projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
sr9700: sanity check for packet length
[linux-block.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>
88ab1932 59#include <linux/rculist_nulls.h>
a57de0b4 60#include <linux/poll.h>
c8c1bbb6 61#include <linux/sockptr.h>
1c5f2ced 62#include <linux/indirect_call_wrapper.h>
c31504dc 63#include <linux/atomic.h>
41c6d650 64#include <linux/refcount.h>
f35f8219 65#include <linux/llist.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;
b6459415 251struct sk_filter;
6ac99e8f 252
1da177e4
LT		 253/**
254 * struct sock - network layer representation of sockets
8feaf0c0 255 * @__sk_common: shared layout with inet_timewait_sock
4dc3b16b
PP		 256 * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
257 * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
258 * @sk_lock: synchronizer
cdfbabfb 259 * @sk_kern_sock: True if sock is using kernel lock classes
4dc3b16b 260 * @sk_rcvbuf: size of receive buffer in bytes
43815482 261 * @sk_wq: sock wait queue and async head
421b3885 262 * @sk_rx_dst: receive input route used by early demux
0c0a5ef8 263 * @sk_rx_dst_ifindex: ifindex for @sk_rx_dst
ef57c161 264 * @sk_rx_dst_cookie: cookie for @sk_rx_dst
4dc3b16b 265 * @sk_dst_cache: destination cache
9b8805a3 266 * @sk_dst_pending_confirm: need to confirm neighbour
4dc3b16b 267 * @sk_policy: flow policy
4dc3b16b
PP		 268 * @sk_receive_queue: incoming packets
269 * @sk_wmem_alloc: transmit queue bytes committed
771edcaf 270 * @sk_tsq_flags: TCP Small Queues flags
4dc3b16b
PP		 271 * @sk_write_queue: Packet sending queue
272 * @sk_omem_alloc: "o" is "option" or "other"
273 * @sk_wmem_queued: persistent queue size
274 * @sk_forward_alloc: space allocated forward
2bb2f5fb 275 * @sk_reserved_mem: space reserved and non-reclaimable for the socket
06021292 276 * @sk_napi_id: id of the last napi context to receive data for sk
dafcc438 277 * @sk_ll_usec: usecs to busypoll when there is no data
4dc3b16b 278 * @sk_allocation: allocation mode
95bd09eb 279 * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
218af599 280 * @sk_pacing_status: Pacing status (requested, handled by sch_fq)
c3f40d7c 281 * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
4dc3b16b 282 * @sk_sndbuf: size of send buffer in bytes
771edcaf 283 * @__sk_flags_offset: empty field used to determine location of bitfield
293de7de 284 * @sk_padding: unused element for alignment
28448b80
TH		 285 * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets
286 * @sk_no_check_rx: allow zero checksum in RX packets
4dc3b16b 287 * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
aba54656 288 * @sk_gso_disabled: if set, NETIF_F_GSO_MASK is forbidden.
bcd76111 289 * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
82cc1a7a 290 * @sk_gso_max_size: Maximum GSO segment size to build
1485348d 291 * @sk_gso_max_segs: Maximum number of GSO segments
3a9b76fd 292 * @sk_pacing_shift: scaling factor for TCP Small Queues
4dc3b16b 293 * @sk_lingertime: %SO_LINGER l_linger setting
4dc3b16b 294 * @sk_backlog: always used with the per-socket spinlock held
df6160de 295 * @defer_list: head of llist storing skbs to be freed
4dc3b16b
PP		 296 * @sk_callback_lock: used with the callbacks in the end of this struct
297 * @sk_error_queue: rarely used
33c732c3
WC		 298 * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
299 * IPV6_ADDRFORM for instance)
4dc3b16b 300 * @sk_err: last error
33c732c3
WC		 301 * @sk_err_soft: errors that don't cause failure but are the cause of a
302 * persistent failure not just 'timed out'
cb61cb9b 303 * @sk_drops: raw/udp drops counter
4dc3b16b
PP		 304 * @sk_ack_backlog: current listen backlog
305 * @sk_max_ack_backlog: listen backlog set in listen()
771edcaf 306 * @sk_uid: user id of owner
7fd3253a 307 * @sk_prefer_busy_poll: prefer busypolling over softirq processing
7c951caf 308 * @sk_busy_poll_budget: napi processing budget when busypolling
4dc3b16b
PP		 309 * @sk_priority: %SO_PRIORITY setting
310 * @sk_type: socket type (%SOCK_STREAM, etc)
311 * @sk_protocol: which protocol this socket belongs in this network family
5fb14d20 312 * @sk_peer_lock: lock protecting @sk_peer_pid and @sk_peer_cred
53c3fa20
RD		 313 * @sk_peer_pid: &struct pid for this socket's peer
314 * @sk_peer_cred: %SO_PEERCRED setting
4dc3b16b
PP		 315 * @sk_rcvlowat: %SO_RCVLOWAT setting
316 * @sk_rcvtimeo: %SO_RCVTIMEO setting
317 * @sk_sndtimeo: %SO_SNDTIMEO setting
b73c3d0e 318 * @sk_txhash: computed flow hash for use on transmit
4dc3b16b 319 * @sk_filter: socket filtering instructions
4dc3b16b
PP		 320 * @sk_timer: sock cleanup timer
321 * @sk_stamp: time stamp of last packet received
3a0ed3e9 322 * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only
d463126e
YL		 323 * @sk_tsflags: SO_TIMESTAMPING flags
324 * @sk_bind_phc: SO_TIMESTAMPING bind PHC index of PTP virtual clock
325 * for timestamping
09c2d251 326 * @sk_tskey: counter to disambiguate concurrent tstamp requests
52267790 327 * @sk_zckey: counter to order MSG_ZEROCOPY notifications
4dc3b16b
PP		 328 * @sk_socket: Identd and reporting IO signals
329 * @sk_user_data: RPC layer private data
5640f768 330 * @sk_frag: cached page frag
d3d4f0a0 331 * @sk_peek_off: current peek_offset value
4dc3b16b 332 * @sk_send_head: front of stuff to transmit
66256e0b 333 * @tcp_rtx_queue: TCP re-transmit queue [union with @sk_send_head]
67be2dd1 334 * @sk_security: used by security modules
31729363 335 * @sk_mark: generic packet mark
2a56a1fe 336 * @sk_cgrp_data: cgroup data for this cgroup
baac50bb 337 * @sk_memcg: this socket's memory cgroup association
4dc3b16b
PP		 338 * @sk_write_pending: a write to stream socket waits to start
339 * @sk_state_change: callback to indicate change in the state of the sock
340 * @sk_data_ready: callback to indicate there is data to be processed
341 * @sk_write_space: callback to indicate there is bf sending space available
342 * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
343 * @sk_backlog_rcv: callback to process the backlog
66256e0b 344 * @sk_validate_xmit_skb: ptr to an optional validate function
4dc3b16b 345 * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
ef456144 346 * @sk_reuseport_cb: reuseport group container
66256e0b 347 * @sk_bpf_storage: ptr to cache and control for bpf_sk_storage
293de7de 348 * @sk_rcu: used during RCU grace period
80b14dee
RC		 349 * @sk_clockid: clockid used by time-based scheduling (SO_TXTIME)
350 * @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME
66256e0b 351 * @sk_txtime_report_errors: set report errors mode for SO_TXTIME
80b14dee 352 * @sk_txtime_unused: unused txtime flags
ffa84b5f 353 * @ns_tracker: tracker for netns reference
293de7de 354 */
1da177e4
LT		 355struct sock {
356 /*
8feaf0c0 357 * Now struct inet_timewait_sock also uses sock_common, so please just
1da177e4
LT		 358 * don't add nothing before this first member (__sk_common) --acme
359 */
360 struct sock_common __sk_common;
4dc6dc71
ED		 361#define sk_node __sk_common.skc_node
362#define sk_nulls_node __sk_common.skc_nulls_node
363#define sk_refcnt __sk_common.skc_refcnt
e022f0b4 364#define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
4e1beecc 365#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
c6345ce7
AN		 366#define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping
367#endif
4dc6dc71 368
68835aba
ED		 369#define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
370#define sk_dontcopy_end __sk_common.skc_dontcopy_end
4dc6dc71 371#define sk_hash __sk_common.skc_hash
50805466 372#define sk_portpair __sk_common.skc_portpair
05dbc7b5
ED		 373#define sk_num __sk_common.skc_num
374#define sk_dport __sk_common.skc_dport
50805466
ED		 375#define sk_addrpair __sk_common.skc_addrpair
376#define sk_daddr __sk_common.skc_daddr
377#define sk_rcv_saddr __sk_common.skc_rcv_saddr
1da177e4
LT		 378#define sk_family __sk_common.skc_family
379#define sk_state __sk_common.skc_state
380#define sk_reuse __sk_common.skc_reuse
055dc21a 381#define sk_reuseport __sk_common.skc_reuseport
9fe516ba 382#define sk_ipv6only __sk_common.skc_ipv6only
26abe143 383#define sk_net_refcnt __sk_common.skc_net_refcnt
1da177e4 384#define sk_bound_dev_if __sk_common.skc_bound_dev_if
1da177e4 385#define sk_bind_node __sk_common.skc_bind_node
8feaf0c0 386#define sk_prot __sk_common.skc_prot
07feaebf 387#define sk_net __sk_common.skc_net
efe4208f
ED		 388#define sk_v6_daddr __sk_common.skc_v6_daddr
389#define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
33cf7c90 390#define sk_cookie __sk_common.skc_cookie
70da268b 391#define sk_incoming_cpu __sk_common.skc_incoming_cpu
8e5eb54d 392#define sk_flags __sk_common.skc_flags
ed53d0ab 393#define sk_rxhash __sk_common.skc_rxhash
efe4208f 394
43f51df4 395 /* early demux fields */
8b3f9133 396 struct dst_entry __rcu *sk_rx_dst;
43f51df4
ED		 397 int sk_rx_dst_ifindex;
398 u32 sk_rx_dst_cookie;
399
1da177e4 400 socket_lock_t sk_lock;
9115e8cd
ED		 401 atomic_t sk_drops;
402 int sk_rcvlowat;
403 struct sk_buff_head sk_error_queue;
b178bb3d 404 struct sk_buff_head sk_receive_queue;
fa438ccf
ED		 405 /*
406 * The backlog queue is special, it is always used with
407 * the per-socket spinlock held and requires low latency
408 * access. Therefore we special case it's implementation.
b178bb3d
ED		 409 * Note : rmem_alloc is in this structure to fill a hole
410 * on 64bit arches, not because its logically part of
411 * backlog.
fa438ccf
ED		 412 */
413 struct {
b178bb3d
ED		 414 atomic_t rmem_alloc;
415 int len;
416 struct sk_buff *head;
417 struct sk_buff *tail;
fa438ccf 418 } sk_backlog;
f35f8219
ED		 419 struct llist_head defer_list;
420
b178bb3d 421#define sk_rmem_alloc sk_backlog.rmem_alloc
2c8c56e1 422
9115e8cd 423 int sk_forward_alloc;
2bb2f5fb 424 u32 sk_reserved_mem;
e0d1095a 425#ifdef CONFIG_NET_RX_BUSY_POLL
dafcc438 426 unsigned int sk_ll_usec;
9115e8cd
ED		 427 /* ===== mostly read cache line ===== */
428 unsigned int sk_napi_id;
b178bb3d 429#endif
b178bb3d
ED		 430 int sk_rcvbuf;
431
432 struct sk_filter __rcu *sk_filter;
ceb5d58b
ED		 433 union {
434 struct socket_wq __rcu *sk_wq;
66256e0b 435 /* private: */
ceb5d58b 436 struct socket_wq *sk_wq_raw;
66256e0b 437 /* public: */
ceb5d58b 438 };
def8b4fa 439#ifdef CONFIG_XFRM
d188ba86 440 struct xfrm_policy __rcu *sk_policy[2];
def8b4fa 441#endif
0c0a5ef8 442
0e36cbb3 443 struct dst_entry __rcu *sk_dst_cache;
1da177e4 444 atomic_t sk_omem_alloc;
4e07a91c 445 int sk_sndbuf;
9115e8cd
ED		 446
447 /* ===== cache line for TX ===== */
448 int sk_wmem_queued;
14afee4b 449 refcount_t sk_wmem_alloc;
9115e8cd 450 unsigned long sk_tsq_flags;
75c119af
ED		 451 union {
452 struct sk_buff *sk_send_head;
453 struct rb_root tcp_rtx_queue;
454 };
1da177e4 455 struct sk_buff_head sk_write_queue;
9115e8cd
ED		 456 __s32 sk_peek_off;
457 int sk_write_pending;
9b8805a3 458 __u32 sk_dst_pending_confirm;
218af599 459 u32 sk_pacing_status; /* see enum sk_pacing */
9115e8cd
ED		 460 long sk_sndtimeo;
461 struct timer_list sk_timer;
462 __u32 sk_priority;
463 __u32 sk_mark;
76a9ebe8
ED		 464 unsigned long sk_pacing_rate; /* bytes per second */
465 unsigned long sk_max_pacing_rate;
9115e8cd
ED		 466 struct page_frag sk_frag;
467 netdev_features_t sk_route_caps;
9115e8cd
ED		 468 int sk_gso_type;
469 unsigned int sk_gso_max_size;
470 gfp_t sk_allocation;
471 __u32 sk_txhash;
fc64869c
AR		 472
473 /*
474 * Because of non atomicity rules, all
475 * changes are protected by socket lock.
