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/timer.h>
106 #include <linux/string.h>
107 #include <linux/sockios.h>
108 #include <linux/net.h>
109 #include <linux/mm.h>
110 #include <linux/slab.h>
111 #include <linux/interrupt.h>
112 #include <linux/poll.h>
113 #include <linux/tcp.h>
114 #include <linux/init.h>
115 #include <linux/highmem.h>
116 #include <linux/user_namespace.h>
117 #include <linux/static_key.h>
118 #include <linux/memcontrol.h>
119 #include <linux/prefetch.h>
121 #include <linux/uaccess.h>
123 #include <linux/netdevice.h>
124 #include <net/protocol.h>
125 #include <linux/skbuff.h>
126 #include <net/net_namespace.h>
127 #include <net/request_sock.h>
128 #include <net/sock.h>
129 #include <linux/net_tstamp.h>
130 #include <net/xfrm.h>
131 #include <linux/ipsec.h>
132 #include <net/cls_cgroup.h>
133 #include <net/netprio_cgroup.h>
134 #include <linux/sock_diag.h>
136 #include <linux/filter.h>
137 #include <net/sock_reuseport.h>
139 #include <trace/events/sock.h>
145 #include <net/busy_poll.h>
147 static DEFINE_MUTEX(proto_list_mutex
);
148 static LIST_HEAD(proto_list
);
151 * sk_ns_capable - General socket capability test
152 * @sk: Socket to use a capability on or through
153 * @user_ns: The user namespace of the capability to use
154 * @cap: The capability to use
156 * Test to see if the opener of the socket had when the socket was
157 * created and the current process has the capability @cap in the user
158 * namespace @user_ns.
160 bool sk_ns_capable(const struct sock
*sk
,
161 struct user_namespace
*user_ns
, int cap
)
163 return file_ns_capable(sk
->sk_socket
->file
, user_ns
, cap
) &&
164 ns_capable(user_ns
, cap
);
166 EXPORT_SYMBOL(sk_ns_capable
);
169 * sk_capable - Socket global capability test
170 * @sk: Socket to use a capability on or through
171 * @cap: The global capability to use
173 * Test to see if the opener of the socket had when the socket was
174 * created and the current process has the capability @cap in all user
177 bool sk_capable(const struct sock
*sk
, int cap
)
179 return sk_ns_capable(sk
, &init_user_ns
, cap
);
181 EXPORT_SYMBOL(sk_capable
);
184 * sk_net_capable - Network namespace socket capability test
185 * @sk: Socket to use a capability on or through
186 * @cap: The capability to use
188 * Test to see if the opener of the socket had when the socket was created
189 * and the current process has the capability @cap over the network namespace
190 * the socket is a member of.
192 bool sk_net_capable(const struct sock
*sk
, int cap
)
194 return sk_ns_capable(sk
, sock_net(sk
)->user_ns
, cap
);
196 EXPORT_SYMBOL(sk_net_capable
);
199 * Each address family might have different locking rules, so we have
200 * one slock key per address family:
202 static struct lock_class_key af_family_keys
[AF_MAX
];
203 static struct lock_class_key af_family_slock_keys
[AF_MAX
];
206 * Make lock validator output more readable. (we pre-construct these
207 * strings build-time, so that runtime initialization of socket
210 static const char *const af_family_key_strings
[AF_MAX
+1] = {
211 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
212 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
213 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
214 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
215 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
216 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
217 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
218 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
219 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
220 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
221 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
222 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
223 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
224 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_KCM" ,
225 "sk_lock-AF_QIPCRTR", "sk_lock-AF_MAX"
227 static const char *const af_family_slock_key_strings
[AF_MAX
+1] = {
228 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
229 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
230 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
231 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
232 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
233 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
234 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
235 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
236 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
237 "slock-27" , "slock-28" , "slock-AF_CAN" ,
238 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
239 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
240 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
241 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_KCM" ,
242 "slock-AF_QIPCRTR", "slock-AF_MAX"
244 static const char *const af_family_clock_key_strings
[AF_MAX
+1] = {
245 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
246 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
247 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
248 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
249 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
250 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
251 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
252 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
253 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
254 "clock-27" , "clock-28" , "clock-AF_CAN" ,
255 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
256 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
257 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
258 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_KCM" ,
259 "clock-AF_QIPCRTR", "clock-AF_MAX"
263 * sk_callback_lock locking rules are per-address-family,
264 * so split the lock classes by using a per-AF key:
266 static struct lock_class_key af_callback_keys
[AF_MAX
];
268 /* Take into consideration the size of the struct sk_buff overhead in the
269 * determination of these values, since that is non-constant across
270 * platforms. This makes socket queueing behavior and performance
271 * not depend upon such differences.
273 #define _SK_MEM_PACKETS 256
274 #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
275 #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
276 #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
278 /* Run time adjustable parameters. */
279 __u32 sysctl_wmem_max __read_mostly
= SK_WMEM_MAX
;
280 EXPORT_SYMBOL(sysctl_wmem_max
);
281 __u32 sysctl_rmem_max __read_mostly
= SK_RMEM_MAX
;
282 EXPORT_SYMBOL(sysctl_rmem_max
);
283 __u32 sysctl_wmem_default __read_mostly
= SK_WMEM_MAX
;
284 __u32 sysctl_rmem_default __read_mostly
= SK_RMEM_MAX
;
286 /* Maximal space eaten by iovec or ancillary data plus some space */
287 int sysctl_optmem_max __read_mostly
= sizeof(unsigned long)*(2*UIO_MAXIOV
+512);
288 EXPORT_SYMBOL(sysctl_optmem_max
);
290 int sysctl_tstamp_allow_data __read_mostly
= 1;
292 struct static_key memalloc_socks
= STATIC_KEY_INIT_FALSE
;
293 EXPORT_SYMBOL_GPL(memalloc_socks
);
296 * sk_set_memalloc - sets %SOCK_MEMALLOC
297 * @sk: socket to set it on
299 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
300 * It's the responsibility of the admin to adjust min_free_kbytes
301 * to meet the requirements
303 void sk_set_memalloc(struct sock
*sk
)
305 sock_set_flag(sk
, SOCK_MEMALLOC
);
306 sk
->sk_allocation
|= __GFP_MEMALLOC
;
307 static_key_slow_inc(&memalloc_socks
);
309 EXPORT_SYMBOL_GPL(sk_set_memalloc
);
311 void sk_clear_memalloc(struct sock
*sk
)
313 sock_reset_flag(sk
, SOCK_MEMALLOC
);
314 sk
->sk_allocation
&= ~__GFP_MEMALLOC
;
315 static_key_slow_dec(&memalloc_socks
);
318 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
319 * progress of swapping. SOCK_MEMALLOC may be cleared while
320 * it has rmem allocations due to the last swapfile being deactivated
321 * but there is a risk that the socket is unusable due to exceeding
322 * the rmem limits. Reclaim the reserves and obey rmem limits again.
326 EXPORT_SYMBOL_GPL(sk_clear_memalloc
);
328 int __sk_backlog_rcv(struct sock
*sk
, struct sk_buff
*skb
)
331 unsigned long pflags
= current
->flags
;
333 /* these should have been dropped before queueing */
334 BUG_ON(!sock_flag(sk
, SOCK_MEMALLOC
));
336 current
->flags
|= PF_MEMALLOC
;
337 ret
= sk
->sk_backlog_rcv(sk
, skb
);
338 tsk_restore_flags(current
, pflags
, PF_MEMALLOC
);
342 EXPORT_SYMBOL(__sk_backlog_rcv
);
344 static int sock_set_timeout(long *timeo_p
, char __user
*optval
, int optlen
)
348 if (optlen
< sizeof(tv
))
350 if (copy_from_user(&tv
, optval
, sizeof(tv
)))
352 if (tv
.tv_usec
< 0 || tv
.tv_usec
>= USEC_PER_SEC
)
356 static int warned __read_mostly
;
359 if (warned
< 10 && net_ratelimit()) {
361 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
362 __func__
, current
->comm
, task_pid_nr(current
));
366 *timeo_p
= MAX_SCHEDULE_TIMEOUT
;
367 if (tv
.tv_sec
== 0 && tv
.tv_usec
== 0)
369 if (tv
.tv_sec
< (MAX_SCHEDULE_TIMEOUT
/HZ
- 1))
370 *timeo_p
= tv
.tv_sec
*HZ
+ (tv
.tv_usec
+(1000000/HZ
-1))/(1000000/HZ
);
374 static void sock_warn_obsolete_bsdism(const char *name
)
377 static char warncomm
[TASK_COMM_LEN
];
378 if (strcmp(warncomm
, current
->comm
) && warned
< 5) {
379 strcpy(warncomm
, current
->comm
);
380 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
386 static bool sock_needs_netstamp(const struct sock
*sk
)
388 switch (sk
->sk_family
) {
397 static void sock_disable_timestamp(struct sock
*sk
, unsigned long flags
)
399 if (sk
->sk_flags
& flags
) {
400 sk
->sk_flags
&= ~flags
;
401 if (sock_needs_netstamp(sk
) &&
402 !(sk
->sk_flags
& SK_FLAGS_TIMESTAMP
))
403 net_disable_timestamp();
408 int __sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
411 struct sk_buff_head
*list
= &sk
->sk_receive_queue
;
413 if (atomic_read(&sk
->sk_rmem_alloc
) >= sk
->sk_rcvbuf
) {
414 atomic_inc(&sk
->sk_drops
);
415 trace_sock_rcvqueue_full(sk
, skb
);
419 if (!sk_rmem_schedule(sk
, skb
, skb
->truesize
)) {
420 atomic_inc(&sk
->sk_drops
);
425 skb_set_owner_r(skb
, sk
);
427 /* we escape from rcu protected region, make sure we dont leak
432 spin_lock_irqsave(&list
->lock
, flags
);
433 sock_skb_set_dropcount(sk
, skb
);
434 __skb_queue_tail(list
, skb
);
435 spin_unlock_irqrestore(&list
->lock
, flags
);
437 if (!sock_flag(sk
, SOCK_DEAD
))
438 sk
->sk_data_ready(sk
);
441 EXPORT_SYMBOL(__sock_queue_rcv_skb
);
443 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
447 err
= sk_filter(sk
, skb
);
451 return __sock_queue_rcv_skb(sk
, skb
);
453 EXPORT_SYMBOL(sock_queue_rcv_skb
);
455 int __sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
,
456 const int nested
, unsigned int trim_cap
, bool refcounted
)
458 int rc
= NET_RX_SUCCESS
;
460 if (sk_filter_trim_cap(sk
, skb
, trim_cap
))
461 goto discard_and_relse
;
465 if (sk_rcvqueues_full(sk
, sk
->sk_rcvbuf
)) {
466 atomic_inc(&sk
->sk_drops
);
467 goto discard_and_relse
;
470 bh_lock_sock_nested(sk
);
473 if (!sock_owned_by_user(sk
)) {
475 * trylock + unlock semantics:
477 mutex_acquire(&sk
->sk_lock
.dep_map
, 0, 1, _RET_IP_
);
479 rc
= sk_backlog_rcv(sk
, skb
);
481 mutex_release(&sk
->sk_lock
.dep_map
, 1, _RET_IP_
);
482 } else if (sk_add_backlog(sk
, skb
, sk
->sk_rcvbuf
)) {
484 atomic_inc(&sk
->sk_drops
);
485 goto discard_and_relse
;
497 EXPORT_SYMBOL(__sk_receive_skb
);
499 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
)
501 struct dst_entry
*dst
= __sk_dst_get(sk
);
503 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, cookie
) == NULL
) {
504 sk_tx_queue_clear(sk
);
505 RCU_INIT_POINTER(sk
->sk_dst_cache
, NULL
);
512 EXPORT_SYMBOL(__sk_dst_check
);
514 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
)
516 struct dst_entry
*dst
= sk_dst_get(sk
);
518 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, cookie
) == NULL
) {
526 EXPORT_SYMBOL(sk_dst_check
);
528 static int sock_setbindtodevice(struct sock
*sk
, char __user
*optval
,
531 int ret
= -ENOPROTOOPT
;
532 #ifdef CONFIG_NETDEVICES
533 struct net
*net
= sock_net(sk
);
534 char devname
[IFNAMSIZ
];
539 if (!ns_capable(net
->user_ns
, CAP_NET_RAW
))
546 /* Bind this socket to a particular device like "eth0",
547 * as specified in the passed interface name. If the
548 * name is "" or the option length is zero the socket
551 if (optlen
> IFNAMSIZ
- 1)
552 optlen
= IFNAMSIZ
- 1;
553 memset(devname
, 0, sizeof(devname
));
556 if (copy_from_user(devname
, optval
, optlen
))
560 if (devname
[0] != '\0') {
561 struct net_device
*dev
;
564 dev
= dev_get_by_name_rcu(net
, devname
);
566 index
= dev
->ifindex
;
574 sk
->sk_bound_dev_if
= index
;
586 static int sock_getbindtodevice(struct sock
*sk
, char __user
*optval
,
587 int __user
*optlen
, int len
)
589 int ret
= -ENOPROTOOPT
;
590 #ifdef CONFIG_NETDEVICES
591 struct net
*net
= sock_net(sk
);
592 char devname
[IFNAMSIZ
];
594 if (sk
->sk_bound_dev_if
== 0) {
603 ret
= netdev_get_name(net
, devname
, sk
->sk_bound_dev_if
);
607 len
= strlen(devname
) + 1;
610 if (copy_to_user(optval
, devname
, len
))
615 if (put_user(len
, optlen
))
626 static inline void sock_valbool_flag(struct sock
*sk
, int bit
, int valbool
)
629 sock_set_flag(sk
, bit
);
631 sock_reset_flag(sk
, bit
);
634 bool sk_mc_loop(struct sock
*sk
)
636 if (dev_recursion_level())
640 switch (sk
->sk_family
) {
642 return inet_sk(sk
)->mc_loop
;
643 #if IS_ENABLED(CONFIG_IPV6)
645 return inet6_sk(sk
)->mc_loop
;
651 EXPORT_SYMBOL(sk_mc_loop
);
654 * This is meant for all protocols to use and covers goings on
655 * at the socket level. Everything here is generic.
