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 #include <linux/capability.h>
93 #include <linux/errno.h>
94 #include <linux/types.h>
95 #include <linux/socket.h>
97 #include <linux/kernel.h>
98 #include <linux/module.h>
99 #include <linux/proc_fs.h>
100 #include <linux/seq_file.h>
101 #include <linux/sched.h>
102 #include <linux/timer.h>
103 #include <linux/string.h>
104 #include <linux/sockios.h>
105 #include <linux/net.h>
106 #include <linux/mm.h>
107 #include <linux/slab.h>
108 #include <linux/interrupt.h>
109 #include <linux/poll.h>
110 #include <linux/tcp.h>
111 #include <linux/init.h>
112 #include <linux/highmem.h>
114 #include <asm/uaccess.h>
115 #include <asm/system.h>
117 #include <linux/netdevice.h>
118 #include <net/protocol.h>
119 #include <linux/skbuff.h>
120 #include <net/net_namespace.h>
121 #include <net/request_sock.h>
122 #include <net/sock.h>
123 #include <linux/net_tstamp.h>
124 #include <net/xfrm.h>
125 #include <linux/ipsec.h>
127 #include <linux/filter.h>
134 * Each address family might have different locking rules, so we have
135 * one slock key per address family:
137 static struct lock_class_key af_family_keys
[AF_MAX
];
138 static struct lock_class_key af_family_slock_keys
[AF_MAX
];
141 * Make lock validator output more readable. (we pre-construct these
142 * strings build-time, so that runtime initialization of socket
145 static const char *af_family_key_strings
[AF_MAX
+1] = {
146 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
147 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
148 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
149 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
150 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
151 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
152 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
153 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
154 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
155 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
156 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
157 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
158 "sk_lock-AF_IEEE802154",
161 static const char *af_family_slock_key_strings
[AF_MAX
+1] = {
162 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
163 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
164 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
165 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
166 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
167 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
168 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
169 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
170 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
171 "slock-27" , "slock-28" , "slock-AF_CAN" ,
172 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
173 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
174 "slock-AF_IEEE802154",
177 static const char *af_family_clock_key_strings
[AF_MAX
+1] = {
178 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
179 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
180 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
181 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
182 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
183 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
184 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
185 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
186 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
187 "clock-27" , "clock-28" , "clock-AF_CAN" ,
188 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
189 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
190 "clock-AF_IEEE802154",
195 * sk_callback_lock locking rules are per-address-family,
196 * so split the lock classes by using a per-AF key:
198 static struct lock_class_key af_callback_keys
[AF_MAX
];
200 /* Take into consideration the size of the struct sk_buff overhead in the
201 * determination of these values, since that is non-constant across
202 * platforms. This makes socket queueing behavior and performance
203 * not depend upon such differences.
205 #define _SK_MEM_PACKETS 256
206 #define _SK_MEM_OVERHEAD (sizeof(struct sk_buff) + 256)
207 #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
208 #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
210 /* Run time adjustable parameters. */
211 __u32 sysctl_wmem_max __read_mostly
= SK_WMEM_MAX
;
212 __u32 sysctl_rmem_max __read_mostly
= SK_RMEM_MAX
;
213 __u32 sysctl_wmem_default __read_mostly
= SK_WMEM_MAX
;
214 __u32 sysctl_rmem_default __read_mostly
= SK_RMEM_MAX
;
216 /* Maximal space eaten by iovec or ancilliary data plus some space */
217 int sysctl_optmem_max __read_mostly
= sizeof(unsigned long)*(2*UIO_MAXIOV
+512);
218 EXPORT_SYMBOL(sysctl_optmem_max
);
220 static int sock_set_timeout(long *timeo_p
, char __user
*optval
, int optlen
)
224 if (optlen
< sizeof(tv
))
226 if (copy_from_user(&tv
, optval
, sizeof(tv
)))
228 if (tv
.tv_usec
< 0 || tv
.tv_usec
>= USEC_PER_SEC
)
232 static int warned __read_mostly
;
235 if (warned
< 10 && net_ratelimit()) {
237 printk(KERN_INFO
"sock_set_timeout: `%s' (pid %d) "
238 "tries to set negative timeout\n",
239 current
->comm
, task_pid_nr(current
));
243 *timeo_p
= MAX_SCHEDULE_TIMEOUT
;
244 if (tv
.tv_sec
== 0 && tv
.tv_usec
== 0)
246 if (tv
.tv_sec
< (MAX_SCHEDULE_TIMEOUT
/HZ
- 1))
247 *timeo_p
= tv
.tv_sec
*HZ
+ (tv
.tv_usec
+(1000000/HZ
-1))/(1000000/HZ
);
251 static void sock_warn_obsolete_bsdism(const char *name
)
254 static char warncomm
[TASK_COMM_LEN
];
255 if (strcmp(warncomm
, current
->comm
) && warned
< 5) {
256 strcpy(warncomm
, current
->comm
);
257 printk(KERN_WARNING
"process `%s' is using obsolete "
258 "%s SO_BSDCOMPAT\n", warncomm
, name
);
263 static void sock_disable_timestamp(struct sock
*sk
, int flag
)
265 if (sock_flag(sk
, flag
)) {
266 sock_reset_flag(sk
, flag
);
267 if (!sock_flag(sk
, SOCK_TIMESTAMP
) &&
268 !sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
)) {
269 net_disable_timestamp();
275 int sock_queue_rcv_skb(struct sock
*sk
, struct sk_buff
*skb
)
280 /* Cast sk->rcvbuf to unsigned... It's pointless, but reduces
281 number of warnings when compiling with -W --ANK
283 if (atomic_read(&sk
->sk_rmem_alloc
) + skb
->truesize
>=
284 (unsigned)sk
->sk_rcvbuf
) {
289 err
= sk_filter(sk
, skb
);
293 if (!sk_rmem_schedule(sk
, skb
->truesize
)) {
299 skb_set_owner_r(skb
, sk
);
301 /* Cache the SKB length before we tack it onto the receive
302 * queue. Once it is added it no longer belongs to us and
303 * may be freed by other threads of control pulling packets
308 skb_queue_tail(&sk
->sk_receive_queue
, skb
);
310 if (!sock_flag(sk
, SOCK_DEAD
))
311 sk
->sk_data_ready(sk
, skb_len
);
315 EXPORT_SYMBOL(sock_queue_rcv_skb
);
317 int sk_receive_skb(struct sock
*sk
, struct sk_buff
*skb
, const int nested
)
319 int rc
= NET_RX_SUCCESS
;
321 if (sk_filter(sk
, skb
))
322 goto discard_and_relse
;
327 bh_lock_sock_nested(sk
);
330 if (!sock_owned_by_user(sk
)) {
332 * trylock + unlock semantics:
334 mutex_acquire(&sk
->sk_lock
.dep_map
, 0, 1, _RET_IP_
);
336 rc
= sk_backlog_rcv(sk
, skb
);
338 mutex_release(&sk
->sk_lock
.dep_map
, 1, _RET_IP_
);
340 sk_add_backlog(sk
, skb
);
349 EXPORT_SYMBOL(sk_receive_skb
);
351 struct dst_entry
*__sk_dst_check(struct sock
*sk
, u32 cookie
)
353 struct dst_entry
*dst
= sk
->sk_dst_cache
;
355 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, cookie
) == NULL
) {
356 sk
->sk_dst_cache
= NULL
;
363 EXPORT_SYMBOL(__sk_dst_check
);
365 struct dst_entry
*sk_dst_check(struct sock
*sk
, u32 cookie
)
367 struct dst_entry
*dst
= sk_dst_get(sk
);
369 if (dst
&& dst
->obsolete
&& dst
->ops
->check(dst
, cookie
) == NULL
) {
377 EXPORT_SYMBOL(sk_dst_check
);
379 static int sock_bindtodevice(struct sock
*sk
, char __user
*optval
, int optlen
)
381 int ret
= -ENOPROTOOPT
;
382 #ifdef CONFIG_NETDEVICES
383 struct net
*net
= sock_net(sk
);
384 char devname
[IFNAMSIZ
];
389 if (!capable(CAP_NET_RAW
))
396 /* Bind this socket to a particular device like "eth0",
397 * as specified in the passed interface name. If the
398 * name is "" or the option length is zero the socket
401 if (optlen
> IFNAMSIZ
- 1)
402 optlen
= IFNAMSIZ
- 1;
403 memset(devname
, 0, sizeof(devname
));
406 if (copy_from_user(devname
, optval
, optlen
))
409 if (devname
[0] == '\0') {
412 struct net_device
*dev
= dev_get_by_name(net
, devname
);
418 index
= dev
->ifindex
;
423 sk
->sk_bound_dev_if
= index
;
435 static inline void sock_valbool_flag(struct sock
*sk
, int bit
, int valbool
)
438 sock_set_flag(sk
, bit
);
440 sock_reset_flag(sk
, bit
);
444 * This is meant for all protocols to use and covers goings on
445 * at the socket level. Everything here is generic.
