]> git.proxmox.com Git - mirror_ubuntu-jammy-kernel.git/blame - include/net/sock.h
projects / mirror_ubuntu-jammy-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
octeontx2-pf: Allow VLAN priority also in ntuple filters
[mirror_ubuntu-jammy-kernel.git] / include / net / sock.h
CommitLineData
2874c5fd 1/* SPDX-License-Identifier: GPL-2.0-or-later */
1da177e4
LT		 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 *
02c30a84 11 * Authors: Ross Biro
1da177e4
LT		 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
1da177e4
LT		 34 */
35#ifndef _SOCK_H
36#define _SOCK_H
37
a6b7a407 38#include <linux/hardirq.h>
172589cc 39#include <linux/kernel.h>
1da177e4 40#include <linux/list.h>
88ab1932 41#include <linux/list_nulls.h>
1da177e4
LT		 42#include <linux/timer.h>
43#include <linux/cache.h>
3f134619 44#include <linux/bitops.h>
a5b5bb9a 45#include <linux/lockdep.h>
1da177e4
LT		 46#include <linux/netdevice.h>
47#include <linux/skbuff.h> /* struct sk_buff */
d7fe0f24 48#include <linux/mm.h>
1da177e4 49#include <linux/security.h>
5a0e3ad6 50#include <linux/slab.h>
c6e1a0d1 51#include <linux/uaccess.h>
3e32cb2e 52#include <linux/page_counter.h>
180d8cd9 53#include <linux/memcontrol.h>
c5905afb 54#include <linux/static_key.h>
40401530 55#include <linux/sched.h>
1ce0bf50 56#include <linux/wait.h>
2a56a1fe 57#include <linux/cgroup-defs.h>
75c119af 58#include <linux/rbtree.h>
1da177e4 59#include <linux/filter.h>
88ab1932 60#include <linux/rculist_nulls.h>
a57de0b4 61#include <linux/poll.h>
c8c1bbb6 62#include <linux/sockptr.h>
1c5f2ced 63#include <linux/indirect_call_wrapper.h>
c31504dc 64#include <linux/atomic.h>
41c6d650 65#include <linux/refcount.h>
1da177e4
LT		 66#include <net/dst.h>
67#include <net/checksum.h>
1d0ab253 68#include <net/tcp_states.h>
b9f40e21 69#include <linux/net_tstamp.h>
54dc3e33 70#include <net/l3mdev.h>
04190bf8 71#include <uapi/linux/socket.h>
1da177e4
LT		 72
73/*
74 * This structure really needs to be cleaned up.
75 * Most of it is for TCP, and not used by any of
76 * the other protocols.
77 */
78
79/* Define this to get the SOCK_DBG debugging facility. */
80#define SOCK_DEBUGGING
81#ifdef SOCK_DEBUGGING
82#define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
83 printk(KERN_DEBUG msg); } while (0)
84#else
4cd9029d 85/* Validate arguments and do nothing */
b9075fa9 86static inline __printf(2, 3)
dc6b9b78 87void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
4cd9029d
SH		 88{
89}
1da177e4
LT		 90#endif
91
92/* This is the per-socket lock. The spinlock provides a synchronization
93 * between user contexts and software interrupt processing, whereas the
94 * mini-semaphore synchronizes multiple users amongst themselves.
95 */
1da177e4
LT		 96typedef struct {
97 spinlock_t slock;
d2e9117c 98 int owned;
1da177e4 99 wait_queue_head_t wq;
a5b5bb9a
IM		 100 /*
101 * We express the mutex-alike socket_lock semantics
102 * to the lock validator by explicitly managing
103 * the slock as a lock variant (in addition to
104 * the slock itself):
105 */
106#ifdef CONFIG_DEBUG_LOCK_ALLOC
107 struct lockdep_map dep_map;
108#endif
1da177e4
LT		 109} socket_lock_t;
110
1da177e4 111struct sock;
8feaf0c0 112struct proto;
0eeb8ffc 113struct net;
1da177e4 114
077b393d
ED		 115typedef __u32 __bitwise __portpair;
116typedef __u64 __bitwise __addrpair;
117
1da177e4 118/**
4dc3b16b 119 * struct sock_common - minimal network layer representation of sockets
68835aba
ED		 120 * @skc_daddr: Foreign IPv4 addr
121 * @skc_rcv_saddr: Bound local IPv4 addr
66256e0b 122 * @skc_addrpair: 8-byte-aligned __u64 union of @skc_daddr & @skc_rcv_saddr
4dc6dc71 123 * @skc_hash: hash value used with various protocol lookup tables
d4cada4a 124 * @skc_u16hashes: two u16 hash values used by UDP lookup tables
ce43b03e
ED		 125 * @skc_dport: placeholder for inet_dport/tw_dport
126 * @skc_num: placeholder for inet_num/tw_num
66256e0b 127 * @skc_portpair: __u32 union of @skc_dport & @skc_num
4dc3b16b
PP		 128 * @skc_family: network address family
129 * @skc_state: Connection state
130 * @skc_reuse: %SO_REUSEADDR setting
055dc21a 131 * @skc_reuseport: %SO_REUSEPORT setting
66256e0b
RD		 132 * @skc_ipv6only: socket is IPV6 only
133 * @skc_net_refcnt: socket is using net ref counting
4dc3b16b 134 * @skc_bound_dev_if: bound device index if != 0
4dc3b16b 135 * @skc_bind_node: bind hash linkage for various protocol lookup tables
512615b6 136 * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
8feaf0c0 137 * @skc_prot: protocol handlers inside a network family
07feaebf 138 * @skc_net: reference to the network namespace of this socket
66256e0b
RD		 139 * @skc_v6_daddr: IPV6 destination address
140 * @skc_v6_rcv_saddr: IPV6 source address
141 * @skc_cookie: socket's cookie value
68835aba
ED		 142 * @skc_node: main hash linkage for various protocol lookup tables
143 * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
144 * @skc_tx_queue_mapping: tx queue number for this connection
c6345ce7 145 * @skc_rx_queue_mapping: rx queue number for this connection
8e5eb54d
ED		 146 * @skc_flags: place holder for sk_flags
147 * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
148 * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
66256e0b
RD		 149 * @skc_listener: connection request listener socket (aka rsk_listener)
150 * [union with @skc_flags]
151 * @skc_tw_dr: (aka tw_dr) ptr to &struct inet_timewait_death_row
152 * [union with @skc_flags]
70da268b 153 * @skc_incoming_cpu: record/match cpu processing incoming packets
66256e0b
RD		 154 * @skc_rcv_wnd: (aka rsk_rcv_wnd) TCP receive window size (possibly scaled)
155 * [union with @skc_incoming_cpu]
156 * @skc_tw_rcv_nxt: (aka tw_rcv_nxt) TCP window next expected seq number
157 * [union with @skc_incoming_cpu]
68835aba 158 * @skc_refcnt: reference count
4dc3b16b
PP		 159 *
160 * This is the minimal network layer representation of sockets, the header
8feaf0c0
ACM		 161 * for struct sock and struct inet_timewait_sock.
162 */
1da177e4 163struct sock_common {
ce43b03e 164 /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
05dbc7b5 165 * address on 64bit arches : cf INET_MATCH()
4dc6dc71 166 */
ce43b03e 167 union {
077b393d 168 __addrpair skc_addrpair;
ce43b03e
ED		 169 struct {
170 __be32 skc_daddr;
171 __be32 skc_rcv_saddr;
172 };
173 };
d4cada4a
ED		 174 union {
175 unsigned int skc_hash;
176 __u16 skc_u16hashes[2];
177 };
ce43b03e
ED		 178 /* skc_dport && skc_num must be grouped as well */
179 union {
077b393d 180 __portpair skc_portpair;
ce43b03e
ED		 181 struct {
182 __be16 skc_dport;
183 __u16 skc_num;
184 };
185 };
186
4dc6dc71
ED		 187 unsigned short skc_family;
188 volatile unsigned char skc_state;
055dc21a 189 unsigned char skc_reuse:4;
9fe516ba
ED		 190 unsigned char skc_reuseport:1;
191 unsigned char skc_ipv6only:1;
26abe143 192 unsigned char skc_net_refcnt:1;
4dc6dc71 193 int skc_bound_dev_if;
512615b6
ED		 194 union {
195 struct hlist_node skc_bind_node;
ca065d0c 196 struct hlist_node skc_portaddr_node;
512615b6 197 };
8feaf0c0 198 struct proto *skc_prot;
0c5c9fb5 199 possible_net_t skc_net;
efe4208f
ED		 200
201#if IS_ENABLED(CONFIG_IPV6)
202 struct in6_addr skc_v6_daddr;
203 struct in6_addr skc_v6_rcv_saddr;
204#endif
205
33cf7c90
ED		 206 atomic64_t skc_cookie;
207
8e5eb54d
ED		 208 /* following fields are padding to force
209 * offset(struct sock, sk_refcnt) == 128 on 64bit arches
210 * assuming IPV6 is enabled. We use this padding differently
211 * for different kind of 'sockets'
212 */
213 union {
214 unsigned long skc_flags;
215 struct sock *skc_listener; /* request_sock */
216 struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */
217 };
68835aba
ED		 218 /*
219 * fields between dontcopy_begin/dontcopy_end
220 * are not copied in sock_copy()
221 */
928c41e7 222 /* private: */
68835aba 223 int skc_dontcopy_begin[0];
928c41e7 224 /* public: */
68835aba
ED		 225 union {
226 struct hlist_node skc_node;
227 struct hlist_nulls_node skc_nulls_node;
228 };
755c31cd 229 unsigned short skc_tx_queue_mapping;
4e1beecc 230#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
c6345ce7
AN		 231 unsigned short skc_rx_queue_mapping;
232#endif
ed53d0ab
ED		 233 union {
234 int skc_incoming_cpu;
235 u32 skc_rcv_wnd;
d475f090 236 u32 skc_tw_rcv_nxt; /* struct tcp_timewait_sock */
ed53d0ab 237 };
70da268b 238
41c6d650 239 refcount_t skc_refcnt;
928c41e7 240 /* private: */
68835aba 241 int skc_dontcopy_end[0];
ed53d0ab
ED		 242 union {
243 u32 skc_rxhash;
244 u32 skc_window_clamp;
d475f090 245 u32 skc_tw_snd_nxt; /* struct tcp_timewait_sock */
ed53d0ab 246 };
928c41e7 247 /* public: */
1da177e4
LT		 248};
249
1f00d375 250struct bpf_local_storage;
6ac99e8f 251
1da177e4
LT		 252/**
253 * struct sock - network layer representation of sockets
8feaf0c0 254 * @__sk_common: shared layout with inet_timewait_sock
4dc3b16b
PP		 255 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
256 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
257 * @sk_lock: synchronizer
cdfbabfb 258 * @sk_kern_sock: True if sock is using kernel lock classes
4dc3b16b 259 * @sk_rcvbuf: size of receive buffer in bytes
43815482 260 * @sk_wq: sock wait queue and async head
421b3885 261 * @sk_rx_dst: receive input route used by early demux
4dc3b16b 262 * @sk_dst_cache: destination cache
9b8805a3 263 * @sk_dst_pending_confirm: need to confirm neighbour
4dc3b16b 264 * @sk_policy: flow policy
66256e0b 265 * @sk_rx_skb_cache: cache copy of recently accessed RX skb
4dc3b16b
PP		 266 * @sk_receive_queue: incoming packets
267 * @sk_wmem_alloc: transmit queue bytes committed
771edcaf 268 * @sk_tsq_flags: TCP Small Queues flags
4dc3b16b
PP		 269 * @sk_write_queue: Packet sending queue
270 * @sk_omem_alloc: "o" is "option" or "other"
271 * @sk_wmem_queued: persistent queue size
272 * @sk_forward_alloc: space allocated forward
06021292 273 * @sk_napi_id: id of the last napi context to receive data for sk
dafcc438 274 * @sk_ll_usec: usecs to busypoll when there is no data
4dc3b16b 275 * @sk_allocation: allocation mode
95bd09eb 276 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
218af599 277 * @sk_pacing_status: Pacing status (requested, handled by sch_fq)
c3f40d7c 278 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
4dc3b16b 279 * @sk_sndbuf: size of send buffer in bytes
771edcaf 280 * @__sk_flags_offset: empty field used to determine location of bitfield
293de7de 281 * @sk_padding: unused element for alignment
28448b80
TH		 282 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
283 * @sk_no_check_rx: allow zero checksum in RX packets
4dc3b16b 284 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
a465419b 285 * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
66256e0b
RD		 286 * @sk_route_forced_caps: static, forced route capabilities
287 * (set in tcp_init_sock())
bcd76111 288 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
82cc1a7a 289 * @sk_gso_max_size: Maximum GSO segment size to build
1485348d 290 * @sk_gso_max_segs: Maximum number of GSO segments
3a9b76fd 291 * @sk_pacing_shift: scaling factor for TCP Small Queues
4dc3b16b 292 * @sk_lingertime: %SO_LINGER l_linger setting
4dc3b16b
PP		 293 * @sk_backlog: always used with the per-socket spinlock held
294 * @sk_callback_lock: used with the callbacks in the end of this struct
295 * @sk_error_queue: rarely used
33c732c3
WC		 296 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
297 * IPV6_ADDRFORM for instance)
4dc3b16b 298 * @sk_err: last error
33c732c3
WC		 299 * @sk_err_soft: errors that don't cause failure but are the cause of a
300 * persistent failure not just 'timed out'
cb61cb9b 301 * @sk_drops: raw/udp drops counter
4dc3b16b
PP		 302 * @sk_ack_backlog: current listen backlog