658 int sock_setsockopt(struct socket
*sock
, int level
, int optname
,
659 char __user
*optval
, unsigned int optlen
)
661 struct sock
*sk
= sock
->sk
;
668 * Options without arguments
671 if (optname
== SO_BINDTODEVICE
)
672 return sock_setbindtodevice(sk
, optval
, optlen
);
674 if (optlen
< sizeof(int))
677 if (get_user(val
, (int __user
*)optval
))
680 valbool
= val
? 1 : 0;
686 if (val
&& !capable(CAP_NET_ADMIN
))
689 sock_valbool_flag(sk
, SOCK_DBG
, valbool
);
692 sk
->sk_reuse
= (valbool
? SK_CAN_REUSE
: SK_NO_REUSE
);
695 sk
->sk_reuseport
= valbool
;
704 sock_valbool_flag(sk
, SOCK_LOCALROUTE
, valbool
);
707 sock_valbool_flag(sk
, SOCK_BROADCAST
, valbool
);
710 /* Don't error on this BSD doesn't and if you think
711 * about it this is right. Otherwise apps have to
712 * play 'guess the biggest size' games. RCVBUF/SNDBUF
713 * are treated in BSD as hints
715 val
= min_t(u32
, val
, sysctl_wmem_max
);
717 sk
->sk_userlocks
|= SOCK_SNDBUF_LOCK
;
718 sk
->sk_sndbuf
= max_t(int, val
* 2, SOCK_MIN_SNDBUF
);
719 /* Wake up sending tasks if we upped the value. */
720 sk
->sk_write_space(sk
);
724 if (!capable(CAP_NET_ADMIN
)) {
731 /* Don't error on this BSD doesn't and if you think
732 * about it this is right. Otherwise apps have to
733 * play 'guess the biggest size' games. RCVBUF/SNDBUF
734 * are treated in BSD as hints
736 val
= min_t(u32
, val
, sysctl_rmem_max
);
738 sk
->sk_userlocks
|= SOCK_RCVBUF_LOCK
;
740 * We double it on the way in to account for
741 * "struct sk_buff" etc. overhead. Applications
742 * assume that the SO_RCVBUF setting they make will
743 * allow that much actual data to be received on that
746 * Applications are unaware that "struct sk_buff" and
747 * other overheads allocate from the receive buffer
748 * during socket buffer allocation.
750 * And after considering the possible alternatives,
751 * returning the value we actually used in getsockopt
752 * is the most desirable behavior.
754 sk
->sk_rcvbuf
= max_t(int, val
* 2, SOCK_MIN_RCVBUF
);
758 if (!capable(CAP_NET_ADMIN
)) {
766 if (sk
->sk_protocol
== IPPROTO_TCP
&&
767 sk
->sk_type
== SOCK_STREAM
)
768 tcp_set_keepalive(sk
, valbool
);
770 sock_valbool_flag(sk
, SOCK_KEEPOPEN
, valbool
);
774 sock_valbool_flag(sk
, SOCK_URGINLINE
, valbool
);
778 sk
->sk_no_check_tx
= valbool
;
782 if ((val
>= 0 && val
<= 6) ||
783 ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
784 sk
->sk_priority
= val
;
790 if (optlen
< sizeof(ling
)) {
791 ret
= -EINVAL
; /* 1003.1g */
794 if (copy_from_user(&ling
, optval
, sizeof(ling
))) {
799 sock_reset_flag(sk
, SOCK_LINGER
);
801 #if (BITS_PER_LONG == 32)
802 if ((unsigned int)ling
.l_linger
>= MAX_SCHEDULE_TIMEOUT
/HZ
)
803 sk
->sk_lingertime
= MAX_SCHEDULE_TIMEOUT
;
806 sk
->sk_lingertime
= (unsigned int)ling
.l_linger
* HZ
;
807 sock_set_flag(sk
, SOCK_LINGER
);
812 sock_warn_obsolete_bsdism("setsockopt");
817 set_bit(SOCK_PASSCRED
, &sock
->flags
);
819 clear_bit(SOCK_PASSCRED
, &sock
->flags
);
825 if (optname
== SO_TIMESTAMP
)
826 sock_reset_flag(sk
, SOCK_RCVTSTAMPNS
);
828 sock_set_flag(sk
, SOCK_RCVTSTAMPNS
);
829 sock_set_flag(sk
, SOCK_RCVTSTAMP
);
830 sock_enable_timestamp(sk
, SOCK_TIMESTAMP
);
832 sock_reset_flag(sk
, SOCK_RCVTSTAMP
);
833 sock_reset_flag(sk
, SOCK_RCVTSTAMPNS
);
837 case SO_TIMESTAMPING
:
838 if (val
& ~SOF_TIMESTAMPING_MASK
) {
843 if (val
& SOF_TIMESTAMPING_OPT_ID
&&
844 !(sk
->sk_tsflags
& SOF_TIMESTAMPING_OPT_ID
)) {
845 if (sk
->sk_protocol
== IPPROTO_TCP
&&
846 sk
->sk_type
== SOCK_STREAM
) {
847 if ((1 << sk
->sk_state
) &
848 (TCPF_CLOSE
| TCPF_LISTEN
)) {
852 sk
->sk_tskey
= tcp_sk(sk
)->snd_una
;
858 if (val
& SOF_TIMESTAMPING_OPT_STATS
&&
859 !(val
& SOF_TIMESTAMPING_OPT_TSONLY
)) {
864 sk
->sk_tsflags
= val
;
865 if (val
& SOF_TIMESTAMPING_RX_SOFTWARE
)
866 sock_enable_timestamp(sk
,
867 SOCK_TIMESTAMPING_RX_SOFTWARE
);
869 sock_disable_timestamp(sk
,
870 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE
));
876 sk
->sk_rcvlowat
= val
? : 1;
880 ret
= sock_set_timeout(&sk
->sk_rcvtimeo
, optval
, optlen
);
884 ret
= sock_set_timeout(&sk
->sk_sndtimeo
, optval
, optlen
);
887 case SO_ATTACH_FILTER
:
889 if (optlen
== sizeof(struct sock_fprog
)) {
890 struct sock_fprog fprog
;
893 if (copy_from_user(&fprog
, optval
, sizeof(fprog
)))
896 ret
= sk_attach_filter(&fprog
, sk
);
902 if (optlen
== sizeof(u32
)) {
906 if (copy_from_user(&ufd
, optval
, sizeof(ufd
)))
909 ret
= sk_attach_bpf(ufd
, sk
);
913 case SO_ATTACH_REUSEPORT_CBPF
:
915 if (optlen
== sizeof(struct sock_fprog
)) {
916 struct sock_fprog fprog
;
919 if (copy_from_user(&fprog
, optval
, sizeof(fprog
)))
922 ret
= sk_reuseport_attach_filter(&fprog
, sk
);
926 case SO_ATTACH_REUSEPORT_EBPF
:
928 if (optlen
== sizeof(u32
)) {
932 if (copy_from_user(&ufd
, optval
, sizeof(ufd
)))
935 ret
= sk_reuseport_attach_bpf(ufd
, sk
);
939 case SO_DETACH_FILTER
:
940 ret
= sk_detach_filter(sk
);
944 if (sock_flag(sk
, SOCK_FILTER_LOCKED
) && !valbool
)
947 sock_valbool_flag(sk
, SOCK_FILTER_LOCKED
, valbool
);
952 set_bit(SOCK_PASSSEC
, &sock
->flags
);
954 clear_bit(SOCK_PASSSEC
, &sock
->flags
);
957 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
964 sock_valbool_flag(sk
, SOCK_RXQ_OVFL
, valbool
);
968 sock_valbool_flag(sk
, SOCK_WIFI_STATUS
, valbool
);
972 if (sock
->ops
->set_peek_off
)
973 ret
= sock
->ops
->set_peek_off(sk
, val
);
979 sock_valbool_flag(sk
, SOCK_NOFCS
, valbool
);
982 case SO_SELECT_ERR_QUEUE
:
983 sock_valbool_flag(sk
, SOCK_SELECT_ERR_QUEUE
, valbool
);
986 #ifdef CONFIG_NET_RX_BUSY_POLL
988 /* allow unprivileged users to decrease the value */
989 if ((val
> sk
->sk_ll_usec
) && !capable(CAP_NET_ADMIN
))
995 sk
->sk_ll_usec
= val
;
1000 case SO_MAX_PACING_RATE
:
1001 sk
->sk_max_pacing_rate
= val
;
1002 sk
->sk_pacing_rate
= min(sk
->sk_pacing_rate
,
1003 sk
->sk_max_pacing_rate
);
1006 case SO_INCOMING_CPU
:
1007 sk
->sk_incoming_cpu
= val
;
1012 dst_negative_advice(sk
);
1021 EXPORT_SYMBOL(sock_setsockopt
);
1024 static void cred_to_ucred(struct pid
*pid
, const struct cred
*cred
,
1025 struct ucred
*ucred
)
1027 ucred
->pid
= pid_vnr(pid
);
1028 ucred
->uid
= ucred
->gid
= -1;
1030 struct user_namespace
*current_ns
= current_user_ns();
1032 ucred
->uid
= from_kuid_munged(current_ns
, cred
->euid
);
1033 ucred
->gid
= from_kgid_munged(current_ns
, cred
->egid
);
1037 int sock_getsockopt(struct socket
