blob: 06e26b77ad9e78437948592eb1781539b6947703 [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 Viroa1f8e7f2006-10-19 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 */
145static const char *af_family_key_strings[AF_MAX+1] = {
146 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
147 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
148 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
149 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
150 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
151 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
152 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
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};
161static const char *af_family_slock_key_strings[AF_MAX+1] = {
162 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
163 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
164 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
165 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
166 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
167 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
168 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
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};
Peter Zijlstra443aef02007-07-19 01:49:00 -0700177static const char *af_family_clock_key_strings[AF_MAX+1] = {
178 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
179 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
180 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
181 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
182 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
183 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
184 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
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{
277 int err = 0;
278 int skb_len;
279
Rami Rosen9ee6b7f2008-05-14 03:50:03 -0700280 /* Cast sk->rcvbuf to unsigned... It's pointless, but reduces
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800281 number of warnings when compiling with -W --ANK
282 */
283 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
284 (unsigned)sk->sk_rcvbuf) {
285 err = -ENOMEM;
286 goto out;
287 }
288
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700289 err = sk_filter(sk, skb);
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800290 if (err)
291 goto out;
292
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800293 if (!sk_rmem_schedule(sk, skb->truesize)) {
294 err = -ENOBUFS;
295 goto out;
296 }
297
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800298 skb->dev = NULL;
299 skb_set_owner_r(skb, sk);
David S. Miller49ad9592008-12-17 22:11:38 -0800300
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800301 /* Cache the SKB length before we tack it onto the receive
302 * queue. Once it is added it no longer belongs to us and
303 * may be freed by other threads of control pulling packets
304 * from the queue.
305 */
306 skb_len = skb->len;
307
308 skb_queue_tail(&sk->sk_receive_queue, skb);
309
310 if (!sock_flag(sk, SOCK_DEAD))
311 sk->sk_data_ready(sk, skb_len);
312out:
313 return err;
314}
315EXPORT_SYMBOL(sock_queue_rcv_skb);
316
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200317int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800318{
319 int rc = NET_RX_SUCCESS;
320
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700321 if (sk_filter(sk, skb))
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800322 goto discard_and_relse;
323
324 skb->dev = NULL;
325
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200326 if (nested)
327 bh_lock_sock_nested(sk);
328 else
329 bh_lock_sock(sk);
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700330 if (!sock_owned_by_user(sk)) {
331 /*
332 * trylock + unlock semantics:
333 */
334 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
335
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700336 rc = sk_backlog_rcv(sk, skb);
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700337
338 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
339 } else
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800340 sk_add_backlog(sk, skb);
341 bh_unlock_sock(sk);
342out:
343 sock_put(sk);
344 return rc;
345discard_and_relse:
346 kfree_skb(skb);
347 goto out;
348}
349EXPORT_SYMBOL(sk_receive_skb);
350
351struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
352{
353 struct dst_entry *dst = sk->sk_dst_cache;
354
355 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
356 sk->sk_dst_cache = NULL;
357 dst_release(dst);
358 return NULL;
359 }
360
361 return dst;
362}
363EXPORT_SYMBOL(__sk_dst_check);
364
365struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
366{
367 struct dst_entry *dst = sk_dst_get(sk);
368
369 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
370 sk_dst_reset(sk);
371 dst_release(dst);
372 return NULL;
373 }
374
375 return dst;
376}
377EXPORT_SYMBOL(sk_dst_check);
378
David S. Miller48788092007-09-14 16:41:03 -0700379static int sock_bindtodevice(struct sock *sk, char __user *optval, int optlen)
380{
381 int ret = -ENOPROTOOPT;
382#ifdef CONFIG_NETDEVICES
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900383 struct net *net = sock_net(sk);
David S. Miller48788092007-09-14 16:41:03 -0700384 char devname[IFNAMSIZ];
385 int index;
386
387 /* Sorry... */
388 ret = -EPERM;
389 if (!capable(CAP_NET_RAW))
390 goto out;
391
392 ret = -EINVAL;
393 if (optlen < 0)
394 goto out;
395
396 /* Bind this socket to a particular device like "eth0",
397 * as specified in the passed interface name. If the
398 * name is "" or the option length is zero the socket
399 * is not bound.
400 */
401 if (optlen > IFNAMSIZ - 1)
402 optlen = IFNAMSIZ - 1;
403 memset(devname, 0, sizeof(devname));
404
405 ret = -EFAULT;
406 if (copy_from_user(devname, optval, optlen))
407 goto out;
408
409 if (devname[0] == '\0') {
410 index = 0;
411 } else {
Eric W. Biederman881d9662007-09-17 11:56:21 -0700412 struct net_device *dev = dev_get_by_name(net, devname);
David S. Miller48788092007-09-14 16:41:03 -0700413
414 ret = -ENODEV;
415 if (!dev)
416 goto out;
417
418 index = dev->ifindex;
419 dev_put(dev);
420 }
421
422 lock_sock(sk);
423 sk->sk_bound_dev_if = index;
424 sk_dst_reset(sk);
425 release_sock(sk);
426
427 ret = 0;
428
429out:
430#endif
431
432 return ret;
433}
434
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800435static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
436{
437 if (valbool)
438 sock_set_flag(sk, bit);
439 else
440 sock_reset_flag(sk, bit);
441}
442
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443/*
444 * This is meant for all protocols to use and covers goings on
445 * at the socket level. Everything here is generic.
446 */
447
448int sock_setsockopt(struct socket *sock, int level, int optname,
449 char __user *optval, int optlen)
450{
Eric Dumazet2a915252009-05-27 11:30:05 +0000451 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452 int val;
453 int valbool;
454 struct linger ling;
455 int ret = 0;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900456
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 /*
458 * Options without arguments
459 */
460
David S. Miller48788092007-09-14 16:41:03 -0700461 if (optname == SO_BINDTODEVICE)
462 return sock_bindtodevice(sk, optval, optlen);
463
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700464 if (optlen < sizeof(int))
465 return -EINVAL;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900466
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 if (get_user(val, (int __user *)optval))
468 return -EFAULT;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900469
Eric Dumazet2a915252009-05-27 11:30:05 +0000470 valbool = val ? 1 : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
472 lock_sock(sk);
473
Eric Dumazet2a915252009-05-27 11:30:05 +0000474 switch (optname) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700475 case SO_DEBUG:
Eric Dumazet2a915252009-05-27 11:30:05 +0000476 if (val && !capable(CAP_NET_ADMIN))
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700477 ret = -EACCES;
Eric Dumazet2a915252009-05-27 11:30:05 +0000478 else
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800479 sock_valbool_flag(sk, SOCK_DBG, valbool);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700480 break;
481 case SO_REUSEADDR:
482 sk->sk_reuse = valbool;
483 break;
484 case SO_TYPE:
485 case SO_ERROR:
486 ret = -ENOPROTOOPT;
487 break;
488 case SO_DONTROUTE:
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800489 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700490 break;
491 case SO_BROADCAST:
492 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
493 break;
494 case SO_SNDBUF:
495 /* Don't error on this BSD doesn't and if you think
496 about it this is right. Otherwise apps have to
497 play 'guess the biggest size' games. RCVBUF/SNDBUF
498 are treated in BSD as hints */
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900499
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700500 if (val > sysctl_wmem_max)
501 val = sysctl_wmem_max;
Patrick McHardyb0573de2005-08-09 19:30:51 -0700502set_sndbuf:
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700503 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
504 if ((val * 2) < SOCK_MIN_SNDBUF)
505 sk->sk_sndbuf = SOCK_MIN_SNDBUF;
506 else
507 sk->sk_sndbuf = val * 2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700509 /*
510 * Wake up sending tasks if we
511 * upped the value.
512 */
513 sk->sk_write_space(sk);
514 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700516 case SO_SNDBUFFORCE:
517 if (!capable(CAP_NET_ADMIN)) {
518 ret = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 break;
520 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700521 goto set_sndbuf;
522
523 case SO_RCVBUF:
524 /* Don't error on this BSD doesn't and if you think
525 about it this is right. Otherwise apps have to
526 play 'guess the biggest size' games. RCVBUF/SNDBUF
527 are treated in BSD as hints */
528
529 if (val > sysctl_rmem_max)
530 val = sysctl_rmem_max;
531set_rcvbuf:
532 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
533 /*
534 * We double it on the way in to account for
535 * "struct sk_buff" etc. overhead. Applications
536 * assume that the SO_RCVBUF setting they make will
537 * allow that much actual data to be received on that
538 * socket.
539 *
540 * Applications are unaware that "struct sk_buff" and
541 * other overheads allocate from the receive buffer
542 * during socket buffer allocation.
543 *
544 * And after considering the possible alternatives,
545 * returning the value we actually used in getsockopt
546 * is the most desirable behavior.
