]>
Commit | Line | Data |
---|---|---|
2874c5fd | 1 | /* SPDX-License-Identifier: GPL-2.0-or-later */ |
1da177e4 LT |
2 | /* |
3 | * INET An implementation of the TCP/IP protocol suite for the LINUX | |
4 | * operating system. INET is implemented using the BSD Socket | |
5 | * interface as the means of communication with the user level. | |
6 | * | |
7 | * Definitions for the AF_INET socket handler. | |
8 | * | |
9 | * Version: @(#)sock.h 1.0.4 05/13/93 | |
10 | * | |
02c30a84 | 11 | * Authors: Ross Biro |
1da177e4 LT |
12 | * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG> |
13 | * Corey Minyard <wf-rch!minyard@relay.EU.net> | |
14 | * Florian La Roche <flla@stud.uni-sb.de> | |
15 | * | |
16 | * Fixes: | |
17 | * Alan Cox : Volatiles in skbuff pointers. See | |
18 | * skbuff comments. May be overdone, | |
19 | * better to prove they can be removed | |
20 | * than the reverse. | |
21 | * Alan Cox : Added a zapped field for tcp to note | |
22 | * a socket is reset and must stay shut up | |
23 | * Alan Cox : New fields for options | |
24 | * Pauline Middelink : identd support | |
25 | * Alan Cox : Eliminate low level recv/recvfrom | |
26 | * David S. Miller : New socket lookup architecture. | |
27 | * Steve Whitehouse: Default routines for sock_ops | |
28 | * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made | |
29 | * protinfo be just a void pointer, as the | |
30 | * protocol specific parts were moved to | |
31 | * respective headers and ipv4/v6, etc now | |
32 | * use private slabcaches for its socks | |
33 | * Pedro Hortas : New flags field for socket options | |
1da177e4 LT |
34 | */ |
35 | #ifndef _SOCK_H | |
36 | #define _SOCK_H | |
37 | ||
a6b7a407 | 38 | #include <linux/hardirq.h> |
172589cc | 39 | #include <linux/kernel.h> |
1da177e4 | 40 | #include <linux/list.h> |
88ab1932 | 41 | #include <linux/list_nulls.h> |
1da177e4 LT |
42 | #include <linux/timer.h> |
43 | #include <linux/cache.h> | |
3f134619 | 44 | #include <linux/bitops.h> |
a5b5bb9a | 45 | #include <linux/lockdep.h> |
1da177e4 LT |
46 | #include <linux/netdevice.h> |
47 | #include <linux/skbuff.h> /* struct sk_buff */ | |
d7fe0f24 | 48 | #include <linux/mm.h> |
1da177e4 | 49 | #include <linux/security.h> |
5a0e3ad6 | 50 | #include <linux/slab.h> |
c6e1a0d1 | 51 | #include <linux/uaccess.h> |
3e32cb2e | 52 | #include <linux/page_counter.h> |
180d8cd9 | 53 | #include <linux/memcontrol.h> |
c5905afb | 54 | #include <linux/static_key.h> |
40401530 | 55 | #include <linux/sched.h> |
1ce0bf50 | 56 | #include <linux/wait.h> |
2a56a1fe | 57 | #include <linux/cgroup-defs.h> |
75c119af | 58 | #include <linux/rbtree.h> |
1da177e4 | 59 | #include <linux/filter.h> |
88ab1932 | 60 | #include <linux/rculist_nulls.h> |
a57de0b4 | 61 | #include <linux/poll.h> |
c8c1bbb6 | 62 | #include <linux/sockptr.h> |
1c5f2ced | 63 | #include <linux/indirect_call_wrapper.h> |
c31504dc | 64 | #include <linux/atomic.h> |
41c6d650 | 65 | #include <linux/refcount.h> |
1da177e4 LT |
66 | #include <net/dst.h> |
67 | #include <net/checksum.h> | |
1d0ab253 | 68 | #include <net/tcp_states.h> |
b9f40e21 | 69 | #include <linux/net_tstamp.h> |
54dc3e33 | 70 | #include <net/l3mdev.h> |
04190bf8 | 71 | #include <uapi/linux/socket.h> |
1da177e4 LT |
72 | |
73 | /* | |
74 | * This structure really needs to be cleaned up. | |
75 | * Most of it is for TCP, and not used by any of | |
76 | * the other protocols. | |
77 | */ | |
78 | ||
79 | /* Define this to get the SOCK_DBG debugging facility. */ | |
80 | #define SOCK_DEBUGGING | |
81 | #ifdef SOCK_DEBUGGING | |
82 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
83 | printk(KERN_DEBUG msg); } while (0) | |
84 | #else | |
4cd9029d | 85 | /* Validate arguments and do nothing */ |
b9075fa9 | 86 | static inline __printf(2, 3) |
dc6b9b78 | 87 | void 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 |
96 | typedef 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 | 111 | struct sock; |
8feaf0c0 | 112 | struct proto; |
0eeb8ffc | 113 | struct net; |
1da177e4 | 114 | |
077b393d ED |
115 | typedef __u32 __bitwise __portpair; |
116 | typedef __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 | 163 | struct 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 | 250 | struct bpf_local_storage; |
6ac99e8f | 251 | |
1da177e4 LT |
252 | /** |
253 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 254 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
255 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
256 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
257 | * @sk_lock: synchronizer | |
cdfbabfb | 258 | * @sk_kern_sock: True if sock is using kernel lock classes |
4dc3b16b | 259 | * @sk_rcvbuf: size of receive buffer in bytes |
43815482 | 260 | * @sk_wq: sock wait queue and async head |
421b3885 | 261 | * @sk_rx_dst: receive input route used by early demux |
4dc3b16b | 262 | * @sk_dst_cache: destination cache |
9b8805a3 | 263 | * @sk_dst_pending_confirm: need to confirm neighbour |
4dc3b16b | 264 | * @sk_policy: flow policy |
66256e0b | 265 | * @sk_rx_skb_cache: cache copy of recently accessed RX skb |
4dc3b16b PP |
266 | * @sk_receive_queue: incoming packets |
267 | * @sk_wmem_alloc: transmit queue bytes committed | |
771edcaf | 268 | * @sk_tsq_flags: TCP Small Queues flags |
4dc3b16b PP |
269 | * @sk_write_queue: Packet sending queue |
270 | * @sk_omem_alloc: "o" is "option" or "other" | |
271 | * @sk_wmem_queued: persistent queue size | |
272 | * @sk_forward_alloc: space allocated forward | |
06021292 | 273 | * @sk_napi_id: id of the last napi context to receive data for sk |
dafcc438 | 274 | * @sk_ll_usec: usecs to busypoll when there is no data |
4dc3b16b | 275 | * @sk_allocation: allocation mode |
95bd09eb | 276 | * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler) |
218af599 | 277 | * @sk_pacing_status: Pacing status (requested, handled by sch_fq) |
c3f40d7c | 278 | * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE) |
4dc3b16b | 279 | * @sk_sndbuf: size of send buffer in bytes |
771edcaf | 280 | * @__sk_flags_offset: empty field used to determine location of bitfield |
293de7de | 281 | * @sk_padding: unused element for alignment |
28448b80 TH |
282 | * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets |
283 | * @sk_no_check_rx: allow zero checksum in RX packets | |
4dc3b16b | 284 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
a465419b | 285 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
66256e0b RD |
286 | * @sk_route_forced_caps: static, forced route capabilities |
287 | * (set in tcp_init_sock()) | |
bcd76111 | 288 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
82cc1a7a | 289 | * @sk_gso_max_size: Maximum GSO segment size to build |
1485348d | 290 | * @sk_gso_max_segs: Maximum number of GSO segments |
3a9b76fd | 291 | * @sk_pacing_shift: scaling factor for TCP Small Queues |
4dc3b16b | 292 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
293 | * @sk_backlog: always used with the per-socket spinlock held |
294 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
295 | * @sk_error_queue: rarely used | |
33c732c3 WC |
296 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
297 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 298 | * @sk_err: last error |
33c732c3 WC |
299 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
300 | * persistent failure not just 'timed out' | |
cb61cb9b | 301 | * @sk_drops: raw/udp drops counter |
4dc3b16b PP |
302 | * @sk_ack_backlog: current listen backlog |
303 | * @sk_max_ack_backlog: listen backlog set in listen() | |
771edcaf | 304 | * @sk_uid: user id of owner |
7fd3253a | 305 | * @sk_prefer_busy_poll: prefer busypolling over softirq processing |
7c951caf | 306 | * @sk_busy_poll_budget: napi processing budget when busypolling |
4dc3b16b PP |
307 | * @sk_priority: %SO_PRIORITY setting |
308 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
309 | * @sk_protocol: which protocol this socket belongs in this network family | |
5fb14d20 | 310 | * @sk_peer_lock: lock protecting @sk_peer_pid and @sk_peer_cred |
53c3fa20 RD |
311 | * @sk_peer_pid: &struct pid for this socket's peer |
312 | * @sk_peer_cred: %SO_PEERCRED setting | |
4dc3b16b PP |
313 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
314 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
315 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
b73c3d0e | 316 | * @sk_txhash: computed flow hash for use on transmit |
4dc3b16b | 317 | * @sk_filter: socket filtering instructions |
4dc3b16b PP |
318 | * @sk_timer: sock cleanup timer |
319 | * @sk_stamp: time stamp of last packet received | |
3a0ed3e9 | 320 | * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only |
d463126e YL |
321 | * @sk_tsflags: SO_TIMESTAMPING flags |
322 | * @sk_bind_phc: SO_TIMESTAMPING bind PHC index of PTP virtual clock | |
323 | * for timestamping | |
09c2d251 | 324 | * @sk_tskey: counter to disambiguate concurrent tstamp requests |
52267790 | 325 | * @sk_zckey: counter to order MSG_ZEROCOPY notifications |
4dc3b16b PP |
326 | * @sk_socket: Identd and reporting IO signals |
327 | * @sk_user_data: RPC layer private data | |
5640f768 | 328 | * @sk_frag: cached page frag |
d3d4f0a0 | 329 | * @sk_peek_off: current peek_offset value |
4dc3b16b | 330 | * @sk_send_head: front of stuff to transmit |
66256e0b RD |
331 | * @tcp_rtx_queue: TCP re-transmit queue [union with @sk_send_head] |
332 | * @sk_tx_skb_cache: cache copy of recently accessed TX skb | |
67be2dd1 | 333 | * @sk_security: used by security modules |
31729363 | 334 | * @sk_mark: generic packet mark |
2a56a1fe | 335 | * @sk_cgrp_data: cgroup data for this cgroup |
baac50bb | 336 | * @sk_memcg: this socket's memory cgroup association |
4dc3b16b PP |
337 | * @sk_write_pending: a write to stream socket waits to start |
338 | * @sk_state_change: callback to indicate change in the state of the sock | |
339 | * @sk_data_ready: callback to indicate there is data to be processed | |
340 | * @sk_write_space: callback to indicate there is bf sending space available | |
341 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
342 | * @sk_backlog_rcv: callback to process the backlog | |
66256e0b | 343 | * @sk_validate_xmit_skb: ptr to an optional validate function |
4dc3b16b | 344 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 |
ef456144 | 345 | * @sk_reuseport_cb: reuseport group container |
66256e0b | 346 | * @sk_bpf_storage: ptr to cache and control for bpf_sk_storage |
293de7de | 347 | * @sk_rcu: used during RCU grace period |
80b14dee RC |
348 | * @sk_clockid: clockid used by time-based scheduling (SO_TXTIME) |
349 | * @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME | |
66256e0b | 350 | * @sk_txtime_report_errors: set report errors mode for SO_TXTIME |
80b14dee | 351 | * @sk_txtime_unused: unused txtime flags |
293de7de | 352 | */ |
1da177e4 LT |
353 | struct sock { |
354 | /* | |
8feaf0c0 | 355 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
356 | * don't add nothing before this first member (__sk_common) --acme |
357 | */ | |
358 | struct sock_common __sk_common; | |
4dc6dc71 ED |
359 | #define sk_node __sk_common.skc_node |
360 | #define sk_nulls_node __sk_common.skc_nulls_node | |
361 | #define sk_refcnt __sk_common.skc_refcnt | |
e022f0b4 | 362 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
4e1beecc | 363 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
c6345ce7 AN |
364 | #define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping |
365 | #endif | |
4dc6dc71 | 366 | |
68835aba ED |
367 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
368 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end | |
4dc6dc71 | 369 | #define sk_hash __sk_common.skc_hash |
50805466 | 370 | #define sk_portpair __sk_common.skc_portpair |
05dbc7b5 ED |
371 | #define sk_num __sk_common.skc_num |
372 | #define sk_dport __sk_common.skc_dport | |
50805466 ED |
373 | #define sk_addrpair __sk_common.skc_addrpair |
374 | #define sk_daddr __sk_common.skc_daddr | |
375 | #define sk_rcv_saddr __sk_common.skc_rcv_saddr | |
1da177e4 LT |
376 | #define sk_family __sk_common.skc_family |
377 | #define sk_state __sk_common.skc_state | |
378 | #define sk_reuse __sk_common.skc_reuse | |
055dc21a | 379 | #define sk_reuseport __sk_common.skc_reuseport |
9fe516ba | 380 | #define sk_ipv6only __sk_common.skc_ipv6only |
26abe143 | 381 | #define sk_net_refcnt __sk_common.skc_net_refcnt |
1da177e4 | 382 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
1da177e4 | 383 | #define sk_bind_node __sk_common.skc_bind_node |
8feaf0c0 | 384 | #define sk_prot __sk_common.skc_prot |
07feaebf | 385 | #define sk_net __sk_common.skc_net |
efe4208f ED |
386 | #define sk_v6_daddr __sk_common.skc_v6_daddr |
387 | #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr | |
33cf7c90 | 388 | #define sk_cookie __sk_common.skc_cookie |
70da268b | 389 | #define sk_incoming_cpu __sk_common.skc_incoming_cpu |
8e5eb54d | 390 | #define sk_flags __sk_common.skc_flags |
ed53d0ab | 391 | #define sk_rxhash __sk_common.skc_rxhash |
efe4208f | 392 | |
1da177e4 | 393 | socket_lock_t sk_lock; |
9115e8cd ED |
394 | atomic_t sk_drops; |
395 | int sk_rcvlowat; | |
396 | struct sk_buff_head sk_error_queue; | |
8b27dae5 | 397 | struct sk_buff *sk_rx_skb_cache; |
b178bb3d | 398 | struct sk_buff_head sk_receive_queue; |
fa438ccf ED |
399 | /* |
400 | * The backlog queue is special, it is always used with | |
401 | * the per-socket spinlock held and requires low latency | |
402 | * access. Therefore we special case it's implementation. | |
b178bb3d ED |
403 | * Note : rmem_alloc is in this structure to fill a hole |
404 | * on 64bit arches, not because its logically part of | |
405 | * backlog. | |
fa438ccf ED |
406 | */ |
407 | struct { | |
b178bb3d ED |
408 | atomic_t rmem_alloc; |
409 | int len; | |
410 | struct sk_buff *head; | |
411 | struct sk_buff *tail; | |
fa438ccf | 412 | } sk_backlog; |
