projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
net-timestamp: extend SCM_TIMESTAMPING ancillary data struct
[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 *
02c30a84 10 * Authors: Ross Biro
1da177e4
LT		 11 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
14 *
15 * Fixes:
16 * Alan Cox : Numerous verify_area() problems
17 * Alan Cox : Connecting on a connecting socket
18 * now returns an error for tcp.
19 * Alan Cox : sock->protocol is set correctly.
20 * and is not sometimes left as 0.
21 * Alan Cox : connect handles icmp errors on a
22 * connect properly. Unfortunately there
23 * is a restart syscall nasty there. I
24 * can't match BSD without hacking the C
25 * library. Ideas urgently sought!
26 * Alan Cox : Disallow bind() to addresses that are
27 * not ours - especially broadcast ones!!
28 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost)
29 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets,
30 * instead they leave that for the DESTROY timer.
31 * Alan Cox : Clean up error flag in accept
32 * Alan Cox : TCP ack handling is buggy, the DESTROY timer
33 * was buggy. Put a remove_sock() in the handler
34 * for memory when we hit 0. Also altered the timer
4ec93edb 35 * code. The ACK stuff can wait and needs major
1da177e4
LT		 36 * TCP layer surgery.
37 * Alan Cox : Fixed TCP ack bug, removed remove sock
38 * and fixed timer/inet_bh race.
39 * Alan Cox : Added zapped flag for TCP
40 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code
41 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb
42 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources
43 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing.
44 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so...
45 * Rick Sladkey : Relaxed UDP rules for matching packets.
46 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support
47 * Pauline Middelink : identd support
48 * Alan Cox : Fixed connect() taking signals I think.
49 * Alan Cox : SO_LINGER supported
50 * Alan Cox : Error reporting fixes
51 * Anonymous : inet_create tidied up (sk->reuse setting)
52 * Alan Cox : inet sockets don't set sk->type!
53 * Alan Cox : Split socket option code
54 * Alan Cox : Callbacks
55 * Alan Cox : Nagle flag for Charles & Johannes stuff
56 * Alex : Removed restriction on inet fioctl
57 * Alan Cox : Splitting INET from NET core
58 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt()
59 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code
60 * Alan Cox : Split IP from generic code
61 * Alan Cox : New kfree_skbmem()
62 * Alan Cox : Make SO_DEBUG superuser only.
63 * Alan Cox : Allow anyone to clear SO_DEBUG
64 * (compatibility fix)
65 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput.
66 * Alan Cox : Allocator for a socket is settable.
67 * Alan Cox : SO_ERROR includes soft errors.
68 * Alan Cox : Allow NULL arguments on some SO_ opts
69 * Alan Cox : Generic socket allocation to make hooks
70 * easier (suggested by Craig Metz).
71 * Michael Pall : SO_ERROR returns positive errno again
72 * Steve Whitehouse: Added default destructor to free
73 * protocol private data.
74 * Steve Whitehouse: Added various other default routines
75 * common to several socket families.
76 * Chris Evans : Call suser() check last on F_SETOWN
77 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER.
78 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s()
79 * Andi Kleen : Fix write_space callback
80 * Chris Evans : Security fixes - signedness again
81 * Arnaldo C. Melo : cleanups, use skb_queue_purge
82 *
83 * To Fix:
84 *
85 *
86 * This program is free software; you can redistribute it and/or
87 * modify it under the terms of the GNU General Public License
88 * as published by the Free Software Foundation; either version
89 * 2 of the License, or (at your option) any later version.
90 */
91
e005d193
JP		 92#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
93
4fc268d2 94#include <linux/capability.h>
1da177e4 95#include <linux/errno.h>
cb820f8e 96#include <linux/errqueue.h>
1da177e4
LT		 97#include <linux/types.h>
98#include <linux/socket.h>
99#include <linux/in.h>
100#include <linux/kernel.h>
1da177e4
LT		 101#include <linux/module.h>
102#include <linux/proc_fs.h>
103#include <linux/seq_file.h>
104#include <linux/sched.h>
105#include <linux/timer.h>
106#include <linux/string.h>
107#include <linux/sockios.h>
108#include <linux/net.h>
109#include <linux/mm.h>
110#include <linux/slab.h>
111#include <linux/interrupt.h>
112#include <linux/poll.h>
113#include <linux/tcp.h>
114#include <linux/init.h>
a1f8e7f7 115#include <linux/highmem.h>
3f551f94 116#include <linux/user_namespace.h>
c5905afb 117#include <linux/static_key.h>
3969eb38 118#include <linux/memcontrol.h>
8c1ae10d 119#include <linux/prefetch.h>
1da177e4
LT		 120
121#include <asm/uaccess.h>
1da177e4
LT		 122
123#include <linux/netdevice.h>
124#include <net/protocol.h>
125#include <linux/skbuff.h>
457c4cbc 126#include <net/net_namespace.h>
2e6599cb 127#include <net/request_sock.h>
1da177e4 128#include <net/sock.h>
20d49473 129#include <linux/net_tstamp.h>
1da177e4
LT		 130#include <net/xfrm.h>
131#include <linux/ipsec.h>
f8451725 132#include <net/cls_cgroup.h>
5bc1421e 133#include <net/netprio_cgroup.h>
1da177e4
LT		 134
135#include <linux/filter.h>
136
3847ce32
SM		 137#include <trace/events/sock.h>
138
1da177e4
LT		 139#ifdef CONFIG_INET
140#include <net/tcp.h>
141#endif
142
076bb0c8 143#include <net/busy_poll.h>
06021292 144
36b77a52 145static DEFINE_MUTEX(proto_list_mutex);
d1a4c0b3
GC		 146static LIST_HEAD(proto_list);
147
a3b299da
EB		 148/**
149 * sk_ns_capable - General socket capability test
150 * @sk: Socket to use a capability on or through
151 * @user_ns: The user namespace of the capability to use
152 * @cap: The capability to use
153 *
154 * Test to see if the opener of the socket had when the socket was
155 * created and the current process has the capability @cap in the user
156 * namespace @user_ns.
157 */
158bool sk_ns_capable(const struct sock *sk,
159 struct user_namespace *user_ns, int cap)
160{
161 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
162 ns_capable(user_ns, cap);
163}
164EXPORT_SYMBOL(sk_ns_capable);
165
166/**
167 * sk_capable - Socket global capability test
168 * @sk: Socket to use a capability on or through
169 * @cap: The global capbility to use
170 *
171 * Test to see if the opener of the socket had when the socket was
172 * created and the current process has the capability @cap in all user
173 * namespaces.
174 */
175bool sk_capable(const struct sock *sk, int cap)
176{
177 return sk_ns_capable(sk, &init_user_ns, cap);
178}
179EXPORT_SYMBOL(sk_capable);
180
181/**
182 * sk_net_capable - Network namespace socket capability test
183 * @sk: Socket to use a capability on or through
184 * @cap: The capability to use
185 *
186 * Test to see if the opener of the socket had when the socke was created
187 * and the current process has the capability @cap over the network namespace
188 * the socket is a member of.
189 */
190bool sk_net_capable(const struct sock *sk, int cap)
191{
192 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
193}
194EXPORT_SYMBOL(sk_net_capable);
195
196
c255a458 197#ifdef CONFIG_MEMCG_KMEM
1d62e436 198int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
d1a4c0b3
GC		 199{
200 struct proto *proto;
201 int ret = 0;
202
36b77a52 203 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 204 list_for_each_entry(proto, &proto_list, node) {
205 if (proto->init_cgroup) {
1d62e436 206 ret = proto->init_cgroup(memcg, ss);
d1a4c0b3
GC		 207 if (ret)
208 goto out;
209 }
210 }
211
36b77a52 212 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 213 return ret;
214out:
215 list_for_each_entry_continue_reverse(proto, &proto_list, node)
216 if (proto->destroy_cgroup)
1d62e436 217 proto->destroy_cgroup(memcg);
36b77a52 218 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 219 return ret;
220}
221
1d62e436 222void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
d1a4c0b3
GC		 223{
224 struct proto *proto;
225
36b77a52 226 mutex_lock(&proto_list_mutex);
d1a4c0b3
GC		 227 list_for_each_entry_reverse(proto, &proto_list, node)
228 if (proto->destroy_cgroup)
1d62e436 229 proto->destroy_cgroup(memcg);
36b77a52 230 mutex_unlock(&proto_list_mutex);
d1a4c0b3
GC		 231}
232#endif
233
da21f24d
IM		 234/*
235 * Each address family might have different locking rules, so we have
236 * one slock key per address family:
237 */
a5b5bb9a
IM		 238static struct lock_class_key af_family_keys[AF_MAX];
239static struct lock_class_key af_family_slock_keys[AF_MAX];
240
cbda4eaf 241#if defined(CONFIG_MEMCG_KMEM)
c5905afb 242struct static_key memcg_socket_limit_enabled;
e1aab161 243EXPORT_SYMBOL(memcg_socket_limit_enabled);
cbda4eaf 244#endif
e1aab161 245
a5b5bb9a
IM		 246/*
247 * Make lock validator output more readable. (we pre-construct these
248 * strings build-time, so that runtime initialization of socket
249 * locks is fast):
250 */
36cbd3dc 251static const char *const af_family_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 252 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
253 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
254 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
255 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
256 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
257 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
258 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
cbd151bf 259 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
a5b5bb9a 260 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
cd05acfe 261 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
17926a79 262 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
bce7b154 263 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
6f107b58 264 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
456db6a4 265 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
a5b5bb9a 266};
36cbd3dc 267static const char *const af_family_slock_key_strings[AF_MAX+1] = {
a5b5bb9a
IM		 268 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
269 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
270 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
271 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
272 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
273 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
274 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
cbd151bf 275 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
a5b5bb9a 276 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
cd05acfe 277 "slock-27" , "slock-28" , "slock-AF_CAN" ,
17926a79 278 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
bce7b154 279 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
6f107b58 280 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
456db6a4 281 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
a5b5bb9a 282};
36cbd3dc 283static const char *const af_family_clock_key_strings[AF_MAX+1] = {
443aef0e
PZ		 284 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
285 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
286 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
287 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
288 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
289 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
290 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
cbd151bf 291 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
443aef0e 292 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
b4942af6 293 "clock-27" , "clock-28" , "clock-AF_CAN" ,
e51f802b 294 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
bce7b154 295 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
6f107b58 296 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
456db6a4 297 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
443aef0e 298};
da21f24d
IM		 299
300/*
301 * sk_callback_lock locking rules are per-address-family,
302 * so split the lock classes by using a per-AF key:
303 */
304static struct lock_class_key af_callback_keys[AF_MAX];
305
1da177e4
LT		 306/* Take into consideration the size of the struct sk_buff overhead in the
307 * determination of these values, since that is non-constant across
308 * platforms. This makes socket queueing behavior and performance
309 * not depend upon such differences.
310 */
311#define _SK_MEM_PACKETS 256
87fb4b7b 312#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
1da177e4
LT		 313#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
314#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
315
316/* Run time adjustable parameters. */
ab32ea5d 317__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
6d8ebc8a 318EXPORT_SYMBOL(sysctl_wmem_max);
ab32ea5d 319__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
6d8ebc8a 320EXPORT_SYMBOL(sysctl_rmem_max);
ab32ea5d
BH		 321__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
322__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
1da177e4 323
25985edc 324/* Maximal space eaten by iovec or ancillary data plus some space */
ab32ea5d 325int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
2a91525c 326EXPORT_SYMBOL(sysctl_optmem_max);
1da177e4 327
c93bdd0e
MG		 328struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
329EXPORT_SYMBOL_GPL(memalloc_socks);
330
7cb02404
MG		 331/**
332 * sk_set_memalloc - sets %SOCK_MEMALLOC
333 * @sk: socket to set it on
334 *
335 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
336 * It's the responsibility of the admin to adjust min_free_kbytes
337 * to meet the requirements
338 */
339void sk_set_memalloc(struct sock *sk)
340{
341 sock_set_flag(sk, SOCK_MEMALLOC);
342 sk->sk_allocation |= __GFP_MEMALLOC;
c93bdd0e 343 static_key_slow_inc(&memalloc_socks);
7cb02404
MG		 344}
345EXPORT_SYMBOL_GPL(sk_set_memalloc);
346
347void sk_clear_memalloc(struct sock *sk)
348{
349 sock_reset_flag(sk, SOCK_MEMALLOC);
350 sk->sk_allocation &= ~__GFP_MEMALLOC;
c93bdd0e 351 static_key_slow_dec(&memalloc_socks);
c76562b6
MG		 352
353 /*
354 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
355 * progress of swapping. However, if SOCK_MEMALLOC is cleared while
356 * it has rmem allocations there is a risk that the user of the
357 * socket cannot make forward progress due to exceeding the rmem
358 * limits. By rights, sk_clear_memalloc() should only be called
359 * on sockets being torn down but warn and reset the accounting if
360 * that assumption breaks.
