| 1 | /* Copyright (c) 2015 Samsung Electronics Co., Ltd. */ |
| 2 | /* |
| 3 | * INET An implementation of the TCP/IP protocol suite for the LINUX |
| 4 | * operating system. INET is implemented using the BSD Socket |
| 5 | * interface as the means of communication with the user level. |
| 6 | * |
| 7 | * Definitions for the AF_INET socket handler. |
| 8 | * |
| 9 | * Version: @(#)sock.h 1.0.4 05/13/93 |
| 10 | * |
| 11 | * Authors: Ross Biro |
| 12 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
| 13 | * Corey Minyard <wf-rch!minyard@relay.EU.net> |
| 14 | * Florian La Roche <flla@stud.uni-sb.de> |
| 15 | * |
| 16 | * Fixes: |
| 17 | * Alan Cox : Volatiles in skbuff pointers. See |
| 18 | * skbuff comments. May be overdone, |
| 19 | * better to prove they can be removed |
| 20 | * than the reverse. |
| 21 | * Alan Cox : Added a zapped field for tcp to note |
| 22 | * a socket is reset and must stay shut up |
| 23 | * Alan Cox : New fields for options |
| 24 | * Pauline Middelink : identd support |
| 25 | * Alan Cox : Eliminate low level recv/recvfrom |
| 26 | * David S. Miller : New socket lookup architecture. |
| 27 | * Steve Whitehouse: Default routines for sock_ops |
| 28 | * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made |
| 29 | * protinfo be just a void pointer, as the |
| 30 | * protocol specific parts were moved to |
| 31 | * respective headers and ipv4/v6, etc now |
| 32 | * use private slabcaches for its socks |
| 33 | * Pedro Hortas : New flags field for socket options |
| 34 | * |
| 35 | * |
| 36 | * This program is free software; you can redistribute it and/or |
| 37 | * modify it under the terms of the GNU General Public License |
| 38 | * as published by the Free Software Foundation; either version |
| 39 | * 2 of the License, or (at your option) any later version. |
| 40 | */ |
| 41 | /* |
| 42 | * Changes: |
| 43 | * KwnagHyun Kim <kh0304.kim@samsung.com> 2015/07/08 |
| 44 | * Baesung Park <baesung.park@samsung.com> 2015/07/08 |
| 45 | * Vignesh Saravanaperumal <vignesh1.s@samsung.com> 2015/07/08 |
| 46 | * Add codes to share UID/PID information |
| 47 | * |
| 48 | */ |
| 49 | |
| 50 | #ifndef _SOCK_H |
| 51 | #define _SOCK_H |
| 52 | |
| 53 | #include <linux/hardirq.h> |
| 54 | #include <linux/kernel.h> |
| 55 | #include <linux/list.h> |
| 56 | #include <linux/list_nulls.h> |
| 57 | #include <linux/timer.h> |
| 58 | #include <linux/cache.h> |
| 59 | #include <linux/bitops.h> |
| 60 | #include <linux/lockdep.h> |
| 61 | #include <linux/netdevice.h> |
| 62 | #include <linux/skbuff.h> /* struct sk_buff */ |
| 63 | #include <linux/mm.h> |
| 64 | #include <linux/security.h> |
| 65 | #include <linux/slab.h> |
| 66 | #include <linux/uaccess.h> |
| 67 | #include <linux/memcontrol.h> |
| 68 | #include <linux/res_counter.h> |
| 69 | #include <linux/static_key.h> |
| 70 | #include <linux/aio.h> |
| 71 | #include <linux/sched.h> |
| 72 | |
| 73 | #include <linux/filter.h> |
| 74 | #include <linux/rculist_nulls.h> |
| 75 | #include <linux/poll.h> |
| 76 | |
| 77 | #include <linux/atomic.h> |
| 78 | #include <net/dst.h> |
| 79 | #include <net/checksum.h> |
| 80 | |
| 81 | struct cgroup; |
| 82 | struct cgroup_subsys; |
| 83 | #ifdef CONFIG_NET |
| 84 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss); |
| 85 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg); |
| 86 | #else |
| 87 | static inline |
| 88 | int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss) |
| 89 | { |
| 90 | return 0; |
| 91 | } |
| 92 | static inline |
| 93 | void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg) |
| 94 | { |
| 95 | } |
| 96 | #endif |
| 97 | /* |
| 98 | * This structure really needs to be cleaned up. |
| 99 | * Most of it is for TCP, and not used by any of |
| 100 | * the other protocols. |
| 101 | */ |
| 102 | |
| 103 | /* Define this to get the SOCK_DBG debugging facility. */ |
| 104 | #define SOCK_DEBUGGING |
| 105 | #ifdef SOCK_DEBUGGING |
| 106 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ |
| 107 | printk(KERN_DEBUG msg); } while (0) |
| 108 | #else |
| 109 | /* Validate arguments and do nothing */ |
| 110 | static inline __printf(2, 3) |
| 111 | void SOCK_DEBUG(const struct sock *sk, const char *msg, ...) |
| 112 | { |
| 113 | } |
| 114 | #endif |
| 115 | |
| 116 | /* This is the per-socket lock. The spinlock provides a synchronization |
| 117 | * between user contexts and software interrupt processing, whereas the |
| 118 | * mini-semaphore synchronizes multiple users amongst themselves. |
| 119 | */ |
| 120 | typedef struct { |
| 121 | spinlock_t slock; |
| 122 | int owned; |
| 123 | wait_queue_head_t wq; |
| 124 | /* |
| 125 | * We express the mutex-alike socket_lock semantics |
| 126 | * to the lock validator by explicitly managing |
| 127 | * the slock as a lock variant (in addition to |
| 128 | * the slock itself): |
| 129 | */ |
| 130 | #ifdef CONFIG_DEBUG_LOCK_ALLOC |
| 131 | struct lockdep_map dep_map; |
| 132 | #endif |
| 133 | } socket_lock_t; |
| 134 | |
| 135 | struct sock; |
| 136 | struct proto; |
| 137 | struct net; |
| 138 | |
| 139 | typedef __u32 __bitwise __portpair; |
| 140 | typedef __u64 __bitwise __addrpair; |
| 141 | |
| 142 | /** |
| 143 | * struct sock_common - minimal network layer representation of sockets |
| 144 | * @skc_daddr: Foreign IPv4 addr |
| 145 | * @skc_rcv_saddr: Bound local IPv4 addr |
| 146 | * @skc_hash: hash value used with various protocol lookup tables |
| 147 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
| 148 | * @skc_dport: placeholder for inet_dport/tw_dport |
| 149 | * @skc_num: placeholder for inet_num/tw_num |
| 150 | * @skc_family: network address family |
| 151 | * @skc_state: Connection state |
| 152 | * @skc_reuse: %SO_REUSEADDR setting |
| 153 | * @skc_reuseport: %SO_REUSEPORT setting |
| 154 | * @skc_bound_dev_if: bound device index if != 0 |
| 155 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
| 156 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol |
| 157 | * @skc_prot: protocol handlers inside a network family |
| 158 | * @skc_net: reference to the network namespace of this socket |
| 159 | * @skc_node: main hash linkage for various protocol lookup tables |
| 160 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol |
| 161 | * @skc_tx_queue_mapping: tx queue number for this connection |
| 162 | * @skc_refcnt: reference count |
| 163 | * |
| 164 | * This is the minimal network layer representation of sockets, the header |
| 165 | * for struct sock and struct inet_timewait_sock. |
| 166 | */ |
| 167 | struct sock_common { |
| 168 | /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned |
| 169 | * address on 64bit arches : cf INET_MATCH() and INET_TW_MATCH() |
| 170 | */ |
| 171 | union { |
| 172 | __addrpair skc_addrpair; |
| 173 | struct { |
| 174 | __be32 skc_daddr; |
| 175 | __be32 skc_rcv_saddr; |
| 176 | }; |
| 177 | }; |
| 178 | union { |
| 179 | unsigned int skc_hash; |
| 180 | __u16 skc_u16hashes[2]; |
| 181 | }; |
| 182 | /* skc_dport && skc_num must be grouped as well */ |
| 183 | union { |
| 184 | __portpair skc_portpair; |
| 185 | struct { |
| 186 | __be16 skc_dport; |
| 187 | __u16 skc_num; |
| 188 | }; |
| 189 | }; |
| 190 | |
| 191 | unsigned short skc_family; |
| 192 | volatile unsigned char skc_state; |
| 193 | unsigned char skc_reuse:4; |
| 194 | unsigned char skc_reuseport:4; |
| 195 | int skc_bound_dev_if; |
| 196 | union { |
| 197 | struct hlist_node skc_bind_node; |
| 198 | struct hlist_nulls_node skc_portaddr_node; |
| 199 | }; |
| 200 | struct proto *skc_prot; |
| 201 | #ifdef CONFIG_NET_NS |
| 202 | struct net *skc_net; |
| 203 | #endif |
| 204 | /* |
| 205 | * fields between dontcopy_begin/dontcopy_end |
| 206 | * are not copied in sock_copy() |
| 207 | */ |
| 208 | /* private: */ |
| 209 | int skc_dontcopy_begin[0]; |
| 210 | /* public: */ |
| 211 | union { |
| 212 | struct hlist_node skc_node; |
| 213 | struct hlist_nulls_node skc_nulls_node; |
| 214 | }; |
| 215 | int skc_tx_queue_mapping; |
| 216 | atomic_t skc_refcnt; |
| 217 | /* private: */ |
| 218 | int skc_dontcopy_end[0]; |
| 219 | /* public: */ |
| 220 | }; |
| 221 | |
| 222 | struct cg_proto; |
| 223 | /** |
| 224 | * struct sock - network layer representation of sockets |
| 225 | * @__sk_common: shared layout with inet_timewait_sock |
| 226 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
| 227 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings |
| 228 | * @sk_lock: synchronizer |
| 229 | * @sk_rcvbuf: size of receive buffer in bytes |
| 230 | * @sk_wq: sock wait queue and async head |
| 231 | * @sk_rx_dst: receive input route used by early tcp demux |
| 232 | * @sk_dst_cache: destination cache |
| 233 | * @sk_dst_lock: destination cache lock |
| 234 | * @sk_policy: flow policy |
| 235 | * @sk_receive_queue: incoming packets |
| 236 | * @sk_wmem_alloc: transmit queue bytes committed |
| 237 | * @sk_write_queue: Packet sending queue |
| 238 | * @sk_async_wait_queue: DMA copied packets |
| 239 | * @sk_omem_alloc: "o" is "option" or "other" |
| 240 | * @sk_wmem_queued: persistent queue size |
| 241 | * @sk_forward_alloc: space allocated forward |
| 242 | * @sk_allocation: allocation mode |
| 243 | * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler) |
| 244 | * @sk_sndbuf: size of send buffer in bytes |
| 245 | * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, |
| 246 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings |
| 247 | * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets |
| 248 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
| 249 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
| 250 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
| 251 | * @sk_gso_max_size: Maximum GSO segment size to build |
| 252 | * @sk_gso_max_segs: Maximum number of GSO segments |
| 253 | * @sk_lingertime: %SO_LINGER l_linger setting |
| 254 | * @sk_backlog: always used with the per-socket spinlock held |
| 255 | * @sk_callback_lock: used with the callbacks in the end of this struct |
| 256 | * @sk_error_queue: rarely used |
| 257 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
| 258 | * IPV6_ADDRFORM for instance) |
| 259 | * @sk_err: last error |
| 260 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
| 261 | * persistent failure not just 'timed out' |
| 262 | * @sk_drops: raw/udp drops counter |
| 263 | * @sk_ack_backlog: current listen backlog |
| 264 | * @sk_max_ack_backlog: listen backlog set in listen() |
| 265 | * @sk_priority: %SO_PRIORITY setting |
| 266 | * @sk_cgrp_prioidx: socket group's priority map index |
| 267 | * @sk_type: socket type (%SOCK_STREAM, etc) |