547 */
548 if ((val * 2) < SOCK_MIN_RCVBUF)
549 sk->sk_rcvbuf = SOCK_MIN_RCVBUF;
550 else
551 sk->sk_rcvbuf = val * 2;
552 break;
553
554 case SO_RCVBUFFORCE:
555 if (!capable(CAP_NET_ADMIN)) {
556 ret = -EPERM;
557 break;
558 }
559 goto set_rcvbuf;
560
561 case SO_KEEPALIVE:
562#ifdef CONFIG_INET
563 if (sk->sk_protocol == IPPROTO_TCP)
564 tcp_set_keepalive(sk, valbool);
565#endif
566 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
567 break;
568
569 case SO_OOBINLINE:
570 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
571 break;
572
573 case SO_NO_CHECK:
574 sk->sk_no_check = valbool;
575 break;
576
577 case SO_PRIORITY:
578 if ((val >= 0 && val <= 6) || capable(CAP_NET_ADMIN))
579 sk->sk_priority = val;
580 else
581 ret = -EPERM;
582 break;
583
584 case SO_LINGER:
585 if (optlen < sizeof(ling)) {
586 ret = -EINVAL; /* 1003.1g */
587 break;
588 }
Eric Dumazet2a915252009-05-27 11:30:05 +0000589 if (copy_from_user(&ling, optval, sizeof(ling))) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700590 ret = -EFAULT;
591 break;
592 }
593 if (!ling.l_onoff)
594 sock_reset_flag(sk, SOCK_LINGER);
595 else {
596#if (BITS_PER_LONG == 32)
597 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
598 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
599 else
600#endif
601 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
602 sock_set_flag(sk, SOCK_LINGER);
603 }
604 break;
605
606 case SO_BSDCOMPAT:
607 sock_warn_obsolete_bsdism("setsockopt");
608 break;
609
610 case SO_PASSCRED:
611 if (valbool)
612 set_bit(SOCK_PASSCRED, &sock->flags);
613 else
614 clear_bit(SOCK_PASSCRED, &sock->flags);
615 break;
616
617 case SO_TIMESTAMP:
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700618 case SO_TIMESTAMPNS:
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700619 if (valbool) {
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700620 if (optname == SO_TIMESTAMP)
621 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
622 else
623 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700624 sock_set_flag(sk, SOCK_RCVTSTAMP);
Patrick Ohly20d49472009-02-12 05:03:38 +0000625 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700626 } else {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700627 sock_reset_flag(sk, SOCK_RCVTSTAMP);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700628 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
629 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700630 break;
631
Patrick Ohly20d49472009-02-12 05:03:38 +0000632 case SO_TIMESTAMPING:
633 if (val & ~SOF_TIMESTAMPING_MASK) {
634 ret = EINVAL;
635 break;
636 }
637 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
638 val & SOF_TIMESTAMPING_TX_HARDWARE);
639 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
640 val & SOF_TIMESTAMPING_TX_SOFTWARE);
641 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
642 val & SOF_TIMESTAMPING_RX_HARDWARE);
643 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
644 sock_enable_timestamp(sk,
645 SOCK_TIMESTAMPING_RX_SOFTWARE);
646 else
647 sock_disable_timestamp(sk,
648 SOCK_TIMESTAMPING_RX_SOFTWARE);
649 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
650 val & SOF_TIMESTAMPING_SOFTWARE);
651 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE,
652 val & SOF_TIMESTAMPING_SYS_HARDWARE);
653 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
654 val & SOF_TIMESTAMPING_RAW_HARDWARE);
655 break;
656
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700657 case SO_RCVLOWAT:
658 if (val < 0)
659 val = INT_MAX;
660 sk->sk_rcvlowat = val ? : 1;
661 break;
662
663 case SO_RCVTIMEO:
664 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
665 break;
666
667 case SO_SNDTIMEO:
668 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
669 break;
670
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700671 case SO_ATTACH_FILTER:
672 ret = -EINVAL;
673 if (optlen == sizeof(struct sock_fprog)) {
674 struct sock_fprog fprog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700676 ret = -EFAULT;
677 if (copy_from_user(&fprog, optval, sizeof(fprog)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700680 ret = sk_attach_filter(&fprog, sk);
681 }
682 break;
683
684 case SO_DETACH_FILTER:
Pavel Emelyanov55b33322007-10-17 21:21:26 -0700685 ret = sk_detach_filter(sk);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700686 break;
687
688 case SO_PASSSEC:
689 if (valbool)
690 set_bit(SOCK_PASSSEC, &sock->flags);
691 else
692 clear_bit(SOCK_PASSSEC, &sock->flags);
693 break;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800694 case SO_MARK:
695 if (!capable(CAP_NET_ADMIN))
696 ret = -EPERM;
Eric Dumazet2a915252009-05-27 11:30:05 +0000697 else
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800698 sk->sk_mark = val;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800699 break;
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700700
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701 /* We implement the SO_SNDLOWAT etc to
702 not be settable (1003.1g 5.3) */
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700703 default:
704 ret = -ENOPROTOOPT;
705 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900706 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707 release_sock(sk);
708 return ret;
709}
Eric Dumazet2a915252009-05-27 11:30:05 +0000710EXPORT_SYMBOL(sock_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
712
713int sock_getsockopt(struct socket *sock, int level, int optname,
714 char __user *optval, int __user *optlen)
715{
716 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900717
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700718 union {
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900719 int val;
720 struct linger ling;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721 struct timeval tm;
722 } v;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900723
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724 unsigned int lv = sizeof(int);
725 int len;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900726
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700727 if (get_user(len, optlen))
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900728 return -EFAULT;
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700729 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730 return -EINVAL;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900731
Eugene Teo50fee1d2009-02-23 15:38:41 -0800732 memset(&v, 0, sizeof(v));
Clément Lecignedf0bca02009-02-12 16:59:09 -0800733
Eric Dumazet2a915252009-05-27 11:30:05 +0000734 switch (optname) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700735 case SO_DEBUG:
736 v.val = sock_flag(sk, SOCK_DBG);
737 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900738
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700739 case SO_DONTROUTE:
740 v.val = sock_flag(sk, SOCK_LOCALROUTE);
741 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900742
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700743 case SO_BROADCAST:
744 v.val = !!sock_flag(sk, SOCK_BROADCAST);
745 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700747 case SO_SNDBUF:
748 v.val = sk->sk_sndbuf;
749 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900750
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700751 case SO_RCVBUF:
752 v.val = sk->sk_rcvbuf;
753 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700755 case SO_REUSEADDR:
756 v.val = sk->sk_reuse;
757 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700759 case SO_KEEPALIVE:
760 v.val = !!sock_flag(sk, SOCK_KEEPOPEN);
761 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700763 case SO_TYPE:
764 v.val = sk->sk_type;
765 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700767 case SO_ERROR:
768 v.val = -sock_error(sk);
Eric Dumazet2a915252009-05-27 11:30:05 +0000769 if (v.val == 0)
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700770 v.val = xchg(&sk->sk_err_soft, 0);
771 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700772
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700773 case SO_OOBINLINE:
774 v.val = !!sock_flag(sk, SOCK_URGINLINE);
775 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900776
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700777 case SO_NO_CHECK:
778 v.val = sk->sk_no_check;
779 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700781 case SO_PRIORITY:
782 v.val = sk->sk_priority;
783 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900784
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700785 case SO_LINGER:
786 lv = sizeof(v.ling);
787 v.ling.l_onoff = !!sock_flag(sk, SOCK_LINGER);
788 v.ling.l_linger = sk->sk_lingertime / HZ;
789 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900790
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700791 case SO_BSDCOMPAT:
792 sock_warn_obsolete_bsdism("getsockopt");
793 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700795 case SO_TIMESTAMP:
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700796 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
797 !sock_flag(sk, SOCK_RCVTSTAMPNS);
798 break;
799
800 case SO_TIMESTAMPNS:
801 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700802 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803
Patrick Ohly20d49472009-02-12 05:03:38 +0000804 case SO_TIMESTAMPING:
805 v.val = 0;
806 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
807 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
808 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
809 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
810 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
811 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
812 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
813 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
814 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
815 v.val |= SOF_TIMESTAMPING_SOFTWARE;
816 if (sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))
817 v.val |= SOF_TIMESTAMPING_SYS_HARDWARE;
818 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
819 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
820 break;
821
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700822 case SO_RCVTIMEO:
Eric Dumazet2a915252009-05-27 11:30:05 +0000823 lv = sizeof(struct timeval);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700824 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
825 v.tm.tv_sec = 0;
826 v.tm.tv_usec = 0;
827 } else {
828 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
829 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700831 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700833 case SO_SNDTIMEO:
Eric Dumazet2a915252009-05-27 11:30:05 +0000834 lv = sizeof(struct timeval);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700835 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
836 v.tm.tv_sec = 0;
837 v.tm.tv_usec = 0;
838 } else {
839 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
840 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
841 }
842 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700844 case SO_RCVLOWAT:
845 v.val = sk->sk_rcvlowat;
846 break;
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700847
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700848 case SO_SNDLOWAT:
Eric Dumazet2a915252009-05-27 11:30:05 +0000849 v.val = 1;
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700850 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700852 case SO_PASSCRED:
853 v.val = test_bit(SOCK_PASSCRED, &sock->flags) ? 1 : 0;
854 break;
855
856 case SO_PEERCRED:
857 if (len > sizeof(sk->sk_peercred))
858 len = sizeof(sk->sk_peercred);
859 if (copy_to_user(optval, &sk->sk_peercred, len))
860 return -EFAULT;
861 goto lenout;
862
863 case SO_PEERNAME:
864 {
865 char address[128];
866
867 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
868 return -ENOTCONN;
869 if (lv < len)
870 return -EINVAL;
871 if (copy_to_user(optval, address, len))
872 return -EFAULT;
873 goto lenout;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700875
876 /* Dubious BSD thing... Probably nobody even uses it, but
877 * the UNIX standard wants it for whatever reason... -DaveM
878 */
879 case SO_ACCEPTCONN:
880 v.val = sk->sk_state == TCP_LISTEN;
881 break;
882
883 case SO_PASSSEC:
884 v.val = test_bit(SOCK_PASSSEC, &sock->flags) ? 1 : 0;
885 break;
886
887 case SO_PEERSEC:
888 return security_socket_getpeersec_stream(sock, optval, optlen, len);
889
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800890 case SO_MARK:
891 v.val = sk->sk_mark;
892 break;
893
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700894 default:
895 return -ENOPROTOOPT;
896 }
897
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 if (len > lv)
899 len = lv;
900 if (copy_to_user(optval, &v, len))
901 return -EFAULT;
902lenout:
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900903 if (put_user(len, optlen))
904 return -EFAULT;
905 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906}
907
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700908/*
909 * Initialize an sk_lock.