361 */
362 if (WARN_ON(sk->sk_forward_alloc))
363 sk_mem_reclaim(sk);
7cb02404
MG		 364}
365EXPORT_SYMBOL_GPL(sk_clear_memalloc);
366
b4b9e355
MG		 367int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
368{
369 int ret;
370 unsigned long pflags = current->flags;
371
372 /* these should have been dropped before queueing */
373 BUG_ON(!sock_flag(sk, SOCK_MEMALLOC));
374
375 current->flags |= PF_MEMALLOC;
376 ret = sk->sk_backlog_rcv(sk, skb);
377 tsk_restore_flags(current, pflags, PF_MEMALLOC);
378
379 return ret;
380}
381EXPORT_SYMBOL(__sk_backlog_rcv);
382
1da177e4
LT		 383static int sock_set_timeout(long *timeo_p, char __user *optval, int optlen)
384{
385 struct timeval tv;
386
387 if (optlen < sizeof(tv))
388 return -EINVAL;
389 if (copy_from_user(&tv, optval, sizeof(tv)))
390 return -EFAULT;
ba78073e
VA		 391 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
392 return -EDOM;
1da177e4 393
ba78073e 394 if (tv.tv_sec < 0) {
6f11df83
AM		 395 static int warned __read_mostly;
396
ba78073e 397 *timeo_p = 0;
50aab54f 398 if (warned < 10 && net_ratelimit()) {
ba78073e 399 warned++;
e005d193
JP		 400 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
401 __func__, current->comm, task_pid_nr(current));
50aab54f 402 }
ba78073e
VA		 403 return 0;
404 }
1da177e4
LT		 405 *timeo_p = MAX_SCHEDULE_TIMEOUT;
406 if (tv.tv_sec == 0 && tv.tv_usec == 0)
407 return 0;
408 if (tv.tv_sec < (MAX_SCHEDULE_TIMEOUT/HZ - 1))
409 *timeo_p = tv.tv_sec*HZ + (tv.tv_usec+(1000000/HZ-1))/(1000000/HZ);
410 return 0;
411}
412
413static void sock_warn_obsolete_bsdism(const char *name)
414{
415 static int warned;
416 static char warncomm[TASK_COMM_LEN];
4ec93edb
YH		 417 if (strcmp(warncomm, current->comm) && warned < 5) {
418 strcpy(warncomm, current->comm);
e005d193
JP		 419 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
420 warncomm, name);
1da177e4
LT		 421 warned++;
422 }
423}
424
08e29af3
ED		 425#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
426
427static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
4ec93edb 428{
08e29af3
ED		 429 if (sk->sk_flags & flags) {
430 sk->sk_flags &= ~flags;
431 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
20d49473 432 net_disable_timestamp();
1da177e4
LT		 433 }
434}
435
436
f0088a50
DV		 437int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
438{
766e9037 439 int err;
f0088a50 440 int skb_len;
3b885787
NH		 441 unsigned long flags;
442 struct sk_buff_head *list = &sk->sk_receive_queue;
f0088a50 443
0fd7bac6 444 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
766e9037 445 atomic_inc(&sk->sk_drops);
3847ce32 446 trace_sock_rcvqueue_full(sk, skb);
766e9037 447 return -ENOMEM;
f0088a50
DV		 448 }
449
fda9ef5d 450 err = sk_filter(sk, skb);
f0088a50 451 if (err)
766e9037 452 return err;
f0088a50 453
c76562b6 454 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
766e9037
ED		 455 atomic_inc(&sk->sk_drops);
456 return -ENOBUFS;
3ab224be
HA		 457 }
458
f0088a50
DV		 459 skb->dev = NULL;
460 skb_set_owner_r(skb, sk);
49ad9599 461
f0088a50
DV		 462 /* Cache the SKB length before we tack it onto the receive
463 * queue. Once it is added it no longer belongs to us and
464 * may be freed by other threads of control pulling packets
465 * from the queue.
466 */
467 skb_len = skb->len;
468
7fee226a
ED		 469 /* we escape from rcu protected region, make sure we dont leak
470 * a norefcounted dst
471 */
472 skb_dst_force(skb);
473
3b885787
NH		 474 spin_lock_irqsave(&list->lock, flags);
475 skb->dropcount = atomic_read(&sk->sk_drops);
476 __skb_queue_tail(list, skb);
477 spin_unlock_irqrestore(&list->lock, flags);
f0088a50
DV		 478
479 if (!sock_flag(sk, SOCK_DEAD))
676d2369 480 sk->sk_data_ready(sk);
766e9037 481 return 0;
f0088a50
DV		 482}
483EXPORT_SYMBOL(sock_queue_rcv_skb);
484
58a5a7b9 485int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
f0088a50
DV		 486{
487 int rc = NET_RX_SUCCESS;
488
fda9ef5d 489 if (sk_filter(sk, skb))
f0088a50
DV		 490 goto discard_and_relse;
491
492 skb->dev = NULL;
493
274f482d 494 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
c377411f
ED		 495 atomic_inc(&sk->sk_drops);
496 goto discard_and_relse;
497 }
58a5a7b9
ACM		 498 if (nested)
499 bh_lock_sock_nested(sk);
500 else
501 bh_lock_sock(sk);
a5b5bb9a
IM		 502 if (!sock_owned_by_user(sk)) {
503 /*
504 * trylock + unlock semantics:
505 */
506 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
507
c57943a1 508 rc = sk_backlog_rcv(sk, skb);
a5b5bb9a
IM		 509
510 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
f545a38f 511 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
8eae939f
ZY		 512 bh_unlock_sock(sk);
513 atomic_inc(&sk->sk_drops);
514 goto discard_and_relse;
515 }
516
f0088a50
DV		 517 bh_unlock_sock(sk);
518out:
519 sock_put(sk);
520 return rc;
521discard_and_relse:
522 kfree_skb(skb);
523 goto out;
524}
525EXPORT_SYMBOL(sk_receive_skb);
526
527struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
528{
b6c6712a 529 struct dst_entry *dst = __sk_dst_get(sk);
f0088a50
DV		 530
531 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
e022f0b4 532 sk_tx_queue_clear(sk);
a9b3cd7f 533 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
f0088a50
DV		 534 dst_release(dst);
535 return NULL;
536 }
537
538 return dst;
539}
540EXPORT_SYMBOL(__sk_dst_check);
541
542struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
543{
544 struct dst_entry *dst = sk_dst_get(sk);
545
546 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
547 sk_dst_reset(sk);
548 dst_release(dst);
549 return NULL;
550 }
551
552 return dst;
553}
554EXPORT_SYMBOL(sk_dst_check);
555
c91f6df2
BH		 556static int sock_setbindtodevice(struct sock *sk, char __user *optval,
557 int optlen)
4878809f
DM		 558{
559 int ret = -ENOPROTOOPT;
560#ifdef CONFIG_NETDEVICES
3b1e0a65 561 struct net *net = sock_net(sk);
4878809f
DM		 562 char devname[IFNAMSIZ];
563 int index;
564
565 /* Sorry... */
566 ret = -EPERM;
5e1fccc0 567 if (!ns_capable(net->user_ns, CAP_NET_RAW))
4878809f
DM		 568 goto out;
569
570 ret = -EINVAL;
571 if (optlen < 0)
572 goto out;
573
574 /* Bind this socket to a particular device like "eth0",
575 * as specified in the passed interface name. If the
576 * name is "" or the option length is zero the socket
577 * is not bound.
578 */
579 if (optlen > IFNAMSIZ - 1)
580 optlen = IFNAMSIZ - 1;
581 memset(devname, 0, sizeof(devname));
582
583 ret = -EFAULT;
584 if (copy_from_user(devname, optval, optlen))
585 goto out;
586
000ba2e4
DM		 587 index = 0;
588 if (devname[0] != '\0') {
bf8e56bf 589 struct net_device *dev;
4878809f 590
bf8e56bf
ED		 591 rcu_read_lock();
592 dev = dev_get_by_name_rcu(net, devname);
593 if (dev)
594 index = dev->ifindex;
595 rcu_read_unlock();
4878809f
DM		 596 ret = -ENODEV;
597 if (!dev)
598 goto out;
4878809f
DM		 599 }
600
601 lock_sock(sk);
602 sk->sk_bound_dev_if = index;
603 sk_dst_reset(sk);
604 release_sock(sk);
605
606 ret = 0;
607
608out:
609#endif
610
611 return ret;
612}
613
c91f6df2
BH		 614static int sock_getbindtodevice(struct sock *sk, char __user *optval,
615 int __user *optlen, int len)
616{
617 int ret = -ENOPROTOOPT;
618#ifdef CONFIG_NETDEVICES
619 struct net *net = sock_net(sk);
c91f6df2 620 char devname[IFNAMSIZ];
c91f6df2
BH		 621
622 if (sk->sk_bound_dev_if == 0) {
623 len = 0;
624 goto zero;
625 }
626
627 ret = -EINVAL;
628 if (len < IFNAMSIZ)
629 goto out;
630
5dbe7c17
NS		 631 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
632 if (ret)
c91f6df2 633 goto out;
c91f6df2
BH		 634
635 len = strlen(devname) + 1;
636
637 ret = -EFAULT;
638 if (copy_to_user(optval, devname, len))
639 goto out;
640
641zero:
642 ret = -EFAULT;
643 if (put_user(len, optlen))
644 goto out;
645
646 ret = 0;
647
648out:
649#endif
650
651 return ret;
652}
653
c0ef877b
PE		 654static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
655{
656 if (valbool)
657 sock_set_flag(sk, bit);
658 else
659 sock_reset_flag(sk, bit);
660}
661
1da177e4
LT		 662/*
663 * This is meant for all protocols to use and covers goings on
664 * at the socket level. Everything here is generic.
665 */
666
667int sock_setsockopt(struct socket *sock, int level, int optname,
b7058842 668 char __user *optval, unsigned int optlen)
1da177e4 669{
2a91525c 670 struct sock *sk = sock->sk;
1da177e4
LT		 671 int val;
672 int valbool;
673 struct linger ling;
674 int ret = 0;
4ec93edb 675
1da177e4
LT		 676 /*
677 * Options without arguments
678 */
679
4878809f 680 if (optname == SO_BINDTODEVICE)
c91f6df2 681 return sock_setbindtodevice(sk, optval, optlen);
4878809f 682
e71a4783
SH		 683 if (optlen < sizeof(int))
684 return -EINVAL;
4ec93edb 685
1da177e4
LT		 686 if (get_user(val, (int __user *)optval))
687 return -EFAULT;
4ec93edb 688
2a91525c 689 valbool = val ? 1 : 0;
1da177e4
LT		 690
691 lock_sock(sk);
692
2a91525c 693 switch (optname) {
e71a4783 694 case SO_DEBUG:
2a91525c 695 if (val && !capable(CAP_NET_ADMIN))
e71a4783 696 ret = -EACCES;
2a91525c 697 else
c0ef877b 698 sock_valbool_flag(sk, SOCK_DBG, valbool);
e71a4783
SH		 699 break;
700 case SO_REUSEADDR:
4a17fd52 701 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
e71a4783 702 break;
055dc21a
TH		 703 case SO_REUSEPORT:
704 sk->sk_reuseport = valbool;
705 break;
e71a4783 706 case SO_TYPE:
49c794e9 707 case SO_PROTOCOL:
0d6038ee 708 case SO_DOMAIN:
e71a4783
SH		 709 case SO_ERROR:
710 ret = -ENOPROTOOPT;
711 break;
712 case SO_DONTROUTE:
c0ef877b 713 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
e71a4783
SH		 714 break;
715 case SO_BROADCAST:
716 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
717 break;
718 case SO_SNDBUF:
719 /* Don't error on this BSD doesn't and if you think
82981930
ED		 720 * about it this is right. Otherwise apps have to
721 * play 'guess the biggest size' games. RCVBUF/SNDBUF
722 * are treated in BSD as hints
723 */
724 val = min_t(u32, val, sysctl_wmem_max);
b0573dea 725set_sndbuf:
e71a4783 726 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
82981930
ED		 727 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
728 /* Wake up sending tasks if we upped the value. */
e71a4783
SH		 729 sk->sk_write_space(sk);
730 break;
1da177e4 731
e71a4783
SH		 732 case SO_SNDBUFFORCE:
733 if (!capable(CAP_NET_ADMIN)) {
734 ret = -EPERM;
735 break;
736 }
737 goto set_sndbuf;
b0573dea 738
e71a4783
SH		 739 case SO_RCVBUF:
740 /* Don't error on this BSD doesn't and if you think
82981930
ED		 741 * about it this is right. Otherwise apps have to
742 * play 'guess the biggest size' games. RCVBUF/SNDBUF
743 * are treated in BSD as hints
744 */
745 val = min_t(u32, val, sysctl_rmem_max);
b0573dea 746set_rcvbuf:
e71a4783
SH		 747 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
748 /*
749 * We double it on the way in to account for
750 * "struct sk_buff" etc. overhead. Applications
751 * assume that the SO_RCVBUF setting they make will
752 * allow that much actual data to be received on that
753 * socket.
754 *
755 * Applications are unaware that "struct sk_buff" and
756 * other overheads allocate from the receive buffer
757 * during socket buffer allocation.
758 *
759 * And after considering the possible alternatives,
760 * returning the value we actually used in getsockopt
761 * is the most desirable behavior.