| 268 | * @sk_protocol: which protocol this socket belongs in this network family |
| 269 | * @sk_peer_pid: &struct pid for this socket's peer |
| 270 | * @sk_peer_cred: %SO_PEERCRED setting |
| 271 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
| 272 | * @sk_rcvtimeo: %SO_RCVTIMEO setting |
| 273 | * @sk_sndtimeo: %SO_SNDTIMEO setting |
| 274 | * @sk_rxhash: flow hash received from netif layer |
| 275 | * @sk_filter: socket filtering instructions |
| 276 | * @sk_protinfo: private area, net family specific, when not using slab |
| 277 | * @sk_timer: sock cleanup timer |
| 278 | * @sk_stamp: time stamp of last packet received |
| 279 | * @sk_socket: Identd and reporting IO signals |
| 280 | * @sk_user_data: RPC layer private data |
| 281 | * @sk_frag: cached page frag |
| 282 | * @sk_peek_off: current peek_offset value |
| 283 | * @sk_send_head: front of stuff to transmit |
| 284 | * @sk_security: used by security modules |
| 285 | * @sk_mark: generic packet mark |
| 286 | * @sk_classid: this socket's cgroup classid |
| 287 | * @sk_cgrp: this socket's cgroup-specific proto data |
| 288 | * @sk_write_pending: a write to stream socket waits to start |
| 289 | * @sk_state_change: callback to indicate change in the state of the sock |
| 290 | * @sk_data_ready: callback to indicate there is data to be processed |
| 291 | * @sk_write_space: callback to indicate there is bf sending space available |
| 292 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) |
| 293 | * @sk_backlog_rcv: callback to process the backlog |
| 294 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 |
| 295 | */ |
| 296 | struct sock { |
| 297 | /* |
| 298 | * Now struct inet_timewait_sock also uses sock_common, so please just |
| 299 | * don't add nothing before this first member (__sk_common) --acme |
| 300 | */ |
| 301 | struct sock_common __sk_common; |
| 302 | #define sk_node __sk_common.skc_node |
| 303 | #define sk_nulls_node __sk_common.skc_nulls_node |
| 304 | #define sk_refcnt __sk_common.skc_refcnt |
| 305 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
| 306 | |
| 307 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
| 308 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end |
| 309 | #define sk_hash __sk_common.skc_hash |
| 310 | #define sk_family __sk_common.skc_family |
| 311 | #define sk_state __sk_common.skc_state |
| 312 | #define sk_reuse __sk_common.skc_reuse |
| 313 | #define sk_reuseport __sk_common.skc_reuseport |
| 314 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
| 315 | #define sk_bind_node __sk_common.skc_bind_node |
| 316 | #define sk_prot __sk_common.skc_prot |
| 317 | #define sk_net __sk_common.skc_net |
| 318 | socket_lock_t sk_lock; |
| 319 | struct sk_buff_head sk_receive_queue; |
| 320 | /* |
| 321 | * The backlog queue is special, it is always used with |
| 322 | * the per-socket spinlock held and requires low latency |
| 323 | * access. Therefore we special case it's implementation. |
| 324 | * Note : rmem_alloc is in this structure to fill a hole |
| 325 | * on 64bit arches, not because its logically part of |
| 326 | * backlog. |
| 327 | */ |
| 328 | struct { |
| 329 | atomic_t rmem_alloc; |
| 330 | int len; |
| 331 | struct sk_buff *head; |
| 332 | struct sk_buff *tail; |
| 333 | } sk_backlog; |
| 334 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
| 335 | int sk_forward_alloc; |
| 336 | #ifdef CONFIG_RPS |
| 337 | __u32 sk_rxhash; |
| 338 | #endif |
| 339 | atomic_t sk_drops; |
| 340 | int sk_rcvbuf; |
| 341 | |
| 342 | struct sk_filter __rcu *sk_filter; |
| 343 | struct socket_wq __rcu *sk_wq; |
| 344 | |
| 345 | #ifdef CONFIG_NET_DMA |
| 346 | struct sk_buff_head sk_async_wait_queue; |
| 347 | #endif |
| 348 | |
| 349 | #ifdef CONFIG_XFRM |
| 350 | struct xfrm_policy *sk_policy[2]; |
| 351 | #endif |
| 352 | unsigned long sk_flags; |
| 353 | struct dst_entry *sk_rx_dst; |
| 354 | struct dst_entry __rcu *sk_dst_cache; |
| 355 | spinlock_t sk_dst_lock; |
| 356 | atomic_t sk_wmem_alloc; |
| 357 | atomic_t sk_omem_alloc; |
| 358 | int sk_sndbuf; |
| 359 | struct sk_buff_head sk_write_queue; |
| 360 | kmemcheck_bitfield_begin(flags); |
| 361 | unsigned int sk_shutdown : 2, |
| 362 | sk_no_check : 2, |
| 363 | sk_userlocks : 4, |
| 364 | sk_protocol : 8, |
| 365 | #define SK_PROTOCOL_MAX U8_MAX |
| 366 | sk_type : 16; |
| 367 | kmemcheck_bitfield_end(flags); |
| 368 | int sk_wmem_queued; |
| 369 | gfp_t sk_allocation; |
| 370 | u32 sk_pacing_rate; /* bytes per second */ |
| 371 | netdev_features_t sk_route_caps; |
| 372 | netdev_features_t sk_route_nocaps; |
| 373 | int sk_gso_type; |
| 374 | unsigned int sk_gso_max_size; |
| 375 | u16 sk_gso_max_segs; |
| 376 | int sk_rcvlowat; |
| 377 | unsigned long sk_lingertime; |
| 378 | struct sk_buff_head sk_error_queue; |
| 379 | struct proto *sk_prot_creator; |
| 380 | rwlock_t sk_callback_lock; |
| 381 | int sk_err, |
| 382 | sk_err_soft; |
| 383 | unsigned short sk_ack_backlog; |
| 384 | unsigned short sk_max_ack_backlog; |
| 385 | __u32 sk_priority; |
| 386 | #if IS_ENABLED(CONFIG_NETPRIO_CGROUP) |
| 387 | __u32 sk_cgrp_prioidx; |
| 388 | #endif |
| 389 | struct pid *sk_peer_pid; |
| 390 | const struct cred *sk_peer_cred; |
| 391 | long sk_rcvtimeo; |
| 392 | long sk_sndtimeo; |
| 393 | void *sk_protinfo; |
| 394 | struct timer_list sk_timer; |
| 395 | ktime_t sk_stamp; |
| 396 | struct socket *sk_socket; |
| 397 | void *sk_user_data; |
| 398 | struct page_frag sk_frag; |
| 399 | struct sk_buff *sk_send_head; |
| 400 | __s32 sk_peek_off; |
| 401 | int sk_write_pending; |
| 402 | #ifdef CONFIG_SECURITY |
| 403 | void *sk_security; |
| 404 | #endif |
| 405 | __u32 sk_mark; |
| 406 | kuid_t sk_uid; |
| 407 | u32 sk_classid; |
| 408 | struct cg_proto *sk_cgrp; |
| 409 | uid_t knox_uid; |
| 410 | pid_t knox_pid; |
| 411 | void (*sk_state_change)(struct sock *sk); |
| 412 | void (*sk_data_ready)(struct sock *sk, int bytes); |
| 413 | void (*sk_write_space)(struct sock *sk); |
| 414 | void (*sk_error_report)(struct sock *sk); |
| 415 | int (*sk_backlog_rcv)(struct sock *sk, |
| 416 | struct sk_buff *skb); |
| 417 | void (*sk_destruct)(struct sock *sk); |
| 418 | }; |
| 419 | |
| 420 | /* |
| 421 | * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK |
| 422 | * or not whether his port will be reused by someone else. SK_FORCE_REUSE |
| 423 | * on a socket means that the socket will reuse everybody else's port |
| 424 | * without looking at the other's sk_reuse value. |
| 425 | */ |
| 426 | |
| 427 | #define SK_NO_REUSE 0 |
| 428 | #define SK_CAN_REUSE 1 |
| 429 | #define SK_FORCE_REUSE 2 |
| 430 | |
| 431 | static inline int sk_peek_offset(struct sock *sk, int flags) |
| 432 | { |
| 433 | if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0)) |
| 434 | return sk->sk_peek_off; |
| 435 | else |
| 436 | return 0; |
| 437 | } |
| 438 | |
| 439 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) |
| 440 | { |
| 441 | if (sk->sk_peek_off >= 0) { |
| 442 | if (sk->sk_peek_off >= val) |
| 443 | sk->sk_peek_off -= val; |
| 444 | else |
| 445 | sk->sk_peek_off = 0; |
| 446 | } |
| 447 | } |
| 448 | |
| 449 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) |
| 450 | { |
| 451 | if (sk->sk_peek_off >= 0) |
| 452 | sk->sk_peek_off += val; |
| 453 | } |
| 454 | |
| 455 | /* |
| 456 | * Hashed lists helper routines |
| 457 | */ |
| 458 | static inline struct sock *sk_entry(const struct hlist_node *node) |
| 459 | { |
| 460 | return hlist_entry(node, struct sock, sk_node); |
| 461 | } |
| 462 | |
| 463 | static inline struct sock *__sk_head(const struct hlist_head *head) |
| 464 | { |
| 465 | return hlist_entry(head->first, struct sock, sk_node); |
| 466 | } |
| 467 | |
| 468 | static inline struct sock *sk_head(const struct hlist_head *head) |
| 469 | { |
| 470 | return hlist_empty(head) ? NULL : __sk_head(head); |
| 471 | } |
| 472 | |
| 473 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
| 474 | { |
| 475 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); |
| 476 | } |
| 477 | |
| 478 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) |
| 479 | { |
| 480 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); |
| 481 | } |
| 482 | |
| 483 | static inline struct sock *sk_next(const struct sock *sk) |
| 484 | { |
| 485 | return sk->sk_node.next ? |
| 486 | hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL; |
| 487 | } |
| 488 | |
| 489 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
| 490 | { |
| 491 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? |
| 492 | hlist_nulls_entry(sk->sk_nulls_node.next, |
| 493 | struct sock, sk_nulls_node) : |
| 494 | NULL; |
| 495 | } |
| 496 | |
| 497 | static inline bool sk_unhashed(const struct sock *sk) |
| 498 | { |
| 499 | return hlist_unhashed(&sk->sk_node); |
| 500 | } |
| 501 | |
| 502 | static inline bool sk_hashed(const struct sock *sk) |
| 503 | { |
| 504 | return !sk_unhashed(sk); |
| 505 | } |
| 506 | |
| 507 | static inline void sk_node_init(struct hlist_node *node) |
| 508 | { |
| 509 | node->pprev = NULL; |
| 510 | } |
| 511 | |
| 512 | static inline void sk_nulls_node_init(struct hlist_nulls_node *node) |
| 513 | { |
| 514 | node->pprev = NULL; |
| 515 | } |
| 516 | |
| 517 | static inline void __sk_del_node(struct sock *sk) |
| 518 | { |
| 519 | __hlist_del(&sk->sk_node); |
| 520 | } |
| 521 | |
| 522 | /* NB: equivalent to hlist_del_init_rcu */ |
| 523 | static inline bool __sk_del_node_init(struct sock *sk) |
| 524 | { |
| 525 | if (sk_hashed(sk)) { |
| 526 | __sk_del_node(sk); |
| 527 | sk_node_init(&sk->sk_node); |
| 528 | return true; |
| 529 | } |
| 530 | return false; |
| 531 | } |
| 532 | |
| 533 | /* Grab socket reference count. This operation is valid only |
| 534 | when sk is ALREADY grabbed f.e. it is found in hash table |
| 535 | or a list and the lookup is made under lock preventing hash table |
| 536 | modifications. |
| 537 | */ |
| 538 | |
| 539 | static inline void sock_hold(struct sock *sk) |
| 540 | { |
| 541 | atomic_inc(&sk->sk_refcnt); |
| 542 | } |
| 543 | |
| 544 | /* Ungrab socket in the context, which assumes that socket refcnt |
| 545 | cannot hit zero, f.e. it is true in context of any socketcall. |
| 546 | */ |
| 547 | static inline void __sock_put(struct sock *sk) |
| 548 | { |
| 549 | atomic_dec(&sk->sk_refcnt); |
| 550 | } |
| 551 | |
| 552 | static inline bool sk_del_node_init(struct sock *sk) |
| 553 | { |
| 554 | bool rc = __sk_del_node_init(sk); |
| 555 | |
| 556 | if (rc) { |
| 557 | /* paranoid for a while -acme */ |
| 558 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
