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> |
88ab1932 | 59 | #include <linux/rculist_nulls.h> |
a57de0b4 | 60 | #include <linux/poll.h> |
c8c1bbb6 | 61 | #include <linux/sockptr.h> |
1c5f2ced | 62 | #include <linux/indirect_call_wrapper.h> |
c31504dc | 63 | #include <linux/atomic.h> |
41c6d650 | 64 | #include <linux/refcount.h> |
f35f8219 | 65 | #include <linux/llist.h> |
1da177e4 LT |
66 | #include <net/dst.h> |
67 | #include <net/checksum.h> | |
1d0ab253 | 68 | #include <net/tcp_states.h> |
b9f40e21 | 69 | #include <linux/net_tstamp.h> |
54dc3e33 | 70 | #include <net/l3mdev.h> |
04190bf8 | 71 | #include <uapi/linux/socket.h> |
1da177e4 LT |
72 | |
73 | /* | |
74 | * This structure really needs to be cleaned up. | |
75 | * Most of it is for TCP, and not used by any of | |
76 | * the other protocols. | |
77 | */ | |
78 | ||
1da177e4 LT |
79 | /* This is the per-socket lock. The spinlock provides a synchronization |
80 | * between user contexts and software interrupt processing, whereas the | |
81 | * mini-semaphore synchronizes multiple users amongst themselves. | |
82 | */ | |
1da177e4 LT |
83 | typedef struct { |
84 | spinlock_t slock; | |
d2e9117c | 85 | int owned; |
1da177e4 | 86 | wait_queue_head_t wq; |
a5b5bb9a IM |
87 | /* |
88 | * We express the mutex-alike socket_lock semantics | |
89 | * to the lock validator by explicitly managing | |
90 | * the slock as a lock variant (in addition to | |
91 | * the slock itself): | |
92 | */ | |
93 | #ifdef CONFIG_DEBUG_LOCK_ALLOC | |
94 | struct lockdep_map dep_map; | |
95 | #endif | |
1da177e4 LT |
96 | } socket_lock_t; |
97 | ||
1da177e4 | 98 | struct sock; |
8feaf0c0 | 99 | struct proto; |
0eeb8ffc | 100 | struct net; |
1da177e4 | 101 | |
077b393d ED |
102 | typedef __u32 __bitwise __portpair; |
103 | typedef __u64 __bitwise __addrpair; | |
104 | ||
1da177e4 | 105 | /** |
4dc3b16b | 106 | * struct sock_common - minimal network layer representation of sockets |
68835aba ED |
107 | * @skc_daddr: Foreign IPv4 addr |
108 | * @skc_rcv_saddr: Bound local IPv4 addr | |
66256e0b | 109 | * @skc_addrpair: 8-byte-aligned __u64 union of @skc_daddr & @skc_rcv_saddr |
4dc6dc71 | 110 | * @skc_hash: hash value used with various protocol lookup tables |
d4cada4a | 111 | * @skc_u16hashes: two u16 hash values used by UDP lookup tables |
ce43b03e ED |
112 | * @skc_dport: placeholder for inet_dport/tw_dport |
113 | * @skc_num: placeholder for inet_num/tw_num | |
66256e0b | 114 | * @skc_portpair: __u32 union of @skc_dport & @skc_num |
4dc3b16b PP |
115 | * @skc_family: network address family |
116 | * @skc_state: Connection state | |
117 | * @skc_reuse: %SO_REUSEADDR setting | |
055dc21a | 118 | * @skc_reuseport: %SO_REUSEPORT setting |
66256e0b RD |
119 | * @skc_ipv6only: socket is IPV6 only |
120 | * @skc_net_refcnt: socket is using net ref counting | |
4dc3b16b | 121 | * @skc_bound_dev_if: bound device index if != 0 |
4dc3b16b | 122 | * @skc_bind_node: bind hash linkage for various protocol lookup tables |
512615b6 | 123 | * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol |
8feaf0c0 | 124 | * @skc_prot: protocol handlers inside a network family |
07feaebf | 125 | * @skc_net: reference to the network namespace of this socket |
66256e0b RD |
126 | * @skc_v6_daddr: IPV6 destination address |
127 | * @skc_v6_rcv_saddr: IPV6 source address | |
128 | * @skc_cookie: socket's cookie value | |
68835aba ED |
129 | * @skc_node: main hash linkage for various protocol lookup tables |
130 | * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol | |
131 | * @skc_tx_queue_mapping: tx queue number for this connection | |
c6345ce7 | 132 | * @skc_rx_queue_mapping: rx queue number for this connection |
8e5eb54d ED |
133 | * @skc_flags: place holder for sk_flags |
134 | * %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE, | |
135 | * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings | |
66256e0b RD |
136 | * @skc_listener: connection request listener socket (aka rsk_listener) |
137 | * [union with @skc_flags] | |
138 | * @skc_tw_dr: (aka tw_dr) ptr to &struct inet_timewait_death_row | |
139 | * [union with @skc_flags] | |
70da268b | 140 | * @skc_incoming_cpu: record/match cpu processing incoming packets |
66256e0b RD |
141 | * @skc_rcv_wnd: (aka rsk_rcv_wnd) TCP receive window size (possibly scaled) |
142 | * [union with @skc_incoming_cpu] | |
143 | * @skc_tw_rcv_nxt: (aka tw_rcv_nxt) TCP window next expected seq number | |
144 | * [union with @skc_incoming_cpu] | |
68835aba | 145 | * @skc_refcnt: reference count |
4dc3b16b PP |
146 | * |
147 | * This is the minimal network layer representation of sockets, the header | |
8feaf0c0 ACM |
148 | * for struct sock and struct inet_timewait_sock. |
149 | */ | |
1da177e4 | 150 | struct sock_common { |
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; |
4e1beecc | 214 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
c6345ce7 AN |
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 | ||
1f00d375 | 234 | struct bpf_local_storage; |
b6459415 | 235 | struct sk_filter; |
6ac99e8f | 236 | |
1da177e4 LT |
237 | /** |
238 | * struct sock - network layer representation of sockets | |
8feaf0c0 | 239 | * @__sk_common: shared layout with inet_timewait_sock |
4dc3b16b PP |
240 | * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN |
241 | * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings | |
242 | * @sk_lock: synchronizer | |
cdfbabfb | 243 | * @sk_kern_sock: True if sock is using kernel lock classes |
4dc3b16b | 244 | * @sk_rcvbuf: size of receive buffer in bytes |
43815482 | 245 | * @sk_wq: sock wait queue and async head |
421b3885 | 246 | * @sk_rx_dst: receive input route used by early demux |
0c0a5ef8 | 247 | * @sk_rx_dst_ifindex: ifindex for @sk_rx_dst |
ef57c161 | 248 | * @sk_rx_dst_cookie: cookie for @sk_rx_dst |
4dc3b16b | 249 | * @sk_dst_cache: destination cache |
9b8805a3 | 250 | * @sk_dst_pending_confirm: need to confirm neighbour |
4dc3b16b | 251 | * @sk_policy: flow policy |
4dc3b16b PP |
252 | * @sk_receive_queue: incoming packets |
253 | * @sk_wmem_alloc: transmit queue bytes committed | |
771edcaf | 254 | * @sk_tsq_flags: TCP Small Queues flags |
4dc3b16b PP |
255 | * @sk_write_queue: Packet sending queue |
256 | * @sk_omem_alloc: "o" is "option" or "other" | |
257 | * @sk_wmem_queued: persistent queue size | |
258 | * @sk_forward_alloc: space allocated forward | |
2bb2f5fb | 259 | * @sk_reserved_mem: space reserved and non-reclaimable for the socket |
06021292 | 260 | * @sk_napi_id: id of the last napi context to receive data for sk |
dafcc438 | 261 | * @sk_ll_usec: usecs to busypoll when there is no data |
4dc3b16b | 262 | * @sk_allocation: allocation mode |
95bd09eb | 263 | * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler) |
218af599 | 264 | * @sk_pacing_status: Pacing status (requested, handled by sch_fq) |
c3f40d7c | 265 | * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE) |
4dc3b16b | 266 | * @sk_sndbuf: size of send buffer in bytes |
28448b80 TH |
267 | * @sk_no_check_tx: %SO_NO_CHECK setting, set checksum in TX packets |
268 | * @sk_no_check_rx: allow zero checksum in RX packets | |
4dc3b16b | 269 | * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO) |
aba54656 | 270 | * @sk_gso_disabled: if set, NETIF_F_GSO_MASK is forbidden. |
bcd76111 | 271 | * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4) |
82cc1a7a | 272 | * @sk_gso_max_size: Maximum GSO segment size to build |
1485348d | 273 | * @sk_gso_max_segs: Maximum number of GSO segments |
3a9b76fd | 274 | * @sk_pacing_shift: scaling factor for TCP Small Queues |
4dc3b16b | 275 | * @sk_lingertime: %SO_LINGER l_linger setting |
4dc3b16b PP |
276 | * @sk_backlog: always used with the per-socket spinlock held |
277 | * @sk_callback_lock: used with the callbacks in the end of this struct | |
278 | * @sk_error_queue: rarely used | |
33c732c3 WC |
279 | * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt, |
280 | * IPV6_ADDRFORM for instance) | |
4dc3b16b | 281 | * @sk_err: last error |
33c732c3 WC |
282 | * @sk_err_soft: errors that don't cause failure but are the cause of a |
283 | * persistent failure not just 'timed out' | |
cb61cb9b | 284 | * @sk_drops: raw/udp drops counter |
4dc3b16b PP |
285 | * @sk_ack_backlog: current listen backlog |
286 | * @sk_max_ack_backlog: listen backlog set in listen() | |
771edcaf | 287 | * @sk_uid: user id of owner |
7fd3253a | 288 | * @sk_prefer_busy_poll: prefer busypolling over softirq processing |
7c951caf | 289 | * @sk_busy_poll_budget: napi processing budget when busypolling |
4dc3b16b PP |
290 | * @sk_priority: %SO_PRIORITY setting |
291 | * @sk_type: socket type (%SOCK_STREAM, etc) | |
292 | * @sk_protocol: which protocol this socket belongs in this network family | |
5fb14d20 | 293 | * @sk_peer_lock: lock protecting @sk_peer_pid and @sk_peer_cred |
53c3fa20 RD |
294 | * @sk_peer_pid: &struct pid for this socket's peer |
295 | * @sk_peer_cred: %SO_PEERCRED setting | |
4dc3b16b PP |
296 | * @sk_rcvlowat: %SO_RCVLOWAT setting |
297 | * @sk_rcvtimeo: %SO_RCVTIMEO setting | |
298 | * @sk_sndtimeo: %SO_SNDTIMEO setting | |
b73c3d0e | 299 | * @sk_txhash: computed flow hash for use on transmit |
26859240 | 300 | * @sk_txrehash: enable TX hash rethink |
4dc3b16b | 301 | * @sk_filter: socket filtering instructions |
4dc3b16b PP |
302 | * @sk_timer: sock cleanup timer |
303 | * @sk_stamp: time stamp of last packet received | |
3a0ed3e9 | 304 | * @sk_stamp_seq: lock for accessing sk_stamp on 32 bit architectures only |
d463126e | 305 | * @sk_tsflags: SO_TIMESTAMPING flags |
fb87bd47 GN |
306 | * @sk_use_task_frag: allow sk_page_frag() to use current->task_frag. |
307 | * Sockets that can be used under memory reclaim should | |
308 | * set this to false. | |
d463126e YL |
309 | * @sk_bind_phc: SO_TIMESTAMPING bind PHC index of PTP virtual clock |
310 | * for timestamping | |
09c2d251 | 311 | * @sk_tskey: counter to disambiguate concurrent tstamp requests |
52267790 | 312 | * @sk_zckey: counter to order MSG_ZEROCOPY notifications |
4dc3b16b | 313 | * @sk_socket: Identd and reporting IO signals |
b68777d5 | 314 | * @sk_user_data: RPC layer private data. Write-protected by @sk_callback_lock. |
5640f768 | 315 | * @sk_frag: cached page frag |
d3d4f0a0 | 316 | * @sk_peek_off: current peek_offset value |
4dc3b16b | 317 | * @sk_send_head: front of stuff to transmit |
66256e0b | 318 | * @tcp_rtx_queue: TCP re-transmit queue [union with @sk_send_head] |
67be2dd1 | 319 | * @sk_security: used by security modules |
31729363 | 320 | * @sk_mark: generic packet mark |
2a56a1fe | 321 | * @sk_cgrp_data: cgroup data for this cgroup |
baac50bb | 322 | * @sk_memcg: this socket's memory cgroup association |
4dc3b16b | 323 | * @sk_write_pending: a write to stream socket waits to start |
419ce133 | 324 | * @sk_disconnects: number of disconnect operations performed on this sock |
4dc3b16b PP |
325 | * @sk_state_change: callback to indicate change in the state of the sock |
326 | * @sk_data_ready: callback to indicate there is data to be processed | |
327 | * @sk_write_space: callback to indicate there is bf sending space available | |
328 | * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE) | |
329 | * @sk_backlog_rcv: callback to process the backlog | |
66256e0b | 330 | * @sk_validate_xmit_skb: ptr to an optional validate function |
4dc3b16b | 331 | * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0 |
ef456144 | 332 | * @sk_reuseport_cb: reuseport group container |
66256e0b | 333 | * @sk_bpf_storage: ptr to cache and control for bpf_sk_storage |
293de7de | 334 | * @sk_rcu: used during RCU grace period |
80b14dee RC |
335 | * @sk_clockid: clockid used by time-based scheduling (SO_TXTIME) |
336 | * @sk_txtime_deadline_mode: set deadline mode for SO_TXTIME | |
66256e0b | 337 | * @sk_txtime_report_errors: set report errors mode for SO_TXTIME |
80b14dee | 338 | * @sk_txtime_unused: unused txtime flags |
ffa84b5f | 339 | * @ns_tracker: tracker for netns reference |
293de7de | 340 | */ |
1da177e4 LT |
341 | struct sock { |
342 | /* | |
8feaf0c0 | 343 | * Now struct inet_timewait_sock also uses sock_common, so please just |
1da177e4 LT |
344 | * don't add nothing before this first member (__sk_common) --acme |
345 | */ | |
346 | struct sock_common __sk_common; | |
4dc6dc71 ED |
347 | #define sk_node __sk_common.skc_node |
348 | #define sk_nulls_node __sk_common.skc_nulls_node | |
349 | #define sk_refcnt __sk_common.skc_refcnt | |
e022f0b4 | 350 | #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping |
4e1beecc | 351 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
c6345ce7 AN |
352 | #define sk_rx_queue_mapping __sk_common.skc_rx_queue_mapping |
353 | #endif | |
4dc6dc71 | 354 | |
68835aba ED |
355 | #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin |
356 | #define sk_dontcopy_end __sk_common.skc_dontcopy_end | |
4dc6dc71 | 357 | #define sk_hash __sk_common.skc_hash |
50805466 | 358 | #define sk_portpair __sk_common.skc_portpair |
05dbc7b5 ED |
359 | #define sk_num __sk_common.skc_num |
360 | #define sk_dport __sk_common.skc_dport | |
50805466 ED |
361 | #define sk_addrpair __sk_common.skc_addrpair |
362 | #define sk_daddr __sk_common.skc_daddr | |
363 | #define sk_rcv_saddr __sk_common.skc_rcv_saddr | |
1da177e4 LT |
364 | #define sk_family __sk_common.skc_family |
365 | #define sk_state __sk_common.skc_state | |
366 | #define sk_reuse __sk_common.skc_reuse | |
055dc21a | 367 | #define sk_reuseport __sk_common.skc_reuseport |
9fe516ba | 368 | #define sk_ipv6only __sk_common.skc_ipv6only |
26abe143 | 369 | #define sk_net_refcnt __sk_common.skc_net_refcnt |
1da177e4 | 370 | #define sk_bound_dev_if __sk_common.skc_bound_dev_if |
1da177e4 | 371 | #define sk_bind_node __sk_common.skc_bind_node |
8feaf0c0 | 372 | #define sk_prot __sk_common.skc_prot |
07feaebf | 373 | #define sk_net __sk_common.skc_net |
efe4208f ED |
374 | #define sk_v6_daddr __sk_common.skc_v6_daddr |
375 | #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr | |
33cf7c90 | 376 | #define sk_cookie __sk_common.skc_cookie |
70da268b | 377 | #define sk_incoming_cpu __sk_common.skc_incoming_cpu |