910 *
911 * (We also register the sk_lock with the lock validator.)
912 */
Dave Jonesb6f99a22007-03-22 12:27:49 -0700913static inline void sock_lock_init(struct sock *sk)
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700914{
Peter Zijlstraed075362006-12-06 20:35:24 -0800915 sock_lock_init_class_and_name(sk,
916 af_family_slock_key_strings[sk->sk_family],
917 af_family_slock_keys + sk->sk_family,
918 af_family_key_strings[sk->sk_family],
919 af_family_keys + sk->sk_family);
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700920}
921
Pavel Emelyanovf1a6c4d2007-11-01 00:29:45 -0700922static void sock_copy(struct sock *nsk, const struct sock *osk)
923{
924#ifdef CONFIG_SECURITY_NETWORK
925 void *sptr = nsk->sk_security;
926#endif
927
928 memcpy(nsk, osk, osk->sk_prot->obj_size);
Pavel Emelyanovf1a6c4d2007-11-01 00:29:45 -0700929#ifdef CONFIG_SECURITY_NETWORK
930 nsk->sk_security = sptr;
931 security_sk_clone(osk, nsk);
932#endif
933}
934
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -0700935static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
936 int family)
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -0700937{
938 struct sock *sk;
939 struct kmem_cache *slab;
940
941 slab = prot->slab;
942 if (slab != NULL)
943 sk = kmem_cache_alloc(slab, priority);
944 else
945 sk = kmalloc(prot->obj_size, priority);
946
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -0700947 if (sk != NULL) {
948 if (security_sk_alloc(sk, family, priority))
949 goto out_free;
950
951 if (!try_module_get(prot->owner))
952 goto out_free_sec;
953 }
954
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -0700955 return sk;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -0700956
957out_free_sec:
958 security_sk_free(sk);
959out_free:
960 if (slab != NULL)
961 kmem_cache_free(slab, sk);
962 else
963 kfree(sk);
964 return NULL;
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -0700965}
966
967static void sk_prot_free(struct proto *prot, struct sock *sk)
968{
969 struct kmem_cache *slab;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -0700970 struct module *owner;
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -0700971
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -0700972 owner = prot->owner;
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -0700973 slab = prot->slab;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -0700974
975 security_sk_free(sk);
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -0700976 if (slab != NULL)
977 kmem_cache_free(slab, sk);
978 else
979 kfree(sk);
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -0700980 module_put(owner);
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -0700981}
982
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983/**
984 * sk_alloc - All socket objects are allocated here
Randy Dunlapc4ea43c2007-10-12 21:17:49 -0700985 * @net: the applicable net namespace
Pavel Pisa4dc3b162005-05-01 08:59:25 -0700986 * @family: protocol family
987 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
988 * @prot: struct proto associated with this new sock instance
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 */
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -0700990struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
Pavel Emelyanov6257ff22007-11-01 00:39:31 -0700991 struct proto *prot)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992{
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -0700993 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994
Pavel Emelyanov154adbc2007-11-01 00:38:43 -0700995 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 if (sk) {
Pavel Emelyanov154adbc2007-11-01 00:38:43 -0700997 sk->sk_family = family;
998 /*
999 * See comment in struct sock definition to understand
1000 * why we need sk_prot_creator -acme
1001 */
1002 sk->sk_prot = sk->sk_prot_creator = prot;
1003 sock_lock_init(sk);
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001004 sock_net_set(sk, get_net(net));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 }
Frank Filza79af592005-09-27 15:23:38 -07001006
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001007 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008}
Eric Dumazet2a915252009-05-27 11:30:05 +00001009EXPORT_SYMBOL(sk_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010
Eric Dumazet2b85a342009-06-11 02:55:43 -07001011static void __sk_free(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012{
1013 struct sk_filter *filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014
1015 if (sk->sk_destruct)
1016 sk->sk_destruct(sk);
1017
Dmitry Mishinfda9ef52006-08-31 15:28:39 -07001018 filter = rcu_dereference(sk->sk_filter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 if (filter) {
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001020 sk_filter_uncharge(sk, filter);
Dmitry Mishinfda9ef52006-08-31 15:28:39 -07001021 rcu_assign_pointer(sk->sk_filter, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 }
1023
Patrick Ohly20d49472009-02-12 05:03:38 +00001024 sock_disable_timestamp(sk, SOCK_TIMESTAMP);
1025 sock_disable_timestamp(sk, SOCK_TIMESTAMPING_RX_SOFTWARE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026
1027 if (atomic_read(&sk->sk_omem_alloc))
1028 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n",
Harvey Harrison0dc47872008-03-05 20:47:47 -08001029 __func__, atomic_read(&sk->sk_omem_alloc));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001031 put_net(sock_net(sk));
Pavel Emelyanovc308c1b2007-11-01 00:33:50 -07001032 sk_prot_free(sk->sk_prot_creator, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033}
Eric Dumazet2b85a342009-06-11 02:55:43 -07001034
1035void sk_free(struct sock *sk)
1036{
1037 /*
1038 * We substract one from sk_wmem_alloc and can know if
1039 * some packets are still in some tx queue.
1040 * If not null, sock_wfree() will call __sk_free(sk) later
1041 */
1042 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1043 __sk_free(sk);
1044}
Eric Dumazet2a915252009-05-27 11:30:05 +00001045EXPORT_SYMBOL(sk_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001046
Denis V. Lunevedf02082008-02-29 11:18:32 -08001047/*
1048 * Last sock_put should drop referrence to sk->sk_net. It has already
1049 * been dropped in sk_change_net. Taking referrence to stopping namespace
1050 * is not an option.
1051 * Take referrence to a socket to remove it from hash _alive_ and after that
1052 * destroy it in the context of init_net.
1053 */
1054void sk_release_kernel(struct sock *sk)
1055{
1056 if (sk == NULL || sk->sk_socket == NULL)
1057 return;
1058
1059 sock_hold(sk);
1060 sock_release(sk->sk_socket);
Denis V. Lunev65a18ec2008-04-16 01:59:46 -07001061 release_net(sock_net(sk));
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001062 sock_net_set(sk, get_net(&init_net));
Denis V. Lunevedf02082008-02-29 11:18:32 -08001063 sock_put(sk);
1064}
David S. Miller45af1752008-02-29 11:33:19 -08001065EXPORT_SYMBOL(sk_release_kernel);
Denis V. Lunevedf02082008-02-29 11:18:32 -08001066
Al Virodd0fc662005-10-07 07:46:04 +01001067struct sock *sk_clone(const struct sock *sk, const gfp_t priority)
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001068{
Pavel Emelyanov8fd1d172007-11-01 00:37:32 -07001069 struct sock *newsk;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001070
Pavel Emelyanov8fd1d172007-11-01 00:37:32 -07001071 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001072 if (newsk != NULL) {
1073 struct sk_filter *filter;
1074
Venkat Yekkirala892c1412006-08-04 23:08:56 -07001075 sock_copy(newsk, sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001076
1077 /* SANITY */
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +09001078 get_net(sock_net(newsk));
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001079 sk_node_init(&newsk->sk_node);
1080 sock_lock_init(newsk);
1081 bh_lock_sock(newsk);
Eric Dumazetfa438cc2007-03-04 16:05:44 -08001082 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001083
1084 atomic_set(&newsk->sk_rmem_alloc, 0);
Eric Dumazet2b85a342009-06-11 02:55:43 -07001085 /*
1086 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1087 */
1088 atomic_set(&newsk->sk_wmem_alloc, 1);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001089 atomic_set(&newsk->sk_omem_alloc, 0);
1090 skb_queue_head_init(&newsk->sk_receive_queue);
1091 skb_queue_head_init(&newsk->sk_write_queue);
Chris Leech97fc2f02006-05-23 17:55:33 -07001092#ifdef CONFIG_NET_DMA
1093 skb_queue_head_init(&newsk->sk_async_wait_queue);
1094#endif
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001095
1096 rwlock_init(&newsk->sk_dst_lock);
1097 rwlock_init(&newsk->sk_callback_lock);
Peter Zijlstra443aef02007-07-19 01:49:00 -07001098 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1099 af_callback_keys + newsk->sk_family,
1100 af_family_clock_key_strings[newsk->sk_family]);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001101
1102 newsk->sk_dst_cache = NULL;
1103 newsk->sk_wmem_queued = 0;
1104 newsk->sk_forward_alloc = 0;
1105 newsk->sk_send_head = NULL;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001106 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1107
1108 sock_reset_flag(newsk, SOCK_DONE);
1109 skb_queue_head_init(&newsk->sk_error_queue);
1110
1111 filter = newsk->sk_filter;
1112 if (filter != NULL)
1113 sk_filter_charge(newsk, filter);
1114
1115 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1116 /* It is still raw copy of parent, so invalidate
1117 * destructor and make plain sk_free() */
1118 newsk->sk_destruct = NULL;
1119 sk_free(newsk);
1120 newsk = NULL;
1121 goto out;
1122 }
1123
1124 newsk->sk_err = 0;
1125 newsk->sk_priority = 0;
1126 atomic_set(&newsk->sk_refcnt, 2);
1127
1128 /*
1129 * Increment the counter in the same struct proto as the master
1130 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1131 * is the same as sk->sk_prot->socks, as this field was copied
1132 * with memcpy).
