Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | /* |
| 2 | * INET An implementation of the TCP/IP protocol suite for the LINUX |
| 3 | * operating system. INET is implemented using the BSD Socket |
| 4 | * interface as the means of communication with the user level. |
| 5 | * |
| 6 | * Definitions for the AF_INET socket handler. |
| 7 | * |
| 8 | * Version: @(#)sock.h 1.0.4 05/13/93 |
| 9 | * |
Jesper Juhl | 02c30a8 | 2005-05-05 16:16:16 -0700 | [diff] [blame] | 10 | * Authors: Ross Biro |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 11 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
| 12 | * Corey Minyard <wf-rch!minyard@relay.EU.net> |
| 13 | * Florian La Roche <flla@stud.uni-sb.de> |
| 14 | * |
| 15 | * Fixes: |
| 16 | * Alan Cox : Volatiles in skbuff pointers. See |
| 17 | * skbuff comments. May be overdone, |
| 18 | * better to prove they can be removed |
| 19 | * than the reverse. |
| 20 | * Alan Cox : Added a zapped field for tcp to note |
| 21 | * a socket is reset and must stay shut up |
| 22 | * Alan Cox : New fields for options |
| 23 | * Pauline Middelink : identd support |
| 24 | * Alan Cox : Eliminate low level recv/recvfrom |
| 25 | * David S. Miller : New socket lookup architecture. |
| 26 | * Steve Whitehouse: Default routines for sock_ops |
| 27 | * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made |
| 28 | * protinfo be just a void pointer, as the |
| 29 | * protocol specific parts were moved to |
| 30 | * respective headers and ipv4/v6, etc now |
| 31 | * use private slabcaches for its socks |
| 32 | * Pedro Hortas : New flags field for socket options |
| 33 | * |
| 34 | * |
| 35 | * This program is free software; you can redistribute it and/or |
| 36 | * modify it under the terms of the GNU General Public License |
| 37 | * as published by the Free Software Foundation; either version |
| 38 | * 2 of the License, or (at your option) any later version. |
| 39 | */ |
| 40 | #ifndef _SOCK_H |
| 41 | #define _SOCK_H |
| 42 | |
Alexey Dobriyan | a6b7a40 | 2011-06-06 10:43:46 +0000 | [diff] [blame] | 43 | #include <linux/hardirq.h> |
Ilpo Järvinen | 172589c | 2007-08-28 15:50:33 -0700 | [diff] [blame] | 44 | #include <linux/kernel.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 45 | #include <linux/list.h> |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 46 | #include <linux/list_nulls.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 47 | #include <linux/timer.h> |
| 48 | #include <linux/cache.h> |
Ingo Molnar | a5b5bb9 | 2006-07-03 00:25:35 -0700 | [diff] [blame] | 49 | #include <linux/lockdep.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 50 | #include <linux/netdevice.h> |
| 51 | #include <linux/skbuff.h> /* struct sk_buff */ |
Al Viro | d7fe0f2 | 2006-12-03 23:15:30 -0500 | [diff] [blame] | 52 | #include <linux/mm.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 53 | #include <linux/security.h> |
Tejun Heo | 5a0e3ad | 2010-03-24 17:04:11 +0900 | [diff] [blame] | 54 | #include <linux/slab.h> |
Tom Herbert | c6e1a0d | 2011-04-04 22:30:30 -0700 | [diff] [blame] | 55 | #include <linux/uaccess.h> |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 56 | #include <linux/memcontrol.h> |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 57 | #include <linux/res_counter.h> |
Ingo Molnar | c5905af | 2012-02-24 08:31:31 +0100 | [diff] [blame] | 58 | #include <linux/static_key.h> |
Al Viro | 4040153 | 2012-02-13 03:58:52 +0000 | [diff] [blame] | 59 | #include <linux/aio.h> |
| 60 | #include <linux/sched.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 61 | |
| 62 | #include <linux/filter.h> |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 63 | #include <linux/rculist_nulls.h> |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 64 | #include <linux/poll.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 65 | |
Eric Dumazet | c31504d | 2010-11-15 19:58:26 +0000 | [diff] [blame] | 66 | #include <linux/atomic.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 67 | #include <net/dst.h> |
| 68 | #include <net/checksum.h> |
| 69 | |
Eric Dumazet | 9f048bf | 2011-12-13 03:59:08 +0000 | [diff] [blame] | 70 | struct cgroup; |
| 71 | struct cgroup_subsys; |
Glauber Costa | c607b2e | 2011-12-16 00:52:00 +0000 | [diff] [blame] | 72 | #ifdef CONFIG_NET |
Glauber Costa | d1a4c0b | 2011-12-11 21:47:04 +0000 | [diff] [blame] | 73 | int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss); |
Li Zefan | 761b3ef | 2012-01-31 13:47:36 +0800 | [diff] [blame] | 74 | void mem_cgroup_sockets_destroy(struct cgroup *cgrp); |
Glauber Costa | c607b2e | 2011-12-16 00:52:00 +0000 | [diff] [blame] | 75 | #else |
| 76 | static inline |
| 77 | int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss) |
| 78 | { |
| 79 | return 0; |
| 80 | } |
| 81 | static inline |
Li Zefan | 761b3ef | 2012-01-31 13:47:36 +0800 | [diff] [blame] | 82 | void mem_cgroup_sockets_destroy(struct cgroup *cgrp) |
Glauber Costa | c607b2e | 2011-12-16 00:52:00 +0000 | [diff] [blame] | 83 | { |
| 84 | } |
| 85 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 86 | /* |
| 87 | * This structure really needs to be cleaned up. |
| 88 | * Most of it is for TCP, and not used by any of |
| 89 | * the other protocols. |
| 90 | */ |
| 91 | |
| 92 | /* Define this to get the SOCK_DBG debugging facility. */ |
| 93 | #define SOCK_DEBUGGING |
| 94 | #ifdef SOCK_DEBUGGING |
| 95 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ |
| 96 | printk(KERN_DEBUG msg); } while (0) |
| 97 | #else |
Stephen Hemminger | 4cd9029 | 2008-03-21 15:54:53 -0700 | [diff] [blame] | 98 | /* Validate arguments and do nothing */ |
Joe Perches | b9075fa | 2011-10-31 17:11:33 -0700 | [diff] [blame] | 99 | static inline __printf(2, 3) |
| 100 | void SOCK_DEBUG(struct sock *sk, const char *msg, ...) |
Stephen Hemminger | 4cd9029 | 2008-03-21 15:54:53 -0700 | [diff] [blame] | 101 | { |
| 102 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 103 | #endif |
| 104 | |
| 105 | /* This is the per-socket lock. The spinlock provides a synchronization |
| 106 | * between user contexts and software interrupt processing, whereas the |
| 107 | * mini-semaphore synchronizes multiple users amongst themselves. |
| 108 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 109 | typedef struct { |
| 110 | spinlock_t slock; |
John Heffner | d2e9117 | 2007-09-12 10:44:19 +0200 | [diff] [blame] | 111 | int owned; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 112 | wait_queue_head_t wq; |
Ingo Molnar | a5b5bb9 | 2006-07-03 00:25:35 -0700 | [diff] [blame] | 113 | /* |
| 114 | * We express the mutex-alike socket_lock semantics |
| 115 | * to the lock validator by explicitly managing |
| 116 | * the slock as a lock variant (in addition to |
| 117 | * the slock itself): |
| 118 | */ |
| 119 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
| 120 | struct lockdep_map dep_map; |
| 121 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 122 | } socket_lock_t; |
| 123 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 124 | struct sock; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 125 | struct proto; |
Denis V. Lunev | 0eeb8ff | 2007-12-04 01:15:45 -0800 | [diff] [blame] | 126 | struct net; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 127 | |
| 128 | /** |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 129 | * struct sock_common - minimal network layer representation of sockets |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 130 | * @skc_daddr: Foreign IPv4 addr |
| 131 | * @skc_rcv_saddr: Bound local IPv4 addr |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 132 | * @skc_hash: hash value used with various protocol lookup tables |
Eric Dumazet | d4cada4 | 2009-11-08 10:17:30 +0000 | [diff] [blame] | 133 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 134 | * @skc_family: network address family |
| 135 | * @skc_state: Connection state |
| 136 | * @skc_reuse: %SO_REUSEADDR setting |
| 137 | * @skc_bound_dev_if: bound device index if != 0 |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 138 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
Eric Dumazet | 512615b | 2009-11-08 10:17:58 +0000 | [diff] [blame] | 139 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 140 | * @skc_prot: protocol handlers inside a network family |
Eric W. Biederman | 07feaeb | 2007-09-12 11:58:02 +0200 | [diff] [blame] | 141 | * @skc_net: reference to the network namespace of this socket |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 142 | * @skc_node: main hash linkage for various protocol lookup tables |
| 143 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol |
| 144 | * @skc_tx_queue_mapping: tx queue number for this connection |
| 145 | * @skc_refcnt: reference count |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 146 | * |
| 147 | * This is the minimal network layer representation of sockets, the header |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 148 | * for struct sock and struct inet_timewait_sock. |
| 149 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 150 | struct sock_common { |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 151 | /* skc_daddr and skc_rcv_saddr must be grouped : |
| 152 | * cf INET_MATCH() and INET_TW_MATCH() |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 153 | */ |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 154 | __be32 skc_daddr; |
| 155 | __be32 skc_rcv_saddr; |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 156 | |
Eric Dumazet | d4cada4 | 2009-11-08 10:17:30 +0000 | [diff] [blame] | 157 | union { |
| 158 | unsigned int skc_hash; |
| 159 | __u16 skc_u16hashes[2]; |
| 160 | }; |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 161 | unsigned short skc_family; |
| 162 | volatile unsigned char skc_state; |
| 163 | unsigned char skc_reuse; |
| 164 | int skc_bound_dev_if; |
Eric Dumazet | 512615b | 2009-11-08 10:17:58 +0000 | [diff] [blame] | 165 | union { |
| 166 | struct hlist_node skc_bind_node; |
| 167 | struct hlist_nulls_node skc_portaddr_node; |
| 168 | }; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 169 | struct proto *skc_prot; |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 170 | #ifdef CONFIG_NET_NS |
Eric W. Biederman | 07feaeb | 2007-09-12 11:58:02 +0200 | [diff] [blame] | 171 | struct net *skc_net; |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 172 | #endif |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 173 | /* |
| 174 | * fields between dontcopy_begin/dontcopy_end |
| 175 | * are not copied in sock_copy() |
| 176 | */ |
Randy Dunlap | 928c41e | 2011-01-08 17:39:21 +0000 | [diff] [blame] | 177 | /* private: */ |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 178 | int skc_dontcopy_begin[0]; |
Randy Dunlap | 928c41e | 2011-01-08 17:39:21 +0000 | [diff] [blame] | 179 | /* public: */ |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 180 | union { |
| 181 | struct hlist_node skc_node; |
| 182 | struct hlist_nulls_node skc_nulls_node; |
| 183 | }; |
| 184 | int skc_tx_queue_mapping; |
| 185 | atomic_t skc_refcnt; |
Randy Dunlap | 928c41e | 2011-01-08 17:39:21 +0000 | [diff] [blame] | 186 | /* private: */ |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 187 | int skc_dontcopy_end[0]; |
Randy Dunlap | 928c41e | 2011-01-08 17:39:21 +0000 | [diff] [blame] | 188 | /* public: */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 189 | }; |
| 190 | |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 191 | struct cg_proto; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 192 | /** |
| 193 | * struct sock - network layer representation of sockets |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 194 | * @__sk_common: shared layout with inet_timewait_sock |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 195 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
| 196 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings |
| 197 | * @sk_lock: synchronizer |
| 198 | * @sk_rcvbuf: size of receive buffer in bytes |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 199 | * @sk_wq: sock wait queue and async head |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 200 | * @sk_dst_cache: destination cache |
| 201 | * @sk_dst_lock: destination cache lock |
| 202 | * @sk_policy: flow policy |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 203 | * @sk_receive_queue: incoming packets |
| 204 | * @sk_wmem_alloc: transmit queue bytes committed |
| 205 | * @sk_write_queue: Packet sending queue |
Chris Leech | 97fc2f0 | 2006-05-23 17:55:33 -0700 | [diff] [blame] | 206 | * @sk_async_wait_queue: DMA copied packets |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 207 | * @sk_omem_alloc: "o" is "option" or "other" |
| 208 | * @sk_wmem_queued: persistent queue size |
| 209 | * @sk_forward_alloc: space allocated forward |
| 210 | * @sk_allocation: allocation mode |
| 211 | * @sk_sndbuf: size of send buffer in bytes |
Wang Chen | 33c732c | 2007-11-13 20:30:01 -0800 | [diff] [blame] | 212 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 213 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 214 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets |
| 215 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
Eric Dumazet | a465419 | 2010-05-16 00:36:33 -0700 | [diff] [blame] | 216 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
Herbert Xu | bcd7611 | 2006-06-30 13:36:35 -0700 | [diff] [blame] | 217 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
Peter P Waskiewicz Jr | 82cc1a7 | 2008-03-21 03:43:19 -0700 | [diff] [blame] | 218 | * @sk_gso_max_size: Maximum GSO segment size to build |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 219 | * @sk_lingertime: %SO_LINGER l_linger setting |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 220 | * @sk_backlog: always used with the per-socket spinlock held |