762 */
82981930 763 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
e71a4783
SH		 764 break;
765
766 case SO_RCVBUFFORCE:
767 if (!capable(CAP_NET_ADMIN)) {
768 ret = -EPERM;
1da177e4 769 break;
e71a4783
SH		 770 }
771 goto set_rcvbuf;
1da177e4 772
e71a4783 773 case SO_KEEPALIVE:
1da177e4 774#ifdef CONFIG_INET
3e10986d
ED		 775 if (sk->sk_protocol == IPPROTO_TCP &&
776 sk->sk_type == SOCK_STREAM)
e71a4783 777 tcp_set_keepalive(sk, valbool);
1da177e4 778#endif
e71a4783
SH		 779 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
780 break;
781
782 case SO_OOBINLINE:
783 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
784 break;
785
786 case SO_NO_CHECK:
28448b80 787 sk->sk_no_check_tx = valbool;
e71a4783
SH		 788 break;
789
790 case SO_PRIORITY:
5e1fccc0
EB		 791 if ((val >= 0 && val <= 6) ||
792 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
e71a4783
SH		 793 sk->sk_priority = val;
794 else
795 ret = -EPERM;
796 break;
797
798 case SO_LINGER:
799 if (optlen < sizeof(ling)) {
800 ret = -EINVAL; /* 1003.1g */
1da177e4 801 break;
e71a4783 802 }
2a91525c 803 if (copy_from_user(&ling, optval, sizeof(ling))) {
e71a4783 804 ret = -EFAULT;
1da177e4 805 break;
e71a4783
SH		 806 }
807 if (!ling.l_onoff)
808 sock_reset_flag(sk, SOCK_LINGER);
809 else {
1da177e4 810#if (BITS_PER_LONG == 32)
e71a4783
SH		 811 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
812 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
1da177e4 813 else
e71a4783
SH		 814#endif
815 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
816 sock_set_flag(sk, SOCK_LINGER);
817 }
818 break;
819
820 case SO_BSDCOMPAT:
821 sock_warn_obsolete_bsdism("setsockopt");
822 break;
823
824 case SO_PASSCRED:
825 if (valbool)
826 set_bit(SOCK_PASSCRED, &sock->flags);
827 else
828 clear_bit(SOCK_PASSCRED, &sock->flags);
829 break;
830
831 case SO_TIMESTAMP:
92f37fd2 832 case SO_TIMESTAMPNS:
e71a4783 833 if (valbool) {
92f37fd2
ED		 834 if (optname == SO_TIMESTAMP)
835 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
836 else
837 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783 838 sock_set_flag(sk, SOCK_RCVTSTAMP);
20d49473 839 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
92f37fd2 840 } else {
e71a4783 841 sock_reset_flag(sk, SOCK_RCVTSTAMP);
92f37fd2
ED		 842 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
843 }
e71a4783
SH		 844 break;
845
20d49473
PO		 846 case SO_TIMESTAMPING:
847 if (val & ~SOF_TIMESTAMPING_MASK) {
f249fb78 848 ret = -EINVAL;
20d49473
PO		 849 break;
850 }
851 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE,
852 val & SOF_TIMESTAMPING_TX_HARDWARE);
853 sock_valbool_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE,
854 val & SOF_TIMESTAMPING_TX_SOFTWARE);
855 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE,
856 val & SOF_TIMESTAMPING_RX_HARDWARE);
857 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
858 sock_enable_timestamp(sk,
859 SOCK_TIMESTAMPING_RX_SOFTWARE);
860 else
861 sock_disable_timestamp(sk,
08e29af3 862 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
20d49473
PO		 863 sock_valbool_flag(sk, SOCK_TIMESTAMPING_SOFTWARE,
864 val & SOF_TIMESTAMPING_SOFTWARE);
20d49473
PO		 865 sock_valbool_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE,
866 val & SOF_TIMESTAMPING_RAW_HARDWARE);
867 break;
868
e71a4783
SH		 869 case SO_RCVLOWAT:
870 if (val < 0)
871 val = INT_MAX;
872 sk->sk_rcvlowat = val ? : 1;
873 break;
874
875 case SO_RCVTIMEO:
876 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
877 break;
878
879 case SO_SNDTIMEO:
880 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
881 break;
1da177e4 882
e71a4783
SH		 883 case SO_ATTACH_FILTER:
884 ret = -EINVAL;
885 if (optlen == sizeof(struct sock_fprog)) {
886 struct sock_fprog fprog;
1da177e4 887
e71a4783
SH		 888 ret = -EFAULT;
889 if (copy_from_user(&fprog, optval, sizeof(fprog)))
1da177e4 890 break;
e71a4783
SH		 891
892 ret = sk_attach_filter(&fprog, sk);
893 }
894 break;
895
896 case SO_DETACH_FILTER:
55b33325 897 ret = sk_detach_filter(sk);
e71a4783 898 break;
1da177e4 899
d59577b6
VB		 900 case SO_LOCK_FILTER:
901 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
902 ret = -EPERM;
903 else
904 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
905 break;
906
e71a4783
SH		 907 case SO_PASSSEC:
908 if (valbool)
909 set_bit(SOCK_PASSSEC, &sock->flags);
910 else
911 clear_bit(SOCK_PASSSEC, &sock->flags);
912 break;
4a19ec58 913 case SO_MARK:
5e1fccc0 914 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
4a19ec58 915 ret = -EPERM;
2a91525c 916 else
4a19ec58 917 sk->sk_mark = val;
4a19ec58 918 break;
877ce7c1 919
1da177e4
LT		 920 /* We implement the SO_SNDLOWAT etc to
921 not be settable (1003.1g 5.3) */
3b885787 922 case SO_RXQ_OVFL:
8083f0fc 923 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
3b885787 924 break;
6e3e939f
JB		 925
926 case SO_WIFI_STATUS:
927 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
928 break;
929
ef64a54f
PE		 930 case SO_PEEK_OFF:
931 if (sock->ops->set_peek_off)
12663bfc 932 ret = sock->ops->set_peek_off(sk, val);
ef64a54f
PE		 933 else
934 ret = -EOPNOTSUPP;
935 break;
3bdc0eba
BG		 936
937 case SO_NOFCS:
938 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
939 break;
940
7d4c04fc
KJ		 941 case SO_SELECT_ERR_QUEUE:
942 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
943 break;
944
e0d1095a 945#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 946 case SO_BUSY_POLL:
dafcc438
ET		 947 /* allow unprivileged users to decrease the value */
948 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
949 ret = -EPERM;
950 else {
951 if (val < 0)
952 ret = -EINVAL;
953 else
954 sk->sk_ll_usec = val;
955 }
956 break;
957#endif
62748f32
ED		 958
959 case SO_MAX_PACING_RATE:
960 sk->sk_max_pacing_rate = val;
961 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
962 sk->sk_max_pacing_rate);
963 break;
964
e71a4783
SH		 965 default:
966 ret = -ENOPROTOOPT;
967 break;
4ec93edb 968 }
1da177e4
LT		 969 release_sock(sk);
970 return ret;
971}
2a91525c 972EXPORT_SYMBOL(sock_setsockopt);
1da177e4
LT		 973
974
8f09898b 975static void cred_to_ucred(struct pid *pid, const struct cred *cred,
976 struct ucred *ucred)
3f551f94
EB		 977{
978 ucred->pid = pid_vnr(pid);
979 ucred->uid = ucred->gid = -1;
980 if (cred) {
981 struct user_namespace *current_ns = current_user_ns();
982
b2e4f544
EB		 983 ucred->uid = from_kuid_munged(current_ns, cred->euid);
984 ucred->gid = from_kgid_munged(current_ns, cred->egid);
3f551f94
EB		 985 }
986}
987
1da177e4
LT		 988int sock_getsockopt(struct socket *sock, int level, int optname,
989 char __user *optval, int __user *optlen)
990{
991 struct sock *sk = sock->sk;
4ec93edb 992
e71a4783 993 union {
4ec93edb
YH		 994 int val;
995 struct linger ling;
1da177e4
LT		 996 struct timeval tm;
997 } v;
4ec93edb 998
4d0392be 999 int lv = sizeof(int);
1da177e4 1000 int len;
4ec93edb 1001
e71a4783 1002 if (get_user(len, optlen))
4ec93edb 1003 return -EFAULT;
e71a4783 1004 if (len < 0)
1da177e4 1005 return -EINVAL;
4ec93edb 1006
50fee1de 1007 memset(&v, 0, sizeof(v));
df0bca04 1008
2a91525c 1009 switch (optname) {
e71a4783
SH		 1010 case SO_DEBUG:
1011 v.val = sock_flag(sk, SOCK_DBG);
1012 break;
1013
1014 case SO_DONTROUTE:
1015 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1016 break;
1017
1018 case SO_BROADCAST:
1b23a5df 1019 v.val = sock_flag(sk, SOCK_BROADCAST);
e71a4783
SH		 1020 break;
1021
1022 case SO_SNDBUF:
1023 v.val = sk->sk_sndbuf;
1024 break;
1025
1026 case SO_RCVBUF:
1027 v.val = sk->sk_rcvbuf;
1028 break;
1029
1030 case SO_REUSEADDR:
1031 v.val = sk->sk_reuse;
1032 break;
1033
055dc21a
TH		 1034 case SO_REUSEPORT:
1035 v.val = sk->sk_reuseport;
1036 break;
1037
e71a4783 1038 case SO_KEEPALIVE:
1b23a5df 1039 v.val = sock_flag(sk, SOCK_KEEPOPEN);
e71a4783
SH		 1040 break;
1041
1042 case SO_TYPE:
1043 v.val = sk->sk_type;
1044 break;
1045
49c794e9
JE		 1046 case SO_PROTOCOL:
1047 v.val = sk->sk_protocol;
1048 break;
1049
0d6038ee
JE		 1050 case SO_DOMAIN:
1051 v.val = sk->sk_family;
1052 break;
1053
e71a4783
SH		 1054 case SO_ERROR:
1055 v.val = -sock_error(sk);
2a91525c 1056 if (v.val == 0)
e71a4783
SH		 1057 v.val = xchg(&sk->sk_err_soft, 0);
1058 break;
1059
1060 case SO_OOBINLINE:
1b23a5df 1061 v.val = sock_flag(sk, SOCK_URGINLINE);
e71a4783
SH		 1062 break;
1063
1064 case SO_NO_CHECK:
28448b80 1065 v.val = sk->sk_no_check_tx;
e71a4783
SH		 1066 break;
1067
1068 case SO_PRIORITY:
1069 v.val = sk->sk_priority;
1070 break;
1071
1072 case SO_LINGER:
1073 lv = sizeof(v.ling);
1b23a5df 1074 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
e71a4783
SH		 1075 v.ling.l_linger = sk->sk_lingertime / HZ;
1076 break;
1077
1078 case SO_BSDCOMPAT:
1079 sock_warn_obsolete_bsdism("getsockopt");
1080 break;
1081
1082 case SO_TIMESTAMP:
92f37fd2
ED		 1083 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1084 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1085 break;
1086
1087 case SO_TIMESTAMPNS:
1088 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
e71a4783
SH		 1089 break;
1090
20d49473
PO		 1091 case SO_TIMESTAMPING:
1092 v.val = 0;
1093 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_HARDWARE))
1094 v.val |= SOF_TIMESTAMPING_TX_HARDWARE;
1095 if (sock_flag(sk, SOCK_TIMESTAMPING_TX_SOFTWARE))
1096 v.val |= SOF_TIMESTAMPING_TX_SOFTWARE;
1097 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_HARDWARE))
1098 v.val |= SOF_TIMESTAMPING_RX_HARDWARE;
1099 if (sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE))
1100 v.val |= SOF_TIMESTAMPING_RX_SOFTWARE;
1101 if (sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE))
1102 v.val |= SOF_TIMESTAMPING_SOFTWARE;
20d49473
PO		 1103 if (sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE))
1104 v.val |= SOF_TIMESTAMPING_RAW_HARDWARE;
1105 break;
1106
e71a4783 1107 case SO_RCVTIMEO:
2a91525c 1108 lv = sizeof(struct timeval);
e71a4783
SH		 1109 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1110 v.tm.tv_sec = 0;
1111 v.tm.tv_usec = 0;
1112 } else {
1113 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1114 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
1115 }
1116 break;
1117
1118 case SO_SNDTIMEO:
2a91525c 1119 lv = sizeof(struct timeval);
e71a4783
SH		 1120 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1121 v.tm.tv_sec = 0;
1122 v.tm.tv_usec = 0;
1123 } else {
1124 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1125 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1126 }
1127 break;
1da177e4 1128
e71a4783
SH		 1129 case SO_RCVLOWAT:
1130 v.val = sk->sk_rcvlowat;
1131 break;
1da177e4 1132
e71a4783 1133 case SO_SNDLOWAT:
2a91525c 1134 v.val = 1;
e71a4783 1135 break;
1da177e4 1136
e71a4783 1137 case SO_PASSCRED:
82981930 1138 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
e71a4783 1139 break;
1da177e4 1140
e71a4783 1141 case SO_PEERCRED:
109f6e39
EB		 1142 {
1143 struct ucred peercred;
1144 if (len > sizeof(peercred))
1145 len = sizeof(peercred);
1146 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1147 if (copy_to_user(optval, &peercred, len))
e71a4783
SH		 1148 return -EFAULT;
1149 goto lenout;
109f6e39 1150 }
1da177e4 1151
e71a4783
SH		 1152 case SO_PEERNAME:
1153 {
1154 char address[128];
1155
1156 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1157 return -ENOTCONN;
1158 if (lv < len)
1159 return -EINVAL;
1160 if (copy_to_user(optval, address, len))
1161 return -EFAULT;
1162 goto lenout;
1163 }
1da177e4 1164
e71a4783
SH		 1165 /* Dubious BSD thing... Probably nobody even uses it, but
1166 * the UNIX standard wants it for whatever reason... -DaveM
1167 */
1168 case SO_ACCEPTCONN:
1169 v.val = sk->sk_state == TCP_LISTEN;
1170 break;
1da177e4 1171
e71a4783 1172 case SO_PASSSEC:
82981930 1173 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
e71a4783 1174 break;
877ce7c1 1175
e71a4783
SH		 1176 case SO_PEERSEC:
1177 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1da177e4 1178
4a19ec58
LAT		 1179 case SO_MARK:
1180 v.val = sk->sk_mark;
1181 break;
1182
3b885787 1183 case SO_RXQ_OVFL:
1b23a5df 1184 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
3b885787
NH		 1185 break;
1186
6e3e939f 1187 case SO_WIFI_STATUS:
1b23a5df 1188 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
6e3e939f
JB		 1189 break;
1190
ef64a54f
PE		 1191 case SO_PEEK_OFF:
1192 if (!sock->ops->set_peek_off)
1193 return -EOPNOTSUPP;
1194
1195 v.val = sk->sk_peek_off;
1196 break;
bc2f7996 1197 case SO_NOFCS:
1b23a5df 1198 v.val = sock_flag(sk, SOCK_NOFCS);
bc2f7996 1199 break;
c91f6df2 1200
f7b86bfe 1201 case SO_BINDTODEVICE:
c91f6df2
BH		 1202 return sock_getbindtodevice(sk, optval, optlen, len);
1203
a8fc9277
PE		 1204 case SO_GET_FILTER:
1205 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1206 if (len < 0)
1207 return len;
1208
1209 goto lenout;
c91f6df2 1210
d59577b6
VB		 1211 case SO_LOCK_FILTER:
1212 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1213 break;
1214
ea02f941
MS		 1215 case SO_BPF_EXTENSIONS:
1216 v.val = bpf_tell_extensions();
1217 break;
1218
7d4c04fc
KJ		 1219 case SO_SELECT_ERR_QUEUE:
1220 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1221 break;
1222
e0d1095a 1223#ifdef CONFIG_NET_RX_BUSY_POLL
64b0dc51 1224 case SO_BUSY_POLL:
dafcc438
ET		 1225 v.val = sk->sk_ll_usec;
1226 break;
1227#endif
1228
62748f32
ED		 1229 case SO_MAX_PACING_RATE:
1230 v.val = sk->sk_max_pacing_rate;
1231 break;
1232
e71a4783
SH		 1233 default:
1234 return -ENOPROTOOPT;
1da177e4 1235 }
e71a4783 1236
1da177e4
LT		 1237 if (len > lv)
1238 len = lv;
1239 if (copy_to_user(optval, &v, len))
1240 return -EFAULT;
1241lenout:
4ec93edb
YH		 1242 if (put_user(len, optlen))
1243 return -EFAULT;
1244 return 0;
1da177e4
LT		 1245}
1246
a5b5bb9a
IM		 1247/*
1248 * Initialize an sk_lock.