1133 *
1134 * This _changes_ the previous behaviour, where
1135 * tcp_create_openreq_child always was incrementing the
1136 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1137 * to be taken into account in all callers. -acme
1138 */
1139 sk_refcnt_debug_inc(newsk);
David S. Miller972692e2008-06-17 22:41:38 -07001140 sk_set_socket(newsk, NULL);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001141 newsk->sk_sleep = NULL;
1142
1143 if (newsk->sk_prot->sockets_allocated)
Eric Dumazet17483762008-11-25 21:16:35 -08001144 percpu_counter_inc(newsk->sk_prot->sockets_allocated);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001145 }
1146out:
1147 return newsk;
1148}
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001149EXPORT_SYMBOL_GPL(sk_clone);
1150
Andi Kleen99580892007-04-20 17:12:43 -07001151void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1152{
1153 __sk_dst_set(sk, dst);
1154 sk->sk_route_caps = dst->dev->features;
1155 if (sk->sk_route_caps & NETIF_F_GSO)
Herbert Xu4fcd6b92007-05-31 22:15:50 -07001156 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
Andi Kleen99580892007-04-20 17:12:43 -07001157 if (sk_can_gso(sk)) {
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001158 if (dst->header_len) {
Andi Kleen99580892007-04-20 17:12:43 -07001159 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001160 } else {
Andi Kleen99580892007-04-20 17:12:43 -07001161 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001162 sk->sk_gso_max_size = dst->dev->gso_max_size;
1163 }
Andi Kleen99580892007-04-20 17:12:43 -07001164 }
1165}
1166EXPORT_SYMBOL_GPL(sk_setup_caps);
1167
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168void __init sk_init(void)
1169{
1170 if (num_physpages <= 4096) {
1171 sysctl_wmem_max = 32767;
1172 sysctl_rmem_max = 32767;
1173 sysctl_wmem_default = 32767;
1174 sysctl_rmem_default = 32767;
1175 } else if (num_physpages >= 131072) {
1176 sysctl_wmem_max = 131071;
1177 sysctl_rmem_max = 131071;
1178 }
1179}
1180
1181/*
1182 * Simple resource managers for sockets.
1183 */
1184
1185
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001186/*
1187 * Write buffer destructor automatically called from kfree_skb.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 */
1189void sock_wfree(struct sk_buff *skb)
1190{
1191 struct sock *sk = skb->sk;
Eric Dumazet2b85a342009-06-11 02:55:43 -07001192 int res;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193
1194 /* In case it might be waiting for more memory. */
Eric Dumazet2b85a342009-06-11 02:55:43 -07001195 res = atomic_sub_return(skb->truesize, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE))
1197 sk->sk_write_space(sk);
Eric Dumazet2b85a342009-06-11 02:55:43 -07001198 /*
1199 * if sk_wmem_alloc reached 0, we are last user and should
1200 * free this sock, as sk_free() call could not do it.
1201 */
1202 if (res == 0)
1203 __sk_free(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204}
Eric Dumazet2a915252009-05-27 11:30:05 +00001205EXPORT_SYMBOL(sock_wfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001207/*
1208 * Read buffer destructor automatically called from kfree_skb.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209 */
1210void sock_rfree(struct sk_buff *skb)
1211{
1212 struct sock *sk = skb->sk;
1213
1214 atomic_sub(skb->truesize, &sk->sk_rmem_alloc);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001215 sk_mem_uncharge(skb->sk, skb->truesize);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216}
Eric Dumazet2a915252009-05-27 11:30:05 +00001217EXPORT_SYMBOL(sock_rfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218
1219
1220int sock_i_uid(struct sock *sk)
1221{
1222 int uid;
1223
1224 read_lock(&sk->sk_callback_lock);
1225 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
1226 read_unlock(&sk->sk_callback_lock);
1227 return uid;
1228}
Eric Dumazet2a915252009-05-27 11:30:05 +00001229EXPORT_SYMBOL(sock_i_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230
1231unsigned long sock_i_ino(struct sock *sk)
1232{
1233 unsigned long ino;
1234
1235 read_lock(&sk->sk_callback_lock);
1236 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
1237 read_unlock(&sk->sk_callback_lock);
1238 return ino;
1239}
Eric Dumazet2a915252009-05-27 11:30:05 +00001240EXPORT_SYMBOL(sock_i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241
1242/*
1243 * Allocate a skb from the socket's send buffer.
1244 */
Victor Fusco86a76ca2005-07-08 14:57:47 -07001245struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001246 gfp_t priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247{
1248 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
Eric Dumazet2a915252009-05-27 11:30:05 +00001249 struct sk_buff *skb = alloc_skb(size, priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250 if (skb) {
1251 skb_set_owner_w(skb, sk);
1252 return skb;
1253 }
1254 }
1255 return NULL;
1256}
Eric Dumazet2a915252009-05-27 11:30:05 +00001257EXPORT_SYMBOL(sock_wmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258
1259/*
1260 * Allocate a skb from the socket's receive buffer.
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001261 */
Victor Fusco86a76ca2005-07-08 14:57:47 -07001262struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001263 gfp_t priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264{
1265 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1266 struct sk_buff *skb = alloc_skb(size, priority);
1267 if (skb) {
1268 skb_set_owner_r(skb, sk);
1269 return skb;
1270 }
1271 }
1272 return NULL;
1273}
1274
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001275/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 * Allocate a memory block from the socket's option memory buffer.
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001277 */
Al Virodd0fc662005-10-07 07:46:04 +01001278void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279{
1280 if ((unsigned)size <= sysctl_optmem_max &&
1281 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1282 void *mem;
1283 /* First do the add, to avoid the race if kmalloc
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001284 * might sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 */
1286 atomic_add(size, &sk->sk_omem_alloc);
1287 mem = kmalloc(size, priority);
1288 if (mem)
1289 return mem;
1290 atomic_sub(size, &sk->sk_omem_alloc);
1291 }
1292 return NULL;
1293}
Eric Dumazet2a915252009-05-27 11:30:05 +00001294EXPORT_SYMBOL(sock_kmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295
1296/*
1297 * Free an option memory block.
1298 */
1299void sock_kfree_s(struct sock *sk, void *mem, int size)
1300{
1301 kfree(mem);
1302 atomic_sub(size, &sk->sk_omem_alloc);
1303}
Eric Dumazet2a915252009-05-27 11:30:05 +00001304EXPORT_SYMBOL(sock_kfree_s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305
1306/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1307 I think, these locks should be removed for datagram sockets.
