include/net/sock.h

/* [<][>][^][v][top][bottom][index][help] */
This source file includes following definitions.
__printf
sk_peek_offset
sk_peek_offset_bwd
sk_peek_offset_fwd
sk_entry
__sk_head
sk_head
__sk_nulls_head
sk_nulls_head
sk_next
sk_nulls_next
sk_unhashed
sk_hashed
sk_node_init
sk_nulls_node_init
__sk_del_node
__sk_del_node_init
sock_hold
__sock_put
sk_del_node_init
__sk_nulls_del_node_init_rcu
sk_nulls_del_node_init_rcu
__sk_add_node
sk_add_node
sk_add_node_rcu
sk_add_node_tail_rcu
__sk_nulls_add_node_rcu
__sk_nulls_add_node_tail_rcu
sk_nulls_add_node_rcu
__sk_del_bind_node
sk_add_bind_node
sk_user_ns
sock_copy_flags
sock_set_flag
sock_reset_flag
sock_flag
sk_memalloc_socks
sk_memalloc_socks
sk_gfp_mask
sk_acceptq_removed
sk_acceptq_added
sk_acceptq_is_full
sk_stream_min_wspace
sk_stream_wspace
sk_wmem_queued_add
__sk_add_backlog
sk_rcvqueues_full
sk_add_backlog
sk_backlog_rcv
sk_incoming_cpu_update
sock_rps_record_flow_hash
sock_rps_record_flow
sock_rps_save_rxhash
sock_rps_reset_rxhash
sk_flush_backlog
sk_prot_clear_nulls
sk_refcnt_debug_inc
sk_refcnt_debug_dec
sk_refcnt_debug_release
__sk_stream_memory_free
sk_stream_memory_free
__sk_stream_is_writeable
sk_stream_is_writeable
sk_under_cgroup_hierarchy
sk_has_memory_pressure
sk_under_memory_pressure
sk_memory_allocated
sk_memory_allocated_add
sk_memory_allocated_sub
sk_sockets_allocated_dec
sk_sockets_allocated_inc
sk_sockets_allocated_read_positive
proto_sockets_allocated_sum_positive
proto_memory_allocated
proto_memory_pressure
sock_prot_inuse_add
__sk_prot_rehash
SOCKET_I
SOCK_INODE
sk_prot_mem_limits
sk_mem_pages
sk_has_account
sk_wmem_schedule
sk_rmem_schedule
sk_mem_reclaim
sk_mem_reclaim_partial
sk_mem_charge
sk_mem_uncharge
sk_wmem_free_skb
sock_release_ownership
lockdep_sock_is_held
lock_sock
unlock_sock_fast
sock_owned_by_me
sock_owned_by_user
sock_owned_by_user_nocheck
sock_allow_reclassification
sockcm_init
sock_put
sk_receive_skb
sk_tx_queue_set
sk_tx_queue_clear
sk_tx_queue_get
sk_rx_queue_set
sk_rx_queue_clear
sk_rx_queue_get
sk_set_socket
sk_sleep
sock_orphan
sock_graft
sock_net_uid
net_tx_rndhash
sk_set_txhash
sk_rethink_txhash
__sk_dst_get
sk_dst_get
dst_negative_advice
__sk_dst_set
sk_dst_set
__sk_dst_reset
sk_dst_reset
sk_dst_confirm
sock_confirm_neigh
sk_can_gso
sk_nocaps_add
skb_do_copy_data_nocache
skb_add_data_nocache
skb_copy_to_page_nocache
sk_wmem_alloc_get
sk_rmem_alloc_get
sk_has_allocations
skwq_has_sleeper
sock_poll_wait
skb_set_hash_from_sk
skb_set_owner_r
sock_error
sock_wspace
sk_set_bit
sk_clear_bit
sk_wake_async
sk_stream_moderate_sndbuf
sk_page_frag
sock_writeable
gfp_any
sock_rcvtimeo
sock_sndtimeo
sock_rcvlowat
sock_intr_errno
sock_skb_set_dropcount
sk_drops_add
sock_read_timestamp
sock_write_timestamp
sock_recv_timestamp
sock_recv_ts_and_drops
_sock_tx_timestamp
sock_tx_timestamp
skb_setup_tx_timestamp
sk_eat_skb
sock_net
sock_net_set
skb_steal_sock
sk_fullsock
sk_validate_xmit_skb
sk_listener
sk_get_wmem0
sk_get_rmem0
sk_pacing_shift_update
sk_dev_equal_l3scope
   1 /* SPDX-License-Identifier: GPL-2.0-or-later */
   2 /*
   3  * INET         An implementation of the TCP/IP protocol suite for the LINUX
   4  *              operating system.  INET is implemented using the  BSD Socket
   5  *              interface as the means of communication with the user level.
   6  *
   7  *              Definitions for the AF_INET socket handler.
   8  *
   9  * Version:     @(#)sock.h      1.0.4   05/13/93
  10  *
  11  * Authors:     Ross Biro
  12  *              Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  13  *              Corey Minyard <wf-rch!minyard@relay.EU.net>
  14  *              Florian La Roche <flla@stud.uni-sb.de>
  15  *
  16  * Fixes:
  17  *              Alan Cox        :       Volatiles in skbuff pointers. See
  18  *                                      skbuff comments. May be overdone,
  19  *                                      better to prove they can be removed
  20  *                                      than the reverse.
  21  *              Alan Cox        :       Added a zapped field for tcp to note
  22  *                                      a socket is reset and must stay shut up
  23  *              Alan Cox        :       New fields for options
  24  *      Pauline Middelink       :       identd support
  25  *              Alan Cox        :       Eliminate low level recv/recvfrom
  26  *              David S. Miller :       New socket lookup architecture.
  27  *              Steve Whitehouse:       Default routines for sock_ops
  28  *              Arnaldo C. Melo :       removed net_pinfo, tp_pinfo and made
  29  *                                      protinfo be just a void pointer, as the
  30  *                                      protocol specific parts were moved to
  31  *                                      respective headers and ipv4/v6, etc now
  32  *                                      use private slabcaches for its socks
  33  *              Pedro Hortas    :       New flags field for socket options
  34  */
  35 #ifndef _SOCK_H
  36 #define _SOCK_H
  37 
  38 #include <linux/hardirq.h>
  39 #include <linux/kernel.h>
  40 #include <linux/list.h>
  41 #include <linux/list_nulls.h>
  42 #include <linux/timer.h>
  43 #include <linux/cache.h>
  44 #include <linux/bitops.h>
  45 #include <linux/lockdep.h>
  46 #include <linux/netdevice.h>
  47 #include <linux/skbuff.h>       /* struct sk_buff */
  48 #include <linux/mm.h>
  49 #include <linux/security.h>
  50 #include <linux/slab.h>
  51 #include <linux/uaccess.h>
  52 #include <linux/page_counter.h>
  53 #include <linux/memcontrol.h>
  54 #include <linux/static_key.h>
  55 #include <linux/sched.h>
  56 #include <linux/wait.h>
  57 #include <linux/cgroup-defs.h>
  58 #include <linux/rbtree.h>
  59 #include <linux/filter.h>
  60 #include <linux/rculist_nulls.h>
  61 #include <linux/poll.h>
  62 
  63 #include <linux/atomic.h>
  64 #include <linux/refcount.h>
  65 #include <net/dst.h>
  66 #include <net/checksum.h>
  67 #include <net/tcp_states.h>
  68 #include <linux/net_tstamp.h>
  69 #include <net/smc.h>
  70 #include <net/l3mdev.h>
  71 
  72 /*
  73  * This structure really needs to be cleaned up.
  74  * Most of it is for TCP, and not used by any of
  75  * the other protocols.
  76  */
  77 
  78 /* Define this to get the SOCK_DBG debugging facility. */
  79 #define SOCK_DEBUGGING
  80 #ifdef SOCK_DEBUGGING
  81 #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
  82                                         printk(KERN_DEBUG msg); } while (0)
  83 #else
  84 /* Validate arguments and do nothing */
  85 static inline __printf(2, 3)
  86 void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
  87 {
  88 }
  89 #endif
  90 
  91 /* This is the per-socket lock.  The spinlock provides a synchronization
  92  * between user contexts and software interrupt processing, whereas the
  93  * mini-semaphore synchronizes multiple users amongst themselves.
  94  */
  95 typedef struct {
  96         spinlock_t              slock;
  97         int                     owned;
  98         wait_queue_head_t       wq;
  99         /*
 100          * We express the mutex-alike socket_lock semantics
 101          * to the lock validator by explicitly managing
 102          * the slock as a lock variant (in addition to
 103          * the slock itself):
 104          */
 105 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 106         struct lockdep_map dep_map;
 107 #endif
 108 } socket_lock_t;
 109 
 110 struct sock;
 111 struct proto;
 112 struct net;
 113 
 114 typedef __u32 __bitwise __portpair;
 115 typedef __u64 __bitwise __addrpair;
 116 
 117 /**
 118  *      struct sock_common - minimal network layer representation of sockets
 119  *      @skc_daddr: Foreign IPv4 addr
 120  *      @skc_rcv_saddr: Bound local IPv4 addr
 121  *      @skc_hash: hash value used with various protocol lookup tables
 122  *      @skc_u16hashes: two u16 hash values used by UDP lookup tables
 123  *      @skc_dport: placeholder for inet_dport/tw_dport
 124  *      @skc_num: placeholder for inet_num/tw_num
 125  *      @skc_family: network address family
 126  *      @skc_state: Connection state
 127  *      @skc_reuse: %SO_REUSEADDR setting
 128  *      @skc_reuseport: %SO_REUSEPORT setting
 129  *      @skc_bound_dev_if: bound device index if != 0
 130  *      @skc_bind_node: bind hash linkage for various protocol lookup tables
 131  *      @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
 132  *      @skc_prot: protocol handlers inside a network family
 133  *      @skc_net: reference to the network namespace of this socket
 134  *      @skc_node: main hash linkage for various protocol lookup tables
 135  *      @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
 136  *      @skc_tx_queue_mapping: tx queue number for this connection
 137  *      @skc_rx_queue_mapping: rx queue number for this connection
 138  *      @skc_flags: place holder for sk_flags
 139  *              %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
 140  *              %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
 141  *      @skc_incoming_cpu: record/match cpu processing incoming packets
 142  *      @skc_refcnt: reference count
 143  *
 144  *      This is the minimal network layer representation of sockets, the header
 145  *      for struct sock and struct inet_timewait_sock.