| 559 | __sock_put(sk); |
| 560 | } |
| 561 | return rc; |
| 562 | } |
| 563 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
| 564 | |
| 565 | static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk) |
| 566 | { |
| 567 | if (sk_hashed(sk)) { |
| 568 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
| 569 | return true; |
| 570 | } |
| 571 | return false; |
| 572 | } |
| 573 | |
| 574 | static inline bool sk_nulls_del_node_init_rcu(struct sock *sk) |
| 575 | { |
| 576 | bool rc = __sk_nulls_del_node_init_rcu(sk); |
| 577 | |
| 578 | if (rc) { |
| 579 | /* paranoid for a while -acme */ |
| 580 | WARN_ON(atomic_read(&sk->sk_refcnt) == 1); |
| 581 | __sock_put(sk); |
| 582 | } |
| 583 | return rc; |
| 584 | } |
| 585 | |
| 586 | static inline void __sk_add_node(struct sock *sk, struct hlist_head *list) |
| 587 | { |
| 588 | hlist_add_head(&sk->sk_node, list); |
| 589 | } |
| 590 | |
| 591 | static inline void sk_add_node(struct sock *sk, struct hlist_head *list) |
| 592 | { |
| 593 | sock_hold(sk); |
| 594 | __sk_add_node(sk, list); |
| 595 | } |
| 596 | |
| 597 | static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
| 598 | { |
| 599 | sock_hold(sk); |
| 600 | hlist_add_head_rcu(&sk->sk_node, list); |
| 601 | } |
| 602 | |
| 603 | static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
| 604 | { |
| 605 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
| 606 | } |
| 607 | |
| 608 | static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
| 609 | { |
| 610 | sock_hold(sk); |
| 611 | __sk_nulls_add_node_rcu(sk, list); |
| 612 | } |
| 613 | |
| 614 | static inline void __sk_del_bind_node(struct sock *sk) |
| 615 | { |
| 616 | __hlist_del(&sk->sk_bind_node); |
| 617 | } |
| 618 | |
| 619 | static inline void sk_add_bind_node(struct sock *sk, |
| 620 | struct hlist_head *list) |
| 621 | { |
| 622 | hlist_add_head(&sk->sk_bind_node, list); |
| 623 | } |
| 624 | |
| 625 | #define sk_for_each(__sk, list) \ |
| 626 | hlist_for_each_entry(__sk, list, sk_node) |
| 627 | #define sk_for_each_rcu(__sk, list) \ |
| 628 | hlist_for_each_entry_rcu(__sk, list, sk_node) |
| 629 | #define sk_nulls_for_each(__sk, node, list) \ |
| 630 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) |
| 631 | #define sk_nulls_for_each_rcu(__sk, node, list) \ |
| 632 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) |
| 633 | #define sk_for_each_from(__sk) \ |
| 634 | hlist_for_each_entry_from(__sk, sk_node) |
| 635 | #define sk_nulls_for_each_from(__sk, node) \ |
| 636 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ |
| 637 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) |
| 638 | #define sk_for_each_safe(__sk, tmp, list) \ |
| 639 | hlist_for_each_entry_safe(__sk, tmp, list, sk_node) |
| 640 | #define sk_for_each_bound(__sk, list) \ |
| 641 | hlist_for_each_entry(__sk, list, sk_bind_node) |
| 642 | |
| 643 | static inline struct user_namespace *sk_user_ns(struct sock *sk) |
| 644 | { |
| 645 | /* Careful only use this in a context where these parameters |
| 646 | * can not change and must all be valid, such as recvmsg from |
| 647 | * userspace. |
| 648 | */ |
| 649 | return sk->sk_socket->file->f_cred->user_ns; |
| 650 | } |
| 651 | |
| 652 | /* Sock flags */ |
| 653 | enum sock_flags { |
| 654 | SOCK_DEAD, |
| 655 | SOCK_DONE, |
| 656 | SOCK_URGINLINE, |
| 657 | SOCK_KEEPOPEN, |
| 658 | SOCK_LINGER, |
| 659 | SOCK_DESTROY, |
| 660 | SOCK_BROADCAST, |
| 661 | SOCK_TIMESTAMP, |
| 662 | SOCK_ZAPPED, |
| 663 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ |
| 664 | SOCK_DBG, /* %SO_DEBUG setting */ |
| 665 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ |
| 666 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
| 667 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
| 668 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ |
| 669 | SOCK_MEMALLOC, /* VM depends on this socket for swapping */ |
| 670 | SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */ |
| 671 | SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */ |
| 672 | SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */ |
| 673 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
| 674 | SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */ |
| 675 | SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */ |
| 676 | SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */ |
| 677 | SOCK_FASYNC, /* fasync() active */ |
| 678 | SOCK_RXQ_OVFL, |
| 679 | SOCK_ZEROCOPY, /* buffers from userspace */ |
| 680 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
| 681 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
| 682 | * Will use last 4 bytes of packet sent from |
| 683 | * user-space instead. |
| 684 | */ |
| 685 | SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */ |
| 686 | SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */ |
| 687 | }; |
| 688 | |
| 689 | #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)) |
| 690 | |
| 691 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
| 692 | { |
| 693 | nsk->sk_flags = osk->sk_flags; |
| 694 | } |
| 695 | |
| 696 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
| 697 | { |
| 698 | __set_bit(flag, &sk->sk_flags); |
| 699 | } |
| 700 | |
| 701 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) |
| 702 | { |
| 703 | __clear_bit(flag, &sk->sk_flags); |
| 704 | } |
| 705 | |
| 706 | static inline bool sock_flag(const struct sock *sk, enum sock_flags flag) |
| 707 | { |
| 708 | return test_bit(flag, &sk->sk_flags); |
| 709 | } |
| 710 | |
| 711 | #ifdef CONFIG_NET |
| 712 | extern struct static_key memalloc_socks; |
| 713 | static inline int sk_memalloc_socks(void) |
| 714 | { |
| 715 | return static_key_false(&memalloc_socks); |
| 716 | } |
| 717 | #else |
| 718 | |
| 719 | static inline int sk_memalloc_socks(void) |
| 720 | { |
| 721 | return 0; |
| 722 | } |
| 723 | |
| 724 | #endif |
| 725 | |
| 726 | static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask) |
| 727 | { |
| 728 | return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC); |
| 729 | } |
| 730 | |
| 731 | static inline void sk_acceptq_removed(struct sock *sk) |
| 732 | { |
| 733 | sk->sk_ack_backlog--; |
| 734 | } |
| 735 | |
| 736 | static inline void sk_acceptq_added(struct sock *sk) |
| 737 | { |
| 738 | sk->sk_ack_backlog++; |
| 739 | } |
| 740 | |
| 741 | static inline bool sk_acceptq_is_full(const struct sock *sk) |
| 742 | { |
| 743 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
| 744 | } |
| 745 | |
| 746 | /* |
| 747 | * Compute minimal free write space needed to queue new packets. |
| 748 | */ |
| 749 | static inline int sk_stream_min_wspace(const struct sock *sk) |
| 750 | { |
| 751 | return sk->sk_wmem_queued >> 1; |
| 752 | } |
| 753 | |
| 754 | static inline int sk_stream_wspace(const struct sock *sk) |
| 755 | { |
| 756 | return sk->sk_sndbuf - sk->sk_wmem_queued; |
| 757 | } |
| 758 | |
| 759 | extern void sk_stream_write_space(struct sock *sk); |
| 760 | |
| 761 | static inline bool sk_stream_memory_free(const struct sock *sk) |
| 762 | { |
| 763 | return sk->sk_wmem_queued < sk->sk_sndbuf; |
| 764 | } |
| 765 | |
| 766 | /* OOB backlog add */ |
| 767 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
| 768 | { |
| 769 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
| 770 | skb_dst_force(skb); |
| 771 | |
| 772 | if (!sk->sk_backlog.tail) |
| 773 | sk->sk_backlog.head = skb; |
| 774 | else |
| 775 | sk->sk_backlog.tail->next = skb; |
| 776 | |
| 777 | sk->sk_backlog.tail = skb; |
| 778 | skb->next = NULL; |
| 779 | } |
| 780 | |
| 781 | /* |
| 782 | * Take into account size of receive queue and backlog queue |
| 783 | * Do not take into account this skb truesize, |
| 784 | * to allow even a single big packet to come. |
| 785 | */ |
| 786 | static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb, |
| 787 | unsigned int limit) |
| 788 | { |
| 789 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); |
| 790 | |
| 791 | return qsize > limit; |
| 792 | } |
| 793 | |
| 794 | /* The per-socket spinlock must be held here. */ |
| 795 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb, |
| 796 | unsigned int limit) |
| 797 | { |
| 798 | if (sk_rcvqueues_full(sk, skb, limit)) |
| 799 | return -ENOBUFS; |
| 800 | |
| 801 | /* |
| 802 | * If the skb was allocated from pfmemalloc reserves, only |
| 803 | * allow SOCK_MEMALLOC sockets to use it as this socket is |
| 804 | * helping free memory |
| 805 | */ |
| 806 | if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC)) |
| 807 | return -ENOMEM; |
| 808 | |
| 809 | __sk_add_backlog(sk, skb); |
| 810 | sk->sk_backlog.len += skb->truesize; |
| 811 | return 0; |
| 812 | } |
| 813 | |
| 814 | extern int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
| 815 | |
| 816 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
| 817 | { |
| 818 | if (sk_memalloc_socks() && skb_pfmemalloc(skb)) |
| 819 | return __sk_backlog_rcv(sk, skb); |
| 820 | |
| 821 | return sk->sk_backlog_rcv(sk, skb); |
| 822 | } |
| 823 | |
| 824 | static inline void sock_rps_record_flow(const struct sock *sk) |
| 825 | { |
| 826 | #ifdef CONFIG_RPS |
| 827 | struct rps_sock_flow_table *sock_flow_table; |
| 828 | |
| 829 | rcu_read_lock(); |
| 830 | sock_flow_table = rcu_dereference(rps_sock_flow_table); |
| 831 | rps_record_sock_flow(sock_flow_table, sk->sk_rxhash); |
| 832 | rcu_read_unlock(); |
| 833 | #endif |
| 834 | } |
| 835 | |
| 836 | static inline void sock_rps_reset_flow(const struct sock *sk) |
| 837 | { |
| 838 | #ifdef CONFIG_RPS |
| 839 | struct rps_sock_flow_table *sock_flow_table; |
| 840 | |
| 841 | rcu_read_lock(); |
| 842 | sock_flow_table = rcu_dereference(rps_sock_flow_table); |
| 843 | rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash); |
| 844 | rcu_read_unlock(); |
| 845 | #endif |
| 846 | } |
| 847 | |
| 848 | static inline void sock_rps_save_rxhash(struct sock *sk, |
| 849 | const struct sk_buff *skb) |
| 850 | { |
| 851 | #ifdef CONFIG_RPS |
| 852 | if (unlikely(sk->sk_rxhash != skb->rxhash)) { |
| 853 | sock_rps_reset_flow(sk); |
| 854 | sk->sk_rxhash = skb->rxhash; |
| 855 | } |
| 856 | #endif |
| 857 | } |
| 858 | |
| 859 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
| 860 | { |
| 861 | #ifdef CONFIG_RPS |
| 862 | sock_rps_reset_flow(sk); |
| 863 | sk->sk_rxhash = 0; |
| 864 | #endif |
| 865 | } |
| 866 | |
| 867 | #define sk_wait_event(__sk, __timeo, __condition) \ |
| 868 | ({ int __rc; \ |
| 869 | release_sock(__sk); \ |
| 870 | __rc = __condition; \ |
| 871 | if (!__rc) { \ |
| 872 | *(__timeo) = schedule_timeout(*(__timeo)); \ |
| 873 | } \ |
| 874 | lock_sock(__sk); \ |