303 * @sk_max_ack_backlog: listen backlog set in listen()
771edcaf 304 * @sk_uid: user id of owner
7fd3253a 305 * @sk_prefer_busy_poll: prefer busypolling over softirq processing
7c951caf 306 * @sk_busy_poll_budget: napi processing budget when busypolling
4dc3b16b
PP		 307 * @sk_priority: %SO_PRIORITY setting
308 * @sk_type: socket type (%SOCK_STREAM, etc)
309 * @sk_protocol: which protocol this socket belongs in this network family
53c3fa20
RD		 310 * @sk_peer_pid: &struct pid for this socket's peer
311 * @sk_peer_cred: %SO_PEERCRED setting
4dc3b16b
PP		 312 * @sk_rcvlowat: %SO_RCVLOWAT setting
313 * @sk_rcvtimeo: %SO_RCVTIMEO setting
314 * @sk_sndtimeo: %SO_SNDTIMEO setting
b73c3d0e 315 * @sk_txhash: computed flow hash for use on transmit
4dc3b16b 316 * @sk_filter: socket filtering instructions
4dc3b16b
PP		 317 * @sk_timer: sock cleanup timer
318 * @sk_stamp: time stamp of last packet received
3a0ed3e9 319 * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
d463126e
YL		 320 * @sk_tsflags: SO_TIMESTAMPING flags
321 * @sk_bind_phc: SO_TIMESTAMPING bind PHC index of PTP virtual clock
322 * for timestamping
09c2d251 323 * @sk_tskey: counter to disambiguate concurrent tstamp requests
52267790 324 * @sk_zckey: counter to order MSG_ZEROCOPY notifications
4dc3b16b
PP		 325 * @sk_socket: Identd and reporting IO signals
326 * @sk_user_data: RPC layer private data
5640f768 327 * @sk_frag: cached page frag
d3d4f0a0 328 * @sk_peek_off: current peek_offset value
4dc3b16b 329 * @sk_send_head: front of stuff to transmit
66256e0b
RD		 330 * @tcp_rtx_queue: TCP re-transmit queue [union with @sk_send_head]
331 * @sk_tx_skb_cache: cache copy of recently accessed TX skb
67be2dd1 332 * @sk_security: used by security modules
31729363 333 * @sk_mark: generic packet mark
2a56a1fe 334 * @sk_cgrp_data: cgroup data for this cgroup
baac50bb 335 * @sk_memcg: this socket's memory cgroup association
4dc3b16b
PP		 336 * @sk_write_pending: a write to stream socket waits to start
337 * @sk_state_change: callback to indicate change in the state of the sock
338 * @sk_data_ready: callback to indicate there is data to be processed
339 * @sk_write_space: callback to indicate there is bf sending space available
340 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
341 * @sk_backlog_rcv: callback to process the backlog
66256e0b 342 * @sk_validate_xmit_skb: ptr to an optional validate function
4dc3b16b 343 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
ef456144 344 * @sk_reuseport_cb: reuseport group container
66256e0b 345 * @sk_bpf_storage: ptr to cache and control for bpf_sk_storage
293de7de 346 * @sk_rcu: used during RCU grace period
80b14dee
RC		 347 * @sk_clockid: clockid used by time-based scheduling (SO_TXTIME)
348 * @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME
66256e0b 349 * @sk_txtime_report_errors: set report errors mode for SO_TXTIME
80b14dee 350 * @sk_txtime_unused: unused txtime flags
293de7de 351 */
1da177e4
LT		 352struct sock {
353 /*
8feaf0c0 354 * Now struct inet_timewait_sock also uses sock_common, so please just
1da177e4
LT		 355 * don't add nothing before this first member (__sk_common) --acme
356 */
357 struct sock_common __sk_common;
4dc6dc71
ED		 358#define sk_node __sk_common.skc_node
359#define sk_nulls_node __sk_common.skc_nulls_node
360#define sk_refcnt __sk_common.skc_refcnt
e022f0b4 361#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
4e1beecc 362#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
c6345ce7
AN		 363#define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping
364#endif
4dc6dc71 365
68835aba
ED		 366#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
367#define sk_dontcopy_end __sk_common.skc_dontcopy_end
4dc6dc71 368#define sk_hash __sk_common.skc_hash
50805466 369#define sk_portpair __sk_common.skc_portpair
05dbc7b5
ED		 370#define sk_num __sk_common.skc_num
371#define sk_dport __sk_common.skc_dport
50805466
ED		 372#define sk_addrpair __sk_common.skc_addrpair
373#define sk_daddr __sk_common.skc_daddr
374#define sk_rcv_saddr __sk_common.skc_rcv_saddr
1da177e4
LT		 375#define sk_family __sk_common.skc_family
376#define sk_state __sk_common.skc_state
377#define sk_reuse __sk_common.skc_reuse
055dc21a 378#define sk_reuseport __sk_common.skc_reuseport
9fe516ba 379#define sk_ipv6only __sk_common.skc_ipv6only
26abe143 380#define sk_net_refcnt __sk_common.skc_net_refcnt
1da177e4 381#define sk_bound_dev_if __sk_common.skc_bound_dev_if
1da177e4 382#define sk_bind_node __sk_common.skc_bind_node
8feaf0c0 383#define sk_prot __sk_common.skc_prot
07feaebf 384#define sk_net __sk_common.skc_net
efe4208f
ED		 385#define sk_v6_daddr __sk_common.skc_v6_daddr
386#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
33cf7c90 387#define sk_cookie __sk_common.skc_cookie
70da268b 388#define sk_incoming_cpu __sk_common.skc_incoming_cpu
8e5eb54d 389#define sk_flags __sk_common.skc_flags
ed53d0ab 390#define sk_rxhash __sk_common.skc_rxhash
efe4208f 391
1da177e4 392 socket_lock_t sk_lock;
9115e8cd
ED		 393 atomic_t sk_drops;
394 int sk_rcvlowat;
395 struct sk_buff_head sk_error_queue;
8b27dae5 396 struct sk_buff *sk_rx_skb_cache;
b178bb3d 397 struct sk_buff_head sk_receive_queue;
fa438ccf
ED		 398 /*
399 * The backlog queue is special, it is always used with
400 * the per-socket spinlock held and requires low latency
401 * access. Therefore we special case it's implementation.
b178bb3d
ED		 402 * Note : rmem_alloc is in this structure to fill a hole
403 * on 64bit arches, not because its logically part of
404 * backlog.
fa438ccf
ED		 405 */
406 struct {
b178bb3d
ED		 407 atomic_t rmem_alloc;
408 int len;
409 struct sk_buff *head;
410 struct sk_buff *tail;
fa438ccf 411 } sk_backlog;
b178bb3d 412#define sk_rmem_alloc sk_backlog.rmem_alloc
2c8c56e1 413
9115e8cd 414 int sk_forward_alloc;
e0d1095a 415#ifdef CONFIG_NET_RX_BUSY_POLL
dafcc438 416 unsigned int sk_ll_usec;
9115e8cd
ED		 417 /* ===== mostly read cache line ===== */
418 unsigned int sk_napi_id;
b178bb3d 419#endif
b178bb3d
ED		 420 int sk_rcvbuf;
421
422 struct sk_filter __rcu *sk_filter;
ceb5d58b
ED		 423 union {
424 struct socket_wq __rcu *sk_wq;
66256e0b 425 /* private: */
ceb5d58b 426 struct socket_wq *sk_wq_raw;
66256e0b 427 /* public: */
ceb5d58b 428 };
def8b4fa 429#ifdef CONFIG_XFRM
d188ba86 430 struct xfrm_policy __rcu *sk_policy[2];
def8b4fa 431#endif
deaa5854 432 struct dst_entry *sk_rx_dst;
0e36cbb3 433 struct dst_entry __rcu *sk_dst_cache;
1da177e4 434 atomic_t sk_omem_alloc;
4e07a91c 435 int sk_sndbuf;
9115e8cd
ED		 436
437 /* ===== cache line for TX ===== */
438 int sk_wmem_queued;
14afee4b 439 refcount_t sk_wmem_alloc;
9115e8cd 440 unsigned long sk_tsq_flags;
75c119af
ED		 441 union {
442 struct sk_buff *sk_send_head;
443 struct rb_root tcp_rtx_queue;
444 };
472c2e07 445 struct sk_buff *sk_tx_skb_cache;
1da177e4 446 struct sk_buff_head sk_write_queue;
9115e8cd
ED		 447 __s32 sk_peek_off;
448 int sk_write_pending;
9b8805a3 449 __u32 sk_dst_pending_confirm;
218af599 450 u32 sk_pacing_status; /* see enum sk_pacing */
9115e8cd
ED		 451 long sk_sndtimeo;
452 struct timer_list sk_timer;
453 __u32 sk_priority;
454 __u32 sk_mark;
76a9ebe8
ED		 455 unsigned long sk_pacing_rate; /* bytes per second */
456 unsigned long sk_max_pacing_rate;
9115e8cd
ED		 457 struct page_frag sk_frag;
458 netdev_features_t sk_route_caps;
459 netdev_features_t sk_route_nocaps;
0a6b2a1d 460 netdev_features_t sk_route_forced_caps;
9115e8cd
ED		 461 int sk_gso_type;
462 unsigned int sk_gso_max_size;
463 gfp_t sk_allocation;
464 __u32 sk_txhash;
fc64869c
AR		 465
466 /*
467 * Because of non atomicity rules, all
468 * changes are protected by socket lock.
469 */
bf976514 470 u8 sk_padding : 1,
cdfbabfb 471 sk_kern_sock : 1,
28448b80
TH		 472 sk_no_check_tx : 1,
473 sk_no_check_rx : 1,
bf976514 474 sk_userlocks : 4;
3a9b76fd 475 u8 sk_pacing_shift;
bf976514
MM		 476 u16 sk_type;
477 u16 sk_protocol;
478 u16 sk_gso_max_segs;
1da177e4 479 unsigned long sk_lingertime;
476e19cf 480 struct proto *sk_prot_creator;
1da177e4
LT		 481 rwlock_t sk_callback_lock;
482 int sk_err,
483 sk_err_soft;
becb74f0
ED		 484 u32 sk_ack_backlog;
485 u32 sk_max_ack_backlog;
86741ec2 486 kuid_t sk_uid;
7fd3253a
BT		 487#ifdef CONFIG_NET_RX_BUSY_POLL
488 u8 sk_prefer_busy_poll;
7c951caf 489 u16 sk_busy_poll_budget;
7fd3253a 490#endif
109f6e39
EB		 491 struct pid *sk_peer_pid;
492 const struct cred *sk_peer_cred;
1da177e4 493 long sk_rcvtimeo;
b7aa0bf7 494 ktime_t sk_stamp;
3a0ed3e9
DD		 495#if BITS_PER_LONG==32
496 seqlock_t sk_stamp_seq;
497#endif
b9f40e21 498 u16 sk_tsflags;
d463126e 499 int sk_bind_phc;
fc64869c 500 u8 sk_shutdown;
09c2d251 501 u32 sk_tskey;
52267790 502 atomic_t sk_zckey;
80b14dee
RC		 503
504 u8 sk_clockid;
505 u8 sk_txtime_deadline_mode : 1,
4b15c707
JSP		 506 sk_txtime_report_errors : 1,
507 sk_txtime_unused : 6;
80b14dee 508
1da177e4
LT		 509 struct socket *sk_socket;
510 void *sk_user_data;
d5f64238 511#ifdef CONFIG_SECURITY
1da177e4 512 void *sk_security;
d5f64238 513#endif
2a56a1fe 514 struct sock_cgroup_data sk_cgrp_data;
baac50bb 515 struct mem_cgroup *sk_memcg;
1da177e4 516 void (*sk_state_change)(struct sock *sk);
676d2369 517 void (*sk_data_ready)(struct sock *sk);
1da177e4
LT		 518 void (*sk_write_space)(struct sock *sk);
519 void (*sk_error_report)(struct sock *sk);
dc6b9b78
ED		 520 int (*sk_backlog_rcv)(struct sock *sk,
521 struct sk_buff *skb);
ebf4e808
IL		 522#ifdef CONFIG_SOCK_VALIDATE_XMIT
523 struct sk_buff* (*sk_validate_xmit_skb)(struct sock *sk,
524 struct net_device *dev,
525 struct sk_buff *skb);
526#endif
1da177e4 527 void (*sk_destruct)(struct sock *sk);
ef456144 528 struct sock_reuseport __rcu *sk_reuseport_cb;
6ac99e8f 529#ifdef CONFIG_BPF_SYSCALL
1f00d375 530 struct bpf_local_storage __rcu *sk_bpf_storage;
6ac99e8f 531#endif
a4298e45 532 struct rcu_head sk_rcu;
1da177e4
LT		 533};
534
218af599
ED		 535enum sk_pacing {
536 SK_PACING_NONE = 0,
537 SK_PACING_NEEDED = 1,
538 SK_PACING_FQ = 2,
539};
540
f1ff5ce2
JS		 541/* Pointer stored in sk_user_data might not be suitable for copying
542 * when cloning the socket. For instance, it can point to a reference
543 * counted object. sk_user_data bottom bit is set if pointer must not
544 * be copied.
545 */
546#define SK_USER_DATA_NOCOPY 1UL
c9a368f1
MKL		 547#define SK_USER_DATA_BPF 2UL /* Managed by BPF */
548#define SK_USER_DATA_PTRMASK ~(SK_USER_DATA_NOCOPY | SK_USER_DATA_BPF)
f1ff5ce2
JS		 549
550/**
551 * sk_user_data_is_nocopy - Test if sk_user_data pointer must not be copied
552 * @sk: socket
553 */
554static inline bool sk_user_data_is_nocopy(const struct sock *sk)
555{
556 return ((uintptr_t)sk->sk_user_data & SK_USER_DATA_NOCOPY);
557}
558
559835ea
PS		 559#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
560
f1ff5ce2
JS		 561#define rcu_dereference_sk_user_data(sk) \
562({ \
563 void *__tmp = rcu_dereference(__sk_user_data((sk))); \
564 (void *)((uintptr_t)__tmp & SK_USER_DATA_PTRMASK); \
565})
566#define rcu_assign_sk_user_data(sk, ptr) \
567({ \
568 uintptr_t __tmp = (uintptr_t)(ptr); \
569 WARN_ON_ONCE(__tmp & ~SK_USER_DATA_PTRMASK); \
570 rcu_assign_pointer(__sk_user_data((sk)), __tmp); \
571})
572#define rcu_assign_sk_user_data_nocopy(sk, ptr) \
573({ \
574 uintptr_t __tmp = (uintptr_t)(ptr); \
575 WARN_ON_ONCE(__tmp & ~SK_USER_DATA_PTRMASK); \
576 rcu_assign_pointer(__sk_user_data((sk)), \
577 __tmp | SK_USER_DATA_NOCOPY); \
578})
559835ea 579
4a17fd52
PE		 580/*
581 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
582 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
583 * on a socket means that the socket will reuse everybody else's port
584 * without looking at the other's sk_reuse value.