448 int sock_setsockopt(struct socket
*sock
, int level
, int optname
,
449 char __user
*optval
, int optlen
)
451 struct sock
*sk
= sock
->sk
;
458 * Options without arguments
461 if (optname
== SO_BINDTODEVICE
)
462 return sock_bindtodevice(sk
, optval
, optlen
);
464 if (optlen
< sizeof(int))
467 if (get_user(val
, (int __user
*)optval
))
470 valbool
= val
? 1 : 0;
476 if (val
&& !capable(CAP_NET_ADMIN
))
479 sock_valbool_flag(sk
, SOCK_DBG
, valbool
);
482 sk
->sk_reuse
= valbool
;
489 sock_valbool_flag(sk
, SOCK_LOCALROUTE
, valbool
);
492 sock_valbool_flag(sk
, SOCK_BROADCAST
, valbool
);
495 /* Don't error on this BSD doesn't and if you think
496 about it this is right. Otherwise apps have to
497 play 'guess the biggest size' games. RCVBUF/SNDBUF
498 are treated in BSD as hints */
500 if (val
> sysctl_wmem_max
)
501 val
= sysctl_wmem_max
;
503 sk
->sk_userlocks
|= SOCK_SNDBUF_LOCK
;
504 if ((val
* 2) < SOCK_MIN_SNDBUF
)
505 sk
->sk_sndbuf
= SOCK_MIN_SNDBUF
;
507 sk
->sk_sndbuf
= val
* 2;
510 * Wake up sending tasks if we
513 sk
->sk_write_space(sk
);
517 if (!capable(CAP_NET_ADMIN
)) {
524 /* Don't error on this BSD doesn't and if you think
525 about it this is right. Otherwise apps have to
526 play 'guess the biggest size' games. RCVBUF/SNDBUF
527 are treated in BSD as hints */
529 if (val
> sysctl_rmem_max
)
530 val
= sysctl_rmem_max
;
532 sk
->sk_userlocks
|= SOCK_RCVBUF_LOCK
;
534 * We double it on the way in to account for
535 * "struct sk_buff" etc. overhead. Applications
536 * assume that the SO_RCVBUF setting they make will
537 * allow that much actual data to be received on that
540 * Applications are unaware that "struct sk_buff" and
541 * other overheads allocate from the receive buffer
542 * during socket buffer allocation.
544 * And after considering the possible alternatives,
545 * returning the value we actually used in getsockopt
546 * is the most desirable behavior.
548 if ((val
* 2) < SOCK_MIN_RCVBUF
)
549 sk
->sk_rcvbuf
= SOCK_MIN_RCVBUF
;
551 sk
->sk_rcvbuf
= val
* 2;
555 if (!capable(CAP_NET_ADMIN
)) {
563 if (sk
->sk_protocol
== IPPROTO_TCP
)
564 tcp_set_keepalive(sk
, valbool
);
566 sock_valbool_flag(sk
, SOCK_KEEPOPEN
, valbool
);
570 sock_valbool_flag(sk
, SOCK_URGINLINE
, valbool
);
574 sk
->sk_no_check
= valbool
;
578 if ((val
>= 0 && val
<= 6) || capable(CAP_NET_ADMIN
))
579 sk
->sk_priority
= val
;
585 if (optlen
< sizeof(ling
)) {
586 ret
= -EINVAL
; /* 1003.1g */
589 if (copy_from_user(&ling
, optval
, sizeof(ling
))) {
594 sock_reset_flag(sk
, SOCK_LINGER
);
596 #if (BITS_PER_LONG == 32)
597 if ((unsigned int)ling
.l_linger
>= MAX_SCHEDULE_TIMEOUT
/HZ
)
598 sk
->sk_lingertime
= MAX_SCHEDULE_TIMEOUT
;
601 sk
->sk_lingertime
= (unsigned int)ling
.l_linger
* HZ
;
602 sock_set_flag(sk
, SOCK_LINGER
);
607 sock_warn_obsolete_bsdism("setsockopt");
612 set_bit(SOCK_PASSCRED
, &sock
->flags
);
614 clear_bit(SOCK_PASSCRED
, &sock
->flags
);
620 if (optname
== SO_TIMESTAMP
)
621 sock_reset_flag(sk
, SOCK_RCVTSTAMPNS
);
623 sock_set_flag(sk
, SOCK_RCVTSTAMPNS
);
624 sock_set_flag(sk
, SOCK_RCVTSTAMP
);
625 sock_enable_timestamp(sk
, SOCK_TIMESTAMP
);
627 sock_reset_flag(sk
, SOCK_RCVTSTAMP
);
628 sock_reset_flag(sk
, SOCK_RCVTSTAMPNS
);
632 case SO_TIMESTAMPING
:
633 if (val
& ~SOF_TIMESTAMPING_MASK
) {
637 sock_valbool_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
,
638 val
& SOF_TIMESTAMPING_TX_HARDWARE
);
639 sock_valbool_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
,
640 val
& SOF_TIMESTAMPING_TX_SOFTWARE
);
641 sock_valbool_flag(sk
, SOCK_TIMESTAMPING_RX_HARDWARE
,
642 val
& SOF_TIMESTAMPING_RX_HARDWARE
);
643 if (val
& SOF_TIMESTAMPING_RX_SOFTWARE
)
644 sock_enable_timestamp(sk
,
645 SOCK_TIMESTAMPING_RX_SOFTWARE
);
647 sock_disable_timestamp(sk
,
648 SOCK_TIMESTAMPING_RX_SOFTWARE
);
649 sock_valbool_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
,
650 val
& SOF_TIMESTAMPING_SOFTWARE
);
651 sock_valbool_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
,
652 val
& SOF_TIMESTAMPING_SYS_HARDWARE
);
653 sock_valbool_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
,
654 val
& SOF_TIMESTAMPING_RAW_HARDWARE
);
660 sk
->sk_rcvlowat
= val
? : 1;
664 ret
= sock_set_timeout(&sk
->sk_rcvtimeo
, optval
, optlen
);
668 ret
= sock_set_timeout(&sk
->sk_sndtimeo
, optval
, optlen
);
671 case SO_ATTACH_FILTER
:
673 if (optlen
== sizeof(struct sock_fprog
)) {
674 struct sock_fprog fprog
;
677 if (copy_from_user(&fprog
, optval
, sizeof(fprog
)))
680 ret
= sk_attach_filter(&fprog
, sk
);
684 case SO_DETACH_FILTER
:
685 ret
= sk_detach_filter(sk
);
690 set_bit(SOCK_PASSSEC
, &sock
->flags
);
692 clear_bit(SOCK_PASSSEC
, &sock
->flags
);
695 if (!capable(CAP_NET_ADMIN
))
701 /* We implement the SO_SNDLOWAT etc to
702 not be settable (1003.1g 5.3) */
710 EXPORT_SYMBOL(sock_setsockopt
);
713 int sock_getsockopt(struct socket
*sock
, int level
, int optname
,
714 char __user
*optval
, int __user
*optlen
)
716 struct sock
*sk
= sock
->sk
;
724 unsigned int lv
= sizeof(int);
727 if (get_user(len
, optlen
))
732 memset(&v
, 0, sizeof(v
));
736 v
.val
= sock_flag(sk
, SOCK_DBG
);
740 v
.val
= sock_flag(sk
, SOCK_LOCALROUTE
);
744 v
.val
= !!sock_flag(sk
, SOCK_BROADCAST
);
748 v
.val
= sk
->sk_sndbuf
;
752 v
.val
= sk
->sk_rcvbuf
;
756 v
.val
= sk
->sk_reuse
;
760 v
.val
= !!sock_flag(sk
, SOCK_KEEPOPEN
);
768 v
.val
= -sock_error(sk
);
770 v
.val
= xchg(&sk
->sk_err_soft
, 0);
774 v
.val
= !!sock_flag(sk
, SOCK_URGINLINE
);
778 v
.val
= sk
->sk_no_check
;
782 v
.val
= sk
->sk_priority
;
787 v
.ling
.l_onoff
= !!sock_flag(sk
, SOCK_LINGER
);