 146  */
 147 struct sock_common {
 148         /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
 149          * address on 64bit arches : cf INET_MATCH()
 150          */
 151         union {
 152                 __addrpair      skc_addrpair;
 153                 struct {
 154                         __be32  skc_daddr;
 155                         __be32  skc_rcv_saddr;
 156                 };
 157         };
 158         union  {
 159                 unsigned int    skc_hash;
 160                 __u16           skc_u16hashes[2];
 161         };
 162         /* skc_dport && skc_num must be grouped as well */
 163         union {
 164                 __portpair      skc_portpair;
 165                 struct {
 166                         __be16  skc_dport;
 167                         __u16   skc_num;
 168                 };
 169         };
 170 
 171         unsigned short          skc_family;
 172         volatile unsigned char  skc_state;
 173         unsigned char           skc_reuse:4;
 174         unsigned char           skc_reuseport:1;
 175         unsigned char           skc_ipv6only:1;
 176         unsigned char           skc_net_refcnt:1;
 177         int                     skc_bound_dev_if;
 178         union {
 179                 struct hlist_node       skc_bind_node;
 180                 struct hlist_node       skc_portaddr_node;
 181         };
 182         struct proto            *skc_prot;
 183         possible_net_t          skc_net;
 184 
 185 #if IS_ENABLED(CONFIG_IPV6)
 186         struct in6_addr         skc_v6_daddr;
 187         struct in6_addr         skc_v6_rcv_saddr;
 188 #endif
 189 
 190         atomic64_t              skc_cookie;
 191 
 192         /* following fields are padding to force
 193          * offset(struct sock, sk_refcnt) == 128 on 64bit arches
 194          * assuming IPV6 is enabled. We use this padding differently
 195          * for different kind of 'sockets'
 196          */
 197         union {
 198                 unsigned long   skc_flags;
 199                 struct sock     *skc_listener; /* request_sock */
 200                 struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
 201         };
 202         /*
 203          * fields between dontcopy_begin/dontcopy_end
 204          * are not copied in sock_copy()
 205          */
 206         /* private: */
 207         int                     skc_dontcopy_begin[0];
 208         /* public: */
 209         union {
 210                 struct hlist_node       skc_node;
 211                 struct hlist_nulls_node skc_nulls_node;
 212         };
 213         unsigned short          skc_tx_queue_mapping;
 214 #ifdef CONFIG_XPS
 215         unsigned short          skc_rx_queue_mapping;
 216 #endif
 217         union {
 218                 int             skc_incoming_cpu;
 219                 u32             skc_rcv_wnd;
 220                 u32             skc_tw_rcv_nxt; /* struct tcp_timewait_sock  */
 221         };
 222 
 223         refcount_t              skc_refcnt;
 224         /* private: */
 225         int                     skc_dontcopy_end[0];
 226         union {
 227                 u32             skc_rxhash;
 228                 u32             skc_window_clamp;
 229                 u32             skc_tw_snd_nxt; /* struct tcp_timewait_sock */
 230         };
 231         /* public: */
 232 };
 233 
 234 struct bpf_sk_storage;
 235 
 236 /**
 237   *     struct sock - network layer representation of sockets
 238   *     @__sk_common: shared layout with inet_timewait_sock
 239   *     @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
 240   *     @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
 241   *     @sk_lock:       synchronizer
 242   *     @sk_kern_sock: True if sock is using kernel lock classes
 243   *     @sk_rcvbuf: size of receive buffer in bytes
 244   *     @sk_wq: sock wait queue and async head
 245   *     @sk_rx_dst: receive input route used by early demux
 246   *     @sk_dst_cache: destination cache
 247   *     @sk_dst_pending_confirm: need to confirm neighbour
 248   *     @sk_policy: flow policy
 249   *     @sk_receive_queue: incoming packets
 250   *     @sk_wmem_alloc: transmit queue bytes committed
 251   *     @sk_tsq_flags: TCP Small Queues flags
 252   *     @sk_write_queue: Packet sending queue
 253   *     @sk_omem_alloc: "o" is "option" or "other"
 254   *     @sk_wmem_queued: persistent queue size
 255   *     @sk_forward_alloc: space allocated forward
 256   *     @sk_napi_id: id of the last napi context to receive data for sk
 257   *     @sk_ll_usec: usecs to busypoll when there is no data
 258   *     @sk_allocation: allocation mode
 259   *     @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
 260   *     @sk_pacing_status: Pacing status (requested, handled by sch_fq)
 261   *     @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
 262   *     @sk_sndbuf: size of send buffer in bytes
 263   *     @__sk_flags_offset: empty field used to determine location of bitfield
 264   *     @sk_padding: unused element for alignment
 265   *     @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
 266   *     @sk_no_check_rx: allow zero checksum in RX packets
 267   *     @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
 268   *     @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
 269   *     @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
 270   *     @sk_gso_max_size: Maximum GSO segment size to build
 271   *     @sk_gso_max_segs: Maximum number of GSO segments
 272   *     @sk_pacing_shift: scaling factor for TCP Small Queues
 273   *     @sk_lingertime: %SO_LINGER l_linger setting
 274   *     @sk_backlog: always used with the per-socket spinlock held
 275   *     @sk_callback_lock: used with the callbacks in the end of this struct
 276   *     @sk_error_queue: rarely used
 277   *     @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
 278   *                       IPV6_ADDRFORM for instance)
 279   *     @sk_err: last error
 280   *     @sk_err_soft: errors that don't cause failure but are the cause of a
 281   *                   persistent failure not just 'timed out'
 282   *     @sk_drops: raw/udp drops counter
 283   *     @sk_ack_backlog: current listen backlog
 284   *     @sk_max_ack_backlog: listen backlog set in listen()
 285   *     @sk_uid: user id of owner
 286   *     @sk_priority: %SO_PRIORITY setting
 287   *     @sk_type: socket type (%SOCK_STREAM, etc)
 288   *     @sk_protocol: which protocol this socket belongs in this network family
 289   *     @sk_peer_pid: &struct pid for this socket's peer
 290   *     @sk_peer_cred: %SO_PEERCRED setting
 291   *     @sk_rcvlowat: %SO_RCVLOWAT setting
 292   *     @sk_rcvtimeo: %SO_RCVTIMEO setting
 293   *     @sk_sndtimeo: %SO_SNDTIMEO setting
 294   *     @sk_txhash: computed flow hash for use on transmit
 295   *     @sk_filter: socket filtering instructions
 296   *     @sk_timer: sock cleanup timer
 297   *     @sk_stamp: time stamp of last packet received
 298   *     @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
 299   *     @sk_tsflags: SO_TIMESTAMPING socket options
 300   *     @sk_tskey: counter to disambiguate concurrent tstamp requests
 301   *     @sk_zckey: counter to order MSG_ZEROCOPY notifications
 302   *     @sk_socket: Identd and reporting IO signals
 303   *     @sk_user_data: RPC layer private data
 304   *     @sk_frag: cached page frag
 305   *     @sk_peek_off: current peek_offset value
 306   *     @sk_send_head: front of stuff to transmit
 307   *     @sk_security: used by security modules
 308   *     @sk_mark: generic packet mark
 309   *     @sk_cgrp_data: cgroup data for this cgroup
 310   *     @sk_memcg: this socket's memory cgroup association
 311   *     @sk_write_pending: a write to stream socket waits to start
 312   *     @sk_state_change: callback to indicate change in the state of the sock
 313   *     @sk_data_ready: callback to indicate there is data to be processed
 314   *     @sk_write_space: callback to indicate there is bf sending space available
 315   *     @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
 316   *     @sk_backlog_rcv: callback to process the backlog
 317   *     @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
 318   *     @sk_reuseport_cb: reuseport group container
 319   *     @sk_rcu: used during RCU grace period
 320   *     @sk_clockid: clockid used by time-based scheduling (SO_TXTIME)
 321   *     @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME
 322   *     @sk_txtime_unused: unused txtime flags
 323   */
 324 struct sock {
 325         /*
 326          * Now struct inet_timewait_sock also uses sock_common, so please just
 327          * don't add nothing before this first member (__sk_common) --acme
 328          */
 329         struct sock_common      __sk_common;
 330 #define sk_node                 __sk_common.skc_node
 331 #define sk_nulls_node           __sk_common.skc_nulls_node
 332 #define sk_refcnt               __sk_common.skc_refcnt
 333 #define sk_tx_queue_mapping     __sk_common.skc_tx_queue_mapping
 334 #ifdef CONFIG_XPS
 335 #define sk_rx_queue_mapping     __sk_common.skc_rx_queue_mapping
 336 #endif
 337 
 338 #define sk_dontcopy_begin       __sk_common.skc_dontcopy_begin
 339 #define sk_dontcopy_end         __sk_common.skc_dontcopy_end
 340 #define sk_hash                 __sk_common.skc_hash
 341 #define sk_portpair             __sk_common.skc_portpair
 342 #define sk_num                  __sk_common.skc_num
 343 #define sk_dport                __sk_common.skc_dport
 344 #define sk_addrpair             __sk_common.skc_addrpair
 345 #define sk_daddr                __sk_common.skc_daddr
 346 #define sk_rcv_saddr            __sk_common.skc_rcv_saddr
 347 #define sk_family               __sk_common.skc_family
 348 #define sk_state                __sk_common.skc_state
 349 #define sk_reuse                __sk_common.skc_reuse
 350 #define sk_reuseport            __sk_common.skc_reuseport
 351 #define sk_ipv6only             __sk_common.skc_ipv6only
 352 #define sk_net_refcnt           __sk_common.skc_net_refcnt
 353 #define sk_bound_dev_if         __sk_common.skc_bound_dev_if
 354 #define sk_bind_node            __sk_common.skc_bind_node
 355 #define sk_prot                 __sk_common.skc_prot
 356 #define sk_net                  __sk_common.skc_net
 357 #define sk_v6_daddr             __sk_common.skc_v6_daddr
 358 #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
 359 #define sk_cookie               __sk_common.skc_cookie
 360 #define sk_incoming_cpu         __sk_common.skc_incoming_cpu
 361 #define sk_flags                __sk_common.skc_flags
 362 #define sk_rxhash               __sk_common.skc_rxhash
 363 
 364         socket_lock_t           sk_lock;
 365         atomic_t                sk_drops;
 366         int                     sk_rcvlowat;
 367         struct sk_buff_head     sk_error_queue;
 368         struct sk_buff          *sk_rx_skb_cache;
 369         struct sk_buff_head     sk_receive_queue;
 370         /*
 371          * The backlog queue is special, it is always used with
 372          * the per-socket spinlock held and requires low latency
 373          * access. Therefore we special case it's implementation.
 374          * Note : rmem_alloc is in this structure to fill a hole
 375          * on 64bit arches, not because its logically part of
 376          * backlog.
 377          */
 378         struct {
 379                 atomic_t        rmem_alloc;
 380                 int             len;
 381                 struct sk_buff  *head;
 382                 struct sk_buff  *tail;
 383         } sk_backlog;
 384 #define sk_rmem_alloc sk_backlog.rmem_alloc
 385 
 386         int                     sk_forward_alloc;
 387 #ifdef CONFIG_NET_RX_BUSY_POLL
 388         unsigned int            sk_ll_usec;
 389         /* ===== mostly read cache line ===== */
 390         unsigned int            sk_napi_id;
 391 #endif
 392         int                     sk_rcvbuf;
 393 
 394         struct sk_filter __rcu  *sk_filter;
 395         union {
 396                 struct socket_wq __rcu  *sk_wq;
 397                 struct socket_wq        *sk_wq_raw;
 398         };
 399 #ifdef CONFIG_XFRM
 400         struct xfrm_policy __rcu *sk_policy[2];
 401 #endif
 402         struct dst_entry        *sk_rx_dst;
 403         struct dst_entry __rcu  *sk_dst_cache;
 404         atomic_t                sk_omem_alloc;
 405         int                     sk_sndbuf;
 406 
 407         /* ===== cache line for TX ===== */
 408         int                     sk_wmem_queued;
 409         refcount_t              sk_wmem_alloc;
 410         unsigned long           sk_tsq_flags;
 411         union {
 412                 struct sk_buff  *sk_send_head;
 413                 struct rb_root  tcp_rtx_queue;
 414         };
 415         struct sk_buff          *sk_tx_skb_cache;
 416         struct sk_buff_head     sk_write_queue;
 417         __s32                   sk_peek_off;
 418         int                     sk_write_pending;
 419         __u32                   sk_dst_pending_confirm;
 420         u32                     sk_pacing_status; /* see enum sk_pacing */
 421         long                    sk_sndtimeo;
 422         struct timer_list       sk_timer;
 423         __u32                   sk_priority;
 424         __u32                   sk_mark;
 425         unsigned long           sk_pacing_rate; /* bytes per second */
 426         unsigned long           sk_max_pacing_rate;
 427         struct page_frag        sk_frag;
 428         netdev_features_t       sk_route_caps;
 429         netdev_features_t       sk_route_nocaps;
 430         netdev_features_t       sk_route_forced_caps;
 431         int                     sk_gso_type;
 432         unsigned int            sk_gso_max_size;
 433         gfp_t                   sk_allocation;
 434         __u32                   sk_txhash;
 435 
 436         /*
 437          * Because of non atomicity rules, all
 438          * changes are protected by socket lock.