| 221 | * @sk_callback_lock: used with the callbacks in the end of this struct |
| 222 | * @sk_error_queue: rarely used |
Wang Chen | 33c732c | 2007-11-13 20:30:01 -0800 | [diff] [blame] | 223 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
| 224 | * IPV6_ADDRFORM for instance) |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 225 | * @sk_err: last error |
Wang Chen | 33c732c | 2007-11-13 20:30:01 -0800 | [diff] [blame] | 226 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
| 227 | * persistent failure not just 'timed out' |
Eric Dumazet | cb61cb9 | 2008-06-17 21:04:56 -0700 | [diff] [blame] | 228 | * @sk_drops: raw/udp drops counter |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 229 | * @sk_ack_backlog: current listen backlog |
| 230 | * @sk_max_ack_backlog: listen backlog set in listen() |
| 231 | * @sk_priority: %SO_PRIORITY setting |
Randy Dunlap | 1a3bc36 | 2012-01-21 09:03:10 +0000 | [diff] [blame] | 232 | * @sk_cgrp_prioidx: socket group's priority map index |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 233 | * @sk_type: socket type (%SOCK_STREAM, etc) |
| 234 | * @sk_protocol: which protocol this socket belongs in this network family |
Randy Dunlap | 53c3fa2 | 2010-08-09 13:41:07 +0000 | [diff] [blame] | 235 | * @sk_peer_pid: &struct pid for this socket's peer |
| 236 | * @sk_peer_cred: %SO_PEERCRED setting |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 237 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
| 238 | * @sk_rcvtimeo: %SO_RCVTIMEO setting |
| 239 | * @sk_sndtimeo: %SO_SNDTIMEO setting |
David S. Miller | c58dc01 | 2010-04-27 15:05:31 -0700 | [diff] [blame] | 240 | * @sk_rxhash: flow hash received from netif layer |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 241 | * @sk_filter: socket filtering instructions |
| 242 | * @sk_protinfo: private area, net family specific, when not using slab |
| 243 | * @sk_timer: sock cleanup timer |
| 244 | * @sk_stamp: time stamp of last packet received |
| 245 | * @sk_socket: Identd and reporting IO signals |
| 246 | * @sk_user_data: RPC layer private data |
| 247 | * @sk_sndmsg_page: cached page for sendmsg |
| 248 | * @sk_sndmsg_off: cached offset for sendmsg |
| 249 | * @sk_send_head: front of stuff to transmit |
Martin Waitz | 67be2dd | 2005-05-01 08:59:26 -0700 | [diff] [blame] | 250 | * @sk_security: used by security modules |
Randy Dunlap | 3172936 | 2008-02-18 20:52:13 -0800 | [diff] [blame] | 251 | * @sk_mark: generic packet mark |
Randy Dunlap | 53c3fa2 | 2010-08-09 13:41:07 +0000 | [diff] [blame] | 252 | * @sk_classid: this socket's cgroup classid |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 253 | * @sk_cgrp: this socket's cgroup-specific proto data |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 254 | * @sk_write_pending: a write to stream socket waits to start |
| 255 | * @sk_state_change: callback to indicate change in the state of the sock |
| 256 | * @sk_data_ready: callback to indicate there is data to be processed |
| 257 | * @sk_write_space: callback to indicate there is bf sending space available |
| 258 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) |
| 259 | * @sk_backlog_rcv: callback to process the backlog |
| 260 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 261 | */ |
| 262 | struct sock { |
| 263 | /* |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 264 | * Now struct inet_timewait_sock also uses sock_common, so please just |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 265 | * don't add nothing before this first member (__sk_common) --acme |
| 266 | */ |
| 267 | struct sock_common __sk_common; |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 268 | #define sk_node __sk_common.skc_node |
| 269 | #define sk_nulls_node __sk_common.skc_nulls_node |
| 270 | #define sk_refcnt __sk_common.skc_refcnt |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 271 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 272 | |
Eric Dumazet | 68835ab | 2010-11-30 19:04:07 +0000 | [diff] [blame] | 273 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
| 274 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 275 | #define sk_hash __sk_common.skc_hash |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 276 | #define sk_family __sk_common.skc_family |
| 277 | #define sk_state __sk_common.skc_state |
| 278 | #define sk_reuse __sk_common.skc_reuse |
| 279 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 280 | #define sk_bind_node __sk_common.skc_bind_node |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 281 | #define sk_prot __sk_common.skc_prot |
Eric W. Biederman | 07feaeb | 2007-09-12 11:58:02 +0200 | [diff] [blame] | 282 | #define sk_net __sk_common.skc_net |
Eric Dumazet | b178bb3 | 2010-11-16 05:56:04 +0000 | [diff] [blame] | 283 | socket_lock_t sk_lock; |
| 284 | struct sk_buff_head sk_receive_queue; |
| 285 | /* |
| 286 | * The backlog queue is special, it is always used with |
| 287 | * the per-socket spinlock held and requires low latency |
| 288 | * access. Therefore we special case it's implementation. |
| 289 | * Note : rmem_alloc is in this structure to fill a hole |
| 290 | * on 64bit arches, not because its logically part of |
| 291 | * backlog. |
| 292 | */ |
| 293 | struct { |
| 294 | atomic_t rmem_alloc; |
| 295 | int len; |
| 296 | struct sk_buff *head; |
| 297 | struct sk_buff *tail; |
| 298 | } sk_backlog; |
| 299 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
| 300 | int sk_forward_alloc; |
| 301 | #ifdef CONFIG_RPS |
| 302 | __u32 sk_rxhash; |
| 303 | #endif |
| 304 | atomic_t sk_drops; |
| 305 | int sk_rcvbuf; |
| 306 | |
| 307 | struct sk_filter __rcu *sk_filter; |
Eric Dumazet | eaefd11 | 2011-02-18 03:26:36 +0000 | [diff] [blame] | 308 | struct socket_wq __rcu *sk_wq; |
Eric Dumazet | b178bb3 | 2010-11-16 05:56:04 +0000 | [diff] [blame] | 309 | |
| 310 | #ifdef CONFIG_NET_DMA |
| 311 | struct sk_buff_head sk_async_wait_queue; |
| 312 | #endif |
| 313 | |
| 314 | #ifdef CONFIG_XFRM |
| 315 | struct xfrm_policy *sk_policy[2]; |
| 316 | #endif |
| 317 | unsigned long sk_flags; |
| 318 | struct dst_entry *sk_dst_cache; |
| 319 | spinlock_t sk_dst_lock; |
| 320 | atomic_t sk_wmem_alloc; |
| 321 | atomic_t sk_omem_alloc; |
| 322 | int sk_sndbuf; |
| 323 | struct sk_buff_head sk_write_queue; |
Vegard Nossum | a98b65a | 2009-02-26 14:46:57 +0100 | [diff] [blame] | 324 | kmemcheck_bitfield_begin(flags); |
Eric Dumazet | 5fdb997 | 2009-10-08 22:50:25 +0000 | [diff] [blame] | 325 | unsigned int sk_shutdown : 2, |
| 326 | sk_no_check : 2, |
| 327 | sk_userlocks : 4, |
| 328 | sk_protocol : 8, |
| 329 | sk_type : 16; |
Vegard Nossum | a98b65a | 2009-02-26 14:46:57 +0100 | [diff] [blame] | 330 | kmemcheck_bitfield_end(flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 331 | int sk_wmem_queued; |
Al Viro | 7d877f3 | 2005-10-21 03:20:43 -0400 | [diff] [blame] | 332 | gfp_t sk_allocation; |
Michał Mirosław | c8f44af | 2011-11-15 15:29:55 +0000 | [diff] [blame] | 333 | netdev_features_t sk_route_caps; |
| 334 | netdev_features_t sk_route_nocaps; |
Herbert Xu | bcd7611 | 2006-06-30 13:36:35 -0700 | [diff] [blame] | 335 | int sk_gso_type; |
Peter P Waskiewicz Jr | 82cc1a7 | 2008-03-21 03:43:19 -0700 | [diff] [blame] | 336 | unsigned int sk_gso_max_size; |
David S. Miller | 9932cf9 | 2006-03-24 15:12:37 -0800 | [diff] [blame] | 337 | int sk_rcvlowat; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 338 | unsigned long sk_lingertime; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 339 | struct sk_buff_head sk_error_queue; |
Arnaldo Carvalho de Melo | 476e19c | 2005-05-05 13:35:15 -0700 | [diff] [blame] | 340 | struct proto *sk_prot_creator; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 341 | rwlock_t sk_callback_lock; |
| 342 | int sk_err, |
| 343 | sk_err_soft; |
| 344 | unsigned short sk_ack_backlog; |
| 345 | unsigned short sk_max_ack_backlog; |
| 346 | __u32 sk_priority; |
Neil Horman | 5bc1421 | 2011-11-22 05:10:51 +0000 | [diff] [blame] | 347 | #ifdef CONFIG_CGROUPS |
| 348 | __u32 sk_cgrp_prioidx; |
| 349 | #endif |
Eric W. Biederman | 109f6e3 | 2010-06-13 03:30:14 +0000 | [diff] [blame] | 350 | struct pid *sk_peer_pid; |
| 351 | const struct cred *sk_peer_cred; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 352 | long sk_rcvtimeo; |
| 353 | long sk_sndtimeo; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 354 | void *sk_protinfo; |
| 355 | struct timer_list sk_timer; |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 356 | ktime_t sk_stamp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 357 | struct socket *sk_socket; |
| 358 | void *sk_user_data; |
| 359 | struct page *sk_sndmsg_page; |
| 360 | struct sk_buff *sk_send_head; |
| 361 | __u32 sk_sndmsg_off; |
Pavel Emelyanov | ef64a54 | 2012-02-21 07:31:34 +0000 | [diff] [blame] | 362 | __s32 sk_peek_off; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 363 | int sk_write_pending; |
Alexey Dobriyan | d5f6423 | 2008-11-04 14:45:58 -0800 | [diff] [blame] | 364 | #ifdef CONFIG_SECURITY |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 365 | void *sk_security; |
Alexey Dobriyan | d5f6423 | 2008-11-04 14:45:58 -0800 | [diff] [blame] | 366 | #endif |
Laszlo Attila Toth | 4a19ec5 | 2008-01-30 19:08:16 -0800 | [diff] [blame] | 367 | __u32 sk_mark; |
Herbert Xu | f845172 | 2010-05-24 00:12:34 -0700 | [diff] [blame] | 368 | u32 sk_classid; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 369 | struct cg_proto *sk_cgrp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 370 | void (*sk_state_change)(struct sock *sk); |
| 371 | void (*sk_data_ready)(struct sock *sk, int bytes); |
| 372 | void (*sk_write_space)(struct sock *sk); |
| 373 | void (*sk_error_report)(struct sock *sk); |
| 374 | int (*sk_backlog_rcv)(struct sock *sk, |
| 375 | struct sk_buff *skb); |
| 376 | void (*sk_destruct)(struct sock *sk); |
| 377 | }; |
| 378 | |
Pavel Emelyanov | ef64a54 | 2012-02-21 07:31:34 +0000 | [diff] [blame] | 379 | static inline int sk_peek_offset(struct sock *sk, int flags) |
| 380 | { |
| 381 | if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0)) |
| 382 | return sk->sk_peek_off; |
| 383 | else |
| 384 | return 0; |
| 385 | } |
| 386 | |
| 387 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) |
| 388 | { |
| 389 | if (sk->sk_peek_off >= 0) { |
| 390 | if (sk->sk_peek_off >= val) |
| 391 | sk->sk_peek_off -= val; |
| 392 | else |
| 393 | sk->sk_peek_off = 0; |
| 394 | } |
| 395 | } |
| 396 | |
| 397 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) |
| 398 | { |
| 399 | if (sk->sk_peek_off >= 0) |
| 400 | sk->sk_peek_off += val; |
| 401 | } |
| 402 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 403 | /* |
| 404 | * Hashed lists helper routines |
| 405 | */ |
Li Zefan | c414664 | 2010-02-08 23:18:45 +0000 | [diff] [blame] | 406 | static inline struct sock *sk_entry(const struct hlist_node *node) |
| 407 | { |
| 408 | return hlist_entry(node, struct sock, sk_node); |
| 409 | } |
| 410 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 411 | static inline struct sock *__sk_head(const struct hlist_head *head) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 412 | { |
| 413 | return hlist_entry(head->first, struct sock, sk_node); |
| 414 | } |
| 415 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 416 | static inline struct sock *sk_head(const struct hlist_head *head) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 417 | { |
| 418 | return hlist_empty(head) ? NULL : __sk_head(head); |
| 419 | } |
| 420 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 421 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
| 422 | { |
| 423 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); |
| 424 | } |
| 425 | |
| 426 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) |
| 427 | { |
| 428 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); |
| 429 | } |
| 430 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 431 | static inline struct sock *sk_next(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 432 | { |
| 433 | return sk->sk_node.next ? |
| 434 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; |
| 435 | } |
| 436 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 437 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
| 438 | { |
| 439 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? |
| 440 | hlist_nulls_entry(sk->sk_nulls_node.next, |
| 441 | struct sock, sk_nulls_node) : |
| 442 | NULL; |
| 443 | } |
| 444 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 445 | static inline int sk_unhashed(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 446 | { |
| 447 | return hlist_unhashed(&sk->sk_node); |
| 448 | } |
| 449 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 450 | static inline int sk_hashed(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 451 | { |
Akinobu Mita | da753be | 2006-04-28 15:21:23 -0700 | [diff] [blame] | 452 | return !sk_unhashed(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 453 | } |
| 454 | |
| 455 | static __inline__ void sk_node_init(struct hlist_node *node) |
| 456 | { |
| 457 | node->pprev = NULL; |
| 458 | } |
| 459 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 460 | static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node) |
| 461 | { |
| 462 | node->pprev = NULL; |
| 463 | } |
| 464 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 465 | static __inline__ void __sk_del_node(struct sock *sk) |
| 466 | { |
| 467 | __hlist_del(&sk->sk_node); |
| 468 | } |
| 469 | |
stephen hemminger | 808f511 | 2010-02-22 07:57:18 +0000 | [diff] [blame] | 470 | /* NB: equivalent to hlist_del_init_rcu */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 471 | static __inline__ int __sk_del_node_init(struct sock *sk) |
| 472 | { |
| 473 | if (sk_hashed(sk)) { |
| 474 | __sk_del_node(sk); |
| 475 | sk_node_init(&sk->sk_node); |