788 v
.ling
.l_linger
= sk
->sk_lingertime
/ HZ
;
792 sock_warn_obsolete_bsdism("getsockopt");
796 v
.val
= sock_flag(sk
, SOCK_RCVTSTAMP
) &&
797 !sock_flag(sk
, SOCK_RCVTSTAMPNS
);
801 v
.val
= sock_flag(sk
, SOCK_RCVTSTAMPNS
);
804 case SO_TIMESTAMPING
:
806 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_HARDWARE
))
807 v
.val
|= SOF_TIMESTAMPING_TX_HARDWARE
;
808 if (sock_flag(sk
, SOCK_TIMESTAMPING_TX_SOFTWARE
))
809 v
.val
|= SOF_TIMESTAMPING_TX_SOFTWARE
;
810 if (sock_flag(sk
, SOCK_TIMESTAMPING_RX_HARDWARE
))
811 v
.val
|= SOF_TIMESTAMPING_RX_HARDWARE
;
812 if (sock_flag(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
))
813 v
.val
|= SOF_TIMESTAMPING_RX_SOFTWARE
;
814 if (sock_flag(sk
, SOCK_TIMESTAMPING_SOFTWARE
))
815 v
.val
|= SOF_TIMESTAMPING_SOFTWARE
;
816 if (sock_flag(sk
, SOCK_TIMESTAMPING_SYS_HARDWARE
))
817 v
.val
|= SOF_TIMESTAMPING_SYS_HARDWARE
;
818 if (sock_flag(sk
, SOCK_TIMESTAMPING_RAW_HARDWARE
))
819 v
.val
|= SOF_TIMESTAMPING_RAW_HARDWARE
;
823 lv
= sizeof(struct timeval
);
824 if (sk
->sk_rcvtimeo
== MAX_SCHEDULE_TIMEOUT
) {
828 v
.tm
.tv_sec
= sk
->sk_rcvtimeo
/ HZ
;
829 v
.tm
.tv_usec
= ((sk
->sk_rcvtimeo
% HZ
) * 1000000) / HZ
;
834 lv
= sizeof(struct timeval
);
835 if (sk
->sk_sndtimeo
== MAX_SCHEDULE_TIMEOUT
) {
839 v
.tm
.tv_sec
= sk
->sk_sndtimeo
/ HZ
;
840 v
.tm
.tv_usec
= ((sk
->sk_sndtimeo
% HZ
) * 1000000) / HZ
;
845 v
.val
= sk
->sk_rcvlowat
;
853 v
.val
= test_bit(SOCK_PASSCRED
, &sock
->flags
) ? 1 : 0;
857 if (len
> sizeof(sk
->sk_peercred
))
858 len
= sizeof(sk
->sk_peercred
);
859 if (copy_to_user(optval
, &sk
->sk_peercred
, len
))
867 if (sock
->ops
->getname(sock
, (struct sockaddr
*)address
, &lv
, 2))
871 if (copy_to_user(optval
, address
, len
))
876 /* Dubious BSD thing... Probably nobody even uses it, but
877 * the UNIX standard wants it for whatever reason... -DaveM
880 v
.val
= sk
->sk_state
== TCP_LISTEN
;
884 v
.val
= test_bit(SOCK_PASSSEC
, &sock
->flags
) ? 1 : 0;
888 return security_socket_getpeersec_stream(sock
, optval
, optlen
, len
);
900 if (copy_to_user(optval
, &v
, len
))
903 if (put_user(len
, optlen
))
909 * Initialize an sk_lock.
911 * (We also register the sk_lock with the lock validator.)
913 static inline void sock_lock_init(struct sock
*sk
)
915 sock_lock_init_class_and_name(sk
,
916 af_family_slock_key_strings
[sk
->sk_family
],
917 af_family_slock_keys
+ sk
->sk_family
,
918 af_family_key_strings
[sk
->sk_family
],
919 af_family_keys
+ sk
->sk_family
);
922 static void sock_copy(struct sock
*nsk
, const struct sock
*osk
)
924 #ifdef CONFIG_SECURITY_NETWORK
925 void *sptr
= nsk
->sk_security
;
928 memcpy(nsk
, osk
, osk
->sk_prot
->obj_size
);
929 #ifdef CONFIG_SECURITY_NETWORK
930 nsk
->sk_security
= sptr
;
931 security_sk_clone(osk
, nsk
);
935 static struct sock
*sk_prot_alloc(struct proto
*prot
, gfp_t priority
,
939 struct kmem_cache
*slab
;
943 sk
= kmem_cache_alloc(slab
, priority
& ~__GFP_ZERO
);
946 if (priority
& __GFP_ZERO
) {
948 * caches using SLAB_DESTROY_BY_RCU should let
949 * sk_node.next un-modified. Special care is taken
950 * when initializing object to zero.
952 if (offsetof(struct sock
, sk_node
.next
) != 0)
953 memset(sk
, 0, offsetof(struct sock
, sk_node
.next
));
954 memset(&sk
->sk_node
.pprev
, 0,
955 prot
->obj_size
- offsetof(struct sock
,
960 sk
= kmalloc(prot
->obj_size
, priority
);
963 kmemcheck_annotate_bitfield(sk
, flags
);
965 if (security_sk_alloc(sk
, family
, priority
))
968 if (!try_module_get(prot
->owner
))
975 security_sk_free(sk
);
978 kmem_cache_free(slab
, sk
);
984 static void sk_prot_free(struct proto
*prot
, struct sock
*sk
)
986 struct kmem_cache
*slab
;
987 struct module
*owner
;
992 security_sk_free(sk
);
994 kmem_cache_free(slab
, sk
);
1001 * sk_alloc - All socket objects are allocated here
1002 * @net: the applicable net namespace
1003 * @family: protocol family
1004 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1005 * @prot: struct proto associated with this new sock instance
1007 struct sock
*sk_alloc(struct net
*net
, int family
, gfp_t priority
,
1012 sk
= sk_prot_alloc(prot
, priority
| __GFP_ZERO
, family
);
1014 sk
->sk_family
= family
;
1016 * See comment in struct sock definition to understand
1017 * why we need sk_prot_creator -acme
1019 sk
->sk_prot
= sk
->sk_prot_creator
= prot
;
1021 sock_net_set(sk
, get_net(net
));
1026 EXPORT_SYMBOL(sk_alloc
);
1028 static void __sk_free(struct sock
*sk
)
1030 struct sk_filter
*filter
;
1032 if (sk
->sk_destruct
)
1033 sk
->sk_destruct(sk
);
1035 filter
= rcu_dereference(sk
->sk_filter
);
1037 sk_filter_uncharge(sk
, filter
);
1038 rcu_assign_pointer(sk
->sk_filter
, NULL
);
1041 sock_disable_timestamp(sk
, SOCK_TIMESTAMP
);
1042 sock_disable_timestamp(sk
, SOCK_TIMESTAMPING_RX_SOFTWARE
);
1044 if (atomic_read(&sk
->sk_omem_alloc
))
1045 printk(KERN_DEBUG
"%s: optmem leakage (%d bytes) detected.\n",
1046 __func__
, atomic_read(&sk
->sk_omem_alloc
));
1048 put_net(sock_net(sk
));
1049 sk_prot_free(sk
->sk_prot_creator
, sk
);
1052 void sk_free(struct sock
*sk
)
1055 * We substract one from sk_wmem_alloc and can know if
1056 * some packets are still in some tx queue.
1057 * If not null, sock_wfree() will call __sk_free(sk) later
1059 if (atomic_dec_and_test(&sk
->sk_wmem_alloc
))
1062 EXPORT_SYMBOL(sk_free
);
1065 * Last sock_put should drop referrence to sk->sk_net. It has already
1066 * been dropped in sk_change_net. Taking referrence to stopping namespace
1068 * Take referrence to a socket to remove it from hash _alive_ and after that
1069 * destroy it in the context of init_net.