585 */
586
587#define SK_NO_REUSE 0
588#define SK_CAN_REUSE 1
589#define SK_FORCE_REUSE 2
590
627d2d6b 591int sk_set_peek_off(struct sock *sk, int val);
592
ef64a54f
PE		 593static inline int sk_peek_offset(struct sock *sk, int flags)
594{
b9bb53f3 595 if (unlikely(flags & MSG_PEEK)) {
a0917e0b 596 return READ_ONCE(sk->sk_peek_off);
b9bb53f3
WB		 597 }
598
599 return 0;
ef64a54f
PE		 600}
601
602static inline void sk_peek_offset_bwd(struct sock *sk, int val)
603{
b9bb53f3
WB		 604 s32 off = READ_ONCE(sk->sk_peek_off);
605
606 if (unlikely(off >= 0)) {
607 off = max_t(s32, off - val, 0);
608 WRITE_ONCE(sk->sk_peek_off, off);
ef64a54f
PE		 609 }
610}
611
612static inline void sk_peek_offset_fwd(struct sock *sk, int val)
613{
b9bb53f3 614 sk_peek_offset_bwd(sk, -val);
ef64a54f
PE		 615}
616
1da177e4
LT		 617/*
618 * Hashed lists helper routines
619 */
c4146644
LZ		 620static inline struct sock *sk_entry(const struct hlist_node *node)
621{
622 return hlist_entry(node, struct sock, sk_node);
623}
624
e48c414e 625static inline struct sock *__sk_head(const struct hlist_head *head)
1da177e4
LT		 626{
627 return hlist_entry(head->first, struct sock, sk_node);
628}
629
e48c414e 630static inline struct sock *sk_head(const struct hlist_head *head)
1da177e4
LT		 631{
632 return hlist_empty(head) ? NULL : __sk_head(head);
633}
634
88ab1932
ED		 635static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
636{
637 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
638}
639
640static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
641{
642 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
643}
644
e48c414e 645static inline struct sock *sk_next(const struct sock *sk)
1da177e4 646{
6c59ebd3 647 return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node);
1da177e4
LT		 648}
649
88ab1932
ED		 650static inline struct sock *sk_nulls_next(const struct sock *sk)
651{
652 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
653 hlist_nulls_entry(sk->sk_nulls_node.next,
654 struct sock, sk_nulls_node) :
655 NULL;
656}
657
dc6b9b78 658static inline bool sk_unhashed(const struct sock *sk)
1da177e4
LT		 659{
660 return hlist_unhashed(&sk->sk_node);
661}
662
dc6b9b78 663static inline bool sk_hashed(const struct sock *sk)
1da177e4 664{
da753bea 665 return !sk_unhashed(sk);
1da177e4
LT		 666}
667
dc6b9b78 668static inline void sk_node_init(struct hlist_node *node)
1da177e4
LT		 669{
670 node->pprev = NULL;
671}
672
dc6b9b78 673static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
88ab1932
ED		 674{
675 node->pprev = NULL;
676}
677
dc6b9b78 678static inline void __sk_del_node(struct sock *sk)
1da177e4
LT		 679{
680 __hlist_del(&sk->sk_node);
681}
682
808f5114 683/* NB: equivalent to hlist_del_init_rcu */
dc6b9b78 684static inline bool __sk_del_node_init(struct sock *sk)
1da177e4
LT		 685{
686 if (sk_hashed(sk)) {
687 __sk_del_node(sk);
688 sk_node_init(&sk->sk_node);
dc6b9b78 689 return true;
1da177e4 690 }
dc6b9b78 691 return false;
1da177e4
LT		 692}
693
694/* Grab socket reference count. This operation is valid only
695 when sk is ALREADY grabbed f.e. it is found in hash table
696 or a list and the lookup is made under lock preventing hash table
697 modifications.
698 */
699
f9a7cbbf 700static __always_inline void sock_hold(struct sock *sk)
1da177e4 701{
41c6d650 702 refcount_inc(&sk->sk_refcnt);
1da177e4
LT		 703}
704
705/* Ungrab socket in the context, which assumes that socket refcnt
706 cannot hit zero, f.e. it is true in context of any socketcall.
707 */
f9a7cbbf 708static __always_inline void __sock_put(struct sock *sk)
1da177e4 709{
41c6d650 710 refcount_dec(&sk->sk_refcnt);
1da177e4
LT		 711}
712
dc6b9b78 713static inline bool sk_del_node_init(struct sock *sk)
1da177e4 714{
dc6b9b78 715 bool rc = __sk_del_node_init(sk);
1da177e4
LT		 716
717 if (rc) {
718 /* paranoid for a while -acme */
41c6d650 719 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
1da177e4
LT		 720 __sock_put(sk);
721 }
722 return rc;
723}
808f5114 724#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
1da177e4 725
dc6b9b78 726static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7
ED		 727{
728 if (sk_hashed(sk)) {
88ab1932 729 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
dc6b9b78 730 return true;
271b72c7 731 }
dc6b9b78 732 return false;
271b72c7
ED		 733}
734
dc6b9b78 735static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7 736{
dc6b9b78 737 bool rc = __sk_nulls_del_node_init_rcu(sk);
271b72c7
ED		 738
739 if (rc) {
740 /* paranoid for a while -acme */
41c6d650 741 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
271b72c7
ED		 742 __sock_put(sk);
743 }
744 return rc;
745}
746
dc6b9b78 747static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT		 748{
749 hlist_add_head(&sk->sk_node, list);
750}
751
dc6b9b78 752static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT		 753{
754 sock_hold(sk);
755 __sk_add_node(sk, list);
756}
757
dc6b9b78 758static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
808f5114 759{
760 sock_hold(sk);
d296ba60
CG		 761 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
762 sk->sk_family == AF_INET6)
763 hlist_add_tail_rcu(&sk->sk_node, list);
764 else
765 hlist_add_head_rcu(&sk->sk_node, list);
808f5114 766}
767
a4dc6a49
MC		 768static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list)
769{
770 sock_hold(sk);
771 hlist_add_tail_rcu(&sk->sk_node, list);
772}
773
dc6b9b78 774static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7 775{
d7efc6c1 776 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
271b72c7
ED		 777}
778
8dbd76e7
ED		 779static inline void __sk_nulls_add_node_tail_rcu(struct sock *sk, struct hlist_nulls_head *list)
780{
781 hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
782}
783
dc6b9b78 784static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7
ED		 785{
786 sock_hold(sk);
88ab1932 787 __sk_nulls_add_node_rcu(sk, list);
271b72c7
ED		 788}
789
dc6b9b78 790static inline void __sk_del_bind_node(struct sock *sk)
1da177e4
LT		 791{
792 __hlist_del(&sk->sk_bind_node);
793}
794
dc6b9b78 795static inline void sk_add_bind_node(struct sock *sk,
1da177e4
LT		 796 struct hlist_head *list)
797{
798 hlist_add_head(&sk->sk_bind_node, list);
799}
800
b67bfe0d
SL		 801#define sk_for_each(__sk, list) \
802 hlist_for_each_entry(__sk, list, sk_node)
803#define sk_for_each_rcu(__sk, list) \
804 hlist_for_each_entry_rcu(__sk, list, sk_node)
88ab1932
ED		 805#define sk_nulls_for_each(__sk, node, list) \
806 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
807#define sk_nulls_for_each_rcu(__sk, node, list) \
808 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
b67bfe0d
SL		 809#define sk_for_each_from(__sk) \
810 hlist_for_each_entry_from(__sk, sk_node)
88ab1932
ED		 811#define sk_nulls_for_each_from(__sk, node) \
812 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
813 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
b67bfe0d
SL		 814#define sk_for_each_safe(__sk, tmp, list) \
815 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
816#define sk_for_each_bound(__sk, list) \
817 hlist_for_each_entry(__sk, list, sk_bind_node)
1da177e4 818
2dc41cff 819/**
ca065d0c 820 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
2dc41cff
DH		 821 * @tpos: the type * to use as a loop cursor.
822 * @pos: the &struct hlist_node to use as a loop cursor.
823 * @head: the head for your list.
824 * @offset: offset of hlist_node within the struct.
825 *
826 */
ca065d0c 827#define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
b6f4f848 828 for (pos = rcu_dereference(hlist_first_rcu(head)); \
ca065d0c 829 pos != NULL && \
2dc41cff 830 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
b6f4f848 831 pos = rcu_dereference(hlist_next_rcu(pos)))
2dc41cff 832
c336d148
EB		 833static inline struct user_namespace *sk_user_ns(struct sock *sk)
834{
835 /* Careful only use this in a context where these parameters
836 * can not change and must all be valid, such as recvmsg from
837 * userspace.
838 */
839 return sk->sk_socket->file->f_cred->user_ns;
840}
841
1da177e4
LT		 842/* Sock flags */
843enum sock_flags {
844 SOCK_DEAD,
845 SOCK_DONE,
846 SOCK_URGINLINE,
847 SOCK_KEEPOPEN,
848 SOCK_LINGER,
849 SOCK_DESTROY,
850 SOCK_BROADCAST,
851 SOCK_TIMESTAMP,
852 SOCK_ZAPPED,
853 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
854 SOCK_DBG, /* %SO_DEBUG setting */
855 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
92f37fd2 856 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
1da177e4 857 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
7cb02404 858 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
20d49473 859 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
bcdce719 860 SOCK_FASYNC, /* fasync() active */
3b885787 861 SOCK_RXQ_OVFL,
1cdebb42 862 SOCK_ZEROCOPY, /* buffers from userspace */
6e3e939f 863 SOCK_WIFI_STATUS, /* push wifi status to userspace */
3bdc0eba
BG		 864 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
865 * Will use last 4 bytes of packet sent from
866 * user-space instead.
867 */
d59577b6 868 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
7d4c04fc 869 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
a4298e45 870 SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
80b14dee 871 SOCK_TXTIME,
e4a2a304 872 SOCK_XDP, /* XDP is attached */
887feae3 873 SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */
1da177e4
LT		 874};
875
01ce63c9
MRL		 876#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
877
53b924b3
RB		 878static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
879{
880 nsk->sk_flags = osk->sk_flags;
881}
882
1da177e4
LT		 883static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
884{
885 __set_bit(flag, &sk->sk_flags);
886}
887
888static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
889{
890 __clear_bit(flag, &sk->sk_flags);
891}
892
dfde1d7d
DY		 893static inline void sock_valbool_flag(struct sock *sk, enum sock_flags bit,
894 int valbool)
895{
896 if (valbool)
897 sock_set_flag(sk, bit);
898 else
899 sock_reset_flag(sk, bit);
900}
901
1b23a5df 902static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
1da177e4
LT		 903{
904 return test_bit(flag, &sk->sk_flags);
905}
906
c93bdd0e 907#ifdef CONFIG_NET
a7950ae8 908DECLARE_STATIC_KEY_FALSE(memalloc_socks_key);
c93bdd0e
MG		 909static inline int sk_memalloc_socks(void)
910{
a7950ae8 911 return static_branch_unlikely(&memalloc_socks_key);
c93bdd0e 912}
d9539752
KC		 913
914void __receive_sock(struct file *file);
c93bdd0e
MG		 915#else
916
917static inline int sk_memalloc_socks(void)
918{
919 return 0;
920}
921
d9539752
KC		 922static inline void __receive_sock(struct file *file)
923{ }
c93bdd0e
MG		 924#endif
925
7450aaf6 926static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
99a1dec7 927{
7450aaf6 928 return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
99a1dec7
MG		 929}
930
1da177e4
LT		 931static inline void sk_acceptq_removed(struct sock *sk)
932{
288efe86 933 WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog - 1);
1da177e4
LT		 934}
935
936static inline void sk_acceptq_added(struct sock *sk)
937{
288efe86 938 WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog + 1);
1da177e4
LT		 939}
940
c609e6aa
ED		 941/* Note: If you think the test should be:
942 * return READ_ONCE(sk->sk_ack_backlog) >= READ_ONCE(sk->sk_max_ack_backlog);
943 * Then please take a look at commit 64a146513f8f ("[NET]: Revert incorrect accept queue backlog changes.")
944 */
dc6b9b78 945static inline bool sk_acceptq_is_full(const struct sock *sk)
1da177e4 946{
c609e6aa 947 return READ_ONCE(sk->sk_ack_backlog) > READ_ONCE(sk->sk_max_ack_backlog);
1da177e4
LT		 948}
949
950/*
951 * Compute minimal free write space needed to queue new packets.
