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 | |
Ilpo Järvinen | 172589c | 2007-08-28 15:50:33 -0700 | [diff] [blame] | 43 | #include <linux/kernel.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 44 | #include <linux/list.h> |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 45 | #include <linux/list_nulls.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 46 | #include <linux/timer.h> |
| 47 | #include <linux/cache.h> |
| 48 | #include <linux/module.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> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 55 | |
| 56 | #include <linux/filter.h> |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 57 | #include <linux/rculist_nulls.h> |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 58 | #include <linux/poll.h> |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 59 | |
| 60 | #include <asm/atomic.h> |
| 61 | #include <net/dst.h> |
| 62 | #include <net/checksum.h> |
| 63 | |
| 64 | /* |
| 65 | * This structure really needs to be cleaned up. |
| 66 | * Most of it is for TCP, and not used by any of |
| 67 | * the other protocols. |
| 68 | */ |
| 69 | |
| 70 | /* Define this to get the SOCK_DBG debugging facility. */ |
| 71 | #define SOCK_DEBUGGING |
| 72 | #ifdef SOCK_DEBUGGING |
| 73 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ |
| 74 | printk(KERN_DEBUG msg); } while (0) |
| 75 | #else |
Stephen Hemminger | 4cd9029 | 2008-03-21 15:54:53 -0700 | [diff] [blame] | 76 | /* Validate arguments and do nothing */ |
Jan Engelhardt | 1183f38 | 2010-05-02 13:42:39 -0700 | [diff] [blame] | 77 | static inline void __attribute__ ((format (printf, 2, 3))) |
Stephen Hemminger | 4cd9029 | 2008-03-21 15:54:53 -0700 | [diff] [blame] | 78 | SOCK_DEBUG(struct sock *sk, const char *msg, ...) |
| 79 | { |
| 80 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 81 | #endif |
| 82 | |
| 83 | /* This is the per-socket lock. The spinlock provides a synchronization |
| 84 | * between user contexts and software interrupt processing, whereas the |
| 85 | * mini-semaphore synchronizes multiple users amongst themselves. |
| 86 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 87 | typedef struct { |
| 88 | spinlock_t slock; |
John Heffner | d2e9117 | 2007-09-12 10:44:19 +0200 | [diff] [blame] | 89 | int owned; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 90 | wait_queue_head_t wq; |
Ingo Molnar | a5b5bb9 | 2006-07-03 00:25:35 -0700 | [diff] [blame] | 91 | /* |
| 92 | * We express the mutex-alike socket_lock semantics |
| 93 | * to the lock validator by explicitly managing |
| 94 | * the slock as a lock variant (in addition to |
| 95 | * the slock itself): |
| 96 | */ |
| 97 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
| 98 | struct lockdep_map dep_map; |
| 99 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 100 | } socket_lock_t; |
| 101 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 102 | struct sock; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 103 | struct proto; |
Denis V. Lunev | 0eeb8ff | 2007-12-04 01:15:45 -0800 | [diff] [blame] | 104 | struct net; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 105 | |
| 106 | /** |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 107 | * struct sock_common - minimal network layer representation of sockets |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 108 | * @skc_node: main hash linkage for various protocol lookup tables |
Eric Dumazet | 512615b | 2009-11-08 10:17:58 +0000 | [diff] [blame] | 109 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 110 | * @skc_refcnt: reference count |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 111 | * @skc_tx_queue_mapping: tx queue number for this connection |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 112 | * @skc_hash: hash value used with various protocol lookup tables |
Eric Dumazet | d4cada4 | 2009-11-08 10:17:30 +0000 | [diff] [blame] | 113 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 114 | * @skc_family: network address family |
| 115 | * @skc_state: Connection state |
| 116 | * @skc_reuse: %SO_REUSEADDR setting |
| 117 | * @skc_bound_dev_if: bound device index if != 0 |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 118 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
Eric Dumazet | 512615b | 2009-11-08 10:17:58 +0000 | [diff] [blame] | 119 | * @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] | 120 | * @skc_prot: protocol handlers inside a network family |
Eric W. Biederman | 07feaeb | 2007-09-12 11:58:02 +0200 | [diff] [blame] | 121 | * @skc_net: reference to the network namespace of this socket |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 122 | * |
| 123 | * 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] | 124 | * for struct sock and struct inet_timewait_sock. |
| 125 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 126 | struct sock_common { |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 127 | /* |
| 128 | * first fields are not copied in sock_copy() |
| 129 | */ |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 130 | union { |
| 131 | struct hlist_node skc_node; |
| 132 | struct hlist_nulls_node skc_nulls_node; |
| 133 | }; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 134 | atomic_t skc_refcnt; |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 135 | int skc_tx_queue_mapping; |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 136 | |
Eric Dumazet | d4cada4 | 2009-11-08 10:17:30 +0000 | [diff] [blame] | 137 | union { |
| 138 | unsigned int skc_hash; |
| 139 | __u16 skc_u16hashes[2]; |
| 140 | }; |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 141 | unsigned short skc_family; |
| 142 | volatile unsigned char skc_state; |
| 143 | unsigned char skc_reuse; |
| 144 | int skc_bound_dev_if; |
Eric Dumazet | 512615b | 2009-11-08 10:17:58 +0000 | [diff] [blame] | 145 | union { |
| 146 | struct hlist_node skc_bind_node; |
| 147 | struct hlist_nulls_node skc_portaddr_node; |
| 148 | }; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 149 | struct proto *skc_prot; |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 150 | #ifdef CONFIG_NET_NS |
Eric W. Biederman | 07feaeb | 2007-09-12 11:58:02 +0200 | [diff] [blame] | 151 | struct net *skc_net; |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 152 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 153 | }; |
| 154 | |
| 155 | /** |
| 156 | * struct sock - network layer representation of sockets |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 157 | * @__sk_common: shared layout with inet_timewait_sock |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 158 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
| 159 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings |
| 160 | * @sk_lock: synchronizer |
| 161 | * @sk_rcvbuf: size of receive buffer in bytes |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 162 | * @sk_wq: sock wait queue and async head |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 163 | * @sk_dst_cache: destination cache |
| 164 | * @sk_dst_lock: destination cache lock |
| 165 | * @sk_policy: flow policy |
| 166 | * @sk_rmem_alloc: receive queue bytes committed |
| 167 | * @sk_receive_queue: incoming packets |
| 168 | * @sk_wmem_alloc: transmit queue bytes committed |
| 169 | * @sk_write_queue: Packet sending queue |
Chris Leech | 97fc2f0 | 2006-05-23 17:55:33 -0700 | [diff] [blame] | 170 | * @sk_async_wait_queue: DMA copied packets |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 171 | * @sk_omem_alloc: "o" is "option" or "other" |
| 172 | * @sk_wmem_queued: persistent queue size |
| 173 | * @sk_forward_alloc: space allocated forward |
| 174 | * @sk_allocation: allocation mode |
| 175 | * @sk_sndbuf: size of send buffer in bytes |
Wang Chen | 33c732c | 2007-11-13 20:30:01 -0800 | [diff] [blame] | 176 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 177 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 178 | * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets |
| 179 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
Eric Dumazet | a465419 | 2010-05-16 00:36:33 -0700 | [diff] [blame] | 180 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
Herbert Xu | bcd7611 | 2006-06-30 13:36:35 -0700 | [diff] [blame] | 181 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
Peter P Waskiewicz Jr | 82cc1a7 | 2008-03-21 03:43:19 -0700 | [diff] [blame] | 182 | * @sk_gso_max_size: Maximum GSO segment size to build |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 183 | * @sk_lingertime: %SO_LINGER l_linger setting |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 184 | * @sk_backlog: always used with the per-socket spinlock held |
| 185 | * @sk_callback_lock: used with the callbacks in the end of this struct |
| 186 | * @sk_error_queue: rarely used |
Wang Chen | 33c732c | 2007-11-13 20:30:01 -0800 | [diff] [blame] | 187 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
| 188 | * IPV6_ADDRFORM for instance) |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 189 | * @sk_err: last error |
Wang Chen | 33c732c | 2007-11-13 20:30:01 -0800 | [diff] [blame] | 190 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
| 191 | * persistent failure not just 'timed out' |
Eric Dumazet | cb61cb9 | 2008-06-17 21:04:56 -0700 | [diff] [blame] | 192 | * @sk_drops: raw/udp drops counter |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 193 | * @sk_ack_backlog: current listen backlog |
| 194 | * @sk_max_ack_backlog: listen backlog set in listen() |
| 195 | * @sk_priority: %SO_PRIORITY setting |
| 196 | * @sk_type: socket type (%SOCK_STREAM, etc) |
| 197 | * @sk_protocol: which protocol this socket belongs in this network family |
| 198 | * @sk_peercred: %SO_PEERCRED setting |
| 199 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
| 200 | * @sk_rcvtimeo: %SO_RCVTIMEO setting |
| 201 | * @sk_sndtimeo: %SO_SNDTIMEO setting |
David S. Miller | c58dc01 | 2010-04-27 15:05:31 -0700 | [diff] [blame] | 202 | * @sk_rxhash: flow hash received from netif layer |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 203 | * @sk_filter: socket filtering instructions |
| 204 | * @sk_protinfo: private area, net family specific, when not using slab |
| 205 | * @sk_timer: sock cleanup timer |
| 206 | * @sk_stamp: time stamp of last packet received |
| 207 | * @sk_socket: Identd and reporting IO signals |
| 208 | * @sk_user_data: RPC layer private data |
| 209 | * @sk_sndmsg_page: cached page for sendmsg |
| 210 | * @sk_sndmsg_off: cached offset for sendmsg |
| 211 | * @sk_send_head: front of stuff to transmit |
Martin Waitz | 67be2dd | 2005-05-01 08:59:26 -0700 | [diff] [blame] | 212 | * @sk_security: used by security modules |
Randy Dunlap | 3172936 | 2008-02-18 20:52:13 -0800 | [diff] [blame] | 213 | * @sk_mark: generic packet mark |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 214 | * @sk_write_pending: a write to stream socket waits to start |
| 215 | * @sk_state_change: callback to indicate change in the state of the sock |