1308 */
Eric Dumazet2a915252009-05-27 11:30:05 +00001309static long sock_wait_for_wmem(struct sock *sk, long timeo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310{
1311 DEFINE_WAIT(wait);
1312
1313 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1314 for (;;) {
1315 if (!timeo)
1316 break;
1317 if (signal_pending(current))
1318 break;
1319 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1320 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1321 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1322 break;
1323 if (sk->sk_shutdown & SEND_SHUTDOWN)
1324 break;
1325 if (sk->sk_err)
1326 break;
1327 timeo = schedule_timeout(timeo);
1328 }
1329 finish_wait(sk->sk_sleep, &wait);
1330 return timeo;
1331}
1332
1333
1334/*
1335 * Generic send/receive buffer handlers
1336 */
1337
Herbert Xu4cc7f682009-02-04 16:55:54 -08001338struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1339 unsigned long data_len, int noblock,
1340 int *errcode)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341{
1342 struct sk_buff *skb;
Al Viro7d877f32005-10-21 03:20:43 -04001343 gfp_t gfp_mask;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344 long timeo;
1345 int err;
1346
1347 gfp_mask = sk->sk_allocation;
1348 if (gfp_mask & __GFP_WAIT)
1349 gfp_mask |= __GFP_REPEAT;
1350
1351 timeo = sock_sndtimeo(sk, noblock);
1352 while (1) {
1353 err = sock_error(sk);
1354 if (err != 0)
1355 goto failure;
1356
1357 err = -EPIPE;
1358 if (sk->sk_shutdown & SEND_SHUTDOWN)
1359 goto failure;
1360
1361 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
Larry Woodmandb38c1792006-11-03 16:05:45 -08001362 skb = alloc_skb(header_len, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363 if (skb) {
1364 int npages;
1365 int i;
1366
1367 /* No pages, we're done... */
1368 if (!data_len)
1369 break;
1370
1371 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1372 skb->truesize += data_len;
1373 skb_shinfo(skb)->nr_frags = npages;
1374 for (i = 0; i < npages; i++) {
1375 struct page *page;
1376 skb_frag_t *frag;
1377
1378 page = alloc_pages(sk->sk_allocation, 0);
1379 if (!page) {
1380 err = -ENOBUFS;
1381 skb_shinfo(skb)->nr_frags = i;
1382 kfree_skb(skb);
1383 goto failure;
1384 }
1385
1386 frag = &skb_shinfo(skb)->frags[i];
1387 frag->page = page;
1388 frag->page_offset = 0;
1389 frag->size = (data_len >= PAGE_SIZE ?
1390 PAGE_SIZE :
1391 data_len);
1392 data_len -= PAGE_SIZE;
1393 }
1394
1395 /* Full success... */
1396 break;
1397 }
1398 err = -ENOBUFS;
1399 goto failure;
1400 }
1401 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1402 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1403 err = -EAGAIN;
1404 if (!timeo)
1405 goto failure;
1406 if (signal_pending(current))
1407 goto interrupted;
1408 timeo = sock_wait_for_wmem(sk, timeo);
1409 }
1410
1411 skb_set_owner_w(skb, sk);
1412 return skb;
1413
1414interrupted:
1415 err = sock_intr_errno(timeo);
1416failure:
1417 *errcode = err;
1418 return NULL;
1419}
Herbert Xu4cc7f682009-02-04 16:55:54 -08001420EXPORT_SYMBOL(sock_alloc_send_pskb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001422struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 int noblock, int *errcode)
1424{
1425 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1426}
Eric Dumazet2a915252009-05-27 11:30:05 +00001427EXPORT_SYMBOL(sock_alloc_send_skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428
1429static void __lock_sock(struct sock *sk)
1430{
1431 DEFINE_WAIT(wait);
1432
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001433 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1435 TASK_UNINTERRUPTIBLE);
1436 spin_unlock_bh(&sk->sk_lock.slock);
1437 schedule();
1438 spin_lock_bh(&sk->sk_lock.slock);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001439 if (!sock_owned_by_user(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 break;
1441 }
1442 finish_wait(&sk->sk_lock.wq, &wait);
1443}
1444
1445static void __release_sock(struct sock *sk)
1446{
1447 struct sk_buff *skb = sk->sk_backlog.head;
1448
1449 do {
1450 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1451 bh_unlock_sock(sk);
1452
1453 do {
1454 struct sk_buff *next = skb->next;
1455
1456 skb->next = NULL;
Peter Zijlstrac57943a2008-10-07 14:18:42 -07001457 sk_backlog_rcv(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458
1459 /*
1460 * We are in process context here with softirqs
1461 * disabled, use cond_resched_softirq() to preempt.
1462 * This is safe to do because we've taken the backlog
1463 * queue private:
1464 */
1465 cond_resched_softirq();
1466
1467 skb = next;
1468 } while (skb != NULL);
1469
1470 bh_lock_sock(sk);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001471 } while ((skb = sk->sk_backlog.head) != NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472}
1473
1474/**
1475 * sk_wait_data - wait for data to arrive at sk_receive_queue
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001476 * @sk: sock to wait on
1477 * @timeo: for how long
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 *
1479 * Now socket state including sk->sk_err is changed only under lock,
1480 * hence we may omit checks after joining wait queue.
1481 * We check receive queue before schedule() only as optimization;
1482 * it is very likely that release_sock() added new data.
1483 */
1484int sk_wait_data(struct sock *sk, long *timeo)
1485{
1486 int rc;
1487 DEFINE_WAIT(wait);
1488
1489 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1490 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1491 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1492 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1493 finish_wait(sk->sk_sleep, &wait);
1494 return rc;
1495}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496EXPORT_SYMBOL(sk_wait_data);
1497
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001498/**
1499 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1500 * @sk: socket
1501 * @size: memory size to allocate
1502 * @kind: allocation type
1503 *
1504 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1505 * rmem allocation. This function assumes that protocols which have
1506 * memory_pressure use sk_wmem_queued as write buffer accounting.
1507 */
1508int __sk_mem_schedule(struct sock *sk, int size, int kind)
1509{
1510 struct proto *prot = sk->sk_prot;
1511 int amt = sk_mem_pages(size);
1512 int allocated;
1513
1514 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
1515 allocated = atomic_add_return(amt, prot->memory_allocated);
1516
1517 /* Under limit. */
1518 if (allocated <= prot->sysctl_mem[0]) {
1519 if (prot->memory_pressure && *prot->memory_pressure)
1520 *prot->memory_pressure = 0;
1521 return 1;
1522 }
1523
1524 /* Under pressure. */
1525 if (allocated > prot->sysctl_mem[1])
1526 if (prot->enter_memory_pressure)
Pavel Emelyanov5c52ba12008-07-16 20:28:10 -07001527 prot->enter_memory_pressure(sk);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001528
1529 /* Over hard limit. */
1530 if (allocated > prot->sysctl_mem[2])
1531 goto suppress_allocation;
1532
1533 /* guarantee minimum buffer size under pressure */
1534 if (kind == SK_MEM_RECV) {
1535 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
1536 return 1;
1537 } else { /* SK_MEM_SEND */
1538 if (sk->sk_type == SOCK_STREAM) {
1539 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
1540 return 1;
1541 } else if (atomic_read(&sk->sk_wmem_alloc) <
1542 prot->sysctl_wmem[0])
1543 return 1;
1544 }
1545
1546 if (prot->memory_pressure) {
Eric Dumazet17483762008-11-25 21:16:35 -08001547 int alloc;
1548
1549 if (!*prot->memory_pressure)
1550 return 1;
1551 alloc = percpu_counter_read_positive(prot->sockets_allocated);
1552 if (prot->sysctl_mem[2] > alloc *
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001553 sk_mem_pages(sk->sk_wmem_queued +
1554 atomic_read(&sk->sk_rmem_alloc) +
1555 sk->sk_forward_alloc))
1556 return 1;
1557 }
1558
1559suppress_allocation:
1560
1561 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
1562 sk_stream_moderate_sndbuf(sk);
1563
1564 /* Fail only if socket is _under_ its sndbuf.
1565 * In this case we cannot block, so that we have to fail.
1566 */
1567 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
1568 return 1;
1569 }
1570
1571 /* Alas. Undo changes. */
1572 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
1573 atomic_sub(amt, prot->memory_allocated);
1574 return 0;
1575}
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001576EXPORT_SYMBOL(__sk_mem_schedule);
1577
1578/**
1579 * __sk_reclaim - reclaim memory_allocated
1580 * @sk: socket
1581 */
1582void __sk_mem_reclaim(struct sock *sk)
1583{
1584 struct proto *prot = sk->sk_prot;
1585
Eric Dumazet680a5a52007-12-31 15:00:50 -08001586 atomic_sub(sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT,
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001587 prot->memory_allocated);
1588 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
1589
1590 if (prot->memory_pressure && *prot->memory_pressure &&
1591 (atomic_read(prot->memory_allocated) < prot->sysctl_mem[0]))
1592 *prot->memory_pressure = 0;
1593}
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001594EXPORT_SYMBOL(__sk_mem_reclaim);
1595
1596
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597/*
1598 * Set of default routines for initialising struct proto_ops when
1599 * the protocol does not support a particular function. In certain
1600 * cases where it makes no sense for a protocol to have a "do nothing"
1601 * function, some default processing is provided.