 439          */
 440         unsigned int            __sk_flags_offset[0];
 441 #ifdef __BIG_ENDIAN_BITFIELD
 442 #define SK_FL_PROTO_SHIFT  16
 443 #define SK_FL_PROTO_MASK   0x00ff0000
 444 
 445 #define SK_FL_TYPE_SHIFT   0
 446 #define SK_FL_TYPE_MASK    0x0000ffff
 447 #else
 448 #define SK_FL_PROTO_SHIFT  8
 449 #define SK_FL_PROTO_MASK   0x0000ff00
 450 
 451 #define SK_FL_TYPE_SHIFT   16
 452 #define SK_FL_TYPE_MASK    0xffff0000
 453 #endif
 454 
 455         unsigned int            sk_padding : 1,
 456                                 sk_kern_sock : 1,
 457                                 sk_no_check_tx : 1,
 458                                 sk_no_check_rx : 1,
 459                                 sk_userlocks : 4,
 460                                 sk_protocol  : 8,
 461                                 sk_type      : 16;
 462 #define SK_PROTOCOL_MAX U8_MAX
 463         u16                     sk_gso_max_segs;
 464         u8                      sk_pacing_shift;
 465         unsigned long           sk_lingertime;
 466         struct proto            *sk_prot_creator;
 467         rwlock_t                sk_callback_lock;
 468         int                     sk_err,
 469                                 sk_err_soft;
 470         u32                     sk_ack_backlog;
 471         u32                     sk_max_ack_backlog;
 472         kuid_t                  sk_uid;
 473         struct pid              *sk_peer_pid;
 474         const struct cred       *sk_peer_cred;
 475         long                    sk_rcvtimeo;
 476         ktime_t                 sk_stamp;
 477 #if BITS_PER_LONG==32
 478         seqlock_t               sk_stamp_seq;
 479 #endif
 480         u16                     sk_tsflags;
 481         u8                      sk_shutdown;
 482         u32                     sk_tskey;
 483         atomic_t                sk_zckey;
 484 
 485         u8                      sk_clockid;
 486         u8                      sk_txtime_deadline_mode : 1,
 487                                 sk_txtime_report_errors : 1,
 488                                 sk_txtime_unused : 6;
 489 
 490         struct socket           *sk_socket;
 491         void                    *sk_user_data;
 492 #ifdef CONFIG_SECURITY
 493         void                    *sk_security;
 494 #endif
 495         struct sock_cgroup_data sk_cgrp_data;
 496         struct mem_cgroup       *sk_memcg;
 497         void                    (*sk_state_change)(struct sock *sk);
 498         void                    (*sk_data_ready)(struct sock *sk);
 499         void                    (*sk_write_space)(struct sock *sk);
 500         void                    (*sk_error_report)(struct sock *sk);
 501         int                     (*sk_backlog_rcv)(struct sock *sk,
 502                                                   struct sk_buff *skb);
 503 #ifdef CONFIG_SOCK_VALIDATE_XMIT
 504         struct sk_buff*         (*sk_validate_xmit_skb)(struct sock *sk,
 505                                                         struct net_device *dev,
 506                                                         struct sk_buff *skb);
 507 #endif
 508         void                    (*sk_destruct)(struct sock *sk);
 509         struct sock_reuseport __rcu     *sk_reuseport_cb;
 510 #ifdef CONFIG_BPF_SYSCALL
 511         struct bpf_sk_storage __rcu     *sk_bpf_storage;
 512 #endif
 513         struct rcu_head         sk_rcu;
 514 };
 515 
 516 enum sk_pacing {
 517         SK_PACING_NONE          = 0,
 518         SK_PACING_NEEDED        = 1,
 519         SK_PACING_FQ            = 2,
 520 };
 521 
 522 #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
 523 
 524 #define rcu_dereference_sk_user_data(sk)        rcu_dereference(__sk_user_data((sk)))
 525 #define rcu_assign_sk_user_data(sk, ptr)        rcu_assign_pointer(__sk_user_data((sk)), ptr)
 526 
 527 /*
 528  * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
 529  * or not whether his port will be reused by someone else. SK_FORCE_REUSE
 530  * on a socket means that the socket will reuse everybody else's port
 531  * without looking at the other's sk_reuse value.
 532  */
 533 
 534 #define SK_NO_REUSE     0
 535 #define SK_CAN_REUSE    1
 536 #define SK_FORCE_REUSE  2
 537 
 538 int sk_set_peek_off(struct sock *sk, int val);
 539 
 540 static inline int sk_peek_offset(struct sock *sk, int flags)
 541 {
 542         if (unlikely(flags & MSG_PEEK)) {
 543                 return READ_ONCE(sk->sk_peek_off);
 544         }
 545 
 546         return 0;
 547 }
 548 
 549 static inline void sk_peek_offset_bwd(struct sock *sk, int val)
 550 {
 551         s32 off = READ_ONCE(sk->sk_peek_off);
 552 
 553         if (unlikely(off >= 0)) {
 554                 off = max_t(s32, off - val, 0);
 555                 WRITE_ONCE(sk->sk_peek_off, off);
 556         }
 557 }
 558 
 559 static inline void sk_peek_offset_fwd(struct sock *sk, int val)
 560 {
 561         sk_peek_offset_bwd(sk, -val);
 562 }
 563 
 564 /*
 565  * Hashed lists helper routines
 566  */
 567 static inline struct sock *sk_entry(const struct hlist_node *node)
 568 {
 569         return hlist_entry(node, struct sock, sk_node);
 570 }
 571 
 572 static inline struct sock *__sk_head(const struct hlist_head *head)
 573 {
 574         return hlist_entry(head->first, struct sock, sk_node);
 575 }
 576 
 577 static inline struct sock *sk_head(const struct hlist_head *head)
 578 {
 579         return hlist_empty(head) ? NULL : __sk_head(head);
 580 }
 581 
 582 static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
 583 {
 584         return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
 585 }
 586 
 587 static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
 588 {
 589         return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
 590 }
 591 
 592 static inline struct sock *sk_next(const struct sock *sk)
 593 {
 594         return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node);
 595 }
 596 
 597 static inline struct sock *sk_nulls_next(const struct sock *sk)
 598 {
 599         return (!is_a_nulls(sk->sk_nulls_node.next)) ?
 600                 hlist_nulls_entry(sk->sk_nulls_node.next,
 601                                   struct sock, sk_nulls_node) :
 602                 NULL;
 603 }
 604 
 605 static inline bool sk_unhashed(const struct sock *sk)
 606 {
 607         return hlist_unhashed(&sk->sk_node);
 608 }
 609 
 610 static inline bool sk_hashed(const struct sock *sk)
 611 {
 612         return !sk_unhashed(sk);
 613 }
 614 
 615 static inline void sk_node_init(struct hlist_node *node)
 616 {
 617         node->pprev = NULL;
 618 }
 619 
 620 static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
 621 {
 622         node->pprev = NULL;
 623 }
 624 
 625 static inline void __sk_del_node(struct sock *sk)
 626 {
 627         __hlist_del(&sk->sk_node);
 628 }
 629 
 630 /* NB: equivalent to hlist_del_init_rcu */
 631 static inline bool __sk_del_node_init(struct sock *sk)
 632 {
 633         if (sk_hashed(sk)) {
 634                 __sk_del_node(sk);
 635                 sk_node_init(&sk->sk_node);
 636                 return true;
 637         }
 638         return false;
 639 }
 640 
 641 /* Grab socket reference count. This operation is valid only
 642    when sk is ALREADY grabbed f.e. it is found in hash table
 643    or a list and the lookup is made under lock preventing hash table
 644    modifications.
 645  */
 646 
 647 static __always_inline void sock_hold(struct sock *sk)
 648 {
 649         refcount_inc(&sk->sk_refcnt);
 650 }
 651 
 652 /* Ungrab socket in the context, which assumes that socket refcnt
 653    cannot hit zero, f.e. it is true in context of any socketcall.
 654  */
 655 static __always_inline void __sock_put(struct sock *sk)
 656 {
 657         refcount_dec(&sk->sk_refcnt);
 658 }
 659 
 660 static inline bool sk_del_node_init(struct sock *sk)
 661 {
 662         bool rc = __sk_del_node_init(sk);
 663 
 664         if (rc) {
 665                 /* paranoid for a while -acme */
 666                 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
 667                 __sock_put(sk);
 668         }
 669         return rc;
 670 }
 671 #define sk_del_node_init_rcu(sk)        sk_del_node_init(sk)
 672 
 673 static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
 674 {
 675         if (sk_hashed(sk)) {
 676                 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
 677                 return true;
 678         }
 679         return false;
 680 }
 681 
 682 static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
 683 {
 684         bool rc = __sk_nulls_del_node_init_rcu(sk);
 685 
 686         if (rc) {
 687                 /* paranoid for a while -acme */
 688                 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
 689                 __sock_put(sk);
 690         }
 691         return rc;
 692 }
 693 
 694 static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
 695 {
 696         hlist_add_head(&sk->sk_node, list);
 697 }
 698 
 699 static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
 700 {
 701         sock_hold(sk);
 702         __sk_add_node(sk, list);
 703 }
 704 
 705 static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
 706 {
 707         sock_hold(sk);
 708         if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
 709             sk->sk_family == AF_INET6)
 710                 hlist_add_tail_rcu(&sk->sk_node, list);
 711         else
 712                 hlist_add_head_rcu(&sk->sk_node, list);
 713 }
 714 
 715 static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list)
 716 {
 717         sock_hold(sk);
 718         hlist_add_tail_rcu(&sk->sk_node, list);
 719 }
 720 
 721 static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
 722 {
 723         hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
 724 }
 725 
 726 static inline void __sk_nulls_add_node_tail_rcu(struct sock *sk, struct hlist_nulls_head *list)
 727 {
 728         hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
 729 }
 730 
 731 static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
 732 {
 733         sock_hold(sk);
 734         __sk_nulls_add_node_rcu(sk, list);
 735 }
 736 
 737 static inline void __sk_del_bind_node(struct sock *sk)
 738 {
 739         __hlist_del(&sk->sk_bind_node);
 740 }
 741 
 742 static inline void sk_add_bind_node(struct sock *sk,
 743                                         struct hlist_head *list)
 744 {
 745         hlist_add_head(&sk->sk_bind_node, list);
 746 }
 747 
 748 #define sk_for_each(__sk, list) \
 749         hlist_for_each_entry(__sk, list, sk_node)
 750 #define sk_for_each_rcu(__sk, list) \
 751         hlist_for_each_entry_rcu(__sk, list, sk_node)
 752 #define sk_nulls_for_each(__sk, node, list) \
 753         hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
 754 #define sk_nulls_for_each_rcu(__sk, node, list) \
 755         hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
 756 #define sk_for_each_from(__sk) \
 757         hlist_for_each_entry_from(__sk, sk_node)
 758 #define sk_nulls_for_each_from(__sk, node) \
 759         if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
 760                 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
 761 #define sk_for_each_safe(__sk, tmp, list) \
 762         hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
 763 #define sk_for_each_bound(__sk, list) \
 764         hlist_for_each_entry(__sk, list, sk_bind_node)
 765 
 766 /**
 767  * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
 768  * @tpos:       the type * to use as a loop cursor.
 769  * @pos:        the &struct hlist_node to use as a loop cursor.
 770  * @head:       the head for your list.
 771  * @offset:     offset of hlist_node within the struct.
 772  *
 773  */
 774 #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset)                  \
 775         for (pos = rcu_dereference(hlist_first_rcu(head));                     \
 776              pos != NULL &&                                                    \
 777                 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;});       \
 778              pos = rcu_dereference(hlist_next_rcu(pos)))
 779 
 780 static inline struct user_namespace *sk_user_ns(struct sock *sk)
 781 {
 782         /* Careful only use this in a context where these parameters
 783          * can not change and must all be valid, such as recvmsg from
 784          * userspace.
 785          */
 786         return sk->sk_socket->file->f_cred->user_ns;
 787 }
 788 
 789 /* Sock flags */
 790 enum sock_flags {
 791         SOCK_DEAD,
 792         SOCK_DONE,
 793         SOCK_URGINLINE,
 794         SOCK_KEEPOPEN,
 795         SOCK_LINGER,
 796         SOCK_DESTROY,
 797         SOCK_BROADCAST,
 798         SOCK_TIMESTAMP,
 799         SOCK_ZAPPED,
 800         SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
 801         SOCK_DBG, /* %SO_DEBUG setting */
 802         SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
 803         SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
 804         SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
 805         SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
 806         SOCK_MEMALLOC, /* VM depends on this socket for swapping */
 807         SOCK_TIMESTAMPING_RX_SOFTWARE,  /* %SOF_TIMESTAMPING_RX_SOFTWARE */
 808         SOCK_FASYNC, /* fasync() active */
 809         SOCK_RXQ_OVFL,
 810         SOCK_ZEROCOPY, /* buffers from userspace */
 811         SOCK_WIFI_STATUS, /* push wifi status to userspace */
 812         SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
 813                      * Will use last 4 bytes of packet sent from
 814                      * user-space instead.