476 */
aba54656 477 u8 sk_gso_disabled : 1,
cdfbabfb 478 sk_kern_sock : 1,
28448b80
TH		 479 sk_no_check_tx : 1,
480 sk_no_check_rx : 1,
bf976514 481 sk_userlocks : 4;
3a9b76fd 482 u8 sk_pacing_shift;
bf976514
MM		 483 u16 sk_type;
484 u16 sk_protocol;
485 u16 sk_gso_max_segs;
1da177e4 486 unsigned long sk_lingertime;
476e19cf 487 struct proto *sk_prot_creator;
1da177e4
LT		 488 rwlock_t sk_callback_lock;
489 int sk_err,
490 sk_err_soft;
becb74f0
ED		 491 u32 sk_ack_backlog;
492 u32 sk_max_ack_backlog;
86741ec2 493 kuid_t sk_uid;
7fd3253a
BT		 494#ifdef CONFIG_NET_RX_BUSY_POLL
495 u8 sk_prefer_busy_poll;
7c951caf 496 u16 sk_busy_poll_budget;
7fd3253a 497#endif
35306eb2 498 spinlock_t sk_peer_lock;
1ace2b4d 499 int sk_bind_phc;
109f6e39
EB		 500 struct pid *sk_peer_pid;
501 const struct cred *sk_peer_cred;
35306eb2 502
1da177e4 503 long sk_rcvtimeo;
b7aa0bf7 504 ktime_t sk_stamp;
3a0ed3e9
DD		 505#if BITS_PER_LONG==32
506 seqlock_t sk_stamp_seq;
507#endif
b9f40e21 508 u16 sk_tsflags;
fc64869c 509 u8 sk_shutdown;
09c2d251 510 u32 sk_tskey;
52267790 511 atomic_t sk_zckey;
80b14dee
RC		 512
513 u8 sk_clockid;
514 u8 sk_txtime_deadline_mode : 1,
4b15c707
JSP		 515 sk_txtime_report_errors : 1,
516 sk_txtime_unused : 6;
80b14dee 517
1da177e4
LT		 518 struct socket *sk_socket;
519 void *sk_user_data;
d5f64238 520#ifdef CONFIG_SECURITY
1da177e4 521 void *sk_security;
d5f64238 522#endif
2a56a1fe 523 struct sock_cgroup_data sk_cgrp_data;
baac50bb 524 struct mem_cgroup *sk_memcg;
1da177e4 525 void (*sk_state_change)(struct sock *sk);
676d2369 526 void (*sk_data_ready)(struct sock *sk);
1da177e4
LT		 527 void (*sk_write_space)(struct sock *sk);
528 void (*sk_error_report)(struct sock *sk);
dc6b9b78
ED		 529 int (*sk_backlog_rcv)(struct sock *sk,
530 struct sk_buff *skb);
ebf4e808
IL		 531#ifdef CONFIG_SOCK_VALIDATE_XMIT
532 struct sk_buff* (*sk_validate_xmit_skb)(struct sock *sk,
533 struct net_device *dev,
534 struct sk_buff *skb);
535#endif
1da177e4 536 void (*sk_destruct)(struct sock *sk);
ef456144 537 struct sock_reuseport __rcu *sk_reuseport_cb;
6ac99e8f 538#ifdef CONFIG_BPF_SYSCALL
1f00d375 539 struct bpf_local_storage __rcu *sk_bpf_storage;
6ac99e8f 540#endif
a4298e45 541 struct rcu_head sk_rcu;
ffa84b5f 542 netns_tracker ns_tracker;
1da177e4
LT		 543};
544
218af599
ED		 545enum sk_pacing {
546 SK_PACING_NONE = 0,
547 SK_PACING_NEEDED = 1,
548 SK_PACING_FQ = 2,
549};
550
f1ff5ce2
JS		 551/* Pointer stored in sk_user_data might not be suitable for copying
552 * when cloning the socket. For instance, it can point to a reference
553 * counted object. sk_user_data bottom bit is set if pointer must not
554 * be copied.
555 */
556#define SK_USER_DATA_NOCOPY 1UL
c9a368f1
MKL		 557#define SK_USER_DATA_BPF 2UL /* Managed by BPF */
558#define SK_USER_DATA_PTRMASK ~(SK_USER_DATA_NOCOPY | SK_USER_DATA_BPF)
f1ff5ce2
JS		 559
560/**
561 * sk_user_data_is_nocopy - Test if sk_user_data pointer must not be copied
562 * @sk: socket
563 */
564static inline bool sk_user_data_is_nocopy(const struct sock *sk)
565{
566 return ((uintptr_t)sk->sk_user_data & SK_USER_DATA_NOCOPY);
567}
568
559835ea
PS		 569#define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
570
f1ff5ce2
JS		 571#define rcu_dereference_sk_user_data(sk) \
572({ \
573 void *__tmp = rcu_dereference(__sk_user_data((sk))); \
574 (void *)((uintptr_t)__tmp & SK_USER_DATA_PTRMASK); \
575})
576#define rcu_assign_sk_user_data(sk, ptr) \
577({ \
578 uintptr_t __tmp = (uintptr_t)(ptr); \
579 WARN_ON_ONCE(__tmp & ~SK_USER_DATA_PTRMASK); \
580 rcu_assign_pointer(__sk_user_data((sk)), __tmp); \
581})
582#define rcu_assign_sk_user_data_nocopy(sk, ptr) \
583({ \
584 uintptr_t __tmp = (uintptr_t)(ptr); \
585 WARN_ON_ONCE(__tmp & ~SK_USER_DATA_PTRMASK); \
586 rcu_assign_pointer(__sk_user_data((sk)), \
587 __tmp | SK_USER_DATA_NOCOPY); \
588})
559835ea 589
4a17fd52
PE		 590/*
591 * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
592 * or not whether his port will be reused by someone else. SK_FORCE_REUSE
593 * on a socket means that the socket will reuse everybody else's port
594 * without looking at the other's sk_reuse value.
595 */
596
597#define SK_NO_REUSE 0
598#define SK_CAN_REUSE 1
599#define SK_FORCE_REUSE 2
600
627d2d6b 601int sk_set_peek_off(struct sock *sk, int val);
602
ef64a54f
PE		 603static inline int sk_peek_offset(struct sock *sk, int flags)
604{
b9bb53f3 605 if (unlikely(flags & MSG_PEEK)) {
a0917e0b 606 return READ_ONCE(sk->sk_peek_off);
b9bb53f3
WB		 607 }
608
609 return 0;
ef64a54f
PE		 610}
611
612static inline void sk_peek_offset_bwd(struct sock *sk, int val)
613{
b9bb53f3
WB		 614 s32 off = READ_ONCE(sk->sk_peek_off);
615
616 if (unlikely(off >= 0)) {
617 off = max_t(s32, off - val, 0);
618 WRITE_ONCE(sk->sk_peek_off, off);
ef64a54f
PE		 619 }
620}
621
622static inline void sk_peek_offset_fwd(struct sock *sk, int val)
623{
b9bb53f3 624 sk_peek_offset_bwd(sk, -val);
ef64a54f
PE		 625}
626
1da177e4
LT		 627/*
628 * Hashed lists helper routines
629 */
c4146644
LZ		 630static inline struct sock *sk_entry(const struct hlist_node *node)
631{
632 return hlist_entry(node, struct sock, sk_node);
633}
634
e48c414e 635static inline struct sock *__sk_head(const struct hlist_head *head)
1da177e4
LT		 636{
637 return hlist_entry(head->first, struct sock, sk_node);
638}
639
e48c414e 640static inline struct sock *sk_head(const struct hlist_head *head)
1da177e4
LT		 641{
642 return hlist_empty(head) ? NULL : __sk_head(head);
643}
644
88ab1932
ED		 645static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
646{
647 return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
648}
649
650static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
651{
652 return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
653}
654
e48c414e 655static inline struct sock *sk_next(const struct sock *sk)
1da177e4 656{
6c59ebd3 657 return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node);
1da177e4
LT		 658}
659
88ab1932
ED		 660static inline struct sock *sk_nulls_next(const struct sock *sk)
661{
662 return (!is_a_nulls(sk->sk_nulls_node.next)) ?
663 hlist_nulls_entry(sk->sk_nulls_node.next,
664 struct sock, sk_nulls_node) :
665 NULL;
666}
667
dc6b9b78 668static inline bool sk_unhashed(const struct sock *sk)
1da177e4
LT		 669{
670 return hlist_unhashed(&sk->sk_node);
671}
672
dc6b9b78 673static inline bool sk_hashed(const struct sock *sk)
1da177e4 674{
da753bea 675 return !sk_unhashed(sk);
1da177e4
LT		 676}
677
dc6b9b78 678static inline void sk_node_init(struct hlist_node *node)
1da177e4
LT		 679{
680 node->pprev = NULL;
681}
682
dc6b9b78 683static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
88ab1932
ED		 684{
685 node->pprev = NULL;
686}
687
dc6b9b78 688static inline void __sk_del_node(struct sock *sk)
1da177e4
LT		 689{
690 __hlist_del(&sk->sk_node);
691}
692
808f5114 693/* NB: equivalent to hlist_del_init_rcu */
dc6b9b78 694static inline bool __sk_del_node_init(struct sock *sk)
1da177e4
LT		 695{
696 if (sk_hashed(sk)) {
697 __sk_del_node(sk);
698 sk_node_init(&sk->sk_node);
dc6b9b78 699 return true;
1da177e4 700 }
dc6b9b78 701 return false;
1da177e4
LT		 702}
703
704/* Grab socket reference count. This operation is valid only
705 when sk is ALREADY grabbed f.e. it is found in hash table
706 or a list and the lookup is made under lock preventing hash table
707 modifications.
708 */
709
f9a7cbbf 710static __always_inline void sock_hold(struct sock *sk)
1da177e4 711{
41c6d650 712 refcount_inc(&sk->sk_refcnt);
1da177e4
LT		 713}
714
715/* Ungrab socket in the context, which assumes that socket refcnt
716 cannot hit zero, f.e. it is true in context of any socketcall.
717 */
f9a7cbbf 718static __always_inline void __sock_put(struct sock *sk)
1da177e4 719{
41c6d650 720 refcount_dec(&sk->sk_refcnt);
1da177e4
LT		 721}
722
dc6b9b78 723static inline bool sk_del_node_init(struct sock *sk)
1da177e4 724{
dc6b9b78 725 bool rc = __sk_del_node_init(sk);
1da177e4
LT		 726
727 if (rc) {
728 /* paranoid for a while -acme */
41c6d650 729 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
1da177e4
LT		 730 __sock_put(sk);
731 }
732 return rc;
733}
808f5114 734#define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
1da177e4 735
dc6b9b78 736static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7
ED		 737{
738 if (sk_hashed(sk)) {
88ab1932 739 hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
dc6b9b78 740 return true;
271b72c7 741 }
dc6b9b78 742 return false;
271b72c7
ED		 743}
744
dc6b9b78 745static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
271b72c7 746{
dc6b9b78 747 bool rc = __sk_nulls_del_node_init_rcu(sk);
271b72c7
ED		 748
749 if (rc) {
750 /* paranoid for a while -acme */
41c6d650 751 WARN_ON(refcount_read(&sk->sk_refcnt) == 1);
271b72c7
ED		 752 __sock_put(sk);
753 }
754 return rc;
755}
756
dc6b9b78 757static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT		 758{
759 hlist_add_head(&sk->sk_node, list);
760}
761
dc6b9b78 762static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
1da177e4
LT		 763{
764 sock_hold(sk);
765 __sk_add_node(sk, list);
766}
767
dc6b9b78 768static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
808f5114 769{
770 sock_hold(sk);
d296ba60
CG		 771 if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport &&
772 sk->sk_family == AF_INET6)
773 hlist_add_tail_rcu(&sk->sk_node, list);
774 else
775 hlist_add_head_rcu(&sk->sk_node, list);
808f5114 776}
777
a4dc6a49
MC		 778static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list)
779{
780 sock_hold(sk);
781 hlist_add_tail_rcu(&sk->sk_node, list);
782}
783
dc6b9b78 784static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7 785{
d7efc6c1 786 hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
271b72c7
ED		 787}
788
8dbd76e7
ED		 789static inline void __sk_nulls_add_node_tail_rcu(struct sock *sk, struct hlist_nulls_head *list)
790{
791 hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list);
792}
793
dc6b9b78 794static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
271b72c7
ED		 795{
796 sock_hold(sk);
88ab1932 797 __sk_nulls_add_node_rcu(sk, list);
271b72c7
ED		 798}
799
dc6b9b78 800static inline void __sk_del_bind_node(struct sock *sk)
1da177e4
LT		 801{
802 __hlist_del(&sk->sk_bind_node);
803}
804
dc6b9b78 805static inline void sk_add_bind_node(struct sock *sk,
1da177e4
LT		 806 struct hlist_head *list)
807{
808 hlist_add_head(&sk->sk_bind_node, list);
809}
810
b67bfe0d
SL		 811#define sk_for_each(__sk, list) \
812 hlist_for_each_entry(__sk, list, sk_node)
813#define sk_for_each_rcu(__sk, list) \
814 hlist_for_each_entry_rcu(__sk, list, sk_node)
88ab1932
ED		 815#define sk_nulls_for_each(__sk, node, list) \
816 hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
817#define sk_nulls_for_each_rcu(__sk, node, list) \
818 hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
b67bfe0d
SL		 819#define sk_for_each_from(__sk) \
820 hlist_for_each_entry_from(__sk, sk_node)
88ab1932
ED		 821#define sk_nulls_for_each_from(__sk, node) \
822 if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
823 hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
b67bfe0d
SL		 824#define sk_for_each_safe(__sk, tmp, list) \
825 hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
826#define sk_for_each_bound(__sk, list) \
827 hlist_for_each_entry(__sk, list, sk_bind_node)
1da177e4 828
2dc41cff 829/**
ca065d0c 830 * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset
2dc41cff
DH		 831 * @tpos: the type * to use as a loop cursor.
832 * @pos: the &struct hlist_node to use as a loop cursor.
833 * @head: the head for your list.
834 * @offset: offset of hlist_node within the struct.
835 *
836 */
ca065d0c 837#define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \
b6f4f848 838 for (pos = rcu_dereference(hlist_first_rcu(head)); \
ca065d0c 839 pos != NULL && \
2dc41cff 840 ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \
b6f4f848 841 pos = rcu_dereference(hlist_next_rcu(pos)))
2dc41cff 842
c336d148
EB		 843static inline struct user_namespace *sk_user_ns(struct sock *sk)
844{
845 /* Careful only use this in a context where these parameters
846 * can not change and must all be valid, such as recvmsg from
847 * userspace.