1071 void sk_release_kernel(struct sock
*sk
)
1073 if (sk
== NULL
|| sk
->sk_socket
== NULL
)
1077 sock_release(sk
->sk_socket
);
1078 release_net(sock_net(sk
));
1079 sock_net_set(sk
, get_net(&init_net
));
1082 EXPORT_SYMBOL(sk_release_kernel
);
1084 struct sock
*sk_clone(const struct sock
*sk
, const gfp_t priority
)
1088 newsk
= sk_prot_alloc(sk
->sk_prot
, priority
, sk
->sk_family
);
1089 if (newsk
!= NULL
) {
1090 struct sk_filter
*filter
;
1092 sock_copy(newsk
, sk
);
1095 get_net(sock_net(newsk
));
1096 sk_node_init(&newsk
->sk_node
);
1097 sock_lock_init(newsk
);
1098 bh_lock_sock(newsk
);
1099 newsk
->sk_backlog
.head
= newsk
->sk_backlog
.tail
= NULL
;
1101 atomic_set(&newsk
->sk_rmem_alloc
, 0);
1103 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1105 atomic_set(&newsk
->sk_wmem_alloc
, 1);
1106 atomic_set(&newsk
->sk_omem_alloc
, 0);
1107 skb_queue_head_init(&newsk
->sk_receive_queue
);
1108 skb_queue_head_init(&newsk
->sk_write_queue
);
1109 #ifdef CONFIG_NET_DMA
1110 skb_queue_head_init(&newsk
->sk_async_wait_queue
);
1113 rwlock_init(&newsk
->sk_dst_lock
);
1114 rwlock_init(&newsk
->sk_callback_lock
);
1115 lockdep_set_class_and_name(&newsk
->sk_callback_lock
,
1116 af_callback_keys
+ newsk
->sk_family
,
1117 af_family_clock_key_strings
[newsk
->sk_family
]);
1119 newsk
->sk_dst_cache
= NULL
;
1120 newsk
->sk_wmem_queued
= 0;
1121 newsk
->sk_forward_alloc
= 0;
1122 newsk
->sk_send_head
= NULL
;
1123 newsk
->sk_userlocks
= sk
->sk_userlocks
& ~SOCK_BINDPORT_LOCK
;
1125 sock_reset_flag(newsk
, SOCK_DONE
);
1126 skb_queue_head_init(&newsk
->sk_error_queue
);
1128 filter
= newsk
->sk_filter
;
1130 sk_filter_charge(newsk
, filter
);
1132 if (unlikely(xfrm_sk_clone_policy(newsk
))) {
1133 /* It is still raw copy of parent, so invalidate
1134 * destructor and make plain sk_free() */
1135 newsk
->sk_destruct
= NULL
;
1142 newsk
->sk_priority
= 0;
1143 atomic_set(&newsk
->sk_refcnt
, 2);
1146 * Increment the counter in the same struct proto as the master
1147 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1148 * is the same as sk->sk_prot->socks, as this field was copied
1151 * This _changes_ the previous behaviour, where
1152 * tcp_create_openreq_child always was incrementing the
1153 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1154 * to be taken into account in all callers. -acme
1156 sk_refcnt_debug_inc(newsk
);
1157 sk_set_socket(newsk
, NULL
);
1158 newsk
->sk_sleep
= NULL
;
1160 if (newsk
->sk_prot
->sockets_allocated
)
1161 percpu_counter_inc(newsk
->sk_prot
->sockets_allocated
);
1166 EXPORT_SYMBOL_GPL(sk_clone
);
1168 void sk_setup_caps(struct sock
*sk
, struct dst_entry
*dst
)
1170 __sk_dst_set(sk
, dst
);
1171 sk
->sk_route_caps
= dst
->dev
->features
;
1172 if (sk
->sk_route_caps
& NETIF_F_GSO
)
1173 sk
->sk_route_caps
|= NETIF_F_GSO_SOFTWARE
;
1174 if (sk_can_gso(sk
)) {
1175 if (dst
->header_len
) {
1176 sk
->sk_route_caps
&= ~NETIF_F_GSO_MASK
;
1178 sk
->sk_route_caps
|= NETIF_F_SG
| NETIF_F_HW_CSUM
;
1179 sk
->sk_gso_max_size
= dst
->dev
->gso_max_size
;
1183 EXPORT_SYMBOL_GPL(sk_setup_caps
);
1185 void __init
sk_init(void)
1187 if (num_physpages
<= 4096) {
1188 sysctl_wmem_max
= 32767;
1189 sysctl_rmem_max
= 32767;
1190 sysctl_wmem_default
= 32767;
1191 sysctl_rmem_default
= 32767;
1192 } else if (num_physpages
>= 131072) {
1193 sysctl_wmem_max
= 131071;
1194 sysctl_rmem_max
= 131071;
1199 * Simple resource managers for sockets.
1204 * Write buffer destructor automatically called from kfree_skb.
1206 void sock_wfree(struct sk_buff
*skb
)
1208 struct sock
*sk
= skb
->sk
;
1211 /* In case it might be waiting for more memory. */
1212 res
= atomic_sub_return(skb
->truesize
, &sk
->sk_wmem_alloc
);
1213 if (!sock_flag(sk
, SOCK_USE_WRITE_QUEUE
))
1214 sk
->sk_write_space(sk
);
1216 * if sk_wmem_alloc reached 0, we are last user and should
1217 * free this sock, as sk_free() call could not do it.
1222 EXPORT_SYMBOL(sock_wfree
);
1225 * Read buffer destructor automatically called from kfree_skb.
1227 void sock_rfree(struct sk_buff
*skb
)
1229 struct sock
*sk
= skb
->sk
;
1231 atomic_sub(skb
->truesize
, &sk
->sk_rmem_alloc
);
1232 sk_mem_uncharge(skb
->sk
, skb
->truesize
);
1234 EXPORT_SYMBOL(sock_rfree
);
1237 int sock_i_uid(struct sock
*sk
)
1241 read_lock(&sk
->sk_callback_lock
);
1242 uid
= sk
->sk_socket
? SOCK_INODE(sk
->sk_socket
)->i_uid
: 0;
1243 read_unlock(&sk
->sk_callback_lock
);
1246 EXPORT_SYMBOL(sock_i_uid
);
1248 unsigned long sock_i_ino(struct sock
*sk
)
1252 read_lock(&sk
->sk_callback_lock
);
1253 ino
= sk
->sk_socket
? SOCK_INODE(sk
->sk_socket
)->i_ino
: 0;
1254 read_unlock(&sk
->sk_callback_lock
);
1257 EXPORT_SYMBOL(sock_i_ino
);
1260 * Allocate a skb from the socket's send buffer.
1262 struct sk_buff
*sock_wmalloc(struct sock
*sk
, unsigned long size
, int force
,
1265 if (force
|| atomic_read(&sk
->sk_wmem_alloc
) < sk
->sk_sndbuf
) {
1266 struct sk_buff
*skb
= alloc_skb(size
, priority
);
1268 skb_set_owner_w(skb
, sk
);
1274 EXPORT_SYMBOL(sock_wmalloc
);
1277 * Allocate a skb from the socket's receive buffer.
1279 struct sk_buff
*sock_rmalloc(struct sock
*sk
, unsigned long size
, int force
,
1282 if (force
|| atomic_read(&sk
->sk_rmem_alloc
) < sk
->sk_rcvbuf
) {
1283 struct sk_buff
*skb
= alloc_skb(size
, priority
);
1285 skb_set_owner_r(skb
, sk
);
1293 * Allocate a memory block from the socket's option memory buffer.
1295 void *sock_kmalloc(struct sock
*sk
, int size
, gfp_t priority
)
1297 if ((unsigned)size
<= sysctl_optmem_max
&&
1298 atomic_read(&sk
->sk_omem_alloc
) + size
< sysctl_optmem_max
) {
1300 /* First do the add, to avoid the race if kmalloc
1303 atomic_add(size
, &sk
->sk_omem_alloc
);
1304 mem
= kmalloc(size
, priority
);
1307 atomic_sub(size
, &sk
->sk_omem_alloc
);
1311 EXPORT_SYMBOL(sock_kmalloc
);
1314 * Free an option memory block.
1316 void sock_kfree_s(struct sock
*sk
, void *mem
, int size
)
1319 atomic_sub(size
, &sk
->sk_omem_alloc
);
1321 EXPORT_SYMBOL(sock_kfree_s
);
1323 /* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1324 I think, these locks should be removed for datagram sockets.
1326 static long sock_wait_for_wmem(struct sock
*sk
, long timeo
)
1330 clear_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
1334 if (signal_pending(current
))
1336 set_bit(SOCK_NOSPACE
, &sk
->sk_socket
->flags
);
1337 prepare_to_wait(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
1338 if (atomic_read(&sk
->sk_wmem_alloc
) < sk
->sk_sndbuf
)
1340 if (sk
->sk_shutdown
& SEND_SHUTDOWN
)
1344 timeo
= schedule_timeout(timeo
);
1346 finish_wait(sk
->sk_sleep
, &wait
);
1352 * Generic send/receive buffer handlers
1355 struct sk_buff
*sock_alloc_send_pskb(struct sock
*sk
, unsigned long header_len
,
1356 unsigned long data_len
, int noblock
,
1359 struct sk_buff
*skb
;
1364 gfp_mask
= sk
->sk_allocation
;
1365 if (gfp_mask
& __GFP_WAIT
)
1366 gfp_mask
|= __GFP_REPEAT
;
1368 timeo
= sock_sndtimeo(sk
, noblock
);
1370 err
= sock_error(sk
);
1375 if (sk
->sk_shutdown
& SEND_SHUTDOWN
)
1378 if (atomic_read(&sk
->sk_wmem_alloc
) < sk
->sk_sndbuf
) {
1379 skb
= alloc_skb(header_len
, gfp_mask
);
1384 /* No pages, we're done... */
1388 npages
= (data_len
+ (PAGE_SIZE
- 1)) >> PAGE_SHIFT
;
1389 skb
->truesize
+= data_len
;
1390 skb_shinfo(skb
)->nr_frags
= npages
;
1391 for (i
= 0; i
< npages
; i
++) {
1395 page
= alloc_pages(sk
->sk_allocation
, 0);
1398 skb_shinfo(skb
)->nr_frags
= i
;
1403 frag
= &skb_shinfo(skb
)->frags
[i
];
1405 frag
->page_offset
= 0;
1406 frag
->size
= (data_len
>= PAGE_SIZE
?
