blob: ca2984afe16ed6545a6d27cdfa91a3c5548bcbf2 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Generic socket support routines. Memory allocators, socket lock/release
7 * handler for protocols to use and generic option handler.
8 *
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Florian La Roche, <flla@stud.uni-sb.de>
13 * Alan Cox, <A.Cox@swansea.ac.uk>
14 *
15 * Fixes:
16 * Alan Cox : Numerous verify_area() problems
17 * Alan Cox : Connecting on a connecting socket
18 * now returns an error for tcp.
19 * Alan Cox : sock->protocol is set correctly.
20 * and is not sometimes left as 0.
21 * Alan Cox : connect handles icmp errors on a
22 * connect properly. Unfortunately there
23 * is a restart syscall nasty there. I
24 * can't match BSD without hacking the C
25 * library. Ideas urgently sought!
26 * Alan Cox : Disallow bind() to addresses that are
27 * not ours - especially broadcast ones!!
28 * Alan Cox : Socket 1024 _IS_ ok for users. (fencepost)
29 * Alan Cox : sock_wfree/sock_rfree don't destroy sockets,
30 * instead they leave that for the DESTROY timer.
31 * Alan Cox : Clean up error flag in accept
32 * Alan Cox : TCP ack handling is buggy, the DESTROY timer
33 * was buggy. Put a remove_sock() in the handler
34 * for memory when we hit 0. Also altered the timer
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +090035 * code. The ACK stuff can wait and needs major
Linus Torvalds1da177e2005-04-16 15:20:36 -070036 * TCP layer surgery.
37 * Alan Cox : Fixed TCP ack bug, removed remove sock
38 * and fixed timer/inet_bh race.
39 * Alan Cox : Added zapped flag for TCP
40 * Alan Cox : Move kfree_skb into skbuff.c and tidied up surplus code
41 * Alan Cox : for new sk_buff allocations wmalloc/rmalloc now call alloc_skb
42 * Alan Cox : kfree_s calls now are kfree_skbmem so we can track skb resources
43 * Alan Cox : Supports socket option broadcast now as does udp. Packet and raw need fixing.
44 * Alan Cox : Added RCVBUF,SNDBUF size setting. It suddenly occurred to me how easy it was so...
45 * Rick Sladkey : Relaxed UDP rules for matching packets.
46 * C.E.Hawkins : IFF_PROMISC/SIOCGHWADDR support
47 * Pauline Middelink : identd support
48 * Alan Cox : Fixed connect() taking signals I think.
49 * Alan Cox : SO_LINGER supported
50 * Alan Cox : Error reporting fixes
51 * Anonymous : inet_create tidied up (sk->reuse setting)
52 * Alan Cox : inet sockets don't set sk->type!
53 * Alan Cox : Split socket option code
54 * Alan Cox : Callbacks
55 * Alan Cox : Nagle flag for Charles & Johannes stuff
56 * Alex : Removed restriction on inet fioctl
57 * Alan Cox : Splitting INET from NET core
58 * Alan Cox : Fixed bogus SO_TYPE handling in getsockopt()
59 * Adam Caldwell : Missing return in SO_DONTROUTE/SO_DEBUG code
60 * Alan Cox : Split IP from generic code
61 * Alan Cox : New kfree_skbmem()
62 * Alan Cox : Make SO_DEBUG superuser only.
63 * Alan Cox : Allow anyone to clear SO_DEBUG
64 * (compatibility fix)
65 * Alan Cox : Added optimistic memory grabbing for AF_UNIX throughput.
66 * Alan Cox : Allocator for a socket is settable.
67 * Alan Cox : SO_ERROR includes soft errors.
68 * Alan Cox : Allow NULL arguments on some SO_ opts
69 * Alan Cox : Generic socket allocation to make hooks
70 * easier (suggested by Craig Metz).
71 * Michael Pall : SO_ERROR returns positive errno again
72 * Steve Whitehouse: Added default destructor to free
73 * protocol private data.
74 * Steve Whitehouse: Added various other default routines
75 * common to several socket families.
76 * Chris Evans : Call suser() check last on F_SETOWN
77 * Jay Schulist : Added SO_ATTACH_FILTER and SO_DETACH_FILTER.
78 * Andi Kleen : Add sock_kmalloc()/sock_kfree_s()
79 * Andi Kleen : Fix write_space callback
80 * Chris Evans : Security fixes - signedness again
81 * Arnaldo C. Melo : cleanups, use skb_queue_purge
82 *
83 * To Fix:
84 *
85 *
86 * This program is free software; you can redistribute it and/or
87 * modify it under the terms of the GNU General Public License
88 * as published by the Free Software Foundation; either version
89 * 2 of the License, or (at your option) any later version.
90 */
91
Joe Perchese005d192012-05-16 19:58:40 +000092#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
93
Randy Dunlap4fc268d2006-01-11 12:17:47 -080094#include <linux/capability.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070095#include <linux/errno.h>
Richard Cochrancb820f82013-07-19 19:40:09 +020096#include <linux/errqueue.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070097#include <linux/types.h>
98#include <linux/socket.h>
99#include <linux/in.h>
100#include <linux/kernel.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101#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>
Al Viroa1f8e7f72006-10-19 16:08:53 -0400115#include <linux/highmem.h>
Eric W. Biederman3f551f92010-06-13 03:28:59 +0000116#include <linux/user_namespace.h>
Ingo Molnarc5905af2012-02-24 08:31:31 +0100117#include <linux/static_key.h>
David S. Miller3969eb32012-01-09 13:44:23 -0800118#include <linux/memcontrol.h>
David S. Miller8c1ae102012-05-03 02:25:55 -0400119#include <linux/prefetch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
121#include <asm/uaccess.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700122
123#include <linux/netdevice.h>
124#include <net/protocol.h>
125#include <linux/skbuff.h>
Eric W. Biederman457c4cb2007-09-12 12:01:34 +0200126#include <net/net_namespace.h>
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -0700127#include <net/request_sock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128#include <net/sock.h>
Patrick Ohly20d49472009-02-12 05:03:38 +0000129#include <linux/net_tstamp.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700130#include <net/xfrm.h>
131#include <linux/ipsec.h>
Herbert Xuf8451722010-05-24 00:12:34 -0700132#include <net/cls_cgroup.h>
Neil Horman5bc14212011-11-22 05:10:51 +0000133#include <net/netprio_cgroup.h>
Craig Gallekeb4cb002015-06-15 11:26:18 -0400134#include <linux/sock_diag.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135
136#include <linux/filter.h>
137
Satoru Moriya3847ce32011-06-17 12:00:03 +0000138#include <trace/events/sock.h>
139
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140#ifdef CONFIG_INET
141#include <net/tcp.h>
142#endif
143
Eliezer Tamir076bb0c2013-07-10 17:13:17 +0300144#include <net/busy_poll.h>
Eliezer Tamir06021292013-06-10 11:39:50 +0300145
Glauber Costa36b77a52011-12-16 00:51:59 +0000146static DEFINE_MUTEX(proto_list_mutex);
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000147static LIST_HEAD(proto_list);
148
Eric W. Biedermana3b299d2014-04-23 14:26:56 -0700149/**
150 * sk_ns_capable - General socket capability test
151 * @sk: Socket to use a capability on or through
152 * @user_ns: The user namespace of the capability to use
153 * @cap: The capability to use
154 *
155 * Test to see if the opener of the socket had when the socket was
156 * created and the current process has the capability @cap in the user
157 * namespace @user_ns.
158 */
159bool sk_ns_capable(const struct sock *sk,
160 struct user_namespace *user_ns, int cap)
161{
162 return file_ns_capable(sk->sk_socket->file, user_ns, cap) &&
163 ns_capable(user_ns, cap);
164}
165EXPORT_SYMBOL(sk_ns_capable);
166
167/**
168 * sk_capable - Socket global capability test
169 * @sk: Socket to use a capability on or through
Masanari Iidae793c0f2014-09-04 23:44:36 +0900170 * @cap: The global capability to use
Eric W. Biedermana3b299d2014-04-23 14:26:56 -0700171 *
172 * Test to see if the opener of the socket had when the socket was
173 * created and the current process has the capability @cap in all user
174 * namespaces.
175 */
176bool sk_capable(const struct sock *sk, int cap)
177{
178 return sk_ns_capable(sk, &init_user_ns, cap);
179}
180EXPORT_SYMBOL(sk_capable);
181
182/**
183 * sk_net_capable - Network namespace socket capability test
184 * @sk: Socket to use a capability on or through
185 * @cap: The capability to use
186 *
Masanari Iidae793c0f2014-09-04 23:44:36 +0900187 * Test to see if the opener of the socket had when the socket was created
Eric W. Biedermana3b299d2014-04-23 14:26:56 -0700188 * and the current process has the capability @cap over the network namespace
189 * the socket is a member of.
190 */
191bool sk_net_capable(const struct sock *sk, int cap)
192{
193 return sk_ns_capable(sk, sock_net(sk)->user_ns, cap);
194}
195EXPORT_SYMBOL(sk_net_capable);
196
197
Andrew Mortonc255a452012-07-31 16:43:02 -0700198#ifdef CONFIG_MEMCG_KMEM
Glauber Costa1d62e432012-04-09 19:36:33 -0300199int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000200{
201 struct proto *proto;
202 int ret = 0;
203
Glauber Costa36b77a52011-12-16 00:51:59 +0000204 mutex_lock(&proto_list_mutex);
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000205 list_for_each_entry(proto, &proto_list, node) {
206 if (proto->init_cgroup) {
Glauber Costa1d62e432012-04-09 19:36:33 -0300207 ret = proto->init_cgroup(memcg, ss);
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000208 if (ret)
209 goto out;
210 }
211 }
212
Glauber Costa36b77a52011-12-16 00:51:59 +0000213 mutex_unlock(&proto_list_mutex);
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000214 return ret;
215out:
216 list_for_each_entry_continue_reverse(proto, &proto_list, node)
217 if (proto->destroy_cgroup)
Glauber Costa1d62e432012-04-09 19:36:33 -0300218 proto->destroy_cgroup(memcg);
Glauber Costa36b77a52011-12-16 00:51:59 +0000219 mutex_unlock(&proto_list_mutex);
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000220 return ret;
221}
222
Glauber Costa1d62e432012-04-09 19:36:33 -0300223void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000224{
225 struct proto *proto;
226
Glauber Costa36b77a52011-12-16 00:51:59 +0000227 mutex_lock(&proto_list_mutex);
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000228 list_for_each_entry_reverse(proto, &proto_list, node)
229 if (proto->destroy_cgroup)
Glauber Costa1d62e432012-04-09 19:36:33 -0300230 proto->destroy_cgroup(memcg);
Glauber Costa36b77a52011-12-16 00:51:59 +0000231 mutex_unlock(&proto_list_mutex);
Glauber Costad1a4c0b2011-12-11 21:47:04 +0000232}
233#endif
234
Ingo Molnarda21f242006-07-03 00:25:12 -0700235/*
236 * Each address family might have different locking rules, so we have
237 * one slock key per address family:
238 */
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700239static struct lock_class_key af_family_keys[AF_MAX];
240static struct lock_class_key af_family_slock_keys[AF_MAX];
241
stephen hemmingercbda4ea2013-02-22 07:59:10 +0000242#if defined(CONFIG_MEMCG_KMEM)
Ingo Molnarc5905af2012-02-24 08:31:31 +0100243struct static_key memcg_socket_limit_enabled;
Glauber Costae1aab162011-12-11 21:47:03 +0000244EXPORT_SYMBOL(memcg_socket_limit_enabled);
stephen hemmingercbda4ea2013-02-22 07:59:10 +0000245#endif
Glauber Costae1aab162011-12-11 21:47:03 +0000246
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700247/*
248 * Make lock validator output more readable. (we pre-construct these
249 * strings build-time, so that runtime initialization of socket
250 * locks is fast):
251 */
Jan Engelhardt36cbd3d2009-08-05 10:42:58 -0700252static const char *const af_family_key_strings[AF_MAX+1] = {
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700253 "sk_lock-AF_UNSPEC", "sk_lock-AF_UNIX" , "sk_lock-AF_INET" ,
254 "sk_lock-AF_AX25" , "sk_lock-AF_IPX" , "sk_lock-AF_APPLETALK",
255 "sk_lock-AF_NETROM", "sk_lock-AF_BRIDGE" , "sk_lock-AF_ATMPVC" ,
256 "sk_lock-AF_X25" , "sk_lock-AF_INET6" , "sk_lock-AF_ROSE" ,
257 "sk_lock-AF_DECnet", "sk_lock-AF_NETBEUI" , "sk_lock-AF_SECURITY" ,
258 "sk_lock-AF_KEY" , "sk_lock-AF_NETLINK" , "sk_lock-AF_PACKET" ,
259 "sk_lock-AF_ASH" , "sk_lock-AF_ECONET" , "sk_lock-AF_ATMSVC" ,
Andy Grovercbd151b2009-02-26 23:43:19 -0800260 "sk_lock-AF_RDS" , "sk_lock-AF_SNA" , "sk_lock-AF_IRDA" ,
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700261 "sk_lock-AF_PPPOX" , "sk_lock-AF_WANPIPE" , "sk_lock-AF_LLC" ,
Oliver Hartkoppcd05acf2007-12-16 15:59:24 -0800262 "sk_lock-27" , "sk_lock-28" , "sk_lock-AF_CAN" ,
David Howells17926a72007-04-26 15:48:28 -0700263 "sk_lock-AF_TIPC" , "sk_lock-AF_BLUETOOTH", "sk_lock-IUCV" ,
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700264 "sk_lock-AF_RXRPC" , "sk_lock-AF_ISDN" , "sk_lock-AF_PHONET" ,
Miloslav Trmač6f107b52010-12-08 14:35:34 +0800265 "sk_lock-AF_IEEE802154", "sk_lock-AF_CAIF" , "sk_lock-AF_ALG" ,
Federico Vaga456db6a2013-05-28 05:02:44 +0000266 "sk_lock-AF_NFC" , "sk_lock-AF_VSOCK" , "sk_lock-AF_MAX"
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700267};
Jan Engelhardt36cbd3d2009-08-05 10:42:58 -0700268static const char *const af_family_slock_key_strings[AF_MAX+1] = {
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700269 "slock-AF_UNSPEC", "slock-AF_UNIX" , "slock-AF_INET" ,
270 "slock-AF_AX25" , "slock-AF_IPX" , "slock-AF_APPLETALK",
271 "slock-AF_NETROM", "slock-AF_BRIDGE" , "slock-AF_ATMPVC" ,
272 "slock-AF_X25" , "slock-AF_INET6" , "slock-AF_ROSE" ,
273 "slock-AF_DECnet", "slock-AF_NETBEUI" , "slock-AF_SECURITY" ,
274 "slock-AF_KEY" , "slock-AF_NETLINK" , "slock-AF_PACKET" ,
275 "slock-AF_ASH" , "slock-AF_ECONET" , "slock-AF_ATMSVC" ,
Andy Grovercbd151b2009-02-26 23:43:19 -0800276 "slock-AF_RDS" , "slock-AF_SNA" , "slock-AF_IRDA" ,
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700277 "slock-AF_PPPOX" , "slock-AF_WANPIPE" , "slock-AF_LLC" ,
Oliver Hartkoppcd05acf2007-12-16 15:59:24 -0800278 "slock-27" , "slock-28" , "slock-AF_CAN" ,
David Howells17926a72007-04-26 15:48:28 -0700279 "slock-AF_TIPC" , "slock-AF_BLUETOOTH", "slock-AF_IUCV" ,
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700280 "slock-AF_RXRPC" , "slock-AF_ISDN" , "slock-AF_PHONET" ,
Miloslav Trmač6f107b52010-12-08 14:35:34 +0800281 "slock-AF_IEEE802154", "slock-AF_CAIF" , "slock-AF_ALG" ,
Federico Vaga456db6a2013-05-28 05:02:44 +0000282 "slock-AF_NFC" , "slock-AF_VSOCK" ,"slock-AF_MAX"
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700283};
Jan Engelhardt36cbd3d2009-08-05 10:42:58 -0700284static const char *const af_family_clock_key_strings[AF_MAX+1] = {
Peter Zijlstra443aef02007-07-19 01:49:00 -0700285 "clock-AF_UNSPEC", "clock-AF_UNIX" , "clock-AF_INET" ,
286 "clock-AF_AX25" , "clock-AF_IPX" , "clock-AF_APPLETALK",
287 "clock-AF_NETROM", "clock-AF_BRIDGE" , "clock-AF_ATMPVC" ,
288 "clock-AF_X25" , "clock-AF_INET6" , "clock-AF_ROSE" ,
289 "clock-AF_DECnet", "clock-AF_NETBEUI" , "clock-AF_SECURITY" ,
290 "clock-AF_KEY" , "clock-AF_NETLINK" , "clock-AF_PACKET" ,
291 "clock-AF_ASH" , "clock-AF_ECONET" , "clock-AF_ATMSVC" ,
Andy Grovercbd151b2009-02-26 23:43:19 -0800292 "clock-AF_RDS" , "clock-AF_SNA" , "clock-AF_IRDA" ,
Peter Zijlstra443aef02007-07-19 01:49:00 -0700293 "clock-AF_PPPOX" , "clock-AF_WANPIPE" , "clock-AF_LLC" ,
Oliver Hartkoppb4942af2008-07-23 14:06:04 -0700294 "clock-27" , "clock-28" , "clock-AF_CAN" ,
David Howellse51f8022007-07-21 19:30:16 -0700295 "clock-AF_TIPC" , "clock-AF_BLUETOOTH", "clock-AF_IUCV" ,
Remi Denis-Courmontbce7b152008-09-22 19:51:15 -0700296 "clock-AF_RXRPC" , "clock-AF_ISDN" , "clock-AF_PHONET" ,
Miloslav Trmač6f107b52010-12-08 14:35:34 +0800297 "clock-AF_IEEE802154", "clock-AF_CAIF" , "clock-AF_ALG" ,
Federico Vaga456db6a2013-05-28 05:02:44 +0000298 "clock-AF_NFC" , "clock-AF_VSOCK" , "clock-AF_MAX"
Peter Zijlstra443aef02007-07-19 01:49:00 -0700299};
Ingo Molnarda21f242006-07-03 00:25:12 -0700300
301/*
302 * sk_callback_lock locking rules are per-address-family,
303 * so split the lock classes by using a per-AF key:
304 */
305static struct lock_class_key af_callback_keys[AF_MAX];
306
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307/* Take into consideration the size of the struct sk_buff overhead in the
308 * determination of these values, since that is non-constant across
309 * platforms. This makes socket queueing behavior and performance
310 * not depend upon such differences.
