projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[NET]: Move some core sock setup into sk_prot_alloc
[deliverable/linux.git] / net / core / sock.c
CommitLineData
1da177e4
LT		 1/*
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 *
10 * Version: $Id: sock.c,v 1.117 2002/02/01 22:01:03 davem Exp $
11 *
02c30a84 12 * Authors: Ross Biro
1da177e4
LT		 13 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
14 * Florian La Roche, <flla@stud.uni-sb.de>
15 * Alan Cox, <A.Cox@swansea.ac.uk>
16 *
17 * Fixes:
18 * Alan Cox : Numerous verify_area() problems
19 * Alan Cox : Connecting on a connecting socket
20 * now returns an error for tcp.
21 * Alan Cox : sock->protocol is set correctly.
22 * and is not sometimes left as 0.
23 * Alan Cox : connect handles icmp errors on a
24 * connect properly. Unfortunately there
25 * is a restart syscall nasty there. I
26 * can't match BSD without hacking the C
27 * library. Ideas urgently sought!
28 * Alan Cox : Disallow bind() to addresses that are
29 * not ours - especially broadcast ones!!
30 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost)
31 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets,
32 * instead they leave that for the DESTROY timer.
33 * Alan Cox : Clean up error flag in accept
34 * Alan Cox : TCP ack handling is buggy, the DESTROY timer
35 * was buggy. Put a remove_sock() in the handler
36 * for memory when we hit 0. Also altered the timer
4ec93edb 37 * code. The ACK stuff can wait and needs major
1da177e4
LT		 38 * TCP layer surgery.
39 * Alan Cox : Fixed TCP ack bug, removed remove sock
40 * and fixed timer/inet_bh race.
41 * Alan Cox : Added zapped flag for TCP
42 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code
43 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb
44 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources
45 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing.
46 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so...
47 * Rick Sladkey : Relaxed UDP rules for matching packets.
48 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support
49 * Pauline Middelink : identd support
50 * Alan Cox : Fixed connect() taking signals I think.
51 * Alan Cox : SO_LINGER supported
52 * Alan Cox : Error reporting fixes
53 * Anonymous : inet_create tidied up (sk->reuse setting)
54 * Alan Cox : inet sockets don't set sk->type!
55 * Alan Cox : Split socket option code
56 * Alan Cox : Callbacks
57 * Alan Cox : Nagle flag for Charles & Johannes stuff
58 * Alex : Removed restriction on inet fioctl
59 * Alan Cox : Splitting INET from NET core
60 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt()
61 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code
62 * Alan Cox : Split IP from generic code
63 * Alan Cox : New kfree_skbmem()
64 * Alan Cox : Make SO_DEBUG superuser only.
65 * Alan Cox : Allow anyone to clear SO_DEBUG
66 * (compatibility fix)
67 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput.
68 * Alan Cox : Allocator for a socket is settable.
69 * Alan Cox : SO_ERROR includes soft errors.
70 * Alan Cox : Allow NULL arguments on some SO_ opts
71 * Alan Cox : Generic socket allocation to make hooks
72 * easier (suggested by Craig Metz).
73 * Michael Pall : SO_ERROR returns positive errno again
74 * Steve Whitehouse: Added default destructor to free
75 * protocol private data.
76 * Steve Whitehouse: Added various other default routines
77 * common to several socket families.
78 * Chris Evans : Call suser() check last on F_SETOWN
79 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER.
80 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s()
81 * Andi Kleen : Fix write_space callback
82 * Chris Evans : Security fixes - signedness again
83 * Arnaldo C. Melo : cleanups, use skb_queue_purge
84 *
85 * To Fix:
86 *
87 *
88 * This program is free software; you can redistribute it and/or
89 * modify it under the terms of the GNU General Public License
90 * as published by the Free Software Foundation; either version
91 * 2 of the License, or (at your option) any later version.
92 */
93
4fc268d2 94#include <linux/capability.h>
1da177e4
LT		 95#include <linux/errno.h>
96#include <linux/types.h>
97#include <linux/socket.h>
98#include <linux/in.h>
99#include <linux/kernel.h>
1da177e4
LT		 100#include <linux/module.h>
101#include <linux/proc_fs.h>
102#include <linux/seq_file.h>
103#include <linux/sched.h>
104#include <linux/timer.h>
105#include <linux/string.h>
106#include <linux/sockios.h>
107#include <linux/net.h>
108#include <linux/mm.h>
109#include <linux/slab.h>
110#include <linux/interrupt.h>
111#include <linux/poll.h>
112#include <linux/tcp.h>
113#include <linux/init.h>
a1f8e7f7 114#include <linux/highmem.h>
1da177e4
LT		 115
116#include <asm/uaccess.h>
117#include <asm/system.h>
118
119#include <linux/netdevice.h>
120#include <net/protocol.h>
121#include <linux/skbuff.h>
457c4cbc 122#include <net/net_namespace.h>
2e6599cb 123#include <net/request_sock.h>
1da177e4
LT		 124#include <net/sock.h>
125#include <net/xfrm.h>
126#include <linux/ipsec.h>
127
128#include <linux/filter.h>
129
130#ifdef CONFIG_INET
131#include <net/tcp.h>
132#endif
133
da21f24d
IM		 134/*
135 * Each address family might have different locking rules, so we have
136 * one slock key per address family:
137 */
a5b5bb9a
IM		 138static struct lock_class_key af_family_keys[AF_MAX];
139static struct lock_class_key af_family_slock_keys[AF_MAX];
140
141#ifdef CONFIG_DEBUG_LOCK_ALLOC
142/*
143 * Make lock validator output more readable. (we pre-construct these
144 * strings build-time, so that runtime initialization of socket
145 * locks is fast):
146 */
147static const char *af_family_key_strings[AF_MAX+1] = {
148 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
149 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
150 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
151 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
152 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
153 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
154 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
155 "sk_lock-21" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
156 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
157 "sk_lock-27" , "sk_lock-28" , "sk_lock-29" ,
17926a79
DH		 158 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
159 "sk_lock-AF_RXRPC" , "sk_lock-AF_MAX"
a5b5bb9a
IM		 160};
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" ,
169 "slock-21" , "slock-AF_SNA" , "slock-AF_IRDA" ,
170 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
171 "slock-27" , "slock-28" , "slock-29" ,
17926a79
DH		 172 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
173 "slock-AF_RXRPC" , "slock-AF_MAX"
a5b5bb9a 174};
443aef0e
PZ		 175static const char *af_family_clock_key_strings[AF_MAX+1] = {
176 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
177 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
178 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
179 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
180 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
181 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
182 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
183 "clock-21" , "clock-AF_SNA" , "clock-AF_IRDA" ,
184 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
185 "clock-27" , "clock-28" , "clock-29" ,
e51f802b
DH		 186 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
187 "clock-AF_RXRPC" , "clock-AF_MAX"
443aef0e 188};
a5b5bb9a 189#endif
da21f24d
IM		 190
191/*
192 * sk_callback_lock locking rules are per-address-family,
193 * so split the lock classes by using a per-AF key:
194 */
195static struct lock_class_key af_callback_keys[AF_MAX];
196
1da177e4
LT		 197/* Take into consideration the size of the struct sk_buff overhead in the
198 * determination of these values, since that is non-constant across
199 * platforms. This makes socket queueing behavior and performance
200 * not depend upon such differences.
