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/page_counter.h>
  58#include <linux/memcontrol.h>
  59#include <linux/static_key.h>
  60#include <linux/sched.h>
  61#include <linux/wait.h>
  62#include <linux/cgroup-defs.h>
  63
  64#include <linux/filter.h>
  65#include <linux/rculist_nulls.h>
  66#include <linux/poll.h>
  67
 
  68#include <linux/atomic.h>
 
 
  69#include <net/dst.h>
  70#include <net/checksum.h>
  71#include <net/tcp_states.h>
  72#include <linux/net_tstamp.h>
 
 
  73
  74/*
  75 * This structure really needs to be cleaned up.
  76 * Most of it is for TCP, and not used by any of
  77 * the other protocols.
  78 */
  79
  80/* Define this to get the SOCK_DBG debugging facility. */
  81#define SOCK_DEBUGGING
  82#ifdef SOCK_DEBUGGING
  83#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
  84					printk(KERN_DEBUG msg); } while (0)
  85#else
  86/* Validate arguments and do nothing */
  87static inline __printf(2, 3)
  88void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
  89{
  90}
  91#endif
  92
  93/* This is the per-socket lock.  The spinlock provides a synchronization
  94 * between user contexts and software interrupt processing, whereas the
  95 * mini-semaphore synchronizes multiple users amongst themselves.
  96 */
  97typedef struct {
  98	spinlock_t		slock;
  99	int			owned;
 100	wait_queue_head_t	wq;
 101	/*
 102	 * We express the mutex-alike socket_lock semantics
 103	 * to the lock validator by explicitly managing
 104	 * the slock as a lock variant (in addition to
 105	 * the slock itself):
 106	 */
 107#ifdef CONFIG_DEBUG_LOCK_ALLOC
 108	struct lockdep_map dep_map;
 109#endif
 110} socket_lock_t;
 111
 112struct sock;
 113struct proto;
 114struct net;
 115
 116typedef __u32 __bitwise __portpair;
 117typedef __u64 __bitwise __addrpair;
 118
 119/**
 120 *	struct sock_common - minimal network layer representation of sockets
 121 *	@skc_daddr: Foreign IPv4 addr
 122 *	@skc_rcv_saddr: Bound local IPv4 addr
 
 123 *	@skc_hash: hash value used with various protocol lookup tables
 124 *	@skc_u16hashes: two u16 hash values used by UDP lookup tables
 125 *	@skc_dport: placeholder for inet_dport/tw_dport
 126 *	@skc_num: placeholder for inet_num/tw_num
 
 127 *	@skc_family: network address family
 128 *	@skc_state: Connection state
 129 *	@skc_reuse: %SO_REUSEADDR setting
 130 *	@skc_reuseport: %SO_REUSEPORT setting
 
 
 131 *	@skc_bound_dev_if: bound device index if != 0
 132 *	@skc_bind_node: bind hash linkage for various protocol lookup tables
 133 *	@skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
 134 *	@skc_prot: protocol handlers inside a network family
 135 *	@skc_net: reference to the network namespace of this socket
 
 
 
 136 *	@skc_node: main hash linkage for various protocol lookup tables
 137 *	@skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
 138 *	@skc_tx_queue_mapping: tx queue number for this connection
 
 139 *	@skc_flags: place holder for sk_flags
 140 *		%SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
 141 *		%SO_OOBINLINE settings, %SO_TIMESTAMPING settings
 
 
 
 
 142 *	@skc_incoming_cpu: record/match cpu processing incoming packets
 
 
 
 
 143 *	@skc_refcnt: reference count
 144 *
 145 *	This is the minimal network layer representation of sockets, the header
 146 *	for struct sock and struct inet_timewait_sock.
 147 */
 148struct sock_common {
 149	/* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
 150	 * address on 64bit arches : cf INET_MATCH()
 151	 */
 152	union {
 153		__addrpair	skc_addrpair;
 154		struct {
 155			__be32	skc_daddr;
 156			__be32	skc_rcv_saddr;
 157		};
 158	};
 159	union  {
 160		unsigned int	skc_hash;
 161		__u16		skc_u16hashes[2];
 162	};
 163	/* skc_dport && skc_num must be grouped as well */
 164	union {
 165		__portpair	skc_portpair;
 166		struct {
 167			__be16	skc_dport;
 168			__u16	skc_num;
 169		};
 170	};
 171
 172	unsigned short		skc_family;
 173	volatile unsigned char	skc_state;
 174	unsigned char		skc_reuse:4;
 175	unsigned char		skc_reuseport:1;
 176	unsigned char		skc_ipv6only:1;
 177	unsigned char		skc_net_refcnt:1;
 178	int			skc_bound_dev_if;
 179	union {
 180		struct hlist_node	skc_bind_node;
 181		struct hlist_nulls_node skc_portaddr_node;
 182	};
 183	struct proto		*skc_prot;
 184	possible_net_t		skc_net;
 185
 186#if IS_ENABLED(CONFIG_IPV6)
 187	struct in6_addr		skc_v6_daddr;
 188	struct in6_addr		skc_v6_rcv_saddr;
 189#endif
 190
 191	atomic64_t		skc_cookie;
 192
 193	/* following fields are padding to force
 194	 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
 195	 * assuming IPV6 is enabled. We use this padding differently
 196	 * for different kind of 'sockets'
 197	 */
 198	union {
 199		unsigned long	skc_flags;
 200		struct sock	*skc_listener; /* request_sock */
 201		struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
 202	};
 203	/*
 204	 * fields between dontcopy_begin/dontcopy_end
 205	 * are not copied in sock_copy()
 206	 */
 207	/* private: */
 208	int			skc_dontcopy_begin[0];
 209	/* public: */
 210	union {
 211		struct hlist_node	skc_node;
 212		struct hlist_nulls_node skc_nulls_node;
 213	};
 214	int			skc_tx_queue_mapping;
 
 
 
 215	union {
 216		int		skc_incoming_cpu;
 217		u32		skc_rcv_wnd;
 218		u32		skc_tw_rcv_nxt; /* struct tcp_timewait_sock  */
 219	};
 220
 221	atomic_t		skc_refcnt;
 222	/* private: */
 223	int                     skc_dontcopy_end[0];
 224	union {
 225		u32		skc_rxhash;
 226		u32		skc_window_clamp;
 227		u32		skc_tw_snd_nxt; /* struct tcp_timewait_sock */
 228	};
 229	/* public: */
 230};
 231
 
 
 
 232/**
 233  *	struct sock - network layer representation of sockets
 234  *	@__sk_common: shared layout with inet_timewait_sock
 235  *	@sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
 236  *	@sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
 237  *	@sk_lock:	synchronizer
 
 238  *	@sk_rcvbuf: size of receive buffer in bytes
 239  *	@sk_wq: sock wait queue and async head
 240  *	@sk_rx_dst: receive input route used by early demux
 
 
 241  *	@sk_dst_cache: destination cache
 
 242  *	@sk_policy: flow policy
 243  *	@sk_receive_queue: incoming packets
 244  *	@sk_wmem_alloc: transmit queue bytes committed
 
 245  *	@sk_write_queue: Packet sending queue
 246  *	@sk_omem_alloc: "o" is "option" or "other"
 247  *	@sk_wmem_queued: persistent queue size
 248  *	@sk_forward_alloc: space allocated forward
 
 249  *	@sk_napi_id: id of the last napi context to receive data for sk
 250  *	@sk_ll_usec: usecs to busypoll when there is no data
 251  *	@sk_allocation: allocation mode
 252  *	@sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
 
 253  *	@sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
 254  *	@sk_sndbuf: size of send buffer in bytes
 255  *	@sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
 256  *	@sk_no_check_rx: allow zero checksum in RX packets
 257  *	@sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
 258  *	@sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
 259  *	@sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
 260  *	@sk_gso_max_size: Maximum GSO segment size to build
 261  *	@sk_gso_max_segs: Maximum number of GSO segments
 
 262  *	@sk_lingertime: %SO_LINGER l_linger setting
 263  *	@sk_backlog: always used with the per-socket spinlock held
 264  *	@sk_callback_lock: used with the callbacks in the end of this struct
 265  *	@sk_error_queue: rarely used
 266  *	@sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
 267  *			  IPV6_ADDRFORM for instance)
 268  *	@sk_err: last error
 269  *	@sk_err_soft: errors that don't cause failure but are the cause of a
 270  *		      persistent failure not just 'timed out'
 271  *	@sk_drops: raw/udp drops counter
 272  *	@sk_ack_backlog: current listen backlog
 273  *	@sk_max_ack_backlog: listen backlog set in listen()
 
 
 
 274  *	@sk_priority: %SO_PRIORITY setting
 275  *	@sk_type: socket type (%SOCK_STREAM, etc)
 276  *	@sk_protocol: which protocol this socket belongs in this network family
 
 277  *	@sk_peer_pid: &struct pid for this socket's peer
 278  *	@sk_peer_cred: %SO_PEERCRED setting
 279  *	@sk_rcvlowat: %SO_RCVLOWAT setting
 280  *	@sk_rcvtimeo: %SO_RCVTIMEO setting
 281  *	@sk_sndtimeo: %SO_SNDTIMEO setting
 282  *	@sk_txhash: computed flow hash for use on transmit
 