311 */
312#define _SK_MEM_PACKETS 256
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000313#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
315#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
316
317/* Run time adjustable parameters. */
Brian Haleyab32ea52006-09-22 14:15:41 -0700318__u32 sysctl_wmem_max __read_mostly = SK_WMEM_MAX;
Hans Schillstrom6d8ebc82012-04-30 08:13:50 +0200319EXPORT_SYMBOL(sysctl_wmem_max);
Brian Haleyab32ea52006-09-22 14:15:41 -0700320__u32 sysctl_rmem_max __read_mostly = SK_RMEM_MAX;
Hans Schillstrom6d8ebc82012-04-30 08:13:50 +0200321EXPORT_SYMBOL(sysctl_rmem_max);
Brian Haleyab32ea52006-09-22 14:15:41 -0700322__u32 sysctl_wmem_default __read_mostly = SK_WMEM_MAX;
323__u32 sysctl_rmem_default __read_mostly = SK_RMEM_MAX;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300325/* Maximal space eaten by iovec or ancillary data plus some space */
Brian Haleyab32ea52006-09-22 14:15:41 -0700326int sysctl_optmem_max __read_mostly = sizeof(unsigned long)*(2*UIO_MAXIOV+512);
Eric Dumazet2a915252009-05-27 11:30:05 +0000327EXPORT_SYMBOL(sysctl_optmem_max);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328
Willem de Bruijnb245be12015-01-30 13:29:32 -0500329int sysctl_tstamp_allow_data __read_mostly = 1;
330
Mel Gormanc93bdd02012-07-31 16:44:19 -0700331struct static_key memalloc_socks = STATIC_KEY_INIT_FALSE;
332EXPORT_SYMBOL_GPL(memalloc_socks);
333
Mel Gorman7cb02402012-07-31 16:44:16 -0700334/**
335 * sk_set_memalloc - sets %SOCK_MEMALLOC
336 * @sk: socket to set it on
337 *
338 * Set %SOCK_MEMALLOC on a socket for access to emergency reserves.
339 * It's the responsibility of the admin to adjust min_free_kbytes
340 * to meet the requirements
341 */
342void sk_set_memalloc(struct sock *sk)
343{
344 sock_set_flag(sk, SOCK_MEMALLOC);
345 sk->sk_allocation |= __GFP_MEMALLOC;
Mel Gormanc93bdd02012-07-31 16:44:19 -0700346 static_key_slow_inc(&memalloc_socks);
Mel Gorman7cb02402012-07-31 16:44:16 -0700347}
348EXPORT_SYMBOL_GPL(sk_set_memalloc);
349
350void sk_clear_memalloc(struct sock *sk)
351{
352 sock_reset_flag(sk, SOCK_MEMALLOC);
353 sk->sk_allocation &= ~__GFP_MEMALLOC;
Mel Gormanc93bdd02012-07-31 16:44:19 -0700354 static_key_slow_dec(&memalloc_socks);
Mel Gormanc76562b2012-07-31 16:44:41 -0700355
356 /*
357 * SOCK_MEMALLOC is allowed to ignore rmem limits to ensure forward
Mel Gorman5d753612015-06-10 21:02:04 -0400358 * progress of swapping. SOCK_MEMALLOC may be cleared while
359 * it has rmem allocations due to the last swapfile being deactivated
360 * but there is a risk that the socket is unusable due to exceeding
361 * the rmem limits. Reclaim the reserves and obey rmem limits again.
Mel Gormanc76562b2012-07-31 16:44:41 -0700362 */
Mel Gorman5d753612015-06-10 21:02:04 -0400363 sk_mem_reclaim(sk);
Mel Gorman7cb02402012-07-31 16:44:16 -0700364}
365EXPORT_SYMBOL_GPL(sk_clear_memalloc);
366
Mel Gormanb4b9e352012-07-31 16:44:26 -0700367int __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
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383static 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;
Vasily Averinba780732007-05-24 16:58:54 -0700391 if (tv.tv_usec < 0 || tv.tv_usec >= USEC_PER_SEC)
392 return -EDOM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393
Vasily Averinba780732007-05-24 16:58:54 -0700394 if (tv.tv_sec < 0) {
Andrew Morton6f11df82007-07-09 13:16:00 -0700395 static int warned __read_mostly;
396
Vasily Averinba780732007-05-24 16:58:54 -0700397 *timeo_p = 0;
Ilpo Järvinen50aab542008-05-02 16:20:10 -0700398 if (warned < 10 && net_ratelimit()) {
Vasily Averinba780732007-05-24 16:58:54 -0700399 warned++;
Joe Perchese005d192012-05-16 19:58:40 +0000400 pr_info("%s: `%s' (pid %d) tries to set negative timeout\n",
401 __func__, current->comm, task_pid_nr(current));
Ilpo Järvinen50aab542008-05-02 16:20:10 -0700402 }
Vasily Averinba780732007-05-24 16:58:54 -0700403 return 0;
404 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405 *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];
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900417 if (strcmp(warncomm, current->comm) && warned < 5) {
418 strcpy(warncomm, current->comm);
Joe Perchese005d192012-05-16 19:58:40 +0000419 pr_warn("process `%s' is using obsolete %s SO_BSDCOMPAT\n",
420 warncomm, name);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 warned++;
422 }
423}
424
Eric Dumazet08e29af2011-11-28 12:04:18 +0000425#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
426
427static void sock_disable_timestamp(struct sock *sk, unsigned long flags)
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900428{
Eric Dumazet08e29af2011-11-28 12:04:18 +0000429 if (sk->sk_flags & flags) {
430 sk->sk_flags &= ~flags;
431 if (!(sk->sk_flags & SK_FLAGS_TIMESTAMP))
Patrick Ohly20d49472009-02-12 05:03:38 +0000432 net_disable_timestamp();
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 }
434}
435
436
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800437int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb)
438{
Eric Dumazet766e90372009-10-14 20:40:11 -0700439 int err;
Neil Horman3b885782009-10-12 13:26:31 -0700440 unsigned long flags;
441 struct sk_buff_head *list = &sk->sk_receive_queue;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800442
Eric Dumazet0fd7bac2011-12-21 07:11:44 +0000443 if (atomic_read(&sk->sk_rmem_alloc) >= sk->sk_rcvbuf) {
Eric Dumazet766e90372009-10-14 20:40:11 -0700444 atomic_inc(&sk->sk_drops);
Satoru Moriya3847ce32011-06-17 12:00:03 +0000445 trace_sock_rcvqueue_full(sk, skb);
Eric Dumazet766e90372009-10-14 20:40:11 -0700446 return -ENOMEM;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800447 }
448
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700449 err = sk_filter(sk, skb);
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800450 if (err)
Eric Dumazet766e90372009-10-14 20:40:11 -0700451 return err;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800452
Mel Gormanc76562b2012-07-31 16:44:41 -0700453 if (!sk_rmem_schedule(sk, skb, skb->truesize)) {
Eric Dumazet766e90372009-10-14 20:40:11 -0700454 atomic_inc(&sk->sk_drops);
455 return -ENOBUFS;
Hideo Aoki3ab224b2007-12-31 00:11:19 -0800456 }
457
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800458 skb->dev = NULL;
459 skb_set_owner_r(skb, sk);
David S. Miller49ad9592008-12-17 22:11:38 -0800460
Eric Dumazet7fee2262010-05-11 23:19:48 +0000461 /* we escape from rcu protected region, make sure we dont leak
462 * a norefcounted dst
463 */
464 skb_dst_force(skb);
465
Neil Horman3b885782009-10-12 13:26:31 -0700466 spin_lock_irqsave(&list->lock, flags);
Eyal Birger3bc3b962015-03-01 14:58:30 +0200467 sock_skb_set_dropcount(sk, skb);
Neil Horman3b885782009-10-12 13:26:31 -0700468 __skb_queue_tail(list, skb);
469 spin_unlock_irqrestore(&list->lock, flags);
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800470
471 if (!sock_flag(sk, SOCK_DEAD))
David S. Miller676d2362014-04-11 16:15:36 -0400472 sk->sk_data_ready(sk);
Eric Dumazet766e90372009-10-14 20:40:11 -0700473 return 0;
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800474}
475EXPORT_SYMBOL(sock_queue_rcv_skb);
476
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200477int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested)
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800478{
479 int rc = NET_RX_SUCCESS;
480
Dmitry Mishinfda9ef52006-08-31 15:28:39 -0700481 if (sk_filter(sk, skb))
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800482 goto discard_and_relse;
483
484 skb->dev = NULL;
485
Sorin Dumitru274f4822014-07-22 21:16:51 +0300486 if (sk_rcvqueues_full(sk, sk->sk_rcvbuf)) {
Eric Dumazetc3774112010-04-27 15:13:20 -0700487 atomic_inc(&sk->sk_drops);
488 goto discard_and_relse;
489 }
Arnaldo Carvalho de Melo58a5a7b2006-11-16 14:06:06 -0200490 if (nested)
491 bh_lock_sock_nested(sk);
492 else
493 bh_lock_sock(sk);
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700494 if (!sock_owned_by_user(sk)) {
495 /*
496 * trylock + unlock semantics:
497 */
498 mutex_acquire(&sk->sk_lock.dep_map, 0, 1, _RET_IP_);
499
Peter Zijlstrac57943a2008-10-07 14:18:42 -0700500 rc = sk_backlog_rcv(sk, skb);
Ingo Molnara5b5bb92006-07-03 00:25:35 -0700501
502 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
Eric Dumazetf545a382012-04-22 23:34:26 +0000503 } else if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) {
Zhu Yi8eae9392010-03-04 18:01:40 +0000504 bh_unlock_sock(sk);
505 atomic_inc(&sk->sk_drops);
506 goto discard_and_relse;
507 }
508
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800509 bh_unlock_sock(sk);
510out:
511 sock_put(sk);
512 return rc;
513discard_and_relse:
514 kfree_skb(skb);
515 goto out;
516}
517EXPORT_SYMBOL(sk_receive_skb);
518
519struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie)
520{
Eric Dumazetb6c67122010-04-08 23:03:29 +0000521 struct dst_entry *dst = __sk_dst_get(sk);
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800522
523 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
Krishna Kumare022f0b2009-10-19 23:46:20 +0000524 sk_tx_queue_clear(sk);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +0000525 RCU_INIT_POINTER(sk->sk_dst_cache, NULL);
Denis Vlasenkof0088a52006-03-28 01:08:21 -0800526 dst_release(dst);
527 return NULL;
528 }
529
530 return dst;
531}
532EXPORT_SYMBOL(__sk_dst_check);
533
534struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie)
535{
536 struct dst_entry *dst = sk_dst_get(sk);
537
538 if (dst && dst->obsolete && dst->ops->check(dst, cookie) == NULL) {
539 sk_dst_reset(sk);
540 dst_release(dst);
541 return NULL;
542 }
543
544 return dst;
545}
546EXPORT_SYMBOL(sk_dst_check);
547
Brian Haleyc91f6df2012-11-26 05:21:08 +0000548static int sock_setbindtodevice(struct sock *sk, char __user *optval,
549 int optlen)
David S. Miller48788092007-09-14 16:41:03 -0700550{
551 int ret = -ENOPROTOOPT;
552#ifdef CONFIG_NETDEVICES
YOSHIFUJI Hideaki3b1e0a62008-03-26 02:26:21 +0900553 struct net *net = sock_net(sk);
David S. Miller48788092007-09-14 16:41:03 -0700554 char devname[IFNAMSIZ];
555 int index;
556
557 /* Sorry... */
558 ret = -EPERM;
Eric W. Biederman5e1fccc2012-11-16 03:03:04 +0000559 if (!ns_capable(net->user_ns, CAP_NET_RAW))
David S. Miller48788092007-09-14 16:41:03 -0700560 goto out;
561
562 ret = -EINVAL;
563 if (optlen < 0)
564 goto out;
565
566 /* Bind this socket to a particular device like "eth0",
567 * as specified in the passed interface name. If the
568 * name is "" or the option length is zero the socket
569 * is not bound.
570 */
571 if (optlen > IFNAMSIZ - 1)
572 optlen = IFNAMSIZ - 1;
573 memset(devname, 0, sizeof(devname));
574
575 ret = -EFAULT;
576 if (copy_from_user(devname, optval, optlen))
577 goto out;
578
David S. Miller000ba2e2009-11-05 22:37:11 -0800579 index = 0;
580 if (devname[0] != '\0') {
Eric Dumazetbf8e56b2009-11-05 21:03:39 -0800581 struct net_device *dev;
David S. Miller48788092007-09-14 16:41:03 -0700582
Eric Dumazetbf8e56b2009-11-05 21:03:39 -0800583 rcu_read_lock();
584 dev = dev_get_by_name_rcu(net, devname);
585 if (dev)
586 index = dev->ifindex;
587 rcu_read_unlock();
David S. Miller48788092007-09-14 16:41:03 -0700588 ret = -ENODEV;
589 if (!dev)
590 goto out;
David S. Miller48788092007-09-14 16:41:03 -0700591 }
592
593 lock_sock(sk);
594 sk->sk_bound_dev_if = index;
595 sk_dst_reset(sk);
596 release_sock(sk);
597
598 ret = 0;
599
600out:
601#endif
602
603 return ret;
604}
605
Brian Haleyc91f6df2012-11-26 05:21:08 +0000606static int sock_getbindtodevice(struct sock *sk, char __user *optval,
607 int __user *optlen, int len)
608{
609 int ret = -ENOPROTOOPT;
610#ifdef CONFIG_NETDEVICES
611 struct net *net = sock_net(sk);
Brian Haleyc91f6df2012-11-26 05:21:08 +0000612 char devname[IFNAMSIZ];
Brian Haleyc91f6df2012-11-26 05:21:08 +0000613
614 if (sk->sk_bound_dev_if == 0) {
615 len = 0;
616 goto zero;
617 }
618
619 ret = -EINVAL;
620 if (len < IFNAMSIZ)
621 goto out;
622
Nicolas Schichan5dbe7c12013-06-26 17:23:42 +0200623 ret = netdev_get_name(net, devname, sk->sk_bound_dev_if);
624 if (ret)
Brian Haleyc91f6df2012-11-26 05:21:08 +0000625 goto out;
Brian Haleyc91f6df2012-11-26 05:21:08 +0000626
627 len = strlen(devname) + 1;
628
629 ret = -EFAULT;
630 if (copy_to_user(optval, devname, len))
631 goto out;
632
633zero:
634 ret = -EFAULT;
635 if (put_user(len, optlen))
636 goto out;
637
638 ret = 0;
639
640out:
641#endif
642
643 return ret;
644}
645
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800646static inline void sock_valbool_flag(struct sock *sk, int bit, int valbool)
647{
648 if (valbool)
649 sock_set_flag(sk, bit);
650 else
651 sock_reset_flag(sk, bit);
652}
653
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +0200654bool sk_mc_loop(struct sock *sk)
655{
656 if (dev_recursion_level())
657 return false;
658 if (!sk)
659 return true;
660 switch (sk->sk_family) {
661 case AF_INET:
662 return inet_sk(sk)->mc_loop;
663#if IS_ENABLED(CONFIG_IPV6)
664 case AF_INET6:
665 return inet6_sk(sk)->mc_loop;
666#endif
667 }
668 WARN_ON(1);
669 return true;
670}
671EXPORT_SYMBOL(sk_mc_loop);
672
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673/*
674 * This is meant for all protocols to use and covers goings on
675 * at the socket level. Everything here is generic.