| 216 | * @sk_data_ready: callback to indicate there is data to be processed |
| 217 | * @sk_write_space: callback to indicate there is bf sending space available |
| 218 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) |
| 219 | * @sk_backlog_rcv: callback to process the backlog |
| 220 | * @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] | 221 | */ |
| 222 | struct sock { |
| 223 | /* |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 224 | * Now struct inet_timewait_sock also uses sock_common, so please just |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 225 | * don't add nothing before this first member (__sk_common) --acme |
| 226 | */ |
| 227 | struct sock_common __sk_common; |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 228 | #define sk_node __sk_common.skc_node |
| 229 | #define sk_nulls_node __sk_common.skc_nulls_node |
| 230 | #define sk_refcnt __sk_common.skc_refcnt |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 231 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
Eric Dumazet | 4dc6dc7 | 2009-07-15 23:13:10 +0000 | [diff] [blame] | 232 | |
| 233 | #define sk_copy_start __sk_common.skc_hash |
| 234 | #define sk_hash __sk_common.skc_hash |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 235 | #define sk_family __sk_common.skc_family |
| 236 | #define sk_state __sk_common.skc_state |
| 237 | #define sk_reuse __sk_common.skc_reuse |
| 238 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 239 | #define sk_bind_node __sk_common.skc_bind_node |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 240 | #define sk_prot __sk_common.skc_prot |
Eric W. Biederman | 07feaeb | 2007-09-12 11:58:02 +0200 | [diff] [blame] | 241 | #define sk_net __sk_common.skc_net |
Vegard Nossum | a98b65a | 2009-02-26 14:46:57 +0100 | [diff] [blame] | 242 | kmemcheck_bitfield_begin(flags); |
Eric Dumazet | 5fdb997 | 2009-10-08 22:50:25 +0000 | [diff] [blame] | 243 | unsigned int sk_shutdown : 2, |
| 244 | sk_no_check : 2, |
| 245 | sk_userlocks : 4, |
| 246 | sk_protocol : 8, |
| 247 | sk_type : 16; |
Vegard Nossum | a98b65a | 2009-02-26 14:46:57 +0100 | [diff] [blame] | 248 | kmemcheck_bitfield_end(flags); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 249 | int sk_rcvbuf; |
| 250 | socket_lock_t sk_lock; |
Eric Dumazet | fa438cc | 2007-03-04 16:05:44 -0800 | [diff] [blame] | 251 | /* |
| 252 | * The backlog queue is special, it is always used with |
| 253 | * the per-socket spinlock held and requires low latency |
| 254 | * access. Therefore we special case it's implementation. |
| 255 | */ |
| 256 | struct { |
| 257 | struct sk_buff *head; |
| 258 | struct sk_buff *tail; |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 259 | int len; |
Eric Dumazet | fa438cc | 2007-03-04 16:05:44 -0800 | [diff] [blame] | 260 | } sk_backlog; |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 261 | struct socket_wq *sk_wq; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 262 | struct dst_entry *sk_dst_cache; |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 263 | #ifdef CONFIG_XFRM |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 264 | struct xfrm_policy *sk_policy[2]; |
Alexey Dobriyan | def8b4f | 2008-10-28 13:24:06 -0700 | [diff] [blame] | 265 | #endif |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 266 | spinlock_t sk_dst_lock; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 267 | atomic_t sk_rmem_alloc; |
| 268 | atomic_t sk_wmem_alloc; |
| 269 | atomic_t sk_omem_alloc; |
Arnaldo Carvalho de Melo | 4e07a91 | 2007-05-29 13:17:47 -0700 | [diff] [blame] | 270 | int sk_sndbuf; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 271 | struct sk_buff_head sk_receive_queue; |
| 272 | struct sk_buff_head sk_write_queue; |
Alexey Dobriyan | 2378982 | 2008-11-12 23:25:32 -0800 | [diff] [blame] | 273 | #ifdef CONFIG_NET_DMA |
Chris Leech | 97fc2f0 | 2006-05-23 17:55:33 -0700 | [diff] [blame] | 274 | struct sk_buff_head sk_async_wait_queue; |
Alexey Dobriyan | 2378982 | 2008-11-12 23:25:32 -0800 | [diff] [blame] | 275 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 276 | int sk_wmem_queued; |
| 277 | int sk_forward_alloc; |
Al Viro | 7d877f3 | 2005-10-21 03:20:43 -0400 | [diff] [blame] | 278 | gfp_t sk_allocation; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 279 | int sk_route_caps; |
Eric Dumazet | a465419 | 2010-05-16 00:36:33 -0700 | [diff] [blame] | 280 | int sk_route_nocaps; |
Herbert Xu | bcd7611 | 2006-06-30 13:36:35 -0700 | [diff] [blame] | 281 | int sk_gso_type; |
Peter P Waskiewicz Jr | 82cc1a7 | 2008-03-21 03:43:19 -0700 | [diff] [blame] | 282 | unsigned int sk_gso_max_size; |
David S. Miller | 9932cf9 | 2006-03-24 15:12:37 -0800 | [diff] [blame] | 283 | int sk_rcvlowat; |
David S. Miller | c58dc01 | 2010-04-27 15:05:31 -0700 | [diff] [blame] | 284 | #ifdef CONFIG_RPS |
| 285 | __u32 sk_rxhash; |
| 286 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 287 | unsigned long sk_flags; |
| 288 | unsigned long sk_lingertime; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 289 | struct sk_buff_head sk_error_queue; |
Arnaldo Carvalho de Melo | 476e19c | 2005-05-05 13:35:15 -0700 | [diff] [blame] | 290 | struct proto *sk_prot_creator; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 291 | rwlock_t sk_callback_lock; |
| 292 | int sk_err, |
| 293 | sk_err_soft; |
Wang Chen | 33c732c | 2007-11-13 20:30:01 -0800 | [diff] [blame] | 294 | atomic_t sk_drops; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 295 | unsigned short sk_ack_backlog; |
| 296 | unsigned short sk_max_ack_backlog; |
| 297 | __u32 sk_priority; |
| 298 | struct ucred sk_peercred; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 299 | long sk_rcvtimeo; |
| 300 | long sk_sndtimeo; |
| 301 | struct sk_filter *sk_filter; |
| 302 | void *sk_protinfo; |
| 303 | struct timer_list sk_timer; |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 304 | ktime_t sk_stamp; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 305 | struct socket *sk_socket; |
| 306 | void *sk_user_data; |
| 307 | struct page *sk_sndmsg_page; |
| 308 | struct sk_buff *sk_send_head; |
| 309 | __u32 sk_sndmsg_off; |
| 310 | int sk_write_pending; |
Alexey Dobriyan | d5f6423 | 2008-11-04 14:45:58 -0800 | [diff] [blame] | 311 | #ifdef CONFIG_SECURITY |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 312 | void *sk_security; |
Alexey Dobriyan | d5f6423 | 2008-11-04 14:45:58 -0800 | [diff] [blame] | 313 | #endif |
Laszlo Attila Toth | 4a19ec5 | 2008-01-30 19:08:16 -0800 | [diff] [blame] | 314 | __u32 sk_mark; |
Herbert Xu | f845172 | 2010-05-24 00:12:34 -0700 | [diff] [blame] | 315 | u32 sk_classid; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 316 | void (*sk_state_change)(struct sock *sk); |
| 317 | void (*sk_data_ready)(struct sock *sk, int bytes); |
| 318 | void (*sk_write_space)(struct sock *sk); |
| 319 | void (*sk_error_report)(struct sock *sk); |
| 320 | int (*sk_backlog_rcv)(struct sock *sk, |
| 321 | struct sk_buff *skb); |
| 322 | void (*sk_destruct)(struct sock *sk); |
| 323 | }; |
| 324 | |
| 325 | /* |
| 326 | * Hashed lists helper routines |
| 327 | */ |
Li Zefan | c414664 | 2010-02-08 23:18:45 +0000 | [diff] [blame] | 328 | static inline struct sock *sk_entry(const struct hlist_node *node) |
| 329 | { |
| 330 | return hlist_entry(node, struct sock, sk_node); |
| 331 | } |
| 332 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 333 | static inline struct sock *__sk_head(const struct hlist_head *head) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 334 | { |
| 335 | return hlist_entry(head->first, struct sock, sk_node); |
| 336 | } |
| 337 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 338 | static inline struct sock *sk_head(const struct hlist_head *head) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 339 | { |
| 340 | return hlist_empty(head) ? NULL : __sk_head(head); |
| 341 | } |
| 342 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 343 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
| 344 | { |
| 345 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); |
| 346 | } |
| 347 | |
| 348 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) |
| 349 | { |
| 350 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); |
| 351 | } |
| 352 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 353 | static inline struct sock *sk_next(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 354 | { |
| 355 | return sk->sk_node.next ? |
| 356 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; |
| 357 | } |
| 358 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 359 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
| 360 | { |
| 361 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? |
| 362 | hlist_nulls_entry(sk->sk_nulls_node.next, |
| 363 | struct sock, sk_nulls_node) : |
| 364 | NULL; |
| 365 | } |
| 366 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 367 | static inline int sk_unhashed(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 368 | { |
| 369 | return hlist_unhashed(&sk->sk_node); |
| 370 | } |
| 371 | |
Arnaldo Carvalho de Melo | e48c414 | 2005-08-09 20:09:46 -0700 | [diff] [blame] | 372 | static inline int sk_hashed(const struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 373 | { |
Akinobu Mita | da753be | 2006-04-28 15:21:23 -0700 | [diff] [blame] | 374 | return !sk_unhashed(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 375 | } |
| 376 | |
| 377 | static __inline__ void sk_node_init(struct hlist_node *node) |
| 378 | { |
| 379 | node->pprev = NULL; |
| 380 | } |
| 381 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 382 | static __inline__ void sk_nulls_node_init(struct hlist_nulls_node *node) |
| 383 | { |
| 384 | node->pprev = NULL; |
| 385 | } |
| 386 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 387 | static __inline__ void __sk_del_node(struct sock *sk) |
| 388 | { |
| 389 | __hlist_del(&sk->sk_node); |
| 390 | } |
| 391 | |
stephen hemminger | 808f511 | 2010-02-22 07:57:18 +0000 | [diff] [blame] | 392 | /* NB: equivalent to hlist_del_init_rcu */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 393 | static __inline__ int __sk_del_node_init(struct sock *sk) |
| 394 | { |
| 395 | if (sk_hashed(sk)) { |
| 396 | __sk_del_node(sk); |
| 397 | sk_node_init(&sk->sk_node); |
| 398 | return 1; |
| 399 | } |
| 400 | return 0; |
| 401 | } |
| 402 | |
| 403 | /* Grab socket reference count. This operation is valid only |
| 404 | when sk is ALREADY grabbed f.e. it is found in hash table |
| 405 | or a list and the lookup is made under lock preventing hash table |
| 406 | modifications. |
| 407 | */ |
| 408 | |
| 409 | static inline void sock_hold(struct sock *sk) |
| 410 | { |
| 411 | atomic_inc(&sk->sk_refcnt); |
| 412 | } |
| 413 | |
| 414 | /* Ungrab socket in the context, which assumes that socket refcnt |
| 415 | cannot hit zero, f.e. it is true in context of any socketcall. |
| 416 | */ |
| 417 | static inline void __sock_put(struct sock *sk) |
| 418 | { |
| 419 | atomic_dec(&sk->sk_refcnt); |
