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> |
f16a7dd5 | 69 | #include <net/smc.h> |
54dc3e33 | 70 | #include <net/l3mdev.h> |
1da177e4 LT |
71 | |
72 | /* | |
73 | * This structure really needs to be cleaned up. | |
74 | * Most of it is for TCP, and not used by any of | |
75 | * the other protocols. | |
76 | */ | |
77 | ||
78 | /* Define this to get the SOCK_DBG debugging facility. */ | |
79 | #define SOCK_DEBUGGING | |
80 | #ifdef SOCK_DEBUGGING | |
81 | #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \ | |
82 | printk(KERN_DEBUG msg); } while (0) | |
83 | #else | |
4cd9029d | 84 | /* Validate arguments and do nothing */ |
b9075fa9 | 85 | static inline __printf(2, 3) |
dc6b9b78 | 86 | void SOCK_DEBUG(const struct sock *sk, const char *msg, ...) |
4cd9029d SH |
87 | { |
88 | } | |
1da177e4 LT |
89 | #endif |
90 | ||
91 | /* This is the per-socket lock. The spinlock provides a synchronization | |
92 | * between user contexts and software interrupt processing, whereas the | |
93 | * mini-semaphore synchronizes multiple users amongst themselves. | |
94 | */ | |
1da177e4 LT |
95 | typedef struct { |
96 | spinlock_t slock; | |
d2e9117c | 97 | int owned; |
1da177e4 | 98 | wait_queue_head_t wq; |
a5b5bb9a IM |
99 | /* |
100 | * We express the mutex-alike socket_lock semantics | |
101 | * to the lock validator by explicitly managing | |
102 | * the slock as a lock variant (in addition to | |
103 | * the slock itself): | |
104 | */ | |
105 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
106 | struct lockdep_map dep_map; | |
107 | #endif | |
1da177e4 LT |
108 | } socket_lock_t; |
109 | ||
1da177e4 | 110 | struct sock; |
8feaf0c0 | 111 | struct proto; |
0eeb8ffc | 112 | struct net; |
1da177e4 | 113 | |
077b393d ED |
114 | typedef __u32 __bitwise __portpair; |
115 | typedef __u64 __bitwise __addrpair; | |
116 | ||
1da177e4 | 117 | /** |
4dc3b16b | 118 | * struct sock_common - minimal network layer representation of sockets |
68835aba ED |
119 | * @skc_daddr: Foreign IPv4 addr |
120 | * @skc_rcv_saddr: Bound local IPv4 addr | |
4dc6dc71 | 121 | * @skc_hash: hash value used with various protocol lookup tables |
d4cada4a | 122 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
ce43b03e ED |
123 | * @skc_dport: placeholder for inet_dport/tw_dport |
124 | * @skc_num: placeholder for inet_num/tw_num | |
4dc3b16b PP |
125 | * @skc_family: network address family |
126 | * @skc_state: Connection state | |
127 | * @skc_reuse: %SO_REUSEADDR setting | |
055dc21a | 128 | * @skc_reuseport: %SO_REUSEPORT setting |
4dc3b16b | 129 | * @skc_bound_dev_if: bound device index if != 0 |
4dc3b16b | 130 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
512615b6 | 131 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol |
8feaf0c0 | 132 | * @skc_prot: protocol handlers inside a network family |
07feaebf | 133 | * @skc_net: reference to the network namespace of this socket |
68835aba ED |
134 | * @skc_node: main hash linkage for various protocol lookup tables |
135 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol | |
136 | * @skc_tx_queue_mapping: tx queue number for this connection | |
c6345ce7 | 137 | * @skc_rx_queue_mapping: rx queue number for this connection |
8e5eb54d ED |
138 | * @skc_flags: place holder for sk_flags |
139 | * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, | |
140 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings | |
70da268b | 141 | * @skc_incoming_cpu: record/match cpu processing incoming packets |
68835aba | 142 | * @skc_refcnt: reference count |
4dc3b16b PP |
143 | * |
144 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
145 | * for struct sock and struct inet_timewait_sock. |
146 | */ | |
1da177e4 | 147 | struct sock_common { |
ce43b03e | 148 | /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned |
05dbc7b5 | 149 | * address on 64bit arches : cf INET_MATCH() |
4dc6dc71 | 150 | */ |
ce43b03e | 151 | union { |
077b393d | 152 | __addrpair skc_addrpair; |
ce43b03e ED |
153 | struct { |
154 | __be32 skc_daddr; | |
155 | __be32 skc_rcv_saddr; | |
156 | }; | |
157 | }; | |
d4cada4a ED |
158 | union { |
159 | unsigned int skc_hash; | |
160 | __u16 skc_u16hashes[2]; | |
161 | }; | |
ce43b03e ED |
162 | /* skc_dport && skc_num must be grouped as well */ |
163 | union { | |
077b393d | 164 | __portpair skc_portpair; |
ce43b03e ED |
165 | struct { |
166 | __be16 skc_dport; | |
167 | __u16 skc_num; | |
168 | }; | |
169 | }; | |
170 | ||
4dc6dc71 ED |
171 | unsigned short skc_family; |
172 | volatile unsigned char skc_state; | |
055dc21a | 173 | unsigned char skc_reuse:4; |
9fe516ba ED |
174 | unsigned char skc_reuseport:1; |
175 | unsigned char skc_ipv6only:1; | |
26abe143 | 176 | unsigned char skc_net_refcnt:1; |
4dc6dc71 | 177 | int skc_bound_dev_if; |
512615b6 ED |
178 | union { |
179 | struct hlist_node skc_bind_node; | |
ca065d0c | 180 | struct hlist_node skc_portaddr_node; |
512615b6 | 181 | }; |
8feaf0c0 | 182 | struct proto *skc_prot; |
0c5c9fb5 | 183 | possible_net_t skc_net; |
efe4208f ED |
184 | |
185 | #if IS_ENABLED(CONFIG_IPV6) | |
186 | struct in6_addr skc_v6_daddr; | |
187 | struct in6_addr skc_v6_rcv_saddr; | |
188 | #endif | |
189 | ||
33cf7c90 ED |
190 | atomic64_t skc_cookie; |
191 | ||
8e5eb54d ED |
192 | /* following fields are padding to force |
193 | * offset(struct sock, sk_refcnt) == 128 on 64bit arches | |
194 | * assuming IPV6 is enabled. We use this padding differently | |
195 | * for different kind of 'sockets' | |
196 | */ | |
197 | union { | |
198 | unsigned long skc_flags; | |
199 | struct sock *skc_listener; /* request_sock */ | |
200 | struct inet_timewait_death_row *skc_tw_dr; /* inet_timewait_sock */ | |
201 | }; | |
68835aba ED |
202 | /* |
203 | * fields between dontcopy_begin/dontcopy_end | |
204 | * are not copied in sock_copy() | |
205 | */ | |
928c41e7 | 206 | /* private: */ |
68835aba | 207 | int skc_dontcopy_begin[0]; |
928c41e7 | 208 | /* public: */ |
68835aba ED |
209 | union { |
210 | struct hlist_node skc_node; | |
211 | struct hlist_nulls_node skc_nulls_node; | |
212 | }; | |
755c31cd | 213 | unsigned short skc_tx_queue_mapping; |
c6345ce7 AN |
214 | #ifdef CONFIG_XPS |
215 | unsigned short skc_rx_queue_mapping; | |
216 | #endif | |
ed53d0ab ED |
217 | union { |
218 | int skc_incoming_cpu; | |
219 | u32 skc_rcv_wnd; | |
d475f090 | 220 | u32 skc_tw_rcv_nxt; /* struct tcp_timewait_sock */ |
ed53d0ab | 221 | }; |
70da268b | 222 | |
41c6d650 | 223 | refcount_t skc_refcnt; |
928c41e7 | 224 | /* private: */ |
68835aba | 225 | int skc_dontcopy_end[0]; |
ed53d0ab ED |
226 | union { |
227 | u32 skc_rxhash; | |
228 | u32 skc_window_clamp; | |
d475f090 | 229 | u32 skc_tw_snd_nxt; /* struct tcp_timewait_sock */ |
ed53d0ab | 230 | }; |
928c41e7 | 231 | /* public: */ |
1da177e4 LT |
232 | }; |
233 | ||
6ac99e8f MKL |
234 | struct bpf_sk_storage; |
235 | ||
1da177e4 LT |
236 | /** |
237 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 238 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
239 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
240 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
241 | * @sk_lock: synchronizer | |
cdfbabfb | 242 | * @sk_kern_sock: True if sock is using kernel lock classes |
4dc3b16b | 243 | * @sk_rcvbuf: size of receive buffer in bytes |
43815482 | 244 | * @sk_wq: sock wait queue and async head |
421b3885 | 245 | * @sk_rx_dst: receive input route used by early demux |
4dc3b16b | 246 | * @sk_dst_cache: destination cache |
9b8805a3 | 247 | * @sk_dst_pending_confirm: need to confirm neighbour |
4dc3b16b | 248 | * @sk_policy: flow policy |
4dc3b16b PP |
249 | * @sk_receive_queue: incoming packets |
250 | * @sk_wmem_alloc: transmit queue bytes committed | |
771edcaf | 251 | * @sk_tsq_flags: TCP Small Queues flags |
4dc3b16b PP |
252 | * @sk_write_queue: Packet sending queue |
253 | * @sk_omem_alloc: "o" is "option" or "other" | |
254 | * @sk_wmem_queued: persistent queue size | |
255 | * @sk_forward_alloc: space allocated forward | |
06021292 | 256 | * @sk_napi_id: id of the last napi context to receive data for sk |
dafcc438 | 257 | * @sk_ll_usec: usecs to busypoll when there is no data |
4dc3b16b | 258 | * @sk_allocation: allocation mode |
95bd09eb | 259 | * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler) |
218af599 | 260 | * @sk_pacing_status: Pacing status (requested, handled by sch_fq) |
c3f40d7c | 261 | * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE) |
4dc3b16b | 262 | * @sk_sndbuf: size of send buffer in bytes |
771edcaf | 263 | * @__sk_flags_offset: empty field used to determine location of bitfield |
293de7de | 264 | * @sk_padding: unused element for alignment |
28448b80 TH |
265 | * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets |
266 | * @sk_no_check_rx: allow zero checksum in RX packets | |
4dc3b16b | 267 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
a465419b | 268 | * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK) |
bcd76111 | 269 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
82cc1a7a | 270 | * @sk_gso_max_size: Maximum GSO segment size to build |
1485348d | 271 | * @sk_gso_max_segs: Maximum number of GSO segments |
3a9b76fd | 272 | * @sk_pacing_shift: scaling factor for TCP Small Queues |
4dc3b16b | 273 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
274 | * @sk_backlog: always used with the per-socket spinlock held |
275 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
276 | * @sk_error_queue: rarely used | |
33c732c3 WC |
277 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
278 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 279 | * @sk_err: last error |
33c732c3 WC |
280 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
281 | * persistent failure not just 'timed out' | |
cb61cb9b | 282 | * @sk_drops: raw/udp drops counter |
4dc3b16b PP |
283 | * @sk_ack_backlog: current listen backlog |
284 | * @sk_max_ack_backlog: listen backlog set in listen() | |
771edcaf | 285 | * @sk_uid: user id of owner |
4dc3b16b PP |
286 | * @sk_priority: %SO_PRIORITY setting |
287 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
288 | * @sk_protocol: which protocol this socket belongs in this network family | |
53c3fa20 RD |
289 | * @sk_peer_pid: &struct pid for this socket's peer |
290 | * @sk_peer_cred: %SO_PEERCRED setting | |
4dc3b16b PP |
291 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
292 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
293 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
b73c3d0e | 294 | * @sk_txhash: computed flow hash for use on transmit |
4dc3b16b | 295 | * @sk_filter: socket filtering instructions |
4dc3b16b PP |
296 | * @sk_timer: sock cleanup timer |
297 | * @sk_stamp: time stamp of last packet received | |
3a0ed3e9 | 298 | * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only |
b9f40e21 | 299 | * @sk_tsflags: SO_TIMESTAMPING socket options |
09c2d251 | 300 | * @sk_tskey: counter to disambiguate concurrent tstamp requests |
52267790 | 301 | * @sk_zckey: counter to order MSG_ZEROCOPY notifications |
4dc3b16b PP |
302 | * @sk_socket: Identd and reporting IO signals |
303 | * @sk_user_data: RPC layer private data | |
5640f768 | 304 | * @sk_frag: cached page frag |
d3d4f0a0 | 305 | * @sk_peek_off: current peek_offset value |
4dc3b16b | 306 | * @sk_send_head: front of stuff to transmit |
67be2dd1 | 307 | * @sk_security: used by security modules |
31729363 | 308 | * @sk_mark: generic packet mark |
2a56a1fe | 309 | * @sk_cgrp_data: cgroup data for this cgroup |
baac50bb | 310 | * @sk_memcg: this socket's memory cgroup association |
4dc3b16b PP |
311 | * @sk_write_pending: a write to stream socket waits to start |
312 | * @sk_state_change: callback to indicate change in the state of the sock | |
313 | * @sk_data_ready: callback to indicate there is data to be processed | |
314 | * @sk_write_space: callback to indicate there is bf sending space available | |
315 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
316 | * @sk_backlog_rcv: callback to process the backlog | |
317 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 | |
ef456144 | 318 | * @sk_reuseport_cb: reuseport group container |
293de7de | 319 | * @sk_rcu: used during RCU grace period |
80b14dee RC |
320 | * @sk_clockid: clockid used by time-based scheduling (SO_TXTIME) |
321 | * @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME | |
322 | * @sk_txtime_unused: unused txtime flags | |
293de7de | 323 | */ |
1da177e4 LT |
324 | struct sock { |
325 | /* | |
8feaf0c0 | 326 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
327 | * don't add nothing before this first member (__sk_common) --acme |
328 | */ | |
329 | struct sock_common __sk_common; | |
4dc6dc71 ED |
330 | #define sk_node __sk_common.skc_node |
331 | #define sk_nulls_node __sk_common.skc_nulls_node | |
332 | #define sk_refcnt __sk_common.skc_refcnt | |
e022f0b4 | 333 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