 815                      */
 816         SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
 817         SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
 818         SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
 819         SOCK_TXTIME,
 820         SOCK_XDP, /* XDP is attached */
 821         SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */
 822 };
 823 
 824 #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
 825 
 826 static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
 827 {
 828         nsk->sk_flags = osk->sk_flags;
 829 }
 830 
 831 static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
 832 {
 833         __set_bit(flag, &sk->sk_flags);
 834 }
 835 
 836 static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
 837 {
 838         __clear_bit(flag, &sk->sk_flags);
 839 }
 840 
 841 static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
 842 {
 843         return test_bit(flag, &sk->sk_flags);
 844 }
 845 
 846 #ifdef CONFIG_NET
 847 DECLARE_STATIC_KEY_FALSE(memalloc_socks_key);
 848 static inline int sk_memalloc_socks(void)
 849 {
 850         return static_branch_unlikely(&memalloc_socks_key);
 851 }
 852 #else
 853 
 854 static inline int sk_memalloc_socks(void)
 855 {
 856         return 0;
 857 }
 858 
 859 #endif
 860 
 861 static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
 862 {
 863         return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
 864 }
 865 
 866 static inline void sk_acceptq_removed(struct sock *sk)
 867 {
 868         sk->sk_ack_backlog--;
 869 }
 870 
 871 static inline void sk_acceptq_added(struct sock *sk)
 872 {
 873         sk->sk_ack_backlog++;
 874 }
 875 
 876 static inline bool sk_acceptq_is_full(const struct sock *sk)
 877 {
 878         return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
 879 }
 880 
 881 /*
 882  * Compute minimal free write space needed to queue new packets.
 883  */
 884 static inline int sk_stream_min_wspace(const struct sock *sk)
 885 {
 886         return READ_ONCE(sk->sk_wmem_queued) >> 1;
 887 }
 888 
 889 static inline int sk_stream_wspace(const struct sock *sk)
 890 {
 891         return READ_ONCE(sk->sk_sndbuf) - READ_ONCE(sk->sk_wmem_queued);
 892 }
 893 
 894 static inline void sk_wmem_queued_add(struct sock *sk, int val)
 895 {
 896         WRITE_ONCE(sk->sk_wmem_queued, sk->sk_wmem_queued + val);
 897 }
 898 
 899 void sk_stream_write_space(struct sock *sk);
 900 
 901 /* OOB backlog add */
 902 static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
 903 {
 904         /* dont let skb dst not refcounted, we are going to leave rcu lock */
 905         skb_dst_force(skb);
 906 
 907         if (!sk->sk_backlog.tail)
 908                 sk->sk_backlog.head = skb;
 909         else
 910                 sk->sk_backlog.tail->next = skb;
 911 
 912         sk->sk_backlog.tail = skb;
 913         skb->next = NULL;
 914 }
 915 
 916 /*
 917  * Take into account size of receive queue and backlog queue
 918  * Do not take into account this skb truesize,
 919  * to allow even a single big packet to come.
 920  */
 921 static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
 922 {
 923         unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
 924 
 925         return qsize > limit;
 926 }
 927 
 928 /* The per-socket spinlock must be held here. */
 929 static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
 930                                               unsigned int limit)
 931 {
 932         if (sk_rcvqueues_full(sk, limit))
 933                 return -ENOBUFS;
 934 
 935         /*
 936          * If the skb was allocated from pfmemalloc reserves, only
 937          * allow SOCK_MEMALLOC sockets to use it as this socket is
 938          * helping free memory
 939          */
 940         if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
 941                 return -ENOMEM;
 942 
 943         __sk_add_backlog(sk, skb);
 944         sk->sk_backlog.len += skb->truesize;
 945         return 0;
 946 }
 947 
 948 int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
 949 
 950 static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
 951 {
 952         if (sk_memalloc_socks() && skb_pfmemalloc(skb))
 953                 return __sk_backlog_rcv(sk, skb);
 954 
 955         return sk->sk_backlog_rcv(sk, skb);
 956 }
 957 
 958 static inline void sk_incoming_cpu_update(struct sock *sk)
 959 {
 960         int cpu = raw_smp_processor_id();
 961 
 962         if (unlikely(READ_ONCE(sk->sk_incoming_cpu) != cpu))
 963                 WRITE_ONCE(sk->sk_incoming_cpu, cpu);
 964 }
 965 
 966 static inline void sock_rps_record_flow_hash(__u32 hash)
 967 {
 968 #ifdef CONFIG_RPS
 969         struct rps_sock_flow_table *sock_flow_table;
 970 
 971         rcu_read_lock();
 972         sock_flow_table = rcu_dereference(rps_sock_flow_table);
 973         rps_record_sock_flow(sock_flow_table, hash);
 974         rcu_read_unlock();
 975 #endif
 976 }
 977 
 978 static inline void sock_rps_record_flow(const struct sock *sk)
 979 {
 980 #ifdef CONFIG_RPS
 981         if (static_branch_unlikely(&rfs_needed)) {
 982                 /* Reading sk->sk_rxhash might incur an expensive cache line
 983                  * miss.
 984                  *
 985                  * TCP_ESTABLISHED does cover almost all states where RFS
 986                  * might be useful, and is cheaper [1] than testing :
 987                  *      IPv4: inet_sk(sk)->inet_daddr
 988                  *      IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
 989                  * OR   an additional socket flag
 990                  * [1] : sk_state and sk_prot are in the same cache line.
 991                  */
 992                 if (sk->sk_state == TCP_ESTABLISHED)
 993                         sock_rps_record_flow_hash(sk->sk_rxhash);
 994         }
 995 #endif
 996 }
 997 
 998 static inline void sock_rps_save_rxhash(struct sock *sk,
 999                                         const struct sk_buff *skb)
1000 {
1001 #ifdef CONFIG_RPS
1002         if (unlikely(sk->sk_rxhash != skb->hash))
1003                 sk->sk_rxhash = skb->hash;
1004 #endif
1005 }
1006 
1007 static inline void sock_rps_reset_rxhash(struct sock *sk)
1008 {
1009 #ifdef CONFIG_RPS
1010         sk->sk_rxhash = 0;
1011 #endif
1012 }
1013 
1014 #define sk_wait_event(__sk, __timeo, __condition, __wait)               \
1015         ({      int __rc;                                               \
1016                 release_sock(__sk);                                     \
1017                 __rc = __condition;                                     \
1018                 if (!__rc) {                                            \
1019                         *(__timeo) = wait_woken(__wait,                 \
1020                                                 TASK_INTERRUPTIBLE,     \
1021                                                 *(__timeo));            \
1022                 }                                                       \
1023                 sched_annotate_sleep();                                 \
1024                 lock_sock(__sk);                                        \
1025                 __rc = __condition;                                     \
1026                 __rc;                                                   \
1027         })
1028 
1029 int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
1030 int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
1031 void sk_stream_wait_close(struct sock *sk, long timeo_p);
1032 int sk_stream_error(struct sock *sk, int flags, int err);
1033 void sk_stream_kill_queues(struct sock *sk);
1034 void sk_set_memalloc(struct sock *sk);
1035 void sk_clear_memalloc(struct sock *sk);
1036 
1037 void __sk_flush_backlog(struct sock *sk);
1038 
1039 static inline bool sk_flush_backlog(struct sock *sk)
1040 {
1041         if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
1042                 __sk_flush_backlog(sk);
1043                 return true;
1044         }
1045         return false;
1046 }
1047 
1048 int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
1049 
1050 struct request_sock_ops;
1051 struct timewait_sock_ops;
1052 struct inet_hashinfo;
1053 struct raw_hashinfo;
1054 struct smc_hashinfo;
1055 struct module;
1056 
1057 /*
1058  * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
1059  * un-modified. Special care is taken when initializing object to zero.
1060  */
1061 static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1062 {
1063         if (offsetof(struct sock, sk_node.next) != 0)
1064                 memset(sk, 0, offsetof(struct sock, sk_node.next));
1065         memset(&sk->sk_node.pprev, 0,
1066                size - offsetof(struct sock, sk_node.pprev));
1067 }
1068 
1069 /* Networking protocol blocks we attach to sockets.
1070  * socket layer -> transport layer interface
1071  */
1072 struct proto {
1073         void                    (*close)(struct sock *sk,
1074                                         long timeout);
1075         int                     (*pre_connect)(struct sock *sk,
1076                                         struct sockaddr *uaddr,
1077                                         int addr_len);
1078         int                     (*connect)(struct sock *sk,
1079                                         struct sockaddr *uaddr,
1080                                         int addr_len);
1081         int                     (*disconnect)(struct sock *sk, int flags);
1082 
1083         struct sock *           (*accept)(struct sock *sk, int flags, int *err,
1084                                           bool kern);
1085 
1086         int                     (*ioctl)(struct sock *sk, int cmd,
1087                                          unsigned long arg);
1088         int                     (*init)(struct sock *sk);
1089         void                    (*destroy)(struct sock *sk);
1090         void                    (*shutdown)(struct sock *sk, int how);
1091         int                     (*setsockopt)(struct sock *sk, int level,
1092                                         int optname, char __user *optval,
1093                                         unsigned int optlen);
1094         int                     (*getsockopt)(struct sock *sk, int level,
1095                                         int optname, char __user *optval,
1096                                         int __user *option);
1097         void                    (*keepalive)(struct sock *sk, int valbool);
1098 #ifdef CONFIG_COMPAT
1099         int                     (*compat_setsockopt)(struct sock *sk,
1100                                         int level,
1101                                         int optname, char __user *optval,
1102                                         unsigned int optlen);
1103         int                     (*compat_getsockopt)(struct sock *sk,
1104                                         int level,
1105                                         int optname, char __user *optval,
1106                                         int __user *option);
1107         int                     (*compat_ioctl)(struct sock *sk,
1108                                         unsigned int cmd, unsigned long arg);
1109 #endif
1110         int                     (*sendmsg)(struct sock *sk, struct msghdr *msg,
1111                                            size_t len);
1112         int                     (*recvmsg)(struct sock *sk, struct msghdr *msg,
1113                                            size_t len, int noblock, int flags,
1114                                            int *addr_len);
1115         int                     (*sendpage)(struct sock *sk, struct page *page,
1116                                         int offset, size_t size, int flags);
1117         int                     (*bind)(struct sock *sk,
1118                                         struct sockaddr *uaddr, int addr_len);
1119 
1120         int                     (*backlog_rcv) (struct sock *sk,
1121                                                 struct sk_buff *skb);
1122 
1123         void            (*release_cb)(struct sock *sk);
1124 
1125         /* Keeping track of sk's, looking them up, and port selection methods. */
1126         int                     (*hash)(struct sock *sk);
1127         void                    (*unhash)(struct sock *sk);
1128         void                    (*rehash)(struct sock *sk);
1129         int                     (*get_port)(struct sock *sk, unsigned short snum);
1130 
1131         /* Keeping track of sockets in use */
1132 #ifdef CONFIG_PROC_FS
1133         unsigned int            inuse_idx;
1134 #endif
1135 
1136         bool                    (*stream_memory_free)(const struct sock *sk, int wake);
1137         bool                    (*stream_memory_read)(const struct sock *sk);
1138         /* Memory pressure */
1139         void                    (*enter_memory_pressure)(struct sock *sk);
1140         void                    (*leave_memory_pressure)(struct sock *sk);
1141         atomic_long_t           *memory_allocated;      /* Current allocated memory. */
1142         struct percpu_counter   *sockets_allocated;     /* Current number of sockets. */
1143         /*
1144          * Pressure flag: try to collapse.
1145          * Technical note: it is used by multiple contexts non atomically.
