]>
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> |
1da177e4 | 62 | |
c31504dc | 63 | #include <linux/atomic.h> |
41c6d650 | 64 | #include <linux/refcount.h> |
1da177e4 LT |
65 | #include <net/dst.h> |
66 | #include <net/checksum.h> | |
1d0ab253 | 67 | #include <net/tcp_states.h> |
b9f40e21 | 68 | #include <linux/net_tstamp.h> |
54dc3e33 | 69 | #include <net/l3mdev.h> |
1da177e4 LT |
70 | |
71 | /* | |
72 | * This structure really needs to be cleaned up. | |
73 | * Most of it is for TCP, and not used by any of | |
74 | * the other protocols. | |
75 | */ | |
76 | ||
77 | /* Define this to get the SOCK_DBG debugging facility. */ | |
78 | #define SOCK_DEBUGGING | |
79 | #ifdef SOCK_DEBUGGING | |
80 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
81 | printk(KERN_DEBUG msg); } while (0) | |
82 | #else | |
4cd9029d | 83 | /* Validate arguments and do nothing */ |
b9075fa9 | 84 | static inline __printf(2, 3) |
dc6b9b78 | 85 | void SOCK_DEBUG(const struct sock *sk, const char *msg, ...) |
4cd9029d SH |
86 | { |
87 | } | |
1da177e4 LT |
88 | #endif |
89 | ||
90 | /* This is the per-socket lock. The spinlock provides a synchronization | |
91 | * between user contexts and software interrupt processing, whereas the | |
92 | * mini-semaphore synchronizes multiple users amongst themselves. | |
93 | */ | |
1da177e4 LT |
94 | typedef struct { |
95 | spinlock_t slock; | |
d2e9117c | 96 | int owned; |
1da177e4 | 97 | wait_queue_head_t wq; |
a5b5bb9a IM |
98 | /* |
99 | * We express the mutex-alike socket_lock semantics | |
100 | * to the lock validator by explicitly managing | |
101 | * the slock as a lock variant (in addition to | |
102 | * the slock itself): | |
103 | */ | |
104 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
105 | struct lockdep_map dep_map; | |
106 | #endif | |
1da177e4 LT |
107 | } socket_lock_t; |
108 | ||
1da177e4 | 109 | struct sock; |
8feaf0c0 | 110 | struct proto; |
0eeb8ffc | 111 | struct net; |
1da177e4 | 112 | |
077b393d ED |
113 | typedef __u32 __bitwise __portpair; |
114 | typedef __u64 __bitwise __addrpair; | |
115 | ||
1da177e4 | 116 | /** |
4dc3b16b | 117 | * struct sock_common - minimal network layer representation of sockets |
68835aba ED |
118 | * @skc_daddr: Foreign IPv4 addr |
119 | * @skc_rcv_saddr: Bound local IPv4 addr | |
4dc6dc71 | 120 | * @skc_hash: hash value used with various protocol lookup tables |
d4cada4a | 121 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
ce43b03e ED |
122 | * @skc_dport: placeholder for inet_dport/tw_dport |
123 | * @skc_num: placeholder for inet_num/tw_num | |
4dc3b16b PP |
124 | * @skc_family: network address family |
125 | * @skc_state: Connection state | |
126 | * @skc_reuse: %SO_REUSEADDR setting | |
055dc21a | 127 | * @skc_reuseport: %SO_REUSEPORT setting |
4dc3b16b | 128 | * @skc_bound_dev_if: bound device index if != 0 |
4dc3b16b | 129 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
512615b6 | 130 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol |
8feaf0c0 | 131 | * @skc_prot: protocol handlers inside a network family |
07feaebf | 132 | * @skc_net: reference to the network namespace of this socket |
68835aba ED |
133 | * @skc_node: main hash linkage for various protocol lookup tables |
134 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol | |
135 | * @skc_tx_queue_mapping: tx queue number for this connection | |
c6345ce7 | 136 | * @skc_rx_queue_mapping: rx queue number for this connection |
8e5eb54d ED |
137 | * @skc_flags: place holder for sk_flags |
138 | * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, | |
139 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings | |
70da268b | 140 | * @skc_incoming_cpu: record/match cpu processing incoming packets |
68835aba | 141 | * @skc_refcnt: reference count |
4dc3b16b PP |
142 | * |
143 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
144 | * for struct sock and struct inet_timewait_sock. |
145 | */ | |
1da177e4 | 146 | struct sock_common { |
ce43b03e | 147 | /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned |
05dbc7b5 | 148 | * address on 64bit arches : cf INET_MATCH() |
4dc6dc71 | 149 | */ |
ce43b03e | 150 | union { |
077b393d | 151 | __addrpair skc_addrpair; |
ce43b03e ED |
152 | struct { |
153 | __be32 skc_daddr; | |
154 | __be32 skc_rcv_saddr; | |
155 | }; | |
156 | }; | |
d4cada4a ED |
157 | union { |
158 | unsigned int skc_hash; | |
159 | __u16 skc_u16hashes[2]; | |
160 | }; | |
ce43b03e ED |
161 | /* skc_dport && skc_num must be grouped as well */ |
162 | union { | |
077b393d | 163 | __portpair skc_portpair; |
ce43b03e ED |
164 | struct { |
165 | __be16 skc_dport; | |
166 | __u16 skc_num; | |
167 | }; | |
168 | }; | |
169 | ||
4dc6dc71 ED |
170 | unsigned short skc_family; |
171 | volatile unsigned char skc_state; | |
055dc21a | 172 | unsigned char skc_reuse:4; |
9fe516ba ED |
173 | unsigned char skc_reuseport:1; |
174 | unsigned char skc_ipv6only:1; | |
26abe143 | 175 | unsigned char skc_net_refcnt:1; |
4dc6dc71 | 176 | int skc_bound_dev_if; |
512615b6 ED |
177 | union { |
178 | struct hlist_node skc_bind_node; | |
ca065d0c | 179 | struct hlist_node skc_portaddr_node; |
512615b6 | 180 | }; |
8feaf0c0 | 181 | struct proto *skc_prot; |
0c5c9fb5 | 182 | possible_net_t skc_net; |
efe4208f ED |
183 | |
184 | #if IS_ENABLED(CONFIG_IPV6) | |
185 | struct in6_addr skc_v6_daddr; | |
186 | struct in6_addr skc_v6_rcv_saddr; | |
187 | #endif | |
188 | ||
33cf7c90 ED |
189 | atomic64_t skc_cookie; |
190 | ||
8e5eb54d ED |
191 | /* following fields are padding to force |
192 | * offset(struct sock, sk_refcnt) == 128 on 64bit arches | |
193 | * assuming IPV6 is enabled. We use this padding differently | |
194 | * for different kind of 'sockets' | |
195 | */ | |
196 | union { | |
197 | unsigned long skc_flags; | |
198 | struct sock *skc_listener; /* request_sock */ | |
199 | struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */ | |
200 | }; | |
68835aba ED |
201 | /* |
202 | * fields between dontcopy_begin/dontcopy_end | |
203 | * are not copied in sock_copy() | |
204 | */ | |
928c41e7 | 205 | /* private: */ |
68835aba | 206 | int skc_dontcopy_begin[0]; |
928c41e7 | 207 | /* public: */ |
68835aba ED |
208 | union { |
209 | struct hlist_node skc_node; | |
210 | struct hlist_nulls_node skc_nulls_node; | |
211 | }; | |
755c31cd | 212 | unsigned short skc_tx_queue_mapping; |
c6345ce7 AN |
213 | #ifdef CONFIG_XPS |
214 | unsigned short skc_rx_queue_mapping; | |
215 | #endif | |
ed53d0ab ED |
216 | union { |
217 | int skc_incoming_cpu; | |
218 | u32 skc_rcv_wnd; | |
d475f090 | 219 | u32 skc_tw_rcv_nxt; /* struct tcp_timewait_sock */ |
ed53d0ab | 220 | }; |
70da268b | 221 | |
41c6d650 | 222 | refcount_t skc_refcnt; |
928c41e7 | 223 | /* private: */ |
68835aba | 224 | int skc_dontcopy_end[0]; |
ed53d0ab ED |
225 | union { |
226 | u32 skc_rxhash; | |
227 | u32 skc_window_clamp; | |
d475f090 | 228 | u32 skc_tw_snd_nxt; /* struct tcp_timewait_sock */ |
ed53d0ab | 229 | }; |
928c41e7 | 230 | /* public: */ |
1da177e4 LT |
231 | }; |
232 | ||
6ac99e8f MKL |
233 | struct bpf_sk_storage; |
234 | ||
1da177e4 LT |
235 | /** |
236 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 237 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
238 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
239 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
240 | * @sk_lock: synchronizer | |
cdfbabfb | 241 | * @sk_kern_sock: True if sock is using kernel lock classes |
4dc3b16b | 242 | * @sk_rcvbuf: size of receive buffer in bytes |
43815482 | 243 | * @sk_wq: sock wait queue and async head |
421b3885 | 244 | * @sk_rx_dst: receive input route used by early demux |
4dc3b16b | 245 | * @sk_dst_cache: destination cache |
9b8805a3 | 246 | * @sk_dst_pending_confirm: need to confirm neighbour |
4dc3b16b | 247 | * @sk_policy: flow policy |
4dc3b16b PP |
248 | * @sk_receive_queue: incoming packets |
249 | * @sk_wmem_alloc: transmit queue bytes committed | |
771edcaf | 250 | * @sk_tsq_flags: TCP Small Queues flags |
4dc3b16b PP |
251 | * @sk_write_queue: Packet sending queue |
252 | * @sk_omem_alloc: "o" is "option" or "other" | |
253 | * @sk_wmem_queued: persistent queue size | |
254 | * @sk_forward_alloc: space allocated forward | |
06021292 | 255 | * @sk_napi_id: id of the last napi context to receive data for sk |
dafcc438 | 256 | * @sk_ll_usec: usecs to busypoll when there is no data |
4dc3b16b | 257 | * @sk_allocation: allocation mode |
95bd09eb | 258 | * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler) |
218af599 | 259 | * @sk_pacing_status: Pacing status (requested, handled by sch_fq) |
c3f40d7c | 260 | * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE) |
4dc3b16b | 261 | * @sk_sndbuf: size of send buffer in bytes |
771edcaf | 262 | * @__sk_flags_offset: empty field used to determine location of bitfield |
293de7de | 263 | * @sk_padding: unused element for alignment |
28448b80 TH |
264 | * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets |
265 | * @sk_no_check_rx: allow zero checksum in RX packets | |
4dc3b16b | 266 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
a465419b | 267 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
bcd76111 | 268 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
82cc1a7a | 269 | * @sk_gso_max_size: Maximum GSO segment size to build |
1485348d | 270 | * @sk_gso_max_segs: Maximum number of GSO segments |
3a9b76fd | 271 | * @sk_pacing_shift: scaling factor for TCP Small Queues |
4dc3b16b | 272 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
273 | * @sk_backlog: always used with the per-socket spinlock held |
274 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
275 | * @sk_error_queue: rarely used | |
33c732c3 WC |
276 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
277 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 278 | * @sk_err: last error |
33c732c3 WC |
279 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
280 | * persistent failure not just 'timed out' | |
cb61cb9b | 281 | * @sk_drops: raw/udp drops counter |
4dc3b16b PP |
282 | * @sk_ack_backlog: current listen backlog |
283 | * @sk_max_ack_backlog: listen backlog set in listen() | |
771edcaf | 284 | * @sk_uid: user id of owner |
4dc3b16b PP |
285 | * @sk_priority: %SO_PRIORITY setting |
286 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
287 | * @sk_protocol: which protocol this socket belongs in this network family | |
53c3fa20 RD |
288 | * @sk_peer_pid: &struct pid for this socket's peer |
289 | * @sk_peer_cred: %SO_PEERCRED setting | |
4dc3b16b PP |
290 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
291 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
292 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
b73c3d0e | 293 | * @sk_txhash: computed flow hash for use on transmit |
4dc3b16b | 294 | * @sk_filter: socket filtering instructions |
4dc3b16b PP |
295 | * @sk_timer: sock cleanup timer |
296 | * @sk_stamp: time stamp of last packet received | |
3a0ed3e9 | 297 | * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only |
b9f40e21 | 298 | * @sk_tsflags: SO_TIMESTAMPING socket options |
09c2d251 | 299 | * @sk_tskey: counter to disambiguate concurrent tstamp requests |
52267790 | 300 | * @sk_zckey: counter to order MSG_ZEROCOPY notifications |
4dc3b16b PP |
301 | * @sk_socket: Identd and reporting IO signals |
302 | * @sk_user_data: RPC layer private data | |
5640f768 | 303 | * @sk_frag: cached page frag |
d3d4f0a0 | 304 | * @sk_peek_off: current peek_offset value |
4dc3b16b | 305 | * @sk_send_head: front of stuff to transmit |
67be2dd1 | 306 | * @sk_security: used by security modules |
31729363 | 307 | * @sk_mark: generic packet mark |
2a56a1fe | 308 | * @sk_cgrp_data: cgroup data for this cgroup |
baac50bb | 309 | * @sk_memcg: this socket's memory cgroup association |
4dc3b16b PP |
310 | * @sk_write_pending: a write to stream socket waits to start |
311 | * @sk_state_change: callback to indicate change in the state of the sock | |
312 | * @sk_data_ready: callback to indicate there is data to be processed | |
313 | * @sk_write_space: callback to indicate there is bf sending space available | |
314 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
315 | * @sk_backlog_rcv: callback to process the backlog | |
316 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
ef456144 | 317 | * @sk_reuseport_cb: reuseport group container |
293de7de | 318 | * @sk_rcu: used during RCU grace period |
80b14dee RC |
319 | * @sk_clockid: clockid used by time-based scheduling (SO_TXTIME) |
320 | * @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME | |
321 | * @sk_txtime_unused: unused txtime flags | |
293de7de | 322 | */ |
1da177e4 LT |
323 | struct sock { |
324 | /* | |
8feaf0c0 | 325 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
326 | * don't add nothing before this first member (__sk_common) --acme |
327 | */ | |
328 | struct sock_common __sk_common; | |
4dc6dc71 ED |
329 | #define sk_node __sk_common.skc_node |
330 | #define sk_nulls_node __sk_common.skc_nulls_node | |
331 | #define sk_refcnt __sk_common.skc_refcnt | |
e022f0b4 | 332 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
