projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[NET]: Move the get_net() from sock_copy()
[linux-2.6-block.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
1da177e4
LT		 873/**
874 * sk_alloc - All socket objects are allocated here
c4ea43c5 875 * @net: the applicable net namespace
4dc3b16b
PP		 876 * @family: protocol family
877 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
878 * @prot: struct proto associated with this new sock instance
879 * @zero_it: if we should zero the newly allocated sock
1da177e4 880 */
1b8d7ae4 881struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
86a76caf 882 struct proto *prot, int zero_it)
1da177e4
LT		 883{
884 struct sock *sk = NULL;
e18b890b 885 struct kmem_cache *slab = prot->slab;
1da177e4
LT		 886
887 if (slab != NULL)
888 sk = kmem_cache_alloc(slab, priority);
889 else
890 sk = kmalloc(prot->obj_size, priority);
891
892 if (sk) {
893 if (zero_it) {
894 memset(sk, 0, prot->obj_size);
895 sk->sk_family = family;
476e19cf
ACM		 896 /*
897 * See comment in struct sock definition to understand
898 * why we need sk_prot_creator -acme
899 */
900 sk->sk_prot = sk->sk_prot_creator = prot;
1da177e4 901 sock_lock_init(sk);
1b8d7ae4 902 sk->sk_net = get_net(net);
1da177e4 903 }
4ec93edb 904
a79af59e
FF		 905 if (security_sk_alloc(sk, family, priority))
906 goto out_free;
907
908 if (!try_module_get(prot->owner))
909 goto out_free;
1da177e4
LT		 910 }
911 return sk;
a79af59e
FF		 912
913out_free:
914 if (slab != NULL)
915 kmem_cache_free(slab, sk);
916 else
917 kfree(sk);
918 return NULL;
1da177e4
LT		 919}
920
921void sk_free(struct sock *sk)
922{
923 struct sk_filter *filter;
476e19cf 924 struct module *owner = sk->sk_prot_creator->owner;
1da177e4
LT		 925
926 if (sk->sk_destruct)
927 sk->sk_destruct(sk);
928
fda9ef5d 929 filter = rcu_dereference(sk->sk_filter);
1da177e4 930 if (filter) {
309dd5fc 931 sk_filter_uncharge(sk, filter);
fda9ef5d 932 rcu_assign_pointer(sk->sk_filter, NULL);
1da177e4
LT		 933 }
934
935 sock_disable_timestamp(sk);
936
937 if (atomic_read(&sk->sk_omem_alloc))
938 printk(KERN_DEBUG "%s: optmem leakage (%d bytes) detected.\n",
939 __FUNCTION__, atomic_read(&sk->sk_omem_alloc));
940
941 security_sk_free(sk);
1b8d7ae4 942 put_net(sk->sk_net);
476e19cf
ACM		 943 if (sk->sk_prot_creator->slab != NULL)
944 kmem_cache_free(sk->sk_prot_creator->slab, sk);
1da177e4
LT		 945 else
946 kfree(sk);
947 module_put(owner);
948}
949
dd0fc66f 950struct sock *sk_clone(const struct sock *sk, const gfp_t priority)
87d11ceb 951{
1b8d7ae4 952 struct sock *newsk = sk_alloc(sk->sk_net, sk->sk_family, priority, sk->sk_prot, 0);
87d11ceb
ACM		 953
954 if (newsk != NULL) {
955 struct sk_filter *filter;
956
892c141e 957 sock_copy(newsk, sk);
87d11ceb
ACM		 958
959 /* SANITY */
1e2e6b89 960 get_net(newsk->sk_net);
87d11ceb
ACM		 961 sk_node_init(&newsk->sk_node);
962 sock_lock_init(newsk);
963 bh_lock_sock(newsk);
fa438ccf 964 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
87d11ceb
ACM		 965
966 atomic_set(&newsk->sk_rmem_alloc, 0);
967 atomic_set(&newsk->sk_wmem_alloc, 0);
968 atomic_set(&newsk->sk_omem_alloc, 0);
969 skb_queue_head_init(&newsk->sk_receive_queue);
970 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 971#ifdef CONFIG_NET_DMA
972 skb_queue_head_init(&newsk->sk_async_wait_queue);
973#endif
87d11ceb
ACM		 974
975 rwlock_init(&newsk->sk_dst_lock);
976 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 977 lockdep_set_class_and_name(&newsk->sk_callback_lock,
978 af_callback_keys + newsk->sk_family,
979 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 980
981 newsk->sk_dst_cache = NULL;
982 newsk->sk_wmem_queued = 0;
983 newsk->sk_forward_alloc = 0;
984 newsk->sk_send_head = NULL;
87d11ceb
ACM		 985 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
986
987 sock_reset_flag(newsk, SOCK_DONE);
988 skb_queue_head_init(&newsk->sk_error_queue);
989
990 filter = newsk->sk_filter;
991 if (filter != NULL)
992 sk_filter_charge(newsk, filter);
993
994 if (unlikely(xfrm_sk_clone_policy(newsk))) {
995 /* It is still raw copy of parent, so invalidate
996 * destructor and make plain sk_free() */
997 newsk->sk_destruct = NULL;
998 sk_free(newsk);
999 newsk = NULL;
1000 goto out;
1001 }
1002
1003 newsk->sk_err = 0;
1004 newsk->sk_priority = 0;
1005 atomic_set(&newsk->sk_refcnt, 2);
1006
1007 /*
1008 * Increment the counter in the same struct proto as the master
1009 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1010 * is the same as sk->sk_prot->socks, as this field was copied
1011 * with memcpy).