676 */
677
678int sock_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -0700679 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680{
Eric Dumazet2a915252009-05-27 11:30:05 +0000681 struct sock *sk = sock->sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 int val;
683 int valbool;
684 struct linger ling;
685 int ret = 0;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900686
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 /*
688 * Options without arguments
689 */
690
David S. Miller48788092007-09-14 16:41:03 -0700691 if (optname == SO_BINDTODEVICE)
Brian Haleyc91f6df2012-11-26 05:21:08 +0000692 return sock_setbindtodevice(sk, optval, optlen);
David S. Miller48788092007-09-14 16:41:03 -0700693
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700694 if (optlen < sizeof(int))
695 return -EINVAL;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900696
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 if (get_user(val, (int __user *)optval))
698 return -EFAULT;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900699
Eric Dumazet2a915252009-05-27 11:30:05 +0000700 valbool = val ? 1 : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
702 lock_sock(sk);
703
Eric Dumazet2a915252009-05-27 11:30:05 +0000704 switch (optname) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700705 case SO_DEBUG:
Eric Dumazet2a915252009-05-27 11:30:05 +0000706 if (val && !capable(CAP_NET_ADMIN))
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700707 ret = -EACCES;
Eric Dumazet2a915252009-05-27 11:30:05 +0000708 else
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800709 sock_valbool_flag(sk, SOCK_DBG, valbool);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700710 break;
711 case SO_REUSEADDR:
Pavel Emelyanov4a17fd52012-04-19 03:39:36 +0000712 sk->sk_reuse = (valbool ? SK_CAN_REUSE : SK_NO_REUSE);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700713 break;
Tom Herbert055dc212013-01-22 09:49:50 +0000714 case SO_REUSEPORT:
715 sk->sk_reuseport = valbool;
716 break;
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700717 case SO_TYPE:
Jan Engelhardt49c794e2009-08-04 07:28:28 +0000718 case SO_PROTOCOL:
Jan Engelhardt0d6038e2009-08-04 07:28:29 +0000719 case SO_DOMAIN:
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700720 case SO_ERROR:
721 ret = -ENOPROTOOPT;
722 break;
723 case SO_DONTROUTE:
Pavel Emelyanovc0ef8772007-11-15 03:03:19 -0800724 sock_valbool_flag(sk, SOCK_LOCALROUTE, valbool);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700725 break;
726 case SO_BROADCAST:
727 sock_valbool_flag(sk, SOCK_BROADCAST, valbool);
728 break;
729 case SO_SNDBUF:
730 /* Don't error on this BSD doesn't and if you think
Eric Dumazet82981932012-04-26 20:07:59 +0000731 * about it this is right. Otherwise apps have to
732 * play 'guess the biggest size' games. RCVBUF/SNDBUF
733 * are treated in BSD as hints
734 */
735 val = min_t(u32, val, sysctl_wmem_max);
Patrick McHardyb0573de2005-08-09 19:30:51 -0700736set_sndbuf:
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700737 sk->sk_userlocks |= SOCK_SNDBUF_LOCK;
Eric Dumazet82981932012-04-26 20:07:59 +0000738 sk->sk_sndbuf = max_t(u32, val * 2, SOCK_MIN_SNDBUF);
739 /* Wake up sending tasks if we upped the value. */
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700740 sk->sk_write_space(sk);
741 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700743 case SO_SNDBUFFORCE:
744 if (!capable(CAP_NET_ADMIN)) {
745 ret = -EPERM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 break;
747 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700748 goto set_sndbuf;
749
750 case SO_RCVBUF:
751 /* Don't error on this BSD doesn't and if you think
Eric Dumazet82981932012-04-26 20:07:59 +0000752 * about it this is right. Otherwise apps have to
753 * play 'guess the biggest size' games. RCVBUF/SNDBUF
754 * are treated in BSD as hints
755 */
756 val = min_t(u32, val, sysctl_rmem_max);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700757set_rcvbuf:
758 sk->sk_userlocks |= SOCK_RCVBUF_LOCK;
759 /*
760 * We double it on the way in to account for
761 * "struct sk_buff" etc. overhead. Applications
762 * assume that the SO_RCVBUF setting they make will
763 * allow that much actual data to be received on that
764 * socket.
765 *
766 * Applications are unaware that "struct sk_buff" and
767 * other overheads allocate from the receive buffer
768 * during socket buffer allocation.
769 *
770 * And after considering the possible alternatives,
771 * returning the value we actually used in getsockopt
772 * is the most desirable behavior.
773 */
Eric Dumazet82981932012-04-26 20:07:59 +0000774 sk->sk_rcvbuf = max_t(u32, val * 2, SOCK_MIN_RCVBUF);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700775 break;
776
777 case SO_RCVBUFFORCE:
778 if (!capable(CAP_NET_ADMIN)) {
779 ret = -EPERM;
780 break;
781 }
782 goto set_rcvbuf;
783
784 case SO_KEEPALIVE:
785#ifdef CONFIG_INET
Eric Dumazet3e109862012-09-24 07:00:11 +0000786 if (sk->sk_protocol == IPPROTO_TCP &&
787 sk->sk_type == SOCK_STREAM)
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700788 tcp_set_keepalive(sk, valbool);
789#endif
790 sock_valbool_flag(sk, SOCK_KEEPOPEN, valbool);
791 break;
792
793 case SO_OOBINLINE:
794 sock_valbool_flag(sk, SOCK_URGINLINE, valbool);
795 break;
796
797 case SO_NO_CHECK:
Tom Herbert28448b82014-05-23 08:47:19 -0700798 sk->sk_no_check_tx = valbool;
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700799 break;
800
801 case SO_PRIORITY:
Eric W. Biederman5e1fccc2012-11-16 03:03:04 +0000802 if ((val >= 0 && val <= 6) ||
803 ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700804 sk->sk_priority = val;
805 else
806 ret = -EPERM;
807 break;
808
809 case SO_LINGER:
810 if (optlen < sizeof(ling)) {
811 ret = -EINVAL; /* 1003.1g */
812 break;
813 }
Eric Dumazet2a915252009-05-27 11:30:05 +0000814 if (copy_from_user(&ling, optval, sizeof(ling))) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700815 ret = -EFAULT;
816 break;
817 }
818 if (!ling.l_onoff)
819 sock_reset_flag(sk, SOCK_LINGER);
820 else {
821#if (BITS_PER_LONG == 32)
822 if ((unsigned int)ling.l_linger >= MAX_SCHEDULE_TIMEOUT/HZ)
823 sk->sk_lingertime = MAX_SCHEDULE_TIMEOUT;
824 else
825#endif
826 sk->sk_lingertime = (unsigned int)ling.l_linger * HZ;
827 sock_set_flag(sk, SOCK_LINGER);
828 }
829 break;
830
831 case SO_BSDCOMPAT:
832 sock_warn_obsolete_bsdism("setsockopt");
833 break;
834
835 case SO_PASSCRED:
836 if (valbool)
837 set_bit(SOCK_PASSCRED, &sock->flags);
838 else
839 clear_bit(SOCK_PASSCRED, &sock->flags);
840 break;
841
842 case SO_TIMESTAMP:
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700843 case SO_TIMESTAMPNS:
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700844 if (valbool) {
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700845 if (optname == SO_TIMESTAMP)
846 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
847 else
848 sock_set_flag(sk, SOCK_RCVTSTAMPNS);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700849 sock_set_flag(sk, SOCK_RCVTSTAMP);
Patrick Ohly20d49472009-02-12 05:03:38 +0000850 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700851 } else {
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700852 sock_reset_flag(sk, SOCK_RCVTSTAMP);
Eric Dumazet92f37fd2007-03-25 22:14:49 -0700853 sock_reset_flag(sk, SOCK_RCVTSTAMPNS);
854 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700855 break;
856
Patrick Ohly20d49472009-02-12 05:03:38 +0000857 case SO_TIMESTAMPING:
858 if (val & ~SOF_TIMESTAMPING_MASK) {
Rémi Denis-Courmontf249fb72009-07-20 00:47:04 +0000859 ret = -EINVAL;
Patrick Ohly20d49472009-02-12 05:03:38 +0000860 break;
861 }
Willem de Bruijnb245be12015-01-30 13:29:32 -0500862
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400863 if (val & SOF_TIMESTAMPING_OPT_ID &&
Willem de Bruijn4ed2d762014-08-04 22:11:49 -0400864 !(sk->sk_tsflags & SOF_TIMESTAMPING_OPT_ID)) {
865 if (sk->sk_protocol == IPPROTO_TCP) {
866 if (sk->sk_state != TCP_ESTABLISHED) {
867 ret = -EINVAL;
868 break;
869 }
870 sk->sk_tskey = tcp_sk(sk)->snd_una;
871 } else {
872 sk->sk_tskey = 0;
873 }
874 }
Willem de Bruijnb9f40e22014-08-04 22:11:46 -0400875 sk->sk_tsflags = val;
Patrick Ohly20d49472009-02-12 05:03:38 +0000876 if (val & SOF_TIMESTAMPING_RX_SOFTWARE)
877 sock_enable_timestamp(sk,
878 SOCK_TIMESTAMPING_RX_SOFTWARE);
879 else
880 sock_disable_timestamp(sk,
Eric Dumazet08e29af2011-11-28 12:04:18 +0000881 (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE));
Patrick Ohly20d49472009-02-12 05:03:38 +0000882 break;
883
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700884 case SO_RCVLOWAT:
885 if (val < 0)
886 val = INT_MAX;
887 sk->sk_rcvlowat = val ? : 1;
888 break;
889
890 case SO_RCVTIMEO:
891 ret = sock_set_timeout(&sk->sk_rcvtimeo, optval, optlen);
892 break;
893
894 case SO_SNDTIMEO:
895 ret = sock_set_timeout(&sk->sk_sndtimeo, optval, optlen);
896 break;
897
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700898 case SO_ATTACH_FILTER:
899 ret = -EINVAL;
900 if (optlen == sizeof(struct sock_fprog)) {
901 struct sock_fprog fprog;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700903 ret = -EFAULT;
904 if (copy_from_user(&fprog, optval, sizeof(fprog)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700907 ret = sk_attach_filter(&fprog, sk);
908 }
909 break;
910
Alexei Starovoitov89aa0752014-12-01 15:06:35 -0800911 case SO_ATTACH_BPF:
912 ret = -EINVAL;
913 if (optlen == sizeof(u32)) {
914 u32 ufd;
915
916 ret = -EFAULT;
917 if (copy_from_user(&ufd, optval, sizeof(ufd)))
918 break;
919
920 ret = sk_attach_bpf(ufd, sk);
921 }
922 break;
923
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700924 case SO_DETACH_FILTER:
Pavel Emelyanov55b33322007-10-17 21:21:26 -0700925 ret = sk_detach_filter(sk);
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700926 break;
927
Vincent Bernatd59577b2013-01-16 22:55:49 +0100928 case SO_LOCK_FILTER:
929 if (sock_flag(sk, SOCK_FILTER_LOCKED) && !valbool)
930 ret = -EPERM;
931 else
932 sock_valbool_flag(sk, SOCK_FILTER_LOCKED, valbool);
933 break;
934
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700935 case SO_PASSSEC:
936 if (valbool)
937 set_bit(SOCK_PASSSEC, &sock->flags);
938 else
939 clear_bit(SOCK_PASSSEC, &sock->flags);
940 break;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800941 case SO_MARK:
Eric W. Biederman5e1fccc2012-11-16 03:03:04 +0000942 if (!ns_capable(sock_net(sk)->user_ns, CAP_NET_ADMIN))
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800943 ret = -EPERM;
Eric Dumazet2a915252009-05-27 11:30:05 +0000944 else
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800945 sk->sk_mark = val;
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -0800946 break;
Catherine Zhang877ce7c2006-06-29 12:27:47 -0700947
Neil Horman3b885782009-10-12 13:26:31 -0700948 case SO_RXQ_OVFL:
Johannes Berg8083f0f2011-10-07 03:30:20 +0000949 sock_valbool_flag(sk, SOCK_RXQ_OVFL, valbool);
Neil Horman3b885782009-10-12 13:26:31 -0700950 break;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100951
952 case SO_WIFI_STATUS:
953 sock_valbool_flag(sk, SOCK_WIFI_STATUS, valbool);
954 break;
955
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000956 case SO_PEEK_OFF:
957 if (sock->ops->set_peek_off)
Sasha Levin12663bf2013-12-07 17:26:27 -0500958 ret = sock->ops->set_peek_off(sk, val);
Pavel Emelyanovef64a542012-02-21 07:31:34 +0000959 else
960 ret = -EOPNOTSUPP;
961 break;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000962
963 case SO_NOFCS:
964 sock_valbool_flag(sk, SOCK_NOFCS, valbool);
965 break;
966
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +0000967 case SO_SELECT_ERR_QUEUE:
968 sock_valbool_flag(sk, SOCK_SELECT_ERR_QUEUE, valbool);
969 break;
970
Cong Wange0d10952013-08-01 11:10:25 +0800971#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +0300972 case SO_BUSY_POLL:
Eliezer Tamirdafcc432013-06-14 16:33:57 +0300973 /* allow unprivileged users to decrease the value */
974 if ((val > sk->sk_ll_usec) && !capable(CAP_NET_ADMIN))
975 ret = -EPERM;
976 else {
977 if (val < 0)
978 ret = -EINVAL;
979 else
980 sk->sk_ll_usec = val;
981 }
982 break;
983#endif
Eric Dumazet62748f32013-09-24 08:20:52 -0700984
985 case SO_MAX_PACING_RATE:
986 sk->sk_max_pacing_rate = val;
987 sk->sk_pacing_rate = min(sk->sk_pacing_rate,
988 sk->sk_max_pacing_rate);
989 break;
990
Stephen Hemmingere71a4782007-04-10 20:10:33 -0700991 default:
992 ret = -ENOPROTOOPT;
993 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +0900994 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 release_sock(sk);
996 return ret;
997}
Eric Dumazet2a915252009-05-27 11:30:05 +0000998EXPORT_SYMBOL(sock_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
1000
stephen hemminger8f098982014-01-03 09:17:14 -08001001static void cred_to_ucred(struct pid *pid, const struct cred *cred,
1002 struct ucred *ucred)
Eric W. Biederman3f551f92010-06-13 03:28:59 +00001003{
1004 ucred->pid = pid_vnr(pid);
1005 ucred->uid = ucred->gid = -1;
1006 if (cred) {
1007 struct user_namespace *current_ns = current_user_ns();
1008
Eric W. Biedermanb2e4f542012-05-23 16:39:45 -06001009 ucred->uid = from_kuid_munged(current_ns, cred->euid);
1010 ucred->gid = from_kgid_munged(current_ns, cred->egid);
Eric W. Biederman3f551f92010-06-13 03:28:59 +00001011 }
1012}
1013
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014int sock_getsockopt(struct socket *sock, int level, int optname,
1015 char __user *optval, int __user *optlen)
1016{
1017 struct sock *sk = sock->sk;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001018
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001019 union {
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001020 int val;
1021 struct linger ling;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022 struct timeval tm;
1023 } v;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001024
H Hartley Sweeten4d0392b2010-01-15 01:08:58 -08001025 int lv = sizeof(int);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 int len;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001027
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001028 if (get_user(len, optlen))
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001029 return -EFAULT;
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001030 if (len < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031 return -EINVAL;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001032
Eugene Teo50fee1d2009-02-23 15:38:41 -08001033 memset(&v, 0, sizeof(v));
Clément Lecignedf0bca02009-02-12 16:59:09 -08001034
Eric Dumazet2a915252009-05-27 11:30:05 +00001035 switch (optname) {
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001036 case SO_DEBUG:
1037 v.val = sock_flag(sk, SOCK_DBG);
1038 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001039
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001040 case SO_DONTROUTE:
1041 v.val = sock_flag(sk, SOCK_LOCALROUTE);
1042 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001043
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001044 case SO_BROADCAST:
Eric Dumazet1b23a5d2012-05-16 05:57:07 +00001045 v.val = sock_flag(sk, SOCK_BROADCAST);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001046 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001048 case SO_SNDBUF:
1049 v.val = sk->sk_sndbuf;
1050 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001051
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001052 case SO_RCVBUF:
1053 v.val = sk->sk_rcvbuf;
1054 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001056 case SO_REUSEADDR:
1057 v.val = sk->sk_reuse;
1058 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059
Tom Herbert055dc212013-01-22 09:49:50 +00001060 case SO_REUSEPORT:
1061 v.val = sk->sk_reuseport;
1062 break;
1063
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001064 case SO_KEEPALIVE:
Eric Dumazet1b23a5d2012-05-16 05:57:07 +00001065 v.val = sock_flag(sk, SOCK_KEEPOPEN);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001066 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001067
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001068 case SO_TYPE:
1069 v.val = sk->sk_type;
1070 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Jan Engelhardt49c794e2009-08-04 07:28:28 +00001072 case SO_PROTOCOL:
1073 v.val = sk->sk_protocol;
1074 break;
1075
Jan Engelhardt0d6038e2009-08-04 07:28:29 +00001076 case SO_DOMAIN:
1077 v.val = sk->sk_family;
1078 break;
1079
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001080 case SO_ERROR:
1081 v.val = -sock_error(sk);
Eric Dumazet2a915252009-05-27 11:30:05 +00001082 if (v.val == 0)
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001083 v.val = xchg(&sk->sk_err_soft, 0);
1084 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001086 case SO_OOBINLINE:
Eric Dumazet1b23a5d2012-05-16 05:57:07 +00001087 v.val = sock_flag(sk, SOCK_URGINLINE);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001088 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001089
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001090 case SO_NO_CHECK:
Tom Herbert28448b82014-05-23 08:47:19 -07001091 v.val = sk->sk_no_check_tx;
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001092 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001094 case SO_PRIORITY:
1095 v.val = sk->sk_priority;
1096 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001097
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001098 case SO_LINGER:
1099 lv = sizeof(v.ling);
Eric Dumazet1b23a5d2012-05-16 05:57:07 +00001100 v.ling.l_onoff = sock_flag(sk, SOCK_LINGER);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001101 v.ling.l_linger = sk->sk_lingertime / HZ;
1102 break;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001103
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001104 case SO_BSDCOMPAT:
1105 sock_warn_obsolete_bsdism("getsockopt");
1106 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001108 case SO_TIMESTAMP:
Eric Dumazet92f37fd2007-03-25 22:14:49 -07001109 v.val = sock_flag(sk, SOCK_RCVTSTAMP) &&
1110 !sock_flag(sk, SOCK_RCVTSTAMPNS);
1111 break;
1112
1113 case SO_TIMESTAMPNS:
1114 v.val = sock_flag(sk, SOCK_RCVTSTAMPNS);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001115 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116
Patrick Ohly20d49472009-02-12 05:03:38 +00001117 case SO_TIMESTAMPING:
Willem de Bruijnb9f40e22014-08-04 22:11:46 -04001118 v.val = sk->sk_tsflags;
Patrick Ohly20d49472009-02-12 05:03:38 +00001119 break;
1120
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001121 case SO_RCVTIMEO:
Eric Dumazet2a915252009-05-27 11:30:05 +00001122 lv = sizeof(struct timeval);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001123 if (sk->sk_rcvtimeo == MAX_SCHEDULE_TIMEOUT) {
1124 v.tm.tv_sec = 0;
1125 v.tm.tv_usec = 0;
1126 } else {
1127 v.tm.tv_sec = sk->sk_rcvtimeo / HZ;
1128 v.tm.tv_usec = ((sk->sk_rcvtimeo % HZ) * 1000000) / HZ;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001130 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001132 case SO_SNDTIMEO:
Eric Dumazet2a915252009-05-27 11:30:05 +00001133 lv = sizeof(struct timeval);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001134 if (sk->sk_sndtimeo == MAX_SCHEDULE_TIMEOUT) {
1135 v.tm.tv_sec = 0;
1136 v.tm.tv_usec = 0;
1137 } else {
1138 v.tm.tv_sec = sk->sk_sndtimeo / HZ;
1139 v.tm.tv_usec = ((sk->sk_sndtimeo % HZ) * 1000000) / HZ;
1140 }
1141 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001143 case SO_RCVLOWAT:
1144 v.val = sk->sk_rcvlowat;
1145 break;
Catherine Zhang877ce7c2006-06-29 12:27:47 -07001146
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001147 case SO_SNDLOWAT:
Eric Dumazet2a915252009-05-27 11:30:05 +00001148 v.val = 1;
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001149 break;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001151 case SO_PASSCRED:
Eric Dumazet82981932012-04-26 20:07:59 +00001152 v.val = !!test_bit(SOCK_PASSCRED, &sock->flags);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001153 break;
1154
1155 case SO_PEERCRED:
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001156 {
1157 struct ucred peercred;
1158 if (len > sizeof(peercred))
1159 len = sizeof(peercred);
1160 cred_to_ucred(sk->sk_peer_pid, sk->sk_peer_cred, &peercred);
1161 if (copy_to_user(optval, &peercred, len))
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001162 return -EFAULT;
1163 goto lenout;
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001164 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001165
1166 case SO_PEERNAME:
1167 {
1168 char address[128];
1169
1170 if (sock->ops->getname(sock, (struct sockaddr *)address, &lv, 2))
1171 return -ENOTCONN;
1172 if (lv < len)
1173 return -EINVAL;
1174 if (copy_to_user(optval, address, len))
1175 return -EFAULT;
1176 goto lenout;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 }
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001178
1179 /* Dubious BSD thing... Probably nobody even uses it, but
1180 * the UNIX standard wants it for whatever reason... -DaveM
1181 */
1182 case SO_ACCEPTCONN:
1183 v.val = sk->sk_state == TCP_LISTEN;
1184 break;
1185
1186 case SO_PASSSEC:
Eric Dumazet82981932012-04-26 20:07:59 +00001187 v.val = !!test_bit(SOCK_PASSSEC, &sock->flags);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001188 break;
1189
1190 case SO_PEERSEC:
1191 return security_socket_getpeersec_stream(sock, optval, optlen, len);
1192
Laszlo Attila Toth4a19ec52008-01-30 19:08:16 -08001193 case SO_MARK:
1194 v.val = sk->sk_mark;
1195 break;
1196
Neil Horman3b885782009-10-12 13:26:31 -07001197 case SO_RXQ_OVFL:
Eric Dumazet1b23a5d2012-05-16 05:57:07 +00001198 v.val = sock_flag(sk, SOCK_RXQ_OVFL);
Neil Horman3b885782009-10-12 13:26:31 -07001199 break;
1200
Johannes Berg6e3e9392011-11-09 10:15:42 +01001201 case SO_WIFI_STATUS:
Eric Dumazet1b23a5d2012-05-16 05:57:07 +00001202 v.val = sock_flag(sk, SOCK_WIFI_STATUS);
Johannes Berg6e3e9392011-11-09 10:15:42 +01001203 break;
1204
Pavel Emelyanovef64a542012-02-21 07:31:34 +00001205 case SO_PEEK_OFF:
1206 if (!sock->ops->set_peek_off)
1207 return -EOPNOTSUPP;
1208
1209 v.val = sk->sk_peek_off;
1210 break;
David S. Millerbc2f7992012-02-24 14:48:34 -05001211 case SO_NOFCS:
Eric Dumazet1b23a5d2012-05-16 05:57:07 +00001212 v.val = sock_flag(sk, SOCK_NOFCS);
David S. Millerbc2f7992012-02-24 14:48:34 -05001213 break;
Brian Haleyc91f6df2012-11-26 05:21:08 +00001214
Pavel Emelyanovf7b86bf2012-10-18 23:55:56 +00001215 case SO_BINDTODEVICE:
Brian Haleyc91f6df2012-11-26 05:21:08 +00001216 return sock_getbindtodevice(sk, optval, optlen, len);
1217
Pavel Emelyanova8fc9272012-11-01 02:01:48 +00001218 case SO_GET_FILTER:
1219 len = sk_get_filter(sk, (struct sock_filter __user *)optval, len);
1220 if (len < 0)
1221 return len;
1222
1223 goto lenout;
Brian Haleyc91f6df2012-11-26 05:21:08 +00001224
Vincent Bernatd59577b2013-01-16 22:55:49 +01001225 case SO_LOCK_FILTER:
1226 v.val = sock_flag(sk, SOCK_FILTER_LOCKED);
1227 break;
1228
Michal Sekletarea02f942014-01-17 17:09:45 +01001229 case SO_BPF_EXTENSIONS:
1230 v.val = bpf_tell_extensions();
1231 break;
1232
Keller, Jacob E7d4c04f2013-03-28 11:19:25 +00001233 case SO_SELECT_ERR_QUEUE:
1234 v.val = sock_flag(sk, SOCK_SELECT_ERR_QUEUE);
1235 break;
1236
Cong Wange0d10952013-08-01 11:10:25 +08001237#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir64b0dc52013-07-10 17:13:36 +03001238 case SO_BUSY_POLL:
Eliezer Tamirdafcc432013-06-14 16:33:57 +03001239 v.val = sk->sk_ll_usec;
1240 break;
1241#endif
1242
Eric Dumazet62748f32013-09-24 08:20:52 -07001243 case SO_MAX_PACING_RATE:
1244 v.val = sk->sk_max_pacing_rate;
1245 break;
1246
Eric Dumazet2c8c56e2014-11-11 05:54:28 -08001247 case SO_INCOMING_CPU:
1248 v.val = sk->sk_incoming_cpu;
1249 break;
1250
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001251 default:
YOSHIFUJI Hideaki/吉藤英明443b5992015-03-23 18:04:13 +09001252 /* We implement the SO_SNDLOWAT etc to not be settable
1253 * (1003.1g 7).
1254 */
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001255 return -ENOPROTOOPT;
1256 }
1257
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 if (len > lv)
1259 len = lv;
1260 if (copy_to_user(optval, &v, len))
1261 return -EFAULT;
1262lenout:
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001263 if (put_user(len, optlen))
1264 return -EFAULT;
1265 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266}
1267
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001268/*
1269 * Initialize an sk_lock.
1270 *
1271 * (We also register the sk_lock with the lock validator.)
1272 */
Dave Jonesb6f99a22007-03-22 12:27:49 -07001273static inline void sock_lock_init(struct sock *sk)
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001274{
Peter Zijlstraed075362006-12-06 20:35:24 -08001275 sock_lock_init_class_and_name(sk,
1276 af_family_slock_key_strings[sk->sk_family],
1277 af_family_slock_keys + sk->sk_family,
1278 af_family_key_strings[sk->sk_family],
1279 af_family_keys + sk->sk_family);
Ingo Molnara5b5bb92006-07-03 00:25:35 -07001280}
1281
Eric Dumazet4dc6dc72009-07-15 23:13:10 +00001282/*
1283 * Copy all fields from osk to nsk but nsk->sk_refcnt must not change yet,
1284 * even temporarly, because of RCU lookups. sk_node should also be left as is.
Eric Dumazet68835ab2010-11-30 19:04:07 +00001285 * We must not copy fields between sk_dontcopy_begin and sk_dontcopy_end
Eric Dumazet4dc6dc72009-07-15 23:13:10 +00001286 */
Pavel Emelyanovf1a6c4d2007-11-01 00:29:45 -07001287static void sock_copy(struct sock *nsk, const struct sock *osk)
1288{
1289#ifdef CONFIG_SECURITY_NETWORK
1290 void *sptr = nsk->sk_security;
1291#endif
Eric Dumazet68835ab2010-11-30 19:04:07 +00001292 memcpy(nsk, osk, offsetof(struct sock, sk_dontcopy_begin));
1293
1294 memcpy(&nsk->sk_dontcopy_end, &osk->sk_dontcopy_end,
1295 osk->sk_prot->obj_size - offsetof(struct sock, sk_dontcopy_end));
1296
Pavel Emelyanovf1a6c4d2007-11-01 00:29:45 -07001297#ifdef CONFIG_SECURITY_NETWORK
1298 nsk->sk_security = sptr;
1299 security_sk_clone(osk, nsk);
1300#endif
1301}
1302
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001303void sk_prot_clear_portaddr_nulls(struct sock *sk, int size)
1304{
1305 unsigned long nulls1, nulls2;
1306
1307 nulls1 = offsetof(struct sock, __sk_common.skc_node.next);
1308 nulls2 = offsetof(struct sock, __sk_common.skc_portaddr_node.next);
1309 if (nulls1 > nulls2)
1310 swap(nulls1, nulls2);
1311
1312 if (nulls1 != 0)
1313 memset((char *)sk, 0, nulls1);
1314 memset((char *)sk + nulls1 + sizeof(void *), 0,
1315 nulls2 - nulls1 - sizeof(void *));
1316 memset((char *)sk + nulls2 + sizeof(void *), 0,
1317 size - nulls2 - sizeof(void *));
1318}
1319EXPORT_SYMBOL(sk_prot_clear_portaddr_nulls);
1320
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001321static struct sock *sk_prot_alloc(struct proto *prot, gfp_t priority,
1322 int family)
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001323{
1324 struct sock *sk;
1325 struct kmem_cache *slab;
1326
1327 slab = prot->slab;
Eric Dumazete912b112009-07-08 19:36:05 +00001328 if (slab != NULL) {
1329 sk = kmem_cache_alloc(slab, priority & ~__GFP_ZERO);
1330 if (!sk)
1331 return sk;
1332 if (priority & __GFP_ZERO) {
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001333 if (prot->clear_sk)
1334 prot->clear_sk(sk, prot->obj_size);
1335 else
1336 sk_prot_clear_nulls(sk, prot->obj_size);
Eric Dumazete912b112009-07-08 19:36:05 +00001337 }
Octavian Purdilafcbdf092010-12-16 14:26:56 -08001338 } else
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001339 sk = kmalloc(prot->obj_size, priority);
1340
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001341 if (sk != NULL) {
Vegard Nossuma98b65a2009-02-26 14:46:57 +01001342 kmemcheck_annotate_bitfield(sk, flags);
1343
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001344 if (security_sk_alloc(sk, family, priority))
1345 goto out_free;
1346
1347 if (!try_module_get(prot->owner))
1348 goto out_free_sec;
Krishna Kumare022f0b2009-10-19 23:46:20 +00001349 sk_tx_queue_clear(sk);
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001350 }
1351
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001352 return sk;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001353
1354out_free_sec:
1355 security_sk_free(sk);
1356out_free:
1357 if (slab != NULL)
1358 kmem_cache_free(slab, sk);
1359 else
1360 kfree(sk);
1361 return NULL;
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001362}
1363
1364static void sk_prot_free(struct proto *prot, struct sock *sk)
1365{
1366 struct kmem_cache *slab;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001367 struct module *owner;
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001368
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001369 owner = prot->owner;
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001370 slab = prot->slab;
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001371
1372 security_sk_free(sk);
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001373 if (slab != NULL)
1374 kmem_cache_free(slab, sk);
1375 else
1376 kfree(sk);
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001377 module_put(owner);
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001378}
1379
Daniel Borkmann86f85152013-12-29 17:27:11 +01001380#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Zefan Li6ffd4642013-04-08 20:03:47 +00001381void sock_update_netprioidx(struct sock *sk)
Neil Horman5bc14212011-11-22 05:10:51 +00001382{
Neil Horman5bc14212011-11-22 05:10:51 +00001383 if (in_interrupt())
1384 return;
Neil Horman2b73bc62012-02-10 05:43:38 +00001385
Zefan Li6ffd4642013-04-08 20:03:47 +00001386 sk->sk_cgrp_prioidx = task_netprioidx(current);
Neil Horman5bc14212011-11-22 05:10:51 +00001387}
1388EXPORT_SYMBOL_GPL(sock_update_netprioidx);
Herbert Xuf8451722010-05-24 00:12:34 -07001389#endif
1390
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391/**
1392 * sk_alloc - All socket objects are allocated here
Randy Dunlapc4ea43c2007-10-12 21:17:49 -07001393 * @net: the applicable net namespace
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001394 * @family: protocol family
1395 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1396 * @prot: struct proto associated with this new sock instance
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05001397 * @kern: is this to be a kernel socket?
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 */
Eric W. Biederman1b8d7ae2007-10-08 23:24:22 -07001399struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
Eric W. Biederman11aa9c22015-05-08 21:09:13 -05001400 struct proto *prot, int kern)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401{
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001402 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403
Pavel Emelyanov154adbc2007-11-01 00:38:43 -07001404 sk = sk_prot_alloc(prot, priority | __GFP_ZERO, family);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405 if (sk) {
Pavel Emelyanov154adbc2007-11-01 00:38:43 -07001406 sk->sk_family = family;
1407 /*
1408 * See comment in struct sock definition to understand
1409 * why we need sk_prot_creator -acme
1410 */
1411 sk->sk_prot = sk->sk_prot_creator = prot;
1412 sock_lock_init(sk);
Eric W. Biederman26abe142015-05-08 21:10:31 -05001413 sk->sk_net_refcnt = kern ? 0 : 1;
1414 if (likely(sk->sk_net_refcnt))
1415 get_net(net);
1416 sock_net_set(sk, net);
Jarek Poplawskid66ee052009-08-30 23:15:36 +00001417 atomic_set(&sk->sk_wmem_alloc, 1);
Herbert Xuf8451722010-05-24 00:12:34 -07001418
Zefan Li211d2f972013-04-08 20:03:35 +00001419 sock_update_classid(sk);
Zefan Li6ffd4642013-04-08 20:03:47 +00001420 sock_update_netprioidx(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 }
Frank Filza79af592005-09-27 15:23:38 -07001422
Pavel Emelyanov2e4afe72007-11-01 00:36:26 -07001423 return sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424}
Eric Dumazet2a915252009-05-27 11:30:05 +00001425EXPORT_SYMBOL(sk_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426
Craig Gallekeb4cb002015-06-15 11:26:18 -04001427void sk_destruct(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428{
1429 struct sk_filter *filter;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430
1431 if (sk->sk_destruct)
1432 sk->sk_destruct(sk);
1433
Paul E. McKenneya898def2010-02-22 17:04:49 -08001434 filter = rcu_dereference_check(sk->sk_filter,
1435 atomic_read(&sk->sk_wmem_alloc) == 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436 if (filter) {
Pavel Emelyanov309dd5f2007-10-17 21:21:51 -07001437 sk_filter_uncharge(sk, filter);
Stephen Hemmingera9b3cd72011-08-01 16:19:00 +00001438 RCU_INIT_POINTER(sk->sk_filter, NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 }
1440
Eric Dumazet08e29af2011-11-28 12:04:18 +00001441 sock_disable_timestamp(sk, SK_FLAGS_TIMESTAMP);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442
1443 if (atomic_read(&sk->sk_omem_alloc))
Joe Perchese005d192012-05-16 19:58:40 +00001444 pr_debug("%s: optmem leakage (%d bytes) detected\n",
1445 __func__, atomic_read(&sk->sk_omem_alloc));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446
Eric W. Biederman109f6e32010-06-13 03:30:14 +00001447 if (sk->sk_peer_cred)
1448 put_cred(sk->sk_peer_cred);
1449 put_pid(sk->sk_peer_pid);
Eric W. Biederman26abe142015-05-08 21:10:31 -05001450 if (likely(sk->sk_net_refcnt))
1451 put_net(sock_net(sk));
Pavel Emelyanovc308c1b22007-11-01 00:33:50 -07001452 sk_prot_free(sk->sk_prot_creator, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453}
Eric Dumazet2b85a342009-06-11 02:55:43 -07001454
Craig Gallekeb4cb002015-06-15 11:26:18 -04001455static void __sk_free(struct sock *sk)
1456{
Craig Gallekb9226222015-06-30 12:49:32 -04001457 if (unlikely(sock_diag_has_destroy_listeners(sk) && sk->sk_net_refcnt))
Craig Gallekeb4cb002015-06-15 11:26:18 -04001458 sock_diag_broadcast_destroy(sk);
1459 else
1460 sk_destruct(sk);
1461}
1462
Eric Dumazet2b85a342009-06-11 02:55:43 -07001463void sk_free(struct sock *sk)
1464{
1465 /*
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001466 * We subtract one from sk_wmem_alloc and can know if
Eric Dumazet2b85a342009-06-11 02:55:43 -07001467 * some packets are still in some tx queue.