8e5eb54d | 378 | #define sk_flags __sk_common.skc_flags |
ed53d0ab | 379 | #define sk_rxhash __sk_common.skc_rxhash |
efe4208f | 380 | |
43f51df4 | 381 | /* early demux fields */ |
8b3f9133 | 382 | struct dst_entry __rcu *sk_rx_dst; |
43f51df4 ED |
383 | int sk_rx_dst_ifindex; |
384 | u32 sk_rx_dst_cookie; | |
385 | ||
1da177e4 | 386 | socket_lock_t sk_lock; |
9115e8cd ED |
387 | atomic_t sk_drops; |
388 | int sk_rcvlowat; | |
389 | struct sk_buff_head sk_error_queue; | |
b178bb3d | 390 | struct sk_buff_head sk_receive_queue; |
fa438ccf ED |
391 | /* |
392 | * The backlog queue is special, it is always used with | |
393 | * the per-socket spinlock held and requires low latency | |
394 | * access. Therefore we special case it's implementation. | |
b178bb3d ED |
395 | * Note : rmem_alloc is in this structure to fill a hole |
396 | * on 64bit arches, not because its logically part of | |
397 | * backlog. | |
fa438ccf ED |
398 | */ |
399 | struct { | |
b178bb3d ED |
400 | atomic_t rmem_alloc; |
401 | int len; | |
402 | struct sk_buff *head; | |
403 | struct sk_buff *tail; | |
fa438ccf | 404 | } sk_backlog; |
f35f8219 | 405 | |
b178bb3d | 406 | #define sk_rmem_alloc sk_backlog.rmem_alloc |
2c8c56e1 | 407 | |
9115e8cd | 408 | int sk_forward_alloc; |
2bb2f5fb | 409 | u32 sk_reserved_mem; |
e0d1095a | 410 | #ifdef CONFIG_NET_RX_BUSY_POLL |
dafcc438 | 411 | unsigned int sk_ll_usec; |
9115e8cd ED |
412 | /* ===== mostly read cache line ===== */ |
413 | unsigned int sk_napi_id; | |
b178bb3d | 414 | #endif |
b178bb3d | 415 | int sk_rcvbuf; |
419ce133 | 416 | int sk_disconnects; |
b178bb3d ED |
417 | |
418 | struct sk_filter __rcu *sk_filter; | |
ceb5d58b ED |
419 | union { |
420 | struct socket_wq __rcu *sk_wq; | |
66256e0b | 421 | /* private: */ |
ceb5d58b | 422 | struct socket_wq *sk_wq_raw; |
66256e0b | 423 | /* public: */ |
ceb5d58b | 424 | }; |
def8b4fa | 425 | #ifdef CONFIG_XFRM |
d188ba86 | 426 | struct xfrm_policy __rcu *sk_policy[2]; |
def8b4fa | 427 | #endif |
0c0a5ef8 | 428 | |
0e36cbb3 | 429 | struct dst_entry __rcu *sk_dst_cache; |
1da177e4 | 430 | atomic_t sk_omem_alloc; |
4e07a91c | 431 | int sk_sndbuf; |
9115e8cd ED |
432 | |
433 | /* ===== cache line for TX ===== */ | |
434 | int sk_wmem_queued; | |
14afee4b | 435 | refcount_t sk_wmem_alloc; |
9115e8cd | 436 | unsigned long sk_tsq_flags; |
75c119af ED |
437 | union { |
438 | struct sk_buff *sk_send_head; | |
439 | struct rb_root tcp_rtx_queue; | |
440 | }; | |
1da177e4 | 441 | struct sk_buff_head sk_write_queue; |
9115e8cd ED |
442 | __s32 sk_peek_off; |
443 | int sk_write_pending; | |
9b8805a3 | 444 | __u32 sk_dst_pending_confirm; |
218af599 | 445 | u32 sk_pacing_status; /* see enum sk_pacing */ |
9115e8cd ED |
446 | long sk_sndtimeo; |
447 | struct timer_list sk_timer; | |
448 | __u32 sk_priority; | |
449 | __u32 sk_mark; | |
76a9ebe8 ED |
450 | unsigned long sk_pacing_rate; /* bytes per second */ |
451 | unsigned long sk_max_pacing_rate; | |
9115e8cd ED |
452 | struct page_frag sk_frag; |
453 | netdev_features_t sk_route_caps; | |
9115e8cd ED |
454 | int sk_gso_type; |
455 | unsigned int sk_gso_max_size; | |
456 | gfp_t sk_allocation; | |
457 | __u32 sk_txhash; | |
fc64869c AR |
458 | |
459 | /* | |
460 | * Because of non atomicity rules, all | |
461 | * changes are protected by socket lock. | |
462 | */ | |
aba54656 | 463 | u8 sk_gso_disabled : 1, |
cdfbabfb | 464 | sk_kern_sock : 1, |
28448b80 TH |
465 | sk_no_check_tx : 1, |
466 | sk_no_check_rx : 1, | |
bf976514 | 467 | sk_userlocks : 4; |
3a9b76fd | 468 | u8 sk_pacing_shift; |
bf976514 MM |
469 | u16 sk_type; |
470 | u16 sk_protocol; | |
471 | u16 sk_gso_max_segs; | |
1da177e4 | 472 | unsigned long sk_lingertime; |
476e19cf | 473 | struct proto *sk_prot_creator; |
1da177e4 LT |
474 | rwlock_t sk_callback_lock; |
475 | int sk_err, | |
476 | sk_err_soft; | |
becb74f0 ED |
477 | u32 sk_ack_backlog; |
478 | u32 sk_max_ack_backlog; | |
86741ec2 | 479 | kuid_t sk_uid; |
26859240 | 480 | u8 sk_txrehash; |
7fd3253a BT |
481 | #ifdef CONFIG_NET_RX_BUSY_POLL |
482 | u8 sk_prefer_busy_poll; | |
7c951caf | 483 | u16 sk_busy_poll_budget; |
7fd3253a | 484 | #endif |
35306eb2 | 485 | spinlock_t sk_peer_lock; |
1ace2b4d | 486 | int sk_bind_phc; |
109f6e39 EB |
487 | struct pid *sk_peer_pid; |
488 | const struct cred *sk_peer_cred; | |
35306eb2 | 489 | |
1da177e4 | 490 | long sk_rcvtimeo; |
b7aa0bf7 | 491 | ktime_t sk_stamp; |
3a0ed3e9 DD |
492 | #if BITS_PER_LONG==32 |
493 | seqlock_t sk_stamp_seq; | |
494 | #endif | |
a1cdec57 | 495 | atomic_t sk_tskey; |
52267790 | 496 | atomic_t sk_zckey; |
b534dc46 WB |
497 | u32 sk_tsflags; |
498 | u8 sk_shutdown; | |
80b14dee RC |
499 | |
500 | u8 sk_clockid; | |
501 | u8 sk_txtime_deadline_mode : 1, | |
4b15c707 JSP |
502 | sk_txtime_report_errors : 1, |
503 | sk_txtime_unused : 6; | |
fb87bd47 | 504 | bool sk_use_task_frag; |
80b14dee | 505 | |
1da177e4 LT |
506 | struct socket *sk_socket; |
507 | void *sk_user_data; | |
d5f64238 | 508 | #ifdef CONFIG_SECURITY |
1da177e4 | 509 | void *sk_security; |
d5f64238 | 510 | #endif |
2a56a1fe | 511 | struct sock_cgroup_data sk_cgrp_data; |
baac50bb | 512 | struct mem_cgroup *sk_memcg; |
1da177e4 | 513 | void (*sk_state_change)(struct sock *sk); |
676d2369 | 514 | void (*sk_data_ready)(struct sock *sk); |
1da177e4 LT |
515 | void (*sk_write_space)(struct sock *sk); |
516 | void (*sk_error_report)(struct sock *sk); | |
dc6b9b78 ED |
517 | int (*sk_backlog_rcv)(struct sock *sk, |
518 | struct sk_buff *skb); | |
ebf4e808 IL |
519 | #ifdef CONFIG_SOCK_VALIDATE_XMIT |
520 | struct sk_buff* (*sk_validate_xmit_skb)(struct sock *sk, | |
521 | struct net_device *dev, | |
522 | struct sk_buff *skb); | |
523 | #endif | |
1da177e4 | 524 | void (*sk_destruct)(struct sock *sk); |
ef456144 | 525 | struct sock_reuseport __rcu *sk_reuseport_cb; |
6ac99e8f | 526 | #ifdef CONFIG_BPF_SYSCALL |
1f00d375 | 527 | struct bpf_local_storage __rcu *sk_bpf_storage; |
6ac99e8f | 528 | #endif |
a4298e45 | 529 | struct rcu_head sk_rcu; |
ffa84b5f | 530 | netns_tracker ns_tracker; |
1da177e4 LT |
531 | }; |
532 | ||
218af599 ED |
533 | enum sk_pacing { |
534 | SK_PACING_NONE = 0, | |
535 | SK_PACING_NEEDED = 1, | |
536 | SK_PACING_FQ = 2, | |
537 | }; | |
538 | ||
2a013372 HJ |
539 | /* flag bits in sk_user_data |
540 | * | |
541 | * - SK_USER_DATA_NOCOPY: Pointer stored in sk_user_data might | |
542 | * not be suitable for copying when cloning the socket. For instance, | |
543 | * it can point to a reference counted object. sk_user_data bottom | |
544 | * bit is set if pointer must not be copied. | |
545 | * | |
546 | * - SK_USER_DATA_BPF: Mark whether sk_user_data field is | |
547 | * managed/owned by a BPF reuseport array. This bit should be set | |
548 | * when sk_user_data's sk is added to the bpf's reuseport_array. | |
549 | * | |
550 | * - SK_USER_DATA_PSOCK: Mark whether pointer stored in | |
551 | * sk_user_data points to psock type. This bit should be set | |
552 | * when sk_user_data is assigned to a psock object. | |
f1ff5ce2 JS |
553 | */ |
554 | #define SK_USER_DATA_NOCOPY 1UL | |
2a013372 HJ |
555 | #define SK_USER_DATA_BPF 2UL |
556 | #define SK_USER_DATA_PSOCK 4UL | |
557 | #define SK_USER_DATA_PTRMASK ~(SK_USER_DATA_NOCOPY | SK_USER_DATA_BPF |\ | |
558 | SK_USER_DATA_PSOCK) | |
f1ff5ce2 JS |
559 | |
560 | /** | |
561 | * sk_user_data_is_nocopy - Test if sk_user_data pointer must not be copied | |
562 | * @sk: socket | |
563 | */ | |
564 | static inline bool sk_user_data_is_nocopy(const struct sock *sk) | |
565 | { | |
566 | return ((uintptr_t)sk->sk_user_data & SK_USER_DATA_NOCOPY); | |
567 | } | |
568 | ||
559835ea PS |
569 | #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data))) |
570 | ||
fc4aaf9f DH |
571 | /** |
572 | * __locked_read_sk_user_data_with_flags - return the pointer | |
573 | * only if argument flags all has been set in sk_user_data. Otherwise | |
574 | * return NULL | |
575 | * | |
576 | * @sk: socket | |
577 | * @flags: flag bits | |
578 | * | |
579 | * The caller must be holding sk->sk_callback_lock. | |
580 | */ | |
581 | static inline void * | |
582 | __locked_read_sk_user_data_with_flags(const struct sock *sk, | |
583 | uintptr_t flags) | |
584 | { | |
585 | uintptr_t sk_user_data = | |
586 | (uintptr_t)rcu_dereference_check(__sk_user_data(sk), | |
587 | lockdep_is_held(&sk->sk_callback_lock)); | |
588 | ||
589 | WARN_ON_ONCE(flags & SK_USER_DATA_PTRMASK); | |
590 | ||
591 | if ((sk_user_data & flags) == flags) | |
592 | return (void *)(sk_user_data & SK_USER_DATA_PTRMASK); | |
593 | return NULL; | |
594 | } | |
595 | ||
2a013372 HJ |
596 | /** |
597 | * __rcu_dereference_sk_user_data_with_flags - return the pointer | |
598 | * only if argument flags all has been set in sk_user_data. Otherwise | |
599 | * return NULL | |
600 | * | |
601 | * @sk: socket | |
602 | * @flags: flag bits | |
603 | */ | |
604 | static inline void * | |
605 | __rcu_dereference_sk_user_data_with_flags(const struct sock *sk, | |
606 | uintptr_t flags) | |
607 | { | |
608 | uintptr_t sk_user_data = (uintptr_t)rcu_dereference(__sk_user_data(sk)); | |
609 | ||
610 | WARN_ON_ONCE(flags & SK_USER_DATA_PTRMASK); | |
611 | ||
612 | if ((sk_user_data & flags) == flags) | |
613 | return (void *)(sk_user_data & SK_USER_DATA_PTRMASK); | |
614 | return NULL; | |
615 | } | |
616 | ||
f1ff5ce2 | 617 | #define rcu_dereference_sk_user_data(sk) \ |
2a013372 HJ |
618 | __rcu_dereference_sk_user_data_with_flags(sk, 0) |
619 | #define __rcu_assign_sk_user_data_with_flags(sk, ptr, flags) \ | |
f1ff5ce2 | 620 | ({ \ |
2a013372 HJ |
621 | uintptr_t __tmp1 = (uintptr_t)(ptr), \ |
622 | __tmp2 = (uintptr_t)(flags); \ | |
623 | WARN_ON_ONCE(__tmp1 & ~SK_USER_DATA_PTRMASK); \ | |
624 | WARN_ON_ONCE(__tmp2 & SK_USER_DATA_PTRMASK); \ | |
f1ff5ce2 | 625 | rcu_assign_pointer(__sk_user_data((sk)), \ |
2a013372 | 626 | __tmp1 | __tmp2); \ |
f1ff5ce2 | 627 | }) |
2a013372 HJ |
628 | #define rcu_assign_sk_user_data(sk, ptr) \ |
629 | __rcu_assign_sk_user_data_with_flags(sk, ptr, 0) | |
559835ea | 630 | |
e187013a AK |
631 | static inline |
632 | struct net *sock_net(const struct sock *sk) | |
633 | { | |
634 | return read_pnet(&sk->sk_net); | |
635 | } | |
636 | ||
637 | static inline | |
638 | void sock_net_set(struct sock *sk, struct net *net) | |
639 | { | |
640 | write_pnet(&sk->sk_net, net); | |
641 | } | |
642 | ||
4a17fd52 PE |
643 | /* |
644 | * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK | |
645 | * or not whether his port will be reused by someone else. SK_FORCE_REUSE | |
646 | * on a socket means that the socket will reuse everybody else's port | |
647 | * without looking at the other's sk_reuse value. | |
648 | */ | |
649 | ||
650 | #define SK_NO_REUSE 0 | |
651 | #define SK_CAN_REUSE 1 | |
652 | #define SK_FORCE_REUSE 2 | |
653 | ||
627d2d6b | 654 | int sk_set_peek_off(struct sock *sk, int val); |
655 | ||
a84a434b | 656 | static inline int sk_peek_offset(const struct sock *sk, int flags) |
ef64a54f | 657 | { |
b9bb53f3 | 658 | if (unlikely(flags & MSG_PEEK)) { |
a0917e0b | 659 | return READ_ONCE(sk->sk_peek_off); |
b9bb53f3 WB |
660 | } |
661 | ||
662 | return 0; | |
ef64a54f PE |
663 | } |
664 | ||
665 | static inline void sk_peek_offset_bwd(struct sock *sk, int val) | |
666 | { | |
b9bb53f3 WB |
667 | s32 off = READ_ONCE(sk->sk_peek_off); |
668 | ||
669 | if (unlikely(off >= 0)) { | |
670 | off = max_t(s32, off - val, 0); | |
671 | WRITE_ONCE(sk->sk_peek_off, off); | |
ef64a54f PE |
672 | } |
673 | } | |
674 | ||
675 | static inline void sk_peek_offset_fwd(struct sock *sk, int val) | |
676 | { | |
b9bb53f3 | 677 | sk_peek_offset_bwd(sk, -val); |
ef64a54f PE |
678 | } |
679 | ||
1da177e4 LT |
680 | /* |
681 | * Hashed lists helper routines | |
682 | */ | |
c4146644 LZ |
683 | static inline struct sock *sk_entry(const struct hlist_node *node) |
684 | { | |
685 | return hlist_entry(node, struct sock, sk_node); | |
686 | } | |
687 | ||
e48c414e | 688 | static inline struct sock *__sk_head(const struct hlist_head *head) |
1da177e4 LT |
689 | { |
690 | return hlist_entry(head->first, struct sock, sk_node); | |
691 | } | |
692 | ||
e48c414e | 693 | static inline struct sock *sk_head(const struct hlist_head *head) |
1da177e4 LT |
694 | { |
695 | return hlist_empty(head) ? NULL : __sk_head(head); | |
696 | } | |
697 | ||
88ab1932 ED |
698 | static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head) |
699 | { | |
700 | return hlist_nulls_entry(head->first, struct sock, sk_nulls_node); | |
701 | } | |
702 | ||
703 | static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head) | |
704 | { | |
705 | return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head); | |
706 | } | |
707 | ||
e48c414e | 708 | static inline struct sock *sk_next(const struct sock *sk) |
1da177e4 | 709 | { |
6c59ebd3 | 710 | return hlist_entry_safe(sk->sk_node.next, struct sock, sk_node); |
1da177e4 LT |
711 | } |
712 | ||
88ab1932 ED |
713 | static inline struct sock *sk_nulls_next(const struct sock *sk) |
714 | { | |
715 | return (!is_a_nulls(sk->sk_nulls_node.next)) ? | |
716 | hlist_nulls_entry(sk->sk_nulls_node.next, | |
717 | struct sock, sk_nulls_node) : | |
718 | NULL; | |
719 | } | |
720 | ||
dc6b9b78 | 721 | static inline bool sk_unhashed(const struct sock *sk) |
1da177e4 LT |
722 | { |
723 | return hlist_unhashed(&sk->sk_node); | |
724 | } | |
725 | ||
dc6b9b78 | 726 | static inline bool sk_hashed(const struct sock *sk) |
1da177e4 | 727 | { |
da753bea | 728 | return !sk_unhashed(sk); |
1da177e4 LT |
729 | } |
730 | ||
dc6b9b78 | 731 | static inline void sk_node_init(struct hlist_node *node) |
1da177e4 LT |
732 | { |
733 | node->pprev = NULL; | |
734 | } | |
735 | ||
dc6b9b78 | 736 | static inline void __sk_del_node(struct sock *sk) |
1da177e4 LT |
737 | { |
738 | __hlist_del(&sk->sk_node); | |
739 | } | |
740 | ||
808f5114 | 741 | /* NB: equivalent to hlist_del_init_rcu */ |
dc6b9b78 | 742 | static inline bool __sk_del_node_init(struct sock *sk) |
1da177e4 LT |
743 | { |
744 | if (sk_hashed(sk)) { | |
745 | __sk_del_node(sk); | |
746 | sk_node_init(&sk->sk_node); | |
dc6b9b78 | 747 | return true; |