| 476 | return 1; |
| 477 | } |
| 478 | return 0; |
| 479 | } |
| 480 | |
| 481 | /* Grab socket reference count. This operation is valid only |
| 482 | when sk is ALREADY grabbed f.e. it is found in hash table |
| 483 | or a list and the lookup is made under lock preventing hash table |
| 484 | modifications. |
| 485 | */ |
| 486 | |
| 487 | static inline void sock_hold(struct sock *sk) |
| 488 | { |
| 489 | atomic_inc(&sk->sk_refcnt); |
| 490 | } |
| 491 | |
| 492 | /* Ungrab socket in the context, which assumes that socket refcnt |
| 493 | cannot hit zero, f.e. it is true in context of any socketcall. |
| 494 | */ |
| 495 | static inline void __sock_put(struct sock *sk) |
| 496 | { |
| 497 | atomic_dec(&sk->sk_refcnt); |
| 498 | } |
| 499 | |
| 500 | static __inline__ int sk_del_node_init(struct sock *sk) |
| 501 | { |
| 502 | int rc = __sk_del_node_init(sk); |
| 503 | |
| 504 | if (rc) { |
| 505 | /* paranoid for a while -acme */ |
| 506 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
| 507 | __sock_put(sk); |
| 508 | } |
| 509 | return rc; |
| 510 | } |
stephen hemminger | 808f511 | 2010-02-22 07:57:18 +0000 | [diff] [blame] | 511 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 512 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 513 | static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk) |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 514 | { |
| 515 | if (sk_hashed(sk)) { |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 516 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 517 | return 1; |
| 518 | } |
| 519 | return 0; |
| 520 | } |
| 521 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 522 | static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk) |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 523 | { |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 524 | int rc = __sk_nulls_del_node_init_rcu(sk); |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 525 | |
| 526 | if (rc) { |
| 527 | /* paranoid for a while -acme */ |
| 528 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
| 529 | __sock_put(sk); |
| 530 | } |
| 531 | return rc; |
| 532 | } |
| 533 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 534 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) |
| 535 | { |
| 536 | hlist_add_head(&sk->sk_node, list); |
| 537 | } |
| 538 | |
| 539 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) |
| 540 | { |
| 541 | sock_hold(sk); |
| 542 | __sk_add_node(sk, list); |
| 543 | } |
| 544 | |
stephen hemminger | 808f511 | 2010-02-22 07:57:18 +0000 | [diff] [blame] | 545 | static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
| 546 | { |
| 547 | sock_hold(sk); |
| 548 | hlist_add_head_rcu(&sk->sk_node, list); |
| 549 | } |
| 550 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 551 | static __inline__ void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 552 | { |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 553 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 554 | } |
| 555 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 556 | static __inline__ void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 557 | { |
| 558 | sock_hold(sk); |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 559 | __sk_nulls_add_node_rcu(sk, list); |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 560 | } |
| 561 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 562 | static __inline__ void __sk_del_bind_node(struct sock *sk) |
| 563 | { |
| 564 | __hlist_del(&sk->sk_bind_node); |
| 565 | } |
| 566 | |
| 567 | static __inline__ void sk_add_bind_node(struct sock *sk, |
| 568 | struct hlist_head *list) |
| 569 | { |
| 570 | hlist_add_head(&sk->sk_bind_node, list); |
| 571 | } |
| 572 | |
| 573 | #define sk_for_each(__sk, node, list) \ |
| 574 | hlist_for_each_entry(__sk, node, list, sk_node) |
stephen hemminger | 808f511 | 2010-02-22 07:57:18 +0000 | [diff] [blame] | 575 | #define sk_for_each_rcu(__sk, node, list) \ |
| 576 | hlist_for_each_entry_rcu(__sk, node, list, sk_node) |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 577 | #define sk_nulls_for_each(__sk, node, list) \ |
| 578 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) |
| 579 | #define sk_nulls_for_each_rcu(__sk, node, list) \ |
| 580 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 581 | #define sk_for_each_from(__sk, node) \ |
| 582 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ |
| 583 | hlist_for_each_entry_from(__sk, node, sk_node) |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 584 | #define sk_nulls_for_each_from(__sk, node) \ |
| 585 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ |
| 586 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 587 | #define sk_for_each_safe(__sk, node, tmp, list) \ |
| 588 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) |
| 589 | #define sk_for_each_bound(__sk, node, list) \ |
| 590 | hlist_for_each_entry(__sk, node, list, sk_bind_node) |
| 591 | |
| 592 | /* Sock flags */ |
| 593 | enum sock_flags { |
| 594 | SOCK_DEAD, |
| 595 | SOCK_DONE, |
| 596 | SOCK_URGINLINE, |
| 597 | SOCK_KEEPOPEN, |
| 598 | SOCK_LINGER, |
| 599 | SOCK_DESTROY, |
| 600 | SOCK_BROADCAST, |
| 601 | SOCK_TIMESTAMP, |
| 602 | SOCK_ZAPPED, |
| 603 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ |
| 604 | SOCK_DBG, /* %SO_DEBUG setting */ |
| 605 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ |
Eric Dumazet | 92f37fd | 2007-03-25 22:14:49 -0700 | [diff] [blame] | 606 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 607 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
| 608 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 609 | SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */ |
| 610 | SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */ |
| 611 | SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */ |
| 612 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
| 613 | SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */ |
| 614 | SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */ |
| 615 | SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */ |
Eric Dumazet | bcdce71 | 2009-10-06 17:28:29 -0700 | [diff] [blame] | 616 | SOCK_FASYNC, /* fasync() active */ |
Neil Horman | 3b88578 | 2009-10-12 13:26:31 -0700 | [diff] [blame] | 617 | SOCK_RXQ_OVFL, |
Shirley Ma | 1cdebb4 | 2011-07-06 12:17:30 +0000 | [diff] [blame] | 618 | SOCK_ZEROCOPY, /* buffers from userspace */ |
Johannes Berg | 6e3e939 | 2011-11-09 10:15:42 +0100 | [diff] [blame] | 619 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
Ben Greear | 3bdc0eb | 2012-02-11 15:39:30 +0000 | [diff] [blame] | 620 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
| 621 | * Will use last 4 bytes of packet sent from |
| 622 | * user-space instead. |
| 623 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 624 | }; |
| 625 | |
Ralf Baechle | 53b924b | 2005-08-23 10:11:30 -0700 | [diff] [blame] | 626 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
| 627 | { |
| 628 | nsk->sk_flags = osk->sk_flags; |
| 629 | } |
| 630 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 631 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
| 632 | { |
| 633 | __set_bit(flag, &sk->sk_flags); |
| 634 | } |
| 635 | |
| 636 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) |
| 637 | { |
| 638 | __clear_bit(flag, &sk->sk_flags); |
| 639 | } |
| 640 | |
| 641 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) |
| 642 | { |
| 643 | return test_bit(flag, &sk->sk_flags); |
| 644 | } |
| 645 | |
| 646 | static inline void sk_acceptq_removed(struct sock *sk) |
| 647 | { |
| 648 | sk->sk_ack_backlog--; |
| 649 | } |
| 650 | |
| 651 | static inline void sk_acceptq_added(struct sock *sk) |
| 652 | { |
| 653 | sk->sk_ack_backlog++; |
| 654 | } |
| 655 | |
| 656 | static inline int sk_acceptq_is_full(struct sock *sk) |
| 657 | { |
David S. Miller | 64a1465 | 2007-03-06 11:21:05 -0800 | [diff] [blame] | 658 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 659 | } |
| 660 | |
| 661 | /* |
| 662 | * Compute minimal free write space needed to queue new packets. |
| 663 | */ |
| 664 | static inline int sk_stream_min_wspace(struct sock *sk) |
| 665 | { |
Eric Dumazet | 8df09ea | 2007-12-21 03:07:41 -0800 | [diff] [blame] | 666 | return sk->sk_wmem_queued >> 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 667 | } |
| 668 | |
| 669 | static inline int sk_stream_wspace(struct sock *sk) |
| 670 | { |
| 671 | return sk->sk_sndbuf - sk->sk_wmem_queued; |
| 672 | } |
| 673 | |
| 674 | extern void sk_stream_write_space(struct sock *sk); |
| 675 | |
| 676 | static inline int sk_stream_memory_free(struct sock *sk) |
| 677 | { |
| 678 | return sk->sk_wmem_queued < sk->sk_sndbuf; |
| 679 | } |
| 680 | |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 681 | /* OOB backlog add */ |
Zhu Yi | a3a858f | 2010-03-04 18:01:47 +0000 | [diff] [blame] | 682 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
Stephen Hemminger | 9ee6b53 | 2005-11-08 09:39:42 -0800 | [diff] [blame] | 683 | { |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 684 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
| 685 | skb_dst_force(skb); |
| 686 | |
| 687 | if (!sk->sk_backlog.tail) |
| 688 | sk->sk_backlog.head = skb; |
| 689 | else |
Stephen Hemminger | 9ee6b53 | 2005-11-08 09:39:42 -0800 | [diff] [blame] | 690 | sk->sk_backlog.tail->next = skb; |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 691 | |
| 692 | sk->sk_backlog.tail = skb; |
Stephen Hemminger | 9ee6b53 | 2005-11-08 09:39:42 -0800 | [diff] [blame] | 693 | skb->next = NULL; |
| 694 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 695 | |
Eric Dumazet | c377411 | 2010-04-27 15:13:20 -0700 | [diff] [blame] | 696 | /* |
| 697 | * Take into account size of receive queue and backlog queue |
Eric Dumazet | 0fd7bac | 2011-12-21 07:11:44 +0000 | [diff] [blame] | 698 | * Do not take into account this skb truesize, |
| 699 | * to allow even a single big packet to come. |
Eric Dumazet | c377411 | 2010-04-27 15:13:20 -0700 | [diff] [blame] | 700 | */ |
| 701 | static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb) |
| 702 | { |
| 703 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); |
| 704 | |
Eric Dumazet | 0fd7bac | 2011-12-21 07:11:44 +0000 | [diff] [blame] | 705 | return qsize > sk->sk_rcvbuf; |
Eric Dumazet | c377411 | 2010-04-27 15:13:20 -0700 | [diff] [blame] | 706 | } |
| 707 | |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 708 | /* The per-socket spinlock must be held here. */ |
Zhu Yi | 4045635 | 2010-03-07 16:21:39 +0000 | [diff] [blame] | 709 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 710 | { |
Eric Dumazet | c377411 | 2010-04-27 15:13:20 -0700 | [diff] [blame] | 711 | if (sk_rcvqueues_full(sk, skb)) |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 712 | return -ENOBUFS; |
| 713 | |
Zhu Yi | a3a858f | 2010-03-04 18:01:47 +0000 | [diff] [blame] | 714 | __sk_add_backlog(sk, skb); |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 715 | sk->sk_backlog.len += skb->truesize; |
| 716 | return 0; |
| 717 | } |
| 718 | |
Peter Zijlstra | c57943a | 2008-10-07 14:18:42 -0700 | [diff] [blame] | 719 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
| 720 | { |
| 721 | return sk->sk_backlog_rcv(sk, skb); |
| 722 | } |
| 723 | |
David S. Miller | c58dc01 | 2010-04-27 15:05:31 -0700 | [diff] [blame] | 724 | static inline void sock_rps_record_flow(const struct sock *sk) |
| 725 | { |
| 726 | #ifdef CONFIG_RPS |
| 727 | struct rps_sock_flow_table *sock_flow_table; |
| 728 | |
| 729 | rcu_read_lock(); |
| 730 | sock_flow_table = rcu_dereference(rps_sock_flow_table); |
| 731 | rps_record_sock_flow(sock_flow_table, sk->sk_rxhash); |
| 732 | rcu_read_unlock(); |
| 733 | #endif |
| 734 | } |
| 735 | |
| 736 | static inline void sock_rps_reset_flow(const struct sock *sk) |
| 737 | { |
| 738 | #ifdef CONFIG_RPS |
| 739 | struct rps_sock_flow_table *sock_flow_table; |
| 740 | |
| 741 | rcu_read_lock(); |
| 742 | sock_flow_table = rcu_dereference(rps_sock_flow_table); |
| 743 | rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash); |
| 744 | rcu_read_unlock(); |
| 745 | #endif |
| 746 | } |
| 747 | |
Tom Herbert | bdeab99 | 2011-08-14 19:45:55 +0000 | [diff] [blame] | 748 | static inline void sock_rps_save_rxhash(struct sock *sk, |
| 749 | const struct sk_buff *skb) |
David S. Miller | c58dc01 | 2010-04-27 15:05:31 -0700 | [diff] [blame] | 750 | { |
| 751 | #ifdef CONFIG_RPS |
Tom Herbert | bdeab99 | 2011-08-14 19:45:55 +0000 | [diff] [blame] | 752 | if (unlikely(sk->sk_rxhash != skb->rxhash)) { |
David S. Miller | c58dc01 | 2010-04-27 15:05:31 -0700 | [diff] [blame] | 753 | sock_rps_reset_flow(sk); |
Tom Herbert | bdeab99 | 2011-08-14 19:45:55 +0000 | [diff] [blame] | 754 | sk->sk_rxhash = skb->rxhash; |
David S. Miller | c58dc01 | 2010-04-27 15:05:31 -0700 | [diff] [blame] | 755 | } |
| 756 | #endif |
| 757 | } |
| 758 | |
Tom Herbert | bdeab99 | 2011-08-14 19:45:55 +0000 | [diff] [blame] | 759 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
| 760 | { |
| 761 | #ifdef CONFIG_RPS |
| 762 | sock_rps_reset_flow(sk); |
| 763 | sk->sk_rxhash = 0; |
| 764 | #endif |
| 765 | } |
| 766 | |
Stephen Hemminger | cfcabdc | 2007-10-09 01:59:42 -0700 | [diff] [blame] | 767 | #define sk_wait_event(__sk, __timeo, __condition) \ |
| 768 | ({ int __rc; \ |
| 769 | release_sock(__sk); \ |
| 770 | __rc = __condition; \ |
| 771 | if (!__rc) { \ |
| 772 | *(__timeo) = schedule_timeout(*(__timeo)); \ |
| 773 | } \ |
| 774 | lock_sock(__sk); \ |
| 775 | __rc = __condition; \ |
| 776 | __rc; \ |
| 777 | }) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 778 | |
| 779 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
| 780 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); |
| 781 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); |
