blob: c5812bbc2cc91128d9797154d43f1cbc104ab979 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
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.
5 *
6 * Generic socket support routines. Memory allocators, socket lock/release
7 * handler for protocols to use and generic option handler.
8 *
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * 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>
14 *
15 * Fixes:
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
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +090035 * code. The ACK stuff can wait and needs major
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 * TCP layer surgery.
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
64 * (compatibility fix)
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
82 *
83 * To Fix:
84 *
85 *
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.
90 */
91
Randy Dunlap4fc268d2006-01-11 12:17:47 -080092#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070093#include <linux/errno.h>
94#include <linux/types.h>
95#include <linux/socket.h>
96#include <linux/in.h>
97#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070098#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>
Al Viroa1f8e7f72006-10-19 16:08:53 -0400112#include <linux/highmem.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
114#include <asm/uaccess.h>
115#include <asm/system.h>
116
117#include <linux/netdevice.h>
118#include <net/protocol.h>
119#include <linux/skbuff.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200120#include <net/net_namespace.h>
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700121#include <net/request_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122#include <net/sock.h>
Patrick Ohly20d49472009-02-12 05:03:38 +0000123#include <linux/net_tstamp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700124#include <net/xfrm.h>
125#include <linux/ipsec.h>
126
127#include <linux/filter.h>
128
129#ifdef CONFIG_INET
130#include <net/tcp.h>
131#endif
132
Ingo Molnarda21f242006-07-03 00:25:12 -0700133/*
134 * Each address family might have different locking rules, so we have
135 * one slock key per address family:
136 */
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700137static struct lock_class_key af_family_keys[AF_MAX];
138static struct lock_class_key af_family_slock_keys[AF_MAX];
139
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700140/*
141 * Make lock validator output more readable. (we pre-construct these
142 * strings build-time, so that runtime initialization of socket
143 * locks is fast):
144 */
Jan Engelhardt36cbd3d2009-08-05 10:42:58 -0700145static const char *const af_family_key_strings[AF_MAX+1] = {
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700146 "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" ,
Andy Grovercbd151b2009-02-26 23:43:19 -0800153 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700154 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
Oliver Hartkoppcd05acf2007-12-16 15:59:24 -0800155 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
David Howells17926a72007-04-26 15:48:28 -0700156 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700157 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
Sergey Lapinfcb94e42009-06-08 12:18:47 +0000158 "sk_lock-AF_IEEE802154",
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700159 "sk_lock-AF_MAX"
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700160};
Jan Engelhardt36cbd3d2009-08-05 10:42:58 -0700161static const char *const af_family_slock_key_strings[AF_MAX+1] = {
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700162 "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" ,
Andy Grovercbd151b2009-02-26 23:43:19 -0800169 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700170 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
Oliver Hartkoppcd05acf2007-12-16 15:59:24 -0800171 "slock-27" , "slock-28" , "slock-AF_CAN" ,
David Howells17926a72007-04-26 15:48:28 -0700172 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700173 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
Sergey Lapinfcb94e42009-06-08 12:18:47 +0000174 "slock-AF_IEEE802154",
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700175 "slock-AF_MAX"
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700176};
Jan Engelhardt36cbd3d2009-08-05 10:42:58 -0700177static const char *const af_family_clock_key_strings[AF_MAX+1] = {
Peter Zijlstra443aef02007-07-19 01:49:00 -0700178 "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" ,
Andy Grovercbd151b2009-02-26 23:43:19 -0800185 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
Peter Zijlstra443aef02007-07-19 01:49:00 -0700186 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
Oliver Hartkoppb4942af2008-07-23 14:06:04 -0700187 "clock-27" , "clock-28" , "clock-AF_CAN" ,
David Howellse51f8022007-07-21 19:30:16 -0700188 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700189 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
Sergey Lapinfcb94e42009-06-08 12:18:47 +0000190 "clock-AF_IEEE802154",
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700191 "clock-AF_MAX"
Peter Zijlstra443aef02007-07-19 01:49:00 -0700192};
Ingo Molnarda21f242006-07-03 00:25:12 -0700193
194/*
195 * sk_callback_lock locking rules are per-address-family,
196 * so split the lock classes by using a per-AF key:
197 */
198static struct lock_class_key af_callback_keys[AF_MAX];
199
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200/* 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.
204 */
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)
209
210/* Run time adjustable parameters. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700211__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;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
216/* Maximal space eaten by iovec or ancilliary data plus some space */
Brian Haleyab32ea52006-09-22 14:15:41 -0700217int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
Eric Dumazet2a915252009-05-27 11:30:05 +0000218EXPORT_SYMBOL(sysctl_optmem_max);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700219
220static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
221{
222 struct timeval tv;
223
224 if (optlen < sizeof(tv))
225 return -EINVAL;
226 if (copy_from_user(&tv, optval, sizeof(tv)))
227 return -EFAULT;
Vasily Averinba780732007-05-24 16:58:54 -0700228 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
229 return -EDOM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230
Vasily Averinba780732007-05-24 16:58:54 -0700231 if (tv.tv_sec < 0) {
Andrew Morton6f11df82007-07-09 13:16:00 -0700232 static int warned __read_mostly;
233
Vasily Averinba780732007-05-24 16:58:54 -0700234 *timeo_p = 0;
Ilpo Järvinen50aab542008-05-02 16:20:10 -0700235 if (warned < 10 && net_ratelimit()) {
Vasily Averinba780732007-05-24 16:58:54 -0700236 warned++;
237 printk(KERN_INFO "sock_set_timeout: `%s' (pid %d) "
238 "tries to set negative timeout\n",
Pavel Emelyanovba25f9d2007-10-18 23:40:40 -0700239 current->comm, task_pid_nr(current));
Ilpo Järvinen50aab542008-05-02 16:20:10 -0700240 }
Vasily Averinba780732007-05-24 16:58:54 -0700241 return 0;
242 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243 *timeo_p = MAX_SCHEDULE_TIMEOUT;
244 if (tv.tv_sec == 0 && tv.tv_usec == 0)
245 return 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);
248 return 0;
249}
250
251static void sock_warn_obsolete_bsdism(const char *name)
252{
253 static int warned;
254 static char warncomm[TASK_COMM_LEN];
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900255 if (strcmp(warncomm, current->comm) && warned < 5) {
256 strcpy(warncomm, current->comm);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 printk(KERN_WARNING "process `%s' is using obsolete "
258 "%s SO_BSDCOMPAT\n", warncomm, name);
259 warned++;
260 }
261}
262
Patrick Ohly20d49472009-02-12 05:03:38 +0000263static void sock_disable_timestamp(struct sock *sk, int flag)
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900264{
Patrick Ohly20d49472009-02-12 05:03:38 +0000265 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();
270 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700271 }
272}
273
274
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800275int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
276{
Eric Dumazet766e90372009-10-14 20:40:11 -0700277 int err;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800278 int skb_len;
Neil Horman3b885782009-10-12 13:26:31 -0700279 unsigned long flags;
280 struct sk_buff_head *list = &sk->sk_receive_queue;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800281
Rami Rosen9ee6b7f2008-05-14 03:50:03 -0700282 /* Cast sk->rcvbuf to unsigned... It's pointless, but reduces
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800283 number of warnings when compiling with -W --ANK
284 */
285 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
286 (unsigned)sk->sk_rcvbuf) {
Eric Dumazet766e90372009-10-14 20:40:11 -0700287 atomic_inc(&sk->sk_drops);
288 return -ENOMEM;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800289 }
290
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700291 err = sk_filter(sk, skb);
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800292 if (err)
Eric Dumazet766e90372009-10-14 20:40:11 -0700293 return err;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800294
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800295 if (!sk_rmem_schedule(sk, skb->truesize)) {
Eric Dumazet766e90372009-10-14 20:40:11 -0700296 atomic_inc(&sk->sk_drops);
297 return -ENOBUFS;
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800298 }
299
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800300 skb->dev = NULL;
301 skb_set_owner_r(skb, sk);
David S. Miller49ad9592008-12-17 22:11:38 -0800302
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800303 /* Cache the SKB length before we tack it onto the receive
304 * queue. Once it is added it no longer belongs to us and
305 * may be freed by other threads of control pulling packets
306 * from the queue.
307 */
308 skb_len = skb->len;
309
Neil Horman3b885782009-10-12 13:26:31 -0700310 spin_lock_irqsave(&list->lock, flags);
311 skb->dropcount = atomic_read(&sk->sk_drops);
312 __skb_queue_tail(list, skb);
313 spin_unlock_irqrestore(&list->lock, flags);
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800314
315 if (!sock_flag(sk, SOCK_DEAD))
316 sk->sk_data_ready(sk, skb_len);
Eric Dumazet766e90372009-10-14 20:40:11 -0700317 return 0;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800318}
319EXPORT_SYMBOL(sock_queue_rcv_skb);
320
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200321int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800322{
323 int rc = NET_RX_SUCCESS;
324
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700325 if (sk_filter(sk, skb))
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800326 goto discard_and_relse;
327
328 skb->dev = NULL;
329
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200330 if (nested)
331 bh_lock_sock_nested(sk);
332 else
333 bh_lock_sock(sk);
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700334 if (!sock_owned_by_user(sk)) {
335 /*
336 * trylock + unlock semantics:
337 */
338 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
339
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700340 rc = sk_backlog_rcv(sk, skb);
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700341
342 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
Zhu Yia3a858f2010-03-04 18:01:47 +0000343 } else if (sk_add_backlog(sk, skb)) {
Zhu Yi8eae9392010-03-04 18:01:40 +0000344 bh_unlock_sock(sk);
345 atomic_inc(&sk->sk_drops);
346 goto discard_and_relse;
347 }
348
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800349 bh_unlock_sock(sk);
350out:
351 sock_put(sk);
352 return rc;
353discard_and_relse:
354 kfree_skb(skb);
355 goto out;
356}
357EXPORT_SYMBOL(sk_receive_skb);
358
Krishna Kumarea94ff32009-10-19 23:46:45 +0000359void sk_reset_txq(struct sock *sk)
360{
361 sk_tx_queue_clear(sk);
362}
363EXPORT_SYMBOL(sk_reset_txq);
364
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800365struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
366{
367 struct dst_entry *dst = sk->sk_dst_cache;
368
369 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
Krishna Kumare022f0b2009-10-19 23:46:20 +0000370 sk_tx_queue_clear(sk);
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800371 sk->sk_dst_cache = NULL;
372 dst_release(dst);
373 return NULL;
374 }
375
376 return dst;
377}
378EXPORT_SYMBOL(__sk_dst_check);
379
380struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
381{
382 struct dst_entry *dst = sk_dst_get(sk);
383
384 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
385 sk_dst_reset(sk);
386 dst_release(dst);
387 return NULL;
388 }
389
390 return dst;
391}
392EXPORT_SYMBOL(sk_dst_check);
393
David S. Miller48788092007-09-14 16:41:03 -0700394static int sock_bindtodevice(struct sock *sk, char __user *optval, int optlen)
395{
396 int ret = -ENOPROTOOPT;
397#ifdef CONFIG_NETDEVICES
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900398 struct net *net = sock_net(sk);
David S. Miller48788092007-09-14 16:41:03 -0700399 char devname[IFNAMSIZ];
400 int index;
401
402 /* Sorry... */
403 ret = -EPERM;
404 if (!capable(CAP_NET_RAW))
405 goto out;
406
407 ret = -EINVAL;
408 if (optlen < 0)
409 goto out;
410
411 /* Bind this socket to a particular device like "eth0",
412 * as specified in the passed interface name. If the
413 * name is "" or the option length is zero the socket
414 * is not bound.