b178bb3d | 413 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
2c8c56e1 | 414 | |
9115e8cd | 415 | int sk_forward_alloc; |
e0d1095a | 416 | #ifdef CONFIG_NET_RX_BUSY_POLL |
dafcc438 | 417 | unsigned int sk_ll_usec; |
9115e8cd ED |
418 | /* ===== mostly read cache line ===== */ |
419 | unsigned int sk_napi_id; | |
b178bb3d | 420 | #endif |
b178bb3d ED |
421 | int sk_rcvbuf; |
422 | ||
423 | struct sk_filter __rcu *sk_filter; | |
ceb5d58b ED |
424 | union { |
425 | struct socket_wq __rcu *sk_wq; | |
66256e0b | 426 | /* private: */ |
ceb5d58b | 427 | struct socket_wq *sk_wq_raw; |
66256e0b | 428 | /* public: */ |
ceb5d58b | 429 | }; |
def8b4fa | 430 | #ifdef CONFIG_XFRM |
d188ba86 | 431 | struct xfrm_policy __rcu *sk_policy[2]; |
def8b4fa | 432 | #endif |
deaa5854 | 433 | struct dst_entry *sk_rx_dst; |
0e36cbb3 | 434 | struct dst_entry __rcu *sk_dst_cache; |
1da177e4 | 435 | atomic_t sk_omem_alloc; |
4e07a91c | 436 | int sk_sndbuf; |
9115e8cd ED |
437 | |
438 | /* ===== cache line for TX ===== */ | |
439 | int sk_wmem_queued; | |
14afee4b | 440 | refcount_t sk_wmem_alloc; |
9115e8cd | 441 | unsigned long sk_tsq_flags; |
75c119af ED |
442 | union { |
443 | struct sk_buff *sk_send_head; | |
444 | struct rb_root tcp_rtx_queue; | |
445 | }; | |
472c2e07 | 446 | struct sk_buff *sk_tx_skb_cache; |
1da177e4 | 447 | struct sk_buff_head sk_write_queue; |
9115e8cd ED |
448 | __s32 sk_peek_off; |
449 | int sk_write_pending; | |
9b8805a3 | 450 | __u32 sk_dst_pending_confirm; |
218af599 | 451 | u32 sk_pacing_status; /* see enum sk_pacing */ |
9115e8cd ED |
452 | long sk_sndtimeo; |
453 | struct timer_list sk_timer; | |
454 | __u32 sk_priority; | |
455 | __u32 sk_mark; | |
76a9ebe8 ED |
456 | unsigned long sk_pacing_rate; /* bytes per second */ |
457 | unsigned long sk_max_pacing_rate; | |
9115e8cd ED |
458 | struct page_frag sk_frag; |
459 | netdev_features_t sk_route_caps; | |
460 | netdev_features_t sk_route_nocaps; | |
0a6b2a1d | 461 | netdev_features_t sk_route_forced_caps; |
9115e8cd ED |
462 | int sk_gso_type; |
463 | unsigned int sk_gso_max_size; | |
464 | gfp_t sk_allocation; | |
465 | __u32 sk_txhash; | |
fc64869c AR |
466 | |
467 | /* | |
468 | * Because of non atomicity rules, all | |
469 | * changes are protected by socket lock. | |
470 | */ | |
bf976514 | 471 | u8 sk_padding : 1, |
cdfbabfb | 472 | sk_kern_sock : 1, |
28448b80 TH |
473 | sk_no_check_tx : 1, |
474 | sk_no_check_rx : 1, | |
bf976514 | 475 | sk_userlocks : 4; |
3a9b76fd | 476 | u8 sk_pacing_shift; |
bf976514 MM |
477 | u16 sk_type; |
478 | u16 sk_protocol; | |
479 | u16 sk_gso_max_segs; | |
1da177e4 | 480 | unsigned long sk_lingertime; |
476e19cf | 481 | struct proto *sk_prot_creator; |
1da177e4 LT |
482 | rwlock_t sk_callback_lock; |
483 | int sk_err, | |
484 | sk_err_soft; | |
becb74f0 ED |
485 | u32 sk_ack_backlog; |
486 | u32 sk_max_ack_backlog; | |
86741ec2 | 487 | kuid_t sk_uid; |
7fd3253a BT |
488 | #ifdef CONFIG_NET_RX_BUSY_POLL |
489 | u8 sk_prefer_busy_poll; | |
7c951caf | 490 | u16 sk_busy_poll_budget; |
7fd3253a | 491 | #endif |
35306eb2 | 492 | spinlock_t sk_peer_lock; |
109f6e39 EB |
493 | struct pid *sk_peer_pid; |
494 | const struct cred *sk_peer_cred; | |
35306eb2 | 495 | |
1da177e4 | 496 | long sk_rcvtimeo; |
b7aa0bf7 | 497 | ktime_t sk_stamp; |
3a0ed3e9 DD |
498 | #if BITS_PER_LONG==32 |
499 | seqlock_t sk_stamp_seq; | |
500 | #endif | |
b9f40e21 | 501 | u16 sk_tsflags; |
d463126e | 502 | int sk_bind_phc; |
fc64869c | 503 | u8 sk_shutdown; |
09c2d251 | 504 | u32 sk_tskey; |
52267790 | 505 | atomic_t sk_zckey; |
80b14dee RC |
506 | |
507 | u8 sk_clockid; | |
508 | u8 sk_txtime_deadline_mode : 1, | |
4b15c707 JSP |
509 | sk_txtime_report_errors : 1, |
510 | sk_txtime_unused : 6; | |
80b14dee | 511 | |
1da177e4 LT |
512 | struct socket *sk_socket; |
513 | void *sk_user_data; | |
d5f64238 | 514 | #ifdef CONFIG_SECURITY |
1da177e4 | 515 | void *sk_security; |
d5f64238 | 516 | #endif |
2a56a1fe | 517 | struct sock_cgroup_data sk_cgrp_data; |
baac50bb | 518 | struct mem_cgroup *sk_memcg; |
1da177e4 | 519 | void (*sk_state_change)(struct sock *sk); |
676d2369 | 520 | void (*sk_data_ready)(struct sock *sk); |
1da177e4 LT |
521 | void (*sk_write_space)(struct sock *sk); |
522 | void (*sk_error_report)(struct sock *sk); | |
dc6b9b78 ED |
523 | int (*sk_backlog_rcv)(struct sock *sk, |
524 | struct sk_buff *skb); | |
ebf4e808 IL |
525 | #ifdef CONFIG_SOCK_VALIDATE_XMIT |
526 | struct sk_buff* (*sk_validate_xmit_skb)(struct sock *sk, | |
527 | struct net_device *dev, | |
528 | struct sk_buff *skb); | |
529 | #endif | |
1da177e4 | 530 | void (*sk_destruct)(struct sock *sk); |
ef456144 | 531 | struct sock_reuseport __rcu *sk_reuseport_cb; |
6ac99e8f | 532 | #ifdef CONFIG_BPF_SYSCALL |
1f00d375 | 533 | struct bpf_local_storage __rcu *sk_bpf_storage; |
6ac99e8f | 534 | #endif |
a4298e45 | 535 | struct rcu_head sk_rcu; |
1da177e4 LT |
536 | }; |
537 | ||
218af599 ED |
538 | enum sk_pacing { |
539 | SK_PACING_NONE = 0, | |
540 | SK_PACING_NEEDED = 1, | |
541 | SK_PACING_FQ = 2, | |
542 | }; | |
543 | ||
f1ff5ce2 JS |
544 | /* Pointer stored in sk_user_data might not be suitable for copying |
545 | * when cloning the socket. For instance, it can point to a reference | |
546 | * counted object. sk_user_data bottom bit is set if pointer must not | |
547 | * be copied. | |
548 | */ | |
549 | #define SK_USER_DATA_NOCOPY 1UL | |
c9a368f1 MKL |
550 | #define SK_USER_DATA_BPF 2UL /* Managed by BPF */ |
551 | #define SK_USER_DATA_PTRMASK ~(SK_USER_DATA_NOCOPY | SK_USER_DATA_BPF) | |
f1ff5ce2 JS |
552 | |
553 | /** | |
554 | * sk_user_data_is_nocopy - Test if sk_user_data pointer must not be copied | |
555 | * @sk: socket | |
556 | */ | |
557 | static inline bool sk_user_data_is_nocopy(const struct sock *sk) | |
558 | { | |
559 | return ((uintptr_t)sk->sk_user_data & SK_USER_DATA_NOCOPY); | |
560 | } | |
561 | ||
559835ea PS |
562 | #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data))) |
563 | ||
f1ff5ce2 JS |
564 | #define rcu_dereference_sk_user_data(sk) \ |
565 | ({ \ | |
566 | void *__tmp = rcu_dereference(__sk_user_data((sk))); \ | |
567 | (void *)((uintptr_t)__tmp & SK_USER_DATA_PTRMASK); \ | |
568 | }) | |
569 | #define rcu_assign_sk_user_data(sk, ptr) \ | |
570 | ({ \ | |
571 | uintptr_t __tmp = (uintptr_t)(ptr); \ | |
572 | WARN_ON_ONCE(__tmp & ~SK_USER_DATA_PTRMASK); \ | |
573 | rcu_assign_pointer(__sk_user_data((sk)), __tmp); \ | |
574 | }) | |
575 | #define rcu_assign_sk_user_data_nocopy(sk, ptr) \ | |
576 | ({ \ | |
577 | uintptr_t __tmp = (uintptr_t)(ptr); \ | |
578 | WARN_ON_ONCE(__tmp & ~SK_USER_DATA_PTRMASK); \ | |
579 | rcu_assign_pointer(__sk_user_data((sk)), \ | |
580 | __tmp | SK_USER_DATA_NOCOPY); \ | |
581 | }) | |
559835ea | 582 | |
4a17fd52 PE |
583 | /* |
584 | * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK | |
585 | * or not whether his port will be reused by someone else. SK_FORCE_REUSE | |
586 | * on a socket means that the socket will reuse everybody else's port | |
587 | * without looking at the other's sk_reuse value. | |
588 | */ | |
589 | ||
590 | #define SK_NO_REUSE 0 | |
591 | #define SK_CAN_REUSE 1 | |
592 | #define SK_FORCE_REUSE 2 | |
593 | ||
627d2d6b | 594 | int sk_set_peek_off(struct sock *sk, int val); |
595 | ||
ef64a54f PE |
596 | static inline int sk_peek_offset(struct sock *sk, int flags) |
597 | { | |
b9bb53f3 | 598 | if (unlikely(flags & MSG_PEEK)) { |
a0917e0b | 599 | return READ_ONCE(sk->sk_peek_off); |
b9bb53f3 WB |
600 | } |
601 | ||
602 | return 0; | |
ef64a54f PE |
603 | } |
604 | ||
605 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) | |
606 | { | |
b9bb53f3 WB |
607 | s32 off = READ_ONCE(sk->sk_peek_off); |
608 | ||
609 | if (unlikely(off >= 0)) { | |
610 | off = max_t(s32, off - val, 0); | |
611 | WRITE_ONCE(sk->sk_peek_off, off); | |
ef64a54f PE |
612 | } |
613 | } | |
614 | ||
615 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) | |
616 | { | |
b9bb53f3 | 617 | sk_peek_offset_bwd(sk, -val); |
ef64a54f PE |
618 | } |
619 | ||
1da177e4 LT |
620 | /* |
621 | * Hashed lists helper routines | |
622 | */ | |
c4146644 LZ |
623 | static inline struct sock *sk_entry(const struct hlist_node *node) |
624 | { | |
625 | return hlist_entry(node, struct sock, sk_node); | |
626 | } | |
627 | ||
e48c414e | 628 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
629 | { |
630 | return hlist_entry(head->first, struct sock, sk_node); | |
631 | } | |
632 | ||
e48c414e | 633 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
634 | { |
635 | return hlist_empty(head) ? NULL : __sk_head(head); | |
636 | } | |
637 | ||
88ab1932 ED |
638 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
639 | { | |
640 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); | |
641 | } | |
642 | ||
643 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) | |
644 | { | |
645 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); | |
646 | } | |
647 | ||
e48c414e | 648 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 | 649 | { |
6c59ebd3 | 650 | return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node); |
1da177e4 LT |
651 | } |
652 | ||
88ab1932 ED |
653 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
654 | { | |
655 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? | |
656 | hlist_nulls_entry(sk->sk_nulls_node.next, | |
657 | struct sock, sk_nulls_node) : | |
658 | NULL; | |
659 | } | |
660 | ||
dc6b9b78 | 661 | static inline bool sk_unhashed(const struct sock *sk) |
1da177e4 LT |
662 | { |
663 | return hlist_unhashed(&sk->sk_node); | |
664 | } | |
665 | ||
dc6b9b78 | 666 | static inline bool sk_hashed(const struct sock *sk) |
1da177e4 | 667 | { |
da753bea | 668 | return !sk_unhashed(sk); |
1da177e4 LT |
669 | } |
670 | ||
dc6b9b78 | 671 | static inline void sk_node_init(struct hlist_node *node) |
1da177e4 LT |
672 | { |
673 | node->pprev = NULL; | |
674 | } | |
675 | ||
dc6b9b78 | 676 | static inline void sk_nulls_node_init(struct hlist_nulls_node *node) |
88ab1932 ED |
677 | { |
678 | node->pprev = NULL; | |
679 | } | |
680 | ||
dc6b9b78 | 681 | static inline void __sk_del_node(struct sock *sk) |
1da177e4 LT |
682 | { |
683 | __hlist_del(&sk->sk_node); | |
684 | } | |
685 | ||
808f5114 | 686 | /* NB: equivalent to hlist_del_init_rcu */ |
dc6b9b78 | 687 | static inline bool __sk_del_node_init(struct sock *sk) |
1da177e4 LT |
688 | { |
689 | if (sk_hashed(sk)) { | |
690 | __sk_del_node(sk); | |
691 | sk_node_init(&sk->sk_node); | |
dc6b9b78 | 692 | return true; |
1da177e4 | 693 | } |
dc6b9b78 | 694 | return false; |
1da177e4 LT |
695 | } |
696 | ||
697 | /* Grab socket reference count. This operation is valid only | |
698 | when sk is ALREADY grabbed f.e. it is found in hash table | |
699 | or a list and the lookup is made under lock preventing hash table | |
700 | modifications. | |
701 | */ | |
702 | ||
f9a7cbbf | 703 | static __always_inline void sock_hold(struct sock *sk) |
1da177e4 | 704 | { |
41c6d650 | 705 | refcount_inc(&sk->sk_refcnt); |
1da177e4 LT |
706 | } |
707 | ||
708 | /* Ungrab socket in the context, which assumes that socket refcnt | |
709 | cannot hit zero, f.e. it is true in context of any socketcall. | |
710 | */ | |
f9a7cbbf | 711 | static __always_inline void __sock_put(struct sock *sk) |
1da177e4 | 712 | { |
41c6d650 | 713 | refcount_dec(&sk->sk_refcnt); |
1da177e4 LT |
714 | } |
715 | ||
dc6b9b78 | 716 | static inline bool sk_del_node_init(struct sock *sk) |
1da177e4 | 717 | { |
dc6b9b78 | 718 | bool rc = __sk_del_node_init(sk); |
1da177e4 LT |
719 | |
720 | if (rc) { | |
721 | /* paranoid for a while -acme */ | |
41c6d650 | 722 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
1da177e4 LT |
723 | __sock_put(sk); |
724 | } | |
725 | return rc; | |
726 | } | |
808f5114 | 727 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
1da177e4 | 728 | |
dc6b9b78 | 729 | static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 ED |
730 | { |
731 | if (sk_hashed(sk)) { | |
88ab1932 | 732 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
dc6b9b78 | 733 | return true; |
271b72c7 | 734 | } |
dc6b9b78 | 735 | return false; |
271b72c7 ED |
736 | } |
737 | ||
dc6b9b78 | 738 | static inline bool sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 | 739 | { |
dc6b9b78 | 740 | bool rc = __sk_nulls_del_node_init_rcu(sk); |
271b72c7 ED |
741 | |
742 | if (rc) { | |
743 | /* paranoid for a while -acme */ | |
41c6d650 | 744 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
271b72c7 ED |
745 | __sock_put(sk); |
746 | } | |
747 | return rc; | |
748 | } | |
749 | ||
dc6b9b78 | 750 | static inline void __sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
751 | { |
752 | hlist_add_head(&sk->sk_node, list); | |
753 | } | |
754 | ||
dc6b9b78 | 755 | static inline void sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
756 | { |
757 | sock_hold(sk); | |
758 | __sk_add_node(sk, list); | |
759 | } | |
760 | ||
dc6b9b78 | 761 | static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
808f5114 | 762 | { |
763 | sock_hold(sk); | |
d296ba60 CG |
764 | if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport && |
765 | sk->sk_family == AF_INET6) | |
766 | hlist_add_tail_rcu(&sk->sk_node, list); | |
767 | else | |
768 | hlist_add_head_rcu(&sk->sk_node, list); | |
808f5114 | 769 | } |
770 | ||
a4dc6a49 MC |
771 | static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list) |
772 | { | |
773 | sock_hold(sk); | |
774 | hlist_add_tail_rcu(&sk->sk_node, list); | |
775 | } | |
776 | ||
dc6b9b78 | 777 | static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 | 778 | { |
d7efc6c1 | 779 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
271b72c7 ED |
780 | } |
781 | ||
8dbd76e7 ED |
782 | static inline void __sk_nulls_add_node_tail_rcu(struct sock *sk, struct hlist_nulls_head *list) |
783 | { | |
784 | hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list); | |