952 */
dc6b9b78 953static inline int sk_stream_min_wspace(const struct sock *sk)
1da177e4 954{
ab4e846a 955 return READ_ONCE(sk->sk_wmem_queued) >> 1;
1da177e4
LT		 956}
957
dc6b9b78 958static inline int sk_stream_wspace(const struct sock *sk)
1da177e4 959{
ab4e846a
ED		 960 return READ_ONCE(sk->sk_sndbuf) - READ_ONCE(sk->sk_wmem_queued);
961}
962
963static inline void sk_wmem_queued_add(struct sock *sk, int val)
964{
965 WRITE_ONCE(sk->sk_wmem_queued, sk->sk_wmem_queued + val);
1da177e4
LT		 966}
967
69336bd2 968void sk_stream_write_space(struct sock *sk);
1da177e4 969
8eae939f 970/* OOB backlog add */
a3a858ff 971static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
9ee6b535 972{
7fee226a 973 /* dont let skb dst not refcounted, we are going to leave rcu lock */
222d7dbd 974 skb_dst_force(skb);
7fee226a
ED		 975
976 if (!sk->sk_backlog.tail)
9ed498c6 977 WRITE_ONCE(sk->sk_backlog.head, skb);
7fee226a 978 else
9ee6b535 979 sk->sk_backlog.tail->next = skb;
7fee226a 980
9ed498c6 981 WRITE_ONCE(sk->sk_backlog.tail, skb);
9ee6b535
SH		 982 skb->next = NULL;
983}
1da177e4 984
c377411f
ED		 985/*
986 * Take into account size of receive queue and backlog queue
0fd7bac6
ED		 987 * Do not take into account this skb truesize,
988 * to allow even a single big packet to come.
c377411f 989 */
274f482d 990static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
c377411f
ED		 991{
992 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
993
f545a38f 994 return qsize > limit;
c377411f
ED		 995}
996
8eae939f 997/* The per-socket spinlock must be held here. */
f545a38f
ED		 998static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
999 unsigned int limit)
8eae939f 1000{
274f482d 1001 if (sk_rcvqueues_full(sk, limit))
8eae939f
ZY		 1002 return -ENOBUFS;
1003
c7c49b8f
ED		 1004 /*
1005 * If the skb was allocated from pfmemalloc reserves, only
1006 * allow SOCK_MEMALLOC sockets to use it as this socket is
1007 * helping free memory
1008 */
1009 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
1010 return -ENOMEM;
1011
a3a858ff 1012 __sk_add_backlog(sk, skb);
8eae939f
ZY		 1013 sk->sk_backlog.len += skb->truesize;
1014 return 0;
1015}
1016
69336bd2 1017int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
b4b9e355 1018
c57943a1
PZ		 1019static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1020{
b4b9e355
MG		 1021 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
1022 return __sk_backlog_rcv(sk, skb);
1023
c57943a1
PZ		 1024 return sk->sk_backlog_rcv(sk, skb);
1025}
1026
2c8c56e1
ED		 1027static inline void sk_incoming_cpu_update(struct sock *sk)
1028{
34cfb542
PA		 1029 int cpu = raw_smp_processor_id();
1030
7170a977
ED		 1031 if (unlikely(READ_ONCE(sk->sk_incoming_cpu) != cpu))
1032 WRITE_ONCE(sk->sk_incoming_cpu, cpu);
2c8c56e1
ED		 1033}
1034
fe477558 1035static inline void sock_rps_record_flow_hash(__u32 hash)
c58dc01b
DM		 1036{
1037#ifdef CONFIG_RPS
1038 struct rps_sock_flow_table *sock_flow_table;
1039
1040 rcu_read_lock();
1041 sock_flow_table = rcu_dereference(rps_sock_flow_table);
fe477558 1042 rps_record_sock_flow(sock_flow_table, hash);
c58dc01b
DM		 1043 rcu_read_unlock();
1044#endif
1045}
1046
fe477558
TH		 1047static inline void sock_rps_record_flow(const struct sock *sk)
1048{
c9d8ca04 1049#ifdef CONFIG_RPS
dc05360f 1050 if (static_branch_unlikely(&rfs_needed)) {
13bfff25
ED		 1051 /* Reading sk->sk_rxhash might incur an expensive cache line
1052 * miss.
1053 *
1054 * TCP_ESTABLISHED does cover almost all states where RFS
1055 * might be useful, and is cheaper [1] than testing :
1056 * IPv4: inet_sk(sk)->inet_daddr
1057 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
1058 * OR an additional socket flag
1059 * [1] : sk_state and sk_prot are in the same cache line.
1060 */
1061 if (sk->sk_state == TCP_ESTABLISHED)
1062 sock_rps_record_flow_hash(sk->sk_rxhash);
1063 }
c9d8ca04 1064#endif
fe477558
TH		 1065}
1066
bdeab991
TH		 1067static inline void sock_rps_save_rxhash(struct sock *sk,
1068 const struct sk_buff *skb)
c58dc01b
DM		 1069{
1070#ifdef CONFIG_RPS
567e4b79 1071 if (unlikely(sk->sk_rxhash != skb->hash))
61b905da 1072 sk->sk_rxhash = skb->hash;
c58dc01b
DM		 1073#endif
1074}
1075
bdeab991
TH		 1076static inline void sock_rps_reset_rxhash(struct sock *sk)
1077{
1078#ifdef CONFIG_RPS
bdeab991
TH		 1079 sk->sk_rxhash = 0;
1080#endif
1081}
1082
d9dc8b0f 1083#define sk_wait_event(__sk, __timeo, __condition, __wait) \
cfcabdcc
SH		 1084 ({ int __rc; \
1085 release_sock(__sk); \
1086 __rc = __condition; \
1087 if (!__rc) { \
d9dc8b0f
WC		 1088 *(__timeo) = wait_woken(__wait, \
1089 TASK_INTERRUPTIBLE, \
1090 *(__timeo)); \
cfcabdcc 1091 } \
d9dc8b0f 1092 sched_annotate_sleep(); \
cfcabdcc
SH		 1093 lock_sock(__sk); \
1094 __rc = __condition; \
1095 __rc; \
1096 })
1da177e4 1097
69336bd2
JP		 1098int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
1099int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
1100void sk_stream_wait_close(struct sock *sk, long timeo_p);
1101int sk_stream_error(struct sock *sk, int flags, int err);
1102void sk_stream_kill_queues(struct sock *sk);
1103void sk_set_memalloc(struct sock *sk);
1104void sk_clear_memalloc(struct sock *sk);
1da177e4 1105
d41a69f1
ED		 1106void __sk_flush_backlog(struct sock *sk);
1107
1108static inline bool sk_flush_backlog(struct sock *sk)
1109{
1110 if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
1111 __sk_flush_backlog(sk);
1112 return true;
1113 }
1114 return false;
1115}
1116
dfbafc99 1117int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
1da177e4 1118
60236fdd 1119struct request_sock_ops;
6d6ee43e 1120struct timewait_sock_ops;
ab1e0a13 1121struct inet_hashinfo;
fc8717ba 1122struct raw_hashinfo;
f16a7dd5 1123struct smc_hashinfo;
de477254 1124struct module;
51e0158a 1125struct sk_psock;
2e6599cb 1126
f77d6021 1127/*
5f0d5a3a 1128 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
f77d6021
ED		 1129 * un-modified. Special care is taken when initializing object to zero.
1130 */
1131static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1132{
1133 if (offsetof(struct sock, sk_node.next) != 0)
1134 memset(sk, 0, offsetof(struct sock, sk_node.next));
1135 memset(&sk->sk_node.pprev, 0,
1136 size - offsetof(struct sock, sk_node.pprev));
1137}
1138
1da177e4
LT		 1139/* Networking protocol blocks we attach to sockets.
1140 * socket layer -> transport layer interface
1da177e4
LT		 1141 */
1142struct proto {
dc6b9b78 1143 void (*close)(struct sock *sk,
1da177e4 1144 long timeout);
d74bad4e
AI		 1145 int (*pre_connect)(struct sock *sk,
1146 struct sockaddr *uaddr,
1147 int addr_len);
1da177e4 1148 int (*connect)(struct sock *sk,
dc6b9b78 1149 struct sockaddr *uaddr,
1da177e4
LT		 1150 int addr_len);
1151 int (*disconnect)(struct sock *sk, int flags);
1152
cdfbabfb
DH		 1153 struct sock * (*accept)(struct sock *sk, int flags, int *err,
1154 bool kern);
1da177e4
LT		 1155
1156 int (*ioctl)(struct sock *sk, int cmd,
1157 unsigned long arg);
1158 int (*init)(struct sock *sk);
7d06b2e0 1159 void (*destroy)(struct sock *sk);
1da177e4 1160 void (*shutdown)(struct sock *sk, int how);
dc6b9b78 1161 int (*setsockopt)(struct sock *sk, int level,
a7b75c5a 1162 int optname, sockptr_t optval,
b7058842 1163 unsigned int optlen);
dc6b9b78
ED		 1164 int (*getsockopt)(struct sock *sk, int level,
1165 int optname, char __user *optval,
1166 int __user *option);
4b9d07a4 1167 void (*keepalive)(struct sock *sk, int valbool);
af01d537 1168#ifdef CONFIG_COMPAT
709b46e8
EB		 1169 int (*compat_ioctl)(struct sock *sk,
1170 unsigned int cmd, unsigned long arg);
af01d537 1171#endif
1b784140
YX		 1172 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
1173 size_t len);
1174 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
dc6b9b78
ED		 1175 size_t len, int noblock, int flags,
1176 int *addr_len);
1da177e4
LT		 1177 int (*sendpage)(struct sock *sk, struct page *page,
1178 int offset, size_t size, int flags);
dc6b9b78 1179 int (*bind)(struct sock *sk,
c0425a42
CH		 1180 struct sockaddr *addr, int addr_len);
1181 int (*bind_add)(struct sock *sk,
1182 struct sockaddr *addr, int addr_len);
1da177e4 1183
dc6b9b78 1184 int (*backlog_rcv) (struct sock *sk,
1da177e4 1185 struct sk_buff *skb);
9cacf81f
SF		 1186 bool (*bpf_bypass_getsockopt)(int level,
1187 int optname);
1da177e4 1188
46d3ceab
ED		 1189 void (*release_cb)(struct sock *sk);
1190
1da177e4 1191 /* Keeping track of sk's, looking them up, and port selection methods. */
086c653f 1192 int (*hash)(struct sock *sk);
1da177e4 1193 void (*unhash)(struct sock *sk);
719f8358 1194 void (*rehash)(struct sock *sk);
1da177e4 1195 int (*get_port)(struct sock *sk, unsigned short snum);
8a59f9d1 1196#ifdef CONFIG_BPF_SYSCALL
51e0158a
CW		 1197 int (*psock_update_sk_prot)(struct sock *sk,
1198 struct sk_psock *psock,
1199 bool restore);
8a59f9d1 1200#endif
1da177e4 1201
286ab3d4 1202 /* Keeping track of sockets in use */
65f76517 1203#ifdef CONFIG_PROC_FS
13ff3d6f 1204 unsigned int inuse_idx;
65f76517 1205#endif
ebb53d75 1206
a74f0fa0 1207 bool (*stream_memory_free)(const struct sock *sk, int wake);
8934ce2f 1208 bool (*stream_memory_read)(const struct sock *sk);
1da177e4 1209 /* Memory pressure */
5c52ba17 1210 void (*enter_memory_pressure)(struct sock *sk);
06044751 1211 void (*leave_memory_pressure)(struct sock *sk);
8d987e5c 1212 atomic_long_t *memory_allocated; /* Current allocated memory. */
1748376b 1213 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
1da177e4
LT		 1214 /*
1215 * Pressure flag: try to collapse.
1216 * Technical note: it is used by multiple contexts non atomically.
3ab224be 1217 * All the __sk_mem_schedule() is of this nature: accounting
1da177e4
LT		 1218 * is strict, actions are advisory and have some latency.
1219 */
06044751 1220 unsigned long *memory_pressure;
8d987e5c 1221 long *sysctl_mem;
a3dcaf17 1222
1da177e4
LT		 1223 int *sysctl_wmem;
1224 int *sysctl_rmem;
a3dcaf17
ED		 1225 u32 sysctl_wmem_offset;
1226 u32 sysctl_rmem_offset;
1227
1da177e4 1228 int max_header;
7ba42910 1229 bool no_autobind;
1da177e4 1230
271b72c7 1231 struct kmem_cache *slab;
1da177e4 1232 unsigned int obj_size;
d50112ed 1233 slab_flags_t slab_flags;
7bbdb81e
AD		 1234 unsigned int useroffset; /* Usercopy region offset */
1235 unsigned int usersize; /* Usercopy region size */
1da177e4 1236
dd24c001 1237 struct percpu_counter *orphan_count;
8feaf0c0 1238
60236fdd 1239 struct request_sock_ops *rsk_prot;
6d6ee43e 1240 struct timewait_sock_ops *twsk_prot;
2e6599cb 1241
39d8cda7
PE		 1242 union {
1243 struct inet_hashinfo *hashinfo;
645ca708 1244 struct udp_table *udp_table;
fc8717ba 1245 struct raw_hashinfo *raw_hash;
f16a7dd5 1246 struct smc_hashinfo *smc_hash;
39d8cda7 1247 } h;
ab1e0a13 1248
1da177e4
LT		 1249 struct module *owner;
1250
1251 char name[32];
1252
1253 struct list_head node;
e6848976
ACM		 1254#ifdef SOCK_REFCNT_DEBUG
1255 atomic_t socks;
e1aab161 1256#endif
64be0aed 1257 int (*diag_destroy)(struct sock *sk, int err);
3859a271 1258} __randomize_layout;
e1aab161 1259
69336bd2
JP		 1260int proto_register(struct proto *prot, int alloc_slab);
1261void proto_unregister(struct proto *prot);
bf2ae2e4 1262int sock_load_diag_module(int family, int protocol);
1da177e4 1263
e6848976
ACM		 1264#ifdef SOCK_REFCNT_DEBUG
1265static inline void sk_refcnt_debug_inc(struct sock *sk)
1266{
1267 atomic_inc(&sk->sk_prot->socks);
1268}
1269
1270static inline void sk_refcnt_debug_dec(struct sock *sk)
1271{
1272 atomic_dec(&sk->sk_prot->socks);
1273 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1274 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1275}
1276
dec34fb0 1277static inline void sk_refcnt_debug_release(const struct sock *sk)
e6848976 1278{
41c6d650 1279 if (refcount_read(&sk->sk_refcnt) != 1)
e6848976 1280 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
41c6d650 1281 sk->sk_prot->name, sk, refcount_read(&sk->sk_refcnt));
e6848976
ACM		 1282}
1283#else /* SOCK_REFCNT_DEBUG */
1284#define sk_refcnt_debug_inc(sk) do { } while (0)
1285#define sk_refcnt_debug_dec(sk) do { } while (0)
1286#define sk_refcnt_debug_release(sk) do { } while (0)
1287#endif /* SOCK_REFCNT_DEBUG */
1288
1c5f2ced
ED		 1289INDIRECT_CALLABLE_DECLARE(bool tcp_stream_memory_free(const struct sock *sk, int wake));
1290
a74f0fa0 1291static inline bool __sk_stream_memory_free(const struct sock *sk, int wake)
c9bee3b7 1292{
ab4e846a 1293 if (READ_ONCE(sk->sk_wmem_queued) >= READ_ONCE(sk->sk_sndbuf))
c9bee3b7
ED		 1294 return false;
1295
fc9840fb 1296#ifdef CONFIG_INET
c9bee3b7 1297 return sk->sk_prot->stream_memory_free ?