1012 *
1013 * This _changes_ the previous behaviour, where
1014 * tcp_create_openreq_child always was incrementing the
1015 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1016 * to be taken into account in all callers. -acme
1017 */
1018 sk_refcnt_debug_inc(newsk);
1019 newsk->sk_socket = NULL;
1020 newsk->sk_sleep = NULL;
1021
1022 if (newsk->sk_prot->sockets_allocated)
1023 atomic_inc(newsk->sk_prot->sockets_allocated);
1024 }
1025out:
1026 return newsk;
1027}
1028
1029EXPORT_SYMBOL_GPL(sk_clone);
1030
9958089a
AK		 1031void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1032{
1033 __sk_dst_set(sk, dst);
1034 sk->sk_route_caps = dst->dev->features;
1035 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1036 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
9958089a
AK		 1037 if (sk_can_gso(sk)) {
1038 if (dst->header_len)
1039 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
1040 else
1041 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
1042 }
1043}
1044EXPORT_SYMBOL_GPL(sk_setup_caps);
1045
1da177e4
LT		 1046void __init sk_init(void)
1047{
1048 if (num_physpages <= 4096) {
1049 sysctl_wmem_max = 32767;
1050 sysctl_rmem_max = 32767;
1051 sysctl_wmem_default = 32767;
1052 sysctl_rmem_default = 32767;
1053 } else if (num_physpages >= 131072) {
1054 sysctl_wmem_max = 131071;
1055 sysctl_rmem_max = 131071;
1056 }
1057}
1058
1059/*
1060 * Simple resource managers for sockets.
1061 */
1062
1063
4ec93edb
YH		 1064/*
1065 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1066 */
1067void sock_wfree(struct sk_buff *skb)
1068{
1069 struct sock *sk = skb->sk;
1070
1071 /* In case it might be waiting for more memory. */
1072 atomic_sub(skb->truesize, &sk->sk_wmem_alloc);
1073 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE))
1074 sk->sk_write_space(sk);
1075 sock_put(sk);
1076}
1077
4ec93edb
YH		 1078/*
1079 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1080 */
1081void sock_rfree(struct sk_buff *skb)
1082{
1083 struct sock *sk = skb->sk;
1084
1085 atomic_sub(skb->truesize, &sk->sk_rmem_alloc);
1086}
1087
1088
1089int sock_i_uid(struct sock *sk)
1090{
1091 int uid;
1092
1093 read_lock(&sk->sk_callback_lock);
1094 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : 0;
1095 read_unlock(&sk->sk_callback_lock);
1096 return uid;
1097}
1098
1099unsigned long sock_i_ino(struct sock *sk)
1100{
1101 unsigned long ino;
1102
1103 read_lock(&sk->sk_callback_lock);
1104 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
1105 read_unlock(&sk->sk_callback_lock);
1106 return ino;
1107}
1108
1109/*
1110 * Allocate a skb from the socket's send buffer.
1111 */
86a76caf 1112struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1113 gfp_t priority)
1da177e4
LT		 1114{
1115 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
1116 struct sk_buff * skb = alloc_skb(size, priority);
1117 if (skb) {
1118 skb_set_owner_w(skb, sk);
1119 return skb;
1120 }
1121 }
1122 return NULL;
1123}
1124
1125/*
1126 * Allocate a skb from the socket's receive buffer.
4ec93edb 1127 */
86a76caf 1128struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1129 gfp_t priority)
1da177e4
LT		 1130{
1131 if (force || atomic_read(&sk->sk_rmem_alloc) < sk->sk_rcvbuf) {
1132 struct sk_buff *skb = alloc_skb(size, priority);
1133 if (skb) {
1134 skb_set_owner_r(skb, sk);
1135 return skb;
1136 }
1137 }
1138 return NULL;
1139}
1140
4ec93edb 1141/*
1da177e4 1142 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1143 */
dd0fc66f 1144void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4
LT		 1145{
1146 if ((unsigned)size <= sysctl_optmem_max &&
1147 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1148 void *mem;
1149 /* First do the add, to avoid the race if kmalloc
4ec93edb 1150 * might sleep.
1da177e4
LT		 1151 */
1152 atomic_add(size, &sk->sk_omem_alloc);
1153 mem = kmalloc(size, priority);
1154 if (mem)
1155 return mem;
1156 atomic_sub(size, &sk->sk_omem_alloc);
1157 }
1158 return NULL;
1159}
1160
1161/*
1162 * Free an option memory block.