1602 */
1603
1604int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1605{
1606 return -EOPNOTSUPP;
1607}
Eric Dumazet2a915252009-05-27 11:30:05 +00001608EXPORT_SYMBOL(sock_no_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001610int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 int len, int flags)
1612{
1613 return -EOPNOTSUPP;
1614}
Eric Dumazet2a915252009-05-27 11:30:05 +00001615EXPORT_SYMBOL(sock_no_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
1617int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1618{
1619 return -EOPNOTSUPP;
1620}
Eric Dumazet2a915252009-05-27 11:30:05 +00001621EXPORT_SYMBOL(sock_no_socketpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622
1623int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1624{
1625 return -EOPNOTSUPP;
1626}
Eric Dumazet2a915252009-05-27 11:30:05 +00001627EXPORT_SYMBOL(sock_no_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001629int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 int *len, int peer)
1631{
1632 return -EOPNOTSUPP;
1633}
Eric Dumazet2a915252009-05-27 11:30:05 +00001634EXPORT_SYMBOL(sock_no_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635
Eric Dumazet2a915252009-05-27 11:30:05 +00001636unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637{
1638 return 0;
1639}
Eric Dumazet2a915252009-05-27 11:30:05 +00001640EXPORT_SYMBOL(sock_no_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641
1642int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1643{
1644 return -EOPNOTSUPP;
1645}
Eric Dumazet2a915252009-05-27 11:30:05 +00001646EXPORT_SYMBOL(sock_no_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
1648int sock_no_listen(struct socket *sock, int backlog)
1649{
1650 return -EOPNOTSUPP;
1651}
Eric Dumazet2a915252009-05-27 11:30:05 +00001652EXPORT_SYMBOL(sock_no_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653
1654int sock_no_shutdown(struct socket *sock, int how)
1655{
1656 return -EOPNOTSUPP;
1657}
Eric Dumazet2a915252009-05-27 11:30:05 +00001658EXPORT_SYMBOL(sock_no_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659
1660int sock_no_setsockopt(struct socket *sock, int level, int optname,
1661 char __user *optval, int optlen)
1662{
1663 return -EOPNOTSUPP;
1664}
Eric Dumazet2a915252009-05-27 11:30:05 +00001665EXPORT_SYMBOL(sock_no_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666
1667int sock_no_getsockopt(struct socket *sock, int level, int optname,
1668 char __user *optval, int __user *optlen)
1669{
1670 return -EOPNOTSUPP;
1671}
Eric Dumazet2a915252009-05-27 11:30:05 +00001672EXPORT_SYMBOL(sock_no_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673
1674int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1675 size_t len)
1676{
1677 return -EOPNOTSUPP;
1678}
Eric Dumazet2a915252009-05-27 11:30:05 +00001679EXPORT_SYMBOL(sock_no_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680
1681int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1682 size_t len, int flags)
1683{
1684 return -EOPNOTSUPP;
1685}
Eric Dumazet2a915252009-05-27 11:30:05 +00001686EXPORT_SYMBOL(sock_no_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
1688int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
1689{
1690 /* Mirror missing mmap method error code */
1691 return -ENODEV;
1692}
Eric Dumazet2a915252009-05-27 11:30:05 +00001693EXPORT_SYMBOL(sock_no_mmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694
1695ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
1696{
1697 ssize_t res;
1698 struct msghdr msg = {.msg_flags = flags};
1699 struct kvec iov;
1700 char *kaddr = kmap(page);
1701 iov.iov_base = kaddr + offset;
1702 iov.iov_len = size;
1703 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
1704 kunmap(page);
1705 return res;
1706}
Eric Dumazet2a915252009-05-27 11:30:05 +00001707EXPORT_SYMBOL(sock_no_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708
1709/*
1710 * Default Socket Callbacks
1711 */
1712
1713static void sock_def_wakeup(struct sock *sk)
1714{
1715 read_lock(&sk->sk_callback_lock);
1716 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1717 wake_up_interruptible_all(sk->sk_sleep);
1718 read_unlock(&sk->sk_callback_lock);
1719}
1720
1721static void sock_def_error_report(struct sock *sk)
1722{
1723 read_lock(&sk->sk_callback_lock);
1724 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
Davide Libenzi37e55402009-03-31 15:24:21 -07001725 wake_up_interruptible_poll(sk->sk_sleep, POLLERR);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001726 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727 read_unlock(&sk->sk_callback_lock);
1728}
1729
1730static void sock_def_readable(struct sock *sk, int len)
1731{
1732 read_lock(&sk->sk_callback_lock);
1733 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
Davide Libenzi37e55402009-03-31 15:24:21 -07001734 wake_up_interruptible_sync_poll(sk->sk_sleep, POLLIN |
1735 POLLRDNORM | POLLRDBAND);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001736 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737 read_unlock(&sk->sk_callback_lock);
1738}
1739
1740static void sock_def_write_space(struct sock *sk)
1741{
1742 read_lock(&sk->sk_callback_lock);
1743
1744 /* Do not wake up a writer until he can make "significant"
1745 * progress. --DaveM
1746 */
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001747 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
Davide Libenzi37e55402009-03-31 15:24:21 -07001749 wake_up_interruptible_sync_poll(sk->sk_sleep, POLLOUT |
1750 POLLWRNORM | POLLWRBAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751
1752 /* Should agree with poll, otherwise some programs break */
1753 if (sock_writeable(sk))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001754 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755 }
1756
1757 read_unlock(&sk->sk_callback_lock);
1758}
1759
1760static void sock_def_destruct(struct sock *sk)
1761{
Jesper Juhla51482b2005-11-08 09:41:34 -08001762 kfree(sk->sk_protinfo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763}
1764
1765void sk_send_sigurg(struct sock *sk)
1766{
1767 if (sk->sk_socket && sk->sk_socket->file)
1768 if (send_sigurg(&sk->sk_socket->file->f_owner))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08001769 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770}
Eric Dumazet2a915252009-05-27 11:30:05 +00001771EXPORT_SYMBOL(sk_send_sigurg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772
1773void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1774 unsigned long expires)
1775{
1776 if (!mod_timer(timer, expires))
1777 sock_hold(sk);
1778}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779EXPORT_SYMBOL(sk_reset_timer);
1780
1781void sk_stop_timer(struct sock *sk, struct timer_list* timer)
1782{
1783 if (timer_pending(timer) && del_timer(timer))
1784 __sock_put(sk);
1785}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786EXPORT_SYMBOL(sk_stop_timer);
1787
1788void sock_init_data(struct socket *sock, struct sock *sk)
1789{
1790 skb_queue_head_init(&sk->sk_receive_queue);
1791 skb_queue_head_init(&sk->sk_write_queue);
1792 skb_queue_head_init(&sk->sk_error_queue);
Chris Leech97fc2f02006-05-23 17:55:33 -07001793#ifdef CONFIG_NET_DMA
1794 skb_queue_head_init(&sk->sk_async_wait_queue);
1795#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796
1797 sk->sk_send_head = NULL;
1798
1799 init_timer(&sk->sk_timer);
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001800
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801 sk->sk_allocation = GFP_KERNEL;
1802 sk->sk_rcvbuf = sysctl_rmem_default;
1803 sk->sk_sndbuf = sysctl_wmem_default;
1804 sk->sk_state = TCP_CLOSE;
David S. Miller972692e2008-06-17 22:41:38 -07001805 sk_set_socket(sk, sock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806
1807 sock_set_flag(sk, SOCK_ZAPPED);
1808
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001809 if (sock) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 sk->sk_type = sock->type;
1811 sk->sk_sleep = &sock->wait;
1812 sock->sk = sk;
1813 } else
1814 sk->sk_sleep = NULL;
1815
1816 rwlock_init(&sk->sk_dst_lock);
1817 rwlock_init(&sk->sk_callback_lock);
Peter Zijlstra443aef02007-07-19 01:49:00 -07001818 lockdep_set_class_and_name(&sk->sk_callback_lock,
1819 af_callback_keys + sk->sk_family,
1820 af_family_clock_key_strings[sk->sk_family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821
1822 sk->sk_state_change = sock_def_wakeup;
1823 sk->sk_data_ready = sock_def_readable;
1824 sk->sk_write_space = sock_def_write_space;
1825 sk->sk_error_report = sock_def_error_report;
1826 sk->sk_destruct = sock_def_destruct;
1827
1828 sk->sk_sndmsg_page = NULL;
1829 sk->sk_sndmsg_off = 0;
1830
1831 sk->sk_peercred.pid = 0;
1832 sk->sk_peercred.uid = -1;
1833 sk->sk_peercred.gid = -1;
1834 sk->sk_write_pending = 0;
1835 sk->sk_rcvlowat = 1;
1836 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
1837 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
1838
Eric Dumazetf37f0af2008-04-13 21:39:26 -07001839 sk->sk_stamp = ktime_set(-1L, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001840
1841 atomic_set(&sk->sk_refcnt, 1);
Eric Dumazet2b85a342009-06-11 02:55:43 -07001842 atomic_set(&sk->sk_wmem_alloc, 1);
Wang Chen33c732c2007-11-13 20:30:01 -08001843 atomic_set(&sk->sk_drops, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844}
Eric Dumazet2a915252009-05-27 11:30:05 +00001845EXPORT_SYMBOL(sock_init_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001846
Harvey Harrisonb5606c22008-02-13 15:03:16 -08001847void lock_sock_nested(struct sock *sk, int subclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848{
1849 might_sleep();
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001850 spin_lock_bh(&sk->sk_lock.slock);
John Heffnerd2e91172007-09-12 10:44:19 +02001851 if (sk->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 __lock_sock(sk);
John Heffnerd2e91172007-09-12 10:44:19 +02001853 sk->sk_lock.owned = 1;
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001854 spin_unlock(&sk->sk_lock.slock);
1855 /*
1856 * The sk_lock has mutex_lock() semantics here:
1857 */
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001858 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001859 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860}
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08001861EXPORT_SYMBOL(lock_sock_nested);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001862