c6345ce7 AN |
333 | #ifdef CONFIG_XPS |
334 | #define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping | |
335 | #endif | |
4dc6dc71 | 336 | |
68835aba ED |
337 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
338 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end | |
4dc6dc71 | 339 | #define sk_hash __sk_common.skc_hash |
50805466 | 340 | #define sk_portpair __sk_common.skc_portpair |
05dbc7b5 ED |
341 | #define sk_num __sk_common.skc_num |
342 | #define sk_dport __sk_common.skc_dport | |
50805466 ED |
343 | #define sk_addrpair __sk_common.skc_addrpair |
344 | #define sk_daddr __sk_common.skc_daddr | |
345 | #define sk_rcv_saddr __sk_common.skc_rcv_saddr | |
1da177e4 LT |
346 | #define sk_family __sk_common.skc_family |
347 | #define sk_state __sk_common.skc_state | |
348 | #define sk_reuse __sk_common.skc_reuse | |
055dc21a | 349 | #define sk_reuseport __sk_common.skc_reuseport |
9fe516ba | 350 | #define sk_ipv6only __sk_common.skc_ipv6only |
26abe143 | 351 | #define sk_net_refcnt __sk_common.skc_net_refcnt |
1da177e4 | 352 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
1da177e4 | 353 | #define sk_bind_node __sk_common.skc_bind_node |
8feaf0c0 | 354 | #define sk_prot __sk_common.skc_prot |
07feaebf | 355 | #define sk_net __sk_common.skc_net |
efe4208f ED |
356 | #define sk_v6_daddr __sk_common.skc_v6_daddr |
357 | #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr | |
33cf7c90 | 358 | #define sk_cookie __sk_common.skc_cookie |
70da268b | 359 | #define sk_incoming_cpu __sk_common.skc_incoming_cpu |
8e5eb54d | 360 | #define sk_flags __sk_common.skc_flags |
ed53d0ab | 361 | #define sk_rxhash __sk_common.skc_rxhash |
efe4208f | 362 | |
1da177e4 | 363 | socket_lock_t sk_lock; |
9115e8cd ED |
364 | atomic_t sk_drops; |
365 | int sk_rcvlowat; | |
366 | struct sk_buff_head sk_error_queue; | |
8b27dae5 | 367 | struct sk_buff *sk_rx_skb_cache; |
b178bb3d | 368 | struct sk_buff_head sk_receive_queue; |
fa438ccf ED |
369 | /* |
370 | * The backlog queue is special, it is always used with | |
371 | * the per-socket spinlock held and requires low latency | |
372 | * access. Therefore we special case it's implementation. | |
b178bb3d ED |
373 | * Note : rmem_alloc is in this structure to fill a hole |
374 | * on 64bit arches, not because its logically part of | |
375 | * backlog. | |
fa438ccf ED |
376 | */ |
377 | struct { | |
b178bb3d ED |
378 | atomic_t rmem_alloc; |
379 | int len; | |
380 | struct sk_buff *head; | |
381 | struct sk_buff *tail; | |
fa438ccf | 382 | } sk_backlog; |
b178bb3d | 383 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
2c8c56e1 | 384 | |
9115e8cd | 385 | int sk_forward_alloc; |
e0d1095a | 386 | #ifdef CONFIG_NET_RX_BUSY_POLL |
dafcc438 | 387 | unsigned int sk_ll_usec; |
9115e8cd ED |
388 | /* ===== mostly read cache line ===== */ |
389 | unsigned int sk_napi_id; | |
b178bb3d | 390 | #endif |
b178bb3d ED |
391 | int sk_rcvbuf; |
392 | ||
393 | struct sk_filter __rcu *sk_filter; | |
ceb5d58b ED |
394 | union { |
395 | struct socket_wq __rcu *sk_wq; | |
396 | struct socket_wq *sk_wq_raw; | |
397 | }; | |
def8b4fa | 398 | #ifdef CONFIG_XFRM |
d188ba86 | 399 | struct xfrm_policy __rcu *sk_policy[2]; |
def8b4fa | 400 | #endif |
deaa5854 | 401 | struct dst_entry *sk_rx_dst; |
0e36cbb3 | 402 | struct dst_entry __rcu *sk_dst_cache; |
1da177e4 | 403 | atomic_t sk_omem_alloc; |
4e07a91c | 404 | int sk_sndbuf; |
9115e8cd ED |
405 | |
406 | /* ===== cache line for TX ===== */ | |
407 | int sk_wmem_queued; | |
14afee4b | 408 | refcount_t sk_wmem_alloc; |
9115e8cd | 409 | unsigned long sk_tsq_flags; |
75c119af ED |
410 | union { |
411 | struct sk_buff *sk_send_head; | |
412 | struct rb_root tcp_rtx_queue; | |
413 | }; | |
472c2e07 | 414 | struct sk_buff *sk_tx_skb_cache; |
1da177e4 | 415 | struct sk_buff_head sk_write_queue; |
9115e8cd ED |
416 | __s32 sk_peek_off; |
417 | int sk_write_pending; | |
9b8805a3 | 418 | __u32 sk_dst_pending_confirm; |
218af599 | 419 | u32 sk_pacing_status; /* see enum sk_pacing */ |
9115e8cd ED |
420 | long sk_sndtimeo; |
421 | struct timer_list sk_timer; | |
422 | __u32 sk_priority; | |
423 | __u32 sk_mark; | |
76a9ebe8 ED |
424 | unsigned long sk_pacing_rate; /* bytes per second */ |
425 | unsigned long sk_max_pacing_rate; | |
9115e8cd ED |
426 | struct page_frag sk_frag; |
427 | netdev_features_t sk_route_caps; | |
428 | netdev_features_t sk_route_nocaps; | |
0a6b2a1d | 429 | netdev_features_t sk_route_forced_caps; |
9115e8cd ED |
430 | int sk_gso_type; |
431 | unsigned int sk_gso_max_size; | |
432 | gfp_t sk_allocation; | |
433 | __u32 sk_txhash; | |
fc64869c AR |
434 | |
435 | /* | |
436 | * Because of non atomicity rules, all | |
437 | * changes are protected by socket lock. | |
438 | */ | |
aa4c1037 DA |
439 | unsigned int __sk_flags_offset[0]; |
440 | #ifdef __BIG_ENDIAN_BITFIELD | |
441 | #define SK_FL_PROTO_SHIFT 16 | |
442 | #define SK_FL_PROTO_MASK 0x00ff0000 | |
443 | ||
444 | #define SK_FL_TYPE_SHIFT 0 | |
445 | #define SK_FL_TYPE_MASK 0x0000ffff | |
446 | #else | |
447 | #define SK_FL_PROTO_SHIFT 8 | |
448 | #define SK_FL_PROTO_MASK 0x0000ff00 | |
449 | ||
450 | #define SK_FL_TYPE_SHIFT 16 | |
451 | #define SK_FL_TYPE_MASK 0xffff0000 | |
452 | #endif | |
453 | ||
cdfbabfb DH |
454 | unsigned int sk_padding : 1, |
455 | sk_kern_sock : 1, | |
28448b80 TH |
456 | sk_no_check_tx : 1, |
457 | sk_no_check_rx : 1, | |
b178bb3d ED |
458 | sk_userlocks : 4, |
459 | sk_protocol : 8, | |
460 | sk_type : 16; | |
1485348d | 461 | u16 sk_gso_max_segs; |
3a9b76fd | 462 | u8 sk_pacing_shift; |
1da177e4 | 463 | unsigned long sk_lingertime; |
476e19cf | 464 | struct proto *sk_prot_creator; |
1da177e4 LT |
465 | rwlock_t sk_callback_lock; |
466 | int sk_err, | |
467 | sk_err_soft; | |
becb74f0 ED |
468 | u32 sk_ack_backlog; |
469 | u32 sk_max_ack_backlog; | |
86741ec2 | 470 | kuid_t sk_uid; |
109f6e39 EB |
471 | struct pid *sk_peer_pid; |
472 | const struct cred *sk_peer_cred; | |
1da177e4 | 473 | long sk_rcvtimeo; |
b7aa0bf7 | 474 | ktime_t sk_stamp; |
3a0ed3e9 DD |
475 | #if BITS_PER_LONG==32 |
476 | seqlock_t sk_stamp_seq; | |
477 | #endif | |
b9f40e21 | 478 | u16 sk_tsflags; |
fc64869c | 479 | u8 sk_shutdown; |
09c2d251 | 480 | u32 sk_tskey; |
52267790 | 481 | atomic_t sk_zckey; |
80b14dee RC |
482 | |
483 | u8 sk_clockid; | |
484 | u8 sk_txtime_deadline_mode : 1, | |
4b15c707 JSP |
485 | sk_txtime_report_errors : 1, |
486 | sk_txtime_unused : 6; | |
80b14dee | 487 | |
1da177e4 LT |
488 | struct socket *sk_socket; |
489 | void *sk_user_data; | |
d5f64238 | 490 | #ifdef CONFIG_SECURITY |
1da177e4 | 491 | void *sk_security; |
d5f64238 | 492 | #endif |
2a56a1fe | 493 | struct sock_cgroup_data sk_cgrp_data; |
baac50bb | 494 | struct mem_cgroup *sk_memcg; |
1da177e4 | 495 | void (*sk_state_change)(struct sock *sk); |
676d2369 | 496 | void (*sk_data_ready)(struct sock *sk); |
1da177e4 LT |
497 | void (*sk_write_space)(struct sock *sk); |
498 | void (*sk_error_report)(struct sock *sk); | |
dc6b9b78 ED |
499 | int (*sk_backlog_rcv)(struct sock *sk, |
500 | struct sk_buff *skb); | |
ebf4e808 IL |
501 | #ifdef CONFIG_SOCK_VALIDATE_XMIT |
502 | struct sk_buff* (*sk_validate_xmit_skb)(struct sock *sk, | |
503 | struct net_device *dev, | |
504 | struct sk_buff *skb); | |
505 | #endif | |
1da177e4 | 506 | void (*sk_destruct)(struct sock *sk); |
ef456144 | 507 | struct sock_reuseport __rcu *sk_reuseport_cb; |
6ac99e8f MKL |
508 | #ifdef CONFIG_BPF_SYSCALL |
509 | struct bpf_sk_storage __rcu *sk_bpf_storage; | |
510 | #endif | |
a4298e45 | 511 | struct rcu_head sk_rcu; |
1da177e4 LT |
512 | }; |
513 | ||
218af599 ED |
514 | enum sk_pacing { |
515 | SK_PACING_NONE = 0, | |
516 | SK_PACING_NEEDED = 1, | |
517 | SK_PACING_FQ = 2, | |
518 | }; | |
519 | ||
559835ea PS |
520 | #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data))) |
521 | ||
522 | #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk))) | |
523 | #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr) | |
524 | ||
4a17fd52 PE |
525 | /* |
526 | * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK | |
527 | * or not whether his port will be reused by someone else. SK_FORCE_REUSE | |
528 | * on a socket means that the socket will reuse everybody else's port | |
529 | * without looking at the other's sk_reuse value. | |
530 | */ | |
531 | ||
532 | #define SK_NO_REUSE 0 | |
533 | #define SK_CAN_REUSE 1 | |
534 | #define SK_FORCE_REUSE 2 | |
535 | ||
627d2d6b | 536 | int sk_set_peek_off(struct sock *sk, int val); |
537 | ||
ef64a54f PE |
538 | static inline int sk_peek_offset(struct sock *sk, int flags) |
539 | { | |
b9bb53f3 | 540 | if (unlikely(flags & MSG_PEEK)) { |
a0917e0b | 541 | return READ_ONCE(sk->sk_peek_off); |
b9bb53f3 WB |
542 | } |
543 | ||
544 | return 0; | |
ef64a54f PE |
545 | } |
546 | ||
547 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) | |
548 | { | |
b9bb53f3 WB |
549 | s32 off = READ_ONCE(sk->sk_peek_off); |
550 | ||
551 | if (unlikely(off >= 0)) { | |
552 | off = max_t(s32, off - val, 0); | |
553 | WRITE_ONCE(sk->sk_peek_off, off); | |
ef64a54f PE |
554 | } |
555 | } | |
556 | ||
557 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) | |
558 | { | |
b9bb53f3 | 559 | sk_peek_offset_bwd(sk, -val); |
ef64a54f PE |
560 | } |
561 | ||
1da177e4 LT |
562 | /* |
563 | * Hashed lists helper routines | |
564 | */ | |
c4146644 LZ |
565 | static inline struct sock *sk_entry(const struct hlist_node *node) |
566 | { | |
567 | return hlist_entry(node, struct sock, sk_node); | |
568 | } | |
569 | ||
e48c414e | 570 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
571 | { |
572 | return hlist_entry(head->first, struct sock, sk_node); | |
573 | } | |
574 | ||
e48c414e | 575 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
576 | { |
577 | return hlist_empty(head) ? NULL : __sk_head(head); | |
578 | } | |
579 | ||
88ab1932 ED |
580 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
581 | { | |
582 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); | |
583 | } | |
584 | ||
585 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) | |
586 | { | |
587 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); | |
588 | } | |
589 | ||
e48c414e | 590 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 | 591 | { |
6c59ebd3 | 592 | return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node); |
1da177e4 LT |
593 | } |
594 | ||
88ab1932 ED |
595 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
596 | { | |
597 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? | |
598 | hlist_nulls_entry(sk->sk_nulls_node.next, | |
599 | struct sock, sk_nulls_node) : | |
600 | NULL; | |
601 | } | |
602 | ||
dc6b9b78 | 603 | static inline bool sk_unhashed(const struct sock *sk) |
1da177e4 LT |
604 | { |
605 | return hlist_unhashed(&sk->sk_node); | |
606 | } | |
607 | ||
dc6b9b78 | 608 | static inline bool sk_hashed(const struct sock *sk) |
1da177e4 | 609 | { |
da753bea | 610 | return !sk_unhashed(sk); |
1da177e4 LT |
611 | } |
612 | ||
dc6b9b78 | 613 | static inline void sk_node_init(struct hlist_node *node) |
1da177e4 LT |
614 | { |
615 | node->pprev = NULL; | |
616 | } | |
617 | ||
dc6b9b78 | 618 | static inline void sk_nulls_node_init(struct hlist_nulls_node *node) |
88ab1932 ED |
619 | { |
620 | node->pprev = NULL; | |
621 | } | |
622 | ||
dc6b9b78 | 623 | static inline void __sk_del_node(struct sock *sk) |
1da177e4 LT |
624 | { |
625 | __hlist_del(&sk->sk_node); | |
626 | } | |
627 | ||
808f5114 | 628 | /* NB: equivalent to hlist_del_init_rcu */ |
dc6b9b78 | 629 | static inline bool __sk_del_node_init(struct sock *sk) |
1da177e4 LT |
630 | { |
631 | if (sk_hashed(sk)) { | |
632 | __sk_del_node(sk); | |
633 | sk_node_init(&sk->sk_node); | |
dc6b9b78 | 634 | return true; |
1da177e4 | 635 | } |
dc6b9b78 | 636 | return false; |
1da177e4 LT |
637 | } |
638 | ||
639 | /* Grab socket reference count. This operation is valid only | |
640 | when sk is ALREADY grabbed f.e. it is found in hash table | |
641 | or a list and the lookup is made under lock preventing hash table | |
642 | modifications. | |
643 | */ | |
644 | ||
f9a7cbbf | 645 | static __always_inline void sock_hold(struct sock *sk) |
1da177e4 | 646 | { |
41c6d650 | 647 | refcount_inc(&sk->sk_refcnt); |
1da177e4 LT |
648 | } |
649 | ||
650 | /* Ungrab socket in the context, which assumes that socket refcnt | |
651 | cannot hit zero, f.e. it is true in context of any socketcall. | |
652 | */ | |
f9a7cbbf | 653 | static __always_inline void __sock_put(struct sock *sk) |
1da177e4 | 654 | { |
41c6d650 | 655 | refcount_dec(&sk->sk_refcnt); |
1da177e4 LT |
656 | } |
657 | ||
dc6b9b78 | 658 | static inline bool sk_del_node_init(struct sock *sk) |
1da177e4 | 659 | { |
dc6b9b78 | 660 | bool rc = __sk_del_node_init(sk); |
1da177e4 LT |
661 | |
662 | if (rc) { | |
663 | /* paranoid for a while -acme */ | |
41c6d650 | 664 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
1da177e4 LT |
665 | __sock_put(sk); |
666 | } | |
667 | return rc; | |
668 | } | |
808f5114 | 669 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
1da177e4 | 670 | |
dc6b9b78 | 671 | static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 ED |
