot)
3231 mutex_lock(&proto_list_mutex);
3232 release_proto_idx(prot);
3233 list_del(&prot->node);
3234 mutex_unlock(&proto_list_mutex);
3236 kmem_cache_destroy(prot->slab);
3239 req_prot_cleanup(prot->rsk_prot);
3241 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
3242 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
3243 kfree(prot->twsk_prot->twsk_slab_name);
3244 prot->twsk_prot->twsk_slab = NULL;
3247 EXPORT_SYMBOL(proto_unregister);
3249 #ifdef CONFIG_PROC_FS
3250 static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
3251 __acquires(proto_list_mutex)
3253 mutex_lock(&proto_list_mutex);
3254 return seq_list_start_head(&proto_list, *pos);
3257 static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
3259 return seq_list_next(v, &proto_list, pos);
3262 static void proto_seq_stop(struct seq_file *seq, void *v)
3263 __releases(proto_list_mutex)
3265 mutex_unlock(&proto_list_mutex);
3268 static char proto_method_implemented(const void *method)
3270 return method == NULL ? 'n' : 'y';
3272 static long sock_prot_memory_allocated(struct proto *proto)
3274 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
3277 static char *sock_prot_memory_pressure(struct proto *proto)
3279 return proto->memory_pressure != NULL ?
3280 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
3283 static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
3286 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
3287 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
3290 sock_prot_inuse_get(seq_file_net(seq), proto),
3291 sock_prot_memory_allocated(proto),
3292 sock_prot_memory_pressure(proto),
3294 proto->slab == NULL ? "no" : "yes",
3295 module_name(proto->owner),
3296 proto_method_implemented(proto->close),
3297 proto_method_implemented(proto->connect),
3298 proto_method_implemented(proto->disconnect),
3299 proto_method_implemented(proto->accept),
3300 proto_method_implemented(proto->ioctl),
3301 proto_method_implemented(proto->init),
3302 proto_method_implemented(proto->destroy),
3303 proto_method_implemented(proto->shutdown),
3304 proto_method_implemented(proto->setsockopt),
3305 proto_method_implemented(proto->getsockopt),
3306 proto_method_implemented(proto->sendmsg),
3307 proto_method_implemented(proto->recvmsg),
3308 proto_method_implemented(proto->sendpage),
3309 proto_method_implemented(proto->bind),
3310 proto_method_implemented(proto->backlog_rcv),
3311 proto_method_implemented(proto->hash),
3312 proto_method_implemented(proto->unhash),
3313 proto_method_implemented(proto->get_port),
3314 proto_method_implemented(proto->enter_memory_pressure));
3317 static int proto_seq_show(struct seq_file *seq, void *v)
3319 if (v == &proto_list)
3320 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3329 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3331 proto_seq_printf(seq, list_entry(v, struct proto, node));
3335 static const struct seq_operations proto_seq_ops = {
3336 .start = proto_seq_start,
3337 .next = proto_seq_next,
3338 .stop = proto_seq_stop,
3339 .show = proto_seq_show,
3342 static int proto_seq_open(struct inode *inode, struct file *file)
3344 return seq_open_net(inode, file, &proto_seq_ops,
3345 sizeof(struct seq_net_private));
3348 static const struct file_operations proto_seq_fops = {
3349 .owner = THIS_MODULE,
3350 .open = proto_seq_open,
3352 .llseek = seq_lseek,
3353 .release = seq_release_net,
3356 static __net_init int proto_init_net(struct net *net)
3358 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
3364 static __net_exit void proto_exit_net(struct net *net)
3366 remove_proc_entry("protocols", net->proc_net);
3370 static __net_initdata struct pernet_operations proto_net_ops = {
3371 .init = proto_init_net,
3372 .exit = proto_exit_net,
3375 static int __init proto_init(void)
3377 return register_pernet_subsys(&proto_net_ops);
3380 subsys_initcall(proto_init);
3382 #endif /* PROC_FS */
3384 #ifdef CONFIG_NET_RX_BUSY_POLL
3385 bool sk_busy_loop_end(void *p, unsigned long start_time)
3387 struct sock *sk = p;
3389 return !skb_queue_empty(&sk->sk_receive_queue) ||
3390 sk_busy_loop_timeout(sk, start_time);
3392 EXPORT_SYMBOL(sk_busy_loop_end);
3393 #endif /* CONFIG_NET_RX_BUSY_POLL */