 283  *	@sk_filter: socket filtering instructions
 284  *	@sk_timer: sock cleanup timer
 285  *	@sk_stamp: time stamp of last packet received
 286  *	@sk_tsflags: SO_TIMESTAMPING socket options
 
 
 
 
 
 
 287  *	@sk_tskey: counter to disambiguate concurrent tstamp requests
 
 288  *	@sk_socket: Identd and reporting IO signals
 289  *	@sk_user_data: RPC layer private data
 290  *	@sk_frag: cached page frag
 291  *	@sk_peek_off: current peek_offset value
 292  *	@sk_send_head: front of stuff to transmit
 
 293  *	@sk_security: used by security modules
 294  *	@sk_mark: generic packet mark
 295  *	@sk_cgrp_data: cgroup data for this cgroup
 296  *	@sk_memcg: this socket's memory cgroup association
 297  *	@sk_write_pending: a write to stream socket waits to start
 
 298  *	@sk_state_change: callback to indicate change in the state of the sock
 299  *	@sk_data_ready: callback to indicate there is data to be processed
 300  *	@sk_write_space: callback to indicate there is bf sending space available
 301  *	@sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
 302  *	@sk_backlog_rcv: callback to process the backlog
 
 303  *	@sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
 304  *	@sk_reuseport_cb: reuseport group container
 305 */
 
 
 
 
 
 
 
 306struct sock {
 307	/*
 308	 * Now struct inet_timewait_sock also uses sock_common, so please just
 309	 * don't add nothing before this first member (__sk_common) --acme
 310	 */
 311	struct sock_common	__sk_common;
 312#define sk_node			__sk_common.skc_node
 313#define sk_nulls_node		__sk_common.skc_nulls_node
 314#define sk_refcnt		__sk_common.skc_refcnt
 315#define sk_tx_queue_mapping	__sk_common.skc_tx_queue_mapping
 
 
 
 316
 317#define sk_dontcopy_begin	__sk_common.skc_dontcopy_begin
 318#define sk_dontcopy_end		__sk_common.skc_dontcopy_end
 319#define sk_hash			__sk_common.skc_hash
 320#define sk_portpair		__sk_common.skc_portpair
 321#define sk_num			__sk_common.skc_num
 322#define sk_dport		__sk_common.skc_dport
 323#define sk_addrpair		__sk_common.skc_addrpair
 324#define sk_daddr		__sk_common.skc_daddr
 325#define sk_rcv_saddr		__sk_common.skc_rcv_saddr
 326#define sk_family		__sk_common.skc_family
 327#define sk_state		__sk_common.skc_state
 328#define sk_reuse		__sk_common.skc_reuse
 329#define sk_reuseport		__sk_common.skc_reuseport
 330#define sk_ipv6only		__sk_common.skc_ipv6only
 331#define sk_net_refcnt		__sk_common.skc_net_refcnt
 332#define sk_bound_dev_if		__sk_common.skc_bound_dev_if
 333#define sk_bind_node		__sk_common.skc_bind_node
 334#define sk_prot			__sk_common.skc_prot
 335#define sk_net			__sk_common.skc_net
 336#define sk_v6_daddr		__sk_common.skc_v6_daddr
 337#define sk_v6_rcv_saddr	__sk_common.skc_v6_rcv_saddr
 338#define sk_cookie		__sk_common.skc_cookie
 339#define sk_incoming_cpu		__sk_common.skc_incoming_cpu
 340#define sk_flags		__sk_common.skc_flags
 341#define sk_rxhash		__sk_common.skc_rxhash
 342
 343	socket_lock_t		sk_lock;
 
 
 
 
 344	struct sk_buff_head	sk_receive_queue;
 345	/*
 346	 * The backlog queue is special, it is always used with
 347	 * the per-socket spinlock held and requires low latency
 348	 * access. Therefore we special case it's implementation.
 349	 * Note : rmem_alloc is in this structure to fill a hole
 350	 * on 64bit arches, not because its logically part of
 351	 * backlog.
 352	 */
 353	struct {
 354		atomic_t	rmem_alloc;
 355		int		len;
 356		struct sk_buff	*head;
 357		struct sk_buff	*tail;
 358	} sk_backlog;
 359#define sk_rmem_alloc sk_backlog.rmem_alloc
 360	int			sk_forward_alloc;
 361
 362	__u32			sk_txhash;
 
 
 
 
 
 
 
 363#ifdef CONFIG_NET_RX_BUSY_POLL
 364	unsigned int		sk_napi_id;
 365	unsigned int		sk_ll_usec;
 
 
 
 366#endif
 367	atomic_t		sk_drops;
 368	int			sk_rcvbuf;
 369
 370	struct sk_filter __rcu	*sk_filter;
 371	union {
 372		struct socket_wq __rcu	*sk_wq;
 
 373		struct socket_wq	*sk_wq_raw;
 
 374	};
 
 
 
 
 
 
 
 
 
 
 375#ifdef CONFIG_XFRM
 376	struct xfrm_policy __rcu *sk_policy[2];
 377#endif
 378	struct dst_entry	*sk_rx_dst;
 379	struct dst_entry __rcu	*sk_dst_cache;
 380	/* Note: 32bit hole on 64bit arches */
 381	atomic_t		sk_wmem_alloc;
 
 
 
 
 
 
 
 382	atomic_t		sk_omem_alloc;
 383	int			sk_sndbuf;
 384	struct sk_buff_head	sk_write_queue;
 385	kmemcheck_bitfield_begin(flags);
 386	unsigned int		sk_shutdown  : 2,
 387				sk_no_check_tx : 1,
 388				sk_no_check_rx : 1,
 389				sk_userlocks : 4,
 390				sk_protocol  : 8,
 391				sk_type      : 16;
 392#define SK_PROTOCOL_MAX U8_MAX
 393	kmemcheck_bitfield_end(flags);
 394	int			sk_wmem_queued;
 395	gfp_t			sk_allocation;
 396	u32			sk_pacing_rate; /* bytes per second */
 397	u32			sk_max_pacing_rate;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 398	netdev_features_t	sk_route_caps;
 399	netdev_features_t	sk_route_nocaps;
 400	int			sk_gso_type;
 401	unsigned int		sk_gso_max_size;
 
 
 
 402	u16			sk_gso_max_segs;
 403	int			sk_rcvlowat;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 404	unsigned long	        sk_lingertime;
 405	struct sk_buff_head	sk_error_queue;
 406	struct proto		*sk_prot_creator;
 407	rwlock_t		sk_callback_lock;
 408	int			sk_err,
 409				sk_err_soft;
 410	u32			sk_ack_backlog;
 411	u32			sk_max_ack_backlog;
 412	__u32			sk_priority;
 413	__u32			sk_mark;
 
 414	struct pid		*sk_peer_pid;
 415	const struct cred	*sk_peer_cred;
 416	long			sk_rcvtimeo;
 417	long			sk_sndtimeo;
 418	struct timer_list	sk_timer;
 419	ktime_t			sk_stamp;
 420	u16			sk_tsflags;
 421	u32			sk_tskey;
 422	struct socket		*sk_socket;
 
 
 
 
 
 
 
 
 423	void			*sk_user_data;
 424	struct page_frag	sk_frag;
 425	struct sk_buff		*sk_send_head;
 426	__s32			sk_peek_off;
 427	int			sk_write_pending;
 428#ifdef CONFIG_SECURITY
 429	void			*sk_security;
 430#endif
 431	struct sock_cgroup_data	sk_cgrp_data;
 432	struct mem_cgroup	*sk_memcg;
 433	void			(*sk_state_change)(struct sock *sk);
 434	void			(*sk_data_ready)(struct sock *sk);
 435	void			(*sk_write_space)(struct sock *sk);
 436	void			(*sk_error_report)(struct sock *sk);
 437	int			(*sk_backlog_rcv)(struct sock *sk,
 438						  struct sk_buff *skb);
 439	void                    (*sk_destruct)(struct sock *sk);
 440	struct sock_reuseport __rcu	*sk_reuseport_cb;
 
 
 
 
 
 441};
 442
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 443#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
 444
 445#define rcu_dereference_sk_user_data(sk)	rcu_dereference(__sk_user_data((sk)))
 446#define rcu_assign_sk_user_data(sk, ptr)	rcu_assign_pointer(__sk_user_data((sk)), ptr)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 447
 448/*
 449 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
 450 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
 451 * on a socket means that the socket will reuse everybody else's port
 452 * without looking at the other's sk_reuse value.
 453 */
 454
 455#define SK_NO_REUSE	0
 456#define SK_CAN_REUSE	1
 457#define SK_FORCE_REUSE	2
 458
 459static inline int sk_peek_offset(struct sock *sk, int flags)
 
 
 460{
 461	if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
 462		return sk->sk_peek_off;
 463	else
 464		return 0;
 