1c5f2ced
ED		 1298 INDIRECT_CALL_1(sk->sk_prot->stream_memory_free,
1299 tcp_stream_memory_free,
1300 sk, wake) : true;
fc9840fb 1301#else
c9bee3b7 1302 return sk->sk_prot->stream_memory_free ?
a74f0fa0 1303 sk->sk_prot->stream_memory_free(sk, wake) : true;
fc9840fb 1304#endif
c9bee3b7
ED		 1305}
1306
a74f0fa0
ED		 1307static inline bool sk_stream_memory_free(const struct sock *sk)
1308{
1309 return __sk_stream_memory_free(sk, 0);
1310}
1311
1312static inline bool __sk_stream_is_writeable(const struct sock *sk, int wake)
64dc6130 1313{
c9bee3b7 1314 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
a74f0fa0
ED		 1315 __sk_stream_memory_free(sk, wake);
1316}
1317
1318static inline bool sk_stream_is_writeable(const struct sock *sk)
1319{
1320 return __sk_stream_is_writeable(sk, 0);
64dc6130 1321}
e1aab161 1322
54fd9c2d
DB		 1323static inline int sk_under_cgroup_hierarchy(struct sock *sk,
1324 struct cgroup *ancestor)
1325{
1326#ifdef CONFIG_SOCK_CGROUP_DATA
1327 return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data),
1328 ancestor);
1329#else
1330 return -ENOTSUPP;
1331#endif
1332}
c9bee3b7 1333
180d8cd9
GC		 1334static inline bool sk_has_memory_pressure(const struct sock *sk)
1335{
1336 return sk->sk_prot->memory_pressure != NULL;
1337}
1338
1339static inline bool sk_under_memory_pressure(const struct sock *sk)
1340{
1341 if (!sk->sk_prot->memory_pressure)
1342 return false;
e1aab161 1343
baac50bb
JW		 1344 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1345 mem_cgroup_under_socket_pressure(sk->sk_memcg))
e805605c 1346 return true;
e1aab161 1347
35b87f6c 1348 return !!*sk->sk_prot->memory_pressure;
180d8cd9
GC		 1349}
1350
180d8cd9
GC		 1351static inline long
1352sk_memory_allocated(const struct sock *sk)
1353{
e805605c 1354 return atomic_long_read(sk->sk_prot->memory_allocated);
180d8cd9
GC		 1355}
1356
1357static inline long
e805605c 1358sk_memory_allocated_add(struct sock *sk, int amt)
180d8cd9 1359{
e805605c 1360 return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
180d8cd9
GC		 1361}
1362
1363static inline void
0e90b31f 1364sk_memory_allocated_sub(struct sock *sk, int amt)
180d8cd9 1365{
e805605c 1366 atomic_long_sub(amt, sk->sk_prot->memory_allocated);
180d8cd9
GC		 1367}
1368
f5a5589c
WW		 1369#define SK_ALLOC_PERCPU_COUNTER_BATCH 16
1370
180d8cd9
GC		 1371static inline void sk_sockets_allocated_dec(struct sock *sk)
1372{
f5a5589c
WW		 1373 percpu_counter_add_batch(sk->sk_prot->sockets_allocated, -1,
1374 SK_ALLOC_PERCPU_COUNTER_BATCH);
180d8cd9
GC		 1375}
1376
1377static inline void sk_sockets_allocated_inc(struct sock *sk)
1378{
f5a5589c
WW		 1379 percpu_counter_add_batch(sk->sk_prot->sockets_allocated, 1,
1380 SK_ALLOC_PERCPU_COUNTER_BATCH);
180d8cd9
GC		 1381}
1382
5bf325a5 1383static inline u64
180d8cd9
GC		 1384sk_sockets_allocated_read_positive(struct sock *sk)
1385{
af95d7df 1386 return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
180d8cd9
GC		 1387}
1388
1389static inline int
1390proto_sockets_allocated_sum_positive(struct proto *prot)
1391{
1392 return percpu_counter_sum_positive(prot->sockets_allocated);
1393}
1394
1395static inline long
1396proto_memory_allocated(struct proto *prot)
1397{
1398 return atomic_long_read(prot->memory_allocated);
1399}
1400
1401static inline bool
1402proto_memory_pressure(struct proto *prot)
1403{
1404 if (!prot->memory_pressure)
1405 return false;
1406 return !!*prot->memory_pressure;
1407}
1408
65f76517
ED		 1409
1410#ifdef CONFIG_PROC_FS
1da177e4 1411/* Called with local bh disabled */
69336bd2
JP		 1412void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
1413int sock_prot_inuse_get(struct net *net, struct proto *proto);
648845ab 1414int sock_inuse_get(struct net *net);
65f76517 1415#else
dc6b9b78 1416static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
c29a0bc4 1417 int inc)
65f76517
ED		 1418{
1419}
65f76517
ED		 1420#endif
1421
1da177e4 1422
614c6cb4
ACM		 1423/* With per-bucket locks this operation is not-atomic, so that
1424 * this version is not worse.
1425 */
086c653f 1426static inline int __sk_prot_rehash(struct sock *sk)
614c6cb4
ACM		 1427{
1428 sk->sk_prot->unhash(sk);
086c653f 1429 return sk->sk_prot->hash(sk);
614c6cb4
ACM		 1430}
1431
1da177e4
LT		 1432/* About 10 seconds */
1433#define SOCK_DESTROY_TIME (10*HZ)
1434
1435/* Sockets 0-1023 can't be bound to unless you are superuser */
1436#define PROT_SOCK 1024
1437
1438#define SHUTDOWN_MASK 3
1439#define RCV_SHUTDOWN 1
1440#define SEND_SHUTDOWN 2
1441
1da177e4
LT		 1442#define SOCK_BINDADDR_LOCK 4
1443#define SOCK_BINDPORT_LOCK 8
1444
1da177e4
LT		 1445struct socket_alloc {
1446 struct socket socket;
1447 struct inode vfs_inode;
1448};
1449
1450static inline struct socket *SOCKET_I(struct inode *inode)
1451{
1452 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1453}
1454
1455static inline struct inode *SOCK_INODE(struct socket *socket)
1456{
1457 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1458}
1459
3ab224be
HA		 1460/*
1461 * Functions for memory accounting
1462 */
f8c3bf00 1463int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind);
69336bd2 1464int __sk_mem_schedule(struct sock *sk, int size, int kind);
f8c3bf00 1465void __sk_mem_reduce_allocated(struct sock *sk, int amount);
1a24e04e 1466void __sk_mem_reclaim(struct sock *sk, int amount);
1da177e4 1467
bd68a2a8
ED		 1468/* We used to have PAGE_SIZE here, but systems with 64KB pages
1469 * do not necessarily have 16x time more memory than 4KB ones.
1470 */
1471#define SK_MEM_QUANTUM 4096
3ab224be
HA		 1472#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1473#define SK_MEM_SEND 0
1474#define SK_MEM_RECV 1
1da177e4 1475
bd68a2a8
ED		 1476/* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1477static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1478{
1479 long val = sk->sk_prot->sysctl_mem[index];
1480
1481#if PAGE_SIZE > SK_MEM_QUANTUM
1482 val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT;
1483#elif PAGE_SIZE < SK_MEM_QUANTUM
1484 val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT;
1485#endif
1486 return val;
1487}
1488
3ab224be 1489static inline int sk_mem_pages(int amt)
1da177e4 1490{
3ab224be 1491 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1da177e4
LT		 1492}
1493
dc6b9b78 1494static inline bool sk_has_account(struct sock *sk)
1da177e4 1495{
3ab224be
HA		 1496 /* return true if protocol supports memory accounting */
1497 return !!sk->sk_prot->memory_allocated;
1da177e4
LT		 1498}
1499
dc6b9b78 1500static inline bool sk_wmem_schedule(struct sock *sk, int size)
1da177e4 1501{
3ab224be 1502 if (!sk_has_account(sk))
dc6b9b78 1503 return true;
3ab224be
HA		 1504 return size <= sk->sk_forward_alloc ||
1505 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1da177e4
LT		 1506}
1507
c76562b6 1508static inline bool
35c448a8 1509sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
d80d99d6 1510{
3ab224be 1511 if (!sk_has_account(sk))
dc6b9b78 1512 return true;
5af68891 1513 return size <= sk->sk_forward_alloc ||
c76562b6
MG		 1514 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1515 skb_pfmemalloc(skb);
3ab224be
HA		 1516}
1517
1518static inline void sk_mem_reclaim(struct sock *sk)
1519{
1520 if (!sk_has_account(sk))
1521 return;
1522 if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
1a24e04e 1523 __sk_mem_reclaim(sk, sk->sk_forward_alloc);
3ab224be
HA		 1524}
1525
9993e7d3
DM		 1526static inline void sk_mem_reclaim_partial(struct sock *sk)
1527{
1528 if (!sk_has_account(sk))
1529 return;
1530 if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
1a24e04e 1531 __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1);
9993e7d3
DM		 1532}
1533
3ab224be
HA		 1534static inline void sk_mem_charge(struct sock *sk, int size)
1535{
1536 if (!sk_has_account(sk))
1537 return;
1538 sk->sk_forward_alloc -= size;
1539}
1540
1541static inline void sk_mem_uncharge(struct sock *sk, int size)
1542{
1543 if (!sk_has_account(sk))
1544 return;
1545 sk->sk_forward_alloc += size;
20c64d5c
ED		 1546
1547 /* Avoid a possible overflow.
1548 * TCP send queues can make this happen, if sk_mem_reclaim()
1549 * is not called and more than 2 GBytes are released at once.
1550 *
1551 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1552 * no need to hold that much forward allocation anyway.
1553 */
1554 if (unlikely(sk->sk_forward_alloc >= 1 << 21))
1555 __sk_mem_reclaim(sk, 1 << 20);
3ab224be
HA		 1556}
1557
0b7d7f6b 1558DECLARE_STATIC_KEY_FALSE(tcp_tx_skb_cache_key);
3ab224be
HA		 1559static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
1560{
ab4e846a 1561 sk_wmem_queued_add(sk, -skb->truesize);
4f661542 1562 sk_mem_uncharge(sk, skb->truesize);
0b7d7f6b
ED		 1563 if (static_branch_unlikely(&tcp_tx_skb_cache_key) &&
1564 !sk->sk_tx_skb_cache && !skb_cloned(skb)) {
e66b2f31 1565 skb_ext_reset(skb);
4f661542 1566 skb_zcopy_clear(skb, true);
472c2e07
ED		 1567 sk->sk_tx_skb_cache = skb;
1568 return;
1569 }
3ab224be 1570 __kfree_skb(skb);
d80d99d6
HX		 1571}
1572
c3f9b018
ED		 1573static inline void sock_release_ownership(struct sock *sk)
1574{
61881cfb
HFS		 1575 if (sk->sk_lock.owned) {
1576 sk->sk_lock.owned = 0;
1577
1578 /* The sk_lock has mutex_unlock() semantics: */
5facae4f 1579 mutex_release(&sk->sk_lock.dep_map, _RET_IP_);
61881cfb 1580 }
c3f9b018
ED		 1581}
1582
ed07536e
PZ		 1583/*
1584 * Macro so as to not evaluate some arguments when
1585 * lockdep is not enabled.
1586 *
1587 * Mark both the sk_lock and the sk_lock.slock as a
1588 * per-address-family lock class.
1589 */
dc6b9b78 1590#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
ed07536e 1591do { \
e8f6fbf6 1592 sk->sk_lock.owned = 0; \
ed07536e
PZ		 1593 init_waitqueue_head(&sk->sk_lock.wq); \
1594 spin_lock_init(&(sk)->sk_lock.slock); \
1595 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1596 sizeof((sk)->sk_lock)); \
1597 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
dc6b9b78 1598 (skey), (sname)); \
ed07536e
PZ		 1599 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1600} while (0)
1601
05b93801 1602static inline bool lockdep_sock_is_held(const struct sock *sk)
1e1d04e6 1603{
1e1d04e6
HFS		 1604 return lockdep_is_held(&sk->sk_lock) ||
1605 lockdep_is_held(&sk->sk_lock.slock);
1606}
1607
69336bd2 1608void lock_sock_nested(struct sock *sk, int subclass);
fcc70d5f
PZ		 1609
1610static inline void lock_sock(struct sock *sk)
1611{
1612 lock_sock_nested(sk, 0);
1613}
1614
ad80b0fc 1615void __lock_sock(struct sock *sk);
8873c064 1616void __release_sock(struct sock *sk);
69336bd2 1617void release_sock(struct sock *sk);
1da177e4
LT		 1618
1619/* BH context may only use the following locking interface. */
1620#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
c6366184
IM		 1621#define bh_lock_sock_nested(__sk) \
1622 spin_lock_nested(&((__sk)->sk_lock.slock), \
1623 SINGLE_DEPTH_NESTING)
1da177e4
LT		 1624#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1625
12f4bd86
PA		 1626bool lock_sock_fast(struct sock *sk) __acquires(&sk->sk_lock.slock);
1627
8a74ad60
ED		 1628/**
1629 * unlock_sock_fast - complement of lock_sock_fast
1630 * @sk: socket
1631 * @slow: slow mode
1632 *
1633 * fast unlock socket for user context.
1634 * If slow mode is on, we call regular release_sock()
1635 */
1636static inline void unlock_sock_fast(struct sock *sk, bool slow)
12f4bd86 1637 __releases(&sk->sk_lock.slock)
4b0b72f7 1638{
12f4bd86 1639 if (slow) {
8a74ad60 1640 release_sock(sk);
12f4bd86
PA		 1641 __release(&sk->sk_lock.slock);
1642 } else {
8a74ad60 1643 spin_unlock_bh(&sk->sk_lock.slock);
12f4bd86 1644 }
4b0b72f7
ED		 1645}
1646
fafc4e1e
HFS		 1647/* Used by processes to "lock" a socket state, so that
1648 * interrupts and bottom half handlers won't change it
1649 * from under us. It essentially blocks any incoming
1650 * packets, so that we won't get any new data or any
1651 * packets that change the state of the socket.
1652 *
1653 * While locked, BH processing will add new packets to
1654 * the backlog queue. This queue is processed by the
1655 * owner of the socket lock right before it is released.
1656 *
1657 * Since ~2.3.5 it is also exclusive sleep lock serializing
1658 * accesses from user process context.