785 | } | |
786 | ||
dc6b9b78 | 787 | static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 ED |
788 | { |
789 | sock_hold(sk); | |
88ab1932 | 790 | __sk_nulls_add_node_rcu(sk, list); |
271b72c7 ED |
791 | } |
792 | ||
dc6b9b78 | 793 | static inline void __sk_del_bind_node(struct sock *sk) |
1da177e4 LT |
794 | { |
795 | __hlist_del(&sk->sk_bind_node); | |
796 | } | |
797 | ||
dc6b9b78 | 798 | static inline void sk_add_bind_node(struct sock *sk, |
1da177e4 LT |
799 | struct hlist_head *list) |
800 | { | |
801 | hlist_add_head(&sk->sk_bind_node, list); | |
802 | } | |
803 | ||
b67bfe0d SL |
804 | #define sk_for_each(__sk, list) \ |
805 | hlist_for_each_entry(__sk, list, sk_node) | |
806 | #define sk_for_each_rcu(__sk, list) \ | |
807 | hlist_for_each_entry_rcu(__sk, list, sk_node) | |
88ab1932 ED |
808 | #define sk_nulls_for_each(__sk, node, list) \ |
809 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) | |
810 | #define sk_nulls_for_each_rcu(__sk, node, list) \ | |
811 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) | |
b67bfe0d SL |
812 | #define sk_for_each_from(__sk) \ |
813 | hlist_for_each_entry_from(__sk, sk_node) | |
88ab1932 ED |
814 | #define sk_nulls_for_each_from(__sk, node) \ |
815 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ | |
816 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) | |
b67bfe0d SL |
817 | #define sk_for_each_safe(__sk, tmp, list) \ |
818 | hlist_for_each_entry_safe(__sk, tmp, list, sk_node) | |
819 | #define sk_for_each_bound(__sk, list) \ | |
820 | hlist_for_each_entry(__sk, list, sk_bind_node) | |
1da177e4 | 821 | |
2dc41cff | 822 | /** |
ca065d0c | 823 | * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset |
2dc41cff DH |
824 | * @tpos: the type * to use as a loop cursor. |
825 | * @pos: the &struct hlist_node to use as a loop cursor. | |
826 | * @head: the head for your list. | |
827 | * @offset: offset of hlist_node within the struct. | |
828 | * | |
829 | */ | |
ca065d0c | 830 | #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \ |
b6f4f848 | 831 | for (pos = rcu_dereference(hlist_first_rcu(head)); \ |
ca065d0c | 832 | pos != NULL && \ |
2dc41cff | 833 | ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \ |
b6f4f848 | 834 | pos = rcu_dereference(hlist_next_rcu(pos))) |
2dc41cff | 835 | |
c336d148 EB |
836 | static inline struct user_namespace *sk_user_ns(struct sock *sk) |
837 | { | |
838 | /* Careful only use this in a context where these parameters | |
839 | * can not change and must all be valid, such as recvmsg from | |
840 | * userspace. | |
841 | */ | |
842 | return sk->sk_socket->file->f_cred->user_ns; | |
843 | } | |
844 | ||
1da177e4 LT |
845 | /* Sock flags */ |
846 | enum sock_flags { | |
847 | SOCK_DEAD, | |
848 | SOCK_DONE, | |
849 | SOCK_URGINLINE, | |
850 | SOCK_KEEPOPEN, | |
851 | SOCK_LINGER, | |
852 | SOCK_DESTROY, | |
853 | SOCK_BROADCAST, | |
854 | SOCK_TIMESTAMP, | |
855 | SOCK_ZAPPED, | |
856 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
857 | SOCK_DBG, /* %SO_DEBUG setting */ | |
858 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 859 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 | 860 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
7cb02404 | 861 | SOCK_MEMALLOC, /* VM depends on this socket for swapping */ |
20d49473 | 862 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
bcdce719 | 863 | SOCK_FASYNC, /* fasync() active */ |
3b885787 | 864 | SOCK_RXQ_OVFL, |
1cdebb42 | 865 | SOCK_ZEROCOPY, /* buffers from userspace */ |
6e3e939f | 866 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
3bdc0eba BG |
867 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
868 | * Will use last 4 bytes of packet sent from | |
869 | * user-space instead. | |
870 | */ | |
d59577b6 | 871 | SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */ |
7d4c04fc | 872 | SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */ |
a4298e45 | 873 | SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */ |
80b14dee | 874 | SOCK_TXTIME, |
e4a2a304 | 875 | SOCK_XDP, /* XDP is attached */ |
887feae3 | 876 | SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */ |
1da177e4 LT |
877 | }; |
878 | ||
01ce63c9 MRL |
879 | #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)) |
880 | ||
53b924b3 RB |
881 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
882 | { | |
883 | nsk->sk_flags = osk->sk_flags; | |
884 | } | |
885 | ||
1da177e4 LT |
886 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
887 | { | |
888 | __set_bit(flag, &sk->sk_flags); | |
889 | } | |
890 | ||
891 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
892 | { | |
893 | __clear_bit(flag, &sk->sk_flags); | |
894 | } | |
895 | ||
dfde1d7d DY |
896 | static inline void sock_valbool_flag(struct sock *sk, enum sock_flags bit, |
897 | int valbool) | |
898 | { | |
899 | if (valbool) | |
900 | sock_set_flag(sk, bit); | |
901 | else | |
902 | sock_reset_flag(sk, bit); | |
903 | } | |
904 | ||
1b23a5df | 905 | static inline bool sock_flag(const struct sock *sk, enum sock_flags flag) |
1da177e4 LT |
906 | { |
907 | return test_bit(flag, &sk->sk_flags); | |
908 | } | |
909 | ||
c93bdd0e | 910 | #ifdef CONFIG_NET |
a7950ae8 | 911 | DECLARE_STATIC_KEY_FALSE(memalloc_socks_key); |
c93bdd0e MG |
912 | static inline int sk_memalloc_socks(void) |
913 | { | |
a7950ae8 | 914 | return static_branch_unlikely(&memalloc_socks_key); |
c93bdd0e | 915 | } |
d9539752 KC |
916 | |
917 | void __receive_sock(struct file *file); | |
c93bdd0e MG |
918 | #else |
919 | ||
920 | static inline int sk_memalloc_socks(void) | |
921 | { | |
922 | return 0; | |
923 | } | |
924 | ||
d9539752 KC |
925 | static inline void __receive_sock(struct file *file) |
926 | { } | |
c93bdd0e MG |
927 | #endif |
928 | ||
7450aaf6 | 929 | static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask) |
99a1dec7 | 930 | { |
7450aaf6 | 931 | return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC); |
99a1dec7 MG |
932 | } |
933 | ||
1da177e4 LT |
934 | static inline void sk_acceptq_removed(struct sock *sk) |
935 | { | |
288efe86 | 936 | WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog - 1); |
1da177e4 LT |
937 | } |
938 | ||
939 | static inline void sk_acceptq_added(struct sock *sk) | |
940 | { | |
288efe86 | 941 | WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog + 1); |
1da177e4 LT |
942 | } |
943 | ||
c609e6aa ED |
944 | /* Note: If you think the test should be: |
945 | * return READ_ONCE(sk->sk_ack_backlog) >= READ_ONCE(sk->sk_max_ack_backlog); | |
946 | * Then please take a look at commit 64a146513f8f ("[NET]: Revert incorrect accept queue backlog changes.") | |
947 | */ | |
dc6b9b78 | 948 | static inline bool sk_acceptq_is_full(const struct sock *sk) |
1da177e4 | 949 | { |
c609e6aa | 950 | return READ_ONCE(sk->sk_ack_backlog) > READ_ONCE(sk->sk_max_ack_backlog); |
1da177e4 LT |
951 | } |
952 | ||
953 | /* | |
954 | * Compute minimal free write space needed to queue new packets. | |
955 | */ | |
dc6b9b78 | 956 | static inline int sk_stream_min_wspace(const struct sock *sk) |
1da177e4 | 957 | { |
ab4e846a | 958 | return READ_ONCE(sk->sk_wmem_queued) >> 1; |
1da177e4 LT |
959 | } |
960 | ||
dc6b9b78 | 961 | static inline int sk_stream_wspace(const struct sock *sk) |
1da177e4 | 962 | { |
ab4e846a ED |
963 | return READ_ONCE(sk->sk_sndbuf) - READ_ONCE(sk->sk_wmem_queued); |
964 | } | |
965 | ||
966 | static inline void sk_wmem_queued_add(struct sock *sk, int val) | |
967 | { | |
968 | WRITE_ONCE(sk->sk_wmem_queued, sk->sk_wmem_queued + val); | |
1da177e4 LT |
969 | } |
970 | ||
69336bd2 | 971 | void sk_stream_write_space(struct sock *sk); |
1da177e4 | 972 | |
8eae939f | 973 | /* OOB backlog add */ |
a3a858ff | 974 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
9ee6b535 | 975 | { |
7fee226a | 976 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
222d7dbd | 977 | skb_dst_force(skb); |
7fee226a ED |
978 | |
979 | if (!sk->sk_backlog.tail) | |
9ed498c6 | 980 | WRITE_ONCE(sk->sk_backlog.head, skb); |
7fee226a | 981 | else |
9ee6b535 | 982 | sk->sk_backlog.tail->next = skb; |
7fee226a | 983 | |
9ed498c6 | 984 | WRITE_ONCE(sk->sk_backlog.tail, skb); |
9ee6b535 SH |
985 | skb->next = NULL; |
986 | } | |
1da177e4 | 987 | |
c377411f ED |
988 | /* |
989 | * Take into account size of receive queue and backlog queue | |
0fd7bac6 ED |
990 | * Do not take into account this skb truesize, |
991 | * to allow even a single big packet to come. | |
c377411f | 992 | */ |
274f482d | 993 | static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit) |
c377411f ED |
994 | { |
995 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); | |
996 | ||
f545a38f | 997 | return qsize > limit; |
c377411f ED |
998 | } |
999 | ||
8eae939f | 1000 | /* The per-socket spinlock must be held here. */ |
f545a38f ED |
1001 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb, |
1002 | unsigned int limit) | |
8eae939f | 1003 | { |
274f482d | 1004 | if (sk_rcvqueues_full(sk, limit)) |
8eae939f ZY |
1005 | return -ENOBUFS; |
1006 | ||
c7c49b8f ED |
1007 | /* |
1008 | * If the skb was allocated from pfmemalloc reserves, only | |
1009 | * allow SOCK_MEMALLOC sockets to use it as this socket is | |
1010 | * helping free memory | |
1011 | */ | |
1012 | if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC)) | |
1013 | return -ENOMEM; | |
1014 | ||
a3a858ff | 1015 | __sk_add_backlog(sk, skb); |
8eae939f ZY |
1016 | sk->sk_backlog.len += skb->truesize; |
1017 | return 0; | |
1018 | } | |
1019 | ||
69336bd2 | 1020 | int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
b4b9e355 | 1021 | |
c57943a1 PZ |
1022 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
1023 | { | |
b4b9e355 MG |
1024 | if (sk_memalloc_socks() && skb_pfmemalloc(skb)) |
1025 | return __sk_backlog_rcv(sk, skb); | |
1026 | ||
c57943a1 PZ |
1027 | return sk->sk_backlog_rcv(sk, skb); |
1028 | } | |
1029 | ||
2c8c56e1 ED |
1030 | static inline void sk_incoming_cpu_update(struct sock *sk) |
1031 | { | |
34cfb542 PA |
1032 | int cpu = raw_smp_processor_id(); |
1033 | ||
7170a977 ED |
1034 | if (unlikely(READ_ONCE(sk->sk_incoming_cpu) != cpu)) |
1035 | WRITE_ONCE(sk->sk_incoming_cpu, cpu); | |
2c8c56e1 ED |
1036 | } |
1037 | ||
fe477558 | 1038 | static inline void sock_rps_record_flow_hash(__u32 hash) |
c58dc01b DM |
1039 | { |
1040 | #ifdef CONFIG_RPS | |
1041 | struct rps_sock_flow_table *sock_flow_table; | |
1042 | ||
1043 | rcu_read_lock(); | |
1044 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
fe477558 | 1045 | rps_record_sock_flow(sock_flow_table, hash); |
c58dc01b DM |
1046 | rcu_read_unlock(); |
1047 | #endif | |
1048 | } | |
1049 | ||
fe477558 TH |
1050 | static inline void sock_rps_record_flow(const struct sock *sk) |
1051 | { | |
c9d8ca04 | 1052 | #ifdef CONFIG_RPS |
dc05360f | 1053 | if (static_branch_unlikely(&rfs_needed)) { |
13bfff25 ED |
1054 | /* Reading sk->sk_rxhash might incur an expensive cache line |
1055 | * miss. | |
1056 | * | |
1057 | * TCP_ESTABLISHED does cover almost all states where RFS | |
1058 | * might be useful, and is cheaper [1] than testing : | |
1059 | * IPv4: inet_sk(sk)->inet_daddr | |
1060 | * IPv6: ipv6_addr_any(&sk->sk_v6_daddr) | |
1061 | * OR an additional socket flag | |
1062 | * [1] : sk_state and sk_prot are in the same cache line. | |
1063 | */ | |
1064 | if (sk->sk_state == TCP_ESTABLISHED) | |
1065 | sock_rps_record_flow_hash(sk->sk_rxhash); | |
1066 | } | |
c9d8ca04 | 1067 | #endif |
fe477558 TH |
1068 | } |
1069 | ||
bdeab991 TH |
1070 | static inline void sock_rps_save_rxhash(struct sock *sk, |
1071 | const struct sk_buff *skb) | |
c58dc01b DM |
1072 | { |
1073 | #ifdef CONFIG_RPS | |
567e4b79 | 1074 | if (unlikely(sk->sk_rxhash != skb->hash)) |
61b905da | 1075 | sk->sk_rxhash = skb->hash; |
c58dc01b DM |
1076 | #endif |
1077 | } | |
1078 | ||
bdeab991 TH |
1079 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
1080 | { | |
1081 | #ifdef CONFIG_RPS | |
bdeab991 TH |
1082 | sk->sk_rxhash = 0; |
1083 | #endif | |
1084 | } | |
1085 | ||
d9dc8b0f | 1086 | #define sk_wait_event(__sk, __timeo, __condition, __wait) \ |
cfcabdcc SH |
1087 | ({ int __rc; \ |
1088 | release_sock(__sk); \ | |
1089 | __rc = __condition; \ | |
1090 | if (!__rc) { \ | |
d9dc8b0f WC |
1091 | *(__timeo) = wait_woken(__wait, \ |
1092 | TASK_INTERRUPTIBLE, \ | |
1093 | *(__timeo)); \ | |
cfcabdcc | 1094 | } \ |
d9dc8b0f | 1095 | sched_annotate_sleep(); \ |
cfcabdcc SH |
1096 | lock_sock(__sk); \ |
1097 | __rc = __condition; \ | |
1098 | __rc; \ | |
1099 | }) | |
1da177e4 | 1100 | |
69336bd2 JP |
1101 | int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
1102 | int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
1103 | void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
1104 | int sk_stream_error(struct sock *sk, int flags, int err); | |
1105 | void sk_stream_kill_queues(struct sock *sk); | |
1106 | void sk_set_memalloc(struct sock *sk); | |
1107 | void sk_clear_memalloc(struct sock *sk); | |
1da177e4 | 1108 | |
d41a69f1 ED |
1109 | void __sk_flush_backlog(struct sock *sk); |
1110 | ||
1111 | static inline bool sk_flush_backlog(struct sock *sk) | |
1112 | { | |
1113 | if (unlikely(READ_ONCE(sk->sk_backlog.tail))) { | |
1114 | __sk_flush_backlog(sk); | |
1115 | return true; | |
1116 | } | |
1117 | return false; | |
1118 | } | |
1119 | ||
dfbafc99 | 1120 | int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb); |
1da177e4 | 1121 | |
60236fdd | 1122 | struct request_sock_ops; |
6d6ee43e | 1123 | struct timewait_sock_ops; |
ab1e0a13 | 1124 | struct inet_hashinfo; |
fc8717ba | 1125 | struct raw_hashinfo; |
f16a7dd5 | 1126 | struct smc_hashinfo; |
de477254 | 1127 | struct module; |
51e0158a | 1128 | struct sk_psock; |
2e6599cb | 1129 | |
f77d6021 | 1130 | /* |
5f0d5a3a | 1131 | * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes |
f77d6021 ED |
1132 | * un-modified. Special care is taken when initializing object to zero. |