*sock
, int level
, int optname
,
1038 char __user
*optval
, int __user
*optlen
)
1040 struct sock
*sk
= sock
->sk
;
1048 int lv
= sizeof(int);
1051 if (get_user(len
, optlen
))
1056 memset(&v
, 0, sizeof(v
));
1060 v
.val
= sock_flag(sk
, SOCK_DBG
);
1064 v
.val
= sock_flag(sk
, SOCK_LOCALROUTE
);
1068 v
.val
= sock_flag(sk
, SOCK_BROADCAST
);
1072 v
.val
= sk
->sk_sndbuf
;
1076 v
.val
= sk
->sk_rcvbuf
;
1080 v
.val
= sk
->sk_reuse
;
1084 v
.val
= sk
->sk_reuseport
;
1088 v
.val
= sock_flag(sk
, SOCK_KEEPOPEN
);
1092 v
.val
= sk
->sk_type
;
1096 v
.val
= sk
->sk_protocol
;
1100 v
.val
= sk
->sk_family
;
1104 v
.val
= -sock_error(sk
);
1106 v
.val
= xchg(&sk
->sk_err_soft
, 0);
1110 v
.val
= sock_flag(sk
, SOCK_URGINLINE
);
1114 v
.val
= sk
->sk_no_check_tx
;
1118 v
.val
= sk
->sk_priority
;
1122 lv
= sizeof(v
.ling
);
1123 v
.ling
.l_onoff
= sock_flag(sk
, SOCK_LINGER
);
1124 v
.ling
.l_linger
= sk
->sk_lingertime
/ HZ
;
1128 sock_warn_obsolete_bsdism("getsockopt");
1132 v
.val
= sock_flag(sk
, SOCK_RCVTSTAMP
) &&
1133 !sock_flag(sk
, SOCK_RCVTSTAMPNS
);
1136 case SO_TIMESTAMPNS
:
1137 v
.val
= sock_flag(sk
, SOCK_RCVTSTAMPNS
);
1140 case SO_TIMESTAMPING
:
1141 v
.val
= sk
->sk_tsflags
;
1145 lv
= sizeof(struct timeval
);
1146 if (sk
->sk_rcvtimeo
== MAX_SCHEDULE_TIMEOUT
) {
1150 v
.tm
.tv_sec
= sk
->sk_rcvtimeo
/ HZ
;
1151 v
.tm
.tv_usec
= ((sk
->sk_rcvtimeo
% HZ
) * 1000000) / HZ
;
1156 lv
= sizeof(struct timeval
);
1157 if (sk
->sk_sndtimeo
== MAX_SCHEDULE_TIMEOUT
) {
1161 v
.tm
.tv_sec
= sk
->sk_sndtimeo
/ HZ
;
1162 v
.tm
.tv_usec
= ((sk
->sk_sndtimeo
% HZ
) * 1000000) / HZ
;
1167 v
.val
= sk
->sk_rcvlowat
;
1175 v
.val
= !!test_bit(SOCK_PASSCRED
, &sock
->flags
);
1180 struct ucred peercred
;
1181 if (len
> sizeof(peercred
))
1182 len
= sizeof(peercred
);
1183 cred_to_ucred(sk
->sk_peer_pid
, sk
->sk_peer_cred
, &peercred
);
1184 if (copy_to_user(optval
, &peercred
, len
))
1193 if (sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &lv
, 2))
1197 if (copy_to_user(optval
, address
, len
))
1202 /* Dubious BSD thing... Probably nobody even uses it, but
1203 * the UNIX standard wants it for whatever reason... -DaveM
1206 v
.val
= sk
->sk_state
== TCP_LISTEN
;
1210 v
.val
= !!test_bit(SOCK_PASSSEC
, &sock
->flags
);
1214 return security_socket_getpeersec_stream(sock
, optval
, optlen
, len
);
1217 v
.val
= sk
->sk_mark
;
1221 v
.val
= sock_flag(sk
, SOCK_RXQ_OVFL
);
1224 case SO_WIFI_STATUS
:
1225 v
.val
= sock_flag(sk
, SOCK_WIFI_STATUS
);
1229 if (!sock
->ops
->set_peek_off
)
1232 v
.val
= sk
->sk_peek_off
;
1235 v
.val
= sock_flag(sk
, SOCK_NOFCS
);
1238 case SO_BINDTODEVICE
:
1239 return sock_getbindtodevice(sk
, optval
, optlen
, len
);
1242 len
= sk_get_filter(sk
, (struct sock_filter __user
*)optval
, len
);
1248 case SO_LOCK_FILTER
:
1249 v
.val
= sock_flag(sk
, SOCK_FILTER_LOCKED
);
1252 case SO_BPF_EXTENSIONS
:
1253 v
.val
= bpf_tell_extensions();
1256 case SO_SELECT_ERR_QUEUE
:
1257 v
.val
= sock_flag(sk
, SOCK_SELECT_ERR_QUEUE
);
1260 #ifdef CONFIG_NET_RX_BUSY_POLL
1262 v
.val
= sk
->sk_ll_usec
;
1266 case SO_MAX_PACING_RATE
:
1267 v
.val
= sk
->sk_max_pacing_rate
;
1270 case SO_INCOMING_CPU
:
1271 v
.val
= sk
->sk_incoming_cpu
;
1275 /* We implement the SO_SNDLOWAT etc to not be settable
1278 return -ENOPROTOOPT
;
1283 if (copy_to_user(optval
, &v
, len
))
1286 if (put_user(len
, optlen
))
1292 * Initialize an sk_lock.
1294 * (We also register the sk_lock with the lock validator.)
1296 static inline void sock_lock_init(struct sock
*sk
)
1298 sock_lock_init_class_and_name(sk
,
1299 af_family_slock_key_strings
[sk
->sk_family
],
1300 af_family_slock_keys
+ sk
->sk_family
,
1301 af_family_key_strings
[sk
->sk_family
],
1302 af_family_keys
+ sk
->sk_family
);
1306 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1307 * even temporarly, because of RCU lookups. sk_node should also be left as is.
1308 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
1310 static void sock_copy(struct sock
*nsk
, const struct sock
*osk
)
1312 #ifdef CONFIG_SECURITY_NETWORK
1313 void *sptr
= nsk
->sk_security
;
1315 memcpy(nsk
, osk
, offsetof(struct sock
, sk_dontcopy_begin
));
1317 memcpy(&nsk
->sk_dontcopy_end
, &osk
->sk_dontcopy_end
,
1318 osk
->sk_prot
->obj_size
- offsetof(struct sock
, sk_dontcopy_end
));
1320 #ifdef CONFIG_SECURITY_NETWORK
1321 nsk
->sk_security
= sptr
;
1322 security_sk_clone(osk
, nsk
);
1326 static struct sock
*sk_prot_alloc(struct proto
*prot
, gfp_t priority
,
1330 struct kmem_cache
*slab
;
1334 sk
= kmem_cache_alloc(slab
, priority
& ~__GFP_ZERO
);
1337 if (priority
& __GFP_ZERO
)
1338 sk_prot_clear_nulls(sk
, prot
->obj_size
);
1340 sk
= kmalloc(prot
->obj_size
, priority
);
1343 kmemcheck_annotate_bitfield(sk
, flags
);
1345 if (security_sk_alloc(sk
, family
, priority
))
1348 if (!try_module_get(prot
->owner
))
1350 sk_tx_queue_clear(sk
);
1356 security_sk_free(sk
);
1359 kmem_cache_free(slab
, sk
);
1365 static void sk_prot_free(struct proto
*prot
, struct sock
*sk
)
1367 struct kmem_cache
*slab
;
1368 struct module
*owner
;
1370 owner
= prot
->owner
;
1373 cgroup_sk_free(&sk
->sk_cgrp_data
);
1374 mem_cgroup_sk_free(sk
);
1375 security_sk_free(sk
);
1377 kmem_cache_free(slab
, sk
);
1384 * sk_alloc - All socket objects are allocated here
1385 * @net: the applicable net namespace
1386 * @family: protocol family
1387 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1388 * @prot: struct proto associated with this new sock instance
1389 * @kern: is this to be a kernel socket?
1391 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1392 struct proto
*prot
, int kern
)
1396 sk
= sk_prot_alloc(prot
, priority
| __GFP_ZERO
, family
);
1398 sk
->sk_family
= family
;
1400 * See comment in struct sock definition to understand
1401 * why we need sk_prot_creator -acme
1403 sk
->sk_prot
= sk
->sk_prot_creator
= prot
;
1405 sk
->sk_net_refcnt
= kern
? 0 : 1;
1406 if (likely(sk
->sk_net_refcnt
))
1408 sock_net_set(sk
, net
);
1409 atomic_set(&sk
->sk_wmem_alloc
, 1);
1411 mem_cgroup_sk_alloc(sk
);
1412 cgroup_sk_alloc(&sk
->sk_cgrp_data
);
1413 sock_update_classid(&sk
->sk_cgrp_data
);
1414 sock_update_netprioidx(&sk
->sk_cgrp_data
);
1419 EXPORT_SYMBOL(sk_alloc
);
1421 /* Sockets having SOCK_RCU_FREE will call this function after one RCU
1422 * grace period. This is the case for UDP sockets and TCP listeners.