1468 * If not null, sock_wfree() will call __sk_free(sk) later
1469 */
1470 if (atomic_dec_and_test(&sk->sk_wmem_alloc))
1471 __sk_free(sk);
1472}
Eric Dumazet2a915252009-05-27 11:30:05 +00001473EXPORT_SYMBOL(sk_free);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474
Stephen Rothwell475f1b52012-01-09 16:33:16 +11001475static void sk_update_clone(const struct sock *sk, struct sock *newsk)
1476{
1477 if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
1478 sock_update_memcg(newsk);
1479}
1480
Eric Dumazete56c57d2011-11-08 17:07:07 -05001481/**
1482 * sk_clone_lock - clone a socket, and lock its clone
1483 * @sk: the socket to clone
1484 * @priority: for allocation (%GFP_KERNEL, %GFP_ATOMIC, etc)
1485 *
1486 * Caller must unlock socket even in error path (bh_unlock_sock(newsk))
1487 */
1488struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority)
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001489{
Pavel Emelyanov8fd1d172007-11-01 00:37:32 -07001490 struct sock *newsk;
Alexei Starovoitov278571b2014-07-30 20:34:12 -07001491 bool is_charged = true;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001492
Pavel Emelyanov8fd1d172007-11-01 00:37:32 -07001493 newsk = sk_prot_alloc(sk->sk_prot, priority, sk->sk_family);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001494 if (newsk != NULL) {
1495 struct sk_filter *filter;
1496
Venkat Yekkirala892c1412006-08-04 23:08:56 -07001497 sock_copy(newsk, sk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001498
1499 /* SANITY */
Sowmini Varadhan8a681732015-07-30 15:50:36 +02001500 if (likely(newsk->sk_net_refcnt))
1501 get_net(sock_net(newsk));
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001502 sk_node_init(&newsk->sk_node);
1503 sock_lock_init(newsk);
1504 bh_lock_sock(newsk);
Eric Dumazetfa438cc2007-03-04 16:05:44 -08001505 newsk->sk_backlog.head = newsk->sk_backlog.tail = NULL;
Zhu Yi8eae9392010-03-04 18:01:40 +00001506 newsk->sk_backlog.len = 0;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001507
1508 atomic_set(&newsk->sk_rmem_alloc, 0);
Eric Dumazet2b85a342009-06-11 02:55:43 -07001509 /*
1510 * sk_wmem_alloc set to one (see sk_free() and sock_wfree())
1511 */
1512 atomic_set(&newsk->sk_wmem_alloc, 1);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001513 atomic_set(&newsk->sk_omem_alloc, 0);
1514 skb_queue_head_init(&newsk->sk_receive_queue);
1515 skb_queue_head_init(&newsk->sk_write_queue);
1516
Eric Dumazetb6c67122010-04-08 23:03:29 +00001517 spin_lock_init(&newsk->sk_dst_lock);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001518 rwlock_init(&newsk->sk_callback_lock);
Peter Zijlstra443aef02007-07-19 01:49:00 -07001519 lockdep_set_class_and_name(&newsk->sk_callback_lock,
1520 af_callback_keys + newsk->sk_family,
1521 af_family_clock_key_strings[newsk->sk_family]);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001522
1523 newsk->sk_dst_cache = NULL;
1524 newsk->sk_wmem_queued = 0;
1525 newsk->sk_forward_alloc = 0;
1526 newsk->sk_send_head = NULL;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001527 newsk->sk_userlocks = sk->sk_userlocks & ~SOCK_BINDPORT_LOCK;
1528
1529 sock_reset_flag(newsk, SOCK_DONE);
1530 skb_queue_head_init(&newsk->sk_error_queue);
1531
Eric Dumazet0d7da9d2010-10-25 03:47:05 +00001532 filter = rcu_dereference_protected(newsk->sk_filter, 1);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001533 if (filter != NULL)
Alexei Starovoitov278571b2014-07-30 20:34:12 -07001534 /* though it's an empty new sock, the charging may fail
1535 * if sysctl_optmem_max was changed between creation of
1536 * original socket and cloning
1537 */
1538 is_charged = sk_filter_charge(newsk, filter);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001539
Alexei Starovoitov278571b2014-07-30 20:34:12 -07001540 if (unlikely(!is_charged || xfrm_sk_clone_policy(newsk))) {
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001541 /* It is still raw copy of parent, so invalidate
1542 * destructor and make plain sk_free() */
1543 newsk->sk_destruct = NULL;
Thomas Gleixnerb0691c82011-10-25 02:30:50 +00001544 bh_unlock_sock(newsk);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001545 sk_free(newsk);
1546 newsk = NULL;
1547 goto out;
1548 }
1549
1550 newsk->sk_err = 0;
1551 newsk->sk_priority = 0;
Eric Dumazet2c8c56e2014-11-11 05:54:28 -08001552 newsk->sk_incoming_cpu = raw_smp_processor_id();
Eric Dumazet33cf7c92015-03-11 18:53:14 -07001553 atomic64_set(&newsk->sk_cookie, 0);
Eric Dumazet4dc6dc72009-07-15 23:13:10 +00001554 /*
1555 * Before updating sk_refcnt, we must commit prior changes to memory
1556 * (Documentation/RCU/rculist_nulls.txt for details)
1557 */
1558 smp_wmb();
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001559 atomic_set(&newsk->sk_refcnt, 2);
1560
1561 /*
1562 * Increment the counter in the same struct proto as the master
1563 * sock (sk_refcnt_debug_inc uses newsk->sk_prot->socks, that
1564 * is the same as sk->sk_prot->socks, as this field was copied
1565 * with memcpy).
1566 *
1567 * This _changes_ the previous behaviour, where
1568 * tcp_create_openreq_child always was incrementing the
1569 * equivalent to tcp_prot->socks (inet_sock_nr), so this have
1570 * to be taken into account in all callers. -acme
1571 */
1572 sk_refcnt_debug_inc(newsk);
David S. Miller972692e2008-06-17 22:41:38 -07001573 sk_set_socket(newsk, NULL);
Eric Dumazet43815482010-04-29 11:01:49 +00001574 newsk->sk_wq = NULL;
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001575
Glauber Costaf3f511e2012-01-05 20:16:39 +00001576 sk_update_clone(sk, newsk);
1577
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001578 if (newsk->sk_prot->sockets_allocated)
Glauber Costa180d8cd2011-12-11 21:47:02 +00001579 sk_sockets_allocated_inc(newsk);
Octavian Purdila704da5602010-01-08 00:00:09 -08001580
Eric Dumazet08e29af2011-11-28 12:04:18 +00001581 if (newsk->sk_flags & SK_FLAGS_TIMESTAMP)
Octavian Purdila704da5602010-01-08 00:00:09 -08001582 net_enable_timestamp();
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001583 }
1584out:
1585 return newsk;
1586}
Eric Dumazete56c57d2011-11-08 17:07:07 -05001587EXPORT_SYMBOL_GPL(sk_clone_lock);
Arnaldo Carvalho de Melo87d11ce2005-08-09 20:10:12 -07001588
Andi Kleen99580892007-04-20 17:12:43 -07001589void sk_setup_caps(struct sock *sk, struct dst_entry *dst)
1590{
Eric Dumazetd6a4e262015-05-26 08:55:28 -07001591 u32 max_segs = 1;
1592
Andi Kleen99580892007-04-20 17:12:43 -07001593 __sk_dst_set(sk, dst);
1594 sk->sk_route_caps = dst->dev->features;
1595 if (sk->sk_route_caps & NETIF_F_GSO)
Herbert Xu4fcd6b92007-05-31 22:15:50 -07001596 sk->sk_route_caps |= NETIF_F_GSO_SOFTWARE;
Eric Dumazeta4654192010-05-16 00:36:33 -07001597 sk->sk_route_caps &= ~sk->sk_route_nocaps;
Andi Kleen99580892007-04-20 17:12:43 -07001598 if (sk_can_gso(sk)) {
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001599 if (dst->header_len) {
Andi Kleen99580892007-04-20 17:12:43 -07001600 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001601 } else {
Andi Kleen99580892007-04-20 17:12:43 -07001602 sk->sk_route_caps |= NETIF_F_SG | NETIF_F_HW_CSUM;
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001603 sk->sk_gso_max_size = dst->dev->gso_max_size;
Eric Dumazetd6a4e262015-05-26 08:55:28 -07001604 max_segs = max_t(u32, dst->dev->gso_max_segs, 1);
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001605 }
Andi Kleen99580892007-04-20 17:12:43 -07001606 }
Eric Dumazetd6a4e262015-05-26 08:55:28 -07001607 sk->sk_gso_max_segs = max_segs;
Andi Kleen99580892007-04-20 17:12:43 -07001608}
1609EXPORT_SYMBOL_GPL(sk_setup_caps);
1610
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611/*
1612 * Simple resource managers for sockets.
1613 */
1614
1615
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001616/*
1617 * Write buffer destructor automatically called from kfree_skb.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 */
1619void sock_wfree(struct sk_buff *skb)
1620{
1621 struct sock *sk = skb->sk;
Eric Dumazetd99927f2009-09-24 10:49:24 +00001622 unsigned int len = skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623
Eric Dumazetd99927f2009-09-24 10:49:24 +00001624 if (!sock_flag(sk, SOCK_USE_WRITE_QUEUE)) {
1625 /*
1626 * Keep a reference on sk_wmem_alloc, this will be released
1627 * after sk_write_space() call
1628 */
1629 atomic_sub(len - 1, &sk->sk_wmem_alloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 sk->sk_write_space(sk);
Eric Dumazetd99927f2009-09-24 10:49:24 +00001631 len = 1;
1632 }
Eric Dumazet2b85a342009-06-11 02:55:43 -07001633 /*
Eric Dumazetd99927f2009-09-24 10:49:24 +00001634 * if sk_wmem_alloc reaches 0, we must finish what sk_free()
1635 * could not do because of in-flight packets
Eric Dumazet2b85a342009-06-11 02:55:43 -07001636 */
Eric Dumazetd99927f2009-09-24 10:49:24 +00001637 if (atomic_sub_and_test(len, &sk->sk_wmem_alloc))
Eric Dumazet2b85a342009-06-11 02:55:43 -07001638 __sk_free(sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639}
Eric Dumazet2a915252009-05-27 11:30:05 +00001640EXPORT_SYMBOL(sock_wfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641
Eric Dumazetf2f872f2013-07-30 17:55:08 -07001642void skb_orphan_partial(struct sk_buff *skb)
1643{
1644 /* TCP stack sets skb->ooo_okay based on sk_wmem_alloc,
1645 * so we do not completely orphan skb, but transfert all
1646 * accounted bytes but one, to avoid unexpected reorders.
1647 */
1648 if (skb->destructor == sock_wfree
1649#ifdef CONFIG_INET
1650 || skb->destructor == tcp_wfree
1651#endif
1652 ) {
1653 atomic_sub(skb->truesize - 1, &skb->sk->sk_wmem_alloc);
1654 skb->truesize = 1;
1655 } else {
1656 skb_orphan(skb);
1657 }
1658}
1659EXPORT_SYMBOL(skb_orphan_partial);
1660
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001661/*
1662 * Read buffer destructor automatically called from kfree_skb.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663 */
1664void sock_rfree(struct sk_buff *skb)
1665{
1666 struct sock *sk = skb->sk;
Eric Dumazetd361fd52010-07-10 22:45:17 +00001667 unsigned int len = skb->truesize;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668
Eric Dumazetd361fd52010-07-10 22:45:17 +00001669 atomic_sub(len, &sk->sk_rmem_alloc);
1670 sk_mem_uncharge(sk, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671}
Eric Dumazet2a915252009-05-27 11:30:05 +00001672EXPORT_SYMBOL(sock_rfree);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673
Oliver Hartkopp7768eed2015-03-10 19:03:46 +01001674/*
1675 * Buffer destructor for skbs that are not used directly in read or write
1676 * path, e.g. for error handler skbs. Automatically called from kfree_skb.
1677 */
Alexander Duyck62bccb82014-09-04 13:31:35 -04001678void sock_efree(struct sk_buff *skb)
1679{
1680 sock_put(skb->sk);
1681}
1682EXPORT_SYMBOL(sock_efree);
1683
Eric W. Biederman976d02012012-05-23 17:16:53 -06001684kuid_t sock_i_uid(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685{
Eric W. Biederman976d02012012-05-23 17:16:53 -06001686 kuid_t uid;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
Eric Dumazetf064af12010-09-22 12:43:39 +00001688 read_lock_bh(&sk->sk_callback_lock);
Eric W. Biederman976d02012012-05-23 17:16:53 -06001689 uid = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_uid : GLOBAL_ROOT_UID;
Eric Dumazetf064af12010-09-22 12:43:39 +00001690 read_unlock_bh(&sk->sk_callback_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691 return uid;
1692}
Eric Dumazet2a915252009-05-27 11:30:05 +00001693EXPORT_SYMBOL(sock_i_uid);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694
1695unsigned long sock_i_ino(struct sock *sk)
1696{
1697 unsigned long ino;
1698
Eric Dumazetf064af12010-09-22 12:43:39 +00001699 read_lock_bh(&sk->sk_callback_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 ino = sk->sk_socket ? SOCK_INODE(sk->sk_socket)->i_ino : 0;
Eric Dumazetf064af12010-09-22 12:43:39 +00001701 read_unlock_bh(&sk->sk_callback_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 return ino;
1703}
Eric Dumazet2a915252009-05-27 11:30:05 +00001704EXPORT_SYMBOL(sock_i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705
1706/*
1707 * Allocate a skb from the socket's send buffer.
1708 */
Victor Fusco86a76ca2005-07-08 14:57:47 -07001709struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
Al Virodd0fc662005-10-07 07:46:04 +01001710 gfp_t priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711{
1712 if (force || atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf) {
Eric Dumazet2a915252009-05-27 11:30:05 +00001713 struct sk_buff *skb = alloc_skb(size, priority);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714 if (skb) {
1715 skb_set_owner_w(skb, sk);
1716 return skb;
1717 }
1718 }
1719 return NULL;
1720}
Eric Dumazet2a915252009-05-27 11:30:05 +00001721EXPORT_SYMBOL(sock_wmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722
1723/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 * Allocate a memory block from the socket's option memory buffer.
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001725 */
Al Virodd0fc662005-10-07 07:46:04 +01001726void *sock_kmalloc(struct sock *sk, int size, gfp_t priority)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001727{
Eric Dumazet95c96172012-04-15 05:58:06 +00001728 if ((unsigned int)size <= sysctl_optmem_max &&
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 atomic_read(&sk->sk_omem_alloc) + size < sysctl_optmem_max) {
1730 void *mem;
1731 /* First do the add, to avoid the race if kmalloc
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001732 * might sleep.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 */
1734 atomic_add(size, &sk->sk_omem_alloc);
1735 mem = kmalloc(size, priority);
1736 if (mem)
1737 return mem;
1738 atomic_sub(size, &sk->sk_omem_alloc);
1739 }
1740 return NULL;
1741}
Eric Dumazet2a915252009-05-27 11:30:05 +00001742EXPORT_SYMBOL(sock_kmalloc);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001743
Daniel Borkmann79e88652014-11-19 17:13:11 +01001744/* Free an option memory block. Note, we actually want the inline
1745 * here as this allows gcc to detect the nullify and fold away the
1746 * condition entirely.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747 */
Daniel Borkmann79e88652014-11-19 17:13:11 +01001748static inline void __sock_kfree_s(struct sock *sk, void *mem, int size,
1749 const bool nullify)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750{
David S. Millere53da5f2014-10-14 17:02:37 -04001751 if (WARN_ON_ONCE(!mem))
1752 return;
Daniel Borkmann79e88652014-11-19 17:13:11 +01001753 if (nullify)
1754 kzfree(mem);
1755 else
1756 kfree(mem);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 atomic_sub(size, &sk->sk_omem_alloc);
1758}
Daniel Borkmann79e88652014-11-19 17:13:11 +01001759
1760void sock_kfree_s(struct sock *sk, void *mem, int size)
1761{
1762 __sock_kfree_s(sk, mem, size, false);
1763}
Eric Dumazet2a915252009-05-27 11:30:05 +00001764EXPORT_SYMBOL(sock_kfree_s);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765
Daniel Borkmann79e88652014-11-19 17:13:11 +01001766void sock_kzfree_s(struct sock *sk, void *mem, int size)
1767{
1768 __sock_kfree_s(sk, mem, size, true);
1769}
1770EXPORT_SYMBOL(sock_kzfree_s);
1771
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772/* It is almost wait_for_tcp_memory minus release_sock/lock_sock.
1773 I think, these locks should be removed for datagram sockets.
1774 */
Eric Dumazet2a915252009-05-27 11:30:05 +00001775static long sock_wait_for_wmem(struct sock *sk, long timeo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776{
1777 DEFINE_WAIT(wait);
1778
1779 clear_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1780 for (;;) {
1781 if (!timeo)
1782 break;
1783 if (signal_pending(current))
1784 break;
1785 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
Eric Dumazetaa395142010-04-20 13:03:51 +00001786 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787 if (atomic_read(&sk->sk_wmem_alloc) < sk->sk_sndbuf)
1788 break;
1789 if (sk->sk_shutdown & SEND_SHUTDOWN)
1790 break;
1791 if (sk->sk_err)
1792 break;
1793 timeo = schedule_timeout(timeo);
1794 }
Eric Dumazetaa395142010-04-20 13:03:51 +00001795 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796 return timeo;
1797}
1798
1799
1800/*
1801 * Generic send/receive buffer handlers
1802 */
1803
Herbert Xu4cc7f682009-02-04 16:55:54 -08001804struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1805 unsigned long data_len, int noblock,
Eric Dumazet28d64272013-08-08 14:38:47 -07001806 int *errcode, int max_page_order)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807{
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001808 struct sk_buff *skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 long timeo;
1810 int err;
1811
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 timeo = sock_sndtimeo(sk, noblock);
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001813 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 err = sock_error(sk);
1815 if (err != 0)
1816 goto failure;
1817
1818 err = -EPIPE;
1819 if (sk->sk_shutdown & SEND_SHUTDOWN)
1820 goto failure;
1821
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001822 if (sk_wmem_alloc_get(sk) < sk->sk_sndbuf)
1823 break;
Eric Dumazet28d64272013-08-08 14:38:47 -07001824
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001825 set_bit(SOCK_ASYNC_NOSPACE, &sk->sk_socket->flags);
1826 set_bit(SOCK_NOSPACE, &sk->sk_socket->flags);
1827 err = -EAGAIN;
1828 if (!timeo)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829 goto failure;
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001830 if (signal_pending(current))
1831 goto interrupted;
1832 timeo = sock_wait_for_wmem(sk, timeo);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 }
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001834 skb = alloc_skb_with_frags(header_len, data_len, max_page_order,
1835 errcode, sk->sk_allocation);
1836 if (skb)
1837 skb_set_owner_w(skb, sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838 return skb;
1839
1840interrupted:
1841 err = sock_intr_errno(timeo);
1842failure:
1843 *errcode = err;
1844 return NULL;
1845}
Herbert Xu4cc7f682009-02-04 16:55:54 -08001846EXPORT_SYMBOL(sock_alloc_send_pskb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09001848struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849 int noblock, int *errcode)
1850{
Eric Dumazet28d64272013-08-08 14:38:47 -07001851 return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852}
Eric Dumazet2a915252009-05-27 11:30:05 +00001853EXPORT_SYMBOL(sock_alloc_send_skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854
Eric Dumazet5640f762012-09-23 23:04:42 +00001855/* On 32bit arches, an skb frag is limited to 2^15 */
1856#define SKB_FRAG_PAGE_ORDER get_order(32768)
1857
Eric Dumazet400dfd32013-10-17 16:27:07 -07001858/**
1859 * skb_page_frag_refill - check that a page_frag contains enough room
1860 * @sz: minimum size of the fragment we want to get
1861 * @pfrag: pointer to page_frag
Eric Dumazet82d5e2b2014-09-08 04:00:00 -07001862 * @gfp: priority for memory allocation
Eric Dumazet400dfd32013-10-17 16:27:07 -07001863 *
1864 * Note: While this allocator tries to use high order pages, there is
1865 * no guarantee that allocations succeed. Therefore, @sz MUST be
1866 * less or equal than PAGE_SIZE.