1659 */
1660
46cc6e49 1661static inline void sock_owned_by_me(const struct sock *sk)
fafc4e1e
HFS		 1662{
1663#ifdef CONFIG_LOCKDEP
5e91f6ce 1664 WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
fafc4e1e 1665#endif
46cc6e49
ED		 1666}
1667
1668static inline bool sock_owned_by_user(const struct sock *sk)
1669{
1670 sock_owned_by_me(sk);
fafc4e1e
HFS		 1671 return sk->sk_lock.owned;
1672}
1673
602f7a27
TH		 1674static inline bool sock_owned_by_user_nocheck(const struct sock *sk)
1675{
1676 return sk->sk_lock.owned;
1677}
1678
fafc4e1e
HFS		 1679/* no reclassification while locks are held */
1680static inline bool sock_allow_reclassification(const struct sock *csk)
1681{
1682 struct sock *sk = (struct sock *)csk;
1683
1684 return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock);
1685}
4b0b72f7 1686
69336bd2 1687struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1688 struct proto *prot, int kern);
69336bd2 1689void sk_free(struct sock *sk);
eb4cb008 1690void sk_destruct(struct sock *sk);
69336bd2 1691struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
94352d45 1692void sk_free_unlock_clone(struct sock *sk);
69336bd2
JP		 1693
1694struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1695 gfp_t priority);
1d2077ac 1696void __sock_wfree(struct sk_buff *skb);
69336bd2 1697void sock_wfree(struct sk_buff *skb);
98ba0bd5
WB		 1698struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1699 gfp_t priority);
69336bd2
JP		 1700void skb_orphan_partial(struct sk_buff *skb);
1701void sock_rfree(struct sk_buff *skb);
62bccb8c 1702void sock_efree(struct sk_buff *skb);
82eabd9e 1703#ifdef CONFIG_INET
69336bd2 1704void sock_edemux(struct sk_buff *skb);
cf7fbe66 1705void sock_pfree(struct sk_buff *skb);
82eabd9e 1706#else
158f323b 1707#define sock_edemux sock_efree
82eabd9e 1708#endif
69336bd2
JP		 1709
1710int sock_setsockopt(struct socket *sock, int level, int op,
c8c1bbb6 1711 sockptr_t optval, unsigned int optlen);
69336bd2
JP		 1712
1713int sock_getsockopt(struct socket *sock, int level, int op,
1714 char __user *optval, int __user *optlen);
c7cbdbf2
AB		 1715int sock_gettstamp(struct socket *sock, void __user *userstamp,
1716 bool timeval, bool time32);
69336bd2
JP		 1717struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1718 int noblock, int *errcode);
1719struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1720 unsigned long data_len, int noblock,
1721 int *errcode, int max_page_order);
1722void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1723void sock_kfree_s(struct sock *sk, void *mem, int size);
79e88659 1724void sock_kzfree_s(struct sock *sk, void *mem, int size);
69336bd2 1725void sk_send_sigurg(struct sock *sk);
1da177e4 1726
f28ea365 1727struct sockcm_cookie {
80b14dee 1728 u64 transmit_time;
f28ea365 1729 u32 mark;
3dd17e63 1730 u16 tsflags;
f28ea365
EJ		 1731};
1732
657a0667
WB		 1733static inline void sockcm_init(struct sockcm_cookie *sockc,
1734 const struct sock *sk)
1735{
1736 *sockc = (struct sockcm_cookie) { .tsflags = sk->sk_tsflags };
1737}
1738
39771b12
WB		 1739int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1740 struct sockcm_cookie *sockc);
f28ea365
EJ		 1741int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1742 struct sockcm_cookie *sockc);
1743
1da177e4
LT		 1744/*
1745 * Functions to fill in entries in struct proto_ops when a protocol
1746 * does not implement a particular function.
1747 */
69336bd2
JP		 1748int sock_no_bind(struct socket *, struct sockaddr *, int);
1749int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1750int sock_no_socketpair(struct socket *, struct socket *);
cdfbabfb 1751int sock_no_accept(struct socket *, struct socket *, int, bool);
9b2c45d4 1752int sock_no_getname(struct socket *, struct sockaddr *, int);
69336bd2
JP		 1753int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1754int sock_no_listen(struct socket *, int);
1755int sock_no_shutdown(struct socket *, int);
1b784140 1756int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
306b13eb 1757int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len);
1b784140 1758int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
69336bd2
JP		 1759int sock_no_mmap(struct file *file, struct socket *sock,
1760 struct vm_area_struct *vma);
1761ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1762 size_t size, int flags);
306b13eb
TH		 1763ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
1764 int offset, size_t size, int flags);
1da177e4
LT		 1765
1766/*
1767 * Functions to fill in entries in struct proto_ops when a protocol
1768 * uses the inet style.
1769 */
69336bd2 1770int sock_common_getsockopt(struct socket *sock, int level, int optname,
1da177e4 1771 char __user *optval, int __user *optlen);
1b784140
YX		 1772int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1773 int flags);
69336bd2 1774int sock_common_setsockopt(struct socket *sock, int level, int optname,
a7b75c5a 1775 sockptr_t optval, unsigned int optlen);
1da177e4 1776
69336bd2 1777void sk_common_release(struct sock *sk);
1da177e4
LT		 1778
1779/*
1780 * Default socket callbacks and setup code
1781 */
dc6b9b78 1782
1da177e4 1783/* Initialise core socket variables */
69336bd2 1784void sock_init_data(struct socket *sock, struct sock *sk);
1da177e4 1785
1da177e4
LT		 1786/*
1787 * Socket reference counting postulates.
1788 *
1789 * * Each user of socket SHOULD hold a reference count.
1790 * * Each access point to socket (an hash table bucket, reference from a list,
1791 * running timer, skb in flight MUST hold a reference count.
1792 * * When reference count hits 0, it means it will never increase back.
1793 * * When reference count hits 0, it means that no references from
1794 * outside exist to this socket and current process on current CPU
1795 * is last user and may/should destroy this socket.
1796 * * sk_free is called from any context: process, BH, IRQ. When
1797 * it is called, socket has no references from outside -> sk_free
1798 * may release descendant resources allocated by the socket, but
1799 * to the time when it is called, socket is NOT referenced by any
1800 * hash tables, lists etc.
1801 * * Packets, delivered from outside (from network or from another process)
1802 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1803 * when they sit in queue. Otherwise, packets will leak to hole, when
1804 * socket is looked up by one cpu and unhasing is made by another CPU.
1805 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1806 * (leak to backlog). Packet socket does all the processing inside
1807 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1808 * use separate SMP lock, so that they are prone too.
1809 */
1810
1811/* Ungrab socket and destroy it, if it was the last reference. */
1812static inline void sock_put(struct sock *sk)
1813{
41c6d650 1814 if (refcount_dec_and_test(&sk->sk_refcnt))
1da177e4
LT		 1815 sk_free(sk);
1816}
05dbc7b5 1817/* Generic version of sock_put(), dealing with all sockets
41b822c5 1818 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
05dbc7b5
ED		 1819 */
1820void sock_gen_put(struct sock *sk);
1da177e4 1821
4f0c40d9 1822int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
c3f24cfb 1823 unsigned int trim_cap, bool refcounted);
4f0c40d9
WB		 1824static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1825 const int nested)
1826{
c3f24cfb 1827 return __sk_receive_skb(sk, skb, nested, 1, true);
4f0c40d9 1828}
25995ff5 1829
e022f0b4
KK		 1830static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1831{
755c31cd
AN		 1832 /* sk_tx_queue_mapping accept only upto a 16-bit value */
1833 if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX))
1834 return;
e022f0b4
KK		 1835 sk->sk_tx_queue_mapping = tx_queue;
1836}
1837
755c31cd
AN		 1838#define NO_QUEUE_MAPPING USHRT_MAX
1839
e022f0b4
KK		 1840static inline void sk_tx_queue_clear(struct sock *sk)
1841{
755c31cd 1842 sk->sk_tx_queue_mapping = NO_QUEUE_MAPPING;
e022f0b4
KK		 1843}
1844
1845static inline int sk_tx_queue_get(const struct sock *sk)
1846{
755c31cd
AN		 1847 if (sk && sk->sk_tx_queue_mapping != NO_QUEUE_MAPPING)
1848 return sk->sk_tx_queue_mapping;
1849
1850 return -1;
e022f0b4
KK		 1851}
1852
c6345ce7
AN		 1853static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb)
1854{
4e1beecc 1855#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
c6345ce7
AN		 1856 if (skb_rx_queue_recorded(skb)) {
1857 u16 rx_queue = skb_get_rx_queue(skb);
1858
1859 if (WARN_ON_ONCE(rx_queue == NO_QUEUE_MAPPING))
1860 return;
1861
1862 sk->sk_rx_queue_mapping = rx_queue;
1863 }
1864#endif
1865}
1866
1867static inline void sk_rx_queue_clear(struct sock *sk)
1868{
4e1beecc 1869#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
c6345ce7
AN		 1870 sk->sk_rx_queue_mapping = NO_QUEUE_MAPPING;
1871#endif
1872}
1873
fc9bab24
AN		 1874static inline int sk_rx_queue_get(const struct sock *sk)
1875{
4e1beecc 1876#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
fc9bab24
AN		 1877 if (sk && sk->sk_rx_queue_mapping != NO_QUEUE_MAPPING)
1878 return sk->sk_rx_queue_mapping;
4e1beecc 1879#endif
fc9bab24
AN		 1880
1881 return -1;
1882}
fc9bab24 1883
972692e0
DM		 1884static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1885{
1886 sk->sk_socket = sock;
1887}
1888
aa395145
ED		 1889static inline wait_queue_head_t *sk_sleep(struct sock *sk)
1890{
eaefd110
ED		 1891 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
1892 return &rcu_dereference_raw(sk->sk_wq)->wait;
aa395145 1893}
1da177e4
LT		 1894/* Detach socket from process context.
1895 * Announce socket dead, detach it from wait queue and inode.
1896 * Note that parent inode held reference count on this struct sock,
1897 * we do not release it in this function, because protocol
1898 * probably wants some additional cleanups or even continuing
1899 * to work with this socket (TCP).
1900 */
1901static inline void sock_orphan(struct sock *sk)
1902{
1903 write_lock_bh(&sk->sk_callback_lock);
1904 sock_set_flag(sk, SOCK_DEAD);
972692e0 1905 sk_set_socket(sk, NULL);
43815482 1906 sk->sk_wq = NULL;
1da177e4
LT		 1907 write_unlock_bh(&sk->sk_callback_lock);
1908}
1909
1910static inline void sock_graft(struct sock *sk, struct socket *parent)
1911{
0ffdaf5b 1912 WARN_ON(parent->sk);
1da177e4 1913 write_lock_bh(&sk->sk_callback_lock);
333f7909 1914 rcu_assign_pointer(sk->sk_wq, &parent->wq);
1da177e4 1915 parent->sk = sk;
972692e0 1916 sk_set_socket(sk, parent);
86741ec2 1917 sk->sk_uid = SOCK_INODE(parent)->i_uid;
4237c75c 1918 security_sock_graft(sk, parent);
1da177e4
LT		 1919 write_unlock_bh(&sk->sk_callback_lock);
1920}
1921
69336bd2
JP		 1922kuid_t sock_i_uid(struct sock *sk);
1923unsigned long sock_i_ino(struct sock *sk);
1da177e4 1924
86741ec2
LC		 1925static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
1926{
1927 return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
1928}
1929
58d607d3 1930static inline u32 net_tx_rndhash(void)
877d1f62 1931{
58d607d3
ED		 1932 u32 v = prandom_u32();
1933
1934 return v ?: 1;
1935}
877d1f62 1936
58d607d3
ED		 1937static inline void sk_set_txhash(struct sock *sk)
1938{
b71eaed8
ED		 1939 /* This pairs with READ_ONCE() in skb_set_hash_from_sk() */
1940 WRITE_ONCE(sk->sk_txhash, net_tx_rndhash());
877d1f62
TH		 1941}
1942
9c30ae83 1943static inline bool sk_rethink_txhash(struct sock *sk)
265f94ff 1944{
9c30ae83 1945 if (sk->sk_txhash) {
265f94ff 1946 sk_set_txhash(sk);
9c30ae83
YC		 1947 return true;
1948 }
1949 return false;
265f94ff
TH		 1950}
1951
1da177e4
LT		 1952static inline struct dst_entry *
1953__sk_dst_get(struct sock *sk)
1954{
1e1d04e6
HFS		 1955 return rcu_dereference_check(sk->sk_dst_cache,
1956 lockdep_sock_is_held(sk));
1da177e4
LT		 1957}
1958
1959static inline struct dst_entry *
1960sk_dst_get(struct sock *sk)
1961{
1962 struct dst_entry *dst;
1963
b6c6712a
ED		 1964 rcu_read_lock();
1965 dst = rcu_dereference(sk->sk_dst_cache);
f8864972
ED		 1966 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
1967 dst = NULL;
b6c6712a 1968 rcu_read_unlock();
1da177e4
LT		 1969 return dst;
1970}
1971
9c30ae83 1972static inline void __dst_negative_advice(struct sock *sk)
b6c6712a
ED		 1973{
1974 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
1975
1976 if (dst && dst->ops->negative_advice) {
1977 ndst = dst->ops->negative_advice(dst);
1978
1979 if (ndst != dst) {
1980 rcu_assign_pointer(sk->sk_dst_cache, ndst);
0a6957e7 1981 sk_tx_queue_clear(sk);
9b8805a3 1982 sk->sk_dst_pending_confirm = 0;
b6c6712a
ED		 1983 }
1984 }
1985}
1986
9c30ae83
YC		 1987static inline void dst_negative_advice(struct sock *sk)
1988{
1989 sk_rethink_txhash(sk);
1990 __dst_negative_advice(sk);
1991}
1992
1da177e4
LT		 1993static inline void
1994__sk_dst_set(struct sock *sk, struct dst_entry *dst)
1995{
1996 struct dst_entry *old_dst;
1997
e022f0b4 1998 sk_tx_queue_clear(sk);
9b8805a3 1999 sk->sk_dst_pending_confirm = 0;
95964c6d
ED		 2000 old_dst = rcu_dereference_protected(sk->sk_dst_cache,
2001 lockdep_sock_is_held(sk));
b6c6712a 2002 rcu_assign_pointer(sk->sk_dst_cache, dst);
1da177e4
LT		 2003 dst_release(old_dst);
2004}
2005
2006static inline void
2007sk_dst_set(struct sock *sk, struct dst_entry *dst)
2008{
7f502361
ED		 2009 struct dst_entry *old_dst;
2010
2011 sk_tx_queue_clear(sk);
9b8805a3 2012 sk->sk_dst_pending_confirm = 0;
5925a055 2013 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
7f502361 2014 dst_release(old_dst);
1da177e4
LT		 2015}
2016