c6345ce7 AN |
334 | #ifdef CONFIG_XPS |
335 | #define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping | |
336 | #endif | |
4dc6dc71 | 337 | |
68835aba ED |
338 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
339 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end | |
4dc6dc71 | 340 | #define sk_hash __sk_common.skc_hash |
50805466 | 341 | #define sk_portpair __sk_common.skc_portpair |
05dbc7b5 ED |
342 | #define sk_num __sk_common.skc_num |
343 | #define sk_dport __sk_common.skc_dport | |
50805466 ED |
344 | #define sk_addrpair __sk_common.skc_addrpair |
345 | #define sk_daddr __sk_common.skc_daddr | |
346 | #define sk_rcv_saddr __sk_common.skc_rcv_saddr | |
1da177e4 LT |
347 | #define sk_family __sk_common.skc_family |
348 | #define sk_state __sk_common.skc_state | |
349 | #define sk_reuse __sk_common.skc_reuse | |
055dc21a | 350 | #define sk_reuseport __sk_common.skc_reuseport |
9fe516ba | 351 | #define sk_ipv6only __sk_common.skc_ipv6only |
26abe143 | 352 | #define sk_net_refcnt __sk_common.skc_net_refcnt |
1da177e4 | 353 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
1da177e4 | 354 | #define sk_bind_node __sk_common.skc_bind_node |
8feaf0c0 | 355 | #define sk_prot __sk_common.skc_prot |
07feaebf | 356 | #define sk_net __sk_common.skc_net |
efe4208f ED |
357 | #define sk_v6_daddr __sk_common.skc_v6_daddr |
358 | #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr | |
33cf7c90 | 359 | #define sk_cookie __sk_common.skc_cookie |
70da268b | 360 | #define sk_incoming_cpu __sk_common.skc_incoming_cpu |
8e5eb54d | 361 | #define sk_flags __sk_common.skc_flags |
ed53d0ab | 362 | #define sk_rxhash __sk_common.skc_rxhash |
efe4208f | 363 | |
1da177e4 | 364 | socket_lock_t sk_lock; |
9115e8cd ED |
365 | atomic_t sk_drops; |
366 | int sk_rcvlowat; | |
367 | struct sk_buff_head sk_error_queue; | |
8b27dae5 | 368 | struct sk_buff *sk_rx_skb_cache; |
b178bb3d | 369 | struct sk_buff_head sk_receive_queue; |
fa438ccf ED |
370 | /* |
371 | * The backlog queue is special, it is always used with | |
372 | * the per-socket spinlock held and requires low latency | |
373 | * access. Therefore we special case it's implementation. | |
b178bb3d ED |
374 | * Note : rmem_alloc is in this structure to fill a hole |
375 | * on 64bit arches, not because its logically part of | |
376 | * backlog. | |
fa438ccf ED |
377 | */ |
378 | struct { | |
b178bb3d ED |
379 | atomic_t rmem_alloc; |
380 | int len; | |
381 | struct sk_buff *head; | |
382 | struct sk_buff *tail; | |
fa438ccf | 383 | } sk_backlog; |
b178bb3d | 384 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
2c8c56e1 | 385 | |
9115e8cd | 386 | int sk_forward_alloc; |
e0d1095a | 387 | #ifdef CONFIG_NET_RX_BUSY_POLL |
dafcc438 | 388 | unsigned int sk_ll_usec; |
9115e8cd ED |
389 | /* ===== mostly read cache line ===== */ |
390 | unsigned int sk_napi_id; | |
b178bb3d | 391 | #endif |
b178bb3d ED |
392 | int sk_rcvbuf; |
393 | ||
394 | struct sk_filter __rcu *sk_filter; | |
ceb5d58b ED |
395 | union { |
396 | struct socket_wq __rcu *sk_wq; | |
397 | struct socket_wq *sk_wq_raw; | |
398 | }; | |
def8b4fa | 399 | #ifdef CONFIG_XFRM |
d188ba86 | 400 | struct xfrm_policy __rcu *sk_policy[2]; |
def8b4fa | 401 | #endif |
deaa5854 | 402 | struct dst_entry *sk_rx_dst; |
0e36cbb3 | 403 | struct dst_entry __rcu *sk_dst_cache; |
1da177e4 | 404 | atomic_t sk_omem_alloc; |
4e07a91c | 405 | int sk_sndbuf; |
9115e8cd ED |
406 | |
407 | /* ===== cache line for TX ===== */ | |
408 | int sk_wmem_queued; | |
14afee4b | 409 | refcount_t sk_wmem_alloc; |
9115e8cd | 410 | unsigned long sk_tsq_flags; |
75c119af ED |
411 | union { |
412 | struct sk_buff *sk_send_head; | |
413 | struct rb_root tcp_rtx_queue; | |
414 | }; | |
472c2e07 | 415 | struct sk_buff *sk_tx_skb_cache; |
1da177e4 | 416 | struct sk_buff_head sk_write_queue; |
9115e8cd ED |
417 | __s32 sk_peek_off; |
418 | int sk_write_pending; | |
9b8805a3 | 419 | __u32 sk_dst_pending_confirm; |
218af599 | 420 | u32 sk_pacing_status; /* see enum sk_pacing */ |
9115e8cd ED |
421 | long sk_sndtimeo; |
422 | struct timer_list sk_timer; | |
423 | __u32 sk_priority; | |
424 | __u32 sk_mark; | |
76a9ebe8 ED |
425 | unsigned long sk_pacing_rate; /* bytes per second */ |
426 | unsigned long sk_max_pacing_rate; | |
9115e8cd ED |
427 | struct page_frag sk_frag; |
428 | netdev_features_t sk_route_caps; | |
429 | netdev_features_t sk_route_nocaps; | |
0a6b2a1d | 430 | netdev_features_t sk_route_forced_caps; |
9115e8cd ED |
431 | int sk_gso_type; |
432 | unsigned int sk_gso_max_size; | |
433 | gfp_t sk_allocation; | |
434 | __u32 sk_txhash; | |
fc64869c AR |
435 | |
436 | /* | |
437 | * Because of non atomicity rules, all | |
438 | * changes are protected by socket lock. | |
439 | */ | |
aa4c1037 DA |
440 | unsigned int __sk_flags_offset[0]; |
441 | #ifdef __BIG_ENDIAN_BITFIELD | |
442 | #define SK_FL_PROTO_SHIFT 16 | |
443 | #define SK_FL_PROTO_MASK 0x00ff0000 | |
444 | ||
445 | #define SK_FL_TYPE_SHIFT 0 | |
446 | #define SK_FL_TYPE_MASK 0x0000ffff | |
447 | #else | |
448 | #define SK_FL_PROTO_SHIFT 8 | |
449 | #define SK_FL_PROTO_MASK 0x0000ff00 | |
450 | ||
451 | #define SK_FL_TYPE_SHIFT 16 | |
452 | #define SK_FL_TYPE_MASK 0xffff0000 | |
453 | #endif | |
454 | ||
cdfbabfb DH |
455 | unsigned int sk_padding : 1, |
456 | sk_kern_sock : 1, | |
28448b80 TH |
457 | sk_no_check_tx : 1, |
458 | sk_no_check_rx : 1, | |
b178bb3d ED |
459 | sk_userlocks : 4, |
460 | sk_protocol : 8, | |
461 | sk_type : 16; | |
7bbadd2d | 462 | #define SK_PROTOCOL_MAX U8_MAX |
1485348d | 463 | u16 sk_gso_max_segs; |
3a9b76fd | 464 | u8 sk_pacing_shift; |
1da177e4 | 465 | unsigned long sk_lingertime; |
476e19cf | 466 | struct proto *sk_prot_creator; |
1da177e4 LT |
467 | rwlock_t sk_callback_lock; |
468 | int sk_err, | |
469 | sk_err_soft; | |
becb74f0 ED |
470 | u32 sk_ack_backlog; |
471 | u32 sk_max_ack_backlog; | |
86741ec2 | 472 | kuid_t sk_uid; |
109f6e39 EB |
473 | struct pid *sk_peer_pid; |
474 | const struct cred *sk_peer_cred; | |
1da177e4 | 475 | long sk_rcvtimeo; |
b7aa0bf7 | 476 | ktime_t sk_stamp; |
3a0ed3e9 DD |
477 | #if BITS_PER_LONG==32 |
478 | seqlock_t sk_stamp_seq; | |
479 | #endif | |
b9f40e21 | 480 | u16 sk_tsflags; |
fc64869c | 481 | u8 sk_shutdown; |
09c2d251 | 482 | u32 sk_tskey; |
52267790 | 483 | atomic_t sk_zckey; |
80b14dee RC |
484 | |
485 | u8 sk_clockid; | |
486 | u8 sk_txtime_deadline_mode : 1, | |
4b15c707 JSP |
487 | sk_txtime_report_errors : 1, |
488 | sk_txtime_unused : 6; | |
80b14dee | 489 | |
1da177e4 LT |
490 | struct socket *sk_socket; |
491 | void *sk_user_data; | |
d5f64238 | 492 | #ifdef CONFIG_SECURITY |
1da177e4 | 493 | void *sk_security; |
d5f64238 | 494 | #endif |
2a56a1fe | 495 | struct sock_cgroup_data sk_cgrp_data; |
baac50bb | 496 | struct mem_cgroup *sk_memcg; |
1da177e4 | 497 | void (*sk_state_change)(struct sock *sk); |
676d2369 | 498 | void (*sk_data_ready)(struct sock *sk); |
1da177e4 LT |
499 | void (*sk_write_space)(struct sock *sk); |
500 | void (*sk_error_report)(struct sock *sk); | |
dc6b9b78 ED |
501 | int (*sk_backlog_rcv)(struct sock *sk, |
502 | struct sk_buff *skb); | |
ebf4e808 IL |
503 | #ifdef CONFIG_SOCK_VALIDATE_XMIT |
504 | struct sk_buff* (*sk_validate_xmit_skb)(struct sock *sk, | |
505 | struct net_device *dev, | |
506 | struct sk_buff *skb); | |
507 | #endif | |
1da177e4 | 508 | void (*sk_destruct)(struct sock *sk); |
ef456144 | 509 | struct sock_reuseport __rcu *sk_reuseport_cb; |
6ac99e8f MKL |
510 | #ifdef CONFIG_BPF_SYSCALL |
511 | struct bpf_sk_storage __rcu *sk_bpf_storage; | |
512 | #endif | |
a4298e45 | 513 | struct rcu_head sk_rcu; |
1da177e4 LT |
514 | }; |
515 | ||
218af599 ED |
516 | enum sk_pacing { |
517 | SK_PACING_NONE = 0, | |
518 | SK_PACING_NEEDED = 1, | |
519 | SK_PACING_FQ = 2, | |
520 | }; | |
521 | ||
559835ea PS |
522 | #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data))) |
523 | ||
524 | #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk))) | |
525 | #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr) | |
526 | ||
4a17fd52 PE |
527 | /* |
528 | * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK | |
529 | * or not whether his port will be reused by someone else. SK_FORCE_REUSE | |
530 | * on a socket means that the socket will reuse everybody else's port | |
531 | * without looking at the other's sk_reuse value. | |
532 | */ | |
533 | ||
534 | #define SK_NO_REUSE 0 | |
535 | #define SK_CAN_REUSE 1 | |
536 | #define SK_FORCE_REUSE 2 | |
537 | ||
627d2d6b | 538 | int sk_set_peek_off(struct sock *sk, int val); |
539 | ||
ef64a54f PE |
540 | static inline int sk_peek_offset(struct sock *sk, int flags) |
541 | { | |
b9bb53f3 | 542 | if (unlikely(flags & MSG_PEEK)) { |
a0917e0b | 543 | return READ_ONCE(sk->sk_peek_off); |
b9bb53f3 WB |
544 | } |
545 | ||
546 | return 0; | |
ef64a54f PE |
547 | } |
548 | ||
549 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) | |
550 | { | |
b9bb53f3 WB |
551 | s32 off = READ_ONCE(sk->sk_peek_off); |
552 | ||
553 | if (unlikely(off >= 0)) { | |
554 | off = max_t(s32, off - val, 0); | |
555 | WRITE_ONCE(sk->sk_peek_off, off); | |
ef64a54f PE |
556 | } |
557 | } | |
558 | ||
559 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) | |
560 | { | |
b9bb53f3 | 561 | sk_peek_offset_bwd(sk, -val); |
ef64a54f PE |
562 | } |
563 | ||
1da177e4 LT |
564 | /* |
565 | * Hashed lists helper routines | |
566 | */ | |
c4146644 LZ |
567 | static inline struct sock *sk_entry(const struct hlist_node *node) |
568 | { | |
569 | return hlist_entry(node, struct sock, sk_node); | |
570 | } | |
571 | ||
e48c414e | 572 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
573 | { |
574 | return hlist_entry(head->first, struct sock, sk_node); | |
575 | } | |
576 | ||
e48c414e | 577 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
578 | { |
579 | return hlist_empty(head) ? NULL : __sk_head(head); | |
580 | } | |
581 | ||
88ab1932 ED |
582 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
583 | { | |
584 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); | |
585 | } | |
586 | ||
587 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) | |
588 | { | |
589 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); | |
590 | } | |
591 | ||
e48c414e | 592 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 | 593 | { |
6c59ebd3 | 594 | return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node); |
1da177e4 LT |
595 | } |
596 | ||
88ab1932 ED |
597 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
598 | { | |
599 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? | |
600 | hlist_nulls_entry(sk->sk_nulls_node.next, | |
601 | struct sock, sk_nulls_node) : | |
602 | NULL; | |
603 | } | |
604 | ||
dc6b9b78 | 605 | static inline bool sk_unhashed(const struct sock *sk) |
1da177e4 LT |
606 | { |
607 | return hlist_unhashed(&sk->sk_node); | |
608 | } | |
609 | ||
dc6b9b78 | 610 | static inline bool sk_hashed(const struct sock *sk) |
1da177e4 | 611 | { |
da753bea | 612 | return !sk_unhashed(sk); |
1da177e4 LT |
613 | } |
614 | ||
dc6b9b78 | 615 | static inline void sk_node_init(struct hlist_node *node) |
1da177e4 LT |
616 | { |
617 | node->pprev = NULL; | |
618 | } | |
619 | ||
dc6b9b78 | 620 | static inline void sk_nulls_node_init(struct hlist_nulls_node *node) |
88ab1932 ED |
621 | { |
622 | node->pprev = NULL; | |
623 | } | |
624 | ||
dc6b9b78 | 625 | static inline void __sk_del_node(struct sock *sk) |
1da177e4 LT |
626 | { |
627 | __hlist_del(&sk->sk_node); | |
628 | } | |
629 | ||
808f5114 | 630 | /* NB: equivalent to hlist_del_init_rcu */ |
dc6b9b78 | 631 | static inline bool __sk_del_node_init(struct sock *sk) |
1da177e4 LT |
632 | { |
633 | if (sk_hashed(sk)) { | |
634 | __sk_del_node(sk); | |
635 | sk_node_init(&sk->sk_node); | |
dc6b9b78 | 636 | return true; |
1da177e4 | 637 | } |
dc6b9b78 | 638 | return false; |
1da177e4 LT |
639 | } |
640 | ||
641 | /* Grab socket reference count. This operation is valid only | |
642 | when sk is ALREADY grabbed f.e. it is found in hash table | |
643 | or a list and the lookup is made under lock preventing hash table | |
644 | modifications. | |
645 | */ | |
646 | ||
f9a7cbbf | 647 | static __always_inline void sock_hold(struct sock *sk) |
1da177e4 | 648 | { |
41c6d650 | 649 | refcount_inc(&sk->sk_refcnt); |
1da177e4 LT |
650 | } |
651 | ||
652 | /* Ungrab socket in the context, which assumes that socket refcnt | |
653 | cannot hit zero, f.e. it is true in context of any socketcall. | |
654 | */ | |
f9a7cbbf | 655 | static __always_inline void __sock_put(struct sock *sk) |
1da177e4 | 656 | { |
41c6d650 | 657 | refcount_dec(&sk->sk_refcnt); |
1da177e4 LT |
658 | } |
659 | ||
dc6b9b78 | 660 | static inline bool sk_del_node_init(struct sock *sk) |
1da177e4 | 661 | { |
dc6b9b78 | 662 | bool rc = __sk_del_node_init(sk); |
1da177e4 LT |
663 | |
664 | if (rc) { | |
665 | /* paranoid for a while -acme */ | |
41c6d650 | 666 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