1249 *
1250 * (We also register the sk_lock with the lock validator.)
1251 */
b6f99a21 1252static inline void sock_lock_init(struct sock *sk)
a5b5bb9a 1253{
ed07536e
PZ		 1254 sock_lock_init_class_and_name(sk,
1255 af_family_slock_key_strings[sk->sk_family],
1256 af_family_slock_keys + sk->sk_family,
1257 af_family_key_strings[sk->sk_family],
1258 af_family_keys + sk->sk_family);
a5b5bb9a
IM		 1259}
1260
4dc6dc71
ED		 1261/*
1262 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1263 * even temporarly, because of RCU lookups. sk_node should also be left as is.
68835aba 1264 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
4dc6dc71 1265 */
f1a6c4da
PE		 1266static void sock_copy(struct sock *nsk, const struct sock *osk)
1267{
1268#ifdef CONFIG_SECURITY_NETWORK
1269 void *sptr = nsk->sk_security;
1270#endif
68835aba
ED		 1271 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1272
1273 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1274 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1275
f1a6c4da
PE		 1276#ifdef CONFIG_SECURITY_NETWORK
1277 nsk->sk_security = sptr;
1278 security_sk_clone(osk, nsk);
1279#endif
1280}
1281
fcbdf09d
OP		 1282void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1283{
1284 unsigned long nulls1, nulls2;
1285
1286 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1287 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1288 if (nulls1 > nulls2)
1289 swap(nulls1, nulls2);
1290
1291 if (nulls1 != 0)
1292 memset((char *)sk, 0, nulls1);
1293 memset((char *)sk + nulls1 + sizeof(void *), 0,
1294 nulls2 - nulls1 - sizeof(void *));
1295 memset((char *)sk + nulls2 + sizeof(void *), 0,
1296 size - nulls2 - sizeof(void *));
1297}
1298EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1299
2e4afe7b
PE		 1300static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1301 int family)
c308c1b2
PE		 1302{
1303 struct sock *sk;
1304 struct kmem_cache *slab;
1305
1306 slab = prot->slab;
e912b114
ED		 1307 if (slab != NULL) {
1308 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1309 if (!sk)
1310 return sk;
1311 if (priority & __GFP_ZERO) {
fcbdf09d
OP		 1312 if (prot->clear_sk)
1313 prot->clear_sk(sk, prot->obj_size);
1314 else
1315 sk_prot_clear_nulls(sk, prot->obj_size);
e912b114 1316 }
fcbdf09d 1317 } else
c308c1b2
PE		 1318 sk = kmalloc(prot->obj_size, priority);
1319
2e4afe7b 1320 if (sk != NULL) {
a98b65a3
VN		 1321 kmemcheck_annotate_bitfield(sk, flags);
1322
2e4afe7b
PE		 1323 if (security_sk_alloc(sk, family, priority))
1324 goto out_free;
1325
1326 if (!try_module_get(prot->owner))
1327 goto out_free_sec;
e022f0b4 1328 sk_tx_queue_clear(sk);
2e4afe7b
PE		 1329 }
1330
c308c1b2 1331 return sk;
2e4afe7b
PE		 1332
1333out_free_sec:
1334 security_sk_free(sk);
1335out_free:
1336 if (slab != NULL)
1337 kmem_cache_free(slab, sk);
1338 else
1339 kfree(sk);
1340 return NULL;
c308c1b2
PE		 1341}
1342
1343static void sk_prot_free(struct proto *prot, struct sock *sk)
1344{
1345 struct kmem_cache *slab;
2e4afe7b 1346 struct module *owner;
c308c1b2 1347
2e4afe7b 1348 owner = prot->owner;
c308c1b2 1349 slab = prot->slab;
2e4afe7b
PE		 1350
1351 security_sk_free(sk);
c308c1b2
PE		 1352 if (slab != NULL)
1353 kmem_cache_free(slab, sk);
1354 else
1355 kfree(sk);
2e4afe7b 1356 module_put(owner);
c308c1b2
PE		 1357}
1358
86f8515f 1359#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
6ffd4641 1360void sock_update_netprioidx(struct sock *sk)
5bc1421e 1361{
5bc1421e
NH		 1362 if (in_interrupt())
1363 return;
2b73bc65 1364
6ffd4641 1365 sk->sk_cgrp_prioidx = task_netprioidx(current);
5bc1421e
NH		 1366}
1367EXPORT_SYMBOL_GPL(sock_update_netprioidx);
f8451725
HX		 1368#endif
1369
1da177e4
LT		 1370/**
1371 * sk_alloc - All socket objects are allocated here
c4ea43c5 1372 * @net: the applicable net namespace
4dc3b16b
PP		 1373 * @family: protocol family
1374 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1375 * @prot: struct proto associated with this new sock instance
1da177e4 1376 */
1b8d7ae4 1377struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
6257ff21 1378 struct proto *prot)
1da177e4 1379{
c308c1b2 1380 struct sock *sk;
1da177e4 1381
154adbc8 1382 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
1da177e4 1383 if (sk) {
154adbc8
PE		 1384 sk->sk_family = family;
1385 /*
1386 * See comment in struct sock definition to understand
1387 * why we need sk_prot_creator -acme
1388 */
1389 sk->sk_prot = sk->sk_prot_creator = prot;
1390 sock_lock_init(sk);
3b1e0a65 1391 sock_net_set(sk, get_net(net));
d66ee058 1392 atomic_set(&sk->sk_wmem_alloc, 1);
f8451725 1393
211d2f97 1394 sock_update_classid(sk);
6ffd4641 1395 sock_update_netprioidx(sk);
1da177e4 1396 }
a79af59e 1397
2e4afe7b 1398 return sk;
1da177e4 1399}
2a91525c 1400EXPORT_SYMBOL(sk_alloc);
1da177e4 1401
2b85a34e 1402static void __sk_free(struct sock *sk)
1da177e4
LT		 1403{
1404 struct sk_filter *filter;
1da177e4
LT		 1405
1406 if (sk->sk_destruct)
1407 sk->sk_destruct(sk);
1408
a898def2
PM		 1409 filter = rcu_dereference_check(sk->sk_filter,
1410 atomic_read(&sk->sk_wmem_alloc) == 0);
1da177e4 1411 if (filter) {
309dd5fc 1412 sk_filter_uncharge(sk, filter);
a9b3cd7f 1413 RCU_INIT_POINTER(sk->sk_filter, NULL);
1da177e4
LT		 1414 }
1415
08e29af3 1416 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
1da177e4
LT		 1417
1418 if (atomic_read(&sk->sk_omem_alloc))
e005d193
JP		 1419 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1420 __func__, atomic_read(&sk->sk_omem_alloc));
1da177e4 1421
109f6e39
EB		 1422 if (sk->sk_peer_cred)
1423 put_cred(sk->sk_peer_cred);
1424 put_pid(sk->sk_peer_pid);
3b1e0a65 1425 put_net(sock_net(sk));
c308c1b2 1426 sk_prot_free(sk->sk_prot_creator, sk);
1da177e4 1427}
2b85a34e
ED		 1428
1429void sk_free(struct sock *sk)
1430{
1431 /*
25985edc 1432 * We subtract one from sk_wmem_alloc and can know if
2b85a34e
ED		 1433 * some packets are still in some tx queue.
1434 * If not null, sock_wfree() will call __sk_free(sk) later
1435 */
1436 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1437 __sk_free(sk);
1438}
2a91525c 1439EXPORT_SYMBOL(sk_free);
1da177e4 1440
edf02087 1441/*
25985edc
LDM		 1442 * Last sock_put should drop reference to sk->sk_net. It has already
1443 * been dropped in sk_change_net. Taking reference to stopping namespace
edf02087 1444 * is not an option.
25985edc 1445 * Take reference to a socket to remove it from hash _alive_ and after that
edf02087
DL		 1446 * destroy it in the context of init_net.
1447 */
1448void sk_release_kernel(struct sock *sk)
1449{
1450 if (sk == NULL || sk->sk_socket == NULL)
1451 return;
1452
1453 sock_hold(sk);
1454 sock_release(sk->sk_socket);
65a18ec5 1455 release_net(sock_net(sk));
3b1e0a65 1456 sock_net_set(sk, get_net(&init_net));
edf02087
DL		 1457 sock_put(sk);
1458}
45af1754 1459EXPORT_SYMBOL(sk_release_kernel);
edf02087 1460
475f1b52
SR		 1461static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1462{
1463 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1464 sock_update_memcg(newsk);
1465}
1466
e56c57d0
ED		 1467/**
1468 * sk_clone_lock - clone a socket, and lock its clone
1469 * @sk: the socket to clone
1470 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1471 *
1472 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1473 */
1474struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
87d11ceb 1475{
8fd1d178 1476 struct sock *newsk;
278571ba 1477 bool is_charged = true;
87d11ceb 1478
8fd1d178 1479 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
87d11ceb
ACM		 1480 if (newsk != NULL) {
1481 struct sk_filter *filter;
1482
892c141e 1483 sock_copy(newsk, sk);
87d11ceb
ACM		 1484
1485 /* SANITY */
3b1e0a65 1486 get_net(sock_net(newsk));
87d11ceb
ACM		 1487 sk_node_init(&newsk->sk_node);
1488 sock_lock_init(newsk);
1489 bh_lock_sock(newsk);
fa438ccf 1490 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
8eae939f 1491 newsk->sk_backlog.len = 0;
87d11ceb
ACM		 1492
1493 atomic_set(&newsk->sk_rmem_alloc, 0);
2b85a34e
ED		 1494 /*
1495 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1496 */
1497 atomic_set(&newsk->sk_wmem_alloc, 1);
87d11ceb
ACM		 1498 atomic_set(&newsk->sk_omem_alloc, 0);
1499 skb_queue_head_init(&newsk->sk_receive_queue);
1500 skb_queue_head_init(&newsk->sk_write_queue);
97fc2f08
CL		 1501#ifdef CONFIG_NET_DMA
1502 skb_queue_head_init(&newsk->sk_async_wait_queue);
1503#endif
87d11ceb 1504
b6c6712a 1505 spin_lock_init(&newsk->sk_dst_lock);
87d11ceb 1506 rwlock_init(&newsk->sk_callback_lock);
443aef0e
PZ		 1507 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1508 af_callback_keys + newsk->sk_family,
1509 af_family_clock_key_strings[newsk->sk_family]);
87d11ceb
ACM		 1510
1511 newsk->sk_dst_cache = NULL;
1512 newsk->sk_wmem_queued = 0;
1513 newsk->sk_forward_alloc = 0;
1514 newsk->sk_send_head = NULL;
87d11ceb
ACM		 1515 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1516
1517 sock_reset_flag(newsk, SOCK_DONE);
1518 skb_queue_head_init(&newsk->sk_error_queue);
1519
0d7da9dd 1520 filter = rcu_dereference_protected(newsk->sk_filter, 1);
87d11ceb 1521 if (filter != NULL)
278571ba
AS		 1522 /* though it's an empty new sock, the charging may fail
1523 * if sysctl_optmem_max was changed between creation of
1524 * original socket and cloning
1525 */
1526 is_charged = sk_filter_charge(newsk, filter);
87d11ceb 1527
278571ba 1528 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk))) {
87d11ceb
ACM		 1529 /* It is still raw copy of parent, so invalidate
1530 * destructor and make plain sk_free() */
1531 newsk->sk_destruct = NULL;
b0691c8e 1532 bh_unlock_sock(newsk);
87d11ceb
ACM		 1533 sk_free(newsk);
1534 newsk = NULL;
1535 goto out;
1536 }
1537
1538 newsk->sk_err = 0;
1539 newsk->sk_priority = 0;
4dc6dc71
ED		 1540 /*
1541 * Before updating sk_refcnt, we must commit prior changes to memory
1542 * (Documentation/RCU/rculist_nulls.txt for details)
1543 */
1544 smp_wmb();
87d11ceb
ACM		 1545 atomic_set(&newsk->sk_refcnt, 2);
1546
1547 /*
1548 * Increment the counter in the same struct proto as the master
1549 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1550 * is the same as sk->sk_prot->socks, as this field was copied
1551 * with memcpy).