848 */
849 return sk->sk_socket->file->f_cred->user_ns;
850}
851
1da177e4
LT		 852/* Sock flags */
853enum sock_flags {
854 SOCK_DEAD,
855 SOCK_DONE,
856 SOCK_URGINLINE,
857 SOCK_KEEPOPEN,
858 SOCK_LINGER,
859 SOCK_DESTROY,
860 SOCK_BROADCAST,
861 SOCK_TIMESTAMP,
862 SOCK_ZAPPED,
863 SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
864 SOCK_DBG, /* %SO_DEBUG setting */
865 SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
92f37fd2 866 SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
1da177e4 867 SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
7cb02404 868 SOCK_MEMALLOC, /* VM depends on this socket for swapping */
20d49473 869 SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
bcdce719 870 SOCK_FASYNC, /* fasync() active */
3b885787 871 SOCK_RXQ_OVFL,
1cdebb42 872 SOCK_ZEROCOPY, /* buffers from userspace */
6e3e939f 873 SOCK_WIFI_STATUS, /* push wifi status to userspace */
3bdc0eba
BG		 874 SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
875 * Will use last 4 bytes of packet sent from
876 * user-space instead.
877 */
d59577b6 878 SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
7d4c04fc 879 SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
a4298e45 880 SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */
80b14dee 881 SOCK_TXTIME,
e4a2a304 882 SOCK_XDP, /* XDP is attached */
887feae3 883 SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */
1da177e4
LT		 884};
885
01ce63c9
MRL		 886#define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE))
887
53b924b3
RB		 888static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
889{
890 nsk->sk_flags = osk->sk_flags;
891}
892
1da177e4
LT		 893static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
894{
895 __set_bit(flag, &sk->sk_flags);
896}
897
898static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
899{
900 __clear_bit(flag, &sk->sk_flags);
901}
902
dfde1d7d
DY		 903static inline void sock_valbool_flag(struct sock *sk, enum sock_flags bit,
904 int valbool)
905{
906 if (valbool)
907 sock_set_flag(sk, bit);
908 else
909 sock_reset_flag(sk, bit);
910}
911
1b23a5df 912static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
1da177e4
LT		 913{
914 return test_bit(flag, &sk->sk_flags);
915}
916
c93bdd0e 917#ifdef CONFIG_NET
a7950ae8 918DECLARE_STATIC_KEY_FALSE(memalloc_socks_key);
c93bdd0e
MG		 919static inline int sk_memalloc_socks(void)
920{
a7950ae8 921 return static_branch_unlikely(&memalloc_socks_key);
c93bdd0e 922}
d9539752
KC		 923
924void __receive_sock(struct file *file);
c93bdd0e
MG		 925#else
926
927static inline int sk_memalloc_socks(void)
928{
929 return 0;
930}
931
d9539752
KC		 932static inline void __receive_sock(struct file *file)
933{ }
c93bdd0e
MG		 934#endif
935
7450aaf6 936static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask)
99a1dec7 937{
7450aaf6 938 return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC);
99a1dec7
MG		 939}
940
1da177e4
LT		 941static inline void sk_acceptq_removed(struct sock *sk)
942{
288efe86 943 WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog - 1);
1da177e4
LT		 944}
945
946static inline void sk_acceptq_added(struct sock *sk)
947{
288efe86 948 WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog + 1);
1da177e4
LT		 949}
950
c609e6aa
ED		 951/* Note: If you think the test should be:
952 * return READ_ONCE(sk->sk_ack_backlog) >= READ_ONCE(sk->sk_max_ack_backlog);
953 * Then please take a look at commit 64a146513f8f ("[NET]: Revert incorrect accept queue backlog changes.")
954 */
dc6b9b78 955static inline bool sk_acceptq_is_full(const struct sock *sk)
1da177e4 956{
c609e6aa 957 return READ_ONCE(sk->sk_ack_backlog) > READ_ONCE(sk->sk_max_ack_backlog);
1da177e4
LT		 958}
959
960/*
961 * Compute minimal free write space needed to queue new packets.
962 */
dc6b9b78 963static inline int sk_stream_min_wspace(const struct sock *sk)
1da177e4 964{
ab4e846a 965 return READ_ONCE(sk->sk_wmem_queued) >> 1;
1da177e4
LT		 966}
967
dc6b9b78 968static inline int sk_stream_wspace(const struct sock *sk)
1da177e4 969{
ab4e846a
ED		 970 return READ_ONCE(sk->sk_sndbuf) - READ_ONCE(sk->sk_wmem_queued);
971}
972
973static inline void sk_wmem_queued_add(struct sock *sk, int val)
974{
975 WRITE_ONCE(sk->sk_wmem_queued, sk->sk_wmem_queued + val);
1da177e4
LT		 976}
977
69336bd2 978void sk_stream_write_space(struct sock *sk);
1da177e4 979
8eae939f 980/* OOB backlog add */
a3a858ff 981static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
9ee6b535 982{
7fee226a 983 /* dont let skb dst not refcounted, we are going to leave rcu lock */
222d7dbd 984 skb_dst_force(skb);
7fee226a
ED		 985
986 if (!sk->sk_backlog.tail)
9ed498c6 987 WRITE_ONCE(sk->sk_backlog.head, skb);
7fee226a 988 else
9ee6b535 989 sk->sk_backlog.tail->next = skb;
7fee226a 990
9ed498c6 991 WRITE_ONCE(sk->sk_backlog.tail, skb);
9ee6b535
SH		 992 skb->next = NULL;
993}
1da177e4 994
c377411f
ED		 995/*
996 * Take into account size of receive queue and backlog queue
0fd7bac6
ED		 997 * Do not take into account this skb truesize,
998 * to allow even a single big packet to come.
c377411f 999 */
274f482d 1000static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit)
c377411f
ED		 1001{
1002 unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
1003
f545a38f 1004 return qsize > limit;
c377411f
ED		 1005}
1006
8eae939f 1007/* The per-socket spinlock must be held here. */
f545a38f
ED		 1008static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
1009 unsigned int limit)
8eae939f 1010{
274f482d 1011 if (sk_rcvqueues_full(sk, limit))
8eae939f
ZY		 1012 return -ENOBUFS;
1013
c7c49b8f
ED		 1014 /*
1015 * If the skb was allocated from pfmemalloc reserves, only
1016 * allow SOCK_MEMALLOC sockets to use it as this socket is
1017 * helping free memory
1018 */
1019 if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC))
1020 return -ENOMEM;
1021
a3a858ff 1022 __sk_add_backlog(sk, skb);
8eae939f
ZY		 1023 sk->sk_backlog.len += skb->truesize;
1024 return 0;
1025}
1026
69336bd2 1027int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
b4b9e355 1028
d2489c7b
ED		 1029INDIRECT_CALLABLE_DECLARE(int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb));
1030INDIRECT_CALLABLE_DECLARE(int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb));
1031
c57943a1
PZ		 1032static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
1033{
b4b9e355
MG		 1034 if (sk_memalloc_socks() && skb_pfmemalloc(skb))
1035 return __sk_backlog_rcv(sk, skb);
1036
d2489c7b
ED		 1037 return INDIRECT_CALL_INET(sk->sk_backlog_rcv,
1038 tcp_v6_do_rcv,
1039 tcp_v4_do_rcv,
1040 sk, skb);
c57943a1
PZ		 1041}
1042
2c8c56e1
ED		 1043static inline void sk_incoming_cpu_update(struct sock *sk)
1044{
34cfb542
PA		 1045 int cpu = raw_smp_processor_id();
1046
7170a977
ED		 1047 if (unlikely(READ_ONCE(sk->sk_incoming_cpu) != cpu))
1048 WRITE_ONCE(sk->sk_incoming_cpu, cpu);
2c8c56e1
ED		 1049}
1050
fe477558 1051static inline void sock_rps_record_flow_hash(__u32 hash)
c58dc01b
DM		 1052{
1053#ifdef CONFIG_RPS
1054 struct rps_sock_flow_table *sock_flow_table;
1055
1056 rcu_read_lock();
1057 sock_flow_table = rcu_dereference(rps_sock_flow_table);
fe477558 1058 rps_record_sock_flow(sock_flow_table, hash);
c58dc01b
DM		 1059 rcu_read_unlock();
1060#endif
1061}
1062
fe477558
TH		 1063static inline void sock_rps_record_flow(const struct sock *sk)
1064{
c9d8ca04 1065#ifdef CONFIG_RPS
dc05360f 1066 if (static_branch_unlikely(&rfs_needed)) {
13bfff25
ED		 1067 /* Reading sk->sk_rxhash might incur an expensive cache line
1068 * miss.
1069 *
1070 * TCP_ESTABLISHED does cover almost all states where RFS
1071 * might be useful, and is cheaper [1] than testing :
1072 * IPv4: inet_sk(sk)->inet_daddr
1073 * IPv6: ipv6_addr_any(&sk->sk_v6_daddr)
1074 * OR an additional socket flag
1075 * [1] : sk_state and sk_prot are in the same cache line.
1076 */
1077 if (sk->sk_state == TCP_ESTABLISHED)
1078 sock_rps_record_flow_hash(sk->sk_rxhash);
1079 }
c9d8ca04 1080#endif
fe477558
TH		 1081}
1082
bdeab991
TH		 1083static inline void sock_rps_save_rxhash(struct sock *sk,
1084 const struct sk_buff *skb)
c58dc01b
DM		 1085{
1086#ifdef CONFIG_RPS
567e4b79 1087 if (unlikely(sk->sk_rxhash != skb->hash))
61b905da 1088 sk->sk_rxhash = skb->hash;
c58dc01b
DM		 1089#endif
1090}
1091
bdeab991
TH		 1092static inline void sock_rps_reset_rxhash(struct sock *sk)
1093{
1094#ifdef CONFIG_RPS
bdeab991
TH		 1095 sk->sk_rxhash = 0;
1096#endif
1097}
1098
d9dc8b0f 1099#define sk_wait_event(__sk, __timeo, __condition, __wait) \
cfcabdcc
SH		 1100 ({ int __rc; \
1101 release_sock(__sk); \
1102 __rc = __condition; \
1103 if (!__rc) { \
d9dc8b0f
WC		 1104 *(__timeo) = wait_woken(__wait, \
1105 TASK_INTERRUPTIBLE, \
1106 *(__timeo)); \
cfcabdcc 1107 } \
d9dc8b0f 1108 sched_annotate_sleep(); \
cfcabdcc
SH		 1109 lock_sock(__sk); \
1110 __rc = __condition; \
1111 __rc; \
1112 })
1da177e4 1113
69336bd2
JP		 1114int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
1115int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
1116void sk_stream_wait_close(struct sock *sk, long timeo_p);
1117int sk_stream_error(struct sock *sk, int flags, int err);
1118void sk_stream_kill_queues(struct sock *sk);
1119void sk_set_memalloc(struct sock *sk);
1120void sk_clear_memalloc(struct sock *sk);
1da177e4 1121
d41a69f1
ED		 1122void __sk_flush_backlog(struct sock *sk);
1123
1124static inline bool sk_flush_backlog(struct sock *sk)
1125{
1126 if (unlikely(READ_ONCE(sk->sk_backlog.tail))) {
1127 __sk_flush_backlog(sk);
1128 return true;
1129 }
1130 return false;
1131}
1132
dfbafc99 1133int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
1da177e4 1134
60236fdd 1135struct request_sock_ops;
6d6ee43e 1136struct timewait_sock_ops;
ab1e0a13 1137struct inet_hashinfo;
fc8717ba 1138struct raw_hashinfo;
f16a7dd5 1139struct smc_hashinfo;
de477254 1140struct module;
51e0158a 1141struct sk_psock;
2e6599cb 1142
f77d6021 1143/*
5f0d5a3a 1144 * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes
f77d6021
ED		 1145 * un-modified. Special care is taken when initializing object to zero.
1146 */
1147static inline void sk_prot_clear_nulls(struct sock *sk, int size)
1148{
1149 if (offsetof(struct sock, sk_node.next) != 0)
1150 memset(sk, 0, offsetof(struct sock, sk_node.next));
1151 memset(&sk->sk_node.pprev, 0,
1152 size - offsetof(struct sock, sk_node.pprev));
1153}
1154
1da177e4
LT		 1155/* Networking protocol blocks we attach to sockets.
1156 * socket layer -> transport layer interface
1da177e4
LT		 1157 */
1158struct proto {
dc6b9b78 1159 void (*close)(struct sock *sk,
1da177e4 1160 long timeout);
d74bad4e
AI		 1161 int (*pre_connect)(struct sock *sk,
1162 struct sockaddr *uaddr,
1163 int addr_len);
1da177e4 1164 int (*connect)(struct sock *sk,
dc6b9b78 1165 struct sockaddr *uaddr,
1da177e4
LT		 1166 int addr_len);
1167 int (*disconnect)(struct sock *sk, int flags);
1168
cdfbabfb
DH		 1169 struct sock * (*accept)(struct sock *sk, int flags, int *err,
1170 bool kern);
1da177e4
LT		 1171
1172 int (*ioctl)(struct sock *sk, int cmd,
1173 unsigned long arg);
1174 int (*init)(struct sock *sk);
7d06b2e0 1175 void (*destroy)(struct sock *sk);
1da177e4 1176 void (*shutdown)(struct sock *sk, int how);
dc6b9b78 1177 int (*setsockopt)(struct sock *sk, int level,
a7b75c5a 1178 int optname, sockptr_t optval,
b7058842 1179 unsigned int optlen);
dc6b9b78
ED		 1180 int (*getsockopt)(struct sock *sk, int level,
1181 int optname, char __user *optval,
1182 int __user *option);
4b9d07a4 1183 void (*keepalive)(struct sock *sk, int valbool);
af01d537 1184#ifdef CONFIG_COMPAT
709b46e8
EB		 1185 int (*compat_ioctl)(struct sock *sk,
1186 unsigned int cmd, unsigned long arg);
af01d537 1187#endif
1b784140
YX		 1188 int (*sendmsg)(struct sock *sk, struct msghdr *msg,
1189 size_t len);
1190 int (*recvmsg)(struct sock *sk, struct msghdr *msg,
dc6b9b78
ED		 1191 size_t len, int noblock, int flags,
1192 int *addr_len);
1da177e4
LT		 1193 int (*sendpage)(struct sock *sk, struct page *page,
1194 int offset, size_t size, int flags);
dc6b9b78 1195 int (*bind)(struct sock *sk,
c0425a42
CH		 1196 struct sockaddr *addr, int addr_len);
1197 int (*bind_add)(struct sock *sk,
1198 struct sockaddr *addr, int addr_len);
1da177e4 1199
dc6b9b78 1200 int (*backlog_rcv) (struct sock *sk,
1da177e4 1201 struct sk_buff *skb);
9cacf81f
SF		 1202 bool (*bpf_bypass_getsockopt)(int level,
1203 int optname);
1da177e4 1204
46d3ceab
ED		 1205 void (*release_cb)(struct sock *sk);
1206
1da177e4 1207 /* Keeping track of sk's, looking them up, and port selection methods. */
086c653f 1208 int (*hash)(struct sock *sk);
1da177e4 1209 void (*unhash)(struct sock *sk);
719f8358 1210 void (*rehash)(struct sock *sk);
1da177e4 1211 int (*get_port)(struct sock *sk, unsigned short snum);
91a760b2 1212 void (*put_port)(struct sock *sk);
8a59f9d1 1213#ifdef CONFIG_BPF_SYSCALL
51e0158a
CW		 1214 int (*psock_update_sk_prot)(struct sock *sk,
1215 struct sk_psock *psock,
1216 bool restore);
8a59f9d1 1217#endif
1da177e4 1218
286ab3d4 1219 /* Keeping track of sockets in use */
65f76517 1220#ifdef CONFIG_PROC_FS
13ff3d6f 1221 unsigned int inuse_idx;
65f76517 1222#endif
ebb53d75 1223
6c302e79 1224#if IS_ENABLED(CONFIG_MPTCP)
292e6077 1225 int (*forward_alloc_get)(const struct sock *sk);
6c302e79 1226#endif
292e6077 1227
a74f0fa0 1228 bool (*stream_memory_free)(const struct sock *sk, int wake);
7b50ecfc 1229 bool (*sock_is_readable)(struct sock *sk);
1da177e4 1230 /* Memory pressure */
5c52ba17 1231 void (*enter_memory_pressure)(struct sock *sk);
06044751 1232 void (*leave_memory_pressure)(struct sock *sk);
8d987e5c 1233 atomic_long_t *memory_allocated; /* Current allocated memory. */
1748376b 1234 struct percpu_counter *sockets_allocated; /* Current number of sockets. */
292e6077 1235
1da177e4
LT		 1236 /*
1237 * Pressure flag: try to collapse.