1409 data_len
-= PAGE_SIZE
;
1412 /* Full success... */
1418 set_bit(SOCK_ASYNC_NOSPACE
, &sk
->sk_socket
->flags
);
1419 set_bit(SOCK_NOSPACE
, &sk
->sk_socket
->flags
);
1423 if (signal_pending(current
))
1425 timeo
= sock_wait_for_wmem(sk
, timeo
);
1428 skb_set_owner_w(skb
, sk
);
1432 err
= sock_intr_errno(timeo
);
1437 EXPORT_SYMBOL(sock_alloc_send_pskb
);
1439 struct sk_buff
*sock_alloc_send_skb(struct sock
*sk
, unsigned long size
,
1440 int noblock
, int *errcode
)
1442 return sock_alloc_send_pskb(sk
, size
, 0, noblock
, errcode
);
1444 EXPORT_SYMBOL(sock_alloc_send_skb
);
1446 static void __lock_sock(struct sock
*sk
)
1451 prepare_to_wait_exclusive(&sk
->sk_lock
.wq
, &wait
,
1452 TASK_UNINTERRUPTIBLE
);
1453 spin_unlock_bh(&sk
->sk_lock
.slock
);
1455 spin_lock_bh(&sk
->sk_lock
.slock
);
1456 if (!sock_owned_by_user(sk
))
1459 finish_wait(&sk
->sk_lock
.wq
, &wait
);
1462 static void __release_sock(struct sock
*sk
)
1464 struct sk_buff
*skb
= sk
->sk_backlog
.head
;
1467 sk
->sk_backlog
.head
= sk
->sk_backlog
.tail
= NULL
;
1471 struct sk_buff
*next
= skb
->next
;
1474 sk_backlog_rcv(sk
, skb
);
1477 * We are in process context here with softirqs
1478 * disabled, use cond_resched_softirq() to preempt.
1479 * This is safe to do because we've taken the backlog
1482 cond_resched_softirq();
1485 } while (skb
!= NULL
);
1488 } while ((skb
= sk
->sk_backlog
.head
) != NULL
);
1492 * sk_wait_data - wait for data to arrive at sk_receive_queue
1493 * @sk: sock to wait on
1494 * @timeo: for how long
1496 * Now socket state including sk->sk_err is changed only under lock,
1497 * hence we may omit checks after joining wait queue.
1498 * We check receive queue before schedule() only as optimization;
1499 * it is very likely that release_sock() added new data.
1501 int sk_wait_data(struct sock
*sk
, long *timeo
)
1506 prepare_to_wait(sk
->sk_sleep
, &wait
, TASK_INTERRUPTIBLE
);
1507 set_bit(SOCK_ASYNC_WAITDATA
, &sk
->sk_socket
->flags
);
1508 rc
= sk_wait_event(sk
, timeo
, !skb_queue_empty(&sk
->sk_receive_queue
));
1509 clear_bit(SOCK_ASYNC_WAITDATA
, &sk
->sk_socket
->flags
);
1510 finish_wait(sk
->sk_sleep
, &wait
);
1513 EXPORT_SYMBOL(sk_wait_data
);
1516 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1518 * @size: memory size to allocate
1519 * @kind: allocation type
1521 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1522 * rmem allocation. This function assumes that protocols which have
1523 * memory_pressure use sk_wmem_queued as write buffer accounting.
1525 int __sk_mem_schedule(struct sock
*sk
, int size
, int kind
)
1527 struct proto
*prot
= sk
->sk_prot
;
1528 int amt
= sk_mem_pages(size
);
1531 sk
->sk_forward_alloc
+= amt
* SK_MEM_QUANTUM
;
1532 allocated
= atomic_add_return(amt
, prot
->memory_allocated
);
1535 if (allocated
<= prot
->sysctl_mem
[0]) {
1536 if (prot
->memory_pressure
&& *prot
->memory_pressure
)
1537 *prot
->memory_pressure
= 0;
1541 /* Under pressure. */
1542 if (allocated
> prot
->sysctl_mem
[1])
1543 if (prot
->enter_memory_pressure
)
1544 prot
->enter_memory_pressure(sk
);
1546 /* Over hard limit. */
1547 if (allocated
> prot
->sysctl_mem
[2])
1548 goto suppress_allocation
;
1550 /* guarantee minimum buffer size under pressure */
1551 if (kind
== SK_MEM_RECV
) {
1552 if (atomic_read(&sk
->sk_rmem_alloc
) < prot
->sysctl_rmem
[0])
1554 } else { /* SK_MEM_SEND */
1555 if (sk
->sk_type
== SOCK_STREAM
) {
1556 if (sk
->sk_wmem_queued
< prot
->sysctl_wmem
[0])
1558 } else if (atomic_read(&sk
->sk_wmem_alloc
) <
1559 prot
->sysctl_wmem
[0])
1563 if (prot
->memory_pressure
) {
1566 if (!*prot
->memory_pressure
)
1568 alloc
= percpu_counter_read_positive(prot
->sockets_allocated
);
1569 if (prot
->sysctl_mem
[2] > alloc
*
1570 sk_mem_pages(sk
->sk_wmem_queued
+
1571 atomic_read(&sk
->sk_rmem_alloc
) +
1572 sk
->sk_forward_alloc
))
1576 suppress_allocation
:
1578 if (kind
== SK_MEM_SEND
&& sk
->sk_type
== SOCK_STREAM
) {
1579 sk_stream_moderate_sndbuf(sk
);
1581 /* Fail only if socket is _under_ its sndbuf.
1582 * In this case we cannot block, so that we have to fail.
1584 if (sk
->sk_wmem_queued
+ size
>= sk
->sk_sndbuf
)
1588 /* Alas. Undo changes. */
1589 sk
->sk_forward_alloc
-= amt
* SK_MEM_QUANTUM
;
1590 atomic_sub(amt
, prot
->memory_allocated
);
1593 EXPORT_SYMBOL(__sk_mem_schedule
);
1596 * __sk_reclaim - reclaim memory_allocated
1599 void __sk_mem_reclaim(struct sock
*sk
)
1601 struct proto
*prot
= sk
->sk_prot
;
1603 atomic_sub(sk
->sk_forward_alloc
>> SK_MEM_QUANTUM_SHIFT
,
1604 prot
->memory_allocated
);
1605 sk
->sk_forward_alloc
&= SK_MEM_QUANTUM
- 1;
1607 if (prot
->memory_pressure
&& *prot
->memory_pressure
&&
1608 (atomic_read(prot
->memory_allocated
) < prot
->sysctl_mem
[0]))
1609 *prot
->memory_pressure
= 0;
1611 EXPORT_SYMBOL(__sk_mem_reclaim
);
1615 * Set of default routines for initialising struct proto_ops when
1616 * the protocol does not support a particular function. In certain
1617 * cases where it makes no sense for a protocol to have a "do nothing"
1618 * function, some default processing is provided.