415 */
416 if (optlen > IFNAMSIZ - 1)
417 optlen = IFNAMSIZ - 1;
418 memset(devname, 0, sizeof(devname));
419
420 ret = -EFAULT;
421 if (copy_from_user(devname, optval, optlen))
422 goto out;
423
David S. Miller000ba2e2009-11-05 22:37:11 -0800424 index = 0;
425 if (devname[0] != '\0') {
Eric Dumazetbf8e56b2009-11-05 21:03:39 -0800426 struct net_device *dev;
David S. Miller48788092007-09-14 16:41:03 -0700427
Eric Dumazetbf8e56b2009-11-05 21:03:39 -0800428 rcu_read_lock();
429 dev = dev_get_by_name_rcu(net, devname);
430 if (dev)
431 index = dev->ifindex;
432 rcu_read_unlock();
David S. Miller48788092007-09-14 16:41:03 -0700433 ret = -ENODEV;
434 if (!dev)
435 goto out;
David S. Miller48788092007-09-14 16:41:03 -0700436 }
437
438 lock_sock(sk);
439 sk->sk_bound_dev_if = index;
440 sk_dst_reset(sk);
441 release_sock(sk);
442
443 ret = 0;
444
445out:
446#endif
447
448 return ret;
449}
450
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800451static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
452{
453 if (valbool)
454 sock_set_flag(sk, bit);
455 else
456 sock_reset_flag(sk, bit);
457}
458
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459/*
460 * This is meant for all protocols to use and covers goings on
461 * at the socket level. Everything here is generic.
462 */
463
464int sock_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -0700465 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466{
Eric Dumazet2a915252009-05-27 11:30:05 +0000467 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468 int val;
469 int valbool;
470 struct linger ling;
471 int ret = 0;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 /*
474 * Options without arguments
475 */
476
David S. Miller48788092007-09-14 16:41:03 -0700477 if (optname == SO_BINDTODEVICE)
478 return sock_bindtodevice(sk, optval, optlen);
479
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700480 if (optlen < sizeof(int))
481 return -EINVAL;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900482
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 if (get_user(val, (int __user *)optval))
484 return -EFAULT;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900485
Eric Dumazet2a915252009-05-27 11:30:05 +0000486 valbool = val ? 1 : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
488 lock_sock(sk);
489
Eric Dumazet2a915252009-05-27 11:30:05 +0000490 switch (optname) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700491 case SO_DEBUG:
Eric Dumazet2a915252009-05-27 11:30:05 +0000492 if (val && !capable(CAP_NET_ADMIN))
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700493 ret = -EACCES;
Eric Dumazet2a915252009-05-27 11:30:05 +0000494 else
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800495 sock_valbool_flag(sk, SOCK_DBG, valbool);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700496 break;
497 case SO_REUSEADDR:
498 sk->sk_reuse = valbool;
499 break;
500 case SO_TYPE:
Jan Engelhardt49c794e2009-08-04 07:28:28 +0000501 case SO_PROTOCOL:
Jan Engelhardt0d6038e2009-08-04 07:28:29 +0000502 case SO_DOMAIN:
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700503 case SO_ERROR:
504 ret = -ENOPROTOOPT;
505 break;
506 case SO_DONTROUTE:
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800507 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700508 break;
509 case SO_BROADCAST:
510 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
511 break;
512 case SO_SNDBUF:
513 /* Don't error on this BSD doesn't and if you think
514 about it this is right. Otherwise apps have to
515 play 'guess the biggest size' games. RCVBUF/SNDBUF
516 are treated in BSD as hints */
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900517
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700518 if (val > sysctl_wmem_max)
519 val = sysctl_wmem_max;
Patrick McHardyb0573de2005-08-09 19:30:51 -0700520set_sndbuf:
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700521 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
522 if ((val * 2) < SOCK_MIN_SNDBUF)
523 sk->sk_sndbuf = SOCK_MIN_SNDBUF;
524 else
525 sk->sk_sndbuf = val * 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700527 /*
528 * Wake up sending tasks if we
529 * upped the value.
530 */
531 sk->sk_write_space(sk);
532 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700534 case SO_SNDBUFFORCE:
535 if (!capable(CAP_NET_ADMIN)) {
536 ret = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 break;
538 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700539 goto set_sndbuf;
540
541 case SO_RCVBUF:
542 /* Don't error on this BSD doesn't and if you think
543 about it this is right. Otherwise apps have to
544 play 'guess the biggest size' games. RCVBUF/SNDBUF
545 are treated in BSD as hints */
546
547 if (val > sysctl_rmem_max)
548 val = sysctl_rmem_max;
549set_rcvbuf:
550 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
551 /*
552 * We double it on the way in to account for
553 * "struct sk_buff" etc. overhead. Applications
554 * assume that the SO_RCVBUF setting they make will
555 * allow that much actual data to be received on that
556 * socket.
557 *
558 * Applications are unaware that "struct sk_buff" and
559 * other overheads allocate from the receive buffer
560 * during socket buffer allocation.
561 *
562 * And after considering the possible alternatives,
563 * returning the value we actually used in getsockopt
564 * is the most desirable behavior.
565 */
566 if ((val * 2) < SOCK_MIN_RCVBUF)
567 sk->sk_rcvbuf = SOCK_MIN_RCVBUF;
568 else
569 sk->sk_rcvbuf = val * 2;
570 break;
571
572 case SO_RCVBUFFORCE:
573 if (!capable(CAP_NET_ADMIN)) {
574 ret = -EPERM;
575 break;
576 }
577 goto set_rcvbuf;
578
579 case SO_KEEPALIVE:
580#ifdef CONFIG_INET
581 if (sk->sk_protocol == IPPROTO_TCP)
582 tcp_set_keepalive(sk, valbool);
583#endif
584 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
585 break;
586
587 case SO_OOBINLINE:
588 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
589 break;
590
591 case SO_NO_CHECK:
592 sk->sk_no_check = valbool;
593 break;
594
595 case SO_PRIORITY:
596 if ((val >= 0 && val <= 6) || capable(CAP_NET_ADMIN))
597 sk->sk_priority = val;
598 else
599 ret = -EPERM;
600 break;
601
602 case SO_LINGER:
603 if (optlen < sizeof(ling)) {
604 ret = -EINVAL; /* 1003.1g */
605 break;
606 }
Eric Dumazet2a915252009-05-27 11:30:05 +0000607 if (copy_from_user(&ling, optval, sizeof(ling))) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700608 ret = -EFAULT;
609 break;
610 }
611 if (!ling.l_onoff)
612 sock_reset_flag(sk, SOCK_LINGER);
613 else {
614#if (BITS_PER_LONG == 32)
615 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
616 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
617 else
618#endif
619 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
620 sock_set_flag(sk, SOCK_LINGER);
621 }
622 break;
623
624 case SO_BSDCOMPAT:
625 sock_warn_obsolete_bsdism("setsockopt");
626 break;
627
628 case SO_PASSCRED:
629 if (valbool)
630 set_bit(SOCK_PASSCRED, &sock->flags);
631 else
632 clear_bit(SOCK_PASSCRED, &sock->flags);
633 break;
634
635 case SO_TIMESTAMP:
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700636 case SO_TIMESTAMPNS:
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700637 if (valbool) {
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700638 if (optname == SO_TIMESTAMP)
639 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
640 else
641 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700642 sock_set_flag(sk, SOCK_RCVTSTAMP);
Patrick Ohly20d49472009-02-12 05:03:38 +0000643 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700644 } else {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700645 sock_reset_flag(sk, SOCK_RCVTSTAMP);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700646 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
647 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700648 break;
649
Patrick Ohly20d49472009-02-12 05:03:38 +0000650 case SO_TIMESTAMPING:
651 if (val & ~SOF_TIMESTAMPING_MASK) {
Rémi Denis-Courmontf249fb72009-07-20 00:47:04 +0000652 ret = -EINVAL;
Patrick Ohly20d49472009-02-12 05:03:38 +0000653 break;
654 }
655 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
656 val & SOF_TIMESTAMPING_TX_HARDWARE);
657 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
658 val & SOF_TIMESTAMPING_TX_SOFTWARE);
659 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
660 val & SOF_TIMESTAMPING_RX_HARDWARE);
661 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
662 sock_enable_timestamp(sk,
663 SOCK_TIMESTAMPING_RX_SOFTWARE);
664 else
665 sock_disable_timestamp(sk,
666 SOCK_TIMESTAMPING_RX_SOFTWARE);
667 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
668 val & SOF_TIMESTAMPING_SOFTWARE);
669 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
670 val & SOF_TIMESTAMPING_SYS_HARDWARE);
671 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
672 val & SOF_TIMESTAMPING_RAW_HARDWARE);
673 break;
674
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700675 case SO_RCVLOWAT:
676 if (val < 0)
677 val = INT_MAX;
678 sk->sk_rcvlowat = val ? : 1;
679 break;
680
681 case SO_RCVTIMEO:
682 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
683 break;
684
685 case SO_SNDTIMEO:
686 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
687 break;
688
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700689 case SO_ATTACH_FILTER:
690 ret = -EINVAL;
691 if (optlen == sizeof(struct sock_fprog)) {
692 struct sock_fprog fprog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700694 ret = -EFAULT;
695 if (copy_from_user(&fprog, optval, sizeof(fprog)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700698 ret = sk_attach_filter(&fprog, sk);
699 }
700 break;
701
702 case SO_DETACH_FILTER:
Pavel Emelyanov55b33322007-10-17 21:21:26 -0700703 ret = sk_detach_filter(sk);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700704 break;
705
706 case SO_PASSSEC:
707 if (valbool)
708 set_bit(SOCK_PASSSEC, &sock->flags);
709 else
710 clear_bit(SOCK_PASSSEC, &sock->flags);
711 break;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800712 case SO_MARK:
713 if (!capable(CAP_NET_ADMIN))
714 ret = -EPERM;
Eric Dumazet2a915252009-05-27 11:30:05 +0000715 else
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800716 sk->sk_mark = val;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800717 break;
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700718
Linus Torvalds1da177e2005-04-16 15:20:36 -0700719 /* We implement the SO_SNDLOWAT etc to
720 not be settable (1003.1g 5.3) */
Neil Horman3b885782009-10-12 13:26:31 -0700721 case SO_RXQ_OVFL:
722 if (valbool)
723 sock_set_flag(sk, SOCK_RXQ_OVFL);
724 else
725 sock_reset_flag(sk, SOCK_RXQ_OVFL);
726 break;