1da177e4 LT |
667 | __sock_put(sk); |
668 | } | |
669 | return rc; | |
670 | } | |
808f5114 | 671 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
1da177e4 | 672 | |
dc6b9b78 | 673 | static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 ED |
674 | { |
675 | if (sk_hashed(sk)) { | |
88ab1932 | 676 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
dc6b9b78 | 677 | return true; |
271b72c7 | 678 | } |
dc6b9b78 | 679 | return false; |
271b72c7 ED |
680 | } |
681 | ||
dc6b9b78 | 682 | static inline bool sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 | 683 | { |
dc6b9b78 | 684 | bool rc = __sk_nulls_del_node_init_rcu(sk); |
271b72c7 ED |
685 | |
686 | if (rc) { | |
687 | /* paranoid for a while -acme */ | |
41c6d650 | 688 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
271b72c7 ED |
689 | __sock_put(sk); |
690 | } | |
691 | return rc; | |
692 | } | |
693 | ||
dc6b9b78 | 694 | static inline void __sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
695 | { |
696 | hlist_add_head(&sk->sk_node, list); | |
697 | } | |
698 | ||
dc6b9b78 | 699 | static inline void sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
700 | { |
701 | sock_hold(sk); | |
702 | __sk_add_node(sk, list); | |
703 | } | |
704 | ||
dc6b9b78 | 705 | static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
808f5114 | 706 | { |
707 | sock_hold(sk); | |
d296ba60 CG |
708 | if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport && |
709 | sk->sk_family == AF_INET6) | |
710 | hlist_add_tail_rcu(&sk->sk_node, list); | |
711 | else | |
712 | hlist_add_head_rcu(&sk->sk_node, list); | |
808f5114 | 713 | } |
714 | ||
a4dc6a49 MC |
715 | static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list) |
716 | { | |
717 | sock_hold(sk); | |
718 | hlist_add_tail_rcu(&sk->sk_node, list); | |
719 | } | |
720 | ||
dc6b9b78 | 721 | static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 | 722 | { |
d7efc6c1 | 723 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
271b72c7 ED |
724 | } |
725 | ||
dc6b9b78 | 726 | static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 ED |
727 | { |
728 | sock_hold(sk); | |
88ab1932 | 729 | __sk_nulls_add_node_rcu(sk, list); |
271b72c7 ED |
730 | } |
731 | ||
dc6b9b78 | 732 | static inline void __sk_del_bind_node(struct sock *sk) |
1da177e4 LT |
733 | { |
734 | __hlist_del(&sk->sk_bind_node); | |
735 | } | |
736 | ||
dc6b9b78 | 737 | static inline void sk_add_bind_node(struct sock *sk, |
1da177e4 LT |
738 | struct hlist_head *list) |
739 | { | |
740 | hlist_add_head(&sk->sk_bind_node, list); | |
741 | } | |
742 | ||
b67bfe0d SL |
743 | #define sk_for_each(__sk, list) \ |
744 | hlist_for_each_entry(__sk, list, sk_node) | |
745 | #define sk_for_each_rcu(__sk, list) \ | |
746 | hlist_for_each_entry_rcu(__sk, list, sk_node) | |
88ab1932 ED |
747 | #define sk_nulls_for_each(__sk, node, list) \ |
748 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) | |
749 | #define sk_nulls_for_each_rcu(__sk, node, list) \ | |
750 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) | |
b67bfe0d SL |
751 | #define sk_for_each_from(__sk) \ |
752 | hlist_for_each_entry_from(__sk, sk_node) | |
88ab1932 ED |
753 | #define sk_nulls_for_each_from(__sk, node) \ |
754 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ | |
755 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) | |
b67bfe0d SL |
756 | #define sk_for_each_safe(__sk, tmp, list) \ |
757 | hlist_for_each_entry_safe(__sk, tmp, list, sk_node) | |
758 | #define sk_for_each_bound(__sk, list) \ | |
759 | hlist_for_each_entry(__sk, list, sk_bind_node) | |
1da177e4 | 760 | |
2dc41cff | 761 | /** |
ca065d0c | 762 | * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset |
2dc41cff DH |
763 | * @tpos: the type * to use as a loop cursor. |
764 | * @pos: the &struct hlist_node to use as a loop cursor. | |
765 | * @head: the head for your list. | |
766 | * @offset: offset of hlist_node within the struct. | |
767 | * | |
768 | */ | |
ca065d0c | 769 | #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \ |
b6f4f848 | 770 | for (pos = rcu_dereference(hlist_first_rcu(head)); \ |
ca065d0c | 771 | pos != NULL && \ |
2dc41cff | 772 | ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \ |
b6f4f848 | 773 | pos = rcu_dereference(hlist_next_rcu(pos))) |
2dc41cff | 774 | |
c336d148 EB |
775 | static inline struct user_namespace *sk_user_ns(struct sock *sk) |
776 | { | |
777 | /* Careful only use this in a context where these parameters | |
778 | * can not change and must all be valid, such as recvmsg from | |
779 | * userspace. | |
780 | */ | |
781 | return sk->sk_socket->file->f_cred->user_ns; | |
782 | } | |
783 | ||
1da177e4 LT |
784 | /* Sock flags */ |
785 | enum sock_flags { | |
786 | SOCK_DEAD, | |
787 | SOCK_DONE, | |
788 | SOCK_URGINLINE, | |
789 | SOCK_KEEPOPEN, | |
790 | SOCK_LINGER, | |
791 | SOCK_DESTROY, | |
792 | SOCK_BROADCAST, | |
793 | SOCK_TIMESTAMP, | |
794 | SOCK_ZAPPED, | |
795 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
796 | SOCK_DBG, /* %SO_DEBUG setting */ | |
797 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 798 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 LT |
799 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
800 | SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */ | |
7cb02404 | 801 | SOCK_MEMALLOC, /* VM depends on this socket for swapping */ |
20d49473 | 802 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
bcdce719 | 803 | SOCK_FASYNC, /* fasync() active */ |
3b885787 | 804 | SOCK_RXQ_OVFL, |
1cdebb42 | 805 | SOCK_ZEROCOPY, /* buffers from userspace */ |
6e3e939f | 806 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
3bdc0eba BG |
807 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
808 | * Will use last 4 bytes of packet sent from | |
809 | * user-space instead. | |
810 | */ | |
d59577b6 | 811 | SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */ |
7d4c04fc | 812 | SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */ |
a4298e45 | 813 | SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */ |
80b14dee | 814 | SOCK_TXTIME, |
e4a2a304 | 815 | SOCK_XDP, /* XDP is attached */ |
887feae3 | 816 | SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */ |
1da177e4 LT |
817 | }; |
818 | ||
01ce63c9 MRL |
819 | #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)) |
820 | ||
53b924b3 RB |
821 | static inline void sock_copy_flags(struct sock *nsk, struct sock *osk) |
822 | { | |
823 | nsk->sk_flags = osk->sk_flags; | |
824 | } | |
825 | ||
1da177e4 LT |
826 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
827 | { | |
828 | __set_bit(flag, &sk->sk_flags); | |
829 | } | |
830 | ||
831 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
832 | { | |
833 | __clear_bit(flag, &sk->sk_flags); | |
834 | } | |
835 | ||
1b23a5df | 836 | static inline bool sock_flag(const struct sock *sk, enum sock_flags flag) |
1da177e4 LT |
837 | { |
838 | return test_bit(flag, &sk->sk_flags); | |
839 | } | |
840 | ||
c93bdd0e | 841 | #ifdef CONFIG_NET |
a7950ae8 | 842 | DECLARE_STATIC_KEY_FALSE(memalloc_socks_key); |
c93bdd0e MG |
843 | static inline int sk_memalloc_socks(void) |
844 | { | |
a7950ae8 | 845 | return static_branch_unlikely(&memalloc_socks_key); |
c93bdd0e MG |
846 | } |
847 | #else | |
848 | ||
849 | static inline int sk_memalloc_socks(void) | |
850 | { | |
851 | return 0; | |
852 | } | |
853 | ||
854 | #endif | |
855 | ||
7450aaf6 | 856 | static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask) |
99a1dec7 | 857 | { |
7450aaf6 | 858 | return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC); |
99a1dec7 MG |
859 | } |
860 | ||
1da177e4 LT |
861 | static inline void sk_acceptq_removed(struct sock *sk) |
862 | { | |
863 | sk->sk_ack_backlog--; | |
864 | } | |
865 | ||
866 | static inline void sk_acceptq_added(struct sock *sk) | |
867 | { | |
868 | sk->sk_ack_backlog++; | |
869 | } | |
870 | ||
dc6b9b78 | 871 | static inline bool sk_acceptq_is_full(const struct sock *sk) |
1da177e4 | 872 | { |
64a14651 | 873 | return sk->sk_ack_backlog > sk->sk_max_ack_backlog; |
1da177e4 LT |
874 | } |
875 | ||
876 | /* | |
877 | * Compute minimal free write space needed to queue new packets. | |
878 | */ | |
dc6b9b78 | 879 | static inline int sk_stream_min_wspace(const struct sock *sk) |
1da177e4 | 880 | { |
8df09ea3 | 881 | return sk->sk_wmem_queued >> 1; |
1da177e4 LT |
882 | } |
883 | ||
dc6b9b78 | 884 | static inline int sk_stream_wspace(const struct sock *sk) |
1da177e4 LT |
885 | { |
886 | return sk->sk_sndbuf - sk->sk_wmem_queued; | |
887 | } | |
888 | ||
69336bd2 | 889 | void sk_stream_write_space(struct sock *sk); |
1da177e4 | 890 | |
8eae939f | 891 | /* OOB backlog add */ |
a3a858ff | 892 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
9ee6b535 | 893 | { |
7fee226a | 894 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
222d7dbd | 895 | skb_dst_force(skb); |
7fee226a ED |
896 | |
897 | if (!sk->sk_backlog.tail) | |
898 | sk->sk_backlog.head = skb; | |
899 | else | |
9ee6b535 | 900 | sk->sk_backlog.tail->next = skb; |
7fee226a ED |
901 | |
902 | sk->sk_backlog.tail = skb; | |
9ee6b535 SH |
903 | skb->next = NULL; |
904 | } | |
1da177e4 | 905 | |
c377411f ED |
906 | /* |
907 | * Take into account size of receive queue and backlog queue | |
0fd7bac6 ED |
908 | * Do not take into account this skb truesize, |
909 | * to allow even a single big packet to come. | |
c377411f | 910 | */ |
274f482d | 911 | static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit) |
c377411f ED |
912 | { |
913 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); | |
914 | ||
f545a38f | 915 | return qsize > limit; |
c377411f ED |
916 | } |
917 | ||
8eae939f | 918 | /* The per-socket spinlock must be held here. */ |
f545a38f ED |
919 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb, |
920 | unsigned int limit) | |
8eae939f | 921 | { |
274f482d | 922 | if (sk_rcvqueues_full(sk, limit)) |
8eae939f ZY |
923 | return -ENOBUFS; |
924 | ||
c7c49b8f ED |
925 | /* |
926 | * If the skb was allocated from pfmemalloc reserves, only | |
927 | * allow SOCK_MEMALLOC sockets to use it as this socket is | |
928 | * helping free memory | |
929 | */ | |
930 | if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC)) | |
931 | return -ENOMEM; | |
932 | ||
a3a858ff | 933 | __sk_add_backlog(sk, skb); |
8eae939f ZY |
934 | sk->sk_backlog.len += skb->truesize; |
935 | return 0; | |
936 | } | |
937 | ||
69336bd2 | 938 | int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
b4b9e355 | 939 | |
c57943a1 PZ |
940 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
941 | { | |
b4b9e355 MG |
942 | if (sk_memalloc_socks() && skb_pfmemalloc(skb)) |
943 | return __sk_backlog_rcv(sk, skb); | |
944 | ||
c57943a1 PZ |
945 | return sk->sk_backlog_rcv(sk, skb); |
946 | } | |
947 | ||
2c8c56e1 ED |
948 | static inline void sk_incoming_cpu_update(struct sock *sk) |
949 | { | |
34cfb542 PA |
950 | int cpu = raw_smp_processor_id(); |
951 | ||
952 | if (unlikely(sk->sk_incoming_cpu != cpu)) | |
953 | sk->sk_incoming_cpu = cpu; | |
2c8c56e1 ED |
954 | } |
955 | ||
fe477558 | 956 | static inline void sock_rps_record_flow_hash(__u32 hash) |
c58dc01b DM |
957 | { |
958 | #ifdef CONFIG_RPS | |
959 | struct rps_sock_flow_table *sock_flow_table; | |
960 | ||
961 | rcu_read_lock(); | |
962 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
fe477558 | 963 | rps_record_sock_flow(sock_flow_table, hash); |
c58dc01b DM |
964 | rcu_read_unlock(); |
965 | #endif | |
966 | } | |
967 | ||
fe477558 TH |
968 | static inline void sock_rps_record_flow(const struct sock *sk) |
969 | { | |
c9d8ca04 | 970 | #ifdef CONFIG_RPS |
dc05360f | 971 | if (static_branch_unlikely(&rfs_needed)) { |
13bfff25 ED |
972 | /* Reading sk->sk_rxhash might incur an expensive cache line |
973 | * miss. | |
974 | * | |
975 | * TCP_ESTABLISHED does cover almost all states where RFS | |
976 | * might be useful, and is cheaper [1] than testing : | |
977 | * IPv4: inet_sk(sk)->inet_daddr | |
978 | * IPv6: ipv6_addr_any(&sk->sk_v6_daddr) | |
979 | * OR an additional socket flag | |
980 | * [1] : sk_state and sk_prot are in the same cache line. | |
981 | */ | |
982 | if (sk->sk_state == TCP_ESTABLISHED) | |
983 | sock_rps_record_flow_hash(sk->sk_rxhash); | |
984 | } | |
c9d8ca04 | 985 | #endif |
fe477558 TH |
986 | } |
987 | ||
bdeab991 TH |
988 | static inline void sock_rps_save_rxhash(struct sock *sk, |
989 | const struct sk_buff *skb) | |
c58dc01b DM |
990 | { |
991 | #ifdef CONFIG_RPS | |
567e4b79 | 992 | if (unlikely(sk->sk_rxhash != skb->hash)) |
61b905da | 993 | sk->sk_rxhash = skb->hash; |
c58dc01b DM |
994 | #endif |
995 | } | |
996 | ||
bdeab991 TH |
997 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
998 | { | |
999 | #ifdef CONFIG_RPS | |
bdeab991 TH |
1000 | sk->sk_rxhash = 0; |
1001 | #endif | |
1002 | } | |
1003 | ||
d9dc8b0f | 1004 | #define sk_wait_event(__sk, __timeo, __condition, __wait) \ |