1621 int sock_no_bind(struct socket
*sock
, struct sockaddr
*saddr
, int len
)
1625 EXPORT_SYMBOL(sock_no_bind
);
1627 int sock_no_connect(struct socket
*sock
, struct sockaddr
*saddr
,
1632 EXPORT_SYMBOL(sock_no_connect
);
1634 int sock_no_socketpair(struct socket
*sock1
, struct socket
*sock2
)
1638 EXPORT_SYMBOL(sock_no_socketpair
);
1640 int sock_no_accept(struct socket
*sock
, struct socket
*newsock
, int flags
)
1644 EXPORT_SYMBOL(sock_no_accept
);
1646 int sock_no_getname(struct socket
*sock
, struct sockaddr
*saddr
,
1651 EXPORT_SYMBOL(sock_no_getname
);
1653 unsigned int sock_no_poll(struct file
*file
, struct socket
*sock
, poll_table
*pt
)
1657 EXPORT_SYMBOL(sock_no_poll
);
1659 int sock_no_ioctl(struct socket
*sock
, unsigned int cmd
, unsigned long arg
)
1663 EXPORT_SYMBOL(sock_no_ioctl
);
1665 int sock_no_listen(struct socket
*sock
, int backlog
)
1669 EXPORT_SYMBOL(sock_no_listen
);
1671 int sock_no_shutdown(struct socket
*sock
, int how
)
1675 EXPORT_SYMBOL(sock_no_shutdown
);
1677 int sock_no_setsockopt(struct socket
*sock
, int level
, int optname
,
1678 char __user
*optval
, int optlen
)
1682 EXPORT_SYMBOL(sock_no_setsockopt
);
1684 int sock_no_getsockopt(struct socket
*sock
, int level
, int optname
,
1685 char __user
*optval
, int __user
*optlen
)
1689 EXPORT_SYMBOL(sock_no_getsockopt
);
1691 int sock_no_sendmsg(struct kiocb
*iocb
, struct socket
*sock
, struct msghdr
*m
,
1696 EXPORT_SYMBOL(sock_no_sendmsg
);
1698 int sock_no_recvmsg(struct kiocb
*iocb
, struct socket
*sock
, struct msghdr
*m
,
1699 size_t len
, int flags
)
1703 EXPORT_SYMBOL(sock_no_recvmsg
);
1705 int sock_no_mmap(struct file
*file
, struct socket
*sock
, struct vm_area_struct
*vma
)
1707 /* Mirror missing mmap method error code */
1710 EXPORT_SYMBOL(sock_no_mmap
);
1712 ssize_t
sock_no_sendpage(struct socket
*sock
, struct page
*page
, int offset
, size_t size
, int flags
)
1715 struct msghdr msg
= {.msg_flags
= flags
};
1717 char *kaddr
= kmap(page
);
1718 iov
.iov_base
= kaddr
+ offset
;
1720 res
= kernel_sendmsg(sock
, &msg
, &iov
, 1, size
);
1724 EXPORT_SYMBOL(sock_no_sendpage
);
1727 * Default Socket Callbacks
1730 static void sock_def_wakeup(struct sock
*sk
)
1732 read_lock(&sk
->sk_callback_lock
);
1733 if (sk_has_sleeper(sk
))
1734 wake_up_interruptible_all(sk
->sk_sleep
);
1735 read_unlock(&sk
->sk_callback_lock
);
1738 static void sock_def_error_report(struct sock
*sk
)
1740 read_lock(&sk
->sk_callback_lock
);
1741 if (sk_has_sleeper(sk
))
1742 wake_up_interruptible_poll(sk
->sk_sleep
, POLLERR
);
1743 sk_wake_async(sk
, SOCK_WAKE_IO
, POLL_ERR
);
1744 read_unlock(&sk
->sk_callback_lock
);
1747 static void sock_def_readable(struct sock
*sk
, int len
)
1749 read_lock(&sk
->sk_callback_lock
);
1750 if (sk_has_sleeper(sk
))
1751 wake_up_interruptible_sync_poll(sk
->sk_sleep
, POLLIN
|
1752 POLLRDNORM
| POLLRDBAND
);
1753 sk_wake_async(sk
, SOCK_WAKE_WAITD
, POLL_IN
);
1754 read_unlock(&sk
->sk_callback_lock
);
1757 static void sock_def_write_space(struct sock
*sk
)
1759 read_lock(&sk
->sk_callback_lock
);
1761 /* Do not wake up a writer until he can make "significant"
1764 if ((atomic_read(&sk
->sk_wmem_alloc
) << 1) <= sk
->sk_sndbuf
) {
1765 if (sk_has_sleeper(sk
))
1766 wake_up_interruptible_sync_poll(sk
->sk_sleep
, POLLOUT
|
1767 POLLWRNORM
| POLLWRBAND
);
1769 /* Should agree with poll, otherwise some programs break */
1770 if (sock_writeable(sk
))
1771 sk_wake_async(sk
, SOCK_WAKE_SPACE
, POLL_OUT
);
1774 read_unlock(&sk
->sk_callback_lock
);
1777 static void sock_def_destruct(struct sock
*sk
)
1779 kfree(sk
->sk_protinfo
);
1782 void sk_send_sigurg(struct sock
*sk
)
1784 if (sk
->sk_socket
&& sk
->sk_socket
->file
)
1785 if (send_sigurg(&sk
->sk_socket
->file
->f_owner
))
1786 sk_wake_async(sk
, SOCK_WAKE_URG
, POLL_PRI
);
1788 EXPORT_SYMBOL(sk_send_sigurg
);
1790 void sk_reset_timer(struct sock
*sk
, struct timer_list
* timer
,
1791 unsigned long expires
)
1793 if (!mod_timer(timer
, expires
))
1796 EXPORT_SYMBOL(sk_reset_timer
);
1798 void sk_stop_timer(struct sock
*sk
, struct timer_list
* timer
)
1800 if (timer_pending(timer
) && del_timer(timer
))
1803 EXPORT_SYMBOL(sk_stop_timer
);
1805 void sock_init_data(struct socket
*sock
, struct sock
*sk
)
1807 skb_queue_head_init(&sk
->sk_receive_queue
);
1808 skb_queue_head_init(&sk
->sk_write_queue
);
1809 skb_queue_head_init(&sk
->sk_error_queue
);
1810 #ifdef CONFIG_NET_DMA
1811 skb_queue_head_init(&sk
->sk_async_wait_queue
);
1814 sk
->sk_send_head
= NULL
;
1816 init_timer(&sk
->sk_timer
);
1818 sk
->sk_allocation
= GFP_KERNEL
;
1819 sk
->sk_rcvbuf
= sysctl_rmem_default
;
1820 sk
->sk_sndbuf
= sysctl_wmem_default
;
1821 sk
->sk_state
= TCP_CLOSE
;
1822 sk_set_socket(sk
, sock
);
1824 sock_set_flag(sk
, SOCK_ZAPPED
);
1827 sk
->sk_type
= sock
->type
;
1828 sk
->sk_sleep
= &sock
->wait
;
1831 sk
->sk_sleep
= NULL
;
1833 rwlock_init(&sk
->sk_dst_lock
);
1834 rwlock_init(&sk
->sk_callback_lock
);
1835 lockdep_set_class_and_name(&sk
->sk_callback_lock
,
1836 af_callback_keys
+ sk
->sk_family
,
1837 af_family_clock_key_strings
[sk
->sk_family
]);
1839 sk
->sk_state_change
= sock_def_wakeup
;
1840 sk
->sk_data_ready
= sock_def_readable
;
1841 sk
->sk_write_space
= sock_def_write_space
;
1842 sk
->sk_error_report
= sock_def_error_report
;
1843 sk
->sk_destruct
= sock_def_destruct
;
1845 sk
->sk_sndmsg_page
= NULL
;
1846 sk
->sk_sndmsg_off
= 0;
1848 sk
->sk_peercred
.pid
= 0;
1849 sk
->sk_peercred
.uid
= -1;
1850 sk
->sk_peercred
.gid
= -1;
1851 sk
->sk_write_pending
= 0;
1852 sk
->sk_rcvlowat
= 1;
1853 sk
->sk_rcvtimeo
= MAX_SCHEDULE_TIMEOUT
;
1854 sk
->sk_sndtimeo
= MAX_SCHEDULE_TIMEOUT
;
1856 sk
->sk_stamp
= ktime_set(-1L, 0);
1858 atomic_set(&sk
->sk_refcnt
, 1);
1859 atomic_set(&sk
->sk_wmem_alloc
, 1);
1860 atomic_set(&sk
->sk_drops
, 0);
1862 EXPORT_SYMBOL(sock_init_data
);
1864 void lock_sock_nested(struct sock
*sk
, int subclass
)
1867 spin_lock_bh(&sk
->sk_lock
.slock
);
1868 if (sk
->sk_lock
.owned
)
1870 sk
->sk_lock
.owned
= 1;
1871 spin_unlock(&sk
->sk_lock
.slock
);
1873 * The sk_lock has mutex_lock() semantics here:
1875 mutex_acquire(&sk
->sk_lock
.dep_map
, subclass
, 0, _RET_IP_
);
1878 EXPORT_SYMBOL(lock_sock_nested
);
1880 void release_sock(struct sock
*sk
)
1883 * The sk_lock has mutex_unlock() semantics:
1885 mutex_release(&sk
->sk_lock
.dep_map
, 1, _RET_IP_
);
1887 spin_lock_bh(&sk
->sk_lock
.slock
);
1888 if (sk
->sk_backlog
.tail
)
1890 sk
->sk_lock
.owned
= 0;
1891 if (waitqueue_active(&sk
->sk_lock
.wq
))
1892 wake_up(&sk
->sk_lock
.wq
);
1893 spin_unlock_bh(&sk
->sk_lock
.slock
);
1895 EXPORT_SYMBOL(release_sock
);
1897 int sock_get_timestamp(struct sock
*sk
, struct timeval __user
*userstamp
)
1900 if (!sock_flag(sk
, SOCK_TIMESTAMP
))
1901 sock_enable_timestamp(sk
, SOCK_TIMESTAMP
);
1902 tv
= ktime_to_timeval(sk
->sk_stamp
);
1903 if (tv
.tv_sec
== -1)
1905 if (tv
.tv_sec
== 0) {
1906 sk
->sk_stamp
= ktime_get_real();
1907 tv
= ktime_to_timeval(sk
->sk_stamp
);
1909 return copy_to_user(userstamp
, &tv
, sizeof(tv
)) ? -EFAULT
: 0;
1911 EXPORT_SYMBOL(sock_get_timestamp
);
1913 int sock_get_timestampns(struct sock
*sk
, struct timespec __user
*userstamp
)
1916 if (!sock_flag(sk
, SOCK_TIMESTAMP
))
1917 sock_enable_timestamp(sk
, SOCK_TIMESTAMP
);
1918 ts
= ktime_to_timespec(sk
->sk_stamp
);
1919 if (ts
.tv_sec
== -1)
1921 if (ts
.tv_sec
== 0) {
1922 sk
->sk_stamp
= ktime_get_real();
1923 ts
= ktime_to_timespec(sk
->sk_stamp
);
1925 return copy_to_user(userstamp
, &ts
, sizeof(ts
)) ? -EFAULT
: 0;
1927 EXPORT_SYMBOL(sock_get_timestampns
);
1929 void sock_enable_timestamp(struct sock
*sk
, int flag
)
1931 if (!sock_flag(sk
, flag
)) {
1932 sock_set_flag(sk
, flag
);
1934 * we just set one of the two flags which require net
1935 * time stamping, but time stamping might have been on
1936 * already because of the other one
1939 flag
== SOCK_TIMESTAMP
?