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700727 default:
728 ret = -ENOPROTOOPT;
729 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900730 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 release_sock(sk);
732 return ret;
733}
Eric Dumazet2a915252009-05-27 11:30:05 +0000734EXPORT_SYMBOL(sock_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735
736
737int sock_getsockopt(struct socket *sock, int level, int optname,
738 char __user *optval, int __user *optlen)
739{
740 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900741
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700742 union {
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900743 int val;
744 struct linger ling;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700745 struct timeval tm;
746 } v;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900747
H Hartley Sweeten4d0392b2010-01-15 01:08:58 -0800748 int lv = sizeof(int);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749 int len;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900750
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700751 if (get_user(len, optlen))
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900752 return -EFAULT;
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700753 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754 return -EINVAL;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900755
Eugene Teo50fee1d2009-02-23 15:38:41 -0800756 memset(&v, 0, sizeof(v));
Clément Lecignedf0bca02009-02-12 16:59:09 -0800757
Eric Dumazet2a915252009-05-27 11:30:05 +0000758 switch (optname) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700759 case SO_DEBUG:
760 v.val = sock_flag(sk, SOCK_DBG);
761 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900762
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700763 case SO_DONTROUTE:
764 v.val = sock_flag(sk, SOCK_LOCALROUTE);
765 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900766
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700767 case SO_BROADCAST:
768 v.val = !!sock_flag(sk, SOCK_BROADCAST);
769 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700771 case SO_SNDBUF:
772 v.val = sk->sk_sndbuf;
773 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900774
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700775 case SO_RCVBUF:
776 v.val = sk->sk_rcvbuf;
777 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700778
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700779 case SO_REUSEADDR:
780 v.val = sk->sk_reuse;
781 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700783 case SO_KEEPALIVE:
784 v.val = !!sock_flag(sk, SOCK_KEEPOPEN);
785 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700787 case SO_TYPE:
788 v.val = sk->sk_type;
789 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790
Jan Engelhardt49c794e2009-08-04 07:28:28 +0000791 case SO_PROTOCOL:
792 v.val = sk->sk_protocol;
793 break;
794
Jan Engelhardt0d6038e2009-08-04 07:28:29 +0000795 case SO_DOMAIN:
796 v.val = sk->sk_family;
797 break;
798
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700799 case SO_ERROR:
800 v.val = -sock_error(sk);
Eric Dumazet2a915252009-05-27 11:30:05 +0000801 if (v.val == 0)
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700802 v.val = xchg(&sk->sk_err_soft, 0);
803 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700804
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700805 case SO_OOBINLINE:
806 v.val = !!sock_flag(sk, SOCK_URGINLINE);
807 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900808
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700809 case SO_NO_CHECK:
810 v.val = sk->sk_no_check;
811 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700813 case SO_PRIORITY:
814 v.val = sk->sk_priority;
815 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900816
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700817 case SO_LINGER:
818 lv = sizeof(v.ling);
819 v.ling.l_onoff = !!sock_flag(sk, SOCK_LINGER);
820 v.ling.l_linger = sk->sk_lingertime / HZ;
821 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900822
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700823 case SO_BSDCOMPAT:
824 sock_warn_obsolete_bsdism("getsockopt");
825 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700827 case SO_TIMESTAMP:
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700828 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
829 !sock_flag(sk, SOCK_RCVTSTAMPNS);
830 break;
831
832 case SO_TIMESTAMPNS:
833 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700834 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835
Patrick Ohly20d49472009-02-12 05:03:38 +0000836 case SO_TIMESTAMPING:
837 v.val = 0;
838 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
839 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
840 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
841 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
842 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
843 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
844 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
845 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
846 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
847 v.val |= SOF_TIMESTAMPING_SOFTWARE;
848 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
849 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
850 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
851 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
852 break;
853
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700854 case SO_RCVTIMEO:
Eric Dumazet2a915252009-05-27 11:30:05 +0000855 lv = sizeof(struct timeval);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700856 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
857 v.tm.tv_sec = 0;
858 v.tm.tv_usec = 0;
859 } else {
860 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
861 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700863 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700865 case SO_SNDTIMEO:
Eric Dumazet2a915252009-05-27 11:30:05 +0000866 lv = sizeof(struct timeval);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700867 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
868 v.tm.tv_sec = 0;
869 v.tm.tv_usec = 0;
870 } else {
871 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
872 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
873 }
874 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700876 case SO_RCVLOWAT:
877 v.val = sk->sk_rcvlowat;
878 break;
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700879
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700880 case SO_SNDLOWAT:
Eric Dumazet2a915252009-05-27 11:30:05 +0000881 v.val = 1;
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700882 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700884 case SO_PASSCRED:
885 v.val = test_bit(SOCK_PASSCRED, &sock->flags) ? 1 : 0;
886 break;
887
888 case SO_PEERCRED:
889 if (len > sizeof(sk->sk_peercred))
890 len = sizeof(sk->sk_peercred);
891 if (copy_to_user(optval, &sk->sk_peercred, len))
892 return -EFAULT;
893 goto lenout;
894
895 case SO_PEERNAME:
896 {
897 char address[128];
898
899 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
900 return -ENOTCONN;
901 if (lv < len)
902 return -EINVAL;
903 if (copy_to_user(optval, address, len))
904 return -EFAULT;
905 goto lenout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700907
908 /* Dubious BSD thing... Probably nobody even uses it, but
909 * the UNIX standard wants it for whatever reason... -DaveM
910 */
911 case SO_ACCEPTCONN:
912 v.val = sk->sk_state == TCP_LISTEN;
913 break;
914
915 case SO_PASSSEC:
916 v.val = test_bit(SOCK_PASSSEC, &sock->flags) ? 1 : 0;
917 break;
918
919 case SO_PEERSEC:
920 return security_socket_getpeersec_stream(sock, optval, optlen, len);
921
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800922 case SO_MARK:
923 v.val = sk->sk_mark;
924 break;
925
Neil Horman3b885782009-10-12 13:26:31 -0700926 case SO_RXQ_OVFL:
927 v.val = !!sock_flag(sk, SOCK_RXQ_OVFL);
928 break;
929
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700930 default:
931 return -ENOPROTOOPT;
932 }
933
Linus Torvalds1da177e2005-04-16 15:20:36 -0700934 if (len > lv)
935 len = lv;
936 if (copy_to_user(optval, &v, len))
937 return -EFAULT;
938lenout:
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900939 if (put_user(len, optlen))
940 return -EFAULT;
941 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942}
943
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700944/*
945 * Initialize an sk_lock.
946 *
947 * (We also register the sk_lock with the lock validator.)
948 */
Dave Jonesb6f99a22007-03-22 12:27:49 -0700949static inline void sock_lock_init(struct sock *sk)
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700950{
Peter Zijlstraed075362006-12-06 20:35:24 -0800951 sock_lock_init_class_and_name(sk,
952 af_family_slock_key_strings[sk->sk_family],
953 af_family_slock_keys + sk->sk_family,
954 af_family_key_strings[sk->sk_family],
955 af_family_keys + sk->sk_family);
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700956}
957
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000958/*
959 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
960 * even temporarly, because of RCU lookups. sk_node should also be left as is.
961 */
Pavel Emelyanovf1a6c4d2007-11-01 00:29:45 -0700962static void sock_copy(struct sock *nsk, const struct sock *osk)
963{
964#ifdef CONFIG_SECURITY_NETWORK
965 void *sptr = nsk->sk_security;
966#endif
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000967 BUILD_BUG_ON(offsetof(struct sock, sk_copy_start) !=
Krishna Kumare022f0b2009-10-19 23:46:20 +0000968 sizeof(osk->sk_node) + sizeof(osk->sk_refcnt) +
969 sizeof(osk->sk_tx_queue_mapping));
Eric Dumazet4dc6dc72009-07-15 23:13:10 +0000970 memcpy(&nsk->sk_copy_start, &osk->sk_copy_start,
971 osk->sk_prot->obj_size - offsetof(struct sock, sk_copy_start));
Pavel Emelyanovf1a6c4d2007-11-01 00:29:45 -0700972#ifdef CONFIG_SECURITY_NETWORK
973 nsk->sk_security = sptr;
974 security_sk_clone(osk, nsk);
975#endif
976}
977
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -0700978static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
979 int family)
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -0700980{
981 struct sock *sk;
982 struct kmem_cache *slab;
983
984 slab = prot->slab;
Eric Dumazete912b112009-07-08 19:36:05 +0000985 if (slab != NULL) {
986 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
987 if (!sk)
988 return sk;
989 if (priority & __GFP_ZERO) {
990 /*
991 * caches using SLAB_DESTROY_BY_RCU should let
992 * sk_node.next un-modified. Special care is taken
993 * when initializing object to zero.