1146          * All the __sk_mem_schedule() is of this nature: accounting
1147          * is strict, actions are advisory and have some latency.
1148          */
1149         unsigned long           *memory_pressure;
1150         long                    *sysctl_mem;
1151 
1152         int                     *sysctl_wmem;
1153         int                     *sysctl_rmem;
1154         u32                     sysctl_wmem_offset;
1155         u32                     sysctl_rmem_offset;
1156 
1157         int                     max_header;
1158         bool                    no_autobind;
1159 
1160         struct kmem_cache       *slab;
1161         unsigned int            obj_size;
1162         slab_flags_t            slab_flags;
1163         unsigned int            useroffset;     /* Usercopy region offset */
1164         unsigned int            usersize;       /* Usercopy region size */
1165 
1166         struct percpu_counter   *orphan_count;
1167 
1168         struct request_sock_ops *rsk_prot;
1169         struct timewait_sock_ops *twsk_prot;
1170 
1171         union {
1172                 struct inet_hashinfo    *hashinfo;
1173                 struct udp_table        *udp_table;
1174                 struct raw_hashinfo     *raw_hash;
1175                 struct smc_hashinfo     *smc_hash;
1176         } h;
1177 
1178         struct module           *owner;
1179 
1180         char                    name[32];
1181 
1182         struct list_head        node;
1183 #ifdef SOCK_REFCNT_DEBUG
1184         atomic_t                socks;
1185 #endif
1186         int                     (*diag_destroy)(struct sock *sk, int err);
1187 } __randomize_layout;
1188 
1189 int proto_register(struct proto *prot, int alloc_slab);
1190 void proto_unregister(struct proto *prot);
1191 int sock_load_diag_module(int family, int protocol);
1192 
1193 #ifdef SOCK_REFCNT_DEBUG
1194 static inline void sk_refcnt_debug_inc(struct sock *sk)
1195 {
1196         atomic_inc(&sk->sk_prot->socks);
1197 }
1198 
1199 static inline void sk_refcnt_debug_dec(struct sock *sk)
1200 {
1201         atomic_dec(&sk->sk_prot->socks);
1202         printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1203                sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1204 }
1205 
1206 static inline void sk_refcnt_debug_release(const struct sock *sk)
1207 {
1208         if (refcount_read(&sk->sk_refcnt) != 1)
1209                 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
1210                        sk->sk_prot->name, sk, refcount_read(&sk->sk_refcnt));
1211 }
1212 #else /* SOCK_REFCNT_DEBUG */
1213 #define sk_refcnt_debug_inc(sk) do { } while (0)
1214 #define sk_refcnt_debug_dec(sk) do { } while (0)
1215 #define sk_refcnt_debug_release(sk) do { } while (0)
1216 #endif /* SOCK_REFCNT_DEBUG */
1217 
1218 static inline bool __sk_stream_memory_free(const struct sock *sk, int wake)
1219 {
1220         if (READ_ONCE(sk->sk_wmem_queued) >= READ_ONCE(sk->sk_sndbuf))
1221                 return false;
1222 
1223         return sk->sk_prot->stream_memory_free ?
1224                 sk->sk_prot->stream_memory_free(sk, wake) : true;
1225 }
1226 
1227 static inline bool sk_stream_memory_free(const struct sock *sk)
1228 {
1229         return __sk_stream_memory_free(sk, 0);
1230 }
1231 
1232 static inline bool __sk_stream_is_writeable(const struct sock *sk, int wake)
1233 {
1234         return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
1235                __sk_stream_memory_free(sk, wake);
1236 }
1237 
1238 static inline bool sk_stream_is_writeable(const struct sock *sk)
1239 {
1240         return __sk_stream_is_writeable(sk, 0);
1241 }
1242 
1243 static inline int sk_under_cgroup_hierarchy(struct sock *sk,
1244                                             struct cgroup *ancestor)
1245 {
1246 #ifdef CONFIG_SOCK_CGROUP_DATA
1247         return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data),
1248                                     ancestor);
1249 #else
1250         return -ENOTSUPP;
1251 #endif
1252 }
1253 
1254 static inline bool sk_has_memory_pressure(const struct sock *sk)
1255 {
1256         return sk->sk_prot->memory_pressure != NULL;
1257 }
1258 
1259 static inline bool sk_under_memory_pressure(const struct sock *sk)
1260 {
1261         if (!sk->sk_prot->memory_pressure)
1262                 return false;
1263 
1264         if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1265             mem_cgroup_under_socket_pressure(sk->sk_memcg))
1266                 return true;
1267 
1268         return !!*sk->sk_prot->memory_pressure;
1269 }
1270 
1271 static inline long
1272 sk_memory_allocated(const struct sock *sk)
1273 {
1274         return atomic_long_read(sk->sk_prot->memory_allocated);
1275 }
1276 
1277 static inline long
1278 sk_memory_allocated_add(struct sock *sk, int amt)
1279 {
1280         return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
1281 }
1282 
1283 static inline void
1284 sk_memory_allocated_sub(struct sock *sk, int amt)
1285 {
1286         atomic_long_sub(amt, sk->sk_prot->memory_allocated);
1287 }
1288 
1289 static inline void sk_sockets_allocated_dec(struct sock *sk)
1290 {
1291         percpu_counter_dec(sk->sk_prot->sockets_allocated);
1292 }
1293 
1294 static inline void sk_sockets_allocated_inc(struct sock *sk)
1295 {
1296         percpu_counter_inc(sk->sk_prot->sockets_allocated);
1297 }
1298 
1299 static inline u64
1300 sk_sockets_allocated_read_positive(struct sock *sk)
1301 {
1302         return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
1303 }
1304 
1305 static inline int
1306 proto_sockets_allocated_sum_positive(struct proto *prot)
1307 {
1308         return percpu_counter_sum_positive(prot->sockets_allocated);
1309 }
1310 
1311 static inline long
1312 proto_memory_allocated(struct proto *prot)
1313 {
1314         return atomic_long_read(prot->memory_allocated);
1315 }
1316 
1317 static inline bool
1318 proto_memory_pressure(struct proto *prot)
1319 {
1320         if (!prot->memory_pressure)
1321                 return false;
1322         return !!*prot->memory_pressure;
1323 }
1324 
1325 
1326 #ifdef CONFIG_PROC_FS
1327 /* Called with local bh disabled */
1328 void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1329 int sock_prot_inuse_get(struct net *net, struct proto *proto);
1330 int sock_inuse_get(struct net *net);
1331 #else
1332 static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
1333                 int inc)
1334 {
1335 }
1336 #endif
1337 
1338 
1339 /* With per-bucket locks this operation is not-atomic, so that
1340  * this version is not worse.
1341  */
1342 static inline int __sk_prot_rehash(struct sock *sk)
1343 {
1344         sk->sk_prot->unhash(sk);
1345         return sk->sk_prot->hash(sk);
1346 }
1347 
1348 /* About 10 seconds */
1349 #define SOCK_DESTROY_TIME (10*HZ)
1350 
1351 /* Sockets 0-1023 can't be bound to unless you are superuser */
1352 #define PROT_SOCK       1024
1353 
1354 #define SHUTDOWN_MASK   3
1355 #define RCV_SHUTDOWN    1
1356 #define SEND_SHUTDOWN   2
1357 
1358 #define SOCK_SNDBUF_LOCK        1
1359 #define SOCK_RCVBUF_LOCK        2
1360 #define SOCK_BINDADDR_LOCK      4
1361 #define SOCK_BINDPORT_LOCK      8
1362 
1363 struct socket_alloc {
1364         struct socket socket;
1365         struct inode vfs_inode;
1366 };
1367 
1368 static inline struct socket *SOCKET_I(struct inode *inode)
1369 {
1370         return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1371 }
1372 
1373 static inline struct inode *SOCK_INODE(struct socket *socket)
1374 {
1375         return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1376 }
1377 
1378 /*
1379  * Functions for memory accounting
1380  */
1381 int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind);
1382 int __sk_mem_schedule(struct sock *sk, int size, int kind);
1383 void __sk_mem_reduce_allocated(struct sock *sk, int amount);
1384 void __sk_mem_reclaim(struct sock *sk, int amount);
1385 
1386 /* We used to have PAGE_SIZE here, but systems with 64KB pages
1387  * do not necessarily have 16x time more memory than 4KB ones.
1388  */
1389 #define SK_MEM_QUANTUM 4096
1390 #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1391 #define SK_MEM_SEND     0
1392 #define SK_MEM_RECV     1
1393 
1394 /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1395 static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1396 {
1397         long val = sk->sk_prot->sysctl_mem[index];
1398 
1399 #if PAGE_SIZE > SK_MEM_QUANTUM
1400         val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT;
1401 #elif PAGE_SIZE < SK_MEM_QUANTUM
1402         val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT;
1403 #endif
1404         return val;
1405 }
1406 
1407 static inline int sk_mem_pages(int amt)
1408 {
1409         return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1410 }
1411 
1412 static inline bool sk_has_account(struct sock *sk)
1413 {
1414         /* return true if protocol supports memory accounting */
1415         return !!sk->sk_prot->memory_allocated;
1416 }
1417 
1418 static inline bool sk_wmem_schedule(struct sock *sk, int size)
1419 {
1420         if (!sk_has_account(sk))
1421                 return true;
1422         return size <= sk->sk_forward_alloc ||
1423                 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1424 }
1425 
1426 static inline bool
1427 sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
1428 {
1429         if (!sk_has_account(sk))
1430                 return true;
1431         return size<= sk->sk_forward_alloc ||
1432                 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1433                 skb_pfmemalloc(skb);
1434 }
1435 
1436 static inline void sk_mem_reclaim(struct sock *sk)
1437 {
1438         if (!sk_has_account(sk))
1439                 return;
1440         if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1441                 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
1442 }
1443 
1444 static inline void sk_mem_reclaim_partial(struct sock *sk)
1445 {
1446         if (!sk_has_account(sk))
1447                 return;
1448         if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1449                 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
1450 }
1451 
1452 static inline void sk_mem_charge(struct sock *sk, int size)
1453 {
1454         if (!sk_has_account(sk))
1455                 return;
1456         sk->sk_forward_alloc -= size;
1457 }
1458 
1459 static inline void sk_mem_uncharge(struct sock *sk, int size)
1460 {
1461         if (!sk_has_account(sk))
1462                 return;
1463         sk->sk_forward_alloc += size;
1464 
1465         /* Avoid a possible overflow.
1466          * TCP send queues can make this happen, if sk_mem_reclaim()
1467          * is not called and more than 2 GBytes are released at once.
1468          *
1469          * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1470          * no need to hold that much forward allocation anyway.
1471          */
1472         if (unlikely(sk->sk_forward_alloc >= 1 << 21))
1473                 __sk_mem_reclaim(sk, 1 << 20);
1474 }
1475 
1476 DECLARE_STATIC_KEY_FALSE(tcp_tx_skb_cache_key);
1477 static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1478 {
1479         sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
1480         sk_wmem_queued_add(sk, -skb->truesize);
1481         sk_mem_uncharge(sk, skb->truesize);
1482         if (static_branch_unlikely(&tcp_tx_skb_cache_key) &&
1483             !sk->sk_tx_skb_cache && !skb_cloned(skb)) {
1484                 skb_zcopy_clear(skb, true);
1485                 sk->sk_tx_skb_cache = skb;
1486                 return;
1487         }
1488         __kfree_skb(skb);
1489 }
1490 
1491 static inline void sock_release_ownership(struct sock *sk)
1492 {
1493         if (sk->sk_lock.owned) {
1494                 sk->sk_lock.owned = 0;
1495 
1496                 /* The sk_lock has mutex_unlock() semantics: */
1497                 mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_);
1498         }
1499 }
1500 
1501 /*
1502  * Macro so as to not evaluate some arguments when
1503  * lockdep is not enabled.
1504  *
1505  * Mark both the sk_lock and the sk_lock.slock as a
1506  * per-address-family lock class.