1133 | */ | |
1134 | static inline void sk_prot_clear_nulls(struct sock *sk, int size) | |
1135 | { | |
1136 | if (offsetof(struct sock, sk_node.next) != 0) | |
1137 | memset(sk, 0, offsetof(struct sock, sk_node.next)); | |
1138 | memset(&sk->sk_node.pprev, 0, | |
1139 | size - offsetof(struct sock, sk_node.pprev)); | |
1140 | } | |
1141 | ||
1da177e4 LT |
1142 | /* Networking protocol blocks we attach to sockets. |
1143 | * socket layer -> transport layer interface | |
1da177e4 LT |
1144 | */ |
1145 | struct proto { | |
dc6b9b78 | 1146 | void (*close)(struct sock *sk, |
1da177e4 | 1147 | long timeout); |
d74bad4e AI |
1148 | int (*pre_connect)(struct sock *sk, |
1149 | struct sockaddr *uaddr, | |
1150 | int addr_len); | |
1da177e4 | 1151 | int (*connect)(struct sock *sk, |
dc6b9b78 | 1152 | struct sockaddr *uaddr, |
1da177e4 LT |
1153 | int addr_len); |
1154 | int (*disconnect)(struct sock *sk, int flags); | |
1155 | ||
cdfbabfb DH |
1156 | struct sock * (*accept)(struct sock *sk, int flags, int *err, |
1157 | bool kern); | |
1da177e4 LT |
1158 | |
1159 | int (*ioctl)(struct sock *sk, int cmd, | |
1160 | unsigned long arg); | |
1161 | int (*init)(struct sock *sk); | |
7d06b2e0 | 1162 | void (*destroy)(struct sock *sk); |
1da177e4 | 1163 | void (*shutdown)(struct sock *sk, int how); |
dc6b9b78 | 1164 | int (*setsockopt)(struct sock *sk, int level, |
a7b75c5a | 1165 | int optname, sockptr_t optval, |
b7058842 | 1166 | unsigned int optlen); |
dc6b9b78 ED |
1167 | int (*getsockopt)(struct sock *sk, int level, |
1168 | int optname, char __user *optval, | |
1169 | int __user *option); | |
4b9d07a4 | 1170 | void (*keepalive)(struct sock *sk, int valbool); |
af01d537 | 1171 | #ifdef CONFIG_COMPAT |
709b46e8 EB |
1172 | int (*compat_ioctl)(struct sock *sk, |
1173 | unsigned int cmd, unsigned long arg); | |
af01d537 | 1174 | #endif |
1b784140 YX |
1175 | int (*sendmsg)(struct sock *sk, struct msghdr *msg, |
1176 | size_t len); | |
1177 | int (*recvmsg)(struct sock *sk, struct msghdr *msg, | |
dc6b9b78 ED |
1178 | size_t len, int noblock, int flags, |
1179 | int *addr_len); | |
1da177e4 LT |
1180 | int (*sendpage)(struct sock *sk, struct page *page, |
1181 | int offset, size_t size, int flags); | |
dc6b9b78 | 1182 | int (*bind)(struct sock *sk, |
c0425a42 CH |
1183 | struct sockaddr *addr, int addr_len); |
1184 | int (*bind_add)(struct sock *sk, | |
1185 | struct sockaddr *addr, int addr_len); | |
1da177e4 | 1186 | |
dc6b9b78 | 1187 | int (*backlog_rcv) (struct sock *sk, |
1da177e4 | 1188 | struct sk_buff *skb); |
9cacf81f SF |
1189 | bool (*bpf_bypass_getsockopt)(int level, |
1190 | int optname); | |
1da177e4 | 1191 | |
46d3ceab ED |
1192 | void (*release_cb)(struct sock *sk); |
1193 | ||
1da177e4 | 1194 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
086c653f | 1195 | int (*hash)(struct sock *sk); |
1da177e4 | 1196 | void (*unhash)(struct sock *sk); |
719f8358 | 1197 | void (*rehash)(struct sock *sk); |
1da177e4 | 1198 | int (*get_port)(struct sock *sk, unsigned short snum); |
8a59f9d1 | 1199 | #ifdef CONFIG_BPF_SYSCALL |
51e0158a CW |
1200 | int (*psock_update_sk_prot)(struct sock *sk, |
1201 | struct sk_psock *psock, | |
1202 | bool restore); | |
8a59f9d1 | 1203 | #endif |
1da177e4 | 1204 | |
286ab3d4 | 1205 | /* Keeping track of sockets in use */ |
65f76517 | 1206 | #ifdef CONFIG_PROC_FS |
13ff3d6f | 1207 | unsigned int inuse_idx; |
65f76517 | 1208 | #endif |
ebb53d75 | 1209 | |
a74f0fa0 | 1210 | bool (*stream_memory_free)(const struct sock *sk, int wake); |
8934ce2f | 1211 | bool (*stream_memory_read)(const struct sock *sk); |
1da177e4 | 1212 | /* Memory pressure */ |
5c52ba17 | 1213 | void (*enter_memory_pressure)(struct sock *sk); |
06044751 | 1214 | void (*leave_memory_pressure)(struct sock *sk); |
8d987e5c | 1215 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
1748376b | 1216 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
1da177e4 LT |
1217 | /* |
1218 | * Pressure flag: try to collapse. | |
1219 | * Technical note: it is used by multiple contexts non atomically. | |
3ab224be | 1220 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
1221 | * is strict, actions are advisory and have some latency. |
1222 | */ | |
06044751 | 1223 | unsigned long *memory_pressure; |
8d987e5c | 1224 | long *sysctl_mem; |
a3dcaf17 | 1225 | |
1da177e4 LT |
1226 | int *sysctl_wmem; |
1227 | int *sysctl_rmem; | |
a3dcaf17 ED |
1228 | u32 sysctl_wmem_offset; |
1229 | u32 sysctl_rmem_offset; | |
1230 | ||
1da177e4 | 1231 | int max_header; |
7ba42910 | 1232 | bool no_autobind; |
1da177e4 | 1233 | |
271b72c7 | 1234 | struct kmem_cache *slab; |
1da177e4 | 1235 | unsigned int obj_size; |
d50112ed | 1236 | slab_flags_t slab_flags; |
7bbdb81e AD |
1237 | unsigned int useroffset; /* Usercopy region offset */ |
1238 | unsigned int usersize; /* Usercopy region size */ | |
1da177e4 | 1239 | |
dd24c001 | 1240 | struct percpu_counter *orphan_count; |
8feaf0c0 | 1241 | |
60236fdd | 1242 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 1243 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 1244 | |
39d8cda7 PE |
1245 | union { |
1246 | struct inet_hashinfo *hashinfo; | |
645ca708 | 1247 | struct udp_table *udp_table; |
fc8717ba | 1248 | struct raw_hashinfo *raw_hash; |
f16a7dd5 | 1249 | struct smc_hashinfo *smc_hash; |
39d8cda7 | 1250 | } h; |
ab1e0a13 | 1251 | |
1da177e4 LT |
1252 | struct module *owner; |
1253 | ||
1254 | char name[32]; | |
1255 | ||
1256 | struct list_head node; | |
e6848976 ACM |
1257 | #ifdef SOCK_REFCNT_DEBUG |
1258 | atomic_t socks; | |
e1aab161 | 1259 | #endif |
64be0aed | 1260 | int (*diag_destroy)(struct sock *sk, int err); |
3859a271 | 1261 | } __randomize_layout; |
e1aab161 | 1262 | |
69336bd2 JP |
1263 | int proto_register(struct proto *prot, int alloc_slab); |
1264 | void proto_unregister(struct proto *prot); | |
bf2ae2e4 | 1265 | int sock_load_diag_module(int family, int protocol); |
1da177e4 | 1266 | |
e6848976 ACM |
1267 | #ifdef SOCK_REFCNT_DEBUG |
1268 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
1269 | { | |
1270 | atomic_inc(&sk->sk_prot->socks); | |
1271 | } | |
1272 | ||
1273 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
1274 | { | |
1275 | atomic_dec(&sk->sk_prot->socks); | |
1276 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
1277 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
1278 | } | |
1279 | ||
dec34fb0 | 1280 | static inline void sk_refcnt_debug_release(const struct sock *sk) |
e6848976 | 1281 | { |
41c6d650 | 1282 | if (refcount_read(&sk->sk_refcnt) != 1) |
e6848976 | 1283 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", |
41c6d650 | 1284 | sk->sk_prot->name, sk, refcount_read(&sk->sk_refcnt)); |
e6848976 ACM |
1285 | } |
1286 | #else /* SOCK_REFCNT_DEBUG */ | |
1287 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
1288 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
1289 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
1290 | #endif /* SOCK_REFCNT_DEBUG */ | |
1291 | ||
1c5f2ced ED |
1292 | INDIRECT_CALLABLE_DECLARE(bool tcp_stream_memory_free(const struct sock *sk, int wake)); |
1293 | ||
a74f0fa0 | 1294 | static inline bool __sk_stream_memory_free(const struct sock *sk, int wake) |
c9bee3b7 | 1295 | { |
ab4e846a | 1296 | if (READ_ONCE(sk->sk_wmem_queued) >= READ_ONCE(sk->sk_sndbuf)) |
c9bee3b7 ED |
1297 | return false; |
1298 | ||
fc9840fb | 1299 | #ifdef CONFIG_INET |
c9bee3b7 | 1300 | return sk->sk_prot->stream_memory_free ? |
1c5f2ced ED |
1301 | INDIRECT_CALL_1(sk->sk_prot->stream_memory_free, |
1302 | tcp_stream_memory_free, | |
1303 | sk, wake) : true; | |
fc9840fb | 1304 | #else |
c9bee3b7 | 1305 | return sk->sk_prot->stream_memory_free ? |
a74f0fa0 | 1306 | sk->sk_prot->stream_memory_free(sk, wake) : true; |
fc9840fb | 1307 | #endif |
c9bee3b7 ED |
1308 | } |
1309 | ||
a74f0fa0 ED |
1310 | static inline bool sk_stream_memory_free(const struct sock *sk) |
1311 | { | |
1312 | return __sk_stream_memory_free(sk, 0); | |
1313 | } | |
1314 | ||
1315 | static inline bool __sk_stream_is_writeable(const struct sock *sk, int wake) | |
64dc6130 | 1316 | { |
c9bee3b7 | 1317 | return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) && |
a74f0fa0 ED |
1318 | __sk_stream_memory_free(sk, wake); |
1319 | } | |
1320 | ||
1321 | static inline bool sk_stream_is_writeable(const struct sock *sk) | |
1322 | { | |
1323 | return __sk_stream_is_writeable(sk, 0); | |
64dc6130 | 1324 | } |
e1aab161 | 1325 | |
54fd9c2d DB |
1326 | static inline int sk_under_cgroup_hierarchy(struct sock *sk, |
1327 | struct cgroup *ancestor) | |
1328 | { | |
1329 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
1330 | return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data), | |
1331 | ancestor); | |
1332 | #else | |
1333 | return -ENOTSUPP; | |
1334 | #endif | |
1335 | } | |
c9bee3b7 | 1336 | |
180d8cd9 GC |
1337 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
1338 | { | |
1339 | return sk->sk_prot->memory_pressure != NULL; | |
1340 | } | |
1341 | ||
1342 | static inline bool sk_under_memory_pressure(const struct sock *sk) | |
1343 | { | |
1344 | if (!sk->sk_prot->memory_pressure) | |
1345 | return false; | |
e1aab161 | 1346 | |
baac50bb JW |
1347 | if (mem_cgroup_sockets_enabled && sk->sk_memcg && |
1348 | mem_cgroup_under_socket_pressure(sk->sk_memcg)) | |
e805605c | 1349 | return true; |
e1aab161 | 1350 | |
35b87f6c | 1351 | return !!*sk->sk_prot->memory_pressure; |
180d8cd9 GC |
1352 | } |
1353 | ||
180d8cd9 GC |
1354 | static inline long |
1355 | sk_memory_allocated(const struct sock *sk) | |
1356 | { | |
e805605c | 1357 | return atomic_long_read(sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1358 | } |
1359 | ||
1360 | static inline long | |
e805605c | 1361 | sk_memory_allocated_add(struct sock *sk, int amt) |
180d8cd9 | 1362 | { |
e805605c | 1363 | return atomic_long_add_return(amt, sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1364 | } |
1365 | ||
1366 | static inline void | |
0e90b31f | 1367 | sk_memory_allocated_sub(struct sock *sk, int amt) |
180d8cd9 | 1368 | { |
e805605c | 1369 | atomic_long_sub(amt, sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1370 | } |
1371 | ||
f5a5589c WW |
1372 | #define SK_ALLOC_PERCPU_COUNTER_BATCH 16 |
1373 | ||
180d8cd9 GC |
1374 | static inline void sk_sockets_allocated_dec(struct sock *sk) |
1375 | { | |
f5a5589c WW |
1376 | percpu_counter_add_batch(sk->sk_prot->sockets_allocated, -1, |
1377 | SK_ALLOC_PERCPU_COUNTER_BATCH); | |
180d8cd9 GC |
1378 | } |
1379 | ||
1380 | static inline void sk_sockets_allocated_inc(struct sock *sk) | |
1381 | { | |
f5a5589c WW |
1382 | percpu_counter_add_batch(sk->sk_prot->sockets_allocated, 1, |
1383 | SK_ALLOC_PERCPU_COUNTER_BATCH); | |
180d8cd9 GC |
1384 | } |
1385 | ||
5bf325a5 | 1386 | static inline u64 |
180d8cd9 GC |
1387 | sk_sockets_allocated_read_positive(struct sock *sk) |
1388 | { | |
af95d7df | 1389 | return percpu_counter_read_positive(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1390 | } |
1391 | ||
1392 | static inline int | |
1393 | proto_sockets_allocated_sum_positive(struct proto *prot) | |
1394 | { | |
1395 | return percpu_counter_sum_positive(prot->sockets_allocated); | |
1396 | } | |
1397 | ||
1398 | static inline long | |
1399 | proto_memory_allocated(struct proto *prot) | |
1400 | { | |
1401 | return atomic_long_read(prot->memory_allocated); | |
1402 | } | |
1403 | ||
1404 | static inline bool | |
1405 | proto_memory_pressure(struct proto *prot) | |
1406 | { | |
1407 | if (!prot->memory_pressure) | |
1408 | return false; | |
1409 | return !!*prot->memory_pressure; | |
1410 | } | |
1411 | ||
65f76517 ED |
1412 | |
1413 | #ifdef CONFIG_PROC_FS | |
1da177e4 | 1414 | /* Called with local bh disabled */ |
69336bd2 JP |
1415 | void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
1416 | int sock_prot_inuse_get(struct net *net, struct proto *proto); | |
648845ab | 1417 | int sock_inuse_get(struct net *net); |
65f76517 | 1418 | #else |
dc6b9b78 | 1419 | static inline void sock_prot_inuse_add(struct net *net, struct proto *prot, |
c29a0bc4 | 1420 | int inc) |
65f76517 ED |
1421 | { |
1422 | } | |
65f76517 ED |
1423 | #endif |
1424 | ||
1da177e4 | 1425 | |
614c6cb4 ACM |
1426 | /* With per-bucket locks this operation is not-atomic, so that |
1427 | * this version is not worse. | |
1428 | */ | |
086c653f | 1429 | static inline int __sk_prot_rehash(struct sock *sk) |
614c6cb4 ACM |
1430 | { |
1431 | sk->sk_prot->unhash(sk); | |
086c653f | 1432 | return sk->sk_prot->hash(sk); |
614c6cb4 ACM |
1433 | } |
1434 | ||
1da177e4 LT |
1435 | /* About 10 seconds */ |
1436 | #define SOCK_DESTROY_TIME (10*HZ) | |
1437 | ||
1438 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
1439 | #define PROT_SOCK 1024 | |
1440 | ||
1441 | #define SHUTDOWN_MASK 3 | |
1442 | #define RCV_SHUTDOWN 1 | |
1443 | #define SEND_SHUTDOWN 2 | |
1444 | ||
1da177e4 LT |
1445 | #define SOCK_BINDADDR_LOCK 4 |
1446 | #define SOCK_BINDPORT_LOCK 8 | |
1447 | ||
1da177e4 LT |
1448 | struct socket_alloc { |
1449 | struct socket socket; | |
1450 | struct inode vfs_inode; | |
1451 | }; | |
1452 | ||
1453 | static inline struct socket *SOCKET_I(struct inode *inode) | |
1454 | { | |
1455 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
1456 | } | |
1457 | ||
1458 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
1459 | { | |
1460 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
1461 | } | |
1462 | ||
3ab224be HA |
1463 | /* |
1464 | * Functions for memory accounting | |
1465 | */ | |
f8c3bf00 | 1466 | int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind); |
69336bd2 | 1467 | int __sk_mem_schedule(struct sock *sk, int size, int kind); |
f8c3bf00 | 1468 | void __sk_mem_reduce_allocated(struct sock *sk, int amount); |
1a24e04e | 1469 | void __sk_mem_reclaim(struct sock *sk, int amount); |
1da177e4 | 1470 | |
bd68a2a8 ED |