| 875 | __rc = __condition; \ |
| 876 | __rc; \ |
| 877 | }) |
| 878 | |
| 879 | extern int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
| 880 | extern int sk_stream_wait_memory(struct sock *sk, long *timeo_p); |
| 881 | extern void sk_stream_wait_close(struct sock *sk, long timeo_p); |
| 882 | extern int sk_stream_error(struct sock *sk, int flags, int err); |
| 883 | extern void sk_stream_kill_queues(struct sock *sk); |
| 884 | extern void sk_set_memalloc(struct sock *sk); |
| 885 | extern void sk_clear_memalloc(struct sock *sk); |
| 886 | |
| 887 | extern int sk_wait_data(struct sock *sk, long *timeo); |
| 888 | |
| 889 | struct request_sock_ops; |
| 890 | struct timewait_sock_ops; |
| 891 | struct inet_hashinfo; |
| 892 | struct raw_hashinfo; |
| 893 | struct module; |
| 894 | |
| 895 | /* |
| 896 | * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes |
| 897 | * un-modified. Special care is taken when initializing object to zero. |
| 898 | */ |
| 899 | static inline void sk_prot_clear_nulls(struct sock *sk, int size) |
| 900 | { |
| 901 | if (offsetof(struct sock, sk_node.next) != 0) |
| 902 | memset(sk, 0, offsetof(struct sock, sk_node.next)); |
| 903 | memset(&sk->sk_node.pprev, 0, |
| 904 | size - offsetof(struct sock, sk_node.pprev)); |
| 905 | } |
| 906 | |
| 907 | /* Networking protocol blocks we attach to sockets. |
| 908 | * socket layer -> transport layer interface |
| 909 | * transport -> network interface is defined by struct inet_proto |
| 910 | */ |
| 911 | struct proto { |
| 912 | void (*close)(struct sock *sk, |
| 913 | long timeout); |
| 914 | int (*connect)(struct sock *sk, |
| 915 | struct sockaddr *uaddr, |
| 916 | int addr_len); |
| 917 | int (*disconnect)(struct sock *sk, int flags); |
| 918 | |
| 919 | struct sock * (*accept)(struct sock *sk, int flags, int *err); |
| 920 | |
| 921 | int (*ioctl)(struct sock *sk, int cmd, |
| 922 | unsigned long arg); |
| 923 | int (*init)(struct sock *sk); |
| 924 | void (*destroy)(struct sock *sk); |
| 925 | void (*shutdown)(struct sock *sk, int how); |
| 926 | int (*setsockopt)(struct sock *sk, int level, |
| 927 | int optname, char __user *optval, |
| 928 | unsigned int optlen); |
| 929 | int (*getsockopt)(struct sock *sk, int level, |
| 930 | int optname, char __user *optval, |
| 931 | int __user *option); |
| 932 | #ifdef CONFIG_COMPAT |
| 933 | int (*compat_setsockopt)(struct sock *sk, |
| 934 | int level, |
| 935 | int optname, char __user *optval, |
| 936 | unsigned int optlen); |
| 937 | int (*compat_getsockopt)(struct sock *sk, |
| 938 | int level, |
| 939 | int optname, char __user *optval, |
| 940 | int __user *option); |
| 941 | int (*compat_ioctl)(struct sock *sk, |
| 942 | unsigned int cmd, unsigned long arg); |
| 943 | #endif |
| 944 | int (*sendmsg)(struct kiocb *iocb, struct sock *sk, |
| 945 | struct msghdr *msg, size_t len); |
| 946 | int (*recvmsg)(struct kiocb *iocb, struct sock *sk, |
| 947 | struct msghdr *msg, |
| 948 | size_t len, int noblock, int flags, |
| 949 | int *addr_len); |
| 950 | int (*sendpage)(struct sock *sk, struct page *page, |
| 951 | int offset, size_t size, int flags); |
| 952 | int (*bind)(struct sock *sk, |
| 953 | struct sockaddr *uaddr, int addr_len); |
| 954 | |
| 955 | int (*backlog_rcv) (struct sock *sk, |
| 956 | struct sk_buff *skb); |
| 957 | |
| 958 | void (*release_cb)(struct sock *sk); |
| 959 | |
| 960 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
| 961 | void (*hash)(struct sock *sk); |
| 962 | void (*unhash)(struct sock *sk); |
| 963 | void (*rehash)(struct sock *sk); |
| 964 | int (*get_port)(struct sock *sk, unsigned short snum); |
| 965 | void (*clear_sk)(struct sock *sk, int size); |
| 966 | |
| 967 | /* Keeping track of sockets in use */ |
| 968 | #ifdef CONFIG_PROC_FS |
| 969 | unsigned int inuse_idx; |
| 970 | #endif |
| 971 | |
| 972 | /* Memory pressure */ |
| 973 | void (*enter_memory_pressure)(struct sock *sk); |
| 974 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
| 975 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
| 976 | /* |
| 977 | * Pressure flag: try to collapse. |
| 978 | * Technical note: it is used by multiple contexts non atomically. |
| 979 | * All the __sk_mem_schedule() is of this nature: accounting |
| 980 | * is strict, actions are advisory and have some latency. |
| 981 | */ |
| 982 | int *memory_pressure; |
| 983 | long *sysctl_mem; |
| 984 | int *sysctl_wmem; |
| 985 | int *sysctl_rmem; |
| 986 | int max_header; |
| 987 | bool no_autobind; |
| 988 | |
| 989 | struct kmem_cache *slab; |
| 990 | unsigned int obj_size; |
| 991 | int slab_flags; |
| 992 | |
| 993 | struct percpu_counter *orphan_count; |
| 994 | |
| 995 | struct request_sock_ops *rsk_prot; |
| 996 | struct timewait_sock_ops *twsk_prot; |
| 997 | |
| 998 | union { |
| 999 | struct inet_hashinfo *hashinfo; |
| 1000 | struct udp_table *udp_table; |
| 1001 | struct raw_hashinfo *raw_hash; |
| 1002 | } h; |
| 1003 | |
| 1004 | struct module *owner; |
| 1005 | |
| 1006 | char name[32]; |
| 1007 | |
| 1008 | struct list_head node; |
| 1009 | #ifdef SOCK_REFCNT_DEBUG |
| 1010 | atomic_t socks; |
| 1011 | #endif |
| 1012 | #ifdef CONFIG_MEMCG_KMEM |
| 1013 | /* |
| 1014 | * cgroup specific init/deinit functions. Called once for all |
| 1015 | * protocols that implement it, from cgroups populate function. |
| 1016 | * This function has to setup any files the protocol want to |
| 1017 | * appear in the kmem cgroup filesystem. |
| 1018 | */ |
| 1019 | int (*init_cgroup)(struct mem_cgroup *memcg, |
| 1020 | struct cgroup_subsys *ss); |
| 1021 | void (*destroy_cgroup)(struct mem_cgroup *memcg); |
| 1022 | struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg); |
| 1023 | #endif |
| 1024 | int (*diag_destroy)(struct sock *sk, int err); |
| 1025 | }; |
| 1026 | |
| 1027 | /* |
| 1028 | * Bits in struct cg_proto.flags |
| 1029 | */ |
| 1030 | enum cg_proto_flags { |
| 1031 | /* Currently active and new sockets should be assigned to cgroups */ |
| 1032 | MEMCG_SOCK_ACTIVE, |
| 1033 | /* It was ever activated; we must disarm static keys on destruction */ |
| 1034 | MEMCG_SOCK_ACTIVATED, |
| 1035 | }; |
| 1036 | |
| 1037 | struct cg_proto { |
| 1038 | void (*enter_memory_pressure)(struct sock *sk); |
| 1039 | struct res_counter *memory_allocated; /* Current allocated memory. */ |
| 1040 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
| 1041 | int *memory_pressure; |
| 1042 | long *sysctl_mem; |
| 1043 | unsigned long flags; |
| 1044 | /* |
| 1045 | * memcg field is used to find which memcg we belong directly |
| 1046 | * Each memcg struct can hold more than one cg_proto, so container_of |
| 1047 | * won't really cut. |
| 1048 | * |
| 1049 | * The elegant solution would be having an inverse function to |
| 1050 | * proto_cgroup in struct proto, but that means polluting the structure |
| 1051 | * for everybody, instead of just for memcg users. |
| 1052 | */ |
| 1053 | struct mem_cgroup *memcg; |
| 1054 | }; |
| 1055 | |
| 1056 | extern int proto_register(struct proto *prot, int alloc_slab); |
| 1057 | extern void proto_unregister(struct proto *prot); |
| 1058 | |
| 1059 | static inline bool memcg_proto_active(struct cg_proto *cg_proto) |
| 1060 | { |
| 1061 | return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags); |
| 1062 | } |
| 1063 | |
| 1064 | static inline bool memcg_proto_activated(struct cg_proto *cg_proto) |
| 1065 | { |
| 1066 | return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags); |
| 1067 | } |
| 1068 | |
| 1069 | #ifdef SOCK_REFCNT_DEBUG |
| 1070 | static inline void sk_refcnt_debug_inc(struct sock *sk) |
| 1071 | { |
| 1072 | atomic_inc(&sk->sk_prot->socks); |
| 1073 | } |
| 1074 | |
| 1075 | static inline void sk_refcnt_debug_dec(struct sock *sk) |
| 1076 | { |
| 1077 | atomic_dec(&sk->sk_prot->socks); |
| 1078 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", |
| 1079 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); |
| 1080 | } |
| 1081 | |
| 1082 | static inline void sk_refcnt_debug_release(const struct sock *sk) |
| 1083 | { |
| 1084 | if (atomic_read(&sk->sk_refcnt) != 1) |
| 1085 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", |
| 1086 | sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt)); |
| 1087 | } |
| 1088 | #else /* SOCK_REFCNT_DEBUG */ |
| 1089 | #define sk_refcnt_debug_inc(sk) do { } while (0) |
| 1090 | #define sk_refcnt_debug_dec(sk) do { } while (0) |
| 1091 | #define sk_refcnt_debug_release(sk) do { } while (0) |
| 1092 | #endif /* SOCK_REFCNT_DEBUG */ |
| 1093 | |
| 1094 | #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET) |
| 1095 | extern struct static_key memcg_socket_limit_enabled; |
| 1096 | static inline struct cg_proto *parent_cg_proto(struct proto *proto, |
| 1097 | struct cg_proto *cg_proto) |
| 1098 | { |
| 1099 | return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg)); |
| 1100 | } |
| 1101 | #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled) |
| 1102 | #else |
| 1103 | #define mem_cgroup_sockets_enabled 0 |
| 1104 | static inline struct cg_proto *parent_cg_proto(struct proto *proto, |
| 1105 | struct cg_proto *cg_proto) |
| 1106 | { |
| 1107 | return NULL; |
| 1108 | } |
| 1109 | #endif |
| 1110 | |
| 1111 | |
| 1112 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
| 1113 | { |
| 1114 | return sk->sk_prot->memory_pressure != NULL; |
| 1115 | } |
| 1116 | |
| 1117 | static inline bool sk_under_memory_pressure(const struct sock *sk) |
| 1118 | { |
| 1119 | if (!sk->sk_prot->memory_pressure) |
| 1120 | return false; |
| 1121 | |
| 1122 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 1123 | return !!*sk->sk_cgrp->memory_pressure; |
| 1124 | |
| 1125 | return !!*sk->sk_prot->memory_pressure; |
| 1126 | } |
| 1127 | |
| 1128 | static inline void sk_leave_memory_pressure(struct sock *sk) |
| 1129 | { |
| 1130 | int *memory_pressure = sk->sk_prot->memory_pressure; |
| 1131 | |
| 1132 | if (!memory_pressure) |
| 1133 | return; |
| 1134 | |
| 1135 | if (*memory_pressure) |
| 1136 | *memory_pressure = 0; |
| 1137 | |
| 1138 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1139 | struct cg_proto *cg_proto = sk->sk_cgrp; |
| 1140 | struct proto *prot = sk->sk_prot; |
| 1141 | |
| 1142 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) |
| 1143 | if (*cg_proto->memory_pressure) |
| 1144 | *cg_proto->memory_pressure = 0; |
| 1145 | } |
| 1146 | |
| 1147 | } |
| 1148 | |
| 1149 | static inline void sk_enter_memory_pressure(struct sock *sk) |
| 1150 | { |
| 1151 | if (!sk->sk_prot->enter_memory_pressure) |
| 1152 | return; |
| 1153 | |