201 */
202#define _SK_MEM_PACKETS 256
203#define _SK_MEM_OVERHEAD (sizeof(struct sk_buff) + 256)
204#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
205#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
206
207/* Run time adjustable parameters. */
ab32ea5d
BH		 208__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
209__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
210__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
211__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4
LT		 212
213/* Maximal space eaten by iovec or ancilliary data plus some space */
ab32ea5d 214int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
1da177e4
LT		 215
216static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
217{
218 struct timeval tv;
219
220 if (optlen < sizeof(tv))
221 return -EINVAL;
222 if (copy_from_user(&tv, optval, sizeof(tv)))
223 return -EFAULT;
ba78073e
VA		 224 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
225 return -EDOM;
1da177e4 226
ba78073e 227 if (tv.tv_sec < 0) {
6f11df83
AM		 228 static int warned __read_mostly;
229
ba78073e
VA		 230 *timeo_p = 0;
231 if (warned < 10 && net_ratelimit())
232 warned++;
233 printk(KERN_INFO "sock_set_timeout: `%s' (pid %d) "
234 "tries to set negative timeout\n",
ba25f9dc 235 current->comm, task_pid_nr(current));
ba78073e
VA		 236 return 0;
237 }
1da177e4
LT		 238 *timeo_p = MAX_SCHEDULE_TIMEOUT;
239 if (tv.tv_sec == 0 && tv.tv_usec == 0)
240 return 0;
241 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
242 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
243 return 0;
244}
245
246static void sock_warn_obsolete_bsdism(const char *name)
247{
248 static int warned;
249 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 250 if (strcmp(warncomm, current->comm) && warned < 5) {
251 strcpy(warncomm, current->comm);
1da177e4
LT		 252 printk(KERN_WARNING "process `%s' is using obsolete "
253 "%s SO_BSDCOMPAT\n", warncomm, name);
254 warned++;
255 }
256}
257
258static void sock_disable_timestamp(struct sock *sk)
4ec93edb
YH		 259{
260 if (sock_flag(sk, SOCK_TIMESTAMP)) {
1da177e4
LT		 261 sock_reset_flag(sk, SOCK_TIMESTAMP);
262 net_disable_timestamp();
263 }
264}
265
266
f0088a50
DV		 267int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
268{
269 int err = 0;
270 int skb_len;
271
272 /* Cast skb->rcvbuf to unsigned... It's pointless, but reduces
273 number of warnings when compiling with -W --ANK
274 */
275 if (atomic_read(&sk->sk_rmem_alloc) + skb->truesize >=
276 (unsigned)sk->sk_rcvbuf) {
277 err = -ENOMEM;
278 goto out;
279 }
280
fda9ef5d 281 err = sk_filter(sk, skb);
f0088a50
DV		 282 if (err)
283 goto out;
284
285 skb->dev = NULL;
286 skb_set_owner_r(skb, sk);
287
288 /* Cache the SKB length before we tack it onto the receive
289 * queue. Once it is added it no longer belongs to us and
290 * may be freed by other threads of control pulling packets
291 * from the queue.
292 */
293 skb_len = skb->len;
294
295 skb_queue_tail(&sk->sk_receive_queue, skb);
296
297 if (!sock_flag(sk, SOCK_DEAD))
298 sk->sk_data_ready(sk, skb_len);
299out:
300 return err;
301}
302EXPORT_SYMBOL(sock_queue_rcv_skb);
303
58a5a7b9 304int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 305{
306 int rc = NET_RX_SUCCESS;
307
fda9ef5d 308 if (sk_filter(sk, skb))
f0088a50
DV		 309 goto discard_and_relse;
310
311 skb->dev = NULL;
312
58a5a7b9
ACM		 313 if (nested)
314 bh_lock_sock_nested(sk);
315 else
316 bh_lock_sock(sk);
a5b5bb9a
IM		 317 if (!sock_owned_by_user(sk)) {
318 /*
319 * trylock + unlock semantics:
320 */
321 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
322
f0088a50 323 rc = sk->sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 324
325 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
326 } else
f0088a50
DV		 327 sk_add_backlog(sk, skb);
328 bh_unlock_sock(sk);
329out:
330 sock_put(sk);
331 return rc;
332discard_and_relse:
333 kfree_skb(skb);
334 goto out;
335}
336EXPORT_SYMBOL(sk_receive_skb);
337
338struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
339{
340 struct dst_entry *dst = sk->sk_dst_cache;
341
342 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
343 sk->sk_dst_cache = NULL;
344 dst_release(dst);
345 return NULL;
346 }
347
348 return dst;
349}
350EXPORT_SYMBOL(__sk_dst_check);
351
352struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
353{
354 struct dst_entry *dst = sk_dst_get(sk);
355
356 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
357 sk_dst_reset(sk);
358 dst_release(dst);
359 return NULL;
360 }
361
362 return dst;
363}
364EXPORT_SYMBOL(sk_dst_check);
365
4878809f
DM		 366static int sock_bindtodevice(struct sock *sk, char __user *optval, int optlen)
367{
368 int ret = -ENOPROTOOPT;
369#ifdef CONFIG_NETDEVICES
881d966b 370 struct net *net = sk->sk_net;
4878809f
DM		 371 char devname[IFNAMSIZ];
372 int index;
373
374 /* Sorry... */
375 ret = -EPERM;
376 if (!capable(CAP_NET_RAW))
377 goto out;
378
379 ret = -EINVAL;
380 if (optlen < 0)
381 goto out;
382
383 /* Bind this socket to a particular device like "eth0",
384 * as specified in the passed interface name. If the
385 * name is "" or the option length is zero the socket
386 * is not bound.
387 */
388 if (optlen > IFNAMSIZ - 1)
389 optlen = IFNAMSIZ - 1;
390 memset(devname, 0, sizeof(devname));
391
392 ret = -EFAULT;
393 if (copy_from_user(devname, optval, optlen))
394 goto out;
395
396 if (devname[0] == '\0') {
397 index = 0;
398 } else {
881d966b 399 struct net_device *dev = dev_get_by_name(net, devname);
4878809f
DM		 400
401 ret = -ENODEV;
402 if (!dev)
403 goto out;
404
405 index = dev->ifindex;
406 dev_put(dev);
407 }
408
409 lock_sock(sk);
410 sk->sk_bound_dev_if = index;
411 sk_dst_reset(sk);
412 release_sock(sk);
413
414 ret = 0;
415
416out:
417#endif
418
419 return ret;
420}
421
1da177e4
LT		 422/*
423 * This is meant for all protocols to use and covers goings on
424 * at the socket level. Everything here is generic.
425 */
426
427int sock_setsockopt(struct socket *sock, int level, int optname,
428 char __user *optval, int optlen)
429{
430 struct sock *sk=sock->sk;
1da177e4
LT		 431 int val;
432 int valbool;
433 struct linger ling;
434 int ret = 0;
4ec93edb 435
1da177e4
LT		 436 /*
437 * Options without arguments
438 */
439
440#ifdef SO_DONTLINGER /* Compatibility item... */
a77be819
KM		 441 if (optname == SO_DONTLINGER) {
442 lock_sock(sk);
443 sock_reset_flag(sk, SOCK_LINGER);
444 release_sock(sk);
445 return 0;
1da177e4 446 }
a77be819 447#endif
4ec93edb 448
4878809f
DM		 449 if (optname == SO_BINDTODEVICE)
450 return sock_bindtodevice(sk, optval, optlen);
451
e71a4783
SH		 452 if (optlen < sizeof(int))
453 return -EINVAL;
4ec93edb 454
1da177e4
LT		 455 if (get_user(val, (int __user *)optval))
456 return -EFAULT;
4ec93edb
YH		 457
458 valbool = val?1:0;
1da177e4
LT		 459
460 lock_sock(sk);
461
e71a4783
SH		 462 switch(optname) {
463 case SO_DEBUG:
464 if (val && !capable(CAP_NET_ADMIN)) {
465 ret = -EACCES;
466 }
467 else if (valbool)
468 sock_set_flag(sk, SOCK_DBG);
469 else
470 sock_reset_flag(sk, SOCK_DBG);
471 break;
472 case SO_REUSEADDR:
473 sk->sk_reuse = valbool;
474 break;
475 case SO_TYPE:
476 case SO_ERROR:
477 ret = -ENOPROTOOPT;
478 break;
479 case SO_DONTROUTE:
480 if (valbool)
481 sock_set_flag(sk, SOCK_LOCALROUTE);
482 else
483 sock_reset_flag(sk, SOCK_LOCALROUTE);
484 break;
485 case SO_BROADCAST:
486 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
487 break;
488 case SO_SNDBUF:
489 /* Don't error on this BSD doesn't and if you think
490 about it this is right. Otherwise apps have to
491 play 'guess the biggest size' games. RCVBUF/SNDBUF
492 are treated in BSD as hints */
493
494 if (val > sysctl_wmem_max)
495 val = sysctl_wmem_max;
b0573dea 496set_sndbuf:
e71a4783
SH		 497 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
498 if ((val * 2) < SOCK_MIN_SNDBUF)
499 sk->sk_sndbuf = SOCK_MIN_SNDBUF;
500 else
501 sk->sk_sndbuf = val * 2;
1da177e4 502
e71a4783
SH		 503 /*
504 * Wake up sending tasks if we
505 * upped the value.