1163 */
1164void sock_kfree_s(struct sock *sk, void *mem, int size)
1165{
1166 kfree(mem);
1167 atomic_sub(size, &sk->sk_omem_alloc);
1168}
1169
1170/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1171 I think, these locks should be removed for datagram sockets.
1172 */
1173static long sock_wait_for_wmem(struct sock * sk, long timeo)
1174{
1175 DEFINE_WAIT(wait);
1176
1177 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1178 for (;;) {
1179 if (!timeo)
1180 break;
1181 if (signal_pending(current))
1182 break;
1183 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1184 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1185 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1186 break;
1187 if (sk->sk_shutdown & SEND_SHUTDOWN)
1188 break;
1189 if (sk->sk_err)
1190 break;
1191 timeo = schedule_timeout(timeo);
1192 }
1193 finish_wait(sk->sk_sleep, &wait);
1194 return timeo;
1195}
1196
1197
1198/*
1199 * Generic send/receive buffer handlers
1200 */
1201
1202static struct sk_buff *sock_alloc_send_pskb(struct sock *sk,
1203 unsigned long header_len,
1204 unsigned long data_len,
1205 int noblock, int *errcode)
1206{
1207 struct sk_buff *skb;
7d877f3b 1208 gfp_t gfp_mask;
1da177e4
LT		 1209 long timeo;
1210 int err;
1211
1212 gfp_mask = sk->sk_allocation;
1213 if (gfp_mask & __GFP_WAIT)
1214 gfp_mask |= __GFP_REPEAT;
1215
1216 timeo = sock_sndtimeo(sk, noblock);
1217 while (1) {
1218 err = sock_error(sk);
1219 if (err != 0)
1220 goto failure;
1221
1222 err = -EPIPE;
1223 if (sk->sk_shutdown & SEND_SHUTDOWN)
1224 goto failure;
1225
1226 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
db38c179 1227 skb = alloc_skb(header_len, gfp_mask);
1da177e4
LT		 1228 if (skb) {
1229 int npages;
1230 int i;
1231
1232 /* No pages, we're done... */
1233 if (!data_len)
1234 break;
1235
1236 npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
1237 skb->truesize += data_len;
1238 skb_shinfo(skb)->nr_frags = npages;
1239 for (i = 0; i < npages; i++) {
1240 struct page *page;
1241 skb_frag_t *frag;
1242
1243 page = alloc_pages(sk->sk_allocation, 0);
1244 if (!page) {
1245 err = -ENOBUFS;
1246 skb_shinfo(skb)->nr_frags = i;
1247 kfree_skb(skb);
1248 goto failure;
1249 }
1250
1251 frag = &skb_shinfo(skb)->frags[i];
1252 frag->page = page;
1253 frag->page_offset = 0;
1254 frag->size = (data_len >= PAGE_SIZE ?
1255 PAGE_SIZE :
1256 data_len);
1257 data_len -= PAGE_SIZE;
1258 }
1259
1260 /* Full success... */
1261 break;
1262 }
1263 err = -ENOBUFS;
1264 goto failure;
1265 }
1266 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1267 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1268 err = -EAGAIN;
1269 if (!timeo)
1270 goto failure;
1271 if (signal_pending(current))
1272 goto interrupted;
1273 timeo = sock_wait_for_wmem(sk, timeo);
1274 }
1275
1276 skb_set_owner_w(skb, sk);
1277 return skb;
1278
1279interrupted:
1280 err = sock_intr_errno(timeo);
1281failure:
1282 *errcode = err;
1283 return NULL;
1284}
1285
4ec93edb 1286struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1287 int noblock, int *errcode)
1288{
1289 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode);
1290}
1291
1292static void __lock_sock(struct sock *sk)
1293{
1294 DEFINE_WAIT(wait);
1295
e71a4783 1296 for (;;) {
1da177e4
LT		 1297 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1298 TASK_UNINTERRUPTIBLE);
1299 spin_unlock_bh(&sk->sk_lock.slock);
1300 schedule();
1301 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1302 if (!sock_owned_by_user(sk))
1da177e4
LT		 1303 break;
1304 }
1305 finish_wait(&sk->sk_lock.wq, &wait);
1306}
1307
1308static void __release_sock(struct sock *sk)
1309{
1310 struct sk_buff *skb = sk->sk_backlog.head;
1311
1312 do {
1313 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1314 bh_unlock_sock(sk);
1315
1316 do {
1317 struct sk_buff *next = skb->next;
1318
1319 skb->next = NULL;
1320 sk->sk_backlog_rcv(sk, skb);
1321
1322 /*
1323 * We are in process context here with softirqs
1324 * disabled, use cond_resched_softirq() to preempt.
1325 * This is safe to do because we've taken the backlog
1326 * queue private:
1327 */
1328 cond_resched_softirq();
1329
1330 skb = next;
1331 } while (skb != NULL);
1332
1333 bh_lock_sock(sk);
e71a4783 1334 } while ((skb = sk->sk_backlog.head) != NULL);
1da177e4
LT		 1335}
1336
1337/**
1338 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1339 * @sk: sock to wait on
1340 * @timeo: for how long
1da177e4
LT		 1341 *
1342 * Now socket state including sk->sk_err is changed only under lock,
1343 * hence we may omit checks after joining wait queue.