1238 * Technical note: it is used by multiple contexts non atomically.
3ab224be 1239 * All the __sk_mem_schedule() is of this nature: accounting
1da177e4
LT		 1240 * is strict, actions are advisory and have some latency.
1241 */
06044751 1242 unsigned long *memory_pressure;
8d987e5c 1243 long *sysctl_mem;
a3dcaf17 1244
1da177e4
LT		 1245 int *sysctl_wmem;
1246 int *sysctl_rmem;
a3dcaf17
ED		 1247 u32 sysctl_wmem_offset;
1248 u32 sysctl_rmem_offset;
1249
1da177e4 1250 int max_header;
7ba42910 1251 bool no_autobind;
1da177e4 1252
271b72c7 1253 struct kmem_cache *slab;
1da177e4 1254 unsigned int obj_size;
d50112ed 1255 slab_flags_t slab_flags;
7bbdb81e
AD		 1256 unsigned int useroffset; /* Usercopy region offset */
1257 unsigned int usersize; /* Usercopy region size */
1da177e4 1258
19757ceb 1259 unsigned int __percpu *orphan_count;
8feaf0c0 1260
60236fdd 1261 struct request_sock_ops *rsk_prot;
6d6ee43e 1262 struct timewait_sock_ops *twsk_prot;
2e6599cb 1263
39d8cda7
PE		 1264 union {
1265 struct inet_hashinfo *hashinfo;
645ca708 1266 struct udp_table *udp_table;
fc8717ba 1267 struct raw_hashinfo *raw_hash;
f16a7dd5 1268 struct smc_hashinfo *smc_hash;
39d8cda7 1269 } h;
ab1e0a13 1270
1da177e4
LT		 1271 struct module *owner;
1272
1273 char name[32];
1274
1275 struct list_head node;
e6848976
ACM		 1276#ifdef SOCK_REFCNT_DEBUG
1277 atomic_t socks;
e1aab161 1278#endif
64be0aed 1279 int (*diag_destroy)(struct sock *sk, int err);
3859a271 1280} __randomize_layout;
e1aab161 1281
69336bd2
JP		 1282int proto_register(struct proto *prot, int alloc_slab);
1283void proto_unregister(struct proto *prot);
bf2ae2e4 1284int sock_load_diag_module(int family, int protocol);
1da177e4 1285
e6848976
ACM		 1286#ifdef SOCK_REFCNT_DEBUG
1287static inline void sk_refcnt_debug_inc(struct sock *sk)
1288{
1289 atomic_inc(&sk->sk_prot->socks);
1290}
1291
1292static inline void sk_refcnt_debug_dec(struct sock *sk)
1293{
1294 atomic_dec(&sk->sk_prot->socks);
1295 printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
1296 sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
1297}
1298
dec34fb0 1299static inline void sk_refcnt_debug_release(const struct sock *sk)
e6848976 1300{
41c6d650 1301 if (refcount_read(&sk->sk_refcnt) != 1)
e6848976 1302 printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
41c6d650 1303 sk->sk_prot->name, sk, refcount_read(&sk->sk_refcnt));
e6848976
ACM		 1304}
1305#else /* SOCK_REFCNT_DEBUG */
1306#define sk_refcnt_debug_inc(sk) do { } while (0)
1307#define sk_refcnt_debug_dec(sk) do { } while (0)
1308#define sk_refcnt_debug_release(sk) do { } while (0)
1309#endif /* SOCK_REFCNT_DEBUG */
1310
1c5f2ced
ED		 1311INDIRECT_CALLABLE_DECLARE(bool tcp_stream_memory_free(const struct sock *sk, int wake));
1312
292e6077
PA		 1313static inline int sk_forward_alloc_get(const struct sock *sk)
1314{
6c302e79
ED		 1315#if IS_ENABLED(CONFIG_MPTCP)
1316 if (sk->sk_prot->forward_alloc_get)
1317 return sk->sk_prot->forward_alloc_get(sk);
1318#endif
1319 return sk->sk_forward_alloc;
292e6077
PA		 1320}
1321
a74f0fa0 1322static inline bool __sk_stream_memory_free(const struct sock *sk, int wake)
c9bee3b7 1323{
ab4e846a 1324 if (READ_ONCE(sk->sk_wmem_queued) >= READ_ONCE(sk->sk_sndbuf))
c9bee3b7
ED		 1325 return false;
1326
1327 return sk->sk_prot->stream_memory_free ?
a406290a
ED		 1328 INDIRECT_CALL_INET_1(sk->sk_prot->stream_memory_free,
1329 tcp_stream_memory_free, sk, wake) : true;
c9bee3b7
ED		 1330}
1331
a74f0fa0
ED		 1332static inline bool sk_stream_memory_free(const struct sock *sk)
1333{
1334 return __sk_stream_memory_free(sk, 0);
1335}
1336
1337static inline bool __sk_stream_is_writeable(const struct sock *sk, int wake)
64dc6130 1338{
c9bee3b7 1339 return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
a74f0fa0
ED		 1340 __sk_stream_memory_free(sk, wake);
1341}
1342
1343static inline bool sk_stream_is_writeable(const struct sock *sk)
1344{
1345 return __sk_stream_is_writeable(sk, 0);
64dc6130 1346}
e1aab161 1347
54fd9c2d
DB		 1348static inline int sk_under_cgroup_hierarchy(struct sock *sk,
1349 struct cgroup *ancestor)
1350{
1351#ifdef CONFIG_SOCK_CGROUP_DATA
1352 return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data),
1353 ancestor);
1354#else
1355 return -ENOTSUPP;
1356#endif
1357}
c9bee3b7 1358
180d8cd9
GC		 1359static inline bool sk_has_memory_pressure(const struct sock *sk)
1360{
1361 return sk->sk_prot->memory_pressure != NULL;
1362}
1363
1364static inline bool sk_under_memory_pressure(const struct sock *sk)
1365{
1366 if (!sk->sk_prot->memory_pressure)
1367 return false;
e1aab161 1368
baac50bb
JW		 1369 if (mem_cgroup_sockets_enabled && sk->sk_memcg &&
1370 mem_cgroup_under_socket_pressure(sk->sk_memcg))
e805605c 1371 return true;
e1aab161 1372
35b87f6c 1373 return !!*sk->sk_prot->memory_pressure;
180d8cd9
GC		 1374}
1375
180d8cd9
GC		 1376static inline long
1377sk_memory_allocated(const struct sock *sk)
1378{
e805605c 1379 return atomic_long_read(sk->sk_prot->memory_allocated);
180d8cd9
GC		 1380}
1381
1382static inline long
e805605c 1383sk_memory_allocated_add(struct sock *sk, int amt)
180d8cd9 1384{
e805605c 1385 return atomic_long_add_return(amt, sk->sk_prot->memory_allocated);
180d8cd9
GC		 1386}
1387
1388static inline void
0e90b31f 1389sk_memory_allocated_sub(struct sock *sk, int amt)
180d8cd9 1390{
e805605c 1391 atomic_long_sub(amt, sk->sk_prot->memory_allocated);
180d8cd9
GC		 1392}
1393
f5a5589c
WW		 1394#define SK_ALLOC_PERCPU_COUNTER_BATCH 16
1395
180d8cd9
GC		 1396static inline void sk_sockets_allocated_dec(struct sock *sk)
1397{
f5a5589c
WW		 1398 percpu_counter_add_batch(sk->sk_prot->sockets_allocated, -1,
1399 SK_ALLOC_PERCPU_COUNTER_BATCH);
180d8cd9
GC		 1400}
1401
1402static inline void sk_sockets_allocated_inc(struct sock *sk)
1403{
f5a5589c
WW		 1404 percpu_counter_add_batch(sk->sk_prot->sockets_allocated, 1,
1405 SK_ALLOC_PERCPU_COUNTER_BATCH);
180d8cd9
GC		 1406}
1407
5bf325a5 1408static inline u64
180d8cd9
GC		 1409sk_sockets_allocated_read_positive(struct sock *sk)
1410{
af95d7df 1411 return percpu_counter_read_positive(sk->sk_prot->sockets_allocated);
180d8cd9
GC		 1412}
1413
1414static inline int
1415proto_sockets_allocated_sum_positive(struct proto *prot)
1416{
1417 return percpu_counter_sum_positive(prot->sockets_allocated);
1418}
1419
1420static inline long
1421proto_memory_allocated(struct proto *prot)
1422{
1423 return atomic_long_read(prot->memory_allocated);
1424}
1425
1426static inline bool
1427proto_memory_pressure(struct proto *prot)
1428{
1429 if (!prot->memory_pressure)
1430 return false;
1431 return !!*prot->memory_pressure;
1432}
1433
65f76517
ED		 1434
1435#ifdef CONFIG_PROC_FS
2a12ae5d
ED		 1436#define PROTO_INUSE_NR 64 /* should be enough for the first time */
1437struct prot_inuse {
4199bae1 1438 int all;
2a12ae5d
ED		 1439 int val[PROTO_INUSE_NR];
1440};
b3cb764a 1441
2a12ae5d
ED		 1442static inline void sock_prot_inuse_add(const struct net *net,
1443 const struct proto *prot, int val)
1444{
b3cb764a 1445 this_cpu_add(net->core.prot_inuse->val[prot->inuse_idx], val);
2a12ae5d 1446}
d477eb90
ED		 1447
1448static inline void sock_inuse_add(const struct net *net, int val)
1449{
4199bae1 1450 this_cpu_add(net->core.prot_inuse->all, val);
d477eb90
ED		 1451}
1452
69336bd2 1453int sock_prot_inuse_get(struct net *net, struct proto *proto);
648845ab 1454int sock_inuse_get(struct net *net);
65f76517 1455#else
2a12ae5d
ED		 1456static inline void sock_prot_inuse_add(const struct net *net,
1457 const struct proto *prot, int val)
65f76517
ED		 1458{
1459}
d477eb90
ED		 1460
1461static inline void sock_inuse_add(const struct net *net, int val)
1462{
1463}
65f76517
ED		 1464#endif
1465
1da177e4 1466
614c6cb4
ACM		 1467/* With per-bucket locks this operation is not-atomic, so that
1468 * this version is not worse.
1469 */
086c653f 1470static inline int __sk_prot_rehash(struct sock *sk)
614c6cb4
ACM		 1471{
1472 sk->sk_prot->unhash(sk);
086c653f 1473 return sk->sk_prot->hash(sk);
614c6cb4
ACM		 1474}
1475
1da177e4
LT		 1476/* About 10 seconds */
1477#define SOCK_DESTROY_TIME (10*HZ)
1478
1479/* Sockets 0-1023 can't be bound to unless you are superuser */
1480#define PROT_SOCK 1024
1481
1482#define SHUTDOWN_MASK 3
1483#define RCV_SHUTDOWN 1
1484#define SEND_SHUTDOWN 2
1485
1da177e4
LT		 1486#define SOCK_BINDADDR_LOCK 4
1487#define SOCK_BINDPORT_LOCK 8
1488
1da177e4
LT		 1489struct socket_alloc {
1490 struct socket socket;
1491 struct inode vfs_inode;
1492};
1493
1494static inline struct socket *SOCKET_I(struct inode *inode)
1495{
1496 return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
1497}
1498
1499static inline struct inode *SOCK_INODE(struct socket *socket)
1500{
1501 return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
1502}
1503
3ab224be
HA		 1504/*
1505 * Functions for memory accounting
1506 */
f8c3bf00 1507int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind);
69336bd2 1508int __sk_mem_schedule(struct sock *sk, int size, int kind);
f8c3bf00 1509void __sk_mem_reduce_allocated(struct sock *sk, int amount);
1a24e04e 1510void __sk_mem_reclaim(struct sock *sk, int amount);
1da177e4 1511
bd68a2a8
ED		 1512/* We used to have PAGE_SIZE here, but systems with 64KB pages
1513 * do not necessarily have 16x time more memory than 4KB ones.