1424 static void __sk_destruct(struct rcu_head
*head
)
1426 struct sock
*sk
= container_of(head
, struct sock
, sk_rcu
);
1427 struct sk_filter
*filter
;
1429 if (sk
->sk_destruct
)
1430 sk
->sk_destruct(sk
);
1432 filter
= rcu_dereference_check(sk
->sk_filter
,
1433 atomic_read(&sk
->sk_wmem_alloc
) == 0);
1435 sk_filter_uncharge(sk
, filter
);
1436 RCU_INIT_POINTER(sk
->sk_filter
, NULL
);
1438 if (rcu_access_pointer(sk
->sk_reuseport_cb
))
1439 reuseport_detach_sock(sk
);
1441 sock_disable_timestamp(sk
, SK_FLAGS_TIMESTAMP
);
1443 if (atomic_read(&sk
->sk_omem_alloc
))
1444 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1445 __func__
, atomic_read(&sk
->sk_omem_alloc
));
1447 if (sk
->sk_frag
.page
) {
1448 put_page(sk
->sk_frag
.page
);
1449 sk
->sk_frag
.page
= NULL
;
1452 if (sk
->sk_peer_cred
)
1453 put_cred(sk
->sk_peer_cred
);
1454 put_pid(sk
->sk_peer_pid
);
1455 if (likely(sk
->sk_net_refcnt
))
1456 put_net(sock_net(sk
));
1457 sk_prot_free(sk
->sk_prot_creator
, sk
);
1460 void sk_destruct(struct sock
*sk
)
1462 if (sock_flag(sk
, SOCK_RCU_FREE
))
1463 call_rcu(&sk
->sk_rcu
, __sk_destruct
);
1465 __sk_destruct(&sk
->sk_rcu
);
1468 static void __sk_free(struct sock
*sk
)
1470 if (unlikely(sock_diag_has_destroy_listeners(sk
) && sk
->sk_net_refcnt
))
1471 sock_diag_broadcast_destroy(sk
);
1476 void sk_free(struct sock
*sk
)
1479 * We subtract one from sk_wmem_alloc and can know if
1480 * some packets are still in some tx queue.
1481 * If not null, sock_wfree() will call __sk_free(sk) later
1483 if (atomic_dec_and_test(&sk
->sk_wmem_alloc
))
1486 EXPORT_SYMBOL(sk_free
);
1489 * sk_clone_lock - clone a socket, and lock its clone
1490 * @sk: the socket to clone
1491 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1493 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1495 struct sock
*sk_clone_lock(const struct sock
*sk
, const gfp_t priority
)
1498 bool is_charged
= true;
1500 newsk
= sk_prot_alloc(sk
->sk_prot
, priority
, sk
->sk_family
);
1501 if (newsk
!= NULL
) {
1502 struct sk_filter
*filter
;
1504 sock_copy(newsk
, sk
);
1507 if (likely(newsk
->sk_net_refcnt
))
1508 get_net(sock_net(newsk
));
1509 sk_node_init(&newsk
->sk_node
);
1510 sock_lock_init(newsk
);
1511 bh_lock_sock(newsk
);
1512 newsk
->sk_backlog
.head
= newsk
->sk_backlog
.tail
= NULL
;
1513 newsk
->sk_backlog
.len
= 0;
1515 atomic_set(&newsk
->sk_rmem_alloc
, 0);
1517 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1519 atomic_set(&newsk
->sk_wmem_alloc
, 1);
1520 atomic_set(&newsk
->sk_omem_alloc
, 0);
1521 skb_queue_head_init(&newsk
->sk_receive_queue
);
1522 skb_queue_head_init(&newsk
->sk_write_queue
);
1524 rwlock_init(&newsk
->sk_callback_lock
);
1525 lockdep_set_class_and_name(&newsk
->sk_callback_lock
,
1526 af_callback_keys
+ newsk
->sk_family
,
1527 af_family_clock_key_strings
[newsk
->sk_family
]);
1529 newsk
->sk_dst_cache
= NULL
;
1530 newsk
->sk_wmem_queued
= 0;
1531 newsk
->sk_forward_alloc
= 0;
1532 atomic_set(&newsk
->sk_drops
, 0);
1533 newsk
->sk_send_head
= NULL
;
1534 newsk
->sk_userlocks
= sk
->sk_userlocks
& ~SOCK_BINDPORT_LOCK
;
1536 sock_reset_flag(newsk
, SOCK_DONE
);
1537 skb_queue_head_init(&newsk
->sk_error_queue
);
1539 filter
= rcu_dereference_protected(newsk
->sk_filter
, 1);
1541 /* though it's an empty new sock, the charging may fail
1542 * if sysctl_optmem_max was changed between creation of
1543 * original socket and cloning
1545 is_charged
= sk_filter_charge(newsk
, filter
);
1547 if (unlikely(!is_charged
|| xfrm_sk_clone_policy(newsk
, sk
))) {
1548 /* We need to make sure that we don't uncharge the new
1549 * socket if we couldn't charge it in the first place
1550 * as otherwise we uncharge the parent's filter.
1553 RCU_INIT_POINTER(newsk
->sk_filter
, NULL
);
1554 /* It is still raw copy of parent, so invalidate
1555 * destructor and make plain sk_free() */
1556 newsk
->sk_destruct
= NULL
;
1557 bh_unlock_sock(newsk
);
1562 RCU_INIT_POINTER(newsk
->sk_reuseport_cb
, NULL
);
1565 newsk
->sk_err_soft
= 0;
1566 newsk
->sk_priority
= 0;
1567 newsk
->sk_incoming_cpu
= raw_smp_processor_id();
1568 atomic64_set(&newsk
->sk_cookie
, 0);
1570 mem_cgroup_sk_alloc(newsk
);
1571 cgroup_sk_alloc(&newsk
->sk_cgrp_data
);
1574 * Before updating sk_refcnt, we must commit prior changes to memory
1575 * (Documentation/RCU/rculist_nulls.txt for details)
1578 atomic_set(&newsk
->sk_refcnt
, 2);
1581 * Increment the counter in the same struct proto as the master
1582 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1583 * is the same as sk->sk_prot->socks, as this field was copied
1586 * This _changes_ the previous behaviour, where
1587 * tcp_create_openreq_child always was incrementing the
1588 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1589 * to be taken into account in all callers. -acme
1591 sk_refcnt_debug_inc(newsk
);
1592 sk_set_socket(newsk
, NULL
);
1593 newsk
->sk_wq
= NULL
;
1595 if (newsk
->sk_prot
->sockets_allocated
)
1596 sk_sockets_allocated_inc(newsk
);
1598 if (sock_needs_netstamp(sk
) &&
1599 newsk
->sk_flags
& SK_FLAGS_TIMESTAMP
)
1600 net_enable_timestamp();
1605 EXPORT_SYMBOL_GPL(sk_clone_lock
);
1607 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
)
1611 sk_dst_set(sk
, dst
);
1612 sk
->sk_route_caps
= dst
->dev
->features
;
1613 if (sk
->sk_route_caps
& NETIF_F_GSO
)
1614 sk
->sk_route_caps
|= NETIF_F_GSO_SOFTWARE
;
1615 sk
->sk_route_caps
&= ~sk
->sk_route_nocaps
;
1616 if (sk_can_gso(sk
)) {
1617 if (dst
->header_len
) {
1618 sk
->sk_route_caps
&= ~NETIF_F_GSO_MASK
;
1620 sk
->sk_route_caps
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
1621 sk
->sk_gso_max_size
= dst
->dev
->gso_max_size
;
1622 max_segs
= max_t(u32
, dst
->dev
->gso_max_segs
, 1);
1625 sk
->sk_gso_max_segs
= max_segs
;
1627 EXPORT_SYMBOL_GPL(sk_setup_caps
);
1630 * Simple resource managers for sockets.
1635 * Write buffer destructor automatically called from kfree_skb.
1637 void sock_wfree(struct sk_buff
*skb
)
1639 struct sock
*sk
= skb
->sk
;
1640 unsigned int len
= skb
->truesize
;
1642 if (!sock_flag(sk
, SOCK_USE_WRITE_QUEUE
)) {
1644 * Keep a reference on sk_wmem_alloc, this will be released
1645 * after sk_write_space() call
1647 atomic_sub(len
- 1, &sk
->sk_wmem_alloc
);
1648 sk
->sk_write_space(sk
);
1652 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1653 * could not do because of in-flight packets
1655 if (atomic_sub_and_test(len
, &sk
->sk_wmem_alloc
))
1658 EXPORT_SYMBOL(sock_wfree
);
1660 /* This variant of sock_wfree() is used by TCP,
1661 * since it sets SOCK_USE_WRITE_QUEUE.
1663 void __sock_wfree(struct sk_buff
*skb
)
1665 struct sock
*sk
= skb
->sk
;
1667 if (atomic_sub_and_test(skb
->truesize
, &sk
->sk_wmem_alloc
))
1671 void skb_set_owner_w(struct sk_buff
*skb
, struct sock
*sk
)
1676 if (unlikely(!sk_fullsock(sk
))) {
1677 skb
->destructor
= sock_edemux
;
1682 skb
->destructor
= sock_wfree
;
1683 skb_set_hash_from_sk(skb
, sk
);
1685 * We used to take a refcount on sk, but following operation
1686 * is enough to guarantee sk_free() wont free this sock until
1687 * all in-flight packets are completed
1689 atomic_add(skb
->truesize
, &sk
->sk_wmem_alloc
);
1691 EXPORT_SYMBOL(skb_set_owner_w
);
1693 /* This helper is used by netem, as it can hold packets in its
1694 * delay queue. We want to allow the owner socket to send more
1695 * packets, as if they were already TX completed by a typical driver.
1696 * But we also want to keep skb->sk set because some packet schedulers
1697 * rely on it (sch_fq for example). So we set skb->truesize to a small
1698 * amount (1) and decrease sk_wmem_alloc accordingly.
1700 void skb_orphan_partial(struct sk_buff
*skb
)
1702 /* If this skb is a TCP pure ACK or already went here,
1703 * we have nothing to do. 2 is already a very small truesize.
1705 if (skb
->truesize
<= 2)
1708 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1709 * so we do not completely orphan skb, but transfert all
1710 * accounted bytes but one, to avoid unexpected reorders.
1712 if (skb
->destructor
== sock_wfree
1714 || skb
->destructor
== tcp_wfree
1717 atomic_sub(skb
->truesize
- 1, &skb
->sk
->sk_wmem_alloc
);
1723 EXPORT_SYMBOL(skb_orphan_partial
);
1726 * Read buffer destructor automatically called from kfree_skb.