672 | { |
673 | if (sk_hashed(sk)) { | |
88ab1932 | 674 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
dc6b9b78 | 675 | return true; |
271b72c7 | 676 | } |
dc6b9b78 | 677 | return false; |
271b72c7 ED |
678 | } |
679 | ||
dc6b9b78 | 680 | static inline bool sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 | 681 | { |
dc6b9b78 | 682 | bool rc = __sk_nulls_del_node_init_rcu(sk); |
271b72c7 ED |
683 | |
684 | if (rc) { | |
685 | /* paranoid for a while -acme */ | |
41c6d650 | 686 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
271b72c7 ED |
687 | __sock_put(sk); |
688 | } | |
689 | return rc; | |
690 | } | |
691 | ||
dc6b9b78 | 692 | static inline void __sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
693 | { |
694 | hlist_add_head(&sk->sk_node, list); | |
695 | } | |
696 | ||
dc6b9b78 | 697 | static inline void sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
698 | { |
699 | sock_hold(sk); | |
700 | __sk_add_node(sk, list); | |
701 | } | |
702 | ||
dc6b9b78 | 703 | static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
808f5114 | 704 | { |
705 | sock_hold(sk); | |
d296ba60 CG |
706 | if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport && |
707 | sk->sk_family == AF_INET6) | |
708 | hlist_add_tail_rcu(&sk->sk_node, list); | |
709 | else | |
710 | hlist_add_head_rcu(&sk->sk_node, list); | |
808f5114 | 711 | } |
712 | ||
a4dc6a49 MC |
713 | static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list) |
714 | { | |
715 | sock_hold(sk); | |
716 | hlist_add_tail_rcu(&sk->sk_node, list); | |
717 | } | |
718 | ||
dc6b9b78 | 719 | static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 | 720 | { |
d7efc6c1 | 721 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
271b72c7 ED |
722 | } |
723 | ||
8dbd76e7 ED |
724 | static inline void __sk_nulls_add_node_tail_rcu(struct sock *sk, struct hlist_nulls_head *list) |
725 | { | |
726 | hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list); | |
727 | } | |
728 | ||
dc6b9b78 | 729 | static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 ED |
730 | { |
731 | sock_hold(sk); | |
88ab1932 | 732 | __sk_nulls_add_node_rcu(sk, list); |
271b72c7 ED |
733 | } |
734 | ||
dc6b9b78 | 735 | static inline void __sk_del_bind_node(struct sock *sk) |
1da177e4 LT |
736 | { |
737 | __hlist_del(&sk->sk_bind_node); | |
738 | } | |
739 | ||
dc6b9b78 | 740 | static inline void sk_add_bind_node(struct sock *sk, |
1da177e4 LT |
741 | struct hlist_head *list) |
742 | { | |
743 | hlist_add_head(&sk->sk_bind_node, list); | |
744 | } | |
745 | ||
b67bfe0d SL |
746 | #define sk_for_each(__sk, list) \ |
747 | hlist_for_each_entry(__sk, list, sk_node) | |
748 | #define sk_for_each_rcu(__sk, list) \ | |
749 | hlist_for_each_entry_rcu(__sk, list, sk_node) | |
88ab1932 ED |
750 | #define sk_nulls_for_each(__sk, node, list) \ |
751 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) | |
752 | #define sk_nulls_for_each_rcu(__sk, node, list) \ | |
753 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) | |
b67bfe0d SL |
754 | #define sk_for_each_from(__sk) \ |
755 | hlist_for_each_entry_from(__sk, sk_node) | |
88ab1932 ED |
756 | #define sk_nulls_for_each_from(__sk, node) \ |
757 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ | |
758 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) | |
b67bfe0d SL |
759 | #define sk_for_each_safe(__sk, tmp, list) \ |
760 | hlist_for_each_entry_safe(__sk, tmp, list, sk_node) | |
761 | #define sk_for_each_bound(__sk, list) \ | |
762 | hlist_for_each_entry(__sk, list, sk_bind_node) | |
1da177e4 | 763 | |
2dc41cff | 764 | /** |
ca065d0c | 765 | * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset |
2dc41cff DH |
766 | * @tpos: the type * to use as a loop cursor. |
767 | * @pos: the &struct hlist_node to use as a loop cursor. | |
768 | * @head: the head for your list. | |
769 | * @offset: offset of hlist_node within the struct. | |
770 | * | |
771 | */ | |
ca065d0c | 772 | #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \ |
b6f4f848 | 773 | for (pos = rcu_dereference(hlist_first_rcu(head)); \ |
ca065d0c | 774 | pos != NULL && \ |
2dc41cff | 775 | ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \ |
b6f4f848 | 776 | pos = rcu_dereference(hlist_next_rcu(pos))) |
2dc41cff | 777 | |
c336d148 EB |
778 | static inline struct user_namespace *sk_user_ns(struct sock *sk) |
779 | { | |
780 | /* Careful only use this in a context where these parameters | |
781 | * can not change and must all be valid, such as recvmsg from | |
782 | * userspace. | |
783 | */ | |
784 | return sk->sk_socket->file->f_cred->user_ns; | |
785 | } | |
786 | ||
1da177e4 LT |
787 | /* Sock flags */ |
788 | enum sock_flags { | |
789 | SOCK_DEAD, | |
790 | SOCK_DONE, | |
791 | SOCK_URGINLINE, | |
792 | SOCK_KEEPOPEN, | |
793 | SOCK_LINGER, | |
794 | SOCK_DESTROY, | |
795 | SOCK_BROADCAST, | |
796 | SOCK_TIMESTAMP, | |
797 | SOCK_ZAPPED, | |
798 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
799 | SOCK_DBG, /* %SO_DEBUG setting */ | |
800 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 801 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 LT |
802 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
803 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
7cb02404 | 804 | SOCK_MEMALLOC, /* VM depends on this socket for swapping */ |
20d49473 | 805 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
bcdce719 | 806 | SOCK_FASYNC, /* fasync() active */ |
3b885787 | 807 | SOCK_RXQ_OVFL, |
1cdebb42 | 808 | SOCK_ZEROCOPY, /* buffers from userspace */ |
6e3e939f | 809 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
3bdc0eba BG |
810 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
811 | * Will use last 4 bytes of packet sent from | |
812 | * user-space instead. | |
813 | */ | |
d59577b6 | 814 | SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */ |
7d4c04fc | 815 | SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */ |
a4298e45 | 816 | SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */ |
80b14dee | 817 | SOCK_TXTIME, |
e4a2a304 | 818 | SOCK_XDP, /* XDP is attached */ |
887feae3 | 819 | SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */ |
1da177e4 LT |
820 | }; |
821 | ||
01ce63c9 MRL |
822 | #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)) |
823 | ||
53b924b3 RB |
824 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
825 | { | |
826 | nsk->sk_flags = osk->sk_flags; | |
827 | } | |
828 | ||
1da177e4 LT |
829 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
830 | { | |
831 | __set_bit(flag, &sk->sk_flags); | |
832 | } | |
833 | ||
834 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
835 | { | |
836 | __clear_bit(flag, &sk->sk_flags); | |
837 | } | |
838 | ||
1b23a5df | 839 | static inline bool sock_flag(const struct sock *sk, enum sock_flags flag) |
1da177e4 LT |
840 | { |
841 | return test_bit(flag, &sk->sk_flags); | |
842 | } | |
843 | ||
c93bdd0e | 844 | #ifdef CONFIG_NET |
a7950ae8 | 845 | DECLARE_STATIC_KEY_FALSE(memalloc_socks_key); |
c93bdd0e MG |
846 | static inline int sk_memalloc_socks(void) |
847 | { | |
a7950ae8 | 848 | return static_branch_unlikely(&memalloc_socks_key); |
c93bdd0e MG |
849 | } |
850 | #else | |
851 | ||
852 | static inline int sk_memalloc_socks(void) | |
853 | { | |
854 | return 0; | |
855 | } | |
856 | ||
857 | #endif | |
858 | ||
7450aaf6 | 859 | static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask) |
99a1dec7 | 860 | { |
7450aaf6 | 861 | return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC); |
99a1dec7 MG |
862 | } |
863 | ||
1da177e4 LT |
864 | static inline void sk_acceptq_removed(struct sock *sk) |
865 | { | |
288efe86 | 866 | WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog - 1); |
1da177e4 LT |
867 | } |
868 | ||
869 | static inline void sk_acceptq_added(struct sock *sk) | |
870 | { | |
288efe86 | 871 | WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog + 1); |
1da177e4 LT |
872 | } |
873 | ||
dc6b9b78 | 874 | static inline bool sk_acceptq_is_full(const struct sock *sk) |
1da177e4 | 875 | { |
099ecf59 | 876 | return READ_ONCE(sk->sk_ack_backlog) > READ_ONCE(sk->sk_max_ack_backlog); |
1da177e4 LT |
877 | } |
878 | ||
879 | /* | |
880 | * Compute minimal free write space needed to queue new packets. | |
881 | */ | |
dc6b9b78 | 882 | static inline int sk_stream_min_wspace(const struct sock *sk) |
1da177e4 | 883 | { |
ab4e846a | 884 | return READ_ONCE(sk->sk_wmem_queued) >> 1; |
1da177e4 LT |
885 | } |
886 | ||
dc6b9b78 | 887 | static inline int sk_stream_wspace(const struct sock *sk) |
1da177e4 | 888 | { |
ab4e846a ED |
889 | return READ_ONCE(sk->sk_sndbuf) - READ_ONCE(sk->sk_wmem_queued); |
890 | } | |
891 | ||
892 | static inline void sk_wmem_queued_add(struct sock *sk, int val) | |
893 | { | |
894 | WRITE_ONCE(sk->sk_wmem_queued, sk->sk_wmem_queued + val); | |
1da177e4 LT |
895 | } |
896 | ||
69336bd2 | 897 | void sk_stream_write_space(struct sock *sk); |
1da177e4 | 898 | |
8eae939f | 899 | /* OOB backlog add */ |
a3a858ff | 900 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
9ee6b535 | 901 | { |
7fee226a | 902 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
222d7dbd | 903 | skb_dst_force(skb); |
7fee226a ED |
904 | |
905 | if (!sk->sk_backlog.tail) | |
9ed498c6 | 906 | WRITE_ONCE(sk->sk_backlog.head, skb); |
7fee226a | 907 | else |
9ee6b535 | 908 | sk->sk_backlog.tail->next = skb; |
7fee226a | 909 | |
9ed498c6 | 910 | WRITE_ONCE(sk->sk_backlog.tail, skb); |
9ee6b535 SH |
911 | skb->next = NULL; |
912 | } | |
1da177e4 | 913 | |
c377411f ED |
914 | /* |
915 | * Take into account size of receive queue and backlog queue | |
0fd7bac6 ED |
916 | * Do not take into account this skb truesize, |
917 | * to allow even a single big packet to come. | |
c377411f | 918 | */ |
274f482d | 919 | static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit) |
c377411f ED |
920 | { |
921 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); | |
922 | ||
f545a38f | 923 | return qsize > limit; |
c377411f ED |
924 | } |
925 | ||
8eae939f | 926 | /* The per-socket spinlock must be held here. */ |
f545a38f ED |
927 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb, |
928 | unsigned int limit) | |
8eae939f | 929 | { |
274f482d | 930 | if (sk_rcvqueues_full(sk, limit)) |
8eae939f ZY |
931 | return -ENOBUFS; |
932 | ||
c7c49b8f ED |
933 | /* |
934 | * If the skb was allocated from pfmemalloc reserves, only | |
935 | * allow SOCK_MEMALLOC sockets to use it as this socket is | |
936 | * helping free memory | |
937 | */ | |
938 | if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC)) | |
939 | return -ENOMEM; | |
940 | ||
a3a858ff | 941 | __sk_add_backlog(sk, skb); |
8eae939f ZY |
942 | sk->sk_backlog.len += skb->truesize; |
943 | return 0; | |
944 | } | |
945 | ||
69336bd2 | 946 | int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
b4b9e355 | 947 | |
c57943a1 PZ |
948 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
949 | { | |
b4b9e355 MG |
950 | if (sk_memalloc_socks() && skb_pfmemalloc(skb)) |
951 | return __sk_backlog_rcv(sk, skb); | |
952 | ||
c57943a1 PZ |
953 | return sk->sk_backlog_rcv(sk, skb); |
954 | } | |
955 | ||
2c8c56e1 ED |
956 | static inline void sk_incoming_cpu_update(struct sock *sk) |
957 | { | |
34cfb542 PA |
958 | int cpu = raw_smp_processor_id(); |
959 | ||
7170a977 ED |
960 | if (unlikely(READ_ONCE(sk->sk_incoming_cpu) != cpu)) |
961 | WRITE_ONCE(sk->sk_incoming_cpu, cpu); | |
2c8c56e1 ED |
962 | } |
963 | ||
fe477558 | 964 | static inline void sock_rps_record_flow_hash(__u32 hash) |
c58dc01b DM |
965 | { |
966 | #ifdef CONFIG_RPS | |
967 | struct rps_sock_flow_table *sock_flow_table; | |
968 | ||
969 | rcu_read_lock(); | |
970 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
fe477558 | 971 | rps_record_sock_flow(sock_flow_table, hash); |
c58dc01b DM |
972 | rcu_read_unlock(); |
973 | #endif | |
974 | } | |
975 | ||
fe477558 TH |
976 | static inline void sock_rps_record_flow(const struct sock *sk) |
977 | { | |
c9d8ca04 | 978 | #ifdef CONFIG_RPS |
dc05360f | 979 | if (static_branch_unlikely(&rfs_needed)) { |
13bfff25 ED |
980 | /* Reading sk->sk_rxhash might incur an expensive cache line |
981 | * miss. | |
982 | * | |
983 | * TCP_ESTABLISHED does cover almost all states where RFS | |
984 | * might be useful, and is cheaper [1] than testing : | |
985 | * IPv4: inet_sk(sk)->inet_daddr | |
986 | * IPv6: ipv6_addr_any(&sk->sk_v6_daddr) | |
987 | * OR an additional socket flag | |
988 | * [1] : sk_state and sk_prot are in the same cache line. | |
989 | */ | |
990 | if (sk->sk_state == TCP_ESTABLISHED) | |
991 | sock_rps_record_flow_hash(sk->sk_rxhash); | |
992 | } | |
c9d8ca04 | 993 | #endif |
fe477558 TH |
994 | } |
995 | ||
bdeab991 TH |
996 | static inline void sock_rps_save_rxhash(struct sock *sk, |
997 | const struct sk_buff *skb) | |
c58dc01b DM |
998 | { |
999 | #ifdef CONFIG_RPS | |
567e4b79 | 1000 | if (unlikely(sk->sk_rxhash != skb->hash)) |
61b905da | 1001 | sk->sk_rxhash = skb->hash; |
c58dc01b DM |
1002 | #endif |
1003 | } | |
1004 | ||
bdeab991 TH |
1005 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