1514 */
1515#define SK_MEM_QUANTUM 4096
3ab224be
HA		 1516#define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
1517#define SK_MEM_SEND 0
1518#define SK_MEM_RECV 1
1da177e4 1519
bd68a2a8
ED		 1520/* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */
1521static inline long sk_prot_mem_limits(const struct sock *sk, int index)
1522{
1523 long val = sk->sk_prot->sysctl_mem[index];
1524
1525#if PAGE_SIZE > SK_MEM_QUANTUM
1526 val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT;
1527#elif PAGE_SIZE < SK_MEM_QUANTUM
1528 val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT;
1529#endif
1530 return val;
1531}
1532
3ab224be 1533static inline int sk_mem_pages(int amt)
1da177e4 1534{
3ab224be 1535 return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
1da177e4
LT		 1536}
1537
dc6b9b78 1538static inline bool sk_has_account(struct sock *sk)
1da177e4 1539{
3ab224be
HA		 1540 /* return true if protocol supports memory accounting */
1541 return !!sk->sk_prot->memory_allocated;
1da177e4
LT		 1542}
1543
dc6b9b78 1544static inline bool sk_wmem_schedule(struct sock *sk, int size)
1da177e4 1545{
3ab224be 1546 if (!sk_has_account(sk))
dc6b9b78 1547 return true;
3ab224be
HA		 1548 return size <= sk->sk_forward_alloc ||
1549 __sk_mem_schedule(sk, size, SK_MEM_SEND);
1da177e4
LT		 1550}
1551
c76562b6 1552static inline bool
35c448a8 1553sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
d80d99d6 1554{
3ab224be 1555 if (!sk_has_account(sk))
dc6b9b78 1556 return true;
5af68891 1557 return size <= sk->sk_forward_alloc ||
c76562b6
MG		 1558 __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
1559 skb_pfmemalloc(skb);
3ab224be
HA		 1560}
1561
2bb2f5fb
WW		 1562static inline int sk_unused_reserved_mem(const struct sock *sk)
1563{
1564 int unused_mem;
1565
1566 if (likely(!sk->sk_reserved_mem))
1567 return 0;
1568
1569 unused_mem = sk->sk_reserved_mem - sk->sk_wmem_queued -
1570 atomic_read(&sk->sk_rmem_alloc);
1571
1572 return unused_mem > 0 ? unused_mem : 0;
1573}
1574
3ab224be
HA		 1575static inline void sk_mem_reclaim(struct sock *sk)
1576{
2bb2f5fb
WW		 1577 int reclaimable;
1578
3ab224be
HA		 1579 if (!sk_has_account(sk))
1580 return;
2bb2f5fb
WW		 1581
1582 reclaimable = sk->sk_forward_alloc - sk_unused_reserved_mem(sk);
1583
1584 if (reclaimable >= SK_MEM_QUANTUM)
1585 __sk_mem_reclaim(sk, reclaimable);
1586}
1587
1588static inline void sk_mem_reclaim_final(struct sock *sk)
1589{
1590 sk->sk_reserved_mem = 0;
1591 sk_mem_reclaim(sk);
3ab224be
HA		 1592}
1593
9993e7d3
DM		 1594static inline void sk_mem_reclaim_partial(struct sock *sk)
1595{
2bb2f5fb
WW		 1596 int reclaimable;
1597
9993e7d3
DM		 1598 if (!sk_has_account(sk))
1599 return;
2bb2f5fb
WW		 1600
1601 reclaimable = sk->sk_forward_alloc - sk_unused_reserved_mem(sk);
1602
1603 if (reclaimable > SK_MEM_QUANTUM)
1604 __sk_mem_reclaim(sk, reclaimable - 1);
9993e7d3
DM		 1605}
1606
3ab224be
HA		 1607static inline void sk_mem_charge(struct sock *sk, int size)
1608{
1609 if (!sk_has_account(sk))
1610 return;
1611 sk->sk_forward_alloc -= size;
1612}
1613
5823fc96
PA		 1614/* the following macros control memory reclaiming in sk_mem_uncharge()
1615 */
1616#define SK_RECLAIM_THRESHOLD (1 << 21)
1617#define SK_RECLAIM_CHUNK (1 << 20)
1618
3ab224be
HA		 1619static inline void sk_mem_uncharge(struct sock *sk, int size)
1620{
2bb2f5fb
WW		 1621 int reclaimable;
1622
3ab224be
HA		 1623 if (!sk_has_account(sk))
1624 return;
1625 sk->sk_forward_alloc += size;
2bb2f5fb 1626 reclaimable = sk->sk_forward_alloc - sk_unused_reserved_mem(sk);
20c64d5c
ED		 1627
1628 /* Avoid a possible overflow.
1629 * TCP send queues can make this happen, if sk_mem_reclaim()
1630 * is not called and more than 2 GBytes are released at once.
1631 *
1632 * If we reach 2 MBytes, reclaim 1 MBytes right now, there is
1633 * no need to hold that much forward allocation anyway.
1634 */
5823fc96
PA		 1635 if (unlikely(reclaimable >= SK_RECLAIM_THRESHOLD))
1636 __sk_mem_reclaim(sk, SK_RECLAIM_CHUNK);
3ab224be
HA		 1637}
1638
ed07536e
PZ		 1639/*
1640 * Macro so as to not evaluate some arguments when
1641 * lockdep is not enabled.
1642 *
1643 * Mark both the sk_lock and the sk_lock.slock as a
1644 * per-address-family lock class.
1645 */
dc6b9b78 1646#define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
ed07536e 1647do { \
e8f6fbf6 1648 sk->sk_lock.owned = 0; \
ed07536e
PZ		 1649 init_waitqueue_head(&sk->sk_lock.wq); \
1650 spin_lock_init(&(sk)->sk_lock.slock); \
1651 debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
1652 sizeof((sk)->sk_lock)); \
1653 lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
dc6b9b78 1654 (skey), (sname)); \
ed07536e
PZ		 1655 lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
1656} while (0)
1657
05b93801 1658static inline bool lockdep_sock_is_held(const struct sock *sk)
1e1d04e6 1659{
1e1d04e6
HFS		 1660 return lockdep_is_held(&sk->sk_lock) ||
1661 lockdep_is_held(&sk->sk_lock.slock);
1662}
1663
69336bd2 1664void lock_sock_nested(struct sock *sk, int subclass);
fcc70d5f
PZ		 1665
1666static inline void lock_sock(struct sock *sk)
1667{
1668 lock_sock_nested(sk, 0);
1669}
1670
ad80b0fc 1671void __lock_sock(struct sock *sk);
8873c064 1672void __release_sock(struct sock *sk);
69336bd2 1673void release_sock(struct sock *sk);
1da177e4
LT		 1674
1675/* BH context may only use the following locking interface. */
1676#define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
c6366184
IM		 1677#define bh_lock_sock_nested(__sk) \
1678 spin_lock_nested(&((__sk)->sk_lock.slock), \
1679 SINGLE_DEPTH_NESTING)
1da177e4
LT		 1680#define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
1681
49054556
PA		 1682bool __lock_sock_fast(struct sock *sk) __acquires(&sk->sk_lock.slock);
1683
1684/**
1685 * lock_sock_fast - fast version of lock_sock
1686 * @sk: socket
1687 *
1688 * This version should be used for very small section, where process wont block
1689 * return false if fast path is taken:
1690 *
1691 * sk_lock.slock locked, owned = 0, BH disabled
1692 *
1693 * return true if slow path is taken:
1694 *
1695 * sk_lock.slock unlocked, owned = 1, BH enabled
1696 */
1697static inline bool lock_sock_fast(struct sock *sk)
1698{
1699 /* The sk_lock has mutex_lock() semantics here. */
1700 mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_);
1701
1702 return __lock_sock_fast(sk);
1703}
1704
1705/* fast socket lock variant for caller already holding a [different] socket lock */
1706static inline bool lock_sock_fast_nested(struct sock *sk)
1707{
1708 mutex_acquire(&sk->sk_lock.dep_map, SINGLE_DEPTH_NESTING, 0, _RET_IP_);
1709
1710 return __lock_sock_fast(sk);
1711}
12f4bd86 1712
8a74ad60
ED		 1713/**
1714 * unlock_sock_fast - complement of lock_sock_fast
1715 * @sk: socket
1716 * @slow: slow mode
1717 *
1718 * fast unlock socket for user context.
1719 * If slow mode is on, we call regular release_sock()
1720 */
1721static inline void unlock_sock_fast(struct sock *sk, bool slow)
12f4bd86 1722 __releases(&sk->sk_lock.slock)
4b0b72f7 1723{
12f4bd86 1724 if (slow) {
8a74ad60 1725 release_sock(sk);
12f4bd86
PA		 1726 __release(&sk->sk_lock.slock);
1727 } else {
2dcb96ba 1728 mutex_release(&sk->sk_lock.dep_map, _RET_IP_);
8a74ad60 1729 spin_unlock_bh(&sk->sk_lock.slock);
12f4bd86 1730 }
4b0b72f7
ED		 1731}
1732
fafc4e1e
HFS		 1733/* Used by processes to "lock" a socket state, so that
1734 * interrupts and bottom half handlers won't change it
1735 * from under us. It essentially blocks any incoming
1736 * packets, so that we won't get any new data or any
1737 * packets that change the state of the socket.
1738 *
1739 * While locked, BH processing will add new packets to
1740 * the backlog queue. This queue is processed by the
1741 * owner of the socket lock right before it is released.
1742 *
1743 * Since ~2.3.5 it is also exclusive sleep lock serializing
1744 * accesses from user process context.
1745 */
1746
46cc6e49 1747static inline void sock_owned_by_me(const struct sock *sk)
fafc4e1e
HFS		 1748{
1749#ifdef CONFIG_LOCKDEP
5e91f6ce 1750 WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks);
fafc4e1e 1751#endif
46cc6e49
ED		 1752}
1753
1754static inline bool sock_owned_by_user(const struct sock *sk)
1755{
1756 sock_owned_by_me(sk);
fafc4e1e
HFS		 1757 return sk->sk_lock.owned;
1758}
1759
602f7a27
TH		 1760static inline bool sock_owned_by_user_nocheck(const struct sock *sk)
1761{
1762 return sk->sk_lock.owned;
1763}
1764
33d60fbd
KI		 1765static inline void sock_release_ownership(struct sock *sk)
1766{
1767 if (sock_owned_by_user_nocheck(sk)) {
1768 sk->sk_lock.owned = 0;
1769
1770 /* The sk_lock has mutex_unlock() semantics: */
1771 mutex_release(&sk->sk_lock.dep_map, _RET_IP_);
1772 }
1773}
1774
fafc4e1e
HFS		 1775/* no reclassification while locks are held */
1776static inline bool sock_allow_reclassification(const struct sock *csk)
1777{
1778 struct sock *sk = (struct sock *)csk;
1779
33d60fbd
KI		 1780 return !sock_owned_by_user_nocheck(sk) &&
1781 !spin_is_locked(&sk->sk_lock.slock);
fafc4e1e 1782}
4b0b72f7 1783
69336bd2 1784struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
11aa9c28 1785 struct proto *prot, int kern);
69336bd2 1786void sk_free(struct sock *sk);
eb4cb008 1787void sk_destruct(struct sock *sk);
69336bd2 1788struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
94352d45 1789void sk_free_unlock_clone(struct sock *sk);
69336bd2
JP		 1790
1791struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
1792 gfp_t priority);
1d2077ac 1793void __sock_wfree(struct sk_buff *skb);
69336bd2 1794void sock_wfree(struct sk_buff *skb);
98ba0bd5
WB		 1795struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size,
1796 gfp_t priority);
69336bd2
JP		 1797void skb_orphan_partial(struct sk_buff *skb);
1798void sock_rfree(struct sk_buff *skb);
62bccb8c 1799void sock_efree(struct sk_buff *skb);
82eabd9e 1800#ifdef CONFIG_INET
69336bd2 1801void sock_edemux(struct sk_buff *skb);
cf7fbe66 1802void sock_pfree(struct sk_buff *skb);
82eabd9e 1803#else
158f323b 1804#define sock_edemux sock_efree
82eabd9e 1805#endif
69336bd2
JP		 1806
1807int sock_setsockopt(struct socket *sock, int level, int op,
c8c1bbb6 1808 sockptr_t optval, unsigned int optlen);
69336bd2
JP		 1809
1810int sock_getsockopt(struct socket *sock, int level, int op,
1811 char __user *optval, int __user *optlen);
c7cbdbf2
AB		 1812int sock_gettstamp(struct socket *sock, void __user *userstamp,
1813 bool timeval, bool time32);
69336bd2
JP		 1814struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
1815 int noblock, int *errcode);
1816struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
1817 unsigned long data_len, int noblock,
1818 int *errcode, int max_page_order);
1819void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
1820void sock_kfree_s(struct sock *sk, void *mem, int size);
79e88659 1821void sock_kzfree_s(struct sock *sk, void *mem, int size);
69336bd2 1822void sk_send_sigurg(struct sock *sk);
1da177e4 1823
f28ea365 1824struct sockcm_cookie {
80b14dee 1825 u64 transmit_time;
f28ea365 1826 u32 mark;
3dd17e63 1827 u16 tsflags;
f28ea365
EJ		 1828};
1829
657a0667
WB		 1830static inline void sockcm_init(struct sockcm_cookie *sockc,
1831 const struct sock *sk)
1832{
1833 *sockc = (struct sockcm_cookie) { .tsflags = sk->sk_tsflags };
1834}
1835
39771b12
WB		 1836int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg,
1837 struct sockcm_cookie *sockc);
f28ea365
EJ		 1838int sock_cmsg_send(struct sock *sk, struct msghdr *msg,
1839 struct sockcm_cookie *sockc);
1840
1da177e4
LT		 1841/*
1842 * Functions to fill in entries in struct proto_ops when a protocol
1843 * does not implement a particular function.
1844 */
69336bd2
JP		 1845int sock_no_bind(struct socket *, struct sockaddr *, int);
1846int sock_no_connect(struct socket *, struct sockaddr *, int, int);
1847int sock_no_socketpair(struct socket *, struct socket *);
cdfbabfb 1848int sock_no_accept(struct socket *, struct socket *, int, bool);
9b2c45d4 1849int sock_no_getname(struct socket *, struct sockaddr *, int);
69336bd2
JP		 1850int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
1851int sock_no_listen(struct socket *, int);
1852int sock_no_shutdown(struct socket *, int);
1b784140 1853int sock_no_sendmsg(struct socket *, struct msghdr *, size_t);
306b13eb 1854int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len);
1b784140 1855int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int);
69336bd2
JP		 1856int sock_no_mmap(struct file *file, struct socket *sock,
1857 struct vm_area_struct *vma);
1858ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
1859 size_t size, int flags);
306b13eb
TH		 1860ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page,
1861 int offset, size_t size, int flags);
1da177e4
LT		 1862
1863/*
1864 * Functions to fill in entries in struct proto_ops when a protocol
1865 * uses the inet style.