1867 */
Eric Dumazetd9b29382014-08-27 20:49:34 -07001868bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t gfp)
Eric Dumazet5640f762012-09-23 23:04:42 +00001869{
Eric Dumazet5640f762012-09-23 23:04:42 +00001870 if (pfrag->page) {
1871 if (atomic_read(&pfrag->page->_count) == 1) {
1872 pfrag->offset = 0;
1873 return true;
1874 }
Eric Dumazet400dfd32013-10-17 16:27:07 -07001875 if (pfrag->offset + sz <= pfrag->size)
Eric Dumazet5640f762012-09-23 23:04:42 +00001876 return true;
1877 put_page(pfrag->page);
1878 }
1879
Eric Dumazetd9b29382014-08-27 20:49:34 -07001880 pfrag->offset = 0;
1881 if (SKB_FRAG_PAGE_ORDER) {
Shaohua Lifb05e7a2015-06-11 16:50:48 -07001882 pfrag->page = alloc_pages((gfp & ~__GFP_WAIT) | __GFP_COMP |
Eric Dumazetd9b29382014-08-27 20:49:34 -07001883 __GFP_NOWARN | __GFP_NORETRY,
1884 SKB_FRAG_PAGE_ORDER);
Eric Dumazet5640f762012-09-23 23:04:42 +00001885 if (likely(pfrag->page)) {
Eric Dumazetd9b29382014-08-27 20:49:34 -07001886 pfrag->size = PAGE_SIZE << SKB_FRAG_PAGE_ORDER;
Eric Dumazet5640f762012-09-23 23:04:42 +00001887 return true;
1888 }
Eric Dumazetd9b29382014-08-27 20:49:34 -07001889 }
1890 pfrag->page = alloc_page(gfp);
1891 if (likely(pfrag->page)) {
1892 pfrag->size = PAGE_SIZE;
1893 return true;
1894 }
Eric Dumazet400dfd32013-10-17 16:27:07 -07001895 return false;
1896}
1897EXPORT_SYMBOL(skb_page_frag_refill);
1898
1899bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag)
1900{
1901 if (likely(skb_page_frag_refill(32U, pfrag, sk->sk_allocation)))
1902 return true;
1903
Eric Dumazet5640f762012-09-23 23:04:42 +00001904 sk_enter_memory_pressure(sk);
1905 sk_stream_moderate_sndbuf(sk);
1906 return false;
1907}
1908EXPORT_SYMBOL(sk_page_frag_refill);
1909
Linus Torvalds1da177e2005-04-16 15:20:36 -07001910static void __lock_sock(struct sock *sk)
Namhyung Kimf39234d2010-09-08 03:48:48 +00001911 __releases(&sk->sk_lock.slock)
1912 __acquires(&sk->sk_lock.slock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913{
1914 DEFINE_WAIT(wait);
1915
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001916 for (;;) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917 prepare_to_wait_exclusive(&sk->sk_lock.wq, &wait,
1918 TASK_UNINTERRUPTIBLE);
1919 spin_unlock_bh(&sk->sk_lock.slock);
1920 schedule();
1921 spin_lock_bh(&sk->sk_lock.slock);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001922 if (!sock_owned_by_user(sk))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923 break;
1924 }
1925 finish_wait(&sk->sk_lock.wq, &wait);
1926}
1927
1928static void __release_sock(struct sock *sk)
Namhyung Kimf39234d2010-09-08 03:48:48 +00001929 __releases(&sk->sk_lock.slock)
1930 __acquires(&sk->sk_lock.slock)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931{
1932 struct sk_buff *skb = sk->sk_backlog.head;
1933
1934 do {
1935 sk->sk_backlog.head = sk->sk_backlog.tail = NULL;
1936 bh_unlock_sock(sk);
1937
1938 do {
1939 struct sk_buff *next = skb->next;
1940
Eric Dumazete4cbb022012-04-30 16:07:09 +00001941 prefetch(next);
Eric Dumazet7fee2262010-05-11 23:19:48 +00001942 WARN_ON_ONCE(skb_dst_is_noref(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 skb->next = NULL;
Peter Zijlstrac57943a2008-10-07 14:18:42 -07001944 sk_backlog_rcv(sk, skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945
1946 /*
1947 * We are in process context here with softirqs
1948 * disabled, use cond_resched_softirq() to preempt.
1949 * This is safe to do because we've taken the backlog
1950 * queue private:
1951 */
1952 cond_resched_softirq();
1953
1954 skb = next;
1955 } while (skb != NULL);
1956
1957 bh_lock_sock(sk);
Stephen Hemmingere71a4782007-04-10 20:10:33 -07001958 } while ((skb = sk->sk_backlog.head) != NULL);
Zhu Yi8eae9392010-03-04 18:01:40 +00001959
1960 /*
1961 * Doing the zeroing here guarantee we can not loop forever
1962 * while a wild producer attempts to flood us.
1963 */
1964 sk->sk_backlog.len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001965}
1966
1967/**
1968 * sk_wait_data - wait for data to arrive at sk_receive_queue
Pavel Pisa4dc3b162005-05-01 08:59:25 -07001969 * @sk: sock to wait on
1970 * @timeo: for how long
Sabrina Dubrocadfbafc92015-07-24 18:19:25 +02001971 * @skb: last skb seen on sk_receive_queue
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 *
1973 * Now socket state including sk->sk_err is changed only under lock,
1974 * hence we may omit checks after joining wait queue.
1975 * We check receive queue before schedule() only as optimization;
1976 * it is very likely that release_sock() added new data.
1977 */
Sabrina Dubrocadfbafc92015-07-24 18:19:25 +02001978int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979{
1980 int rc;
1981 DEFINE_WAIT(wait);
1982
Eric Dumazetaa395142010-04-20 13:03:51 +00001983 prepare_to_wait(sk_sleep(sk), &wait, TASK_INTERRUPTIBLE);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 set_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
Sabrina Dubrocadfbafc92015-07-24 18:19:25 +02001985 rc = sk_wait_event(sk, timeo, skb_peek_tail(&sk->sk_receive_queue) != skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 clear_bit(SOCK_ASYNC_WAITDATA, &sk->sk_socket->flags);
Eric Dumazetaa395142010-04-20 13:03:51 +00001987 finish_wait(sk_sleep(sk), &wait);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988 return rc;
1989}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990EXPORT_SYMBOL(sk_wait_data);
1991
Hideo Aoki3ab224b2007-12-31 00:11:19 -08001992/**
1993 * __sk_mem_schedule - increase sk_forward_alloc and memory_allocated
1994 * @sk: socket
1995 * @size: memory size to allocate
1996 * @kind: allocation type
1997 *
1998 * If kind is SK_MEM_SEND, it means wmem allocation. Otherwise it means
1999 * rmem allocation. This function assumes that protocols which have
2000 * memory_pressure use sk_wmem_queued as write buffer accounting.
2001 */
2002int __sk_mem_schedule(struct sock *sk, int size, int kind)
2003{
2004 struct proto *prot = sk->sk_prot;
2005 int amt = sk_mem_pages(size);
Eric Dumazet8d987e52010-11-09 23:24:26 +00002006 long allocated;
Glauber Costae1aab162011-12-11 21:47:03 +00002007 int parent_status = UNDER_LIMIT;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002008
2009 sk->sk_forward_alloc += amt * SK_MEM_QUANTUM;
Glauber Costa180d8cd2011-12-11 21:47:02 +00002010
Glauber Costae1aab162011-12-11 21:47:03 +00002011 allocated = sk_memory_allocated_add(sk, amt, &parent_status);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002012
2013 /* Under limit. */
Glauber Costae1aab162011-12-11 21:47:03 +00002014 if (parent_status == UNDER_LIMIT &&
2015 allocated <= sk_prot_mem_limits(sk, 0)) {
Glauber Costa180d8cd2011-12-11 21:47:02 +00002016 sk_leave_memory_pressure(sk);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002017 return 1;
2018 }
2019
Glauber Costae1aab162011-12-11 21:47:03 +00002020 /* Under pressure. (we or our parents) */
2021 if ((parent_status > SOFT_LIMIT) ||
2022 allocated > sk_prot_mem_limits(sk, 1))
Glauber Costa180d8cd2011-12-11 21:47:02 +00002023 sk_enter_memory_pressure(sk);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002024
Glauber Costae1aab162011-12-11 21:47:03 +00002025 /* Over hard limit (we or our parents) */
2026 if ((parent_status == OVER_LIMIT) ||
2027 (allocated > sk_prot_mem_limits(sk, 2)))
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002028 goto suppress_allocation;
2029
2030 /* guarantee minimum buffer size under pressure */
2031 if (kind == SK_MEM_RECV) {
2032 if (atomic_read(&sk->sk_rmem_alloc) < prot->sysctl_rmem[0])
2033 return 1;
Glauber Costa180d8cd2011-12-11 21:47:02 +00002034
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002035 } else { /* SK_MEM_SEND */
2036 if (sk->sk_type == SOCK_STREAM) {
2037 if (sk->sk_wmem_queued < prot->sysctl_wmem[0])
2038 return 1;
2039 } else if (atomic_read(&sk->sk_wmem_alloc) <
2040 prot->sysctl_wmem[0])
2041 return 1;
2042 }
2043
Glauber Costa180d8cd2011-12-11 21:47:02 +00002044 if (sk_has_memory_pressure(sk)) {
Eric Dumazet17483762008-11-25 21:16:35 -08002045 int alloc;
2046
Glauber Costa180d8cd2011-12-11 21:47:02 +00002047 if (!sk_under_memory_pressure(sk))
Eric Dumazet17483762008-11-25 21:16:35 -08002048 return 1;
Glauber Costa180d8cd2011-12-11 21:47:02 +00002049 alloc = sk_sockets_allocated_read_positive(sk);
2050 if (sk_prot_mem_limits(sk, 2) > alloc *
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002051 sk_mem_pages(sk->sk_wmem_queued +
2052 atomic_read(&sk->sk_rmem_alloc) +
2053 sk->sk_forward_alloc))
2054 return 1;
2055 }
2056
2057suppress_allocation:
2058
2059 if (kind == SK_MEM_SEND && sk->sk_type == SOCK_STREAM) {
2060 sk_stream_moderate_sndbuf(sk);
2061
2062 /* Fail only if socket is _under_ its sndbuf.
2063 * In this case we cannot block, so that we have to fail.
2064 */
2065 if (sk->sk_wmem_queued + size >= sk->sk_sndbuf)
2066 return 1;
2067 }
2068
Satoru Moriya3847ce32011-06-17 12:00:03 +00002069 trace_sock_exceed_buf_limit(sk, prot, allocated);
2070
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002071 /* Alas. Undo changes. */
2072 sk->sk_forward_alloc -= amt * SK_MEM_QUANTUM;
Glauber Costa180d8cd2011-12-11 21:47:02 +00002073
Glauber Costa0e90b312012-01-20 04:57:16 +00002074 sk_memory_allocated_sub(sk, amt);
Glauber Costa180d8cd2011-12-11 21:47:02 +00002075
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002076 return 0;
2077}
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002078EXPORT_SYMBOL(__sk_mem_schedule);
2079
2080/**
Jean Sacren69dba9b2015-08-27 18:05:49 -06002081 * __sk_mem_reclaim - reclaim memory_allocated
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002082 * @sk: socket
Eric Dumazet1a24e042015-05-15 12:39:25 -07002083 * @amount: number of bytes (rounded down to a SK_MEM_QUANTUM multiple)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002084 */
Eric Dumazet1a24e042015-05-15 12:39:25 -07002085void __sk_mem_reclaim(struct sock *sk, int amount)
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002086{
Eric Dumazet1a24e042015-05-15 12:39:25 -07002087 amount >>= SK_MEM_QUANTUM_SHIFT;
2088 sk_memory_allocated_sub(sk, amount);
2089 sk->sk_forward_alloc -= amount << SK_MEM_QUANTUM_SHIFT;
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002090
Glauber Costa180d8cd2011-12-11 21:47:02 +00002091 if (sk_under_memory_pressure(sk) &&
2092 (sk_memory_allocated(sk) < sk_prot_mem_limits(sk, 0)))
2093 sk_leave_memory_pressure(sk);
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002094}
Hideo Aoki3ab224b2007-12-31 00:11:19 -08002095EXPORT_SYMBOL(__sk_mem_reclaim);
2096
2097
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098/*
2099 * Set of default routines for initialising struct proto_ops when
2100 * the protocol does not support a particular function. In certain
2101 * cases where it makes no sense for a protocol to have a "do nothing"
2102 * function, some default processing is provided.