994 */
995 if (offsetof(struct sock, sk_node.next) != 0)
996 memset(sk, 0, offsetof(struct sock, sk_node.next));
997 memset(&sk->sk_node.pprev, 0,
998 prot->obj_size - offsetof(struct sock,
999 sk_node.pprev));
1000 }
1001 }
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001002 else
1003 sk = kmalloc(prot->obj_size, priority);
1004
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001005 if (sk != NULL) {
Vegard Nossuma98b65a2009-02-26 14:46:57 +01001006 kmemcheck_annotate_bitfield(sk, flags);
1007
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001008 if (security_sk_alloc(sk, family, priority))
1009 goto out_free;
1010
1011 if (!try_module_get(prot->owner))
1012 goto out_free_sec;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001013 sk_tx_queue_clear(sk);
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001014 }
1015
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001016 return sk;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001017
1018out_free_sec:
1019 security_sk_free(sk);
1020out_free:
1021 if (slab != NULL)
1022 kmem_cache_free(slab, sk);
1023 else
1024 kfree(sk);
1025 return NULL;
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001026}
1027
1028static void sk_prot_free(struct proto *prot, struct sock *sk)
1029{
1030 struct kmem_cache *slab;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001031 struct module *owner;
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001032
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001033 owner = prot->owner;
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001034 slab = prot->slab;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001035
1036 security_sk_free(sk);
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001037 if (slab != NULL)
1038 kmem_cache_free(slab, sk);
1039 else
1040 kfree(sk);
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001041 module_put(owner);
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001042}
1043
Linus Torvalds1da177e2005-04-16 15:20:36 -07001044/**
1045 * sk_alloc - All socket objects are allocated here
Randy Dunlapc4ea43c2007-10-12 21:17:49 -07001046 * @net: the applicable net namespace
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001047 * @family: protocol family
1048 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1049 * @prot: struct proto associated with this new sock instance
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050 */
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001051struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -07001052 struct proto *prot)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001053{
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001054 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055
Pavel Emelyanov154adbc2007-11-01 00:38:43 -07001056 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057 if (sk) {
Pavel Emelyanov154adbc2007-11-01 00:38:43 -07001058 sk->sk_family = family;
1059 /*
1060 * See comment in struct sock definition to understand
1061 * why we need sk_prot_creator -acme
1062 */
1063 sk->sk_prot = sk->sk_prot_creator = prot;
1064 sock_lock_init(sk);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001065 sock_net_set(sk, get_net(net));
Jarek Poplawskid66ee052009-08-30 23:15:36 +00001066 atomic_set(&sk->sk_wmem_alloc, 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067 }
Frank Filza79af592005-09-27 15:23:38 -07001068
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001069 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070}
Eric Dumazet2a915252009-05-27 11:30:05 +00001071EXPORT_SYMBOL(sk_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072
Eric Dumazet2b85a342009-06-11 02:55:43 -07001073static void __sk_free(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074{
1075 struct sk_filter *filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001076
1077 if (sk->sk_destruct)
1078 sk->sk_destruct(sk);
1079
Paul E. McKenneya898def2010-02-22 17:04:49 -08001080 filter = rcu_dereference_check(sk->sk_filter,
1081 atomic_read(&sk->sk_wmem_alloc) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 if (filter) {
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001083 sk_filter_uncharge(sk, filter);
Dmitry Mishinfda9ef52006-08-31 15:28:39 -07001084 rcu_assign_pointer(sk->sk_filter, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085 }
1086
Patrick Ohly20d49472009-02-12 05:03:38 +00001087 sock_disable_timestamp(sk, SOCK_TIMESTAMP);
1088 sock_disable_timestamp(sk, SOCK_TIMESTAMPING_RX_SOFTWARE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089
1090 if (atomic_read(&sk->sk_omem_alloc))
1091 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08001092 __func__, atomic_read(&sk->sk_omem_alloc));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001094 put_net(sock_net(sk));
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001095 sk_prot_free(sk->sk_prot_creator, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096}
Eric Dumazet2b85a342009-06-11 02:55:43 -07001097
1098void sk_free(struct sock *sk)
1099{
1100 /*
1101 * We substract one from sk_wmem_alloc and can know if
1102 * some packets are still in some tx queue.
1103 * If not null, sock_wfree() will call __sk_free(sk) later
1104 */
1105 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1106 __sk_free(sk);
1107}
Eric Dumazet2a915252009-05-27 11:30:05 +00001108EXPORT_SYMBOL(sk_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109
Denis V. Lunevedf02082008-02-29 11:18:32 -08001110/*
1111 * Last sock_put should drop referrence to sk->sk_net. It has already
1112 * been dropped in sk_change_net. Taking referrence to stopping namespace
1113 * is not an option.
1114 * Take referrence to a socket to remove it from hash _alive_ and after that
1115 * destroy it in the context of init_net.
1116 */
1117void sk_release_kernel(struct sock *sk)
1118{
1119 if (sk == NULL || sk->sk_socket == NULL)
1120 return;
1121
1122 sock_hold(sk);
1123 sock_release(sk->sk_socket);
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07001124 release_net(sock_net(sk));
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001125 sock_net_set(sk, get_net(&init_net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08001126 sock_put(sk);
1127}
David S. Miller45af1752008-02-29 11:33:19 -08001128EXPORT_SYMBOL(sk_release_kernel);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001129
Al Virodd0fc662005-10-07 07:46:04 +01001130struct sock *sk_clone(const struct sock *sk, const gfp_t priority)
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001131{
Pavel Emelyanov8fd1d172007-11-01 00:37:32 -07001132 struct sock *newsk;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001133
Pavel Emelyanov8fd1d172007-11-01 00:37:32 -07001134 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001135 if (newsk != NULL) {
1136 struct sk_filter *filter;
1137
Venkat Yekkirala892c1412006-08-04 23:08:56 -07001138 sock_copy(newsk, sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001139
1140 /* SANITY */
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001141 get_net(sock_net(newsk));
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001142 sk_node_init(&newsk->sk_node);
1143 sock_lock_init(newsk);
1144 bh_lock_sock(newsk);
Eric Dumazetfa438cc2007-03-04 16:05:44 -08001145 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
Zhu Yi8eae9392010-03-04 18:01:40 +00001146 newsk->sk_backlog.len = 0;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001147
1148 atomic_set(&newsk->sk_rmem_alloc, 0);
Eric Dumazet2b85a342009-06-11 02:55:43 -07001149 /*
1150 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1151 */
1152 atomic_set(&newsk->sk_wmem_alloc, 1);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001153 atomic_set(&newsk->sk_omem_alloc, 0);
1154 skb_queue_head_init(&newsk->sk_receive_queue);
1155 skb_queue_head_init(&newsk->sk_write_queue);
Chris Leech97fc2f02006-05-23 17:55:33 -07001156#ifdef CONFIG_NET_DMA
1157 skb_queue_head_init(&newsk->sk_async_wait_queue);
1158#endif
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001159
1160 rwlock_init(&newsk->sk_dst_lock);
1161 rwlock_init(&newsk->sk_callback_lock);
Peter Zijlstra443aef02007-07-19 01:49:00 -07001162 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1163 af_callback_keys + newsk->sk_family,
1164 af_family_clock_key_strings[newsk->sk_family]);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001165
1166 newsk->sk_dst_cache = NULL;
1167 newsk->sk_wmem_queued = 0;
1168 newsk->sk_forward_alloc = 0;
1169 newsk->sk_send_head = NULL;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001170 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1171
1172 sock_reset_flag(newsk, SOCK_DONE);
1173 skb_queue_head_init(&newsk->sk_error_queue);
1174
1175 filter = newsk->sk_filter;
1176 if (filter != NULL)
1177 sk_filter_charge(newsk, filter);
1178
1179 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1180 /* It is still raw copy of parent, so invalidate
1181 * destructor and make plain sk_free() */
1182 newsk->sk_destruct = NULL;
1183 sk_free(newsk);
1184 newsk = NULL;
1185 goto out;
1186 }
1187
1188 newsk->sk_err = 0;
1189 newsk->sk_priority = 0;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +00001190 /*
1191 * Before updating sk_refcnt, we must commit prior changes to memory
1192 * (Documentation/RCU/rculist_nulls.txt for details)
1193 */
1194 smp_wmb();
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001195 atomic_set(&newsk->sk_refcnt, 2);
1196
1197 /*
1198 * Increment the counter in the same struct proto as the master
1199 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1200 * is the same as sk->sk_prot->socks, as this field was copied
1201 * with memcpy).
1202 *
1203 * This _changes_ the previous behaviour, where
1204 * tcp_create_openreq_child always was incrementing the
1205 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1206 * to be taken into account in all callers. -acme
1207 */
1208 sk_refcnt_debug_inc(newsk);
David S. Miller972692e2008-06-17 22:41:38 -07001209 sk_set_socket(newsk, NULL);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001210 newsk->sk_sleep = NULL;
1211
1212 if (newsk->sk_prot->sockets_allocated)
Eric Dumazet17483762008-11-25 21:16:35 -08001213 percpu_counter_inc(newsk->sk_prot->sockets_allocated);
Octavian Purdila704da5602010-01-08 00:00:09 -08001214
1215 if (sock_flag(newsk, SOCK_TIMESTAMP) ||
1216 sock_flag(newsk, SOCK_TIMESTAMPING_RX_SOFTWARE))
1217 net_enable_timestamp();
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001218 }
1219out:
1220 return newsk;
1221}
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001222EXPORT_SYMBOL_GPL(sk_clone);
1223
Andi Kleen99580892007-04-20 17:12:43 -07001224void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1225{
1226 __sk_dst_set(sk, dst);
1227 sk->sk_route_caps = dst->dev->features;
1228 if (sk->sk_route_caps & NETIF_F_GSO)
Herbert Xu4fcd6b92007-05-31 22:15:50 -07001229 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
Andi Kleen99580892007-04-20 17:12:43 -07001230 if (sk_can_gso(sk)) {
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001231 if (dst->header_len) {
Andi Kleen99580892007-04-20 17:12:43 -07001232 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001233 } else {
Andi Kleen99580892007-04-20 17:12:43 -07001234 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001235 sk->sk_gso_max_size = dst->dev->gso_max_size;
1236 }
Andi Kleen99580892007-04-20 17:12:43 -07001237 }
1238}
1239EXPORT_SYMBOL_GPL(sk_setup_caps);
1240
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241void __init sk_init(void)
1242{
Jan Beulich44813742009-09-21 17:03:05 -07001243 if (totalram_pages <= 4096) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244 sysctl_wmem_max = 32767;
1245 sysctl_rmem_max = 32767;
1246 sysctl_wmem_default = 32767;
1247 sysctl_rmem_default = 32767;
Jan Beulich44813742009-09-21 17:03:05 -07001248 } else if (totalram_pages >= 131072) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249 sysctl_wmem_max = 131071;
1250 sysctl_rmem_max = 131071;
1251 }
1252}
1253
1254/*
1255 * Simple resource managers for sockets.