1866 */
69336bd2 1867int sock_common_getsockopt(struct socket *sock, int level, int optname,
1da177e4 1868 char __user *optval, int __user *optlen);
1b784140
YX		 1869int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size,
1870 int flags);
69336bd2 1871int sock_common_setsockopt(struct socket *sock, int level, int optname,
a7b75c5a 1872 sockptr_t optval, unsigned int optlen);
1da177e4 1873
69336bd2 1874void sk_common_release(struct sock *sk);
1da177e4
LT		 1875
1876/*
1877 * Default socket callbacks and setup code
1878 */
dc6b9b78 1879
1da177e4 1880/* Initialise core socket variables */
69336bd2 1881void sock_init_data(struct socket *sock, struct sock *sk);
1da177e4 1882
1da177e4
LT		 1883/*
1884 * Socket reference counting postulates.
1885 *
1886 * * Each user of socket SHOULD hold a reference count.
1887 * * Each access point to socket (an hash table bucket, reference from a list,
1888 * running timer, skb in flight MUST hold a reference count.
1889 * * When reference count hits 0, it means it will never increase back.
1890 * * When reference count hits 0, it means that no references from
1891 * outside exist to this socket and current process on current CPU
1892 * is last user and may/should destroy this socket.
1893 * * sk_free is called from any context: process, BH, IRQ. When
1894 * it is called, socket has no references from outside -> sk_free
1895 * may release descendant resources allocated by the socket, but
1896 * to the time when it is called, socket is NOT referenced by any
1897 * hash tables, lists etc.
1898 * * Packets, delivered from outside (from network or from another process)
1899 * and enqueued on receive/error queues SHOULD NOT grab reference count,
1900 * when they sit in queue. Otherwise, packets will leak to hole, when
1901 * socket is looked up by one cpu and unhasing is made by another CPU.
1902 * It is true for udp/raw, netlink (leak to receive and error queues), tcp
1903 * (leak to backlog). Packet socket does all the processing inside
1904 * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
1905 * use separate SMP lock, so that they are prone too.
1906 */
1907
1908/* Ungrab socket and destroy it, if it was the last reference. */
1909static inline void sock_put(struct sock *sk)
1910{
41c6d650 1911 if (refcount_dec_and_test(&sk->sk_refcnt))
1da177e4
LT		 1912 sk_free(sk);
1913}
05dbc7b5 1914/* Generic version of sock_put(), dealing with all sockets
41b822c5 1915 * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...)
05dbc7b5
ED		 1916 */
1917void sock_gen_put(struct sock *sk);
1da177e4 1918
4f0c40d9 1919int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested,
c3f24cfb 1920 unsigned int trim_cap, bool refcounted);
4f0c40d9
WB		 1921static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb,
1922 const int nested)
1923{
c3f24cfb 1924 return __sk_receive_skb(sk, skb, nested, 1, true);
4f0c40d9 1925}
25995ff5 1926
e022f0b4
KK		 1927static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
1928{
755c31cd
AN		 1929 /* sk_tx_queue_mapping accept only upto a 16-bit value */
1930 if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX))
1931 return;
e022f0b4
KK		 1932 sk->sk_tx_queue_mapping = tx_queue;
1933}
1934
755c31cd
AN		 1935#define NO_QUEUE_MAPPING USHRT_MAX
1936
e022f0b4
KK		 1937static inline void sk_tx_queue_clear(struct sock *sk)
1938{
755c31cd 1939 sk->sk_tx_queue_mapping = NO_QUEUE_MAPPING;
e022f0b4
KK		 1940}
1941
1942static inline int sk_tx_queue_get(const struct sock *sk)
1943{
755c31cd
AN		 1944 if (sk && sk->sk_tx_queue_mapping != NO_QUEUE_MAPPING)
1945 return sk->sk_tx_queue_mapping;
1946
1947 return -1;
e022f0b4
KK		 1948}
1949
a37a0ee4
ED		 1950static inline void __sk_rx_queue_set(struct sock *sk,
1951 const struct sk_buff *skb,
1952 bool force_set)
c6345ce7 1953{
4e1beecc 1954#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
c6345ce7
AN		 1955 if (skb_rx_queue_recorded(skb)) {
1956 u16 rx_queue = skb_get_rx_queue(skb);
1957
a37a0ee4
ED		 1958 if (force_set ||
1959 unlikely(READ_ONCE(sk->sk_rx_queue_mapping) != rx_queue))
342159ee 1960 WRITE_ONCE(sk->sk_rx_queue_mapping, rx_queue);
c6345ce7
AN		 1961 }
1962#endif
1963}
1964
a37a0ee4
ED		 1965static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb)
1966{
1967 __sk_rx_queue_set(sk, skb, true);
1968}
1969
1970static inline void sk_rx_queue_update(struct sock *sk, const struct sk_buff *skb)
1971{
1972 __sk_rx_queue_set(sk, skb, false);
1973}
1974
c6345ce7
AN		 1975static inline void sk_rx_queue_clear(struct sock *sk)
1976{
4e1beecc 1977#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
09b89846 1978 WRITE_ONCE(sk->sk_rx_queue_mapping, NO_QUEUE_MAPPING);
c6345ce7
AN		 1979#endif
1980}
1981
fc9bab24
AN		 1982static inline int sk_rx_queue_get(const struct sock *sk)
1983{
4e1beecc 1984#ifdef CONFIG_SOCK_RX_QUEUE_MAPPING
09b89846
ED		 1985 if (sk) {
1986 int res = READ_ONCE(sk->sk_rx_queue_mapping);
1987
1988 if (res != NO_QUEUE_MAPPING)
1989 return res;
1990 }
4e1beecc 1991#endif
fc9bab24
AN		 1992
1993 return -1;
1994}
fc9bab24 1995
972692e0
DM		 1996static inline void sk_set_socket(struct sock *sk, struct socket *sock)
1997{
1998 sk->sk_socket = sock;
1999}
2000
aa395145
ED		 2001static inline wait_queue_head_t *sk_sleep(struct sock *sk)
2002{
eaefd110
ED		 2003 BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
2004 return &rcu_dereference_raw(sk->sk_wq)->wait;
aa395145 2005}
1da177e4
LT		 2006/* Detach socket from process context.
2007 * Announce socket dead, detach it from wait queue and inode.
2008 * Note that parent inode held reference count on this struct sock,
2009 * we do not release it in this function, because protocol
2010 * probably wants some additional cleanups or even continuing
2011 * to work with this socket (TCP).
2012 */
2013static inline void sock_orphan(struct sock *sk)
2014{
2015 write_lock_bh(&sk->sk_callback_lock);
2016 sock_set_flag(sk, SOCK_DEAD);
972692e0 2017 sk_set_socket(sk, NULL);
43815482 2018 sk->sk_wq = NULL;
1da177e4
LT		 2019 write_unlock_bh(&sk->sk_callback_lock);
2020}
2021
2022static inline void sock_graft(struct sock *sk, struct socket *parent)
2023{
0ffdaf5b 2024 WARN_ON(parent->sk);
1da177e4 2025 write_lock_bh(&sk->sk_callback_lock);
333f7909 2026 rcu_assign_pointer(sk->sk_wq, &parent->wq);
1da177e4 2027 parent->sk = sk;
972692e0 2028 sk_set_socket(sk, parent);
86741ec2 2029 sk->sk_uid = SOCK_INODE(parent)->i_uid;
4237c75c 2030 security_sock_graft(sk, parent);
1da177e4
LT		 2031 write_unlock_bh(&sk->sk_callback_lock);
2032}
2033
69336bd2
JP		 2034kuid_t sock_i_uid(struct sock *sk);
2035unsigned long sock_i_ino(struct sock *sk);
1da177e4 2036
86741ec2
LC		 2037static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
2038{
2039 return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
2040}
2041
58d607d3 2042static inline u32 net_tx_rndhash(void)
877d1f62 2043{
58d607d3
ED		 2044 u32 v = prandom_u32();
2045
2046 return v ?: 1;
2047}
877d1f62 2048
58d607d3
ED		 2049static inline void sk_set_txhash(struct sock *sk)
2050{
b71eaed8
ED		 2051 /* This pairs with READ_ONCE() in skb_set_hash_from_sk() */
2052 WRITE_ONCE(sk->sk_txhash, net_tx_rndhash());
877d1f62
TH		 2053}
2054
9c30ae83 2055static inline bool sk_rethink_txhash(struct sock *sk)
265f94ff 2056{
9c30ae83 2057 if (sk->sk_txhash) {
265f94ff 2058 sk_set_txhash(sk);
9c30ae83
YC		 2059 return true;
2060 }
2061 return false;
265f94ff
TH		 2062}
2063
1da177e4
LT		 2064static inline struct dst_entry *
2065__sk_dst_get(struct sock *sk)
2066{
1e1d04e6
HFS		 2067 return rcu_dereference_check(sk->sk_dst_cache,
2068 lockdep_sock_is_held(sk));
1da177e4
LT		 2069}
2070
2071static inline struct dst_entry *
2072sk_dst_get(struct sock *sk)
2073{
2074 struct dst_entry *dst;
2075
b6c6712a
ED		 2076 rcu_read_lock();
2077 dst = rcu_dereference(sk->sk_dst_cache);
f8864972
ED		 2078 if (dst && !atomic_inc_not_zero(&dst->__refcnt))
2079 dst = NULL;
b6c6712a 2080 rcu_read_unlock();
1da177e4
LT		 2081 return dst;
2082}
2083
9c30ae83 2084static inline void __dst_negative_advice(struct sock *sk)
b6c6712a
ED		 2085{
2086 struct dst_entry *ndst, *dst = __sk_dst_get(sk);
2087
2088 if (dst && dst->ops->negative_advice) {
2089 ndst = dst->ops->negative_advice(dst);
2090
2091 if (ndst != dst) {
2092 rcu_assign_pointer(sk->sk_dst_cache, ndst);
0a6957e7 2093 sk_tx_queue_clear(sk);
9b8805a3 2094 sk->sk_dst_pending_confirm = 0;
b6c6712a
ED		 2095 }
2096 }
2097}
2098
9c30ae83
YC		 2099static inline void dst_negative_advice(struct sock *sk)
2100{
2101 sk_rethink_txhash(sk);
2102 __dst_negative_advice(sk);
2103}
2104
1da177e4
LT		 2105static inline void
2106__sk_dst_set(struct sock *sk, struct dst_entry *dst)
2107{
2108 struct dst_entry *old_dst;
2109
e022f0b4 2110 sk_tx_queue_clear(sk);
9b8805a3 2111 sk->sk_dst_pending_confirm = 0;
95964c6d
ED		 2112 old_dst = rcu_dereference_protected(sk->sk_dst_cache,
2113 lockdep_sock_is_held(sk));
b6c6712a 2114 rcu_assign_pointer(sk->sk_dst_cache, dst);
1da177e4
LT		 2115 dst_release(old_dst);
2116}
2117
2118static inline void
2119sk_dst_set(struct sock *sk, struct dst_entry *dst)
2120{
7f502361
ED		 2121 struct dst_entry *old_dst;
2122
2123 sk_tx_queue_clear(sk);
9b8805a3 2124 sk->sk_dst_pending_confirm = 0;
5925a055 2125 old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst);
7f502361 2126 dst_release(old_dst);
1da177e4
LT		 2127}
2128
2129static inline void
2130__sk_dst_reset(struct sock *sk)
2131{
b6c6712a 2132 __sk_dst_set(sk, NULL);
1da177e4
LT		 2133}
2134
2135static inline void
2136sk_dst_reset(struct sock *sk)
2137{
7f502361 2138 sk_dst_set(sk, NULL);
1da177e4
LT		 2139}
2140
69336bd2 2141struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 2142
69336bd2 2143struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
1da177e4 2144
9b8805a3
JA		 2145static inline void sk_dst_confirm(struct sock *sk)
2146{
25c7a6d1
ED		 2147 if (!READ_ONCE(sk->sk_dst_pending_confirm))
2148 WRITE_ONCE(sk->sk_dst_pending_confirm, 1);
9b8805a3
JA		 2149}
2150
4ff06203
JA		 2151static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n)
2152{
2153 if (skb_get_dst_pending_confirm(skb)) {
2154 struct sock *sk = skb->sk;
4ff06203 2155
25c7a6d1
ED		 2156 if (sk && READ_ONCE(sk->sk_dst_pending_confirm))
2157 WRITE_ONCE(sk->sk_dst_pending_confirm, 0);
1e84dc6b 2158 neigh_confirm(n);
4ff06203
JA		 2159 }
2160}
2161
f60e5990 2162bool sk_mc_loop(struct sock *sk);
2163
dc6b9b78 2164static inline bool sk_can_gso(const struct sock *sk)
bcd76111
HX		 2165{
2166 return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
2167}
2168
69336bd2 2169void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
6cbb0df7 2170
aba54656 2171static inline void sk_gso_disable(struct sock *sk)
a465419b 2172{
aba54656
ED		 2173 sk->sk_gso_disabled = 1;
2174 sk->sk_route_caps &= ~NETIF_F_GSO_MASK;
a465419b
ED		 2175}
2176
c6e1a0d1 2177static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
57be5bda 2178 struct iov_iter *from, char *to,
912d398d 2179 int copy, int offset)
c6e1a0d1
TH		 2180{
2181 if (skb->ip_summed == CHECKSUM_NONE) {
57be5bda 2182 __wsum csum = 0;
15e6cb46 2183 if (!csum_and_copy_from_iter_full(to, copy, &csum, from))
57be5bda 2184 return -EFAULT;
912d398d 2185 skb->csum = csum_block_add(skb->csum, csum, offset);
c6e1a0d1 2186 } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
15e6cb46 2187 if (!copy_from_iter_full_nocache(to, copy, from))
c6e1a0d1 2188 return -EFAULT;
15e6cb46 2189 } else if (!copy_from_iter_full(to, copy, from))
c6e1a0d1
TH		 2190 return -EFAULT;
2191
2192 return 0;
2193}
2194
2195static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
57be5bda 2196 struct iov_iter *from, int copy)
c6e1a0d1 2197{
912d398d 2198 int err, offset = skb->len;
c6e1a0d1 2199
912d398d
WY		 2200 err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
2201 copy, offset);
c6e1a0d1 2202 if (err)
912d398d 2203 __skb_trim(skb, offset);
c6e1a0d1
TH		 2204
2205 return err;
2206}
2207
57be5bda 2208static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from,
c6e1a0d1
TH		 2209 struct sk_buff *skb,
2210 struct page *page,
2211 int off, int copy)
2212{
2213 int err;
2214
912d398d
WY		 2215 err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
2216 copy, skb->len);
c6e1a0d1
TH		 2217 if (err)
2218 return err;
2219
2220 skb->len += copy;
2221 skb->data_len += copy;
2222 skb->truesize += copy;
ab4e846a 2223 sk_wmem_queued_add(sk, copy);
c6e1a0d1
TH		 2224 sk_mem_charge(sk, copy);
2225 return 0;
2226}
2227
c564039f
ED		 2228/**
2229 * sk_wmem_alloc_get - returns write allocations
2230 * @sk: socket
2231 *
66256e0b 2232 * Return: sk_wmem_alloc minus initial offset of one
c564039f
ED		 2233 */
2234static inline int sk_wmem_alloc_get(const struct sock *sk)
2235{
14afee4b 2236 return refcount_read(&sk->sk_wmem_alloc) - 1;
c564039f
ED		 2237}
2238
2239/**
2240 * sk_rmem_alloc_get - returns read allocations
2241 * @sk: socket
2242 *
66256e0b 2243 * Return: sk_rmem_alloc
c564039f
ED		 2244 */
2245static inline int sk_rmem_alloc_get(const struct sock *sk)
2246{
2247 return atomic_read(&sk->sk_rmem_alloc);
2248}
2249
2250/**
2251 * sk_has_allocations - check if allocations are outstanding
2252 * @sk: socket
2253 *
66256e0b 2254 * Return: true if socket has write or read allocations
c564039f 2255 */
dc6b9b78 2256static inline bool sk_has_allocations(const struct sock *sk)
c564039f
ED		 2257{
2258 return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
2259}
2260
a57de0b4 2261/**
1ce0bf50 2262 * skwq_has_sleeper - check if there are any waiting processes
acfbe96a 2263 * @wq: struct socket_wq
a57de0b4 2264 *
66256e0b 2265 * Return: true if socket_wq has waiting processes
a57de0b4 2266 *
1ce0bf50 2267 * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory
a57de0b4
JO		 2268 * barrier call. They were added due to the race found within the tcp code.