506 */
507 sk->sk_write_space(sk);
508 break;
1da177e4 509
e71a4783
SH		 510 case SO_SNDBUFFORCE:
511 if (!capable(CAP_NET_ADMIN)) {
512 ret = -EPERM;
513 break;
514 }
515 goto set_sndbuf;
b0573dea 516
e71a4783
SH		 517 case SO_RCVBUF:
518 /* Don't error on this BSD doesn't and if you think
519 about it this is right. Otherwise apps have to
520 play 'guess the biggest size' games. RCVBUF/SNDBUF
521 are treated in BSD as hints */
4ec93edb 522
e71a4783
SH		 523 if (val > sysctl_rmem_max)
524 val = sysctl_rmem_max;
b0573dea 525set_rcvbuf:
e71a4783
SH		 526 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
527 /*
528 * We double it on the way in to account for
529 * "struct sk_buff" etc. overhead. Applications
530 * assume that the SO_RCVBUF setting they make will
531 * allow that much actual data to be received on that
532 * socket.
533 *
534 * Applications are unaware that "struct sk_buff" and
535 * other overheads allocate from the receive buffer
536 * during socket buffer allocation.
537 *
538 * And after considering the possible alternatives,
539 * returning the value we actually used in getsockopt
540 * is the most desirable behavior.
541 */
542 if ((val * 2) < SOCK_MIN_RCVBUF)
543 sk->sk_rcvbuf = SOCK_MIN_RCVBUF;
544 else
545 sk->sk_rcvbuf = val * 2;
546 break;
547
548 case SO_RCVBUFFORCE:
549 if (!capable(CAP_NET_ADMIN)) {
550 ret = -EPERM;
1da177e4 551 break;
e71a4783
SH		 552 }
553 goto set_rcvbuf;
1da177e4 554
e71a4783 555 case SO_KEEPALIVE:
1da177e4 556#ifdef CONFIG_INET
e71a4783
SH		 557 if (sk->sk_protocol == IPPROTO_TCP)
558 tcp_set_keepalive(sk, valbool);
1da177e4 559#endif
e71a4783
SH		 560 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
561 break;
562
563 case SO_OOBINLINE:
564 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
565 break;
566
567 case SO_NO_CHECK:
568 sk->sk_no_check = valbool;
569 break;
570
571 case SO_PRIORITY:
572 if ((val >= 0 && val <= 6) || capable(CAP_NET_ADMIN))
573 sk->sk_priority = val;
574 else
575 ret = -EPERM;
576 break;
577
578 case SO_LINGER:
579 if (optlen < sizeof(ling)) {
580 ret = -EINVAL; /* 1003.1g */
1da177e4 581 break;
e71a4783
SH		 582 }
583 if (copy_from_user(&ling,optval,sizeof(ling))) {
584 ret = -EFAULT;
1da177e4 585 break;
e71a4783
SH		 586 }
587 if (!ling.l_onoff)
588 sock_reset_flag(sk, SOCK_LINGER);
589 else {
1da177e4 590#if (BITS_PER_LONG == 32)
e71a4783
SH		 591 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
592 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 593 else
e71a4783
SH		 594#endif
595 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
596 sock_set_flag(sk, SOCK_LINGER);
597 }
598 break;
599
600 case SO_BSDCOMPAT:
601 sock_warn_obsolete_bsdism("setsockopt");
602 break;
603
604 case SO_PASSCRED:
605 if (valbool)
606 set_bit(SOCK_PASSCRED, &sock->flags);
607 else
608 clear_bit(SOCK_PASSCRED, &sock->flags);
609 break;
610
611 case SO_TIMESTAMP:
92f37fd2 612 case SO_TIMESTAMPNS:
e71a4783 613 if (valbool) {
92f37fd2
ED		 614 if (optname == SO_TIMESTAMP)
615 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
616 else
617 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 618 sock_set_flag(sk, SOCK_RCVTSTAMP);
619 sock_enable_timestamp(sk);
92f37fd2 620 } else {
e71a4783 621 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 622 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
623 }
e71a4783
SH		 624 break;
625
626 case SO_RCVLOWAT:
627 if (val < 0)
628 val = INT_MAX;
629 sk->sk_rcvlowat = val ? : 1;
630 break;
631
632 case SO_RCVTIMEO:
633 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
634 break;
635
636 case SO_SNDTIMEO:
637 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
638 break;
1da177e4 639
e71a4783
SH		 640 case SO_ATTACH_FILTER:
641 ret = -EINVAL;
642 if (optlen == sizeof(struct sock_fprog)) {
643 struct sock_fprog fprog;
1da177e4 644
e71a4783
SH		 645 ret = -EFAULT;
646 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 647 break;
e71a4783
SH		 648
649 ret = sk_attach_filter(&fprog, sk);
650 }
651 break;
652
653 case SO_DETACH_FILTER:
55b33325 654 ret = sk_detach_filter(sk);
e71a4783 655 break;
1da177e4 656
e71a4783
SH		 657 case SO_PASSSEC:
658 if (valbool)
659 set_bit(SOCK_PASSSEC, &sock->flags);
660 else
661 clear_bit(SOCK_PASSSEC, &sock->flags);
662 break;
877ce7c1 663
1da177e4
LT		 664 /* We implement the SO_SNDLOWAT etc to
665 not be settable (1003.1g 5.3) */
e71a4783
SH		 666 default:
667 ret = -ENOPROTOOPT;
668 break;
4ec93edb 669 }
1da177e4
LT		 670 release_sock(sk);
671 return ret;
672}
673
674
675int sock_getsockopt(struct socket *sock, int level, int optname,
676 char __user *optval, int __user *optlen)
677{
678 struct sock *sk = sock->sk;
4ec93edb 679
e71a4783 680 union {
4ec93edb
YH		 681 int val;
682 struct linger ling;
1da177e4
LT		 683 struct timeval tm;
684 } v;
4ec93edb 685
1da177e4
LT		 686 unsigned int lv = sizeof(int);
687 int len;
4ec93edb 688
e71a4783 689 if (get_user(len, optlen))
4ec93edb 690 return -EFAULT;
e71a4783 691 if (len < 0)
1da177e4 692 return -EINVAL;
4ec93edb 693
e71a4783
SH		 694 switch(optname) {
695 case SO_DEBUG:
696 v.val = sock_flag(sk, SOCK_DBG);
697 break;
698
699 case SO_DONTROUTE:
700 v.val = sock_flag(sk, SOCK_LOCALROUTE);
701 break;
702
703 case SO_BROADCAST:
704 v.val = !!sock_flag(sk, SOCK_BROADCAST);
705 break;
706
707 case SO_SNDBUF:
708 v.val = sk->sk_sndbuf;
709 break;
710
711 case SO_RCVBUF:
712 v.val = sk->sk_rcvbuf;
713 break;
714
715 case SO_REUSEADDR:
716 v.val = sk->sk_reuse;
717 break;
718
719 case SO_KEEPALIVE:
720 v.val = !!sock_flag(sk, SOCK_KEEPOPEN);
721 break;
722
723 case SO_TYPE:
724 v.val = sk->sk_type;
725 break;
726
727 case SO_ERROR:
728 v.val = -sock_error(sk);
729 if (v.val==0)
730 v.val = xchg(&sk->sk_err_soft, 0);
731 break;
732
733 case SO_OOBINLINE:
734 v.val = !!sock_flag(sk, SOCK_URGINLINE);
735 break;
736
737 case SO_NO_CHECK:
738 v.val = sk->sk_no_check;
739 break;
740
741 case SO_PRIORITY:
742 v.val = sk->sk_priority;
743 break;
744
745 case SO_LINGER:
746 lv = sizeof(v.ling);
747 v.ling.l_onoff = !!sock_flag(sk, SOCK_LINGER);
748 v.ling.l_linger = sk->sk_lingertime / HZ;
749 break;
750
751 case SO_BSDCOMPAT:
752 sock_warn_obsolete_bsdism("getsockopt");
753 break;
754
755 case SO_TIMESTAMP:
92f37fd2
ED		 756 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
757 !sock_flag(sk, SOCK_RCVTSTAMPNS);
758 break;
759
760 case SO_TIMESTAMPNS:
761 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 762 break;
763
764 case SO_RCVTIMEO:
765 lv=sizeof(struct timeval);
766 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
767 v.tm.tv_sec = 0;
768 v.tm.tv_usec = 0;
769 } else {
770 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
771 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
772 }
773 break;
774
775 case SO_SNDTIMEO:
776 lv=sizeof(struct timeval);
777 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
778 v.tm.tv_sec = 0;
779 v.tm.tv_usec = 0;
780 } else {
781 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
782 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
783 }
784 break;
1da177e4 785
e71a4783
SH		 786 case SO_RCVLOWAT:
787 v.val = sk->sk_rcvlowat;
788 break;
1da177e4 789
e71a4783
SH		 790 case SO_SNDLOWAT:
791 v.val=1;
792 break;
1da177e4 793
e71a4783
SH		 794 case SO_PASSCRED:
795 v.val = test_bit(SOCK_PASSCRED, &sock->flags) ? 1 : 0;
796 break;
1da177e4 797
e71a4783
SH		 798 case SO_PEERCRED:
799 if (len > sizeof(sk->sk_peercred))
800 len = sizeof(sk->sk_peercred);
801 if (copy_to_user(optval, &sk->sk_peercred, len))
802 return -EFAULT;
803 goto lenout;
1da177e4 804
e71a4783
SH		 805 case SO_PEERNAME:
806 {
807 char address[128];
808
809 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
810 return -ENOTCONN;
811 if (lv < len)
812 return -EINVAL;
813 if (copy_to_user(optval, address, len))
814 return -EFAULT;
815 goto lenout;
816 }
1da177e4 817
e71a4783
SH		 818 /* Dubious BSD thing... Probably nobody even uses it, but
819 * the UNIX standard wants it for whatever reason... -DaveM
820 */
821 case SO_ACCEPTCONN:
822 v.val = sk->sk_state == TCP_LISTEN;
823 break;
1da177e4 824
e71a4783
SH		 825 case SO_PASSSEC:
826 v.val = test_bit(SOCK_PASSSEC, &sock->flags) ? 1 : 0;
827 break;
877ce7c1 828
e71a4783
SH		 829 case SO_PEERSEC:
830 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 831
e71a4783
SH		 832 default:
833 return -ENOPROTOOPT;
1da177e4 834 }
e71a4783 835
1da177e4
LT		 836 if (len > lv)
837 len = lv;
838 if (copy_to_user(optval, &v, len))
839 return -EFAULT;
840lenout:
4ec93edb
YH		 841 if (put_user(len, optlen))
842 return -EFAULT;
843 return 0;
1da177e4
LT		 844}
845
a5b5bb9a
IM		 846/*
847 * Initialize an sk_lock.