| 782 | extern int sk_stream_error(struct sock *sk, int flags, int err); |
| 783 | extern void sk_stream_kill_queues(struct sock *sk); |
| 784 | |
| 785 | extern int sk_wait_data(struct sock *sk, long *timeo); |
| 786 | |
Arnaldo Carvalho de Melo | 60236fd | 2005-06-18 22:47:21 -0700 | [diff] [blame] | 787 | struct request_sock_ops; |
Arnaldo Carvalho de Melo | 6d6ee43 | 2005-12-13 23:25:19 -0800 | [diff] [blame] | 788 | struct timewait_sock_ops; |
Arnaldo Carvalho de Melo | ab1e0a1 | 2008-02-03 04:06:04 -0800 | [diff] [blame] | 789 | struct inet_hashinfo; |
Pavel Emelyanov | fc8717b | 2008-03-22 16:56:51 -0700 | [diff] [blame] | 790 | struct raw_hashinfo; |
Paul Gortmaker | de47725 | 2011-05-26 13:46:22 -0400 | [diff] [blame] | 791 | struct module; |
Arnaldo Carvalho de Melo | 2e6599c | 2005-06-18 22:46:52 -0700 | [diff] [blame] | 792 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 793 | /* Networking protocol blocks we attach to sockets. |
| 794 | * socket layer -> transport layer interface |
| 795 | * transport -> network interface is defined by struct inet_proto |
| 796 | */ |
| 797 | struct proto { |
| 798 | void (*close)(struct sock *sk, |
| 799 | long timeout); |
| 800 | int (*connect)(struct sock *sk, |
| 801 | struct sockaddr *uaddr, |
| 802 | int addr_len); |
| 803 | int (*disconnect)(struct sock *sk, int flags); |
| 804 | |
| 805 | struct sock * (*accept) (struct sock *sk, int flags, int *err); |
| 806 | |
| 807 | int (*ioctl)(struct sock *sk, int cmd, |
| 808 | unsigned long arg); |
| 809 | int (*init)(struct sock *sk); |
Brian Haley | 7d06b2e | 2008-06-14 17:04:49 -0700 | [diff] [blame] | 810 | void (*destroy)(struct sock *sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 811 | void (*shutdown)(struct sock *sk, int how); |
| 812 | int (*setsockopt)(struct sock *sk, int level, |
| 813 | int optname, char __user *optval, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 814 | unsigned int optlen); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 815 | int (*getsockopt)(struct sock *sk, int level, |
| 816 | int optname, char __user *optval, |
| 817 | int __user *option); |
Alexey Dobriyan | af01d53 | 2008-08-28 02:53:51 -0700 | [diff] [blame] | 818 | #ifdef CONFIG_COMPAT |
Dmitry Mishin | 3fdadf7 | 2006-03-20 22:45:21 -0800 | [diff] [blame] | 819 | int (*compat_setsockopt)(struct sock *sk, |
| 820 | int level, |
| 821 | int optname, char __user *optval, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 822 | unsigned int optlen); |
Dmitry Mishin | 3fdadf7 | 2006-03-20 22:45:21 -0800 | [diff] [blame] | 823 | int (*compat_getsockopt)(struct sock *sk, |
| 824 | int level, |
| 825 | int optname, char __user *optval, |
| 826 | int __user *option); |
Eric W. Biederman | 709b46e | 2011-01-29 16:15:56 +0000 | [diff] [blame] | 827 | int (*compat_ioctl)(struct sock *sk, |
| 828 | unsigned int cmd, unsigned long arg); |
Alexey Dobriyan | af01d53 | 2008-08-28 02:53:51 -0700 | [diff] [blame] | 829 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 830 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
| 831 | struct msghdr *msg, size_t len); |
| 832 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, |
| 833 | struct msghdr *msg, |
| 834 | size_t len, int noblock, int flags, |
| 835 | int *addr_len); |
| 836 | int (*sendpage)(struct sock *sk, struct page *page, |
| 837 | int offset, size_t size, int flags); |
| 838 | int (*bind)(struct sock *sk, |
| 839 | struct sockaddr *uaddr, int addr_len); |
| 840 | |
| 841 | int (*backlog_rcv) (struct sock *sk, |
| 842 | struct sk_buff *skb); |
| 843 | |
| 844 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
| 845 | void (*hash)(struct sock *sk); |
| 846 | void (*unhash)(struct sock *sk); |
Eric Dumazet | 719f835 | 2010-09-08 05:08:44 +0000 | [diff] [blame] | 847 | void (*rehash)(struct sock *sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 848 | int (*get_port)(struct sock *sk, unsigned short snum); |
Octavian Purdila | fcbdf09 | 2010-12-16 14:26:56 -0800 | [diff] [blame] | 849 | void (*clear_sk)(struct sock *sk, int size); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 850 | |
Eric Dumazet | 286ab3d | 2007-11-05 23:38:39 -0800 | [diff] [blame] | 851 | /* Keeping track of sockets in use */ |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 852 | #ifdef CONFIG_PROC_FS |
Pavel Emelyanov | 13ff3d6 | 2008-03-28 16:38:17 -0700 | [diff] [blame] | 853 | unsigned int inuse_idx; |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 854 | #endif |
Arnaldo Carvalho de Melo | ebb53d7 | 2007-11-21 22:08:50 +0800 | [diff] [blame] | 855 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 856 | /* Memory pressure */ |
Pavel Emelyanov | 5c52ba1 | 2008-07-16 20:28:10 -0700 | [diff] [blame] | 857 | void (*enter_memory_pressure)(struct sock *sk); |
Eric Dumazet | 8d987e5 | 2010-11-09 23:24:26 +0000 | [diff] [blame] | 858 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
Eric Dumazet | 1748376 | 2008-11-25 21:16:35 -0800 | [diff] [blame] | 859 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 860 | /* |
| 861 | * Pressure flag: try to collapse. |
| 862 | * Technical note: it is used by multiple contexts non atomically. |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 863 | * All the __sk_mem_schedule() is of this nature: accounting |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 864 | * is strict, actions are advisory and have some latency. |
| 865 | */ |
| 866 | int *memory_pressure; |
Eric Dumazet | 8d987e5 | 2010-11-09 23:24:26 +0000 | [diff] [blame] | 867 | long *sysctl_mem; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 868 | int *sysctl_wmem; |
| 869 | int *sysctl_rmem; |
| 870 | int max_header; |
Changli Gao | 7ba4291 | 2010-07-10 20:41:55 +0000 | [diff] [blame] | 871 | bool no_autobind; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 872 | |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 873 | struct kmem_cache *slab; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 874 | unsigned int obj_size; |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 875 | int slab_flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 876 | |
Eric Dumazet | dd24c00 | 2008-11-25 21:17:14 -0800 | [diff] [blame] | 877 | struct percpu_counter *orphan_count; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 878 | |
Arnaldo Carvalho de Melo | 60236fd | 2005-06-18 22:47:21 -0700 | [diff] [blame] | 879 | struct request_sock_ops *rsk_prot; |
Arnaldo Carvalho de Melo | 6d6ee43 | 2005-12-13 23:25:19 -0800 | [diff] [blame] | 880 | struct timewait_sock_ops *twsk_prot; |
Arnaldo Carvalho de Melo | 2e6599c | 2005-06-18 22:46:52 -0700 | [diff] [blame] | 881 | |
Pavel Emelyanov | 39d8cda | 2008-03-22 16:50:58 -0700 | [diff] [blame] | 882 | union { |
| 883 | struct inet_hashinfo *hashinfo; |
Eric Dumazet | 645ca70 | 2008-10-29 01:41:45 -0700 | [diff] [blame] | 884 | struct udp_table *udp_table; |
Pavel Emelyanov | fc8717b | 2008-03-22 16:56:51 -0700 | [diff] [blame] | 885 | struct raw_hashinfo *raw_hash; |
Pavel Emelyanov | 39d8cda | 2008-03-22 16:50:58 -0700 | [diff] [blame] | 886 | } h; |
Arnaldo Carvalho de Melo | ab1e0a1 | 2008-02-03 04:06:04 -0800 | [diff] [blame] | 887 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 888 | struct module *owner; |
| 889 | |
| 890 | char name[32]; |
| 891 | |
| 892 | struct list_head node; |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 893 | #ifdef SOCK_REFCNT_DEBUG |
| 894 | atomic_t socks; |
| 895 | #endif |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 896 | #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM |
| 897 | /* |
| 898 | * cgroup specific init/deinit functions. Called once for all |
| 899 | * protocols that implement it, from cgroups populate function. |
| 900 | * This function has to setup any files the protocol want to |
| 901 | * appear in the kmem cgroup filesystem. |
| 902 | */ |
| 903 | int (*init_cgroup)(struct cgroup *cgrp, |
| 904 | struct cgroup_subsys *ss); |
Li Zefan | 761b3ef | 2012-01-31 13:47:36 +0800 | [diff] [blame] | 905 | void (*destroy_cgroup)(struct cgroup *cgrp); |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 906 | struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg); |
| 907 | #endif |
| 908 | }; |
| 909 | |
| 910 | struct cg_proto { |
| 911 | void (*enter_memory_pressure)(struct sock *sk); |
| 912 | struct res_counter *memory_allocated; /* Current allocated memory. */ |
| 913 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
| 914 | int *memory_pressure; |
| 915 | long *sysctl_mem; |
| 916 | /* |
| 917 | * memcg field is used to find which memcg we belong directly |
| 918 | * Each memcg struct can hold more than one cg_proto, so container_of |
| 919 | * won't really cut. |
| 920 | * |
| 921 | * The elegant solution would be having an inverse function to |
| 922 | * proto_cgroup in struct proto, but that means polluting the structure |
| 923 | * for everybody, instead of just for memcg users. |
| 924 | */ |
| 925 | struct mem_cgroup *memcg; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 926 | }; |
| 927 | |
| 928 | extern int proto_register(struct proto *prot, int alloc_slab); |
| 929 | extern void proto_unregister(struct proto *prot); |
| 930 | |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 931 | #ifdef SOCK_REFCNT_DEBUG |
| 932 | static inline void sk_refcnt_debug_inc(struct sock *sk) |
| 933 | { |
| 934 | atomic_inc(&sk->sk_prot->socks); |
| 935 | } |
| 936 | |
| 937 | static inline void sk_refcnt_debug_dec(struct sock *sk) |
| 938 | { |
| 939 | atomic_dec(&sk->sk_prot->socks); |
| 940 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", |
| 941 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); |
| 942 | } |
| 943 | |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 944 | inline void sk_refcnt_debug_release(const struct sock *sk) |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 945 | { |
| 946 | if (atomic_read(&sk->sk_refcnt) != 1) |
| 947 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", |
| 948 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); |
| 949 | } |
| 950 | #else /* SOCK_REFCNT_DEBUG */ |
| 951 | #define sk_refcnt_debug_inc(sk) do { } while (0) |
| 952 | #define sk_refcnt_debug_dec(sk) do { } while (0) |
| 953 | #define sk_refcnt_debug_release(sk) do { } while (0) |
| 954 | #endif /* SOCK_REFCNT_DEBUG */ |
| 955 | |
Glauber Costa | 376be5f | 2012-01-20 04:57:14 +0000 | [diff] [blame] | 956 | #if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET) |
Ingo Molnar | c5905af | 2012-02-24 08:31:31 +0100 | [diff] [blame] | 957 | extern struct static_key memcg_socket_limit_enabled; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 958 | static inline struct cg_proto *parent_cg_proto(struct proto *proto, |
| 959 | struct cg_proto *cg_proto) |
| 960 | { |
| 961 | return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg)); |
| 962 | } |
Ingo Molnar | c5905af | 2012-02-24 08:31:31 +0100 | [diff] [blame] | 963 | #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled) |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 964 | #else |
| 965 | #define mem_cgroup_sockets_enabled 0 |
| 966 | static inline struct cg_proto *parent_cg_proto(struct proto *proto, |
| 967 | struct cg_proto *cg_proto) |
| 968 | { |
| 969 | return NULL; |
| 970 | } |
| 971 | #endif |
| 972 | |
| 973 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 974 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
| 975 | { |
| 976 | return sk->sk_prot->memory_pressure != NULL; |
| 977 | } |
| 978 | |
| 979 | static inline bool sk_under_memory_pressure(const struct sock *sk) |
| 980 | { |
| 981 | if (!sk->sk_prot->memory_pressure) |
| 982 | return false; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 983 | |
| 984 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 985 | return !!*sk->sk_cgrp->memory_pressure; |
| 986 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 987 | return !!*sk->sk_prot->memory_pressure; |
| 988 | } |
| 989 | |
| 990 | static inline void sk_leave_memory_pressure(struct sock *sk) |
| 991 | { |
| 992 | int *memory_pressure = sk->sk_prot->memory_pressure; |
| 993 | |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 994 | if (!memory_pressure) |
| 995 | return; |
| 996 | |
| 997 | if (*memory_pressure) |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 998 | *memory_pressure = 0; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 999 | |
| 1000 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1001 | struct cg_proto *cg_proto = sk->sk_cgrp; |
| 1002 | struct proto *prot = sk->sk_prot; |
| 1003 | |
| 1004 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) |
| 1005 | if (*cg_proto->memory_pressure) |
| 1006 | *cg_proto->memory_pressure = 0; |
| 1007 | } |
| 1008 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1009 | } |
| 1010 | |
| 1011 | static inline void sk_enter_memory_pressure(struct sock *sk) |
| 1012 | { |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1013 | if (!sk->sk_prot->enter_memory_pressure) |
| 1014 | return; |
| 1015 | |
| 1016 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1017 | struct cg_proto *cg_proto = sk->sk_cgrp; |
| 1018 | struct proto *prot = sk->sk_prot; |
| 1019 | |
| 1020 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) |
| 1021 | cg_proto->enter_memory_pressure(sk); |
| 1022 | } |
| 1023 | |
| 1024 | sk->sk_prot->enter_memory_pressure(sk); |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1025 | } |
| 1026 | |
| 1027 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) |
| 1028 | { |
| 1029 | long *prot = sk->sk_prot->sysctl_mem; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1030 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 1031 | prot = sk->sk_cgrp->sysctl_mem; |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1032 | return prot[index]; |
| 1033 | } |