1256 */
1257
1258
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001259/*
1260 * Write buffer destructor automatically called from kfree_skb.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 */
1262void sock_wfree(struct sk_buff *skb)
1263{
1264 struct sock *sk = skb->sk;
Eric Dumazetd99927f2009-09-24 10:49:24 +00001265 unsigned int len = skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266
Eric Dumazetd99927f2009-09-24 10:49:24 +00001267 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1268 /*
1269 * Keep a reference on sk_wmem_alloc, this will be released
1270 * after sk_write_space() call
1271 */
1272 atomic_sub(len - 1, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273 sk->sk_write_space(sk);
Eric Dumazetd99927f2009-09-24 10:49:24 +00001274 len = 1;
1275 }
Eric Dumazet2b85a342009-06-11 02:55:43 -07001276 /*
Eric Dumazetd99927f2009-09-24 10:49:24 +00001277 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1278 * could not do because of in-flight packets
Eric Dumazet2b85a342009-06-11 02:55:43 -07001279 */
Eric Dumazetd99927f2009-09-24 10:49:24 +00001280 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
Eric Dumazet2b85a342009-06-11 02:55:43 -07001281 __sk_free(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282}
Eric Dumazet2a915252009-05-27 11:30:05 +00001283EXPORT_SYMBOL(sock_wfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001285/*
1286 * Read buffer destructor automatically called from kfree_skb.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 */
1288void sock_rfree(struct sk_buff *skb)
1289{
1290 struct sock *sk = skb->sk;
1291
1292 atomic_sub(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001293 sk_mem_uncharge(skb->sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294}
Eric Dumazet2a915252009-05-27 11:30:05 +00001295EXPORT_SYMBOL(sock_rfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296
1297
1298int sock_i_uid(struct sock *sk)
1299{
1300 int uid;
1301
1302 read_lock(&sk->sk_callback_lock);
1303 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
1304 read_unlock(&sk->sk_callback_lock);
1305 return uid;
1306}
Eric Dumazet2a915252009-05-27 11:30:05 +00001307EXPORT_SYMBOL(sock_i_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308
1309unsigned long sock_i_ino(struct sock *sk)
1310{
1311 unsigned long ino;
1312
1313 read_lock(&sk->sk_callback_lock);
1314 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
1315 read_unlock(&sk->sk_callback_lock);
1316 return ino;
1317}
Eric Dumazet2a915252009-05-27 11:30:05 +00001318EXPORT_SYMBOL(sock_i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319
1320/*
1321 * Allocate a skb from the socket's send buffer.
1322 */
Victor Fusco86a76ca2005-07-08 14:57:47 -07001323struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001324 gfp_t priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325{
1326 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
Eric Dumazet2a915252009-05-27 11:30:05 +00001327 struct sk_buff *skb = alloc_skb(size, priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328 if (skb) {
1329 skb_set_owner_w(skb, sk);
1330 return skb;
1331 }
1332 }
1333 return NULL;
1334}
Eric Dumazet2a915252009-05-27 11:30:05 +00001335EXPORT_SYMBOL(sock_wmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336
1337/*
1338 * Allocate a skb from the socket's receive buffer.
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001339 */
Victor Fusco86a76ca2005-07-08 14:57:47 -07001340struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001341 gfp_t priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342{
1343 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1344 struct sk_buff *skb = alloc_skb(size, priority);
1345 if (skb) {
1346 skb_set_owner_r(skb, sk);
1347 return skb;
1348 }
1349 }
1350 return NULL;
1351}
1352
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001353/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 * Allocate a memory block from the socket's option memory buffer.
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001355 */
Al Virodd0fc662005-10-07 07:46:04 +01001356void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357{
1358 if ((unsigned)size <= sysctl_optmem_max &&
1359 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1360 void *mem;
1361 /* First do the add, to avoid the race if kmalloc
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001362 * might sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 */
1364 atomic_add(size, &sk->sk_omem_alloc);
1365 mem = kmalloc(size, priority);
1366 if (mem)
1367 return mem;
1368 atomic_sub(size, &sk->sk_omem_alloc);
1369 }
1370 return NULL;
1371}
Eric Dumazet2a915252009-05-27 11:30:05 +00001372EXPORT_SYMBOL(sock_kmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373
1374/*
1375 * Free an option memory block.
1376 */
1377void sock_kfree_s(struct sock *sk, void *mem, int size)
1378{
1379 kfree(mem);
1380 atomic_sub(size, &sk->sk_omem_alloc);
1381}
Eric Dumazet2a915252009-05-27 11:30:05 +00001382EXPORT_SYMBOL(sock_kfree_s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383
1384/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1385 I think, these locks should be removed for datagram sockets.
1386 */
Eric Dumazet2a915252009-05-27 11:30:05 +00001387static long sock_wait_for_wmem(struct sock *sk, long timeo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388{
1389 DEFINE_WAIT(wait);
1390
1391 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1392 for (;;) {
1393 if (!timeo)
1394 break;
1395 if (signal_pending(current))
1396 break;
1397 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1398 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1399 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1400 break;
1401 if (sk->sk_shutdown & SEND_SHUTDOWN)
1402 break;
1403 if (sk->sk_err)
1404 break;
1405 timeo = schedule_timeout(timeo);
1406 }
1407 finish_wait(sk->sk_sleep, &wait);
1408 return timeo;
1409}
1410
1411
1412/*
1413 * Generic send/receive buffer handlers
1414 */
1415
Herbert Xu4cc7f682009-02-04 16:55:54 -08001416struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1417 unsigned long data_len, int noblock,
1418 int *errcode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419{
1420 struct sk_buff *skb;
Al Viro7d877f32005-10-21 03:20:43 -04001421 gfp_t gfp_mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 long timeo;
1423 int err;
1424
1425 gfp_mask = sk->sk_allocation;
1426 if (gfp_mask & __GFP_WAIT)
1427 gfp_mask |= __GFP_REPEAT;
1428
1429 timeo = sock_sndtimeo(sk, noblock);
1430 while (1) {
1431 err = sock_error(sk);
1432 if (err != 0)
1433 goto failure;
1434
1435 err = -EPIPE;
1436 if (sk->sk_shutdown & SEND_SHUTDOWN)
1437 goto failure;
1438
1439 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
Larry Woodmandb38c1792006-11-03 16:05:45 -08001440 skb = alloc_skb(header_len, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 if (skb) {
1442 int npages;
1443 int i;
1444
1445 /* No pages, we're done... */
1446 if (!data_len)
1447 break;
1448
1449 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1450 skb->truesize += data_len;
1451 skb_shinfo(skb)->nr_frags = npages;
1452 for (i = 0; i < npages; i++) {
1453 struct page *page;
1454 skb_frag_t *frag;
1455
1456 page = alloc_pages(sk->sk_allocation, 0);
1457 if (!page) {
1458 err = -ENOBUFS;
1459 skb_shinfo(skb)->nr_frags = i;
1460 kfree_skb(skb);
1461 goto failure;
1462 }
1463
1464 frag = &skb_shinfo(skb)->frags[i];
1465 frag->page = page;
1466 frag->page_offset = 0;
1467 frag->size = (data_len >= PAGE_SIZE ?
1468 PAGE_SIZE :
1469 data_len);
1470 data_len -= PAGE_SIZE;
1471 }
1472
1473 /* Full success... */
1474 break;
1475 }
1476 err = -ENOBUFS;
1477 goto failure;
1478 }
1479 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1480 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1481 err = -EAGAIN;
1482 if (!timeo)
1483 goto failure;
1484 if (signal_pending(current))
1485 goto interrupted;
1486 timeo = sock_wait_for_wmem(sk, timeo);
1487 }
1488
1489 skb_set_owner_w(skb, sk);
1490 return skb;
1491
1492interrupted:
1493 err = sock_intr_errno(timeo);
1494failure:
1495 *errcode = err;
1496 return NULL;
1497}
Herbert Xu4cc7f682009-02-04 16:55:54 -08001498EXPORT_SYMBOL(sock_alloc_send_pskb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001500struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501 int noblock, int *errcode)
1502{
1503 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1504}
Eric Dumazet2a915252009-05-27 11:30:05 +00001505EXPORT_SYMBOL(sock_alloc_send_skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506
1507static void __lock_sock(struct sock *sk)
1508{
1509 DEFINE_WAIT(wait);
1510
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001511 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1513 TASK_UNINTERRUPTIBLE);
1514 spin_unlock_bh(&sk->sk_lock.slock);
1515 schedule();
1516 spin_lock_bh(&sk->sk_lock.slock);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001517 if (!sock_owned_by_user(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 break;
1519 }
1520 finish_wait(&sk->sk_lock.wq, &wait);
1521}
1522
1523static void __release_sock(struct sock *sk)
1524{
1525 struct sk_buff *skb = sk->sk_backlog.head;
1526
1527 do {
1528 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1529 bh_unlock_sock(sk);
1530
1531 do {
1532 struct sk_buff *next = skb->next;
1533
1534 skb->next = NULL;
Peter Zijlstrac57943a2008-10-07 14:18:42 -07001535 sk_backlog_rcv(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536
1537 /*
1538 * We are in process context here with softirqs
1539 * disabled, use cond_resched_softirq() to preempt.
1540 * This is safe to do because we've taken the backlog
1541 * queue private:
1542 */
1543 cond_resched_softirq();
1544
1545 skb = next;
1546 } while (skb != NULL);
1547
1548 bh_lock_sock(sk);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001549 } while ((skb = sk->sk_backlog.head) != NULL);
Zhu Yi8eae9392010-03-04 18:01:40 +00001550
1551 /*
1552 * Doing the zeroing here guarantee we can not loop forever
1553 * while a wild producer attempts to flood us.
1554 */
1555 sk->sk_backlog.len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556}
1557
1558/**
1559 * sk_wait_data - wait for data to arrive at sk_receive_queue
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001560 * @sk: sock to wait on
1561 * @timeo: for how long
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 *
1563 * Now socket state including sk->sk_err is changed only under lock,
1564 * hence we may omit checks after joining wait queue.