848 *
849 * (We also register the sk_lock with the lock validator.)
850 */
b6f99a21 851static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 852{
ed07536e
PZ		 853 sock_lock_init_class_and_name(sk,
854 af_family_slock_key_strings[sk->sk_family],
855 af_family_slock_keys + sk->sk_family,
856 af_family_key_strings[sk->sk_family],
857 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 858}
859
f1a6c4da
PE		 860static void sock_copy(struct sock *nsk, const struct sock *osk)
861{
862#ifdef CONFIG_SECURITY_NETWORK
863 void *sptr = nsk->sk_security;
864#endif
865
866 memcpy(nsk, osk, osk->sk_prot->obj_size);
f1a6c4da
PE		 867#ifdef CONFIG_SECURITY_NETWORK
868 nsk->sk_security = sptr;
869 security_sk_clone(osk, nsk);
870#endif
871}
872
2e4afe7b
PE		 873static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
874 int family)
c308c1b2
PE		 875{
876 struct sock *sk;
877 struct kmem_cache *slab;
878
879 slab = prot->slab;
880 if (slab != NULL)
881 sk = kmem_cache_alloc(slab, priority);
882 else
883 sk = kmalloc(prot->obj_size, priority);
884
2e4afe7b
PE		 885 if (sk != NULL) {
886 if (security_sk_alloc(sk, family, priority))
887 goto out_free;
888
889 if (!try_module_get(prot->owner))
890 goto out_free_sec;
891 }
892
c308c1b2 893 return sk;
2e4afe7b
PE		 894
895out_free_sec:
896 security_sk_free(sk);
897out_free:
898 if (slab != NULL)
899 kmem_cache_free(slab, sk);
900 else
901 kfree(sk);
902 return NULL;
c308c1b2
PE		 903}
904
905static void sk_prot_free(struct proto *prot, struct sock *sk)
906{
907 struct kmem_cache *slab;
2e4afe7b 908 struct module *owner;
c308c1b2 909
2e4afe7b 910 owner = prot->owner;
c308c1b2 911 slab = prot->slab;
2e4afe7b
PE		 912
913 security_sk_free(sk);
c308c1b2
PE		 914 if (slab != NULL)
915 kmem_cache_free(slab, sk);
916 else
917 kfree(sk);
2e4afe7b 918 module_put(owner);
c308c1b2
PE		 919}
920
1da177e4
LT		 921/**
922 * sk_alloc - All socket objects are allocated here
c4ea43c5 923 * @net: the applicable net namespace
4dc3b16b
PP		 924 * @family: protocol family
925 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
926 * @prot: struct proto associated with this new sock instance
927 * @zero_it: if we should zero the newly allocated sock
1da177e4 928 */
1b8d7ae4 929struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
86a76caf 930 struct proto *prot, int zero_it)
1da177e4 931{
c308c1b2 932 struct sock *sk;
1da177e4 933
3f0666ee
PE		 934 if (zero_it)
935 priority |= __GFP_ZERO;
936
2e4afe7b 937 sk = sk_prot_alloc(prot, priority, family);
1da177e4
LT		 938 if (sk) {
939 if (zero_it) {
1da177e4 940 sk->sk_family = family;
476e19cf
ACM		 941 /*
942 * See comment in struct sock definition to understand
943 * why we need sk_prot_creator -acme
944 */
945 sk->sk_prot = sk->sk_prot_creator = prot;
1da177e4 946 sock_lock_init(sk);
1b8d7ae4 947 sk->sk_net = get_net(net);
1da177e4 948 }
1da177e4 949 }
a79af59e 950
2e4afe7b 951 return sk;
1da177e4
LT		 952}
953
954void sk_free(struct sock *sk)
955{
956 struct sk_filter *filter;
1da177e4
LT		 957
958 if (sk->sk_destruct)
959 sk->sk_destruct(sk);
960
fda9ef5d 961 filter = rcu_dereference(sk->sk_filter);
1da177e4 962 if (filter) {
309dd5fc 963 sk_filter_uncharge(sk, filter);
fda9ef5d 964 rcu_assign_pointer(sk->sk_filter, NULL);
1da177e4
LT		 965 }
966
967 sock_disable_timestamp(sk);
968
969 if (atomic_read(&sk->sk_omem_alloc))
970 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n",
971 __FUNCTION__, atomic_read(&sk->sk_omem_alloc));
972
1b8d7ae4 973 put_net(sk->sk_net);
c308c1b2 974 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4
LT		 975}
976
dd0fc66f 977struct sock *sk_clone(const struct sock *sk, const gfp_t priority)
87d11ceb 978{
1b8d7ae4 979 struct sock *newsk = sk_alloc(sk->sk_net, sk->sk_family, priority, sk->sk_prot, 0);
87d11ceb
ACM		 980
981 if (newsk != NULL) {
982 struct sk_filter *filter;
983
892c141e 984 sock_copy(newsk, sk);
87d11ceb
ACM		 985
986 /* SANITY */
1e2e6b89 987 get_net(newsk->sk_net);
87d11ceb
ACM		 988 sk_node_init(&newsk->sk_node);
989 sock_lock_init(newsk);
990 bh_lock_sock(newsk);
fa438ccf 991 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
87d11ceb
ACM		 992
993 atomic_set(&newsk->sk_rmem_alloc, 0);
994 atomic_set(&newsk->sk_wmem_alloc, 0);
995 atomic_set(&newsk->sk_omem_alloc, 0);
996 skb_queue_head_init(&newsk->sk_receive_queue);
997 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 998#ifdef CONFIG_NET_DMA
999 skb_queue_head_init(&newsk->sk_async_wait_queue);
1000#endif
87d11ceb
ACM		 1001
1002 rwlock_init(&newsk->sk_dst_lock);
1003 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1004 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1005 af_callback_keys + newsk->sk_family,
1006 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1007
1008 newsk->sk_dst_cache = NULL;
1009 newsk->sk_wmem_queued = 0;
1010 newsk->sk_forward_alloc = 0;
1011 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1012 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1013
1014 sock_reset_flag(newsk, SOCK_DONE);
1015 skb_queue_head_init(&newsk->sk_error_queue);
1016
1017 filter = newsk->sk_filter;
1018 if (filter != NULL)
1019 sk_filter_charge(newsk, filter);
1020
1021 if (unlikely(xfrm_sk_clone_policy(newsk))) {
1022 /* It is still raw copy of parent, so invalidate
1023 * destructor and make plain sk_free() */
1024 newsk->sk_destruct = NULL;
1025 sk_free(newsk);
1026 newsk = NULL;
1027 goto out;
1028 }
1029
1030 newsk->sk_err = 0;
1031 newsk->sk_priority = 0;
1032 atomic_set(&newsk->sk_refcnt, 2);
1033
1034 /*
1035 * Increment the counter in the same struct proto as the master
1036 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1037 * is the same as sk->sk_prot->socks, as this field was copied
1038 * with memcpy).