| 1034 | |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1035 | static inline void memcg_memory_allocated_add(struct cg_proto *prot, |
| 1036 | unsigned long amt, |
| 1037 | int *parent_status) |
| 1038 | { |
| 1039 | struct res_counter *fail; |
| 1040 | int ret; |
| 1041 | |
Glauber Costa | 0e90b31 | 2012-01-20 04:57:16 +0000 | [diff] [blame] | 1042 | ret = res_counter_charge_nofail(prot->memory_allocated, |
| 1043 | amt << PAGE_SHIFT, &fail); |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1044 | if (ret < 0) |
| 1045 | *parent_status = OVER_LIMIT; |
| 1046 | } |
| 1047 | |
| 1048 | static inline void memcg_memory_allocated_sub(struct cg_proto *prot, |
| 1049 | unsigned long amt) |
| 1050 | { |
| 1051 | res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT); |
| 1052 | } |
| 1053 | |
| 1054 | static inline u64 memcg_memory_allocated_read(struct cg_proto *prot) |
| 1055 | { |
| 1056 | u64 ret; |
| 1057 | ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE); |
| 1058 | return ret >> PAGE_SHIFT; |
| 1059 | } |
| 1060 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1061 | static inline long |
| 1062 | sk_memory_allocated(const struct sock *sk) |
| 1063 | { |
| 1064 | struct proto *prot = sk->sk_prot; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1065 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 1066 | return memcg_memory_allocated_read(sk->sk_cgrp); |
| 1067 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1068 | return atomic_long_read(prot->memory_allocated); |
| 1069 | } |
| 1070 | |
| 1071 | static inline long |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1072 | sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status) |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1073 | { |
| 1074 | struct proto *prot = sk->sk_prot; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1075 | |
| 1076 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1077 | memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status); |
| 1078 | /* update the root cgroup regardless */ |
| 1079 | atomic_long_add_return(amt, prot->memory_allocated); |
| 1080 | return memcg_memory_allocated_read(sk->sk_cgrp); |
| 1081 | } |
| 1082 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1083 | return atomic_long_add_return(amt, prot->memory_allocated); |
| 1084 | } |
| 1085 | |
| 1086 | static inline void |
Glauber Costa | 0e90b31 | 2012-01-20 04:57:16 +0000 | [diff] [blame] | 1087 | sk_memory_allocated_sub(struct sock *sk, int amt) |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1088 | { |
| 1089 | struct proto *prot = sk->sk_prot; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1090 | |
Glauber Costa | 0e90b31 | 2012-01-20 04:57:16 +0000 | [diff] [blame] | 1091 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1092 | memcg_memory_allocated_sub(sk->sk_cgrp, amt); |
| 1093 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1094 | atomic_long_sub(amt, prot->memory_allocated); |
| 1095 | } |
| 1096 | |
| 1097 | static inline void sk_sockets_allocated_dec(struct sock *sk) |
| 1098 | { |
| 1099 | struct proto *prot = sk->sk_prot; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1100 | |
| 1101 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1102 | struct cg_proto *cg_proto = sk->sk_cgrp; |
| 1103 | |
| 1104 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) |
| 1105 | percpu_counter_dec(cg_proto->sockets_allocated); |
| 1106 | } |
| 1107 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1108 | percpu_counter_dec(prot->sockets_allocated); |
| 1109 | } |
| 1110 | |
| 1111 | static inline void sk_sockets_allocated_inc(struct sock *sk) |
| 1112 | { |
| 1113 | struct proto *prot = sk->sk_prot; |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1114 | |
| 1115 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1116 | struct cg_proto *cg_proto = sk->sk_cgrp; |
| 1117 | |
| 1118 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) |
| 1119 | percpu_counter_inc(cg_proto->sockets_allocated); |
| 1120 | } |
| 1121 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1122 | percpu_counter_inc(prot->sockets_allocated); |
| 1123 | } |
| 1124 | |
| 1125 | static inline int |
| 1126 | sk_sockets_allocated_read_positive(struct sock *sk) |
| 1127 | { |
| 1128 | struct proto *prot = sk->sk_prot; |
| 1129 | |
Glauber Costa | e1aab16 | 2011-12-11 21:47:03 +0000 | [diff] [blame] | 1130 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 1131 | return percpu_counter_sum_positive(sk->sk_cgrp->sockets_allocated); |
| 1132 | |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1133 | return percpu_counter_sum_positive(prot->sockets_allocated); |
| 1134 | } |
| 1135 | |
| 1136 | static inline int |
| 1137 | proto_sockets_allocated_sum_positive(struct proto *prot) |
| 1138 | { |
| 1139 | return percpu_counter_sum_positive(prot->sockets_allocated); |
| 1140 | } |
| 1141 | |
| 1142 | static inline long |
| 1143 | proto_memory_allocated(struct proto *prot) |
| 1144 | { |
| 1145 | return atomic_long_read(prot->memory_allocated); |
| 1146 | } |
| 1147 | |
| 1148 | static inline bool |
| 1149 | proto_memory_pressure(struct proto *prot) |
| 1150 | { |
| 1151 | if (!prot->memory_pressure) |
| 1152 | return false; |
| 1153 | return !!*prot->memory_pressure; |
| 1154 | } |
| 1155 | |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 1156 | |
| 1157 | #ifdef CONFIG_PROC_FS |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1158 | /* Called with local bh disabled */ |
Pavel Emelyanov | c29a0bc | 2008-03-31 19:41:46 -0700 | [diff] [blame] | 1159 | extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
| 1160 | extern int sock_prot_inuse_get(struct net *net, struct proto *proto); |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 1161 | #else |
Pavel Emelyanov | c29a0bc | 2008-03-31 19:41:46 -0700 | [diff] [blame] | 1162 | static void inline sock_prot_inuse_add(struct net *net, struct proto *prot, |
| 1163 | int inc) |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 1164 | { |
| 1165 | } |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 1166 | #endif |
| 1167 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1168 | |
Arnaldo Carvalho de Melo | 614c6cb | 2005-08-09 19:47:37 -0700 | [diff] [blame] | 1169 | /* With per-bucket locks this operation is not-atomic, so that |
| 1170 | * this version is not worse. |
| 1171 | */ |
| 1172 | static inline void __sk_prot_rehash(struct sock *sk) |
| 1173 | { |
| 1174 | sk->sk_prot->unhash(sk); |
| 1175 | sk->sk_prot->hash(sk); |
| 1176 | } |
| 1177 | |
Octavian Purdila | fcbdf09 | 2010-12-16 14:26:56 -0800 | [diff] [blame] | 1178 | void sk_prot_clear_portaddr_nulls(struct sock *sk, int size); |
| 1179 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1180 | /* About 10 seconds */ |
| 1181 | #define SOCK_DESTROY_TIME (10*HZ) |
| 1182 | |
| 1183 | /* Sockets 0-1023 can't be bound to unless you are superuser */ |
| 1184 | #define PROT_SOCK 1024 |
| 1185 | |
| 1186 | #define SHUTDOWN_MASK 3 |
| 1187 | #define RCV_SHUTDOWN 1 |
| 1188 | #define SEND_SHUTDOWN 2 |
| 1189 | |
| 1190 | #define SOCK_SNDBUF_LOCK 1 |
| 1191 | #define SOCK_RCVBUF_LOCK 2 |
| 1192 | #define SOCK_BINDADDR_LOCK 4 |
| 1193 | #define SOCK_BINDPORT_LOCK 8 |
| 1194 | |
| 1195 | /* sock_iocb: used to kick off async processing of socket ios */ |
| 1196 | struct sock_iocb { |
| 1197 | struct list_head list; |
| 1198 | |
| 1199 | int flags; |
| 1200 | int size; |
| 1201 | struct socket *sock; |
| 1202 | struct sock *sk; |
| 1203 | struct scm_cookie *scm; |
| 1204 | struct msghdr *msg, async_msg; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1205 | struct kiocb *kiocb; |
| 1206 | }; |
| 1207 | |
| 1208 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) |
| 1209 | { |
| 1210 | return (struct sock_iocb *)iocb->private; |
| 1211 | } |
| 1212 | |
| 1213 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) |
| 1214 | { |
| 1215 | return si->kiocb; |
| 1216 | } |
| 1217 | |
| 1218 | struct socket_alloc { |
| 1219 | struct socket socket; |
| 1220 | struct inode vfs_inode; |
| 1221 | }; |
| 1222 | |
| 1223 | static inline struct socket *SOCKET_I(struct inode *inode) |
| 1224 | { |
| 1225 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; |
| 1226 | } |
| 1227 | |
| 1228 | static inline struct inode *SOCK_INODE(struct socket *socket) |
| 1229 | { |
| 1230 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; |
| 1231 | } |
| 1232 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1233 | /* |
| 1234 | * Functions for memory accounting |
| 1235 | */ |
| 1236 | extern int __sk_mem_schedule(struct sock *sk, int size, int kind); |
| 1237 | extern void __sk_mem_reclaim(struct sock *sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1238 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1239 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) |
| 1240 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) |
| 1241 | #define SK_MEM_SEND 0 |
| 1242 | #define SK_MEM_RECV 1 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1243 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1244 | static inline int sk_mem_pages(int amt) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1245 | { |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1246 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1247 | } |
| 1248 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1249 | static inline int sk_has_account(struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1250 | { |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1251 | /* return true if protocol supports memory accounting */ |
| 1252 | return !!sk->sk_prot->memory_allocated; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1253 | } |
| 1254 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1255 | static inline int sk_wmem_schedule(struct sock *sk, int size) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1256 | { |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1257 | if (!sk_has_account(sk)) |
| 1258 | return 1; |
Herbert Xu | d80d99d | 2005-09-01 17:48:23 -0700 | [diff] [blame] | 1259 | return size <= sk->sk_forward_alloc || |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1260 | __sk_mem_schedule(sk, size, SK_MEM_SEND); |
| 1261 | } |
| 1262 | |
| 1263 | static inline int sk_rmem_schedule(struct sock *sk, int size) |
| 1264 | { |
| 1265 | if (!sk_has_account(sk)) |
| 1266 | return 1; |
| 1267 | return size <= sk->sk_forward_alloc || |
| 1268 | __sk_mem_schedule(sk, size, SK_MEM_RECV); |
| 1269 | } |
| 1270 | |
| 1271 | static inline void sk_mem_reclaim(struct sock *sk) |
| 1272 | { |
| 1273 | if (!sk_has_account(sk)) |
| 1274 | return; |
| 1275 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) |
| 1276 | __sk_mem_reclaim(sk); |
| 1277 | } |
| 1278 | |
David S. Miller | 9993e7d | 2008-01-10 21:56:38 -0800 | [diff] [blame] | 1279 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
| 1280 | { |
| 1281 | if (!sk_has_account(sk)) |
| 1282 | return; |
| 1283 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) |
| 1284 | __sk_mem_reclaim(sk); |
| 1285 | } |
| 1286 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1287 | static inline void sk_mem_charge(struct sock *sk, int size) |
| 1288 | { |
| 1289 | if (!sk_has_account(sk)) |
| 1290 | return; |
| 1291 | sk->sk_forward_alloc -= size; |
| 1292 | } |
| 1293 | |
| 1294 | static inline void sk_mem_uncharge(struct sock *sk, int size) |
| 1295 | { |
| 1296 | if (!sk_has_account(sk)) |
| 1297 | return; |
| 1298 | sk->sk_forward_alloc += size; |
| 1299 | } |
| 1300 | |
| 1301 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) |
| 1302 | { |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1303 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
| 1304 | sk->sk_wmem_queued -= skb->truesize; |
| 1305 | sk_mem_uncharge(sk, skb->truesize); |
| 1306 | __kfree_skb(skb); |
Herbert Xu | d80d99d | 2005-09-01 17:48:23 -0700 | [diff] [blame] | 1307 | } |
| 1308 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1309 | /* Used by processes to "lock" a socket state, so that |
| 1310 | * interrupts and bottom half handlers won't change it |
| 1311 | * from under us. It essentially blocks any incoming |
| 1312 | * packets, so that we won't get any new data or any |
| 1313 | * packets that change the state of the socket. |
| 1314 | * |
| 1315 | * While locked, BH processing will add new packets to |
| 1316 | * the backlog queue. This queue is processed by the |
| 1317 | * owner of the socket lock right before it is released. |
| 1318 | * |
| 1319 | * Since ~2.3.5 it is also exclusive sleep lock serializing |
| 1320 | * accesses from user process context. |
| 1321 | */ |
John Heffner | d2e9117 | 2007-09-12 10:44:19 +0200 | [diff] [blame] | 1322 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1323 | |
Peter Zijlstra | ed07536 | 2006-12-06 20:35:24 -0800 | [diff] [blame] | 1324 | /* |
| 1325 | * Macro so as to not evaluate some arguments when |
| 1326 | * lockdep is not enabled. |
| 1327 | * |
| 1328 | * Mark both the sk_lock and the sk_lock.slock as a |
| 1329 | * per-address-family lock class. |
| 1330 | */ |
| 1331 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
| 1332 | do { \ |
Ingo Molnar | e8f6fbf | 2008-11-12 01:38:36 +0000 | [diff] [blame] | 1333 | sk->sk_lock.owned = 0; \ |
Peter Zijlstra | ed07536 | 2006-12-06 20:35:24 -0800 | [diff] [blame] | 1334 | init_waitqueue_head(&sk->sk_lock.wq); \ |
| 1335 | spin_lock_init(&(sk)->sk_lock.slock); \ |
| 1336 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ |
| 1337 | sizeof((sk)->sk_lock)); \ |
| 1338 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ |
| 1339 | (skey), (sname)); \ |
| 1340 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
| 1341 | } while (0) |
| 1342 | |
Harvey Harrison | 4138093 | 2007-12-12 10:46:51 -0800 | [diff] [blame] | 1343 | extern void lock_sock_nested(struct sock *sk, int subclass); |