1344 * We check receive queue before schedule() only as optimization;
1345 * it is very likely that release_sock() added new data.
1346 */
1347int sk_wait_data(struct sock *sk, long *timeo)
1348{
1349 int rc;
1350 DEFINE_WAIT(wait);
1351
1352 prepare_to_wait(sk->sk_sleep, &wait, TASK_INTERRUPTIBLE);
1353 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1354 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1355 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1356 finish_wait(sk->sk_sleep, &wait);
1357 return rc;
1358}
1359
1360EXPORT_SYMBOL(sk_wait_data);
1361
1362/*
1363 * Set of default routines for initialising struct proto_ops when
1364 * the protocol does not support a particular function. In certain
1365 * cases where it makes no sense for a protocol to have a "do nothing"
1366 * function, some default processing is provided.
1367 */
1368
1369int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
1370{
1371 return -EOPNOTSUPP;
1372}
1373
4ec93edb 1374int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1375 int len, int flags)
1376{
1377 return -EOPNOTSUPP;
1378}
1379
1380int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
1381{
1382 return -EOPNOTSUPP;
1383}
1384
1385int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
1386{
1387 return -EOPNOTSUPP;
1388}
1389
4ec93edb 1390int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 1391 int *len, int peer)
1392{
1393 return -EOPNOTSUPP;
1394}
1395
1396unsigned int sock_no_poll(struct file * file, struct socket *sock, poll_table *pt)
1397{
1398 return 0;
1399}
1400
1401int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
1402{
1403 return -EOPNOTSUPP;
1404}
1405
1406int sock_no_listen(struct socket *sock, int backlog)
1407{
1408 return -EOPNOTSUPP;
1409}
1410
1411int sock_no_shutdown(struct socket *sock, int how)
1412{
1413 return -EOPNOTSUPP;
1414}
1415
1416int sock_no_setsockopt(struct socket *sock, int level, int optname,
1417 char __user *optval, int optlen)
1418{
1419 return -EOPNOTSUPP;
1420}
1421
1422int sock_no_getsockopt(struct socket *sock, int level, int optname,
1423 char __user *optval, int __user *optlen)
1424{
1425 return -EOPNOTSUPP;
1426}
1427
1428int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1429 size_t len)
1430{
1431 return -EOPNOTSUPP;
1432}
1433
1434int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
1435 size_t len, int flags)
1436{
1437 return -EOPNOTSUPP;
1438}
1439
1440int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
1441{
1442 /* Mirror missing mmap method error code */
1443 return -ENODEV;
1444}
1445
1446ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
1447{
1448 ssize_t res;
1449 struct msghdr msg = {.msg_flags = flags};
1450 struct kvec iov;
1451 char *kaddr = kmap(page);
1452 iov.iov_base = kaddr + offset;
1453 iov.iov_len = size;
1454 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
1455 kunmap(page);
1456 return res;
1457}
1458
1459/*
1460 * Default Socket Callbacks
1461 */
1462
1463static void sock_def_wakeup(struct sock *sk)
1464{
1465 read_lock(&sk->sk_callback_lock);
1466 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1467 wake_up_interruptible_all(sk->sk_sleep);
1468 read_unlock(&sk->sk_callback_lock);
1469}
1470
1471static void sock_def_error_report(struct sock *sk)
1472{
1473 read_lock(&sk->sk_callback_lock);
1474 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1475 wake_up_interruptible(sk->sk_sleep);
4ec93edb 1476 sk_wake_async(sk,0,POLL_ERR);
1da177e4
LT		 1477 read_unlock(&sk->sk_callback_lock);
1478}
1479
1480static void sock_def_readable(struct sock *sk, int len)
1481{
1482 read_lock(&sk->sk_callback_lock);
1483 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1484 wake_up_interruptible(sk->sk_sleep);
1485 sk_wake_async(sk,1,POLL_IN);
1486 read_unlock(&sk->sk_callback_lock);
1487}
1488
1489static void sock_def_write_space(struct sock *sk)
1490{
1491 read_lock(&sk->sk_callback_lock);
1492
1493 /* Do not wake up a writer until he can make "significant"
1494 * progress. --DaveM
1495 */
e71a4783 1496 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
1da177e4
LT		 1497 if (sk->sk_sleep && waitqueue_active(sk->sk_sleep))
1498 wake_up_interruptible(sk->sk_sleep);
1499
1500 /* Should agree with poll, otherwise some programs break */
1501 if (sock_writeable(sk))
1502 sk_wake_async(sk, 2, POLL_OUT);
1503 }
1504
1505 read_unlock(&sk->sk_callback_lock);
1506}
1507
1508static void sock_def_destruct(struct sock *sk)
1509{
a51482bd 1510 kfree(sk->sk_protinfo);
1da177e4
LT		 1511}
1512
1513void sk_send_sigurg(struct sock *sk)
1514{
1515 if (sk->sk_socket && sk->sk_socket->file)
1516 if (send_sigurg(&sk->sk_socket->file->f_owner))