| 1154 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1155 | struct cg_proto *cg_proto = sk->sk_cgrp; |
| 1156 | struct proto *prot = sk->sk_prot; |
| 1157 | |
| 1158 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) |
| 1159 | cg_proto->enter_memory_pressure(sk); |
| 1160 | } |
| 1161 | |
| 1162 | sk->sk_prot->enter_memory_pressure(sk); |
| 1163 | } |
| 1164 | |
| 1165 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) |
| 1166 | { |
| 1167 | long *prot = sk->sk_prot->sysctl_mem; |
| 1168 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 1169 | prot = sk->sk_cgrp->sysctl_mem; |
| 1170 | return prot[index]; |
| 1171 | } |
| 1172 | |
| 1173 | static inline void memcg_memory_allocated_add(struct cg_proto *prot, |
| 1174 | unsigned long amt, |
| 1175 | int *parent_status) |
| 1176 | { |
| 1177 | struct res_counter *fail; |
| 1178 | int ret; |
| 1179 | |
| 1180 | ret = res_counter_charge_nofail(prot->memory_allocated, |
| 1181 | amt << PAGE_SHIFT, &fail); |
| 1182 | if (ret < 0) |
| 1183 | *parent_status = OVER_LIMIT; |
| 1184 | } |
| 1185 | |
| 1186 | static inline void memcg_memory_allocated_sub(struct cg_proto *prot, |
| 1187 | unsigned long amt) |
| 1188 | { |
| 1189 | res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT); |
| 1190 | } |
| 1191 | |
| 1192 | static inline u64 memcg_memory_allocated_read(struct cg_proto *prot) |
| 1193 | { |
| 1194 | u64 ret; |
| 1195 | ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE); |
| 1196 | return ret >> PAGE_SHIFT; |
| 1197 | } |
| 1198 | |
| 1199 | static inline long |
| 1200 | sk_memory_allocated(const struct sock *sk) |
| 1201 | { |
| 1202 | struct proto *prot = sk->sk_prot; |
| 1203 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 1204 | return memcg_memory_allocated_read(sk->sk_cgrp); |
| 1205 | |
| 1206 | return atomic_long_read(prot->memory_allocated); |
| 1207 | } |
| 1208 | |
| 1209 | static inline long |
| 1210 | sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status) |
| 1211 | { |
| 1212 | struct proto *prot = sk->sk_prot; |
| 1213 | |
| 1214 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1215 | memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status); |
| 1216 | /* update the root cgroup regardless */ |
| 1217 | atomic_long_add_return(amt, prot->memory_allocated); |
| 1218 | return memcg_memory_allocated_read(sk->sk_cgrp); |
| 1219 | } |
| 1220 | |
| 1221 | return atomic_long_add_return(amt, prot->memory_allocated); |
| 1222 | } |
| 1223 | |
| 1224 | static inline void |
| 1225 | sk_memory_allocated_sub(struct sock *sk, int amt) |
| 1226 | { |
| 1227 | struct proto *prot = sk->sk_prot; |
| 1228 | |
| 1229 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 1230 | memcg_memory_allocated_sub(sk->sk_cgrp, amt); |
| 1231 | |
| 1232 | atomic_long_sub(amt, prot->memory_allocated); |
| 1233 | } |
| 1234 | |
| 1235 | static inline void sk_sockets_allocated_dec(struct sock *sk) |
| 1236 | { |
| 1237 | struct proto *prot = sk->sk_prot; |
| 1238 | |
| 1239 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1240 | struct cg_proto *cg_proto = sk->sk_cgrp; |
| 1241 | |
| 1242 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) |
| 1243 | percpu_counter_dec(cg_proto->sockets_allocated); |
| 1244 | } |
| 1245 | |
| 1246 | percpu_counter_dec(prot->sockets_allocated); |
| 1247 | } |
| 1248 | |
| 1249 | static inline void sk_sockets_allocated_inc(struct sock *sk) |
| 1250 | { |
| 1251 | struct proto *prot = sk->sk_prot; |
| 1252 | |
| 1253 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) { |
| 1254 | struct cg_proto *cg_proto = sk->sk_cgrp; |
| 1255 | |
| 1256 | for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto)) |
| 1257 | percpu_counter_inc(cg_proto->sockets_allocated); |
| 1258 | } |
| 1259 | |
| 1260 | percpu_counter_inc(prot->sockets_allocated); |
| 1261 | } |
| 1262 | |
| 1263 | static inline int |
| 1264 | sk_sockets_allocated_read_positive(struct sock *sk) |
| 1265 | { |
| 1266 | struct proto *prot = sk->sk_prot; |
| 1267 | |
| 1268 | if (mem_cgroup_sockets_enabled && sk->sk_cgrp) |
| 1269 | return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated); |
| 1270 | |
| 1271 | return percpu_counter_read_positive(prot->sockets_allocated); |
| 1272 | } |
| 1273 | |
| 1274 | static inline int |
| 1275 | proto_sockets_allocated_sum_positive(struct proto *prot) |
| 1276 | { |
| 1277 | return percpu_counter_sum_positive(prot->sockets_allocated); |
| 1278 | } |
| 1279 | |
| 1280 | static inline long |
| 1281 | proto_memory_allocated(struct proto *prot) |
| 1282 | { |
| 1283 | return atomic_long_read(prot->memory_allocated); |
| 1284 | } |
| 1285 | |
| 1286 | static inline bool |
| 1287 | proto_memory_pressure(struct proto *prot) |
| 1288 | { |
| 1289 | if (!prot->memory_pressure) |
| 1290 | return false; |
| 1291 | return !!*prot->memory_pressure; |
| 1292 | } |
| 1293 | |
| 1294 | |
| 1295 | #ifdef CONFIG_PROC_FS |
| 1296 | /* Called with local bh disabled */ |
| 1297 | extern void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
| 1298 | extern int sock_prot_inuse_get(struct net *net, struct proto *proto); |
| 1299 | #else |
| 1300 | static inline void sock_prot_inuse_add(struct net *net, struct proto *prot, |
| 1301 | int inc) |
| 1302 | { |
| 1303 | } |
| 1304 | #endif |
| 1305 | |
| 1306 | |
| 1307 | /* With per-bucket locks this operation is not-atomic, so that |
| 1308 | * this version is not worse. |
| 1309 | */ |
| 1310 | static inline void __sk_prot_rehash(struct sock *sk) |
| 1311 | { |
| 1312 | sk->sk_prot->unhash(sk); |
| 1313 | sk->sk_prot->hash(sk); |
| 1314 | } |
| 1315 | |
| 1316 | void sk_prot_clear_portaddr_nulls(struct sock *sk, int size); |
| 1317 | |
| 1318 | /* About 10 seconds */ |
| 1319 | #define SOCK_DESTROY_TIME (10*HZ) |
| 1320 | |
| 1321 | /* Sockets 0-1023 can't be bound to unless you are superuser */ |
| 1322 | #define PROT_SOCK 1024 |
| 1323 | |
| 1324 | #define SHUTDOWN_MASK 3 |
| 1325 | #define RCV_SHUTDOWN 1 |
| 1326 | #define SEND_SHUTDOWN 2 |
| 1327 | |
| 1328 | #define SOCK_SNDBUF_LOCK 1 |
| 1329 | #define SOCK_RCVBUF_LOCK 2 |
| 1330 | #define SOCK_BINDADDR_LOCK 4 |
| 1331 | #define SOCK_BINDPORT_LOCK 8 |
| 1332 | |
| 1333 | /* sock_iocb: used to kick off async processing of socket ios */ |
| 1334 | struct sock_iocb { |
| 1335 | struct list_head list; |
| 1336 | |
| 1337 | int flags; |
| 1338 | int size; |
| 1339 | struct socket *sock; |
| 1340 | struct sock *sk; |
| 1341 | struct scm_cookie *scm; |
| 1342 | struct msghdr *msg, async_msg; |
| 1343 | struct kiocb *kiocb; |
| 1344 | }; |
| 1345 | |
| 1346 | static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb) |
| 1347 | { |
| 1348 | return (struct sock_iocb *)iocb->private; |
| 1349 | } |
| 1350 | |
| 1351 | static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si) |
| 1352 | { |
| 1353 | return si->kiocb; |
| 1354 | } |
| 1355 | |
| 1356 | struct socket_alloc { |
| 1357 | struct socket socket; |
| 1358 | struct inode vfs_inode; |
| 1359 | }; |
| 1360 | |
| 1361 | static inline struct socket *SOCKET_I(struct inode *inode) |
| 1362 | { |
| 1363 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; |
| 1364 | } |
| 1365 | |
| 1366 | static inline struct inode *SOCK_INODE(struct socket *socket) |
| 1367 | { |
| 1368 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; |
| 1369 | } |
| 1370 | |
| 1371 | /* |
| 1372 | * Functions for memory accounting |
| 1373 | */ |
| 1374 | extern int __sk_mem_schedule(struct sock *sk, int size, int kind); |
| 1375 | void __sk_mem_reclaim(struct sock *sk, int amount); |
| 1376 | |
| 1377 | #define SK_MEM_QUANTUM ((int)PAGE_SIZE) |
| 1378 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) |
| 1379 | #define SK_MEM_SEND 0 |
| 1380 | #define SK_MEM_RECV 1 |
| 1381 | |
| 1382 | static inline int sk_mem_pages(int amt) |
| 1383 | { |
| 1384 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
| 1385 | } |
| 1386 | |
| 1387 | static inline bool sk_has_account(struct sock *sk) |
| 1388 | { |
| 1389 | /* return true if protocol supports memory accounting */ |
| 1390 | return !!sk->sk_prot->memory_allocated; |
| 1391 | } |
| 1392 | |
| 1393 | static inline bool sk_wmem_schedule(struct sock *sk, int size) |
| 1394 | { |
| 1395 | if (!sk_has_account(sk)) |
| 1396 | return true; |
| 1397 | return size <= sk->sk_forward_alloc || |
| 1398 | __sk_mem_schedule(sk, size, SK_MEM_SEND); |
| 1399 | } |
| 1400 | |
| 1401 | static inline bool |
| 1402 | sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size) |
| 1403 | { |
| 1404 | if (!sk_has_account(sk)) |
| 1405 | return true; |
| 1406 | return size<= sk->sk_forward_alloc || |
| 1407 | __sk_mem_schedule(sk, size, SK_MEM_RECV) || |
| 1408 | skb_pfmemalloc(skb); |
| 1409 | } |
| 1410 | |
| 1411 | static inline void sk_mem_reclaim(struct sock *sk) |
| 1412 | { |
| 1413 | if (!sk_has_account(sk)) |
| 1414 | return; |
| 1415 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) |
| 1416 | __sk_mem_reclaim(sk, sk->sk_forward_alloc); |
| 1417 | } |
| 1418 | |
| 1419 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
| 1420 | { |
| 1421 | if (!sk_has_account(sk)) |
| 1422 | return; |
| 1423 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) |
| 1424 | __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1); |
| 1425 | } |
| 1426 | |
| 1427 | static inline void sk_mem_charge(struct sock *sk, int size) |
| 1428 | { |
| 1429 | if (!sk_has_account(sk)) |
| 1430 | return; |
| 1431 | sk->sk_forward_alloc -= size; |
| 1432 | } |
| 1433 | |
| 1434 | static inline void sk_mem_uncharge(struct sock *sk, int size) |
| 1435 | { |
| 1436 | if (!sk_has_account(sk)) |
| 1437 | return; |
| 1438 | sk->sk_forward_alloc += size; |
| 1439 | |
| 1440 | /* Avoid a possible overflow. |
| 1441 | * TCP send queues can make this happen, if sk_mem_reclaim() |
| 1442 | * is not called and more than 2 GBytes are released at once. |
| 1443 | * |
| 1444 | * If we reach 2 MBytes, reclaim 1 MBytes right now, there is |
| 1445 | * no need to hold that much forward allocation anyway. |
| 1446 | */ |
| 1447 | if (unlikely(sk->sk_forward_alloc >= 1 << 21)) |
| 1448 | __sk_mem_reclaim(sk, 1 << 20); |
| 1449 | } |
| 1450 | |
| 1451 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) |
| 1452 | { |
| 1453 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
| 1454 | sk->sk_wmem_queued -= skb->truesize; |
| 1455 | sk_mem_uncharge(sk, skb->truesize); |
| 1456 | __kfree_skb(skb); |
| 1457 | } |
| 1458 | |
| 1459 | /* Used by processes to "lock" a socket state, so that |
| 1460 | * interrupts and bottom half handlers won't change it |
| 1461 | * from under us. It essentially blocks any incoming |