| 420 | } |
| 421 | |
| 422 | static __inline__ int sk_del_node_init(struct sock *sk) |
| 423 | { |
| 424 | int rc = __sk_del_node_init(sk); |
| 425 | |
| 426 | if (rc) { |
| 427 | /* paranoid for a while -acme */ |
| 428 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
| 429 | __sock_put(sk); |
| 430 | } |
| 431 | return rc; |
| 432 | } |
stephen hemminger | 808f511 | 2010-02-22 07:57:18 +0000 | [diff] [blame] | 433 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 434 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 435 | static __inline__ int __sk_nulls_del_node_init_rcu(struct sock *sk) |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 436 | { |
| 437 | if (sk_hashed(sk)) { |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 438 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 439 | return 1; |
| 440 | } |
| 441 | return 0; |
| 442 | } |
| 443 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 444 | static __inline__ int sk_nulls_del_node_init_rcu(struct sock *sk) |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 445 | { |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 446 | int rc = __sk_nulls_del_node_init_rcu(sk); |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 447 | |
| 448 | if (rc) { |
| 449 | /* paranoid for a while -acme */ |
| 450 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
| 451 | __sock_put(sk); |
| 452 | } |
| 453 | return rc; |
| 454 | } |
| 455 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 456 | static __inline__ void __sk_add_node(struct sock *sk, struct hlist_head *list) |
| 457 | { |
| 458 | hlist_add_head(&sk->sk_node, list); |
| 459 | } |
| 460 | |
| 461 | static __inline__ void sk_add_node(struct sock *sk, struct hlist_head *list) |
| 462 | { |
| 463 | sock_hold(sk); |
| 464 | __sk_add_node(sk, list); |
| 465 | } |
| 466 | |
stephen hemminger | 808f511 | 2010-02-22 07:57:18 +0000 | [diff] [blame] | 467 | static __inline__ void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
| 468 | { |
| 469 | sock_hold(sk); |
| 470 | hlist_add_head_rcu(&sk->sk_node, list); |
| 471 | } |
| 472 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 473 | 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] | 474 | { |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 475 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 476 | } |
| 477 | |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 478 | 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] | 479 | { |
| 480 | sock_hold(sk); |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 481 | __sk_nulls_add_node_rcu(sk, list); |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 482 | } |
| 483 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 484 | static __inline__ void __sk_del_bind_node(struct sock *sk) |
| 485 | { |
| 486 | __hlist_del(&sk->sk_bind_node); |
| 487 | } |
| 488 | |
| 489 | static __inline__ void sk_add_bind_node(struct sock *sk, |
| 490 | struct hlist_head *list) |
| 491 | { |
| 492 | hlist_add_head(&sk->sk_bind_node, list); |
| 493 | } |
| 494 | |
| 495 | #define sk_for_each(__sk, node, list) \ |
| 496 | hlist_for_each_entry(__sk, node, list, sk_node) |
stephen hemminger | 808f511 | 2010-02-22 07:57:18 +0000 | [diff] [blame] | 497 | #define sk_for_each_rcu(__sk, node, list) \ |
| 498 | hlist_for_each_entry_rcu(__sk, node, list, sk_node) |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 499 | #define sk_nulls_for_each(__sk, node, list) \ |
| 500 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) |
| 501 | #define sk_nulls_for_each_rcu(__sk, node, list) \ |
| 502 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 503 | #define sk_for_each_from(__sk, node) \ |
| 504 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ |
| 505 | hlist_for_each_entry_from(__sk, node, sk_node) |
Eric Dumazet | 88ab193 | 2008-11-16 19:39:21 -0800 | [diff] [blame] | 506 | #define sk_nulls_for_each_from(__sk, node) \ |
| 507 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ |
| 508 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 509 | #define sk_for_each_continue(__sk, node) \ |
| 510 | if (__sk && ({ node = &(__sk)->sk_node; 1; })) \ |
| 511 | hlist_for_each_entry_continue(__sk, node, sk_node) |
| 512 | #define sk_for_each_safe(__sk, node, tmp, list) \ |
| 513 | hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node) |
| 514 | #define sk_for_each_bound(__sk, node, list) \ |
| 515 | hlist_for_each_entry(__sk, node, list, sk_bind_node) |
| 516 | |
| 517 | /* Sock flags */ |
| 518 | enum sock_flags { |
| 519 | SOCK_DEAD, |
| 520 | SOCK_DONE, |
| 521 | SOCK_URGINLINE, |
| 522 | SOCK_KEEPOPEN, |
| 523 | SOCK_LINGER, |
| 524 | SOCK_DESTROY, |
| 525 | SOCK_BROADCAST, |
| 526 | SOCK_TIMESTAMP, |
| 527 | SOCK_ZAPPED, |
| 528 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ |
| 529 | SOCK_DBG, /* %SO_DEBUG setting */ |
| 530 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ |
Eric Dumazet | 92f37fd | 2007-03-25 22:14:49 -0700 | [diff] [blame] | 531 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 532 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
| 533 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 534 | SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */ |
| 535 | SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */ |
| 536 | SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */ |
| 537 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
| 538 | SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */ |
| 539 | SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */ |
| 540 | SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */ |
Eric Dumazet | bcdce71 | 2009-10-06 17:28:29 -0700 | [diff] [blame] | 541 | SOCK_FASYNC, /* fasync() active */ |
Neil Horman | 3b88578 | 2009-10-12 13:26:31 -0700 | [diff] [blame] | 542 | SOCK_RXQ_OVFL, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 543 | }; |
| 544 | |
Ralf Baechle | 53b924b | 2005-08-23 10:11:30 -0700 | [diff] [blame] | 545 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
| 546 | { |
| 547 | nsk->sk_flags = osk->sk_flags; |
| 548 | } |
| 549 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 550 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
| 551 | { |
| 552 | __set_bit(flag, &sk->sk_flags); |
| 553 | } |
| 554 | |
| 555 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) |
| 556 | { |
| 557 | __clear_bit(flag, &sk->sk_flags); |
| 558 | } |
| 559 | |
| 560 | static inline int sock_flag(struct sock *sk, enum sock_flags flag) |
| 561 | { |
| 562 | return test_bit(flag, &sk->sk_flags); |
| 563 | } |
| 564 | |
| 565 | static inline void sk_acceptq_removed(struct sock *sk) |
| 566 | { |
| 567 | sk->sk_ack_backlog--; |
| 568 | } |
| 569 | |
| 570 | static inline void sk_acceptq_added(struct sock *sk) |
| 571 | { |
| 572 | sk->sk_ack_backlog++; |
| 573 | } |
| 574 | |
| 575 | static inline int sk_acceptq_is_full(struct sock *sk) |
| 576 | { |
David S. Miller | 64a1465 | 2007-03-06 11:21:05 -0800 | [diff] [blame] | 577 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 578 | } |
| 579 | |
| 580 | /* |
| 581 | * Compute minimal free write space needed to queue new packets. |
| 582 | */ |
| 583 | static inline int sk_stream_min_wspace(struct sock *sk) |
| 584 | { |
Eric Dumazet | 8df09ea | 2007-12-21 03:07:41 -0800 | [diff] [blame] | 585 | return sk->sk_wmem_queued >> 1; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 586 | } |
| 587 | |
| 588 | static inline int sk_stream_wspace(struct sock *sk) |
| 589 | { |
| 590 | return sk->sk_sndbuf - sk->sk_wmem_queued; |
| 591 | } |
| 592 | |
| 593 | extern void sk_stream_write_space(struct sock *sk); |
| 594 | |
| 595 | static inline int sk_stream_memory_free(struct sock *sk) |
| 596 | { |
| 597 | return sk->sk_wmem_queued < sk->sk_sndbuf; |
| 598 | } |
| 599 | |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 600 | /* OOB backlog add */ |
Zhu Yi | a3a858f | 2010-03-04 18:01:47 +0000 | [diff] [blame] | 601 | 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] | 602 | { |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 603 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
| 604 | skb_dst_force(skb); |
| 605 | |
| 606 | if (!sk->sk_backlog.tail) |
| 607 | sk->sk_backlog.head = skb; |
| 608 | else |
Stephen Hemminger | 9ee6b53 | 2005-11-08 09:39:42 -0800 | [diff] [blame] | 609 | sk->sk_backlog.tail->next = skb; |
Eric Dumazet | 7fee226 | 2010-05-11 23:19:48 +0000 | [diff] [blame] | 610 | |
| 611 | sk->sk_backlog.tail = skb; |
Stephen Hemminger | 9ee6b53 | 2005-11-08 09:39:42 -0800 | [diff] [blame] | 612 | skb->next = NULL; |
| 613 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 614 | |
Eric Dumazet | c377411 | 2010-04-27 15:13:20 -0700 | [diff] [blame] | 615 | /* |
| 616 | * Take into account size of receive queue and backlog queue |
| 617 | */ |
| 618 | static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb) |
| 619 | { |
| 620 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); |
| 621 | |
| 622 | return qsize + skb->truesize > sk->sk_rcvbuf; |
| 623 | } |
| 624 | |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 625 | /* The per-socket spinlock must be held here. */ |
Zhu Yi | 4045635 | 2010-03-07 16:21:39 +0000 | [diff] [blame] | 626 | 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] | 627 | { |
Eric Dumazet | c377411 | 2010-04-27 15:13:20 -0700 | [diff] [blame] | 628 | if (sk_rcvqueues_full(sk, skb)) |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 629 | return -ENOBUFS; |
| 630 | |
Zhu Yi | a3a858f | 2010-03-04 18:01:47 +0000 | [diff] [blame] | 631 | __sk_add_backlog(sk, skb); |
Zhu Yi | 8eae939 | 2010-03-04 18:01:40 +0000 | [diff] [blame] | 632 | sk->sk_backlog.len += skb->truesize; |
| 633 | return 0; |
| 634 | } |
| 635 | |
Peter Zijlstra | c57943a | 2008-10-07 14:18:42 -0700 | [diff] [blame] | 636 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
| 637 | { |
| 638 | return sk->sk_backlog_rcv(sk, skb); |
| 639 | } |
| 640 | |
David S. Miller | c58dc01 | 2010-04-27 15:05:31 -0700 | [diff] [blame] | 641 | static inline void sock_rps_record_flow(const struct sock *sk) |
| 642 | { |
| 643 | #ifdef CONFIG_RPS |
| 644 | struct rps_sock_flow_table *sock_flow_table; |
| 645 | |
| 646 | rcu_read_lock(); |
| 647 | sock_flow_table = rcu_dereference(rps_sock_flow_table); |
| 648 | rps_record_sock_flow(sock_flow_table, sk->sk_rxhash); |
| 649 | rcu_read_unlock(); |
| 650 | #endif |
| 651 | } |
| 652 | |
| 653 | static inline void sock_rps_reset_flow(const struct sock *sk) |
| 654 | { |
| 655 | #ifdef CONFIG_RPS |
| 656 | struct rps_sock_flow_table *sock_flow_table; |
| 657 | |
| 658 | rcu_read_lock(); |
| 659 | sock_flow_table = rcu_dereference(rps_sock_flow_table); |
| 660 | rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash); |
| 661 | rcu_read_unlock(); |
| 662 | #endif |
| 663 | } |
| 664 | |
| 665 | static inline void sock_rps_save_rxhash(struct sock *sk, u32 rxhash) |
| 666 | { |
| 667 | #ifdef CONFIG_RPS |
| 668 | if (unlikely(sk->sk_rxhash != rxhash)) { |