1517 sk_wake_async(sk, 3, POLL_PRI);
1518}
1519
1520void sk_reset_timer(struct sock *sk, struct timer_list* timer,
1521 unsigned long expires)
1522{
1523 if (!mod_timer(timer, expires))
1524 sock_hold(sk);
1525}
1526
1527EXPORT_SYMBOL(sk_reset_timer);
1528
1529void sk_stop_timer(struct sock *sk, struct timer_list* timer)
1530{
1531 if (timer_pending(timer) && del_timer(timer))
1532 __sock_put(sk);
1533}
1534
1535EXPORT_SYMBOL(sk_stop_timer);
1536
1537void sock_init_data(struct socket *sock, struct sock *sk)
1538{
1539 skb_queue_head_init(&sk->sk_receive_queue);
1540 skb_queue_head_init(&sk->sk_write_queue);
1541 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 1542#ifdef CONFIG_NET_DMA
1543 skb_queue_head_init(&sk->sk_async_wait_queue);
1544#endif
1da177e4
LT		 1545
1546 sk->sk_send_head = NULL;
1547
1548 init_timer(&sk->sk_timer);
4ec93edb 1549
1da177e4
LT		 1550 sk->sk_allocation = GFP_KERNEL;
1551 sk->sk_rcvbuf = sysctl_rmem_default;
1552 sk->sk_sndbuf = sysctl_wmem_default;
1553 sk->sk_state = TCP_CLOSE;
1554 sk->sk_socket = sock;
1555
1556 sock_set_flag(sk, SOCK_ZAPPED);
1557
e71a4783 1558 if (sock) {
1da177e4
LT		 1559 sk->sk_type = sock->type;
1560 sk->sk_sleep = &sock->wait;
1561 sock->sk = sk;
1562 } else
1563 sk->sk_sleep = NULL;
1564
1565 rwlock_init(&sk->sk_dst_lock);
1566 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 1567 lockdep_set_class_and_name(&sk->sk_callback_lock,
1568 af_callback_keys + sk->sk_family,
1569 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 1570
1571 sk->sk_state_change = sock_def_wakeup;
1572 sk->sk_data_ready = sock_def_readable;
1573 sk->sk_write_space = sock_def_write_space;
1574 sk->sk_error_report = sock_def_error_report;
1575 sk->sk_destruct = sock_def_destruct;
1576
1577 sk->sk_sndmsg_page = NULL;
1578 sk->sk_sndmsg_off = 0;
1579
1580 sk->sk_peercred.pid = 0;
1581 sk->sk_peercred.uid = -1;
1582 sk->sk_peercred.gid = -1;
1583 sk->sk_write_pending = 0;
1584 sk->sk_rcvlowat = 1;
1585 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
1586 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
1587
b7aa0bf7 1588 sk->sk_stamp = ktime_set(-1L, -1L);
1da177e4
LT		 1589
1590 atomic_set(&sk->sk_refcnt, 1);
1591}
1592
fcc70d5f 1593void fastcall lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 1594{
1595 might_sleep();
a5b5bb9a 1596 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 1597 if (sk->sk_lock.owned)
1da177e4 1598 __lock_sock(sk);
d2e9117c 1599 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 1600 spin_unlock(&sk->sk_lock.slock);
1601 /*
1602 * The sk_lock has mutex_lock() semantics here:
1603 */
fcc70d5f 1604 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 1605 local_bh_enable();
1da177e4
LT		 1606}
1607
fcc70d5f 1608EXPORT_SYMBOL(lock_sock_nested);
1da177e4
LT		 1609
1610void fastcall release_sock(struct sock *sk)
1611{
a5b5bb9a
IM		 1612 /*
1613 * The sk_lock has mutex_unlock() semantics:
1614 */
1615 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1616
1617 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 1618 if (sk->sk_backlog.tail)
1619 __release_sock(sk);
d2e9117c 1620 sk->sk_lock.owned = 0;
a5b5bb9a
IM		 1621 if (waitqueue_active(&sk->sk_lock.wq))
1622 wake_up(&sk->sk_lock.wq);
1623 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 1624}
1625EXPORT_SYMBOL(release_sock);
1626
1627int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 1628{
b7aa0bf7 1629 struct timeval tv;
1da177e4
LT		 1630 if (!sock_flag(sk, SOCK_TIMESTAMP))
1631 sock_enable_timestamp(sk);
b7aa0bf7
ED		 1632 tv = ktime_to_timeval(sk->sk_stamp);
1633 if (tv.tv_sec == -1)
1da177e4 1634 return -ENOENT;
b7aa0bf7
ED		 1635 if (tv.tv_sec == 0) {
1636 sk->sk_stamp = ktime_get_real();
1637 tv = ktime_to_timeval(sk->sk_stamp);
1638 }
1639 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 1640}
1da177e4
LT		 1641EXPORT_SYMBOL(sock_get_timestamp);
1642
ae40eb1e
ED		 1643int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
1644{
1645 struct timespec ts;
1646 if (!sock_flag(sk, SOCK_TIMESTAMP))
1647 sock_enable_timestamp(sk);
1648 ts = ktime_to_timespec(sk->sk_stamp);
1649 if (ts.tv_sec == -1)
1650 return -ENOENT;
1651 if (ts.tv_sec == 0) {
1652 sk->sk_stamp = ktime_get_real();
1653 ts = ktime_to_timespec(sk->sk_stamp);
1654 }
1655 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
1656}
1657EXPORT_SYMBOL(sock_get_timestampns);
1658
1da177e4 1659void sock_enable_timestamp(struct sock *sk)
4ec93edb
YH		 1660{
1661 if (!sock_flag(sk, SOCK_TIMESTAMP)) {
1da177e4
LT		 1662 sock_set_flag(sk, SOCK_TIMESTAMP);
1663 net_enable_timestamp();
1664 }
1665}
1da177e4
LT		 1666
1667/*
1668 * Get a socket option on an socket.