1728 void sock_rfree(struct sk_buff
*skb
)
1730 struct sock
*sk
= skb
->sk
;
1731 unsigned int len
= skb
->truesize
;
1733 atomic_sub(len
, &sk
->sk_rmem_alloc
);
1734 sk_mem_uncharge(sk
, len
);
1736 EXPORT_SYMBOL(sock_rfree
);
1739 * Buffer destructor for skbs that are not used directly in read or write
1740 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1742 void sock_efree(struct sk_buff
*skb
)
1746 EXPORT_SYMBOL(sock_efree
);
1748 kuid_t
sock_i_uid(struct sock
*sk
)
1752 read_lock_bh(&sk
->sk_callback_lock
);
1753 uid
= sk
->sk_socket
? SOCK_INODE(sk
->sk_socket
)->i_uid
: GLOBAL_ROOT_UID
;
1754 read_unlock_bh(&sk
->sk_callback_lock
);
1757 EXPORT_SYMBOL(sock_i_uid
);
1759 unsigned long sock_i_ino(struct sock
*sk
)
1763 read_lock_bh(&sk
->sk_callback_lock
);
1764 ino
= sk
->sk_socket
? SOCK_INODE(sk
->sk_socket
)->i_ino
: 0;
1765 read_unlock_bh(&sk
->sk_callback_lock
);
1768 EXPORT_SYMBOL(sock_i_ino
);
1771 * Allocate a skb from the socket's send buffer.
1773 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1776 if (force
|| atomic_read(&sk
->sk_wmem_alloc
) < sk
->sk_sndbuf
) {
1777 struct sk_buff
*skb
= alloc_skb(size
, priority
);
1779 skb_set_owner_w(skb
, sk
);
1785 EXPORT_SYMBOL(sock_wmalloc
);
1788 * Allocate a memory block from the socket's option memory buffer.
1790 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
)
1792 if ((unsigned int)size
<= sysctl_optmem_max
&&
1793 atomic_read(&sk
->sk_omem_alloc
) + size
< sysctl_optmem_max
) {
1795 /* First do the add, to avoid the race if kmalloc
1798 atomic_add(size
, &sk
->sk_omem_alloc
);
1799 mem
= kmalloc(size
, priority
);
1802 atomic_sub(size
, &sk
->sk_omem_alloc
);
1806 EXPORT_SYMBOL(sock_kmalloc
);
1808 /* Free an option memory block. Note, we actually want the inline
1809 * here as this allows gcc to detect the nullify and fold away the
1810 * condition entirely.
1812 static inline void __sock_kfree_s(struct sock
*sk
, void *mem
, int size
,
1815 if (WARN_ON_ONCE(!mem
))
1821 atomic_sub(size
, &sk
->sk_omem_alloc
);
1824 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
)
1826 __sock_kfree_s(sk
, mem
, size
, false);
1828 EXPORT_SYMBOL(sock_kfree_s
);
1830 void sock_kzfree_s(struct sock
*sk
, void *mem
, int size
)
1832 __sock_kfree_s(sk
, mem
, size
, true);
1834 EXPORT_SYMBOL(sock_kzfree_s
);
1836 /* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1837 I think, these locks should be removed for datagram sockets.
1839 static long sock_wait_for_wmem(struct sock
*sk
, long timeo
)
1843 sk_clear_bit(SOCKWQ_ASYNC_NOSPACE
, sk
);
1847 if (signal_pending(current
))
1849 set_bit(SOCK_NOSPACE
, &sk
->sk_socket
->flags
);
1850 prepare_to_wait(sk_sleep(sk
), &wait
, TASK_INTERRUPTIBLE
);
1851 if (atomic_read(&sk
->sk_wmem_alloc
) < sk
->sk_sndbuf
)
1853 if (sk
->sk_shutdown
& SEND_SHUTDOWN
)
1857 timeo
= schedule_timeout(timeo
);
1859 finish_wait(sk_sleep(sk
), &wait
);
1865 * Generic send/receive buffer handlers
1868 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1869 unsigned long data_len
, int noblock
,
1870 int *errcode
, int max_page_order
)
1872 struct sk_buff
*skb
;
1876 timeo
= sock_sndtimeo(sk
, noblock
);
1878 err
= sock_error(sk
);
1883 if (sk
->sk_shutdown
& SEND_SHUTDOWN
)
1886 if (sk_wmem_alloc_get(sk
) < sk
->sk_sndbuf
)
1889 sk_set_bit(SOCKWQ_ASYNC_NOSPACE
, sk
);
1890 set_bit(SOCK_NOSPACE
, &sk
->sk_socket
->flags
);
1894 if (signal_pending(current
))
1896 timeo
= sock_wait_for_wmem(sk
, timeo
);
1898 skb
= alloc_skb_with_frags(header_len
, data_len
, max_page_order
,
1899 errcode
, sk
->sk_allocation
);
1901 skb_set_owner_w(skb
, sk
);
1905 err
= sock_intr_errno(timeo
);
1910 EXPORT_SYMBOL(sock_alloc_send_pskb
);
1912 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1913 int noblock
, int *errcode
)
1915 return sock_alloc_send_pskb(sk
, size
, 0, noblock
, errcode
, 0);
1917 EXPORT_SYMBOL(sock_alloc_send_skb
);
1919 int __sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
, struct cmsghdr
*cmsg
,
1920 struct sockcm_cookie
*sockc
)
1924 switch (cmsg
->cmsg_type
) {
1926 if (!ns_capable(sock_net(sk
)->user_ns
, CAP_NET_ADMIN
))
1928 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(u32
)))
1930 sockc
->mark
= *(u32
*)CMSG_DATA(cmsg
);
1932 case SO_TIMESTAMPING
:
1933 if (cmsg
->cmsg_len
!= CMSG_LEN(sizeof(u32
)))
1936 tsflags
= *(u32
*)CMSG_DATA(cmsg
);
1937 if (tsflags
& ~SOF_TIMESTAMPING_TX_RECORD_MASK
)
1940 sockc
->tsflags
&= ~SOF_TIMESTAMPING_TX_RECORD_MASK
;
1941 sockc
->tsflags
|= tsflags
;
1943 /* SCM_RIGHTS and SCM_CREDENTIALS are semantically in SOL_UNIX. */
1945 case SCM_CREDENTIALS
:
1952 EXPORT_SYMBOL(__sock_cmsg_send
);
1954 int sock_cmsg_send(struct sock
*sk
, struct msghdr
*msg
,
1955 struct sockcm_cookie
*sockc
)
1957 struct cmsghdr
*cmsg
;
1960 for_each_cmsghdr(cmsg
, msg
) {
1961 if (!CMSG_OK(msg
, cmsg
))
1963 if (cmsg
->cmsg_level
!= SOL_SOCKET
)
1965 ret
= __sock_cmsg_send(sk
, msg
, cmsg
, sockc
);
1971 EXPORT_SYMBOL(sock_cmsg_send
);
1973 /* On 32bit arches, an skb frag is limited to 2^15 */
1974 #define SKB_FRAG_PAGE_ORDER get_order(32768)
1977 * skb_page_frag_refill - check that a page_frag contains enough room
1978 * @sz: minimum size of the fragment we want to get
1979 * @pfrag: pointer to page_frag
1980 * @gfp: priority for memory allocation
1982 * Note: While this allocator tries to use high order pages, there is
1983 * no guarantee that allocations succeed. Therefore, @sz MUST be
1984 * less or equal than PAGE_SIZE.
1986 bool skb_page_frag_refill(unsigned int sz
, struct page_frag
*pfrag
, gfp_t gfp
)
1989 if (page_ref_count(pfrag
->page
) == 1) {
1993 if (pfrag
->offset
+ sz
<= pfrag
->size
)
1995 put_page(pfrag
->page
);
1999 if (SKB_FRAG_PAGE_ORDER
) {
2000 /* Avoid direct reclaim but allow kswapd to wake */
2001 pfrag
->page
= alloc_pages((gfp
& ~__GFP_DIRECT_RECLAIM
) |
2002 __GFP_COMP
| __GFP_NOWARN
|
2004 SKB_FRAG_PAGE_ORDER
);
2005 if (likely(pfrag
->page
)) {
2006 pfrag
->size
= PAGE_SIZE
<< SKB_FRAG_PAGE_ORDER
;
2010 pfrag
->page
= alloc_page(gfp
);
2011 if (likely(pfrag
->page
)) {
2012 pfrag
->size
= PAGE_SIZE
;
2017 EXPORT_SYMBOL(skb_page_frag_refill
);
2019 bool sk_page_frag_refill(struct sock
*sk
, struct page_frag
*pfrag
)
2021 if (likely(skb_page_frag_refill(32U, pfrag
, sk
->sk_allocation
)))
2024 sk_enter_memory_pressure(sk
);
2025 sk_stream_moderate_sndbuf(sk
);
2028 EXPORT_SYMBOL(sk_page_frag_refill
);
2030 static void __lock_sock(struct sock
*sk
)
2031 __releases(&sk
->sk_lock
.slock
)
2032 __acquires(&sk
->sk_lock
.slock
)
2037 prepare_to_wait_exclusive(&sk
->sk_lock
.wq
, &wait
,
2038 TASK_UNINTERRUPTIBLE
);
2039 spin_unlock_bh(&sk
->sk_lock
.slock
);
2041 spin_lock_bh(&sk
->sk_lock
.slock
);
2042 if (!sock_owned_by_user(sk
))
2045 finish_wait(&sk
->sk_lock
.wq
, &wait
);
2048 static void __release_sock(struct sock
*sk
)
2049 __releases(&sk
->sk_lock
.slock
)
2050 __acquires(&sk
->sk_lock
.slock
)
2052 struct sk_buff
*skb
, *next
;
2054 while ((skb
= sk
->sk_backlog
.head
) != NULL
) {
2055 sk
->sk_backlog
.head
= sk
->sk_backlog
.tail
= NULL
;
2057 spin_unlock_bh(&sk
->sk_lock
.slock
);
2062 WARN_ON_ONCE(skb_dst_is_noref(skb
));
2064 sk_backlog_rcv(sk
, skb
);
2069 } while (skb
!= NULL
);
2071 spin_lock_bh(&sk
->sk_lock
.slock
);
2075 * Doing the zeroing here guarantee we can not loop forever
2076 * while a wild producer attempts to flood us.
2078 sk
->sk_backlog
.len
= 0;
2081 void __sk_flush_backlog(struct sock
*sk
)
2083 spin_lock_bh(&sk
->sk_lock
.slock
);
2085 spin_unlock_bh(&sk
->sk_lock
.slock
);
2089 * sk_wait_data - wait for data to arrive at sk_receive_queue
2090 * @sk: sock to wait on
2091 * @timeo: for how long
2092 * @skb: last skb seen on sk_receive_queue
2094 * Now socket state including sk->sk_err is changed only under lock,
2095 * hence we may omit checks after joining wait queue.
2096 * We check receive queue before schedule() only as optimization;
2097 * it is very likely that release_sock() added new data.