1da177e4 | 748 | } |
dc6b9b78 | 749 | return false; |
1da177e4 LT |
750 | } |
751 | ||
752 | /* Grab socket reference count. This operation is valid only | |
753 | when sk is ALREADY grabbed f.e. it is found in hash table | |
754 | or a list and the lookup is made under lock preventing hash table | |
755 | modifications. | |
756 | */ | |
757 | ||
f9a7cbbf | 758 | static __always_inline void sock_hold(struct sock *sk) |
1da177e4 | 759 | { |
41c6d650 | 760 | refcount_inc(&sk->sk_refcnt); |
1da177e4 LT |
761 | } |
762 | ||
763 | /* Ungrab socket in the context, which assumes that socket refcnt | |
764 | cannot hit zero, f.e. it is true in context of any socketcall. | |
765 | */ | |
f9a7cbbf | 766 | static __always_inline void __sock_put(struct sock *sk) |
1da177e4 | 767 | { |
41c6d650 | 768 | refcount_dec(&sk->sk_refcnt); |
1da177e4 LT |
769 | } |
770 | ||
dc6b9b78 | 771 | static inline bool sk_del_node_init(struct sock *sk) |
1da177e4 | 772 | { |
dc6b9b78 | 773 | bool rc = __sk_del_node_init(sk); |
1da177e4 LT |
774 | |
775 | if (rc) { | |
776 | /* paranoid for a while -acme */ | |
41c6d650 | 777 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
1da177e4 LT |
778 | __sock_put(sk); |
779 | } | |
780 | return rc; | |
781 | } | |
808f5114 | 782 | #define sk_del_node_init_rcu(sk) sk_del_node_init(sk) |
1da177e4 | 783 | |
dc6b9b78 | 784 | static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 ED |
785 | { |
786 | if (sk_hashed(sk)) { | |
88ab1932 | 787 | hlist_nulls_del_init_rcu(&sk->sk_nulls_node); |
dc6b9b78 | 788 | return true; |
271b72c7 | 789 | } |
dc6b9b78 | 790 | return false; |
271b72c7 ED |
791 | } |
792 | ||
dc6b9b78 | 793 | static inline bool sk_nulls_del_node_init_rcu(struct sock *sk) |
271b72c7 | 794 | { |
dc6b9b78 | 795 | bool rc = __sk_nulls_del_node_init_rcu(sk); |
271b72c7 ED |
796 | |
797 | if (rc) { | |
798 | /* paranoid for a while -acme */ | |
41c6d650 | 799 | WARN_ON(refcount_read(&sk->sk_refcnt) == 1); |
271b72c7 ED |
800 | __sock_put(sk); |
801 | } | |
802 | return rc; | |
803 | } | |
804 | ||
dc6b9b78 | 805 | static inline void __sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
806 | { |
807 | hlist_add_head(&sk->sk_node, list); | |
808 | } | |
809 | ||
dc6b9b78 | 810 | static inline void sk_add_node(struct sock *sk, struct hlist_head *list) |
1da177e4 LT |
811 | { |
812 | sock_hold(sk); | |
813 | __sk_add_node(sk, list); | |
814 | } | |
815 | ||
dc6b9b78 | 816 | static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list) |
808f5114 | 817 | { |
818 | sock_hold(sk); | |
d296ba60 CG |
819 | if (IS_ENABLED(CONFIG_IPV6) && sk->sk_reuseport && |
820 | sk->sk_family == AF_INET6) | |
821 | hlist_add_tail_rcu(&sk->sk_node, list); | |
822 | else | |
823 | hlist_add_head_rcu(&sk->sk_node, list); | |
808f5114 | 824 | } |
825 | ||
a4dc6a49 MC |
826 | static inline void sk_add_node_tail_rcu(struct sock *sk, struct hlist_head *list) |
827 | { | |
828 | sock_hold(sk); | |
829 | hlist_add_tail_rcu(&sk->sk_node, list); | |
830 | } | |
831 | ||
dc6b9b78 | 832 | static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 | 833 | { |
d7efc6c1 | 834 | hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list); |
271b72c7 ED |
835 | } |
836 | ||
8dbd76e7 ED |
837 | static inline void __sk_nulls_add_node_tail_rcu(struct sock *sk, struct hlist_nulls_head *list) |
838 | { | |
839 | hlist_nulls_add_tail_rcu(&sk->sk_nulls_node, list); | |
840 | } | |
841 | ||
dc6b9b78 | 842 | static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list) |
271b72c7 ED |
843 | { |
844 | sock_hold(sk); | |
88ab1932 | 845 | __sk_nulls_add_node_rcu(sk, list); |
271b72c7 ED |
846 | } |
847 | ||
dc6b9b78 | 848 | static inline void __sk_del_bind_node(struct sock *sk) |
1da177e4 LT |
849 | { |
850 | __hlist_del(&sk->sk_bind_node); | |
851 | } | |
852 | ||
dc6b9b78 | 853 | static inline void sk_add_bind_node(struct sock *sk, |
1da177e4 LT |
854 | struct hlist_head *list) |
855 | { | |
856 | hlist_add_head(&sk->sk_bind_node, list); | |
857 | } | |
858 | ||
b67bfe0d SL |
859 | #define sk_for_each(__sk, list) \ |
860 | hlist_for_each_entry(__sk, list, sk_node) | |
861 | #define sk_for_each_rcu(__sk, list) \ | |
862 | hlist_for_each_entry_rcu(__sk, list, sk_node) | |
88ab1932 ED |
863 | #define sk_nulls_for_each(__sk, node, list) \ |
864 | hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node) | |
865 | #define sk_nulls_for_each_rcu(__sk, node, list) \ | |
866 | hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node) | |
b67bfe0d SL |
867 | #define sk_for_each_from(__sk) \ |
868 | hlist_for_each_entry_from(__sk, sk_node) | |
88ab1932 ED |
869 | #define sk_nulls_for_each_from(__sk, node) \ |
870 | if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \ | |
871 | hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node) | |
b67bfe0d SL |
872 | #define sk_for_each_safe(__sk, tmp, list) \ |
873 | hlist_for_each_entry_safe(__sk, tmp, list, sk_node) | |
874 | #define sk_for_each_bound(__sk, list) \ | |
875 | hlist_for_each_entry(__sk, list, sk_bind_node) | |
1da177e4 | 876 | |
2dc41cff | 877 | /** |
ca065d0c | 878 | * sk_for_each_entry_offset_rcu - iterate over a list at a given struct offset |
2dc41cff DH |
879 | * @tpos: the type * to use as a loop cursor. |
880 | * @pos: the &struct hlist_node to use as a loop cursor. | |
881 | * @head: the head for your list. | |
882 | * @offset: offset of hlist_node within the struct. | |
883 | * | |
884 | */ | |
ca065d0c | 885 | #define sk_for_each_entry_offset_rcu(tpos, pos, head, offset) \ |
b6f4f848 | 886 | for (pos = rcu_dereference(hlist_first_rcu(head)); \ |
ca065d0c | 887 | pos != NULL && \ |
2dc41cff | 888 | ({ tpos = (typeof(*tpos) *)((void *)pos - offset); 1;}); \ |
b6f4f848 | 889 | pos = rcu_dereference(hlist_next_rcu(pos))) |
2dc41cff | 890 | |
a84a434b | 891 | static inline struct user_namespace *sk_user_ns(const struct sock *sk) |
c336d148 EB |
892 | { |
893 | /* Careful only use this in a context where these parameters | |
894 | * can not change and must all be valid, such as recvmsg from | |
895 | * userspace. | |
896 | */ | |
897 | return sk->sk_socket->file->f_cred->user_ns; | |
898 | } | |
899 | ||
1da177e4 LT |
900 | /* Sock flags */ |
901 | enum sock_flags { | |
902 | SOCK_DEAD, | |
903 | SOCK_DONE, | |
904 | SOCK_URGINLINE, | |
905 | SOCK_KEEPOPEN, | |
906 | SOCK_LINGER, | |
907 | SOCK_DESTROY, | |
908 | SOCK_BROADCAST, | |
909 | SOCK_TIMESTAMP, | |
910 | SOCK_ZAPPED, | |
911 | SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */ | |
912 | SOCK_DBG, /* %SO_DEBUG setting */ | |
913 | SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */ | |
92f37fd2 | 914 | SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */ |
1da177e4 | 915 | SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */ |
7cb02404 | 916 | SOCK_MEMALLOC, /* VM depends on this socket for swapping */ |
20d49473 | 917 | SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */ |
bcdce719 | 918 | SOCK_FASYNC, /* fasync() active */ |
3b885787 | 919 | SOCK_RXQ_OVFL, |
1cdebb42 | 920 | SOCK_ZEROCOPY, /* buffers from userspace */ |
6e3e939f | 921 | SOCK_WIFI_STATUS, /* push wifi status to userspace */ |
3bdc0eba BG |
922 | SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS. |
923 | * Will use last 4 bytes of packet sent from | |
924 | * user-space instead. | |
925 | */ | |
d59577b6 | 926 | SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */ |
7d4c04fc | 927 | SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */ |
a4298e45 | 928 | SOCK_RCU_FREE, /* wait rcu grace period in sk_destruct() */ |
80b14dee | 929 | SOCK_TXTIME, |
e4a2a304 | 930 | SOCK_XDP, /* XDP is attached */ |
887feae3 | 931 | SOCK_TSTAMP_NEW, /* Indicates 64 bit timestamps always */ |
6fd1d51c | 932 | SOCK_RCVMARK, /* Receive SO_MARK ancillary data with packet */ |
1da177e4 LT |
933 | }; |
934 | ||
01ce63c9 MRL |
935 | #define SK_FLAGS_TIMESTAMP ((1UL << SOCK_TIMESTAMP) | (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE)) |
936 | ||
a84a434b | 937 | static inline void sock_copy_flags(struct sock *nsk, const struct sock *osk) |
53b924b3 RB |
938 | { |
939 | nsk->sk_flags = osk->sk_flags; | |
940 | } | |
941 | ||
1da177e4 LT |
942 | static inline void sock_set_flag(struct sock *sk, enum sock_flags flag) |
943 | { | |
944 | __set_bit(flag, &sk->sk_flags); | |
945 | } | |
946 | ||
947 | static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag) | |
948 | { | |
949 | __clear_bit(flag, &sk->sk_flags); | |
950 | } | |
951 | ||
dfde1d7d DY |
952 | static inline void sock_valbool_flag(struct sock *sk, enum sock_flags bit, |
953 | int valbool) | |
954 | { | |
955 | if (valbool) | |
956 | sock_set_flag(sk, bit); | |
957 | else | |
958 | sock_reset_flag(sk, bit); | |
959 | } | |
960 | ||
1b23a5df | 961 | static inline bool sock_flag(const struct sock *sk, enum sock_flags flag) |
1da177e4 LT |
962 | { |
963 | return test_bit(flag, &sk->sk_flags); | |
964 | } | |
965 | ||
c93bdd0e | 966 | #ifdef CONFIG_NET |
a7950ae8 | 967 | DECLARE_STATIC_KEY_FALSE(memalloc_socks_key); |
c93bdd0e MG |
968 | static inline int sk_memalloc_socks(void) |
969 | { | |
a7950ae8 | 970 | return static_branch_unlikely(&memalloc_socks_key); |
c93bdd0e | 971 | } |
d9539752 KC |
972 | |
973 | void __receive_sock(struct file *file); | |
c93bdd0e MG |
974 | #else |
975 | ||
976 | static inline int sk_memalloc_socks(void) | |
977 | { | |
978 | return 0; | |
979 | } | |
980 | ||
d9539752 KC |
981 | static inline void __receive_sock(struct file *file) |
982 | { } | |
c93bdd0e MG |
983 | #endif |
984 | ||
7450aaf6 | 985 | static inline gfp_t sk_gfp_mask(const struct sock *sk, gfp_t gfp_mask) |
99a1dec7 | 986 | { |
7450aaf6 | 987 | return gfp_mask | (sk->sk_allocation & __GFP_MEMALLOC); |
99a1dec7 MG |
988 | } |
989 | ||
1da177e4 LT |
990 | static inline void sk_acceptq_removed(struct sock *sk) |
991 | { | |
288efe86 | 992 | WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog - 1); |
1da177e4 LT |
993 | } |
994 | ||
995 | static inline void sk_acceptq_added(struct sock *sk) | |
996 | { | |
288efe86 | 997 | WRITE_ONCE(sk->sk_ack_backlog, sk->sk_ack_backlog + 1); |
1da177e4 LT |
998 | } |
999 | ||
c609e6aa ED |
1000 | /* Note: If you think the test should be: |
1001 | * return READ_ONCE(sk->sk_ack_backlog) >= READ_ONCE(sk->sk_max_ack_backlog); | |
1002 | * Then please take a look at commit 64a146513f8f ("[NET]: Revert incorrect accept queue backlog changes.") | |
1003 | */ | |
dc6b9b78 | 1004 | static inline bool sk_acceptq_is_full(const struct sock *sk) |
1da177e4 | 1005 | { |
c609e6aa | 1006 | return READ_ONCE(sk->sk_ack_backlog) > READ_ONCE(sk->sk_max_ack_backlog); |
1da177e4 LT |
1007 | } |
1008 | ||
1009 | /* | |
1010 | * Compute minimal free write space needed to queue new packets. | |
1011 | */ | |
dc6b9b78 | 1012 | static inline int sk_stream_min_wspace(const struct sock *sk) |
1da177e4 | 1013 | { |
ab4e846a | 1014 | return READ_ONCE(sk->sk_wmem_queued) >> 1; |
1da177e4 LT |
1015 | } |
1016 | ||
dc6b9b78 | 1017 | static inline int sk_stream_wspace(const struct sock *sk) |
1da177e4 | 1018 | { |
ab4e846a ED |
1019 | return READ_ONCE(sk->sk_sndbuf) - READ_ONCE(sk->sk_wmem_queued); |
1020 | } | |
1021 | ||
1022 | static inline void sk_wmem_queued_add(struct sock *sk, int val) | |
1023 | { | |
1024 | WRITE_ONCE(sk->sk_wmem_queued, sk->sk_wmem_queued + val); | |
1da177e4 LT |
1025 | } |
1026 | ||
5e6300e7 ED |
1027 | static inline void sk_forward_alloc_add(struct sock *sk, int val) |
1028 | { | |
1029 | /* Paired with lockless reads of sk->sk_forward_alloc */ | |
1030 | WRITE_ONCE(sk->sk_forward_alloc, sk->sk_forward_alloc + val); | |
1031 | } | |
1032 | ||
69336bd2 | 1033 | void sk_stream_write_space(struct sock *sk); |
1da177e4 | 1034 | |
8eae939f | 1035 | /* OOB backlog add */ |
a3a858ff | 1036 | static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb) |
9ee6b535 | 1037 | { |
7fee226a | 1038 | /* dont let skb dst not refcounted, we are going to leave rcu lock */ |
222d7dbd | 1039 | skb_dst_force(skb); |
7fee226a ED |
1040 | |
1041 | if (!sk->sk_backlog.tail) | |
9ed498c6 | 1042 | WRITE_ONCE(sk->sk_backlog.head, skb); |
7fee226a | 1043 | else |
9ee6b535 | 1044 | sk->sk_backlog.tail->next = skb; |
7fee226a | 1045 | |
9ed498c6 | 1046 | WRITE_ONCE(sk->sk_backlog.tail, skb); |
9ee6b535 SH |
1047 | skb->next = NULL; |
1048 | } | |
1da177e4 | 1049 | |
c377411f ED |
1050 | /* |
1051 | * Take into account size of receive queue and backlog queue | |
0fd7bac6 ED |
1052 | * Do not take into account this skb truesize, |
1053 | * to allow even a single big packet to come. | |
c377411f | 1054 | */ |
274f482d | 1055 | static inline bool sk_rcvqueues_full(const struct sock *sk, unsigned int limit) |
c377411f ED |
1056 | { |
1057 | unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc); | |
1058 | ||
f545a38f | 1059 | return qsize > limit; |
c377411f ED |
1060 | } |
1061 | ||
8eae939f | 1062 | /* The per-socket spinlock must be held here. */ |
f545a38f ED |
1063 | static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb, |
1064 | unsigned int limit) | |
8eae939f | 1065 | { |
274f482d | 1066 | if (sk_rcvqueues_full(sk, limit)) |
8eae939f ZY |
1067 | return -ENOBUFS; |
1068 | ||
c7c49b8f ED |
1069 | /* |
1070 | * If the skb was allocated from pfmemalloc reserves, only | |
1071 | * allow SOCK_MEMALLOC sockets to use it as this socket is | |
1072 | * helping free memory | |
1073 | */ | |
1074 | if (skb_pfmemalloc(skb) && !sock_flag(sk, SOCK_MEMALLOC)) | |
1075 | return -ENOMEM; | |
1076 | ||
a3a858ff | 1077 | __sk_add_backlog(sk, skb); |
8eae939f ZY |
1078 | sk->sk_backlog.len += skb->truesize; |
1079 | return 0; | |
1080 | } | |
1081 | ||
69336bd2 | 1082 | int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb); |
b4b9e355 | 1083 | |
d2489c7b ED |
1084 | INDIRECT_CALLABLE_DECLARE(int tcp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)); |
1085 | INDIRECT_CALLABLE_DECLARE(int tcp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)); | |
1086 | ||
c57943a1 PZ |
1087 | static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb) |
1088 | { | |
b4b9e355 MG |