Peter Zijlstra | fcc70d5 | 2006-11-08 22:44:35 -0800 | [diff] [blame] | 1344 | |
| 1345 | static inline void lock_sock(struct sock *sk) |
| 1346 | { |
| 1347 | lock_sock_nested(sk, 0); |
| 1348 | } |
| 1349 | |
Harvey Harrison | 4138093 | 2007-12-12 10:46:51 -0800 | [diff] [blame] | 1350 | extern void release_sock(struct sock *sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1351 | |
| 1352 | /* BH context may only use the following locking interface. */ |
| 1353 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) |
Ingo Molnar | c636618 | 2006-07-03 00:25:13 -0700 | [diff] [blame] | 1354 | #define bh_lock_sock_nested(__sk) \ |
| 1355 | spin_lock_nested(&((__sk)->sk_lock.slock), \ |
| 1356 | SINGLE_DEPTH_NESTING) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1357 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
| 1358 | |
Eric Dumazet | 8a74ad6 | 2010-05-26 19:20:18 +0000 | [diff] [blame] | 1359 | extern bool lock_sock_fast(struct sock *sk); |
| 1360 | /** |
| 1361 | * unlock_sock_fast - complement of lock_sock_fast |
| 1362 | * @sk: socket |
| 1363 | * @slow: slow mode |
| 1364 | * |
| 1365 | * fast unlock socket for user context. |
| 1366 | * If slow mode is on, we call regular release_sock() |
| 1367 | */ |
| 1368 | static inline void unlock_sock_fast(struct sock *sk, bool slow) |
Eric Dumazet | 4b0b72f | 2010-04-28 14:35:48 -0700 | [diff] [blame] | 1369 | { |
Eric Dumazet | 8a74ad6 | 2010-05-26 19:20:18 +0000 | [diff] [blame] | 1370 | if (slow) |
| 1371 | release_sock(sk); |
| 1372 | else |
| 1373 | spin_unlock_bh(&sk->sk_lock.slock); |
Eric Dumazet | 4b0b72f | 2010-04-28 14:35:48 -0700 | [diff] [blame] | 1374 | } |
| 1375 | |
Eric Dumazet | 4b0b72f | 2010-04-28 14:35:48 -0700 | [diff] [blame] | 1376 | |
Eric W. Biederman | 1b8d7ae | 2007-10-08 23:24:22 -0700 | [diff] [blame] | 1377 | extern struct sock *sk_alloc(struct net *net, int family, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1378 | gfp_t priority, |
Pavel Emelyanov | 6257ff2 | 2007-11-01 00:39:31 -0700 | [diff] [blame] | 1379 | struct proto *prot); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1380 | extern void sk_free(struct sock *sk); |
Denis V. Lunev | edf0208 | 2008-02-29 11:18:32 -0800 | [diff] [blame] | 1381 | extern void sk_release_kernel(struct sock *sk); |
Eric Dumazet | e56c57d | 2011-11-08 17:07:07 -0500 | [diff] [blame] | 1382 | extern struct sock *sk_clone_lock(const struct sock *sk, |
| 1383 | const gfp_t priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1384 | |
| 1385 | extern struct sk_buff *sock_wmalloc(struct sock *sk, |
| 1386 | unsigned long size, int force, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1387 | gfp_t priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1388 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
| 1389 | unsigned long size, int force, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1390 | gfp_t priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1391 | extern void sock_wfree(struct sk_buff *skb); |
| 1392 | extern void sock_rfree(struct sk_buff *skb); |
| 1393 | |
| 1394 | extern int sock_setsockopt(struct socket *sock, int level, |
| 1395 | int op, char __user *optval, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 1396 | unsigned int optlen); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1397 | |
| 1398 | extern int sock_getsockopt(struct socket *sock, int level, |
| 1399 | int op, char __user *optval, |
| 1400 | int __user *optlen); |
| 1401 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, |
| 1402 | unsigned long size, |
| 1403 | int noblock, |
| 1404 | int *errcode); |
Herbert Xu | 4cc7f68 | 2009-02-04 16:55:54 -0800 | [diff] [blame] | 1405 | extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk, |
| 1406 | unsigned long header_len, |
| 1407 | unsigned long data_len, |
| 1408 | int noblock, |
| 1409 | int *errcode); |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 1410 | extern void *sock_kmalloc(struct sock *sk, int size, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1411 | gfp_t priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1412 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
| 1413 | extern void sk_send_sigurg(struct sock *sk); |
| 1414 | |
Herbert Xu | f845172 | 2010-05-24 00:12:34 -0700 | [diff] [blame] | 1415 | #ifdef CONFIG_CGROUPS |
| 1416 | extern void sock_update_classid(struct sock *sk); |
| 1417 | #else |
| 1418 | static inline void sock_update_classid(struct sock *sk) |
| 1419 | { |
| 1420 | } |
| 1421 | #endif |
| 1422 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1423 | /* |
| 1424 | * Functions to fill in entries in struct proto_ops when a protocol |
| 1425 | * does not implement a particular function. |
| 1426 | */ |
| 1427 | extern int sock_no_bind(struct socket *, |
| 1428 | struct sockaddr *, int); |
| 1429 | extern int sock_no_connect(struct socket *, |
| 1430 | struct sockaddr *, int, int); |
| 1431 | extern int sock_no_socketpair(struct socket *, |
| 1432 | struct socket *); |
| 1433 | extern int sock_no_accept(struct socket *, |
| 1434 | struct socket *, int); |
| 1435 | extern int sock_no_getname(struct socket *, |
| 1436 | struct sockaddr *, int *, int); |
| 1437 | extern unsigned int sock_no_poll(struct file *, struct socket *, |
| 1438 | struct poll_table_struct *); |
| 1439 | extern int sock_no_ioctl(struct socket *, unsigned int, |
| 1440 | unsigned long); |
| 1441 | extern int sock_no_listen(struct socket *, int); |
| 1442 | extern int sock_no_shutdown(struct socket *, int); |
| 1443 | extern int sock_no_getsockopt(struct socket *, int , int, |
| 1444 | char __user *, int __user *); |
| 1445 | extern int sock_no_setsockopt(struct socket *, int, int, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 1446 | char __user *, unsigned int); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1447 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, |
| 1448 | struct msghdr *, size_t); |
| 1449 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, |
| 1450 | struct msghdr *, size_t, int); |
| 1451 | extern int sock_no_mmap(struct file *file, |
| 1452 | struct socket *sock, |
| 1453 | struct vm_area_struct *vma); |
| 1454 | extern ssize_t sock_no_sendpage(struct socket *sock, |
| 1455 | struct page *page, |
| 1456 | int offset, size_t size, |
| 1457 | int flags); |
| 1458 | |
| 1459 | /* |
| 1460 | * Functions to fill in entries in struct proto_ops when a protocol |
| 1461 | * uses the inet style. |
| 1462 | */ |
| 1463 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, |
| 1464 | char __user *optval, int __user *optlen); |
| 1465 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, |
| 1466 | struct msghdr *msg, size_t size, int flags); |
| 1467 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 1468 | char __user *optval, unsigned int optlen); |
Dmitry Mishin | 3fdadf7 | 2006-03-20 22:45:21 -0800 | [diff] [blame] | 1469 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
| 1470 | int optname, char __user *optval, int __user *optlen); |
| 1471 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 1472 | int optname, char __user *optval, unsigned int optlen); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1473 | |
| 1474 | extern void sk_common_release(struct sock *sk); |
| 1475 | |
| 1476 | /* |
| 1477 | * Default socket callbacks and setup code |
| 1478 | */ |
| 1479 | |
| 1480 | /* Initialise core socket variables */ |
| 1481 | extern void sock_init_data(struct socket *sock, struct sock *sk); |
| 1482 | |
Eric Dumazet | 46bcf14 | 2010-12-06 09:29:43 -0800 | [diff] [blame] | 1483 | extern void sk_filter_release_rcu(struct rcu_head *rcu); |
| 1484 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1485 | /** |
Ben Hutchings | 1a5778a | 2010-02-14 22:35:47 -0800 | [diff] [blame] | 1486 | * sk_filter_release - release a socket filter |
Paul Bonser | dc9b334 | 2006-11-23 17:56:13 -0800 | [diff] [blame] | 1487 | * @fp: filter to remove |
| 1488 | * |
| 1489 | * Remove a filter from a socket and release its resources. |
| 1490 | */ |
| 1491 | |
Pavel Emelyanov | 309dd5f | 2007-10-17 21:21:51 -0700 | [diff] [blame] | 1492 | static inline void sk_filter_release(struct sk_filter *fp) |
| 1493 | { |
| 1494 | if (atomic_dec_and_test(&fp->refcnt)) |
Eric Dumazet | 80f8f10 | 2011-01-18 07:46:52 +0000 | [diff] [blame] | 1495 | call_rcu(&fp->rcu, sk_filter_release_rcu); |
Pavel Emelyanov | 309dd5f | 2007-10-17 21:21:51 -0700 | [diff] [blame] | 1496 | } |
| 1497 | |
| 1498 | static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1499 | { |
| 1500 | unsigned int size = sk_filter_len(fp); |
| 1501 | |
| 1502 | atomic_sub(size, &sk->sk_omem_alloc); |
Pavel Emelyanov | 309dd5f | 2007-10-17 21:21:51 -0700 | [diff] [blame] | 1503 | sk_filter_release(fp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1504 | } |
| 1505 | |
| 1506 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) |
| 1507 | { |
| 1508 | atomic_inc(&fp->refcnt); |
| 1509 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); |
| 1510 | } |
| 1511 | |
| 1512 | /* |
| 1513 | * Socket reference counting postulates. |
| 1514 | * |
| 1515 | * * Each user of socket SHOULD hold a reference count. |
| 1516 | * * Each access point to socket (an hash table bucket, reference from a list, |
| 1517 | * running timer, skb in flight MUST hold a reference count. |
| 1518 | * * When reference count hits 0, it means it will never increase back. |
| 1519 | * * When reference count hits 0, it means that no references from |
| 1520 | * outside exist to this socket and current process on current CPU |
| 1521 | * is last user and may/should destroy this socket. |
| 1522 | * * sk_free is called from any context: process, BH, IRQ. When |
| 1523 | * it is called, socket has no references from outside -> sk_free |
| 1524 | * may release descendant resources allocated by the socket, but |
| 1525 | * to the time when it is called, socket is NOT referenced by any |
| 1526 | * hash tables, lists etc. |
| 1527 | * * Packets, delivered from outside (from network or from another process) |
| 1528 | * and enqueued on receive/error queues SHOULD NOT grab reference count, |
| 1529 | * when they sit in queue. Otherwise, packets will leak to hole, when |
| 1530 | * socket is looked up by one cpu and unhasing is made by another CPU. |
| 1531 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp |
| 1532 | * (leak to backlog). Packet socket does all the processing inside |
| 1533 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets |
| 1534 | * use separate SMP lock, so that they are prone too. |
| 1535 | */ |
| 1536 | |
| 1537 | /* Ungrab socket and destroy it, if it was the last reference. */ |
| 1538 | static inline void sock_put(struct sock *sk) |
| 1539 | { |
| 1540 | if (atomic_dec_and_test(&sk->sk_refcnt)) |
| 1541 | sk_free(sk); |
| 1542 | } |
| 1543 | |
Arnaldo Carvalho de Melo | 58a5a7b | 2006-11-16 14:06:06 -0200 | [diff] [blame] | 1544 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
| 1545 | const int nested); |
Arnaldo Carvalho de Melo | 25995ff | 2005-12-27 02:42:22 -0200 | [diff] [blame] | 1546 | |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 1547 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
| 1548 | { |
| 1549 | sk->sk_tx_queue_mapping = tx_queue; |
| 1550 | } |
| 1551 | |
| 1552 | static inline void sk_tx_queue_clear(struct sock *sk) |
| 1553 | { |
| 1554 | sk->sk_tx_queue_mapping = -1; |
| 1555 | } |
| 1556 | |
| 1557 | static inline int sk_tx_queue_get(const struct sock *sk) |
| 1558 | { |
Tom Herbert | b0f77d0 | 2010-07-14 20:50:29 -0700 | [diff] [blame] | 1559 | return sk ? sk->sk_tx_queue_mapping : -1; |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 1560 | } |
| 1561 | |
David S. Miller | 972692e | 2008-06-17 22:41:38 -0700 | [diff] [blame] | 1562 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
| 1563 | { |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 1564 | sk_tx_queue_clear(sk); |
David S. Miller | 972692e | 2008-06-17 22:41:38 -0700 | [diff] [blame] | 1565 | sk->sk_socket = sock; |
| 1566 | } |
| 1567 | |
Eric Dumazet | aa39514 | 2010-04-20 13:03:51 +0000 | [diff] [blame] | 1568 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
| 1569 | { |
Eric Dumazet | eaefd11 | 2011-02-18 03:26:36 +0000 | [diff] [blame] | 1570 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
| 1571 | return &rcu_dereference_raw(sk->sk_wq)->wait; |
Eric Dumazet | aa39514 | 2010-04-20 13:03:51 +0000 | [diff] [blame] | 1572 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1573 | /* Detach socket from process context. |
| 1574 | * Announce socket dead, detach it from wait queue and inode. |
| 1575 | * Note that parent inode held reference count on this struct sock, |
| 1576 | * we do not release it in this function, because protocol |
| 1577 | * probably wants some additional cleanups or even continuing |
| 1578 | * to work with this socket (TCP). |
| 1579 | */ |
| 1580 | static inline void sock_orphan(struct sock *sk) |
| 1581 | { |
| 1582 | write_lock_bh(&sk->sk_callback_lock); |
| 1583 | sock_set_flag(sk, SOCK_DEAD); |
David S. Miller | 972692e | 2008-06-17 22:41:38 -0700 | [diff] [blame] | 1584 | sk_set_socket(sk, NULL); |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1585 | sk->sk_wq = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1586 | write_unlock_bh(&sk->sk_callback_lock); |
| 1587 | } |
| 1588 | |
| 1589 | static inline void sock_graft(struct sock *sk, struct socket *parent) |
| 1590 | { |
| 1591 | write_lock_bh(&sk->sk_callback_lock); |
Eric Dumazet | eaefd11 | 2011-02-18 03:26:36 +0000 | [diff] [blame] | 1592 | sk->sk_wq = parent->wq; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1593 | parent->sk = sk; |
David S. Miller | 972692e | 2008-06-17 22:41:38 -0700 | [diff] [blame] | 1594 | sk_set_socket(sk, parent); |