| 669 | sock_rps_reset_flow(sk); |
| 670 | sk->sk_rxhash = rxhash; |
| 671 | } |
| 672 | #endif |
| 673 | } |
| 674 | |
Stephen Hemminger | cfcabdc | 2007-10-09 01:59:42 -0700 | [diff] [blame] | 675 | #define sk_wait_event(__sk, __timeo, __condition) \ |
| 676 | ({ int __rc; \ |
| 677 | release_sock(__sk); \ |
| 678 | __rc = __condition; \ |
| 679 | if (!__rc) { \ |
| 680 | *(__timeo) = schedule_timeout(*(__timeo)); \ |
| 681 | } \ |
| 682 | lock_sock(__sk); \ |
| 683 | __rc = __condition; \ |
| 684 | __rc; \ |
| 685 | }) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 686 | |
| 687 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
| 688 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); |
| 689 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); |
| 690 | extern int sk_stream_error(struct sock *sk, int flags, int err); |
| 691 | extern void sk_stream_kill_queues(struct sock *sk); |
| 692 | |
| 693 | extern int sk_wait_data(struct sock *sk, long *timeo); |
| 694 | |
Arnaldo Carvalho de Melo | 60236fd | 2005-06-18 22:47:21 -0700 | [diff] [blame] | 695 | struct request_sock_ops; |
Arnaldo Carvalho de Melo | 6d6ee43 | 2005-12-13 23:25:19 -0800 | [diff] [blame] | 696 | struct timewait_sock_ops; |
Arnaldo Carvalho de Melo | ab1e0a1 | 2008-02-03 04:06:04 -0800 | [diff] [blame] | 697 | struct inet_hashinfo; |
Pavel Emelyanov | fc8717b | 2008-03-22 16:56:51 -0700 | [diff] [blame] | 698 | struct raw_hashinfo; |
Arnaldo Carvalho de Melo | 2e6599c | 2005-06-18 22:46:52 -0700 | [diff] [blame] | 699 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 700 | /* Networking protocol blocks we attach to sockets. |
| 701 | * socket layer -> transport layer interface |
| 702 | * transport -> network interface is defined by struct inet_proto |
| 703 | */ |
| 704 | struct proto { |
| 705 | void (*close)(struct sock *sk, |
| 706 | long timeout); |
| 707 | int (*connect)(struct sock *sk, |
| 708 | struct sockaddr *uaddr, |
| 709 | int addr_len); |
| 710 | int (*disconnect)(struct sock *sk, int flags); |
| 711 | |
| 712 | struct sock * (*accept) (struct sock *sk, int flags, int *err); |
| 713 | |
| 714 | int (*ioctl)(struct sock *sk, int cmd, |
| 715 | unsigned long arg); |
| 716 | int (*init)(struct sock *sk); |
Brian Haley | 7d06b2e | 2008-06-14 17:04:49 -0700 | [diff] [blame] | 717 | void (*destroy)(struct sock *sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 718 | void (*shutdown)(struct sock *sk, int how); |
| 719 | int (*setsockopt)(struct sock *sk, int level, |
| 720 | int optname, char __user *optval, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 721 | unsigned int optlen); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 722 | int (*getsockopt)(struct sock *sk, int level, |
| 723 | int optname, char __user *optval, |
| 724 | int __user *option); |
Alexey Dobriyan | af01d53 | 2008-08-28 02:53:51 -0700 | [diff] [blame] | 725 | #ifdef CONFIG_COMPAT |
Dmitry Mishin | 3fdadf7 | 2006-03-20 22:45:21 -0800 | [diff] [blame] | 726 | int (*compat_setsockopt)(struct sock *sk, |
| 727 | int level, |
| 728 | int optname, char __user *optval, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 729 | unsigned int optlen); |
Dmitry Mishin | 3fdadf7 | 2006-03-20 22:45:21 -0800 | [diff] [blame] | 730 | int (*compat_getsockopt)(struct sock *sk, |
| 731 | int level, |
| 732 | int optname, char __user *optval, |
| 733 | int __user *option); |
Alexey Dobriyan | af01d53 | 2008-08-28 02:53:51 -0700 | [diff] [blame] | 734 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 735 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
| 736 | struct msghdr *msg, size_t len); |
| 737 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, |
| 738 | struct msghdr *msg, |
| 739 | size_t len, int noblock, int flags, |
| 740 | int *addr_len); |
| 741 | int (*sendpage)(struct sock *sk, struct page *page, |
| 742 | int offset, size_t size, int flags); |
| 743 | int (*bind)(struct sock *sk, |
| 744 | struct sockaddr *uaddr, int addr_len); |
| 745 | |
| 746 | int (*backlog_rcv) (struct sock *sk, |
| 747 | struct sk_buff *skb); |
| 748 | |
| 749 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
| 750 | void (*hash)(struct sock *sk); |
| 751 | void (*unhash)(struct sock *sk); |
| 752 | int (*get_port)(struct sock *sk, unsigned short snum); |
| 753 | |
Eric Dumazet | 286ab3d | 2007-11-05 23:38:39 -0800 | [diff] [blame] | 754 | /* Keeping track of sockets in use */ |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 755 | #ifdef CONFIG_PROC_FS |
Pavel Emelyanov | 13ff3d6 | 2008-03-28 16:38:17 -0700 | [diff] [blame] | 756 | unsigned int inuse_idx; |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 757 | #endif |
Arnaldo Carvalho de Melo | ebb53d7 | 2007-11-21 22:08:50 +0800 | [diff] [blame] | 758 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 759 | /* Memory pressure */ |
Pavel Emelyanov | 5c52ba1 | 2008-07-16 20:28:10 -0700 | [diff] [blame] | 760 | void (*enter_memory_pressure)(struct sock *sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 761 | atomic_t *memory_allocated; /* Current allocated memory. */ |
Eric Dumazet | 1748376 | 2008-11-25 21:16:35 -0800 | [diff] [blame] | 762 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 763 | /* |
| 764 | * Pressure flag: try to collapse. |
| 765 | * Technical note: it is used by multiple contexts non atomically. |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 766 | * All the __sk_mem_schedule() is of this nature: accounting |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 767 | * is strict, actions are advisory and have some latency. |
| 768 | */ |
| 769 | int *memory_pressure; |
| 770 | int *sysctl_mem; |
| 771 | int *sysctl_wmem; |
| 772 | int *sysctl_rmem; |
| 773 | int max_header; |
| 774 | |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 775 | struct kmem_cache *slab; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 776 | unsigned int obj_size; |
Eric Dumazet | 271b72c | 2008-10-29 02:11:14 -0700 | [diff] [blame] | 777 | int slab_flags; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 778 | |
Eric Dumazet | dd24c00 | 2008-11-25 21:17:14 -0800 | [diff] [blame] | 779 | struct percpu_counter *orphan_count; |
Arnaldo Carvalho de Melo | 8feaf0c0 | 2005-08-09 20:09:30 -0700 | [diff] [blame] | 780 | |
Arnaldo Carvalho de Melo | 60236fd | 2005-06-18 22:47:21 -0700 | [diff] [blame] | 781 | struct request_sock_ops *rsk_prot; |
Arnaldo Carvalho de Melo | 6d6ee43 | 2005-12-13 23:25:19 -0800 | [diff] [blame] | 782 | struct timewait_sock_ops *twsk_prot; |
Arnaldo Carvalho de Melo | 2e6599c | 2005-06-18 22:46:52 -0700 | [diff] [blame] | 783 | |
Pavel Emelyanov | 39d8cda | 2008-03-22 16:50:58 -0700 | [diff] [blame] | 784 | union { |
| 785 | struct inet_hashinfo *hashinfo; |
Eric Dumazet | 645ca70 | 2008-10-29 01:41:45 -0700 | [diff] [blame] | 786 | struct udp_table *udp_table; |
Pavel Emelyanov | fc8717b | 2008-03-22 16:56:51 -0700 | [diff] [blame] | 787 | struct raw_hashinfo *raw_hash; |
Pavel Emelyanov | 39d8cda | 2008-03-22 16:50:58 -0700 | [diff] [blame] | 788 | } h; |
Arnaldo Carvalho de Melo | ab1e0a1 | 2008-02-03 04:06:04 -0800 | [diff] [blame] | 789 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 790 | struct module *owner; |
| 791 | |
| 792 | char name[32]; |
| 793 | |
| 794 | struct list_head node; |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 795 | #ifdef SOCK_REFCNT_DEBUG |
| 796 | atomic_t socks; |
| 797 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 798 | }; |
| 799 | |
| 800 | extern int proto_register(struct proto *prot, int alloc_slab); |
| 801 | extern void proto_unregister(struct proto *prot); |
| 802 | |
Arnaldo Carvalho de Melo | e684897 | 2005-08-09 19:45:38 -0700 | [diff] [blame] | 803 | #ifdef SOCK_REFCNT_DEBUG |
| 804 | static inline void sk_refcnt_debug_inc(struct sock *sk) |
| 805 | { |
| 806 | atomic_inc(&sk->sk_prot->socks); |
| 807 | } |
| 808 | |
| 809 | static inline void sk_refcnt_debug_dec(struct sock *sk) |
| 810 | { |
| 811 | atomic_dec(&sk->sk_prot->socks); |
| 812 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", |
| 813 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); |
| 814 | } |
| 815 | |
| 816 | static inline void sk_refcnt_debug_release(const struct sock *sk) |
| 817 | { |
| 818 | if (atomic_read(&sk->sk_refcnt) != 1) |
| 819 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", |
| 820 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); |
| 821 | } |
| 822 | #else /* SOCK_REFCNT_DEBUG */ |
| 823 | #define sk_refcnt_debug_inc(sk) do { } while (0) |
| 824 | #define sk_refcnt_debug_dec(sk) do { } while (0) |
| 825 | #define sk_refcnt_debug_release(sk) do { } while (0) |
| 826 | #endif /* SOCK_REFCNT_DEBUG */ |
| 827 | |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 828 | |
| 829 | #ifdef CONFIG_PROC_FS |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 830 | /* Called with local bh disabled */ |
Pavel Emelyanov | c29a0bc | 2008-03-31 19:41:46 -0700 | [diff] [blame] | 831 | extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
| 832 | extern int sock_prot_inuse_get(struct net *net, struct proto *proto); |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 833 | #else |
Pavel Emelyanov | c29a0bc | 2008-03-31 19:41:46 -0700 | [diff] [blame] | 834 | static void inline sock_prot_inuse_add(struct net *net, struct proto *prot, |
| 835 | int inc) |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 836 | { |
| 837 | } |
Eric Dumazet | 65f7651 | 2008-01-03 20:46:48 -0800 | [diff] [blame] | 838 | #endif |
| 839 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 840 | |
Arnaldo Carvalho de Melo | 614c6cb | 2005-08-09 19:47:37 -0700 | [diff] [blame] | 841 | /* With per-bucket locks this operation is not-atomic, so that |
| 842 | * this version is not worse. |
| 843 | */ |
| 844 | static inline void __sk_prot_rehash(struct sock *sk) |
| 845 | { |
| 846 | sk->sk_prot->unhash(sk); |
| 847 | sk->sk_prot->hash(sk); |
| 848 | } |
| 849 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 850 | /* About 10 seconds */ |
| 851 | #define SOCK_DESTROY_TIME (10*HZ) |
| 852 | |
| 853 | /* Sockets 0-1023 can't be bound to unless you are superuser */ |
| 854 | #define PROT_SOCK 1024 |
| 855 | |
| 856 | #define SHUTDOWN_MASK 3 |
| 857 | #define RCV_SHUTDOWN 1 |
| 858 | #define SEND_SHUTDOWN 2 |
| 859 | |
| 860 | #define SOCK_SNDBUF_LOCK 1 |
| 861 | #define SOCK_RCVBUF_LOCK 2 |
| 862 | #define SOCK_BINDADDR_LOCK 4 |
| 863 | #define SOCK_BINDPORT_LOCK 8 |
| 864 | |
| 865 | /* sock_iocb: used to kick off async processing of socket ios */ |
| 866 | struct sock_iocb { |
| 867 | struct list_head list; |
| 868 | |
| 869 | int flags; |
| 870 | int size; |
| 871 | struct socket *sock; |
| 872 | struct sock *sk; |
| 873 | struct scm_cookie *scm; |
| 874 | struct msghdr *msg, async_msg; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 875 | struct kiocb *kiocb; |
| 876 | }; |
| 877 | |
| 878 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) |
| 879 | { |
| 880 | return (struct sock_iocb *)iocb->private; |