2017static inline void
2018__sk_dst_reset(struct sock *sk)
2019{
b6c6712a 2020 __sk_dst_set(sk, NULL);
1da177e4
LT		 2021}
2022
2023static inline void
2024sk_dst_reset(struct sock *sk)
2025{
7f502361 2026 sk_dst_set(sk, NULL);
1da177e4
LT		 2027}
2028
69336bd2 2029struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 2030
69336bd2 2031struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 2032
9b8805a3
JA		 2033static inline void sk_dst_confirm(struct sock *sk)
2034{
25c7a6d1
ED		 2035 if (!READ_ONCE(sk->sk_dst_pending_confirm))
2036 WRITE_ONCE(sk->sk_dst_pending_confirm, 1);
9b8805a3
JA		 2037}
2038
4ff06203
JA		 2039static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
2040{
2041 if (skb_get_dst_pending_confirm(skb)) {
2042 struct sock *sk = skb->sk;
2043 unsigned long now = jiffies;
2044
2045 /* avoid dirtying neighbour */
25c7a6d1
ED		 2046 if (READ_ONCE(n->confirmed) != now)
2047 WRITE_ONCE(n->confirmed, now);
2048 if (sk && READ_ONCE(sk->sk_dst_pending_confirm))
2049 WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
4ff06203
JA		 2050 }
2051}
2052
f60e5990 2053bool sk_mc_loop(struct sock *sk);
2054
dc6b9b78 2055static inline bool sk_can_gso(const struct sock *sk)
bcd76111
HX		 2056{
2057 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
2058}
2059
69336bd2 2060void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
6cbb0df7 2061
c8f44aff 2062static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
a465419b
ED		 2063{
2064 sk->sk_route_nocaps |= flags;
2065 sk->sk_route_caps &= ~flags;
2066}
2067
c6e1a0d1 2068static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
57be5bda 2069 struct iov_iter *from, char *to,
912d398d 2070 int copy, int offset)
c6e1a0d1
TH		 2071{
2072 if (skb->ip_summed == CHECKSUM_NONE) {
57be5bda 2073 __wsum csum = 0;
15e6cb46 2074 if (!csum_and_copy_from_iter_full(to, copy, &csum, from))
57be5bda 2075 return -EFAULT;
912d398d 2076 skb->csum = csum_block_add(skb->csum, csum, offset);
c6e1a0d1 2077 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
15e6cb46 2078 if (!copy_from_iter_full_nocache(to, copy, from))
c6e1a0d1 2079 return -EFAULT;
15e6cb46 2080 } else if (!copy_from_iter_full(to, copy, from))
c6e1a0d1
TH		 2081 return -EFAULT;
2082
2083 return 0;
2084}
2085
2086static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
57be5bda 2087 struct iov_iter *from, int copy)
c6e1a0d1 2088{
912d398d 2089 int err, offset = skb->len;
c6e1a0d1 2090
912d398d
WY		 2091 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
2092 copy, offset);
c6e1a0d1 2093 if (err)
912d398d 2094 __skb_trim(skb, offset);
c6e1a0d1
TH		 2095
2096 return err;
2097}
2098
57be5bda 2099static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
c6e1a0d1
TH		 2100 struct sk_buff *skb,
2101 struct page *page,
2102 int off, int copy)
2103{
2104 int err;
2105
912d398d
WY		 2106 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
2107 copy, skb->len);
c6e1a0d1
TH		 2108 if (err)
2109 return err;
2110
2111 skb->len += copy;
2112 skb->data_len += copy;
2113 skb->truesize += copy;
ab4e846a 2114 sk_wmem_queued_add(sk, copy);
c6e1a0d1
TH		 2115 sk_mem_charge(sk, copy);
2116 return 0;
2117}
2118
c564039f
ED		 2119/**
2120 * sk_wmem_alloc_get - returns write allocations
2121 * @sk: socket
2122 *
66256e0b 2123 * Return: sk_wmem_alloc minus initial offset of one
c564039f
ED		 2124 */
2125static inline int sk_wmem_alloc_get(const struct sock *sk)
2126{
14afee4b 2127 return refcount_read(&sk->sk_wmem_alloc) - 1;
c564039f
ED		 2128}
2129
2130/**
2131 * sk_rmem_alloc_get - returns read allocations
2132 * @sk: socket
2133 *
66256e0b 2134 * Return: sk_rmem_alloc
c564039f
ED		 2135 */
2136static inline int sk_rmem_alloc_get(const struct sock *sk)
2137{
2138 return atomic_read(&sk->sk_rmem_alloc);
2139}
2140
2141/**
2142 * sk_has_allocations - check if allocations are outstanding
2143 * @sk: socket
2144 *
66256e0b 2145 * Return: true if socket has write or read allocations
c564039f 2146 */
dc6b9b78 2147static inline bool sk_has_allocations(const struct sock *sk)
c564039f
ED		 2148{
2149 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
2150}
2151
a57de0b4 2152/**
1ce0bf50 2153 * skwq_has_sleeper - check if there are any waiting processes
acfbe96a 2154 * @wq: struct socket_wq
a57de0b4 2155 *
66256e0b 2156 * Return: true if socket_wq has waiting processes
a57de0b4 2157 *
1ce0bf50 2158 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
a57de0b4
JO		 2159 * barrier call. They were added due to the race found within the tcp code.
2160 *
d651983d 2161 * Consider following tcp code paths::
a57de0b4 2162 *
d651983d
MCC		 2163 * CPU1 CPU2
2164 * sys_select receive packet
a57de0b4
JO		 2165 * ... ...
2166 * __add_wait_queue update tp->rcv_nxt
2167 * ... ...
2168 * tp->rcv_nxt check sock_def_readable
2169 * ... {
43815482
ED		 2170 * schedule rcu_read_lock();
2171 * wq = rcu_dereference(sk->sk_wq);
2172 * if (wq && waitqueue_active(&wq->wait))
2173 * wake_up_interruptible(&wq->wait)
a57de0b4
JO		 2174 * ...
2175 * }
2176 *
2177 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
2178 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
2179 * could then endup calling schedule and sleep forever if there are no more
2180 * data on the socket.
ad462769 2181 *
a57de0b4 2182 */
1ce0bf50 2183static inline bool skwq_has_sleeper(struct socket_wq *wq)
a57de0b4 2184{
1ce0bf50 2185 return wq && wq_has_sleeper(&wq->wait);
a57de0b4
JO		 2186}
2187
2188/**
2189 * sock_poll_wait - place memory barrier behind the poll_wait call.
2190 * @filp: file
89ab066d 2191 * @sock: socket to wait on
a57de0b4
JO		 2192 * @p: poll_table
2193 *
43815482 2194 * See the comments in the wq_has_sleeper function.
a57de0b4 2195 */
89ab066d
KG		 2196static inline void sock_poll_wait(struct file *filp, struct socket *sock,
2197 poll_table *p)
a57de0b4 2198{
d8bbd13b 2199 if (!poll_does_not_wait(p)) {
333f7909 2200 poll_wait(filp, &sock->wq.wait, p);
dc6b9b78 2201 /* We need to be sure we are in sync with the
a57de0b4
JO		 2202 * socket flags modification.
2203 *
43815482 2204 * This memory barrier is paired in the wq_has_sleeper.
dc6b9b78 2205 */
a57de0b4
JO		 2206 smp_mb();
2207 }
2208}
2209
b73c3d0e
TH		 2210static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
2211{
b71eaed8
ED		 2212 /* This pairs with WRITE_ONCE() in sk_set_txhash() */
2213 u32 txhash = READ_ONCE(sk->sk_txhash);
2214
2215 if (txhash) {
b73c3d0e 2216 skb->l4_hash = 1;
b71eaed8 2217 skb->hash = txhash;
b73c3d0e
TH		 2218 }
2219}
2220
9e17f8a4
ED		 2221void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
2222
1da177e4 2223/*
dc6b9b78 2224 * Queue a received datagram if it will fit. Stream and sequenced
1da177e4
LT		 2225 * protocols can't normally use this as they need to fit buffers in
2226 * and play with them.
2227 *
dc6b9b78 2228 * Inlined as it's very short and called for pretty much every
1da177e4
LT		 2229 * packet ever received.
2230 */
1da177e4
LT		 2231static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2232{
d55d87fd 2233 skb_orphan(skb);
1da177e4
LT		 2234 skb->sk = sk;
2235 skb->destructor = sock_rfree;
2236 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
3ab224be 2237 sk_mem_charge(sk, skb->truesize);
1da177e4
LT		 2238}
2239
098116e7 2240static inline __must_check bool skb_set_owner_sk_safe(struct sk_buff *skb, struct sock *sk)
9adc89af
PA		 2241{
2242 if (sk && refcount_inc_not_zero(&sk->sk_refcnt)) {
2243 skb_orphan(skb);
2244 skb->destructor = sock_efree;
2245 skb->sk = sk;
098116e7 2246 return true;
9adc89af 2247 }
098116e7 2248 return false;
9adc89af
PA		 2249}
2250
5e10da53
PA		 2251static inline void skb_prepare_for_gro(struct sk_buff *skb)
2252{
2253 if (skb->destructor != sock_wfree) {
2254 skb_orphan(skb);
2255 return;
2256 }
2257 skb->slow_gro = 1;
2258}
2259
69336bd2
JP		 2260void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2261 unsigned long expires);
1da177e4 2262
69336bd2 2263void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1da177e4 2264
08b81d87
GT		 2265void sk_stop_timer_sync(struct sock *sk, struct timer_list *timer);
2266
65101aec
PA		 2267int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue,
2268 struct sk_buff *skb, unsigned int flags,
69629464
ED		 2269 void (*destructor)(struct sock *sk,
2270 struct sk_buff *skb));
e6afc8ac 2271int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
69336bd2 2272int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1da177e4 2273
69336bd2 2274int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
364a9e93 2275struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
1da177e4
LT		 2276
2277/*
2278 * Recover an error report and clear atomically
2279 */
dc6b9b78 2280
1da177e4
LT		 2281static inline int sock_error(struct sock *sk)
2282{
c1cbe4b7 2283 int err;
f13ef100
ED		 2284
2285 /* Avoid an atomic operation for the common case.
2286 * This is racy since another cpu/thread can change sk_err under us.
2287 */
2288 if (likely(data_race(!sk->sk_err)))
c1cbe4b7 2289 return 0;
f13ef100 2290
c1cbe4b7 2291 err = xchg(&sk->sk_err, 0);
1da177e4
LT		 2292 return -err;
2293}
2294
e3ae2365
AA		 2295void sk_error_report(struct sock *sk);
2296
1da177e4
LT		 2297static inline unsigned long sock_wspace(struct sock *sk)
2298{
2299 int amt = 0;
2300
2301 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
14afee4b 2302 amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc);
dc6b9b78 2303 if (amt < 0)
1da177e4
LT		 2304 amt = 0;
2305 }
2306 return amt;
2307}
2308
ceb5d58b
ED		 2309/* Note:
2310 * We use sk->sk_wq_raw, from contexts knowing this
2311 * pointer is not NULL and cannot disappear/change.
2312 */
9cd3e072 2313static inline void sk_set_bit(int nr, struct sock *sk)
1da177e4 2314{
4be73522
ED		 2315 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2316 !sock_flag(sk, SOCK_FASYNC))
9317bb69
ED		 2317 return;
2318
ceb5d58b 2319 set_bit(nr, &sk->sk_wq_raw->flags);
9cd3e072
ED		 2320}
2321
2322static inline void sk_clear_bit(int nr, struct sock *sk)
2323{
4be73522
ED		 2324 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2325 !sock_flag(sk, SOCK_FASYNC))
9317bb69
ED		 2326 return;
2327
ceb5d58b 2328 clear_bit(nr, &sk->sk_wq_raw->flags);
9cd3e072
ED		 2329}
2330
ceb5d58b 2331static inline void sk_wake_async(const struct sock *sk, int how, int band)
1da177e4 2332{
ceb5d58b
ED		 2333 if (sock_flag(sk, SOCK_FASYNC)) {
2334 rcu_read_lock();
2335 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
2336 rcu_read_unlock();
2337 }
1da177e4
LT		 2338}
2339
eea86af6
DB		 2340/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2341 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2342 * Note: for send buffers, TCP works better if we can build two skbs at
2343 * minimum.
7a91b434 2344 */
9eb5bf83 2345#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
eea86af6
DB		 2346
2347#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2348#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
1da177e4
LT		 2349
2350static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2351{
e292f05e
ED		 2352 u32 val;
2353
2354 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
2355 return;
2356
2357 val = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
2358
2359 WRITE_ONCE(sk->sk_sndbuf, max_t(u32, val, SOCK_MIN_SNDBUF));
1da177e4
LT		 2360}
2361
eb934478
ED		 2362struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp,
2363 bool force_schedule);
1da177e4 2364
5640f768
ED		 2365/**
2366 * sk_page_frag - return an appropriate page_frag
2367 * @sk: socket
2368 *
20eb4f29
TH		 2369 * Use the per task page_frag instead of the per socket one for
2370 * optimization when we know that we're in the normal context and owns
2371 * everything that's associated with %current.
2372 *
2373 * gfpflags_allow_blocking() isn't enough here as direct reclaim may nest
2374 * inside other socket operations and end up recursing into sk_page_frag()
2375 * while it's already in use.
66256e0b
RD		 2376 *
2377 * Return: a per task page_frag if context allows that,
2378 * otherwise a per socket one.
5640f768
ED		 2379 */
2380static inline struct page_frag *sk_page_frag(struct sock *sk)
1da177e4 2381{
20eb4f29 2382 if (gfpflags_normal_context(sk->sk_allocation))
5640f768 2383 return &current->task_frag;
1da177e4 2384
5640f768 2385 return &sk->sk_frag;
1da177e4
LT		 2386}
2387
69336bd2 2388bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
5640f768 2389
1da177e4
LT		 2390/*
2391 * Default write policy as shown to user space via poll/select/SIGIO
2392 */
dc6b9b78 2393static inline bool sock_writeable(const struct sock *sk)
1da177e4 2394{
e292f05e 2395 return refcount_read(&sk->sk_wmem_alloc) < (READ_ONCE(sk->sk_sndbuf) >> 1);
1da177e4
LT		 2396}
2397
dd0fc66f 2398static inline gfp_t gfp_any(void)
1da177e4 2399{
99709372 2400 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1da177e4
LT		 2401}
2402
dc6b9b78 2403static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
1da177e4
LT		 2404{
2405 return noblock ? 0 : sk->sk_rcvtimeo;
2406}
2407
dc6b9b78 2408static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
1da177e4
LT		 2409{
2410 return noblock ? 0 : sk->sk_sndtimeo;
2411}
2412
2413static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2414{
eac66402
ED		 2415 int v = waitall ? len : min_t(int, READ_ONCE(sk->sk_rcvlowat), len);
2416
2417 return v ?: 1;
1da177e4
LT		 2418}
2419
2420/* Alas, with timeout socket operations are not restartable.