2269 *
d651983d 2270 * Consider following tcp code paths::
a57de0b4 2271 *
d651983d
MCC		 2272 * CPU1 CPU2
2273 * sys_select receive packet
a57de0b4
JO		 2274 * ... ...
2275 * __add_wait_queue update tp->rcv_nxt
2276 * ... ...
2277 * tp->rcv_nxt check sock_def_readable
2278 * ... {
43815482
ED		 2279 * schedule rcu_read_lock();
2280 * wq = rcu_dereference(sk->sk_wq);
2281 * if (wq && waitqueue_active(&wq->wait))
2282 * wake_up_interruptible(&wq->wait)
a57de0b4
JO		 2283 * ...
2284 * }
2285 *
2286 * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
2287 * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
2288 * could then endup calling schedule and sleep forever if there are no more
2289 * data on the socket.
ad462769 2290 *
a57de0b4 2291 */
1ce0bf50 2292static inline bool skwq_has_sleeper(struct socket_wq *wq)
a57de0b4 2293{
1ce0bf50 2294 return wq && wq_has_sleeper(&wq->wait);
a57de0b4
JO		 2295}
2296
2297/**
2298 * sock_poll_wait - place memory barrier behind the poll_wait call.
2299 * @filp: file
89ab066d 2300 * @sock: socket to wait on
a57de0b4
JO		 2301 * @p: poll_table
2302 *
43815482 2303 * See the comments in the wq_has_sleeper function.
a57de0b4 2304 */
89ab066d
KG		 2305static inline void sock_poll_wait(struct file *filp, struct socket *sock,
2306 poll_table *p)
a57de0b4 2307{
d8bbd13b 2308 if (!poll_does_not_wait(p)) {
333f7909 2309 poll_wait(filp, &sock->wq.wait, p);
dc6b9b78 2310 /* We need to be sure we are in sync with the
a57de0b4
JO		 2311 * socket flags modification.
2312 *
43815482 2313 * This memory barrier is paired in the wq_has_sleeper.
dc6b9b78 2314 */
a57de0b4
JO		 2315 smp_mb();
2316 }
2317}
2318
b73c3d0e
TH		 2319static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk)
2320{
b71eaed8
ED		 2321 /* This pairs with WRITE_ONCE() in sk_set_txhash() */
2322 u32 txhash = READ_ONCE(sk->sk_txhash);
2323
2324 if (txhash) {
b73c3d0e 2325 skb->l4_hash = 1;
b71eaed8 2326 skb->hash = txhash;
b73c3d0e
TH		 2327 }
2328}
2329
9e17f8a4
ED		 2330void skb_set_owner_w(struct sk_buff *skb, struct sock *sk);
2331
1da177e4 2332/*
dc6b9b78 2333 * Queue a received datagram if it will fit. Stream and sequenced
1da177e4
LT		 2334 * protocols can't normally use this as they need to fit buffers in
2335 * and play with them.
2336 *
dc6b9b78 2337 * Inlined as it's very short and called for pretty much every
1da177e4
LT		 2338 * packet ever received.
2339 */
1da177e4
LT		 2340static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
2341{
d55d87fd 2342 skb_orphan(skb);
1da177e4
LT		 2343 skb->sk = sk;
2344 skb->destructor = sock_rfree;
2345 atomic_add(skb->truesize, &sk->sk_rmem_alloc);
3ab224be 2346 sk_mem_charge(sk, skb->truesize);
1da177e4
LT		 2347}
2348
098116e7 2349static inline __must_check bool skb_set_owner_sk_safe(struct sk_buff *skb, struct sock *sk)
9adc89af
PA		 2350{
2351 if (sk && refcount_inc_not_zero(&sk->sk_refcnt)) {
2352 skb_orphan(skb);
2353 skb->destructor = sock_efree;
2354 skb->sk = sk;
098116e7 2355 return true;
9adc89af 2356 }
098116e7 2357 return false;
9adc89af
PA		 2358}
2359
5e10da53
PA		 2360static inline void skb_prepare_for_gro(struct sk_buff *skb)
2361{
2362 if (skb->destructor != sock_wfree) {
2363 skb_orphan(skb);
2364 return;
2365 }
2366 skb->slow_gro = 1;
2367}
2368
69336bd2
JP		 2369void sk_reset_timer(struct sock *sk, struct timer_list *timer,
2370 unsigned long expires);
1da177e4 2371
69336bd2 2372void sk_stop_timer(struct sock *sk, struct timer_list *timer);
1da177e4 2373
08b81d87
GT		 2374void sk_stop_timer_sync(struct sock *sk, struct timer_list *timer);
2375
65101aec
PA		 2376int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue,
2377 struct sk_buff *skb, unsigned int flags,
69629464
ED		 2378 void (*destructor)(struct sock *sk,
2379 struct sk_buff *skb));
e6afc8ac 2380int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
69336bd2 2381int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
1da177e4 2382
69336bd2 2383int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
364a9e93 2384struct sk_buff *sock_dequeue_err_skb(struct sock *sk);
1da177e4
LT		 2385
2386/*
2387 * Recover an error report and clear atomically
2388 */
dc6b9b78 2389
1da177e4
LT		 2390static inline int sock_error(struct sock *sk)
2391{
c1cbe4b7 2392 int err;
f13ef100
ED		 2393
2394 /* Avoid an atomic operation for the common case.
2395 * This is racy since another cpu/thread can change sk_err under us.
2396 */
2397 if (likely(data_race(!sk->sk_err)))
c1cbe4b7 2398 return 0;
f13ef100 2399
c1cbe4b7 2400 err = xchg(&sk->sk_err, 0);
1da177e4
LT		 2401 return -err;
2402}
2403
e3ae2365
AA		 2404void sk_error_report(struct sock *sk);
2405
1da177e4
LT		 2406static inline unsigned long sock_wspace(struct sock *sk)
2407{
2408 int amt = 0;
2409
2410 if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
14afee4b 2411 amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc);
dc6b9b78 2412 if (amt < 0)
1da177e4
LT		 2413 amt = 0;
2414 }
2415 return amt;
2416}
2417
ceb5d58b
ED		 2418/* Note:
2419 * We use sk->sk_wq_raw, from contexts knowing this
2420 * pointer is not NULL and cannot disappear/change.
2421 */
9cd3e072 2422static inline void sk_set_bit(int nr, struct sock *sk)
1da177e4 2423{
4be73522
ED		 2424 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2425 !sock_flag(sk, SOCK_FASYNC))
9317bb69
ED		 2426 return;
2427
ceb5d58b 2428 set_bit(nr, &sk->sk_wq_raw->flags);
9cd3e072
ED		 2429}
2430
2431static inline void sk_clear_bit(int nr, struct sock *sk)
2432{
4be73522
ED		 2433 if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) &&
2434 !sock_flag(sk, SOCK_FASYNC))
9317bb69
ED		 2435 return;
2436
ceb5d58b 2437 clear_bit(nr, &sk->sk_wq_raw->flags);
9cd3e072
ED		 2438}
2439
ceb5d58b 2440static inline void sk_wake_async(const struct sock *sk, int how, int band)
1da177e4 2441{
ceb5d58b
ED		 2442 if (sock_flag(sk, SOCK_FASYNC)) {
2443 rcu_read_lock();
2444 sock_wake_async(rcu_dereference(sk->sk_wq), how, band);
2445 rcu_read_unlock();
2446 }
1da177e4
LT		 2447}
2448
eea86af6
DB		 2449/* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
2450 * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
2451 * Note: for send buffers, TCP works better if we can build two skbs at
2452 * minimum.
7a91b434 2453 */
9eb5bf83 2454#define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
eea86af6
DB		 2455
2456#define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
2457#define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
1da177e4
LT		 2458
2459static inline void sk_stream_moderate_sndbuf(struct sock *sk)
2460{
e292f05e
ED		 2461 u32 val;
2462
2463 if (sk->sk_userlocks & SOCK_SNDBUF_LOCK)
2464 return;
2465
2466 val = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
ca057051 2467 val = max_t(u32, val, sk_unused_reserved_mem(sk));
e292f05e
ED		 2468
2469 WRITE_ONCE(sk->sk_sndbuf, max_t(u32, val, SOCK_MIN_SNDBUF));
1da177e4
LT		 2470}
2471
5640f768
ED		 2472/**
2473 * sk_page_frag - return an appropriate page_frag
2474 * @sk: socket
2475 *
20eb4f29 2476 * Use the per task page_frag instead of the per socket one for
dacb5d88 2477 * optimization when we know that we're in process context and own
20eb4f29
TH		 2478 * everything that's associated with %current.
2479 *
dacb5d88
PA		 2480 * Both direct reclaim and page faults can nest inside other
2481 * socket operations and end up recursing into sk_page_frag()
2482 * while it's already in use: explicitly avoid task page_frag
2483 * usage if the caller is potentially doing any of them.
2484 * This assumes that page fault handlers use the GFP_NOFS flags.
66256e0b
RD		 2485 *
2486 * Return: a per task page_frag if context allows that,
2487 * otherwise a per socket one.
5640f768
ED		 2488 */
2489static inline struct page_frag *sk_page_frag(struct sock *sk)
1da177e4 2490{
dacb5d88
PA		 2491 if ((sk->sk_allocation & (__GFP_DIRECT_RECLAIM | __GFP_MEMALLOC | __GFP_FS)) ==
2492 (__GFP_DIRECT_RECLAIM | __GFP_FS))
5640f768 2493 return &current->task_frag;
1da177e4 2494
5640f768 2495 return &sk->sk_frag;
1da177e4
LT		 2496}
2497
69336bd2 2498bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
5640f768 2499
1da177e4
LT		 2500/*
2501 * Default write policy as shown to user space via poll/select/SIGIO
2502 */
dc6b9b78 2503static inline bool sock_writeable(const struct sock *sk)
1da177e4 2504{
e292f05e 2505 return refcount_read(&sk->sk_wmem_alloc) < (READ_ONCE(sk->sk_sndbuf) >> 1);
1da177e4
LT		 2506}
2507
dd0fc66f 2508static inline gfp_t gfp_any(void)
1da177e4 2509{
99709372 2510 return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
1da177e4
LT		 2511}
2512
4b1327be
WW		 2513static inline gfp_t gfp_memcg_charge(void)
2514{
2515 return in_softirq() ? GFP_NOWAIT : GFP_KERNEL;
2516}
2517
dc6b9b78 2518static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
1da177e4
LT		 2519{
2520 return noblock ? 0 : sk->sk_rcvtimeo;
2521}
2522
dc6b9b78 2523static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
1da177e4
LT		 2524{
2525 return noblock ? 0 : sk->sk_sndtimeo;
2526}
2527
2528static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
2529{
eac66402
ED		 2530 int v = waitall ? len : min_t(int, READ_ONCE(sk->sk_rcvlowat), len);
2531
2532 return v ?: 1;
1da177e4
LT		 2533}
2534
2535/* Alas, with timeout socket operations are not restartable.
2536 * Compare this to poll().
2537 */
2538static inline int sock_intr_errno(long timeo)
2539{
2540 return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
2541}
2542
744d5a3e
EB		 2543struct sock_skb_cb {
2544 u32 dropcount;
2545};
2546
2547/* Store sock_skb_cb at the end of skb->cb[] so protocol families
2548 * using skb->cb[] would keep using it directly and utilize its
2549 * alignement guarantee.