1552 *
1553 * This _changes_ the previous behaviour, where
1554 * tcp_create_openreq_child always was incrementing the
1555 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1556 * to be taken into account in all callers. -acme
1557 */
1558 sk_refcnt_debug_inc(newsk);
972692e0 1559 sk_set_socket(newsk, NULL);
43815482 1560 newsk->sk_wq = NULL;
87d11ceb 1561
f3f511e1
GC		 1562 sk_update_clone(sk, newsk);
1563
87d11ceb 1564 if (newsk->sk_prot->sockets_allocated)
180d8cd9 1565 sk_sockets_allocated_inc(newsk);
704da560 1566
08e29af3 1567 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
704da560 1568 net_enable_timestamp();
87d11ceb
ACM		 1569 }
1570out:
1571 return newsk;
1572}
e56c57d0 1573EXPORT_SYMBOL_GPL(sk_clone_lock);
87d11ceb 1574
9958089a
AK		 1575void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1576{
1577 __sk_dst_set(sk, dst);
1578 sk->sk_route_caps = dst->dev->features;
1579 if (sk->sk_route_caps & NETIF_F_GSO)
4fcd6b99 1580 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
a465419b 1581 sk->sk_route_caps &= ~sk->sk_route_nocaps;
9958089a 1582 if (sk_can_gso(sk)) {
82cc1a7a 1583 if (dst->header_len) {
9958089a 1584 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
82cc1a7a 1585 } else {
9958089a 1586 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
82cc1a7a 1587 sk->sk_gso_max_size = dst->dev->gso_max_size;
1485348d 1588 sk->sk_gso_max_segs = dst->dev->gso_max_segs;
82cc1a7a 1589 }
9958089a
AK		 1590 }
1591}
1592EXPORT_SYMBOL_GPL(sk_setup_caps);
1593
1da177e4
LT		 1594/*
1595 * Simple resource managers for sockets.
1596 */
1597
1598
4ec93edb
YH		 1599/*
1600 * Write buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1601 */
1602void sock_wfree(struct sk_buff *skb)
1603{
1604 struct sock *sk = skb->sk;
d99927f4 1605 unsigned int len = skb->truesize;
1da177e4 1606
d99927f4
ED		 1607 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1608 /*
1609 * Keep a reference on sk_wmem_alloc, this will be released
1610 * after sk_write_space() call
1611 */
1612 atomic_sub(len - 1, &sk->sk_wmem_alloc);
1da177e4 1613 sk->sk_write_space(sk);
d99927f4
ED		 1614 len = 1;
1615 }
2b85a34e 1616 /*
d99927f4
ED		 1617 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1618 * could not do because of in-flight packets
2b85a34e 1619 */
d99927f4 1620 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
2b85a34e 1621 __sk_free(sk);
1da177e4 1622}
2a91525c 1623EXPORT_SYMBOL(sock_wfree);
1da177e4 1624
f2f872f9
ED		 1625void skb_orphan_partial(struct sk_buff *skb)
1626{
1627 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1628 * so we do not completely orphan skb, but transfert all
1629 * accounted bytes but one, to avoid unexpected reorders.
1630 */
1631 if (skb->destructor == sock_wfree
1632#ifdef CONFIG_INET
1633 || skb->destructor == tcp_wfree
1634#endif
1635 ) {
1636 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1637 skb->truesize = 1;
1638 } else {
1639 skb_orphan(skb);
1640 }
1641}
1642EXPORT_SYMBOL(skb_orphan_partial);
1643
4ec93edb
YH		 1644/*
1645 * Read buffer destructor automatically called from kfree_skb.
1da177e4
LT		 1646 */
1647void sock_rfree(struct sk_buff *skb)
1648{
1649 struct sock *sk = skb->sk;
d361fd59 1650 unsigned int len = skb->truesize;
1da177e4 1651
d361fd59
ED		 1652 atomic_sub(len, &sk->sk_rmem_alloc);
1653 sk_mem_uncharge(sk, len);
1da177e4 1654}
2a91525c 1655EXPORT_SYMBOL(sock_rfree);
1da177e4 1656
41063e9d
DM		 1657void sock_edemux(struct sk_buff *skb)
1658{
e812347c
ED		 1659 struct sock *sk = skb->sk;
1660
1c463e57 1661#ifdef CONFIG_INET
e812347c
ED		 1662 if (sk->sk_state == TCP_TIME_WAIT)
1663 inet_twsk_put(inet_twsk(sk));
1664 else
1c463e57 1665#endif
e812347c 1666 sock_put(sk);
41063e9d
DM		 1667}
1668EXPORT_SYMBOL(sock_edemux);
1da177e4 1669
976d0201 1670kuid_t sock_i_uid(struct sock *sk)
1da177e4 1671{
976d0201 1672 kuid_t uid;
1da177e4 1673
f064af1e 1674 read_lock_bh(&sk->sk_callback_lock);
976d0201 1675 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
f064af1e 1676 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1677 return uid;
1678}
2a91525c 1679EXPORT_SYMBOL(sock_i_uid);
1da177e4
LT		 1680
1681unsigned long sock_i_ino(struct sock *sk)
1682{
1683 unsigned long ino;
1684
f064af1e 1685 read_lock_bh(&sk->sk_callback_lock);
1da177e4 1686 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
f064af1e 1687 read_unlock_bh(&sk->sk_callback_lock);
1da177e4
LT		 1688 return ino;
1689}
2a91525c 1690EXPORT_SYMBOL(sock_i_ino);
1da177e4
LT		 1691
1692/*
1693 * Allocate a skb from the socket's send buffer.
1694 */
86a76caf 1695struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
dd0fc66f 1696 gfp_t priority)
1da177e4
LT		 1697{
1698 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
2a91525c 1699 struct sk_buff *skb = alloc_skb(size, priority);
1da177e4
LT		 1700 if (skb) {
1701 skb_set_owner_w(skb, sk);
1702 return skb;
1703 }
1704 }
1705 return NULL;
1706}
2a91525c 1707EXPORT_SYMBOL(sock_wmalloc);
1da177e4 1708
4ec93edb 1709/*
1da177e4 1710 * Allocate a memory block from the socket's option memory buffer.
4ec93edb 1711 */
dd0fc66f 1712void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
1da177e4 1713{
95c96174 1714 if ((unsigned int)size <= sysctl_optmem_max &&
1da177e4
LT		 1715 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1716 void *mem;
1717 /* First do the add, to avoid the race if kmalloc
4ec93edb 1718 * might sleep.
1da177e4
LT		 1719 */
1720 atomic_add(size, &sk->sk_omem_alloc);
1721 mem = kmalloc(size, priority);
1722 if (mem)
1723 return mem;
1724 atomic_sub(size, &sk->sk_omem_alloc);
1725 }
1726 return NULL;
1727}
2a91525c 1728EXPORT_SYMBOL(sock_kmalloc);
1da177e4
LT		 1729
1730/*
1731 * Free an option memory block.
1732 */
1733void sock_kfree_s(struct sock *sk, void *mem, int size)
1734{
1735 kfree(mem);
1736 atomic_sub(size, &sk->sk_omem_alloc);
1737}
2a91525c 1738EXPORT_SYMBOL(sock_kfree_s);
1da177e4
LT		 1739
1740/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1741 I think, these locks should be removed for datagram sockets.
1742 */
2a91525c 1743static long sock_wait_for_wmem(struct sock *sk, long timeo)
1da177e4
LT		 1744{
1745 DEFINE_WAIT(wait);
1746
1747 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1748 for (;;) {
1749 if (!timeo)
1750 break;
1751 if (signal_pending(current))
1752 break;
1753 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
aa395145 1754 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1755 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1756 break;
1757 if (sk->sk_shutdown & SEND_SHUTDOWN)
1758 break;
1759 if (sk->sk_err)
1760 break;
1761 timeo = schedule_timeout(timeo);
1762 }
aa395145 1763 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 1764 return timeo;
1765}
1766
1767
1768/*
1769 * Generic send/receive buffer handlers
1770 */
1771
4cc7f68d
HX		 1772struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1773 unsigned long data_len, int noblock,
28d64271 1774 int *errcode, int max_page_order)
1da177e4 1775{
28d64271
ED		 1776 struct sk_buff *skb = NULL;
1777 unsigned long chunk;
7d877f3b 1778 gfp_t gfp_mask;
1da177e4
LT		 1779 long timeo;
1780 int err;
cc9b17ad 1781 int npages = (data_len + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
28d64271
ED		 1782 struct page *page;
1783 int i;
cc9b17ad
JW		 1784
1785 err = -EMSGSIZE;
1786 if (npages > MAX_SKB_FRAGS)
1787 goto failure;
1da177e4 1788
1da177e4 1789 timeo = sock_sndtimeo(sk, noblock);
28d64271 1790 while (!skb) {
1da177e4
LT		 1791 err = sock_error(sk);
1792 if (err != 0)
1793 goto failure;
1794
1795 err = -EPIPE;
1796 if (sk->sk_shutdown & SEND_SHUTDOWN)
1797 goto failure;
1798
28d64271
ED		 1799 if (atomic_read(&sk->sk_wmem_alloc) >= sk->sk_sndbuf) {
1800 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1801 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1802 err = -EAGAIN;
1803 if (!timeo)
1804 goto failure;
1805 if (signal_pending(current))
1806 goto interrupted;
1807 timeo = sock_wait_for_wmem(sk, timeo);
1808 continue;
1809 }
1da177e4 1810
28d64271
ED		 1811 err = -ENOBUFS;
1812 gfp_mask = sk->sk_allocation;
1813 if (gfp_mask & __GFP_WAIT)
1814 gfp_mask |= __GFP_REPEAT;
1815
1816 skb = alloc_skb(header_len, gfp_mask);
1817 if (!skb)
1da177e4 1818 goto failure;
28d64271
ED		 1819
1820 skb->truesize += data_len;
1821
1822 for (i = 0; npages > 0; i++) {
1823 int order = max_page_order;
1824
1825 while (order) {
1826 if (npages >= 1 << order) {
1827 page = alloc_pages(sk->sk_allocation |
ed98df33
ED		 1828 __GFP_COMP |
1829 __GFP_NOWARN |
1830 __GFP_NORETRY,
28d64271
ED		 1831 order);
1832 if (page)
1833 goto fill_page;
1834 }
1835 order--;
1836 }
1837 page = alloc_page(sk->sk_allocation);
1838 if (!page)
1839 goto failure;
1840fill_page:
1841 chunk = min_t(unsigned long, data_len,
1842 PAGE_SIZE << order);
1843 skb_fill_page_desc(skb, i, page, 0, chunk);
1844 data_len -= chunk;
1845 npages -= 1 << order;
1da177e4 1846 }
1da177e4
LT		 1847 }
1848
1849 skb_set_owner_w(skb, sk);
1850 return skb;
1851
1852interrupted:
1853 err = sock_intr_errno(timeo);
1854failure:
28d64271 1855 kfree_skb(skb);
1da177e4
LT		 1856 *errcode = err;
1857 return NULL;
1858}
4cc7f68d 1859EXPORT_SYMBOL(sock_alloc_send_pskb);
1da177e4 1860
4ec93edb 1861struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1da177e4
LT		 1862 int noblock, int *errcode)
1863{
28d64271 1864 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
1da177e4 1865}
2a91525c 1866EXPORT_SYMBOL(sock_alloc_send_skb);
1da177e4 1867
5640f768
ED		 1868/* On 32bit arches, an skb frag is limited to 2^15 */
1869#define SKB_FRAG_PAGE_ORDER get_order(32768)
1870
400dfd3a
ED		 1871/**
1872 * skb_page_frag_refill - check that a page_frag contains enough room
1873 * @sz: minimum size of the fragment we want to get
1874 * @pfrag: pointer to page_frag
1875 * @prio: priority for memory allocation
1876 *
1877 * Note: While this allocator tries to use high order pages, there is
1878 * no guarantee that allocations succeed. Therefore, @sz MUST be
1879 * less or equal than PAGE_SIZE.