cfcabdcc SH |
1005 | ({ int __rc; \ |
1006 | release_sock(__sk); \ | |
1007 | __rc = __condition; \ | |
1008 | if (!__rc) { \ | |
d9dc8b0f WC |
1009 | *(__timeo) = wait_woken(__wait, \ |
1010 | TASK_INTERRUPTIBLE, \ | |
1011 | *(__timeo)); \ | |
cfcabdcc | 1012 | } \ |
d9dc8b0f | 1013 | sched_annotate_sleep(); \ |
cfcabdcc SH |
1014 | lock_sock(__sk); \ |
1015 | __rc = __condition; \ | |
1016 | __rc; \ | |
1017 | }) | |
1da177e4 | 1018 | |
69336bd2 JP |
1019 | int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
1020 | int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
1021 | void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
1022 | int sk_stream_error(struct sock *sk, int flags, int err); | |
1023 | void sk_stream_kill_queues(struct sock *sk); | |
1024 | void sk_set_memalloc(struct sock *sk); | |
1025 | void sk_clear_memalloc(struct sock *sk); | |
1da177e4 | 1026 | |
d41a69f1 ED |
1027 | void __sk_flush_backlog(struct sock *sk); |
1028 | ||
1029 | static inline bool sk_flush_backlog(struct sock *sk) | |
1030 | { | |
1031 | if (unlikely(READ_ONCE(sk->sk_backlog.tail))) { | |
1032 | __sk_flush_backlog(sk); | |
1033 | return true; | |
1034 | } | |
1035 | return false; | |
1036 | } | |
1037 | ||
dfbafc99 | 1038 | int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb); |
1da177e4 | 1039 | |
60236fdd | 1040 | struct request_sock_ops; |
6d6ee43e | 1041 | struct timewait_sock_ops; |
ab1e0a13 | 1042 | struct inet_hashinfo; |
fc8717ba | 1043 | struct raw_hashinfo; |
f16a7dd5 | 1044 | struct smc_hashinfo; |
de477254 | 1045 | struct module; |
2e6599cb | 1046 | |
f77d6021 | 1047 | /* |
5f0d5a3a | 1048 | * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes |
f77d6021 ED |
1049 | * un-modified. Special care is taken when initializing object to zero. |
1050 | */ | |
1051 | static inline void sk_prot_clear_nulls(struct sock *sk, int size) | |
1052 | { | |
1053 | if (offsetof(struct sock, sk_node.next) != 0) | |
1054 | memset(sk, 0, offsetof(struct sock, sk_node.next)); | |
1055 | memset(&sk->sk_node.pprev, 0, | |
1056 | size - offsetof(struct sock, sk_node.pprev)); | |
1057 | } | |
1058 | ||
1da177e4 LT |
1059 | /* Networking protocol blocks we attach to sockets. |
1060 | * socket layer -> transport layer interface | |
1da177e4 LT |
1061 | */ |
1062 | struct proto { | |
dc6b9b78 | 1063 | void (*close)(struct sock *sk, |
1da177e4 | 1064 | long timeout); |
d74bad4e AI |
1065 | int (*pre_connect)(struct sock *sk, |
1066 | struct sockaddr *uaddr, | |
1067 | int addr_len); | |
1da177e4 | 1068 | int (*connect)(struct sock *sk, |
dc6b9b78 | 1069 | struct sockaddr *uaddr, |
1da177e4 LT |
1070 | int addr_len); |
1071 | int (*disconnect)(struct sock *sk, int flags); | |
1072 | ||
cdfbabfb DH |
1073 | struct sock * (*accept)(struct sock *sk, int flags, int *err, |
1074 | bool kern); | |
1da177e4 LT |
1075 | |
1076 | int (*ioctl)(struct sock *sk, int cmd, | |
1077 | unsigned long arg); | |
1078 | int (*init)(struct sock *sk); | |
7d06b2e0 | 1079 | void (*destroy)(struct sock *sk); |
1da177e4 | 1080 | void (*shutdown)(struct sock *sk, int how); |
dc6b9b78 | 1081 | int (*setsockopt)(struct sock *sk, int level, |
1da177e4 | 1082 | int optname, char __user *optval, |
b7058842 | 1083 | unsigned int optlen); |
dc6b9b78 ED |
1084 | int (*getsockopt)(struct sock *sk, int level, |
1085 | int optname, char __user *optval, | |
1086 | int __user *option); | |
4b9d07a4 | 1087 | void (*keepalive)(struct sock *sk, int valbool); |
af01d537 | 1088 | #ifdef CONFIG_COMPAT |
3fdadf7d DM |
1089 | int (*compat_setsockopt)(struct sock *sk, |
1090 | int level, | |
1091 | int optname, char __user *optval, | |
b7058842 | 1092 | unsigned int optlen); |
3fdadf7d DM |
1093 | int (*compat_getsockopt)(struct sock *sk, |
1094 | int level, | |
1095 | int optname, char __user *optval, | |
1096 | int __user *option); | |
709b46e8 EB |
1097 | int (*compat_ioctl)(struct sock *sk, |
1098 | unsigned int cmd, unsigned long arg); | |
af01d537 | 1099 | #endif |
1b784140 YX |
1100 | int (*sendmsg)(struct sock *sk, struct msghdr *msg, |
1101 | size_t len); | |
1102 | int (*recvmsg)(struct sock *sk, struct msghdr *msg, | |
dc6b9b78 ED |
1103 | size_t len, int noblock, int flags, |
1104 | int *addr_len); | |
1da177e4 LT |
1105 | int (*sendpage)(struct sock *sk, struct page *page, |
1106 | int offset, size_t size, int flags); | |
dc6b9b78 | 1107 | int (*bind)(struct sock *sk, |
1da177e4 LT |
1108 | struct sockaddr *uaddr, int addr_len); |
1109 | ||
dc6b9b78 | 1110 | int (*backlog_rcv) (struct sock *sk, |
1da177e4 LT |
1111 | struct sk_buff *skb); |
1112 | ||
46d3ceab ED |
1113 | void (*release_cb)(struct sock *sk); |
1114 | ||
1da177e4 | 1115 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
086c653f | 1116 | int (*hash)(struct sock *sk); |
1da177e4 | 1117 | void (*unhash)(struct sock *sk); |
719f8358 | 1118 | void (*rehash)(struct sock *sk); |
1da177e4 LT |
1119 | int (*get_port)(struct sock *sk, unsigned short snum); |
1120 | ||
286ab3d4 | 1121 | /* Keeping track of sockets in use */ |
65f76517 | 1122 | #ifdef CONFIG_PROC_FS |
13ff3d6f | 1123 | unsigned int inuse_idx; |
65f76517 | 1124 | #endif |
ebb53d75 | 1125 | |
a74f0fa0 | 1126 | bool (*stream_memory_free)(const struct sock *sk, int wake); |
8934ce2f | 1127 | bool (*stream_memory_read)(const struct sock *sk); |
1da177e4 | 1128 | /* Memory pressure */ |
5c52ba17 | 1129 | void (*enter_memory_pressure)(struct sock *sk); |
06044751 | 1130 | void (*leave_memory_pressure)(struct sock *sk); |
8d987e5c | 1131 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
1748376b | 1132 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
1da177e4 LT |
1133 | /* |
1134 | * Pressure flag: try to collapse. | |
1135 | * Technical note: it is used by multiple contexts non atomically. | |
3ab224be | 1136 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
1137 | * is strict, actions are advisory and have some latency. |
1138 | */ | |
06044751 | 1139 | unsigned long *memory_pressure; |
8d987e5c | 1140 | long *sysctl_mem; |
a3dcaf17 | 1141 | |
1da177e4 LT |
1142 | int *sysctl_wmem; |
1143 | int *sysctl_rmem; | |
a3dcaf17 ED |
1144 | u32 sysctl_wmem_offset; |
1145 | u32 sysctl_rmem_offset; | |
1146 | ||
1da177e4 | 1147 | int max_header; |
7ba42910 | 1148 | bool no_autobind; |
1da177e4 | 1149 | |
271b72c7 | 1150 | struct kmem_cache *slab; |
1da177e4 | 1151 | unsigned int obj_size; |
d50112ed | 1152 | slab_flags_t slab_flags; |
7bbdb81e AD |
1153 | unsigned int useroffset; /* Usercopy region offset */ |
1154 | unsigned int usersize; /* Usercopy region size */ | |
1da177e4 | 1155 | |
dd24c001 | 1156 | struct percpu_counter *orphan_count; |
8feaf0c0 | 1157 | |
60236fdd | 1158 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 1159 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 1160 | |
39d8cda7 PE |
1161 | union { |
1162 | struct inet_hashinfo *hashinfo; | |
645ca708 | 1163 | struct udp_table *udp_table; |
fc8717ba | 1164 | struct raw_hashinfo *raw_hash; |
f16a7dd5 | 1165 | struct smc_hashinfo *smc_hash; |
39d8cda7 | 1166 | } h; |
ab1e0a13 | 1167 | |
1da177e4 LT |
1168 | struct module *owner; |
1169 | ||
1170 | char name[32]; | |
1171 | ||
1172 | struct list_head node; | |
e6848976 ACM |
1173 | #ifdef SOCK_REFCNT_DEBUG |
1174 | atomic_t socks; | |
e1aab161 | 1175 | #endif |
64be0aed | 1176 | int (*diag_destroy)(struct sock *sk, int err); |
3859a271 | 1177 | } __randomize_layout; |
e1aab161 | 1178 | |
69336bd2 JP |
1179 | int proto_register(struct proto *prot, int alloc_slab); |
1180 | void proto_unregister(struct proto *prot); | |
bf2ae2e4 | 1181 | int sock_load_diag_module(int family, int protocol); |
1da177e4 | 1182 | |
e6848976 ACM |
1183 | #ifdef SOCK_REFCNT_DEBUG |
1184 | static inline void sk_refcnt_debug_inc(struct sock *sk) | |
1185 | { | |
1186 | atomic_inc(&sk->sk_prot->socks); | |
1187 | } | |
1188 | ||
1189 | static inline void sk_refcnt_debug_dec(struct sock *sk) | |
1190 | { | |
1191 | atomic_dec(&sk->sk_prot->socks); | |
1192 | printk(KERN_DEBUG "%s socket %p released, %d are still alive\n", | |
1193 | sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks)); | |
1194 | } | |
1195 | ||
dec34fb0 | 1196 | static inline void sk_refcnt_debug_release(const struct sock *sk) |
e6848976 | 1197 | { |
41c6d650 | 1198 | if (refcount_read(&sk->sk_refcnt) != 1) |
e6848976 | 1199 | printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n", |
41c6d650 | 1200 | sk->sk_prot->name, sk, refcount_read(&sk->sk_refcnt)); |
e6848976 ACM |
1201 | } |
1202 | #else /* SOCK_REFCNT_DEBUG */ | |
1203 | #define sk_refcnt_debug_inc(sk) do { } while (0) | |
1204 | #define sk_refcnt_debug_dec(sk) do { } while (0) | |
1205 | #define sk_refcnt_debug_release(sk) do { } while (0) | |
1206 | #endif /* SOCK_REFCNT_DEBUG */ | |
1207 | ||
a74f0fa0 | 1208 | static inline bool __sk_stream_memory_free(const struct sock *sk, int wake) |
c9bee3b7 ED |
1209 | { |
1210 | if (sk->sk_wmem_queued >= sk->sk_sndbuf) | |
1211 | return false; | |
1212 | ||
1213 | return sk->sk_prot->stream_memory_free ? | |
a74f0fa0 | 1214 | sk->sk_prot->stream_memory_free(sk, wake) : true; |
c9bee3b7 ED |
1215 | } |
1216 | ||
a74f0fa0 ED |
1217 | static inline bool sk_stream_memory_free(const struct sock *sk) |
1218 | { | |
1219 | return __sk_stream_memory_free(sk, 0); | |
1220 | } | |
1221 | ||
1222 | static inline bool __sk_stream_is_writeable(const struct sock *sk, int wake) | |
64dc6130 | 1223 | { |
c9bee3b7 | 1224 | return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) && |
a74f0fa0 ED |
1225 | __sk_stream_memory_free(sk, wake); |
1226 | } | |
1227 | ||
1228 | static inline bool sk_stream_is_writeable(const struct sock *sk) | |
1229 | { | |
1230 | return __sk_stream_is_writeable(sk, 0); | |
64dc6130 | 1231 | } |
e1aab161 | 1232 | |
54fd9c2d DB |
1233 | static inline int sk_under_cgroup_hierarchy(struct sock *sk, |
1234 | struct cgroup *ancestor) | |
1235 | { | |
1236 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
1237 | return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data), | |
1238 | ancestor); | |
1239 | #else | |
1240 | return -ENOTSUPP; | |
1241 | #endif | |
1242 | } | |
c9bee3b7 | 1243 | |
180d8cd9 GC |
1244 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
1245 | { | |
1246 | return sk->sk_prot->memory_pressure != NULL; | |
1247 | } | |
1248 | ||
1249 | static inline bool sk_under_memory_pressure(const struct sock *sk) | |
1250 | { | |
1251 | if (!sk->sk_prot->memory_pressure) | |
1252 | return false; | |
e1aab161 | 1253 | |
baac50bb JW |
1254 | if (mem_cgroup_sockets_enabled && sk->sk_memcg && |
1255 | mem_cgroup_under_socket_pressure(sk->sk_memcg)) | |
e805605c | 1256 | return true; |
e1aab161 | 1257 | |
35b87f6c | 1258 | return !!*sk->sk_prot->memory_pressure; |
180d8cd9 GC |
1259 | } |
1260 | ||
180d8cd9 GC |
1261 | static inline long |
1262 | sk_memory_allocated(const struct sock *sk) | |
1263 | { | |
e805605c | 1264 | return atomic_long_read(sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1265 | } |
1266 | ||
1267 | static inline long | |
e805605c | 1268 | sk_memory_allocated_add(struct sock *sk, int amt) |
180d8cd9 | 1269 | { |
e805605c | 1270 | return atomic_long_add_return(amt, sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1271 | } |
1272 | ||
1273 | static inline void | |
0e90b31f | 1274 | sk_memory_allocated_sub(struct sock *sk, int amt) |
180d8cd9 | 1275 | { |
e805605c | 1276 | atomic_long_sub(amt, sk->sk_prot->memory_allocated); |
180d8cd9 GC |
1277 | } |
1278 | ||
1279 | static inline void sk_sockets_allocated_dec(struct sock *sk) | |
1280 | { | |
af95d7df | 1281 | percpu_counter_dec(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1282 | } |
1283 | ||
1284 | static inline void sk_sockets_allocated_inc(struct sock *sk) | |
1285 | { | |
af95d7df | 1286 | percpu_counter_inc(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1287 | } |
1288 | ||
5bf325a5 | 1289 | static inline u64 |
180d8cd9 GC |
1290 | sk_sockets_allocated_read_positive(struct sock *sk) |
1291 | { | |
af95d7df | 1292 | return percpu_counter_read_positive(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1293 | } |
1294 | ||
1295 | static inline int | |
1296 | proto_sockets_allocated_sum_positive(struct proto *prot) | |
1297 | { | |
1298 | return percpu_counter_sum_positive(prot->sockets_allocated); | |
1299 | } | |
1300 | ||
1301 | static inline long | |
1302 | proto_memory_allocated(struct proto *prot) | |
1303 | { | |
1304 | return atomic_long_read(prot->memory_allocated); | |
1305 | } | |
1306 | ||
1307 | static inline bool | |
1308 | proto_memory_pressure(struct proto *prot) | |
1309 | { | |
1310 | if (!prot->memory_pressure) | |
1311 | return false; | |
1312 | return !!*prot->memory_pressure; | |
1313 | } | |
1314 | ||
65f76517 ED |
1315 | |
1316 | #ifdef CONFIG_PROC_FS | |
1da177e4 | 1317 | /* Called with local bh disabled */ |
69336bd2 JP |
1318 | void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc); |