2103 */
2104
2105int sock_no_bind(struct socket *sock, struct sockaddr *saddr, int len)
2106{
2107 return -EOPNOTSUPP;
2108}
Eric Dumazet2a915252009-05-27 11:30:05 +00002109EXPORT_SYMBOL(sock_no_bind);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09002111int sock_no_connect(struct socket *sock, struct sockaddr *saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002112 int len, int flags)
2113{
2114 return -EOPNOTSUPP;
2115}
Eric Dumazet2a915252009-05-27 11:30:05 +00002116EXPORT_SYMBOL(sock_no_connect);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117
2118int sock_no_socketpair(struct socket *sock1, struct socket *sock2)
2119{
2120 return -EOPNOTSUPP;
2121}
Eric Dumazet2a915252009-05-27 11:30:05 +00002122EXPORT_SYMBOL(sock_no_socketpair);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123
2124int sock_no_accept(struct socket *sock, struct socket *newsock, int flags)
2125{
2126 return -EOPNOTSUPP;
2127}
Eric Dumazet2a915252009-05-27 11:30:05 +00002128EXPORT_SYMBOL(sock_no_accept);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09002130int sock_no_getname(struct socket *sock, struct sockaddr *saddr,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131 int *len, int peer)
2132{
2133 return -EOPNOTSUPP;
2134}
Eric Dumazet2a915252009-05-27 11:30:05 +00002135EXPORT_SYMBOL(sock_no_getname);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136
Eric Dumazet2a915252009-05-27 11:30:05 +00002137unsigned int sock_no_poll(struct file *file, struct socket *sock, poll_table *pt)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138{
2139 return 0;
2140}
Eric Dumazet2a915252009-05-27 11:30:05 +00002141EXPORT_SYMBOL(sock_no_poll);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142
2143int sock_no_ioctl(struct socket *sock, unsigned int cmd, unsigned long arg)
2144{
2145 return -EOPNOTSUPP;
2146}
Eric Dumazet2a915252009-05-27 11:30:05 +00002147EXPORT_SYMBOL(sock_no_ioctl);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148
2149int sock_no_listen(struct socket *sock, int backlog)
2150{
2151 return -EOPNOTSUPP;
2152}
Eric Dumazet2a915252009-05-27 11:30:05 +00002153EXPORT_SYMBOL(sock_no_listen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154
2155int sock_no_shutdown(struct socket *sock, int how)
2156{
2157 return -EOPNOTSUPP;
2158}
Eric Dumazet2a915252009-05-27 11:30:05 +00002159EXPORT_SYMBOL(sock_no_shutdown);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002160
2161int sock_no_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002162 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163{
2164 return -EOPNOTSUPP;
2165}
Eric Dumazet2a915252009-05-27 11:30:05 +00002166EXPORT_SYMBOL(sock_no_setsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167
2168int sock_no_getsockopt(struct socket *sock, int level, int optname,
2169 char __user *optval, int __user *optlen)
2170{
2171 return -EOPNOTSUPP;
2172}
Eric Dumazet2a915252009-05-27 11:30:05 +00002173EXPORT_SYMBOL(sock_no_getsockopt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002174
Ying Xue1b784142015-03-02 15:37:48 +08002175int sock_no_sendmsg(struct socket *sock, struct msghdr *m, size_t len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176{
2177 return -EOPNOTSUPP;
2178}
Eric Dumazet2a915252009-05-27 11:30:05 +00002179EXPORT_SYMBOL(sock_no_sendmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002180
Ying Xue1b784142015-03-02 15:37:48 +08002181int sock_no_recvmsg(struct socket *sock, struct msghdr *m, size_t len,
2182 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183{
2184 return -EOPNOTSUPP;
2185}
Eric Dumazet2a915252009-05-27 11:30:05 +00002186EXPORT_SYMBOL(sock_no_recvmsg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002187
2188int sock_no_mmap(struct file *file, struct socket *sock, struct vm_area_struct *vma)
2189{
2190 /* Mirror missing mmap method error code */
2191 return -ENODEV;
2192}
Eric Dumazet2a915252009-05-27 11:30:05 +00002193EXPORT_SYMBOL(sock_no_mmap);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194
2195ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags)
2196{
2197 ssize_t res;
2198 struct msghdr msg = {.msg_flags = flags};
2199 struct kvec iov;
2200 char *kaddr = kmap(page);
2201 iov.iov_base = kaddr + offset;
2202 iov.iov_len = size;
2203 res = kernel_sendmsg(sock, &msg, &iov, 1, size);
2204 kunmap(page);
2205 return res;
2206}
Eric Dumazet2a915252009-05-27 11:30:05 +00002207EXPORT_SYMBOL(sock_no_sendpage);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208
2209/*
2210 * Default Socket Callbacks
2211 */
2212
2213static void sock_def_wakeup(struct sock *sk)
2214{
Eric Dumazet43815482010-04-29 11:01:49 +00002215 struct socket_wq *wq;
2216
2217 rcu_read_lock();
2218 wq = rcu_dereference(sk->sk_wq);
2219 if (wq_has_sleeper(wq))
2220 wake_up_interruptible_all(&wq->wait);
2221 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002222}
2223
2224static void sock_def_error_report(struct sock *sk)
2225{
Eric Dumazet43815482010-04-29 11:01:49 +00002226 struct socket_wq *wq;
2227
2228 rcu_read_lock();
2229 wq = rcu_dereference(sk->sk_wq);
2230 if (wq_has_sleeper(wq))
2231 wake_up_interruptible_poll(&wq->wait, POLLERR);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08002232 sk_wake_async(sk, SOCK_WAKE_IO, POLL_ERR);
Eric Dumazet43815482010-04-29 11:01:49 +00002233 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002234}
2235
David S. Miller676d2362014-04-11 16:15:36 -04002236static void sock_def_readable(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237{
Eric Dumazet43815482010-04-29 11:01:49 +00002238 struct socket_wq *wq;
2239
2240 rcu_read_lock();
2241 wq = rcu_dereference(sk->sk_wq);
2242 if (wq_has_sleeper(wq))
Eric Dumazet2c6607c2011-01-06 10:54:29 -08002243 wake_up_interruptible_sync_poll(&wq->wait, POLLIN | POLLPRI |
Davide Libenzi37e55402009-03-31 15:24:21 -07002244 POLLRDNORM | POLLRDBAND);
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08002245 sk_wake_async(sk, SOCK_WAKE_WAITD, POLL_IN);
Eric Dumazet43815482010-04-29 11:01:49 +00002246 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247}
2248
2249static void sock_def_write_space(struct sock *sk)
2250{
Eric Dumazet43815482010-04-29 11:01:49 +00002251 struct socket_wq *wq;
2252
2253 rcu_read_lock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254
2255 /* Do not wake up a writer until he can make "significant"
2256 * progress. --DaveM
2257 */
Stephen Hemmingere71a4782007-04-10 20:10:33 -07002258 if ((atomic_read(&sk->sk_wmem_alloc) << 1) <= sk->sk_sndbuf) {
Eric Dumazet43815482010-04-29 11:01:49 +00002259 wq = rcu_dereference(sk->sk_wq);
2260 if (wq_has_sleeper(wq))
2261 wake_up_interruptible_sync_poll(&wq->wait, POLLOUT |
Davide Libenzi37e55402009-03-31 15:24:21 -07002262 POLLWRNORM | POLLWRBAND);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263
2264 /* Should agree with poll, otherwise some programs break */
2265 if (sock_writeable(sk))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08002266 sk_wake_async(sk, SOCK_WAKE_SPACE, POLL_OUT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267 }
2268
Eric Dumazet43815482010-04-29 11:01:49 +00002269 rcu_read_unlock();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270}
2271
2272static void sock_def_destruct(struct sock *sk)
2273{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274}
2275
2276void sk_send_sigurg(struct sock *sk)
2277{
2278 if (sk->sk_socket && sk->sk_socket->file)
2279 if (send_sigurg(&sk->sk_socket->file->f_owner))
Pavel Emelyanov8d8ad9d2007-11-26 20:10:50 +08002280 sk_wake_async(sk, SOCK_WAKE_URG, POLL_PRI);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281}
Eric Dumazet2a915252009-05-27 11:30:05 +00002282EXPORT_SYMBOL(sk_send_sigurg);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283
2284void sk_reset_timer(struct sock *sk, struct timer_list* timer,
2285 unsigned long expires)
2286{
2287 if (!mod_timer(timer, expires))
2288 sock_hold(sk);
2289}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290EXPORT_SYMBOL(sk_reset_timer);
2291
2292void sk_stop_timer(struct sock *sk, struct timer_list* timer)
2293{
Ying Xue25cc4ae2013-02-03 20:32:57 +00002294 if (del_timer(timer))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295 __sock_put(sk);
2296}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297EXPORT_SYMBOL(sk_stop_timer);
2298
2299void sock_init_data(struct socket *sock, struct sock *sk)
2300{
2301 skb_queue_head_init(&sk->sk_receive_queue);
2302 skb_queue_head_init(&sk->sk_write_queue);
2303 skb_queue_head_init(&sk->sk_error_queue);
2304
2305 sk->sk_send_head = NULL;
2306
2307 init_timer(&sk->sk_timer);
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09002308
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309 sk->sk_allocation = GFP_KERNEL;
2310 sk->sk_rcvbuf = sysctl_rmem_default;
2311 sk->sk_sndbuf = sysctl_wmem_default;
2312 sk->sk_state = TCP_CLOSE;
David S. Miller972692e2008-06-17 22:41:38 -07002313 sk_set_socket(sk, sock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314
2315 sock_set_flag(sk, SOCK_ZAPPED);
2316
Stephen Hemmingere71a4782007-04-10 20:10:33 -07002317 if (sock) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318 sk->sk_type = sock->type;
Eric Dumazet43815482010-04-29 11:01:49 +00002319 sk->sk_wq = sock->wq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002320 sock->sk = sk;
2321 } else
Eric Dumazet43815482010-04-29 11:01:49 +00002322 sk->sk_wq = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323
Eric Dumazetb6c67122010-04-08 23:03:29 +00002324 spin_lock_init(&sk->sk_dst_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325 rwlock_init(&sk->sk_callback_lock);
Peter Zijlstra443aef02007-07-19 01:49:00 -07002326 lockdep_set_class_and_name(&sk->sk_callback_lock,
2327 af_callback_keys + sk->sk_family,
2328 af_family_clock_key_strings[sk->sk_family]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329
2330 sk->sk_state_change = sock_def_wakeup;
2331 sk->sk_data_ready = sock_def_readable;
2332 sk->sk_write_space = sock_def_write_space;
2333 sk->sk_error_report = sock_def_error_report;
2334 sk->sk_destruct = sock_def_destruct;
2335
Eric Dumazet5640f762012-09-23 23:04:42 +00002336 sk->sk_frag.page = NULL;
2337 sk->sk_frag.offset = 0;
Pavel Emelyanovef64a542012-02-21 07:31:34 +00002338 sk->sk_peek_off = -1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002339
Eric W. Biederman109f6e32010-06-13 03:30:14 +00002340 sk->sk_peer_pid = NULL;
2341 sk->sk_peer_cred = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342 sk->sk_write_pending = 0;
2343 sk->sk_rcvlowat = 1;
2344 sk->sk_rcvtimeo = MAX_SCHEDULE_TIMEOUT;
2345 sk->sk_sndtimeo = MAX_SCHEDULE_TIMEOUT;
2346
Eric Dumazetf37f0af2008-04-13 21:39:26 -07002347 sk->sk_stamp = ktime_set(-1L, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348
Cong Wange0d10952013-08-01 11:10:25 +08002349#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +03002350 sk->sk_napi_id = 0;
Eliezer Tamir64b0dc52013-07-10 17:13:36 +03002351 sk->sk_ll_usec = sysctl_net_busy_read;
Eliezer Tamir06021292013-06-10 11:39:50 +03002352#endif
2353
Eric Dumazet62748f32013-09-24 08:20:52 -07002354 sk->sk_max_pacing_rate = ~0U;
Eric Dumazet7eec4172013-10-08 15:16:00 -07002355 sk->sk_pacing_rate = ~0U;
Eric Dumazet4dc6dc72009-07-15 23:13:10 +00002356 /*
2357 * Before updating sk_refcnt, we must commit prior changes to memory
2358 * (Documentation/RCU/rculist_nulls.txt for details)
2359 */
2360 smp_wmb();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361 atomic_set(&sk->sk_refcnt, 1);
Wang Chen33c732c2007-11-13 20:30:01 -08002362 atomic_set(&sk->sk_drops, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363}
Eric Dumazet2a915252009-05-27 11:30:05 +00002364EXPORT_SYMBOL(sock_init_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365
Harvey Harrisonb5606c22008-02-13 15:03:16 -08002366void lock_sock_nested(struct sock *sk, int subclass)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367{
2368 might_sleep();
Ingo Molnara5b5bb92006-07-03 00:25:35 -07002369 spin_lock_bh(&sk->sk_lock.slock);
John Heffnerd2e91172007-09-12 10:44:19 +02002370 if (sk->sk_lock.owned)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371 __lock_sock(sk);
John Heffnerd2e91172007-09-12 10:44:19 +02002372 sk->sk_lock.owned = 1;
Ingo Molnara5b5bb92006-07-03 00:25:35 -07002373 spin_unlock(&sk->sk_lock.slock);
2374 /*
2375 * The sk_lock has mutex_lock() semantics here:
2376 */
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08002377 mutex_acquire(&sk->sk_lock.dep_map, subclass, 0, _RET_IP_);
Ingo Molnara5b5bb92006-07-03 00:25:35 -07002378 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379}
Peter Zijlstrafcc70d52006-11-08 22:44:35 -08002380EXPORT_SYMBOL(lock_sock_nested);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381
Harvey Harrisonb5606c22008-02-13 15:03:16 -08002382void release_sock(struct sock *sk)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383{
Ingo Molnara5b5bb92006-07-03 00:25:35 -07002384 /*
2385 * The sk_lock has mutex_unlock() semantics:
2386 */
2387 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
2388
2389 spin_lock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 if (sk->sk_backlog.tail)
2391 __release_sock(sk);
Eric Dumazet46d3cea2012-07-11 05:50:31 +00002392
Eric Dumazetc3f9b012014-03-10 09:50:11 -07002393 /* Warning : release_cb() might need to release sk ownership,
2394 * ie call sock_release_ownership(sk) before us.
2395 */
Eric Dumazet46d3cea2012-07-11 05:50:31 +00002396 if (sk->sk_prot->release_cb)
2397 sk->sk_prot->release_cb(sk);
2398
Eric Dumazetc3f9b012014-03-10 09:50:11 -07002399 sock_release_ownership(sk);
Ingo Molnara5b5bb92006-07-03 00:25:35 -07002400 if (waitqueue_active(&sk->sk_lock.wq))
2401 wake_up(&sk->sk_lock.wq);
2402 spin_unlock_bh(&sk->sk_lock.slock);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002403}
2404EXPORT_SYMBOL(release_sock);
2405
Eric Dumazet8a74ad62010-05-26 19:20:18 +00002406/**
2407 * lock_sock_fast - fast version of lock_sock
2408 * @sk: socket
2409 *
2410 * This version should be used for very small section, where process wont block
2411 * return false if fast path is taken
2412 * sk_lock.slock locked, owned = 0, BH disabled
2413 * return true if slow path is taken
2414 * sk_lock.slock unlocked, owned = 1, BH enabled
2415 */
2416bool lock_sock_fast(struct sock *sk)
2417{
2418 might_sleep();
2419 spin_lock_bh(&sk->sk_lock.slock);
2420
2421 if (!sk->sk_lock.owned)
2422 /*
2423 * Note : We must disable BH
2424 */
2425 return false;
2426
2427 __lock_sock(sk);
2428 sk->sk_lock.owned = 1;
2429 spin_unlock(&sk->sk_lock.slock);
2430 /*
2431 * The sk_lock has mutex_lock() semantics here:
2432 */
2433 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
2434 local_bh_enable();
2435 return true;
2436}
2437EXPORT_SYMBOL(lock_sock_fast);
2438
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439int sock_get_timestamp(struct sock *sk, struct timeval __user *userstamp)
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09002440{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002441 struct timeval tv;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442 if (!sock_flag(sk, SOCK_TIMESTAMP))
Patrick Ohly20d49472009-02-12 05:03:38 +00002443 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002444 tv = ktime_to_timeval(sk->sk_stamp);
2445 if (tv.tv_sec == -1)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002446 return -ENOENT;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002447 if (tv.tv_sec == 0) {
2448 sk->sk_stamp = ktime_get_real();
2449 tv = ktime_to_timeval(sk->sk_stamp);
2450 }
2451 return copy_to_user(userstamp, &tv, sizeof(tv)) ? -EFAULT : 0;
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09002452}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453EXPORT_SYMBOL(sock_get_timestamp);
2454
Eric Dumazetae40eb12007-03-18 17:33:16 -07002455int sock_get_timestampns(struct sock *sk, struct timespec __user *userstamp)
2456{
2457 struct timespec ts;
2458 if (!sock_flag(sk, SOCK_TIMESTAMP))
Patrick Ohly20d49472009-02-12 05:03:38 +00002459 sock_enable_timestamp(sk, SOCK_TIMESTAMP);
Eric Dumazetae40eb12007-03-18 17:33:16 -07002460 ts = ktime_to_timespec(sk->sk_stamp);
2461 if (ts.tv_sec == -1)
2462 return -ENOENT;
2463 if (ts.tv_sec == 0) {
2464 sk->sk_stamp = ktime_get_real();
2465 ts = ktime_to_timespec(sk->sk_stamp);
2466 }
2467 return copy_to_user(userstamp, &ts, sizeof(ts)) ? -EFAULT : 0;
2468}
2469EXPORT_SYMBOL(sock_get_timestampns);
2470
Patrick Ohly20d49472009-02-12 05:03:38 +00002471void sock_enable_timestamp(struct sock *sk, int flag)
YOSHIFUJI Hideaki4ec93ed2007-02-09 23:24:36 +09002472{
Patrick Ohly20d49472009-02-12 05:03:38 +00002473 if (!sock_flag(sk, flag)) {
Eric Dumazet08e29af2011-11-28 12:04:18 +00002474 unsigned long previous_flags = sk->sk_flags;
2475
Patrick Ohly20d49472009-02-12 05:03:38 +00002476 sock_set_flag(sk, flag);
2477 /*
2478 * we just set one of the two flags which require net
2479 * time stamping, but time stamping might have been on
2480 * already because of the other one
2481 */
Eric Dumazet08e29af2011-11-28 12:04:18 +00002482 if (!(previous_flags & SK_FLAGS_TIMESTAMP))
Patrick Ohly20d49472009-02-12 05:03:38 +00002483 net_enable_timestamp();
Linus Torvalds1da177e2005-04-16 15:20:36 -07002484 }
2485}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002486
Richard Cochrancb820f82013-07-19 19:40:09 +02002487int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len,
2488 int level, int type)
2489{
2490 struct sock_exterr_skb *serr;
Willem de Bruijn364a9e92014-08-31 21:30:27 -04002491 struct sk_buff *skb;
Richard Cochrancb820f82013-07-19 19:40:09 +02002492 int copied, err;
2493
2494 err = -EAGAIN;
Willem de Bruijn364a9e92014-08-31 21:30:27 -04002495 skb = sock_dequeue_err_skb(sk);
Richard Cochrancb820f82013-07-19 19:40:09 +02002496 if (skb == NULL)
2497 goto out;
2498
2499 copied = skb->len;
2500 if (copied > len) {
2501 msg->msg_flags |= MSG_TRUNC;
2502 copied = len;
2503 }
David S. Miller51f3d022014-11-05 16:46:40 -05002504 err = skb_copy_datagram_msg(skb, 0, msg, copied);
Richard Cochrancb820f82013-07-19 19:40:09 +02002505 if (err)
2506 goto out_free_skb;
2507
2508 sock_recv_timestamp(msg, sk, skb);
2509
2510 serr = SKB_EXT_ERR(skb);
2511 put_cmsg(msg, level, type, sizeof(serr->ee), &serr->ee);
2512
2513 msg->msg_flags |= MSG_ERRQUEUE;
2514 err = copied;
2515
Richard Cochrancb820f82013-07-19 19:40:09 +02002516out_free_skb:
2517 kfree_skb(skb);
2518out:
2519 return err;
2520}
2521EXPORT_SYMBOL(sock_recv_errqueue);
2522
Linus Torvalds1da177e2005-04-16 15:20:36 -07002523/*
2524 * Get a socket option on an socket.
2525 *
2526 * FIX: POSIX 1003.1g is very ambiguous here. It states that
2527 * asynchronous errors should be reported by getsockopt. We assume
2528 * this means if you specify SO_ERROR (otherwise whats the point of it).
2529 */
2530int sock_common_getsockopt(struct socket *sock, int level, int optname,
2531 char __user *optval, int __user *optlen)
2532{
2533 struct sock *sk = sock->sk;
2534
2535 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2536}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002537EXPORT_SYMBOL(sock_common_getsockopt);
2538
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002539#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002540int compat_sock_common_getsockopt(struct socket *sock, int level, int optname,
2541 char __user *optval, int __user *optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002542{
2543 struct sock *sk = sock->sk;
2544
Johannes Berg1e51f952007-03-06 13:44:06 -08002545 if (sk->sk_prot->compat_getsockopt != NULL)
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002546 return sk->sk_prot->compat_getsockopt(sk, level, optname,
2547 optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002548 return sk->sk_prot->getsockopt(sk, level, optname, optval, optlen);
2549}
2550EXPORT_SYMBOL(compat_sock_common_getsockopt);
2551#endif
2552
Ying Xue1b784142015-03-02 15:37:48 +08002553int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
2554 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555{
2556 struct sock *sk = sock->sk;
2557 int addr_len = 0;
2558 int err;
2559
Ying Xue1b784142015-03-02 15:37:48 +08002560 err = sk->sk_prot->recvmsg(sk, msg, size, flags & MSG_DONTWAIT,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002561 flags & ~MSG_DONTWAIT, &addr_len);
2562 if (err >= 0)
2563 msg->msg_namelen = addr_len;
2564 return err;
2565}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002566EXPORT_SYMBOL(sock_common_recvmsg);
2567
2568/*
2569 * Set socket options on an inet socket.
2570 */
2571int sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002572 char __user *optval, unsigned int optlen)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002573{
2574 struct sock *sk = sock->sk;
2575
2576 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2577}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578EXPORT_SYMBOL(sock_common_setsockopt);
2579
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002580#ifdef CONFIG_COMPAT
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002581int compat_sock_common_setsockopt(struct socket *sock, int level, int optname,
David S. Millerb7058842009-09-30 16:12:20 -07002582 char __user *optval, unsigned int optlen)
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002583{
2584 struct sock *sk = sock->sk;
2585
Arnaldo Carvalho de Melo543d9cf2006-03-20 22:48:35 -08002586 if (sk->sk_prot->compat_setsockopt != NULL)
2587 return sk->sk_prot->compat_setsockopt(sk, level, optname,
2588 optval, optlen);
Dmitry Mishin3fdadf72006-03-20 22:45:21 -08002589 return sk->sk_prot->setsockopt(sk, level, optname, optval, optlen);
2590}
2591EXPORT_SYMBOL(compat_sock_common_setsockopt);
2592#endif
2593
Linus Torvalds1da177e2005-04-16 15:20:36 -07002594void sk_common_release(struct sock *sk)
2595{
2596 if (sk->sk_prot->destroy)
2597 sk->sk_prot->destroy(sk);
2598
2599 /*
2600 * Observation: when sock_common_release is called, processes have
2601 * no access to socket. But net still has.