1565 * We check receive queue before schedule() only as optimization;
1566 * it is very likely that release_sock() added new data.
1567 */
1568int sk_wait_data(struct sock *sk, long *timeo)
1569{
1570 int rc;
1571 DEFINE_WAIT(wait);
1572
1573 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1574 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1575 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1576 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1577 finish_wait(sk->sk_sleep, &wait);
1578 return rc;
1579}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580EXPORT_SYMBOL(sk_wait_data);
1581
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001582/**
1583 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1584 * @sk: socket
1585 * @size: memory size to allocate
1586 * @kind: allocation type
1587 *
1588 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1589 * rmem allocation. This function assumes that protocols which have
1590 * memory_pressure use sk_wmem_queued as write buffer accounting.
1591 */
1592int __sk_mem_schedule(struct sock *sk, int size, int kind)
1593{
1594 struct proto *prot = sk->sk_prot;
1595 int amt = sk_mem_pages(size);
1596 int allocated;
1597
1598 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
1599 allocated = atomic_add_return(amt, prot->memory_allocated);
1600
1601 /* Under limit. */
1602 if (allocated <= prot->sysctl_mem[0]) {
1603 if (prot->memory_pressure && *prot->memory_pressure)
1604 *prot->memory_pressure = 0;
1605 return 1;
1606 }
1607
1608 /* Under pressure. */
1609 if (allocated > prot->sysctl_mem[1])
1610 if (prot->enter_memory_pressure)
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001611 prot->enter_memory_pressure(sk);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001612
1613 /* Over hard limit. */
1614 if (allocated > prot->sysctl_mem[2])
1615 goto suppress_allocation;
1616
1617 /* guarantee minimum buffer size under pressure */
1618 if (kind == SK_MEM_RECV) {
1619 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1620 return 1;
1621 } else { /* SK_MEM_SEND */
1622 if (sk->sk_type == SOCK_STREAM) {
1623 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1624 return 1;
1625 } else if (atomic_read(&sk->sk_wmem_alloc) <
1626 prot->sysctl_wmem[0])
1627 return 1;
1628 }
1629
1630 if (prot->memory_pressure) {
Eric Dumazet17483762008-11-25 21:16:35 -08001631 int alloc;
1632
1633 if (!*prot->memory_pressure)
1634 return 1;
1635 alloc = percpu_counter_read_positive(prot->sockets_allocated);
1636 if (prot->sysctl_mem[2] > alloc *
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001637 sk_mem_pages(sk->sk_wmem_queued +
1638 atomic_read(&sk->sk_rmem_alloc) +
1639 sk->sk_forward_alloc))
1640 return 1;
1641 }
1642
1643suppress_allocation:
1644
1645 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1646 sk_stream_moderate_sndbuf(sk);
1647
1648 /* Fail only if socket is _under_ its sndbuf.
1649 * In this case we cannot block, so that we have to fail.
1650 */
1651 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1652 return 1;
1653 }
1654
1655 /* Alas. Undo changes. */
1656 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
1657 atomic_sub(amt, prot->memory_allocated);
1658 return 0;
1659}
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001660EXPORT_SYMBOL(__sk_mem_schedule);
1661
1662/**
1663 * __sk_reclaim - reclaim memory_allocated
1664 * @sk: socket
1665 */
1666void __sk_mem_reclaim(struct sock *sk)
1667{
1668 struct proto *prot = sk->sk_prot;
1669
Eric Dumazet680a5a52007-12-31 15:00:50 -08001670 atomic_sub(sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT,
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001671 prot->memory_allocated);
1672 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
1673
1674 if (prot->memory_pressure && *prot->memory_pressure &&
1675 (atomic_read(prot->memory_allocated) < prot->sysctl_mem[0]))
1676 *prot->memory_pressure = 0;
1677}
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001678EXPORT_SYMBOL(__sk_mem_reclaim);
1679
1680
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681/*
1682 * Set of default routines for initialising struct proto_ops when
1683 * the protocol does not support a particular function. In certain
1684 * cases where it makes no sense for a protocol to have a "do nothing"
1685 * function, some default processing is provided.
1686 */
1687
1688int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1689{
1690 return -EOPNOTSUPP;
1691}
Eric Dumazet2a915252009-05-27 11:30:05 +00001692EXPORT_SYMBOL(sock_no_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001694int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695 int len, int flags)
1696{
1697 return -EOPNOTSUPP;
1698}
Eric Dumazet2a915252009-05-27 11:30:05 +00001699EXPORT_SYMBOL(sock_no_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700
1701int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1702{
1703 return -EOPNOTSUPP;
1704}
Eric Dumazet2a915252009-05-27 11:30:05 +00001705EXPORT_SYMBOL(sock_no_socketpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706
1707int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1708{
1709 return -EOPNOTSUPP;
1710}
Eric Dumazet2a915252009-05-27 11:30:05 +00001711EXPORT_SYMBOL(sock_no_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001713int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 int *len, int peer)
1715{
1716 return -EOPNOTSUPP;
1717}
Eric Dumazet2a915252009-05-27 11:30:05 +00001718EXPORT_SYMBOL(sock_no_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719
Eric Dumazet2a915252009-05-27 11:30:05 +00001720unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721{
1722 return 0;
1723}
Eric Dumazet2a915252009-05-27 11:30:05 +00001724EXPORT_SYMBOL(sock_no_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725
1726int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1727{
1728 return -EOPNOTSUPP;
1729}
Eric Dumazet2a915252009-05-27 11:30:05 +00001730EXPORT_SYMBOL(sock_no_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
1732int sock_no_listen(struct socket *sock, int backlog)
1733{
1734 return -EOPNOTSUPP;
1735}
Eric Dumazet2a915252009-05-27 11:30:05 +00001736EXPORT_SYMBOL(sock_no_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737
1738int sock_no_shutdown(struct socket *sock, int how)
1739{
1740 return -EOPNOTSUPP;
1741}
Eric Dumazet2a915252009-05-27 11:30:05 +00001742EXPORT_SYMBOL(sock_no_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743
1744int sock_no_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07001745 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746{
1747 return -EOPNOTSUPP;
1748}
Eric Dumazet2a915252009-05-27 11:30:05 +00001749EXPORT_SYMBOL(sock_no_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750
1751int sock_no_getsockopt(struct socket *sock, int level, int optname,
1752 char __user *optval, int __user *optlen)
1753{
1754 return -EOPNOTSUPP;
1755}
Eric Dumazet2a915252009-05-27 11:30:05 +00001756EXPORT_SYMBOL(sock_no_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757
1758int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1759 size_t len)
1760{
1761 return -EOPNOTSUPP;
1762}
Eric Dumazet2a915252009-05-27 11:30:05 +00001763EXPORT_SYMBOL(sock_no_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764
1765int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1766 size_t len, int flags)
1767{
1768 return -EOPNOTSUPP;
1769}
Eric Dumazet2a915252009-05-27 11:30:05 +00001770EXPORT_SYMBOL(sock_no_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771
1772int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
1773{
1774 /* Mirror missing mmap method error code */
1775 return -ENODEV;
1776}
Eric Dumazet2a915252009-05-27 11:30:05 +00001777EXPORT_SYMBOL(sock_no_mmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001778
1779ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
1780{
1781 ssize_t res;
1782 struct msghdr msg = {.msg_flags = flags};
1783 struct kvec iov;
1784 char *kaddr = kmap(page);
1785 iov.iov_base = kaddr + offset;
1786 iov.iov_len = size;
1787 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
1788 kunmap(page);
1789 return res;
1790}
Eric Dumazet2a915252009-05-27 11:30:05 +00001791EXPORT_SYMBOL(sock_no_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792
1793/*
1794 * Default Socket Callbacks
1795 */
1796
1797static void sock_def_wakeup(struct sock *sk)
1798{
1799 read_lock(&sk->sk_callback_lock);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001800 if (sk_has_sleeper(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 wake_up_interruptible_all(sk->sk_sleep);
1802 read_unlock(&sk->sk_callback_lock);
1803}
1804
1805static void sock_def_error_report(struct sock *sk)
1806{
1807 read_lock(&sk->sk_callback_lock);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001808 if (sk_has_sleeper(sk))
Davide Libenzi37e55402009-03-31 15:24:21 -07001809 wake_up_interruptible_poll(sk->sk_sleep, POLLERR);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001810 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811 read_unlock(&sk->sk_callback_lock);
1812}
1813
1814static void sock_def_readable(struct sock *sk, int len)
1815{
1816 read_lock(&sk->sk_callback_lock);
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001817 if (sk_has_sleeper(sk))
Davide Libenzi37e55402009-03-31 15:24:21 -07001818 wake_up_interruptible_sync_poll(sk->sk_sleep, POLLIN |
1819 POLLRDNORM | POLLRDBAND);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001820 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 read_unlock(&sk->sk_callback_lock);
1822}
1823
1824static void sock_def_write_space(struct sock *sk)
1825{
1826 read_lock(&sk->sk_callback_lock);
1827
1828 /* Do not wake up a writer until he can make "significant"
1829 * progress. --DaveM
1830 */
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001831 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
Jiri Olsaa57de0b2009-07-08 12:09:13 +00001832 if (sk_has_sleeper(sk))
Davide Libenzi37e55402009-03-31 15:24:21 -07001833 wake_up_interruptible_sync_poll(sk->sk_sleep, POLLOUT |
1834 POLLWRNORM | POLLWRBAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001835
1836 /* Should agree with poll, otherwise some programs break */
1837 if (sock_writeable(sk))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001838 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839 }
1840
1841 read_unlock(&sk->sk_callback_lock);
1842}
1843
1844static void sock_def_destruct(struct sock *sk)
1845{
Jesper Juhla51482b2005-11-08 09:41:34 -08001846 kfree(sk->sk_protinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847}
1848
1849void sk_send_sigurg(struct sock *sk)
1850{
1851 if (sk->sk_socket && sk->sk_socket->file)
1852 if (send_sigurg(&sk->sk_socket->file->f_owner))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001853 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854}
Eric Dumazet2a915252009-05-27 11:30:05 +00001855EXPORT_SYMBOL(sk_send_sigurg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856