1089 | if (sk_memalloc_socks() && skb_pfmemalloc(skb)) |
1090 | return __sk_backlog_rcv(sk, skb); | |
1091 | ||
d2489c7b ED |
1092 | return INDIRECT_CALL_INET(sk->sk_backlog_rcv, |
1093 | tcp_v6_do_rcv, | |
1094 | tcp_v4_do_rcv, | |
1095 | sk, skb); | |
c57943a1 PZ |
1096 | } |
1097 | ||
2c8c56e1 ED |
1098 | static inline void sk_incoming_cpu_update(struct sock *sk) |
1099 | { | |
34cfb542 PA |
1100 | int cpu = raw_smp_processor_id(); |
1101 | ||
7170a977 ED |
1102 | if (unlikely(READ_ONCE(sk->sk_incoming_cpu) != cpu)) |
1103 | WRITE_ONCE(sk->sk_incoming_cpu, cpu); | |
2c8c56e1 ED |
1104 | } |
1105 | ||
fe477558 | 1106 | static inline void sock_rps_record_flow_hash(__u32 hash) |
c58dc01b DM |
1107 | { |
1108 | #ifdef CONFIG_RPS | |
1109 | struct rps_sock_flow_table *sock_flow_table; | |
1110 | ||
1111 | rcu_read_lock(); | |
1112 | sock_flow_table = rcu_dereference(rps_sock_flow_table); | |
fe477558 | 1113 | rps_record_sock_flow(sock_flow_table, hash); |
c58dc01b DM |
1114 | rcu_read_unlock(); |
1115 | #endif | |
1116 | } | |
1117 | ||
fe477558 TH |
1118 | static inline void sock_rps_record_flow(const struct sock *sk) |
1119 | { | |
c9d8ca04 | 1120 | #ifdef CONFIG_RPS |
dc05360f | 1121 | if (static_branch_unlikely(&rfs_needed)) { |
13bfff25 ED |
1122 | /* Reading sk->sk_rxhash might incur an expensive cache line |
1123 | * miss. | |
1124 | * | |
1125 | * TCP_ESTABLISHED does cover almost all states where RFS | |
1126 | * might be useful, and is cheaper [1] than testing : | |
1127 | * IPv4: inet_sk(sk)->inet_daddr | |
1128 | * IPv6: ipv6_addr_any(&sk->sk_v6_daddr) | |
1129 | * OR an additional socket flag | |
1130 | * [1] : sk_state and sk_prot are in the same cache line. | |
1131 | */ | |
1e5c647c ED |
1132 | if (sk->sk_state == TCP_ESTABLISHED) { |
1133 | /* This READ_ONCE() is paired with the WRITE_ONCE() | |
1134 | * from sock_rps_save_rxhash() and sock_rps_reset_rxhash(). | |
1135 | */ | |
1136 | sock_rps_record_flow_hash(READ_ONCE(sk->sk_rxhash)); | |
1137 | } | |
13bfff25 | 1138 | } |
c9d8ca04 | 1139 | #endif |
fe477558 TH |
1140 | } |
1141 | ||
bdeab991 TH |
1142 | static inline void sock_rps_save_rxhash(struct sock *sk, |
1143 | const struct sk_buff *skb) | |
c58dc01b DM |
1144 | { |
1145 | #ifdef CONFIG_RPS | |
1e5c647c ED |
1146 | /* The following WRITE_ONCE() is paired with the READ_ONCE() |
1147 | * here, and another one in sock_rps_record_flow(). | |
1148 | */ | |
1149 | if (unlikely(READ_ONCE(sk->sk_rxhash) != skb->hash)) | |
1150 | WRITE_ONCE(sk->sk_rxhash, skb->hash); | |
c58dc01b DM |
1151 | #endif |
1152 | } | |
1153 | ||
bdeab991 TH |
1154 | static inline void sock_rps_reset_rxhash(struct sock *sk) |
1155 | { | |
1156 | #ifdef CONFIG_RPS | |
1e5c647c ED |
1157 | /* Paired with READ_ONCE() in sock_rps_record_flow() */ |
1158 | WRITE_ONCE(sk->sk_rxhash, 0); | |
bdeab991 TH |
1159 | #endif |
1160 | } | |
1161 | ||
d9dc8b0f | 1162 | #define sk_wait_event(__sk, __timeo, __condition, __wait) \ |
419ce133 | 1163 | ({ int __rc, __dis = __sk->sk_disconnects; \ |
cfcabdcc SH |
1164 | release_sock(__sk); \ |
1165 | __rc = __condition; \ | |
1166 | if (!__rc) { \ | |
d9dc8b0f WC |
1167 | *(__timeo) = wait_woken(__wait, \ |
1168 | TASK_INTERRUPTIBLE, \ | |
1169 | *(__timeo)); \ | |
cfcabdcc | 1170 | } \ |
d9dc8b0f | 1171 | sched_annotate_sleep(); \ |
cfcabdcc | 1172 | lock_sock(__sk); \ |
419ce133 | 1173 | __rc = __dis == __sk->sk_disconnects ? __condition : -EPIPE; \ |
cfcabdcc SH |
1174 | __rc; \ |
1175 | }) | |
1da177e4 | 1176 | |
69336bd2 JP |
1177 | int sk_stream_wait_connect(struct sock *sk, long *timeo_p); |
1178 | int sk_stream_wait_memory(struct sock *sk, long *timeo_p); | |
1179 | void sk_stream_wait_close(struct sock *sk, long timeo_p); | |
1180 | int sk_stream_error(struct sock *sk, int flags, int err); | |
1181 | void sk_stream_kill_queues(struct sock *sk); | |
1182 | void sk_set_memalloc(struct sock *sk); | |
1183 | void sk_clear_memalloc(struct sock *sk); | |
1da177e4 | 1184 | |
d41a69f1 ED |
1185 | void __sk_flush_backlog(struct sock *sk); |
1186 | ||
1187 | static inline bool sk_flush_backlog(struct sock *sk) | |
1188 | { | |
1189 | if (unlikely(READ_ONCE(sk->sk_backlog.tail))) { | |
1190 | __sk_flush_backlog(sk); | |
1191 | return true; | |
1192 | } | |
1193 | return false; | |
1194 | } | |
1195 | ||
dfbafc99 | 1196 | int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb); |
1da177e4 | 1197 | |
60236fdd | 1198 | struct request_sock_ops; |
6d6ee43e | 1199 | struct timewait_sock_ops; |
ab1e0a13 | 1200 | struct inet_hashinfo; |
fc8717ba | 1201 | struct raw_hashinfo; |
f16a7dd5 | 1202 | struct smc_hashinfo; |
de477254 | 1203 | struct module; |
51e0158a | 1204 | struct sk_psock; |
2e6599cb | 1205 | |
f77d6021 | 1206 | /* |
5f0d5a3a | 1207 | * caches using SLAB_TYPESAFE_BY_RCU should let .next pointer from nulls nodes |
f77d6021 ED |
1208 | * un-modified. Special care is taken when initializing object to zero. |
1209 | */ | |
1210 | static inline void sk_prot_clear_nulls(struct sock *sk, int size) | |
1211 | { | |
1212 | if (offsetof(struct sock, sk_node.next) != 0) | |
1213 | memset(sk, 0, offsetof(struct sock, sk_node.next)); | |
1214 | memset(&sk->sk_node.pprev, 0, | |
1215 | size - offsetof(struct sock, sk_node.pprev)); | |
1216 | } | |
1217 | ||
1da177e4 LT |
1218 | /* Networking protocol blocks we attach to sockets. |
1219 | * socket layer -> transport layer interface | |
1da177e4 LT |
1220 | */ |
1221 | struct proto { | |
dc6b9b78 | 1222 | void (*close)(struct sock *sk, |
1da177e4 | 1223 | long timeout); |
d74bad4e AI |
1224 | int (*pre_connect)(struct sock *sk, |
1225 | struct sockaddr *uaddr, | |
1226 | int addr_len); | |
1da177e4 | 1227 | int (*connect)(struct sock *sk, |
dc6b9b78 | 1228 | struct sockaddr *uaddr, |
1da177e4 LT |
1229 | int addr_len); |
1230 | int (*disconnect)(struct sock *sk, int flags); | |
1231 | ||
cdfbabfb DH |
1232 | struct sock * (*accept)(struct sock *sk, int flags, int *err, |
1233 | bool kern); | |
1da177e4 LT |
1234 | |
1235 | int (*ioctl)(struct sock *sk, int cmd, | |
e1d001fa | 1236 | int *karg); |
1da177e4 | 1237 | int (*init)(struct sock *sk); |
7d06b2e0 | 1238 | void (*destroy)(struct sock *sk); |
1da177e4 | 1239 | void (*shutdown)(struct sock *sk, int how); |
dc6b9b78 | 1240 | int (*setsockopt)(struct sock *sk, int level, |
a7b75c5a | 1241 | int optname, sockptr_t optval, |
b7058842 | 1242 | unsigned int optlen); |
dc6b9b78 ED |
1243 | int (*getsockopt)(struct sock *sk, int level, |
1244 | int optname, char __user *optval, | |
1245 | int __user *option); | |
4b9d07a4 | 1246 | void (*keepalive)(struct sock *sk, int valbool); |
af01d537 | 1247 | #ifdef CONFIG_COMPAT |
709b46e8 EB |
1248 | int (*compat_ioctl)(struct sock *sk, |
1249 | unsigned int cmd, unsigned long arg); | |
af01d537 | 1250 | #endif |
1b784140 YX |
1251 | int (*sendmsg)(struct sock *sk, struct msghdr *msg, |
1252 | size_t len); | |
1253 | int (*recvmsg)(struct sock *sk, struct msghdr *msg, | |
ec095263 | 1254 | size_t len, int flags, int *addr_len); |
2bfc6685 | 1255 | void (*splice_eof)(struct socket *sock); |
dc6b9b78 | 1256 | int (*bind)(struct sock *sk, |
c0425a42 CH |
1257 | struct sockaddr *addr, int addr_len); |
1258 | int (*bind_add)(struct sock *sk, | |
1259 | struct sockaddr *addr, int addr_len); | |
1da177e4 | 1260 | |
dc6b9b78 | 1261 | int (*backlog_rcv) (struct sock *sk, |
1da177e4 | 1262 | struct sk_buff *skb); |
9cacf81f SF |
1263 | bool (*bpf_bypass_getsockopt)(int level, |
1264 | int optname); | |
1da177e4 | 1265 | |
46d3ceab ED |
1266 | void (*release_cb)(struct sock *sk); |
1267 | ||
1da177e4 | 1268 | /* Keeping track of sk's, looking them up, and port selection methods. */ |
086c653f | 1269 | int (*hash)(struct sock *sk); |
1da177e4 | 1270 | void (*unhash)(struct sock *sk); |
719f8358 | 1271 | void (*rehash)(struct sock *sk); |
1da177e4 | 1272 | int (*get_port)(struct sock *sk, unsigned short snum); |
91a760b2 | 1273 | void (*put_port)(struct sock *sk); |
8a59f9d1 | 1274 | #ifdef CONFIG_BPF_SYSCALL |
51e0158a CW |
1275 | int (*psock_update_sk_prot)(struct sock *sk, |
1276 | struct sk_psock *psock, | |
1277 | bool restore); | |
8a59f9d1 | 1278 | #endif |
1da177e4 | 1279 | |
286ab3d4 | 1280 | /* Keeping track of sockets in use */ |
65f76517 | 1281 | #ifdef CONFIG_PROC_FS |
13ff3d6f | 1282 | unsigned int inuse_idx; |
65f76517 | 1283 | #endif |
ebb53d75 | 1284 | |
6c302e79 | 1285 | #if IS_ENABLED(CONFIG_MPTCP) |
292e6077 | 1286 | int (*forward_alloc_get)(const struct sock *sk); |
6c302e79 | 1287 | #endif |
292e6077 | 1288 | |
a74f0fa0 | 1289 | bool (*stream_memory_free)(const struct sock *sk, int wake); |
7b50ecfc | 1290 | bool (*sock_is_readable)(struct sock *sk); |
1da177e4 | 1291 | /* Memory pressure */ |
5c52ba17 | 1292 | void (*enter_memory_pressure)(struct sock *sk); |
06044751 | 1293 | void (*leave_memory_pressure)(struct sock *sk); |
8d987e5c | 1294 | atomic_long_t *memory_allocated; /* Current allocated memory. */ |
0defbb0a | 1295 | int __percpu *per_cpu_fw_alloc; |
1748376b | 1296 | struct percpu_counter *sockets_allocated; /* Current number of sockets. */ |
292e6077 | 1297 | |
1da177e4 LT |
1298 | /* |
1299 | * Pressure flag: try to collapse. | |
1300 | * Technical note: it is used by multiple contexts non atomically. | |
76f33296 | 1301 | * Make sure to use READ_ONCE()/WRITE_ONCE() for all reads/writes. |
3ab224be | 1302 | * All the __sk_mem_schedule() is of this nature: accounting |
1da177e4 LT |
1303 | * is strict, actions are advisory and have some latency. |
1304 | */ | |
06044751 | 1305 | unsigned long *memory_pressure; |
8d987e5c | 1306 | long *sysctl_mem; |
a3dcaf17 | 1307 | |
1da177e4 LT |
1308 | int *sysctl_wmem; |
1309 | int *sysctl_rmem; | |
a3dcaf17 ED |
1310 | u32 sysctl_wmem_offset; |
1311 | u32 sysctl_rmem_offset; | |
1312 | ||
1da177e4 | 1313 | int max_header; |
7ba42910 | 1314 | bool no_autobind; |
1da177e4 | 1315 | |
271b72c7 | 1316 | struct kmem_cache *slab; |
1da177e4 | 1317 | unsigned int obj_size; |
f5f80e32 | 1318 | unsigned int ipv6_pinfo_offset; |
d50112ed | 1319 | slab_flags_t slab_flags; |
7bbdb81e AD |
1320 | unsigned int useroffset; /* Usercopy region offset */ |
1321 | unsigned int usersize; /* Usercopy region size */ | |
1da177e4 | 1322 | |
19757ceb | 1323 | unsigned int __percpu *orphan_count; |
8feaf0c0 | 1324 | |
60236fdd | 1325 | struct request_sock_ops *rsk_prot; |
6d6ee43e | 1326 | struct timewait_sock_ops *twsk_prot; |
2e6599cb | 1327 | |
39d8cda7 PE |
1328 | union { |
1329 | struct inet_hashinfo *hashinfo; | |
645ca708 | 1330 | struct udp_table *udp_table; |
fc8717ba | 1331 | struct raw_hashinfo *raw_hash; |
f16a7dd5 | 1332 | struct smc_hashinfo *smc_hash; |
39d8cda7 | 1333 | } h; |
ab1e0a13 | 1334 | |
1da177e4 LT |
1335 | struct module *owner; |
1336 | ||
1337 | char name[32]; | |
1338 | ||
1339 | struct list_head node; | |
64be0aed | 1340 | int (*diag_destroy)(struct sock *sk, int err); |
3859a271 | 1341 | } __randomize_layout; |
e1aab161 | 1342 | |
69336bd2 JP |
1343 | int proto_register(struct proto *prot, int alloc_slab); |
1344 | void proto_unregister(struct proto *prot); | |
bf2ae2e4 | 1345 | int sock_load_diag_module(int family, int protocol); |
1da177e4 | 1346 | |
1c5f2ced ED |
1347 | INDIRECT_CALLABLE_DECLARE(bool tcp_stream_memory_free(const struct sock *sk, int wake)); |
1348 | ||
292e6077 PA |
1349 | static inline int sk_forward_alloc_get(const struct sock *sk) |
1350 | { | |
6c302e79 ED |
1351 | #if IS_ENABLED(CONFIG_MPTCP) |
1352 | if (sk->sk_prot->forward_alloc_get) | |
1353 | return sk->sk_prot->forward_alloc_get(sk); | |
1354 | #endif | |
5e6300e7 | 1355 | return READ_ONCE(sk->sk_forward_alloc); |
292e6077 PA |
1356 | } |
1357 | ||
a74f0fa0 | 1358 | static inline bool __sk_stream_memory_free(const struct sock *sk, int wake) |
c9bee3b7 | 1359 | { |
ab4e846a | 1360 | if (READ_ONCE(sk->sk_wmem_queued) >= READ_ONCE(sk->sk_sndbuf)) |
c9bee3b7 ED |
1361 | return false; |
1362 | ||
1363 | return sk->sk_prot->stream_memory_free ? | |
a406290a ED |
1364 | INDIRECT_CALL_INET_1(sk->sk_prot->stream_memory_free, |
1365 | tcp_stream_memory_free, sk, wake) : true; | |
c9bee3b7 ED |
1366 | } |
1367 | ||
a74f0fa0 ED |
1368 | static inline bool sk_stream_memory_free(const struct sock *sk) |
1369 | { | |
1370 | return __sk_stream_memory_free(sk, 0); | |
1371 | } | |
1372 | ||
1373 | static inline bool __sk_stream_is_writeable(const struct sock *sk, int wake) | |
64dc6130 | 1374 | { |
c9bee3b7 | 1375 | return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) && |
a74f0fa0 ED |
1376 | __sk_stream_memory_free(sk, wake); |
1377 | } | |
1378 | ||
1379 | static inline bool sk_stream_is_writeable(const struct sock *sk) | |
1380 | { | |
1381 | return __sk_stream_is_writeable(sk, 0); | |
64dc6130 | 1382 | } |
e1aab161 | 1383 | |
54fd9c2d DB |
1384 | static inline int sk_under_cgroup_hierarchy(struct sock *sk, |
1385 | struct cgroup *ancestor) | |
1386 | { | |
1387 | #ifdef CONFIG_SOCK_CGROUP_DATA | |
1388 | return cgroup_is_descendant(sock_cgroup_ptr(&sk->sk_cgrp_data), | |
1389 | ancestor); | |
1390 | #else | |
1391 | return -ENOTSUPP; | |
1392 | #endif | |
1393 | } | |
c9bee3b7 | 1394 | |
180d8cd9 GC |
1395 | static inline bool sk_has_memory_pressure(const struct sock *sk) |
1396 | { | |
1397 | return sk->sk_prot->memory_pressure != NULL; | |
1398 | } | |
1399 | ||
2d0c88e8 AW |
1400 | static inline bool sk_under_global_memory_pressure(const struct sock *sk) |
1401 | { | |
1402 | return sk->sk_prot->memory_pressure && | |
76f33296 | 1403 | !!READ_ONCE(*sk->sk_prot->memory_pressure); |
2d0c88e8 AW |
1404 | } |
1405 | ||
180d8cd9 GC |
1406 | static inline bool sk_under_memory_pressure(const struct sock *sk) |
1407 | { | |
1408 | if (!sk->sk_prot->memory_pressure) | |
1409 | return false; | |
e1aab161 | 1410 | |
baac50bb JW |
1411 | if (mem_cgroup_sockets_enabled && sk->sk_memcg && |
1412 | mem_cgroup_under_socket_pressure(sk->sk_memcg)) | |
e805605c | 1413 | return true; |
e1aab161 | 1414 | |
76f33296 | 1415 | return !!READ_ONCE(*sk->sk_prot->memory_pressure); |
180d8cd9 GC |
1416 | } |
1417 | ||
3cd3399d ED |
1418 | static inline long |
1419 | proto_memory_allocated(const struct proto *prot) | |