1880 */
1881bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio)
5640f768
ED		 1882{
1883 int order;
1884
1885 if (pfrag->page) {
1886 if (atomic_read(&pfrag->page->_count) == 1) {
1887 pfrag->offset = 0;
1888 return true;
1889 }
400dfd3a 1890 if (pfrag->offset + sz <= pfrag->size)
5640f768
ED		 1891 return true;
1892 put_page(pfrag->page);
1893 }
1894
097b4f19 1895 order = SKB_FRAG_PAGE_ORDER;
5640f768 1896 do {
400dfd3a 1897 gfp_t gfp = prio;
5640f768
ED		 1898
1899 if (order)
ed98df33 1900 gfp |= __GFP_COMP | __GFP_NOWARN | __GFP_NORETRY;
5640f768
ED		 1901 pfrag->page = alloc_pages(gfp, order);
1902 if (likely(pfrag->page)) {
1903 pfrag->offset = 0;
1904 pfrag->size = PAGE_SIZE << order;
1905 return true;
1906 }
1907 } while (--order >= 0);
1908
400dfd3a
ED		 1909 return false;
1910}
1911EXPORT_SYMBOL(skb_page_frag_refill);
1912
1913bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1914{
1915 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1916 return true;
1917
5640f768
ED		 1918 sk_enter_memory_pressure(sk);
1919 sk_stream_moderate_sndbuf(sk);
1920 return false;
1921}
1922EXPORT_SYMBOL(sk_page_frag_refill);
1923
1da177e4 1924static void __lock_sock(struct sock *sk)
f39234d6
NK		 1925 __releases(&sk->sk_lock.slock)
1926 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1927{
1928 DEFINE_WAIT(wait);
1929
e71a4783 1930 for (;;) {
1da177e4
LT		 1931 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1932 TASK_UNINTERRUPTIBLE);
1933 spin_unlock_bh(&sk->sk_lock.slock);
1934 schedule();
1935 spin_lock_bh(&sk->sk_lock.slock);
e71a4783 1936 if (!sock_owned_by_user(sk))
1da177e4
LT		 1937 break;
1938 }
1939 finish_wait(&sk->sk_lock.wq, &wait);
1940}
1941
1942static void __release_sock(struct sock *sk)
f39234d6
NK		 1943 __releases(&sk->sk_lock.slock)
1944 __acquires(&sk->sk_lock.slock)
1da177e4
LT		 1945{
1946 struct sk_buff *skb = sk->sk_backlog.head;
1947
1948 do {
1949 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1950 bh_unlock_sock(sk);
1951
1952 do {
1953 struct sk_buff *next = skb->next;
1954
e4cbb02a 1955 prefetch(next);
7fee226a 1956 WARN_ON_ONCE(skb_dst_is_noref(skb));
1da177e4 1957 skb->next = NULL;
c57943a1 1958 sk_backlog_rcv(sk, skb);
1da177e4
LT		 1959
1960 /*
1961 * We are in process context here with softirqs
1962 * disabled, use cond_resched_softirq() to preempt.
1963 * This is safe to do because we've taken the backlog
1964 * queue private:
1965 */
1966 cond_resched_softirq();
1967
1968 skb = next;
1969 } while (skb != NULL);
1970
1971 bh_lock_sock(sk);
e71a4783 1972 } while ((skb = sk->sk_backlog.head) != NULL);
8eae939f
ZY		 1973
1974 /*
1975 * Doing the zeroing here guarantee we can not loop forever
1976 * while a wild producer attempts to flood us.
1977 */
1978 sk->sk_backlog.len = 0;
1da177e4
LT		 1979}
1980
1981/**
1982 * sk_wait_data - wait for data to arrive at sk_receive_queue
4dc3b16b
PP		 1983 * @sk: sock to wait on
1984 * @timeo: for how long
1da177e4
LT		 1985 *
1986 * Now socket state including sk->sk_err is changed only under lock,
1987 * hence we may omit checks after joining wait queue.
1988 * We check receive queue before schedule() only as optimization;
1989 * it is very likely that release_sock() added new data.
1990 */
1991int sk_wait_data(struct sock *sk, long *timeo)
1992{
1993 int rc;
1994 DEFINE_WAIT(wait);
1995
aa395145 1996 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
1da177e4
LT		 1997 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
1998 rc = sk_wait_event(sk, timeo, !skb_queue_empty(&sk->sk_receive_queue));
1999 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
aa395145 2000 finish_wait(sk_sleep(sk), &wait);
1da177e4
LT		 2001 return rc;
2002}
1da177e4
LT		 2003EXPORT_SYMBOL(sk_wait_data);
2004
3ab224be
HA		 2005/**
2006 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
2007 * @sk: socket
2008 * @size: memory size to allocate
2009 * @kind: allocation type
2010 *
2011 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
2012 * rmem allocation. This function assumes that protocols which have
2013 * memory_pressure use sk_wmem_queued as write buffer accounting.
2014 */
2015int __sk_mem_schedule(struct sock *sk, int size, int kind)
2016{
2017 struct proto *prot = sk->sk_prot;
2018 int amt = sk_mem_pages(size);
8d987e5c 2019 long allocated;
e1aab161 2020 int parent_status = UNDER_LIMIT;
3ab224be
HA		 2021
2022 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
180d8cd9 2023
e1aab161 2024 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
3ab224be
HA		 2025
2026 /* Under limit. */
e1aab161
GC		 2027 if (parent_status == UNDER_LIMIT &&
2028 allocated <= sk_prot_mem_limits(sk, 0)) {
180d8cd9 2029 sk_leave_memory_pressure(sk);
3ab224be
HA		 2030 return 1;
2031 }
2032
e1aab161
GC		 2033 /* Under pressure. (we or our parents) */
2034 if ((parent_status > SOFT_LIMIT) ||
2035 allocated > sk_prot_mem_limits(sk, 1))
180d8cd9 2036 sk_enter_memory_pressure(sk);
3ab224be 2037
e1aab161
GC		 2038 /* Over hard limit (we or our parents) */
2039 if ((parent_status == OVER_LIMIT) ||
2040 (allocated > sk_prot_mem_limits(sk, 2)))
3ab224be
HA		 2041 goto suppress_allocation;
2042
2043 /* guarantee minimum buffer size under pressure */
2044 if (kind == SK_MEM_RECV) {
2045 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2046 return 1;
180d8cd9 2047
3ab224be
HA		 2048 } else { /* SK_MEM_SEND */
2049 if (sk->sk_type == SOCK_STREAM) {
2050 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2051 return 1;
2052 } else if (atomic_read(&sk->sk_wmem_alloc) <
2053 prot->sysctl_wmem[0])
2054 return 1;
2055 }
2056
180d8cd9 2057 if (sk_has_memory_pressure(sk)) {
1748376b
ED		 2058 int alloc;
2059
180d8cd9 2060 if (!sk_under_memory_pressure(sk))
1748376b 2061 return 1;
180d8cd9
GC		 2062 alloc = sk_sockets_allocated_read_positive(sk);
2063 if (sk_prot_mem_limits(sk, 2) > alloc *
3ab224be
HA		 2064 sk_mem_pages(sk->sk_wmem_queued +
2065 atomic_read(&sk->sk_rmem_alloc) +
2066 sk->sk_forward_alloc))
2067 return 1;
2068 }
2069
2070suppress_allocation:
2071
2072 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2073 sk_stream_moderate_sndbuf(sk);
2074
2075 /* Fail only if socket is _under_ its sndbuf.
2076 * In this case we cannot block, so that we have to fail.
2077 */
2078 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2079 return 1;
2080 }
2081
3847ce32
SM		 2082 trace_sock_exceed_buf_limit(sk, prot, allocated);
2083
3ab224be
HA		 2084 /* Alas. Undo changes. */
2085 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
180d8cd9 2086
0e90b31f 2087 sk_memory_allocated_sub(sk, amt);
180d8cd9 2088
3ab224be
HA		 2089 return 0;
2090}
3ab224be
HA		 2091EXPORT_SYMBOL(__sk_mem_schedule);
2092
2093/**
2094 * __sk_reclaim - reclaim memory_allocated
2095 * @sk: socket
2096 */
2097void __sk_mem_reclaim(struct sock *sk)
2098{
180d8cd9 2099 sk_memory_allocated_sub(sk,
0e90b31f 2100 sk->sk_forward_alloc >> SK_MEM_QUANTUM_SHIFT);
3ab224be
HA		 2101 sk->sk_forward_alloc &= SK_MEM_QUANTUM - 1;
2102
180d8cd9
GC		 2103 if (sk_under_memory_pressure(sk) &&
2104 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2105 sk_leave_memory_pressure(sk);
3ab224be 2106}
3ab224be
HA		 2107EXPORT_SYMBOL(__sk_mem_reclaim);
2108
2109
1da177e4
LT		 2110/*
2111 * Set of default routines for initialising struct proto_ops when
2112 * the protocol does not support a particular function. In certain
2113 * cases where it makes no sense for a protocol to have a "do nothing"
2114 * function, some default processing is provided.
2115 */
2116
2117int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2118{
2119 return -EOPNOTSUPP;
2120}
2a91525c 2121EXPORT_SYMBOL(sock_no_bind);
1da177e4 2122
4ec93edb 2123int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2124 int len, int flags)
2125{
2126 return -EOPNOTSUPP;
2127}
2a91525c 2128EXPORT_SYMBOL(sock_no_connect);
1da177e4
LT		 2129
2130int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2131{
2132 return -EOPNOTSUPP;
2133}
2a91525c 2134EXPORT_SYMBOL(sock_no_socketpair);
1da177e4
LT		 2135
2136int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2137{
2138 return -EOPNOTSUPP;
2139}
2a91525c 2140EXPORT_SYMBOL(sock_no_accept);
1da177e4 2141
4ec93edb 2142int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
1da177e4
LT		 2143 int *len, int peer)
2144{
2145 return -EOPNOTSUPP;
2146}
2a91525c 2147EXPORT_SYMBOL(sock_no_getname);
1da177e4 2148
2a91525c 2149unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
1da177e4
LT		 2150{
2151 return 0;
2152}
2a91525c 2153EXPORT_SYMBOL(sock_no_poll);
1da177e4
LT		 2154
2155int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2156{
2157 return -EOPNOTSUPP;
2158}
2a91525c 2159EXPORT_SYMBOL(sock_no_ioctl);
1da177e4
LT		 2160
2161int sock_no_listen(struct socket *sock, int backlog)
2162{
2163 return -EOPNOTSUPP;
2164}
2a91525c 2165EXPORT_SYMBOL(sock_no_listen);
1da177e4
LT		 2166
2167int sock_no_shutdown(struct socket *sock, int how)
2168{
2169 return -EOPNOTSUPP;
2170}
2a91525c 2171EXPORT_SYMBOL(sock_no_shutdown);
1da177e4
LT		 2172
2173int sock_no_setsockopt(struct socket *sock, int level, int optname,
b7058842 2174 char __user *optval, unsigned int optlen)
1da177e4
LT		 2175{
2176 return -EOPNOTSUPP;
2177}
2a91525c 2178EXPORT_SYMBOL(sock_no_setsockopt);
1da177e4
LT		 2179
2180int sock_no_getsockopt(struct socket *sock, int level, int optname,
2181 char __user *optval, int __user *optlen)
2182{
2183 return -EOPNOTSUPP;
2184}
2a91525c 2185EXPORT_SYMBOL(sock_no_getsockopt);
1da177e4
LT		 2186
2187int sock_no_sendmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2188 size_t len)
2189{
2190 return -EOPNOTSUPP;
2191}
2a91525c 2192EXPORT_SYMBOL(sock_no_sendmsg);
1da177e4
LT		 2193
2194int sock_no_recvmsg(struct kiocb *iocb, struct socket *sock, struct msghdr *m,
2195 size_t len, int flags)
2196{
2197 return -EOPNOTSUPP;
2198}
2a91525c 2199EXPORT_SYMBOL(sock_no_recvmsg);
1da177e4
LT		 2200
2201int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2202{
2203 /* Mirror missing mmap method error code */
2204 return -ENODEV;
2205}
2a91525c 2206EXPORT_SYMBOL(sock_no_mmap);
1da177e4
LT		 2207
2208ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2209{
2210 ssize_t res;
2211 struct msghdr msg = {.msg_flags = flags};
2212 struct kvec iov;
2213 char *kaddr = kmap(page);
2214 iov.iov_base = kaddr + offset;
2215 iov.iov_len = size;
2216 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2217 kunmap(page);
2218 return res;
2219}
2a91525c 2220EXPORT_SYMBOL(sock_no_sendpage);
1da177e4
LT		 2221
2222/*
2223 * Default Socket Callbacks
2224 */
2225
2226static void sock_def_wakeup(struct sock *sk)
2227{
43815482
ED		 2228 struct socket_wq *wq;
2229
2230 rcu_read_lock();
2231 wq = rcu_dereference(sk->sk_wq);
2232 if (wq_has_sleeper(wq))
2233 wake_up_interruptible_all(&wq->wait);
2234 rcu_read_unlock();
1da177e4
LT		 2235}
2236
2237static void sock_def_error_report(struct sock *sk)
2238{
43815482
ED		 2239 struct socket_wq *wq;
2240
2241 rcu_read_lock();
2242 wq = rcu_dereference(sk->sk_wq);
2243 if (wq_has_sleeper(wq))
2244 wake_up_interruptible_poll(&wq->wait, POLLERR);
8d8ad9d7 2245 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
43815482 2246 rcu_read_unlock();
1da177e4
LT		 2247}
2248
676d2369 2249static void sock_def_readable(struct sock *sk)
1da177e4 2250{
43815482
ED		 2251 struct socket_wq *wq;
2252
2253 rcu_read_lock();
2254 wq = rcu_dereference(sk->sk_wq);
2255 if (wq_has_sleeper(wq))
2c6607c6 2256 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
37e5540b 2257 POLLRDNORM | POLLRDBAND);
8d8ad9d7 2258 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
43815482 2259 rcu_read_unlock();
1da177e4
LT		 2260}
2261
2262static void sock_def_write_space(struct sock *sk)
2263{
43815482