1857void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1858 unsigned long expires)
1859{
1860 if (!mod_timer(timer, expires))
1861 sock_hold(sk);
1862}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863EXPORT_SYMBOL(sk_reset_timer);
1864
1865void sk_stop_timer(struct sock *sk, struct timer_list* timer)
1866{
1867 if (timer_pending(timer) && del_timer(timer))
1868 __sock_put(sk);
1869}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870EXPORT_SYMBOL(sk_stop_timer);
1871
1872void sock_init_data(struct socket *sock, struct sock *sk)
1873{
1874 skb_queue_head_init(&sk->sk_receive_queue);
1875 skb_queue_head_init(&sk->sk_write_queue);
1876 skb_queue_head_init(&sk->sk_error_queue);
Chris Leech97fc2f02006-05-23 17:55:33 -07001877#ifdef CONFIG_NET_DMA
1878 skb_queue_head_init(&sk->sk_async_wait_queue);
1879#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001880
1881 sk->sk_send_head = NULL;
1882
1883 init_timer(&sk->sk_timer);
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001884
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885 sk->sk_allocation = GFP_KERNEL;
1886 sk->sk_rcvbuf = sysctl_rmem_default;
1887 sk->sk_sndbuf = sysctl_wmem_default;
Zhu Yi8eae9392010-03-04 18:01:40 +00001888 sk->sk_backlog.limit = sk->sk_rcvbuf << 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889 sk->sk_state = TCP_CLOSE;
David S. Miller972692e2008-06-17 22:41:38 -07001890 sk_set_socket(sk, sock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001891
1892 sock_set_flag(sk, SOCK_ZAPPED);
1893
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001894 if (sock) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895 sk->sk_type = sock->type;
1896 sk->sk_sleep = &sock->wait;
1897 sock->sk = sk;
1898 } else
1899 sk->sk_sleep = NULL;
1900
1901 rwlock_init(&sk->sk_dst_lock);
1902 rwlock_init(&sk->sk_callback_lock);
Peter Zijlstra443aef02007-07-19 01:49:00 -07001903 lockdep_set_class_and_name(&sk->sk_callback_lock,
1904 af_callback_keys + sk->sk_family,
1905 af_family_clock_key_strings[sk->sk_family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906
1907 sk->sk_state_change = sock_def_wakeup;
1908 sk->sk_data_ready = sock_def_readable;
1909 sk->sk_write_space = sock_def_write_space;
1910 sk->sk_error_report = sock_def_error_report;
1911 sk->sk_destruct = sock_def_destruct;
1912
1913 sk->sk_sndmsg_page = NULL;
1914 sk->sk_sndmsg_off = 0;
1915
1916 sk->sk_peercred.pid = 0;
1917 sk->sk_peercred.uid = -1;
1918 sk->sk_peercred.gid = -1;
1919 sk->sk_write_pending = 0;
1920 sk->sk_rcvlowat = 1;
1921 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
1922 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
1923
Eric Dumazetf37f0af2008-04-13 21:39:26 -07001924 sk->sk_stamp = ktime_set(-1L, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925
Eric Dumazet4dc6dc72009-07-15 23:13:10 +00001926 /*
1927 * Before updating sk_refcnt, we must commit prior changes to memory
1928 * (Documentation/RCU/rculist_nulls.txt for details)
1929 */
1930 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931 atomic_set(&sk->sk_refcnt, 1);
Wang Chen33c732c2007-11-13 20:30:01 -08001932 atomic_set(&sk->sk_drops, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933}
Eric Dumazet2a915252009-05-27 11:30:05 +00001934EXPORT_SYMBOL(sock_init_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935
Harvey Harrisonb5606c22008-02-13 15:03:16 -08001936void lock_sock_nested(struct sock *sk, int subclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937{
1938 might_sleep();
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001939 spin_lock_bh(&sk->sk_lock.slock);
John Heffnerd2e91172007-09-12 10:44:19 +02001940 if (sk->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001941 __lock_sock(sk);
John Heffnerd2e91172007-09-12 10:44:19 +02001942 sk->sk_lock.owned = 1;
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001943 spin_unlock(&sk->sk_lock.slock);
1944 /*
1945 * The sk_lock has mutex_lock() semantics here:
1946 */
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001947 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001948 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949}
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001950EXPORT_SYMBOL(lock_sock_nested);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951
Harvey Harrisonb5606c22008-02-13 15:03:16 -08001952void release_sock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001953{
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001954 /*
1955 * The sk_lock has mutex_unlock() semantics:
1956 */
1957 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1958
1959 spin_lock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 if (sk->sk_backlog.tail)
1961 __release_sock(sk);
John Heffnerd2e91172007-09-12 10:44:19 +02001962 sk->sk_lock.owned = 0;
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001963 if (waitqueue_active(&sk->sk_lock.wq))
1964 wake_up(&sk->sk_lock.wq);
1965 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966}
1967EXPORT_SYMBOL(release_sock);
1968
1969int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001970{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001971 struct timeval tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 if (!sock_flag(sk, SOCK_TIMESTAMP))
Patrick Ohly20d49472009-02-12 05:03:38 +00001973 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001974 tv = ktime_to_timeval(sk->sk_stamp);
1975 if (tv.tv_sec == -1)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 return -ENOENT;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001977 if (tv.tv_sec == 0) {
1978 sk->sk_stamp = ktime_get_real();
1979 tv = ktime_to_timeval(sk->sk_stamp);
1980 }
1981 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001982}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983EXPORT_SYMBOL(sock_get_timestamp);
1984
Eric Dumazetae40eb12007-03-18 17:33:16 -07001985int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
1986{
1987 struct timespec ts;
1988 if (!sock_flag(sk, SOCK_TIMESTAMP))
Patrick Ohly20d49472009-02-12 05:03:38 +00001989 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazetae40eb12007-03-18 17:33:16 -07001990 ts = ktime_to_timespec(sk->sk_stamp);
1991 if (ts.tv_sec == -1)
1992 return -ENOENT;
1993 if (ts.tv_sec == 0) {
1994 sk->sk_stamp = ktime_get_real();
1995 ts = ktime_to_timespec(sk->sk_stamp);
1996 }
1997 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
1998}
1999EXPORT_SYMBOL(sock_get_timestampns);
2000
Patrick Ohly20d49472009-02-12 05:03:38 +00002001void sock_enable_timestamp(struct sock *sk, int flag)
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09002002{
Patrick Ohly20d49472009-02-12 05:03:38 +00002003 if (!sock_flag(sk, flag)) {
2004 sock_set_flag(sk, flag);
2005 /*
2006 * we just set one of the two flags which require net
2007 * time stamping, but time stamping might have been on
2008 * already because of the other one
2009 */
2010 if (!sock_flag(sk,
2011 flag == SOCK_TIMESTAMP ?
2012 SOCK_TIMESTAMPING_RX_SOFTWARE :
2013 SOCK_TIMESTAMP))
2014 net_enable_timestamp();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 }
2016}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017
2018/*
2019 * Get a socket option on an socket.
2020 *
2021 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2022 * asynchronous errors should be reported by getsockopt. We assume
2023 * this means if you specify SO_ERROR (otherwise whats the point of it).
2024 */
2025int sock_common_getsockopt(struct socket *sock, int level, int optname,
2026 char __user *optval, int __user *optlen)
2027{
2028 struct sock *sk = sock->sk;
2029
2030 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2031}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032EXPORT_SYMBOL(sock_common_getsockopt);
2033
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002034#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002035int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2036 char __user *optval, int __user *optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002037{
2038 struct sock *sk = sock->sk;
2039
Johannes Berg1e51f952007-03-06 13:44:06 -08002040 if (sk->sk_prot->compat_getsockopt != NULL)
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002041 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2042 optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002043 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2044}
2045EXPORT_SYMBOL(compat_sock_common_getsockopt);
2046#endif
2047
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2049 struct msghdr *msg, size_t size, int flags)
2050{
2051 struct sock *sk = sock->sk;
2052 int addr_len = 0;
2053 int err;
2054
2055 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2056 flags & ~MSG_DONTWAIT, &addr_len);
2057 if (err >= 0)
2058 msg->msg_namelen = addr_len;
2059 return err;
2060}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061EXPORT_SYMBOL(sock_common_recvmsg);
2062
2063/*
2064 * Set socket options on an inet socket.
2065 */
2066int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002067 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002068{
2069 struct sock *sk = sock->sk;
2070
2071 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2072}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073EXPORT_SYMBOL(sock_common_setsockopt);
2074
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002075#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002076int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002077 char __user *optval, unsigned int optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002078{
2079 struct sock *sk = sock->sk;
2080
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002081 if (sk->sk_prot->compat_setsockopt != NULL)
2082 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2083 optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002084 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2085}
2086EXPORT_SYMBOL(compat_sock_common_setsockopt);
2087#endif
2088
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089void sk_common_release(struct sock *sk)
2090{
2091 if (sk->sk_prot->destroy)
2092 sk->sk_prot->destroy(sk);
2093
2094 /*
2095 * Observation: when sock_common_release is called, processes have
2096 * no access to socket. But net still has.
2097 * Step one, detach it from networking:
2098 *
2099 * A. Remove from hash tables.
2100 */
2101
2102 sk->sk_prot->unhash(sk);
2103
2104 /*
2105 * In this point socket cannot receive new packets, but it is possible
2106 * that some packets are in flight because some CPU runs receiver and
2107 * did hash table lookup before we unhashed socket. They will achieve
2108 * receive queue and will be purged by socket destructor.
2109 *
2110 * Also we still have packets pending on receive queue and probably,
2111 * our own packets waiting in device queues. sock_destroy will drain
2112 * receive queue, but transmitted packets will delay socket destruction
2113 * until the last reference will be released.