1319 | int sock_prot_inuse_get(struct net *net, struct proto *proto); | |
648845ab | 1320 | int sock_inuse_get(struct net *net); |
65f76517 | 1321 | #else |
dc6b9b78 | 1322 | static inline void sock_prot_inuse_add(struct net *net, struct proto *prot, |
c29a0bc4 | 1323 | int inc) |
65f76517 ED |
1324 | { |
1325 | } | |
65f76517 ED |
1326 | #endif |
1327 | ||
1da177e4 | 1328 | |
614c6cb4 ACM |
1329 | /* With per-bucket locks this operation is not-atomic, so that |
1330 | * this version is not worse. | |
1331 | */ | |
086c653f | 1332 | static inline int __sk_prot_rehash(struct sock *sk) |
614c6cb4 ACM |
1333 | { |
1334 | sk->sk_prot->unhash(sk); | |
086c653f | 1335 | return sk->sk_prot->hash(sk); |
614c6cb4 ACM |
1336 | } |
1337 | ||
1da177e4 LT |
1338 | /* About 10 seconds */ |
1339 | #define SOCK_DESTROY_TIME (10*HZ) | |
1340 | ||
1341 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
1342 | #define PROT_SOCK 1024 | |
1343 | ||
1344 | #define SHUTDOWN_MASK 3 | |
1345 | #define RCV_SHUTDOWN 1 | |
1346 | #define SEND_SHUTDOWN 2 | |
1347 | ||
1348 | #define SOCK_SNDBUF_LOCK 1 | |
1349 | #define SOCK_RCVBUF_LOCK 2 | |
1350 | #define SOCK_BINDADDR_LOCK 4 | |
1351 | #define SOCK_BINDPORT_LOCK 8 | |
1352 | ||
1da177e4 LT |
1353 | struct socket_alloc { |
1354 | struct socket socket; | |
1355 | struct inode vfs_inode; | |
1356 | }; | |
1357 | ||
1358 | static inline struct socket *SOCKET_I(struct inode *inode) | |
1359 | { | |
1360 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
1361 | } | |
1362 | ||
1363 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
1364 | { | |
1365 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
1366 | } | |
1367 | ||
3ab224be HA |
1368 | /* |
1369 | * Functions for memory accounting | |
1370 | */ | |
f8c3bf00 | 1371 | int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind); |
69336bd2 | 1372 | int __sk_mem_schedule(struct sock *sk, int size, int kind); |
f8c3bf00 | 1373 | void __sk_mem_reduce_allocated(struct sock *sk, int amount); |
1a24e04e | 1374 | void __sk_mem_reclaim(struct sock *sk, int amount); |
1da177e4 | 1375 | |
bd68a2a8 ED |
1376 | /* We used to have PAGE_SIZE here, but systems with 64KB pages |
1377 | * do not necessarily have 16x time more memory than 4KB ones. | |
1378 | */ | |
1379 | #define SK_MEM_QUANTUM 4096 | |
3ab224be HA |
1380 | #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM) |
1381 | #define SK_MEM_SEND 0 | |
1382 | #define SK_MEM_RECV 1 | |
1da177e4 | 1383 | |
bd68a2a8 ED |
1384 | /* sysctl_mem values are in pages, we convert them in SK_MEM_QUANTUM units */ |
1385 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) | |
1386 | { | |
1387 | long val = sk->sk_prot->sysctl_mem[index]; | |
1388 | ||
1389 | #if PAGE_SIZE > SK_MEM_QUANTUM | |
1390 | val <<= PAGE_SHIFT - SK_MEM_QUANTUM_SHIFT; | |
1391 | #elif PAGE_SIZE < SK_MEM_QUANTUM | |
1392 | val >>= SK_MEM_QUANTUM_SHIFT - PAGE_SHIFT; | |
1393 | #endif | |
1394 | return val; | |
1395 | } | |
1396 | ||
3ab224be | 1397 | static inline int sk_mem_pages(int amt) |
1da177e4 | 1398 | { |
3ab224be | 1399 | return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT; |
1da177e4 LT |
1400 | } |
1401 | ||
dc6b9b78 | 1402 | static inline bool sk_has_account(struct sock *sk) |
1da177e4 | 1403 | { |
3ab224be HA |
1404 | /* return true if protocol supports memory accounting */ |
1405 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
1406 | } |
1407 | ||
dc6b9b78 | 1408 | static inline bool sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 1409 | { |
3ab224be | 1410 | if (!sk_has_account(sk)) |
dc6b9b78 | 1411 | return true; |
3ab224be HA |
1412 | return size <= sk->sk_forward_alloc || |
1413 | __sk_mem_schedule(sk, size, SK_MEM_SEND); | |
1da177e4 LT |
1414 | } |
1415 | ||
c76562b6 | 1416 | static inline bool |
35c448a8 | 1417 | sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size) |
d80d99d6 | 1418 | { |
3ab224be | 1419 | if (!sk_has_account(sk)) |
dc6b9b78 | 1420 | return true; |
c76562b6 MG |
1421 | return size<= sk->sk_forward_alloc || |
1422 | __sk_mem_schedule(sk, size, SK_MEM_RECV) || | |
1423 | skb_pfmemalloc(skb); | |
3ab224be HA |
1424 | } |
1425 | ||
1426 | static inline void sk_mem_reclaim(struct sock *sk) | |
1427 | { | |
1428 | if (!sk_has_account(sk)) | |
1429 | return; | |
1430 | if (sk->sk_forward_alloc >= SK_MEM_QUANTUM) | |
1a24e04e | 1431 | __sk_mem_reclaim(sk, sk->sk_forward_alloc); |
3ab224be HA |
1432 | } |
1433 | ||
9993e7d3 DM |
1434 | static inline void sk_mem_reclaim_partial(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 - 1); |
9993e7d3 DM |
1440 | } |
1441 | ||
3ab224be HA |
1442 | static inline void sk_mem_charge(struct sock *sk, int size) |
1443 | { | |
1444 | if (!sk_has_account(sk)) | |
1445 | return; | |
1446 | sk->sk_forward_alloc -= size; | |
1447 | } | |
1448 | ||
1449 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
1450 | { | |
1451 | if (!sk_has_account(sk)) | |
1452 | return; | |
1453 | sk->sk_forward_alloc += size; | |
20c64d5c ED |
1454 | |
1455 | /* Avoid a possible overflow. | |
1456 | * TCP send queues can make this happen, if sk_mem_reclaim() | |
1457 | * is not called and more than 2 GBytes are released at once. | |
1458 | * | |
1459 | * If we reach 2 MBytes, reclaim 1 MBytes right now, there is | |
1460 | * no need to hold that much forward allocation anyway. | |
1461 | */ | |
1462 | if (unlikely(sk->sk_forward_alloc >= 1 << 21)) | |
1463 | __sk_mem_reclaim(sk, 1 << 20); | |
3ab224be HA |
1464 | } |
1465 | ||
0b7d7f6b | 1466 | DECLARE_STATIC_KEY_FALSE(tcp_tx_skb_cache_key); |
3ab224be HA |
1467 | static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb) |
1468 | { | |
4f661542 ED |
1469 | sock_set_flag(sk, SOCK_QUEUE_SHRUNK); |
1470 | sk->sk_wmem_queued -= skb->truesize; | |
1471 | sk_mem_uncharge(sk, skb->truesize); | |
0b7d7f6b ED |
1472 | if (static_branch_unlikely(&tcp_tx_skb_cache_key) && |
1473 | !sk->sk_tx_skb_cache && !skb_cloned(skb)) { | |
4f661542 | 1474 | skb_zcopy_clear(skb, true); |
472c2e07 ED |
1475 | sk->sk_tx_skb_cache = skb; |
1476 | return; | |
1477 | } | |
3ab224be | 1478 | __kfree_skb(skb); |
d80d99d6 HX |
1479 | } |
1480 | ||
c3f9b018 ED |
1481 | static inline void sock_release_ownership(struct sock *sk) |
1482 | { | |
61881cfb HFS |
1483 | if (sk->sk_lock.owned) { |
1484 | sk->sk_lock.owned = 0; | |
1485 | ||
1486 | /* The sk_lock has mutex_unlock() semantics: */ | |
1487 | mutex_release(&sk->sk_lock.dep_map, 1, _RET_IP_); | |
1488 | } | |
c3f9b018 ED |
1489 | } |
1490 | ||
ed07536e PZ |
1491 | /* |
1492 | * Macro so as to not evaluate some arguments when | |
1493 | * lockdep is not enabled. | |
1494 | * | |
1495 | * Mark both the sk_lock and the sk_lock.slock as a | |
1496 | * per-address-family lock class. | |
1497 | */ | |
dc6b9b78 | 1498 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
ed07536e | 1499 | do { \ |
e8f6fbf6 | 1500 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
1501 | init_waitqueue_head(&sk->sk_lock.wq); \ |
1502 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
1503 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
1504 | sizeof((sk)->sk_lock)); \ | |
1505 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
dc6b9b78 | 1506 | (skey), (sname)); \ |
ed07536e PZ |
1507 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
1508 | } while (0) | |
1509 | ||
b33b0a1b | 1510 | #ifdef CONFIG_LOCKDEP |
05b93801 | 1511 | static inline bool lockdep_sock_is_held(const struct sock *sk) |
1e1d04e6 | 1512 | { |
1e1d04e6 HFS |
1513 | return lockdep_is_held(&sk->sk_lock) || |
1514 | lockdep_is_held(&sk->sk_lock.slock); | |
1515 | } | |
b33b0a1b | 1516 | #endif |
1e1d04e6 | 1517 | |
69336bd2 | 1518 | void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
1519 | |
1520 | static inline void lock_sock(struct sock *sk) | |
1521 | { | |
1522 | lock_sock_nested(sk, 0); | |
1523 | } | |
1524 | ||
8873c064 | 1525 | void __release_sock(struct sock *sk); |
69336bd2 | 1526 | void release_sock(struct sock *sk); |
1da177e4 LT |
1527 | |
1528 | /* BH context may only use the following locking interface. */ | |
1529 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
1530 | #define bh_lock_sock_nested(__sk) \ |
1531 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
1532 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
1533 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
1534 | ||
69336bd2 | 1535 | bool lock_sock_fast(struct sock *sk); |
8a74ad60 ED |
1536 | /** |
1537 | * unlock_sock_fast - complement of lock_sock_fast | |
1538 | * @sk: socket | |
1539 | * @slow: slow mode | |
1540 | * | |
1541 | * fast unlock socket for user context. | |
1542 | * If slow mode is on, we call regular release_sock() | |
1543 | */ | |
1544 | static inline void unlock_sock_fast(struct sock *sk, bool slow) | |
4b0b72f7 | 1545 | { |
8a74ad60 ED |
1546 | if (slow) |
1547 | release_sock(sk); | |
1548 | else | |
1549 | spin_unlock_bh(&sk->sk_lock.slock); | |
4b0b72f7 ED |
1550 | } |
1551 | ||
fafc4e1e HFS |
1552 | /* Used by processes to "lock" a socket state, so that |
1553 | * interrupts and bottom half handlers won't change it | |
1554 | * from under us. It essentially blocks any incoming | |
1555 | * packets, so that we won't get any new data or any | |
1556 | * packets that change the state of the socket. | |
1557 | * | |
1558 | * While locked, BH processing will add new packets to | |
1559 | * the backlog queue. This queue is processed by the | |
1560 | * owner of the socket lock right before it is released. | |
1561 | * | |
1562 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
1563 | * accesses from user process context. | |
1564 | */ | |
1565 | ||
46cc6e49 | 1566 | static inline void sock_owned_by_me(const struct sock *sk) |
fafc4e1e HFS |
1567 | { |
1568 | #ifdef CONFIG_LOCKDEP | |
5e91f6ce | 1569 | WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks); |
fafc4e1e | 1570 | #endif |
46cc6e49 ED |
1571 | } |
1572 | ||
1573 | static inline bool sock_owned_by_user(const struct sock *sk) | |
1574 | { | |
1575 | sock_owned_by_me(sk); | |
fafc4e1e HFS |
1576 | return sk->sk_lock.owned; |
1577 | } | |
1578 | ||
602f7a27 TH |
1579 | static inline bool sock_owned_by_user_nocheck(const struct sock *sk) |
1580 | { | |
1581 | return sk->sk_lock.owned; | |
1582 | } | |
1583 | ||
fafc4e1e HFS |
1584 | /* no reclassification while locks are held */ |
1585 | static inline bool sock_allow_reclassification(const struct sock *csk) | |
1586 | { | |
1587 | struct sock *sk = (struct sock *)csk; | |
1588 | ||
1589 | return !sk->sk_lock.owned && !spin_is_locked(&sk->sk_lock.slock); | |
1590 | } | |
4b0b72f7 | 1591 | |
69336bd2 | 1592 | struct sock *sk_alloc(struct net *net, int family, gfp_t priority, |
11aa9c28 | 1593 | struct proto *prot, int kern); |
69336bd2 | 1594 | void sk_free(struct sock *sk); |
eb4cb008 | 1595 | void sk_destruct(struct sock *sk); |
69336bd2 | 1596 | struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority); |
94352d45 | 1597 | void sk_free_unlock_clone(struct sock *sk); |
69336bd2 JP |
1598 | |
1599 | struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force, | |
1600 | gfp_t priority); | |
1d2077ac | 1601 | void __sock_wfree(struct sk_buff *skb); |
69336bd2 | 1602 | void sock_wfree(struct sk_buff *skb); |
98ba0bd5 WB |
1603 | struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size, |
1604 | gfp_t priority); | |
69336bd2 JP |
1605 | void skb_orphan_partial(struct sk_buff *skb); |
1606 | void sock_rfree(struct sk_buff *skb); | |
62bccb8c | 1607 | void sock_efree(struct sk_buff *skb); |
82eabd9e | 1608 | #ifdef CONFIG_INET |
69336bd2 | 1609 | void sock_edemux(struct sk_buff *skb); |
82eabd9e | 1610 | #else |
158f323b | 1611 | #define sock_edemux sock_efree |
82eabd9e | 1612 | #endif |
69336bd2 JP |
1613 | |
1614 | int sock_setsockopt(struct socket *sock, int level, int op, | |
1615 | char __user *optval, unsigned int optlen); | |
1616 | ||
1617 | int sock_getsockopt(struct socket *sock, int level, int op, | |
1618 | char __user *optval, int __user *optlen); | |
c7cbdbf2 AB |
1619 | int sock_gettstamp(struct socket *sock, void __user *userstamp, |
1620 | bool timeval, bool time32); | |
69336bd2 JP |
1621 | struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size, |
1622 | int noblock, int *errcode); | |
1623 | struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, | |
1624 | unsigned long data_len, int noblock, | |
1625 | int *errcode, int max_page_order); | |
1626 | void *sock_kmalloc(struct sock *sk, int size, gfp_t priority); | |
1627 | void sock_kfree_s(struct sock *sk, void *mem, int size); | |
79e88659 | 1628 | void sock_kzfree_s(struct sock *sk, void *mem, int size); |
69336bd2 | 1629 | void sk_send_sigurg(struct sock *sk); |
1da177e4 | 1630 | |
f28ea365 | 1631 | struct sockcm_cookie { |
80b14dee | 1632 | u64 transmit_time; |
f28ea365 | 1633 | u32 mark; |
3dd17e63 | 1634 | u16 tsflags; |
f28ea365 EJ |
1635 | }; |