1471 | /* We used to have PAGE_SIZE here, but systems with 64KB pages |
1472 | * do not necessarily have 16x time more memory than 4KB ones. | |
1473 | */ | |
1474 | #define SK_MEM_QUANTUM 4096 | |
3ab224be HA |
1475 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) |
1476 | #define SK_MEM_SEND 0 | |
1477 | #define SK_MEM_RECV 1 | |
1da177e4 | 1478 | |
bd68a2a8 ED |
1479 | /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */ |
1480 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) | |
1481 | { | |
1482 | long val = sk->sk_prot->sysctl_mem[index]; | |
1483 | ||
1484 | #if PAGE_SIZE > SK_MEM_QUANTUM | |
1485 | val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT; | |
1486 | #elif PAGE_SIZE < SK_MEM_QUANTUM | |
1487 | val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT; | |
1488 | #endif | |
1489 | return val; | |
1490 | } | |
1491 | ||
3ab224be | 1492 | static inline int sk_mem_pages(int amt) |
1da177e4 | 1493 | { |
3ab224be | 1494 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
1da177e4 LT |
1495 | } |
1496 | ||
dc6b9b78 | 1497 | static inline bool sk_has_account(struct sock *sk) |
1da177e4 | 1498 | { |
3ab224be HA |
1499 | /* return true if protocol supports memory accounting */ |
1500 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
1501 | } |
1502 | ||
dc6b9b78 | 1503 | static inline bool sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 1504 | { |
3ab224be | 1505 | if (!sk_has_account(sk)) |
dc6b9b78 | 1506 | return true; |
3ab224be HA |
1507 | return size <= sk->sk_forward_alloc || |
1508 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
1da177e4 LT |
1509 | } |
1510 | ||
c76562b6 | 1511 | static inline bool |
35c448a8 | 1512 | sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size) |
d80d99d6 | 1513 | { |
3ab224be | 1514 | if (!sk_has_account(sk)) |
dc6b9b78 | 1515 | return true; |
5af68891 | 1516 | return size <= sk->sk_forward_alloc || |
c76562b6 MG |
1517 | __sk_mem_schedule(sk, size, SK_MEM_RECV) || |
1518 | skb_pfmemalloc(skb); | |
3ab224be HA |
1519 | } |
1520 | ||
1521 | static inline void sk_mem_reclaim(struct sock *sk) | |
1522 | { | |
1523 | if (!sk_has_account(sk)) | |
1524 | return; | |
1525 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
1a24e04e | 1526 | __sk_mem_reclaim(sk, sk->sk_forward_alloc); |
3ab224be HA |
1527 | } |
1528 | ||
9993e7d3 DM |
1529 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
1530 | { | |
1531 | if (!sk_has_account(sk)) | |
1532 | return; | |
1533 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) | |
1a24e04e | 1534 | __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1); |
9993e7d3 DM |
1535 | } |
1536 | ||
3ab224be HA |
1537 | static inline void sk_mem_charge(struct sock *sk, int size) |
1538 | { | |
1539 | if (!sk_has_account(sk)) | |
1540 | return; | |
1541 | sk->sk_forward_alloc -= size; | |
1542 | } | |
1543 | ||
1544 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
1545 | { | |
1546 | if (!sk_has_account(sk)) | |
1547 | return; | |
1548 | sk->sk_forward_alloc += size; | |
20c64d5c ED |
1549 | |
1550 | /* Avoid a possible overflow. | |
1551 | * TCP send queues can make this happen, if sk_mem_reclaim() | |
1552 | * is not called and more than 2 GBytes are released at once. | |
1553 | * | |
1554 | * If we reach 2 MBytes, reclaim 1 MBytes right now, there is | |
1555 | * no need to hold that much forward allocation anyway. | |
1556 | */ | |
1557 | if (unlikely(sk->sk_forward_alloc >= 1 << 21)) | |
1558 | __sk_mem_reclaim(sk, 1 << 20); | |
3ab224be HA |
1559 | } |
1560 | ||
0b7d7f6b | 1561 | DECLARE_STATIC_KEY_FALSE(tcp_tx_skb_cache_key); |
3ab224be HA |
1562 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) |
1563 | { | |
ab4e846a | 1564 | sk_wmem_queued_add(sk, -skb->truesize); |
4f661542 | 1565 | sk_mem_uncharge(sk, skb->truesize); |
0b7d7f6b ED |
1566 | if (static_branch_unlikely(&tcp_tx_skb_cache_key) && |
1567 | !sk->sk_tx_skb_cache && !skb_cloned(skb)) { | |
e66b2f31 | 1568 | skb_ext_reset(skb); |
4f661542 | 1569 | skb_zcopy_clear(skb, true); |
472c2e07 ED |
1570 | sk->sk_tx_skb_cache = skb; |
1571 | return; | |
1572 | } | |
3ab224be | 1573 | __kfree_skb(skb); |
d80d99d6 HX |
1574 | } |
1575 | ||
c3f9b018 ED |
1576 | static inline void sock_release_ownership(struct sock *sk) |
1577 | { | |
61881cfb HFS |
1578 | if (sk->sk_lock.owned) { |
1579 | sk->sk_lock.owned = 0; | |
1580 | ||
1581 | /* The sk_lock has mutex_unlock() semantics: */ | |
5facae4f | 1582 | mutex_release(&sk->sk_lock.dep_map, _RET_IP_); |
61881cfb | 1583 | } |
c3f9b018 ED |
1584 | } |
1585 | ||
ed07536e PZ |
1586 | /* |
1587 | * Macro so as to not evaluate some arguments when | |
1588 | * lockdep is not enabled. | |
1589 | * | |
1590 | * Mark both the sk_lock and the sk_lock.slock as a | |
1591 | * per-address-family lock class. | |
1592 | */ | |
dc6b9b78 | 1593 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
ed07536e | 1594 | do { \ |
e8f6fbf6 | 1595 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
1596 | init_waitqueue_head(&sk->sk_lock.wq); \ |
1597 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
1598 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
1599 | sizeof((sk)->sk_lock)); \ | |
1600 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
dc6b9b78 | 1601 | (skey), (sname)); \ |
ed07536e PZ |
1602 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
1603 | } while (0) | |
1604 | ||
05b93801 | 1605 | static inline bool lockdep_sock_is_held(const struct sock *sk) |
1e1d04e6 | 1606 | { |
1e1d04e6 HFS |
1607 | return lockdep_is_held(&sk->sk_lock) || |
1608 | lockdep_is_held(&sk->sk_lock.slock); | |
1609 | } | |
1610 | ||
69336bd2 | 1611 | void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
1612 | |
1613 | static inline void lock_sock(struct sock *sk) | |
1614 | { | |
1615 | lock_sock_nested(sk, 0); | |
1616 | } | |
1617 | ||
ad80b0fc | 1618 | void __lock_sock(struct sock *sk); |
8873c064 | 1619 | void __release_sock(struct sock *sk); |
69336bd2 | 1620 | void release_sock(struct sock *sk); |
1da177e4 LT |
1621 | |
1622 | /* BH context may only use the following locking interface. */ | |
1623 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
1624 | #define bh_lock_sock_nested(__sk) \ |
1625 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
1626 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
1627 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
1628 | ||
49054556 PA |
1629 | bool __lock_sock_fast(struct sock *sk) __acquires(&sk->sk_lock.slock); |
1630 | ||
1631 | /** | |
1632 | * lock_sock_fast - fast version of lock_sock | |
1633 | * @sk: socket | |
1634 | * | |
1635 | * This version should be used for very small section, where process wont block | |
1636 | * return false if fast path is taken: | |
1637 | * | |
1638 | * sk_lock.slock locked, owned = 0, BH disabled | |
1639 | * | |
1640 | * return true if slow path is taken: | |
1641 | * | |
1642 | * sk_lock.slock unlocked, owned = 1, BH enabled | |
1643 | */ | |
1644 | static inline bool lock_sock_fast(struct sock *sk) | |
1645 | { | |
1646 | /* The sk_lock has mutex_lock() semantics here. */ | |
1647 | mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_); | |
1648 | ||
1649 | return __lock_sock_fast(sk); | |
1650 | } | |
1651 | ||
1652 | /* fast socket lock variant for caller already holding a [different] socket lock */ | |
1653 | static inline bool lock_sock_fast_nested(struct sock *sk) | |
1654 | { | |
1655 | mutex_acquire(&sk->sk_lock.dep_map, SINGLE_DEPTH_NESTING, 0, _RET_IP_); | |
1656 | ||
1657 | return __lock_sock_fast(sk); | |
1658 | } | |
12f4bd86 | 1659 | |
8a74ad60 ED |
1660 | /** |
1661 | * unlock_sock_fast - complement of lock_sock_fast | |
1662 | * @sk: socket | |
1663 | * @slow: slow mode | |
1664 | * | |
1665 | * fast unlock socket for user context. | |
1666 | * If slow mode is on, we call regular release_sock() | |
1667 | */ | |
1668 | static inline void unlock_sock_fast(struct sock *sk, bool slow) | |
12f4bd86 | 1669 | __releases(&sk->sk_lock.slock) |
4b0b72f7 | 1670 | { |
12f4bd86 | 1671 | if (slow) { |
8a74ad60 | 1672 | release_sock(sk); |
12f4bd86 PA |
1673 | __release(&sk->sk_lock.slock); |
1674 | } else { | |
2dcb96ba | 1675 | mutex_release(&sk->sk_lock.dep_map, _RET_IP_); |
8a74ad60 | 1676 | spin_unlock_bh(&sk->sk_lock.slock); |
12f4bd86 | 1677 | } |
4b0b72f7 ED |
1678 | } |
1679 | ||
fafc4e1e HFS |
1680 | /* Used by processes to "lock" a socket state, so that |
1681 | * interrupts and bottom half handlers won't change it | |
1682 | * from under us. It essentially blocks any incoming | |
1683 | * packets, so that we won't get any new data or any | |
1684 | * packets that change the state of the socket. | |
1685 | * | |
1686 | * While locked, BH processing will add new packets to | |
1687 | * the backlog queue. This queue is processed by the | |
1688 | * owner of the socket lock right before it is released. | |
1689 | * | |
1690 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
1691 | * accesses from user process context. | |
1692 | */ | |
1693 | ||
46cc6e49 | 1694 | static inline void sock_owned_by_me(const struct sock *sk) |
fafc4e1e HFS |
1695 | { |
1696 | #ifdef CONFIG_LOCKDEP | |
5e91f6ce | 1697 | WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks); |
fafc4e1e | 1698 | #endif |
46cc6e49 ED |
1699 | } |
1700 | ||
1701 | static inline bool sock_owned_by_user(const struct sock *sk) | |
1702 | { | |
1703 | sock_owned_by_me(sk); | |
fafc4e1e HFS |
1704 | return sk->sk_lock.owned; |
1705 | } | |
1706 | ||
602f7a27 TH |
1707 | static inline bool sock_owned_by_user_nocheck(const struct sock *sk) |
1708 | { | |
1709 | return sk->sk_lock.owned; | |
1710 | } | |
1711 | ||
fafc4e1e HFS |
1712 | /* no reclassification while locks are held */ |
1713 | static inline bool sock_allow_reclassification(const struct sock *csk) | |
1714 | { | |
1715 | struct sock *sk = (struct sock *)csk; | |
1716 | ||
1717 | return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock); | |
1718 | } | |
4b0b72f7 | 1719 | |
69336bd2 | 1720 | struct sock *sk_alloc(struct net *net, int family, gfp_t priority, |
11aa9c28 | 1721 | struct proto *prot, int kern); |
69336bd2 | 1722 | void sk_free(struct sock *sk); |
eb4cb008 | 1723 | void sk_destruct(struct sock *sk); |
69336bd2 | 1724 | struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority); |
94352d45 | 1725 | void sk_free_unlock_clone(struct sock *sk); |
69336bd2 JP |
1726 | |
1727 | struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force, | |
1728 | gfp_t priority); | |
1d2077ac | 1729 | void __sock_wfree(struct sk_buff *skb); |
69336bd2 | 1730 | void sock_wfree(struct sk_buff *skb); |
98ba0bd5 WB |
1731 | struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size, |
1732 | gfp_t priority); | |
69336bd2 JP |
1733 | void skb_orphan_partial(struct sk_buff *skb); |
1734 | void sock_rfree(struct sk_buff *skb); | |
62bccb8c | 1735 | void sock_efree(struct sk_buff *skb); |
82eabd9e | 1736 | #ifdef CONFIG_INET |
69336bd2 | 1737 | void sock_edemux(struct sk_buff *skb); |
cf7fbe66 | 1738 | void sock_pfree(struct sk_buff *skb); |
82eabd9e | 1739 | #else |
158f323b | 1740 | #define sock_edemux sock_efree |
82eabd9e | 1741 | #endif |
69336bd2 JP |
1742 | |
1743 | int sock_setsockopt(struct socket *sock, int level, int op, | |
c8c1bbb6 | 1744 | sockptr_t optval, unsigned int optlen); |
69336bd2 JP |
1745 | |
1746 | int sock_getsockopt(struct socket *sock, int level, int op, | |
1747 | char __user *optval, int __user *optlen); | |
c7cbdbf2 AB |
1748 | int sock_gettstamp(struct socket *sock, void __user *userstamp, |
1749 | bool timeval, bool time32); | |
69336bd2 JP |
1750 | struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size, |
1751 | int noblock, int *errcode); | |
1752 | struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, | |
1753 | unsigned long data_len, int noblock, | |
1754 | int *errcode, int max_page_order); | |
1755 | void *sock_kmalloc(struct sock *sk, int size, gfp_t priority); | |
1756 | void sock_kfree_s(struct sock *sk, void *mem, int size); | |
79e88659 | 1757 | void sock_kzfree_s(struct sock *sk, void *mem, int size); |
69336bd2 | 1758 | void sk_send_sigurg(struct sock *sk); |
1da177e4 | 1759 | |
f28ea365 | 1760 | struct sockcm_cookie { |
80b14dee | 1761 | u64 transmit_time; |
f28ea365 | 1762 | u32 mark; |
3dd17e63 | 1763 | u16 tsflags; |
f28ea365 EJ |
1764 | }; |
1765 | ||
657a0667 WB |
1766 | static inline void sockcm_init(struct sockcm_cookie *sockc, |
1767 | const struct sock *sk) | |
1768 | { | |
1769 | *sockc = (struct sockcm_cookie) { .tsflags = sk->sk_tsflags }; | |
1770 | } | |
1771 | ||
39771b12 WB |
1772 | int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg, |
1773 | struct sockcm_cookie *sockc); | |
f28ea365 EJ |
1774 | int sock_cmsg_send(struct sock *sk, struct msghdr *msg, |
1775 | struct sockcm_cookie *sockc); | |
1776 | ||
1da177e4 LT |
1777 | /* |
1778 | * Functions to fill in entries in struct proto_ops when a protocol | |
1779 | * does not implement a particular function. | |
1780 | */ | |
69336bd2 JP |
1781 | int sock_no_bind(struct socket *, struct sockaddr *, int); |
1782 | int sock_no_connect(struct socket *, struct sockaddr *, int, int); | |
1783 | int sock_no_socketpair(struct socket *, struct socket *); | |
cdfbabfb | 1784 | int sock_no_accept(struct socket *, struct socket *, int, bool); |
9b2c45d4 | 1785 | int sock_no_getname(struct socket *, struct sockaddr *, int); |
69336bd2 JP |
1786 | int sock_no_ioctl(struct socket *, unsigned int, unsigned long); |
1787 | int sock_no_listen(struct socket *, int); | |
1788 | int sock_no_shutdown(struct socket *, int); | |
1b784140 | 1789 | int sock_no_sendmsg(struct socket *, struct msghdr *, size_t); |
306b13eb | 1790 | int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len); |
1b784140 | 1791 | int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int); |
69336bd2 JP |