1420 | { | |
1421 | return max(0L, atomic_long_read(prot->memory_allocated)); | |
1422 | } | |
1423 | ||
180d8cd9 GC |
1424 | static inline long |
1425 | sk_memory_allocated(const struct sock *sk) | |
1426 | { | |
3cd3399d | 1427 | return proto_memory_allocated(sk->sk_prot); |
180d8cd9 GC |
1428 | } |
1429 | ||
3cd3399d ED |
1430 | /* 1 MB per cpu, in page units */ |
1431 | #define SK_MEMORY_PCPU_RESERVE (1 << (20 - PAGE_SHIFT)) | |
1432 | ||
219160be | 1433 | static inline void |
e805605c | 1434 | sk_memory_allocated_add(struct sock *sk, int amt) |
180d8cd9 | 1435 | { |
3cd3399d ED |
1436 | int local_reserve; |
1437 | ||
1438 | preempt_disable(); | |
1439 | local_reserve = __this_cpu_add_return(*sk->sk_prot->per_cpu_fw_alloc, amt); | |
1440 | if (local_reserve >= SK_MEMORY_PCPU_RESERVE) { | |
1441 | __this_cpu_sub(*sk->sk_prot->per_cpu_fw_alloc, local_reserve); | |
1442 | atomic_long_add(local_reserve, sk->sk_prot->memory_allocated); | |
1443 | } | |
1444 | preempt_enable(); | |
180d8cd9 GC |
1445 | } |
1446 | ||
1447 | static inline void | |
0e90b31f | 1448 | sk_memory_allocated_sub(struct sock *sk, int amt) |
180d8cd9 | 1449 | { |
3cd3399d ED |
1450 | int local_reserve; |
1451 | ||
1452 | preempt_disable(); | |
1453 | local_reserve = __this_cpu_sub_return(*sk->sk_prot->per_cpu_fw_alloc, amt); | |
1454 | if (local_reserve <= -SK_MEMORY_PCPU_RESERVE) { | |
1455 | __this_cpu_sub(*sk->sk_prot->per_cpu_fw_alloc, local_reserve); | |
1456 | atomic_long_add(local_reserve, sk->sk_prot->memory_allocated); | |
1457 | } | |
1458 | preempt_enable(); | |
180d8cd9 GC |
1459 | } |
1460 | ||
f5a5589c WW |
1461 | #define SK_ALLOC_PERCPU_COUNTER_BATCH 16 |
1462 | ||
180d8cd9 GC |
1463 | static inline void sk_sockets_allocated_dec(struct sock *sk) |
1464 | { | |
f5a5589c WW |
1465 | percpu_counter_add_batch(sk->sk_prot->sockets_allocated, -1, |
1466 | SK_ALLOC_PERCPU_COUNTER_BATCH); | |
180d8cd9 GC |
1467 | } |
1468 | ||
1469 | static inline void sk_sockets_allocated_inc(struct sock *sk) | |
1470 | { | |
f5a5589c WW |
1471 | percpu_counter_add_batch(sk->sk_prot->sockets_allocated, 1, |
1472 | SK_ALLOC_PERCPU_COUNTER_BATCH); | |
180d8cd9 GC |
1473 | } |
1474 | ||
5bf325a5 | 1475 | static inline u64 |
180d8cd9 GC |
1476 | sk_sockets_allocated_read_positive(struct sock *sk) |
1477 | { | |
af95d7df | 1478 | return percpu_counter_read_positive(sk->sk_prot->sockets_allocated); |
180d8cd9 GC |
1479 | } |
1480 | ||
1481 | static inline int | |
1482 | proto_sockets_allocated_sum_positive(struct proto *prot) | |
1483 | { | |
1484 | return percpu_counter_sum_positive(prot->sockets_allocated); | |
1485 | } | |
1486 | ||
180d8cd9 GC |
1487 | static inline bool |
1488 | proto_memory_pressure(struct proto *prot) | |
1489 | { | |
1490 | if (!prot->memory_pressure) | |
1491 | return false; | |
76f33296 | 1492 | return !!READ_ONCE(*prot->memory_pressure); |
180d8cd9 GC |
1493 | } |
1494 | ||
65f76517 ED |
1495 | |
1496 | #ifdef CONFIG_PROC_FS | |
2a12ae5d ED |
1497 | #define PROTO_INUSE_NR 64 /* should be enough for the first time */ |
1498 | struct prot_inuse { | |
4199bae1 | 1499 | int all; |
2a12ae5d ED |
1500 | int val[PROTO_INUSE_NR]; |
1501 | }; | |
b3cb764a | 1502 | |
2a12ae5d ED |
1503 | static inline void sock_prot_inuse_add(const struct net *net, |
1504 | const struct proto *prot, int val) | |
1505 | { | |
b3cb764a | 1506 | this_cpu_add(net->core.prot_inuse->val[prot->inuse_idx], val); |
2a12ae5d | 1507 | } |
d477eb90 ED |
1508 | |
1509 | static inline void sock_inuse_add(const struct net *net, int val) | |
1510 | { | |
4199bae1 | 1511 | this_cpu_add(net->core.prot_inuse->all, val); |
d477eb90 ED |
1512 | } |
1513 | ||
69336bd2 | 1514 | int sock_prot_inuse_get(struct net *net, struct proto *proto); |
648845ab | 1515 | int sock_inuse_get(struct net *net); |
65f76517 | 1516 | #else |
2a12ae5d ED |
1517 | static inline void sock_prot_inuse_add(const struct net *net, |
1518 | const struct proto *prot, int val) | |
65f76517 ED |
1519 | { |
1520 | } | |
d477eb90 ED |
1521 | |
1522 | static inline void sock_inuse_add(const struct net *net, int val) | |
1523 | { | |
1524 | } | |
65f76517 ED |
1525 | #endif |
1526 | ||
1da177e4 | 1527 | |
614c6cb4 ACM |
1528 | /* With per-bucket locks this operation is not-atomic, so that |
1529 | * this version is not worse. | |
1530 | */ | |
086c653f | 1531 | static inline int __sk_prot_rehash(struct sock *sk) |
614c6cb4 ACM |
1532 | { |
1533 | sk->sk_prot->unhash(sk); | |
086c653f | 1534 | return sk->sk_prot->hash(sk); |
614c6cb4 ACM |
1535 | } |
1536 | ||
1da177e4 LT |
1537 | /* About 10 seconds */ |
1538 | #define SOCK_DESTROY_TIME (10*HZ) | |
1539 | ||
1540 | /* Sockets 0-1023 can't be bound to unless you are superuser */ | |
1541 | #define PROT_SOCK 1024 | |
1542 | ||
1543 | #define SHUTDOWN_MASK 3 | |
1544 | #define RCV_SHUTDOWN 1 | |
1545 | #define SEND_SHUTDOWN 2 | |
1546 | ||
1da177e4 LT |
1547 | #define SOCK_BINDADDR_LOCK 4 |
1548 | #define SOCK_BINDPORT_LOCK 8 | |
1549 | ||
1da177e4 LT |
1550 | struct socket_alloc { |
1551 | struct socket socket; | |
1552 | struct inode vfs_inode; | |
1553 | }; | |
1554 | ||
1555 | static inline struct socket *SOCKET_I(struct inode *inode) | |
1556 | { | |
1557 | return &container_of(inode, struct socket_alloc, vfs_inode)->socket; | |
1558 | } | |
1559 | ||
1560 | static inline struct inode *SOCK_INODE(struct socket *socket) | |
1561 | { | |
1562 | return &container_of(socket, struct socket_alloc, socket)->vfs_inode; | |
1563 | } | |
1564 | ||
3ab224be HA |
1565 | /* |
1566 | * Functions for memory accounting | |
1567 | */ | |
f8c3bf00 | 1568 | int __sk_mem_raise_allocated(struct sock *sk, int size, int amt, int kind); |
69336bd2 | 1569 | int __sk_mem_schedule(struct sock *sk, int size, int kind); |
f8c3bf00 | 1570 | void __sk_mem_reduce_allocated(struct sock *sk, int amount); |
1a24e04e | 1571 | void __sk_mem_reclaim(struct sock *sk, int amount); |
1da177e4 | 1572 | |
3ab224be HA |
1573 | #define SK_MEM_SEND 0 |
1574 | #define SK_MEM_RECV 1 | |
1da177e4 | 1575 | |
e70f3c70 | 1576 | /* sysctl_mem values are in pages */ |
bd68a2a8 ED |
1577 | static inline long sk_prot_mem_limits(const struct sock *sk, int index) |
1578 | { | |
816cd168 | 1579 | return READ_ONCE(sk->sk_prot->sysctl_mem[index]); |
bd68a2a8 ED |
1580 | } |
1581 | ||
3ab224be | 1582 | static inline int sk_mem_pages(int amt) |
1da177e4 | 1583 | { |
100fdd1f | 1584 | return (amt + PAGE_SIZE - 1) >> PAGE_SHIFT; |
1da177e4 LT |
1585 | } |
1586 | ||
dc6b9b78 | 1587 | static inline bool sk_has_account(struct sock *sk) |
1da177e4 | 1588 | { |
3ab224be HA |
1589 | /* return true if protocol supports memory accounting */ |
1590 | return !!sk->sk_prot->memory_allocated; | |
1da177e4 LT |
1591 | } |
1592 | ||
dc6b9b78 | 1593 | static inline bool sk_wmem_schedule(struct sock *sk, int size) |
1da177e4 | 1594 | { |
7c80b038 ED |
1595 | int delta; |
1596 | ||
3ab224be | 1597 | if (!sk_has_account(sk)) |
dc6b9b78 | 1598 | return true; |
7c80b038 ED |
1599 | delta = size - sk->sk_forward_alloc; |
1600 | return delta <= 0 || __sk_mem_schedule(sk, delta, SK_MEM_SEND); | |
1da177e4 LT |
1601 | } |
1602 | ||
c76562b6 | 1603 | static inline bool |
35c448a8 | 1604 | sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size) |
d80d99d6 | 1605 | { |
7c80b038 ED |
1606 | int delta; |
1607 | ||
3ab224be | 1608 | if (!sk_has_account(sk)) |
dc6b9b78 | 1609 | return true; |
7c80b038 ED |
1610 | delta = size - sk->sk_forward_alloc; |
1611 | return delta <= 0 || __sk_mem_schedule(sk, delta, SK_MEM_RECV) || | |
c76562b6 | 1612 | skb_pfmemalloc(skb); |
3ab224be HA |
1613 | } |
1614 | ||
2bb2f5fb WW |
1615 | static inline int sk_unused_reserved_mem(const struct sock *sk) |
1616 | { | |
1617 | int unused_mem; | |
1618 | ||
1619 | if (likely(!sk->sk_reserved_mem)) | |
1620 | return 0; | |
1621 | ||
1622 | unused_mem = sk->sk_reserved_mem - sk->sk_wmem_queued - | |
1623 | atomic_read(&sk->sk_rmem_alloc); | |
1624 | ||
1625 | return unused_mem > 0 ? unused_mem : 0; | |
1626 | } | |
1627 | ||
3ab224be HA |
1628 | static inline void sk_mem_reclaim(struct sock *sk) |
1629 | { | |
2bb2f5fb WW |
1630 | int reclaimable; |
1631 | ||
3ab224be HA |
1632 | if (!sk_has_account(sk)) |
1633 | return; | |
2bb2f5fb WW |
1634 | |
1635 | reclaimable = sk->sk_forward_alloc - sk_unused_reserved_mem(sk); | |
1636 | ||
100fdd1f | 1637 | if (reclaimable >= (int)PAGE_SIZE) |
2bb2f5fb WW |
1638 | __sk_mem_reclaim(sk, reclaimable); |
1639 | } | |
1640 | ||
1641 | static inline void sk_mem_reclaim_final(struct sock *sk) | |
1642 | { | |
1643 | sk->sk_reserved_mem = 0; | |
1644 | sk_mem_reclaim(sk); | |
3ab224be HA |
1645 | } |
1646 | ||
1647 | static inline void sk_mem_charge(struct sock *sk, int size) | |
1648 | { | |
1649 | if (!sk_has_account(sk)) | |
1650 | return; | |
5e6300e7 | 1651 | sk_forward_alloc_add(sk, -size); |
3ab224be HA |
1652 | } |
1653 | ||
1654 | static inline void sk_mem_uncharge(struct sock *sk, int size) | |
1655 | { | |
1656 | if (!sk_has_account(sk)) | |
1657 | return; | |
5e6300e7 | 1658 | sk_forward_alloc_add(sk, size); |
4890b686 | 1659 | sk_mem_reclaim(sk); |
3ab224be HA |
1660 | } |
1661 | ||
ed07536e PZ |
1662 | /* |
1663 | * Macro so as to not evaluate some arguments when | |
1664 | * lockdep is not enabled. | |
1665 | * | |
1666 | * Mark both the sk_lock and the sk_lock.slock as a | |
1667 | * per-address-family lock class. | |
1668 | */ | |
dc6b9b78 | 1669 | #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \ |
ed07536e | 1670 | do { \ |
e8f6fbf6 | 1671 | sk->sk_lock.owned = 0; \ |
ed07536e PZ |
1672 | init_waitqueue_head(&sk->sk_lock.wq); \ |
1673 | spin_lock_init(&(sk)->sk_lock.slock); \ | |
1674 | debug_check_no_locks_freed((void *)&(sk)->sk_lock, \ | |
1675 | sizeof((sk)->sk_lock)); \ | |
1676 | lockdep_set_class_and_name(&(sk)->sk_lock.slock, \ | |
dc6b9b78 | 1677 | (skey), (sname)); \ |
ed07536e PZ |
1678 | lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \ |
1679 | } while (0) | |
1680 | ||
05b93801 | 1681 | static inline bool lockdep_sock_is_held(const struct sock *sk) |
1e1d04e6 | 1682 | { |
1e1d04e6 HFS |
1683 | return lockdep_is_held(&sk->sk_lock) || |
1684 | lockdep_is_held(&sk->sk_lock.slock); | |
1685 | } | |
1686 | ||
69336bd2 | 1687 | void lock_sock_nested(struct sock *sk, int subclass); |
fcc70d5f PZ |
1688 | |
1689 | static inline void lock_sock(struct sock *sk) | |
1690 | { | |
1691 | lock_sock_nested(sk, 0); | |
1692 | } | |
1693 | ||
ad80b0fc | 1694 | void __lock_sock(struct sock *sk); |
8873c064 | 1695 | void __release_sock(struct sock *sk); |
69336bd2 | 1696 | void release_sock(struct sock *sk); |
1da177e4 LT |
1697 | |
1698 | /* BH context may only use the following locking interface. */ | |
1699 | #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock)) | |
c6366184 IM |
1700 | #define bh_lock_sock_nested(__sk) \ |
1701 | spin_lock_nested(&((__sk)->sk_lock.slock), \ | |
1702 | SINGLE_DEPTH_NESTING) | |
1da177e4 LT |
1703 | #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock)) |
1704 | ||
49054556 PA |
1705 | bool __lock_sock_fast(struct sock *sk) __acquires(&sk->sk_lock.slock); |
1706 | ||
1707 | /** | |
1708 | * lock_sock_fast - fast version of lock_sock | |
1709 | * @sk: socket | |
1710 | * | |
1711 | * This version should be used for very small section, where process wont block | |
1712 | * return false if fast path is taken: | |
1713 | * | |
1714 | * sk_lock.slock locked, owned = 0, BH disabled | |
1715 | * | |
1716 | * return true if slow path is taken: | |
1717 | * | |
1718 | * sk_lock.slock unlocked, owned = 1, BH enabled | |
1719 | */ | |
1720 | static inline bool lock_sock_fast(struct sock *sk) | |
1721 | { | |
1722 | /* The sk_lock has mutex_lock() semantics here. */ | |
1723 | mutex_acquire(&sk->sk_lock.dep_map, 0, 0, _RET_IP_); | |
1724 | ||
1725 | return __lock_sock_fast(sk); | |
1726 | } | |
1727 | ||
1728 | /* fast socket lock variant for caller already holding a [different] socket lock */ | |
1729 | static inline bool lock_sock_fast_nested(struct sock *sk) | |
1730 | { | |
1731 | mutex_acquire(&sk->sk_lock.dep_map, SINGLE_DEPTH_NESTING, 0, _RET_IP_); | |
1732 | ||
1733 | return __lock_sock_fast(sk); | |
1734 | } | |
12f4bd86 | 1735 | |
8a74ad60 ED |
1736 | /** |
1737 | * unlock_sock_fast - complement of lock_sock_fast | |
1738 | * @sk: socket | |
1739 | * @slow: slow mode | |
1740 | * | |
1741 | * fast unlock socket for user context. | |
1742 | * If slow mode is on, we call regular release_sock() | |
1743 | */ | |
1744 | static inline void unlock_sock_fast(struct sock *sk, bool slow) | |
12f4bd86 | 1745 | __releases(&sk->sk_lock.slock) |
4b0b72f7 | 1746 | { |
12f4bd86 | 1747 | if (slow) { |
8a74ad60 | 1748 | release_sock(sk); |
12f4bd86 PA |
1749 | __release(&sk->sk_lock.slock); |
1750 | } else { | |
2dcb96ba | 1751 | mutex_release(&sk->sk_lock.dep_map, _RET_IP_); |
8a74ad60 | 1752 | spin_unlock_bh(&sk->sk_lock.slock); |
12f4bd86 | 1753 | } |
4b0b72f7 ED |
1754 | } |
1755 | ||
24426654 MKL |
1756 | void sockopt_lock_sock(struct sock *sk); |
1757 | void sockopt_release_sock(struct sock *sk); | |
e42c7bee MKL |
1758 | bool sockopt_ns_capable(struct user_namespace *ns, int cap); |
1759 | bool sockopt_capable(int cap); | |
24426654 | 1760 | |
fafc4e1e HFS |
1761 | /* Used by processes to "lock" a socket state, so that |
1762 | * interrupts and bottom half handlers won't change it | |
1763 | * from under us. It essentially blocks any incoming | |
1764 | * packets, so that we won't get any new data or any | |
1765 | * packets that change the state of the socket. | |
1766 | * | |
1767 | * While locked, BH processing will add new packets to | |
1768 | * the backlog queue. This queue is processed by the | |
1769 | * owner of the socket lock right before it is released. | |
1770 | * | |
1771 | * Since ~2.3.5 it is also exclusive sleep lock serializing | |
1772 | * accesses from user process context. | |
1773 | */ | |
1774 | ||
46cc6e49 | 1775 | static inline void sock_owned_by_me(const struct sock *sk) |
fafc4e1e HFS |
1776 | { |
1777 | #ifdef CONFIG_LOCKDEP | |
5e91f6ce | 1778 | WARN_ON_ONCE(!lockdep_sock_is_held(sk) && debug_locks); |