1636 | ||
657a0667 WB |
1637 | static inline void sockcm_init(struct sockcm_cookie *sockc, |
1638 | const struct sock *sk) | |
1639 | { | |
1640 | *sockc = (struct sockcm_cookie) { .tsflags = sk->sk_tsflags }; | |
1641 | } | |
1642 | ||
39771b12 WB |
1643 | int __sock_cmsg_send(struct sock *sk, struct msghdr *msg, struct cmsghdr *cmsg, |
1644 | struct sockcm_cookie *sockc); | |
f28ea365 EJ |
1645 | int sock_cmsg_send(struct sock *sk, struct msghdr *msg, |
1646 | struct sockcm_cookie *sockc); | |
1647 | ||
1da177e4 LT |
1648 | /* |
1649 | * Functions to fill in entries in struct proto_ops when a protocol | |
1650 | * does not implement a particular function. | |
1651 | */ | |
69336bd2 JP |
1652 | int sock_no_bind(struct socket *, struct sockaddr *, int); |
1653 | int sock_no_connect(struct socket *, struct sockaddr *, int, int); | |
1654 | int sock_no_socketpair(struct socket *, struct socket *); | |
cdfbabfb | 1655 | int sock_no_accept(struct socket *, struct socket *, int, bool); |
9b2c45d4 | 1656 | int sock_no_getname(struct socket *, struct sockaddr *, int); |
69336bd2 JP |
1657 | int sock_no_ioctl(struct socket *, unsigned int, unsigned long); |
1658 | int sock_no_listen(struct socket *, int); | |
1659 | int sock_no_shutdown(struct socket *, int); | |
1660 | int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *); | |
1661 | int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int); | |
1b784140 | 1662 | int sock_no_sendmsg(struct socket *, struct msghdr *, size_t); |
306b13eb | 1663 | int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len); |
1b784140 | 1664 | int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int); |
69336bd2 JP |
1665 | int sock_no_mmap(struct file *file, struct socket *sock, |
1666 | struct vm_area_struct *vma); | |
1667 | ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset, | |
1668 | size_t size, int flags); | |
306b13eb TH |
1669 | ssize_t sock_no_sendpage_locked(struct sock *sk, struct page *page, |
1670 | int offset, size_t size, int flags); | |
1da177e4 LT |
1671 | |
1672 | /* | |
1673 | * Functions to fill in entries in struct proto_ops when a protocol | |
1674 | * uses the inet style. | |
1675 | */ | |
69336bd2 | 1676 | int sock_common_getsockopt(struct socket *sock, int level, int optname, |
1da177e4 | 1677 | char __user *optval, int __user *optlen); |
1b784140 YX |
1678 | int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
1679 | int flags); | |
69336bd2 | 1680 | int sock_common_setsockopt(struct socket *sock, int level, int optname, |
b7058842 | 1681 | char __user *optval, unsigned int optlen); |
69336bd2 | 1682 | int compat_sock_common_getsockopt(struct socket *sock, int level, |
3fdadf7d | 1683 | int optname, char __user *optval, int __user *optlen); |
69336bd2 | 1684 | int compat_sock_common_setsockopt(struct socket *sock, int level, |
b7058842 | 1685 | int optname, char __user *optval, unsigned int optlen); |
1da177e4 | 1686 | |
69336bd2 | 1687 | void sk_common_release(struct sock *sk); |
1da177e4 LT |
1688 | |
1689 | /* | |
1690 | * Default socket callbacks and setup code | |
1691 | */ | |
dc6b9b78 | 1692 | |
1da177e4 | 1693 | /* Initialise core socket variables */ |
69336bd2 | 1694 | void sock_init_data(struct socket *sock, struct sock *sk); |
1da177e4 | 1695 | |
1da177e4 LT |
1696 | /* |
1697 | * Socket reference counting postulates. | |
1698 | * | |
1699 | * * Each user of socket SHOULD hold a reference count. | |
1700 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1701 | * running timer, skb in flight MUST hold a reference count. | |
1702 | * * When reference count hits 0, it means it will never increase back. | |
1703 | * * When reference count hits 0, it means that no references from | |
1704 | * outside exist to this socket and current process on current CPU | |
1705 | * is last user and may/should destroy this socket. | |
1706 | * * sk_free is called from any context: process, BH, IRQ. When | |
1707 | * it is called, socket has no references from outside -> sk_free | |
1708 | * may release descendant resources allocated by the socket, but | |
1709 | * to the time when it is called, socket is NOT referenced by any | |
1710 | * hash tables, lists etc. | |
1711 | * * Packets, delivered from outside (from network or from another process) | |
1712 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1713 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1714 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1715 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1716 | * (leak to backlog). Packet socket does all the processing inside | |
1717 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1718 | * use separate SMP lock, so that they are prone too. | |
1719 | */ | |
1720 | ||
1721 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1722 | static inline void sock_put(struct sock *sk) | |
1723 | { | |
41c6d650 | 1724 | if (refcount_dec_and_test(&sk->sk_refcnt)) |
1da177e4 LT |
1725 | sk_free(sk); |
1726 | } | |
05dbc7b5 | 1727 | /* Generic version of sock_put(), dealing with all sockets |
41b822c5 | 1728 | * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...) |
05dbc7b5 ED |
1729 | */ |
1730 | void sock_gen_put(struct sock *sk); | |
1da177e4 | 1731 | |
4f0c40d9 | 1732 | int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested, |
c3f24cfb | 1733 | unsigned int trim_cap, bool refcounted); |
4f0c40d9 WB |
1734 | static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
1735 | const int nested) | |
1736 | { | |
c3f24cfb | 1737 | return __sk_receive_skb(sk, skb, nested, 1, true); |
4f0c40d9 | 1738 | } |
25995ff5 | 1739 | |
e022f0b4 KK |
1740 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
1741 | { | |
755c31cd AN |
1742 | /* sk_tx_queue_mapping accept only upto a 16-bit value */ |
1743 | if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX)) | |
1744 | return; | |
e022f0b4 KK |
1745 | sk->sk_tx_queue_mapping = tx_queue; |
1746 | } | |
1747 | ||
755c31cd AN |
1748 | #define NO_QUEUE_MAPPING USHRT_MAX |
1749 | ||
e022f0b4 KK |
1750 | static inline void sk_tx_queue_clear(struct sock *sk) |
1751 | { | |
755c31cd | 1752 | sk->sk_tx_queue_mapping = NO_QUEUE_MAPPING; |
e022f0b4 KK |
1753 | } |
1754 | ||
1755 | static inline int sk_tx_queue_get(const struct sock *sk) | |
1756 | { | |
755c31cd AN |
1757 | if (sk && sk->sk_tx_queue_mapping != NO_QUEUE_MAPPING) |
1758 | return sk->sk_tx_queue_mapping; | |
1759 | ||
1760 | return -1; | |
e022f0b4 KK |
1761 | } |
1762 | ||
c6345ce7 AN |
1763 | static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb) |
1764 | { | |
1765 | #ifdef CONFIG_XPS | |
1766 | if (skb_rx_queue_recorded(skb)) { | |
1767 | u16 rx_queue = skb_get_rx_queue(skb); | |
1768 | ||
1769 | if (WARN_ON_ONCE(rx_queue == NO_QUEUE_MAPPING)) | |
1770 | return; | |
1771 | ||
1772 | sk->sk_rx_queue_mapping = rx_queue; | |
1773 | } | |
1774 | #endif | |
1775 | } | |
1776 | ||
1777 | static inline void sk_rx_queue_clear(struct sock *sk) | |
1778 | { | |
1779 | #ifdef CONFIG_XPS | |
1780 | sk->sk_rx_queue_mapping = NO_QUEUE_MAPPING; | |
1781 | #endif | |
1782 | } | |
1783 | ||
fc9bab24 AN |
1784 | #ifdef CONFIG_XPS |
1785 | static inline int sk_rx_queue_get(const struct sock *sk) | |
1786 | { | |
1787 | if (sk && sk->sk_rx_queue_mapping != NO_QUEUE_MAPPING) | |
1788 | return sk->sk_rx_queue_mapping; | |
1789 | ||
1790 | return -1; | |
1791 | } | |
1792 | #endif | |
1793 | ||
972692e0 DM |
1794 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
1795 | { | |
e022f0b4 | 1796 | sk_tx_queue_clear(sk); |
972692e0 DM |
1797 | sk->sk_socket = sock; |
1798 | } | |
1799 | ||
aa395145 ED |
1800 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
1801 | { | |
eaefd110 ED |
1802 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
1803 | return &rcu_dereference_raw(sk->sk_wq)->wait; | |
aa395145 | 1804 | } |
1da177e4 LT |
1805 | /* Detach socket from process context. |
1806 | * Announce socket dead, detach it from wait queue and inode. | |
1807 | * Note that parent inode held reference count on this struct sock, | |
1808 | * we do not release it in this function, because protocol | |
1809 | * probably wants some additional cleanups or even continuing | |
1810 | * to work with this socket (TCP). | |
1811 | */ | |
1812 | static inline void sock_orphan(struct sock *sk) | |
1813 | { | |
1814 | write_lock_bh(&sk->sk_callback_lock); | |
1815 | sock_set_flag(sk, SOCK_DEAD); | |
972692e0 | 1816 | sk_set_socket(sk, NULL); |
43815482 | 1817 | sk->sk_wq = NULL; |
1da177e4 LT |
1818 | write_unlock_bh(&sk->sk_callback_lock); |
1819 | } | |
1820 | ||
1821 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
1822 | { | |
0ffdaf5b | 1823 | WARN_ON(parent->sk); |
1da177e4 | 1824 | write_lock_bh(&sk->sk_callback_lock); |
333f7909 | 1825 | rcu_assign_pointer(sk->sk_wq, &parent->wq); |
1da177e4 | 1826 | parent->sk = sk; |
972692e0 | 1827 | sk_set_socket(sk, parent); |
86741ec2 | 1828 | sk->sk_uid = SOCK_INODE(parent)->i_uid; |
4237c75c | 1829 | security_sock_graft(sk, parent); |
1da177e4 LT |
1830 | write_unlock_bh(&sk->sk_callback_lock); |
1831 | } | |
1832 | ||
69336bd2 JP |
1833 | kuid_t sock_i_uid(struct sock *sk); |
1834 | unsigned long sock_i_ino(struct sock *sk); | |
1da177e4 | 1835 | |
86741ec2 LC |
1836 | static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk) |
1837 | { | |
1838 | return sk ? sk->sk_uid : make_kuid(net->user_ns, 0); | |
1839 | } | |
1840 | ||
58d607d3 | 1841 | static inline u32 net_tx_rndhash(void) |
877d1f62 | 1842 | { |
58d607d3 ED |
1843 | u32 v = prandom_u32(); |
1844 | ||
1845 | return v ?: 1; | |
1846 | } | |
877d1f62 | 1847 | |
58d607d3 ED |
1848 | static inline void sk_set_txhash(struct sock *sk) |
1849 | { | |
1850 | sk->sk_txhash = net_tx_rndhash(); | |
877d1f62 TH |
1851 | } |
1852 | ||
265f94ff TH |
1853 | static inline void sk_rethink_txhash(struct sock *sk) |
1854 | { | |
1855 | if (sk->sk_txhash) | |
1856 | sk_set_txhash(sk); | |
1857 | } | |
1858 | ||
1da177e4 LT |
1859 | static inline struct dst_entry * |
1860 | __sk_dst_get(struct sock *sk) | |
1861 | { | |
1e1d04e6 HFS |
1862 | return rcu_dereference_check(sk->sk_dst_cache, |
1863 | lockdep_sock_is_held(sk)); | |
1da177e4 LT |
1864 | } |
1865 | ||
1866 | static inline struct dst_entry * | |
1867 | sk_dst_get(struct sock *sk) | |
1868 | { | |
1869 | struct dst_entry *dst; | |
1870 | ||
b6c6712a ED |
1871 | rcu_read_lock(); |
1872 | dst = rcu_dereference(sk->sk_dst_cache); | |
f8864972 ED |
1873 | if (dst && !atomic_inc_not_zero(&dst->__refcnt)) |
1874 | dst = NULL; | |
b6c6712a | 1875 | rcu_read_unlock(); |
1da177e4 LT |
1876 | return dst; |
1877 | } | |
1878 | ||
b6c6712a ED |
1879 | static inline void dst_negative_advice(struct sock *sk) |
1880 | { | |
1881 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); | |
1882 | ||
265f94ff TH |
1883 | sk_rethink_txhash(sk); |
1884 | ||
b6c6712a ED |
1885 | if (dst && dst->ops->negative_advice) { |
1886 | ndst = dst->ops->negative_advice(dst); | |
1887 | ||
1888 | if (ndst != dst) { | |
1889 | rcu_assign_pointer(sk->sk_dst_cache, ndst); | |
0a6957e7 | 1890 | sk_tx_queue_clear(sk); |
9b8805a3 | 1891 | sk->sk_dst_pending_confirm = 0; |
b6c6712a ED |
1892 | } |
1893 | } | |
1894 | } | |
1895 | ||
1da177e4 LT |
1896 | static inline void |
1897 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1898 | { | |
1899 | struct dst_entry *old_dst; | |
1900 | ||
e022f0b4 | 1901 | sk_tx_queue_clear(sk); |
9b8805a3 | 1902 | sk->sk_dst_pending_confirm = 0; |
95964c6d ED |
1903 | old_dst = rcu_dereference_protected(sk->sk_dst_cache, |
1904 | lockdep_sock_is_held(sk)); | |
b6c6712a | 1905 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
1da177e4 LT |
1906 | dst_release(old_dst); |
1907 | } | |
1908 | ||
1909 | static inline void | |
1910 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
1911 | { | |
7f502361 ED |
1912 | struct dst_entry *old_dst; |
1913 | ||
1914 | sk_tx_queue_clear(sk); | |
9b8805a3 | 1915 | sk->sk_dst_pending_confirm = 0; |
5925a055 | 1916 | old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst); |
7f502361 | 1917 | dst_release(old_dst); |
1da177e4 LT |
1918 | } |
1919 | ||
1920 | static inline void | |
1921 | __sk_dst_reset(struct sock *sk) | |
1922 | { | |
b6c6712a | 1923 | __sk_dst_set(sk, NULL); |
1da177e4 LT |
1924 | } |
1925 | ||
1926 | static inline void | |
1927 | sk_dst_reset(struct sock *sk) | |
1928 | { | |
7f502361 | 1929 | sk_dst_set(sk, NULL); |
1da177e4 LT |
1930 | } |
1931 | ||
69336bd2 | 1932 | struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1933 | |
69336bd2 | 1934 | struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 1935 | |
9b8805a3 JA |
1936 | static inline void sk_dst_confirm(struct sock *sk) |
1937 | { | |
1938 | if (!sk->sk_dst_pending_confirm) | |
1939 | sk->sk_dst_pending_confirm = 1; | |
1940 | } | |
1941 | ||
4ff06203 JA |
1942 | static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n) |
1943 | { | |