| 1462 | * packets, so that we won't get any new data or any |
| 1463 | * packets that change the state of the socket. |
| 1464 | * |
| 1465 | * While locked, BH processing will add new packets to |
| 1466 | * the backlog queue. This queue is processed by the |
| 1467 | * owner of the socket lock right before it is released. |
| 1468 | * |
| 1469 | * Since ~2.3.5 it is also exclusive sleep lock serializing |
| 1470 | * accesses from user process context. |
| 1471 | */ |
| 1472 | #define sock_owned_by_user(sk) ((sk)->sk_lock.owned) |
| 1473 | |
| 1474 | static inline void sock_release_ownership(struct sock *sk) |
| 1475 | { |
| 1476 | sk->sk_lock.owned = 0; |
| 1477 | } |
| 1478 | |
| 1479 | /* |
| 1480 | * Macro so as to not evaluate some arguments when |
| 1481 | * lockdep is not enabled. |
| 1482 | * |
| 1483 | * Mark both the sk_lock and the sk_lock.slock as a |
| 1484 | * per-address-family lock class. |
| 1485 | */ |
| 1486 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
| 1487 | do { \ |
| 1488 | sk->sk_lock.owned = 0; \ |
| 1489 | init_waitqueue_head(&sk->sk_lock.wq); \ |
| 1490 | spin_lock_init(&(sk)->sk_lock.slock); \ |
| 1491 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ |
| 1492 | sizeof((sk)->sk_lock)); \ |
| 1493 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ |
| 1494 | (skey), (sname)); \ |
| 1495 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
| 1496 | } while (0) |
| 1497 | |
| 1498 | extern void lock_sock_nested(struct sock *sk, int subclass); |
| 1499 | |
| 1500 | static inline void lock_sock(struct sock *sk) |
| 1501 | { |
| 1502 | lock_sock_nested(sk, 0); |
| 1503 | } |
| 1504 | |
| 1505 | extern void release_sock(struct sock *sk); |
| 1506 | |
| 1507 | /* BH context may only use the following locking interface. */ |
| 1508 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) |
| 1509 | #define bh_lock_sock_nested(__sk) \ |
| 1510 | spin_lock_nested(&((__sk)->sk_lock.slock), \ |
| 1511 | SINGLE_DEPTH_NESTING) |
| 1512 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
| 1513 | |
| 1514 | extern bool lock_sock_fast(struct sock *sk); |
| 1515 | /** |
| 1516 | * unlock_sock_fast - complement of lock_sock_fast |
| 1517 | * @sk: socket |
| 1518 | * @slow: slow mode |
| 1519 | * |
| 1520 | * fast unlock socket for user context. |
| 1521 | * If slow mode is on, we call regular release_sock() |
| 1522 | */ |
| 1523 | static inline void unlock_sock_fast(struct sock *sk, bool slow) |
| 1524 | { |
| 1525 | if (slow) |
| 1526 | release_sock(sk); |
| 1527 | else |
| 1528 | spin_unlock_bh(&sk->sk_lock.slock); |
| 1529 | } |
| 1530 | |
| 1531 | |
| 1532 | extern struct sock *sk_alloc(struct net *net, int family, |
| 1533 | gfp_t priority, |
| 1534 | struct proto *prot); |
| 1535 | extern void sk_free(struct sock *sk); |
| 1536 | extern void sk_release_kernel(struct sock *sk); |
| 1537 | extern struct sock *sk_clone_lock(const struct sock *sk, |
| 1538 | const gfp_t priority); |
| 1539 | |
| 1540 | extern struct sk_buff *sock_wmalloc(struct sock *sk, |
| 1541 | unsigned long size, int force, |
| 1542 | gfp_t priority); |
| 1543 | extern struct sk_buff *sock_rmalloc(struct sock *sk, |
| 1544 | unsigned long size, int force, |
| 1545 | gfp_t priority); |
| 1546 | extern void sock_wfree(struct sk_buff *skb); |
| 1547 | extern void sock_rfree(struct sk_buff *skb); |
| 1548 | extern void sock_edemux(struct sk_buff *skb); |
| 1549 | |
| 1550 | extern int sock_setsockopt(struct socket *sock, int level, |
| 1551 | int op, char __user *optval, |
| 1552 | unsigned int optlen); |
| 1553 | |
| 1554 | extern int sock_getsockopt(struct socket *sock, int level, |
| 1555 | int op, char __user *optval, |
| 1556 | int __user *optlen); |
| 1557 | extern struct sk_buff *sock_alloc_send_skb(struct sock *sk, |
| 1558 | unsigned long size, |
| 1559 | int noblock, |
| 1560 | int *errcode); |
| 1561 | extern struct sk_buff *sock_alloc_send_pskb(struct sock *sk, |
| 1562 | unsigned long header_len, |
| 1563 | unsigned long data_len, |
| 1564 | int noblock, |
| 1565 | int *errcode); |
| 1566 | extern void *sock_kmalloc(struct sock *sk, int size, |
| 1567 | gfp_t priority); |
| 1568 | extern void sock_kfree_s(struct sock *sk, void *mem, int size); |
| 1569 | extern void sk_send_sigurg(struct sock *sk); |
| 1570 | |
| 1571 | /* |
| 1572 | * Functions to fill in entries in struct proto_ops when a protocol |
| 1573 | * does not implement a particular function. |
| 1574 | */ |
| 1575 | extern int sock_no_bind(struct socket *, |
| 1576 | struct sockaddr *, int); |
| 1577 | extern int sock_no_connect(struct socket *, |
| 1578 | struct sockaddr *, int, int); |
| 1579 | extern int sock_no_socketpair(struct socket *, |
| 1580 | struct socket *); |
| 1581 | extern int sock_no_accept(struct socket *, |
| 1582 | struct socket *, int); |
| 1583 | extern int sock_no_getname(struct socket *, |
| 1584 | struct sockaddr *, int *, int); |
| 1585 | extern unsigned int sock_no_poll(struct file *, struct socket *, |
| 1586 | struct poll_table_struct *); |
| 1587 | extern int sock_no_ioctl(struct socket *, unsigned int, |
| 1588 | unsigned long); |
| 1589 | extern int sock_no_listen(struct socket *, int); |
| 1590 | extern int sock_no_shutdown(struct socket *, int); |
| 1591 | extern int sock_no_getsockopt(struct socket *, int , int, |
| 1592 | char __user *, int __user *); |
| 1593 | extern int sock_no_setsockopt(struct socket *, int, int, |
| 1594 | char __user *, unsigned int); |
| 1595 | extern int sock_no_sendmsg(struct kiocb *, struct socket *, |
| 1596 | struct msghdr *, size_t); |
| 1597 | extern int sock_no_recvmsg(struct kiocb *, struct socket *, |
| 1598 | struct msghdr *, size_t, int); |
| 1599 | extern int sock_no_mmap(struct file *file, |
| 1600 | struct socket *sock, |
| 1601 | struct vm_area_struct *vma); |
| 1602 | extern ssize_t sock_no_sendpage(struct socket *sock, |
| 1603 | struct page *page, |
| 1604 | int offset, size_t size, |
| 1605 | int flags); |
| 1606 | |
| 1607 | /* |
| 1608 | * Functions to fill in entries in struct proto_ops when a protocol |
| 1609 | * uses the inet style. |
| 1610 | */ |
| 1611 | extern int sock_common_getsockopt(struct socket *sock, int level, int optname, |
| 1612 | char __user *optval, int __user *optlen); |
| 1613 | extern int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock, |
| 1614 | struct msghdr *msg, size_t size, int flags); |
| 1615 | extern int sock_common_setsockopt(struct socket *sock, int level, int optname, |
| 1616 | char __user *optval, unsigned int optlen); |
| 1617 | extern int compat_sock_common_getsockopt(struct socket *sock, int level, |
| 1618 | int optname, char __user *optval, int __user *optlen); |
| 1619 | extern int compat_sock_common_setsockopt(struct socket *sock, int level, |
| 1620 | int optname, char __user *optval, unsigned int optlen); |
| 1621 | |
| 1622 | extern void sk_common_release(struct sock *sk); |
| 1623 | |
| 1624 | /* |
| 1625 | * Default socket callbacks and setup code |
| 1626 | */ |
| 1627 | |
| 1628 | /* Initialise core socket variables */ |
| 1629 | extern void sock_init_data(struct socket *sock, struct sock *sk); |
| 1630 | |
| 1631 | extern void sk_filter_release_rcu(struct rcu_head *rcu); |
| 1632 | |
| 1633 | /** |
| 1634 | * sk_filter_release - release a socket filter |
| 1635 | * @fp: filter to remove |
| 1636 | * |
| 1637 | * Remove a filter from a socket and release its resources. |
| 1638 | */ |
| 1639 | |
| 1640 | static inline void sk_filter_release(struct sk_filter *fp) |
| 1641 | { |
| 1642 | if (atomic_dec_and_test(&fp->refcnt)) |
| 1643 | call_rcu(&fp->rcu, sk_filter_release_rcu); |
| 1644 | } |
| 1645 | |
| 1646 | static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp) |
| 1647 | { |
| 1648 | unsigned int size = sk_filter_len(fp); |
| 1649 | |
| 1650 | atomic_sub(size, &sk->sk_omem_alloc); |
| 1651 | sk_filter_release(fp); |
| 1652 | } |
| 1653 | |
| 1654 | static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp) |
| 1655 | { |
| 1656 | atomic_inc(&fp->refcnt); |
| 1657 | atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc); |
| 1658 | } |
| 1659 | |
| 1660 | /* |
| 1661 | * Socket reference counting postulates. |
| 1662 | * |
| 1663 | * * Each user of socket SHOULD hold a reference count. |
| 1664 | * * Each access point to socket (an hash table bucket, reference from a list, |
| 1665 | * running timer, skb in flight MUST hold a reference count. |
| 1666 | * * When reference count hits 0, it means it will never increase back. |
| 1667 | * * When reference count hits 0, it means that no references from |
| 1668 | * outside exist to this socket and current process on current CPU |
| 1669 | * is last user and may/should destroy this socket. |
| 1670 | * * sk_free is called from any context: process, BH, IRQ. When |
| 1671 | * it is called, socket has no references from outside -> sk_free |
| 1672 | * may release descendant resources allocated by the socket, but |
| 1673 | * to the time when it is called, socket is NOT referenced by any |
| 1674 | * hash tables, lists etc. |
| 1675 | * * Packets, delivered from outside (from network or from another process) |
| 1676 | * and enqueued on receive/error queues SHOULD NOT grab reference count, |
| 1677 | * when they sit in queue. Otherwise, packets will leak to hole, when |
| 1678 | * socket is looked up by one cpu and unhasing is made by another CPU. |
| 1679 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp |
| 1680 | * (leak to backlog). Packet socket does all the processing inside |
| 1681 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets |
| 1682 | * use separate SMP lock, so that they are prone too. |
| 1683 | */ |
| 1684 | |
| 1685 | /* Ungrab socket and destroy it, if it was the last reference. */ |
| 1686 | static inline void sock_put(struct sock *sk) |
| 1687 | { |
| 1688 | if (atomic_dec_and_test(&sk->sk_refcnt)) |
| 1689 | sk_free(sk); |
| 1690 | } |
| 1691 | |
| 1692 | extern int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
| 1693 | const int nested); |
| 1694 | |
| 1695 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
| 1696 | { |
| 1697 | sk->sk_tx_queue_mapping = tx_queue; |
| 1698 | } |
| 1699 | |
| 1700 | static inline void sk_tx_queue_clear(struct sock *sk) |
| 1701 | { |
| 1702 | sk->sk_tx_queue_mapping = -1; |
| 1703 | } |
| 1704 | |
| 1705 | static inline int sk_tx_queue_get(const struct sock *sk) |
| 1706 | { |
| 1707 | return sk ? sk->sk_tx_queue_mapping : -1; |
| 1708 | } |
| 1709 | |
| 1710 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
| 1711 | { |
| 1712 | sk_tx_queue_clear(sk); |