1507  */
1508 #define sock_lock_init_class_and_name(sk, sname, skey, name, key)       \
1509 do {                                                                    \
1510         sk->sk_lock.owned = 0;                                          \
1511         init_waitqueue_head(&sk->sk_lock.wq);                           \
1512         spin_lock_init(&(sk)->sk_lock.slock);                           \
1513         debug_check_no_locks_freed((void *)&(sk)->sk_lock,              \
1514                         sizeof((sk)->sk_lock));                         \
1515         lockdep_set_class_and_name(&(sk)->sk_lock.slock,                \
1516                                 (skey), (sname));                               \
1517         lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0);     \
1518 } while (0)
1519 
1520 #ifdef CONFIG_LOCKDEP
1521 static inline bool lockdep_sock_is_held(const struct sock *sk)
1522 {
1523         return lockdep_is_held(&sk->sk_lock) ||
1524                lockdep_is_held(&sk->sk_lock.slock);
1525 }
1526 #endif
1527 
1528 void lock_sock_nested(struct sock *sk, int subclass);
1529 
1530 static inline void lock_sock(struct sock *sk)
1531 {
1532         lock_sock_nested(sk, 0);
1533 }
1534 
1535 void __release_sock(struct sock *sk);
1536 void release_sock(struct sock *sk);
1537 
1538 /* BH context may only use the following locking interface. */
1539 #define bh_lock_sock(__sk)      spin_lock(&((__sk)->sk_lock.slock))
1540 #define bh_lock_sock_nested(__sk) \
1541                                 spin_lock_nested(&((__sk)->sk_lock.slock), \
1542                                 SINGLE_DEPTH_NESTING)
1543 #define bh_unlock_sock(__sk)    spin_unlock(&((__sk)->sk_lock.slock))
1544 
1545 bool lock_sock_fast(struct sock *sk);
1546 /**
1547  * unlock_sock_fast - complement of lock_sock_fast
1548  * @sk: socket
1549  * @slow: slow mode
1550  *
1551  * fast unlock socket for user context.
1552  * If slow mode is on, we call regular release_sock()
1553  */
1554 static inline void unlock_sock_fast(struct sock *sk, bool slow)
1555 {
1556         if (slow)
1557                 release_sock(sk);
1558         else
1559                 spin_unlock_bh(&sk->sk_lock.slock);
1560 }
1561 
1562 /* Used by processes to "lock" a socket state, so that
1563  * interrupts and bottom half handlers won't change it
1564  * from under us. It essentially blocks any incoming
1565  * packets, so that we won't get any new data or any
1566  * packets that change the state of the socket.
1567  *
1568  * While locked, BH processing will add new packets to
1569  * the backlog queue.  This queue is processed by the
1570  * owner of the socket lock right before it is released.
1571  *
1572  * Since ~2.3.5 it is also exclusive sleep lock serializing
1573  * accesses from user process context.
1574  */
1575 
1576 static inline void sock_owned_by_me(const struct sock *sk)
1577 {
1578 #ifdef CONFIG_LOCKDEP
1579         WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
1580 #endif
1581 }
1582 
1583 static inline bool sock_owned_by_user(const struct sock *sk)
1584 {
1585         sock_owned_by_me(sk);
1586         return sk->sk_lock.owned;
1587 }
1588 
1589 static inline bool sock_owned_by_user_nocheck(const struct sock *sk)
1590 {
1591         return sk->sk_lock.owned;
1592 }
1593 
1594 /* no reclassification while locks are held */
1595 static inline bool sock_allow_reclassification(const struct sock *csk)
1596 {
1597         struct sock *sk = (struct sock *)csk;
1598 
1599         return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock);
1600 }
1601 
1602 struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
1603                       struct proto *prot, int kern);
1604 void sk_free(struct sock *sk);
1605 void sk_destruct(struct sock *sk);
1606 struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
1607 void sk_free_unlock_clone(struct sock *sk);
1608 
1609 struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1610                              gfp_t priority);
1611 void __sock_wfree(struct sk_buff *skb);
1612 void sock_wfree(struct sk_buff *skb);
1613 struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1614                              gfp_t priority);
1615 void skb_orphan_partial(struct sk_buff *skb);
1616 void sock_rfree(struct sk_buff *skb);
1617 void sock_efree(struct sk_buff *skb);
1618 #ifdef CONFIG_INET
1619 void sock_edemux(struct sk_buff *skb);
1620 #else
1621 #define sock_edemux sock_efree
1622 #endif
1623 
1624 int sock_setsockopt(struct socket *sock, int level, int op,
1625                     char __user *optval, unsigned int optlen);
1626 
1627 int sock_getsockopt(struct socket *sock, int level, int op,
1628                     char __user *optval, int __user *optlen);
1629 int sock_gettstamp(struct socket *sock, void __user *userstamp,
1630                    bool timeval, bool time32);
1631 struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1632                                     int noblock, int *errcode);
1633 struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1634                                      unsigned long data_len, int noblock,
1635                                      int *errcode, int max_page_order);
1636 void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1637 void sock_kfree_s(struct sock *sk, void *mem, int size);
1638 void sock_kzfree_s(struct sock *sk, void *mem, int size);
1639 void sk_send_sigurg(struct sock *sk);
1640 
1641 struct sockcm_cookie {
1642         u64 transmit_time;
1643         u32 mark;
1644         u16 tsflags;
1645 };
1646 
1647 static inline void sockcm_init(struct sockcm_cookie *sockc,
1648                                const struct sock *sk)
1649 {
1650         *sockc = (struct sockcm_cookie) { .tsflags = sk->sk_tsflags };
1651 }
1652 
1653 int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1654                      struct sockcm_cookie *sockc);
1655 int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1656                    struct sockcm_cookie *sockc);
1657 
1658 /*
1659  * Functions to fill in entries in struct proto_ops when a protocol
1660  * does not implement a particular function.
1661  */
1662 int sock_no_bind(struct socket *, struct sockaddr *, int);
1663 int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1664 int sock_no_socketpair(struct socket *, struct socket *);
1665 int sock_no_accept(struct socket *, struct socket *, int, bool);
1666 int sock_no_getname(struct socket *, struct sockaddr *, int);
1667 int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1668 int sock_no_listen(struct socket *, int);
1669 int sock_no_shutdown(struct socket *, int);
1670 int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
1671 int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
1672 int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
1673 int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len);
1674 int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
1675 int sock_no_mmap(struct file *file, struct socket *sock,
1676                  struct vm_area_struct *vma);
1677 ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1678                          size_t size, int flags);
1679 ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
1680                                 int offset, size_t size, int flags);
1681 
1682 /*
1683  * Functions to fill in entries in struct proto_ops when a protocol
1684  * uses the inet style.
1685  */
1686 int sock_common_getsockopt(struct socket *sock, int level, int optname,
1687                                   char __user *optval, int __user *optlen);
1688 int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1689                         int flags);
1690 int sock_common_setsockopt(struct socket *sock, int level, int optname,
1691                                   char __user *optval, unsigned int optlen);
1692 int compat_sock_common_getsockopt(struct socket *sock, int level,
1693                 int optname, char __user *optval, int __user *optlen);
1694 int compat_sock_common_setsockopt(struct socket *sock, int level,
1695                 int optname, char __user *optval, unsigned int optlen);
1696 
1697 void sk_common_release(struct sock *sk);
1698 
1699 /*
1700  *      Default socket callbacks and setup code
1701  */
1702 
1703 /* Initialise core socket variables */
1704 void sock_init_data(struct socket *sock, struct sock *sk);
1705 
1706 /*
1707  * Socket reference counting postulates.
1708  *
1709  * * Each user of socket SHOULD hold a reference count.
1710  * * Each access point to socket (an hash table bucket, reference from a list,
1711  *   running timer, skb in flight MUST hold a reference count.
1712  * * When reference count hits 0, it means it will never increase back.
1713  * * When reference count hits 0, it means that no references from
1714  *   outside exist to this socket and current process on current CPU
1715  *   is last user and may/should destroy this socket.
1716  * * sk_free is called from any context: process, BH, IRQ. When
1717  *   it is called, socket has no references from outside -> sk_free
1718  *   may release descendant resources allocated by the socket, but
1719  *   to the time when it is called, socket is NOT referenced by any
1720  *   hash tables, lists etc.
1721  * * Packets, delivered from outside (from network or from another process)
1722  *   and enqueued on receive/error queues SHOULD NOT grab reference count,
1723  *   when they sit in queue. Otherwise, packets will leak to hole, when
1724  *   socket is looked up by one cpu and unhasing is made by another CPU.
1725  *   It is true for udp/raw, netlink (leak to receive and error queues), tcp
1726  *   (leak to backlog). Packet socket does all the processing inside
1727  *   BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1728  *   use separate SMP lock, so that they are prone too.
1729  */
1730 
1731 /* Ungrab socket and destroy it, if it was the last reference. */
1732 static inline void sock_put(struct sock *sk)
1733 {
1734         if (refcount_dec_and_test(&sk->sk_refcnt))
1735                 sk_free(sk);
1736 }
1737 /* Generic version of sock_put(), dealing with all sockets
1738  * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
1739  */
1740 void sock_gen_put(struct sock *sk);
1741 
1742 int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
1743                      unsigned int trim_cap, bool refcounted);
1744 static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1745                                  const int nested)
1746 {
1747         return __sk_receive_skb(sk, skb, nested, 1, true);
1748 }
1749 
1750 static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1751 {
1752         /* sk_tx_queue_mapping accept only upto a 16-bit value */
1753         if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX))
1754                 return;
1755         sk->sk_tx_queue_mapping = tx_queue;
1756 }
1757 
1758 #define NO_QUEUE_MAPPING        USHRT_MAX
1759 
1760 static inline void sk_tx_queue_clear(struct sock *sk)
1761 {
1762         sk->sk_tx_queue_mapping = NO_QUEUE_MAPPING;
1763 }
1764 
1765 static inline int sk_tx_queue_get(const struct sock *sk)
1766 {
1767         if (sk && sk->sk_tx_queue_mapping != NO_QUEUE_MAPPING)
1768                 return sk->sk_tx_queue_mapping;
1769 
1770         return -1;
1771 }
1772 
1773 static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb)
1774 {
1775 #ifdef CONFIG_XPS
1776         if (skb_rx_queue_recorded(skb)) {
1777                 u16 rx_queue = skb_get_rx_queue(skb);
1778 
1779                 if (WARN_ON_ONCE(rx_queue == NO_QUEUE_MAPPING))
1780                         return;
1781 
1782                 sk->sk_rx_queue_mapping = rx_queue;
1783         }
1784 #endif
1785 }
1786 
1787 static inline void sk_rx_queue_clear(struct sock *sk)
1788 {
1789 #ifdef CONFIG_XPS
1790         sk->sk_rx_queue_mapping = NO_QUEUE_MAPPING;
1791 #endif
1792 }
1793 
1794 #ifdef CONFIG_XPS
1795 static inline int sk_rx_queue_get(const struct sock *sk)
1796 {
1797         if (sk && sk->sk_rx_queue_mapping != NO_QUEUE_MAPPING)
1798                 return sk->sk_rx_queue_mapping;
1799 
1800         return -1;
1801 }
1802 #endif
1803 
1804 static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1805 {
1806         sk_tx_queue_clear(sk);
1807         sk->sk_socket = sock;
1808 }
1809 
1810 static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1811 {
1812         BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1813         return &rcu_dereference_raw(sk->sk_wq)->wait;
1814 }
1815 /* Detach socket from process context.
1816  * Announce socket dead, detach it from wait queue and inode.
1817  * Note that parent inode held reference count on this struct sock,
1818  * we do not release it in this function, because protocol
1819  * probably wants some additional cleanups or even continuing
1820  * to work with this socket (TCP).
