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