 465}
 466
 467static inline void sk_peek_offset_bwd(struct sock *sk, int val)
 468{
 469	if (sk->sk_peek_off >= 0) {
 470		if (sk->sk_peek_off >= val)
 471			sk->sk_peek_off -= val;
 472		else
 473			sk->sk_peek_off = 0;
 474	}
 475}
 476
 477static inline void sk_peek_offset_fwd(struct sock *sk, int val)
 478{
 479	if (sk->sk_peek_off >= 0)
 480		sk->sk_peek_off += val;
 481}
 482
 483/*
 484 * Hashed lists helper routines
 485 */
 486static inline struct sock *sk_entry(const struct hlist_node *node)
 487{
 488	return hlist_entry(node, struct sock, sk_node);
 489}
 490
 491static inline struct sock *__sk_head(const struct hlist_head *head)
 492{
 493	return hlist_entry(head->first, struct sock, sk_node);
 494}
 495
 496static inline struct sock *sk_head(const struct hlist_head *head)
 497{
 498	return hlist_empty(head) ? NULL : __sk_head(head);
 499}
 500
 501static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
 502{
 503	return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
 504}
 505
 506static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
 507{
 508	return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
 509}
 510
 511static inline struct sock *sk_next(const struct sock *sk)
 512{
 513	return sk->sk_node.next ?
 514		hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
 515}
 516
 517static inline struct sock *sk_nulls_next(const struct sock *sk)
 518{
 519	return (!is_a_nulls(sk->sk_nulls_node.next)) ?
 520		hlist_nulls_entry(sk->sk_nulls_node.next,
 521				  struct sock, sk_nulls_node) :
 522		NULL;
 523}
 524
 525static inline bool sk_unhashed(const struct sock *sk)
 526{
 527	return hlist_unhashed(&sk->sk_node);
 528}
 529
 530static inline bool sk_hashed(const struct sock *sk)
 531{
 532	return !sk_unhashed(sk);
 533}
 534
 535static inline void sk_node_init(struct hlist_node *node)
 536{
 537	node->pprev = NULL;
 538}
 539
 540static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
 541{
 542	node->pprev = NULL;
 543}
 544
 545static inline void __sk_del_node(struct sock *sk)
 546{
 547	__hlist_del(&sk->sk_node);
 548}
 549
 550/* NB: equivalent to hlist_del_init_rcu */
 551static inline bool __sk_del_node_init(struct sock *sk)
 552{
 553	if (sk_hashed(sk)) {
 554		__sk_del_node(sk);
 555		sk_node_init(&sk->sk_node);
 556		return true;
 557	}
 558	return false;
 559}
 560
 561/* Grab socket reference count. This operation is valid only
 562   when sk is ALREADY grabbed f.e. it is found in hash table
 563   or a list and the lookup is made under lock preventing hash table
 564   modifications.
 565 */
 566
 567static inline void sock_hold(struct sock *sk)
 568{
 569	atomic_inc(&sk->sk_refcnt);
 570}
 571
 572/* Ungrab socket in the context, which assumes that socket refcnt
 573   cannot hit zero, f.e. it is true in context of any socketcall.
 574 */
 575static inline void __sock_put(struct sock *sk)
 576{
 577	atomic_dec(&sk->sk_refcnt);
 578}
 579
 580static inline bool sk_del_node_init(struct sock *sk)
 581{
 582	bool rc = __sk_del_node_init(sk);
 583
 584	if (rc) {
 585		/* paranoid for a while -acme */
 586		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
 587		__sock_put(sk);
 588	}
 589	return rc;
 590}
 591#define sk_del_node_init_rcu(sk)	sk_del_node_init(sk)
 592
 593static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
 594{
 595	if (sk_hashed(sk)) {
 596		hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
 597		return true;
 598	}
 599	return false;
 600}
 601
 602static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
 603{
 604	bool rc = __sk_nulls_del_node_init_rcu(sk);
 605
 606	if (rc) {
 607		/* paranoid for a while -acme */
 608		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
 609		__sock_put(sk);
 610	}
 611	return rc;
 612}
 613
 614static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
 615{
 616	hlist_add_head(&sk->sk_node, list);
 617}
 618
 619static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
 620{
 621	sock_hold(sk);
 622	__sk_add_node(sk, list);
 623}
 624
 625static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
 626{
 627	sock_hold(sk);
 628	hlist_add_head_rcu(&sk->sk_node, list);
 
 
 
 
 
 
 
 
 
 
 629}
 630
 631static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
 632{
 633	if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
 634	    sk->sk_family == AF_INET6)
 635		hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
 636	else
 637		hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
 
 638}
 639
 640static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
 641{
 642	sock_hold(sk);
 643	__sk_nulls_add_node_rcu(sk, list);
 644}
 645
 646static inline void __sk_del_bind_node(struct sock *sk)
 647{
 648	__hlist_del(&sk->sk_bind_node);
 649}
 650
 651static inline void sk_add_bind_node(struct sock *sk,
 652					struct hlist_head *list)
 653{
 654	hlist_add_head(&sk->sk_bind_node, list);
 655}
 656
 657#define sk_for_each(__sk, list) \
 658	hlist_for_each_entry(__sk, list, sk_node)
 659#define sk_for_each_rcu(__sk, list) \
 660	hlist_for_each_entry_rcu(__sk, list, sk_node)
 661#define sk_nulls_for_each(__sk, node, list) \
 662	hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
 663#define sk_nulls_for_each_rcu(__sk, node, list) \
 664	hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
 665#define sk_for_each_from(__sk) \
 666	hlist_for_each_entry_from(__sk, sk_node)
 667#define sk_nulls_for_each_from(__sk, node) \
 668	if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
 669		hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
 670#define sk_for_each_safe(__sk, tmp, list) \
 671	hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
 672#define sk_for_each_bound(__sk, list) \
 673	hlist_for_each_entry(__sk, list, sk_bind_node)
 674
 675/**
 676 * sk_nulls_for_each_entry_offset - iterate over a list at a given struct offset
 677 * @tpos:	the type * to use as a loop cursor.
 678 * @pos:	the &struct hlist_node to use as a loop cursor.
 679 * @head:	the head for your list.
 680 * @offset:	offset of hlist_node within the struct.
 681 *
 682 */
 683#define sk_nulls_for_each_entry_offset(tpos, pos, head, offset)		       \
 684	for (pos = (head)->first;					       \
 685	     (!is_a_nulls(pos)) &&					       \
 686		({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;});       \
 687	     pos = pos->next)
 688
 689static inline struct user_namespace *sk_user_ns(struct sock *sk)
 690{
 691	/* Careful only use this in a context where these parameters
 692	 * can not change and must all be valid, such as recvmsg from
 693	 * userspace.
 694	 */
 695	return sk->sk_socket->file->f_cred->user_ns;
 696}
 697
 698/* Sock flags */
 699enum sock_flags {
 700	SOCK_DEAD,
 701	SOCK_DONE,
 702	SOCK_URGINLINE,
 703	SOCK_KEEPOPEN,
 704	SOCK_LINGER,
 705	SOCK_DESTROY,
 706	SOCK_BROADCAST,
 707	SOCK_TIMESTAMP,
 708	SOCK_ZAPPED,
 709	SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
 710	SOCK_DBG, /* %SO_DEBUG setting */
 711	SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
 712	SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
 713	SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
 714	SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
 715	SOCK_MEMALLOC, /* VM depends on this socket for swapping */
 716	SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
 717	SOCK_FASYNC, /* fasync() active */
 718	SOCK_RXQ_OVFL,
 719	SOCK_ZEROCOPY, /* buffers from userspace */
 720	SOCK_WIFI_STATUS, /* push wifi status to userspace */
 721	SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
 722		     * Will use last 4 bytes of packet sent from
 723		     * user-space instead.
 724		     */
 725	SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
 726	SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
 
 
 
 
 
 727};
 728
 729#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
 730
 731static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
 732{
 733	nsk->sk_flags = osk->sk_flags;
 734}
 735
 736static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
 737{
 738	__set_bit(flag, &sk->sk_flags);
 739}
 740
 741static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
 742{
 743	__clear_bit(flag, &sk->sk_flags);
 744}
 745
 
 
 
 
 
 
 
 
 
 746static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
 747{
 748	return test_bit(flag, &sk->sk_flags);
 749}
 750
 751#ifdef CONFIG_NET
 752extern struct static_key memalloc_socks;
 753static inline int sk_memalloc_socks(void)
 754{
 755	return static_key_false(&memalloc_socks);
 756}
 
 
 757#else
 758
 759static inline int sk_memalloc_socks(void)
 760{
 761	return 0;
 762}
 763
 
 
 764#endif
 765
 766static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
 767{
 768	return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
 769}
 770
 771static inline void sk_acceptq_removed(struct sock *sk)
 772{
 773	sk->sk_ack_backlog--;
 774}
 775
 776static inline void sk_acceptq_added(struct sock *sk)
 777{
 778	sk->sk_ack_backlog++;
 779}
 780
 
 
 