1821  */
1822 static inline void sock_orphan(struct sock *sk)
1823 {
1824         write_lock_bh(&sk->sk_callback_lock);
1825         sock_set_flag(sk, SOCK_DEAD);
1826         sk_set_socket(sk, NULL);
1827         sk->sk_wq  = NULL;
1828         write_unlock_bh(&sk->sk_callback_lock);
1829 }
1830 
1831 static inline void sock_graft(struct sock *sk, struct socket *parent)
1832 {
1833         WARN_ON(parent->sk);
1834         write_lock_bh(&sk->sk_callback_lock);
1835         rcu_assign_pointer(sk->sk_wq, &parent->wq);
1836         parent->sk = sk;
1837         sk_set_socket(sk, parent);
1838         sk->sk_uid = SOCK_INODE(parent)->i_uid;
1839         security_sock_graft(sk, parent);
1840         write_unlock_bh(&sk->sk_callback_lock);
1841 }
1842 
1843 kuid_t sock_i_uid(struct sock *sk);
1844 unsigned long sock_i_ino(struct sock *sk);
1845 
1846 static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
1847 {
1848         return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
1849 }
1850 
1851 static inline u32 net_tx_rndhash(void)
1852 {
1853         u32 v = prandom_u32();
1854 
1855         return v ?: 1;
1856 }
1857 
1858 static inline void sk_set_txhash(struct sock *sk)
1859 {
1860         sk->sk_txhash = net_tx_rndhash();
1861 }
1862 
1863 static inline void sk_rethink_txhash(struct sock *sk)
1864 {
1865         if (sk->sk_txhash)
1866                 sk_set_txhash(sk);
1867 }
1868 
1869 static inline struct dst_entry *
1870 __sk_dst_get(struct sock *sk)
1871 {
1872         return rcu_dereference_check(sk->sk_dst_cache,
1873                                      lockdep_sock_is_held(sk));
1874 }
1875 
1876 static inline struct dst_entry *
1877 sk_dst_get(struct sock *sk)
1878 {
1879         struct dst_entry *dst;
1880 
1881         rcu_read_lock();
1882         dst = rcu_dereference(sk->sk_dst_cache);
1883         if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1884                 dst = NULL;
1885         rcu_read_unlock();
1886         return dst;
1887 }
1888 
1889 static inline void dst_negative_advice(struct sock *sk)
1890 {
1891         struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1892 
1893         sk_rethink_txhash(sk);
1894 
1895         if (dst && dst->ops->negative_advice) {
1896                 ndst = dst->ops->negative_advice(dst);
1897 
1898                 if (ndst != dst) {
1899                         rcu_assign_pointer(sk->sk_dst_cache, ndst);
1900                         sk_tx_queue_clear(sk);
1901                         sk->sk_dst_pending_confirm = 0;
1902                 }
1903         }
1904 }
1905 
1906 static inline void
1907 __sk_dst_set(struct sock *sk, struct dst_entry *dst)
1908 {
1909         struct dst_entry *old_dst;
1910 
1911         sk_tx_queue_clear(sk);
1912         sk->sk_dst_pending_confirm = 0;
1913         old_dst = rcu_dereference_protected(sk->sk_dst_cache,
1914                                             lockdep_sock_is_held(sk));
1915         rcu_assign_pointer(sk->sk_dst_cache, dst);
1916         dst_release(old_dst);
1917 }
1918 
1919 static inline void
1920 sk_dst_set(struct sock *sk, struct dst_entry *dst)
1921 {
1922         struct dst_entry *old_dst;
1923 
1924         sk_tx_queue_clear(sk);
1925         sk->sk_dst_pending_confirm = 0;
1926         old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
1927         dst_release(old_dst);
1928 }
1929 
1930 static inline void
1931 __sk_dst_reset(struct sock *sk)
1932 {
1933         __sk_dst_set(sk, NULL);
1934 }
1935 
1936 static inline void
1937 sk_dst_reset(struct sock *sk)
1938 {
1939         sk_dst_set(sk, NULL);
1940 }
1941 
1942 struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1943 
1944 struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1945 
1946 static inline void sk_dst_confirm(struct sock *sk)
1947 {
1948         if (!READ_ONCE(sk->sk_dst_pending_confirm))
1949                 WRITE_ONCE(sk->sk_dst_pending_confirm, 1);
1950 }
1951 
1952 static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
1953 {
1954         if (skb_get_dst_pending_confirm(skb)) {
1955                 struct sock *sk = skb->sk;
1956                 unsigned long now = jiffies;
1957 
1958                 /* avoid dirtying neighbour */
1959                 if (READ_ONCE(n->confirmed) != now)
1960                         WRITE_ONCE(n->confirmed, now);
1961                 if (sk && READ_ONCE(sk->sk_dst_pending_confirm))
1962                         WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
1963         }
1964 }
1965 
1966 bool sk_mc_loop(struct sock *sk);
1967 
1968 static inline bool sk_can_gso(const struct sock *sk)
1969 {
1970         return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
1971 }
1972 
1973 void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
1974 
1975 static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
1976 {
1977         sk->sk_route_nocaps |= flags;
1978         sk->sk_route_caps &= ~flags;
1979 }
1980 
1981 static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
1982                                            struct iov_iter *from, char *to,
1983                                            int copy, int offset)
1984 {
1985         if (skb->ip_summed == CHECKSUM_NONE) {
1986                 __wsum csum = 0;
1987                 if (!csum_and_copy_from_iter_full(to, copy, &csum, from))
1988                         return -EFAULT;
1989                 skb->csum = csum_block_add(skb->csum, csum, offset);
1990         } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
1991                 if (!copy_from_iter_full_nocache(to, copy, from))
1992                         return -EFAULT;
1993         } else if (!copy_from_iter_full(to, copy, from))
1994                 return -EFAULT;
1995 
1996         return 0;
1997 }
1998 
1999 static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
2000                                        struct iov_iter *from, int copy)
2001 {
2002         int err, offset = skb->len;
2003 
2004         err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
2005                                        copy, offset);
2006         if (err)
2007                 __skb_trim(skb, offset);
2008 
2009         return err;
2010 }
2011 
2012 static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
2013                                            struct sk_buff *skb,
2014                                            struct page *page,
2015                                            int off, int copy)
2016 {
2017         int err;
2018 
2019         err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
2020                                        copy, skb->len);
2021         if (err)
2022                 return err;
2023 
2024         skb->len             += copy;
2025         skb->data_len        += copy;
2026         skb->truesize        += copy;
2027         sk_wmem_queued_add(sk, copy);
2028         sk_mem_charge(sk, copy);
2029         return 0;
2030 }
2031 
2032 /**
2033  * sk_wmem_alloc_get - returns write allocations
2034  * @sk: socket
2035  *
2036  * Returns sk_wmem_alloc minus initial offset of one
2037  */
2038 static inline int sk_wmem_alloc_get(const struct sock *sk)
2039 {
2040         return refcount_read(&sk->sk_wmem_alloc) - 1;
2041 }
2042 
2043 /**
2044  * sk_rmem_alloc_get - returns read allocations
2045  * @sk: socket
2046  *
2047  * Returns sk_rmem_alloc
2048  */
2049 static inline int sk_rmem_alloc_get(const struct sock *sk)
2050 {
2051         return atomic_read(&sk->sk_rmem_alloc);
2052 }
2053 
2054 /**
2055  * sk_has_allocations - check if allocations are outstanding
2056  * @sk: socket
2057  *
2058  * Returns true if socket has write or read allocations
2059  */
2060 static inline bool sk_has_allocations(const struct sock *sk)
2061 {
2062         return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
2063 }
2064 
2065 /**
2066  * skwq_has_sleeper - check if there are any waiting processes
2067  * @wq: struct socket_wq
2068  *
2069  * Returns true if socket_wq has waiting processes
2070  *
2071  * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
2072  * barrier call. They were added due to the race found within the tcp code.
2073  *
2074  * Consider following tcp code paths::
2075  *
2076  *   CPU1                CPU2
2077  *   sys_select          receive packet
2078  *   ...                 ...
2079  *   __add_wait_queue    update tp->rcv_nxt
2080  *   ...                 ...
2081  *   tp->rcv_nxt check   sock_def_readable
2082  *   ...                 {
2083  *   schedule               rcu_read_lock();
2084  *                          wq = rcu_dereference(sk->sk_wq);
2085  *                          if (wq && waitqueue_active(&wq->wait))
2086  *                              wake_up_interruptible(&wq->wait)
2087  *                          ...
2088  *                       }
2089  *
2090  * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
2091  * in its cache, and so does the tp->rcv_nxt update on CPU2 side.  The CPU1
2092  * could then endup calling schedule and sleep forever if there are no more
2093  * data on the socket.
2094  *
2095  */
2096 static inline bool skwq_has_sleeper(struct socket_wq *wq)
2097 {
2098         return wq && wq_has_sleeper(&wq->wait);
2099 }
2100 
2101 /**
2102  * sock_poll_wait - place memory barrier behind the poll_wait call.
2103  * @filp:           file
2104  * @sock:           socket to wait on
2105  * @p:              poll_table
2106  *
2107  * See the comments in the wq_has_sleeper function.
2108  */
2109 static inline void sock_poll_wait(struct file *filp, struct socket *sock,
2110                                   poll_table *p)
2111 {
2112         if (!poll_does_not_wait(p)) {
2113                 poll_wait(filp, &sock->wq.wait, p);
2114                 /* We need to be sure we are in sync with the
2115                  * socket flags modification.
2116                  *
2117                  * This memory barrier is paired in the wq_has_sleeper.
2118                  */
2119                 smp_mb();
2120         }
2121 }
2122 
2123 static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
2124 {
2125         if (sk->sk_txhash) {
2126                 skb->l4_hash = 1;
2127                 skb->hash = sk->sk_txhash;
2128         }
2129 }
2130 
2131 void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
2132 
2133 /*
2134  *      Queue a received datagram if it will fit. Stream and sequenced
2135  *      protocols can't normally use this as they need to fit buffers in
2136  *      and play with them.
2137  *
2138  *      Inlined as it's very short and called for pretty much every
2139  *      packet ever received.
2140  */
2141 static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2142 {
2143         skb_orphan(skb);
2144         skb->sk = sk;
2145         skb->destructor = sock_rfree;
2146         atomic_add(skb->truesize, &sk->sk_rmem_alloc);
2147         sk_mem_charge(sk, skb->truesize);
2148 }
2149 
2150 void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2151                     unsigned long expires);
2152 
2153 void sk_stop_timer(struct sock *sk, struct timer_list *timer);
2154 
2155 int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue,
2156                         struct sk_buff *skb, unsigned int flags,
2157                         void (*destructor)(struct sock *sk,
2158                                            struct sk_buff *skb));
2159 int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2160 int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
2161 
2162 int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
2163 struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
2164 
2165 /*
2166  *      Recover an error report and clear atomically
2167  */
2168 
2169 static inline int sock_error(struct sock *sk)
2170 {
2171         int err;
2172         if (likely(!sk->sk_err))
2173                 return 0;
2174         err = xchg(&sk->sk_err, 0);
2175         return -err;
2176 }
2177 
2178 static inline unsigned long sock_wspace(struct sock *sk)
2179 {
2180         int amt = 0;
2181 
2182         if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
2183                 amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc);
2184                 if (amt < 0)
2185                         amt = 0;
2186         }
2187         return amt;
2188 }
2189 
2190 /* Note:
2191  *  We use sk->sk_wq_raw, from contexts knowing this
2192  *  pointer is not NULL and cannot disappear/change.
2193  */
2194 static inline void sk_set_bit(int nr, struct sock *sk)
2195 {
2196         if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2197             !sock_flag(sk, SOCK_FASYNC))
2198                 return;
2199 
2200         set_bit(nr, &sk->sk_wq_raw->flags);
2201 }
2202 
2203 static inline void sk_clear_bit(int nr, struct sock *sk)
2204 {
2205         if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2206             !sock_flag(sk, SOCK_FASYNC))
2207                 return;
2208 
2209         clear_bit(nr, &sk->sk_wq_raw->flags);
2210 }
2211 
2212 static inline void sk_wake_async(const struct sock *sk, int how, int band)
2213 {
2214         if (sock_flag(sk, SOCK_FASYNC)) {
2215                 rcu_read_lock();
2216                 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
2217                 rcu_read_unlock();
2218         }
2219 }
2220 
2221 /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2222  * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2223  * Note: for send buffers, TCP works better if we can build two skbs at
2224  * minimum.