1039 *
1040 * This _changes_ the previous behaviour, where
1041 * tcp_create_openreq_child always was incrementing the
1042 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1043 * to be taken into account in all callers. -acme
1044 */
1045 sk_refcnt_debug_inc(newsk);
1046 newsk->sk_socket = NULL;
1047 newsk->sk_sleep = NULL;
1048
1049 if (newsk->sk_prot->sockets_allocated)
1050 atomic_inc(newsk->sk_prot->sockets_allocated);
1051 }
1052out:
1053 return newsk;
1054}
1055
1056EXPORT_SYMBOL_GPL(sk_clone);
1057
9958089a
AK		 1058void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1059{
1060 __sk_dst_set(sk, dst);
1061 sk->sk_route_caps = dst->dev->features;
1062 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1063 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
9958089a
AK		 1064 if (sk_can_gso(sk)) {
1065 if (dst->header_len)
1066 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
1067 else
1068 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
1069 }
1070}
1071EXPORT_SYMBOL_GPL(sk_setup_caps);
1072
1da177e4
LT		 1073void __init sk_init(void)
1074{
1075 if (num_physpages <= 4096) {
1076 sysctl_wmem_max = 32767;
1077 sysctl_rmem_max = 32767;
1078 sysctl_wmem_default = 32767;
1079 sysctl_rmem_default = 32767;
1080 } else if (num_physpages >= 131072) {
1081 sysctl_wmem_max = 131071;
1082 sysctl_rmem_max = 131071;
1083 }
1084}
1085
1086/*
1087 * Simple resource managers for sockets.
1088 */
1089
1090
4ec93edb
YH		 1091/*
1092 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1093 */
1094void sock_wfree(struct sk_buff *skb)
1095{
1096 struct sock *sk = skb->sk;
1097
1098 /* In case it might be waiting for more memory. */
1099 atomic_sub(skb->truesize, &sk->sk_wmem_alloc);
1100 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE))
1101 sk->sk_write_space(sk);
1102 sock_put(sk);
1103}
1104
4ec93edb
YH		 1105/*
1106 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1107 */
1108void sock_rfree(struct sk_buff *skb)
1109{
1110 struct sock *sk = skb->sk;
1111
1112 atomic_sub(skb->truesize, &sk->sk_rmem_alloc);
1113}
1114
1115
1116int sock_i_uid(struct sock *sk)
1117{
1118 int uid;
1119
1120 read_lock(&sk->sk_callback_lock);
1121 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
1122 read_unlock(&sk->sk_callback_lock);
1123 return uid;
1124}
1125
1126unsigned long sock_i_ino(struct sock *sk)
1127{
1128 unsigned long ino;
1129
1130 read_lock(&sk->sk_callback_lock);
1131 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
1132 read_unlock(&sk->sk_callback_lock);
1133 return ino;
1134}
1135
1136/*
1137 * Allocate a skb from the socket's send buffer.
1138 */
86a76caf 1139struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1140 gfp_t priority)
1da177e4
LT		 1141{
1142 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
1143 struct sk_buff * skb = alloc_skb(size, priority);
1144 if (skb) {
1145 skb_set_owner_w(skb, sk);
1146 return skb;
1147 }
1148 }
1149 return NULL;
1150}
1151
1152/*
1153 * Allocate a skb from the socket's receive buffer.
4ec93edb 1154 */
86a76caf 1155struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1156 gfp_t priority)
1da177e4
LT		 1157{
1158 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1159 struct sk_buff *skb = alloc_skb(size, priority);
1160 if (skb) {
1161 skb_set_owner_r(skb, sk);
1162 return skb;
1163 }
1164 }
1165 return NULL;
1166}
1167
4ec93edb 1168/*
1da177e4 1169 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1170 */
dd0fc66f 1171void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4
LT		 1172{
1173 if ((unsigned)size <= sysctl_optmem_max &&
1174 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1175 void *mem;
1176 /* First do the add, to avoid the race if kmalloc
4ec93edb 1177 * might sleep.
1da177e4
LT		 1178 */
1179 atomic_add(size, &sk->sk_omem_alloc);
1180 mem = kmalloc(size, priority);
1181 if (mem)
1182 return mem;
1183 atomic_sub(size, &sk->sk_omem_alloc);
1184 }
1185 return NULL;
1186}
1187
1188/*
1189 * Free an option memory block.
1190 */
1191void sock_kfree_s(struct sock *sk, void *mem, int size)
1192{
1193 kfree(mem);
1194 atomic_sub(size, &sk->sk_omem_alloc);
1195}
1196
1197/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1198 I think, these locks should be removed for datagram sockets.
1199 */
1200static long sock_wait_for_wmem(struct sock * sk, long timeo)
1201{
1202 DEFINE_WAIT(wait);
1203
1204 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1205 for (;;) {
1206 if (!timeo)
1207 break;
1208 if (signal_pending(current))
1209 break;
1210 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1211 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1212 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1213 break;
1214 if (sk->sk_shutdown & SEND_SHUTDOWN)
1215 break;
1216 if (sk->sk_err)
1217 break;
1218 timeo = schedule_timeout(timeo);
1219 }
1220 finish_wait(sk->sk_sleep, &wait);
1221 return timeo;
1222}
1223
1224
1225/*
1226 * Generic send/receive buffer handlers
1227 */
1228
1229static struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1230 unsigned long header_len,
1231 unsigned long data_len,
1232 int noblock, int *errcode)
1233{
1234 struct sk_buff *skb;
7d877f3b 1235 gfp_t gfp_mask;
1da177e4
LT		 1236 long timeo;
1237 int err;
1238
1239 gfp_mask = sk->sk_allocation;
1240 if (gfp_mask & __GFP_WAIT)
1241 gfp_mask |= __GFP_REPEAT;
1242
1243 timeo = sock_sndtimeo(sk, noblock);
1244 while (1) {
1245 err = sock_error(sk);
1246 if (err != 0)
1247 goto failure;
1248
1249 err = -EPIPE;
1250 if (sk->sk_shutdown & SEND_SHUTDOWN)
1251 goto failure;
1252
1253 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1254 skb = alloc_skb(header_len, gfp_mask);
1da177e4
LT		 1255 if (skb) {
1256 int npages;
1257 int i;
1258
1259 /* No pages, we're done... */
1260 if (!data_len)
1261 break;
1262
1263 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1264 skb->truesize += data_len;
1265 skb_shinfo(skb)->nr_frags = npages;
1266 for (i = 0; i < npages; i++) {
1267 struct page *page;
1268 skb_frag_t *frag;
1269
1270 page = alloc_pages(sk->sk_allocation, 0);
1271 if (!page) {
1272 err = -ENOBUFS;
1273 skb_shinfo(skb)->nr_frags = i;
1274 kfree_skb(skb);
1275 goto failure;
1276 }
1277
1278 frag = &skb_shinfo(skb)->frags[i];
1279 frag->page = page;
1280 frag->page_offset = 0;
1281 frag->size = (data_len >= PAGE_SIZE ?
1282 PAGE_SIZE :
1283 data_len);
1284 data_len -= PAGE_SIZE;
1285 }
1286
1287 /* Full success... */
1288 break;
1289 }
1290 err = -ENOBUFS;
1291 goto failure;
1292 }
1293 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1294 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1295 err = -EAGAIN;
1296 if (!timeo)
1297 goto failure;
1298 if (signal_pending(current))
1299 goto interrupted;
1300 timeo = sock_wait_for_wmem(sk, timeo);
1301 }
1302
1303 skb_set_owner_w(skb, sk);
1304 return skb;
1305
1306interrupted:
1307 err = sock_intr_errno(timeo);
1308failure:
1309 *errcode = err;
1310 return NULL;
1311}
1312
4ec93edb 1313struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1314 int noblock, int *errcode)
1315{
1316 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1317}
1318
1319static void __lock_sock(struct sock *sk)
1320{
1321 DEFINE_WAIT(wait);
1322
e71a4783 1323 for (;;) {
1da177e4
LT		 1324 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1325 TASK_UNINTERRUPTIBLE);
1326 spin_unlock_bh(&sk->sk_lock.slock);
1327 schedule();
1328 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1329 if (!sock_owned_by_user(sk))
1da177e4
LT		 1330 break;
1331 }
1332 finish_wait(&sk->sk_lock.wq, &wait);
1333}
1334
1335static void __release_sock(struct sock *sk)
1336{
1337 struct sk_buff *skb = sk->sk_backlog.head;
1338
1339 do {
1340 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1341 bh_unlock_sock(sk);
1342
1343 do {
1344 struct sk_buff *next = skb->next;
1345
1346 skb->next = NULL;
1347 sk->sk_backlog_rcv(sk, skb);
1348
1349 /*
1350 * We are in process context here with softirqs
1351 * disabled, use cond_resched_softirq() to preempt.