| 881 | } |
| 882 | |
| 883 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) |
| 884 | { |
| 885 | return si->kiocb; |
| 886 | } |
| 887 | |
| 888 | struct socket_alloc { |
| 889 | struct socket socket; |
| 890 | struct inode vfs_inode; |
| 891 | }; |
| 892 | |
| 893 | static inline struct socket *SOCKET_I(struct inode *inode) |
| 894 | { |
| 895 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; |
| 896 | } |
| 897 | |
| 898 | static inline struct inode *SOCK_INODE(struct socket *socket) |
| 899 | { |
| 900 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; |
| 901 | } |
| 902 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 903 | /* |
| 904 | * Functions for memory accounting |
| 905 | */ |
| 906 | extern int __sk_mem_schedule(struct sock *sk, int size, int kind); |
| 907 | extern void __sk_mem_reclaim(struct sock *sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 908 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 909 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) |
| 910 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) |
| 911 | #define SK_MEM_SEND 0 |
| 912 | #define SK_MEM_RECV 1 |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 913 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 914 | static inline int sk_mem_pages(int amt) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 915 | { |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 916 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 917 | } |
| 918 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 919 | static inline int sk_has_account(struct sock *sk) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 920 | { |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 921 | /* return true if protocol supports memory accounting */ |
| 922 | return !!sk->sk_prot->memory_allocated; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 923 | } |
| 924 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 925 | static inline int sk_wmem_schedule(struct sock *sk, int size) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 926 | { |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 927 | if (!sk_has_account(sk)) |
| 928 | return 1; |
Herbert Xu | d80d99d | 2005-09-01 17:48:23 -0700 | [diff] [blame] | 929 | return size <= sk->sk_forward_alloc || |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 930 | __sk_mem_schedule(sk, size, SK_MEM_SEND); |
| 931 | } |
| 932 | |
| 933 | static inline int sk_rmem_schedule(struct sock *sk, int size) |
| 934 | { |
| 935 | if (!sk_has_account(sk)) |
| 936 | return 1; |
| 937 | return size <= sk->sk_forward_alloc || |
| 938 | __sk_mem_schedule(sk, size, SK_MEM_RECV); |
| 939 | } |
| 940 | |
| 941 | static inline void sk_mem_reclaim(struct sock *sk) |
| 942 | { |
| 943 | if (!sk_has_account(sk)) |
| 944 | return; |
| 945 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) |
| 946 | __sk_mem_reclaim(sk); |
| 947 | } |
| 948 | |
David S. Miller | 9993e7d | 2008-01-10 21:56:38 -0800 | [diff] [blame] | 949 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
| 950 | { |
| 951 | if (!sk_has_account(sk)) |
| 952 | return; |
| 953 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) |
| 954 | __sk_mem_reclaim(sk); |
| 955 | } |
| 956 | |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 957 | static inline void sk_mem_charge(struct sock *sk, int size) |
| 958 | { |
| 959 | if (!sk_has_account(sk)) |
| 960 | return; |
| 961 | sk->sk_forward_alloc -= size; |
| 962 | } |
| 963 | |
| 964 | static inline void sk_mem_uncharge(struct sock *sk, int size) |
| 965 | { |
| 966 | if (!sk_has_account(sk)) |
| 967 | return; |
| 968 | sk->sk_forward_alloc += size; |
| 969 | } |
| 970 | |
| 971 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) |
| 972 | { |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 973 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
| 974 | sk->sk_wmem_queued -= skb->truesize; |
| 975 | sk_mem_uncharge(sk, skb->truesize); |
| 976 | __kfree_skb(skb); |
Herbert Xu | d80d99d | 2005-09-01 17:48:23 -0700 | [diff] [blame] | 977 | } |
| 978 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 979 | /* Used by processes to "lock" a socket state, so that |
| 980 | * interrupts and bottom half handlers won't change it |
| 981 | * from under us. It essentially blocks any incoming |
| 982 | * packets, so that we won't get any new data or any |
| 983 | * packets that change the state of the socket. |
| 984 | * |
| 985 | * While locked, BH processing will add new packets to |
| 986 | * the backlog queue. This queue is processed by the |
| 987 | * owner of the socket lock right before it is released. |
| 988 | * |
| 989 | * Since ~2.3.5 it is also exclusive sleep lock serializing |
| 990 | * accesses from user process context. |
| 991 | */ |
John Heffner | d2e9117 | 2007-09-12 10:44:19 +0200 | [diff] [blame] | 992 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 993 | |
Peter Zijlstra | ed07536 | 2006-12-06 20:35:24 -0800 | [diff] [blame] | 994 | /* |
| 995 | * Macro so as to not evaluate some arguments when |
| 996 | * lockdep is not enabled. |
| 997 | * |
| 998 | * Mark both the sk_lock and the sk_lock.slock as a |
| 999 | * per-address-family lock class. |
| 1000 | */ |
| 1001 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
| 1002 | do { \ |
Ingo Molnar | e8f6fbf | 2008-11-12 01:38:36 +0000 | [diff] [blame] | 1003 | sk->sk_lock.owned = 0; \ |
Peter Zijlstra | ed07536 | 2006-12-06 20:35:24 -0800 | [diff] [blame] | 1004 | init_waitqueue_head(&sk->sk_lock.wq); \ |
| 1005 | spin_lock_init(&(sk)->sk_lock.slock); \ |
| 1006 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ |
| 1007 | sizeof((sk)->sk_lock)); \ |
| 1008 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ |
| 1009 | (skey), (sname)); \ |
| 1010 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
| 1011 | } while (0) |
| 1012 | |
Harvey Harrison | 4138093 | 2007-12-12 10:46:51 -0800 | [diff] [blame] | 1013 | extern void lock_sock_nested(struct sock *sk, int subclass); |
Peter Zijlstra | fcc70d5 | 2006-11-08 22:44:35 -0800 | [diff] [blame] | 1014 | |
| 1015 | static inline void lock_sock(struct sock *sk) |
| 1016 | { |
| 1017 | lock_sock_nested(sk, 0); |
| 1018 | } |
| 1019 | |
Harvey Harrison | 4138093 | 2007-12-12 10:46:51 -0800 | [diff] [blame] | 1020 | extern void release_sock(struct sock *sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1021 | |
| 1022 | /* BH context may only use the following locking interface. */ |
| 1023 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) |
Ingo Molnar | c636618 | 2006-07-03 00:25:13 -0700 | [diff] [blame] | 1024 | #define bh_lock_sock_nested(__sk) \ |
| 1025 | spin_lock_nested(&((__sk)->sk_lock.slock), \ |
| 1026 | SINGLE_DEPTH_NESTING) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1027 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
| 1028 | |
Eric Dumazet | 8a74ad6 | 2010-05-26 19:20:18 +0000 | [diff] [blame] | 1029 | extern bool lock_sock_fast(struct sock *sk); |
| 1030 | /** |
| 1031 | * unlock_sock_fast - complement of lock_sock_fast |
| 1032 | * @sk: socket |
| 1033 | * @slow: slow mode |
| 1034 | * |
| 1035 | * fast unlock socket for user context. |
| 1036 | * If slow mode is on, we call regular release_sock() |
| 1037 | */ |
| 1038 | static inline void unlock_sock_fast(struct sock *sk, bool slow) |
Eric Dumazet | 4b0b72f | 2010-04-28 14:35:48 -0700 | [diff] [blame] | 1039 | { |
Eric Dumazet | 8a74ad6 | 2010-05-26 19:20:18 +0000 | [diff] [blame] | 1040 | if (slow) |
| 1041 | release_sock(sk); |
| 1042 | else |
| 1043 | spin_unlock_bh(&sk->sk_lock.slock); |
Eric Dumazet | 4b0b72f | 2010-04-28 14:35:48 -0700 | [diff] [blame] | 1044 | } |
| 1045 | |
Eric Dumazet | 4b0b72f | 2010-04-28 14:35:48 -0700 | [diff] [blame] | 1046 | |
Eric W. Biederman | 1b8d7ae | 2007-10-08 23:24:22 -0700 | [diff] [blame] | 1047 | extern struct sock *sk_alloc(struct net *net, int family, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1048 | gfp_t priority, |
Pavel Emelyanov | 6257ff2 | 2007-11-01 00:39:31 -0700 | [diff] [blame] | 1049 | struct proto *prot); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1050 | extern void sk_free(struct sock *sk); |
Denis V. Lunev | edf0208 | 2008-02-29 11:18:32 -0800 | [diff] [blame] | 1051 | extern void sk_release_kernel(struct sock *sk); |
Arnaldo Carvalho de Melo | 87d11ce | 2005-08-09 20:10:12 -0700 | [diff] [blame] | 1052 | extern struct sock *sk_clone(const struct sock *sk, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1053 | const gfp_t priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1054 | |
| 1055 | extern struct sk_buff *sock_wmalloc(struct sock *sk, |
| 1056 | unsigned long size, int force, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1057 | gfp_t priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1058 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
| 1059 | unsigned long size, int force, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1060 | gfp_t priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1061 | extern void sock_wfree(struct sk_buff *skb); |
| 1062 | extern void sock_rfree(struct sk_buff *skb); |
| 1063 | |
| 1064 | extern int sock_setsockopt(struct socket *sock, int level, |
| 1065 | int op, char __user *optval, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 1066 | unsigned int optlen); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1067 | |
| 1068 | extern int sock_getsockopt(struct socket *sock, int level, |
| 1069 | int op, char __user *optval, |
| 1070 | int __user *optlen); |
| 1071 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, |
| 1072 | unsigned long size, |
| 1073 | int noblock, |
| 1074 | int *errcode); |
Herbert Xu | 4cc7f68 | 2009-02-04 16:55:54 -0800 | [diff] [blame] | 1075 | extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk, |
| 1076 | unsigned long header_len, |
| 1077 | unsigned long data_len, |
| 1078 | int noblock, |
| 1079 | int *errcode); |
Victor Fusco | 86a76ca | 2005-07-08 14:57:47 -0700 | [diff] [blame] | 1080 | extern void *sock_kmalloc(struct sock *sk, int size, |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1081 | gfp_t priority); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1082 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
| 1083 | extern void sk_send_sigurg(struct sock *sk); |
| 1084 | |
Herbert Xu | f845172 | 2010-05-24 00:12:34 -0700 | [diff] [blame] | 1085 | #ifdef CONFIG_CGROUPS |
| 1086 | extern void sock_update_classid(struct sock *sk); |
| 1087 | #else |
| 1088 | static inline void sock_update_classid(struct sock *sk) |
| 1089 | { |
| 1090 | } |
| 1091 | #endif |
| 1092 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1093 | /* |
| 1094 | * Functions to fill in entries in struct proto_ops when a protocol |
| 1095 | * does not implement a particular function. |
| 1096 | */ |
| 1097 | extern int sock_no_bind(struct socket *, |
| 1098 | struct sockaddr *, int); |
| 1099 | extern int sock_no_connect(struct socket *, |
| 1100 | struct sockaddr *, int, int); |