1792 | int sock_no_mmap(struct file *file, struct socket *sock, |
1793 | struct vm_area_struct *vma); | |
1794 | ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, | |
1795 | size_t size, int flags); | |
306b13eb TH |
1796 | ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page, |
1797 | int offset, size_t size, int flags); | |
1da177e4 LT |
1798 | |
1799 | /* | |
1800 | * Functions to fill in entries in struct proto_ops when a protocol | |
1801 | * uses the inet style. | |
1802 | */ | |
69336bd2 | 1803 | int sock_common_getsockopt(struct socket *sock, int level, int optname, |
1da177e4 | 1804 | char __user *optval, int __user *optlen); |
1b784140 YX |
1805 | int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
1806 | int flags); | |
69336bd2 | 1807 | int sock_common_setsockopt(struct socket *sock, int level, int optname, |
a7b75c5a | 1808 | sockptr_t optval, unsigned int optlen); |
1da177e4 | 1809 | |
69336bd2 | 1810 | void sk_common_release(struct sock *sk); |
1da177e4 LT |
1811 | |
1812 | /* | |
1813 | * Default socket callbacks and setup code | |
1814 | */ | |
dc6b9b78 | 1815 | |
1da177e4 | 1816 | /* Initialise core socket variables */ |
69336bd2 | 1817 | void sock_init_data(struct socket *sock, struct sock *sk); |
1da177e4 | 1818 | |
1da177e4 LT |
1819 | /* |
1820 | * Socket reference counting postulates. | |
1821 | * | |
1822 | * * Each user of socket SHOULD hold a reference count. | |
1823 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1824 | * running timer, skb in flight MUST hold a reference count. | |
1825 | * * When reference count hits 0, it means it will never increase back. | |
1826 | * * When reference count hits 0, it means that no references from | |
1827 | * outside exist to this socket and current process on current CPU | |
1828 | * is last user and may/should destroy this socket. | |
1829 | * * sk_free is called from any context: process, BH, IRQ. When | |
1830 | * it is called, socket has no references from outside -> sk_free | |
1831 | * may release descendant resources allocated by the socket, but | |
1832 | * to the time when it is called, socket is NOT referenced by any | |
1833 | * hash tables, lists etc. | |
1834 | * * Packets, delivered from outside (from network or from another process) | |
1835 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1836 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1837 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1838 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1839 | * (leak to backlog). Packet socket does all the processing inside | |
1840 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1841 | * use separate SMP lock, so that they are prone too. | |
1842 | */ | |
1843 | ||
1844 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1845 | static inline void sock_put(struct sock *sk) | |
1846 | { | |
41c6d650 | 1847 | if (refcount_dec_and_test(&sk->sk_refcnt)) |
1da177e4 LT |
1848 | sk_free(sk); |
1849 | } | |
05dbc7b5 | 1850 | /* Generic version of sock_put(), dealing with all sockets |
41b822c5 | 1851 | * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...) |
05dbc7b5 ED |
1852 | */ |
1853 | void sock_gen_put(struct sock *sk); | |
1da177e4 | 1854 | |
4f0c40d9 | 1855 | int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested, |
c3f24cfb | 1856 | unsigned int trim_cap, bool refcounted); |
4f0c40d9 WB |
1857 | static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
1858 | const int nested) | |
1859 | { | |
c3f24cfb | 1860 | return __sk_receive_skb(sk, skb, nested, 1, true); |
4f0c40d9 | 1861 | } |
25995ff5 | 1862 | |
e022f0b4 KK |
1863 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
1864 | { | |
755c31cd AN |
1865 | /* sk_tx_queue_mapping accept only upto a 16-bit value */ |
1866 | if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX)) | |
1867 | return; | |
e022f0b4 KK |
1868 | sk->sk_tx_queue_mapping = tx_queue; |
1869 | } | |
1870 | ||
755c31cd AN |
1871 | #define NO_QUEUE_MAPPING USHRT_MAX |
1872 | ||
e022f0b4 KK |
1873 | static inline void sk_tx_queue_clear(struct sock *sk) |
1874 | { | |
755c31cd | 1875 | sk->sk_tx_queue_mapping = NO_QUEUE_MAPPING; |
e022f0b4 KK |
1876 | } |
1877 | ||
1878 | static inline int sk_tx_queue_get(const struct sock *sk) | |
1879 | { | |
755c31cd AN |
1880 | if (sk && sk->sk_tx_queue_mapping != NO_QUEUE_MAPPING) |
1881 | return sk->sk_tx_queue_mapping; | |
1882 | ||
1883 | return -1; | |
e022f0b4 KK |
1884 | } |
1885 | ||
c6345ce7 AN |
1886 | static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb) |
1887 | { | |
4e1beecc | 1888 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
c6345ce7 AN |
1889 | if (skb_rx_queue_recorded(skb)) { |
1890 | u16 rx_queue = skb_get_rx_queue(skb); | |
1891 | ||
1892 | if (WARN_ON_ONCE(rx_queue == NO_QUEUE_MAPPING)) | |
1893 | return; | |
1894 | ||
1895 | sk->sk_rx_queue_mapping = rx_queue; | |
1896 | } | |
1897 | #endif | |
1898 | } | |
1899 | ||
1900 | static inline void sk_rx_queue_clear(struct sock *sk) | |
1901 | { | |
4e1beecc | 1902 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
c6345ce7 AN |
1903 | sk->sk_rx_queue_mapping = NO_QUEUE_MAPPING; |
1904 | #endif | |
1905 | } | |
1906 | ||
fc9bab24 AN |
1907 | static inline int sk_rx_queue_get(const struct sock *sk) |
1908 | { | |
4e1beecc | 1909 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
fc9bab24 AN |
1910 | if (sk && sk->sk_rx_queue_mapping != NO_QUEUE_MAPPING) |
1911 | return sk->sk_rx_queue_mapping; | |
4e1beecc | 1912 | #endif |
fc9bab24 AN |
1913 | |
1914 | return -1; | |
1915 | } | |
fc9bab24 | 1916 | |
972692e0 DM |
1917 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
1918 | { | |
1919 | sk->sk_socket = sock; | |
1920 | } | |
1921 | ||
aa395145 ED |
1922 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
1923 | { | |
eaefd110 ED |
1924 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
1925 | return &rcu_dereference_raw(sk->sk_wq)->wait; | |
aa395145 | 1926 | } |
1da177e4 LT |
1927 | /* Detach socket from process context. |
1928 | * Announce socket dead, detach it from wait queue and inode. | |
1929 | * Note that parent inode held reference count on this struct sock, | |
1930 | * we do not release it in this function, because protocol | |
1931 | * probably wants some additional cleanups or even continuing | |
1932 | * to work with this socket (TCP). | |
1933 | */ | |
1934 | static inline void sock_orphan(struct sock *sk) | |
1935 | { | |
1936 | write_lock_bh(&sk->sk_callback_lock); | |
1937 | sock_set_flag(sk, SOCK_DEAD); | |
972692e0 | 1938 | sk_set_socket(sk, NULL); |
43815482 | 1939 | sk->sk_wq = NULL; |
1da177e4 LT |
1940 | write_unlock_bh(&sk->sk_callback_lock); |
1941 | } | |
1942 | ||
1943 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1944 | { | |
0ffdaf5b | 1945 | WARN_ON(parent->sk); |
1da177e4 | 1946 | write_lock_bh(&sk->sk_callback_lock); |
333f7909 | 1947 | rcu_assign_pointer(sk->sk_wq, &parent->wq); |
1da177e4 | 1948 | parent->sk = sk; |
972692e0 | 1949 | sk_set_socket(sk, parent); |
86741ec2 | 1950 | sk->sk_uid = SOCK_INODE(parent)->i_uid; |
4237c75c | 1951 | security_sock_graft(sk, parent); |
1da177e4 LT |
1952 | write_unlock_bh(&sk->sk_callback_lock); |
1953 | } | |
1954 | ||
69336bd2 JP |
1955 | kuid_t sock_i_uid(struct sock *sk); |
1956 | unsigned long sock_i_ino(struct sock *sk); | |
1da177e4 | 1957 | |
86741ec2 LC |
1958 | static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk) |
1959 | { | |
1960 | return sk ? sk->sk_uid : make_kuid(net->user_ns, 0); | |
1961 | } | |
1962 | ||
58d607d3 | 1963 | static inline u32 net_tx_rndhash(void) |
877d1f62 | 1964 | { |
58d607d3 ED |
1965 | u32 v = prandom_u32(); |
1966 | ||
1967 | return v ?: 1; | |
1968 | } | |
877d1f62 | 1969 | |
58d607d3 ED |
1970 | static inline void sk_set_txhash(struct sock *sk) |
1971 | { | |
b71eaed8 ED |
1972 | /* This pairs with READ_ONCE() in skb_set_hash_from_sk() */ |
1973 | WRITE_ONCE(sk->sk_txhash, net_tx_rndhash()); | |
877d1f62 TH |
1974 | } |
1975 | ||
9c30ae83 | 1976 | static inline bool sk_rethink_txhash(struct sock *sk) |
265f94ff | 1977 | { |
9c30ae83 | 1978 | if (sk->sk_txhash) { |
265f94ff | 1979 | sk_set_txhash(sk); |
9c30ae83 YC |
1980 | return true; |
1981 | } | |
1982 | return false; | |
265f94ff TH |
1983 | } |
1984 | ||
1da177e4 LT |
1985 | static inline struct dst_entry * |
1986 | __sk_dst_get(struct sock *sk) | |
1987 | { | |
1e1d04e6 HFS |
1988 | return rcu_dereference_check(sk->sk_dst_cache, |
1989 | lockdep_sock_is_held(sk)); | |
1da177e4 LT |
1990 | } |
1991 | ||
1992 | static inline struct dst_entry * | |
1993 | sk_dst_get(struct sock *sk) | |
1994 | { | |
1995 | struct dst_entry *dst; | |
1996 | ||
b6c6712a ED |
1997 | rcu_read_lock(); |
1998 | dst = rcu_dereference(sk->sk_dst_cache); | |
f8864972 ED |
1999 | if (dst && !atomic_inc_not_zero(&dst->__refcnt)) |
2000 | dst = NULL; | |
b6c6712a | 2001 | rcu_read_unlock(); |
1da177e4 LT |
2002 | return dst; |
2003 | } | |
2004 | ||
9c30ae83 | 2005 | static inline void __dst_negative_advice(struct sock *sk) |
b6c6712a ED |
2006 | { |
2007 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); | |
2008 | ||
2009 | if (dst && dst->ops->negative_advice) { | |
2010 | ndst = dst->ops->negative_advice(dst); | |
2011 | ||
2012 | if (ndst != dst) { | |
2013 | rcu_assign_pointer(sk->sk_dst_cache, ndst); | |
0a6957e7 | 2014 | sk_tx_queue_clear(sk); |
9b8805a3 | 2015 | sk->sk_dst_pending_confirm = 0; |
b6c6712a ED |
2016 | } |
2017 | } | |
2018 | } | |
2019 | ||
9c30ae83 YC |
2020 | static inline void dst_negative_advice(struct sock *sk) |
2021 | { | |
2022 | sk_rethink_txhash(sk); | |
2023 | __dst_negative_advice(sk); | |
2024 | } | |
2025 | ||
1da177e4 LT |
2026 | static inline void |
2027 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
2028 | { | |
2029 | struct dst_entry *old_dst; | |
2030 | ||
e022f0b4 | 2031 | sk_tx_queue_clear(sk); |
9b8805a3 | 2032 | sk->sk_dst_pending_confirm = 0; |
95964c6d ED |
2033 | old_dst = rcu_dereference_protected(sk->sk_dst_cache, |
2034 | lockdep_sock_is_held(sk)); | |
b6c6712a | 2035 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
1da177e4 LT |
2036 | dst_release(old_dst); |
2037 | } | |
2038 | ||
2039 | static inline void | |
2040 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
2041 | { | |
7f502361 ED |
2042 | struct dst_entry *old_dst; |
2043 | ||
2044 | sk_tx_queue_clear(sk); | |
9b8805a3 | 2045 | sk->sk_dst_pending_confirm = 0; |
5925a055 | 2046 | old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst); |
7f502361 | 2047 | dst_release(old_dst); |
1da177e4 LT |
2048 | } |
2049 | ||
2050 | static inline void | |
2051 | __sk_dst_reset(struct sock *sk) | |
2052 | { | |
b6c6712a | 2053 | __sk_dst_set(sk, NULL); |
1da177e4 LT |
2054 | } |
2055 | ||
2056 | static inline void | |
2057 | sk_dst_reset(struct sock *sk) | |
2058 | { | |
7f502361 | 2059 | sk_dst_set(sk, NULL); |
1da177e4 LT |
2060 | } |
2061 | ||
69336bd2 | 2062 | struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 2063 | |
69336bd2 | 2064 | struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 2065 | |
9b8805a3 JA |
2066 | static inline void sk_dst_confirm(struct sock *sk) |
2067 | { | |
25c7a6d1 ED |
2068 | if (!READ_ONCE(sk->sk_dst_pending_confirm)) |
2069 | WRITE_ONCE(sk->sk_dst_pending_confirm, 1); | |
9b8805a3 JA |
2070 | } |
2071 | ||
4ff06203 JA |
2072 | static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n) |
2073 | { | |
2074 | if (skb_get_dst_pending_confirm(skb)) { | |
2075 | struct sock *sk = skb->sk; | |
2076 | unsigned long now = jiffies; | |
2077 | ||
2078 | /* avoid dirtying neighbour */ | |
25c7a6d1 ED |
2079 | if (READ_ONCE(n->confirmed) != now) |
2080 | WRITE_ONCE(n->confirmed, now); | |
2081 | if (sk && READ_ONCE(sk->sk_dst_pending_confirm)) | |
2082 | WRITE_ONCE(sk->sk_dst_pending_confirm, 0); | |
4ff06203 JA |
2083 | } |
2084 | } | |
2085 | ||
f60e5990 | 2086 | bool sk_mc_loop(struct sock *sk); |
2087 | ||
dc6b9b78 | 2088 | static inline bool sk_can_gso(const struct sock *sk) |
bcd76111 HX |
2089 | { |
2090 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
2091 | } | |
2092 | ||
69336bd2 | 2093 | void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 2094 | |
c8f44aff | 2095 | static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) |
a465419b ED |
2096 | { |
2097 | sk->sk_route_nocaps |= flags; | |
2098 | sk->sk_route_caps &= ~flags; | |
2099 | } | |
2100 | ||
c6e1a0d1 | 2101 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
57be5bda | 2102 | struct iov_iter *from, char *to, |
912d398d | 2103 | int copy, int offset) |
c6e1a0d1 TH |
2104 | { |
2105 | if (skb->ip_summed == CHECKSUM_NONE) { | |
57be5bda | 2106 | __wsum csum = 0; |
15e6cb46 | 2107 | if (!csum_and_copy_from_iter_full(to, copy, &csum, from)) |
57be5bda | 2108 | return -EFAULT; |
912d398d | 2109 | skb->csum = csum_block_add(skb->csum, csum, offset); |
c6e1a0d1 | 2110 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
15e6cb46 | 2111 | if (!copy_from_iter_full_nocache(to, copy, from)) |
c6e1a0d1 | 2112 | return -EFAULT; |
15e6cb46 | 2113 | } else if (!copy_from_iter_full(to, copy, from)) |
c6e1a0d1 TH |
2114 | return -EFAULT; |
2115 | ||
2116 | return 0; | |
2117 | } | |
2118 | ||
2119 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, | |
57be5bda | 2120 | struct iov_iter *from, int copy) |
c6e1a0d1 | 2121 | { |
912d398d | 2122 | int err, offset = skb->len; |
c6e1a0d1 | 2123 | |
912d398d WY |
2124 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
2125 | copy, offset); | |
c6e1a0d1 | 2126 | if (err) |
912d398d | 2127 | __skb_trim(skb, offset); |
c6e1a0d1 TH |
2128 | |
2129 | return err; | |
2130 | } | |
2131 | ||
57be5bda | 2132 | static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from, |
c6e1a0d1 TH |
2133 | struct sk_buff *skb, |
2134 | struct page *page, | |
2135 | int off, int copy) | |
2136 | { | |
2137 | int err; | |
2138 | ||
912d398d WY |