1352 * This is safe to do because we've taken the backlog
1353 * queue private:
1354 */
1355 cond_resched_softirq();
1356
1357 skb = next;
1358 } while (skb != NULL);
1359
1360 bh_lock_sock(sk);
e71a4783 1361 } while ((skb = sk->sk_backlog.head) != NULL);
1da177e4
LT		 1362}
1363
1364/**
1365 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1366 * @sk: sock to wait on
1367 * @timeo: for how long
1da177e4
LT		 1368 *
1369 * Now socket state including sk->sk_err is changed only under lock,
1370 * hence we may omit checks after joining wait queue.
1371 * We check receive queue before schedule() only as optimization;
1372 * it is very likely that release_sock() added new data.
1373 */
1374int sk_wait_data(struct sock *sk, long *timeo)
1375{
1376 int rc;
1377 DEFINE_WAIT(wait);
1378
1379 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1380 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1381 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1382 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1383 finish_wait(sk->sk_sleep, &wait);
1384 return rc;
1385}
1386
1387EXPORT_SYMBOL(sk_wait_data);
1388
1389/*
1390 * Set of default routines for initialising struct proto_ops when
1391 * the protocol does not support a particular function. In certain
1392 * cases where it makes no sense for a protocol to have a "do nothing"
1393 * function, some default processing is provided.
1394 */
1395
1396int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1397{
1398 return -EOPNOTSUPP;
1399}
1400
4ec93edb 1401int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1402 int len, int flags)
1403{
1404 return -EOPNOTSUPP;
1405}
1406
1407int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1408{
1409 return -EOPNOTSUPP;
1410}
1411
1412int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1413{
1414 return -EOPNOTSUPP;
1415}
1416
4ec93edb 1417int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1418 int *len, int peer)
1419{
1420 return -EOPNOTSUPP;
1421}
1422
1423unsigned int sock_no_poll(struct file * file, struct socket *sock, poll_table *pt)
1424{
1425 return 0;
1426}
1427
1428int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1429{
1430 return -EOPNOTSUPP;
1431}
1432
1433int sock_no_listen(struct socket *sock, int backlog)
1434{
1435 return -EOPNOTSUPP;
1436}
1437
1438int sock_no_shutdown(struct socket *sock, int how)
1439{
1440 return -EOPNOTSUPP;
1441}
1442
1443int sock_no_setsockopt(struct socket *sock, int level, int optname,
1444 char __user *optval, int optlen)
1445{
1446 return -EOPNOTSUPP;
1447}
1448
1449int sock_no_getsockopt(struct socket *sock, int level, int optname,
1450 char __user *optval, int __user *optlen)
1451{
1452 return -EOPNOTSUPP;
1453}
1454
1455int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1456 size_t len)
1457{
1458 return -EOPNOTSUPP;
1459}
1460
1461int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1462 size_t len, int flags)
1463{
1464 return -EOPNOTSUPP;
1465}
1466
1467int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
1468{
1469 /* Mirror missing mmap method error code */
1470 return -ENODEV;
1471}
1472
1473ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
1474{
1475 ssize_t res;
1476 struct msghdr msg = {.msg_flags = flags};
1477 struct kvec iov;
1478 char *kaddr = kmap(page);
1479 iov.iov_base = kaddr + offset;
1480 iov.iov_len = size;
1481 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
1482 kunmap(page);
1483 return res;
1484}
1485
1486/*
1487 * Default Socket Callbacks
1488 */
1489
1490static void sock_def_wakeup(struct sock *sk)
1491{
1492 read_lock(&sk->sk_callback_lock);
1493 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1494 wake_up_interruptible_all(sk->sk_sleep);
1495 read_unlock(&sk->sk_callback_lock);
1496}
1497
1498static void sock_def_error_report(struct sock *sk)
1499{
1500 read_lock(&sk->sk_callback_lock);
1501 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1502 wake_up_interruptible(sk->sk_sleep);
4ec93edb 1503 sk_wake_async(sk,0,POLL_ERR);
1da177e4
LT		 1504 read_unlock(&sk->sk_callback_lock);
1505}
1506
1507static void sock_def_readable(struct sock *sk, int len)
1508{
1509 read_lock(&sk->sk_callback_lock);
1510 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1511 wake_up_interruptible(sk->sk_sleep);
1512 sk_wake_async(sk,1,POLL_IN);
1513 read_unlock(&sk->sk_callback_lock);
1514}
1515
1516static void sock_def_write_space(struct sock *sk)
1517{
1518 read_lock(&sk->sk_callback_lock);
1519
1520 /* Do not wake up a writer until he can make "significant"
1521 * progress. --DaveM
1522 */
e71a4783 1523 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
1da177e4
LT		 1524 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1525 wake_up_interruptible(sk->sk_sleep);
1526
1527 /* Should agree with poll, otherwise some programs break */
1528 if (sock_writeable(sk))
1529 sk_wake_async(sk, 2, POLL_OUT);
1530 }
1531
1532 read_unlock(&sk->sk_callback_lock);
1533}
1534
1535static void sock_def_destruct(struct sock *sk)
1536{
a51482bd 1537 kfree(sk->sk_protinfo);
1da177e4
LT		 1538}
1539
1540void sk_send_sigurg(struct sock *sk)
1541{
1542 if (sk->sk_socket && sk->sk_socket->file)
1543 if (send_sigurg(&sk->sk_socket->file->f_owner))
1544 sk_wake_async(sk, 3, POLL_PRI);
1545}
1546
1547void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1548 unsigned long expires)
1549{
1550 if (!mod_timer(timer, expires))
1551 sock_hold(sk);
1552}
1553
1554EXPORT_SYMBOL(sk_reset_timer);
1555
1556void sk_stop_timer(struct sock *sk, struct timer_list* timer)
1557{
1558 if (timer_pending(timer) && del_timer(timer))
1559 __sock_put(sk);
1560}
1561
1562EXPORT_SYMBOL(sk_stop_timer);
1563
1564void sock_init_data(struct socket *sock, struct sock *sk)
1565{
1566 skb_queue_head_init(&sk->sk_receive_queue);
1567 skb_queue_head_init(&sk->sk_write_queue);
1568 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 1569#ifdef CONFIG_NET_DMA
1570 skb_queue_head_init(&sk->sk_async_wait_queue);
1571#endif
1da177e4
LT		 1572
1573 sk->sk_send_head = NULL;
1574
1575 init_timer(&sk->sk_timer);
4ec93edb 1576
1da177e4
LT		 1577 sk->sk_allocation = GFP_KERNEL;
1578 sk->sk_rcvbuf = sysctl_rmem_default;
1579 sk->sk_sndbuf = sysctl_wmem_default;
1580 sk->sk_state = TCP_CLOSE;
1581 sk->sk_socket = sock;
1582
1583 sock_set_flag(sk, SOCK_ZAPPED);
1584
e71a4783 1585 if (sock) {
1da177e4
LT		 1586 sk->sk_type = sock->type;
1587 sk->sk_sleep = &sock->wait;
1588 sock->sk = sk;
1589 } else
1590 sk->sk_sleep = NULL;
1591
1592 rwlock_init(&sk->sk_dst_lock);