2421 * Compare this to poll().
2422 */
2423static inline int sock_intr_errno(long timeo)
2424{
2425 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2426}
2427
744d5a3e
EB		 2428struct sock_skb_cb {
2429 u32 dropcount;
2430};
2431
2432/* Store sock_skb_cb at the end of skb->cb[] so protocol families
2433 * using skb->cb[] would keep using it directly and utilize its
2434 * alignement guarantee.
2435 */
c593642c 2436#define SOCK_SKB_CB_OFFSET ((sizeof_field(struct sk_buff, cb) - \
744d5a3e
EB		 2437 sizeof(struct sock_skb_cb)))
2438
2439#define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2440 SOCK_SKB_CB_OFFSET))
2441
b4772ef8 2442#define sock_skb_cb_check_size(size) \
744d5a3e 2443 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
b4772ef8 2444
3bc3b96f
EB		 2445static inline void
2446sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2447{
3665f381
ED		 2448 SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ?
2449 atomic_read(&sk->sk_drops) : 0;
3bc3b96f
EB		 2450}
2451
532182cd
ED		 2452static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2453{
2454 int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2455
2456 atomic_add(segs, &sk->sk_drops);
2457}
2458
3a0ed3e9
DD		 2459static inline ktime_t sock_read_timestamp(struct sock *sk)
2460{
2461#if BITS_PER_LONG==32
2462 unsigned int seq;
2463 ktime_t kt;
2464
2465 do {
2466 seq = read_seqbegin(&sk->sk_stamp_seq);
2467 kt = sk->sk_stamp;
2468 } while (read_seqretry(&sk->sk_stamp_seq, seq));
2469
2470 return kt;
2471#else
f75359f3 2472 return READ_ONCE(sk->sk_stamp);
3a0ed3e9
DD		 2473#endif
2474}
2475
2476static inline void sock_write_timestamp(struct sock *sk, ktime_t kt)
2477{
2478#if BITS_PER_LONG==32
2479 write_seqlock(&sk->sk_stamp_seq);
2480 sk->sk_stamp = kt;
2481 write_sequnlock(&sk->sk_stamp_seq);
2482#else
f75359f3 2483 WRITE_ONCE(sk->sk_stamp, kt);
3a0ed3e9
DD		 2484#endif
2485}
2486
69336bd2
JP		 2487void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2488 struct sk_buff *skb);
2489void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2490 struct sk_buff *skb);
92f37fd2 2491
dc6b9b78 2492static inline void
1da177e4
LT		 2493sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2494{
b7aa0bf7 2495 ktime_t kt = skb->tstamp;
20d49473 2496 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
a61bbcf2 2497
20d49473
PO		 2498 /*
2499 * generate control messages if
b9f40e21 2500 * - receive time stamping in software requested
20d49473 2501 * - software time stamp available and wanted
20d49473 2502 * - hardware time stamps available and wanted
20d49473
PO		 2503 */
2504 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
b9f40e21 2505 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2456e855
TG		 2506 (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2507 (hwtstamps->hwtstamp &&
b9f40e21 2508 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
92f37fd2
ED		 2509 __sock_recv_timestamp(msg, sk, skb);
2510 else
3a0ed3e9 2511 sock_write_timestamp(sk, kt);
6e3e939f
JB		 2512
2513 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2514 __sock_recv_wifi_status(msg, sk, skb);
1da177e4
LT		 2515}
2516
69336bd2
JP		 2517void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2518 struct sk_buff *skb);
767dd033 2519
6c7c98ba 2520#define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
767dd033
ED		 2521static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2522 struct sk_buff *skb)
2523{
2524#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
b9f40e21
WB		 2525 (1UL << SOCK_RCVTSTAMP))
2526#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2527 SOF_TIMESTAMPING_RAW_HARDWARE)
767dd033 2528
b9f40e21 2529 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
767dd033 2530 __sock_recv_ts_and_drops(msg, sk, skb);
d3fbff30 2531 else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP)))
3a0ed3e9 2532 sock_write_timestamp(sk, skb->tstamp);
6c7c98ba 2533 else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP))
3a0ed3e9 2534 sock_write_timestamp(sk, 0);
767dd033 2535}
3b885787 2536
c14ac945 2537void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
67cc0d40 2538
20d49473 2539/**
8f932f76 2540 * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
20d49473 2541 * @sk: socket sending this packet
c14ac945 2542 * @tsflags: timestamping flags to use
140c55d4 2543 * @tx_flags: completed with instructions for time stamping
8f932f76 2544 * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno)
140c55d4 2545 *
d651983d 2546 * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
20d49473 2547 */
8f932f76
WB		 2548static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2549 __u8 *tx_flags, __u32 *tskey)
67cc0d40 2550{
8f932f76 2551 if (unlikely(tsflags)) {
c14ac945 2552 __sock_tx_timestamp(tsflags, tx_flags);
8f932f76
WB		 2553 if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey &&
2554 tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK)
2555 *tskey = sk->sk_tskey++;
2556 }
67cc0d40
WB		 2557 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2558 *tx_flags |= SKBTX_WIFI_STATUS;
2559}
20d49473 2560
8f932f76
WB		 2561static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2562 __u8 *tx_flags)
2563{
2564 _sock_tx_timestamp(sk, tsflags, tx_flags, NULL);
2565}
2566
2567static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags)
2568{
2569 _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags,
2570 &skb_shinfo(skb)->tskey);
2571}
2572
66256e0b 2573DECLARE_STATIC_KEY_FALSE(tcp_rx_skb_cache_key);
1da177e4
LT		 2574/**
2575 * sk_eat_skb - Release a skb if it is no longer needed
4dc3b16b
PP		 2576 * @sk: socket to eat this skb from
2577 * @skb: socket buffer to eat
1da177e4
LT		 2578 *
2579 * This routine must be called with interrupts disabled or with the socket
2580 * locked so that the sk_buff queue operation is ok.
2581*/
7bced397 2582static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1da177e4
LT		 2583{
2584 __skb_unlink(skb, &sk->sk_receive_queue);
ede61ca4 2585 if (static_branch_unlikely(&tcp_rx_skb_cache_key) &&
8b27dae5
ED		 2586 !sk->sk_rx_skb_cache) {
2587 sk->sk_rx_skb_cache = skb;
2588 skb_orphan(skb);
2589 return;
2590 }
1da177e4
LT		 2591 __kfree_skb(skb);
2592}
2593
3b1e0a65
YH		 2594static inline
2595struct net *sock_net(const struct sock *sk)
2596{
c2d9ba9b 2597 return read_pnet(&sk->sk_net);
3b1e0a65
YH		 2598}
2599
2600static inline
f5aa23fd 2601void sock_net_set(struct sock *sk, struct net *net)
3b1e0a65 2602{
c2d9ba9b 2603 write_pnet(&sk->sk_net, net);
3b1e0a65
YH		 2604}
2605
cf7fbe66
JS		 2606static inline bool
2607skb_sk_is_prefetched(struct sk_buff *skb)
2608{
2609#ifdef CONFIG_INET
2610 return skb->destructor == sock_pfree;
2611#else
2612 return false;
2613#endif /* CONFIG_INET */
2614}
2615
7ae215d2
JS		 2616/* This helper checks if a socket is a full socket,
2617 * ie _not_ a timewait or request socket.
2618 */
2619static inline bool sk_fullsock(const struct sock *sk)
2620{
2621 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2622}
2623
2624static inline bool
2625sk_is_refcounted(struct sock *sk)
2626{
2627 /* Only full sockets have sk->sk_flags. */
2628 return !sk_fullsock(sk) || !sock_flag(sk, SOCK_RCU_FREE);
2629}
2630
71489e21 2631/**
045065f0
LR		 2632 * skb_steal_sock - steal a socket from an sk_buff
2633 * @skb: sk_buff to steal the socket from
2634 * @refcounted: is set to true if the socket is reference-counted
71489e21
JS		 2635 */
2636static inline struct sock *
2637skb_steal_sock(struct sk_buff *skb, bool *refcounted)
23542618 2638{
efc27f8c 2639 if (skb->sk) {
23542618
KK		 2640 struct sock *sk = skb->sk;
2641
71489e21 2642 *refcounted = true;
7ae215d2
JS		 2643 if (skb_sk_is_prefetched(skb))
2644 *refcounted = sk_is_refcounted(sk);
23542618
KK		 2645 skb->destructor = NULL;
2646 skb->sk = NULL;
2647 return sk;
2648 }
71489e21 2649 *refcounted = false;
23542618
KK		 2650 return NULL;
2651}
2652
ebf4e808
IL		 2653/* Checks if this SKB belongs to an HW offloaded socket
2654 * and whether any SW fallbacks are required based on dev.
41477662 2655 * Check decrypted mark in case skb_orphan() cleared socket.
ebf4e808
IL		 2656 */
2657static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb,
2658 struct net_device *dev)
2659{
2660#ifdef CONFIG_SOCK_VALIDATE_XMIT
2661 struct sock *sk = skb->sk;
2662
41477662 2663 if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb) {
ebf4e808 2664 skb = sk->sk_validate_xmit_skb(sk, dev, skb);
41477662
JK		 2665#ifdef CONFIG_TLS_DEVICE
2666 } else if (unlikely(skb->decrypted)) {
2667 pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n");
2668 kfree_skb(skb);
2669 skb = NULL;
2670#endif
2671 }
ebf4e808
IL		 2672#endif
2673
2674 return skb;
2675}
2676
e446f9df
ED		 2677/* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2678 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2679 */
2680static inline bool sk_listener(const struct sock *sk)
2681{
2682 return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2683}
2684
193d357d 2685void sock_enable_timestamp(struct sock *sk, enum sock_flags flag);
69336bd2
JP		 2686int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2687 int type);
1da177e4 2688
a3b299da
EB		 2689bool sk_ns_capable(const struct sock *sk,
2690 struct user_namespace *user_ns, int cap);
2691bool sk_capable(const struct sock *sk, int cap);
2692bool sk_net_capable(const struct sock *sk, int cap);
2693
a2d133b1
JH		 2694void sk_get_meminfo(const struct sock *sk, u32 *meminfo);
2695
eaa72dc4
ED		 2696/* Take into consideration the size of the struct sk_buff overhead in the
2697 * determination of these values, since that is non-constant across
2698 * platforms. This makes socket queueing behavior and performance
2699 * not depend upon such differences.
2700 */
2701#define _SK_MEM_PACKETS 256
2702#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
2703#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2704#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2705
1da177e4
LT		 2706extern __u32 sysctl_wmem_max;
2707extern __u32 sysctl_rmem_max;
2708
b245be1f 2709extern int sysctl_tstamp_allow_data;
6baf1f41
DM		 2710extern int sysctl_optmem_max;
2711
20380731
ACM		 2712extern __u32 sysctl_wmem_default;
2713extern __u32 sysctl_rmem_default;
20380731 2714
ce27ec60
ED		 2715DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key);
2716
a3dcaf17
ED		 2717static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto)
2718{
2719 /* Does this proto have per netns sysctl_wmem ? */
2720 if (proto->sysctl_wmem_offset)
2721 return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset);
2722
2723 return *proto->sysctl_wmem;
2724}
2725
2726static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto)
2727{
2728 /* Does this proto have per netns sysctl_rmem ? */
2729 if (proto->sysctl_rmem_offset)
2730 return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset);
2731
2732 return *proto->sysctl_rmem;
2733}
2734
c9f1f58d
ED		 2735/* Default TCP Small queue budget is ~1 ms of data (1sec >> 10)
2736 * Some wifi drivers need to tweak it to get more chunks.
2737 * They can use this helper from their ndo_start_xmit()
2738 */
2739static inline void sk_pacing_shift_update(struct sock *sk, int val)
2740{
7c68fa2b 2741 if (!sk || !sk_fullsock(sk) || READ_ONCE(sk->sk_pacing_shift) == val)
c9f1f58d 2742 return;
7c68fa2b 2743 WRITE_ONCE(sk->sk_pacing_shift, val);
c9f1f58d
ED		 2744}
2745
54dc3e33
DA		 2746/* if a socket is bound to a device, check that the given device
2747 * index is either the same or that the socket is bound to an L3
2748 * master device and the given device index is also enslaved to
2749 * that L3 master
2750 */
2751static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
2752{
2753 int mdif;
2754
2755 if (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif)
2756 return true;
2757
2758 mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif);
2759 if (mdif && mdif == sk->sk_bound_dev_if)
2760 return true;
2761
2762 return false;
2763}
2764
43a825af
BT		 2765void sock_def_readable(struct sock *sk);
2766
8ea204c2 2767int sock_bindtoindex(struct sock *sk, int ifindex, bool lock_sk);
371087aa 2768void sock_set_timestamp(struct sock *sk, int optname, bool valbool);
d463126e
YL		 2769int sock_set_timestamping(struct sock *sk, int optname,
2770 struct so_timestamping timestamping);
ced122d9 2771
783da70e 2772void sock_enable_timestamps(struct sock *sk);
c433594c 2773void sock_no_linger(struct sock *sk);
ce3d9544 2774void sock_set_keepalive(struct sock *sk);
6e434967 2775void sock_set_priority(struct sock *sk, u32 priority);
26cfabf9 2776void sock_set_rcvbuf(struct sock *sk, int val);
84d1c617 2777void sock_set_mark(struct sock *sk, u32 val);
b58f0e8f 2778void sock_set_reuseaddr(struct sock *sk);
fe31a326 2779void sock_set_reuseport(struct sock *sk);
76ee0785 2780void sock_set_sndtimeo(struct sock *sk, s64 secs);
b58f0e8f 2781
c0425a42
CH		 2782int sock_bind_add(struct sock *sk, struct sockaddr *addr, int addr_len);
2783
1da177e4 2784#endif /* _SOCK_H */
Ubuntu Jammy Linux kernel mirror