ED		 2264 struct socket_wq *wq;
2265
2266 rcu_read_lock();
1da177e4
LT		 2267
2268 /* Do not wake up a writer until he can make "significant"
2269 * progress. --DaveM
2270 */
e71a4783 2271 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
43815482
ED		 2272 wq = rcu_dereference(sk->sk_wq);
2273 if (wq_has_sleeper(wq))
2274 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
37e5540b 2275 POLLWRNORM | POLLWRBAND);
1da177e4
LT		 2276
2277 /* Should agree with poll, otherwise some programs break */
2278 if (sock_writeable(sk))
8d8ad9d7 2279 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
1da177e4
LT		 2280 }
2281
43815482 2282 rcu_read_unlock();
1da177e4
LT		 2283}
2284
2285static void sock_def_destruct(struct sock *sk)
2286{
a51482bd 2287 kfree(sk->sk_protinfo);
1da177e4
LT		 2288}
2289
2290void sk_send_sigurg(struct sock *sk)
2291{
2292 if (sk->sk_socket && sk->sk_socket->file)
2293 if (send_sigurg(&sk->sk_socket->file->f_owner))
8d8ad9d7 2294 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
1da177e4 2295}
2a91525c 2296EXPORT_SYMBOL(sk_send_sigurg);
1da177e4
LT		 2297
2298void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2299 unsigned long expires)
2300{
2301 if (!mod_timer(timer, expires))
2302 sock_hold(sk);
2303}
1da177e4
LT		 2304EXPORT_SYMBOL(sk_reset_timer);
2305
2306void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2307{
25cc4ae9 2308 if (del_timer(timer))
1da177e4
LT		 2309 __sock_put(sk);
2310}
1da177e4
LT		 2311EXPORT_SYMBOL(sk_stop_timer);
2312
2313void sock_init_data(struct socket *sock, struct sock *sk)
2314{
2315 skb_queue_head_init(&sk->sk_receive_queue);
2316 skb_queue_head_init(&sk->sk_write_queue);
2317 skb_queue_head_init(&sk->sk_error_queue);
97fc2f08
CL		 2318#ifdef CONFIG_NET_DMA
2319 skb_queue_head_init(&sk->sk_async_wait_queue);
2320#endif
1da177e4
LT		 2321
2322 sk->sk_send_head = NULL;
2323
2324 init_timer(&sk->sk_timer);
4ec93edb 2325
1da177e4
LT		 2326 sk->sk_allocation = GFP_KERNEL;
2327 sk->sk_rcvbuf = sysctl_rmem_default;
2328 sk->sk_sndbuf = sysctl_wmem_default;
2329 sk->sk_state = TCP_CLOSE;
972692e0 2330 sk_set_socket(sk, sock);
1da177e4
LT		 2331
2332 sock_set_flag(sk, SOCK_ZAPPED);
2333
e71a4783 2334 if (sock) {
1da177e4 2335 sk->sk_type = sock->type;
43815482 2336 sk->sk_wq = sock->wq;
1da177e4
LT		 2337 sock->sk = sk;
2338 } else
43815482 2339 sk->sk_wq = NULL;
1da177e4 2340
b6c6712a 2341 spin_lock_init(&sk->sk_dst_lock);
1da177e4 2342 rwlock_init(&sk->sk_callback_lock);
443aef0e
PZ		 2343 lockdep_set_class_and_name(&sk->sk_callback_lock,
2344 af_callback_keys + sk->sk_family,
2345 af_family_clock_key_strings[sk->sk_family]);
1da177e4
LT		 2346
2347 sk->sk_state_change = sock_def_wakeup;
2348 sk->sk_data_ready = sock_def_readable;
2349 sk->sk_write_space = sock_def_write_space;
2350 sk->sk_error_report = sock_def_error_report;
2351 sk->sk_destruct = sock_def_destruct;
2352
5640f768
ED		 2353 sk->sk_frag.page = NULL;
2354 sk->sk_frag.offset = 0;
ef64a54f 2355 sk->sk_peek_off = -1;
1da177e4 2356
109f6e39
EB		 2357 sk->sk_peer_pid = NULL;
2358 sk->sk_peer_cred = NULL;
1da177e4
LT		 2359 sk->sk_write_pending = 0;
2360 sk->sk_rcvlowat = 1;
2361 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2362 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2363
f37f0afb 2364 sk->sk_stamp = ktime_set(-1L, 0);
1da177e4 2365
e0d1095a 2366#ifdef CONFIG_NET_RX_BUSY_POLL
06021292 2367 sk->sk_napi_id = 0;
64b0dc51 2368 sk->sk_ll_usec = sysctl_net_busy_read;
06021292
ET		 2369#endif
2370
62748f32 2371 sk->sk_max_pacing_rate = ~0U;
7eec4174 2372 sk->sk_pacing_rate = ~0U;
4dc6dc71
ED		 2373 /*
2374 * Before updating sk_refcnt, we must commit prior changes to memory
2375 * (Documentation/RCU/rculist_nulls.txt for details)
2376 */
2377 smp_wmb();
1da177e4 2378 atomic_set(&sk->sk_refcnt, 1);
33c732c3 2379 atomic_set(&sk->sk_drops, 0);
1da177e4 2380}
2a91525c 2381EXPORT_SYMBOL(sock_init_data);
1da177e4 2382
b5606c2d 2383void lock_sock_nested(struct sock *sk, int subclass)
1da177e4
LT		 2384{
2385 might_sleep();
a5b5bb9a 2386 spin_lock_bh(&sk->sk_lock.slock);
d2e9117c 2387 if (sk->sk_lock.owned)
1da177e4 2388 __lock_sock(sk);
d2e9117c 2389 sk->sk_lock.owned = 1;
a5b5bb9a
IM		 2390 spin_unlock(&sk->sk_lock.slock);
2391 /*
2392 * The sk_lock has mutex_lock() semantics here:
2393 */
fcc70d5f 2394 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
a5b5bb9a 2395 local_bh_enable();
1da177e4 2396}
fcc70d5f 2397EXPORT_SYMBOL(lock_sock_nested);
1da177e4 2398
b5606c2d 2399void release_sock(struct sock *sk)
1da177e4 2400{
a5b5bb9a
IM		 2401 /*
2402 * The sk_lock has mutex_unlock() semantics:
2403 */
2404 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2405
2406 spin_lock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2407 if (sk->sk_backlog.tail)
2408 __release_sock(sk);
46d3ceab 2409
c3f9b018
ED		 2410 /* Warning : release_cb() might need to release sk ownership,
2411 * ie call sock_release_ownership(sk) before us.
2412 */
46d3ceab
ED		 2413 if (sk->sk_prot->release_cb)
2414 sk->sk_prot->release_cb(sk);
2415
c3f9b018 2416 sock_release_ownership(sk);
a5b5bb9a
IM		 2417 if (waitqueue_active(&sk->sk_lock.wq))
2418 wake_up(&sk->sk_lock.wq);
2419 spin_unlock_bh(&sk->sk_lock.slock);
1da177e4
LT		 2420}
2421EXPORT_SYMBOL(release_sock);
2422
8a74ad60
ED		 2423/**
2424 * lock_sock_fast - fast version of lock_sock
2425 * @sk: socket
2426 *
2427 * This version should be used for very small section, where process wont block
2428 * return false if fast path is taken
2429 * sk_lock.slock locked, owned = 0, BH disabled
2430 * return true if slow path is taken
2431 * sk_lock.slock unlocked, owned = 1, BH enabled
2432 */
2433bool lock_sock_fast(struct sock *sk)
2434{
2435 might_sleep();
2436 spin_lock_bh(&sk->sk_lock.slock);
2437
2438 if (!sk->sk_lock.owned)
2439 /*
2440 * Note : We must disable BH
2441 */
2442 return false;
2443
2444 __lock_sock(sk);
2445 sk->sk_lock.owned = 1;
2446 spin_unlock(&sk->sk_lock.slock);
2447 /*
2448 * The sk_lock has mutex_lock() semantics here:
2449 */
2450 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2451 local_bh_enable();
2452 return true;
2453}
2454EXPORT_SYMBOL(lock_sock_fast);
2455
1da177e4 2456int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
4ec93edb 2457{
b7aa0bf7 2458 struct timeval tv;
1da177e4 2459 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2460 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
b7aa0bf7
ED		 2461 tv = ktime_to_timeval(sk->sk_stamp);
2462 if (tv.tv_sec == -1)
1da177e4 2463 return -ENOENT;
b7aa0bf7
ED		 2464 if (tv.tv_sec == 0) {
2465 sk->sk_stamp = ktime_get_real();
2466 tv = ktime_to_timeval(sk->sk_stamp);
2467 }
2468 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
4ec93edb 2469}
1da177e4
LT		 2470EXPORT_SYMBOL(sock_get_timestamp);
2471
ae40eb1e
ED		 2472int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2473{
2474 struct timespec ts;
2475 if (!sock_flag(sk, SOCK_TIMESTAMP))
20d49473 2476 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
ae40eb1e
ED		 2477 ts = ktime_to_timespec(sk->sk_stamp);
2478 if (ts.tv_sec == -1)
2479 return -ENOENT;
2480 if (ts.tv_sec == 0) {
2481 sk->sk_stamp = ktime_get_real();
2482 ts = ktime_to_timespec(sk->sk_stamp);
2483 }
2484 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2485}
2486EXPORT_SYMBOL(sock_get_timestampns);
2487
20d49473 2488void sock_enable_timestamp(struct sock *sk, int flag)
4ec93edb 2489{
20d49473 2490 if (!sock_flag(sk, flag)) {
08e29af3
ED		 2491 unsigned long previous_flags = sk->sk_flags;
2492
20d49473
PO		 2493 sock_set_flag(sk, flag);
2494 /*
2495 * we just set one of the two flags which require net
2496 * time stamping, but time stamping might have been on
2497 * already because of the other one
2498 */
08e29af3 2499 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
20d49473 2500 net_enable_timestamp();
1da177e4
LT		 2501 }
2502}
1da177e4 2503
cb820f8e
RC		 2504int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2505 int level, int type)
2506{
2507 struct sock_exterr_skb *serr;
2508 struct sk_buff *skb, *skb2;
2509 int copied, err;
2510
2511 err = -EAGAIN;
2512 skb = skb_dequeue(&sk->sk_error_queue);
2513 if (skb == NULL)
2514 goto out;
2515
2516 copied = skb->len;
2517 if (copied > len) {
2518 msg->msg_flags |= MSG_TRUNC;
2519 copied = len;
2520 }
2521 err = skb_copy_datagram_iovec(skb, 0, msg->msg_iov, copied);
2522 if (err)
2523 goto out_free_skb;
2524
2525 sock_recv_timestamp(msg, sk, skb);
2526
2527 serr = SKB_EXT_ERR(skb);
2528 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2529
2530 msg->msg_flags |= MSG_ERRQUEUE;
2531 err = copied;
2532
2533 /* Reset and regenerate socket error */
2534 spin_lock_bh(&sk->sk_error_queue.lock);
2535 sk->sk_err = 0;
2536 if ((skb2 = skb_peek(&sk->sk_error_queue)) != NULL) {
2537 sk->sk_err = SKB_EXT_ERR(skb2)->ee.ee_errno;
2538 spin_unlock_bh(&sk->sk_error_queue.lock);
2539 sk->sk_error_report(sk);
2540 } else
2541 spin_unlock_bh(&sk->sk_error_queue.lock);
2542
2543out_free_skb:
2544 kfree_skb(skb);
2545out:
2546 return err;
2547}
2548EXPORT_SYMBOL(sock_recv_errqueue);
2549
1da177e4
LT		 2550/*
2551 * Get a socket option on an socket.
2552 *
2553 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2554 * asynchronous errors should be reported by getsockopt. We assume
2555 * this means if you specify SO_ERROR (otherwise whats the point of it).
2556 */
2557int sock_common_getsockopt(struct socket *sock, int level, int optname,
2558 char __user *optval, int __user *optlen)
2559{
2560 struct sock *sk = sock->sk;
2561
2562 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2563}
1da177e4
LT		 2564EXPORT_SYMBOL(sock_common_getsockopt);
2565
3fdadf7d 2566#ifdef CONFIG_COMPAT
543d9cfe
ACM		 2567int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2568 char __user *optval, int __user *optlen)
3fdadf7d
DM		 2569{
2570 struct sock *sk = sock->sk;
2571
1e51f951 2572 if (sk->sk_prot->compat_getsockopt != NULL)
543d9cfe
ACM		 2573 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2574 optval, optlen);
3fdadf7d
DM		 2575 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2576}
2577EXPORT_SYMBOL(compat_sock_common_getsockopt);
2578#endif
2579
1da177e4
LT		 2580int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
2581 struct msghdr *msg, size_t size, int flags)
2582{
2583 struct sock *sk = sock->sk;
2584 int addr_len = 0;
2585 int err;
2586
2587 err = sk->sk_prot->recvmsg(iocb, sk, msg, size, flags & MSG_DONTWAIT,
2588 flags & ~MSG_DONTWAIT, &addr_len);
2589 if (err >= 0)
2590 msg->msg_namelen = addr_len;
2591 return err;
2592}
1da177e4
LT		 2593EXPORT_SYMBOL(sock_common_recvmsg);
2594
2595/*
2596 * Set socket options on an inet socket.
2597 */
2598int sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2599 char __user *optval, unsigned int optlen)
1da177e4
LT		 2600{
2601 struct sock *sk = sock->sk;
2602
2603 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2604}
1da177e4
LT		 2605EXPORT_SYMBOL(sock_common_setsockopt);
2606
3fdadf7d 2607#ifdef CONFIG_COMPAT
543d9cfe 2608int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
b7058842 2609 char __user *optval, unsigned int optlen)
3fdadf7d
DM		 2610{
2611 struct sock *sk = sock->sk;
2612
543d9cfe
ACM		 2613 if (sk->sk_prot->compat_setsockopt != NULL)
2614 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2615 optval, optlen);
3fdadf7d
DM		 2616 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2617}
2618EXPORT_SYMBOL(compat_sock_common_setsockopt);
2619#endif
2620
1da177e4
LT		 2621void sk_common_release(struct sock *sk)
2622{
2623 if (sk->sk_prot->destroy)
2624 sk->sk_prot->destroy(sk);
2625
2626 /*
2627 * Observation: when sock_common_release is called, processes have
2628 * no access to socket. But net still has.