1006 | { | |
1007 | #ifdef CONFIG_RPS | |
bdeab991 TH |
1008 | sk->sk_rxhash = 0; |
1009 | #endif | |
1010 | } | |
1011 | ||
d9dc8b0f | 1012 | #define sk_wait_event(__sk, __timeo, __condition, __wait) \ |
cfcabdcc SH |
1013 | ({ int __rc; \ |
1014 | release_sock(__sk); \ | |
1015 | __rc = __condition; \ | |
1016 | if (!__rc) { \ | |
d9dc8b0f WC |
1017 | *(__timeo) = wait_woken(__wait, \ |
1018 | TASK_INTERRUPTIBLE, \ | |
1019 | *(__timeo)); \ | |
cfcabdcc | 1020 | } \ |
d9dc8b0f | 1021 | sched_annotate_sleep(); \ |
cfcabdcc SH |
1022 | lock_sock(__sk); \ |
1023 | __rc = __condition; \ | |
1024 | __rc; \ | |
1025 | }) | |
1da177e4 | 1026 | |
69336bd2 JP |
1027 | int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
1028 | int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
1029 | void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
1030 | int sk_stream_error(struct sock *sk, int flags, int err); | |
1031 | void sk_stream_kill_queues(struct sock *sk); | |
1032 | void sk_set_memalloc(struct sock *sk); | |
1033 | void sk_clear_memalloc(struct sock *sk); | |
1da177e4 | 1034 | |
d41a69f1 ED |
1035 | void __sk_flush_backlog(struct sock *sk); |
1036 | ||
1037 | static inline bool sk_flush_backlog(struct sock *sk) | |
1038 | { | |
1039 | if (unlikely(READ_ONCE(sk->sk_backlog.tail))) { | |
1040 | __sk_flush_backlog(sk); | |
1041 | return true; | |
1042 | } | |
1043 | return false; | |
1044 | } | |
1045 | ||
dfbafc99 | 1046 | int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb); |
1da177e4 | 1047 | |
60236fdd | 1048 | struct request_sock_ops; |
6d6ee43e | 1049 | struct timewait_sock_ops; |
ab1e0a13 | 1050 | struct inet_hashinfo; |
fc8717ba | 1051 | struct raw_hashinfo; |
f16a7dd5 | 1052 | struct smc_hashinfo; |
de477254 | 1053 | struct module; |
2e6599cb | 1054 | |
f77d6021 | 1055 | /* |
5f0d5a3a | 1056 | * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes |
f77d6021 ED |
1057 | * un-modified. Special care is taken when initializing object to zero. |
1058 | */ | |
1059 | static inline void sk_prot_clear_nulls(struct sock *sk, int size) | |
1060 | { | |
1061 | if (offsetof(struct sock, sk_node.next) != 0) | |
1062 | memset(sk, 0, offsetof(struct sock, sk_node.next)); | |
1063 | memset(&sk->sk_node.pprev, 0, | |
1064 | size - offsetof(struct sock, sk_node.pprev)); | |
1065 | } | |
1066 | ||
1da177e4 LT |
1067 | /* Networking protocol blocks we attach to sockets. |
1068 | * socket layer -> transport layer interface | |
1da177e4 LT |
1069 | */ |
1070 | struct proto { | |
dc6b9b78 | 1071 | void (*close)(struct sock *sk, |
1da177e4 | 1072 | long timeout); |
d74bad4e AI |
1073 | int (*pre_connect)(struct sock *sk, |
1074 | struct sockaddr *uaddr, | |
1075 | int addr_len); | |
1da177e4 | 1076 | int (*connect)(struct sock *sk, |
dc6b9b78 | 1077 | struct sockaddr *uaddr, |
1da177e4 LT |
1078 | int addr_len); |
1079 | int (*disconnect)(struct sock *sk, int flags); | |
1080 | ||
cdfbabfb DH |
1081 | struct sock * (*accept)(struct sock *sk, int flags, int *err, |
1082 | bool kern); | |
1da177e4 LT |
1083 | |
1084 | int (*ioctl)(struct sock *sk, int cmd, | |
1085 | unsigned long arg); | |
1086 | int (*init)(struct sock *sk); | |
7d06b2e0 | 1087 | void (*destroy)(struct sock *sk); |
1da177e4 | 1088 | void (*shutdown)(struct sock *sk, int how); |
dc6b9b78 | 1089 | int (*setsockopt)(struct sock *sk, int level, |
1da177e4 | 1090 | int optname, char __user *optval, |
b7058842 | 1091 | unsigned int optlen); |
dc6b9b78 ED |
1092 | int (*getsockopt)(struct sock *sk, int level, |
1093 | int optname, char __user *optval, | |
1094 | int __user *option); | |
4b9d07a4 | 1095 | void (*keepalive)(struct sock *sk, int valbool); |
af01d537 | 1096 | #ifdef CONFIG_COMPAT |
3fdadf7d DM |
1097 | int (*compat_setsockopt)(struct sock *sk, |
1098 | int level, | |
1099 | int optname, char __user *optval, | |
b7058842 | 1100 | unsigned int optlen); |
3fdadf7d DM |
1101 | int (*compat_getsockopt)(struct sock *sk, |
1102 | int level, | |
1103 | int optname, char __user *optval, | |
1104 | int __user *option); | |
709b46e8 EB |
1105 | int (*compat_ioctl)(struct sock *sk, |
1106 | unsigned int cmd, unsigned long arg); | |
af01d537 | 1107 | #endif |
1b784140 YX |
1108 | int (*sendmsg)(struct sock *sk, struct msghdr *msg, |
1109 | size_t len); | |
1110 | int (*recvmsg)(struct sock *sk, struct msghdr *msg, | |
dc6b9b78 ED |
1111 | size_t len, int noblock, int flags, |
1112 | int *addr_len); | |
1da177e4 LT |
1113 | int (*sendpage)(struct sock *sk, struct page *page, |
1114 | int offset, size_t size, int flags); | |
dc6b9b78 | 1115 | int (*bind)(struct sock *sk, |
1da177e4 LT |
1116 | struct sockaddr *uaddr, int addr_len); |
1117 | ||
dc6b9b78 | 1118 | int (*backlog_rcv) (struct sock *sk, |
1da177e4 LT |
1119 | struct sk_buff *skb); |
1120 | ||
46d3ceab ED |
1121 | void (*release_cb)(struct sock *sk); |
1122 | ||
1da177e4 | 1123 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
086c653f | 1124 | int (*hash)(struct sock *sk); |
1da177e4 | 1125 | void (*unhash)(struct sock *sk); |
719f8358 | 1126 | void (*rehash)(struct sock *sk); |
1da177e4 LT |
1127 | int (*get_port)(struct sock *sk, unsigned short snum); |
1128 | ||
286ab3d4 | 1129 | /* Keeping track of sockets in use */ |
65f76517 | 1130 | #ifdef CONFIG_PROC_FS |
13ff3d6f | 1131 | unsigned int inuse_idx; |
65f76517 | 1132 | #endif |
ebb53d75 | 1133 | |
a74f0fa0 | 1134 | bool (*stream_memory_free)(const struct sock *sk, int wake); |
8934ce2f | 1135 | bool (*stream_memory_read)(const struct sock *sk); |
1da177e4 | 1136 | /* Memory pressure */ |
5c52ba17 | 1137 | void (*enter_memory_pressure)(struct sock *sk); |
06044751 | 1138 | void (*leave_memory_pressure)(struct sock *sk); |
8d987e5c | 1139 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
1748376b | 1140 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
1da177e4 LT |
1141 | /* |
1142 | * Pressure flag: try to collapse. | |
1143 | * Technical note: it is used by multiple contexts non atomically. | |
3ab224be | 1144 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
1145 | * is strict, actions are advisory and have some latency. |
1146 | */ | |
06044751 | 1147 | unsigned long *memory_pressure; |
8d987e5c | 1148 | long *sysctl_mem; |
a3dcaf17 | 1149 | |
1da177e4 LT |
1150 | int *sysctl_wmem; |
1151 | int *sysctl_rmem; | |
a3dcaf17 ED |
1152 | u32 sysctl_wmem_offset; |
1153 | u32 sysctl_rmem_offset; | |
1154 | ||
1da177e4 | 1155 | int max_header; |
7ba42910 | 1156 | bool no_autobind; |
1da177e4 | 1157 | |
271b72c7 | 1158 | struct kmem_cache *slab; |
1da177e4 | 1159 | unsigned int obj_size; |
d50112ed | 1160 | slab_flags_t slab_flags; |
7bbdb81e AD |
1161 | unsigned int useroffset; /* Usercopy region offset */ |
1162 | unsigned int usersize; /* Usercopy region size */ | |
1da177e4 | 1163 | |
dd24c001 | 1164 | struct percpu_counter *orphan_count; |
8feaf0c0 | 1165 | |
60236fdd | 1166 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 1167 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 1168 | |
39d8cda7 PE |
1169 | union { |
1170 | struct inet_hashinfo *hashinfo; | |
645ca708 | 1171 | struct udp_table *udp_table; |
fc8717ba | 1172 | struct raw_hashinfo *raw_hash; |
f16a7dd5 | 1173 | struct smc_hashinfo *smc_hash; |
39d8cda7 | 1174 | } h; |
ab1e0a13 | 1175 | |
1da177e4 LT |
1176 | struct module *owner; |
1177 | ||
1178 | char name[32]; | |
1179 | ||
1180 | struct list_head node; | |
e6848976 ACM |
1181 | #ifdef SOCK_REFCNT_DEBUG |
1182 | atomic_t socks; | |
e1aab161 | 1183 | #endif |
64be0aed | 1184 | int (*diag_destroy)(struct sock *sk, int err); |
3859a271 | 1185 | } __randomize_layout; |
e1aab161 | 1186 | |
69336bd2 JP |
1187 | int proto_register(struct proto *prot, int alloc_slab); |
1188 | void proto_unregister(struct proto *prot); | |
bf2ae2e4 | 1189 | int sock_load_diag_module(int family, int protocol); |
1da177e4 | 1190 | |
e6848976 ACM |
1191 | #ifdef SOCK_REFCNT_DEBUG |
1192 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
1193 | { | |
1194 | atomic_inc(&sk->sk_prot->socks); | |
1195 | } | |
1196 | ||
1197 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
1198 | { | |
1199 | atomic_dec(&sk->sk_prot->socks); | |
1200 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
1201 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
1202 | } | |
1203 | ||
dec34fb0 | 1204 | static inline void sk_refcnt_debug_release(const struct sock *sk) |
e6848976 | 1205 | { |
41c6d650 | 1206 | if (refcount_read(&sk->sk_refcnt) != 1) |
e6848976 | 1207 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", |
41c6d650 | 1208 | sk->sk_prot->name, sk, refcount_read(&sk->sk_refcnt)); |
e6848976 ACM |
1209 | } |
1210 | #else /* SOCK_REFCNT_DEBUG */ | |
1211 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
1212 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
1213 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
1214 | #endif /* SOCK_REFCNT_DEBUG */ | |
1215 | ||
a74f0fa0 | 1216 | static inline bool __sk_stream_memory_free(const struct sock *sk, int wake) |
c9bee3b7 | 1217 | { |
ab4e846a | 1218 | if (READ_ONCE(sk->sk_wmem_queued) >= READ_ONCE(sk->sk_sndbuf)) |
c9bee3b7 ED |
1219 | return false; |
1220 | ||
1221 | return sk->sk_prot->stream_memory_free ? | |
a74f0fa0 | 1222 | sk->sk_prot->stream_memory_free(sk, wake) : true; |
c9bee3b7 ED |
1223 | } |
1224 | ||
a74f0fa0 ED |
1225 | static inline bool sk_stream_memory_free(const struct sock *sk) |
1226 | { | |
1227 | return __sk_stream_memory_free(sk, 0); | |
1228 | } | |
1229 | ||
1230 | static inline bool __sk_stream_is_writeable(const struct sock *sk, int wake) | |
64dc6130 | 1231 | { |
c9bee3b7 | 1232 | return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) && |
a74f0fa0 ED |
1233 | __sk_stream_memory_free(sk, wake); |
1234 | } | |
1235 | ||
1236 | static inline bool sk_stream_is_writeable(const struct sock *sk) | |
1237 | { | |
1238 | return __sk_stream_is_writeable(sk, 0); | |
64dc6130 | 1239 | } |
e1aab161 | 1240 | |
54fd9c2d DB |
1241 | static inline int sk_under_cgroup_hierarchy(struct sock *sk, |
1242 | struct cgroup *ancestor) | |
1243 | { | |
1244 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
1245 | return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data), | |
1246 | ancestor); | |
1247 | #else | |
1248 | return -ENOTSUPP; | |
1249 | #endif | |
1250 | } | |
c9bee3b7 | 1251 | |
180d8cd9 GC |
1252 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
1253 | { | |
1254 | return sk->sk_prot->memory_pressure != NULL; | |
1255 | } | |
1256 | ||
1257 | static inline bool sk_under_memory_pressure(const struct sock *sk) | |
1258 | { | |
1259 | if (!sk->sk_prot->memory_pressure) | |
1260 | return false; | |
e1aab161 | 1261 | |
baac50bb JW |
1262 | if (mem_cgroup_sockets_enabled && sk->sk_memcg && |
1263 | mem_cgroup_under_socket_pressure(sk->sk_memcg)) | |
e805605c | 1264 | return true; |
e1aab161 | 1265 | |
35b87f6c | 1266 | return !!*sk->sk_prot->memory_pressure; |
180d8cd9 GC |
1267 | } |
1268 | ||
180d8cd9 GC |
1269 | static inline long |
1270 | sk_memory_allocated(const struct sock *sk) | |
1271 | { | |
e805605c | 1272 | return atomic_long_read(sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1273 | } |
1274 | ||
1275 | static inline long | |
e805605c | 1276 | sk_memory_allocated_add(struct sock *sk, int amt) |
180d8cd9 | 1277 | { |
e805605c | 1278 | return atomic_long_add_return(amt, sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1279 | } |
1280 | ||
1281 | static inline void | |
0e90b31f | 1282 | sk_memory_allocated_sub(struct sock *sk, int amt) |
180d8cd9 | 1283 | { |
e805605c | 1284 | atomic_long_sub(amt, sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1285 | } |
1286 | ||
1287 | static inline void sk_sockets_allocated_dec(struct sock *sk) | |
1288 | { | |
af95d7df | 1289 | percpu_counter_dec(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1290 | } |
1291 | ||
1292 | static inline void sk_sockets_allocated_inc(struct sock *sk) | |
1293 | { | |
af95d7df | 1294 | percpu_counter_inc(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1295 | } |
1296 | ||
5bf325a5 | 1297 | static inline u64 |
180d8cd9 GC |
1298 | sk_sockets_allocated_read_positive(struct sock *sk) |
1299 | { | |
af95d7df | 1300 | return percpu_counter_read_positive(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1301 | } |
1302 | ||
1303 | static inline int | |
1304 | proto_sockets_allocated_sum_positive(struct proto *prot) | |
1305 | { | |
1306 | return percpu_counter_sum_positive(prot->sockets_allocated); | |
1307 | } | |
1308 | ||
1309 | static inline long | |
1310 | proto_memory_allocated(struct proto *prot) | |
1311 | { | |
1312 | return atomic_long_read(prot->memory_allocated); | |
1313 | } | |
1314 | ||
1315 | static inline bool | |
1316 | proto_memory_pressure(struct proto *prot) | |
1317 | { | |
1318 | if (!prot->memory_pressure) | |
1319 | return false; | |
1320 | return !!*prot->memory_pressure; | |
1321 | } | |
1322 | ||
65f76517 ED |
1323 | |
1324 | #ifdef CONFIG_PROC_FS | |
1da177e4 | 1325 | /* Called with local bh disabled */ |
69336bd2 JP |
1326 | void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