Harvey Harrisonb5606c22008-02-13 15:03:16 -08001863void release_sock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864{
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001865 /*
1866 * The sk_lock has mutex_unlock() semantics:
1867 */
1868 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1869
1870 spin_lock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 if (sk->sk_backlog.tail)
1872 __release_sock(sk);
John Heffnerd2e91172007-09-12 10:44:19 +02001873 sk->sk_lock.owned = 0;
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001874 if (waitqueue_active(&sk->sk_lock.wq))
1875 wake_up(&sk->sk_lock.wq);
1876 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877}
1878EXPORT_SYMBOL(release_sock);
1879
1880int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001881{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001882 struct timeval tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883 if (!sock_flag(sk, SOCK_TIMESTAMP))
Patrick Ohly20d49472009-02-12 05:03:38 +00001884 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001885 tv = ktime_to_timeval(sk->sk_stamp);
1886 if (tv.tv_sec == -1)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 return -ENOENT;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001888 if (tv.tv_sec == 0) {
1889 sk->sk_stamp = ktime_get_real();
1890 tv = ktime_to_timeval(sk->sk_stamp);
1891 }
1892 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001893}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894EXPORT_SYMBOL(sock_get_timestamp);
1895
Eric Dumazetae40eb12007-03-18 17:33:16 -07001896int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
1897{
1898 struct timespec ts;
1899 if (!sock_flag(sk, SOCK_TIMESTAMP))
Patrick Ohly20d49472009-02-12 05:03:38 +00001900 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazetae40eb12007-03-18 17:33:16 -07001901 ts = ktime_to_timespec(sk->sk_stamp);
1902 if (ts.tv_sec == -1)
1903 return -ENOENT;
1904 if (ts.tv_sec == 0) {
1905 sk->sk_stamp = ktime_get_real();
1906 ts = ktime_to_timespec(sk->sk_stamp);
1907 }
1908 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
1909}
1910EXPORT_SYMBOL(sock_get_timestampns);
1911
Patrick Ohly20d49472009-02-12 05:03:38 +00001912void sock_enable_timestamp(struct sock *sk, int flag)
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001913{
Patrick Ohly20d49472009-02-12 05:03:38 +00001914 if (!sock_flag(sk, flag)) {
1915 sock_set_flag(sk, flag);
1916 /*
1917 * we just set one of the two flags which require net
1918 * time stamping, but time stamping might have been on
1919 * already because of the other one
1920 */
1921 if (!sock_flag(sk,
1922 flag == SOCK_TIMESTAMP ?
1923 SOCK_TIMESTAMPING_RX_SOFTWARE :
1924 SOCK_TIMESTAMP))
1925 net_enable_timestamp();
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926 }
1927}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928
1929/*
1930 * Get a socket option on an socket.
1931 *
1932 * FIX: POSIX 1003.1g is very ambiguous here. It states that
1933 * asynchronous errors should be reported by getsockopt. We assume
1934 * this means if you specify SO_ERROR (otherwise whats the point of it).
1935 */
1936int sock_common_getsockopt(struct socket *sock, int level, int optname,
1937 char __user *optval, int __user *optlen)
1938{
1939 struct sock *sk = sock->sk;
1940
1941 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
1942}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943EXPORT_SYMBOL(sock_common_getsockopt);
1944
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001945#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001946int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
1947 char __user *optval, int __user *optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001948{
1949 struct sock *sk = sock->sk;
1950
Johannes Berg1e51f952007-03-06 13:44:06 -08001951 if (sk->sk_prot->compat_getsockopt != NULL)
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001952 return sk->sk_prot->compat_getsockopt(sk, level, optname,
1953 optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001954 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
1955}
1956EXPORT_SYMBOL(compat_sock_common_getsockopt);
1957#endif
1958
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1960 struct msghdr *msg, size_t size, int flags)
1961{
1962 struct sock *sk = sock->sk;
1963 int addr_len = 0;
1964 int err;
1965
1966 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
1967 flags & ~MSG_DONTWAIT, &addr_len);
1968 if (err >= 0)
1969 msg->msg_namelen = addr_len;
1970 return err;
1971}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972EXPORT_SYMBOL(sock_common_recvmsg);
1973
1974/*
1975 * Set socket options on an inet socket.
1976 */
1977int sock_common_setsockopt(struct socket *sock, int level, int optname,
1978 char __user *optval, int optlen)
1979{
1980 struct sock *sk = sock->sk;
1981
1982 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
1983}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984EXPORT_SYMBOL(sock_common_setsockopt);
1985
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001986#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001987int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
1988 char __user *optval, int optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001989{
1990 struct sock *sk = sock->sk;
1991
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08001992 if (sk->sk_prot->compat_setsockopt != NULL)
1993 return sk->sk_prot->compat_setsockopt(sk, level, optname,
1994 optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08001995 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
1996}
1997EXPORT_SYMBOL(compat_sock_common_setsockopt);
1998#endif
1999
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000void sk_common_release(struct sock *sk)
2001{
2002 if (sk->sk_prot->destroy)
2003 sk->sk_prot->destroy(sk);
2004
2005 /*
2006 * Observation: when sock_common_release is called, processes have
2007 * no access to socket. But net still has.
2008 * Step one, detach it from networking:
2009 *
2010 * A. Remove from hash tables.
2011 */
2012
2013 sk->sk_prot->unhash(sk);
2014
2015 /*
2016 * In this point socket cannot receive new packets, but it is possible
2017 * that some packets are in flight because some CPU runs receiver and
2018 * did hash table lookup before we unhashed socket. They will achieve
2019 * receive queue and will be purged by socket destructor.
2020 *
2021 * Also we still have packets pending on receive queue and probably,
2022 * our own packets waiting in device queues. sock_destroy will drain
2023 * receive queue, but transmitted packets will delay socket destruction
2024 * until the last reference will be released.
2025 */
2026
2027 sock_orphan(sk);
2028
2029 xfrm_sk_free_policy(sk);
2030
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07002031 sk_refcnt_debug_release(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032 sock_put(sk);
2033}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034EXPORT_SYMBOL(sk_common_release);
2035
2036static DEFINE_RWLOCK(proto_list_lock);
2037static LIST_HEAD(proto_list);
2038
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002039#ifdef CONFIG_PROC_FS
2040#define PROTO_INUSE_NR 64 /* should be enough for the first time */
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002041struct prot_inuse {
2042 int val[PROTO_INUSE_NR];
2043};
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002044
2045static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002046
2047#ifdef CONFIG_NET_NS
2048void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2049{
2050 int cpu = smp_processor_id();
2051 per_cpu_ptr(net->core.inuse, cpu)->val[prot->inuse_idx] += val;
2052}
2053EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2054
2055int sock_prot_inuse_get(struct net *net, struct proto *prot)
2056{
2057 int cpu, idx = prot->inuse_idx;
2058 int res = 0;
2059
2060 for_each_possible_cpu(cpu)
2061 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2062
2063 return res >= 0 ? res : 0;
2064}
2065EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2066
2067static int sock_inuse_init_net(struct net *net)
2068{
2069 net->core.inuse = alloc_percpu(struct prot_inuse);
2070 return net->core.inuse ? 0 : -ENOMEM;
2071}
2072
2073static void sock_inuse_exit_net(struct net *net)
2074{
2075 free_percpu(net->core.inuse);
2076}
2077
2078static struct pernet_operations net_inuse_ops = {
2079 .init = sock_inuse_init_net,
2080 .exit = sock_inuse_exit_net,
2081};
2082
2083static __init int net_inuse_init(void)
2084{
2085 if (register_pernet_subsys(&net_inuse_ops))
2086 panic("Cannot initialize net inuse counters");
2087
2088 return 0;
2089}
2090
2091core_initcall(net_inuse_init);
2092#else
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002093static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2094
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07002095void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002096{
2097 __get_cpu_var(prot_inuse).val[prot->inuse_idx] += val;
2098}
2099EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2100
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07002101int sock_prot_inuse_get(struct net *net, struct proto *prot)
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002102{
2103 int cpu, idx = prot->inuse_idx;
2104 int res = 0;
2105
2106 for_each_possible_cpu(cpu)
2107 res += per_cpu(prot_inuse, cpu).val[idx];
2108
2109 return res >= 0 ? res : 0;
2110}
2111EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002112#endif
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002113
2114static void assign_proto_idx(struct proto *prot)
2115{
2116 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2117
2118 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
2119 printk(KERN_ERR "PROTO_INUSE_NR exhausted\n");
2120 return;
2121 }
2122
2123 set_bit(prot->inuse_idx, proto_inuse_idx);
2124}
2125
2126static void release_proto_idx(struct proto *prot)
2127{
2128 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2129 clear_bit(prot->inuse_idx, proto_inuse_idx);
2130}
2131#else
2132static inline void assign_proto_idx(struct proto *prot)
2133{
2134}
2135
2136static inline void release_proto_idx(struct proto *prot)
2137{
2138}
2139#endif
2140