2139 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
2140 | copy, skb->len); | |
c6e1a0d1 TH |
2141 | if (err) |
2142 | return err; | |
2143 | ||
2144 | skb->len += copy; | |
2145 | skb->data_len += copy; | |
2146 | skb->truesize += copy; | |
ab4e846a | 2147 | sk_wmem_queued_add(sk, copy); |
c6e1a0d1 TH |
2148 | sk_mem_charge(sk, copy); |
2149 | return 0; | |
2150 | } | |
2151 | ||
c564039f ED |
2152 | /** |
2153 | * sk_wmem_alloc_get - returns write allocations | |
2154 | * @sk: socket | |
2155 | * | |
66256e0b | 2156 | * Return: sk_wmem_alloc minus initial offset of one |
c564039f ED |
2157 | */ |
2158 | static inline int sk_wmem_alloc_get(const struct sock *sk) | |
2159 | { | |
14afee4b | 2160 | return refcount_read(&sk->sk_wmem_alloc) - 1; |
c564039f ED |
2161 | } |
2162 | ||
2163 | /** | |
2164 | * sk_rmem_alloc_get - returns read allocations | |
2165 | * @sk: socket | |
2166 | * | |
66256e0b | 2167 | * Return: sk_rmem_alloc |
c564039f ED |
2168 | */ |
2169 | static inline int sk_rmem_alloc_get(const struct sock *sk) | |
2170 | { | |
2171 | return atomic_read(&sk->sk_rmem_alloc); | |
2172 | } | |
2173 | ||
2174 | /** | |
2175 | * sk_has_allocations - check if allocations are outstanding | |
2176 | * @sk: socket | |
2177 | * | |
66256e0b | 2178 | * Return: true if socket has write or read allocations |
c564039f | 2179 | */ |
dc6b9b78 | 2180 | static inline bool sk_has_allocations(const struct sock *sk) |
c564039f ED |
2181 | { |
2182 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); | |
2183 | } | |
2184 | ||
a57de0b4 | 2185 | /** |
1ce0bf50 | 2186 | * skwq_has_sleeper - check if there are any waiting processes |
acfbe96a | 2187 | * @wq: struct socket_wq |
a57de0b4 | 2188 | * |
66256e0b | 2189 | * Return: true if socket_wq has waiting processes |
a57de0b4 | 2190 | * |
1ce0bf50 | 2191 | * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory |
a57de0b4 JO |
2192 | * barrier call. They were added due to the race found within the tcp code. |
2193 | * | |
d651983d | 2194 | * Consider following tcp code paths:: |
a57de0b4 | 2195 | * |
d651983d MCC |
2196 | * CPU1 CPU2 |
2197 | * sys_select receive packet | |
a57de0b4 JO |
2198 | * ... ... |
2199 | * __add_wait_queue update tp->rcv_nxt | |
2200 | * ... ... | |
2201 | * tp->rcv_nxt check sock_def_readable | |
2202 | * ... { | |
43815482 ED |
2203 | * schedule rcu_read_lock(); |
2204 | * wq = rcu_dereference(sk->sk_wq); | |
2205 | * if (wq && waitqueue_active(&wq->wait)) | |
2206 | * wake_up_interruptible(&wq->wait) | |
a57de0b4 JO |
2207 | * ... |
2208 | * } | |
2209 | * | |
2210 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay | |
2211 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 | |
2212 | * could then endup calling schedule and sleep forever if there are no more | |
2213 | * data on the socket. | |
ad462769 | 2214 | * |
a57de0b4 | 2215 | */ |
1ce0bf50 | 2216 | static inline bool skwq_has_sleeper(struct socket_wq *wq) |
a57de0b4 | 2217 | { |
1ce0bf50 | 2218 | return wq && wq_has_sleeper(&wq->wait); |
a57de0b4 JO |
2219 | } |
2220 | ||
2221 | /** | |
2222 | * sock_poll_wait - place memory barrier behind the poll_wait call. | |
2223 | * @filp: file | |
89ab066d | 2224 | * @sock: socket to wait on |
a57de0b4 JO |
2225 | * @p: poll_table |
2226 | * | |
43815482 | 2227 | * See the comments in the wq_has_sleeper function. |
a57de0b4 | 2228 | */ |
89ab066d KG |
2229 | static inline void sock_poll_wait(struct file *filp, struct socket *sock, |
2230 | poll_table *p) | |
a57de0b4 | 2231 | { |
d8bbd13b | 2232 | if (!poll_does_not_wait(p)) { |
333f7909 | 2233 | poll_wait(filp, &sock->wq.wait, p); |
dc6b9b78 | 2234 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
2235 | * socket flags modification. |
2236 | * | |
43815482 | 2237 | * This memory barrier is paired in the wq_has_sleeper. |
dc6b9b78 | 2238 | */ |
a57de0b4 JO |
2239 | smp_mb(); |
2240 | } | |
2241 | } | |
2242 | ||
b73c3d0e TH |
2243 | static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk) |
2244 | { | |
b71eaed8 ED |
2245 | /* This pairs with WRITE_ONCE() in sk_set_txhash() */ |
2246 | u32 txhash = READ_ONCE(sk->sk_txhash); | |
2247 | ||
2248 | if (txhash) { | |
b73c3d0e | 2249 | skb->l4_hash = 1; |
b71eaed8 | 2250 | skb->hash = txhash; |
b73c3d0e TH |
2251 | } |
2252 | } | |
2253 | ||
9e17f8a4 ED |
2254 | void skb_set_owner_w(struct sk_buff *skb, struct sock *sk); |
2255 | ||
1da177e4 | 2256 | /* |
dc6b9b78 | 2257 | * Queue a received datagram if it will fit. Stream and sequenced |
1da177e4 LT |
2258 | * protocols can't normally use this as they need to fit buffers in |
2259 | * and play with them. | |
2260 | * | |
dc6b9b78 | 2261 | * Inlined as it's very short and called for pretty much every |
1da177e4 LT |
2262 | * packet ever received. |
2263 | */ | |
1da177e4 LT |
2264 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
2265 | { | |
d55d87fd | 2266 | skb_orphan(skb); |
1da177e4 LT |
2267 | skb->sk = sk; |
2268 | skb->destructor = sock_rfree; | |
2269 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 2270 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
2271 | } |
2272 | ||
098116e7 | 2273 | static inline __must_check bool skb_set_owner_sk_safe(struct sk_buff *skb, struct sock *sk) |
9adc89af PA |
2274 | { |
2275 | if (sk && refcount_inc_not_zero(&sk->sk_refcnt)) { | |
2276 | skb_orphan(skb); | |
2277 | skb->destructor = sock_efree; | |
2278 | skb->sk = sk; | |
098116e7 | 2279 | return true; |
9adc89af | 2280 | } |
098116e7 | 2281 | return false; |
9adc89af PA |
2282 | } |
2283 | ||
5e10da53 PA |
2284 | static inline void skb_prepare_for_gro(struct sk_buff *skb) |
2285 | { | |
2286 | if (skb->destructor != sock_wfree) { | |
2287 | skb_orphan(skb); | |
2288 | return; | |
2289 | } | |
2290 | skb->slow_gro = 1; | |
2291 | } | |
2292 | ||
69336bd2 JP |
2293 | void sk_reset_timer(struct sock *sk, struct timer_list *timer, |
2294 | unsigned long expires); | |
1da177e4 | 2295 | |
69336bd2 | 2296 | void sk_stop_timer(struct sock *sk, struct timer_list *timer); |
1da177e4 | 2297 | |
08b81d87 GT |
2298 | void sk_stop_timer_sync(struct sock *sk, struct timer_list *timer); |
2299 | ||
65101aec PA |
2300 | int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue, |
2301 | struct sk_buff *skb, unsigned int flags, | |
69629464 ED |
2302 | void (*destructor)(struct sock *sk, |
2303 | struct sk_buff *skb)); | |
e6afc8ac | 2304 | int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
69336bd2 | 2305 | int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 | 2306 | |
69336bd2 | 2307 | int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
364a9e93 | 2308 | struct sk_buff *sock_dequeue_err_skb(struct sock *sk); |
1da177e4 LT |
2309 | |
2310 | /* | |
2311 | * Recover an error report and clear atomically | |
2312 | */ | |
dc6b9b78 | 2313 | |
1da177e4 LT |
2314 | static inline int sock_error(struct sock *sk) |
2315 | { | |
c1cbe4b7 | 2316 | int err; |
f13ef100 ED |
2317 | |
2318 | /* Avoid an atomic operation for the common case. | |
2319 | * This is racy since another cpu/thread can change sk_err under us. | |
2320 | */ | |
2321 | if (likely(data_race(!sk->sk_err))) | |
c1cbe4b7 | 2322 | return 0; |
f13ef100 | 2323 | |
c1cbe4b7 | 2324 | err = xchg(&sk->sk_err, 0); |
1da177e4 LT |
2325 | return -err; |
2326 | } | |
2327 | ||
e3ae2365 AA |
2328 | void sk_error_report(struct sock *sk); |
2329 | ||
1da177e4 LT |
2330 | static inline unsigned long sock_wspace(struct sock *sk) |
2331 | { | |
2332 | int amt = 0; | |
2333 | ||
2334 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
14afee4b | 2335 | amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc); |
dc6b9b78 | 2336 | if (amt < 0) |
1da177e4 LT |
2337 | amt = 0; |
2338 | } | |
2339 | return amt; | |
2340 | } | |
2341 | ||
ceb5d58b ED |
2342 | /* Note: |
2343 | * We use sk->sk_wq_raw, from contexts knowing this | |
2344 | * pointer is not NULL and cannot disappear/change. | |
2345 | */ | |
9cd3e072 | 2346 | static inline void sk_set_bit(int nr, struct sock *sk) |
1da177e4 | 2347 | { |
4be73522 ED |
2348 | if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) && |
2349 | !sock_flag(sk, SOCK_FASYNC)) | |
9317bb69 ED |
2350 | return; |
2351 | ||
ceb5d58b | 2352 | set_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
2353 | } |
2354 | ||
2355 | static inline void sk_clear_bit(int nr, struct sock *sk) | |
2356 | { | |
4be73522 ED |
2357 | if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) && |
2358 | !sock_flag(sk, SOCK_FASYNC)) | |
9317bb69 ED |
2359 | return; |
2360 | ||
ceb5d58b | 2361 | clear_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
2362 | } |
2363 | ||
ceb5d58b | 2364 | static inline void sk_wake_async(const struct sock *sk, int how, int band) |
1da177e4 | 2365 | { |
ceb5d58b ED |
2366 | if (sock_flag(sk, SOCK_FASYNC)) { |
2367 | rcu_read_lock(); | |
2368 | sock_wake_async(rcu_dereference(sk->sk_wq), how, band); | |
2369 | rcu_read_unlock(); | |
2370 | } | |
1da177e4 LT |
2371 | } |
2372 | ||
eea86af6 DB |
2373 | /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might |
2374 | * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak. | |
2375 | * Note: for send buffers, TCP works better if we can build two skbs at | |
2376 | * minimum. | |
7a91b434 | 2377 | */ |
9eb5bf83 | 2378 | #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff))) |
eea86af6 DB |
2379 | |
2380 | #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2) | |
2381 | #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE | |
1da177e4 LT |
2382 | |
2383 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
2384 | { | |
e292f05e ED |
2385 | u32 val; |
2386 | ||
2387 | if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) | |
2388 | return; | |
2389 | ||
2390 | val = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); | |
2391 | ||
2392 | WRITE_ONCE(sk->sk_sndbuf, max_t(u32, val, SOCK_MIN_SNDBUF)); | |
1da177e4 LT |
2393 | } |
2394 | ||
eb934478 ED |
2395 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp, |
2396 | bool force_schedule); | |
1da177e4 | 2397 | |
5640f768 ED |
2398 | /** |
2399 | * sk_page_frag - return an appropriate page_frag | |
2400 | * @sk: socket | |
2401 | * | |
20eb4f29 TH |
2402 | * Use the per task page_frag instead of the per socket one for |
2403 | * optimization when we know that we're in the normal context and owns | |
2404 | * everything that's associated with %current. | |
2405 | * | |
2406 | * gfpflags_allow_blocking() isn't enough here as direct reclaim may nest | |
2407 | * inside other socket operations and end up recursing into sk_page_frag() | |
2408 | * while it's already in use. | |
66256e0b RD |
2409 | * |
2410 | * Return: a per task page_frag if context allows that, | |
2411 | * otherwise a per socket one. | |
5640f768 ED |
2412 | */ |
2413 | static inline struct page_frag *sk_page_frag(struct sock *sk) | |
1da177e4 | 2414 | { |
20eb4f29 | 2415 | if (gfpflags_normal_context(sk->sk_allocation)) |
5640f768 | 2416 | return ¤t->task_frag; |
1da177e4 | 2417 | |
5640f768 | 2418 | return &sk->sk_frag; |
1da177e4 LT |
2419 | } |
2420 | ||
69336bd2 | 2421 | bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag); |
5640f768 | 2422 | |
1da177e4 LT |
2423 | /* |
2424 | * Default write policy as shown to user space via poll/select/SIGIO | |
2425 | */ | |
dc6b9b78 | 2426 | static inline bool sock_writeable(const struct sock *sk) |
1da177e4 | 2427 | { |
e292f05e | 2428 | return refcount_read(&sk->sk_wmem_alloc) < (READ_ONCE(sk->sk_sndbuf) >> 1); |
1da177e4 LT |
2429 | } |
2430 | ||
dd0fc66f | 2431 | static inline gfp_t gfp_any(void) |
1da177e4 | 2432 | { |
99709372 | 2433 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
2434 | } |
2435 | ||
4b1327be WW |
2436 | static inline gfp_t gfp_memcg_charge(void) |
2437 | { | |
2438 | return in_softirq() ? GFP_NOWAIT : GFP_KERNEL; | |
2439 | } | |
2440 | ||
dc6b9b78 | 2441 | static inline long sock_rcvtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2442 | { |
2443 | return noblock ? 0 : sk->sk_rcvtimeo; | |
2444 | } | |
2445 | ||
dc6b9b78 | 2446 | static inline long sock_sndtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2447 | { |
2448 | return noblock ? 0 : sk->sk_sndtimeo; | |
2449 | } | |
2450 | ||
2451 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
2452 | { | |
eac66402 ED |
2453 | int v = waitall ? len : min_t(int, READ_ONCE(sk->sk_rcvlowat), len); |
2454 | ||
2455 | return v ?: 1; | |
1da177e4 LT |
2456 | } |
2457 | ||
2458 | /* Alas, with timeout socket operations are not restartable. | |
2459 | * Compare this to poll(). | |
2460 | */ | |
2461 | static inline int sock_intr_errno(long timeo) | |
2462 | { | |
2463 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
2464 | } | |
2465 | ||
744d5a3e EB |
2466 | struct sock_skb_cb { |
2467 | u32 dropcount; | |
2468 | }; | |
2469 | ||
2470 | /* Store sock_skb_cb at the end of skb->cb[] so protocol families | |
2471 | * using skb->cb[] would keep using it directly and utilize its | |
2472 | * alignement guarantee. | |
2473 | */ | |
c593642c | 2474 | #define SOCK_SKB_CB_OFFSET ((sizeof_field(struct sk_buff, cb) - \ |
744d5a3e EB |
2475 | sizeof(struct sock_skb_cb))) |
2476 | ||
2477 | #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \ | |
2478 | SOCK_SKB_CB_OFFSET)) | |
2479 | ||
b4772ef8 | 2480 | #define sock_skb_cb_check_size(size) \ |
744d5a3e | 2481 | BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET) |
b4772ef8 | 2482 | |
3bc3b96f EB |
2483 | static inline void |
2484 | sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb) | |
2485 | { | |
3665f381 ED |
2486 | SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ? |
2487 | atomic_read(&sk->sk_drops) : 0; | |
3bc3b96f EB |
2488 | } |
2489 | ||
532182cd ED |
2490 | static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb) |
2491 | { | |
2492 | int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs); | |
2493 | ||