1593 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 1594 lockdep_set_class_and_name(&sk->sk_callback_lock,
1595 af_callback_keys + sk->sk_family,
1596 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 1597
1598 sk->sk_state_change = sock_def_wakeup;
1599 sk->sk_data_ready = sock_def_readable;
1600 sk->sk_write_space = sock_def_write_space;
1601 sk->sk_error_report = sock_def_error_report;
1602 sk->sk_destruct = sock_def_destruct;
1603
1604 sk->sk_sndmsg_page = NULL;
1605 sk->sk_sndmsg_off = 0;
1606
1607 sk->sk_peercred.pid = 0;
1608 sk->sk_peercred.uid = -1;
1609 sk->sk_peercred.gid = -1;
1610 sk->sk_write_pending = 0;
1611 sk->sk_rcvlowat = 1;
1612 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
1613 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
1614
b7aa0bf7 1615 sk->sk_stamp = ktime_set(-1L, -1L);
1da177e4
LT		 1616
1617 atomic_set(&sk->sk_refcnt, 1);
1618}
1619
fcc70d5f 1620void fastcall lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 1621{
1622 might_sleep();
a5b5bb9a 1623 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 1624 if (sk->sk_lock.owned)
1da177e4 1625 __lock_sock(sk);
d2e9117c 1626 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 1627 spin_unlock(&sk->sk_lock.slock);
1628 /*
1629 * The sk_lock has mutex_lock() semantics here:
1630 */
fcc70d5f 1631 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 1632 local_bh_enable();
1da177e4
LT		 1633}
1634
fcc70d5f 1635EXPORT_SYMBOL(lock_sock_nested);
1da177e4
LT		 1636
1637void fastcall release_sock(struct sock *sk)
1638{
a5b5bb9a
IM		 1639 /*
1640 * The sk_lock has mutex_unlock() semantics:
1641 */
1642 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1643
1644 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 1645 if (sk->sk_backlog.tail)
1646 __release_sock(sk);
d2e9117c 1647 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 1648 if (waitqueue_active(&sk->sk_lock.wq))
1649 wake_up(&sk->sk_lock.wq);
1650 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 1651}
1652EXPORT_SYMBOL(release_sock);
1653
1654int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 1655{
b7aa0bf7 1656 struct timeval tv;
1da177e4
LT		 1657 if (!sock_flag(sk, SOCK_TIMESTAMP))
1658 sock_enable_timestamp(sk);
b7aa0bf7
ED		 1659 tv = ktime_to_timeval(sk->sk_stamp);
1660 if (tv.tv_sec == -1)
1da177e4 1661 return -ENOENT;
b7aa0bf7
ED		 1662 if (tv.tv_sec == 0) {
1663 sk->sk_stamp = ktime_get_real();
1664 tv = ktime_to_timeval(sk->sk_stamp);
1665 }
1666 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 1667}
1da177e4
LT		 1668EXPORT_SYMBOL(sock_get_timestamp);
1669
ae40eb1e
ED		 1670int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
1671{
1672 struct timespec ts;
1673 if (!sock_flag(sk, SOCK_TIMESTAMP))
1674 sock_enable_timestamp(sk);
1675 ts = ktime_to_timespec(sk->sk_stamp);
1676 if (ts.tv_sec == -1)
1677 return -ENOENT;
1678 if (ts.tv_sec == 0) {
1679 sk->sk_stamp = ktime_get_real();
1680 ts = ktime_to_timespec(sk->sk_stamp);
1681 }
1682 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
1683}
1684EXPORT_SYMBOL(sock_get_timestampns);
1685
1da177e4 1686void sock_enable_timestamp(struct sock *sk)
4ec93edb
YH		 1687{
1688 if (!sock_flag(sk, SOCK_TIMESTAMP)) {
1da177e4
LT		 1689 sock_set_flag(sk, SOCK_TIMESTAMP);
1690 net_enable_timestamp();
1691 }
1692}
1da177e4
LT		 1693
1694/*
1695 * Get a socket option on an socket.
1696 *
1697 * FIX: POSIX 1003.1g is very ambiguous here. It states that
1698 * asynchronous errors should be reported by getsockopt. We assume
1699 * this means if you specify SO_ERROR (otherwise whats the point of it).
1700 */
1701int sock_common_getsockopt(struct socket *sock, int level, int optname,
1702 char __user *optval, int __user *optlen)
1703{
1704 struct sock *sk = sock->sk;
1705
1706 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
1707}
1708
1709EXPORT_SYMBOL(sock_common_getsockopt);
1710
3fdadf7d 1711#ifdef CONFIG_COMPAT
543d9cfe
ACM		 1712int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
1713 char __user *optval, int __user *optlen)
3fdadf7d
DM		 1714{
1715 struct sock *sk = sock->sk;
1716
1e51f951 1717 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 1718 return sk->sk_prot->compat_getsockopt(sk, level, optname,
1719 optval, optlen);
3fdadf7d
DM		 1720 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
1721}
1722EXPORT_SYMBOL(compat_sock_common_getsockopt);
1723#endif
1724
1da177e4
LT		 1725int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1726 struct msghdr *msg, size_t size, int flags)
1727{
1728 struct sock *sk = sock->sk;
1729 int addr_len = 0;
1730 int err;
1731
1732 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
1733 flags & ~MSG_DONTWAIT, &addr_len);
1734 if (err >= 0)
1735 msg->msg_namelen = addr_len;
1736 return err;
1737}
1738
1739EXPORT_SYMBOL(sock_common_recvmsg);
1740
1741/*
1742 * Set socket options on an inet socket.
1743 */
1744int sock_common_setsockopt(struct socket *sock, int level, int optname,
1745 char __user *optval, int optlen)
1746{
1747 struct sock *sk = sock->sk;
1748
1749 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
1750}
1751
1752EXPORT_SYMBOL(sock_common_setsockopt);
1753
3fdadf7d 1754#ifdef CONFIG_COMPAT
543d9cfe
ACM		 1755int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
1756 char __user *optval, int optlen)
3fdadf7d
DM		 1757{
1758 struct sock *sk = sock->sk;
1759
543d9cfe
ACM		 1760 if (sk->sk_prot->compat_setsockopt != NULL)
1761 return sk->sk_prot->compat_setsockopt(sk, level, optname,
1762 optval, optlen);
3fdadf7d
DM		 1763 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
1764}
1765EXPORT_SYMBOL(compat_sock_common_setsockopt);
1766#endif
1767
1da177e4
LT		 1768void sk_common_release(struct sock *sk)
1769{
1770 if (sk->sk_prot->destroy)
1771 sk->sk_prot->destroy(sk);
1772
1773 /*
1774 * Observation: when sock_common_release is called, processes have
1775 * no access to socket. But net still has.
1776 * Step one, detach it from networking:
1777 *
1778 * A. Remove from hash tables.
1779 */
1780
1781 sk->sk_prot->unhash(sk);
1782
1783 /*
1784 * In this point socket cannot receive new packets, but it is possible
1785 * that some packets are in flight because some CPU runs receiver and
1786 * did hash table lookup before we unhashed socket. They will achieve
1787 * receive queue and will be purged by socket destructor.
1788 *
1789 * Also we still have packets pending on receive queue and probably,
1790 * our own packets waiting in device queues. sock_destroy will drain
1791 * receive queue, but transmitted packets will delay socket destruction
1792 * until the last reference will be released.