1327 | int sock_prot_inuse_get(struct net *net, struct proto *proto); | |
648845ab | 1328 | int sock_inuse_get(struct net *net); |
65f76517 | 1329 | #else |
dc6b9b78 | 1330 | static inline void sock_prot_inuse_add(struct net *net, struct proto *prot, |
c29a0bc4 | 1331 | int inc) |
65f76517 ED |
1332 | { |
1333 | } | |
65f76517 ED |
1334 | #endif |
1335 | ||
1da177e4 | 1336 | |
614c6cb4 ACM |
1337 | /* With per-bucket locks this operation is not-atomic, so that |
1338 | * this version is not worse. | |
1339 | */ | |
086c653f | 1340 | static inline int __sk_prot_rehash(struct sock *sk) |
614c6cb4 ACM |
1341 | { |
1342 | sk->sk_prot->unhash(sk); | |
086c653f | 1343 | return sk->sk_prot->hash(sk); |
614c6cb4 ACM |
1344 | } |
1345 | ||
1da177e4 LT |
1346 | /* About 10 seconds */ |
1347 | #define SOCK_DESTROY_TIME (10*HZ) | |
1348 | ||
1349 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
1350 | #define PROT_SOCK 1024 | |
1351 | ||
1352 | #define SHUTDOWN_MASK 3 | |
1353 | #define RCV_SHUTDOWN 1 | |
1354 | #define SEND_SHUTDOWN 2 | |
1355 | ||
1356 | #define SOCK_SNDBUF_LOCK 1 | |
1357 | #define SOCK_RCVBUF_LOCK 2 | |
1358 | #define SOCK_BINDADDR_LOCK 4 | |
1359 | #define SOCK_BINDPORT_LOCK 8 | |
1360 | ||
1da177e4 LT |
1361 | struct socket_alloc { |
1362 | struct socket socket; | |
1363 | struct inode vfs_inode; | |
1364 | }; | |
1365 | ||
1366 | static inline struct socket *SOCKET_I(struct inode *inode) | |
1367 | { | |
1368 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
1369 | } | |
1370 | ||
1371 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
1372 | { | |
1373 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
1374 | } | |
1375 | ||
3ab224be HA |
1376 | /* |
1377 | * Functions for memory accounting | |
1378 | */ | |
f8c3bf00 | 1379 | int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind); |
69336bd2 | 1380 | int __sk_mem_schedule(struct sock *sk, int size, int kind); |
f8c3bf00 | 1381 | void __sk_mem_reduce_allocated(struct sock *sk, int amount); |
1a24e04e | 1382 | void __sk_mem_reclaim(struct sock *sk, int amount); |
1da177e4 | 1383 | |
bd68a2a8 ED |
1384 | /* We used to have PAGE_SIZE here, but systems with 64KB pages |
1385 | * do not necessarily have 16x time more memory than 4KB ones. | |
1386 | */ | |
1387 | #define SK_MEM_QUANTUM 4096 | |
3ab224be HA |
1388 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) |
1389 | #define SK_MEM_SEND 0 | |
1390 | #define SK_MEM_RECV 1 | |
1da177e4 | 1391 | |
bd68a2a8 ED |
1392 | /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */ |
1393 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) | |
1394 | { | |
1395 | long val = sk->sk_prot->sysctl_mem[index]; | |
1396 | ||
1397 | #if PAGE_SIZE > SK_MEM_QUANTUM | |
1398 | val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT; | |
1399 | #elif PAGE_SIZE < SK_MEM_QUANTUM | |
1400 | val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT; | |
1401 | #endif | |
1402 | return val; | |
1403 | } | |
1404 | ||
3ab224be | 1405 | static inline int sk_mem_pages(int amt) |
1da177e4 | 1406 | { |
3ab224be | 1407 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
1da177e4 LT |
1408 | } |
1409 | ||
dc6b9b78 | 1410 | static inline bool sk_has_account(struct sock *sk) |
1da177e4 | 1411 | { |
3ab224be HA |
1412 | /* return true if protocol supports memory accounting */ |
1413 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
1414 | } |
1415 | ||
dc6b9b78 | 1416 | static inline bool sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 1417 | { |
3ab224be | 1418 | if (!sk_has_account(sk)) |
dc6b9b78 | 1419 | return true; |
3ab224be HA |
1420 | return size <= sk->sk_forward_alloc || |
1421 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
1da177e4 LT |
1422 | } |
1423 | ||
c76562b6 | 1424 | static inline bool |
35c448a8 | 1425 | sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size) |
d80d99d6 | 1426 | { |
3ab224be | 1427 | if (!sk_has_account(sk)) |
dc6b9b78 | 1428 | return true; |
c76562b6 MG |
1429 | return size<= sk->sk_forward_alloc || |
1430 | __sk_mem_schedule(sk, size, SK_MEM_RECV) || | |
1431 | skb_pfmemalloc(skb); | |
3ab224be HA |
1432 | } |
1433 | ||
1434 | static inline void sk_mem_reclaim(struct sock *sk) | |
1435 | { | |
1436 | if (!sk_has_account(sk)) | |
1437 | return; | |
1438 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
1a24e04e | 1439 | __sk_mem_reclaim(sk, sk->sk_forward_alloc); |
3ab224be HA |
1440 | } |
1441 | ||
9993e7d3 DM |
1442 | static inline void sk_mem_reclaim_partial(struct sock *sk) |
1443 | { | |
1444 | if (!sk_has_account(sk)) | |
1445 | return; | |
1446 | if (sk->sk_forward_alloc > SK_MEM_QUANTUM) | |
1a24e04e | 1447 | __sk_mem_reclaim(sk, sk->sk_forward_alloc - 1); |
9993e7d3 DM |
1448 | } |
1449 | ||
3ab224be HA |
1450 | static inline void sk_mem_charge(struct sock *sk, int size) |
1451 | { | |
1452 | if (!sk_has_account(sk)) | |
1453 | return; | |
1454 | sk->sk_forward_alloc -= size; | |
1455 | } | |
1456 | ||
1457 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
1458 | { | |
1459 | if (!sk_has_account(sk)) | |
1460 | return; | |
1461 | sk->sk_forward_alloc += size; | |
20c64d5c ED |
1462 | |
1463 | /* Avoid a possible overflow. | |
1464 | * TCP send queues can make this happen, if sk_mem_reclaim() | |
1465 | * is not called and more than 2 GBytes are released at once. | |
1466 | * | |
1467 | * If we reach 2 MBytes, reclaim 1 MBytes right now, there is | |
1468 | * no need to hold that much forward allocation anyway. | |
1469 | */ | |
1470 | if (unlikely(sk->sk_forward_alloc >= 1 << 21)) | |
1471 | __sk_mem_reclaim(sk, 1 << 20); | |
3ab224be HA |
1472 | } |
1473 | ||
0b7d7f6b | 1474 | DECLARE_STATIC_KEY_FALSE(tcp_tx_skb_cache_key); |
3ab224be HA |
1475 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) |
1476 | { | |
4f661542 | 1477 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
ab4e846a | 1478 | sk_wmem_queued_add(sk, -skb->truesize); |
4f661542 | 1479 | sk_mem_uncharge(sk, skb->truesize); |
0b7d7f6b ED |
1480 | if (static_branch_unlikely(&tcp_tx_skb_cache_key) && |
1481 | !sk->sk_tx_skb_cache && !skb_cloned(skb)) { | |
4f661542 | 1482 | skb_zcopy_clear(skb, true); |
472c2e07 ED |
1483 | sk->sk_tx_skb_cache = skb; |
1484 | return; | |
1485 | } | |
3ab224be | 1486 | __kfree_skb(skb); |
d80d99d6 HX |
1487 | } |
1488 | ||
c3f9b018 ED |
1489 | static inline void sock_release_ownership(struct sock *sk) |
1490 | { | |
61881cfb HFS |
1491 | if (sk->sk_lock.owned) { |
1492 | sk->sk_lock.owned = 0; | |
1493 | ||
1494 | /* The sk_lock has mutex_unlock() semantics: */ | |
5facae4f | 1495 | mutex_release(&sk->sk_lock.dep_map, _RET_IP_); |
61881cfb | 1496 | } |
c3f9b018 ED |
1497 | } |
1498 | ||
ed07536e PZ |
1499 | /* |
1500 | * Macro so as to not evaluate some arguments when | |
1501 | * lockdep is not enabled. | |
1502 | * | |
1503 | * Mark both the sk_lock and the sk_lock.slock as a | |
1504 | * per-address-family lock class. | |
1505 | */ | |
dc6b9b78 | 1506 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
ed07536e | 1507 | do { \ |
e8f6fbf6 | 1508 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
1509 | init_waitqueue_head(&sk->sk_lock.wq); \ |
1510 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
1511 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
1512 | sizeof((sk)->sk_lock)); \ | |
1513 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
dc6b9b78 | 1514 | (skey), (sname)); \ |
ed07536e PZ |
1515 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
1516 | } while (0) | |
1517 | ||
b33b0a1b | 1518 | #ifdef CONFIG_LOCKDEP |
05b93801 | 1519 | static inline bool lockdep_sock_is_held(const struct sock *sk) |
1e1d04e6 | 1520 | { |
1e1d04e6 HFS |
1521 | return lockdep_is_held(&sk->sk_lock) || |
1522 | lockdep_is_held(&sk->sk_lock.slock); | |
1523 | } | |
b33b0a1b | 1524 | #endif |
1e1d04e6 | 1525 | |
69336bd2 | 1526 | void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
1527 | |
1528 | static inline void lock_sock(struct sock *sk) | |
1529 | { | |
1530 | lock_sock_nested(sk, 0); | |
1531 | } | |
1532 | ||
8873c064 | 1533 | void __release_sock(struct sock *sk); |
69336bd2 | 1534 | void release_sock(struct sock *sk); |
1da177e4 LT |
1535 | |
1536 | /* BH context may only use the following locking interface. */ | |
1537 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
1538 | #define bh_lock_sock_nested(__sk) \ |
1539 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
1540 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
1541 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
1542 | ||
69336bd2 | 1543 | bool lock_sock_fast(struct sock *sk); |
8a74ad60 ED |
1544 | /** |
1545 | * unlock_sock_fast - complement of lock_sock_fast | |
1546 | * @sk: socket | |
1547 | * @slow: slow mode | |
1548 | * | |
1549 | * fast unlock socket for user context. | |
1550 | * If slow mode is on, we call regular release_sock() | |
1551 | */ | |
1552 | static inline void unlock_sock_fast(struct sock *sk, bool slow) | |
4b0b72f7 | 1553 | { |
8a74ad60 ED |
1554 | if (slow) |
1555 | release_sock(sk); | |
1556 | else | |
1557 | spin_unlock_bh(&sk->sk_lock.slock); | |
4b0b72f7 ED |
1558 | } |
1559 | ||
fafc4e1e HFS |
1560 | /* Used by processes to "lock" a socket state, so that |
1561 | * interrupts and bottom half handlers won't change it | |
1562 | * from under us. It essentially blocks any incoming | |
1563 | * packets, so that we won't get any new data or any | |
1564 | * packets that change the state of the socket. | |
1565 | * | |
1566 | * While locked, BH processing will add new packets to | |
1567 | * the backlog queue. This queue is processed by the | |
1568 | * owner of the socket lock right before it is released. | |
1569 | * | |
1570 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
1571 | * accesses from user process context. | |
1572 | */ | |
1573 | ||
46cc6e49 | 1574 | static inline void sock_owned_by_me(const struct sock *sk) |
fafc4e1e HFS |
1575 | { |
1576 | #ifdef CONFIG_LOCKDEP | |
5e91f6ce | 1577 | WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks); |
fafc4e1e | 1578 | #endif |
46cc6e49 ED |
1579 | } |
1580 | ||
1581 | static inline bool sock_owned_by_user(const struct sock *sk) | |
1582 | { | |
1583 | sock_owned_by_me(sk); | |
fafc4e1e HFS |
1584 | return sk->sk_lock.owned; |
1585 | } | |
1586 | ||
602f7a27 TH |
1587 | static inline bool sock_owned_by_user_nocheck(const struct sock *sk) |
1588 | { | |
1589 | return sk->sk_lock.owned; | |
1590 | } | |
1591 | ||
fafc4e1e HFS |
1592 | /* no reclassification while locks are held */ |
1593 | static inline bool sock_allow_reclassification(const struct sock *csk) | |
1594 | { | |
1595 | struct sock *sk = (struct sock *)csk; | |
1596 | ||
1597 | return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock); | |
1598 | } | |
4b0b72f7 | 1599 | |
69336bd2 | 1600 | struct sock *sk_alloc(struct net *net, int family, gfp_t priority, |
11aa9c28 | 1601 | struct proto *prot, int kern); |
69336bd2 | 1602 | void sk_free(struct sock *sk); |
eb4cb008 | 1603 | void sk_destruct(struct sock *sk); |
69336bd2 | 1604 | struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority); |
94352d45 | 1605 | void sk_free_unlock_clone(struct sock *sk); |
69336bd2 JP |
1606 | |
1607 | struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force, | |
1608 | gfp_t priority); | |
1d2077ac | 1609 | void __sock_wfree(struct sk_buff *skb); |
69336bd2 | 1610 | void sock_wfree(struct sk_buff *skb); |
98ba0bd5 WB |
1611 | struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size, |
1612 | gfp_t priority); | |
69336bd2 JP |
1613 | void skb_orphan_partial(struct sk_buff *skb); |
1614 | void sock_rfree(struct sk_buff *skb); | |
62bccb8c | 1615 | void sock_efree(struct sk_buff *skb); |
82eabd9e | 1616 | #ifdef CONFIG_INET |
69336bd2 | 1617 | void sock_edemux(struct sk_buff *skb); |
82eabd9e | 1618 | #else |
158f323b | 1619 | #define sock_edemux sock_efree |
82eabd9e | 1620 | #endif |
69336bd2 JP |
1621 | |
1622 | int sock_setsockopt(struct socket *sock, int level, int op, | |
1623 | char __user *optval, unsigned int optlen); | |
1624 | ||
1625 | int sock_getsockopt(struct socket *sock, int level, int op, | |
1626 | char __user *optval, int __user *optlen); | |
c7cbdbf2 AB |
1627 | int sock_gettstamp(struct socket *sock, void __user *userstamp, |
1628 | bool timeval, bool time32); | |
69336bd2 JP |
1629 | struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size, |
1630 | int noblock, int *errcode); | |
1631 | struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, | |
1632 | unsigned long data_len, int noblock, | |
1633 | int *errcode, int max_page_order); | |
1634 | void *sock_kmalloc(struct sock *sk, int size, gfp_t priority); | |
1635 | void sock_kfree_s(struct sock *sk, void *mem, int size); | |
79e88659 | 1636 | void sock_kzfree_s(struct sock *sk, void *mem, int size); |
69336bd2 | 1637 | void sk_send_sigurg(struct sock *sk); |
1da177e4 | 1638 | |
f28ea365 | 1639 | struct sockcm_cookie { |
80b14dee | 1640 | u64 transmit_time; |
f28ea365 | 1641 | u32 mark; |
3dd17e63 | 1642 | u16 tsflags; |
f28ea365 EJ |
1643 | }; |
1644 | ||
657a0667 WB |