2225  */
2226 #define TCP_SKB_MIN_TRUESIZE    (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
2227 
2228 #define SOCK_MIN_SNDBUF         (TCP_SKB_MIN_TRUESIZE * 2)
2229 #define SOCK_MIN_RCVBUF          TCP_SKB_MIN_TRUESIZE
2230 
2231 static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2232 {
2233         u32 val;
2234 
2235         if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
2236                 return;
2237 
2238         val = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2239 
2240         WRITE_ONCE(sk->sk_sndbuf, max_t(u32, val, SOCK_MIN_SNDBUF));
2241 }
2242 
2243 struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
2244                                     bool force_schedule);
2245 
2246 /**
2247  * sk_page_frag - return an appropriate page_frag
2248  * @sk: socket
2249  *
2250  * Use the per task page_frag instead of the per socket one for
2251  * optimization when we know that we're in the normal context and owns
2252  * everything that's associated with %current.
2253  *
2254  * gfpflags_allow_blocking() isn't enough here as direct reclaim may nest
2255  * inside other socket operations and end up recursing into sk_page_frag()
2256  * while it's already in use.
2257  */
2258 static inline struct page_frag *sk_page_frag(struct sock *sk)
2259 {
2260         if (gfpflags_normal_context(sk->sk_allocation))
2261                 return &current->task_frag;
2262 
2263         return &sk->sk_frag;
2264 }
2265 
2266 bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
2267 
2268 /*
2269  *      Default write policy as shown to user space via poll/select/SIGIO
2270  */
2271 static inline bool sock_writeable(const struct sock *sk)
2272 {
2273         return refcount_read(&sk->sk_wmem_alloc) < (READ_ONCE(sk->sk_sndbuf) >> 1);
2274 }
2275 
2276 static inline gfp_t gfp_any(void)
2277 {
2278         return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
2279 }
2280 
2281 static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
2282 {
2283         return noblock ? 0 : sk->sk_rcvtimeo;
2284 }
2285 
2286 static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
2287 {
2288         return noblock ? 0 : sk->sk_sndtimeo;
2289 }
2290 
2291 static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2292 {
2293         int v = waitall ? len : min_t(int, READ_ONCE(sk->sk_rcvlowat), len);
2294 
2295         return v ?: 1;
2296 }
2297 
2298 /* Alas, with timeout socket operations are not restartable.
2299  * Compare this to poll().
2300  */
2301 static inline int sock_intr_errno(long timeo)
2302 {
2303         return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2304 }
2305 
2306 struct sock_skb_cb {
2307         u32 dropcount;
2308 };
2309 
2310 /* Store sock_skb_cb at the end of skb->cb[] so protocol families
2311  * using skb->cb[] would keep using it directly and utilize its
2312  * alignement guarantee.
2313  */
2314 #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \
2315                             sizeof(struct sock_skb_cb)))
2316 
2317 #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2318                             SOCK_SKB_CB_OFFSET))
2319 
2320 #define sock_skb_cb_check_size(size) \
2321         BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
2322 
2323 static inline void
2324 sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2325 {
2326         SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ?
2327                                                 atomic_read(&sk->sk_drops) : 0;
2328 }
2329 
2330 static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2331 {
2332         int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2333 
2334         atomic_add(segs, &sk->sk_drops);
2335 }
2336 
2337 static inline ktime_t sock_read_timestamp(struct sock *sk)
2338 {
2339 #if BITS_PER_LONG==32
2340         unsigned int seq;
2341         ktime_t kt;
2342 
2343         do {
2344                 seq = read_seqbegin(&sk->sk_stamp_seq);
2345                 kt = sk->sk_stamp;
2346         } while (read_seqretry(&sk->sk_stamp_seq, seq));
2347 
2348         return kt;
2349 #else
2350         return READ_ONCE(sk->sk_stamp);
2351 #endif
2352 }
2353 
2354 static inline void sock_write_timestamp(struct sock *sk, ktime_t kt)
2355 {
2356 #if BITS_PER_LONG==32
2357         write_seqlock(&sk->sk_stamp_seq);
2358         sk->sk_stamp = kt;
2359         write_sequnlock(&sk->sk_stamp_seq);
2360 #else
2361         WRITE_ONCE(sk->sk_stamp, kt);
2362 #endif
2363 }
2364 
2365 void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2366                            struct sk_buff *skb);
2367 void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2368                              struct sk_buff *skb);
2369 
2370 static inline void
2371 sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2372 {
2373         ktime_t kt = skb->tstamp;
2374         struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
2375 
2376         /*
2377          * generate control messages if
2378          * - receive time stamping in software requested
2379          * - software time stamp available and wanted
2380          * - hardware time stamps available and wanted
2381          */
2382         if (sock_flag(sk, SOCK_RCVTSTAMP) ||
2383             (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2384             (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2385             (hwtstamps->hwtstamp &&
2386              (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
2387                 __sock_recv_timestamp(msg, sk, skb);
2388         else
2389                 sock_write_timestamp(sk, kt);
2390 
2391         if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2392                 __sock_recv_wifi_status(msg, sk, skb);
2393 }
2394 
2395 void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2396                               struct sk_buff *skb);
2397 
2398 #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
2399 static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2400                                           struct sk_buff *skb)
2401 {
2402 #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL)                       | \
2403                            (1UL << SOCK_RCVTSTAMP))
2404 #define TSFLAGS_ANY       (SOF_TIMESTAMPING_SOFTWARE                    | \
2405                            SOF_TIMESTAMPING_RAW_HARDWARE)
2406 
2407         if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
2408                 __sock_recv_ts_and_drops(msg, sk, skb);
2409         else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP)))
2410                 sock_write_timestamp(sk, skb->tstamp);
2411         else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP))
2412                 sock_write_timestamp(sk, 0);
2413 }
2414 
2415 void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
2416 
2417 /**
2418  * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
2419  * @sk:         socket sending this packet
2420  * @tsflags:    timestamping flags to use
2421  * @tx_flags:   completed with instructions for time stamping
2422  * @tskey:      filled in with next sk_tskey (not for TCP, which uses seqno)
2423  *
2424  * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
2425  */
2426 static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2427                                       __u8 *tx_flags, __u32 *tskey)
2428 {
2429         if (unlikely(tsflags)) {
2430                 __sock_tx_timestamp(tsflags, tx_flags);
2431                 if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey &&
2432                     tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK)
2433                         *tskey = sk->sk_tskey++;
2434         }
2435         if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2436                 *tx_flags |= SKBTX_WIFI_STATUS;
2437 }
2438 
2439 static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2440                                      __u8 *tx_flags)
2441 {
2442         _sock_tx_timestamp(sk, tsflags, tx_flags, NULL);
2443 }
2444 
2445 static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags)
2446 {
2447         _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags,
2448                            &skb_shinfo(skb)->tskey);
2449 }
2450 
2451 /**
2452  * sk_eat_skb - Release a skb if it is no longer needed
2453  * @sk: socket to eat this skb from
2454  * @skb: socket buffer to eat
2455  *
2456  * This routine must be called with interrupts disabled or with the socket
2457  * locked so that the sk_buff queue operation is ok.
2458 */
2459 DECLARE_STATIC_KEY_FALSE(tcp_rx_skb_cache_key);
2460 static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
2461 {
2462         __skb_unlink(skb, &sk->sk_receive_queue);
2463         if (static_branch_unlikely(&tcp_rx_skb_cache_key) &&
2464             !sk->sk_rx_skb_cache) {
2465                 sk->sk_rx_skb_cache = skb;
2466                 skb_orphan(skb);
2467                 return;
2468         }
2469         __kfree_skb(skb);
2470 }
2471 
2472 static inline
2473 struct net *sock_net(const struct sock *sk)
2474 {
2475         return read_pnet(&sk->sk_net);
2476 }
2477 
2478 static inline
2479 void sock_net_set(struct sock *sk, struct net *net)
2480 {
2481         write_pnet(&sk->sk_net, net);
2482 }
2483 
2484 static inline struct sock *skb_steal_sock(struct sk_buff *skb)
2485 {
2486         if (skb->sk) {
2487                 struct sock *sk = skb->sk;
2488 
2489                 skb->destructor = NULL;
2490                 skb->sk = NULL;
2491                 return sk;
2492         }
2493         return NULL;
2494 }
2495 
2496 /* This helper checks if a socket is a full socket,
2497  * ie _not_ a timewait or request socket.
2498  */
2499 static inline bool sk_fullsock(const struct sock *sk)
2500 {
2501         return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2502 }
2503 
2504 /* Checks if this SKB belongs to an HW offloaded socket
2505  * and whether any SW fallbacks are required based on dev.
2506  * Check decrypted mark in case skb_orphan() cleared socket.
2507  */
2508 static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb,
2509                                                    struct net_device *dev)
2510 {
2511 #ifdef CONFIG_SOCK_VALIDATE_XMIT
2512         struct sock *sk = skb->sk;
2513 
2514         if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb) {
2515                 skb = sk->sk_validate_xmit_skb(sk, dev, skb);
2516 #ifdef CONFIG_TLS_DEVICE
2517         } else if (unlikely(skb->decrypted)) {
2518                 pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n");
2519                 kfree_skb(skb);
2520                 skb = NULL;
2521 #endif
2522         }
2523 #endif
2524 
2525         return skb;
2526 }
2527 
2528 /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2529  * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2530  */
2531 static inline bool sk_listener(const struct sock *sk)
2532 {
2533         return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2534 }
2535 
2536 void sock_enable_timestamp(struct sock *sk, int flag);
2537 int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2538                        int type);
2539 
2540 bool sk_ns_capable(const struct sock *sk,
2541                    struct user_namespace *user_ns, int cap);
2542 bool sk_capable(const struct sock *sk, int cap);
2543 bool sk_net_capable(const struct sock *sk, int cap);
2544 
2545 void sk_get_meminfo(const struct sock *sk, u32 *meminfo);
2546 
2547 /* Take into consideration the size of the struct sk_buff overhead in the
2548  * determination of these values, since that is non-constant across
2549  * platforms.  This makes socket queueing behavior and performance
2550  * not depend upon such differences.
2551  */
2552 #define _SK_MEM_PACKETS         256
2553 #define _SK_MEM_OVERHEAD        SKB_TRUESIZE(256)
2554 #define SK_WMEM_MAX             (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2555 #define SK_RMEM_MAX             (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2556 
2557 extern __u32 sysctl_wmem_max;
2558 extern __u32 sysctl_rmem_max;
2559 
2560 extern int sysctl_tstamp_allow_data;
2561 extern int sysctl_optmem_max;
2562 
2563 extern __u32 sysctl_wmem_default;
2564 extern __u32 sysctl_rmem_default;
2565 
2566 DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key);
2567 
2568 static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto)
2569 {
2570         /* Does this proto have per netns sysctl_wmem ? */
2571         if (proto->sysctl_wmem_offset)
2572                 return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset);
2573 
2574         return *proto->sysctl_wmem;
2575 }
2576 
2577 static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto)
2578 {
2579         /* Does this proto have per netns sysctl_rmem ? */
2580         if (proto->sysctl_rmem_offset)
2581                 return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset);
2582 
2583         return *proto->sysctl_rmem;
2584 }
2585 
2586 /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10)
2587  * Some wifi drivers need to tweak it to get more chunks.
2588  * They can use this helper from their ndo_start_xmit()
2589  */
2590 static inline void sk_pacing_shift_update(struct sock *sk, int val)
2591 {
2592         if (!sk || !sk_fullsock(sk) || READ_ONCE(sk->sk_pacing_shift) == val)
2593                 return;
2594         WRITE_ONCE(sk->sk_pacing_shift, val);
2595 }
2596 
2597 /* if a socket is bound to a device, check that the given device
2598  * index is either the same or that the socket is bound to an L3
2599  * master device and the given device index is also enslaved to
2600  * that L3 master
2601  */
2602 static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
2603 {
2604         int mdif;
2605 
2606         if (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif)
2607                 return true;
2608 
2609         mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif);
2610         if (mdif && mdif == sk->sk_bound_dev_if)
2611                 return true;
2612 
2613         return false;
2614 }
2615 
2616 #endif  /* _SOCK_H */
/* [<][>][^][v][top][bottom][index][help] */
root/include/net/sock.h

INCLUDED FROM

DEFINITIONS