Venkat Yekkirala | 4237c75 | 2006-07-24 23:32:50 -0700 | [diff] [blame] | 1595 | security_sock_graft(sk, parent); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1596 | write_unlock_bh(&sk->sk_callback_lock); |
| 1597 | } |
| 1598 | |
| 1599 | extern int sock_i_uid(struct sock *sk); |
| 1600 | extern unsigned long sock_i_ino(struct sock *sk); |
| 1601 | |
| 1602 | static inline struct dst_entry * |
| 1603 | __sk_dst_get(struct sock *sk) |
| 1604 | { |
Michal Hocko | d8bf4ca | 2011-07-08 14:39:41 +0200 | [diff] [blame] | 1605 | return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) || |
Eric Dumazet | f68c224 | 2010-04-22 16:06:59 -0700 | [diff] [blame] | 1606 | lockdep_is_held(&sk->sk_lock.slock)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1607 | } |
| 1608 | |
| 1609 | static inline struct dst_entry * |
| 1610 | sk_dst_get(struct sock *sk) |
| 1611 | { |
| 1612 | struct dst_entry *dst; |
| 1613 | |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1614 | rcu_read_lock(); |
| 1615 | dst = rcu_dereference(sk->sk_dst_cache); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1616 | if (dst) |
| 1617 | dst_hold(dst); |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1618 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1619 | return dst; |
| 1620 | } |
| 1621 | |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1622 | extern void sk_reset_txq(struct sock *sk); |
| 1623 | |
| 1624 | static inline void dst_negative_advice(struct sock *sk) |
| 1625 | { |
| 1626 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); |
| 1627 | |
| 1628 | if (dst && dst->ops->negative_advice) { |
| 1629 | ndst = dst->ops->negative_advice(dst); |
| 1630 | |
| 1631 | if (ndst != dst) { |
| 1632 | rcu_assign_pointer(sk->sk_dst_cache, ndst); |
| 1633 | sk_reset_txq(sk); |
| 1634 | } |
| 1635 | } |
| 1636 | } |
| 1637 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1638 | static inline void |
| 1639 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) |
| 1640 | { |
| 1641 | struct dst_entry *old_dst; |
| 1642 | |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 1643 | sk_tx_queue_clear(sk); |
Eric Dumazet | 0b53ff2 | 2010-04-26 20:40:43 +0000 | [diff] [blame] | 1644 | /* |
| 1645 | * This can be called while sk is owned by the caller only, |
| 1646 | * with no state that can be checked in a rcu_dereference_check() cond |
| 1647 | */ |
| 1648 | old_dst = rcu_dereference_raw(sk->sk_dst_cache); |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1649 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1650 | dst_release(old_dst); |
| 1651 | } |
| 1652 | |
| 1653 | static inline void |
| 1654 | sk_dst_set(struct sock *sk, struct dst_entry *dst) |
| 1655 | { |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1656 | spin_lock(&sk->sk_dst_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1657 | __sk_dst_set(sk, dst); |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1658 | spin_unlock(&sk->sk_dst_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1659 | } |
| 1660 | |
| 1661 | static inline void |
| 1662 | __sk_dst_reset(struct sock *sk) |
| 1663 | { |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1664 | __sk_dst_set(sk, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1665 | } |
| 1666 | |
| 1667 | static inline void |
| 1668 | sk_dst_reset(struct sock *sk) |
| 1669 | { |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1670 | spin_lock(&sk->sk_dst_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1671 | __sk_dst_reset(sk); |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1672 | spin_unlock(&sk->sk_dst_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1673 | } |
| 1674 | |
Denis Vlasenko | f0088a5 | 2006-03-28 01:08:21 -0800 | [diff] [blame] | 1675 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1676 | |
Denis Vlasenko | f0088a5 | 2006-03-28 01:08:21 -0800 | [diff] [blame] | 1677 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1678 | |
Herbert Xu | bcd7611 | 2006-06-30 13:36:35 -0700 | [diff] [blame] | 1679 | static inline int sk_can_gso(const struct sock *sk) |
| 1680 | { |
| 1681 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); |
| 1682 | } |
| 1683 | |
Andi Kleen | 9958089 | 2007-04-20 17:12:43 -0700 | [diff] [blame] | 1684 | extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
Arnaldo Carvalho de Melo | 6cbb0df | 2005-08-09 19:49:02 -0700 | [diff] [blame] | 1685 | |
Michał Mirosław | c8f44af | 2011-11-15 15:29:55 +0000 | [diff] [blame] | 1686 | static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) |
Eric Dumazet | a465419 | 2010-05-16 00:36:33 -0700 | [diff] [blame] | 1687 | { |
| 1688 | sk->sk_route_nocaps |= flags; |
| 1689 | sk->sk_route_caps &= ~flags; |
| 1690 | } |
| 1691 | |
Tom Herbert | c6e1a0d | 2011-04-04 22:30:30 -0700 | [diff] [blame] | 1692 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
| 1693 | char __user *from, char *to, |
Wei Yongjun | 912d398 | 2011-04-06 18:40:12 +0000 | [diff] [blame] | 1694 | int copy, int offset) |
Tom Herbert | c6e1a0d | 2011-04-04 22:30:30 -0700 | [diff] [blame] | 1695 | { |
| 1696 | if (skb->ip_summed == CHECKSUM_NONE) { |
| 1697 | int err = 0; |
| 1698 | __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err); |
| 1699 | if (err) |
| 1700 | return err; |
Wei Yongjun | 912d398 | 2011-04-06 18:40:12 +0000 | [diff] [blame] | 1701 | skb->csum = csum_block_add(skb->csum, csum, offset); |
Tom Herbert | c6e1a0d | 2011-04-04 22:30:30 -0700 | [diff] [blame] | 1702 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
| 1703 | if (!access_ok(VERIFY_READ, from, copy) || |
| 1704 | __copy_from_user_nocache(to, from, copy)) |
| 1705 | return -EFAULT; |
| 1706 | } else if (copy_from_user(to, from, copy)) |
| 1707 | return -EFAULT; |
| 1708 | |
| 1709 | return 0; |
| 1710 | } |
| 1711 | |
| 1712 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, |
| 1713 | char __user *from, int copy) |
| 1714 | { |
Wei Yongjun | 912d398 | 2011-04-06 18:40:12 +0000 | [diff] [blame] | 1715 | int err, offset = skb->len; |
Tom Herbert | c6e1a0d | 2011-04-04 22:30:30 -0700 | [diff] [blame] | 1716 | |
Wei Yongjun | 912d398 | 2011-04-06 18:40:12 +0000 | [diff] [blame] | 1717 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
| 1718 | copy, offset); |
Tom Herbert | c6e1a0d | 2011-04-04 22:30:30 -0700 | [diff] [blame] | 1719 | if (err) |
Wei Yongjun | 912d398 | 2011-04-06 18:40:12 +0000 | [diff] [blame] | 1720 | __skb_trim(skb, offset); |
Tom Herbert | c6e1a0d | 2011-04-04 22:30:30 -0700 | [diff] [blame] | 1721 | |
| 1722 | return err; |
| 1723 | } |
| 1724 | |
| 1725 | static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from, |
| 1726 | struct sk_buff *skb, |
| 1727 | struct page *page, |
| 1728 | int off, int copy) |
| 1729 | { |
| 1730 | int err; |
| 1731 | |
Wei Yongjun | 912d398 | 2011-04-06 18:40:12 +0000 | [diff] [blame] | 1732 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
| 1733 | copy, skb->len); |
Tom Herbert | c6e1a0d | 2011-04-04 22:30:30 -0700 | [diff] [blame] | 1734 | if (err) |
| 1735 | return err; |
| 1736 | |
| 1737 | skb->len += copy; |
| 1738 | skb->data_len += copy; |
| 1739 | skb->truesize += copy; |
| 1740 | sk->sk_wmem_queued += copy; |
| 1741 | sk_mem_charge(sk, copy); |
| 1742 | return 0; |
| 1743 | } |
| 1744 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1745 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, |
| 1746 | struct sk_buff *skb, struct page *page, |
| 1747 | int off, int copy) |
| 1748 | { |
| 1749 | if (skb->ip_summed == CHECKSUM_NONE) { |
| 1750 | int err = 0; |
Al Viro | 5084205 | 2006-11-14 21:36:34 -0800 | [diff] [blame] | 1751 | __wsum csum = csum_and_copy_from_user(from, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1752 | page_address(page) + off, |
| 1753 | copy, 0, &err); |
| 1754 | if (err) |
| 1755 | return err; |
| 1756 | skb->csum = csum_block_add(skb->csum, csum, skb->len); |
| 1757 | } else if (copy_from_user(page_address(page) + off, from, copy)) |
| 1758 | return -EFAULT; |
| 1759 | |
| 1760 | skb->len += copy; |
| 1761 | skb->data_len += copy; |
| 1762 | skb->truesize += copy; |
| 1763 | sk->sk_wmem_queued += copy; |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1764 | sk_mem_charge(sk, copy); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1765 | return 0; |
| 1766 | } |
| 1767 | |
Eric Dumazet | c564039 | 2009-06-16 10:12:03 +0000 | [diff] [blame] | 1768 | /** |
| 1769 | * sk_wmem_alloc_get - returns write allocations |
| 1770 | * @sk: socket |
| 1771 | * |
| 1772 | * Returns sk_wmem_alloc minus initial offset of one |
| 1773 | */ |
| 1774 | static inline int sk_wmem_alloc_get(const struct sock *sk) |
| 1775 | { |
| 1776 | return atomic_read(&sk->sk_wmem_alloc) - 1; |
| 1777 | } |
| 1778 | |
| 1779 | /** |
| 1780 | * sk_rmem_alloc_get - returns read allocations |
| 1781 | * @sk: socket |
| 1782 | * |
| 1783 | * Returns sk_rmem_alloc |
| 1784 | */ |
| 1785 | static inline int sk_rmem_alloc_get(const struct sock *sk) |
| 1786 | { |
| 1787 | return atomic_read(&sk->sk_rmem_alloc); |
| 1788 | } |
| 1789 | |
| 1790 | /** |
| 1791 | * sk_has_allocations - check if allocations are outstanding |
| 1792 | * @sk: socket |
| 1793 | * |
| 1794 | * Returns true if socket has write or read allocations |
| 1795 | */ |
| 1796 | static inline int sk_has_allocations(const struct sock *sk) |
| 1797 | { |
| 1798 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); |
| 1799 | } |
| 1800 | |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1801 | /** |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1802 | * wq_has_sleeper - check if there are any waiting processes |
Randy Dunlap | acfbe96 | 2010-05-24 23:54:18 -0700 | [diff] [blame] | 1803 | * @wq: struct socket_wq |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1804 | * |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1805 | * Returns true if socket_wq has waiting processes |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1806 | * |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1807 | * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1808 | * barrier call. They were added due to the race found within the tcp code. |
| 1809 | * |
| 1810 | * Consider following tcp code paths: |
| 1811 | * |
| 1812 | * CPU1 CPU2 |
| 1813 | * |
| 1814 | * sys_select receive packet |
| 1815 | * ... ... |
| 1816 | * __add_wait_queue update tp->rcv_nxt |
| 1817 | * ... ... |
| 1818 | * tp->rcv_nxt check sock_def_readable |
| 1819 | * ... { |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1820 | * schedule rcu_read_lock(); |
| 1821 | * wq = rcu_dereference(sk->sk_wq); |
| 1822 | * if (wq && waitqueue_active(&wq->wait)) |
| 1823 | * wake_up_interruptible(&wq->wait) |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1824 | * ... |
| 1825 | * } |
| 1826 | * |
| 1827 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay |
| 1828 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 |
| 1829 | * could then endup calling schedule and sleep forever if there are no more |
| 1830 | * data on the socket. |
Jiri Olsa | ad46276 | 2009-07-08 12:10:31 +0000 | [diff] [blame] | 1831 | * |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1832 | */ |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1833 | static inline bool wq_has_sleeper(struct socket_wq *wq) |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1834 | { |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1835 | |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1836 | /* |
| 1837 | * We need to be sure we are in sync with the |
| 1838 | * add_wait_queue modifications to the wait queue. |
| 1839 | * |
| 1840 | * This memory barrier is paired in the sock_poll_wait. |
| 1841 | */ |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1842 | smp_mb(); |
| 1843 | return wq && waitqueue_active(&wq->wait); |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1844 | } |
| 1845 | |
| 1846 | /** |
| 1847 | * sock_poll_wait - place memory barrier behind the poll_wait call. |
| 1848 | * @filp: file |
| 1849 | * @wait_address: socket wait queue |
| 1850 | * @p: poll_table |
| 1851 | * |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1852 | * See the comments in the wq_has_sleeper function. |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1853 | */ |
| 1854 | static inline void sock_poll_wait(struct file *filp, |
| 1855 | wait_queue_head_t *wait_address, poll_table *p) |
| 1856 | { |
Hans Verkuil | 626cf23 | 2012-03-23 15:02:27 -0700 | [diff] [blame] | 1857 | if (!poll_does_not_wait(p) && wait_address) { |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1858 | poll_wait(filp, wait_address, p); |
| 1859 | /* |
| 1860 | * We need to be sure we are in sync with the |
| 1861 | * socket flags modification. |
| 1862 | * |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1863 | * This memory barrier is paired in the wq_has_sleeper. |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1864 | */ |
| 1865 | smp_mb(); |
| 1866 | } |
| 1867 | } |
| 1868 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1869 | /* |
| 1870 | * Queue a received datagram if it will fit. Stream and sequenced |
| 1871 | * protocols can't normally use this as they need to fit buffers in |
| 1872 | * and play with them. |
| 1873 | * |
| 1874 | * Inlined as it's very short and called for pretty much every |
| 1875 | * packet ever received. |
| 1876 | */ |
| 1877 | |
| 1878 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) |
| 1879 | { |
Herbert Xu | d55d87f | 2009-06-22 02:25:25 +0000 | [diff] [blame] | 1880 | skb_orphan(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1881 | skb->sk = sk; |
| 1882 | skb->destructor = sock_wfree; |
Eric Dumazet | 2b85a34 | 2009-06-11 02:55:43 -0700 | [diff] [blame] | 1883 | /* |
| 1884 | * We used to take a refcount on sk, but following operation |
| 1885 | * is enough to guarantee sk_free() wont free this sock until |