1645 | static inline void sockcm_init(struct sockcm_cookie *sockc, |
1646 | const struct sock *sk) | |
1647 | { | |
1648 | *sockc = (struct sockcm_cookie) { .tsflags = sk->sk_tsflags }; | |
1649 | } | |
1650 | ||
39771b12 WB |
1651 | int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg, |
1652 | struct sockcm_cookie *sockc); | |
f28ea365 EJ |
1653 | int sock_cmsg_send(struct sock *sk, struct msghdr *msg, |
1654 | struct sockcm_cookie *sockc); | |
1655 | ||
1da177e4 LT |
1656 | /* |
1657 | * Functions to fill in entries in struct proto_ops when a protocol | |
1658 | * does not implement a particular function. | |
1659 | */ | |
69336bd2 JP |
1660 | int sock_no_bind(struct socket *, struct sockaddr *, int); |
1661 | int sock_no_connect(struct socket *, struct sockaddr *, int, int); | |
1662 | int sock_no_socketpair(struct socket *, struct socket *); | |
cdfbabfb | 1663 | int sock_no_accept(struct socket *, struct socket *, int, bool); |
9b2c45d4 | 1664 | int sock_no_getname(struct socket *, struct sockaddr *, int); |
69336bd2 JP |
1665 | int sock_no_ioctl(struct socket *, unsigned int, unsigned long); |
1666 | int sock_no_listen(struct socket *, int); | |
1667 | int sock_no_shutdown(struct socket *, int); | |
1668 | int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *); | |
1669 | int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int); | |
1b784140 | 1670 | int sock_no_sendmsg(struct socket *, struct msghdr *, size_t); |
306b13eb | 1671 | int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len); |
1b784140 | 1672 | int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int); |
69336bd2 JP |
1673 | int sock_no_mmap(struct file *file, struct socket *sock, |
1674 | struct vm_area_struct *vma); | |
1675 | ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, | |
1676 | size_t size, int flags); | |
306b13eb TH |
1677 | ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page, |
1678 | int offset, size_t size, int flags); | |
1da177e4 LT |
1679 | |
1680 | /* | |
1681 | * Functions to fill in entries in struct proto_ops when a protocol | |
1682 | * uses the inet style. | |
1683 | */ | |
69336bd2 | 1684 | int sock_common_getsockopt(struct socket *sock, int level, int optname, |
1da177e4 | 1685 | char __user *optval, int __user *optlen); |
1b784140 YX |
1686 | int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
1687 | int flags); | |
69336bd2 | 1688 | int sock_common_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1689 | char __user *optval, unsigned int optlen); |
69336bd2 | 1690 | int compat_sock_common_getsockopt(struct socket *sock, int level, |
3fdadf7d | 1691 | int optname, char __user *optval, int __user *optlen); |
69336bd2 | 1692 | int compat_sock_common_setsockopt(struct socket *sock, int level, |
b7058842 | 1693 | int optname, char __user *optval, unsigned int optlen); |
1da177e4 | 1694 | |
69336bd2 | 1695 | void sk_common_release(struct sock *sk); |
1da177e4 LT |
1696 | |
1697 | /* | |
1698 | * Default socket callbacks and setup code | |
1699 | */ | |
dc6b9b78 | 1700 | |
1da177e4 | 1701 | /* Initialise core socket variables */ |
69336bd2 | 1702 | void sock_init_data(struct socket *sock, struct sock *sk); |
1da177e4 | 1703 | |
1da177e4 LT |
1704 | /* |
1705 | * Socket reference counting postulates. | |
1706 | * | |
1707 | * * Each user of socket SHOULD hold a reference count. | |
1708 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1709 | * running timer, skb in flight MUST hold a reference count. | |
1710 | * * When reference count hits 0, it means it will never increase back. | |
1711 | * * When reference count hits 0, it means that no references from | |
1712 | * outside exist to this socket and current process on current CPU | |
1713 | * is last user and may/should destroy this socket. | |
1714 | * * sk_free is called from any context: process, BH, IRQ. When | |
1715 | * it is called, socket has no references from outside -> sk_free | |
1716 | * may release descendant resources allocated by the socket, but | |
1717 | * to the time when it is called, socket is NOT referenced by any | |
1718 | * hash tables, lists etc. | |
1719 | * * Packets, delivered from outside (from network or from another process) | |
1720 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1721 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1722 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1723 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1724 | * (leak to backlog). Packet socket does all the processing inside | |
1725 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1726 | * use separate SMP lock, so that they are prone too. | |
1727 | */ | |
1728 | ||
1729 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1730 | static inline void sock_put(struct sock *sk) | |
1731 | { | |
41c6d650 | 1732 | if (refcount_dec_and_test(&sk->sk_refcnt)) |
1da177e4 LT |
1733 | sk_free(sk); |
1734 | } | |
05dbc7b5 | 1735 | /* Generic version of sock_put(), dealing with all sockets |
41b822c5 | 1736 | * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...) |
05dbc7b5 ED |
1737 | */ |
1738 | void sock_gen_put(struct sock *sk); | |
1da177e4 | 1739 | |
4f0c40d9 | 1740 | int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested, |
c3f24cfb | 1741 | unsigned int trim_cap, bool refcounted); |
4f0c40d9 WB |
1742 | static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
1743 | const int nested) | |
1744 | { | |
c3f24cfb | 1745 | return __sk_receive_skb(sk, skb, nested, 1, true); |
4f0c40d9 | 1746 | } |
25995ff5 | 1747 | |
e022f0b4 KK |
1748 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
1749 | { | |
755c31cd AN |
1750 | /* sk_tx_queue_mapping accept only upto a 16-bit value */ |
1751 | if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX)) | |
1752 | return; | |
e022f0b4 KK |
1753 | sk->sk_tx_queue_mapping = tx_queue; |
1754 | } | |
1755 | ||
755c31cd AN |
1756 | #define NO_QUEUE_MAPPING USHRT_MAX |
1757 | ||
e022f0b4 KK |
1758 | static inline void sk_tx_queue_clear(struct sock *sk) |
1759 | { | |
755c31cd | 1760 | sk->sk_tx_queue_mapping = NO_QUEUE_MAPPING; |
e022f0b4 KK |
1761 | } |
1762 | ||
1763 | static inline int sk_tx_queue_get(const struct sock *sk) | |
1764 | { | |
755c31cd AN |
1765 | if (sk && sk->sk_tx_queue_mapping != NO_QUEUE_MAPPING) |
1766 | return sk->sk_tx_queue_mapping; | |
1767 | ||
1768 | return -1; | |
e022f0b4 KK |
1769 | } |
1770 | ||
c6345ce7 AN |
1771 | static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb) |
1772 | { | |
1773 | #ifdef CONFIG_XPS | |
1774 | if (skb_rx_queue_recorded(skb)) { | |
1775 | u16 rx_queue = skb_get_rx_queue(skb); | |
1776 | ||
1777 | if (WARN_ON_ONCE(rx_queue == NO_QUEUE_MAPPING)) | |
1778 | return; | |
1779 | ||
1780 | sk->sk_rx_queue_mapping = rx_queue; | |
1781 | } | |
1782 | #endif | |
1783 | } | |
1784 | ||
1785 | static inline void sk_rx_queue_clear(struct sock *sk) | |
1786 | { | |
1787 | #ifdef CONFIG_XPS | |
1788 | sk->sk_rx_queue_mapping = NO_QUEUE_MAPPING; | |
1789 | #endif | |
1790 | } | |
1791 | ||
fc9bab24 AN |
1792 | #ifdef CONFIG_XPS |
1793 | static inline int sk_rx_queue_get(const struct sock *sk) | |
1794 | { | |
1795 | if (sk && sk->sk_rx_queue_mapping != NO_QUEUE_MAPPING) | |
1796 | return sk->sk_rx_queue_mapping; | |
1797 | ||
1798 | return -1; | |
1799 | } | |
1800 | #endif | |
1801 | ||
972692e0 DM |
1802 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
1803 | { | |
e022f0b4 | 1804 | sk_tx_queue_clear(sk); |
972692e0 DM |
1805 | sk->sk_socket = sock; |
1806 | } | |
1807 | ||
aa395145 ED |
1808 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
1809 | { | |
eaefd110 ED |
1810 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
1811 | return &rcu_dereference_raw(sk->sk_wq)->wait; | |
aa395145 | 1812 | } |
1da177e4 LT |
1813 | /* Detach socket from process context. |
1814 | * Announce socket dead, detach it from wait queue and inode. | |
1815 | * Note that parent inode held reference count on this struct sock, | |
1816 | * we do not release it in this function, because protocol | |
1817 | * probably wants some additional cleanups or even continuing | |
1818 | * to work with this socket (TCP). | |
1819 | */ | |
1820 | static inline void sock_orphan(struct sock *sk) | |
1821 | { | |
1822 | write_lock_bh(&sk->sk_callback_lock); | |
1823 | sock_set_flag(sk, SOCK_DEAD); | |
972692e0 | 1824 | sk_set_socket(sk, NULL); |
43815482 | 1825 | sk->sk_wq = NULL; |
1da177e4 LT |
1826 | write_unlock_bh(&sk->sk_callback_lock); |
1827 | } | |
1828 | ||
1829 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1830 | { | |
0ffdaf5b | 1831 | WARN_ON(parent->sk); |
1da177e4 | 1832 | write_lock_bh(&sk->sk_callback_lock); |
333f7909 | 1833 | rcu_assign_pointer(sk->sk_wq, &parent->wq); |
1da177e4 | 1834 | parent->sk = sk; |
972692e0 | 1835 | sk_set_socket(sk, parent); |
86741ec2 | 1836 | sk->sk_uid = SOCK_INODE(parent)->i_uid; |
4237c75c | 1837 | security_sock_graft(sk, parent); |
1da177e4 LT |
1838 | write_unlock_bh(&sk->sk_callback_lock); |
1839 | } | |
1840 | ||
69336bd2 JP |
1841 | kuid_t sock_i_uid(struct sock *sk); |
1842 | unsigned long sock_i_ino(struct sock *sk); | |
1da177e4 | 1843 | |
86741ec2 LC |
1844 | static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk) |
1845 | { | |
1846 | return sk ? sk->sk_uid : make_kuid(net->user_ns, 0); | |
1847 | } | |
1848 | ||
58d607d3 | 1849 | static inline u32 net_tx_rndhash(void) |
877d1f62 | 1850 | { |
58d607d3 ED |
1851 | u32 v = prandom_u32(); |
1852 | ||
1853 | return v ?: 1; | |
1854 | } | |
877d1f62 | 1855 | |
58d607d3 ED |
1856 | static inline void sk_set_txhash(struct sock *sk) |
1857 | { | |
1858 | sk->sk_txhash = net_tx_rndhash(); | |
877d1f62 TH |
1859 | } |
1860 | ||
265f94ff TH |
1861 | static inline void sk_rethink_txhash(struct sock *sk) |
1862 | { | |
1863 | if (sk->sk_txhash) | |
1864 | sk_set_txhash(sk); | |
1865 | } | |
1866 | ||
1da177e4 LT |
1867 | static inline struct dst_entry * |
1868 | __sk_dst_get(struct sock *sk) | |
1869 | { | |
1e1d04e6 HFS |
1870 | return rcu_dereference_check(sk->sk_dst_cache, |
1871 | lockdep_sock_is_held(sk)); | |
1da177e4 LT |
1872 | } |
1873 | ||
1874 | static inline struct dst_entry * | |
1875 | sk_dst_get(struct sock *sk) | |
1876 | { | |
1877 | struct dst_entry *dst; | |
1878 | ||
b6c6712a ED |
1879 | rcu_read_lock(); |
1880 | dst = rcu_dereference(sk->sk_dst_cache); | |
f8864972 ED |
1881 | if (dst && !atomic_inc_not_zero(&dst->__refcnt)) |
1882 | dst = NULL; | |
b6c6712a | 1883 | rcu_read_unlock(); |
1da177e4 LT |
1884 | return dst; |
1885 | } | |
1886 | ||
b6c6712a ED |
1887 | static inline void dst_negative_advice(struct sock *sk) |
1888 | { | |
1889 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); | |
1890 | ||
265f94ff TH |
1891 | sk_rethink_txhash(sk); |
1892 | ||
b6c6712a ED |
1893 | if (dst && dst->ops->negative_advice) { |
1894 | ndst = dst->ops->negative_advice(dst); | |
1895 | ||
1896 | if (ndst != dst) { | |
1897 | rcu_assign_pointer(sk->sk_dst_cache, ndst); | |
0a6957e7 | 1898 | sk_tx_queue_clear(sk); |
9b8805a3 | 1899 | sk->sk_dst_pending_confirm = 0; |
b6c6712a ED |
1900 | } |
1901 | } | |
1902 | } | |
1903 | ||
1da177e4 LT |
1904 | static inline void |
1905 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1906 | { | |
1907 | struct dst_entry *old_dst; | |
1908 | ||
e022f0b4 | 1909 | sk_tx_queue_clear(sk); |
9b8805a3 | 1910 | sk->sk_dst_pending_confirm = 0; |
95964c6d ED |
1911 | old_dst = rcu_dereference_protected(sk->sk_dst_cache, |
1912 | lockdep_sock_is_held(sk)); | |
b6c6712a | 1913 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
1da177e4 LT |
1914 | dst_release(old_dst); |
1915 | } | |
1916 | ||
1917 | static inline void | |
1918 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1919 | { | |
7f502361 ED |
1920 | struct dst_entry *old_dst; |
1921 | ||
1922 | sk_tx_queue_clear(sk); | |
9b8805a3 | 1923 | sk->sk_dst_pending_confirm = 0; |
5925a055 | 1924 | old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst); |
7f502361 | 1925 | dst_release(old_dst); |
1da177e4 LT |
1926 | } |
1927 | ||
1928 | static inline void | |
1929 | __sk_dst_reset(struct sock *sk) | |
1930 | { | |
b6c6712a | 1931 | __sk_dst_set(sk, NULL); |
1da177e4 LT |
1932 | } |
1933 | ||
1934 | static inline void | |
1935 | sk_dst_reset(struct sock *sk) | |
1936 | { | |
7f502361 | 1937 | sk_dst_set(sk, NULL); |
1da177e4 LT |
1938 | } |
1939 | ||
69336bd2 | 1940 | struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1941 | |
69336bd2 | 1942 | struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1943 | |
9b8805a3 JA |
1944 | static inline void sk_dst_confirm(struct sock *sk) |
1945 | { | |
25c7a6d1 ED |
1946 | if (!READ_ONCE(sk->sk_dst_pending_confirm)) |
1947 | WRITE_ONCE(sk->sk_dst_pending_confirm, 1); | |
9b8805a3 JA |
1948 | } |
1949 | ||
4ff06203 JA |
1950 | static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n) |
1951 | { | |
1952 | if (skb_get_dst_pending_confirm(skb)) { | |
1953 | struct sock *sk = skb->sk; | |
1954 | unsigned long now = jiffies; | |
1955 | ||
1956 | /* avoid dirtying neighbour */ | |
25c7a6d1 ED |
1957 | if (READ_ONCE(n->confirmed) != now) |