2099 int sk_wait_data(struct sock
*sk
, long *timeo
, const struct sk_buff
*skb
)
2101 DEFINE_WAIT_FUNC(wait
, woken_wake_function
);
2104 add_wait_queue(sk_sleep(sk
), &wait
);
2105 sk_set_bit(SOCKWQ_ASYNC_WAITDATA
, sk
);
2106 rc
= sk_wait_event(sk
, timeo
, skb_peek_tail(&sk
->sk_receive_queue
) != skb
, &wait
);
2107 sk_clear_bit(SOCKWQ_ASYNC_WAITDATA
, sk
);
2108 remove_wait_queue(sk_sleep(sk
), &wait
);
2111 EXPORT_SYMBOL(sk_wait_data
);
2114 * __sk_mem_raise_allocated - increase memory_allocated
2116 * @size: memory size to allocate
2117 * @amt: pages to allocate
2118 * @kind: allocation type
2120 * Similar to __sk_mem_schedule(), but does not update sk_forward_alloc
2122 int __sk_mem_raise_allocated(struct sock
*sk
, int size
, int amt
, int kind
)
2124 struct proto
*prot
= sk
->sk_prot
;
2125 long allocated
= sk_memory_allocated_add(sk
, amt
);
2127 if (mem_cgroup_sockets_enabled
&& sk
->sk_memcg
&&
2128 !mem_cgroup_charge_skmem(sk
->sk_memcg
, amt
))
2129 goto suppress_allocation
;
2132 if (allocated
<= sk_prot_mem_limits(sk
, 0)) {
2133 sk_leave_memory_pressure(sk
);
2137 /* Under pressure. */
2138 if (allocated
> sk_prot_mem_limits(sk
, 1))
2139 sk_enter_memory_pressure(sk
);
2141 /* Over hard limit. */
2142 if (allocated
> sk_prot_mem_limits(sk
, 2))
2143 goto suppress_allocation
;
2145 /* guarantee minimum buffer size under pressure */
2146 if (kind
== SK_MEM_RECV
) {
2147 if (atomic_read(&sk
->sk_rmem_alloc
) < prot
->sysctl_rmem
[0])
2150 } else { /* SK_MEM_SEND */
2151 if (sk
->sk_type
== SOCK_STREAM
) {
2152 if (sk
->sk_wmem_queued
< prot
->sysctl_wmem
[0])
2154 } else if (atomic_read(&sk
->sk_wmem_alloc
) <
2155 prot
->sysctl_wmem
[0])
2159 if (sk_has_memory_pressure(sk
)) {
2162 if (!sk_under_memory_pressure(sk
))
2164 alloc
= sk_sockets_allocated_read_positive(sk
);
2165 if (sk_prot_mem_limits(sk
, 2) > alloc
*
2166 sk_mem_pages(sk
->sk_wmem_queued
+
2167 atomic_read(&sk
->sk_rmem_alloc
) +
2168 sk
->sk_forward_alloc
))
2172 suppress_allocation
:
2174 if (kind
== SK_MEM_SEND
&& sk
->sk_type
== SOCK_STREAM
) {
2175 sk_stream_moderate_sndbuf(sk
);
2177 /* Fail only if socket is _under_ its sndbuf.
2178 * In this case we cannot block, so that we have to fail.
2180 if (sk
->sk_wmem_queued
+ size
>= sk
->sk_sndbuf
)
2184 trace_sock_exceed_buf_limit(sk
, prot
, allocated
);
2186 sk_memory_allocated_sub(sk
, amt
);
2188 if (mem_cgroup_sockets_enabled
&& sk
->sk_memcg
)
2189 mem_cgroup_uncharge_skmem(sk
->sk_memcg
, amt
);
2193 EXPORT_SYMBOL(__sk_mem_raise_allocated
);
2196 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2198 * @size: memory size to allocate
2199 * @kind: allocation type
2201 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2202 * rmem allocation. This function assumes that protocols which have
2203 * memory_pressure use sk_wmem_queued as write buffer accounting.
2205 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
)
2207 int ret
, amt
= sk_mem_pages(size
);
2209 sk
->sk_forward_alloc
+= amt
<< SK_MEM_QUANTUM_SHIFT
;
2210 ret
= __sk_mem_raise_allocated(sk
, size
, amt
, kind
);
2212 sk
->sk_forward_alloc
-= amt
<< SK_MEM_QUANTUM_SHIFT
;
2215 EXPORT_SYMBOL(__sk_mem_schedule
);
2218 * __sk_mem_reduce_allocated - reclaim memory_allocated
2220 * @amount: number of quanta
2222 * Similar to __sk_mem_reclaim(), but does not update sk_forward_alloc
2224 void __sk_mem_reduce_allocated(struct sock
*sk
, int amount
)
2226 sk_memory_allocated_sub(sk
, amount
);
2228 if (mem_cgroup_sockets_enabled
&& sk
->sk_memcg
)
2229 mem_cgroup_uncharge_skmem(sk
->sk_memcg
, amount
);
2231 if (sk_under_memory_pressure(sk
) &&
2232 (sk_memory_allocated(sk
) < sk_prot_mem_limits(sk
, 0)))
2233 sk_leave_memory_pressure(sk
);
2235 EXPORT_SYMBOL(__sk_mem_reduce_allocated
);
2238 * __sk_mem_reclaim - reclaim sk_forward_alloc and memory_allocated
2240 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
2242 void __sk_mem_reclaim(struct sock
*sk
, int amount
)
2244 amount
>>= SK_MEM_QUANTUM_SHIFT
;
2245 sk
->sk_forward_alloc
-= amount
<< SK_MEM_QUANTUM_SHIFT
;
2246 __sk_mem_reduce_allocated(sk
, amount
);
2248 EXPORT_SYMBOL(__sk_mem_reclaim
);
2250 int sk_set_peek_off(struct sock
*sk
, int val
)
2255 sk
->sk_peek_off
= val
;
2258 EXPORT_SYMBOL_GPL(sk_set_peek_off
);
2261 * Set of default routines for initialising struct proto_ops when
2262 * the protocol does not support a particular function. In certain
2263 * cases where it makes no sense for a protocol to have a "do nothing"
2264 * function, some default processing is provided.
2267 int sock_no_bind(struct socket
*sock
, struct sockaddr
*saddr
, int len
)
2271 EXPORT_SYMBOL(sock_no_bind
);
2273 int sock_no_connect(struct socket
*sock
, struct sockaddr
*saddr
,
2278 EXPORT_SYMBOL(sock_no_connect
);
2280 int sock_no_socketpair(struct socket
*sock1
, struct socket
*sock2
)
2284 EXPORT_SYMBOL(sock_no_socketpair
);
2286 int sock_no_accept(struct socket
*sock
, struct socket
*newsock
, int flags
)
2290 EXPORT_SYMBOL(sock_no_accept
);
2292 int sock_no_getname(struct socket
*sock
, struct sockaddr
*saddr
,
2297 EXPORT_SYMBOL(sock_no_getname
);
2299 unsigned int sock_no_poll(struct file
*file
, struct socket
*sock
, poll_table
*pt
)
2303 EXPORT_SYMBOL(sock_no_poll
);
2305 int sock_no_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
2309 EXPORT_SYMBOL(sock_no_ioctl
);
2311 int sock_no_listen(struct socket
*sock
, int backlog
)
2315 EXPORT_SYMBOL(sock_no_listen
);
2317 int sock_no_shutdown(struct socket
*sock
, int how
)
2321 EXPORT_SYMBOL(sock_no_shutdown
);
2323 int sock_no_setsockopt(struct socket
*sock
, int level
, int optname
,
2324 char __user
*optval
, unsigned int optlen
)
2328 EXPORT_SYMBOL(sock_no_setsockopt
);
2330 int sock_no_getsockopt(struct socket
*sock
, int level
, int optname
,
2331 char __user
*optval
, int __user
*optlen
)
2335 EXPORT_SYMBOL(sock_no_getsockopt
);
2337 int sock_no_sendmsg(struct socket
*sock
, struct msghdr
*m
, size_t len
)
2341 EXPORT_SYMBOL(sock_no_sendmsg
);
2343 int sock_no_recvmsg(struct socket
*sock
, struct msghdr
*m
, size_t len
,
2348 EXPORT_SYMBOL(sock_no_recvmsg
);
2350 int sock_no_mmap(struct file
*file
, struct socket
*sock
, struct vm_area_struct
*vma
)
2352 /* Mirror missing mmap method error code */
2355 EXPORT_SYMBOL(sock_no_mmap
);
2357 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
, size_t size
, int flags
)
2360 struct msghdr msg
= {.msg_flags
= flags
};
2362 char *kaddr
= kmap(page
);
2363 iov
.iov_base
= kaddr
+ offset
;
2365 res
= kernel_sendmsg(sock
, &msg
, &iov
, 1, size
);
2369 EXPORT_SYMBOL(sock_no_sendpage
);
2372 * Default Socket Callbacks
2375 static void sock_def_wakeup(struct sock
*sk
)
2377 struct socket_wq
*wq
;
2380 wq
= rcu_dereference(sk
->sk_wq
);
2381 if (skwq_has_sleeper(wq
))
2382 wake_up_interruptible_all(&wq
->wait
);
2386 static void sock_def_error_report(struct sock
*sk
)
2388 struct socket_wq
*wq
;
2391 wq
= rcu_dereference(sk
->sk_wq
);
2392 if (skwq_has_sleeper(wq
))
2393 wake_up_interruptible_poll(&wq
->wait
, POLLERR
);
2394 sk_wake_async(sk
, SOCK_WAKE_IO
, POLL_ERR
);
2398 static void sock_def_readable(struct sock
*sk
)
2400 struct socket_wq
*wq
;
2403 wq
= rcu_dereference(sk
->sk_wq
);
2404 if (skwq_has_sleeper(wq
))
2405 wake_up_interruptible_sync_poll(&wq
->wait
, POLLIN
| POLLPRI
|
2406 POLLRDNORM
| POLLRDBAND
);
2407 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
2411 static void sock_def_write_space(struct sock
*sk
)
2413 struct socket_wq
*wq
;
2417 /* Do not wake up a writer until he can make "significant"
2420 if ((atomic_read(&sk
->sk_wmem_alloc
) << 1) <= sk
->sk_sndbuf
) {
2421 wq
= rcu_dereference(sk
->sk_wq
);
2422 if (skwq_has_sleeper(wq
))
2423 wake_up_interruptible_sync_poll(&wq
->wait
, POLLOUT
|
2424 POLLWRNORM
| POLLWRBAND
);
2426 /* Should agree with poll, otherwise some programs break */
2427 if (sock_writeable(sk
))
2428 sk_wake_async(sk
, SOCK_WAKE_SPACE
, POLL_OUT
);
2434 static void sock_def_destruct(struct sock
*sk
)
2438 void sk_send_sigurg(struct sock
*sk
)
2440 if (sk
->sk_socket
&& sk
->sk_socket
->file
)
2441 if (send_sigurg(&sk
->sk_socket
->file
->f_owner
))
2442 sk_wake_async(sk
, SOCK_WAKE_URG
, POLL_PRI
);
2444 EXPORT_SYMBOL(sk_send_sigurg
);
2446 void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
2447 unsigned long expires
)
2449 if (!mod_timer(timer
, expires
))
2452 EXPORT_SYMBOL(sk_reset_timer
);