1940 SOCK_TIMESTAMPING_RX_SOFTWARE
:
1942 net_enable_timestamp();
1947 * Get a socket option on an socket.
1949 * FIX: POSIX 1003.1g is very ambiguous here. It states that
1950 * asynchronous errors should be reported by getsockopt. We assume
1951 * this means if you specify SO_ERROR (otherwise whats the point of it).
1953 int sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1954 char __user
*optval
, int __user
*optlen
)
1956 struct sock
*sk
= sock
->sk
;
1958 return sk
->sk_prot
->getsockopt(sk
, level
, optname
, optval
, optlen
);
1960 EXPORT_SYMBOL(sock_common_getsockopt
);
1962 #ifdef CONFIG_COMPAT
1963 int compat_sock_common_getsockopt(struct socket
*sock
, int level
, int optname
,
1964 char __user
*optval
, int __user
*optlen
)
1966 struct sock
*sk
= sock
->sk
;
1968 if (sk
->sk_prot
->compat_getsockopt
!= NULL
)
1969 return sk
->sk_prot
->compat_getsockopt(sk
, level
, optname
,
1971 return sk
->sk_prot
->getsockopt(sk
, level
, optname
, optval
, optlen
);
1973 EXPORT_SYMBOL(compat_sock_common_getsockopt
);
1976 int sock_common_recvmsg(struct kiocb
*iocb
, struct socket
*sock
,
1977 struct msghdr
*msg
, size_t size
, int flags
)
1979 struct sock
*sk
= sock
->sk
;
1983 err
= sk
->sk_prot
->recvmsg(iocb
, sk
, msg
, size
, flags
& MSG_DONTWAIT
,
1984 flags
& ~MSG_DONTWAIT
, &addr_len
);
1986 msg
->msg_namelen
= addr_len
;
1989 EXPORT_SYMBOL(sock_common_recvmsg
);
1992 * Set socket options on an inet socket.
1994 int sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
1995 char __user
*optval
, int optlen
)
1997 struct sock
*sk
= sock
->sk
;
1999 return sk
->sk_prot
->setsockopt(sk
, level
, optname
, optval
, optlen
);
2001 EXPORT_SYMBOL(sock_common_setsockopt
);
2003 #ifdef CONFIG_COMPAT
2004 int compat_sock_common_setsockopt(struct socket
*sock
, int level
, int optname
,
2005 char __user
*optval
, int optlen
)
2007 struct sock
*sk
= sock
->sk
;
2009 if (sk
->sk_prot
->compat_setsockopt
!= NULL
)
2010 return sk
->sk_prot
->compat_setsockopt(sk
, level
, optname
,
2012 return sk
->sk_prot
->setsockopt(sk
, level
, optname
, optval
, optlen
);
2014 EXPORT_SYMBOL(compat_sock_common_setsockopt
);
2017 void sk_common_release(struct sock
*sk
)
2019 if (sk
->sk_prot
->destroy
)
2020 sk
->sk_prot
->destroy(sk
);
2023 * Observation: when sock_common_release is called, processes have
2024 * no access to socket. But net still has.
2025 * Step one, detach it from networking:
2027 * A. Remove from hash tables.
2030 sk
->sk_prot
->unhash(sk
);
2033 * In this point socket cannot receive new packets, but it is possible
2034 * that some packets are in flight because some CPU runs receiver and
2035 * did hash table lookup before we unhashed socket. They will achieve
2036 * receive queue and will be purged by socket destructor.
2038 * Also we still have packets pending on receive queue and probably,
2039 * our own packets waiting in device queues. sock_destroy will drain
2040 * receive queue, but transmitted packets will delay socket destruction
2041 * until the last reference will be released.
2046 xfrm_sk_free_policy(sk
);
2048 sk_refcnt_debug_release(sk
);
2051 EXPORT_SYMBOL(sk_common_release
);
2053 static DEFINE_RWLOCK(proto_list_lock
);
2054 static LIST_HEAD(proto_list
);
2056 #ifdef CONFIG_PROC_FS
2057 #define PROTO_INUSE_NR 64 /* should be enough for the first time */
2059 int val
[PROTO_INUSE_NR
];
2062 static DECLARE_BITMAP(proto_inuse_idx
, PROTO_INUSE_NR
);
2064 #ifdef CONFIG_NET_NS
2065 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int val
)
2067 int cpu
= smp_processor_id();
2068 per_cpu_ptr(net
->core
.inuse
, cpu
)->val
[prot
->inuse_idx
] += val
;
2070 EXPORT_SYMBOL_GPL(sock_prot_inuse_add
);
2072 int sock_prot_inuse_get(struct net
*net
, struct proto
*prot
)
2074 int cpu
, idx
= prot
->inuse_idx
;
2077 for_each_possible_cpu(cpu
)
2078 res
+= per_cpu_ptr(net
->core
.inuse
, cpu
)->val
[idx
];
2080 return res
>= 0 ? res
: 0;
2082 EXPORT_SYMBOL_GPL(sock_prot_inuse_get
);
2084 static int sock_inuse_init_net(struct net
*net
)
2086 net
->core
.inuse
= alloc_percpu(struct prot_inuse
);
2087 return net
->core
.inuse
? 0 : -ENOMEM
;
2090 static void sock_inuse_exit_net(struct net
*net
)
2092 free_percpu(net
->core
.inuse
);
2095 static struct pernet_operations net_inuse_ops
= {
2096 .init
= sock_inuse_init_net
,
2097 .exit
= sock_inuse_exit_net
,
2100 static __init
int net_inuse_init(void)
2102 if (register_pernet_subsys(&net_inuse_ops
))
2103 panic("Cannot initialize net inuse counters");
2108 core_initcall(net_inuse_init
);
2110 static DEFINE_PER_CPU(struct prot_inuse
, prot_inuse
);
2112 void sock_prot_inuse_add(struct net
*net
, struct proto
*prot
, int val
)
2114 __get_cpu_var(prot_inuse
).val
[prot
->inuse_idx
] += val
;
2116 EXPORT_SYMBOL_GPL(sock_prot_inuse_add
);
2118 int sock_prot_inuse_get(struct net
*net
, struct proto
*prot
)
2120 int cpu
, idx
= prot
->inuse_idx
;
2123 for_each_possible_cpu(cpu
)
2124 res
+= per_cpu(prot_inuse
, cpu
).val
[idx
];
2126 return res
>= 0 ? res
: 0;
2128 EXPORT_SYMBOL_GPL(sock_prot_inuse_get
);
2131 static void assign_proto_idx(struct proto
*prot
)
2133 prot
->inuse_idx
= find_first_zero_bit(proto_inuse_idx
, PROTO_INUSE_NR
);
2135 if (unlikely(prot
->inuse_idx
== PROTO_INUSE_NR
- 1)) {
2136 printk(KERN_ERR
"PROTO_INUSE_NR exhausted\n");
2140 set_bit(prot
->inuse_idx
, proto_inuse_idx
);
2143 static void release_proto_idx(struct proto
*prot
)
2145 if (prot
->inuse_idx
!= PROTO_INUSE_NR
- 1)
2146 clear_bit(prot
->inuse_idx
, proto_inuse_idx
);
2149 static inline void assign_proto_idx(struct proto
*prot
)
2153 static inline void release_proto_idx(struct proto
*prot
)
2158 int proto_register(struct proto
*prot
, int alloc_slab
)
2161 prot
->slab
= kmem_cache_create(prot
->name
, prot
->obj_size
, 0,
2162 SLAB_HWCACHE_ALIGN
| prot
->slab_flags
,
2165 if (prot
->slab
== NULL
) {
2166 printk(KERN_CRIT
"%s: Can't create sock SLAB cache!\n",