2629 * Step one, detach it from networking:
2630 *
2631 * A. Remove from hash tables.
2632 */
2633
2634 sk->sk_prot->unhash(sk);
2635
2636 /*
2637 * In this point socket cannot receive new packets, but it is possible
2638 * that some packets are in flight because some CPU runs receiver and
2639 * did hash table lookup before we unhashed socket. They will achieve
2640 * receive queue and will be purged by socket destructor.
2641 *
2642 * Also we still have packets pending on receive queue and probably,
2643 * our own packets waiting in device queues. sock_destroy will drain
2644 * receive queue, but transmitted packets will delay socket destruction
2645 * until the last reference will be released.
2646 */
2647
2648 sock_orphan(sk);
2649
2650 xfrm_sk_free_policy(sk);
2651
e6848976 2652 sk_refcnt_debug_release(sk);
5640f768
ED		 2653
2654 if (sk->sk_frag.page) {
2655 put_page(sk->sk_frag.page);
2656 sk->sk_frag.page = NULL;
2657 }
2658
1da177e4
LT		 2659 sock_put(sk);
2660}
1da177e4
LT		 2661EXPORT_SYMBOL(sk_common_release);
2662
13ff3d6f
PE		 2663#ifdef CONFIG_PROC_FS
2664#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1338d466
PE		 2665struct prot_inuse {
2666 int val[PROTO_INUSE_NR];
2667};
13ff3d6f
PE		 2668
2669static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
70ee1159
PE		 2670
2671#ifdef CONFIG_NET_NS
2672void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2673{
d6d9ca0f 2674 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
70ee1159
PE		 2675}
2676EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2677
2678int sock_prot_inuse_get(struct net *net, struct proto *prot)
2679{
2680 int cpu, idx = prot->inuse_idx;
2681 int res = 0;
2682
2683 for_each_possible_cpu(cpu)
2684 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2685
2686 return res >= 0 ? res : 0;
2687}
2688EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2689
2c8c1e72 2690static int __net_init sock_inuse_init_net(struct net *net)
70ee1159
PE		 2691{
2692 net->core.inuse = alloc_percpu(struct prot_inuse);
2693 return net->core.inuse ? 0 : -ENOMEM;
2694}
2695
2c8c1e72 2696static void __net_exit sock_inuse_exit_net(struct net *net)
70ee1159
PE		 2697{
2698 free_percpu(net->core.inuse);
2699}
2700
2701static struct pernet_operations net_inuse_ops = {
2702 .init = sock_inuse_init_net,
2703 .exit = sock_inuse_exit_net,
2704};
2705
2706static __init int net_inuse_init(void)
2707{
2708 if (register_pernet_subsys(&net_inuse_ops))
2709 panic("Cannot initialize net inuse counters");
2710
2711 return 0;
2712}
2713
2714core_initcall(net_inuse_init);
2715#else
1338d466
PE		 2716static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2717
c29a0bc4 2718void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
1338d466 2719{
d6d9ca0f 2720 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
1338d466
PE		 2721}
2722EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2723
c29a0bc4 2724int sock_prot_inuse_get(struct net *net, struct proto *prot)
1338d466
PE		 2725{
2726 int cpu, idx = prot->inuse_idx;
2727 int res = 0;
2728
2729 for_each_possible_cpu(cpu)
2730 res += per_cpu(prot_inuse, cpu).val[idx];
2731
2732 return res >= 0 ? res : 0;
2733}
2734EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
70ee1159 2735#endif
13ff3d6f
PE		 2736
2737static void assign_proto_idx(struct proto *prot)
2738{
2739 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2740
2741 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
e005d193 2742 pr_err("PROTO_INUSE_NR exhausted\n");
13ff3d6f
PE		 2743 return;
2744 }
2745
2746 set_bit(prot->inuse_idx, proto_inuse_idx);
2747}
2748
2749static void release_proto_idx(struct proto *prot)
2750{
2751 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2752 clear_bit(prot->inuse_idx, proto_inuse_idx);
2753}
2754#else
2755static inline void assign_proto_idx(struct proto *prot)
2756{
2757}
2758
2759static inline void release_proto_idx(struct proto *prot)
2760{
2761}
2762#endif
2763
b733c007
PE		 2764int proto_register(struct proto *prot, int alloc_slab)
2765{
1da177e4
LT		 2766 if (alloc_slab) {
2767 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
271b72c7
ED		 2768 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2769 NULL);
1da177e4
LT		 2770
2771 if (prot->slab == NULL) {
e005d193
JP		 2772 pr_crit("%s: Can't create sock SLAB cache!\n",
2773 prot->name);
60e7663d 2774 goto out;
1da177e4 2775 }
2e6599cb
ACM		 2776
2777 if (prot->rsk_prot != NULL) {
faf23422 2778 prot->rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s", prot->name);
7e56b5d6 2779 if (prot->rsk_prot->slab_name == NULL)
2e6599cb
ACM		 2780 goto out_free_sock_slab;
2781
7e56b5d6 2782 prot->rsk_prot->slab = kmem_cache_create(prot->rsk_prot->slab_name,
2e6599cb 2783 prot->rsk_prot->obj_size, 0,
20c2df83 2784 SLAB_HWCACHE_ALIGN, NULL);
2e6599cb
ACM		 2785
2786 if (prot->rsk_prot->slab == NULL) {
e005d193
JP		 2787 pr_crit("%s: Can't create request sock SLAB cache!\n",
2788 prot->name);
2e6599cb
ACM		 2789 goto out_free_request_sock_slab_name;
2790 }
2791 }
8feaf0c0 2792
6d6ee43e 2793 if (prot->twsk_prot != NULL) {
faf23422 2794 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
8feaf0c0 2795
7e56b5d6 2796 if (prot->twsk_prot->twsk_slab_name == NULL)
8feaf0c0
ACM		 2797 goto out_free_request_sock_slab;
2798
6d6ee43e 2799 prot->twsk_prot->twsk_slab =
7e56b5d6 2800 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
6d6ee43e 2801 prot->twsk_prot->twsk_obj_size,
3ab5aee7
ED		 2802 0,
2803 SLAB_HWCACHE_ALIGN |
2804 prot->slab_flags,
20c2df83 2805 NULL);
6d6ee43e 2806 if (prot->twsk_prot->twsk_slab == NULL)
8feaf0c0
ACM		 2807 goto out_free_timewait_sock_slab_name;
2808 }
1da177e4
LT		 2809 }
2810
36b77a52 2811 mutex_lock(&proto_list_mutex);
1da177e4 2812 list_add(&prot->node, &proto_list);
13ff3d6f 2813 assign_proto_idx(prot);
36b77a52 2814 mutex_unlock(&proto_list_mutex);
b733c007
PE		 2815 return 0;
2816
8feaf0c0 2817out_free_timewait_sock_slab_name:
7e56b5d6 2818 kfree(prot->twsk_prot->twsk_slab_name);
8feaf0c0
ACM		 2819out_free_request_sock_slab:
2820 if (prot->rsk_prot && prot->rsk_prot->slab) {
2821 kmem_cache_destroy(prot->rsk_prot->slab);
2822 prot->rsk_prot->slab = NULL;
2823 }
2e6599cb 2824out_free_request_sock_slab_name:
72150e9b
DC		 2825 if (prot->rsk_prot)
2826 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2827out_free_sock_slab:
2828 kmem_cache_destroy(prot->slab);
2829 prot->slab = NULL;
b733c007
PE		 2830out:
2831 return -ENOBUFS;
1da177e4 2832}
1da177e4
LT		 2833EXPORT_SYMBOL(proto_register);
2834
2835void proto_unregister(struct proto *prot)
2836{
36b77a52 2837 mutex_lock(&proto_list_mutex);
13ff3d6f 2838 release_proto_idx(prot);
0a3f4358 2839 list_del(&prot->node);
36b77a52 2840 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2841
2842 if (prot->slab != NULL) {
2843 kmem_cache_destroy(prot->slab);
2844 prot->slab = NULL;
2845 }
2846
2e6599cb 2847 if (prot->rsk_prot != NULL && prot->rsk_prot->slab != NULL) {
2e6599cb 2848 kmem_cache_destroy(prot->rsk_prot->slab);
7e56b5d6 2849 kfree(prot->rsk_prot->slab_name);
2e6599cb
ACM		 2850 prot->rsk_prot->slab = NULL;
2851 }
2852
6d6ee43e 2853 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
6d6ee43e 2854 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
7e56b5d6 2855 kfree(prot->twsk_prot->twsk_slab_name);
6d6ee43e 2856 prot->twsk_prot->twsk_slab = NULL;
8feaf0c0 2857 }
1da177e4 2858}
1da177e4
LT		 2859EXPORT_SYMBOL(proto_unregister);
2860
2861#ifdef CONFIG_PROC_FS
1da177e4 2862static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
36b77a52 2863 __acquires(proto_list_mutex)
1da177e4 2864{
36b77a52 2865 mutex_lock(&proto_list_mutex);
60f0438a 2866 return seq_list_start_head(&proto_list, *pos);
1da177e4
LT		 2867}
2868
2869static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2870{
60f0438a 2871 return seq_list_next(v, &proto_list, pos);
1da177e4
LT		 2872}
2873
2874static void proto_seq_stop(struct seq_file *seq, void *v)
36b77a52 2875 __releases(proto_list_mutex)
1da177e4 2876{
36b77a52 2877 mutex_unlock(&proto_list_mutex);
1da177e4
LT		 2878}
2879
2880static char proto_method_implemented(const void *method)
2881{
2882 return method == NULL ? 'n' : 'y';
2883}
180d8cd9
GC		 2884static long sock_prot_memory_allocated(struct proto *proto)
2885{
cb75a36c 2886 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
180d8cd9
GC		 2887}
2888
2889static char *sock_prot_memory_pressure(struct proto *proto)
2890{
2891 return proto->memory_pressure != NULL ?
2892 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2893}
1da177e4
LT		 2894
2895static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2896{
180d8cd9 2897
8d987e5c 2898 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
1da177e4
LT		 2899 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2900 proto->name,
2901 proto->obj_size,
14e943db 2902 sock_prot_inuse_get(seq_file_net(seq), proto),
180d8cd9
GC		 2903 sock_prot_memory_allocated(proto),
2904 sock_prot_memory_pressure(proto),
1da177e4
LT		 2905 proto->max_header,
2906 proto->slab == NULL ? "no" : "yes",
2907 module_name(proto->owner),
2908 proto_method_implemented(proto->close),
2909 proto_method_implemented(proto->connect),
2910 proto_method_implemented(proto->disconnect),
2911 proto_method_implemented(proto->accept),
2912 proto_method_implemented(proto->ioctl),
2913 proto_method_implemented(proto->init),
2914 proto_method_implemented(proto->destroy),
2915 proto_method_implemented(proto->shutdown),
2916 proto_method_implemented(proto->setsockopt),
2917 proto_method_implemented(proto->getsockopt),
2918 proto_method_implemented(proto->sendmsg),
2919 proto_method_implemented(proto->recvmsg),
2920 proto_method_implemented(proto->sendpage),
2921 proto_method_implemented(proto->bind),
2922 proto_method_implemented(proto->backlog_rcv),
2923 proto_method_implemented(proto->hash),
2924 proto_method_implemented(proto->unhash),
2925 proto_method_implemented(proto->get_port),
2926 proto_method_implemented(proto->enter_memory_pressure));
2927}
2928
2929static int proto_seq_show(struct seq_file *seq, void *v)
2930{
60f0438a 2931 if (v == &proto_list)
1da177e4
LT		 2932 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2933 "protocol",
2934 "size",
2935 "sockets",
2936 "memory",
2937 "press",
2938 "maxhdr",
2939 "slab",
2940 "module",
2941 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2942 else
60f0438a 2943 proto_seq_printf(seq, list_entry(v, struct proto, node));
1da177e4
LT		 2944 return 0;
2945}
2946
f690808e 2947static const struct seq_operations proto_seq_ops = {
1da177e4
LT		 2948 .start = proto_seq_start,
2949 .next = proto_seq_next,
2950 .stop = proto_seq_stop,
2951 .show = proto_seq_show,
2952};
2953
2954static int proto_seq_open(struct inode *inode, struct file *file)
2955{
14e943db
ED		 2956 return seq_open_net(inode, file, &proto_seq_ops,
2957 sizeof(struct seq_net_private));
1da177e4
LT		 2958}
2959
9a32144e 2960static const struct file_operations proto_seq_fops = {
1da177e4
LT		 2961 .owner = THIS_MODULE,
2962 .open = proto_seq_open,
2963 .read = seq_read,
2964 .llseek = seq_lseek,
14e943db
ED		 2965 .release = seq_release_net,
2966};
2967
2968static __net_init int proto_init_net(struct net *net)
2969{
d4beaa66 2970 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
14e943db
ED		 2971 return -ENOMEM;
2972
2973 return 0;
2974}
2975
2976static __net_exit void proto_exit_net(struct net *net)
2977{
ece31ffd 2978 remove_proc_entry("protocols", net->proc_net);
14e943db
ED		 2979}
2980
2981
2982static __net_initdata struct pernet_operations proto_net_ops = {
2983 .init = proto_init_net,
2984 .exit = proto_exit_net,
1da177e4
LT		 2985};
2986
2987static int __init proto_init(void)
2988{
14e943db 2989 return register_pernet_subsys(&proto_net_ops);
1da177e4
LT		 2990}
2991
2992subsys_initcall(proto_init);
2993
2994#endif /* PROC_FS */
Linux kernel - Deliverables
This page took 1.007548 seconds and 5 git commands to generate.