2454 void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
)
2456 if (del_timer(timer
))
2459 EXPORT_SYMBOL(sk_stop_timer
);
2461 void sock_init_data(struct socket
*sock
, struct sock
*sk
)
2463 skb_queue_head_init(&sk
->sk_receive_queue
);
2464 skb_queue_head_init(&sk
->sk_write_queue
);
2465 skb_queue_head_init(&sk
->sk_error_queue
);
2467 sk
->sk_send_head
= NULL
;
2469 init_timer(&sk
->sk_timer
);
2471 sk
->sk_allocation
= GFP_KERNEL
;
2472 sk
->sk_rcvbuf
= sysctl_rmem_default
;
2473 sk
->sk_sndbuf
= sysctl_wmem_default
;
2474 sk
->sk_state
= TCP_CLOSE
;
2475 sk_set_socket(sk
, sock
);
2477 sock_set_flag(sk
, SOCK_ZAPPED
);
2480 sk
->sk_type
= sock
->type
;
2481 sk
->sk_wq
= sock
->wq
;
2483 sk
->sk_uid
= SOCK_INODE(sock
)->i_uid
;
2486 sk
->sk_uid
= make_kuid(sock_net(sk
)->user_ns
, 0);
2489 rwlock_init(&sk
->sk_callback_lock
);
2490 lockdep_set_class_and_name(&sk
->sk_callback_lock
,
2491 af_callback_keys
+ sk
->sk_family
,
2492 af_family_clock_key_strings
[sk
->sk_family
]);
2494 sk
->sk_state_change
= sock_def_wakeup
;
2495 sk
->sk_data_ready
= sock_def_readable
;
2496 sk
->sk_write_space
= sock_def_write_space
;
2497 sk
->sk_error_report
= sock_def_error_report
;
2498 sk
->sk_destruct
= sock_def_destruct
;
2500 sk
->sk_frag
.page
= NULL
;
2501 sk
->sk_frag
.offset
= 0;
2502 sk
->sk_peek_off
= -1;
2504 sk
->sk_peer_pid
= NULL
;
2505 sk
->sk_peer_cred
= NULL
;
2506 sk
->sk_write_pending
= 0;
2507 sk
->sk_rcvlowat
= 1;
2508 sk
->sk_rcvtimeo
= MAX_SCHEDULE_TIMEOUT
;
2509 sk
->sk_sndtimeo
= MAX_SCHEDULE_TIMEOUT
;
2511 sk
->sk_stamp
= ktime_set(-1L, 0);
2513 #ifdef CONFIG_NET_RX_BUSY_POLL
2515 sk
->sk_ll_usec
= sysctl_net_busy_read
;
2518 sk
->sk_max_pacing_rate
= ~0U;
2519 sk
->sk_pacing_rate
= ~0U;
2520 sk
->sk_incoming_cpu
= -1;
2522 * Before updating sk_refcnt, we must commit prior changes to memory
2523 * (Documentation/RCU/rculist_nulls.txt for details)
2526 atomic_set(&sk
->sk_refcnt
, 1);
2527 atomic_set(&sk
->sk_drops
, 0);
2529 EXPORT_SYMBOL(sock_init_data
);
2531 void lock_sock_nested(struct sock
*sk
, int subclass
)
2534 spin_lock_bh(&sk
->sk_lock
.slock
);
2535 if (sk
->sk_lock
.owned
)
2537 sk
->sk_lock
.owned
= 1;
2538 spin_unlock(&sk
->sk_lock
.slock
);
2540 * The sk_lock has mutex_lock() semantics here:
2542 mutex_acquire(&sk
->sk_lock
.dep_map
, subclass
, 0, _RET_IP_
);
2545 EXPORT_SYMBOL(lock_sock_nested
);
2547 void release_sock(struct sock
*sk
)
2549 spin_lock_bh(&sk
->sk_lock
.slock
);
2550 if (sk
->sk_backlog
.tail
)
2553 /* Warning : release_cb() might need to release sk ownership,
2554 * ie call sock_release_ownership(sk) before us.
2556 if (sk
->sk_prot
->release_cb
)
2557 sk
->sk_prot
->release_cb(sk
);
2559 sock_release_ownership(sk
);
2560 if (waitqueue_active(&sk
->sk_lock
.wq
))
2561 wake_up(&sk
->sk_lock
.wq
);
2562 spin_unlock_bh(&sk
->sk_lock
.slock
);
2564 EXPORT_SYMBOL(release_sock
);
2567 * lock_sock_fast - fast version of lock_sock
2570 * This version should be used for very small section, where process wont block
2571 * return false if fast path is taken
2572 * sk_lock.slock locked, owned = 0, BH disabled
2573 * return true if slow path is taken
2574 * sk_lock.slock unlocked, owned = 1, BH enabled
2576 bool lock_sock_fast(struct sock
*sk
)
2579 spin_lock_bh(&sk
->sk_lock
.slock
);
2581 if (!sk
->sk_lock
.owned
)
2583 * Note : We must disable BH
2588 sk
->sk_lock
.owned
= 1;
2589 spin_unlock(&sk
->sk_lock
.slock
);
2591 * The sk_lock has mutex_lock() semantics here:
2593 mutex_acquire(&sk
->sk_lock
.dep_map
, 0, 0, _RET_IP_
);
2597 EXPORT_SYMBOL(lock_sock_fast
);
2599 int sock_get_timestamp(struct sock
*sk
, struct timeval __user
*userstamp
)
2602 if (!sock_flag(sk
, SOCK_TIMESTAMP
))
2603 sock_enable_timestamp(sk
, SOCK_TIMESTAMP
);
2604 tv
= ktime_to_timeval(sk
->sk_stamp
);
2605 if (tv
.tv_sec
== -1)
2607 if (tv
.tv_sec
== 0) {
2608 sk
->sk_stamp
= ktime_get_real();
2609 tv
= ktime_to_timeval(sk
->sk_stamp
);
2611 return copy_to_user(userstamp
, &tv
, sizeof(tv
)) ? -EFAULT
: 0;
2613 EXPORT_SYMBOL(sock_get_timestamp
);
2615 int sock_get_timestampns(struct sock
*sk
, struct timespec __user
*userstamp
)
2618 if (!sock_flag(sk
, SOCK_TIMESTAMP
))
2619 sock_enable_timestamp(sk
, SOCK_TIMESTAMP
);
2620 ts
= ktime_to_timespec(sk
->sk_stamp
);
2621 if (ts
.tv_sec
== -1)
2623 if (ts
.tv_sec
== 0) {
2624 sk
->sk_stamp
= ktime_get_real();
2625 ts
= ktime_to_timespec(sk
->sk_stamp
);
2627 return copy_to_user(userstamp
, &ts
, sizeof(ts
)) ? -EFAULT
: 0;
2629 EXPORT_SYMBOL(sock_get_timestampns
);
2631 void sock_enable_timestamp(struct sock
*sk
, int flag
)
2633 if (!sock_flag(sk
, flag
)) {
2634 unsigned long previous_flags
= sk
->sk_flags
;
2636 sock_set_flag(sk
, flag
);
2638 * we just set one of the two flags which require net
2639 * time stamping, but time stamping might have been on
2640 * already because of the other one
2642 if (sock_needs_netstamp(sk
) &&
2643 !(previous_flags
& SK_FLAGS_TIMESTAMP
))
2644 net_enable_timestamp();
2648 int sock_recv_errqueue(struct sock
*sk
, struct msghdr
*msg
, int len
,
2649 int level
, int type
)
2651 struct sock_exterr_skb
*serr
;
2652 struct sk_buff
*skb
;
2656 skb
= sock_dequeue_err_skb(sk
);
2662 msg
->msg_flags
|= MSG_TRUNC
;
2665 err
= skb_copy_datagram_msg(skb
, 0, msg
, copied
);
2669 sock_recv_timestamp(msg
, sk
, skb
);
2671 serr
= SKB_EXT_ERR(skb
);
2672 put_cmsg(msg
, level
, type
, sizeof(serr
->ee
), &serr
->ee
);
2674 msg
->msg_flags
|= MSG_ERRQUEUE
;
2682 EXPORT_SYMBOL(sock_recv_errqueue
);
2685 * Get a socket option on an socket.
2687 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2688 * asynchronous errors should be reported by getsockopt. We assume
2689 * this means if you specify SO_ERROR (otherwise whats the point of it).
2691 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
2692 char __user
*optval
, int __user
*optlen
)
2694 struct sock
*sk
= sock
->sk
;
2696 return sk
->sk_prot
->getsockopt(sk
, level
, optname
, optval
, optlen
);
2698 EXPORT_SYMBOL(sock_common_getsockopt
);
2700 #ifdef CONFIG_COMPAT
2701 int compat_sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
2702 char __user
*optval
, int __user
*optlen
)
2704 struct sock
*sk
= sock
->sk
;
2706 if (sk
->sk_prot
->compat_getsockopt
!= NULL
)
2707 return sk
->sk_prot
->compat_getsockopt(sk
, level
, optname
,
2709 return sk
->sk_prot
->getsockopt(sk
, level
, optname
, optval
, optlen
);
2711 EXPORT_SYMBOL(compat_sock_common_getsockopt
);
2714 int sock_common_recvmsg(struct socket
*sock
, struct msghdr
*msg
, size_t size
,
2717 struct sock
*sk
= sock
->sk
;
2721 err
= sk
->sk_prot
->recvmsg(sk
, msg
, size
, flags
& MSG_DONTWAIT
,
2722 flags
& ~MSG_DONTWAIT
, &addr_len
);
2724 msg
->msg_namelen
= addr_len
;
2727 EXPORT_SYMBOL(sock_common_recvmsg
);
2730 * Set socket options on an inet socket.
2732 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
2733 char __user
*optval
, unsigned int optlen
)
2735 struct sock
*sk
= sock
->sk
;
2737 return sk
->sk_prot
->setsockopt(sk
, level
, optname
, optval
, optlen
);
2739 EXPORT_SYMBOL(sock_common_setsockopt
);
2741 #ifdef CONFIG_COMPAT
2742 int compat_sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
2743 char __user
*optval
, unsigned int optlen
)
2745 struct sock
*sk
= sock
->sk
;
2747 if (sk
->sk_prot
->compat_setsockopt
!= NULL
)
2748 return sk
->sk_prot
->compat_setsockopt(sk
, level
, optname
,
2750 return sk
->sk_prot
->setsockopt(sk
, level
, optname
, optval
, optlen
);
2752 EXPORT_SYMBOL(compat_sock_common_setsockopt
);
2755 void sk_common_release(struct sock
*sk
)
2757 if (sk
->sk_prot
->destroy
)
2758 sk
->sk_prot
->destroy(sk
);
2761 * Observation: when sock_common_release is called, processes have
2762 * no access to socket. But net still has.
2763 * Step one, detach it from networking:
2765 * A. Remove from hash tables.
2768 sk
->sk_prot
->unhash(sk
);
2771 * In this point socket cannot receive new packets, but it is possible
2772 * that some packets are in flight because some CPU runs receiver and
2773 * did hash table lookup before we unhashed socket. They will achieve
2774 * receive queue and will be purged by socket destructor.
2776 * Also we still have packets pending on receive queue and probably,
2777 * our own packets waiting in device queues. sock_destroy will drain
2778 * receive queue, but transmitted packets will delay socket destruction
2779 * until the last reference will be released.