fafc4e1e | 1779 | #endif |
46cc6e49 ED |
1780 | } |
1781 | ||
1782 | static inline bool sock_owned_by_user(const struct sock *sk) | |
1783 | { | |
1784 | sock_owned_by_me(sk); | |
fafc4e1e HFS |
1785 | return sk->sk_lock.owned; |
1786 | } | |
1787 | ||
602f7a27 TH |
1788 | static inline bool sock_owned_by_user_nocheck(const struct sock *sk) |
1789 | { | |
1790 | return sk->sk_lock.owned; | |
1791 | } | |
1792 | ||
33d60fbd KI |
1793 | static inline void sock_release_ownership(struct sock *sk) |
1794 | { | |
11445469 ED |
1795 | DEBUG_NET_WARN_ON_ONCE(!sock_owned_by_user_nocheck(sk)); |
1796 | sk->sk_lock.owned = 0; | |
33d60fbd | 1797 | |
11445469 ED |
1798 | /* The sk_lock has mutex_unlock() semantics: */ |
1799 | mutex_release(&sk->sk_lock.dep_map, _RET_IP_); | |
33d60fbd KI |
1800 | } |
1801 | ||
fafc4e1e HFS |
1802 | /* no reclassification while locks are held */ |
1803 | static inline bool sock_allow_reclassification(const struct sock *csk) | |
1804 | { | |
1805 | struct sock *sk = (struct sock *)csk; | |
1806 | ||
33d60fbd KI |
1807 | return !sock_owned_by_user_nocheck(sk) && |
1808 | !spin_is_locked(&sk->sk_lock.slock); | |
fafc4e1e | 1809 | } |
4b0b72f7 | 1810 | |
69336bd2 | 1811 | struct sock *sk_alloc(struct net *net, int family, gfp_t priority, |
11aa9c28 | 1812 | struct proto *prot, int kern); |
69336bd2 | 1813 | void sk_free(struct sock *sk); |
eb4cb008 | 1814 | void sk_destruct(struct sock *sk); |
69336bd2 | 1815 | struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority); |
94352d45 | 1816 | void sk_free_unlock_clone(struct sock *sk); |
69336bd2 JP |
1817 | |
1818 | struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force, | |
1819 | gfp_t priority); | |
1d2077ac | 1820 | void __sock_wfree(struct sk_buff *skb); |
69336bd2 | 1821 | void sock_wfree(struct sk_buff *skb); |
98ba0bd5 WB |
1822 | struct sk_buff *sock_omalloc(struct sock *sk, unsigned long size, |
1823 | gfp_t priority); | |
69336bd2 JP |
1824 | void skb_orphan_partial(struct sk_buff *skb); |
1825 | void sock_rfree(struct sk_buff *skb); | |
62bccb8c | 1826 | void sock_efree(struct sk_buff *skb); |
82eabd9e | 1827 | #ifdef CONFIG_INET |
69336bd2 | 1828 | void sock_edemux(struct sk_buff *skb); |
cf7fbe66 | 1829 | void sock_pfree(struct sk_buff *skb); |
82eabd9e | 1830 | #else |
158f323b | 1831 | #define sock_edemux sock_efree |
82eabd9e | 1832 | #endif |
69336bd2 | 1833 | |
29003875 MKL |
1834 | int sk_setsockopt(struct sock *sk, int level, int optname, |
1835 | sockptr_t optval, unsigned int optlen); | |
69336bd2 | 1836 | int sock_setsockopt(struct socket *sock, int level, int op, |
c8c1bbb6 | 1837 | sockptr_t optval, unsigned int optlen); |
1406245c BL |
1838 | int do_sock_setsockopt(struct socket *sock, bool compat, int level, |
1839 | int optname, sockptr_t optval, int optlen); | |
0b05b0cd BL |
1840 | int do_sock_getsockopt(struct socket *sock, bool compat, int level, |
1841 | int optname, sockptr_t optval, sockptr_t optlen); | |
69336bd2 | 1842 | |
65ddc82d MKL |
1843 | int sk_getsockopt(struct sock *sk, int level, int optname, |
1844 | sockptr_t optval, sockptr_t optlen); | |
c7cbdbf2 AB |
1845 | int sock_gettstamp(struct socket *sock, void __user *userstamp, |
1846 | bool timeval, bool time32); | |
69336bd2 JP |
1847 | struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len, |
1848 | unsigned long data_len, int noblock, | |
1849 | int *errcode, int max_page_order); | |
de32bc6a PB |
1850 | |
1851 | static inline struct sk_buff *sock_alloc_send_skb(struct sock *sk, | |
1852 | unsigned long size, | |
1853 | int noblock, int *errcode) | |
1854 | { | |
1855 | return sock_alloc_send_pskb(sk, size, 0, noblock, errcode, 0); | |
1856 | } | |
1857 | ||
69336bd2 JP |
1858 | void *sock_kmalloc(struct sock *sk, int size, gfp_t priority); |
1859 | void sock_kfree_s(struct sock *sk, void *mem, int size); | |
79e88659 | 1860 | void sock_kzfree_s(struct sock *sk, void *mem, int size); |
69336bd2 | 1861 | void sk_send_sigurg(struct sock *sk); |
1da177e4 | 1862 | |
fee9ac06 PB |
1863 | static inline void sock_replace_proto(struct sock *sk, struct proto *proto) |
1864 | { | |
1865 | if (sk->sk_socket) | |
1866 | clear_bit(SOCK_SUPPORT_ZC, &sk->sk_socket->flags); | |
1867 | WRITE_ONCE(sk->sk_prot, proto); | |
1868 | } | |
1869 | ||
f28ea365 | 1870 | struct sockcm_cookie { |
80b14dee | 1871 | u64 transmit_time; |
f28ea365 | 1872 | u32 mark; |
b534dc46 | 1873 | u32 tsflags; |
f28ea365 EJ |
1874 | }; |
1875 | ||
657a0667 WB |
1876 | static inline void sockcm_init(struct sockcm_cookie *sockc, |
1877 | const struct sock *sk) | |
1878 | { | |
e3390b30 ED |
1879 | *sockc = (struct sockcm_cookie) { |
1880 | .tsflags = READ_ONCE(sk->sk_tsflags) | |
1881 | }; | |
657a0667 WB |
1882 | } |
1883 | ||
233baf9a | 1884 | int __sock_cmsg_send(struct sock *sk, struct cmsghdr *cmsg, |
39771b12 | 1885 | struct sockcm_cookie *sockc); |
f28ea365 EJ |
1886 | int sock_cmsg_send(struct sock *sk, struct msghdr *msg, |
1887 | struct sockcm_cookie *sockc); | |
1888 | ||
1da177e4 LT |
1889 | /* |
1890 | * Functions to fill in entries in struct proto_ops when a protocol | |
1891 | * does not implement a particular function. | |
1892 | */ | |
69336bd2 JP |
1893 | int sock_no_bind(struct socket *, struct sockaddr *, int); |
1894 | int sock_no_connect(struct socket *, struct sockaddr *, int, int); | |
1895 | int sock_no_socketpair(struct socket *, struct socket *); | |
cdfbabfb | 1896 | int sock_no_accept(struct socket *, struct socket *, int, bool); |
9b2c45d4 | 1897 | int sock_no_getname(struct socket *, struct sockaddr *, int); |
69336bd2 JP |
1898 | int sock_no_ioctl(struct socket *, unsigned int, unsigned long); |
1899 | int sock_no_listen(struct socket *, int); | |
1900 | int sock_no_shutdown(struct socket *, int); | |
1b784140 | 1901 | int sock_no_sendmsg(struct socket *, struct msghdr *, size_t); |
306b13eb | 1902 | int sock_no_sendmsg_locked(struct sock *sk, struct msghdr *msg, size_t len); |
1b784140 | 1903 | int sock_no_recvmsg(struct socket *, struct msghdr *, size_t, int); |
69336bd2 JP |
1904 | int sock_no_mmap(struct file *file, struct socket *sock, |
1905 | struct vm_area_struct *vma); | |
1da177e4 LT |
1906 | |
1907 | /* | |
1908 | * Functions to fill in entries in struct proto_ops when a protocol | |
1909 | * uses the inet style. | |
1910 | */ | |
69336bd2 | 1911 | int sock_common_getsockopt(struct socket *sock, int level, int optname, |
1da177e4 | 1912 | char __user *optval, int __user *optlen); |
1b784140 YX |
1913 | int sock_common_recvmsg(struct socket *sock, struct msghdr *msg, size_t size, |
1914 | int flags); | |
69336bd2 | 1915 | int sock_common_setsockopt(struct socket *sock, int level, int optname, |
a7b75c5a | 1916 | sockptr_t optval, unsigned int optlen); |
1da177e4 | 1917 | |
69336bd2 | 1918 | void sk_common_release(struct sock *sk); |
1da177e4 LT |
1919 | |
1920 | /* | |
1921 | * Default socket callbacks and setup code | |
1922 | */ | |
dc6b9b78 | 1923 | |
584f3742 PB |
1924 | /* Initialise core socket variables using an explicit uid. */ |
1925 | void sock_init_data_uid(struct socket *sock, struct sock *sk, kuid_t uid); | |
1926 | ||
1927 | /* Initialise core socket variables. | |
1928 | * Assumes struct socket *sock is embedded in a struct socket_alloc. | |
1929 | */ | |
69336bd2 | 1930 | void sock_init_data(struct socket *sock, struct sock *sk); |
1da177e4 | 1931 | |
1da177e4 LT |
1932 | /* |
1933 | * Socket reference counting postulates. | |
1934 | * | |
1935 | * * Each user of socket SHOULD hold a reference count. | |
1936 | * * Each access point to socket (an hash table bucket, reference from a list, | |
1937 | * running timer, skb in flight MUST hold a reference count. | |
1938 | * * When reference count hits 0, it means it will never increase back. | |
1939 | * * When reference count hits 0, it means that no references from | |
1940 | * outside exist to this socket and current process on current CPU | |
1941 | * is last user and may/should destroy this socket. | |
1942 | * * sk_free is called from any context: process, BH, IRQ. When | |
1943 | * it is called, socket has no references from outside -> sk_free | |
1944 | * may release descendant resources allocated by the socket, but | |
1945 | * to the time when it is called, socket is NOT referenced by any | |
1946 | * hash tables, lists etc. | |
1947 | * * Packets, delivered from outside (from network or from another process) | |
1948 | * and enqueued on receive/error queues SHOULD NOT grab reference count, | |
1949 | * when they sit in queue. Otherwise, packets will leak to hole, when | |
1950 | * socket is looked up by one cpu and unhasing is made by another CPU. | |
1951 | * It is true for udp/raw, netlink (leak to receive and error queues), tcp | |
1952 | * (leak to backlog). Packet socket does all the processing inside | |
1953 | * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets | |
1954 | * use separate SMP lock, so that they are prone too. | |
1955 | */ | |
1956 | ||
1957 | /* Ungrab socket and destroy it, if it was the last reference. */ | |
1958 | static inline void sock_put(struct sock *sk) | |
1959 | { | |
41c6d650 | 1960 | if (refcount_dec_and_test(&sk->sk_refcnt)) |
1da177e4 LT |
1961 | sk_free(sk); |
1962 | } | |
05dbc7b5 | 1963 | /* Generic version of sock_put(), dealing with all sockets |
41b822c5 | 1964 | * (TCP_TIMEWAIT, TCP_NEW_SYN_RECV, ESTABLISHED...) |
05dbc7b5 ED |
1965 | */ |
1966 | void sock_gen_put(struct sock *sk); | |
1da177e4 | 1967 | |
4f0c40d9 | 1968 | int __sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested, |
c3f24cfb | 1969 | unsigned int trim_cap, bool refcounted); |
4f0c40d9 WB |
1970 | static inline int sk_receive_skb(struct sock *sk, struct sk_buff *skb, |
1971 | const int nested) | |
1972 | { | |
c3f24cfb | 1973 | return __sk_receive_skb(sk, skb, nested, 1, true); |
4f0c40d9 | 1974 | } |
25995ff5 | 1975 | |
e022f0b4 KK |
1976 | static inline void sk_tx_queue_set(struct sock *sk, int tx_queue) |
1977 | { | |
755c31cd AN |
1978 | /* sk_tx_queue_mapping accept only upto a 16-bit value */ |
1979 | if (WARN_ON_ONCE((unsigned short)tx_queue >= USHRT_MAX)) | |
1980 | return; | |
0bb4d124 ED |
1981 | /* Paired with READ_ONCE() in sk_tx_queue_get() and |
1982 | * other WRITE_ONCE() because socket lock might be not held. | |
1983 | */ | |
1984 | WRITE_ONCE(sk->sk_tx_queue_mapping, tx_queue); | |
e022f0b4 KK |
1985 | } |
1986 | ||
755c31cd AN |
1987 | #define NO_QUEUE_MAPPING USHRT_MAX |
1988 | ||
e022f0b4 KK |
1989 | static inline void sk_tx_queue_clear(struct sock *sk) |
1990 | { | |
0bb4d124 ED |
1991 | /* Paired with READ_ONCE() in sk_tx_queue_get() and |
1992 | * other WRITE_ONCE() because socket lock might be not held. | |
1993 | */ | |
1994 | WRITE_ONCE(sk->sk_tx_queue_mapping, NO_QUEUE_MAPPING); | |
e022f0b4 KK |
1995 | } |
1996 | ||
1997 | static inline int sk_tx_queue_get(const struct sock *sk) | |
1998 | { | |
0bb4d124 ED |
1999 | if (sk) { |
2000 | /* Paired with WRITE_ONCE() in sk_tx_queue_clear() | |
2001 | * and sk_tx_queue_set(). | |
2002 | */ | |
2003 | int val = READ_ONCE(sk->sk_tx_queue_mapping); | |
755c31cd | 2004 | |
0bb4d124 ED |
2005 | if (val != NO_QUEUE_MAPPING) |
2006 | return val; | |
2007 | } | |
755c31cd | 2008 | return -1; |
e022f0b4 KK |
2009 | } |
2010 | ||
a37a0ee4 ED |
2011 | static inline void __sk_rx_queue_set(struct sock *sk, |
2012 | const struct sk_buff *skb, | |
2013 | bool force_set) | |
c6345ce7 | 2014 | { |
4e1beecc | 2015 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
c6345ce7 AN |
2016 | if (skb_rx_queue_recorded(skb)) { |
2017 | u16 rx_queue = skb_get_rx_queue(skb); | |
2018 | ||
a37a0ee4 ED |
2019 | if (force_set || |
2020 | unlikely(READ_ONCE(sk->sk_rx_queue_mapping) != rx_queue)) | |
342159ee | 2021 | WRITE_ONCE(sk->sk_rx_queue_mapping, rx_queue); |
c6345ce7 AN |
2022 | } |
2023 | #endif | |
2024 | } | |
2025 | ||
a37a0ee4 ED |
2026 | static inline void sk_rx_queue_set(struct sock *sk, const struct sk_buff *skb) |
2027 | { | |
2028 | __sk_rx_queue_set(sk, skb, true); | |
2029 | } | |
2030 | ||
2031 | static inline void sk_rx_queue_update(struct sock *sk, const struct sk_buff *skb) | |
2032 | { | |
2033 | __sk_rx_queue_set(sk, skb, false); | |
2034 | } | |
2035 | ||
c6345ce7 AN |
2036 | static inline void sk_rx_queue_clear(struct sock *sk) |
2037 | { | |
4e1beecc | 2038 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
09b89846 | 2039 | WRITE_ONCE(sk->sk_rx_queue_mapping, NO_QUEUE_MAPPING); |
c6345ce7 AN |
2040 | #endif |
2041 | } | |
2042 | ||
fc9bab24 AN |
2043 | static inline int sk_rx_queue_get(const struct sock *sk) |
2044 | { | |
4e1beecc | 2045 | #ifdef CONFIG_SOCK_RX_QUEUE_MAPPING |
09b89846 ED |
2046 | if (sk) { |
2047 | int res = READ_ONCE(sk->sk_rx_queue_mapping); | |
2048 | ||
2049 | if (res != NO_QUEUE_MAPPING) | |
2050 | return res; | |
2051 | } | |
4e1beecc | 2052 | #endif |
fc9bab24 AN |
2053 | |
2054 | return -1; | |
2055 | } | |
fc9bab24 | 2056 | |
972692e0 DM |
2057 | static inline void sk_set_socket(struct sock *sk, struct socket *sock) |
2058 | { | |
2059 | sk->sk_socket = sock; | |
2060 | } | |
2061 | ||
aa395145 ED |
2062 | static inline wait_queue_head_t *sk_sleep(struct sock *sk) |
2063 | { | |
eaefd110 ED |
2064 | BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0); |
2065 | return &rcu_dereference_raw(sk->sk_wq)->wait; | |
aa395145 | 2066 | } |
1da177e4 LT |
2067 | /* Detach socket from process context. |
2068 | * Announce socket dead, detach it from wait queue and inode. | |
2069 | * Note that parent inode held reference count on this struct sock, | |
2070 | * we do not release it in this function, because protocol | |
2071 | * probably wants some additional cleanups or even continuing | |
2072 | * to work with this socket (TCP). | |
2073 | */ | |
2074 | static inline void sock_orphan(struct sock *sk) | |
2075 | { | |
2076 | write_lock_bh(&sk->sk_callback_lock); | |
2077 | sock_set_flag(sk, SOCK_DEAD); | |
972692e0 | 2078 | sk_set_socket(sk, NULL); |
43815482 | 2079 | sk->sk_wq = NULL; |
1da177e4 LT |
2080 | write_unlock_bh(&sk->sk_callback_lock); |
2081 | } | |
2082 | ||