1944 | if (skb_get_dst_pending_confirm(skb)) { | |
1945 | struct sock *sk = skb->sk; | |
1946 | unsigned long now = jiffies; | |
1947 | ||
1948 | /* avoid dirtying neighbour */ | |
1949 | if (n->confirmed != now) | |
1950 | n->confirmed = now; | |
1951 | if (sk && sk->sk_dst_pending_confirm) | |
1952 | sk->sk_dst_pending_confirm = 0; | |
1953 | } | |
1954 | } | |
1955 | ||
f60e5990 | 1956 | bool sk_mc_loop(struct sock *sk); |
1957 | ||
dc6b9b78 | 1958 | static inline bool sk_can_gso(const struct sock *sk) |
bcd76111 HX |
1959 | { |
1960 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
1961 | } | |
1962 | ||
69336bd2 | 1963 | void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 1964 | |
c8f44aff | 1965 | static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags) |
a465419b ED |
1966 | { |
1967 | sk->sk_route_nocaps |= flags; | |
1968 | sk->sk_route_caps &= ~flags; | |
1969 | } | |
1970 | ||
c6e1a0d1 | 1971 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
57be5bda | 1972 | struct iov_iter *from, char *to, |
912d398d | 1973 | int copy, int offset) |
c6e1a0d1 TH |
1974 | { |
1975 | if (skb->ip_summed == CHECKSUM_NONE) { | |
57be5bda | 1976 | __wsum csum = 0; |
15e6cb46 | 1977 | if (!csum_and_copy_from_iter_full(to, copy, &csum, from)) |
57be5bda | 1978 | return -EFAULT; |
912d398d | 1979 | skb->csum = csum_block_add(skb->csum, csum, offset); |
c6e1a0d1 | 1980 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
15e6cb46 | 1981 | if (!copy_from_iter_full_nocache(to, copy, from)) |
c6e1a0d1 | 1982 | return -EFAULT; |
15e6cb46 | 1983 | } else if (!copy_from_iter_full(to, copy, from)) |
c6e1a0d1 TH |
1984 | return -EFAULT; |
1985 | ||
1986 | return 0; | |
1987 | } | |
1988 | ||
1989 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, | |
57be5bda | 1990 | struct iov_iter *from, int copy) |
c6e1a0d1 | 1991 | { |
912d398d | 1992 | int err, offset = skb->len; |
c6e1a0d1 | 1993 | |
912d398d WY |
1994 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
1995 | copy, offset); | |
c6e1a0d1 | 1996 | if (err) |
912d398d | 1997 | __skb_trim(skb, offset); |
c6e1a0d1 TH |
1998 | |
1999 | return err; | |
2000 | } | |
2001 | ||
57be5bda | 2002 | static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from, |
c6e1a0d1 TH |
2003 | struct sk_buff *skb, |
2004 | struct page *page, | |
2005 | int off, int copy) | |
2006 | { | |
2007 | int err; | |
2008 | ||
912d398d WY |
2009 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
2010 | copy, skb->len); | |
c6e1a0d1 TH |
2011 | if (err) |
2012 | return err; | |
2013 | ||
2014 | skb->len += copy; | |
2015 | skb->data_len += copy; | |
2016 | skb->truesize += copy; | |
2017 | sk->sk_wmem_queued += copy; | |
2018 | sk_mem_charge(sk, copy); | |
2019 | return 0; | |
2020 | } | |
2021 | ||
c564039f ED |
2022 | /** |
2023 | * sk_wmem_alloc_get - returns write allocations | |
2024 | * @sk: socket | |
2025 | * | |
2026 | * Returns sk_wmem_alloc minus initial offset of one | |
2027 | */ | |
2028 | static inline int sk_wmem_alloc_get(const struct sock *sk) | |
2029 | { | |
14afee4b | 2030 | return refcount_read(&sk->sk_wmem_alloc) - 1; |
c564039f ED |
2031 | } |
2032 | ||
2033 | /** | |
2034 | * sk_rmem_alloc_get - returns read allocations | |
2035 | * @sk: socket | |
2036 | * | |
2037 | * Returns sk_rmem_alloc | |
2038 | */ | |
2039 | static inline int sk_rmem_alloc_get(const struct sock *sk) | |
2040 | { | |
2041 | return atomic_read(&sk->sk_rmem_alloc); | |
2042 | } | |
2043 | ||
2044 | /** | |
2045 | * sk_has_allocations - check if allocations are outstanding | |
2046 | * @sk: socket | |
2047 | * | |
2048 | * Returns true if socket has write or read allocations | |
2049 | */ | |
dc6b9b78 | 2050 | static inline bool sk_has_allocations(const struct sock *sk) |
c564039f ED |
2051 | { |
2052 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); | |
2053 | } | |
2054 | ||
a57de0b4 | 2055 | /** |
1ce0bf50 | 2056 | * skwq_has_sleeper - check if there are any waiting processes |
acfbe96a | 2057 | * @wq: struct socket_wq |
a57de0b4 | 2058 | * |
43815482 | 2059 | * Returns true if socket_wq has waiting processes |
a57de0b4 | 2060 | * |
1ce0bf50 | 2061 | * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory |
a57de0b4 JO |
2062 | * barrier call. They were added due to the race found within the tcp code. |
2063 | * | |
d651983d | 2064 | * Consider following tcp code paths:: |
a57de0b4 | 2065 | * |
d651983d MCC |
2066 | * CPU1 CPU2 |
2067 | * sys_select receive packet | |
a57de0b4 JO |
2068 | * ... ... |
2069 | * __add_wait_queue update tp->rcv_nxt | |
2070 | * ... ... | |
2071 | * tp->rcv_nxt check sock_def_readable | |
2072 | * ... { | |
43815482 ED |
2073 | * schedule rcu_read_lock(); |
2074 | * wq = rcu_dereference(sk->sk_wq); | |
2075 | * if (wq && waitqueue_active(&wq->wait)) | |
2076 | * wake_up_interruptible(&wq->wait) | |
a57de0b4 JO |
2077 | * ... |
2078 | * } | |
2079 | * | |
2080 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay | |
2081 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 | |
2082 | * could then endup calling schedule and sleep forever if there are no more | |
2083 | * data on the socket. | |
ad462769 | 2084 | * |
a57de0b4 | 2085 | */ |
1ce0bf50 | 2086 | static inline bool skwq_has_sleeper(struct socket_wq *wq) |
a57de0b4 | 2087 | { |
1ce0bf50 | 2088 | return wq && wq_has_sleeper(&wq->wait); |
a57de0b4 JO |
2089 | } |
2090 | ||
2091 | /** | |
2092 | * sock_poll_wait - place memory barrier behind the poll_wait call. | |
2093 | * @filp: file | |
89ab066d | 2094 | * @sock: socket to wait on |
a57de0b4 JO |
2095 | * @p: poll_table |
2096 | * | |
43815482 | 2097 | * See the comments in the wq_has_sleeper function. |
a57de0b4 | 2098 | */ |
89ab066d KG |
2099 | static inline void sock_poll_wait(struct file *filp, struct socket *sock, |
2100 | poll_table *p) | |
a57de0b4 | 2101 | { |
d8bbd13b | 2102 | if (!poll_does_not_wait(p)) { |
333f7909 | 2103 | poll_wait(filp, &sock->wq.wait, p); |
dc6b9b78 | 2104 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
2105 | * socket flags modification. |
2106 | * | |
43815482 | 2107 | * This memory barrier is paired in the wq_has_sleeper. |
dc6b9b78 | 2108 | */ |
a57de0b4 JO |
2109 | smp_mb(); |
2110 | } | |
2111 | } | |
2112 | ||
b73c3d0e TH |
2113 | static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk) |
2114 | { | |
2115 | if (sk->sk_txhash) { | |
2116 | skb->l4_hash = 1; | |
2117 | skb->hash = sk->sk_txhash; | |
2118 | } | |
2119 | } | |
2120 | ||
9e17f8a4 ED |
2121 | void skb_set_owner_w(struct sk_buff *skb, struct sock *sk); |
2122 | ||
1da177e4 | 2123 | /* |
dc6b9b78 | 2124 | * Queue a received datagram if it will fit. Stream and sequenced |
1da177e4 LT |
2125 | * protocols can't normally use this as they need to fit buffers in |
2126 | * and play with them. | |
2127 | * | |
dc6b9b78 | 2128 | * Inlined as it's very short and called for pretty much every |
1da177e4 LT |
2129 | * packet ever received. |
2130 | */ | |
1da177e4 LT |
2131 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
2132 | { | |
d55d87fd | 2133 | skb_orphan(skb); |
1da177e4 LT |
2134 | skb->sk = sk; |
2135 | skb->destructor = sock_rfree; | |
2136 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 2137 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
2138 | } |
2139 | ||
69336bd2 JP |
2140 | void sk_reset_timer(struct sock *sk, struct timer_list *timer, |
2141 | unsigned long expires); | |
1da177e4 | 2142 | |
69336bd2 | 2143 | void sk_stop_timer(struct sock *sk, struct timer_list *timer); |
1da177e4 | 2144 | |
65101aec PA |
2145 | int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue, |
2146 | struct sk_buff *skb, unsigned int flags, | |
69629464 ED |
2147 | void (*destructor)(struct sock *sk, |
2148 | struct sk_buff *skb)); | |
e6afc8ac | 2149 | int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
69336bd2 | 2150 | int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
1da177e4 | 2151 | |
69336bd2 | 2152 | int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
364a9e93 | 2153 | struct sk_buff *sock_dequeue_err_skb(struct sock *sk); |
1da177e4 LT |
2154 | |
2155 | /* | |
2156 | * Recover an error report and clear atomically | |
2157 | */ | |
dc6b9b78 | 2158 | |
1da177e4 LT |
2159 | static inline int sock_error(struct sock *sk) |
2160 | { | |
c1cbe4b7 BL |
2161 | int err; |
2162 | if (likely(!sk->sk_err)) | |
2163 | return 0; | |
2164 | err = xchg(&sk->sk_err, 0); | |
1da177e4 LT |
2165 | return -err; |
2166 | } | |
2167 | ||
2168 | static inline unsigned long sock_wspace(struct sock *sk) | |
2169 | { | |
2170 | int amt = 0; | |
2171 | ||
2172 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
14afee4b | 2173 | amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc); |
dc6b9b78 | 2174 | if (amt < 0) |
1da177e4 LT |
2175 | amt = 0; |
2176 | } | |
2177 | return amt; | |
2178 | } | |
2179 | ||
ceb5d58b ED |
2180 | /* Note: |
2181 | * We use sk->sk_wq_raw, from contexts knowing this | |
2182 | * pointer is not NULL and cannot disappear/change. | |
2183 | */ | |
9cd3e072 | 2184 | static inline void sk_set_bit(int nr, struct sock *sk) |
1da177e4 | 2185 | { |
4be73522 ED |
2186 | if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) && |
2187 | !sock_flag(sk, SOCK_FASYNC)) | |
9317bb69 ED |
2188 | return; |
2189 | ||
ceb5d58b | 2190 | set_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
2191 | } |
2192 | ||
2193 | static inline void sk_clear_bit(int nr, struct sock *sk) | |
2194 | { | |
4be73522 ED |
2195 | if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) && |
2196 | !sock_flag(sk, SOCK_FASYNC)) | |
9317bb69 ED |
2197 | return; |
2198 | ||
ceb5d58b | 2199 | clear_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
2200 | } |
2201 | ||
ceb5d58b | 2202 | static inline void sk_wake_async(const struct sock *sk, int how, int band) |
1da177e4 | 2203 | { |
ceb5d58b ED |
2204 | if (sock_flag(sk, SOCK_FASYNC)) { |
2205 | rcu_read_lock(); | |
2206 | sock_wake_async(rcu_dereference(sk->sk_wq), how, band); | |
2207 | rcu_read_unlock(); | |
2208 | } | |
1da177e4 LT |
2209 | } |
2210 | ||
eea86af6 DB |
2211 | /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might |
2212 | * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak. | |
2213 | * Note: for send buffers, TCP works better if we can build two skbs at | |
2214 | * minimum. | |
7a91b434 | 2215 | */ |
9eb5bf83 | 2216 | #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff))) |
eea86af6 DB |
2217 | |
2218 | #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2) | |
2219 | #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE | |
1da177e4 LT |
2220 | |
2221 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
2222 | { | |
2223 | if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) { | |
8df09ea3 | 2224 | sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); |
eea86af6 | 2225 | sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF); |
1da177e4 LT |
2226 | } |
2227 | } | |
2228 | ||
eb934478 ED |
2229 | struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp, |
2230 | bool force_schedule); | |
1da177e4 | 2231 | |
5640f768 ED |
2232 | /** |
2233 | * sk_page_frag - return an appropriate page_frag | |
2234 | * @sk: socket | |
2235 | * | |
2236 | * If socket allocation mode allows current thread to sleep, it means its | |
2237 | * safe to use the per task page_frag instead of the per socket one. | |
2238 | */ | |
2239 | static inline struct page_frag *sk_page_frag(struct sock *sk) | |
1da177e4 | 2240 | { |
d0164adc | 2241 | if (gfpflags_allow_blocking(sk->sk_allocation)) |
5640f768 | 2242 | return ¤t->task_frag; |
1da177e4 | 2243 | |
5640f768 | 2244 | return &sk->sk_frag; |
1da177e4 LT |
2245 | } |
2246 | ||
69336bd2 | 2247 | bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag); |
5640f768 | 2248 | |
1da177e4 LT |
2249 | /* |
2250 | * Default write policy as shown to user space via poll/select/SIGIO | |
2251 | */ | |
dc6b9b78 | 2252 | static inline bool sock_writeable(const struct sock *sk) |
1da177e4 | 2253 | { |
14afee4b | 2254 | return refcount_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1); |
1da177e4 LT |
2255 | } |
2256 | ||
dd0fc66f | 2257 | static inline gfp_t gfp_any(void) |
1da177e4 | 2258 | { |
99709372 | 2259 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
2260 | } |
2261 | ||
dc6b9b78 | 2262 | static inline long sock_rcvtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2263 | { |
2264 | return noblock ? 0 : sk->sk_rcvtimeo; | |
2265 | } | |
2266 | ||
dc6b9b78 | 2267 | static inline long sock_sndtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2268 | { |
2269 | return noblock ? 0 : sk->sk_sndtimeo; | |
2270 | } | |
2271 | ||
2272 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
2273 | { | |
2274 | return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1; | |
2275 | } | |
2276 | ||
2277 | /* Alas, with timeout socket operations are not restartable. | |