2171 if (prot
->rsk_prot
!= NULL
) {
2172 static const char mask
[] = "request_sock_%s";
2174 prot
->rsk_prot
->slab_name
= kmalloc(strlen(prot
->name
) + sizeof(mask
) - 1, GFP_KERNEL
);
2175 if (prot
->rsk_prot
->slab_name
== NULL
)
2176 goto out_free_sock_slab
;
2178 sprintf(prot
->rsk_prot
->slab_name
, mask
, prot
->name
);
2179 prot
->rsk_prot
->slab
= kmem_cache_create(prot
->rsk_prot
->slab_name
,
2180 prot
->rsk_prot
->obj_size
, 0,
2181 SLAB_HWCACHE_ALIGN
, NULL
);
2183 if (prot
->rsk_prot
->slab
== NULL
) {
2184 printk(KERN_CRIT
"%s: Can't create request sock SLAB cache!\n",
2186 goto out_free_request_sock_slab_name
;
2190 if (prot
->twsk_prot
!= NULL
) {
2191 static const char mask
[] = "tw_sock_%s";
2193 prot
->twsk_prot
->twsk_slab_name
= kmalloc(strlen(prot
->name
) + sizeof(mask
) - 1, GFP_KERNEL
);
2195 if (prot
->twsk_prot
->twsk_slab_name
== NULL
)
2196 goto out_free_request_sock_slab
;
2198 sprintf(prot
->twsk_prot
->twsk_slab_name
, mask
, prot
->name
);
2199 prot
->twsk_prot
->twsk_slab
=
2200 kmem_cache_create(prot
->twsk_prot
->twsk_slab_name
,
2201 prot
->twsk_prot
->twsk_obj_size
,
2203 SLAB_HWCACHE_ALIGN
|
2206 if (prot
->twsk_prot
->twsk_slab
== NULL
)
2207 goto out_free_timewait_sock_slab_name
;
2211 write_lock(&proto_list_lock
);
2212 list_add(&prot
->node
, &proto_list
);
2213 assign_proto_idx(prot
);
2214 write_unlock(&proto_list_lock
);
2217 out_free_timewait_sock_slab_name
:
2218 kfree(prot
->twsk_prot
->twsk_slab_name
);
2219 out_free_request_sock_slab
:
2220 if (prot
->rsk_prot
&& prot
->rsk_prot
->slab
) {
2221 kmem_cache_destroy(prot
->rsk_prot
->slab
);
2222 prot
->rsk_prot
->slab
= NULL
;
2224 out_free_request_sock_slab_name
:
2225 kfree(prot
->rsk_prot
->slab_name
);
2227 kmem_cache_destroy(prot
->slab
);
2232 EXPORT_SYMBOL(proto_register
);
2234 void proto_unregister(struct proto
*prot
)
2236 write_lock(&proto_list_lock
);
2237 release_proto_idx(prot
);
2238 list_del(&prot
->node
);
2239 write_unlock(&proto_list_lock
);
2241 if (prot
->slab
!= NULL
) {
2242 kmem_cache_destroy(prot
->slab
);
2246 if (prot
->rsk_prot
!= NULL
&& prot
->rsk_prot
->slab
!= NULL
) {
2247 kmem_cache_destroy(prot
->rsk_prot
->slab
);
2248 kfree(prot
->rsk_prot
->slab_name
);
2249 prot
->rsk_prot
->slab
= NULL
;
2252 if (prot
->twsk_prot
!= NULL
&& prot
->twsk_prot
->twsk_slab
!= NULL
) {
2253 kmem_cache_destroy(prot
->twsk_prot
->twsk_slab
);
2254 kfree(prot
->twsk_prot
->twsk_slab_name
);
2255 prot
->twsk_prot
->twsk_slab
= NULL
;
2258 EXPORT_SYMBOL(proto_unregister
);
2260 #ifdef CONFIG_PROC_FS
2261 static void *proto_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2262 __acquires(proto_list_lock
)
2264 read_lock(&proto_list_lock
);
2265 return seq_list_start_head(&proto_list
, *pos
);
2268 static void *proto_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2270 return seq_list_next(v
, &proto_list
, pos
);
2273 static void proto_seq_stop(struct seq_file
*seq
, void *v
)
2274 __releases(proto_list_lock
)
2276 read_unlock(&proto_list_lock
);
2279 static char proto_method_implemented(const void *method
)
2281 return method
== NULL
? 'n' : 'y';
2284 static void proto_seq_printf(struct seq_file
*seq
, struct proto
*proto
)
2286 seq_printf(seq
, "%-9s %4u %6d %6d %-3s %6u %-3s %-10s "
2287 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2290 sock_prot_inuse_get(seq_file_net(seq
), proto
),
2291 proto
->memory_allocated
!= NULL
? atomic_read(proto
->memory_allocated
) : -1,
2292 proto
->memory_pressure
!= NULL
? *proto
->memory_pressure
? "yes" : "no" : "NI",
2294 proto
->slab
== NULL
? "no" : "yes",
2295 module_name(proto
->owner
),
2296 proto_method_implemented(proto
->close
),
2297 proto_method_implemented(proto
->connect
),
2298 proto_method_implemented(proto
->disconnect
),
2299 proto_method_implemented(proto
->accept
),
2300 proto_method_implemented(proto
->ioctl
),
2301 proto_method_implemented(proto
->init
),
2302 proto_method_implemented(proto
->destroy
),
2303 proto_method_implemented(proto
->shutdown
),
2304 proto_method_implemented(proto
->setsockopt
),
2305 proto_method_implemented(proto
->getsockopt
),
2306 proto_method_implemented(proto
->sendmsg
),
2307 proto_method_implemented(proto
->recvmsg
),
2308 proto_method_implemented(proto
->sendpage
),
2309 proto_method_implemented(proto
->bind
),
2310 proto_method_implemented(proto
->backlog_rcv
),
2311 proto_method_implemented(proto
->hash
),
2312 proto_method_implemented(proto
->unhash
),
2313 proto_method_implemented(proto
->get_port
),
2314 proto_method_implemented(proto
->enter_memory_pressure
));
2317 static int proto_seq_show(struct seq_file
*seq
, void *v
)
2319 if (v
== &proto_list
)
2320 seq_printf(seq
, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2329 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2331 proto_seq_printf(seq
, list_entry(v
, struct proto
, node
));
2335 static const struct seq_operations proto_seq_ops
= {
2336 .start
= proto_seq_start
,
2337 .next
= proto_seq_next
,
2338 .stop
= proto_seq_stop
,
2339 .show
= proto_seq_show
,
2342 static int proto_seq_open(struct inode
*inode
, struct file
*file
)
2344 return seq_open_net(inode
, file
, &proto_seq_ops
,
2345 sizeof(struct seq_net_private
));
2348 static const struct file_operations proto_seq_fops
= {
2349 .owner
= THIS_MODULE
,
2350 .open
= proto_seq_open
,
2352 .llseek
= seq_lseek
,
2353 .release
= seq_release_net
,
2356 static __net_init
int proto_init_net(struct net
*net
)
2358 if (!proc_net_fops_create(net
, "protocols", S_IRUGO
, &proto_seq_fops
))
2364 static __net_exit
void proto_exit_net(struct net
*net
)
2366 proc_net_remove(net
, "protocols");
2370 static __net_initdata
struct pernet_operations proto_net_ops
= {
2371 .init
= proto_init_net
,
2372 .exit
= proto_exit_net
,
2375 static int __init
proto_init(void)
2377 return register_pernet_subsys(&proto_net_ops
);
2380 subsys_initcall(proto_init
);
2382 #endif /* PROC_FS */