1669 *
1670 * FIX: POSIX 1003.1g is very ambiguous here. It states that
1671 * asynchronous errors should be reported by getsockopt. We assume
1672 * this means if you specify SO_ERROR (otherwise whats the point of it).
1673 */
1674int sock_common_getsockopt(struct socket *sock, int level, int optname,
1675 char __user *optval, int __user *optlen)
1676{
1677 struct sock *sk = sock->sk;
1678
1679 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
1680}
1681
1682EXPORT_SYMBOL(sock_common_getsockopt);
1683
3fdadf7d 1684#ifdef CONFIG_COMPAT
543d9cfe
ACM		 1685int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
1686 char __user *optval, int __user *optlen)
3fdadf7d
DM		 1687{
1688 struct sock *sk = sock->sk;
1689
1e51f951 1690 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 1691 return sk->sk_prot->compat_getsockopt(sk, level, optname,
1692 optval, optlen);
3fdadf7d
DM		 1693 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
1694}
1695EXPORT_SYMBOL(compat_sock_common_getsockopt);
1696#endif
1697
1da177e4
LT		 1698int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
1699 struct msghdr *msg, size_t size, int flags)
1700{
1701 struct sock *sk = sock->sk;
1702 int addr_len = 0;
1703 int err;
1704
1705 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
1706 flags & ~MSG_DONTWAIT, &addr_len);
1707 if (err >= 0)
1708 msg->msg_namelen = addr_len;
1709 return err;
1710}
1711
1712EXPORT_SYMBOL(sock_common_recvmsg);
1713
1714/*
1715 * Set socket options on an inet socket.
1716 */
1717int sock_common_setsockopt(struct socket *sock, int level, int optname,
1718 char __user *optval, int optlen)
1719{
1720 struct sock *sk = sock->sk;
1721
1722 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
1723}
1724
1725EXPORT_SYMBOL(sock_common_setsockopt);
1726
3fdadf7d 1727#ifdef CONFIG_COMPAT
543d9cfe
ACM		 1728int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
1729 char __user *optval, int optlen)
3fdadf7d
DM		 1730{
1731 struct sock *sk = sock->sk;
1732
543d9cfe
ACM		 1733 if (sk->sk_prot->compat_setsockopt != NULL)
1734 return sk->sk_prot->compat_setsockopt(sk, level, optname,
1735 optval, optlen);
3fdadf7d
DM		 1736 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
1737}
1738EXPORT_SYMBOL(compat_sock_common_setsockopt);
1739#endif
1740
1da177e4
LT		 1741void sk_common_release(struct sock *sk)
1742{
1743 if (sk->sk_prot->destroy)
1744 sk->sk_prot->destroy(sk);
1745
1746 /*
1747 * Observation: when sock_common_release is called, processes have
1748 * no access to socket. But net still has.
1749 * Step one, detach it from networking:
1750 *
1751 * A. Remove from hash tables.
1752 */
1753
1754 sk->sk_prot->unhash(sk);
1755
1756 /*
1757 * In this point socket cannot receive new packets, but it is possible
1758 * that some packets are in flight because some CPU runs receiver and
1759 * did hash table lookup before we unhashed socket. They will achieve
1760 * receive queue and will be purged by socket destructor.
1761 *
1762 * Also we still have packets pending on receive queue and probably,
1763 * our own packets waiting in device queues. sock_destroy will drain
1764 * receive queue, but transmitted packets will delay socket destruction
1765 * until the last reference will be released.