1958 | WRITE_ONCE(n->confirmed, now); | |
1959 | if (sk && READ_ONCE(sk->sk_dst_pending_confirm)) | |
1960 | WRITE_ONCE(sk->sk_dst_pending_confirm, 0); | |
4ff06203 JA |
1961 | } |
1962 | } | |
1963 | ||
f60e5990 | 1964 | bool sk_mc_loop(struct sock *sk); |
1965 | ||
dc6b9b78 | 1966 | static inline bool sk_can_gso(const struct sock *sk) |
bcd76111 HX |
1967 | { |
1968 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1969 | } | |
1970 | ||
69336bd2 | 1971 | void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 1972 | |
c8f44aff | 1973 | static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) |
a465419b ED |
1974 | { |
1975 | sk->sk_route_nocaps |= flags; | |
1976 | sk->sk_route_caps &= ~flags; | |
1977 | } | |
1978 | ||
c6e1a0d1 | 1979 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
57be5bda | 1980 | struct iov_iter *from, char *to, |
912d398d | 1981 | int copy, int offset) |
c6e1a0d1 TH |
1982 | { |
1983 | if (skb->ip_summed == CHECKSUM_NONE) { | |
57be5bda | 1984 | __wsum csum = 0; |
15e6cb46 | 1985 | if (!csum_and_copy_from_iter_full(to, copy, &csum, from)) |
57be5bda | 1986 | return -EFAULT; |
912d398d | 1987 | skb->csum = csum_block_add(skb->csum, csum, offset); |
c6e1a0d1 | 1988 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
15e6cb46 | 1989 | if (!copy_from_iter_full_nocache(to, copy, from)) |
c6e1a0d1 | 1990 | return -EFAULT; |
15e6cb46 | 1991 | } else if (!copy_from_iter_full(to, copy, from)) |
c6e1a0d1 TH |
1992 | return -EFAULT; |
1993 | ||
1994 | return 0; | |
1995 | } | |
1996 | ||
1997 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, | |
57be5bda | 1998 | struct iov_iter *from, int copy) |
c6e1a0d1 | 1999 | { |
912d398d | 2000 | int err, offset = skb->len; |
c6e1a0d1 | 2001 | |
912d398d WY |
2002 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
2003 | copy, offset); | |
c6e1a0d1 | 2004 | if (err) |
912d398d | 2005 | __skb_trim(skb, offset); |
c6e1a0d1 TH |
2006 | |
2007 | return err; | |
2008 | } | |
2009 | ||
57be5bda | 2010 | static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from, |
c6e1a0d1 TH |
2011 | struct sk_buff *skb, |
2012 | struct page *page, | |
2013 | int off, int copy) | |
2014 | { | |
2015 | int err; | |
2016 | ||
912d398d WY |
2017 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
2018 | copy, skb->len); | |
c6e1a0d1 TH |
2019 | if (err) |
2020 | return err; | |
2021 | ||
2022 | skb->len += copy; | |
2023 | skb->data_len += copy; | |
2024 | skb->truesize += copy; | |
ab4e846a | 2025 | sk_wmem_queued_add(sk, copy); |
c6e1a0d1 TH |
2026 | sk_mem_charge(sk, copy); |
2027 | return 0; | |
2028 | } | |
2029 | ||
c564039f ED |
2030 | /** |
2031 | * sk_wmem_alloc_get - returns write allocations | |
2032 | * @sk: socket | |
2033 | * | |
2034 | * Returns sk_wmem_alloc minus initial offset of one | |
2035 | */ | |
2036 | static inline int sk_wmem_alloc_get(const struct sock *sk) | |
2037 | { | |
14afee4b | 2038 | return refcount_read(&sk->sk_wmem_alloc) - 1; |
c564039f ED |
2039 | } |
2040 | ||
2041 | /** | |
2042 | * sk_rmem_alloc_get - returns read allocations | |
2043 | * @sk: socket | |
2044 | * | |
2045 | * Returns sk_rmem_alloc | |
2046 | */ | |
2047 | static inline int sk_rmem_alloc_get(const struct sock *sk) | |
2048 | { | |
2049 | return atomic_read(&sk->sk_rmem_alloc); | |
2050 | } | |
2051 | ||
2052 | /** | |
2053 | * sk_has_allocations - check if allocations are outstanding | |
2054 | * @sk: socket | |
2055 | * | |
2056 | * Returns true if socket has write or read allocations | |
2057 | */ | |
dc6b9b78 | 2058 | static inline bool sk_has_allocations(const struct sock *sk) |
c564039f ED |
2059 | { |
2060 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); | |
2061 | } | |
2062 | ||
a57de0b4 | 2063 | /** |
1ce0bf50 | 2064 | * skwq_has_sleeper - check if there are any waiting processes |
acfbe96a | 2065 | * @wq: struct socket_wq |
a57de0b4 | 2066 | * |
43815482 | 2067 | * Returns true if socket_wq has waiting processes |
a57de0b4 | 2068 | * |
1ce0bf50 | 2069 | * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory |
a57de0b4 JO |
2070 | * barrier call. They were added due to the race found within the tcp code. |
2071 | * | |
d651983d | 2072 | * Consider following tcp code paths:: |
a57de0b4 | 2073 | * |
d651983d MCC |
2074 | * CPU1 CPU2 |
2075 | * sys_select receive packet | |
a57de0b4 JO |
2076 | * ... ... |
2077 | * __add_wait_queue update tp->rcv_nxt | |
2078 | * ... ... | |
2079 | * tp->rcv_nxt check sock_def_readable | |
2080 | * ... { | |
43815482 ED |
2081 | * schedule rcu_read_lock(); |
2082 | * wq = rcu_dereference(sk->sk_wq); | |
2083 | * if (wq && waitqueue_active(&wq->wait)) | |
2084 | * wake_up_interruptible(&wq->wait) | |
a57de0b4 JO |
2085 | * ... |
2086 | * } | |
2087 | * | |
2088 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay | |
2089 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 | |
2090 | * could then endup calling schedule and sleep forever if there are no more | |
2091 | * data on the socket. | |
ad462769 | 2092 | * |
a57de0b4 | 2093 | */ |
1ce0bf50 | 2094 | static inline bool skwq_has_sleeper(struct socket_wq *wq) |
a57de0b4 | 2095 | { |
1ce0bf50 | 2096 | return wq && wq_has_sleeper(&wq->wait); |
a57de0b4 JO |
2097 | } |
2098 | ||
2099 | /** | |
2100 | * sock_poll_wait - place memory barrier behind the poll_wait call. | |
2101 | * @filp: file | |
89ab066d | 2102 | * @sock: socket to wait on |
a57de0b4 JO |
2103 | * @p: poll_table |
2104 | * | |
43815482 | 2105 | * See the comments in the wq_has_sleeper function. |
a57de0b4 | 2106 | */ |
89ab066d KG |
2107 | static inline void sock_poll_wait(struct file *filp, struct socket *sock, |
2108 | poll_table *p) | |
a57de0b4 | 2109 | { |
d8bbd13b | 2110 | if (!poll_does_not_wait(p)) { |
333f7909 | 2111 | poll_wait(filp, &sock->wq.wait, p); |
dc6b9b78 | 2112 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
2113 | * socket flags modification. |
2114 | * | |
43815482 | 2115 | * This memory barrier is paired in the wq_has_sleeper. |
dc6b9b78 | 2116 | */ |
a57de0b4 JO |
2117 | smp_mb(); |
2118 | } | |
2119 | } | |
2120 | ||
b73c3d0e TH |
2121 | static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk) |
2122 | { | |
2123 | if (sk->sk_txhash) { | |
2124 | skb->l4_hash = 1; | |
2125 | skb->hash = sk->sk_txhash; | |
2126 | } | |
2127 | } | |
2128 | ||
9e17f8a4 ED |
2129 | void skb_set_owner_w(struct sk_buff *skb, struct sock *sk); |
2130 | ||
1da177e4 | 2131 | /* |
dc6b9b78 | 2132 | * Queue a received datagram if it will fit. Stream and sequenced |
1da177e4 LT |
2133 | * protocols can't normally use this as they need to fit buffers in |
2134 | * and play with them. | |
2135 | * | |
dc6b9b78 | 2136 | * Inlined as it's very short and called for pretty much every |
1da177e4 LT |
2137 | * packet ever received. |
2138 | */ | |
1da177e4 LT |
2139 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
2140 | { | |
d55d87fd | 2141 | skb_orphan(skb); |
1da177e4 LT |
2142 | skb->sk = sk; |
2143 | skb->destructor = sock_rfree; | |
2144 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 2145 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
2146 | } |
2147 | ||
69336bd2 JP |
2148 | void sk_reset_timer(struct sock *sk, struct timer_list *timer, |
2149 | unsigned long expires); | |
1da177e4 | 2150 | |
69336bd2 | 2151 | void sk_stop_timer(struct sock *sk, struct timer_list *timer); |
1da177e4 | 2152 | |
65101aec PA |
2153 | int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue, |
2154 | struct sk_buff *skb, unsigned int flags, | |
69629464 ED |
2155 | void (*destructor)(struct sock *sk, |
2156 | struct sk_buff *skb)); | |
e6afc8ac | 2157 | int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
69336bd2 | 2158 | int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 | 2159 | |
69336bd2 | 2160 | int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
364a9e93 | 2161 | struct sk_buff *sock_dequeue_err_skb(struct sock *sk); |
1da177e4 LT |
2162 | |
2163 | /* | |
2164 | * Recover an error report and clear atomically | |
2165 | */ | |
dc6b9b78 | 2166 | |
1da177e4 LT |
2167 | static inline int sock_error(struct sock *sk) |
2168 | { | |
c1cbe4b7 BL |
2169 | int err; |
2170 | if (likely(!sk->sk_err)) | |
2171 | return 0; | |
2172 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
2173 | return -err; |
2174 | } | |
2175 | ||
2176 | static inline unsigned long sock_wspace(struct sock *sk) | |
2177 | { | |
2178 | int amt = 0; | |
2179 | ||
2180 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
14afee4b | 2181 | amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc); |
dc6b9b78 | 2182 | if (amt < 0) |
1da177e4 LT |
2183 | amt = 0; |
2184 | } | |
2185 | return amt; | |
2186 | } | |
2187 | ||
ceb5d58b ED |
2188 | /* Note: |
2189 | * We use sk->sk_wq_raw, from contexts knowing this | |
2190 | * pointer is not NULL and cannot disappear/change. | |
2191 | */ | |
9cd3e072 | 2192 | static inline void sk_set_bit(int nr, struct sock *sk) |
1da177e4 | 2193 | { |
4be73522 ED |
2194 | if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) && |
2195 | !sock_flag(sk, SOCK_FASYNC)) | |
9317bb69 ED |
2196 | return; |
2197 | ||
ceb5d58b | 2198 | set_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
2199 | } |
2200 | ||
2201 | static inline void sk_clear_bit(int nr, struct sock *sk) | |
2202 | { | |
4be73522 ED |
2203 | if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) && |
2204 | !sock_flag(sk, SOCK_FASYNC)) | |
9317bb69 ED |
2205 | return; |
2206 | ||
ceb5d58b | 2207 | clear_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
2208 | } |
2209 | ||
ceb5d58b | 2210 | static inline void sk_wake_async(const struct sock *sk, int how, int band) |
1da177e4 | 2211 | { |
ceb5d58b ED |
2212 | if (sock_flag(sk, SOCK_FASYNC)) { |
2213 | rcu_read_lock(); | |
2214 | sock_wake_async(rcu_dereference(sk->sk_wq), how, band); | |
2215 | rcu_read_unlock(); | |
2216 | } | |
1da177e4 LT |
2217 | } |
2218 | ||
eea86af6 DB |
2219 | /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might |
2220 | * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak. | |
2221 | * Note: for send buffers, TCP works better if we can build two skbs at | |
2222 | * minimum. | |
7a91b434 | 2223 | */ |
9eb5bf83 | 2224 | #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff))) |
eea86af6 DB |
2225 | |
2226 | #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2) | |
2227 | #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE | |
1da177e4 LT |
2228 | |
2229 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
2230 | { | |
e292f05e ED |
2231 | u32 val; |
2232 | ||
2233 | if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) | |
2234 | return; | |
2235 | ||
2236 | val = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); | |
2237 | ||
2238 | WRITE_ONCE(sk->sk_sndbuf, max_t(u32, val, SOCK_MIN_SNDBUF)); | |
1da177e4 LT |
2239 | } |
2240 | ||
eb934478 ED |
2241 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp, |
2242 | bool force_schedule); | |
1da177e4 | 2243 | |
5640f768 ED |
2244 | /** |
2245 | * sk_page_frag - return an appropriate page_frag | |
2246 | * @sk: socket | |
2247 | * | |
20eb4f29 TH |
2248 | * Use the per task page_frag instead of the per socket one for |
2249 | * optimization when we know that we're in the normal context and owns | |
2250 | * everything that's associated with %current. | |
2251 | * | |
2252 | * gfpflags_allow_blocking() isn't enough here as direct reclaim may nest | |
2253 | * inside other socket operations and end up recursing into sk_page_frag() | |
2254 | * while it's already in use. | |
5640f768 ED |
2255 | */ |
2256 | static inline struct page_frag *sk_page_frag(struct sock *sk) | |
1da177e4 | 2257 | { |
20eb4f29 | 2258 | if (gfpflags_normal_context(sk->sk_allocation)) |
5640f768 | 2259 | return ¤t->task_frag; |
1da177e4 | 2260 | |
5640f768 | 2261 | return &sk->sk_frag; |
1da177e4 LT |
2262 | } |
2263 | ||
69336bd2 | 2264 | bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag); |
5640f768 | 2265 | |
1da177e4 LT |
2266 | /* |
2267 | * Default write policy as shown to user space via poll/select/SIGIO | |
2268 | */ | |
dc6b9b78 | 2269 | static inline bool sock_writeable(const struct sock *sk) |
1da177e4 | 2270 | { |
e292f05e | 2271 | return refcount_read(&sk->sk_wmem_alloc) < (READ_ONCE(sk->sk_sndbuf) >> 1); |
1da177e4 LT |
2272 | } |
2273 | ||
dd0fc66f | 2274 | static inline gfp_t gfp_any(void) |
1da177e4 | 2275 | { |
99709372 | 2276 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
2277 | } |
2278 | ||
dc6b9b78 | 2279 | static inline long sock_rcvtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2280 | { |
2281 | return noblock ? 0 : sk->sk_rcvtimeo; | |
2282 | } | |
2283 | ||
dc6b9b78 | 2284 | static inline long sock_sndtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2285 | { |
2286 | return noblock ? 0 : sk->sk_sndtimeo; | |
2287 | } | |
2288 | ||
2289 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
2290 | { | |
eac66402 ED |
2291 | int v = waitall ? len : min_t(int, READ_ONCE(sk->sk_rcvlowat), len); |
2292 | ||
2293 | return v ?: 1; | |
1da177e4 LT |
2294 | } |
2295 | ||