2550 */
c593642c 2551#define SOCK_SKB_CB_OFFSET ((sizeof_field(struct sk_buff, cb) - \
744d5a3e
EB		 2552 sizeof(struct sock_skb_cb)))
2553
2554#define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \
2555 SOCK_SKB_CB_OFFSET))
2556
b4772ef8 2557#define sock_skb_cb_check_size(size) \
744d5a3e 2558 BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET)
b4772ef8 2559
3bc3b96f
EB		 2560static inline void
2561sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb)
2562{
3665f381
ED		 2563 SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ?
2564 atomic_read(&sk->sk_drops) : 0;
3bc3b96f
EB		 2565}
2566
532182cd
ED		 2567static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb)
2568{
2569 int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs);
2570
2571 atomic_add(segs, &sk->sk_drops);
2572}
2573
3a0ed3e9
DD		 2574static inline ktime_t sock_read_timestamp(struct sock *sk)
2575{
2576#if BITS_PER_LONG==32
2577 unsigned int seq;
2578 ktime_t kt;
2579
2580 do {
2581 seq = read_seqbegin(&sk->sk_stamp_seq);
2582 kt = sk->sk_stamp;
2583 } while (read_seqretry(&sk->sk_stamp_seq, seq));
2584
2585 return kt;
2586#else
f75359f3 2587 return READ_ONCE(sk->sk_stamp);
3a0ed3e9
DD		 2588#endif
2589}
2590
2591static inline void sock_write_timestamp(struct sock *sk, ktime_t kt)
2592{
2593#if BITS_PER_LONG==32
2594 write_seqlock(&sk->sk_stamp_seq);
2595 sk->sk_stamp = kt;
2596 write_sequnlock(&sk->sk_stamp_seq);
2597#else
f75359f3 2598 WRITE_ONCE(sk->sk_stamp, kt);
3a0ed3e9
DD		 2599#endif
2600}
2601
69336bd2
JP		 2602void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
2603 struct sk_buff *skb);
2604void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
2605 struct sk_buff *skb);
92f37fd2 2606
dc6b9b78 2607static inline void
1da177e4
LT		 2608sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
2609{
b7aa0bf7 2610 ktime_t kt = skb->tstamp;
20d49473 2611 struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
a61bbcf2 2612
20d49473
PO		 2613 /*
2614 * generate control messages if
b9f40e21 2615 * - receive time stamping in software requested
20d49473 2616 * - software time stamp available and wanted
20d49473 2617 * - hardware time stamps available and wanted
20d49473
PO		 2618 */
2619 if (sock_flag(sk, SOCK_RCVTSTAMP) ||
b9f40e21 2620 (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) ||
2456e855
TG		 2621 (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) ||
2622 (hwtstamps->hwtstamp &&
b9f40e21 2623 (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE)))
92f37fd2
ED		 2624 __sock_recv_timestamp(msg, sk, skb);
2625 else
3a0ed3e9 2626 sock_write_timestamp(sk, kt);
6e3e939f
JB		 2627
2628 if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
2629 __sock_recv_wifi_status(msg, sk, skb);
1da177e4
LT		 2630}
2631
69336bd2
JP		 2632void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2633 struct sk_buff *skb);
767dd033 2634
6c7c98ba 2635#define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC)
767dd033
ED		 2636static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
2637 struct sk_buff *skb)
2638{
2639#define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
b9f40e21
WB		 2640 (1UL << SOCK_RCVTSTAMP))
2641#define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \
2642 SOF_TIMESTAMPING_RAW_HARDWARE)
767dd033 2643
b9f40e21 2644 if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY)
767dd033 2645 __sock_recv_ts_and_drops(msg, sk, skb);
d3fbff30 2646 else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP)))
3a0ed3e9 2647 sock_write_timestamp(sk, skb->tstamp);
6c7c98ba 2648 else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP))
3a0ed3e9 2649 sock_write_timestamp(sk, 0);
767dd033 2650}
3b885787 2651
c14ac945 2652void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags);
67cc0d40 2653
20d49473 2654/**
8f932f76 2655 * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
20d49473 2656 * @sk: socket sending this packet
c14ac945 2657 * @tsflags: timestamping flags to use
140c55d4 2658 * @tx_flags: completed with instructions for time stamping
8f932f76 2659 * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno)
140c55d4 2660 *
d651983d 2661 * Note: callers should take care of initial ``*tx_flags`` value (usually 0)
20d49473 2662 */
8f932f76
WB		 2663static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2664 __u8 *tx_flags, __u32 *tskey)
67cc0d40 2665{
8f932f76 2666 if (unlikely(tsflags)) {
c14ac945 2667 __sock_tx_timestamp(tsflags, tx_flags);
8f932f76
WB		 2668 if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey &&
2669 tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK)
2670 *tskey = sk->sk_tskey++;
2671 }
67cc0d40
WB		 2672 if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS)))
2673 *tx_flags |= SKBTX_WIFI_STATUS;
2674}
20d49473 2675
8f932f76
WB		 2676static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags,
2677 __u8 *tx_flags)
2678{
2679 _sock_tx_timestamp(sk, tsflags, tx_flags, NULL);
2680}
2681
2682static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags)
2683{
2684 _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags,
2685 &skb_shinfo(skb)->tskey);
2686}
2687
42f67eea
ED		 2688static inline bool sk_is_tcp(const struct sock *sk)
2689{
2690 return sk->sk_type == SOCK_STREAM && sk->sk_protocol == IPPROTO_TCP;
2691}
2692
1da177e4
LT		 2693/**
2694 * sk_eat_skb - Release a skb if it is no longer needed
4dc3b16b
PP		 2695 * @sk: socket to eat this skb from
2696 * @skb: socket buffer to eat
1da177e4
LT		 2697 *
2698 * This routine must be called with interrupts disabled or with the socket
2699 * locked so that the sk_buff queue operation is ok.
2700*/
7bced397 2701static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb)
1da177e4
LT		 2702{
2703 __skb_unlink(skb, &sk->sk_receive_queue);
2704 __kfree_skb(skb);
2705}
2706
3b1e0a65
YH		 2707static inline
2708struct net *sock_net(const struct sock *sk)
2709{
c2d9ba9b 2710 return read_pnet(&sk->sk_net);
3b1e0a65
YH		 2711}
2712
2713static inline
f5aa23fd 2714void sock_net_set(struct sock *sk, struct net *net)
3b1e0a65 2715{
c2d9ba9b 2716 write_pnet(&sk->sk_net, net);
3b1e0a65
YH		 2717}
2718
cf7fbe66
JS		 2719static inline bool
2720skb_sk_is_prefetched(struct sk_buff *skb)
2721{
2722#ifdef CONFIG_INET
2723 return skb->destructor == sock_pfree;
2724#else
2725 return false;
2726#endif /* CONFIG_INET */
2727}
2728
7ae215d2
JS		 2729/* This helper checks if a socket is a full socket,
2730 * ie _not_ a timewait or request socket.
2731 */
2732static inline bool sk_fullsock(const struct sock *sk)
2733{
2734 return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV);
2735}
2736
2737static inline bool
2738sk_is_refcounted(struct sock *sk)
2739{
2740 /* Only full sockets have sk->sk_flags. */
2741 return !sk_fullsock(sk) || !sock_flag(sk, SOCK_RCU_FREE);
2742}
2743
71489e21 2744/**
045065f0
LR		 2745 * skb_steal_sock - steal a socket from an sk_buff
2746 * @skb: sk_buff to steal the socket from
2747 * @refcounted: is set to true if the socket is reference-counted
71489e21
JS		 2748 */
2749static inline struct sock *
2750skb_steal_sock(struct sk_buff *skb, bool *refcounted)
23542618 2751{
efc27f8c 2752 if (skb->sk) {
23542618
KK		 2753 struct sock *sk = skb->sk;
2754
71489e21 2755 *refcounted = true;
7ae215d2
JS		 2756 if (skb_sk_is_prefetched(skb))
2757 *refcounted = sk_is_refcounted(sk);
23542618
KK		 2758 skb->destructor = NULL;
2759 skb->sk = NULL;
2760 return sk;
2761 }
71489e21 2762 *refcounted = false;
23542618
KK		 2763 return NULL;
2764}
2765
ebf4e808
IL		 2766/* Checks if this SKB belongs to an HW offloaded socket
2767 * and whether any SW fallbacks are required based on dev.
41477662 2768 * Check decrypted mark in case skb_orphan() cleared socket.
ebf4e808
IL		 2769 */
2770static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb,
2771 struct net_device *dev)
2772{
2773#ifdef CONFIG_SOCK_VALIDATE_XMIT
2774 struct sock *sk = skb->sk;
2775
41477662 2776 if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb) {
ebf4e808 2777 skb = sk->sk_validate_xmit_skb(sk, dev, skb);
41477662
JK		 2778#ifdef CONFIG_TLS_DEVICE
2779 } else if (unlikely(skb->decrypted)) {
2780 pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n");
2781 kfree_skb(skb);
2782 skb = NULL;
2783#endif
2784 }
ebf4e808
IL		 2785#endif
2786
2787 return skb;
2788}
2789
e446f9df
ED		 2790/* This helper checks if a socket is a LISTEN or NEW_SYN_RECV
2791 * SYNACK messages can be attached to either ones (depending on SYNCOOKIE)
2792 */
2793static inline bool sk_listener(const struct sock *sk)
2794{
2795 return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV);
2796}
2797
193d357d 2798void sock_enable_timestamp(struct sock *sk, enum sock_flags flag);
69336bd2
JP		 2799int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
2800 int type);
1da177e4 2801
a3b299da
EB		 2802bool sk_ns_capable(const struct sock *sk,
2803 struct user_namespace *user_ns, int cap);
2804bool sk_capable(const struct sock *sk, int cap);
2805bool sk_net_capable(const struct sock *sk, int cap);
2806
a2d133b1
JH		 2807void sk_get_meminfo(const struct sock *sk, u32 *meminfo);
2808
eaa72dc4
ED		 2809/* Take into consideration the size of the struct sk_buff overhead in the
2810 * determination of these values, since that is non-constant across
2811 * platforms. This makes socket queueing behavior and performance
2812 * not depend upon such differences.
2813 */
2814#define _SK_MEM_PACKETS 256
2815#define _SK_MEM_OVERHEAD SKB_TRUESIZE(256)
2816#define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2817#define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS)
2818
1da177e4
LT		 2819extern __u32 sysctl_wmem_max;
2820extern __u32 sysctl_rmem_max;
2821
b245be1f 2822extern int sysctl_tstamp_allow_data;
6baf1f41
DM		 2823extern int sysctl_optmem_max;
2824
20380731
ACM		 2825extern __u32 sysctl_wmem_default;
2826extern __u32 sysctl_rmem_default;
20380731 2827
723783d0 2828#define SKB_FRAG_PAGE_ORDER get_order(32768)
ce27ec60
ED		 2829DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key);
2830
a3dcaf17
ED		 2831static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto)
2832{
2833 /* Does this proto have per netns sysctl_wmem ? */
2834 if (proto->sysctl_wmem_offset)
2835 return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset);
2836
2837 return *proto->sysctl_wmem;
2838}
2839
2840static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto)
2841{
2842 /* Does this proto have per netns sysctl_rmem ? */
2843 if (proto->sysctl_rmem_offset)
2844 return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset);
2845
2846 return *proto->sysctl_rmem;
2847}
2848
c9f1f58d
ED		 2849/* Default TCP Small queue budget is ~1 ms of data (1sec >> 10)
2850 * Some wifi drivers need to tweak it to get more chunks.
2851 * They can use this helper from their ndo_start_xmit()
2852 */
2853static inline void sk_pacing_shift_update(struct sock *sk, int val)
2854{
7c68fa2b 2855 if (!sk || !sk_fullsock(sk) || READ_ONCE(sk->sk_pacing_shift) == val)
c9f1f58d 2856 return;
7c68fa2b 2857 WRITE_ONCE(sk->sk_pacing_shift, val);
c9f1f58d
ED		 2858}
2859
54dc3e33
DA		 2860/* if a socket is bound to a device, check that the given device
2861 * index is either the same or that the socket is bound to an L3
2862 * master device and the given device index is also enslaved to
2863 * that L3 master
2864 */
2865static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif)
2866{
2867 int mdif;
2868
2869 if (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif)
2870 return true;
2871
2872 mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif);
2873 if (mdif && mdif == sk->sk_bound_dev_if)
2874 return true;
2875
2876 return false;
2877}
2878
43a825af
BT		 2879void sock_def_readable(struct sock *sk);
2880
8ea204c2 2881int sock_bindtoindex(struct sock *sk, int ifindex, bool lock_sk);
371087aa 2882void sock_set_timestamp(struct sock *sk, int optname, bool valbool);
d463126e
YL		 2883int sock_set_timestamping(struct sock *sk, int optname,
2884 struct so_timestamping timestamping);
ced122d9 2885
783da70e 2886void sock_enable_timestamps(struct sock *sk);
c433594c 2887void sock_no_linger(struct sock *sk);
ce3d9544 2888void sock_set_keepalive(struct sock *sk);
6e434967 2889void sock_set_priority(struct sock *sk, u32 priority);
26cfabf9 2890void sock_set_rcvbuf(struct sock *sk, int val);
84d1c617 2891void sock_set_mark(struct sock *sk, u32 val);
b58f0e8f 2892void sock_set_reuseaddr(struct sock *sk);
fe31a326 2893void sock_set_reuseport(struct sock *sk);
76ee0785 2894void sock_set_sndtimeo(struct sock *sk, s64 secs);
b58f0e8f 2895
c0425a42
CH		 2896int sock_bind_add(struct sock *sk, struct sockaddr *addr, int addr_len);
2897
4c1e34c0
RP		 2898int sock_get_timeout(long timeo, void *optval, bool old_timeval);
2899int sock_copy_user_timeval(struct __kernel_sock_timeval *tv,
2900 sockptr_t optval, int optlen, bool old_timeval);
2901
7b50ecfc
CW		 2902static inline bool sk_is_readable(struct sock *sk)
2903{
2904 if (sk->sk_prot->sock_is_readable)
2905 return sk->sk_prot->sock_is_readable(sk);
2906 return false;
2907}
1da177e4 2908#endif /* _SOCK_H */
Linux 6.x block layer and io_uring tree(s)