2602 * Step one, detach it from networking:
2603 *
2604 * A. Remove from hash tables.
2605 */
2606
2607 sk->sk_prot->unhash(sk);
2608
2609 /*
2610 * In this point socket cannot receive new packets, but it is possible
2611 * that some packets are in flight because some CPU runs receiver and
2612 * did hash table lookup before we unhashed socket. They will achieve
2613 * receive queue and will be purged by socket destructor.
2614 *
2615 * Also we still have packets pending on receive queue and probably,
2616 * our own packets waiting in device queues. sock_destroy will drain
2617 * receive queue, but transmitted packets will delay socket destruction
2618 * until the last reference will be released.
2619 */
2620
2621 sock_orphan(sk);
2622
2623 xfrm_sk_free_policy(sk);
2624
Arnaldo Carvalho de Meloe6848972005-08-09 19:45:38 -07002625 sk_refcnt_debug_release(sk);
Eric Dumazet5640f762012-09-23 23:04:42 +00002626
2627 if (sk->sk_frag.page) {
2628 put_page(sk->sk_frag.page);
2629 sk->sk_frag.page = NULL;
2630 }
2631
Linus Torvalds1da177e2005-04-16 15:20:36 -07002632 sock_put(sk);
2633}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002634EXPORT_SYMBOL(sk_common_release);
2635
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002636#ifdef CONFIG_PROC_FS
2637#define PROTO_INUSE_NR 64 /* should be enough for the first time */
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002638struct prot_inuse {
2639 int val[PROTO_INUSE_NR];
2640};
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002641
2642static DECLARE_BITMAP(proto_inuse_idx, PROTO_INUSE_NR);
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002643
2644#ifdef CONFIG_NET_NS
2645void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
2646{
Eric Dumazetd6d9ca02010-07-19 10:48:49 +00002647 __this_cpu_add(net->core.inuse->val[prot->inuse_idx], val);
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002648}
2649EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2650
2651int sock_prot_inuse_get(struct net *net, struct proto *prot)
2652{
2653 int cpu, idx = prot->inuse_idx;
2654 int res = 0;
2655
2656 for_each_possible_cpu(cpu)
2657 res += per_cpu_ptr(net->core.inuse, cpu)->val[idx];
2658
2659 return res >= 0 ? res : 0;
2660}
2661EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
2662
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002663static int __net_init sock_inuse_init_net(struct net *net)
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002664{
2665 net->core.inuse = alloc_percpu(struct prot_inuse);
2666 return net->core.inuse ? 0 : -ENOMEM;
2667}
2668
Alexey Dobriyan2c8c1e72010-01-17 03:35:32 +00002669static void __net_exit sock_inuse_exit_net(struct net *net)
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002670{
2671 free_percpu(net->core.inuse);
2672}
2673
2674static struct pernet_operations net_inuse_ops = {
2675 .init = sock_inuse_init_net,
2676 .exit = sock_inuse_exit_net,
2677};
2678
2679static __init int net_inuse_init(void)
2680{
2681 if (register_pernet_subsys(&net_inuse_ops))
2682 panic("Cannot initialize net inuse counters");
2683
2684 return 0;
2685}
2686
2687core_initcall(net_inuse_init);
2688#else
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002689static DEFINE_PER_CPU(struct prot_inuse, prot_inuse);
2690
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07002691void sock_prot_inuse_add(struct net *net, struct proto *prot, int val)
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002692{
Eric Dumazetd6d9ca02010-07-19 10:48:49 +00002693 __this_cpu_add(prot_inuse.val[prot->inuse_idx], val);
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002694}
2695EXPORT_SYMBOL_GPL(sock_prot_inuse_add);
2696
Pavel Emelyanovc29a0bc2008-03-31 19:41:46 -07002697int sock_prot_inuse_get(struct net *net, struct proto *prot)
Pavel Emelyanov1338d462008-03-28 16:38:43 -07002698{
2699 int cpu, idx = prot->inuse_idx;
2700 int res = 0;
2701
2702 for_each_possible_cpu(cpu)
2703 res += per_cpu(prot_inuse, cpu).val[idx];
2704
2705 return res >= 0 ? res : 0;
2706}
2707EXPORT_SYMBOL_GPL(sock_prot_inuse_get);
Pavel Emelyanov70ee1152008-03-31 19:42:16 -07002708#endif
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002709
2710static void assign_proto_idx(struct proto *prot)
2711{
2712 prot->inuse_idx = find_first_zero_bit(proto_inuse_idx, PROTO_INUSE_NR);
2713
2714 if (unlikely(prot->inuse_idx == PROTO_INUSE_NR - 1)) {
Joe Perchese005d192012-05-16 19:58:40 +00002715 pr_err("PROTO_INUSE_NR exhausted\n");
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002716 return;
2717 }
2718
2719 set_bit(prot->inuse_idx, proto_inuse_idx);
2720}
2721
2722static void release_proto_idx(struct proto *prot)
2723{
2724 if (prot->inuse_idx != PROTO_INUSE_NR - 1)
2725 clear_bit(prot->inuse_idx, proto_inuse_idx);
2726}
2727#else
2728static inline void assign_proto_idx(struct proto *prot)
2729{
2730}
2731
2732static inline void release_proto_idx(struct proto *prot)
2733{
2734}
2735#endif
2736
Eric Dumazet0159dfd2015-03-12 16:44:07 -07002737static void req_prot_cleanup(struct request_sock_ops *rsk_prot)
2738{
2739 if (!rsk_prot)
2740 return;
2741 kfree(rsk_prot->slab_name);
2742 rsk_prot->slab_name = NULL;
2743 if (rsk_prot->slab) {
2744 kmem_cache_destroy(rsk_prot->slab);
2745 rsk_prot->slab = NULL;
2746 }
2747}
2748
2749static int req_prot_init(const struct proto *prot)
2750{
2751 struct request_sock_ops *rsk_prot = prot->rsk_prot;
2752
2753 if (!rsk_prot)
2754 return 0;
2755
2756 rsk_prot->slab_name = kasprintf(GFP_KERNEL, "request_sock_%s",
2757 prot->name);
2758 if (!rsk_prot->slab_name)
2759 return -ENOMEM;
2760
2761 rsk_prot->slab = kmem_cache_create(rsk_prot->slab_name,
2762 rsk_prot->obj_size, 0,
Eric Dumazetfa76ce732015-03-19 19:04:20 -07002763 0, NULL);
Eric Dumazet0159dfd2015-03-12 16:44:07 -07002764
2765 if (!rsk_prot->slab) {
2766 pr_crit("%s: Can't create request sock SLAB cache!\n",
2767 prot->name);
2768 return -ENOMEM;
2769 }
2770 return 0;
2771}
2772
Linus Torvalds1da177e2005-04-16 15:20:36 -07002773int proto_register(struct proto *prot, int alloc_slab)
2774{
Linus Torvalds1da177e2005-04-16 15:20:36 -07002775 if (alloc_slab) {
2776 prot->slab = kmem_cache_create(prot->name, prot->obj_size, 0,
Eric Dumazet271b72c2008-10-29 02:11:14 -07002777 SLAB_HWCACHE_ALIGN | prot->slab_flags,
2778 NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002779
2780 if (prot->slab == NULL) {
Joe Perchese005d192012-05-16 19:58:40 +00002781 pr_crit("%s: Can't create sock SLAB cache!\n",
2782 prot->name);
Pavel Emelyanov60e76632008-03-28 16:39:10 -07002783 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002784 }
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002785
Eric Dumazet0159dfd2015-03-12 16:44:07 -07002786 if (req_prot_init(prot))
2787 goto out_free_request_sock_slab;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002788
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002789 if (prot->twsk_prot != NULL) {
Alexey Dobriyanfaf23422010-02-17 09:34:12 +00002790 prot->twsk_prot->twsk_slab_name = kasprintf(GFP_KERNEL, "tw_sock_%s", prot->name);
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002791
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002792 if (prot->twsk_prot->twsk_slab_name == NULL)
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002793 goto out_free_request_sock_slab;
2794
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002795 prot->twsk_prot->twsk_slab =
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002796 kmem_cache_create(prot->twsk_prot->twsk_slab_name,
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002797 prot->twsk_prot->twsk_obj_size,
Eric Dumazet3ab5aee2008-11-16 19:40:17 -08002798 0,
Eric Dumazet52db70d2015-04-10 06:07:18 -07002799 prot->slab_flags,
Paul Mundt20c2df82007-07-20 10:11:58 +09002800 NULL);
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002801 if (prot->twsk_prot->twsk_slab == NULL)
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002802 goto out_free_timewait_sock_slab_name;
2803 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804 }
2805
Glauber Costa36b77a52011-12-16 00:51:59 +00002806 mutex_lock(&proto_list_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002807 list_add(&prot->node, &proto_list);
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002808 assign_proto_idx(prot);
Glauber Costa36b77a52011-12-16 00:51:59 +00002809 mutex_unlock(&proto_list_mutex);
Pavel Emelyanovb733c002007-11-07 02:23:38 -08002810 return 0;
2811
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002812out_free_timewait_sock_slab_name:
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002813 kfree(prot->twsk_prot->twsk_slab_name);
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002814out_free_request_sock_slab:
Eric Dumazet0159dfd2015-03-12 16:44:07 -07002815 req_prot_cleanup(prot->rsk_prot);
2816
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002817 kmem_cache_destroy(prot->slab);
2818 prot->slab = NULL;
Pavel Emelyanovb733c002007-11-07 02:23:38 -08002819out:
2820 return -ENOBUFS;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002821}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002822EXPORT_SYMBOL(proto_register);
2823
2824void proto_unregister(struct proto *prot)
2825{
Glauber Costa36b77a52011-12-16 00:51:59 +00002826 mutex_lock(&proto_list_mutex);
Pavel Emelyanov13ff3d62008-03-28 16:38:17 -07002827 release_proto_idx(prot);
Patrick McHardy0a3f4352005-09-06 19:47:50 -07002828 list_del(&prot->node);
Glauber Costa36b77a52011-12-16 00:51:59 +00002829 mutex_unlock(&proto_list_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002830
2831 if (prot->slab != NULL) {
2832 kmem_cache_destroy(prot->slab);
2833 prot->slab = NULL;
2834 }
2835
Eric Dumazet0159dfd2015-03-12 16:44:07 -07002836 req_prot_cleanup(prot->rsk_prot);
Arnaldo Carvalho de Melo2e6599c2005-06-18 22:46:52 -07002837
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002838 if (prot->twsk_prot != NULL && prot->twsk_prot->twsk_slab != NULL) {
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002839 kmem_cache_destroy(prot->twsk_prot->twsk_slab);
Catalin Marinas7e56b5d2008-11-21 16:45:22 -08002840 kfree(prot->twsk_prot->twsk_slab_name);
Arnaldo Carvalho de Melo6d6ee432005-12-13 23:25:19 -08002841 prot->twsk_prot->twsk_slab = NULL;
Arnaldo Carvalho de Melo8feaf0c02005-08-09 20:09:30 -07002842 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002843}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002844EXPORT_SYMBOL(proto_unregister);
2845
2846#ifdef CONFIG_PROC_FS
Linus Torvalds1da177e2005-04-16 15:20:36 -07002847static void *proto_seq_start(struct seq_file *seq, loff_t *pos)
Glauber Costa36b77a52011-12-16 00:51:59 +00002848 __acquires(proto_list_mutex)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002849{
Glauber Costa36b77a52011-12-16 00:51:59 +00002850 mutex_lock(&proto_list_mutex);
Pavel Emelianov60f04382007-07-09 13:15:14 -07002851 return seq_list_start_head(&proto_list, *pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002852}
2853
2854static void *proto_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2855{
Pavel Emelianov60f04382007-07-09 13:15:14 -07002856 return seq_list_next(v, &proto_list, pos);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002857}
2858
2859static void proto_seq_stop(struct seq_file *seq, void *v)
Glauber Costa36b77a52011-12-16 00:51:59 +00002860 __releases(proto_list_mutex)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002861{
Glauber Costa36b77a52011-12-16 00:51:59 +00002862 mutex_unlock(&proto_list_mutex);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002863}
2864
2865static char proto_method_implemented(const void *method)
2866{
2867 return method == NULL ? 'n' : 'y';
2868}
Glauber Costa180d8cd2011-12-11 21:47:02 +00002869static long sock_prot_memory_allocated(struct proto *proto)
2870{
Jeffrin Josecb75a362012-04-25 19:17:29 +05302871 return proto->memory_allocated != NULL ? proto_memory_allocated(proto) : -1L;
Glauber Costa180d8cd2011-12-11 21:47:02 +00002872}
2873
2874static char *sock_prot_memory_pressure(struct proto *proto)
2875{
2876 return proto->memory_pressure != NULL ?
2877 proto_memory_pressure(proto) ? "yes" : "no" : "NI";
2878}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002879
2880static void proto_seq_printf(struct seq_file *seq, struct proto *proto)
2881{
Glauber Costa180d8cd2011-12-11 21:47:02 +00002882
Eric Dumazet8d987e52010-11-09 23:24:26 +00002883 seq_printf(seq, "%-9s %4u %6d %6ld %-3s %6u %-3s %-10s "
Linus Torvalds1da177e2005-04-16 15:20:36 -07002884 "%2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c %2c\n",
2885 proto->name,
2886 proto->obj_size,
Eric Dumazet14e943d2008-11-19 15:14:01 -08002887 sock_prot_inuse_get(seq_file_net(seq), proto),
Glauber Costa180d8cd2011-12-11 21:47:02 +00002888 sock_prot_memory_allocated(proto),
2889 sock_prot_memory_pressure(proto),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002890 proto->max_header,
2891 proto->slab == NULL ? "no" : "yes",
2892 module_name(proto->owner),
2893 proto_method_implemented(proto->close),
2894 proto_method_implemented(proto->connect),
2895 proto_method_implemented(proto->disconnect),
2896 proto_method_implemented(proto->accept),
2897 proto_method_implemented(proto->ioctl),
2898 proto_method_implemented(proto->init),
2899 proto_method_implemented(proto->destroy),
2900 proto_method_implemented(proto->shutdown),
2901 proto_method_implemented(proto->setsockopt),
2902 proto_method_implemented(proto->getsockopt),
2903 proto_method_implemented(proto->sendmsg),
2904 proto_method_implemented(proto->recvmsg),
2905 proto_method_implemented(proto->sendpage),
2906 proto_method_implemented(proto->bind),
2907 proto_method_implemented(proto->backlog_rcv),
2908 proto_method_implemented(proto->hash),
2909 proto_method_implemented(proto->unhash),
2910 proto_method_implemented(proto->get_port),
2911 proto_method_implemented(proto->enter_memory_pressure));
2912}
2913
2914static int proto_seq_show(struct seq_file *seq, void *v)
2915{
Pavel Emelianov60f04382007-07-09 13:15:14 -07002916 if (v == &proto_list)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002917 seq_printf(seq, "%-9s %-4s %-8s %-6s %-5s %-7s %-4s %-10s %s",
2918 "protocol",
2919 "size",
2920 "sockets",
2921 "memory",
2922 "press",
2923 "maxhdr",
2924 "slab",
2925 "module",
2926 "cl co di ac io in de sh ss gs se re sp bi br ha uh gp em\n");
2927 else
Pavel Emelianov60f04382007-07-09 13:15:14 -07002928 proto_seq_printf(seq, list_entry(v, struct proto, node));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002929 return 0;
2930}
2931
Stephen Hemmingerf6908082007-03-12 14:34:29 -07002932static const struct seq_operations proto_seq_ops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002933 .start = proto_seq_start,
2934 .next = proto_seq_next,
2935 .stop = proto_seq_stop,
2936 .show = proto_seq_show,
2937};
2938
2939static int proto_seq_open(struct inode *inode, struct file *file)
2940{
Eric Dumazet14e943d2008-11-19 15:14:01 -08002941 return seq_open_net(inode, file, &proto_seq_ops,
2942 sizeof(struct seq_net_private));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002943}
2944
Arjan van de Ven9a321442007-02-12 00:55:35 -08002945static const struct file_operations proto_seq_fops = {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002946 .owner = THIS_MODULE,
2947 .open = proto_seq_open,
2948 .read = seq_read,
2949 .llseek = seq_lseek,
Eric Dumazet14e943d2008-11-19 15:14:01 -08002950 .release = seq_release_net,
2951};
2952
2953static __net_init int proto_init_net(struct net *net)
2954{
Gao fengd4beaa62013-02-18 01:34:54 +00002955 if (!proc_create("protocols", S_IRUGO, net->proc_net, &proto_seq_fops))
Eric Dumazet14e943d2008-11-19 15:14:01 -08002956 return -ENOMEM;
2957
2958 return 0;
2959}
2960
2961static __net_exit void proto_exit_net(struct net *net)
2962{
Gao fengece31ff2013-02-18 01:34:56 +00002963 remove_proc_entry("protocols", net->proc_net);
Eric Dumazet14e943d2008-11-19 15:14:01 -08002964}
2965
2966
2967static __net_initdata struct pernet_operations proto_net_ops = {
2968 .init = proto_init_net,
2969 .exit = proto_exit_net,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002970};
2971
2972static int __init proto_init(void)
2973{
Eric Dumazet14e943d2008-11-19 15:14:01 -08002974 return register_pernet_subsys(&proto_net_ops);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002975}
2976
2977subsys_initcall(proto_init);
2978
2979#endif /* PROC_FS */