 
 781static inline bool sk_acceptq_is_full(const struct sock *sk)
 782{
 783	return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
 784}
 785
 786/*
 787 * Compute minimal free write space needed to queue new packets.
 788 */
 789static inline int sk_stream_min_wspace(const struct sock *sk)
 790{
 791	return sk->sk_wmem_queued >> 1;
 792}
 793
 794static inline int sk_stream_wspace(const struct sock *sk)
 795{
 796	return sk->sk_sndbuf - sk->sk_wmem_queued;
 
 
 
 
 
 
 
 
 
 
 
 797}
 798
 799void sk_stream_write_space(struct sock *sk);
 800
 801/* OOB backlog add */
 802static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
 803{
 804	/* dont let skb dst not refcounted, we are going to leave rcu lock */
 805	skb_dst_force_safe(skb);
 806
 807	if (!sk->sk_backlog.tail)
 808		sk->sk_backlog.head = skb;
 809	else
 810		sk->sk_backlog.tail->next = skb;
 811
 812	sk->sk_backlog.tail = skb;
 813	skb->next = NULL;
 814}
 815
 816/*
 817 * Take into account size of receive queue and backlog queue
 818 * Do not take into account this skb truesize,
 819 * to allow even a single big packet to come.
 820 */
 821static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
 822{
 823	unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
 824
 825	return qsize > limit;
 826}
 827
 828/* The per-socket spinlock must be held here. */
 829static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
 830					      unsigned int limit)
 831{
 832	if (sk_rcvqueues_full(sk, limit))
 833		return -ENOBUFS;
 834
 835	/*
 836	 * If the skb was allocated from pfmemalloc reserves, only
 837	 * allow SOCK_MEMALLOC sockets to use it as this socket is
 838	 * helping free memory
 839	 */
 840	if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
 841		return -ENOMEM;
 842
 843	__sk_add_backlog(sk, skb);
 844	sk->sk_backlog.len += skb->truesize;
 845	return 0;
 846}
 847
 848int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
 849
 
 
 
 850static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
 851{
 852	if (sk_memalloc_socks() && skb_pfmemalloc(skb))
 853		return __sk_backlog_rcv(sk, skb);
 854
 855	return sk->sk_backlog_rcv(sk, skb);
 
 
 
 856}
 857
 858static inline void sk_incoming_cpu_update(struct sock *sk)
 859{
 860	sk->sk_incoming_cpu = raw_smp_processor_id();
 861}
 862
 863static inline void sock_rps_record_flow_hash(__u32 hash)
 864{
 865#ifdef CONFIG_RPS
 866	struct rps_sock_flow_table *sock_flow_table;
 867
 868	rcu_read_lock();
 869	sock_flow_table = rcu_dereference(rps_sock_flow_table);
 870	rps_record_sock_flow(sock_flow_table, hash);
 871	rcu_read_unlock();
 872#endif
 873}
 874
 875static inline void sock_rps_record_flow(const struct sock *sk)
 876{
 877#ifdef CONFIG_RPS
 878	sock_rps_record_flow_hash(sk->sk_rxhash);
 879#endif
 880}
 881
 882static inline void sock_rps_save_rxhash(struct sock *sk,
 883					const struct sk_buff *skb)
 884{
 885#ifdef CONFIG_RPS
 886	if (unlikely(sk->sk_rxhash != skb->hash))
 887		sk->sk_rxhash = skb->hash;
 
 
 
 888#endif
 889}
 890
 891static inline void sock_rps_reset_rxhash(struct sock *sk)
 892{
 893#ifdef CONFIG_RPS
 894	sk->sk_rxhash = 0;
 
 895#endif
 896}
 897
 898#define sk_wait_event(__sk, __timeo, __condition)			\
 899	({	int __rc;						\
 900		release_sock(__sk);					\
 901		__rc = __condition;					\
 902		if (!__rc) {						\
 903			*(__timeo) = schedule_timeout(*(__timeo));	\
 
 
 904		}							\
 905		sched_annotate_sleep();						\
 906		lock_sock(__sk);					\
 907		__rc = __condition;					\
 908		__rc;							\
 909	})
 910
 911int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
 912int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
 913void sk_stream_wait_close(struct sock *sk, long timeo_p);
 914int sk_stream_error(struct sock *sk, int flags, int err);
 915void sk_stream_kill_queues(struct sock *sk);
 916void sk_set_memalloc(struct sock *sk);
 917void sk_clear_memalloc(struct sock *sk);
 918
 
 
 
 
 
 
 
 
 
 
 
 919int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
 920
 921struct request_sock_ops;
 922struct timewait_sock_ops;
 923struct inet_hashinfo;
 924struct raw_hashinfo;
 
 925struct module;
 
 926
 927/*
 928 * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
 929 * un-modified. Special care is taken when initializing object to zero.
 930 */
 931static inline void sk_prot_clear_nulls(struct sock *sk, int size)
 932{
 933	if (offsetof(struct sock, sk_node.next) != 0)
 934		memset(sk, 0, offsetof(struct sock, sk_node.next));
 935	memset(&sk->sk_node.pprev, 0,
 936	       size - offsetof(struct sock, sk_node.pprev));
 937}
 938
 939/* Networking protocol blocks we attach to sockets.
 940 * socket layer -> transport layer interface
 941 */
 942struct proto {
 943	void			(*close)(struct sock *sk,
 944					long timeout);
 
 
 
 945	int			(*connect)(struct sock *sk,
 946					struct sockaddr *uaddr,
 947					int addr_len);
 948	int			(*disconnect)(struct sock *sk, int flags);
 949
 950	struct sock *		(*accept)(struct sock *sk, int flags, int *err);
 
 951
 952	int			(*ioctl)(struct sock *sk, int cmd,
 953					 unsigned long arg);
 954	int			(*init)(struct sock *sk);
 955	void			(*destroy)(struct sock *sk);
 956	void			(*shutdown)(struct sock *sk, int how);
 957	int			(*setsockopt)(struct sock *sk, int level,
 958					int optname, char __user *optval,
 959					unsigned int optlen);
 960	int			(*getsockopt)(struct sock *sk, int level,
 961					int optname, char __user *optval,
 962					int __user *option);
 
 963#ifdef CONFIG_COMPAT
 964	int			(*compat_setsockopt)(struct sock *sk,
 965					int level,
 966					int optname, char __user *optval,
 967					unsigned int optlen);
 968	int			(*compat_getsockopt)(struct sock *sk,
 969					int level,
 970					int optname, char __user *optval,
 971					int __user *option);
 972	int			(*compat_ioctl)(struct sock *sk,
 973					unsigned int cmd, unsigned long arg);
 974#endif
 975	int			(*sendmsg)(struct sock *sk, struct msghdr *msg,
 976					   size_t len);
 977	int			(*recvmsg)(struct sock *sk, struct msghdr *msg,
 978					   size_t len, int noblock, int flags,
 979					   int *addr_len);
 980	int			(*sendpage)(struct sock *sk, struct page *page,
 981					int offset, size_t size, int flags);
 982	int			(*bind)(struct sock *sk,
 983					struct sockaddr *uaddr, int addr_len);
 
 
 984
 985	int			(*backlog_rcv) (struct sock *sk,
 986						struct sk_buff *skb);
 
 
 987
 988	void		(*release_cb)(struct sock *sk);
 989
 990	/* Keeping track of sk's, looking them up, and port selection methods. */
 991	int			(*hash)(struct sock *sk);
 992	void			(*unhash)(struct sock *sk);
 993	void			(*rehash)(struct sock *sk);
 994	int			(*get_port)(struct sock *sk, unsigned short snum);
 995	void			(*clear_sk)(struct sock *sk, int size);
 
 
 
 
 
 996
 997	/* Keeping track of sockets in use */
 998#ifdef CONFIG_PROC_FS
 999	unsigned int		inuse_idx;
1000#endif
1001
1002	bool			(*stream_memory_free)(const struct sock *sk);
 
 
 
 
 
1003	/* Memory pressure */
1004	void			(*enter_memory_pressure)(struct sock *sk);
 
1005	atomic_long_t		*memory_allocated;	/* Current allocated memory. */
 
1006	struct percpu_counter	*sockets_allocated;	/* Current number of sockets. */
 
1007	/*
1008	 * Pressure flag: try to collapse.
1009	 * Technical note: it is used by multiple contexts non atomically.
 