2784 xfrm_sk_free_policy(sk
);
2786 sk_refcnt_debug_release(sk
);
2790 EXPORT_SYMBOL(sk_common_release
);
2792 #ifdef CONFIG_PROC_FS
2793 #define PROTO_INUSE_NR 64 /* should be enough for the first time */
2795 int val
[PROTO_INUSE_NR
];
2798 static DECLARE_BITMAP(proto_inuse_idx
, PROTO_INUSE_NR
);
2800 #ifdef CONFIG_NET_NS
2801 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int val
)
2803 __this_cpu_add(net
->core
.inuse
->val
[prot
->inuse_idx
], val
);
2805 EXPORT_SYMBOL_GPL(sock_prot_inuse_add
);
2807 int sock_prot_inuse_get(struct net
*net
, struct proto
*prot
)
2809 int cpu
, idx
= prot
->inuse_idx
;
2812 for_each_possible_cpu(cpu
)
2813 res
+= per_cpu_ptr(net
->core
.inuse
, cpu
)->val
[idx
];
2815 return res
>= 0 ? res
: 0;
2817 EXPORT_SYMBOL_GPL(sock_prot_inuse_get
);
2819 static int __net_init
sock_inuse_init_net(struct net
*net
)
2821 net
->core
.inuse
= alloc_percpu(struct prot_inuse
);
2822 return net
->core
.inuse
? 0 : -ENOMEM
;
2825 static void __net_exit
sock_inuse_exit_net(struct net
*net
)
2827 free_percpu(net
->core
.inuse
);
2830 static struct pernet_operations net_inuse_ops
= {
2831 .init
= sock_inuse_init_net
,
2832 .exit
= sock_inuse_exit_net
,
2835 static __init
int net_inuse_init(void)
2837 if (register_pernet_subsys(&net_inuse_ops
))
2838 panic("Cannot initialize net inuse counters");
2843 core_initcall(net_inuse_init
);
2845 static DEFINE_PER_CPU(struct prot_inuse
, prot_inuse
);
2847 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int val
)
2849 __this_cpu_add(prot_inuse
.val
[prot
->inuse_idx
], val
);
2851 EXPORT_SYMBOL_GPL(sock_prot_inuse_add
);
2853 int sock_prot_inuse_get(struct net
*net
, struct proto
*prot
)
2855 int cpu
, idx
= prot
->inuse_idx
;
2858 for_each_possible_cpu(cpu
)
2859 res
+= per_cpu(prot_inuse
, cpu
).val
[idx
];
2861 return res
>= 0 ? res
: 0;
2863 EXPORT_SYMBOL_GPL(sock_prot_inuse_get
);
2866 static void assign_proto_idx(struct proto
*prot
)
2868 prot
->inuse_idx
= find_first_zero_bit(proto_inuse_idx
, PROTO_INUSE_NR
);
2870 if (unlikely(prot
->inuse_idx
== PROTO_INUSE_NR
- 1)) {
2871 pr_err("PROTO_INUSE_NR exhausted\n");
2875 set_bit(prot
->inuse_idx
, proto_inuse_idx
);
2878 static void release_proto_idx(struct proto
*prot
)
2880 if (prot
->inuse_idx
!= PROTO_INUSE_NR
- 1)
2881 clear_bit(prot
->inuse_idx
, proto_inuse_idx
);
2884 static inline void assign_proto_idx(struct proto
*prot
)
2888 static inline void release_proto_idx(struct proto
*prot
)
2893 static void req_prot_cleanup(struct request_sock_ops
*rsk_prot
)
2897 kfree(rsk_prot
->slab_name
);
2898 rsk_prot
->slab_name
= NULL
;
2899 kmem_cache_destroy(rsk_prot
->slab
);
2900 rsk_prot
->slab
= NULL
;
2903 static int req_prot_init(const struct proto
*prot
)
2905 struct request_sock_ops
*rsk_prot
= prot
->rsk_prot
;
2910 rsk_prot
->slab_name
= kasprintf(GFP_KERNEL
, "request_sock_%s",
2912 if (!rsk_prot
->slab_name
)
2915 rsk_prot
->slab
= kmem_cache_create(rsk_prot
->slab_name
,
2916 rsk_prot
->obj_size
, 0,
2917 prot
->slab_flags
, NULL
);
2919 if (!rsk_prot
->slab
) {
2920 pr_crit("%s: Can't create request sock SLAB cache!\n",
2927 int proto_register(struct proto
*prot
, int alloc_slab
)
2930 prot
->slab
= kmem_cache_create(prot
->name
, prot
->obj_size
, 0,
2931 SLAB_HWCACHE_ALIGN
| prot
->slab_flags
,
2934 if (prot
->slab
== NULL
) {
2935 pr_crit("%s: Can't create sock SLAB cache!\n",
2940 if (req_prot_init(prot
))
2941 goto out_free_request_sock_slab
;
2943 if (prot
->twsk_prot
!= NULL
) {
2944 prot
->twsk_prot
->twsk_slab_name
= kasprintf(GFP_KERNEL
, "tw_sock_%s", prot
->name
);
2946 if (prot
->twsk_prot
->twsk_slab_name
== NULL
)
2947 goto out_free_request_sock_slab
;
2949 prot
->twsk_prot
->twsk_slab
=
2950 kmem_cache_create(prot
->twsk_prot
->twsk_slab_name
,
2951 prot
->twsk_prot
->twsk_obj_size
,
2955 if (prot
->twsk_prot
->twsk_slab
== NULL
)
2956 goto out_free_timewait_sock_slab_name
;
2960 mutex_lock(&proto_list_mutex
);
2961 list_add(&prot
->node
, &proto_list
);
2962 assign_proto_idx(prot
);
2963 mutex_unlock(&proto_list_mutex
);
2966 out_free_timewait_sock_slab_name
:
2967 kfree(prot
->twsk_prot
->twsk_slab_name
);
2968 out_free_request_sock_slab
:
2969 req_prot_cleanup(prot
->rsk_prot
);
2971 kmem_cache_destroy(prot
->slab
);
2976 EXPORT_SYMBOL(proto_register
);
2978 void proto_unregister(struct proto
*prot
)
2980 mutex_lock(&proto_list_mutex
);
2981 release_proto_idx(prot
);
2982 list_del(&prot
->node
);
2983 mutex_unlock(&proto_list_mutex
);
2985 kmem_cache_destroy(prot
->slab
);
2988 req_prot_cleanup(prot
->rsk_prot
);
2990 if (prot
->twsk_prot
!= NULL
&& prot
->twsk_prot
->twsk_slab
!= NULL
) {
2991 kmem_cache_destroy(prot
->twsk_prot
->twsk_slab
);
2992 kfree(prot
->twsk_prot
->twsk_slab_name
);
2993 prot
->twsk_prot
->twsk_slab
= NULL
;
2996 EXPORT_SYMBOL(proto_unregister
);
2998 #ifdef CONFIG_PROC_FS
2999 static void *proto_seq_start(struct seq_file
*seq
, loff_t
*pos
)
3000 __acquires(proto_list_mutex
)
3002 mutex_lock(&proto_list_mutex
);
3003 return seq_list_start_head(&proto_list
, *pos
);
3006 static void *proto_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
3008 return seq_list_next(v
, &proto_list
, pos
);
3011 static void proto_seq_stop(struct seq_file
*seq
, void *v
)
3012 __releases(proto_list_mutex
)
3014 mutex_unlock(&proto_list_mutex
);
3017 static char proto_method_implemented(const void *method
)
3019 return method
== NULL
? 'n' : 'y';
3021 static long sock_prot_memory_allocated(struct proto
*proto
)
3023 return proto
->memory_allocated
!= NULL
? proto_memory_allocated(proto
) : -1L;
3026 static char *sock_prot_memory_pressure(struct proto
*proto
)
3028 return proto
->memory_pressure
!= NULL
?
3029 proto_memory_pressure(proto
) ? "yes" : "no" : "NI";
3032 static void proto_seq_printf(struct seq_file
*seq
, struct proto
*proto
)
3035 seq_printf(seq
, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
3036 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
3039 sock_prot_inuse_get(seq_file_net(seq
), proto
),
3040 sock_prot_memory_allocated(proto
),
3041 sock_prot_memory_pressure(proto
),
3043 proto
->slab
== NULL
? "no" : "yes",
3044 module_name(proto
->owner
),
3045 proto_method_implemented(proto
->close
),
3046 proto_method_implemented(proto
->connect
),
3047 proto_method_implemented(proto
->disconnect
),
3048 proto_method_implemented(proto
->accept
),
3049 proto_method_implemented(proto
->ioctl
),
3050 proto_method_implemented(proto
->init
),
3051 proto_method_implemented(proto
->destroy
),
3052 proto_method_implemented(proto
->shutdown
),
3053 proto_method_implemented(proto
->setsockopt
),
3054 proto_method_implemented(proto
->getsockopt
),
3055 proto_method_implemented(proto
->sendmsg
),
3056 proto_method_implemented(proto
->recvmsg
),
3057 proto_method_implemented(proto
->sendpage
),
3058 proto_method_implemented(proto
->bind
),
3059 proto_method_implemented(proto
->backlog_rcv
),
3060 proto_method_implemented(proto
->hash
),
3061 proto_method_implemented(proto
->unhash
),
3062 proto_method_implemented(proto
->get_port
),
3063 proto_method_implemented(proto
->enter_memory_pressure
));
3066 static int proto_seq_show(struct seq_file
*seq
, void *v
)
3068 if (v
== &proto_list
)
3069 seq_printf(seq
, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
3078 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
3080 proto_seq_printf(seq
, list_entry(v
, struct proto
, node
));
3084 static const struct seq_operations proto_seq_ops
= {
3085 .start
= proto_seq_start
,
3086 .next
= proto_seq_next
,
3087 .stop
= proto_seq_stop
,
3088 .show
= proto_seq_show
,
3091 static int proto_seq_open(struct inode
*inode
, struct file
*file
)
3093 return seq_open_net(inode
, file
, &proto_seq_ops
,
3094 sizeof(struct seq_net_private
));
3097 static const struct file_operations proto_seq_fops
= {
3098 .owner
= THIS_MODULE
,
3099 .open
= proto_seq_open
,
3101 .llseek
= seq_lseek
,
3102 .release
= seq_release_net
,
3105 static __net_init
int proto_init_net(struct net
*net
)
3107 if (!proc_create("protocols", S_IRUGO
, net
->proc_net
, &proto_seq_fops
))
3113 static __net_exit
void proto_exit_net(struct net
*net
)
3115 remove_proc_entry("protocols", net
->proc_net
);
3119 static __net_initdata
struct pernet_operations proto_net_ops
= {
3120 .init
= proto_init_net
,
3121 .exit
= proto_exit_net
,
3124 static int __init
proto_init(void)
3126 return register_pernet_subsys(&proto_net_ops
);
3129 subsys_initcall(proto_init
);
3131 #endif /* PROC_FS */