| 1101 | extern int sock_no_socketpair(struct socket *, |
| 1102 | struct socket *); |
| 1103 | extern int sock_no_accept(struct socket *, |
| 1104 | struct socket *, int); |
| 1105 | extern int sock_no_getname(struct socket *, |
| 1106 | struct sockaddr *, int *, int); |
| 1107 | extern unsigned int sock_no_poll(struct file *, struct socket *, |
| 1108 | struct poll_table_struct *); |
| 1109 | extern int sock_no_ioctl(struct socket *, unsigned int, |
| 1110 | unsigned long); |
| 1111 | extern int sock_no_listen(struct socket *, int); |
| 1112 | extern int sock_no_shutdown(struct socket *, int); |
| 1113 | extern int sock_no_getsockopt(struct socket *, int , int, |
| 1114 | char __user *, int __user *); |
| 1115 | extern int sock_no_setsockopt(struct socket *, int, int, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 1116 | char __user *, unsigned int); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1117 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, |
| 1118 | struct msghdr *, size_t); |
| 1119 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, |
| 1120 | struct msghdr *, size_t, int); |
| 1121 | extern int sock_no_mmap(struct file *file, |
| 1122 | struct socket *sock, |
| 1123 | struct vm_area_struct *vma); |
| 1124 | extern ssize_t sock_no_sendpage(struct socket *sock, |
| 1125 | struct page *page, |
| 1126 | int offset, size_t size, |
| 1127 | int flags); |
| 1128 | |
| 1129 | /* |
| 1130 | * Functions to fill in entries in struct proto_ops when a protocol |
| 1131 | * uses the inet style. |
| 1132 | */ |
| 1133 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, |
| 1134 | char __user *optval, int __user *optlen); |
| 1135 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, |
| 1136 | struct msghdr *msg, size_t size, int flags); |
| 1137 | 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] | 1138 | char __user *optval, unsigned int optlen); |
Dmitry Mishin | 3fdadf7 | 2006-03-20 22:45:21 -0800 | [diff] [blame] | 1139 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
| 1140 | int optname, char __user *optval, int __user *optlen); |
| 1141 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, |
David S. Miller | b705884 | 2009-09-30 16:12:20 -0700 | [diff] [blame] | 1142 | int optname, char __user *optval, unsigned int optlen); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1143 | |
| 1144 | extern void sk_common_release(struct sock *sk); |
| 1145 | |
| 1146 | /* |
| 1147 | * Default socket callbacks and setup code |
| 1148 | */ |
| 1149 | |
| 1150 | /* Initialise core socket variables */ |
| 1151 | extern void sock_init_data(struct socket *sock, struct sock *sk); |
| 1152 | |
| 1153 | /** |
Ben Hutchings | 1a5778a | 2010-02-14 22:35:47 -0800 | [diff] [blame] | 1154 | * sk_filter_release - release a socket filter |
Paul Bonser | dc9b334 | 2006-11-23 17:56:13 -0800 | [diff] [blame] | 1155 | * @fp: filter to remove |
| 1156 | * |
| 1157 | * Remove a filter from a socket and release its resources. |
| 1158 | */ |
| 1159 | |
Pavel Emelyanov | 309dd5f | 2007-10-17 21:21:51 -0700 | [diff] [blame] | 1160 | static inline void sk_filter_release(struct sk_filter *fp) |
| 1161 | { |
| 1162 | if (atomic_dec_and_test(&fp->refcnt)) |
Pavel Emelyanov | 47e958e | 2007-10-17 21:22:42 -0700 | [diff] [blame] | 1163 | kfree(fp); |
Pavel Emelyanov | 309dd5f | 2007-10-17 21:21:51 -0700 | [diff] [blame] | 1164 | } |
| 1165 | |
| 1166 | 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] | 1167 | { |
| 1168 | unsigned int size = sk_filter_len(fp); |
| 1169 | |
| 1170 | atomic_sub(size, &sk->sk_omem_alloc); |
Pavel Emelyanov | 309dd5f | 2007-10-17 21:21:51 -0700 | [diff] [blame] | 1171 | sk_filter_release(fp); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1172 | } |
| 1173 | |
| 1174 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) |
| 1175 | { |
| 1176 | atomic_inc(&fp->refcnt); |
| 1177 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); |
| 1178 | } |
| 1179 | |
| 1180 | /* |
| 1181 | * Socket reference counting postulates. |
| 1182 | * |
| 1183 | * * Each user of socket SHOULD hold a reference count. |
| 1184 | * * Each access point to socket (an hash table bucket, reference from a list, |
| 1185 | * running timer, skb in flight MUST hold a reference count. |
| 1186 | * * When reference count hits 0, it means it will never increase back. |
| 1187 | * * When reference count hits 0, it means that no references from |
| 1188 | * outside exist to this socket and current process on current CPU |
| 1189 | * is last user and may/should destroy this socket. |
| 1190 | * * sk_free is called from any context: process, BH, IRQ. When |
| 1191 | * it is called, socket has no references from outside -> sk_free |
| 1192 | * may release descendant resources allocated by the socket, but |
| 1193 | * to the time when it is called, socket is NOT referenced by any |
| 1194 | * hash tables, lists etc. |
| 1195 | * * Packets, delivered from outside (from network or from another process) |
| 1196 | * and enqueued on receive/error queues SHOULD NOT grab reference count, |
| 1197 | * when they sit in queue. Otherwise, packets will leak to hole, when |
| 1198 | * socket is looked up by one cpu and unhasing is made by another CPU. |
| 1199 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp |
| 1200 | * (leak to backlog). Packet socket does all the processing inside |
| 1201 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets |
| 1202 | * use separate SMP lock, so that they are prone too. |
| 1203 | */ |
| 1204 | |
| 1205 | /* Ungrab socket and destroy it, if it was the last reference. */ |
| 1206 | static inline void sock_put(struct sock *sk) |
| 1207 | { |
| 1208 | if (atomic_dec_and_test(&sk->sk_refcnt)) |
| 1209 | sk_free(sk); |
| 1210 | } |
| 1211 | |
Arnaldo Carvalho de Melo | 58a5a7b | 2006-11-16 14:06:06 -0200 | [diff] [blame] | 1212 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
| 1213 | const int nested); |
Arnaldo Carvalho de Melo | 25995ff | 2005-12-27 02:42:22 -0200 | [diff] [blame] | 1214 | |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 1215 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
| 1216 | { |
| 1217 | sk->sk_tx_queue_mapping = tx_queue; |
| 1218 | } |
| 1219 | |
| 1220 | static inline void sk_tx_queue_clear(struct sock *sk) |
| 1221 | { |
| 1222 | sk->sk_tx_queue_mapping = -1; |
| 1223 | } |
| 1224 | |
| 1225 | static inline int sk_tx_queue_get(const struct sock *sk) |
| 1226 | { |
| 1227 | return sk->sk_tx_queue_mapping; |
| 1228 | } |
| 1229 | |
| 1230 | static inline bool sk_tx_queue_recorded(const struct sock *sk) |
| 1231 | { |
| 1232 | return (sk && sk->sk_tx_queue_mapping >= 0); |
| 1233 | } |
| 1234 | |
David S. Miller | 972692e | 2008-06-17 22:41:38 -0700 | [diff] [blame] | 1235 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
| 1236 | { |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 1237 | sk_tx_queue_clear(sk); |
David S. Miller | 972692e | 2008-06-17 22:41:38 -0700 | [diff] [blame] | 1238 | sk->sk_socket = sock; |
| 1239 | } |
| 1240 | |
Eric Dumazet | aa39514 | 2010-04-20 13:03:51 +0000 | [diff] [blame] | 1241 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
| 1242 | { |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1243 | return &sk->sk_wq->wait; |
Eric Dumazet | aa39514 | 2010-04-20 13:03:51 +0000 | [diff] [blame] | 1244 | } |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1245 | /* Detach socket from process context. |
| 1246 | * Announce socket dead, detach it from wait queue and inode. |
| 1247 | * Note that parent inode held reference count on this struct sock, |
| 1248 | * we do not release it in this function, because protocol |
| 1249 | * probably wants some additional cleanups or even continuing |
| 1250 | * to work with this socket (TCP). |
| 1251 | */ |
| 1252 | static inline void sock_orphan(struct sock *sk) |
| 1253 | { |
| 1254 | write_lock_bh(&sk->sk_callback_lock); |
| 1255 | sock_set_flag(sk, SOCK_DEAD); |
David S. Miller | 972692e | 2008-06-17 22:41:38 -0700 | [diff] [blame] | 1256 | sk_set_socket(sk, NULL); |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1257 | sk->sk_wq = NULL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1258 | write_unlock_bh(&sk->sk_callback_lock); |
| 1259 | } |
| 1260 | |
| 1261 | static inline void sock_graft(struct sock *sk, struct socket *parent) |
| 1262 | { |
| 1263 | write_lock_bh(&sk->sk_callback_lock); |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1264 | rcu_assign_pointer(sk->sk_wq, parent->wq); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1265 | parent->sk = sk; |
David S. Miller | 972692e | 2008-06-17 22:41:38 -0700 | [diff] [blame] | 1266 | sk_set_socket(sk, parent); |
Venkat Yekkirala | 4237c75 | 2006-07-24 23:32:50 -0700 | [diff] [blame] | 1267 | security_sock_graft(sk, parent); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1268 | write_unlock_bh(&sk->sk_callback_lock); |
| 1269 | } |
| 1270 | |
| 1271 | extern int sock_i_uid(struct sock *sk); |
| 1272 | extern unsigned long sock_i_ino(struct sock *sk); |
| 1273 | |
| 1274 | static inline struct dst_entry * |
| 1275 | __sk_dst_get(struct sock *sk) |
| 1276 | { |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1277 | return rcu_dereference_check(sk->sk_dst_cache, rcu_read_lock_held() || |
Eric Dumazet | f68c224 | 2010-04-22 16:06:59 -0700 | [diff] [blame] | 1278 | sock_owned_by_user(sk) || |
| 1279 | lockdep_is_held(&sk->sk_lock.slock)); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1280 | } |
| 1281 | |
| 1282 | static inline struct dst_entry * |
| 1283 | sk_dst_get(struct sock *sk) |
| 1284 | { |
| 1285 | struct dst_entry *dst; |
| 1286 | |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1287 | rcu_read_lock(); |
| 1288 | dst = rcu_dereference(sk->sk_dst_cache); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1289 | if (dst) |
| 1290 | dst_hold(dst); |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1291 | rcu_read_unlock(); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1292 | return dst; |
| 1293 | } |
| 1294 | |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1295 | extern void sk_reset_txq(struct sock *sk); |
| 1296 | |
| 1297 | static inline void dst_negative_advice(struct sock *sk) |
| 1298 | { |
| 1299 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); |
| 1300 | |
| 1301 | if (dst && dst->ops->negative_advice) { |
| 1302 | ndst = dst->ops->negative_advice(dst); |
| 1303 | |
| 1304 | if (ndst != dst) { |
| 1305 | rcu_assign_pointer(sk->sk_dst_cache, ndst); |
| 1306 | sk_reset_txq(sk); |
| 1307 | } |
| 1308 | } |
| 1309 | } |
| 1310 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1311 | static inline void |
| 1312 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) |
| 1313 | { |
| 1314 | struct dst_entry *old_dst; |
| 1315 | |
Krishna Kumar | e022f0b | 2009-10-19 23:46:20 +0000 | [diff] [blame] | 1316 | sk_tx_queue_clear(sk); |
Eric Dumazet | 0b53ff2 | 2010-04-26 20:40:43 +0000 | [diff] [blame] | 1317 | /* |