1010	 * All the __sk_mem_schedule() is of this nature: accounting
1011	 * is strict, actions are advisory and have some latency.
1012	 */
1013	int			*memory_pressure;
1014	long			*sysctl_mem;
 
1015	int			*sysctl_wmem;
1016	int			*sysctl_rmem;
 
 
 
1017	int			max_header;
1018	bool			no_autobind;
1019
1020	struct kmem_cache	*slab;
1021	unsigned int		obj_size;
1022	int			slab_flags;
 
 
 
1023
1024	struct percpu_counter	*orphan_count;
1025
1026	struct request_sock_ops	*rsk_prot;
1027	struct timewait_sock_ops *twsk_prot;
1028
1029	union {
1030		struct inet_hashinfo	*hashinfo;
1031		struct udp_table	*udp_table;
1032		struct raw_hashinfo	*raw_hash;
 
1033	} h;
1034
1035	struct module		*owner;
1036
1037	char			name[32];
1038
1039	struct list_head	node;
1040#ifdef SOCK_REFCNT_DEBUG
1041	atomic_t		socks;
1042#endif
1043	int			(*diag_destroy)(struct sock *sk, int err);
1044};
1045
1046int proto_register(struct proto *prot, int alloc_slab);
1047void proto_unregister(struct proto *prot);
 
1048
1049#ifdef SOCK_REFCNT_DEBUG
1050static inline void sk_refcnt_debug_inc(struct sock *sk)
1051{
1052	atomic_inc(&sk->sk_prot->socks);
1053}
1054
1055static inline void sk_refcnt_debug_dec(struct sock *sk)
1056{
1057	atomic_dec(&sk->sk_prot->socks);
1058	printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1059	       sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
 
 
1060}
1061
1062static inline void sk_refcnt_debug_release(const struct sock *sk)
1063{
1064	if (atomic_read(&sk->sk_refcnt) != 1)
1065		printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1066		       sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
 
 
 
1067}
1068#else /* SOCK_REFCNT_DEBUG */
1069#define sk_refcnt_debug_inc(sk) do { } while (0)
1070#define sk_refcnt_debug_dec(sk) do { } while (0)
1071#define sk_refcnt_debug_release(sk) do { } while (0)
1072#endif /* SOCK_REFCNT_DEBUG */
1073
1074static inline bool sk_stream_memory_free(const struct sock *sk)
1075{
1076	if (sk->sk_wmem_queued >= sk->sk_sndbuf)
1077		return false;
1078
1079	return sk->sk_prot->stream_memory_free ?
1080		sk->sk_prot->stream_memory_free(sk) : true;
 
 
1081}
1082
1083static inline bool sk_stream_is_writeable(const struct sock *sk)
1084{
1085	return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1086	       sk_stream_memory_free(sk);
1087}
1088
 
 
 
 
 
 
 
 
 
 
1089
1090static inline bool sk_has_memory_pressure(const struct sock *sk)
1091{
1092	return sk->sk_prot->memory_pressure != NULL;
1093}
1094
 
 
 
 
 
 
1095static inline bool sk_under_memory_pressure(const struct sock *sk)
1096{
1097	if (!sk->sk_prot->memory_pressure)
1098		return false;
1099
1100	if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1101	    mem_cgroup_under_socket_pressure(sk->sk_memcg))
1102		return true;
1103
1104	return !!*sk->sk_prot->memory_pressure;
1105}
1106
1107static inline void sk_leave_memory_pressure(struct sock *sk)
 
1108{
1109	int *memory_pressure = sk->sk_prot->memory_pressure;
1110
1111	if (!memory_pressure)
1112		return;
1113
1114	if (*memory_pressure)
1115		*memory_pressure = 0;
1116}
1117
1118static inline void sk_enter_memory_pressure(struct sock *sk)
 
1119{
1120	if (!sk->sk_prot->enter_memory_pressure)
1121		return;
1122
1123	sk->sk_prot->enter_memory_pressure(sk);
1124}
1125
1126static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1127{
1128	return sk->sk_prot->sysctl_mem[index];
1129}
1130
1131static inline long
1132sk_memory_allocated(const struct sock *sk)
1133{
1134	return atomic_long_read(sk->sk_prot->memory_allocated);
 
 
 
1135}
1136
1137static inline long
1138sk_memory_allocated_add(struct sock *sk, int amt)
1139{
1140	return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
 
 
 
 
 
1141}
1142
1143static inline void
1144sk_memory_allocated_sub(struct sock *sk, int amt)
1145{
1146	atomic_long_sub(amt, sk->sk_prot->memory_allocated);
 
 
 
 
 
1147}
1148
 
 
1149static inline void sk_sockets_allocated_dec(struct sock *sk)
1150{
1151	percpu_counter_dec(sk->sk_prot->sockets_allocated);
 
1152}
1153
1154static inline void sk_sockets_allocated_inc(struct sock *sk)
1155{
1156	percpu_counter_inc(sk->sk_prot->sockets_allocated);
 
1157}
1158
1159static inline int
1160sk_sockets_allocated_read_positive(struct sock *sk)
1161{
1162	return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
1163}
1164
1165static inline int
1166proto_sockets_allocated_sum_positive(struct proto *prot)
1167{
1168	return percpu_counter_sum_positive(prot->sockets_allocated);
1169}
1170
1171static inline long
1172proto_memory_allocated(struct proto *prot)
1173{
1174	return atomic_long_read(prot->memory_allocated);
1175}
1176
1177static inline bool
1178proto_memory_pressure(struct proto *prot)
1179{
1180	if (!prot->memory_pressure)
1181		return false;
1182	return !!*prot->memory_pressure;
1183}
1184
1185
1186#ifdef CONFIG_PROC_FS
1187/* Called with local bh disabled */
1188void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1189int sock_prot_inuse_get(struct net *net, struct proto *proto);
 
1190#else
1191static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1192		int inc)
 
 
 
 
1193{
1194}
1195#endif
1196
1197
1198/* With per-bucket locks this operation is not-atomic, so that
1199 * this version is not worse.
1200 */
1201static inline int __sk_prot_rehash(struct sock *sk)
1202{
1203	sk->sk_prot->unhash(sk);
1204	return sk->sk_prot->hash(sk);
1205}
1206
1207void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
1208
1209/* About 10 seconds */
1210#define SOCK_DESTROY_TIME (10*HZ)
1211
1212/* Sockets 0-1023 can't be bound to unless you are superuser */
1213#define PROT_SOCK	1024
1214
1215#define SHUTDOWN_MASK	3
1216#define RCV_SHUTDOWN	1
1217#define SEND_SHUTDOWN	2
1218
1219#define SOCK_SNDBUF_LOCK	1
1220#define SOCK_RCVBUF_LOCK	2
1221#define SOCK_BINDADDR_LOCK	4
1222#define SOCK_BINDPORT_LOCK	8
1223
1224struct socket_alloc {
1225	struct socket socket;
1226	struct inode vfs_inode;
1227};
1228
1229static inline struct socket *SOCKET_I(struct inode *inode)
1230{
1231	return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1232}
1233
1234static inline struct inode *SOCK_INODE(struct socket *socket)
1235{
1236	return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1237}
1238
1239/*
1240 * Functions for memory accounting
1241 */
 
1242int __sk_mem_schedule(struct sock *sk, int size, int kind);
 
1243void __sk_mem_reclaim(struct sock *sk, int amount);
1244
1245#define SK_MEM_QUANTUM ((int)PAGE_SIZE)
1246#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1247#define SK_MEM_SEND	0
1248#define SK_MEM_RECV	1
1249
 
 
 
 
 
 
1250static inline int sk_mem_pages(int amt)
1251{
1252	return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1253}
1254
1255static inline bool sk_has_account(struct sock *sk)
1256{
1257	/* return true if protocol supports memory accounting */
1258	return !!sk->sk_prot->memory_allocated;
1259}
1260
1261static inline bool sk_wmem_schedule(struct sock *sk, int size)
1262{
 
 
1263	if (!sk_has_account(sk))
1264		return true;
1265	return size <= sk->sk_forward_alloc ||
1266		__sk_mem_schedule(sk, size, SK_MEM_SEND);
1267}
1268
1269static inline bool
1270sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1271{
 
 
1272	if (!sk_has_account(sk))
1273		return true;
1274	return size<= sk->sk_forward_alloc ||
1275		__sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1276		skb_pfmemalloc(skb);
1277}
1278
 
 
 
 
 
 
 
 
 
 
 
 
 
1279static inline void sk_mem_reclaim(struct sock *sk)
1280{
 
 
1281	if (!sk_has_account(sk))
1282		return;
1283	if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1284		__sk_mem_reclaim(sk, sk->sk_forward_alloc);
 
 
 