1766 */
1767
1768 sock_orphan(sk);
1769
1770 xfrm_sk_free_policy(sk);
1771
e6848976 1772 sk_refcnt_debug_release(sk);
1da177e4
LT		 1773 sock_put(sk);
1774}
1775
1776EXPORT_SYMBOL(sk_common_release);
1777
1778static DEFINE_RWLOCK(proto_list_lock);
1779static LIST_HEAD(proto_list);
1780
1781int proto_register(struct proto *prot, int alloc_slab)
1782{
8feaf0c0
ACM		 1783 char *request_sock_slab_name = NULL;
1784 char *timewait_sock_slab_name;
1da177e4
LT		 1785 int rc = -ENOBUFS;
1786
1da177e4
LT		 1787 if (alloc_slab) {
1788 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
20c2df83 1789 SLAB_HWCACHE_ALIGN, NULL);
1da177e4
LT		 1790
1791 if (prot->slab == NULL) {
1792 printk(KERN_CRIT "%s: Can't create sock SLAB cache!\n",
1793 prot->name);
2a278051 1794 goto out;
1da177e4 1795 }
2e6599cb
ACM		 1796
1797 if (prot->rsk_prot != NULL) {
1798 static const char mask[] = "request_sock_%s";
1799
1800 request_sock_slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL);
1801 if (request_sock_slab_name == NULL)
1802 goto out_free_sock_slab;
1803
1804 sprintf(request_sock_slab_name, mask, prot->name);
1805 prot->rsk_prot->slab = kmem_cache_create(request_sock_slab_name,
1806 prot->rsk_prot->obj_size, 0,
20c2df83 1807 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 1808
1809 if (prot->rsk_prot->slab == NULL) {
1810 printk(KERN_CRIT "%s: Can't create request sock SLAB cache!\n",
1811 prot->name);
1812 goto out_free_request_sock_slab_name;
1813 }
1814 }
8feaf0c0 1815
6d6ee43e 1816 if (prot->twsk_prot != NULL) {
8feaf0c0
ACM		 1817 static const char mask[] = "tw_sock_%s";
1818
1819 timewait_sock_slab_name = kmalloc(strlen(prot->name) + sizeof(mask) - 1, GFP_KERNEL);
1820
1821 if (timewait_sock_slab_name == NULL)
1822 goto out_free_request_sock_slab;
1823
1824 sprintf(timewait_sock_slab_name, mask, prot->name);
6d6ee43e
ACM		 1825 prot->twsk_prot->twsk_slab =
1826 kmem_cache_create(timewait_sock_slab_name,
1827 prot->twsk_prot->twsk_obj_size,
1828 0, SLAB_HWCACHE_ALIGN,
20c2df83 1829 NULL);
6d6ee43e 1830 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 1831 goto out_free_timewait_sock_slab_name;
1832 }
1da177e4
LT		 1833 }
1834
2a278051 1835 write_lock(&proto_list_lock);
1da177e4 1836 list_add(&prot->node, &proto_list);
1da177e4 1837 write_unlock(&proto_list_lock);
2a278051
ACM		 1838 rc = 0;
1839out:
1da177e4 1840 return rc;
8feaf0c0
ACM		 1841out_free_timewait_sock_slab_name:
1842 kfree(timewait_sock_slab_name);
1843out_free_request_sock_slab:
1844 if (prot->rsk_prot && prot->rsk_prot->slab) {
1845 kmem_cache_destroy(prot->rsk_prot->slab);
1846 prot->rsk_prot->slab = NULL;
1847 }
2e6599cb
ACM		 1848out_free_request_sock_slab_name:
1849 kfree(request_sock_slab_name);
1850out_free_sock_slab:
1851 kmem_cache_destroy(prot->slab);
1852 prot->slab = NULL;
1853 goto out;
1da177e4
LT		 1854}
1855
1856EXPORT_SYMBOL(proto_register);
1857
1858void proto_unregister(struct proto *prot)
1859{
1860 write_lock(&proto_list_lock);
0a3f4358
PM		 1861 list_del(&prot->node);
1862 write_unlock(&proto_list_lock);
1da177e4
LT		 1863
1864 if (prot->slab != NULL) {
1865 kmem_cache_destroy(prot->slab);
1866 prot->slab = NULL;
1867 }
1868
2e6599cb
ACM		 1869 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
1870 const char *name = kmem_cache_name(prot->rsk_prot->slab);
1871
1872 kmem_cache_destroy(prot->rsk_prot->slab);
1873 kfree(name);
1874 prot->rsk_prot->slab = NULL;
1875 }
1876
6d6ee43e
ACM		 1877 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
1878 const char *name = kmem_cache_name(prot->twsk_prot->twsk_slab);
8feaf0c0 1879
6d6ee43e 1880 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
8feaf0c0 1881 kfree(name);
6d6ee43e 1882 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 1883 }
1da177e4
LT		 1884}
1885
1886EXPORT_SYMBOL(proto_unregister);
1887
1888#ifdef CONFIG_PROC_FS
1da177e4
LT		 1889static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
1890{
1891 read_lock(&proto_list_lock);
60f0438a 1892 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 1893}
1894
1895static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1896{
60f0438a 1897 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 1898}
1899