2494 | atomic_add(segs, &sk->sk_drops); | |
2495 | } | |
2496 | ||
3a0ed3e9 DD |
2497 | static inline ktime_t sock_read_timestamp(struct sock *sk) |
2498 | { | |
2499 | #if BITS_PER_LONG==32 | |
2500 | unsigned int seq; | |
2501 | ktime_t kt; | |
2502 | ||
2503 | do { | |
2504 | seq = read_seqbegin(&sk->sk_stamp_seq); | |
2505 | kt = sk->sk_stamp; | |
2506 | } while (read_seqretry(&sk->sk_stamp_seq, seq)); | |
2507 | ||
2508 | return kt; | |
2509 | #else | |
f75359f3 | 2510 | return READ_ONCE(sk->sk_stamp); |
3a0ed3e9 DD |
2511 | #endif |
2512 | } | |
2513 | ||
2514 | static inline void sock_write_timestamp(struct sock *sk, ktime_t kt) | |
2515 | { | |
2516 | #if BITS_PER_LONG==32 | |
2517 | write_seqlock(&sk->sk_stamp_seq); | |
2518 | sk->sk_stamp = kt; | |
2519 | write_sequnlock(&sk->sk_stamp_seq); | |
2520 | #else | |
f75359f3 | 2521 | WRITE_ONCE(sk->sk_stamp, kt); |
3a0ed3e9 DD |
2522 | #endif |
2523 | } | |
2524 | ||
69336bd2 JP |
2525 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
2526 | struct sk_buff *skb); | |
2527 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, | |
2528 | struct sk_buff *skb); | |
92f37fd2 | 2529 | |
dc6b9b78 | 2530 | static inline void |
1da177e4 LT |
2531 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
2532 | { | |
b7aa0bf7 | 2533 | ktime_t kt = skb->tstamp; |
20d49473 | 2534 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
a61bbcf2 | 2535 | |
20d49473 PO |
2536 | /* |
2537 | * generate control messages if | |
b9f40e21 | 2538 | * - receive time stamping in software requested |
20d49473 | 2539 | * - software time stamp available and wanted |
20d49473 | 2540 | * - hardware time stamps available and wanted |
20d49473 PO |
2541 | */ |
2542 | if (sock_flag(sk, SOCK_RCVTSTAMP) || | |
b9f40e21 | 2543 | (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) || |
2456e855 TG |
2544 | (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) || |
2545 | (hwtstamps->hwtstamp && | |
b9f40e21 | 2546 | (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE))) |
92f37fd2 ED |
2547 | __sock_recv_timestamp(msg, sk, skb); |
2548 | else | |
3a0ed3e9 | 2549 | sock_write_timestamp(sk, kt); |
6e3e939f JB |
2550 | |
2551 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) | |
2552 | __sock_recv_wifi_status(msg, sk, skb); | |
1da177e4 LT |
2553 | } |
2554 | ||
69336bd2 JP |
2555 | void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2556 | struct sk_buff *skb); | |
767dd033 | 2557 | |
6c7c98ba | 2558 | #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC) |
767dd033 ED |
2559 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2560 | struct sk_buff *skb) | |
2561 | { | |
2562 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ | |
b9f40e21 WB |
2563 | (1UL << SOCK_RCVTSTAMP)) |
2564 | #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \ | |
2565 | SOF_TIMESTAMPING_RAW_HARDWARE) | |
767dd033 | 2566 | |
b9f40e21 | 2567 | if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY) |
767dd033 | 2568 | __sock_recv_ts_and_drops(msg, sk, skb); |
d3fbff30 | 2569 | else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP))) |
3a0ed3e9 | 2570 | sock_write_timestamp(sk, skb->tstamp); |
6c7c98ba | 2571 | else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP)) |
3a0ed3e9 | 2572 | sock_write_timestamp(sk, 0); |
767dd033 | 2573 | } |
3b885787 | 2574 | |
c14ac945 | 2575 | void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags); |
67cc0d40 | 2576 | |
20d49473 | 2577 | /** |
8f932f76 | 2578 | * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped |
20d49473 | 2579 | * @sk: socket sending this packet |
c14ac945 | 2580 | * @tsflags: timestamping flags to use |
140c55d4 | 2581 | * @tx_flags: completed with instructions for time stamping |
8f932f76 | 2582 | * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno) |
140c55d4 | 2583 | * |
d651983d | 2584 | * Note: callers should take care of initial ``*tx_flags`` value (usually 0) |
20d49473 | 2585 | */ |
8f932f76 WB |
2586 | static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags, |
2587 | __u8 *tx_flags, __u32 *tskey) | |
67cc0d40 | 2588 | { |
8f932f76 | 2589 | if (unlikely(tsflags)) { |
c14ac945 | 2590 | __sock_tx_timestamp(tsflags, tx_flags); |
8f932f76 WB |
2591 | if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey && |
2592 | tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK) | |
2593 | *tskey = sk->sk_tskey++; | |
2594 | } | |
67cc0d40 WB |
2595 | if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS))) |
2596 | *tx_flags |= SKBTX_WIFI_STATUS; | |
2597 | } | |
20d49473 | 2598 | |
8f932f76 WB |
2599 | static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags, |
2600 | __u8 *tx_flags) | |
2601 | { | |
2602 | _sock_tx_timestamp(sk, tsflags, tx_flags, NULL); | |
2603 | } | |
2604 | ||
2605 | static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags) | |
2606 | { | |
2607 | _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags, | |
2608 | &skb_shinfo(skb)->tskey); | |
2609 | } | |
2610 | ||
66256e0b | 2611 | DECLARE_STATIC_KEY_FALSE(tcp_rx_skb_cache_key); |
1da177e4 LT |
2612 | /** |
2613 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
2614 | * @sk: socket to eat this skb from |
2615 | * @skb: socket buffer to eat | |
1da177e4 LT |
2616 | * |
2617 | * This routine must be called with interrupts disabled or with the socket | |
2618 | * locked so that the sk_buff queue operation is ok. | |
2619 | */ | |
7bced397 | 2620 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
2621 | { |
2622 | __skb_unlink(skb, &sk->sk_receive_queue); | |
ede61ca4 | 2623 | if (static_branch_unlikely(&tcp_rx_skb_cache_key) && |
8b27dae5 ED |
2624 | !sk->sk_rx_skb_cache) { |
2625 | sk->sk_rx_skb_cache = skb; | |
2626 | skb_orphan(skb); | |
2627 | return; | |
2628 | } | |
1da177e4 LT |
2629 | __kfree_skb(skb); |
2630 | } | |
2631 | ||
3b1e0a65 YH |
2632 | static inline |
2633 | struct net *sock_net(const struct sock *sk) | |
2634 | { | |
c2d9ba9b | 2635 | return read_pnet(&sk->sk_net); |
3b1e0a65 YH |
2636 | } |
2637 | ||
2638 | static inline | |
f5aa23fd | 2639 | void sock_net_set(struct sock *sk, struct net *net) |
3b1e0a65 | 2640 | { |
c2d9ba9b | 2641 | write_pnet(&sk->sk_net, net); |
3b1e0a65 YH |
2642 | } |
2643 | ||
cf7fbe66 JS |
2644 | static inline bool |
2645 | skb_sk_is_prefetched(struct sk_buff *skb) | |
2646 | { | |
2647 | #ifdef CONFIG_INET | |
2648 | return skb->destructor == sock_pfree; | |
2649 | #else | |
2650 | return false; | |
2651 | #endif /* CONFIG_INET */ | |
2652 | } | |
2653 | ||
7ae215d2 JS |
2654 | /* This helper checks if a socket is a full socket, |
2655 | * ie _not_ a timewait or request socket. | |
2656 | */ | |
2657 | static inline bool sk_fullsock(const struct sock *sk) | |
2658 | { | |
2659 | return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV); | |
2660 | } | |
2661 | ||
2662 | static inline bool | |
2663 | sk_is_refcounted(struct sock *sk) | |
2664 | { | |
2665 | /* Only full sockets have sk->sk_flags. */ | |
2666 | return !sk_fullsock(sk) || !sock_flag(sk, SOCK_RCU_FREE); | |
2667 | } | |
2668 | ||
71489e21 | 2669 | /** |
045065f0 LR |
2670 | * skb_steal_sock - steal a socket from an sk_buff |
2671 | * @skb: sk_buff to steal the socket from | |
2672 | * @refcounted: is set to true if the socket is reference-counted | |
71489e21 JS |
2673 | */ |
2674 | static inline struct sock * | |
2675 | skb_steal_sock(struct sk_buff *skb, bool *refcounted) | |
23542618 | 2676 | { |
efc27f8c | 2677 | if (skb->sk) { |
23542618 KK |
2678 | struct sock *sk = skb->sk; |
2679 | ||
71489e21 | 2680 | *refcounted = true; |
7ae215d2 JS |
2681 | if (skb_sk_is_prefetched(skb)) |
2682 | *refcounted = sk_is_refcounted(sk); | |
23542618 KK |
2683 | skb->destructor = NULL; |
2684 | skb->sk = NULL; | |
2685 | return sk; | |
2686 | } | |
71489e21 | 2687 | *refcounted = false; |
23542618 KK |
2688 | return NULL; |
2689 | } | |
2690 | ||
ebf4e808 IL |
2691 | /* Checks if this SKB belongs to an HW offloaded socket |
2692 | * and whether any SW fallbacks are required based on dev. | |
41477662 | 2693 | * Check decrypted mark in case skb_orphan() cleared socket. |
ebf4e808 IL |
2694 | */ |
2695 | static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb, | |
2696 | struct net_device *dev) | |
2697 | { | |
2698 | #ifdef CONFIG_SOCK_VALIDATE_XMIT | |
2699 | struct sock *sk = skb->sk; | |
2700 | ||
41477662 | 2701 | if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb) { |
ebf4e808 | 2702 | skb = sk->sk_validate_xmit_skb(sk, dev, skb); |
41477662 JK |
2703 | #ifdef CONFIG_TLS_DEVICE |
2704 | } else if (unlikely(skb->decrypted)) { | |
2705 | pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n"); | |
2706 | kfree_skb(skb); | |
2707 | skb = NULL; | |
2708 | #endif | |
2709 | } | |
ebf4e808 IL |
2710 | #endif |
2711 | ||
2712 | return skb; | |
2713 | } | |
2714 | ||
e446f9df ED |
2715 | /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV |
2716 | * SYNACK messages can be attached to either ones (depending on SYNCOOKIE) | |
2717 | */ | |
2718 | static inline bool sk_listener(const struct sock *sk) | |
2719 | { | |
2720 | return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV); | |
2721 | } | |
2722 | ||
193d357d | 2723 | void sock_enable_timestamp(struct sock *sk, enum sock_flags flag); |
69336bd2 JP |
2724 | int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level, |
2725 | int type); | |
1da177e4 | 2726 | |
a3b299da EB |
2727 | bool sk_ns_capable(const struct sock *sk, |
2728 | struct user_namespace *user_ns, int cap); | |
2729 | bool sk_capable(const struct sock *sk, int cap); | |
2730 | bool sk_net_capable(const struct sock *sk, int cap); | |
2731 | ||
a2d133b1 JH |
2732 | void sk_get_meminfo(const struct sock *sk, u32 *meminfo); |
2733 | ||
eaa72dc4 ED |
2734 | /* Take into consideration the size of the struct sk_buff overhead in the |
2735 | * determination of these values, since that is non-constant across | |
2736 | * platforms. This makes socket queueing behavior and performance | |
2737 | * not depend upon such differences. | |
2738 | */ | |
2739 | #define _SK_MEM_PACKETS 256 | |
2740 | #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256) | |
2741 | #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
2742 | #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
2743 | ||
1da177e4 LT |
2744 | extern __u32 sysctl_wmem_max; |
2745 | extern __u32 sysctl_rmem_max; | |
2746 | ||
b245be1f | 2747 | extern int sysctl_tstamp_allow_data; |
6baf1f41 DM |
2748 | extern int sysctl_optmem_max; |
2749 | ||
20380731 ACM |
2750 | extern __u32 sysctl_wmem_default; |
2751 | extern __u32 sysctl_rmem_default; | |
20380731 | 2752 | |
723783d0 | 2753 | #define SKB_FRAG_PAGE_ORDER get_order(32768) |
ce27ec60 ED |
2754 | DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key); |
2755 | ||
a3dcaf17 ED |
2756 | static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto) |
2757 | { | |
2758 | /* Does this proto have per netns sysctl_wmem ? */ | |
2759 | if (proto->sysctl_wmem_offset) | |
2760 | return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset); | |
2761 | ||
2762 | return *proto->sysctl_wmem; | |
2763 | } | |
2764 | ||
2765 | static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto) | |
2766 | { | |
2767 | /* Does this proto have per netns sysctl_rmem ? */ | |
2768 | if (proto->sysctl_rmem_offset) | |
2769 | return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset); | |
2770 | ||
2771 | return *proto->sysctl_rmem; | |
2772 | } | |
2773 | ||
c9f1f58d ED |
2774 | /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10) |
2775 | * Some wifi drivers need to tweak it to get more chunks. | |
2776 | * They can use this helper from their ndo_start_xmit() | |
2777 | */ | |
2778 | static inline void sk_pacing_shift_update(struct sock *sk, int val) | |
2779 | { | |
7c68fa2b | 2780 | if (!sk || !sk_fullsock(sk) || READ_ONCE(sk->sk_pacing_shift) == val) |
c9f1f58d | 2781 | return; |
7c68fa2b | 2782 | WRITE_ONCE(sk->sk_pacing_shift, val); |
c9f1f58d ED |
2783 | } |
2784 | ||
54dc3e33 DA |
2785 | /* if a socket is bound to a device, check that the given device |
2786 | * index is either the same or that the socket is bound to an L3 | |
2787 | * master device and the given device index is also enslaved to | |
2788 | * that L3 master | |
2789 | */ | |
2790 | static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif) | |
2791 | { | |
2792 | int mdif; | |
2793 | ||
2794 | if (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif) | |
2795 | return true; | |
2796 | ||
2797 | mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif); | |
2798 | if (mdif && mdif == sk->sk_bound_dev_if) | |
2799 | return true; | |
2800 | ||
2801 | return false; | |
2802 | } | |
2803 | ||
43a825af BT |
2804 | void sock_def_readable(struct sock *sk); |
2805 | ||
8ea204c2 | 2806 | int sock_bindtoindex(struct sock *sk, int ifindex, bool lock_sk); |
371087aa | 2807 | void sock_set_timestamp(struct sock *sk, int optname, bool valbool); |
d463126e YL |
2808 | int sock_set_timestamping(struct sock *sk, int optname, |
2809 | struct so_timestamping timestamping); | |
ced122d9 | 2810 | |
783da70e | 2811 | void sock_enable_timestamps(struct sock *sk); |
c433594c | 2812 | void sock_no_linger(struct sock *sk); |
ce3d9544 | 2813 | void sock_set_keepalive(struct sock *sk); |
6e434967 | 2814 | void sock_set_priority(struct sock *sk, u32 priority); |
26cfabf9 | 2815 | void sock_set_rcvbuf(struct sock *sk, int val); |
84d1c617 | 2816 | void sock_set_mark(struct sock *sk, u32 val); |
b58f0e8f | 2817 | void sock_set_reuseaddr(struct sock *sk); |
fe31a326 | 2818 | void sock_set_reuseport(struct sock *sk); |
76ee0785 | 2819 | void sock_set_sndtimeo(struct sock *sk, s64 secs); |
b58f0e8f | 2820 | |
c0425a42 CH |
2821 | int sock_bind_add(struct sock *sk, struct sockaddr *addr, int addr_len); |
2822 | ||
1da177e4 | 2823 | #endif /* _SOCK_H */ |