1285}
1286
1287static inline void sk_mem_reclaim_partial(struct sock *sk)
1288{
1289	if (!sk_has_account(sk))
1290		return;
1291	if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1292		__sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
1293}
1294
1295static inline void sk_mem_charge(struct sock *sk, int size)
1296{
1297	if (!sk_has_account(sk))
1298		return;
1299	sk->sk_forward_alloc -= size;
1300}
1301
1302static inline void sk_mem_uncharge(struct sock *sk, int size)
1303{
1304	if (!sk_has_account(sk))
1305		return;
1306	sk->sk_forward_alloc += size;
1307}
1308
1309static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1310{
1311	sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1312	sk->sk_wmem_queued -= skb->truesize;
1313	sk_mem_uncharge(sk, skb->truesize);
1314	__kfree_skb(skb);
1315}
1316
1317/* Used by processes to "lock" a socket state, so that
1318 * interrupts and bottom half handlers won't change it
1319 * from under us. It essentially blocks any incoming
1320 * packets, so that we won't get any new data or any
1321 * packets that change the state of the socket.
1322 *
1323 * While locked, BH processing will add new packets to
1324 * the backlog queue.  This queue is processed by the
1325 * owner of the socket lock right before it is released.
1326 *
1327 * Since ~2.3.5 it is also exclusive sleep lock serializing
1328 * accesses from user process context.
1329 */
1330#define sock_owned_by_user(sk)	((sk)->sk_lock.owned)
1331
1332static inline void sock_release_ownership(struct sock *sk)
1333{
1334	sk->sk_lock.owned = 0;
1335}
1336
1337/*
1338 * Macro so as to not evaluate some arguments when
1339 * lockdep is not enabled.
1340 *
1341 * Mark both the sk_lock and the sk_lock.slock as a
1342 * per-address-family lock class.
1343 */
1344#define sock_lock_init_class_and_name(sk, sname, skey, name, key)	\
1345do {									\
1346	sk->sk_lock.owned = 0;						\
1347	init_waitqueue_head(&sk->sk_lock.wq);				\
1348	spin_lock_init(&(sk)->sk_lock.slock);				\
1349	debug_check_no_locks_freed((void *)&(sk)->sk_lock,		\
1350			sizeof((sk)->sk_lock));				\
1351	lockdep_set_class_and_name(&(sk)->sk_lock.slock,		\
1352				(skey), (sname));				\
1353	lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);	\
1354} while (0)
1355
 
 
 
 
 
 
1356void lock_sock_nested(struct sock *sk, int subclass);
1357
1358static inline void lock_sock(struct sock *sk)
1359{
1360	lock_sock_nested(sk, 0);
1361}
1362
 
 
1363void release_sock(struct sock *sk);
1364
1365/* BH context may only use the following locking interface. */
1366#define bh_lock_sock(__sk)	spin_lock(&((__sk)->sk_lock.slock))
1367#define bh_lock_sock_nested(__sk) \
1368				spin_lock_nested(&((__sk)->sk_lock.slock), \
1369				SINGLE_DEPTH_NESTING)
1370#define bh_unlock_sock(__sk)	spin_unlock(&((__sk)->sk_lock.slock))
1371
1372bool lock_sock_fast(struct sock *sk);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1373/**
1374 * unlock_sock_fast - complement of lock_sock_fast
1375 * @sk: socket
1376 * @slow: slow mode
1377 *
1378 * fast unlock socket for user context.
1379 * If slow mode is on, we call regular release_sock()
1380 */
1381static inline void unlock_sock_fast(struct sock *sk, bool slow)
 
1382{
1383	if (slow)
1384		release_sock(sk);
1385	else
 
 
1386		spin_unlock_bh(&sk->sk_lock.slock);
 
1387}
1388
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1389
1390struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1391		      struct proto *prot, int kern);
1392void sk_free(struct sock *sk);
1393void sk_destruct(struct sock *sk);
1394struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
 
1395
1396struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1397			     gfp_t priority);
 
1398void sock_wfree(struct sk_buff *skb);
 
 
1399void skb_orphan_partial(struct sk_buff *skb);
1400void sock_rfree(struct sk_buff *skb);
1401void sock_efree(struct sk_buff *skb);
1402#ifdef CONFIG_INET
1403void sock_edemux(struct sk_buff *skb);
 
1404#else
1405#define sock_edemux(skb) sock_efree(skb)
1406#endif
1407
 
 
1408int sock_setsockopt(struct socket *sock, int level, int op,
1409		    char __user *optval, unsigned int optlen);
1410
1411int sock_getsockopt(struct socket *sock, int level, int op,
1412		    char __user *optval, int __user *optlen);
1413struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1414				    int noblock, int *errcode);
 
 
 
 
1415struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1416				     unsigned long data_len, int noblock,
1417				     int *errcode, int max_page_order);
 
 
 
 
 
 
 
 
1418void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1419void sock_kfree_s(struct sock *sk, void *mem, int size);
1420void sock_kzfree_s(struct sock *sk, void *mem, int size);
1421void sk_send_sigurg(struct sock *sk);
1422
 
 
 
 
 
 
 
1423struct sockcm_cookie {
 
1424	u32 mark;
 
1425};
1426
 
 
 
 
 
 
 
 
 
 
1427int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1428		   struct sockcm_cookie *sockc);
1429
1430/*
1431 * Functions to fill in entries in struct proto_ops when a protocol
1432 * does not implement a particular function.
1433 */
1434int sock_no_bind(struct socket *, struct sockaddr *, int);
1435int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1436int sock_no_socketpair(struct socket *, struct socket *);
1437int sock_no_accept(struct socket *, struct socket *, int);
1438int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
1439unsigned int sock_no_poll(struct file *, struct socket *,
1440			  struct poll_table_struct *);
1441int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1442int sock_no_listen(struct socket *, int);
1443int sock_no_shutdown(struct socket *, int);
1444int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1445int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1446int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
 
1447int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
1448int sock_no_mmap(struct file *file, struct socket *sock,
1449		 struct vm_area_struct *vma);
1450ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1451			 size_t size, int flags);
1452
1453/*
1454 * Functions to fill in entries in struct proto_ops when a protocol
1455 * uses the inet style.
1456 */
1457int sock_common_getsockopt(struct socket *sock, int level, int optname,
1458				  char __user *optval, int __user *optlen);
1459int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1460			int flags);
1461int sock_common_setsockopt(struct socket *sock, int level, int optname,
1462				  char __user *optval, unsigned int optlen);
1463int compat_sock_common_getsockopt(struct socket *sock, int level,
1464		int optname, char __user *optval, int __user *optlen);
1465int compat_sock_common_setsockopt(struct socket *sock, int level,
1466		int optname, char __user *optval, unsigned int optlen);
1467
1468void sk_common_release(struct sock *sk);
1469
1470/*
1471 *	Default socket callbacks and setup code
1472 */
1473
1474/* Initialise core socket variables */
 
 
 
 
 
1475void sock_init_data(struct socket *sock, struct sock *sk);
1476
1477/*
1478 * Socket reference counting postulates.
1479 *
1480 * * Each user of socket SHOULD hold a reference count.
1481 * * Each access point to socket (an hash table bucket, reference from a list,
1482 *   running timer, skb in flight MUST hold a reference count.
1483 * * When reference count hits 0, it means it will never increase back.
1484 * * When reference count hits 0, it means that no references from
1485 *   outside exist to this socket and current process on current CPU
1486 *   is last user and may/should destroy this socket.
1487 * * sk_free is called from any context: process, BH, IRQ. When
1488 *   it is called, socket has no references from outside -> sk_free
1489 *   may release descendant resources allocated by the socket, but
1490 *   to the time when it is called, socket is NOT referenced by any
1491 *   hash tables, lists etc.
1492 * * Packets, delivered from outside (from network or from another process)
1493 *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1494 *   when they sit in queue. Otherwise, packets will leak to hole, when
1495 *   socket is looked up by one cpu and unhasing is made by another CPU.
1496 *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1497 *   (leak to backlog). Packet socket does all the processing inside
1498 *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1499 *   use separate SMP lock, so that they are prone too.
1500 */
1501
1502/* Ungrab socket and destroy it, if it was the last reference. */
1503static inline void sock_put(struct sock *sk)
1504{
1505	if (atomic_dec_and_test(&sk->sk_refcnt))
1506		sk_free(sk);
1507}
1508/* Generic version of sock_put(), dealing with all sockets
1509 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1510 */
1511void sock_gen_put(struct sock *sk);
1512
1513int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
 
 
 
 
 
 
1514
1515static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1516{
1517	sk->sk_tx_queue_mapping = tx_queue;
 
 
 
 
 
 
1518}
1519
 
 
1520static inline void sk_tx_queue_clear(struct sock *sk)
1521{
1522	sk->sk_tx_queue_mapping = -1;
 
 
 
1523}
1524
1525static inline int sk_tx_queue_get(const struct sock *sk)
1526{
1527	return sk ? sk->sk_tx_queue_mapping : -1;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1528}
1529
1530static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1531{
1532	sk_tx_queue_clear(sk);
1533	sk->sk_socket = sock;
1534}
1535
1536static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1537{
1538	BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1539	return &rcu_dereference_raw(sk->sk_wq)->wait;
1540}
1541/* Detach socket from process context.
1542 * Announce socket dead, detach it from wait queue and inode.
1543 * Note that parent inode held reference count on this struct sock,
1544 * we do not release it in this function, because protocol
1545 * probably wants some additional cleanups or even continuing
1546 * to work with this socket (TCP).
1547 */
1548static inline void sock_orphan(struct sock *sk)
1549{
1550	write_lock_bh(&sk->sk_callback_lock);
1551	sock_set_flag(sk, SOCK_DEAD);
1552	sk_set_socket(sk, NULL);
1553	sk->sk_wq  = NULL;
1554	write_unlock_bh(&sk->sk_callback_lock);
1555}
1556
1557static inline void sock_graft(struct sock *sk, struct socket *parent)
1558{
 
1559	write_lock_bh(&sk->sk_callback_lock);
1560	sk->sk_wq = parent->wq;
1561	parent->sk = sk;
1562	sk_set_socket(sk, parent);
 