| 1318 | * This can be called while sk is owned by the caller only, |
| 1319 | * with no state that can be checked in a rcu_dereference_check() cond |
| 1320 | */ |
| 1321 | old_dst = rcu_dereference_raw(sk->sk_dst_cache); |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1322 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1323 | dst_release(old_dst); |
| 1324 | } |
| 1325 | |
| 1326 | static inline void |
| 1327 | sk_dst_set(struct sock *sk, struct dst_entry *dst) |
| 1328 | { |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1329 | spin_lock(&sk->sk_dst_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1330 | __sk_dst_set(sk, dst); |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1331 | spin_unlock(&sk->sk_dst_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1332 | } |
| 1333 | |
| 1334 | static inline void |
| 1335 | __sk_dst_reset(struct sock *sk) |
| 1336 | { |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1337 | __sk_dst_set(sk, NULL); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1338 | } |
| 1339 | |
| 1340 | static inline void |
| 1341 | sk_dst_reset(struct sock *sk) |
| 1342 | { |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1343 | spin_lock(&sk->sk_dst_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1344 | __sk_dst_reset(sk); |
Eric Dumazet | b6c6712 | 2010-04-08 23:03:29 +0000 | [diff] [blame] | 1345 | spin_unlock(&sk->sk_dst_lock); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1346 | } |
| 1347 | |
Denis Vlasenko | f0088a5 | 2006-03-28 01:08:21 -0800 | [diff] [blame] | 1348 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1349 | |
Denis Vlasenko | f0088a5 | 2006-03-28 01:08:21 -0800 | [diff] [blame] | 1350 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1351 | |
Herbert Xu | bcd7611 | 2006-06-30 13:36:35 -0700 | [diff] [blame] | 1352 | static inline int sk_can_gso(const struct sock *sk) |
| 1353 | { |
| 1354 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); |
| 1355 | } |
| 1356 | |
Andi Kleen | 9958089 | 2007-04-20 17:12:43 -0700 | [diff] [blame] | 1357 | 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] | 1358 | |
Eric Dumazet | a465419 | 2010-05-16 00:36:33 -0700 | [diff] [blame] | 1359 | static inline void sk_nocaps_add(struct sock *sk, int flags) |
| 1360 | { |
| 1361 | sk->sk_route_nocaps |= flags; |
| 1362 | sk->sk_route_caps &= ~flags; |
| 1363 | } |
| 1364 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1365 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, |
| 1366 | struct sk_buff *skb, struct page *page, |
| 1367 | int off, int copy) |
| 1368 | { |
| 1369 | if (skb->ip_summed == CHECKSUM_NONE) { |
| 1370 | int err = 0; |
Al Viro | 5084205 | 2006-11-14 21:36:34 -0800 | [diff] [blame] | 1371 | __wsum csum = csum_and_copy_from_user(from, |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1372 | page_address(page) + off, |
| 1373 | copy, 0, &err); |
| 1374 | if (err) |
| 1375 | return err; |
| 1376 | skb->csum = csum_block_add(skb->csum, csum, skb->len); |
| 1377 | } else if (copy_from_user(page_address(page) + off, from, copy)) |
| 1378 | return -EFAULT; |
| 1379 | |
| 1380 | skb->len += copy; |
| 1381 | skb->data_len += copy; |
| 1382 | skb->truesize += copy; |
| 1383 | sk->sk_wmem_queued += copy; |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1384 | sk_mem_charge(sk, copy); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1385 | return 0; |
| 1386 | } |
| 1387 | |
Eric Dumazet | c564039 | 2009-06-16 10:12:03 +0000 | [diff] [blame] | 1388 | /** |
| 1389 | * sk_wmem_alloc_get - returns write allocations |
| 1390 | * @sk: socket |
| 1391 | * |
| 1392 | * Returns sk_wmem_alloc minus initial offset of one |
| 1393 | */ |
| 1394 | static inline int sk_wmem_alloc_get(const struct sock *sk) |
| 1395 | { |
| 1396 | return atomic_read(&sk->sk_wmem_alloc) - 1; |
| 1397 | } |
| 1398 | |
| 1399 | /** |
| 1400 | * sk_rmem_alloc_get - returns read allocations |
| 1401 | * @sk: socket |
| 1402 | * |
| 1403 | * Returns sk_rmem_alloc |
| 1404 | */ |
| 1405 | static inline int sk_rmem_alloc_get(const struct sock *sk) |
| 1406 | { |
| 1407 | return atomic_read(&sk->sk_rmem_alloc); |
| 1408 | } |
| 1409 | |
| 1410 | /** |
| 1411 | * sk_has_allocations - check if allocations are outstanding |
| 1412 | * @sk: socket |
| 1413 | * |
| 1414 | * Returns true if socket has write or read allocations |
| 1415 | */ |
| 1416 | static inline int sk_has_allocations(const struct sock *sk) |
| 1417 | { |
| 1418 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); |
| 1419 | } |
| 1420 | |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1421 | /** |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1422 | * wq_has_sleeper - check if there are any waiting processes |
Randy Dunlap | acfbe96 | 2010-05-24 23:54:18 -0700 | [diff] [blame] | 1423 | * @wq: struct socket_wq |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1424 | * |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1425 | * Returns true if socket_wq has waiting processes |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1426 | * |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1427 | * 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] | 1428 | * barrier call. They were added due to the race found within the tcp code. |
| 1429 | * |
| 1430 | * Consider following tcp code paths: |
| 1431 | * |
| 1432 | * CPU1 CPU2 |
| 1433 | * |
| 1434 | * sys_select receive packet |
| 1435 | * ... ... |
| 1436 | * __add_wait_queue update tp->rcv_nxt |
| 1437 | * ... ... |
| 1438 | * tp->rcv_nxt check sock_def_readable |
| 1439 | * ... { |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1440 | * schedule rcu_read_lock(); |
| 1441 | * wq = rcu_dereference(sk->sk_wq); |
| 1442 | * if (wq && waitqueue_active(&wq->wait)) |
| 1443 | * wake_up_interruptible(&wq->wait) |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1444 | * ... |
| 1445 | * } |
| 1446 | * |
| 1447 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay |
| 1448 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 |
| 1449 | * could then endup calling schedule and sleep forever if there are no more |
| 1450 | * data on the socket. |
Jiri Olsa | ad46276 | 2009-07-08 12:10:31 +0000 | [diff] [blame] | 1451 | * |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1452 | */ |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1453 | static inline bool wq_has_sleeper(struct socket_wq *wq) |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1454 | { |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1455 | |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1456 | /* |
| 1457 | * We need to be sure we are in sync with the |
| 1458 | * add_wait_queue modifications to the wait queue. |
| 1459 | * |
| 1460 | * This memory barrier is paired in the sock_poll_wait. |
| 1461 | */ |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1462 | smp_mb(); |
| 1463 | return wq && waitqueue_active(&wq->wait); |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1464 | } |
| 1465 | |
| 1466 | /** |
| 1467 | * sock_poll_wait - place memory barrier behind the poll_wait call. |
| 1468 | * @filp: file |
| 1469 | * @wait_address: socket wait queue |
| 1470 | * @p: poll_table |
| 1471 | * |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1472 | * See the comments in the wq_has_sleeper function. |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1473 | */ |
| 1474 | static inline void sock_poll_wait(struct file *filp, |
| 1475 | wait_queue_head_t *wait_address, poll_table *p) |
| 1476 | { |
| 1477 | if (p && wait_address) { |
| 1478 | poll_wait(filp, wait_address, p); |
| 1479 | /* |
| 1480 | * We need to be sure we are in sync with the |
| 1481 | * socket flags modification. |
| 1482 | * |
Eric Dumazet | 4381548 | 2010-04-29 11:01:49 +0000 | [diff] [blame] | 1483 | * This memory barrier is paired in the wq_has_sleeper. |
Jiri Olsa | a57de0b | 2009-07-08 12:09:13 +0000 | [diff] [blame] | 1484 | */ |
| 1485 | smp_mb(); |
| 1486 | } |
| 1487 | } |
| 1488 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1489 | /* |
| 1490 | * Queue a received datagram if it will fit. Stream and sequenced |
| 1491 | * protocols can't normally use this as they need to fit buffers in |
| 1492 | * and play with them. |
| 1493 | * |
| 1494 | * Inlined as it's very short and called for pretty much every |
| 1495 | * packet ever received. |
| 1496 | */ |
| 1497 | |
| 1498 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) |
| 1499 | { |
Herbert Xu | d55d87f | 2009-06-22 02:25:25 +0000 | [diff] [blame] | 1500 | skb_orphan(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1501 | skb->sk = sk; |
| 1502 | skb->destructor = sock_wfree; |
Eric Dumazet | 2b85a34 | 2009-06-11 02:55:43 -0700 | [diff] [blame] | 1503 | /* |
| 1504 | * We used to take a refcount on sk, but following operation |
| 1505 | * is enough to guarantee sk_free() wont free this sock until |
| 1506 | * all in-flight packets are completed |
| 1507 | */ |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1508 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); |
| 1509 | } |
| 1510 | |
| 1511 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
| 1512 | { |
Herbert Xu | d55d87f | 2009-06-22 02:25:25 +0000 | [diff] [blame] | 1513 | skb_orphan(skb); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1514 | skb->sk = sk; |
| 1515 | skb->destructor = sock_rfree; |
| 1516 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); |
Hideo Aoki | 3ab224b | 2007-12-31 00:11:19 -0800 | [diff] [blame] | 1517 | sk_mem_charge(sk, skb->truesize); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1518 | } |
| 1519 | |
| 1520 | extern void sk_reset_timer(struct sock *sk, struct timer_list* timer, |
| 1521 | unsigned long expires); |
| 1522 | |
| 1523 | extern void sk_stop_timer(struct sock *sk, struct timer_list* timer); |
| 1524 | |
Denis Vlasenko | f0088a5 | 2006-03-28 01:08:21 -0800 | [diff] [blame] | 1525 | 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] | 1526 | |
Eric Dumazet | b1faf56 | 2010-05-31 23:44:05 -0700 | [diff] [blame] | 1527 | 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] | 1528 | |
| 1529 | /* |
| 1530 | * Recover an error report and clear atomically |
| 1531 | */ |
| 1532 | |
| 1533 | static inline int sock_error(struct sock *sk) |
| 1534 | { |
Benjamin LaHaise | c1cbe4b | 2005-12-13 23:22:19 -0800 | [diff] [blame] | 1535 | int err; |
| 1536 | if (likely(!sk->sk_err)) |
| 1537 | return 0; |
| 1538 | err = xchg(&sk->sk_err, 0); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1539 | return -err; |
| 1540 | } |
| 1541 | |
| 1542 | static inline unsigned long sock_wspace(struct sock *sk) |
| 1543 | { |
| 1544 | int amt = 0; |
| 1545 | |
| 1546 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { |
| 1547 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); |
| 1548 | if (amt < 0) |
| 1549 | amt = 0; |
| 1550 | } |
| 1551 | return amt; |
| 1552 | } |
| 1553 | |
| 1554 | static inline void sk_wake_async(struct sock *sk, int how, int band) |
| 1555 | { |