2278 | * Compare this to poll(). | |
2279 | */ | |
2280 | static inline int sock_intr_errno(long timeo) | |
2281 | { | |
2282 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
2283 | } | |
2284 | ||
744d5a3e EB |
2285 | struct sock_skb_cb { |
2286 | u32 dropcount; | |
2287 | }; | |
2288 | ||
2289 | /* Store sock_skb_cb at the end of skb->cb[] so protocol families | |
2290 | * using skb->cb[] would keep using it directly and utilize its | |
2291 | * alignement guarantee. | |
2292 | */ | |
2293 | #define SOCK_SKB_CB_OFFSET ((FIELD_SIZEOF(struct sk_buff, cb) - \ | |
2294 | sizeof(struct sock_skb_cb))) | |
2295 | ||
2296 | #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \ | |
2297 | SOCK_SKB_CB_OFFSET)) | |
2298 | ||
b4772ef8 | 2299 | #define sock_skb_cb_check_size(size) \ |
744d5a3e | 2300 | BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET) |
b4772ef8 | 2301 | |
3bc3b96f EB |
2302 | static inline void |
2303 | sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb) | |
2304 | { | |
3665f381 ED |
2305 | SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ? |
2306 | atomic_read(&sk->sk_drops) : 0; | |
3bc3b96f EB |
2307 | } |
2308 | ||
532182cd ED |
2309 | static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb) |
2310 | { | |
2311 | int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs); | |
2312 | ||
2313 | atomic_add(segs, &sk->sk_drops); | |
2314 | } | |
2315 | ||
3a0ed3e9 DD |
2316 | static inline ktime_t sock_read_timestamp(struct sock *sk) |
2317 | { | |
2318 | #if BITS_PER_LONG==32 | |
2319 | unsigned int seq; | |
2320 | ktime_t kt; | |
2321 | ||
2322 | do { | |
2323 | seq = read_seqbegin(&sk->sk_stamp_seq); | |
2324 | kt = sk->sk_stamp; | |
2325 | } while (read_seqretry(&sk->sk_stamp_seq, seq)); | |
2326 | ||
2327 | return kt; | |
2328 | #else | |
2329 | return sk->sk_stamp; | |
2330 | #endif | |
2331 | } | |
2332 | ||
2333 | static inline void sock_write_timestamp(struct sock *sk, ktime_t kt) | |
2334 | { | |
2335 | #if BITS_PER_LONG==32 | |
2336 | write_seqlock(&sk->sk_stamp_seq); | |
2337 | sk->sk_stamp = kt; | |
2338 | write_sequnlock(&sk->sk_stamp_seq); | |
2339 | #else | |
2340 | sk->sk_stamp = kt; | |
2341 | #endif | |
2342 | } | |
2343 | ||
69336bd2 JP |
2344 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
2345 | struct sk_buff *skb); | |
2346 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, | |
2347 | struct sk_buff *skb); | |
92f37fd2 | 2348 | |
dc6b9b78 | 2349 | static inline void |
1da177e4 LT |
2350 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
2351 | { | |
b7aa0bf7 | 2352 | ktime_t kt = skb->tstamp; |
20d49473 | 2353 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
a61bbcf2 | 2354 | |
20d49473 PO |
2355 | /* |
2356 | * generate control messages if | |
b9f40e21 | 2357 | * - receive time stamping in software requested |
20d49473 | 2358 | * - software time stamp available and wanted |
20d49473 | 2359 | * - hardware time stamps available and wanted |
20d49473 PO |
2360 | */ |
2361 | if (sock_flag(sk, SOCK_RCVTSTAMP) || | |
b9f40e21 | 2362 | (sk->sk_tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) || |
2456e855 TG |
2363 | (kt && sk->sk_tsflags & SOF_TIMESTAMPING_SOFTWARE) || |
2364 | (hwtstamps->hwtstamp && | |
b9f40e21 | 2365 | (sk->sk_tsflags & SOF_TIMESTAMPING_RAW_HARDWARE))) |
92f37fd2 ED |
2366 | __sock_recv_timestamp(msg, sk, skb); |
2367 | else | |
3a0ed3e9 | 2368 | sock_write_timestamp(sk, kt); |
6e3e939f JB |
2369 | |
2370 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid) | |
2371 | __sock_recv_wifi_status(msg, sk, skb); | |
1da177e4 LT |
2372 | } |
2373 | ||
69336bd2 JP |
2374 | void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2375 | struct sk_buff *skb); | |
767dd033 | 2376 | |
6c7c98ba | 2377 | #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC) |
767dd033 ED |
2378 | static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk, |
2379 | struct sk_buff *skb) | |
2380 | { | |
2381 | #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \ | |
b9f40e21 WB |
2382 | (1UL << SOCK_RCVTSTAMP)) |
2383 | #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \ | |
2384 | SOF_TIMESTAMPING_RAW_HARDWARE) | |
767dd033 | 2385 | |
b9f40e21 | 2386 | if (sk->sk_flags & FLAGS_TS_OR_DROPS || sk->sk_tsflags & TSFLAGS_ANY) |
767dd033 | 2387 | __sock_recv_ts_and_drops(msg, sk, skb); |
d3fbff30 | 2388 | else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP))) |
3a0ed3e9 | 2389 | sock_write_timestamp(sk, skb->tstamp); |
6c7c98ba | 2390 | else if (unlikely(sk->sk_stamp == SK_DEFAULT_STAMP)) |
3a0ed3e9 | 2391 | sock_write_timestamp(sk, 0); |
767dd033 | 2392 | } |
3b885787 | 2393 | |
c14ac945 | 2394 | void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags); |
67cc0d40 | 2395 | |
20d49473 | 2396 | /** |
8f932f76 | 2397 | * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped |
20d49473 | 2398 | * @sk: socket sending this packet |
c14ac945 | 2399 | * @tsflags: timestamping flags to use |
140c55d4 | 2400 | * @tx_flags: completed with instructions for time stamping |
8f932f76 | 2401 | * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno) |
140c55d4 | 2402 | * |
d651983d | 2403 | * Note: callers should take care of initial ``*tx_flags`` value (usually 0) |
20d49473 | 2404 | */ |
8f932f76 WB |
2405 | static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags, |
2406 | __u8 *tx_flags, __u32 *tskey) | |
67cc0d40 | 2407 | { |
8f932f76 | 2408 | if (unlikely(tsflags)) { |
c14ac945 | 2409 | __sock_tx_timestamp(tsflags, tx_flags); |
8f932f76 WB |
2410 | if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey && |
2411 | tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK) | |
2412 | *tskey = sk->sk_tskey++; | |
2413 | } | |
67cc0d40 WB |
2414 | if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS))) |
2415 | *tx_flags |= SKBTX_WIFI_STATUS; | |
2416 | } | |
20d49473 | 2417 | |
8f932f76 WB |
2418 | static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags, |
2419 | __u8 *tx_flags) | |
2420 | { | |
2421 | _sock_tx_timestamp(sk, tsflags, tx_flags, NULL); | |
2422 | } | |
2423 | ||
2424 | static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags) | |
2425 | { | |
2426 | _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags, | |
2427 | &skb_shinfo(skb)->tskey); | |
2428 | } | |
2429 | ||
1da177e4 LT |
2430 | /** |
2431 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
2432 | * @sk: socket to eat this skb from |
2433 | * @skb: socket buffer to eat | |
1da177e4 LT |
2434 | * |
2435 | * This routine must be called with interrupts disabled or with the socket | |
2436 | * locked so that the sk_buff queue operation is ok. | |
2437 | */ | |
ede61ca4 | 2438 | DECLARE_STATIC_KEY_FALSE(tcp_rx_skb_cache_key); |
7bced397 | 2439 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
2440 | { |
2441 | __skb_unlink(skb, &sk->sk_receive_queue); | |
ede61ca4 | 2442 | if (static_branch_unlikely(&tcp_rx_skb_cache_key) && |
8b27dae5 ED |
2443 | !sk->sk_rx_skb_cache) { |
2444 | sk->sk_rx_skb_cache = skb; | |
2445 | skb_orphan(skb); | |
2446 | return; | |
2447 | } | |
1da177e4 LT |
2448 | __kfree_skb(skb); |
2449 | } | |
2450 | ||
3b1e0a65 YH |
2451 | static inline |
2452 | struct net *sock_net(const struct sock *sk) | |
2453 | { | |
c2d9ba9b | 2454 | return read_pnet(&sk->sk_net); |
3b1e0a65 YH |
2455 | } |
2456 | ||
2457 | static inline | |
f5aa23fd | 2458 | void sock_net_set(struct sock *sk, struct net *net) |
3b1e0a65 | 2459 | { |
c2d9ba9b | 2460 | write_pnet(&sk->sk_net, net); |
3b1e0a65 YH |
2461 | } |
2462 | ||
23542618 KK |
2463 | static inline struct sock *skb_steal_sock(struct sk_buff *skb) |
2464 | { | |
efc27f8c | 2465 | if (skb->sk) { |
23542618 KK |
2466 | struct sock *sk = skb->sk; |
2467 | ||
2468 | skb->destructor = NULL; | |
2469 | skb->sk = NULL; | |
2470 | return sk; | |
2471 | } | |
2472 | return NULL; | |
2473 | } | |
2474 | ||
1d0ab253 ED |
2475 | /* This helper checks if a socket is a full socket, |
2476 | * ie _not_ a timewait or request socket. | |
2477 | */ | |
2478 | static inline bool sk_fullsock(const struct sock *sk) | |
2479 | { | |
2480 | return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV); | |
2481 | } | |
2482 | ||
ebf4e808 IL |
2483 | /* Checks if this SKB belongs to an HW offloaded socket |
2484 | * and whether any SW fallbacks are required based on dev. | |
41477662 | 2485 | * Check decrypted mark in case skb_orphan() cleared socket. |
ebf4e808 IL |
2486 | */ |
2487 | static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb, | |
2488 | struct net_device *dev) | |
2489 | { | |
2490 | #ifdef CONFIG_SOCK_VALIDATE_XMIT | |
2491 | struct sock *sk = skb->sk; | |
2492 | ||
41477662 | 2493 | if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb) { |
ebf4e808 | 2494 | skb = sk->sk_validate_xmit_skb(sk, dev, skb); |
41477662 JK |
2495 | #ifdef CONFIG_TLS_DEVICE |
2496 | } else if (unlikely(skb->decrypted)) { | |
2497 | pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n"); | |
2498 | kfree_skb(skb); | |
2499 | skb = NULL; | |
2500 | #endif | |
2501 | } | |
ebf4e808 IL |
2502 | #endif |
2503 | ||
2504 | return skb; | |
2505 | } | |
2506 | ||
e446f9df ED |
2507 | /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV |
2508 | * SYNACK messages can be attached to either ones (depending on SYNCOOKIE) | |
2509 | */ | |
2510 | static inline bool sk_listener(const struct sock *sk) | |
2511 | { | |
2512 | return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV); | |
2513 | } | |
2514 | ||
69336bd2 | 2515 | void sock_enable_timestamp(struct sock *sk, int flag); |
69336bd2 JP |
2516 | int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level, |
2517 | int type); | |
1da177e4 | 2518 | |
a3b299da EB |
2519 | bool sk_ns_capable(const struct sock *sk, |
2520 | struct user_namespace *user_ns, int cap); | |
2521 | bool sk_capable(const struct sock *sk, int cap); | |
2522 | bool sk_net_capable(const struct sock *sk, int cap); | |
2523 | ||
a2d133b1 JH |
2524 | void sk_get_meminfo(const struct sock *sk, u32 *meminfo); |
2525 | ||
eaa72dc4 ED |
2526 | /* Take into consideration the size of the struct sk_buff overhead in the |
2527 | * determination of these values, since that is non-constant across | |
2528 | * platforms. This makes socket queueing behavior and performance | |
2529 | * not depend upon such differences. | |
2530 | */ | |
2531 | #define _SK_MEM_PACKETS 256 | |
2532 | #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256) | |
2533 | #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
2534 | #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
2535 | ||
1da177e4 LT |
2536 | extern __u32 sysctl_wmem_max; |
2537 | extern __u32 sysctl_rmem_max; | |
2538 | ||
b245be1f | 2539 | extern int sysctl_tstamp_allow_data; |
6baf1f41 DM |
2540 | extern int sysctl_optmem_max; |
2541 | ||
20380731 ACM |
2542 | extern __u32 sysctl_wmem_default; |
2543 | extern __u32 sysctl_rmem_default; | |
20380731 | 2544 | |
ce27ec60 ED |
2545 | DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key); |
2546 | ||
a3dcaf17 ED |
2547 | static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto) |
2548 | { | |
2549 | /* Does this proto have per netns sysctl_wmem ? */ | |
2550 | if (proto->sysctl_wmem_offset) | |
2551 | return *(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset); | |
2552 | ||
2553 | return *proto->sysctl_wmem; | |
2554 | } | |
2555 | ||
2556 | static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto) | |
2557 | { | |
2558 | /* Does this proto have per netns sysctl_rmem ? */ | |
2559 | if (proto->sysctl_rmem_offset) | |
2560 | return *(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset); | |
2561 | ||
2562 | return *proto->sysctl_rmem; | |
2563 | } | |
2564 | ||
c9f1f58d ED |
2565 | /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10) |
2566 | * Some wifi drivers need to tweak it to get more chunks. | |
2567 | * They can use this helper from their ndo_start_xmit() | |
2568 | */ | |
2569 | static inline void sk_pacing_shift_update(struct sock *sk, int val) | |
2570 | { | |
2571 | if (!sk || !sk_fullsock(sk) || sk->sk_pacing_shift == val) | |
2572 | return; | |
2573 | sk->sk_pacing_shift = val; | |
2574 | } | |
2575 | ||
54dc3e33 DA |
2576 | /* if a socket is bound to a device, check that the given device |
2577 | * index is either the same or that the socket is bound to an L3 | |
2578 | * master device and the given device index is also enslaved to | |
2579 | * that L3 master | |
2580 | */ | |
2581 | static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif) | |
2582 | { | |
2583 | int mdif; | |
2584 | ||
2585 | if (!sk->sk_bound_dev_if || sk->sk_bound_dev_if == dif) | |
2586 | return true; | |
2587 | ||
2588 | mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif); | |
2589 | if (mdif && mdif == sk->sk_bound_dev_if) | |
2590 | return true; | |
2591 | ||
2592 | return false; | |
2593 | } | |
2594 | ||
1da177e4 | 2595 | #endif /* _SOCK_H */ |