1563	security_sock_graft(sk, parent);
1564	write_unlock_bh(&sk->sk_callback_lock);
1565}
1566
1567kuid_t sock_i_uid(struct sock *sk);
 
1568unsigned long sock_i_ino(struct sock *sk);
1569
 
 
 
 
 
1570static inline u32 net_tx_rndhash(void)
1571{
1572	u32 v = prandom_u32();
1573
1574	return v ?: 1;
1575}
1576
1577static inline void sk_set_txhash(struct sock *sk)
1578{
1579	sk->sk_txhash = net_tx_rndhash();
 
1580}
1581
1582static inline void sk_rethink_txhash(struct sock *sk)
1583{
1584	if (sk->sk_txhash)
1585		sk_set_txhash(sk);
 
 
 
1586}
1587
1588static inline struct dst_entry *
1589__sk_dst_get(struct sock *sk)
1590{
1591	return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
1592						       lockdep_is_held(&sk->sk_lock.slock));
1593}
1594
1595static inline struct dst_entry *
1596sk_dst_get(struct sock *sk)
1597{
1598	struct dst_entry *dst;
1599
1600	rcu_read_lock();
1601	dst = rcu_dereference(sk->sk_dst_cache);
1602	if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1603		dst = NULL;
1604	rcu_read_unlock();
1605	return dst;
1606}
1607
1608static inline void dst_negative_advice(struct sock *sk)
1609{
1610	struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1611
1612	sk_rethink_txhash(sk);
1613
1614	if (dst && dst->ops->negative_advice) {
1615		ndst = dst->ops->negative_advice(dst);
 
1616
1617		if (ndst != dst) {
1618			rcu_assign_pointer(sk->sk_dst_cache, ndst);
1619			sk_tx_queue_clear(sk);
1620		}
1621	}
1622}
1623
1624static inline void
1625__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1626{
1627	struct dst_entry *old_dst;
1628
1629	sk_tx_queue_clear(sk);
1630	/*
1631	 * This can be called while sk is owned by the caller only,
1632	 * with no state that can be checked in a rcu_dereference_check() cond
1633	 */
1634	old_dst = rcu_dereference_raw(sk->sk_dst_cache);
1635	rcu_assign_pointer(sk->sk_dst_cache, dst);
1636	dst_release(old_dst);
1637}
1638
1639static inline void
1640sk_dst_set(struct sock *sk, struct dst_entry *dst)
1641{
1642	struct dst_entry *old_dst;
1643
1644	sk_tx_queue_clear(sk);
 
1645	old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1646	dst_release(old_dst);
1647}
1648
1649static inline void
1650__sk_dst_reset(struct sock *sk)
1651{
1652	__sk_dst_set(sk, NULL);
1653}
1654
1655static inline void
1656sk_dst_reset(struct sock *sk)
1657{
1658	sk_dst_set(sk, NULL);
1659}
1660
1661struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1662
1663struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1664
1665bool sk_mc_loop(struct sock *sk);
 
 
 
 
1666
1667static inline bool sk_can_gso(const struct sock *sk)
1668{
1669	return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
 
 
 
 
 
 
1670}
1671
1672void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1673
1674static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1675{
1676	sk->sk_route_nocaps |= flags;
1677	sk->sk_route_caps &= ~flags;
1678}
1679
1680static inline bool sk_check_csum_caps(struct sock *sk)
 
 
1681{
1682	return (sk->sk_route_caps & NETIF_F_HW_CSUM) ||
1683	       (sk->sk_family == PF_INET &&
1684		(sk->sk_route_caps & NETIF_F_IP_CSUM)) ||
1685	       (sk->sk_family == PF_INET6 &&
1686		(sk->sk_route_caps & NETIF_F_IPV6_CSUM));
1687}
1688
1689static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1690					   struct iov_iter *from, char *to,
1691					   int copy, int offset)
1692{
1693	if (skb->ip_summed == CHECKSUM_NONE) {
1694		__wsum csum = 0;
1695		if (csum_and_copy_from_iter(to, copy, &csum, from) != copy)
1696			return -EFAULT;
1697		skb->csum = csum_block_add(skb->csum, csum, offset);
1698	} else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1699		if (copy_from_iter_nocache(to, copy, from) != copy)
1700			return -EFAULT;
1701	} else if (copy_from_iter(to, copy, from) != copy)
1702		return -EFAULT;
1703
1704	return 0;
1705}
1706
1707static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
1708				       struct iov_iter *from, int copy)
1709{
1710	int err, offset = skb->len;
1711
1712	err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
1713				       copy, offset);
1714	if (err)
1715		__skb_trim(skb, offset);
1716
1717	return err;
1718}
1719
1720static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
1721					   struct sk_buff *skb,
1722					   struct page *page,
1723					   int off, int copy)
1724{
1725	int err;
1726
1727	err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
1728				       copy, skb->len);
1729	if (err)
1730		return err;
1731
1732	skb->len	     += copy;
1733	skb->data_len	     += copy;
1734	skb->truesize	     += copy;
1735	sk->sk_wmem_queued   += copy;
1736	sk_mem_charge(sk, copy);
1737	return 0;
1738}
1739
1740/**
1741 * sk_wmem_alloc_get - returns write allocations
1742 * @sk: socket
1743 *
1744 * Returns sk_wmem_alloc minus initial offset of one
1745 */
1746static inline int sk_wmem_alloc_get(const struct sock *sk)
1747{
1748	return atomic_read(&sk->sk_wmem_alloc) - 1;
1749}
1750
1751/**
1752 * sk_rmem_alloc_get - returns read allocations
1753 * @sk: socket
1754 *
1755 * Returns sk_rmem_alloc
1756 */
1757static inline int sk_rmem_alloc_get(const struct sock *sk)
1758{
1759	return atomic_read(&sk->sk_rmem_alloc);
1760}
1761
1762/**
1763 * sk_has_allocations - check if allocations are outstanding
1764 * @sk: socket
1765 *
1766 * Returns true if socket has write or read allocations
1767 */
1768static inline bool sk_has_allocations(const struct sock *sk)
1769{
1770	return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
1771}
1772
1773/**
1774 * skwq_has_sleeper - check if there are any waiting processes
1775 * @wq: struct socket_wq
1776 *
1777 * Returns true if socket_wq has waiting processes
1778 *
1779 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
1780 * barrier call. They were added due to the race found within the tcp code.
1781 *
1782 * Consider following tcp code paths:
1783 *
1784 * CPU1                  CPU2
1785 *
1786 * sys_select            receive packet
1787 *   ...                 ...
1788 *   __add_wait_queue    update tp->rcv_nxt
1789 *   ...                 ...
1790 *   tp->rcv_nxt check   sock_def_readable
1791 *   ...                 {
1792 *   schedule               rcu_read_lock();
1793 *                          wq = rcu_dereference(sk->sk_wq);
1794 *                          if (wq && waitqueue_active(&wq->wait))
1795 *                              wake_up_interruptible(&wq->wait)
1796 *                          ...
1797 *                       }
1798 *
1799 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
1800 * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
1801 * could then endup calling schedule and sleep forever if there are no more
1802 * data on the socket.
1803 *
1804 */
1805static inline bool skwq_has_sleeper(struct socket_wq *wq)
1806{
1807	return wq && wq_has_sleeper(&wq->wait);
1808}
1809
1810/**
1811 * sock_poll_wait - place memory barrier behind the poll_wait call.
1812 * @filp:           file
1813 * @wait_address:   socket wait queue
1814 * @p:              poll_table
1815 *
1816 * See the comments in the wq_has_sleeper function.
1817 */
1818static inline void sock_poll_wait(struct file *filp,
1819		wait_queue_head_t *wait_address, poll_table *p)
1820{
1821	if (!poll_does_not_wait(p) && wait_address) {
1822		poll_wait(filp, wait_address, p);
1823		/* We need to be sure we are in sync with the
1824		 * socket flags modification.
1825		 *
1826		 * This memory barrier is paired in the wq_has_sleeper.
1827		 */
1828		smp_mb();
1829	}
1830}
1831
1832static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
1833{
1834	if (sk->sk_txhash) {
 
 
 
1835		skb->l4_hash = 1;
1836		skb->hash = sk->sk_txhash;
1837	}
1838}
1839
1840void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
1841
1842/*
1843 *	Queue a received datagram if it will fit. Stream and sequenced
1844 *	protocols can't normally use this as they need to fit buffers in
1845 *	and play with them.
1846 *
1847 *	Inlined as it's very short and called for pretty much every
1848 *	packet ever received.
1849 */
1850static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
1851{
1852	skb_orphan(skb);
1853	skb->sk = sk;
1854	skb->destructor = sock_rfree;
1855	atomic_add(skb->truesize, &sk->sk_rmem_alloc);
1856	sk_mem_charge(sk, skb->truesize);
1857}
1858
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1859void sk_reset_timer(struct sock *sk, struct timer_list *timer,
1860		    unsigned long expires);
1861
1862void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1863
1864int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1865
1866int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
1867struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
1868
1869/*
1870 *	Recover an error report and clear atomically
1871 */
1872
1873static inline int sock_error(struct sock *sk)
1874{
1875	int err;
1876	if (likely(!sk->sk_err))
 