2083 | static inline void sock_graft(struct sock *sk, struct socket *parent) | |
2084 | { | |
0ffdaf5b | 2085 | WARN_ON(parent->sk); |
1da177e4 | 2086 | write_lock_bh(&sk->sk_callback_lock); |
333f7909 | 2087 | rcu_assign_pointer(sk->sk_wq, &parent->wq); |
1da177e4 | 2088 | parent->sk = sk; |
972692e0 | 2089 | sk_set_socket(sk, parent); |
86741ec2 | 2090 | sk->sk_uid = SOCK_INODE(parent)->i_uid; |
4237c75c | 2091 | security_sock_graft(sk, parent); |
1da177e4 LT |
2092 | write_unlock_bh(&sk->sk_callback_lock); |
2093 | } | |
2094 | ||
69336bd2 | 2095 | kuid_t sock_i_uid(struct sock *sk); |
25a9c8a4 | 2096 | unsigned long __sock_i_ino(struct sock *sk); |
69336bd2 | 2097 | unsigned long sock_i_ino(struct sock *sk); |
1da177e4 | 2098 | |
86741ec2 LC |
2099 | static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk) |
2100 | { | |
2101 | return sk ? sk->sk_uid : make_kuid(net->user_ns, 0); | |
2102 | } | |
2103 | ||
58d607d3 | 2104 | static inline u32 net_tx_rndhash(void) |
877d1f62 | 2105 | { |
a251c17a | 2106 | u32 v = get_random_u32(); |
58d607d3 ED |
2107 | |
2108 | return v ?: 1; | |
2109 | } | |
877d1f62 | 2110 | |
58d607d3 ED |
2111 | static inline void sk_set_txhash(struct sock *sk) |
2112 | { | |
b71eaed8 ED |
2113 | /* This pairs with READ_ONCE() in skb_set_hash_from_sk() */ |
2114 | WRITE_ONCE(sk->sk_txhash, net_tx_rndhash()); | |
877d1f62 TH |
2115 | } |
2116 | ||
9c30ae83 | 2117 | static inline bool sk_rethink_txhash(struct sock *sk) |
265f94ff | 2118 | { |
26859240 | 2119 | if (sk->sk_txhash && sk->sk_txrehash == SOCK_TXREHASH_ENABLED) { |
265f94ff | 2120 | sk_set_txhash(sk); |
9c30ae83 YC |
2121 | return true; |
2122 | } | |
2123 | return false; | |
265f94ff TH |
2124 | } |
2125 | ||
1da177e4 | 2126 | static inline struct dst_entry * |
5033f58d | 2127 | __sk_dst_get(const struct sock *sk) |
1da177e4 | 2128 | { |
1e1d04e6 HFS |
2129 | return rcu_dereference_check(sk->sk_dst_cache, |
2130 | lockdep_sock_is_held(sk)); | |
1da177e4 LT |
2131 | } |
2132 | ||
2133 | static inline struct dst_entry * | |
5033f58d | 2134 | sk_dst_get(const struct sock *sk) |
1da177e4 LT |
2135 | { |
2136 | struct dst_entry *dst; | |
2137 | ||
b6c6712a ED |
2138 | rcu_read_lock(); |
2139 | dst = rcu_dereference(sk->sk_dst_cache); | |
bc9d3a9f | 2140 | if (dst && !rcuref_get(&dst->__rcuref)) |
f8864972 | 2141 | dst = NULL; |
b6c6712a | 2142 | rcu_read_unlock(); |
1da177e4 LT |
2143 | return dst; |
2144 | } | |
2145 | ||
9c30ae83 | 2146 | static inline void __dst_negative_advice(struct sock *sk) |
b6c6712a ED |
2147 | { |
2148 | struct dst_entry *ndst, *dst = __sk_dst_get(sk); | |
2149 | ||
2150 | if (dst && dst->ops->negative_advice) { | |
2151 | ndst = dst->ops->negative_advice(dst); | |
2152 | ||
2153 | if (ndst != dst) { | |
2154 | rcu_assign_pointer(sk->sk_dst_cache, ndst); | |
0a6957e7 | 2155 | sk_tx_queue_clear(sk); |
eb44ad4e | 2156 | WRITE_ONCE(sk->sk_dst_pending_confirm, 0); |
b6c6712a ED |
2157 | } |
2158 | } | |
2159 | } | |
2160 | ||
9c30ae83 YC |
2161 | static inline void dst_negative_advice(struct sock *sk) |
2162 | { | |
2163 | sk_rethink_txhash(sk); | |
2164 | __dst_negative_advice(sk); | |
2165 | } | |
2166 | ||
1da177e4 LT |
2167 | static inline void |
2168 | __sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
2169 | { | |
2170 | struct dst_entry *old_dst; | |
2171 | ||
e022f0b4 | 2172 | sk_tx_queue_clear(sk); |
eb44ad4e | 2173 | WRITE_ONCE(sk->sk_dst_pending_confirm, 0); |
95964c6d ED |
2174 | old_dst = rcu_dereference_protected(sk->sk_dst_cache, |
2175 | lockdep_sock_is_held(sk)); | |
b6c6712a | 2176 | rcu_assign_pointer(sk->sk_dst_cache, dst); |
1da177e4 LT |
2177 | dst_release(old_dst); |
2178 | } | |
2179 | ||
2180 | static inline void | |
2181 | sk_dst_set(struct sock *sk, struct dst_entry *dst) | |
2182 | { | |
7f502361 ED |
2183 | struct dst_entry *old_dst; |
2184 | ||
2185 | sk_tx_queue_clear(sk); | |
eb44ad4e | 2186 | WRITE_ONCE(sk->sk_dst_pending_confirm, 0); |
5925a055 | 2187 | old_dst = xchg((__force struct dst_entry **)&sk->sk_dst_cache, dst); |
7f502361 | 2188 | dst_release(old_dst); |
1da177e4 LT |
2189 | } |
2190 | ||
2191 | static inline void | |
2192 | __sk_dst_reset(struct sock *sk) | |
2193 | { | |
b6c6712a | 2194 | __sk_dst_set(sk, NULL); |
1da177e4 LT |
2195 | } |
2196 | ||
2197 | static inline void | |
2198 | sk_dst_reset(struct sock *sk) | |
2199 | { | |
7f502361 | 2200 | sk_dst_set(sk, NULL); |
1da177e4 LT |
2201 | } |
2202 | ||
69336bd2 | 2203 | struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 2204 | |
69336bd2 | 2205 | struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie); |
1da177e4 | 2206 | |
9b8805a3 JA |
2207 | static inline void sk_dst_confirm(struct sock *sk) |
2208 | { | |
25c7a6d1 ED |
2209 | if (!READ_ONCE(sk->sk_dst_pending_confirm)) |
2210 | WRITE_ONCE(sk->sk_dst_pending_confirm, 1); | |
9b8805a3 JA |
2211 | } |
2212 | ||
4ff06203 JA |
2213 | static inline void sock_confirm_neigh(struct sk_buff *skb, struct neighbour *n) |
2214 | { | |
2215 | if (skb_get_dst_pending_confirm(skb)) { | |
2216 | struct sock *sk = skb->sk; | |
4ff06203 | 2217 | |
25c7a6d1 ED |
2218 | if (sk && READ_ONCE(sk->sk_dst_pending_confirm)) |
2219 | WRITE_ONCE(sk->sk_dst_pending_confirm, 0); | |
1e84dc6b | 2220 | neigh_confirm(n); |
4ff06203 JA |
2221 | } |
2222 | } | |
2223 | ||
d986f521 | 2224 | bool sk_mc_loop(const struct sock *sk); |
f60e5990 | 2225 | |
dc6b9b78 | 2226 | static inline bool sk_can_gso(const struct sock *sk) |
bcd76111 HX |
2227 | { |
2228 | return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type); | |
2229 | } | |
2230 | ||
69336bd2 | 2231 | void sk_setup_caps(struct sock *sk, struct dst_entry *dst); |
6cbb0df7 | 2232 | |
aba54656 | 2233 | static inline void sk_gso_disable(struct sock *sk) |
a465419b | 2234 | { |
aba54656 ED |
2235 | sk->sk_gso_disabled = 1; |
2236 | sk->sk_route_caps &= ~NETIF_F_GSO_MASK; | |
a465419b ED |
2237 | } |
2238 | ||
c6e1a0d1 | 2239 | static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb, |
57be5bda | 2240 | struct iov_iter *from, char *to, |
912d398d | 2241 | int copy, int offset) |
c6e1a0d1 TH |
2242 | { |
2243 | if (skb->ip_summed == CHECKSUM_NONE) { | |
57be5bda | 2244 | __wsum csum = 0; |
15e6cb46 | 2245 | if (!csum_and_copy_from_iter_full(to, copy, &csum, from)) |
57be5bda | 2246 | return -EFAULT; |
912d398d | 2247 | skb->csum = csum_block_add(skb->csum, csum, offset); |
c6e1a0d1 | 2248 | } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) { |
15e6cb46 | 2249 | if (!copy_from_iter_full_nocache(to, copy, from)) |
c6e1a0d1 | 2250 | return -EFAULT; |
15e6cb46 | 2251 | } else if (!copy_from_iter_full(to, copy, from)) |
c6e1a0d1 TH |
2252 | return -EFAULT; |
2253 | ||
2254 | return 0; | |
2255 | } | |
2256 | ||
2257 | static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb, | |
57be5bda | 2258 | struct iov_iter *from, int copy) |
c6e1a0d1 | 2259 | { |
912d398d | 2260 | int err, offset = skb->len; |
c6e1a0d1 | 2261 | |
912d398d WY |
2262 | err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy), |
2263 | copy, offset); | |
c6e1a0d1 | 2264 | if (err) |
912d398d | 2265 | __skb_trim(skb, offset); |
c6e1a0d1 TH |
2266 | |
2267 | return err; | |
2268 | } | |
2269 | ||
57be5bda | 2270 | static inline int skb_copy_to_page_nocache(struct sock *sk, struct iov_iter *from, |
c6e1a0d1 TH |
2271 | struct sk_buff *skb, |
2272 | struct page *page, | |
2273 | int off, int copy) | |
2274 | { | |
2275 | int err; | |
2276 | ||
912d398d WY |
2277 | err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off, |
2278 | copy, skb->len); | |
c6e1a0d1 TH |
2279 | if (err) |
2280 | return err; | |
2281 | ||
ede57d58 | 2282 | skb_len_add(skb, copy); |
ab4e846a | 2283 | sk_wmem_queued_add(sk, copy); |
c6e1a0d1 TH |
2284 | sk_mem_charge(sk, copy); |
2285 | return 0; | |
2286 | } | |
2287 | ||
c564039f ED |
2288 | /** |
2289 | * sk_wmem_alloc_get - returns write allocations | |
2290 | * @sk: socket | |
2291 | * | |
66256e0b | 2292 | * Return: sk_wmem_alloc minus initial offset of one |
c564039f ED |
2293 | */ |
2294 | static inline int sk_wmem_alloc_get(const struct sock *sk) | |
2295 | { | |
14afee4b | 2296 | return refcount_read(&sk->sk_wmem_alloc) - 1; |
c564039f ED |
2297 | } |
2298 | ||
2299 | /** | |
2300 | * sk_rmem_alloc_get - returns read allocations | |
2301 | * @sk: socket | |
2302 | * | |
66256e0b | 2303 | * Return: sk_rmem_alloc |
c564039f ED |
2304 | */ |
2305 | static inline int sk_rmem_alloc_get(const struct sock *sk) | |
2306 | { | |
2307 | return atomic_read(&sk->sk_rmem_alloc); | |
2308 | } | |
2309 | ||
2310 | /** | |
2311 | * sk_has_allocations - check if allocations are outstanding | |
2312 | * @sk: socket | |
2313 | * | |
66256e0b | 2314 | * Return: true if socket has write or read allocations |
c564039f | 2315 | */ |
dc6b9b78 | 2316 | static inline bool sk_has_allocations(const struct sock *sk) |
c564039f ED |
2317 | { |
2318 | return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk); | |
2319 | } | |
2320 | ||
a57de0b4 | 2321 | /** |
1ce0bf50 | 2322 | * skwq_has_sleeper - check if there are any waiting processes |
acfbe96a | 2323 | * @wq: struct socket_wq |
a57de0b4 | 2324 | * |
66256e0b | 2325 | * Return: true if socket_wq has waiting processes |
a57de0b4 | 2326 | * |
1ce0bf50 | 2327 | * The purpose of the skwq_has_sleeper and sock_poll_wait is to wrap the memory |
a57de0b4 JO |
2328 | * barrier call. They were added due to the race found within the tcp code. |
2329 | * | |
d651983d | 2330 | * Consider following tcp code paths:: |
a57de0b4 | 2331 | * |
d651983d MCC |
2332 | * CPU1 CPU2 |
2333 | * sys_select receive packet | |
a57de0b4 JO |
2334 | * ... ... |
2335 | * __add_wait_queue update tp->rcv_nxt | |
2336 | * ... ... | |
2337 | * tp->rcv_nxt check sock_def_readable | |
2338 | * ... { | |
43815482 ED |
2339 | * schedule rcu_read_lock(); |
2340 | * wq = rcu_dereference(sk->sk_wq); | |
2341 | * if (wq && waitqueue_active(&wq->wait)) | |
2342 | * wake_up_interruptible(&wq->wait) | |
a57de0b4 JO |
2343 | * ... |
2344 | * } | |
2345 | * | |
2346 | * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay | |
2347 | * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1 | |
2348 | * could then endup calling schedule and sleep forever if there are no more | |
2349 | * data on the socket. | |
ad462769 | 2350 | * |
a57de0b4 | 2351 | */ |
1ce0bf50 | 2352 | static inline bool skwq_has_sleeper(struct socket_wq *wq) |
a57de0b4 | 2353 | { |
1ce0bf50 | 2354 | return wq && wq_has_sleeper(&wq->wait); |
a57de0b4 JO |
2355 | } |
2356 | ||
2357 | /** | |
2358 | * sock_poll_wait - place memory barrier behind the poll_wait call. | |
2359 | * @filp: file | |
89ab066d | 2360 | * @sock: socket to wait on |
a57de0b4 JO |
2361 | * @p: poll_table |
2362 | * | |
43815482 | 2363 | * See the comments in the wq_has_sleeper function. |
a57de0b4 | 2364 | */ |
89ab066d KG |
2365 | static inline void sock_poll_wait(struct file *filp, struct socket *sock, |
2366 | poll_table *p) | |
a57de0b4 | 2367 | { |
d8bbd13b | 2368 | if (!poll_does_not_wait(p)) { |
333f7909 | 2369 | poll_wait(filp, &sock->wq.wait, p); |
dc6b9b78 | 2370 | /* We need to be sure we are in sync with the |
a57de0b4 JO |
2371 | * socket flags modification. |
2372 | * | |
43815482 | 2373 | * This memory barrier is paired in the wq_has_sleeper. |
dc6b9b78 | 2374 | */ |
a57de0b4 JO |
2375 | smp_mb(); |
2376 | } | |
2377 | } | |
2378 | ||
b73c3d0e TH |
2379 | static inline void skb_set_hash_from_sk(struct sk_buff *skb, struct sock *sk) |
2380 | { | |
b71eaed8 ED |
2381 | /* This pairs with WRITE_ONCE() in sk_set_txhash() */ |
2382 | u32 txhash = READ_ONCE(sk->sk_txhash); | |
2383 | ||
2384 | if (txhash) { | |
b73c3d0e | 2385 | skb->l4_hash = 1; |
b71eaed8 | 2386 | skb->hash = txhash; |
b73c3d0e TH |
2387 | } |
2388 | } | |
2389 | ||
9e17f8a4 ED |
2390 | void skb_set_owner_w(struct sk_buff *skb, struct sock *sk); |
2391 | ||
1da177e4 | 2392 | /* |
dc6b9b78 | 2393 | * Queue a received datagram if it will fit. Stream and sequenced |
1da177e4 LT |
2394 | * protocols can't normally use this as they need to fit buffers in |
2395 | * and play with them. | |
2396 | * | |
dc6b9b78 | 2397 | * Inlined as it's very short and called for pretty much every |
1da177e4 LT |
2398 | * packet ever received. |
2399 | */ | |
1da177e4 LT |
2400 | static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk) |
2401 | { | |
d55d87fd | 2402 | skb_orphan(skb); |
1da177e4 LT |
2403 | skb->sk = sk; |
2404 | skb->destructor = sock_rfree; | |
2405 | atomic_add(skb->truesize, &sk->sk_rmem_alloc); | |
3ab224be | 2406 | sk_mem_charge(sk, skb->truesize); |
1da177e4 LT |
2407 | } |
2408 | ||
098116e7 | 2409 | static inline __must_check bool skb_set_owner_sk_safe(struct sk_buff *skb, struct sock *sk) |
9adc89af PA |
2410 | { |
2411 | if (sk && refcount_inc_not_zero(&sk->sk_refcnt)) { | |
2412 | skb_orphan(skb); | |
2413 | skb->destructor = sock_efree; | |
2414 | skb->sk = sk; | |
098116e7 | 2415 | return true; |
9adc89af | 2416 | } |
098116e7 | 2417 | return false; |
9adc89af PA |
2418 | } |
2419 | ||
ca43ccf4 KI |
2420 | static inline struct sk_buff *skb_clone_and_charge_r(struct sk_buff *skb, struct sock *sk) |
2421 | { | |
2422 | skb = skb_clone(skb, sk_gfp_mask(sk, GFP_ATOMIC)); | |
2423 | if (skb) { | |
2424 | if (sk_rmem_schedule(sk, skb, skb->truesize)) { | |
2425 | skb_set_owner_r(skb, sk); | |
2426 | return skb; | |
2427 | } | |
2428 | __kfree_skb(skb); | |
2429 | } | |
2430 | return NULL; | |
2431 | } | |
2432 | ||
5e10da53 PA |
2433 | static inline void skb_prepare_for_gro(struct sk_buff *skb) |
2434 | { | |
2435 | if (skb->destructor != sock_wfree) { | |
2436 | skb_orphan(skb); | |
2437 | return; | |
2438 | } | |
2439 | skb->slow_gro = 1; | |
2440 | } | |
2441 | ||
69336bd2 JP |
2442 | void sk_reset_timer(struct sock *sk, struct timer_list *timer, |
2443 | unsigned long expires); | |
1da177e4 | 2444 | |
69336bd2 | 2445 | void sk_stop_timer(struct sock *sk, struct timer_list *timer); |
1da177e4 | 2446 | |
08b81d87 GT |
2447 | void sk_stop_timer_sync(struct sock *sk, struct timer_list *timer); |
2448 | ||
65101aec PA |