2296 | /* Alas, with timeout socket operations are not restartable. | |
2297 | * Compare this to poll(). | |
2298 | */ | |
2299 | static inline int sock_intr_errno(long timeo) | |
2300 | { | |
2301 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
2302 | } | |
2303 | ||
744d5a3e EB |
2304 | struct sock_skb_cb { |
2305 | u32 dropcount; | |
2306 | }; | |
2307 | ||
2308 | /* Store sock_skb_cb at the end of skb->cb[] so protocol families | |
2309 | * using skb->cb[] would keep using it directly and utilize its | |
2310 | * alignement guarantee. | |
2311 | */ | |
c593642c | 2312 | #define SOCK_SKB_CB_OFFSET ((sizeof_field(struct sk_buff, cb) - \ |
744d5a3e EB |
2313 | sizeof(struct sock_skb_cb))) |
2314 | ||
2315 | #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \ | |
2316 | SOCK_SKB_CB_OFFSET)) | |
2317 | ||
b4772ef8 | 2318 | #define sock_skb_cb_check_size(size) \ |
744d5a3e | 2319 | BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET) |
b4772ef8 | 2320 | |
3bc3b96f EB |
2321 | static inline void |
2322 | sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb) | |
2323 | { | |
3665f381 ED |
2324 | SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ? |
2325 | atomic_read(&sk->sk_drops) : 0; | |
3bc3b96f EB |
2326 | } |
2327 | ||
532182cd ED |
2328 | static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb) |
2329 | { | |
2330 | int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs); | |
2331 | ||
2332 | atomic_add(segs, &sk->sk_drops); | |
2333 | } | |
2334 | ||
3a0ed3e9 DD |
2335 | static inline ktime_t sock_read_timestamp(struct sock *sk) |
2336 | { | |
2337 | #if BITS_PER_LONG==32 | |
2338 | unsigned int seq; | |
2339 | ktime_t kt; | |
2340 | ||
2341 | do { | |
2342 | seq = read_seqbegin(&sk->sk_stamp_seq); | |
2343 | kt = sk->sk_stamp; | |
2344 | } while (read_seqretry(&sk->sk_stamp_seq, seq)); | |
2345 | ||
2346 | return kt; | |
2347 | #else | |
f75359f3 | 2348 | return READ_ONCE(sk->sk_stamp); |
3a0ed3e9 DD |
2349 | #endif |
2350 | } | |
2351 | ||
2352 | static inline void sock_write_timestamp(struct sock *sk, ktime_t kt) | |
2353 | { | |
2354 | #if BITS_PER_LONG==32 | |
2355 | write_seqlock(&sk->sk_stamp_seq); | |
2356 | sk->sk_stamp = kt; | |
2357 | write_sequnlock(&sk->sk_stamp_seq); | |
2358 | #else | |
f75359f3 | 2359 | WRITE_ONCE(sk->sk_stamp, kt); |
3a0ed3e9 DD |
2360 | #endif |
2361 | } | |
2362 | ||
69336bd2 JP |
2363 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
2364 | struct sk_buff *skb); | |
2365 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, | |
2366 | struct sk_buff *skb); | |
92f37fd2 | 2367 | |
dc6b9b78 | 2368 | static inline void |
1da177e4 LT |
2369 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
2370 | { | |
b7aa0bf7 | 2371 | ktime_t kt = skb->tstamp; |
20d49473 | 2372 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
a61bbcf2 | 2373 | |
20d49473 PO |
2374 | /* |
2375 | * generate control messages if | |
b9f40e21 | 2376 | * - receive time stamping in software requested |
20d49473 | 2377 | * - software time stamp available and wanted |
20d49473 | 2378 | * - hardware time stamps available and wanted |
20d49473 PO |
2379 | */ |
2380 | if (sock_flag(sk, SOCK_RCVTSTAMP) || | |
b9f40e21 | 2381 | (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) || |
2456e855 TG |
2382 | (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) || |
2383 | (hwtstamps->hwtstamp && | |
b9f40e21 | 2384 | (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE))) |
92f37fd2 ED |
2385 | __sock_recv_timestamp(msg, sk, skb); |
2386 | else | |
3a0ed3e9 | 2387 | sock_write_timestamp(sk, kt); |
6e3e939f JB |
2388 | |
2389 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) | |
2390 | __sock_recv_wifi_status(msg, sk, skb); | |
1da177e4 LT |
2391 | } |
2392 | ||
69336bd2 JP |
2393 | void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2394 | struct sk_buff *skb); | |
767dd033 | 2395 | |
6c7c98ba | 2396 | #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC) |
767dd033 ED |
2397 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2398 | struct sk_buff *skb) | |
2399 | { | |
2400 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ | |
b9f40e21 WB |
2401 | (1UL << SOCK_RCVTSTAMP)) |
2402 | #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \ | |
2403 | SOF_TIMESTAMPING_RAW_HARDWARE) | |
767dd033 | 2404 | |
b9f40e21 | 2405 | if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY) |
767dd033 | 2406 | __sock_recv_ts_and_drops(msg, sk, skb); |
d3fbff30 | 2407 | else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP))) |
3a0ed3e9 | 2408 | sock_write_timestamp(sk, skb->tstamp); |
6c7c98ba | 2409 | else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP)) |
3a0ed3e9 | 2410 | sock_write_timestamp(sk, 0); |
767dd033 | 2411 | } |
3b885787 | 2412 | |
c14ac945 | 2413 | void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags); |
67cc0d40 | 2414 | |
20d49473 | 2415 | /** |
8f932f76 | 2416 | * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped |
20d49473 | 2417 | * @sk: socket sending this packet |
c14ac945 | 2418 | * @tsflags: timestamping flags to use |
140c55d4 | 2419 | * @tx_flags: completed with instructions for time stamping |
8f932f76 | 2420 | * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno) |
140c55d4 | 2421 | * |
d651983d | 2422 | * Note: callers should take care of initial ``*tx_flags`` value (usually 0) |
20d49473 | 2423 | */ |
8f932f76 WB |
2424 | static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags, |
2425 | __u8 *tx_flags, __u32 *tskey) | |
67cc0d40 | 2426 | { |
8f932f76 | 2427 | if (unlikely(tsflags)) { |
c14ac945 | 2428 | __sock_tx_timestamp(tsflags, tx_flags); |
8f932f76 WB |
2429 | if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey && |
2430 | tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK) | |
2431 | *tskey = sk->sk_tskey++; | |
2432 | } | |
67cc0d40 WB |
2433 | if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS))) |
2434 | *tx_flags |= SKBTX_WIFI_STATUS; | |
2435 | } | |
20d49473 | 2436 | |
8f932f76 WB |
2437 | static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags, |
2438 | __u8 *tx_flags) | |
2439 | { | |
2440 | _sock_tx_timestamp(sk, tsflags, tx_flags, NULL); | |
2441 | } | |
2442 | ||
2443 | static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags) | |
2444 | { | |
2445 | _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags, | |
2446 | &skb_shinfo(skb)->tskey); | |
2447 | } | |
2448 | ||
1da177e4 LT |
2449 | /** |
2450 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
2451 | * @sk: socket to eat this skb from |
2452 | * @skb: socket buffer to eat | |
1da177e4 LT |
2453 | * |
2454 | * This routine must be called with interrupts disabled or with the socket | |
2455 | * locked so that the sk_buff queue operation is ok. | |
2456 | */ | |
ede61ca4 | 2457 | DECLARE_STATIC_KEY_FALSE(tcp_rx_skb_cache_key); |
7bced397 | 2458 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
2459 | { |
2460 | __skb_unlink(skb, &sk->sk_receive_queue); | |
ede61ca4 | 2461 | if (static_branch_unlikely(&tcp_rx_skb_cache_key) && |
8b27dae5 ED |
2462 | !sk->sk_rx_skb_cache) { |
2463 | sk->sk_rx_skb_cache = skb; | |
2464 | skb_orphan(skb); | |
2465 | return; | |
2466 | } | |
1da177e4 LT |
2467 | __kfree_skb(skb); |
2468 | } | |
2469 | ||
3b1e0a65 YH |
2470 | static inline |
2471 | struct net *sock_net(const struct sock *sk) | |
2472 | { | |
c2d9ba9b | 2473 | return read_pnet(&sk->sk_net); |
3b1e0a65 YH |
2474 | } |
2475 | ||
2476 | static inline | |
f5aa23fd | 2477 | void sock_net_set(struct sock *sk, struct net *net) |
3b1e0a65 | 2478 | { |
c2d9ba9b | 2479 | write_pnet(&sk->sk_net, net); |
3b1e0a65 YH |
2480 | } |
2481 | ||
23542618 KK |
2482 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) |
2483 | { | |
efc27f8c | 2484 | if (skb->sk) { |
23542618 KK |
2485 | struct sock *sk = skb->sk; |
2486 | ||
2487 | skb->destructor = NULL; | |
2488 | skb->sk = NULL; | |
2489 | return sk; | |
2490 | } | |
2491 | return NULL; | |
2492 | } | |
2493 | ||
1d0ab253 ED |
2494 | /* This helper checks if a socket is a full socket, |
2495 | * ie _not_ a timewait or request socket. | |
2496 | */ | |
2497 | static inline bool sk_fullsock(const struct sock *sk) | |
2498 | { | |
2499 | return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV); | |
2500 | } | |
2501 | ||
ebf4e808 IL |
2502 | /* Checks if this SKB belongs to an HW offloaded socket |
2503 | * and whether any SW fallbacks are required based on dev. | |
41477662 | 2504 | * Check decrypted mark in case skb_orphan() cleared socket. |
ebf4e808 IL |
2505 | */ |
2506 | static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb, | |
2507 | struct net_device *dev) | |
2508 | { | |
2509 | #ifdef CONFIG_SOCK_VALIDATE_XMIT | |
2510 | struct sock *sk = skb->sk; | |
2511 | ||
41477662 | 2512 | if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb) { |
ebf4e808 | 2513 | skb = sk->sk_validate_xmit_skb(sk, dev, skb); |
41477662 JK |
2514 | #ifdef CONFIG_TLS_DEVICE |
2515 | } else if (unlikely(skb->decrypted)) { | |
2516 | pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n"); | |
2517 | kfree_skb(skb); | |
2518 | skb = NULL; | |
2519 | #endif | |
2520 | } | |
ebf4e808 IL |
2521 | #endif |
2522 | ||
2523 | return skb; | |
2524 | } | |
2525 | ||
e446f9df ED |
2526 | /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV |
2527 | * SYNACK messages can be attached to either ones (depending on SYNCOOKIE) | |
2528 | */ | |
2529 | static inline bool sk_listener(const struct sock *sk) | |
2530 | { | |
2531 | return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV); | |
2532 | } | |
2533 | ||
193d357d | 2534 | void sock_enable_timestamp(struct sock *sk, enum sock_flags flag); |
69336bd2 JP |
2535 | int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level, |
2536 | int type); | |
1da177e4 | 2537 | |
a3b299da EB |
2538 | bool sk_ns_capable(const struct sock *sk, |
2539 | struct user_namespace *user_ns, int cap); | |
2540 | bool sk_capable(const struct sock *sk, int cap); | |
2541 | bool sk_net_capable(const struct sock *sk, int cap); | |
2542 | ||
a2d133b1 JH |
2543 | void sk_get_meminfo(const struct sock *sk, u32 *meminfo); |
2544 | ||
eaa72dc4 ED |
2545 | /* Take into consideration the size of the struct sk_buff overhead in the |
2546 | * determination of these values, since that is non-constant across | |
2547 | * platforms. This makes socket queueing behavior and performance | |
2548 | * not depend upon such differences. | |
2549 | */ | |
2550 | #define _SK_MEM_PACKETS 256 | |
2551 | #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256) | |
2552 | #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
2553 | #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
2554 | ||
1da177e4 LT |
2555 | extern __u32 sysctl_wmem_max; |
2556 | extern __u32 sysctl_rmem_max; | |
2557 | ||
b245be1f | 2558 | extern int sysctl_tstamp_allow_data; |
6baf1f41 DM |
2559 | extern int sysctl_optmem_max; |
2560 | ||
20380731 ACM |
2561 | extern __u32 sysctl_wmem_default; |
2562 | extern __u32 sysctl_rmem_default; | |
20380731 | 2563 | |
ce27ec60 ED |
2564 | DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key); |
2565 | ||
a3dcaf17 ED |
2566 | static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto) |
2567 | { | |
2568 | /* Does this proto have per netns sysctl_wmem ? */ | |
2569 | if (proto->sysctl_wmem_offset) | |
2570 | return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset); | |
2571 | ||
2572 | return *proto->sysctl_wmem; | |
2573 | } | |
2574 | ||
2575 | static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto) | |
2576 | { | |
2577 | /* Does this proto have per netns sysctl_rmem ? */ | |
2578 | if (proto->sysctl_rmem_offset) | |
2579 | return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset); | |
2580 | ||
2581 | return *proto->sysctl_rmem; | |
2582 | } | |
2583 | ||
c9f1f58d ED |
2584 | /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10) |
2585 | * Some wifi drivers need to tweak it to get more chunks. | |
2586 | * They can use this helper from their ndo_start_xmit() | |
2587 | */ | |
2588 | static inline void sk_pacing_shift_update(struct sock *sk, int val) | |
2589 | { | |
7c68fa2b | 2590 | if (!sk || !sk_fullsock(sk) || READ_ONCE(sk->sk_pacing_shift) == val) |
c9f1f58d | 2591 | return; |
7c68fa2b | 2592 | WRITE_ONCE(sk->sk_pacing_shift, val); |
c9f1f58d ED |
2593 | } |
2594 | ||
54dc3e33 DA |
2595 | /* if a socket is bound to a device, check that the given device |
2596 | * index is either the same or that the socket is bound to an L3 | |
2597 | * master device and the given device index is also enslaved to | |
2598 | * that L3 master | |
2599 | */ | |
2600 | static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif) | |
2601 | { | |
2602 | int mdif; | |
2603 | ||
2604 | if (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif) | |
2605 | return true; | |
2606 | ||
2607 | mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif); | |
2608 | if (mdif && mdif == sk->sk_bound_dev_if) | |
2609 | return true; | |
2610 | ||
2611 | return false; | |
2612 | } | |
2613 | ||
1da177e4 | 2614 | #endif /* _SOCK_H */ |