| 1713 | sk->sk_socket = sock; |
| 1714 | } |
| 1715 | |
| 1716 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
| 1717 | { |
| 1718 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
| 1719 | return &rcu_dereference_raw(sk->sk_wq)->wait; |
| 1720 | } |
| 1721 | /* Detach socket from process context. |
| 1722 | * Announce socket dead, detach it from wait queue and inode. |
| 1723 | * Note that parent inode held reference count on this struct sock, |
| 1724 | * we do not release it in this function, because protocol |
| 1725 | * probably wants some additional cleanups or even continuing |
| 1726 | * to work with this socket (TCP). |
| 1727 | */ |
| 1728 | static inline void sock_orphan(struct sock *sk) |
| 1729 | { |
| 1730 | write_lock_bh(&sk->sk_callback_lock); |
| 1731 | sock_set_flag(sk, SOCK_DEAD); |
| 1732 | sk_set_socket(sk, NULL); |
| 1733 | sk->sk_wq = NULL; |
| 1734 | write_unlock_bh(&sk->sk_callback_lock); |
| 1735 | } |
| 1736 | |
| 1737 | static inline void sock_graft(struct sock *sk, struct socket *parent) |
| 1738 | { |
| 1739 | write_lock_bh(&sk->sk_callback_lock); |
| 1740 | sk->sk_wq = parent->wq; |
| 1741 | parent->sk = sk; |
| 1742 | sk_set_socket(sk, parent); |
| 1743 | sk->sk_uid = SOCK_INODE(parent)->i_uid; |
| 1744 | security_sock_graft(sk, parent); |
| 1745 | write_unlock_bh(&sk->sk_callback_lock); |
| 1746 | } |
| 1747 | |
| 1748 | extern kuid_t sock_i_uid(struct sock *sk); |
| 1749 | extern unsigned long sock_i_ino(struct sock *sk); |
| 1750 | |
| 1751 | static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk) |
| 1752 | { |
| 1753 | return sk ? sk->sk_uid : make_kuid(net->user_ns, 0); |
| 1754 | } |
| 1755 | |
| 1756 | static inline struct dst_entry * |
| 1757 | __sk_dst_get(struct sock *sk) |
| 1758 | { |
| 1759 | return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) || |
| 1760 | lockdep_is_held(&sk->sk_lock.slock)); |
| 1761 | } |
| 1762 | |
| 1763 | static inline struct dst_entry * |
| 1764 | sk_dst_get(struct sock *sk) |
| 1765 | { |
| 1766 | struct dst_entry *dst; |
| 1767 | |
| 1768 | rcu_read_lock(); |
| 1769 | dst = rcu_dereference(sk->sk_dst_cache); |
| 1770 | if (dst && !atomic_inc_not_zero(&dst->__refcnt)) |
| 1771 | dst = NULL; |
| 1772 | rcu_read_unlock(); |
| 1773 | return dst; |
| 1774 | } |
| 1775 | |
| 1776 | extern void sk_reset_txq(struct sock *sk); |
| 1777 | |
| 1778 | static inline void dst_negative_advice(struct sock *sk) |
| 1779 | { |
| 1780 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); |
| 1781 | |
| 1782 | if (dst && dst->ops->negative_advice) { |
| 1783 | ndst = dst->ops->negative_advice(dst); |
| 1784 | |
| 1785 | if (ndst != dst) { |
| 1786 | rcu_assign_pointer(sk->sk_dst_cache, ndst); |
| 1787 | sk_reset_txq(sk); |
| 1788 | } |
| 1789 | } |
| 1790 | } |
| 1791 | |
| 1792 | static inline void |
| 1793 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) |
| 1794 | { |
| 1795 | struct dst_entry *old_dst; |
| 1796 | |
| 1797 | sk_tx_queue_clear(sk); |
| 1798 | /* |
| 1799 | * This can be called while sk is owned by the caller only, |
| 1800 | * with no state that can be checked in a rcu_dereference_check() cond |
| 1801 | */ |
| 1802 | old_dst = rcu_dereference_raw(sk->sk_dst_cache); |
| 1803 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
| 1804 | dst_release(old_dst); |
| 1805 | } |
| 1806 | |
| 1807 | static inline void |
| 1808 | sk_dst_set(struct sock *sk, struct dst_entry *dst) |
| 1809 | { |
| 1810 | struct dst_entry *old_dst; |
| 1811 | |
| 1812 | sk_tx_queue_clear(sk); |
| 1813 | old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst); |
| 1814 | dst_release(old_dst); |
| 1815 | } |
| 1816 | |
| 1817 | static inline void |
| 1818 | __sk_dst_reset(struct sock *sk) |
| 1819 | { |
| 1820 | __sk_dst_set(sk, NULL); |
| 1821 | } |
| 1822 | |
| 1823 | static inline void |
| 1824 | sk_dst_reset(struct sock *sk) |
| 1825 | { |
| 1826 | sk_dst_set(sk, NULL); |
| 1827 | } |
| 1828 | |
| 1829 | extern struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
| 1830 | |
| 1831 | extern struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
| 1832 | |
| 1833 | static inline bool sk_can_gso(const struct sock *sk) |
| 1834 | { |
| 1835 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); |
| 1836 | } |
| 1837 | |
| 1838 | extern void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
| 1839 | |
| 1840 | static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) |
| 1841 | { |
| 1842 | sk->sk_route_nocaps |= flags; |
| 1843 | sk->sk_route_caps &= ~flags; |
| 1844 | } |
| 1845 | |
| 1846 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
| 1847 | char __user *from, char *to, |
| 1848 | int copy, int offset) |
| 1849 | { |
| 1850 | if (skb->ip_summed == CHECKSUM_NONE) { |
| 1851 | int err = 0; |
| 1852 | __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err); |
| 1853 | if (err) |
| 1854 | return err; |
| 1855 | skb->csum = csum_block_add(skb->csum, csum, offset); |
| 1856 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
| 1857 | if (!access_ok(VERIFY_READ, from, copy) || |
| 1858 | __copy_from_user_nocache(to, from, copy)) |
| 1859 | return -EFAULT; |
| 1860 | } else if (copy_from_user(to, from, copy)) |
| 1861 | return -EFAULT; |
| 1862 | |
| 1863 | return 0; |
| 1864 | } |
| 1865 | |
| 1866 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, |
| 1867 | char __user *from, int copy) |
| 1868 | { |
| 1869 | int err, offset = skb->len; |
| 1870 | |
| 1871 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
| 1872 | copy, offset); |
| 1873 | if (err) |
| 1874 | __skb_trim(skb, offset); |
| 1875 | |
| 1876 | return err; |
| 1877 | } |
| 1878 | |
| 1879 | static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from, |
| 1880 | struct sk_buff *skb, |
| 1881 | struct page *page, |
| 1882 | int off, int copy) |
| 1883 | { |
| 1884 | int err; |
| 1885 | |
| 1886 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
| 1887 | copy, skb->len); |
| 1888 | if (err) |
| 1889 | return err; |
| 1890 | |
| 1891 | skb->len += copy; |
| 1892 | skb->data_len += copy; |
| 1893 | skb->truesize += copy; |
| 1894 | sk->sk_wmem_queued += copy; |
| 1895 | sk_mem_charge(sk, copy); |
| 1896 | return 0; |
| 1897 | } |
| 1898 | |
| 1899 | static inline int skb_copy_to_page(struct sock *sk, char __user *from, |
| 1900 | struct sk_buff *skb, struct page *page, |
| 1901 | int off, int copy) |
| 1902 | { |
| 1903 | if (skb->ip_summed == CHECKSUM_NONE) { |
| 1904 | int err = 0; |
| 1905 | __wsum csum = csum_and_copy_from_user(from, |
| 1906 | page_address(page) + off, |
| 1907 | copy, 0, &err); |
| 1908 | if (err) |
| 1909 | return err; |
| 1910 | skb->csum = csum_block_add(skb->csum, csum, skb->len); |
| 1911 | } else if (copy_from_user(page_address(page) + off, from, copy)) |
| 1912 | return -EFAULT; |
| 1913 | |
| 1914 | skb->len += copy; |
| 1915 | skb->data_len += copy; |
| 1916 | skb->truesize += copy; |
| 1917 | sk->sk_wmem_queued += copy; |
| 1918 | sk_mem_charge(sk, copy); |
| 1919 | return 0; |
| 1920 | } |
| 1921 | |
| 1922 | /** |
| 1923 | * sk_wmem_alloc_get - returns write allocations |
| 1924 | * @sk: socket |
| 1925 | * |
| 1926 | * Returns sk_wmem_alloc minus initial offset of one |
| 1927 | */ |
| 1928 | static inline int sk_wmem_alloc_get(const struct sock *sk) |
| 1929 | { |
| 1930 | return atomic_read(&sk->sk_wmem_alloc) - 1; |
| 1931 | } |
| 1932 | |
| 1933 | /** |
| 1934 | * sk_rmem_alloc_get - returns read allocations |
| 1935 | * @sk: socket |
| 1936 | * |
| 1937 | * Returns sk_rmem_alloc |
| 1938 | */ |
| 1939 | static inline int sk_rmem_alloc_get(const struct sock *sk) |
| 1940 | { |
| 1941 | return atomic_read(&sk->sk_rmem_alloc); |
| 1942 | } |
| 1943 | |
| 1944 | /** |
| 1945 | * sk_has_allocations - check if allocations are outstanding |
| 1946 | * @sk: socket |
| 1947 | * |
| 1948 | * Returns true if socket has write or read allocations |
| 1949 | */ |
| 1950 | static inline bool sk_has_allocations(const struct sock *sk) |
| 1951 | { |
| 1952 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); |
| 1953 | } |
| 1954 | |
| 1955 | /** |
| 1956 | * wq_has_sleeper - check if there are any waiting processes |
| 1957 | * @wq: struct socket_wq |
| 1958 | * |
| 1959 | * Returns true if socket_wq has waiting processes |
| 1960 | * |
| 1961 | * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory |
| 1962 | * barrier call. They were added due to the race found within the tcp code. |
| 1963 | * |
| 1964 | * Consider following tcp code paths: |
| 1965 | * |
| 1966 | * CPU1 CPU2 |
| 1967 | * |
| 1968 | * sys_select receive packet |
| 1969 | * ... ... |
| 1970 | * __add_wait_queue update tp->rcv_nxt |
| 1971 | * ... ... |
| 1972 | * tp->rcv_nxt check sock_def_readable |
| 1973 | * ... { |
| 1974 | * schedule rcu_read_lock(); |
| 1975 | * wq = rcu_dereference(sk->sk_wq); |
| 1976 | * if (wq && waitqueue_active(&wq->wait)) |
| 1977 | * wake_up_interruptible(&wq->wait) |
| 1978 | * ... |
| 1979 | * } |
| 1980 | * |
| 1981 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay |
| 1982 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 |
| 1983 | * could then endup calling schedule and sleep forever if there are no more |
| 1984 | * data on the socket. |
| 1985 | * |
| 1986 | */ |
| 1987 | static inline bool wq_has_sleeper(struct socket_wq *wq) |
| 1988 | { |
| 1989 | /* We need to be sure we are in sync with the |
| 1990 | * add_wait_queue modifications to the wait queue. |
| 1991 | * |
| 1992 | * This memory barrier is paired in the sock_poll_wait. |
| 1993 | */ |
| 1994 | smp_mb(); |
| 1995 | return wq && waitqueue_active(&wq->wait); |
| 1996 | } |
| 1997 | |
| 1998 | /** |
| 1999 | * sock_poll_wait - place memory barrier behind the poll_wait call. |
| 2000 | * @filp: file |
| 2001 | * @wait_address: socket wait queue |
| 2002 | * @p: poll_table |
| 2003 | * |
| 2004 | * See the comments in the wq_has_sleeper function. |
| 2005 | */ |
| 2006 | static inline void sock_poll_wait(struct file *filp, |
| 2007 | wait_queue_head_t *wait_address, poll_table *p) |
| 2008 | { |
| 2009 | if (!poll_does_not_wait(p) && wait_address) { |
| 2010 | poll_wait(filp, wait_address, p); |
| 2011 | /* We need to be sure we are in sync with the |
| 2012 | * socket flags modification. |
| 2013 | * |
| 2014 | * This memory barrier is paired in the wq_has_sleeper. |
| 2015 | */ |
| 2016 | smp_mb(); |
| 2017 | } |
| 2018 | } |
| 2019 | |
| 2020 | /* |
| 2021 | * Queue a received datagram if it will fit. Stream and sequenced |
| 2022 | * protocols can't normally use this as they need to fit buffers in |
| 2023 | * and play with them. |