 
 
 
1877		return 0;
 
1878	err = xchg(&sk->sk_err, 0);
1879	return -err;
1880}
1881
 
 
1882static inline unsigned long sock_wspace(struct sock *sk)
1883{
1884	int amt = 0;
1885
1886	if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
1887		amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
1888		if (amt < 0)
1889			amt = 0;
1890	}
1891	return amt;
1892}
1893
1894/* Note:
1895 *  We use sk->sk_wq_raw, from contexts knowing this
1896 *  pointer is not NULL and cannot disappear/change.
1897 */
1898static inline void sk_set_bit(int nr, struct sock *sk)
1899{
 
 
 
 
1900	set_bit(nr, &sk->sk_wq_raw->flags);
1901}
1902
1903static inline void sk_clear_bit(int nr, struct sock *sk)
1904{
 
 
 
 
1905	clear_bit(nr, &sk->sk_wq_raw->flags);
1906}
1907
1908static inline void sk_wake_async(const struct sock *sk, int how, int band)
1909{
1910	if (sock_flag(sk, SOCK_FASYNC)) {
1911		rcu_read_lock();
1912		sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
1913		rcu_read_unlock();
1914	}
1915}
1916
1917/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
1918 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
1919 * Note: for send buffers, TCP works better if we can build two skbs at
1920 * minimum.
1921 */
1922#define TCP_SKB_MIN_TRUESIZE	(2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
1923
1924#define SOCK_MIN_SNDBUF		(TCP_SKB_MIN_TRUESIZE * 2)
1925#define SOCK_MIN_RCVBUF		 TCP_SKB_MIN_TRUESIZE
1926
1927static inline void sk_stream_moderate_sndbuf(struct sock *sk)
1928{
1929	if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
1930		sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
1931		sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
1932	}
1933}
1934
1935struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
1936				    bool force_schedule);
 
 
 
 
 
 
1937
1938/**
1939 * sk_page_frag - return an appropriate page_frag
1940 * @sk: socket
1941 *
1942 * If socket allocation mode allows current thread to sleep, it means its
1943 * safe to use the per task page_frag instead of the per socket one.
 
 
 
 
 
 
 
 
 
1944 */
1945static inline struct page_frag *sk_page_frag(struct sock *sk)
1946{
1947	if (gfpflags_allow_blocking(sk->sk_allocation))
1948		return &current->task_frag;
1949
1950	return &sk->sk_frag;
1951}
1952
1953bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
1954
1955/*
1956 *	Default write policy as shown to user space via poll/select/SIGIO
1957 */
1958static inline bool sock_writeable(const struct sock *sk)
1959{
1960	return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
1961}
1962
1963static inline gfp_t gfp_any(void)
1964{
1965	return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1966}
1967
 
 
 
 
 
1968static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
1969{
1970	return noblock ? 0 : sk->sk_rcvtimeo;
1971}
1972
1973static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
1974{
1975	return noblock ? 0 : sk->sk_sndtimeo;
1976}
1977
1978static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
1979{
1980	return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
 
 
1981}
1982
1983/* Alas, with timeout socket operations are not restartable.
1984 * Compare this to poll().
1985 */
1986static inline int sock_intr_errno(long timeo)
1987{
1988	return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
1989}
1990
1991struct sock_skb_cb {
1992	u32 dropcount;
1993};
1994
1995/* Store sock_skb_cb at the end of skb->cb[] so protocol families
1996 * using skb->cb[] would keep using it directly and utilize its
1997 * alignement guarantee.
1998 */
1999#define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2000			    sizeof(struct sock_skb_cb)))
2001
2002#define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2003			    SOCK_SKB_CB_OFFSET))
2004
2005#define sock_skb_cb_check_size(size) \
2006	BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2007
2008static inline void
2009sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2010{
2011	SOCK_SKB_CB(skb)->dropcount = atomic_read(&sk->sk_drops);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2012}
2013
2014void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2015			   struct sk_buff *skb);
2016void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2017			     struct sk_buff *skb);
2018
2019static inline void
2020sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2021{
2022	ktime_t kt = skb->tstamp;
2023	struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2024
 
2025	/*
2026	 * generate control messages if
2027	 * - receive time stamping in software requested
2028	 * - software time stamp available and wanted
2029	 * - hardware time stamps available and wanted
2030	 */
2031	if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2032	    (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2033	    (kt.tv64 && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2034	    (hwtstamps->hwtstamp.tv64 &&
2035	     (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2036		__sock_recv_timestamp(msg, sk, skb);
2037	else
2038		sk->sk_stamp = kt;
2039
2040	if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2041		__sock_recv_wifi_status(msg, sk, skb);
2042}
2043
2044void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2045			      struct sk_buff *skb);
2046
2047static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2048					  struct sk_buff *skb)
2049{
2050#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)			| \
2051			   (1UL << SOCK_RCVTSTAMP))
 
 
2052#define TSFLAGS_ANY	  (SOF_TIMESTAMPING_SOFTWARE			| \
2053			   SOF_TIMESTAMPING_RAW_HARDWARE)
2054
2055	if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
2056		__sock_recv_ts_and_drops(msg, sk, skb);
2057	else
2058		sk->sk_stamp = skb->tstamp;
 
 
 
2059}
2060
2061void __sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags);
2062
2063/**
2064 * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2065 * @sk:		socket sending this packet
 
2066 * @tx_flags:	completed with instructions for time stamping
 
2067 *
2068 * Note : callers should take care of initial *tx_flags value (usually 0)
2069 */
2070static inline void sock_tx_timestamp(const struct sock *sk, __u8 *tx_flags)
 
2071{
2072	if (unlikely(sk->sk_tsflags))
2073		__sock_tx_timestamp(sk, tx_flags);
 
 
 
 
2074	if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2075		*tx_flags |= SKBTX_WIFI_STATUS;
2076}
2077
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2078/**
2079 * sk_eat_skb - Release a skb if it is no longer needed
2080 * @sk: socket to eat this skb from
2081 * @skb: socket buffer to eat
2082 *
2083 * This routine must be called with interrupts disabled or with the socket
2084 * locked so that the sk_buff queue operation is ok.
2085*/
2086static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
2087{
2088	__skb_unlink(skb, &sk->sk_receive_queue);
2089	__kfree_skb(skb);
2090}
2091
2092static inline
2093struct net *sock_net(const struct sock *sk)
2094{
2095	return read_pnet(&sk->sk_net);
2096}
2097
2098static inline
2099void sock_net_set(struct sock *sk, struct net *net)
2100{
2101	write_pnet(&sk->sk_net, net);
2102}
2103
2104static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2105{
2106	if (skb->sk) {
2107		struct sock *sk = skb->sk;
2108
2109		skb->destructor = NULL;
2110		skb->sk = NULL;
2111		return sk;
2112	}
2113	return NULL;
2114}
2115
2116/* This helper checks if a socket is a full socket,
2117 * ie _not_ a timewait or request socket.
2118 */
2119static inline bool sk_fullsock(const struct sock *sk)
2120{
2121	return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2122}
2123
2124/* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2125 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2126 */
2127static inline bool sk_listener(const struct sock *sk)
2128{
2129	return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
 
2130}
2131
2132/**
2133 * sk_state_load - read sk->sk_state for lockless contexts
2134 * @sk: socket pointer
2135 *
2136 * Paired with sk_state_store(). Used in places we do not hold socket lock :
2137 * tcp_diag_get_info(), tcp_get_info(), tcp_poll(), get_tcp4_sock() ...
2138 */
2139static inline int sk_state_load(const struct sock *sk)
 
2140{
2141	return smp_load_acquire(&sk->sk_state);
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2142}
2143
2144/**
2145 * sk_state_store - update sk->sk_state
2146 * @sk: socket pointer
2147 * @newstate: new state
2148 *
2149 * Paired with sk_state_load(). Should be used in contexts where
2150 * state change might impact lockless readers.
2151 */
2152static inline void sk_state_store(struct sock *sk, int newstate)
2153{
2154	smp_store_release(&sk->sk_state, newstate);
2155}
2156
2157void sock_enable_timestamp(struct sock *sk, int flag);
2158int sock_get_timestamp(struct sock *, struct timeval __user *);
2159int sock_get_timestampns(struct sock *, struct timespec __user *);
2160int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2161		       int type);
2162
2163bool sk_ns_capable(const struct sock *sk,
2164		   struct user_namespace *user_ns, int cap);
2165bool sk_capable(const struct sock *sk, int cap);
2166bool sk_net_capable(const struct sock *sk, int cap);
2167
 
 
 
 
 
 
 
 
 
 
 
 
2168extern __u32 sysctl_wmem_max;
2169extern __u32 sysctl_rmem_max;
2170
2171extern int sysctl_tstamp_allow_data;
2172extern int sysctl_optmem_max;
2173
2174extern __u32 sysctl_wmem_default;
2175extern __u32 sysctl_rmem_default;
2176
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
2177#endif	/* _SOCK_H */