2114 */
2115
2116 sock_orphan(sk);
2117
2118 xfrm_sk_free_policy(sk);
2119
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07002120 sk_refcnt_debug_release(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121 sock_put(sk);
2122}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123EXPORT_SYMBOL(sk_common_release);
2124
2125static DEFINE_RWLOCK(proto_list_lock);
2126static LIST_HEAD(proto_list);
2127
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002128#ifdef CONFIG_PROC_FS
2129#define PROTO_INUSE_NR 64 /* should be enough for the first time */
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002130struct prot_inuse {
2131 int val[PROTO_INUSE_NR];
2132};
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002133
2134static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002135
2136#ifdef CONFIG_NET_NS
2137void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2138{
2139 int cpu = smp_processor_id();
2140 per_cpu_ptr(net->core.inuse, cpu)->val[prot->inuse_idx] += val;
2141}
2142EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2143
2144int sock_prot_inuse_get(struct net *net, struct proto *prot)
2145{
2146 int cpu, idx = prot->inuse_idx;
2147 int res = 0;
2148
2149 for_each_possible_cpu(cpu)
2150 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2151
2152 return res >= 0 ? res : 0;
2153}
2154EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2155
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002156static int __net_init sock_inuse_init_net(struct net *net)
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002157{
2158 net->core.inuse = alloc_percpu(struct prot_inuse);
2159 return net->core.inuse ? 0 : -ENOMEM;
2160}
2161
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002162static void __net_exit sock_inuse_exit_net(struct net *net)
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002163{
2164 free_percpu(net->core.inuse);
2165}
2166
2167static struct pernet_operations net_inuse_ops = {
2168 .init = sock_inuse_init_net,
2169 .exit = sock_inuse_exit_net,
2170};
2171
2172static __init int net_inuse_init(void)
2173{
2174 if (register_pernet_subsys(&net_inuse_ops))
2175 panic("Cannot initialize net inuse counters");
2176
2177 return 0;
2178}
2179
2180core_initcall(net_inuse_init);
2181#else
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002182static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2183
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07002184void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002185{
2186 __get_cpu_var(prot_inuse).val[prot->inuse_idx] += val;
2187}
2188EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2189
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07002190int sock_prot_inuse_get(struct net *net, struct proto *prot)
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002191{
2192 int cpu, idx = prot->inuse_idx;
2193 int res = 0;
2194
2195 for_each_possible_cpu(cpu)
2196 res += per_cpu(prot_inuse, cpu).val[idx];
2197
2198 return res >= 0 ? res : 0;
2199}
2200EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002201#endif
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002202
2203static void assign_proto_idx(struct proto *prot)
2204{
2205 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2206
2207 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
2208 printk(KERN_ERR "PROTO_INUSE_NR exhausted\n");
2209 return;
2210 }
2211
2212 set_bit(prot->inuse_idx, proto_inuse_idx);
2213}
2214
2215static void release_proto_idx(struct proto *prot)
2216{
2217 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2218 clear_bit(prot->inuse_idx, proto_inuse_idx);
2219}
2220#else
2221static inline void assign_proto_idx(struct proto *prot)
2222{
2223}
2224
2225static inline void release_proto_idx(struct proto *prot)
2226{
2227}
2228#endif
2229
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230int proto_register(struct proto *prot, int alloc_slab)
2231{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232 if (alloc_slab) {
2233 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
Eric Dumazet271b72c2008-10-29 02:11:14 -07002234 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2235 NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236
2237 if (prot->slab == NULL) {
2238 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n",
2239 prot->name);
Pavel Emelyanov60e76632008-03-28 16:39:10 -07002240 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241 }
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002242
2243 if (prot->rsk_prot != NULL) {
Alexey Dobriyanfaf23422010-02-17 09:34:12 +00002244 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002245 if (prot->rsk_prot->slab_name == NULL)
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002246 goto out_free_sock_slab;
2247
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002248 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002249 prot->rsk_prot->obj_size, 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09002250 SLAB_HWCACHE_ALIGN, NULL);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002251
2252 if (prot->rsk_prot->slab == NULL) {
2253 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n",
2254 prot->name);
2255 goto out_free_request_sock_slab_name;
2256 }
2257 }
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002258
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002259 if (prot->twsk_prot != NULL) {
Alexey Dobriyanfaf23422010-02-17 09:34:12 +00002260 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002261
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002262 if (prot->twsk_prot->twsk_slab_name == NULL)
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002263 goto out_free_request_sock_slab;
2264
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002265 prot->twsk_prot->twsk_slab =
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002266 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002267 prot->twsk_prot->twsk_obj_size,
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002268 0,
2269 SLAB_HWCACHE_ALIGN |
2270 prot->slab_flags,
Paul Mundt20c2df82007-07-20 10:11:58 +09002271 NULL);
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002272 if (prot->twsk_prot->twsk_slab == NULL)
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002273 goto out_free_timewait_sock_slab_name;
2274 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275 }
2276
Arnaldo Carvalho de Melo2a278052005-04-16 15:24:09 -07002277 write_lock(&proto_list_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278 list_add(&prot->node, &proto_list);
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002279 assign_proto_idx(prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280 write_unlock(&proto_list_lock);
Pavel Emelyanovb733c002007-11-07 02:23:38 -08002281 return 0;
2282
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002283out_free_timewait_sock_slab_name:
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002284 kfree(prot->twsk_prot->twsk_slab_name);
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002285out_free_request_sock_slab:
2286 if (prot->rsk_prot && prot->rsk_prot->slab) {
2287 kmem_cache_destroy(prot->rsk_prot->slab);
2288 prot->rsk_prot->slab = NULL;
2289 }
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002290out_free_request_sock_slab_name:
Dan Carpenter72150e92010-03-06 01:04:45 +00002291 if (prot->rsk_prot)
2292 kfree(prot->rsk_prot->slab_name);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002293out_free_sock_slab:
2294 kmem_cache_destroy(prot->slab);
2295 prot->slab = NULL;
Pavel Emelyanovb733c002007-11-07 02:23:38 -08002296out:
2297 return -ENOBUFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299EXPORT_SYMBOL(proto_register);
2300
2301void proto_unregister(struct proto *prot)
2302{
2303 write_lock(&proto_list_lock);
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002304 release_proto_idx(prot);
Patrick McHardy0a3f4352005-09-06 19:47:50 -07002305 list_del(&prot->node);
2306 write_unlock(&proto_list_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002307
2308 if (prot->slab != NULL) {
2309 kmem_cache_destroy(prot->slab);
2310 prot->slab = NULL;
2311 }
2312
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002313 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002314 kmem_cache_destroy(prot->rsk_prot->slab);
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002315 kfree(prot->rsk_prot->slab_name);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002316 prot->rsk_prot->slab = NULL;
2317 }
2318
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002319 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002320 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002321 kfree(prot->twsk_prot->twsk_slab_name);
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002322 prot->twsk_prot->twsk_slab = NULL;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002323 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325EXPORT_SYMBOL(proto_unregister);
2326
2327#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002329 __acquires(proto_list_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002330{
2331 read_lock(&proto_list_lock);
Pavel Emelianov60f04382007-07-09 13:15:14 -07002332 return seq_list_start_head(&proto_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333}
2334
2335static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2336{
Pavel Emelianov60f04382007-07-09 13:15:14 -07002337 return seq_list_next(v, &proto_list, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338}
2339
2340static void proto_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002341 __releases(proto_list_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342{
2343 read_unlock(&proto_list_lock);
2344}
2345
2346static char proto_method_implemented(const void *method)
2347{
2348 return method == NULL ? 'n' : 'y';
2349}
2350
2351static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2352{
2353 seq_printf(seq, "%-9s %4u %6d %6d %-3s %6u %-3s %-10s "
2354 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2355 proto->name,
2356 proto->obj_size,
Eric Dumazet14e943d2008-11-19 15:14:01 -08002357 sock_prot_inuse_get(seq_file_net(seq), proto),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358 proto->memory_allocated != NULL ? atomic_read(proto->memory_allocated) : -1,
2359 proto->memory_pressure != NULL ? *proto->memory_pressure ? "yes" : "no" : "NI",
2360 proto->max_header,
2361 proto->slab == NULL ? "no" : "yes",
2362 module_name(proto->owner),
2363 proto_method_implemented(proto->close),
2364 proto_method_implemented(proto->connect),
2365 proto_method_implemented(proto->disconnect),
2366 proto_method_implemented(proto->accept),
2367 proto_method_implemented(proto->ioctl),
2368 proto_method_implemented(proto->init),
2369 proto_method_implemented(proto->destroy),
2370 proto_method_implemented(proto->shutdown),
2371 proto_method_implemented(proto->setsockopt),
2372 proto_method_implemented(proto->getsockopt),
2373 proto_method_implemented(proto->sendmsg),
2374 proto_method_implemented(proto->recvmsg),
2375 proto_method_implemented(proto->sendpage),
2376 proto_method_implemented(proto->bind),
2377 proto_method_implemented(proto->backlog_rcv),
2378 proto_method_implemented(proto->hash),
2379 proto_method_implemented(proto->unhash),
2380 proto_method_implemented(proto->get_port),
2381 proto_method_implemented(proto->enter_memory_pressure));
2382}
2383
2384static int proto_seq_show(struct seq_file *seq, void *v)
2385{
Pavel Emelianov60f04382007-07-09 13:15:14 -07002386 if (v == &proto_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2388 "protocol",
2389 "size",
2390 "sockets",
2391 "memory",
2392 "press",
2393 "maxhdr",
2394 "slab",
2395 "module",
2396 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2397 else
Pavel Emelianov60f04382007-07-09 13:15:14 -07002398 proto_seq_printf(seq, list_entry(v, struct proto, node));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399 return 0;
2400}
2401
Stephen Hemmingerf6908082007-03-12 14:34:29 -07002402static const struct seq_operations proto_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002403 .start = proto_seq_start,
2404 .next = proto_seq_next,
2405 .stop = proto_seq_stop,
2406 .show = proto_seq_show,
2407};
2408
2409static int proto_seq_open(struct inode *inode, struct file *file)
2410{
Eric Dumazet14e943d2008-11-19 15:14:01 -08002411 return seq_open_net(inode, file, &proto_seq_ops,
2412 sizeof(struct seq_net_private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002413}
2414
Arjan van de Ven9a321442007-02-12 00:55:35 -08002415static const struct file_operations proto_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002416 .owner = THIS_MODULE,
2417 .open = proto_seq_open,
2418 .read = seq_read,
2419 .llseek = seq_lseek,
Eric Dumazet14e943d2008-11-19 15:14:01 -08002420 .release = seq_release_net,
2421};
2422
2423static __net_init int proto_init_net(struct net *net)
2424{
2425 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2426 return -ENOMEM;
2427
2428 return 0;
2429}
2430
2431static __net_exit void proto_exit_net(struct net *net)
2432{
2433 proc_net_remove(net, "protocols");
2434}
2435
2436
2437static __net_initdata struct pernet_operations proto_net_ops = {
2438 .init = proto_init_net,
2439 .exit = proto_exit_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440};
2441
2442static int __init proto_init(void)
2443{
Eric Dumazet14e943d2008-11-19 15:14:01 -08002444 return register_pernet_subsys(&proto_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445}
2446
2447subsys_initcall(proto_init);
2448
2449#endif /* PROC_FS */