1900static void proto_seq_stop(struct seq_file *seq, void *v)
1901{
1902 read_unlock(&proto_list_lock);
1903}
1904
1905static char proto_method_implemented(const void *method)
1906{
1907 return method == NULL ? 'n' : 'y';
1908}
1909
1910static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
1911{
1912 seq_printf(seq, "%-9s %4u %6d %6d %-3s %6u %-3s %-10s "
1913 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
1914 proto->name,
1915 proto->obj_size,
1916 proto->sockets_allocated != NULL ? atomic_read(proto->sockets_allocated) : -1,
1917 proto->memory_allocated != NULL ? atomic_read(proto->memory_allocated) : -1,
1918 proto->memory_pressure != NULL ? *proto->memory_pressure ? "yes" : "no" : "NI",
1919 proto->max_header,
1920 proto->slab == NULL ? "no" : "yes",
1921 module_name(proto->owner),
1922 proto_method_implemented(proto->close),
1923 proto_method_implemented(proto->connect),
1924 proto_method_implemented(proto->disconnect),
1925 proto_method_implemented(proto->accept),
1926 proto_method_implemented(proto->ioctl),
1927 proto_method_implemented(proto->init),
1928 proto_method_implemented(proto->destroy),
1929 proto_method_implemented(proto->shutdown),
1930 proto_method_implemented(proto->setsockopt),
1931 proto_method_implemented(proto->getsockopt),
1932 proto_method_implemented(proto->sendmsg),
1933 proto_method_implemented(proto->recvmsg),
1934 proto_method_implemented(proto->sendpage),
1935 proto_method_implemented(proto->bind),
1936 proto_method_implemented(proto->backlog_rcv),
1937 proto_method_implemented(proto->hash),
1938 proto_method_implemented(proto->unhash),
1939 proto_method_implemented(proto->get_port),
1940 proto_method_implemented(proto->enter_memory_pressure));
1941}
1942
1943static int proto_seq_show(struct seq_file *seq, void *v)
1944{
60f0438a 1945 if (v == &proto_list)
1da177e4
LT		 1946 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
1947 "protocol",
1948 "size",
1949 "sockets",
1950 "memory",
1951 "press",
1952 "maxhdr",
1953 "slab",
1954 "module",
1955 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
1956 else
60f0438a 1957 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 1958 return 0;
1959}
1960
f690808e 1961static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 1962 .start = proto_seq_start,
1963 .next = proto_seq_next,
1964 .stop = proto_seq_stop,
1965 .show = proto_seq_show,
1966};
1967
1968static int proto_seq_open(struct inode *inode, struct file *file)
1969{
1970 return seq_open(file, &proto_seq_ops);
1971}
1972
9a32144e 1973static const struct file_operations proto_seq_fops = {
1da177e4
LT		 1974 .owner = THIS_MODULE,
1975 .open = proto_seq_open,
1976 .read = seq_read,
1977 .llseek = seq_lseek,
1978 .release = seq_release,
1979};
1980
1981static int __init proto_init(void)
1982{
1983 /* register /proc/net/protocols */
457c4cbc 1984 return proc_net_fops_create(&init_net, "protocols", S_IRUGO, &proto_seq_fops) == NULL ? -ENOBUFS : 0;
1da177e4
LT		 1985}
1986
1987subsys_initcall(proto_init);
1988
1989#endif /* PROC_FS */
1990
1991EXPORT_SYMBOL(sk_alloc);
1992EXPORT_SYMBOL(sk_free);
1993EXPORT_SYMBOL(sk_send_sigurg);
1994EXPORT_SYMBOL(sock_alloc_send_skb);
1995EXPORT_SYMBOL(sock_init_data);
1996EXPORT_SYMBOL(sock_kfree_s);
1997EXPORT_SYMBOL(sock_kmalloc);
1998EXPORT_SYMBOL(sock_no_accept);
1999EXPORT_SYMBOL(sock_no_bind);
2000EXPORT_SYMBOL(sock_no_connect);
2001EXPORT_SYMBOL(sock_no_getname);
2002EXPORT_SYMBOL(sock_no_getsockopt);
2003EXPORT_SYMBOL(sock_no_ioctl);
2004EXPORT_SYMBOL(sock_no_listen);
2005EXPORT_SYMBOL(sock_no_mmap);
2006EXPORT_SYMBOL(sock_no_poll);
2007EXPORT_SYMBOL(sock_no_recvmsg);
2008EXPORT_SYMBOL(sock_no_sendmsg);
2009EXPORT_SYMBOL(sock_no_sendpage);
2010EXPORT_SYMBOL(sock_no_setsockopt);
2011EXPORT_SYMBOL(sock_no_shutdown);
2012EXPORT_SYMBOL(sock_no_socketpair);
2013EXPORT_SYMBOL(sock_rfree);
2014EXPORT_SYMBOL(sock_setsockopt);
2015EXPORT_SYMBOL(sock_wfree);
2016EXPORT_SYMBOL(sock_wmalloc);
2017EXPORT_SYMBOL(sock_i_uid);
2018EXPORT_SYMBOL(sock_i_ino);
1da177e4 2019EXPORT_SYMBOL(sysctl_optmem_max);
6baf1f41 2020#ifdef CONFIG_SYSCTL
1da177e4
LT		 2021EXPORT_SYMBOL(sysctl_rmem_max);
2022EXPORT_SYMBOL(sysctl_wmem_max);
2023#endif
Linux 6.x block layer and io_uring tree(s)