Eric Dumazet | bcdce71 | 2009-10-06 17:28:29 -0700 | [diff] [blame] | 1556 | if (sock_flag(sk, SOCK_FASYNC)) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1557 | sock_wake_async(sk->sk_socket, how, band); |
| 1558 | } |
| 1559 | |
| 1560 | #define SOCK_MIN_SNDBUF 2048 |
| 1561 | #define SOCK_MIN_RCVBUF 256 |
| 1562 | |
| 1563 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) |
| 1564 | { |
| 1565 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { |
Eric Dumazet | 8df09ea | 2007-12-21 03:07:41 -0800 | [diff] [blame] | 1566 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1567 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
| 1568 | } |
| 1569 | } |
| 1570 | |
Pavel Emelyanov | df97c70 | 2007-11-29 21:22:33 +1100 | [diff] [blame] | 1571 | 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] | 1572 | |
| 1573 | static inline struct page *sk_stream_alloc_page(struct sock *sk) |
| 1574 | { |
| 1575 | struct page *page = NULL; |
| 1576 | |
Herbert Xu | ef01578 | 2005-09-01 17:48:59 -0700 | [diff] [blame] | 1577 | page = alloc_pages(sk->sk_allocation, 0); |
| 1578 | if (!page) { |
Pavel Emelyanov | 5c52ba1 | 2008-07-16 20:28:10 -0700 | [diff] [blame] | 1579 | sk->sk_prot->enter_memory_pressure(sk); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1580 | sk_stream_moderate_sndbuf(sk); |
| 1581 | } |
| 1582 | return page; |
| 1583 | } |
| 1584 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1585 | /* |
| 1586 | * Default write policy as shown to user space via poll/select/SIGIO |
| 1587 | */ |
| 1588 | static inline int sock_writeable(const struct sock *sk) |
| 1589 | { |
Eric Dumazet | 8df09ea | 2007-12-21 03:07:41 -0800 | [diff] [blame] | 1590 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1591 | } |
| 1592 | |
Al Viro | dd0fc66 | 2005-10-07 07:46:04 +0100 | [diff] [blame] | 1593 | static inline gfp_t gfp_any(void) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1594 | { |
Andrew Morton | 9970937 | 2009-02-12 16:43:17 -0800 | [diff] [blame] | 1595 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1596 | } |
| 1597 | |
| 1598 | static inline long sock_rcvtimeo(const struct sock *sk, int noblock) |
| 1599 | { |
| 1600 | return noblock ? 0 : sk->sk_rcvtimeo; |
| 1601 | } |
| 1602 | |
| 1603 | static inline long sock_sndtimeo(const struct sock *sk, int noblock) |
| 1604 | { |
| 1605 | return noblock ? 0 : sk->sk_sndtimeo; |
| 1606 | } |
| 1607 | |
| 1608 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) |
| 1609 | { |
| 1610 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; |
| 1611 | } |
| 1612 | |
| 1613 | /* Alas, with timeout socket operations are not restartable. |
| 1614 | * Compare this to poll(). |
| 1615 | */ |
| 1616 | static inline int sock_intr_errno(long timeo) |
| 1617 | { |
| 1618 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; |
| 1619 | } |
| 1620 | |
Eric Dumazet | 92f37fd | 2007-03-25 22:14:49 -0700 | [diff] [blame] | 1621 | extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
| 1622 | struct sk_buff *skb); |
| 1623 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1624 | static __inline__ void |
| 1625 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
| 1626 | { |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 1627 | ktime_t kt = skb->tstamp; |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 1628 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
Patrick McHardy | a61bbcf | 2005-08-14 17:24:31 -0700 | [diff] [blame] | 1629 | |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 1630 | /* |
| 1631 | * generate control messages if |
| 1632 | * - receive time stamping in software requested (SOCK_RCVTSTAMP |
| 1633 | * or SOCK_TIMESTAMPING_RX_SOFTWARE) |
| 1634 | * - software time stamp available and wanted |
| 1635 | * (SOCK_TIMESTAMPING_SOFTWARE) |
| 1636 | * - hardware time stamps available and wanted |
| 1637 | * (SOCK_TIMESTAMPING_SYS_HARDWARE or |
| 1638 | * SOCK_TIMESTAMPING_RAW_HARDWARE) |
| 1639 | */ |
| 1640 | if (sock_flag(sk, SOCK_RCVTSTAMP) || |
| 1641 | sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) || |
| 1642 | (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) || |
| 1643 | (hwtstamps->hwtstamp.tv64 && |
| 1644 | sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) || |
| 1645 | (hwtstamps->syststamp.tv64 && |
| 1646 | sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))) |
Eric Dumazet | 92f37fd | 2007-03-25 22:14:49 -0700 | [diff] [blame] | 1647 | __sock_recv_timestamp(msg, sk, skb); |
| 1648 | else |
Eric Dumazet | b7aa0bf | 2007-04-19 16:16:32 -0700 | [diff] [blame] | 1649 | sk->sk_stamp = kt; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1650 | } |
| 1651 | |
Eric Dumazet | 767dd03 | 2010-04-28 19:14:43 +0000 | [diff] [blame] | 1652 | extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
| 1653 | struct sk_buff *skb); |
| 1654 | |
| 1655 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
| 1656 | struct sk_buff *skb) |
| 1657 | { |
| 1658 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ |
| 1659 | (1UL << SOCK_RCVTSTAMP) | \ |
| 1660 | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \ |
| 1661 | (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \ |
| 1662 | (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \ |
| 1663 | (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE)) |
| 1664 | |
| 1665 | if (sk->sk_flags & FLAGS_TS_OR_DROPS) |
| 1666 | __sock_recv_ts_and_drops(msg, sk, skb); |
| 1667 | else |
| 1668 | sk->sk_stamp = skb->tstamp; |
| 1669 | } |
Neil Horman | 3b88578 | 2009-10-12 13:26:31 -0700 | [diff] [blame] | 1670 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1671 | /** |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 1672 | * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped |
| 1673 | * @msg: outgoing packet |
| 1674 | * @sk: socket sending this packet |
| 1675 | * @shtx: filled with instructions for time stamping |
| 1676 | * |
| 1677 | * Currently only depends on SOCK_TIMESTAMPING* flags. Returns error code if |
| 1678 | * parameters are invalid. |
| 1679 | */ |
| 1680 | extern int sock_tx_timestamp(struct msghdr *msg, |
| 1681 | struct sock *sk, |
| 1682 | union skb_shared_tx *shtx); |
| 1683 | |
| 1684 | |
| 1685 | /** |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1686 | * sk_eat_skb - Release a skb if it is no longer needed |
Pavel Pisa | 4dc3b16 | 2005-05-01 08:59:25 -0700 | [diff] [blame] | 1687 | * @sk: socket to eat this skb from |
| 1688 | * @skb: socket buffer to eat |
Randy Dunlap | f4b8ea7 | 2006-06-22 16:00:11 -0700 | [diff] [blame] | 1689 | * @copied_early: flag indicating whether DMA operations copied this data early |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1690 | * |
| 1691 | * This routine must be called with interrupts disabled or with the socket |
| 1692 | * locked so that the sk_buff queue operation is ok. |
| 1693 | */ |
Chris Leech | 624d116 | 2006-05-23 18:01:28 -0700 | [diff] [blame] | 1694 | #ifdef CONFIG_NET_DMA |
| 1695 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, int copied_early) |
| 1696 | { |
| 1697 | __skb_unlink(skb, &sk->sk_receive_queue); |
| 1698 | if (!copied_early) |
| 1699 | __kfree_skb(skb); |
| 1700 | else |
| 1701 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); |
| 1702 | } |
| 1703 | #else |
| 1704 | 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] | 1705 | { |
| 1706 | __skb_unlink(skb, &sk->sk_receive_queue); |
| 1707 | __kfree_skb(skb); |
| 1708 | } |
Chris Leech | 624d116 | 2006-05-23 18:01:28 -0700 | [diff] [blame] | 1709 | #endif |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1710 | |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 1711 | static inline |
| 1712 | struct net *sock_net(const struct sock *sk) |
| 1713 | { |
| 1714 | #ifdef CONFIG_NET_NS |
| 1715 | return sk->sk_net; |
| 1716 | #else |
| 1717 | return &init_net; |
| 1718 | #endif |
| 1719 | } |
| 1720 | |
| 1721 | static inline |
Denis V. Lunev | f5aa23f | 2008-03-26 00:48:17 -0700 | [diff] [blame] | 1722 | void sock_net_set(struct sock *sk, struct net *net) |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 1723 | { |
| 1724 | #ifdef CONFIG_NET_NS |
| 1725 | sk->sk_net = net; |
| 1726 | #endif |
| 1727 | } |
| 1728 | |
Denis V. Lunev | edf0208 | 2008-02-29 11:18:32 -0800 | [diff] [blame] | 1729 | /* |
| 1730 | * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. |
| 1731 | * They should not hold a referrence to a namespace in order to allow |
| 1732 | * to stop it. |
| 1733 | * Sockets after sk_change_net should be released using sk_release_kernel |
| 1734 | */ |
| 1735 | static inline void sk_change_net(struct sock *sk, struct net *net) |
| 1736 | { |
YOSHIFUJI Hideaki | 3b1e0a6 | 2008-03-26 02:26:21 +0900 | [diff] [blame] | 1737 | put_net(sock_net(sk)); |
Denis V. Lunev | 65a18ec | 2008-04-16 01:59:46 -0700 | [diff] [blame] | 1738 | sock_net_set(sk, hold_net(net)); |
Denis V. Lunev | edf0208 | 2008-02-29 11:18:32 -0800 | [diff] [blame] | 1739 | } |
| 1740 | |
KOVACS Krisztian | 2354261 | 2008-10-07 12:41:01 -0700 | [diff] [blame] | 1741 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) |
| 1742 | { |
| 1743 | if (unlikely(skb->sk)) { |
| 1744 | struct sock *sk = skb->sk; |
| 1745 | |
| 1746 | skb->destructor = NULL; |
| 1747 | skb->sk = NULL; |
| 1748 | return sk; |
| 1749 | } |
| 1750 | return NULL; |
| 1751 | } |
| 1752 | |
Patrick Ohly | 20d4947 | 2009-02-12 05:03:38 +0000 | [diff] [blame] | 1753 | extern void sock_enable_timestamp(struct sock *sk, int flag); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1754 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); |
Eric Dumazet | ae40eb1 | 2007-03-18 17:33:16 -0700 | [diff] [blame] | 1755 | extern int sock_get_timestampns(struct sock *, struct timespec __user *); |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1756 | |
| 1757 | /* |
| 1758 | * Enable debug/info messages |
| 1759 | */ |
Stephen Hemminger | a2a316f | 2007-03-08 20:41:08 -0800 | [diff] [blame] | 1760 | extern int net_msg_warn; |
| 1761 | #define NETDEBUG(fmt, args...) \ |
| 1762 | do { if (net_msg_warn) printk(fmt,##args); } while (0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1763 | |
Stephen Hemminger | a2a316f | 2007-03-08 20:41:08 -0800 | [diff] [blame] | 1764 | #define LIMIT_NETDEBUG(fmt, args...) \ |
| 1765 | do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1766 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1767 | extern __u32 sysctl_wmem_max; |
| 1768 | extern __u32 sysctl_rmem_max; |
| 1769 | |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 1770 | extern void sk_init(void); |
| 1771 | |
David S. Miller | 6baf1f4 | 2005-09-05 18:14:11 -0700 | [diff] [blame] | 1772 | extern int sysctl_optmem_max; |
| 1773 | |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 1774 | extern __u32 sysctl_wmem_default; |
| 1775 | extern __u32 sysctl_rmem_default; |
Arnaldo Carvalho de Melo | 2038073 | 2005-08-16 02:18:02 -0300 | [diff] [blame] | 1776 | |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1777 | #endif /* _SOCK_H */ |