| 2024 | * |
| 2025 | * Inlined as it's very short and called for pretty much every |
| 2026 | * packet ever received. |
| 2027 | */ |
| 2028 | |
| 2029 | static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk) |
| 2030 | { |
| 2031 | skb_orphan(skb); |
| 2032 | skb->sk = sk; |
| 2033 | skb->destructor = sock_wfree; |
| 2034 | /* |
| 2035 | * We used to take a refcount on sk, but following operation |
| 2036 | * is enough to guarantee sk_free() wont free this sock until |
| 2037 | * all in-flight packets are completed |
| 2038 | */ |
| 2039 | atomic_add(skb->truesize, &sk->sk_wmem_alloc); |
| 2040 | } |
| 2041 | |
| 2042 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
| 2043 | { |
| 2044 | skb_orphan(skb); |
| 2045 | skb->sk = sk; |
| 2046 | skb->destructor = sock_rfree; |
| 2047 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); |
| 2048 | sk_mem_charge(sk, skb->truesize); |
| 2049 | } |
| 2050 | |
| 2051 | extern void sk_reset_timer(struct sock *sk, struct timer_list *timer, |
| 2052 | unsigned long expires); |
| 2053 | |
| 2054 | extern void sk_stop_timer(struct sock *sk, struct timer_list *timer); |
| 2055 | |
| 2056 | extern int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
| 2057 | |
| 2058 | extern int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
| 2059 | |
| 2060 | /* |
| 2061 | * Recover an error report and clear atomically |
| 2062 | */ |
| 2063 | |
| 2064 | static inline int sock_error(struct sock *sk) |
| 2065 | { |
| 2066 | int err; |
| 2067 | if (likely(!sk->sk_err)) |
| 2068 | return 0; |
| 2069 | err = xchg(&sk->sk_err, 0); |
| 2070 | return -err; |
| 2071 | } |
| 2072 | |
| 2073 | static inline unsigned long sock_wspace(struct sock *sk) |
| 2074 | { |
| 2075 | int amt = 0; |
| 2076 | |
| 2077 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { |
| 2078 | amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc); |
| 2079 | if (amt < 0) |
| 2080 | amt = 0; |
| 2081 | } |
| 2082 | return amt; |
| 2083 | } |
| 2084 | |
| 2085 | static inline void sk_wake_async(struct sock *sk, int how, int band) |
| 2086 | { |
| 2087 | if (sock_flag(sk, SOCK_FASYNC)) |
| 2088 | sock_wake_async(sk->sk_socket, how, band); |
| 2089 | } |
| 2090 | |
| 2091 | #define SOCK_MIN_SNDBUF 2048 |
| 2092 | /* |
| 2093 | * Since sk_rmem_alloc sums skb->truesize, even a small frame might need |
| 2094 | * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak |
| 2095 | */ |
| 2096 | #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff)) |
| 2097 | |
| 2098 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) |
| 2099 | { |
| 2100 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { |
| 2101 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
| 2102 | sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
| 2103 | } |
| 2104 | } |
| 2105 | |
| 2106 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp); |
| 2107 | |
| 2108 | /** |
| 2109 | * sk_page_frag - return an appropriate page_frag |
| 2110 | * @sk: socket |
| 2111 | * |
| 2112 | * If socket allocation mode allows current thread to sleep, it means its |
| 2113 | * safe to use the per task page_frag instead of the per socket one. |
| 2114 | */ |
| 2115 | static inline struct page_frag *sk_page_frag(struct sock *sk) |
| 2116 | { |
| 2117 | if (sk->sk_allocation & __GFP_WAIT) |
| 2118 | return ¤t->task_frag; |
| 2119 | |
| 2120 | return &sk->sk_frag; |
| 2121 | } |
| 2122 | |
| 2123 | extern bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag); |
| 2124 | |
| 2125 | /* |
| 2126 | * Default write policy as shown to user space via poll/select/SIGIO |
| 2127 | */ |
| 2128 | static inline bool sock_writeable(const struct sock *sk) |
| 2129 | { |
| 2130 | return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
| 2131 | } |
| 2132 | |
| 2133 | static inline gfp_t gfp_any(void) |
| 2134 | { |
| 2135 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
| 2136 | } |
| 2137 | |
| 2138 | static inline long sock_rcvtimeo(const struct sock *sk, bool noblock) |
| 2139 | { |
| 2140 | return noblock ? 0 : sk->sk_rcvtimeo; |
| 2141 | } |
| 2142 | |
| 2143 | static inline long sock_sndtimeo(const struct sock *sk, bool noblock) |
| 2144 | { |
| 2145 | return noblock ? 0 : sk->sk_sndtimeo; |
| 2146 | } |
| 2147 | |
| 2148 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) |
| 2149 | { |
| 2150 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; |
| 2151 | } |
| 2152 | |
| 2153 | /* Alas, with timeout socket operations are not restartable. |
| 2154 | * Compare this to poll(). |
| 2155 | */ |
| 2156 | static inline int sock_intr_errno(long timeo) |
| 2157 | { |
| 2158 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; |
| 2159 | } |
| 2160 | |
| 2161 | extern void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
| 2162 | struct sk_buff *skb); |
| 2163 | extern void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, |
| 2164 | struct sk_buff *skb); |
| 2165 | |
| 2166 | static inline void |
| 2167 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
| 2168 | { |
| 2169 | ktime_t kt = skb->tstamp; |
| 2170 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
| 2171 | |
| 2172 | /* |
| 2173 | * generate control messages if |
| 2174 | * - receive time stamping in software requested (SOCK_RCVTSTAMP |
| 2175 | * or SOCK_TIMESTAMPING_RX_SOFTWARE) |
| 2176 | * - software time stamp available and wanted |
| 2177 | * (SOCK_TIMESTAMPING_SOFTWARE) |
| 2178 | * - hardware time stamps available and wanted |
| 2179 | * (SOCK_TIMESTAMPING_SYS_HARDWARE or |
| 2180 | * SOCK_TIMESTAMPING_RAW_HARDWARE) |
| 2181 | */ |
| 2182 | if (sock_flag(sk, SOCK_RCVTSTAMP) || |
| 2183 | sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) || |
| 2184 | (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) || |
| 2185 | (hwtstamps->hwtstamp.tv64 && |
| 2186 | sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) || |
| 2187 | (hwtstamps->syststamp.tv64 && |
| 2188 | sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE))) |
| 2189 | __sock_recv_timestamp(msg, sk, skb); |
| 2190 | else |
| 2191 | sk->sk_stamp = kt; |
| 2192 | |
| 2193 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) |
| 2194 | __sock_recv_wifi_status(msg, sk, skb); |
| 2195 | } |
| 2196 | |
| 2197 | extern void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
| 2198 | struct sk_buff *skb); |
| 2199 | |
| 2200 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
| 2201 | struct sk_buff *skb) |
| 2202 | { |
| 2203 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ |
| 2204 | (1UL << SOCK_RCVTSTAMP) | \ |
| 2205 | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \ |
| 2206 | (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \ |
| 2207 | (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \ |
| 2208 | (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE)) |
| 2209 | |
| 2210 | if (sk->sk_flags & FLAGS_TS_OR_DROPS) |
| 2211 | __sock_recv_ts_and_drops(msg, sk, skb); |
| 2212 | else |
| 2213 | sk->sk_stamp = skb->tstamp; |
| 2214 | } |
| 2215 | |
| 2216 | /** |
| 2217 | * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped |
| 2218 | * @sk: socket sending this packet |
| 2219 | * @tx_flags: filled with instructions for time stamping |
| 2220 | * |
| 2221 | * Currently only depends on SOCK_TIMESTAMPING* flags. |
| 2222 | */ |
| 2223 | extern void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags); |
| 2224 | |
| 2225 | /** |
| 2226 | * sk_eat_skb - Release a skb if it is no longer needed |
| 2227 | * @sk: socket to eat this skb from |
| 2228 | * @skb: socket buffer to eat |
| 2229 | * @copied_early: flag indicating whether DMA operations copied this data early |
| 2230 | * |
| 2231 | * This routine must be called with interrupts disabled or with the socket |
| 2232 | * locked so that the sk_buff queue operation is ok. |
| 2233 | */ |
| 2234 | #ifdef CONFIG_NET_DMA |
| 2235 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early) |
| 2236 | { |
| 2237 | __skb_unlink(skb, &sk->sk_receive_queue); |
| 2238 | if (!copied_early) |
| 2239 | __kfree_skb(skb); |
| 2240 | else |
| 2241 | __skb_queue_tail(&sk->sk_async_wait_queue, skb); |
| 2242 | } |
| 2243 | #else |
| 2244 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early) |
| 2245 | { |
| 2246 | __skb_unlink(skb, &sk->sk_receive_queue); |
| 2247 | __kfree_skb(skb); |
| 2248 | } |
| 2249 | #endif |
| 2250 | |
| 2251 | static inline |
| 2252 | struct net *sock_net(const struct sock *sk) |
| 2253 | { |
| 2254 | return read_pnet(&sk->sk_net); |
| 2255 | } |
| 2256 | |
| 2257 | static inline |
| 2258 | void sock_net_set(struct sock *sk, struct net *net) |
| 2259 | { |
| 2260 | write_pnet(&sk->sk_net, net); |
| 2261 | } |
| 2262 | |
| 2263 | /* |
| 2264 | * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace. |
| 2265 | * They should not hold a reference to a namespace in order to allow |
| 2266 | * to stop it. |
| 2267 | * Sockets after sk_change_net should be released using sk_release_kernel |
| 2268 | */ |
| 2269 | static inline void sk_change_net(struct sock *sk, struct net *net) |
| 2270 | { |
| 2271 | put_net(sock_net(sk)); |
| 2272 | sock_net_set(sk, hold_net(net)); |
| 2273 | } |
| 2274 | |
| 2275 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) |
| 2276 | { |
| 2277 | if (skb->sk) { |
| 2278 | struct sock *sk = skb->sk; |
| 2279 | |
| 2280 | skb->destructor = NULL; |
| 2281 | skb->sk = NULL; |
| 2282 | return sk; |
| 2283 | } |
| 2284 | return NULL; |
| 2285 | } |
| 2286 | |
| 2287 | extern void sock_enable_timestamp(struct sock *sk, int flag); |
| 2288 | extern int sock_get_timestamp(struct sock *, struct timeval __user *); |
| 2289 | extern int sock_get_timestampns(struct sock *, struct timespec __user *); |
| 2290 | |
| 2291 | bool sk_ns_capable(const struct sock *sk, |
| 2292 | struct user_namespace *user_ns, int cap); |
| 2293 | bool sk_capable(const struct sock *sk, int cap); |
| 2294 | bool sk_net_capable(const struct sock *sk, int cap); |
| 2295 | |
| 2296 | /* |
| 2297 | * Enable debug/info messages |
| 2298 | */ |
| 2299 | extern int net_msg_warn; |
| 2300 | #define NETDEBUG(fmt, args...) \ |
| 2301 | do { if (net_msg_warn) printk(fmt,##args); } while (0) |
| 2302 | |
| 2303 | #define LIMIT_NETDEBUG(fmt, args...) \ |
| 2304 | do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0) |
| 2305 | |
| 2306 | extern __u32 sysctl_wmem_max; |
| 2307 | extern __u32 sysctl_rmem_max; |
| 2308 | |
| 2309 | extern int sysctl_optmem_max; |
| 2310 | |
| 2311 | extern __u32 sysctl_wmem_default; |
| 2312 | extern __u32 sysctl_rmem_default; |
| 2313 | |
| 2314 | #endif /* _SOCK_H */ |