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141int proto_register(struct proto *prot, int alloc_slab)
2142{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 if (alloc_slab) {
2144 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
Eric Dumazet271b72c2008-10-29 02:11:14 -07002145 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2146 NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147
2148 if (prot->slab == NULL) {
2149 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n",
2150 prot->name);
Pavel Emelyanov60e76632008-03-28 16:39:10 -07002151 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152 }
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002153
2154 if (prot->rsk_prot != NULL) {
2155 static const char mask[] = "request_sock_%s";
2156
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002157 prot->rsk_prot->slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL);
2158 if (prot->rsk_prot->slab_name == NULL)
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002159 goto out_free_sock_slab;
2160
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002161 sprintf(prot->rsk_prot->slab_name, mask, prot->name);
2162 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002163 prot->rsk_prot->obj_size, 0,
Paul Mundt20c2df82007-07-20 10:11:58 +09002164 SLAB_HWCACHE_ALIGN, NULL);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002165
2166 if (prot->rsk_prot->slab == NULL) {
2167 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n",
2168 prot->name);
2169 goto out_free_request_sock_slab_name;
2170 }
2171 }
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -07002172
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002173 if (prot->twsk_prot != NULL) {
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -07002174 static const char mask[] = "tw_sock_%s";
2175
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002176 prot->twsk_prot->twsk_slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL);
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -07002177
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002178 if (prot->twsk_prot->twsk_slab_name == NULL)
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -07002179 goto out_free_request_sock_slab;
2180
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002181 sprintf(prot->twsk_prot->twsk_slab_name, mask, prot->name);
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002182 prot->twsk_prot->twsk_slab =
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002183 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002184 prot->twsk_prot->twsk_obj_size,
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002185 0,
2186 SLAB_HWCACHE_ALIGN |
2187 prot->slab_flags,
Paul Mundt20c2df82007-07-20 10:11:58 +09002188 NULL);
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002189 if (prot->twsk_prot->twsk_slab == NULL)
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -07002190 goto out_free_timewait_sock_slab_name;
2191 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192 }
2193
Arnaldo Carvalho de Melo2a278052005-04-16 15:24:09 -07002194 write_lock(&proto_list_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195 list_add(&prot->node, &proto_list);
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002196 assign_proto_idx(prot);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197 write_unlock(&proto_list_lock);
Pavel Emelyanovb733c002007-11-07 02:23:38 -08002198 return 0;
2199
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -07002200out_free_timewait_sock_slab_name:
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002201 kfree(prot->twsk_prot->twsk_slab_name);
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -07002202out_free_request_sock_slab:
2203 if (prot->rsk_prot && prot->rsk_prot->slab) {
2204 kmem_cache_destroy(prot->rsk_prot->slab);
2205 prot->rsk_prot->slab = NULL;
2206 }
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002207out_free_request_sock_slab_name:
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002208 kfree(prot->rsk_prot->slab_name);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002209out_free_sock_slab:
2210 kmem_cache_destroy(prot->slab);
2211 prot->slab = NULL;
Pavel Emelyanovb733c002007-11-07 02:23:38 -08002212out:
2213 return -ENOBUFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215EXPORT_SYMBOL(proto_register);
2216
2217void proto_unregister(struct proto *prot)
2218{
2219 write_lock(&proto_list_lock);
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002220 release_proto_idx(prot);
Patrick McHardy0a3f4352005-09-06 19:47:50 -07002221 list_del(&prot->node);
2222 write_unlock(&proto_list_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223
2224 if (prot->slab != NULL) {
2225 kmem_cache_destroy(prot->slab);
2226 prot->slab = NULL;
2227 }
2228
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002229 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002230 kmem_cache_destroy(prot->rsk_prot->slab);
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002231 kfree(prot->rsk_prot->slab_name);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002232 prot->rsk_prot->slab = NULL;
2233 }
2234
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002235 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002236 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002237 kfree(prot->twsk_prot->twsk_slab_name);
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002238 prot->twsk_prot->twsk_slab = NULL;
Arnaldo Carvalho de Melo8feaf0c2005-08-09 20:09:30 -07002239 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002240}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241EXPORT_SYMBOL(proto_unregister);
2242
2243#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002245 __acquires(proto_list_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246{
2247 read_lock(&proto_list_lock);
Pavel Emelianov60f04382007-07-09 13:15:14 -07002248 return seq_list_start_head(&proto_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002249}
2250
2251static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2252{
Pavel Emelianov60f04382007-07-09 13:15:14 -07002253 return seq_list_next(v, &proto_list, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254}
2255
2256static void proto_seq_stop(struct seq_file *seq, void *v)
Eric Dumazet9a429c42008-01-01 21:58:02 -08002257 __releases(proto_list_lock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002258{
2259 read_unlock(&proto_list_lock);
2260}
2261
2262static char proto_method_implemented(const void *method)
2263{
2264 return method == NULL ? 'n' : 'y';
2265}
2266
2267static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2268{
2269 seq_printf(seq, "%-9s %4u %6d %6d %-3s %6u %-3s %-10s "
2270 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2271 proto->name,
2272 proto->obj_size,
Eric Dumazet14e943d2008-11-19 15:14:01 -08002273 sock_prot_inuse_get(seq_file_net(seq), proto),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274 proto->memory_allocated != NULL ? atomic_read(proto->memory_allocated) : -1,
2275 proto->memory_pressure != NULL ? *proto->memory_pressure ? "yes" : "no" : "NI",
2276 proto->max_header,
2277 proto->slab == NULL ? "no" : "yes",
2278 module_name(proto->owner),
2279 proto_method_implemented(proto->close),
2280 proto_method_implemented(proto->connect),
2281 proto_method_implemented(proto->disconnect),
2282 proto_method_implemented(proto->accept),
2283 proto_method_implemented(proto->ioctl),
2284 proto_method_implemented(proto->init),
2285 proto_method_implemented(proto->destroy),
2286 proto_method_implemented(proto->shutdown),
2287 proto_method_implemented(proto->setsockopt),
2288 proto_method_implemented(proto->getsockopt),
2289 proto_method_implemented(proto->sendmsg),
2290 proto_method_implemented(proto->recvmsg),
2291 proto_method_implemented(proto->sendpage),
2292 proto_method_implemented(proto->bind),
2293 proto_method_implemented(proto->backlog_rcv),
2294 proto_method_implemented(proto->hash),
2295 proto_method_implemented(proto->unhash),
2296 proto_method_implemented(proto->get_port),
2297 proto_method_implemented(proto->enter_memory_pressure));
2298}
2299
2300static int proto_seq_show(struct seq_file *seq, void *v)
2301{
Pavel Emelianov60f04382007-07-09 13:15:14 -07002302 if (v == &proto_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002303 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2304 "protocol",
2305 "size",
2306 "sockets",
2307 "memory",
2308 "press",
2309 "maxhdr",
2310 "slab",
2311 "module",
2312 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2313 else
Pavel Emelianov60f04382007-07-09 13:15:14 -07002314 proto_seq_printf(seq, list_entry(v, struct proto, node));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315 return 0;
2316}
2317
Stephen Hemmingerf6908082007-03-12 14:34:29 -07002318static const struct seq_operations proto_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319 .start = proto_seq_start,
2320 .next = proto_seq_next,
2321 .stop = proto_seq_stop,
2322 .show = proto_seq_show,
2323};
2324
2325static int proto_seq_open(struct inode *inode, struct file *file)
2326{
Eric Dumazet14e943d2008-11-19 15:14:01 -08002327 return seq_open_net(inode, file, &proto_seq_ops,
2328 sizeof(struct seq_net_private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329}
2330
Arjan van de Ven9a321442007-02-12 00:55:35 -08002331static const struct file_operations proto_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002332 .owner = THIS_MODULE,
2333 .open = proto_seq_open,
2334 .read = seq_read,
2335 .llseek = seq_lseek,
Eric Dumazet14e943d2008-11-19 15:14:01 -08002336 .release = seq_release_net,
2337};
2338
2339static __net_init int proto_init_net(struct net *net)
2340{
2341 if (!proc_net_fops_create(net, "protocols", S_IRUGO, &proto_seq_fops))
2342 return -ENOMEM;
2343
2344 return 0;
2345}
2346
2347static __net_exit void proto_exit_net(struct net *net)
2348{
2349 proc_net_remove(net, "protocols");
2350}
2351
2352
2353static __net_initdata struct pernet_operations proto_net_ops = {
2354 .init = proto_init_net,
2355 .exit = proto_exit_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356};
2357
2358static int __init proto_init(void)
2359{
Eric Dumazet14e943d2008-11-19 15:14:01 -08002360 return register_pernet_subsys(&proto_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361}
2362
2363subsys_initcall(proto_init);
2364
2365#endif /* PROC_FS */