| 1886 | * all in-flight packets are completed |
| 1887 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1888 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); |
| 1889 | } |
| 1890 | |
| 1891 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
| 1892 | { |
Herbert Xu | d55d87f | 2009-06-22 02:25:25 +0000 | [diff] [blame] | 1893 | skb_orphan(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1894 | skb->sk = sk; |
| 1895 | skb->destructor = sock_rfree; |
| 1896 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1897 | sk_mem_charge(sk, skb->truesize); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1898 | } |
| 1899 | |
| 1900 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, |
| 1901 | unsigned long expires); |
| 1902 | |
| 1903 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); |
| 1904 | |
Denis Vlasenko | f0088a5 | 2006-03-28 01:08:21 -0800 | [diff] [blame] | 1905 | extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1906 | |
Eric Dumazet | b1faf56 | 2010-05-31 23:44:05 -0700 | [diff] [blame] | 1907 | extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1908 | |
| 1909 | /* |
| 1910 | * Recover an error report and clear atomically |
| 1911 | */ |
| 1912 | |
| 1913 | static inline int sock_error(struct sock *sk) |
| 1914 | { |
Benjamin LaHaise | c1cbe4b | 2005-12-13 23:22:19 -0800 | [diff] [blame] | 1915 | int err; |
| 1916 | if (likely(!sk->sk_err)) |
| 1917 | return 0; |
| 1918 | err = xchg(&sk->sk_err, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1919 | return -err; |
| 1920 | } |
| 1921 | |
| 1922 | static inline unsigned long sock_wspace(struct sock *sk) |
| 1923 | { |
| 1924 | int amt = 0; |
| 1925 | |
| 1926 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { |
| 1927 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); |
| 1928 | if (amt < 0) |
| 1929 | amt = 0; |
| 1930 | } |
| 1931 | return amt; |
| 1932 | } |
| 1933 | |
| 1934 | static inline void sk_wake_async(struct sock *sk, int how, int band) |
| 1935 | { |
Eric Dumazet | bcdce71 | 2009-10-06 17:28:29 -0700 | [diff] [blame] | 1936 | if (sock_flag(sk, SOCK_FASYNC)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1937 | sock_wake_async(sk->sk_socket, how, band); |
| 1938 | } |
| 1939 | |
| 1940 | #define SOCK_MIN_SNDBUF 2048 |
Eric Dumazet | 7a91b434 | 2010-09-26 18:53:07 -0700 | [diff] [blame] | 1941 | /* |
| 1942 | * Since sk_rmem_alloc sums skb->truesize, even a small frame might need |
| 1943 | * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak |
| 1944 | */ |
| 1945 | #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1946 | |
| 1947 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) |
| 1948 | { |
| 1949 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { |
Eric Dumazet | 8df09ea | 2007-12-21 03:07:41 -0800 | [diff] [blame] | 1950 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1951 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
| 1952 | } |
| 1953 | } |
| 1954 | |
Pavel Emelyanov | df97c70 | 2007-11-29 21:22:33 +1100 | [diff] [blame] | 1955 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1956 | |
| 1957 | static inline struct page *sk_stream_alloc_page(struct sock *sk) |
| 1958 | { |
| 1959 | struct page *page = NULL; |
| 1960 | |
Herbert Xu | ef01578 | 2005-09-01 17:48:59 -0700 | [diff] [blame] | 1961 | page = alloc_pages(sk->sk_allocation, 0); |
| 1962 | if (!page) { |
Glauber Costa | 180d8cd | 2011-12-11 21:47:02 +0000 | [diff] [blame] | 1963 | sk_enter_memory_pressure(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1964 | sk_stream_moderate_sndbuf(sk); |
| 1965 | } |
| 1966 | return page; |
| 1967 | } |
| 1968 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1969 | /* |
| 1970 | * Default write policy as shown to user space via poll/select/SIGIO |
| 1971 | */ |
| 1972 | static inline int sock_writeable(const struct sock *sk) |
| 1973 | { |
Eric Dumazet | 8df09ea | 2007-12-21 03:07:41 -0800 | [diff] [blame] | 1974 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1975 | } |
| 1976 | |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1977 | static inline gfp_t gfp_any(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1978 | { |
Andrew Morton | 9970937 | 2009-02-12 16:43:17 -0800 | [diff] [blame] | 1979 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1980 | } |
| 1981 | |
| 1982 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) |
| 1983 | { |
| 1984 | return noblock ? 0 : sk->sk_rcvtimeo; |
| 1985 | } |
| 1986 | |
| 1987 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) |
| 1988 | { |
| 1989 | return noblock ? 0 : sk->sk_sndtimeo; |
| 1990 | } |
| 1991 | |
| 1992 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) |
| 1993 | { |
| 1994 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; |
| 1995 | } |
| 1996 | |
| 1997 | /* Alas, with timeout socket operations are not restartable. |
| 1998 | * Compare this to poll(). |
| 1999 | */ |
| 2000 | static inline int sock_intr_errno(long timeo) |
| 2001 | { |
| 2002 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; |
| 2003 | } |
| 2004 | |
Eric Dumazet | 92f37fd | 2007-03-25 22:14:49 -0700 | [diff] [blame] | 2005 | extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
| 2006 | struct sk_buff *skb); |
Johannes Berg | 6e3e939 | 2011-11-09 10:15:42 +0100 | [diff] [blame] | 2007 | extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, |
| 2008 | struct sk_buff *skb); |
Eric Dumazet | 92f37fd | 2007-03-25 22:14:49 -0700 | [diff] [blame] | 2009 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2010 | static __inline__ void |
| 2011 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
| 2012 | { |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 2013 | ktime_t kt = skb->tstamp; |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 2014 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 2015 | |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 2016 | /* |
| 2017 | * generate control messages if |
| 2018 | * - receive time stamping in software requested (SOCK_RCVTSTAMP |
| 2019 | * or SOCK_TIMESTAMPING_RX_SOFTWARE) |
| 2020 | * - software time stamp available and wanted |
| 2021 | * (SOCK_TIMESTAMPING_SOFTWARE) |
| 2022 | * - hardware time stamps available and wanted |
| 2023 | * (SOCK_TIMESTAMPING_SYS_HARDWARE or |
| 2024 | * SOCK_TIMESTAMPING_RAW_HARDWARE) |
| 2025 | */ |
| 2026 | if (sock_flag(sk, SOCK_RCVTSTAMP) || |
| 2027 | sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) || |
| 2028 | (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) || |
| 2029 | (hwtstamps->hwtstamp.tv64 && |
| 2030 | sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) || |
| 2031 | (hwtstamps->syststamp.tv64 && |
| 2032 | sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))) |
Eric Dumazet | 92f37fd | 2007-03-25 22:14:49 -0700 | [diff] [blame] | 2033 | __sock_recv_timestamp(msg, sk, skb); |
| 2034 | else |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 2035 | sk->sk_stamp = kt; |
Johannes Berg | 6e3e939 | 2011-11-09 10:15:42 +0100 | [diff] [blame] | 2036 | |
| 2037 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) |
| 2038 | __sock_recv_wifi_status(msg, sk, skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2039 | } |
| 2040 | |
Eric Dumazet | 767dd03 | 2010-04-28 19:14:43 +0000 | [diff] [blame] | 2041 | extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
| 2042 | struct sk_buff *skb); |
| 2043 | |
| 2044 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
| 2045 | struct sk_buff *skb) |
| 2046 | { |
| 2047 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ |
| 2048 | (1UL << SOCK_RCVTSTAMP) | \ |
| 2049 | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \ |
| 2050 | (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \ |
| 2051 | (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \ |
| 2052 | (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE)) |
| 2053 | |
| 2054 | if (sk->sk_flags & FLAGS_TS_OR_DROPS) |
| 2055 | __sock_recv_ts_and_drops(msg, sk, skb); |
| 2056 | else |
| 2057 | sk->sk_stamp = skb->tstamp; |
| 2058 | } |
Neil Horman | 3b88578 | 2009-10-12 13:26:31 -0700 | [diff] [blame] | 2059 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2060 | /** |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 2061 | * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 2062 | * @sk: socket sending this packet |
Oliver Hartkopp | 2244d07 | 2010-08-17 08:59:14 +0000 | [diff] [blame] | 2063 | * @tx_flags: filled with instructions for time stamping |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 2064 | * |
| 2065 | * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if |
| 2066 | * parameters are invalid. |
| 2067 | */ |
Oliver Hartkopp | 2244d07 | 2010-08-17 08:59:14 +0000 | [diff] [blame] | 2068 | extern int sock_tx_timestamp(struct sock *sk, __u8 *tx_flags); |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 2069 | |
| 2070 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2071 | * sk_eat_skb - Release a skb if it is no longer needed |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 2072 | * @sk: socket to eat this skb from |
| 2073 | * @skb: socket buffer to eat |
Randy Dunlap | f4b8ea7 | 2006-06-22 16:00:11 -0700 | [diff] [blame] | 2074 | * @copied_early: flag indicating whether DMA operations copied this data early |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2075 | * |
| 2076 | * This routine must be called with interrupts disabled or with the socket |
| 2077 | * locked so that the sk_buff queue operation is ok. |
| 2078 | */ |
Chris Leech | 624d116 | 2006-05-23 18:01:28 -0700 | [diff] [blame] | 2079 | #ifdef CONFIG_NET_DMA |
| 2080 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) |
| 2081 | { |
| 2082 | __skb_unlink(skb, &sk->sk_receive_queue); |
| 2083 | if (!copied_early) |
| 2084 | __kfree_skb(skb); |
| 2085 | else |
| 2086 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); |
| 2087 | } |
| 2088 | #else |
| 2089 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2090 | { |
| 2091 | __skb_unlink(skb, &sk->sk_receive_queue); |
| 2092 | __kfree_skb(skb); |
| 2093 | } |
Chris Leech | 624d116 | 2006-05-23 18:01:28 -0700 | [diff] [blame] | 2094 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2095 | |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 2096 | static inline |
| 2097 | struct net *sock_net(const struct sock *sk) |
| 2098 | { |
Eric Dumazet | c2d9ba9 | 2010-06-01 06:51:19 +0000 | [diff] [blame] | 2099 | return read_pnet(&sk->sk_net); |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 2100 | } |
| 2101 | |
| 2102 | static inline |
Denis V. Lunev | f5aa23f | 2008-03-26 00:48:17 -0700 | [diff] [blame] | 2103 | void sock_net_set(struct sock *sk, struct net *net) |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 2104 | { |
Eric Dumazet | c2d9ba9 | 2010-06-01 06:51:19 +0000 | [diff] [blame] | 2105 | write_pnet(&sk->sk_net, net); |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 2106 | } |
| 2107 | |
Denis V. Lunev | edf0208 | 2008-02-29 11:18:32 -0800 | [diff] [blame] | 2108 | /* |
| 2109 | * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. |
Lucas De Marchi | 25985ed | 2011-03-30 22:57:33 -0300 | [diff] [blame] | 2110 | * They should not hold a reference to a namespace in order to allow |
Denis V. Lunev | edf0208 | 2008-02-29 11:18:32 -0800 | [diff] [blame] | 2111 | * to stop it. |
| 2112 | * Sockets after sk_change_net should be released using sk_release_kernel |
| 2113 | */ |
| 2114 | static inline void sk_change_net(struct sock *sk, struct net *net) |
| 2115 | { |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 2116 | put_net(sock_net(sk)); |
Denis V. Lunev | 65a18ec | 2008-04-16 01:59:46 -0700 | [diff] [blame] | 2117 | sock_net_set(sk, hold_net(net)); |
Denis V. Lunev | edf0208 | 2008-02-29 11:18:32 -0800 | [diff] [blame] | 2118 | } |
| 2119 | |
KOVACS Krisztian | 2354261 | 2008-10-07 12:41:01 -0700 | [diff] [blame] | 2120 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) |
| 2121 | { |
| 2122 | if (unlikely(skb->sk)) { |
| 2123 | struct sock *sk = skb->sk; |
| 2124 | |
| 2125 | skb->destructor = NULL; |
| 2126 | skb->sk = NULL; |
| 2127 | return sk; |
| 2128 | } |
| 2129 | return NULL; |
| 2130 | } |
| 2131 | |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 2132 | extern void sock_enable_timestamp(struct sock *sk, int flag); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2133 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); |
Eric Dumazet | ae40eb1 | 2007-03-18 17:33:16 -0700 | [diff] [blame] | 2134 | extern int sock_get_timestampns(struct sock *, struct timespec __user *); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2135 | |
| 2136 | /* |
| 2137 | * Enable debug/info messages |
| 2138 | */ |
Stephen Hemminger | a2a316f | 2007-03-08 20:41:08 -0800 | [diff] [blame] | 2139 | extern int net_msg_warn; |
| 2140 | #define NETDEBUG(fmt, args...) \ |
| 2141 | do { if (net_msg_warn) printk(fmt,##args); } while (0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2142 | |
Stephen Hemminger | a2a316f | 2007-03-08 20:41:08 -0800 | [diff] [blame] | 2143 | #define LIMIT_NETDEBUG(fmt, args...) \ |
| 2144 | do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2145 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2146 | extern __u32 sysctl_wmem_max; |
| 2147 | extern __u32 sysctl_rmem_max; |
| 2148 | |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 2149 | extern void sk_init(void); |
| 2150 | |
David S. Miller | 6baf1f4 | 2005-09-05 18:14:11 -0700 | [diff] [blame] | 2151 | extern int sysctl_optmem_max; |
| 2152 | |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 2153 | extern __u32 sysctl_wmem_default; |
| 2154 | extern __u32 sysctl_rmem_default; |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 2155 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 2156 | #endif /* _SOCK_H */ |