1793 */
1794
1795 sock_orphan(sk);
1796
1797 xfrm_sk_free_policy(sk);
1798
e6848976 1799 sk_refcnt_debug_release(sk);
1da177e4
LT		 1800 sock_put(sk);
1801}
1802
1803EXPORT_SYMBOL(sk_common_release);
1804
1805static DEFINE_RWLOCK(proto_list_lock);
1806static LIST_HEAD(proto_list);
1807
1808int proto_register(struct proto *prot, int alloc_slab)
1809{
8feaf0c0
ACM		 1810 char *request_sock_slab_name = NULL;
1811 char *timewait_sock_slab_name;
1da177e4
LT		 1812 int rc = -ENOBUFS;
1813
1da177e4
LT		 1814 if (alloc_slab) {
1815 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
20c2df83 1816 SLAB_HWCACHE_ALIGN, NULL);
1da177e4
LT		 1817
1818 if (prot->slab == NULL) {
1819 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n",
1820 prot->name);
2a278051 1821 goto out;
1da177e4 1822 }
2e6599cb
ACM		 1823
1824 if (prot->rsk_prot != NULL) {
1825 static const char mask[] = "request_sock_%s";
1826
1827 request_sock_slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL);
1828 if (request_sock_slab_name == NULL)
1829 goto out_free_sock_slab;
1830
1831 sprintf(request_sock_slab_name, mask, prot->name);
1832 prot->rsk_prot->slab = kmem_cache_create(request_sock_slab_name,
1833 prot->rsk_prot->obj_size, 0,
20c2df83 1834 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 1835
1836 if (prot->rsk_prot->slab == NULL) {
1837 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n",
1838 prot->name);
1839 goto out_free_request_sock_slab_name;
1840 }
1841 }
8feaf0c0 1842
6d6ee43e 1843 if (prot->twsk_prot != NULL) {
8feaf0c0
ACM		 1844 static const char mask[] = "tw_sock_%s";
1845
1846 timewait_sock_slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL);
1847
1848 if (timewait_sock_slab_name == NULL)
1849 goto out_free_request_sock_slab;
1850
1851 sprintf(timewait_sock_slab_name, mask, prot->name);
6d6ee43e
ACM		 1852 prot->twsk_prot->twsk_slab =
1853 kmem_cache_create(timewait_sock_slab_name,
1854 prot->twsk_prot->twsk_obj_size,
1855 0, SLAB_HWCACHE_ALIGN,
20c2df83 1856 NULL);
6d6ee43e 1857 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 1858 goto out_free_timewait_sock_slab_name;
1859 }
1da177e4
LT		 1860 }
1861
2a278051 1862 write_lock(&proto_list_lock);
1da177e4 1863 list_add(&prot->node, &proto_list);
1da177e4 1864 write_unlock(&proto_list_lock);
2a278051
ACM		 1865 rc = 0;
1866out:
1da177e4 1867 return rc;
8feaf0c0
ACM		 1868out_free_timewait_sock_slab_name:
1869 kfree(timewait_sock_slab_name);
1870out_free_request_sock_slab:
1871 if (prot->rsk_prot && prot->rsk_prot->slab) {
1872 kmem_cache_destroy(prot->rsk_prot->slab);
1873 prot->rsk_prot->slab = NULL;
1874 }
2e6599cb
ACM		 1875out_free_request_sock_slab_name:
1876 kfree(request_sock_slab_name);
1877out_free_sock_slab:
1878 kmem_cache_destroy(prot->slab);
1879 prot->slab = NULL;
1880 goto out;
1da177e4
LT		 1881}
1882
1883EXPORT_SYMBOL(proto_register);
1884
1885void proto_unregister(struct proto *prot)
1886{
1887 write_lock(&proto_list_lock);
0a3f4358
PM		 1888 list_del(&prot->node);
1889 write_unlock(&proto_list_lock);
1da177e4
LT		 1890
1891 if (prot->slab != NULL) {
1892 kmem_cache_destroy(prot->slab);
1893 prot->slab = NULL;
1894 }
1895
2e6599cb
ACM		 1896 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
1897 const char *name = kmem_cache_name(prot->rsk_prot->slab);
1898
1899 kmem_cache_destroy(prot->rsk_prot->slab);
1900 kfree(name);
1901 prot->rsk_prot->slab = NULL;
1902 }
1903
6d6ee43e
ACM		 1904 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
1905 const char *name = kmem_cache_name(prot->twsk_prot->twsk_slab);
8feaf0c0 1906
6d6ee43e 1907 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
8feaf0c0 1908 kfree(name);
6d6ee43e 1909 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 1910 }
1da177e4
LT		 1911}
1912
1913EXPORT_SYMBOL(proto_unregister);
1914
1915#ifdef CONFIG_PROC_FS
1da177e4
LT		 1916static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
1917{
1918 read_lock(&proto_list_lock);
60f0438a 1919 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 1920}
1921
1922static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1923{
60f0438a 1924 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 1925}
1926
1927static void proto_seq_stop(struct seq_file *seq, void *v)
1928{
1929 read_unlock(&proto_list_lock);
1930}
1931
1932static char proto_method_implemented(const void *method)
1933{
1934 return method == NULL ? 'n' : 'y';
1935}
1936
1937static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
1938{
1939 seq_printf(seq, "%-9s %4u %6d %6d %-3s %6u %-3s %-10s "
1940 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
1941 proto->name,
1942 proto->obj_size,
1943 proto->sockets_allocated != NULL ? atomic_read(proto->sockets_allocated) : -1,
1944 proto->memory_allocated != NULL ? atomic_read(proto->memory_allocated) : -1,
1945 proto->memory_pressure != NULL ? *proto->memory_pressure ? "yes" : "no" : "NI",
1946 proto->max_header,
1947 proto->slab == NULL ? "no" : "yes",
1948 module_name(proto->owner),
1949 proto_method_implemented(proto->close),
1950 proto_method_implemented(proto->connect),
1951 proto_method_implemented(proto->disconnect),
1952 proto_method_implemented(proto->accept),
1953 proto_method_implemented(proto->ioctl),
1954 proto_method_implemented(proto->init),
1955 proto_method_implemented(proto->destroy),
1956 proto_method_implemented(proto->shutdown),
1957 proto_method_implemented(proto->setsockopt),
1958 proto_method_implemented(proto->getsockopt),
1959 proto_method_implemented(proto->sendmsg),
1960 proto_method_implemented(proto->recvmsg),
1961 proto_method_implemented(proto->sendpage),
1962 proto_method_implemented(proto->bind),
1963 proto_method_implemented(proto->backlog_rcv),
1964 proto_method_implemented(proto->hash),
1965 proto_method_implemented(proto->unhash),
1966 proto_method_implemented(proto->get_port),
1967 proto_method_implemented(proto->enter_memory_pressure));
1968}
1969
1970static int proto_seq_show(struct seq_file *seq, void *v)
1971{
60f0438a 1972 if (v == &proto_list)
1da177e4
LT		 1973 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
1974 "protocol",
1975 "size",
1976 "sockets",
1977 "memory",
1978 "press",
1979 "maxhdr",
1980 "slab",
1981 "module",
1982 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
1983 else
60f0438a 1984 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 1985 return 0;
1986}
1987
f690808e 1988static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 1989 .start = proto_seq_start,
1990 .next = proto_seq_next,
1991 .stop = proto_seq_stop,
1992 .show = proto_seq_show,
1993};
1994
1995static int proto_seq_open(struct inode *inode, struct file *file)
1996{
1997 return seq_open(file, &proto_seq_ops);
1998}
1999
9a32144e 2000static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2001 .owner = THIS_MODULE,
2002 .open = proto_seq_open,
2003 .read = seq_read,
2004 .llseek = seq_lseek,
2005 .release = seq_release,
2006};
2007
2008static int __init proto_init(void)
2009{
2010 /* register /proc/net/protocols */
457c4cbc 2011 return proc_net_fops_create(&init_net, "protocols", S_IRUGO, &proto_seq_fops) == NULL ? -ENOBUFS : 0;
1da177e4
LT		 2012}
2013
2014subsys_initcall(proto_init);
2015
2016#endif /* PROC_FS */
2017
2018EXPORT_SYMBOL(sk_alloc);
2019EXPORT_SYMBOL(sk_free);
2020EXPORT_SYMBOL(sk_send_sigurg);
2021EXPORT_SYMBOL(sock_alloc_send_skb);
2022EXPORT_SYMBOL(sock_init_data);
2023EXPORT_SYMBOL(sock_kfree_s);
2024EXPORT_SYMBOL(sock_kmalloc);
2025EXPORT_SYMBOL(sock_no_accept);
2026EXPORT_SYMBOL(sock_no_bind);
2027EXPORT_SYMBOL(sock_no_connect);
2028EXPORT_SYMBOL(sock_no_getname);
2029EXPORT_SYMBOL(sock_no_getsockopt);
2030EXPORT_SYMBOL(sock_no_ioctl);
2031EXPORT_SYMBOL(sock_no_listen);
2032EXPORT_SYMBOL(sock_no_mmap);
2033EXPORT_SYMBOL(sock_no_poll);
2034EXPORT_SYMBOL(sock_no_recvmsg);
2035EXPORT_SYMBOL(sock_no_sendmsg);
2036EXPORT_SYMBOL(sock_no_sendpage);
2037EXPORT_SYMBOL(sock_no_setsockopt);
2038EXPORT_SYMBOL(sock_no_shutdown);
2039EXPORT_SYMBOL(sock_no_socketpair);
2040EXPORT_SYMBOL(sock_rfree);
2041EXPORT_SYMBOL(sock_setsockopt);
2042EXPORT_SYMBOL(sock_wfree);
2043EXPORT_SYMBOL(sock_wmalloc);
2044EXPORT_SYMBOL(sock_i_uid);
2045EXPORT_SYMBOL(sock_i_ino);
1da177e4 2046EXPORT_SYMBOL(sysctl_optmem_max);
6baf1f41 2047#ifdef CONFIG_SYSCTL
1da177e4
LT		 2048EXPORT_SYMBOL(sysctl_rmem_max);
2049EXPORT_SYMBOL(sysctl_wmem_max);
2050#endif
Linux kernel - Deliverables
This page took 0.407861 seconds and 5 git commands to generate.