2449 | int __sk_queue_drop_skb(struct sock *sk, struct sk_buff_head *sk_queue, |
2450 | struct sk_buff *skb, unsigned int flags, | |
69629464 ED |
2451 | void (*destructor)(struct sock *sk, |
2452 | struct sk_buff *skb)); | |
e6afc8ac | 2453 | int __sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb); |
c1b8a567 MD |
2454 | |
2455 | int sock_queue_rcv_skb_reason(struct sock *sk, struct sk_buff *skb, | |
2456 | enum skb_drop_reason *reason); | |
2457 | ||
2458 | static inline int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb) | |
2459 | { | |
2460 | return sock_queue_rcv_skb_reason(sk, skb, NULL); | |
2461 | } | |
1da177e4 | 2462 | |
69336bd2 | 2463 | int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb); |
364a9e93 | 2464 | struct sk_buff *sock_dequeue_err_skb(struct sock *sk); |
1da177e4 LT |
2465 | |
2466 | /* | |
2467 | * Recover an error report and clear atomically | |
2468 | */ | |
dc6b9b78 | 2469 | |
1da177e4 LT |
2470 | static inline int sock_error(struct sock *sk) |
2471 | { | |
c1cbe4b7 | 2472 | int err; |
f13ef100 ED |
2473 | |
2474 | /* Avoid an atomic operation for the common case. | |
2475 | * This is racy since another cpu/thread can change sk_err under us. | |
2476 | */ | |
2477 | if (likely(data_race(!sk->sk_err))) | |
c1cbe4b7 | 2478 | return 0; |
f13ef100 | 2479 | |
c1cbe4b7 | 2480 | err = xchg(&sk->sk_err, 0); |
1da177e4 LT |
2481 | return -err; |
2482 | } | |
2483 | ||
e3ae2365 AA |
2484 | void sk_error_report(struct sock *sk); |
2485 | ||
1da177e4 LT |
2486 | static inline unsigned long sock_wspace(struct sock *sk) |
2487 | { | |
2488 | int amt = 0; | |
2489 | ||
2490 | if (!(sk->sk_shutdown & SEND_SHUTDOWN)) { | |
14afee4b | 2491 | amt = sk->sk_sndbuf - refcount_read(&sk->sk_wmem_alloc); |
dc6b9b78 | 2492 | if (amt < 0) |
1da177e4 LT |
2493 | amt = 0; |
2494 | } | |
2495 | return amt; | |
2496 | } | |
2497 | ||
ceb5d58b ED |
2498 | /* Note: |
2499 | * We use sk->sk_wq_raw, from contexts knowing this | |
2500 | * pointer is not NULL and cannot disappear/change. | |
2501 | */ | |
9cd3e072 | 2502 | static inline void sk_set_bit(int nr, struct sock *sk) |
1da177e4 | 2503 | { |
4be73522 ED |
2504 | if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) && |
2505 | !sock_flag(sk, SOCK_FASYNC)) | |
9317bb69 ED |
2506 | return; |
2507 | ||
ceb5d58b | 2508 | set_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
2509 | } |
2510 | ||
2511 | static inline void sk_clear_bit(int nr, struct sock *sk) | |
2512 | { | |
4be73522 ED |
2513 | if ((nr == SOCKWQ_ASYNC_NOSPACE || nr == SOCKWQ_ASYNC_WAITDATA) && |
2514 | !sock_flag(sk, SOCK_FASYNC)) | |
9317bb69 ED |
2515 | return; |
2516 | ||
ceb5d58b | 2517 | clear_bit(nr, &sk->sk_wq_raw->flags); |
9cd3e072 ED |
2518 | } |
2519 | ||
ceb5d58b | 2520 | static inline void sk_wake_async(const struct sock *sk, int how, int band) |
1da177e4 | 2521 | { |
ceb5d58b ED |
2522 | if (sock_flag(sk, SOCK_FASYNC)) { |
2523 | rcu_read_lock(); | |
2524 | sock_wake_async(rcu_dereference(sk->sk_wq), how, band); | |
2525 | rcu_read_unlock(); | |
2526 | } | |
1da177e4 LT |
2527 | } |
2528 | ||
eea86af6 DB |
2529 | /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might |
2530 | * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak. | |
2531 | * Note: for send buffers, TCP works better if we can build two skbs at | |
2532 | * minimum. | |
7a91b434 | 2533 | */ |
9eb5bf83 | 2534 | #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff))) |
eea86af6 DB |
2535 | |
2536 | #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2) | |
2537 | #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE | |
1da177e4 LT |
2538 | |
2539 | static inline void sk_stream_moderate_sndbuf(struct sock *sk) | |
2540 | { | |
e292f05e ED |
2541 | u32 val; |
2542 | ||
2543 | if (sk->sk_userlocks & SOCK_SNDBUF_LOCK) | |
2544 | return; | |
2545 | ||
2546 | val = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1); | |
ca057051 | 2547 | val = max_t(u32, val, sk_unused_reserved_mem(sk)); |
e292f05e ED |
2548 | |
2549 | WRITE_ONCE(sk->sk_sndbuf, max_t(u32, val, SOCK_MIN_SNDBUF)); | |
1da177e4 LT |
2550 | } |
2551 | ||
5640f768 ED |
2552 | /** |
2553 | * sk_page_frag - return an appropriate page_frag | |
2554 | * @sk: socket | |
2555 | * | |
20eb4f29 | 2556 | * Use the per task page_frag instead of the per socket one for |
dacb5d88 | 2557 | * optimization when we know that we're in process context and own |
20eb4f29 TH |
2558 | * everything that's associated with %current. |
2559 | * | |
dacb5d88 PA |
2560 | * Both direct reclaim and page faults can nest inside other |
2561 | * socket operations and end up recursing into sk_page_frag() | |
2562 | * while it's already in use: explicitly avoid task page_frag | |
08f65892 | 2563 | * when users disable sk_use_task_frag. |
66256e0b RD |
2564 | * |
2565 | * Return: a per task page_frag if context allows that, | |
2566 | * otherwise a per socket one. | |
5640f768 ED |
2567 | */ |
2568 | static inline struct page_frag *sk_page_frag(struct sock *sk) | |
1da177e4 | 2569 | { |
08f65892 | 2570 | if (sk->sk_use_task_frag) |
5640f768 | 2571 | return ¤t->task_frag; |
1da177e4 | 2572 | |
5640f768 | 2573 | return &sk->sk_frag; |
1da177e4 LT |
2574 | } |
2575 | ||
69336bd2 | 2576 | bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag); |
5640f768 | 2577 | |
1da177e4 LT |
2578 | /* |
2579 | * Default write policy as shown to user space via poll/select/SIGIO | |
2580 | */ | |
dc6b9b78 | 2581 | static inline bool sock_writeable(const struct sock *sk) |
1da177e4 | 2582 | { |
e292f05e | 2583 | return refcount_read(&sk->sk_wmem_alloc) < (READ_ONCE(sk->sk_sndbuf) >> 1); |
1da177e4 LT |
2584 | } |
2585 | ||
dd0fc66f | 2586 | static inline gfp_t gfp_any(void) |
1da177e4 | 2587 | { |
99709372 | 2588 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
1da177e4 LT |
2589 | } |
2590 | ||
4b1327be WW |
2591 | static inline gfp_t gfp_memcg_charge(void) |
2592 | { | |
720ca52b | 2593 | return in_softirq() ? GFP_ATOMIC : GFP_KERNEL; |
4b1327be WW |
2594 | } |
2595 | ||
dc6b9b78 | 2596 | static inline long sock_rcvtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2597 | { |
2598 | return noblock ? 0 : sk->sk_rcvtimeo; | |
2599 | } | |
2600 | ||
dc6b9b78 | 2601 | static inline long sock_sndtimeo(const struct sock *sk, bool noblock) |
1da177e4 LT |
2602 | { |
2603 | return noblock ? 0 : sk->sk_sndtimeo; | |
2604 | } | |
2605 | ||
2606 | static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len) | |
2607 | { | |
eac66402 ED |
2608 | int v = waitall ? len : min_t(int, READ_ONCE(sk->sk_rcvlowat), len); |
2609 | ||
2610 | return v ?: 1; | |
1da177e4 LT |
2611 | } |
2612 | ||
2613 | /* Alas, with timeout socket operations are not restartable. | |
2614 | * Compare this to poll(). | |
2615 | */ | |
2616 | static inline int sock_intr_errno(long timeo) | |
2617 | { | |
2618 | return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR; | |
2619 | } | |
2620 | ||
744d5a3e EB |
2621 | struct sock_skb_cb { |
2622 | u32 dropcount; | |
2623 | }; | |
2624 | ||
2625 | /* Store sock_skb_cb at the end of skb->cb[] so protocol families | |
2626 | * using skb->cb[] would keep using it directly and utilize its | |
2627 | * alignement guarantee. | |
2628 | */ | |
c593642c | 2629 | #define SOCK_SKB_CB_OFFSET ((sizeof_field(struct sk_buff, cb) - \ |
744d5a3e EB |
2630 | sizeof(struct sock_skb_cb))) |
2631 | ||
2632 | #define SOCK_SKB_CB(__skb) ((struct sock_skb_cb *)((__skb)->cb + \ | |
2633 | SOCK_SKB_CB_OFFSET)) | |
2634 | ||
b4772ef8 | 2635 | #define sock_skb_cb_check_size(size) \ |
744d5a3e | 2636 | BUILD_BUG_ON((size) > SOCK_SKB_CB_OFFSET) |
b4772ef8 | 2637 | |
3bc3b96f EB |
2638 | static inline void |
2639 | sock_skb_set_dropcount(const struct sock *sk, struct sk_buff *skb) | |
2640 | { | |
3665f381 ED |
2641 | SOCK_SKB_CB(skb)->dropcount = sock_flag(sk, SOCK_RXQ_OVFL) ? |
2642 | atomic_read(&sk->sk_drops) : 0; | |
3bc3b96f EB |
2643 | } |
2644 | ||
532182cd ED |
2645 | static inline void sk_drops_add(struct sock *sk, const struct sk_buff *skb) |
2646 | { | |
2647 | int segs = max_t(u16, 1, skb_shinfo(skb)->gso_segs); | |
2648 | ||
2649 | atomic_add(segs, &sk->sk_drops); | |
2650 | } | |
2651 | ||
3a0ed3e9 DD |
2652 | static inline ktime_t sock_read_timestamp(struct sock *sk) |
2653 | { | |
2654 | #if BITS_PER_LONG==32 | |
2655 | unsigned int seq; | |
2656 | ktime_t kt; | |
2657 | ||
2658 | do { | |
2659 | seq = read_seqbegin(&sk->sk_stamp_seq); | |
2660 | kt = sk->sk_stamp; | |
2661 | } while (read_seqretry(&sk->sk_stamp_seq, seq)); | |
2662 | ||
2663 | return kt; | |
2664 | #else | |
f75359f3 | 2665 | return READ_ONCE(sk->sk_stamp); |
3a0ed3e9 DD |
2666 | #endif |
2667 | } | |
2668 | ||
2669 | static inline void sock_write_timestamp(struct sock *sk, ktime_t kt) | |
2670 | { | |
2671 | #if BITS_PER_LONG==32 | |
2672 | write_seqlock(&sk->sk_stamp_seq); | |
2673 | sk->sk_stamp = kt; | |
2674 | write_sequnlock(&sk->sk_stamp_seq); | |
2675 | #else | |
f75359f3 | 2676 | WRITE_ONCE(sk->sk_stamp, kt); |
3a0ed3e9 DD |
2677 | #endif |
2678 | } | |
2679 | ||
69336bd2 JP |
2680 | void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk, |
2681 | struct sk_buff *skb); | |
2682 | void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk, | |
2683 | struct sk_buff *skb); | |
92f37fd2 | 2684 | |
dc6b9b78 | 2685 | static inline void |
1da177e4 LT |
2686 | sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb) |
2687 | { | |
20d49473 | 2688 | struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb); |
e3390b30 ED |
2689 | u32 tsflags = READ_ONCE(sk->sk_tsflags); |
2690 | ktime_t kt = skb->tstamp; | |
20d49473 PO |
2691 | /* |
2692 | * generate control messages if | |
b9f40e21 | 2693 | * - receive time stamping in software requested |
20d49473 | 2694 | * - software time stamp available and wanted |
20d49473 | 2695 | * - hardware time stamps available and wanted |
20d49473 PO |
2696 | */ |
2697 | if (sock_flag(sk, SOCK_RCVTSTAMP) || | |
e3390b30 ED |
2698 | (tsflags & SOF_TIMESTAMPING_RX_SOFTWARE) || |
2699 | (kt && tsflags & SOF_TIMESTAMPING_SOFTWARE) || | |
2456e855 | 2700 | (hwtstamps->hwtstamp && |
e3390b30 | 2701 | (tsflags & SOF_TIMESTAMPING_RAW_HARDWARE))) |
92f37fd2 ED |
2702 | __sock_recv_timestamp(msg, sk, skb); |
2703 | else | |
3a0ed3e9 | 2704 | sock_write_timestamp(sk, kt); |
6e3e939f | 2705 | |
eb6fba75 | 2706 | if (sock_flag(sk, SOCK_WIFI_STATUS) && skb_wifi_acked_valid(skb)) |
6e3e939f | 2707 | __sock_recv_wifi_status(msg, sk, skb); |
1da177e4 LT |
2708 | } |
2709 | ||
6fd1d51c EM |
2710 | void __sock_recv_cmsgs(struct msghdr *msg, struct sock *sk, |
2711 | struct sk_buff *skb); | |
767dd033 | 2712 | |
6c7c98ba | 2713 | #define SK_DEFAULT_STAMP (-1L * NSEC_PER_SEC) |
6fd1d51c EM |
2714 | static inline void sock_recv_cmsgs(struct msghdr *msg, struct sock *sk, |
2715 | struct sk_buff *skb) | |
767dd033 | 2716 | { |
6fd1d51c EM |
2717 | #define FLAGS_RECV_CMSGS ((1UL << SOCK_RXQ_OVFL) | \ |
2718 | (1UL << SOCK_RCVTSTAMP) | \ | |
2719 | (1UL << SOCK_RCVMARK)) | |
b9f40e21 WB |
2720 | #define TSFLAGS_ANY (SOF_TIMESTAMPING_SOFTWARE | \ |
2721 | SOF_TIMESTAMPING_RAW_HARDWARE) | |
767dd033 | 2722 | |
e3390b30 ED |
2723 | if (sk->sk_flags & FLAGS_RECV_CMSGS || |
2724 | READ_ONCE(sk->sk_tsflags) & TSFLAGS_ANY) | |
6fd1d51c | 2725 | __sock_recv_cmsgs(msg, sk, skb); |
d3fbff30 | 2726 | else if (unlikely(sock_flag(sk, SOCK_TIMESTAMP))) |
3a0ed3e9 | 2727 | sock_write_timestamp(sk, skb->tstamp); |
dfd9248c | 2728 | else if (unlikely(sock_read_timestamp(sk) == SK_DEFAULT_STAMP)) |
3a0ed3e9 | 2729 | sock_write_timestamp(sk, 0); |
767dd033 | 2730 | } |
3b885787 | 2731 | |
c14ac945 | 2732 | void __sock_tx_timestamp(__u16 tsflags, __u8 *tx_flags); |
67cc0d40 | 2733 | |
20d49473 | 2734 | /** |
8f932f76 | 2735 | * _sock_tx_timestamp - checks whether the outgoing packet is to be time stamped |
20d49473 | 2736 | * @sk: socket sending this packet |
c14ac945 | 2737 | * @tsflags: timestamping flags to use |
140c55d4 | 2738 | * @tx_flags: completed with instructions for time stamping |
8f932f76 | 2739 | * @tskey: filled in with next sk_tskey (not for TCP, which uses seqno) |
140c55d4 | 2740 | * |
d651983d | 2741 | * Note: callers should take care of initial ``*tx_flags`` value (usually 0) |
20d49473 | 2742 | */ |
8f932f76 WB |
2743 | static inline void _sock_tx_timestamp(struct sock *sk, __u16 tsflags, |
2744 | __u8 *tx_flags, __u32 *tskey) | |
67cc0d40 | 2745 | { |
8f932f76 | 2746 | if (unlikely(tsflags)) { |
c14ac945 | 2747 | __sock_tx_timestamp(tsflags, tx_flags); |
8f932f76 WB |
2748 | if (tsflags & SOF_TIMESTAMPING_OPT_ID && tskey && |
2749 | tsflags & SOF_TIMESTAMPING_TX_RECORD_MASK) | |
a1cdec57 | 2750 | *tskey = atomic_inc_return(&sk->sk_tskey) - 1; |
8f932f76 | 2751 | } |
67cc0d40 WB |
2752 | if (unlikely(sock_flag(sk, SOCK_WIFI_STATUS))) |
2753 | *tx_flags |= SKBTX_WIFI_STATUS; | |
2754 | } | |
20d49473 | 2755 | |
8f932f76 WB |
2756 | static inline void sock_tx_timestamp(struct sock *sk, __u16 tsflags, |
2757 | __u8 *tx_flags) | |
2758 | { | |
2759 | _sock_tx_timestamp(sk, tsflags, tx_flags, NULL); | |
2760 | } | |
2761 | ||
2762 | static inline void skb_setup_tx_timestamp(struct sk_buff *skb, __u16 tsflags) | |
2763 | { | |
2764 | _sock_tx_timestamp(skb->sk, tsflags, &skb_shinfo(skb)->tx_flags, | |
2765 | &skb_shinfo(skb)->tskey); | |
2766 | } | |
2767 | ||
a54d51fb ED |
2768 | static inline bool sk_is_inet(const struct sock *sk) |
2769 | { | |
2770 | int family = READ_ONCE(sk->sk_family); | |
2771 | ||
2772 | return family == AF_INET || family == AF_INET6; | |
2773 | } | |
2774 | ||
42f67eea ED |
2775 | static inline bool sk_is_tcp(const struct sock *sk) |
2776 | { | |
a54d51fb ED |
2777 | return sk_is_inet(sk) && |
2778 | sk->sk_type == SOCK_STREAM && | |
2779 | sk->sk_protocol == IPPROTO_TCP; | |
2780 | } | |
2781 | ||
2782 | static inline bool sk_is_udp(const struct sock *sk) | |
2783 | { | |
2784 | return sk_is_inet(sk) && | |
2785 | sk->sk_type == SOCK_DGRAM && | |
2786 | sk->sk_protocol == IPPROTO_UDP; | |
42f67eea ED |
2787 | } |
2788 | ||
8d665064 JF |
2789 | static inline bool sk_is_stream_unix(const struct sock *sk) |
2790 | { | |
2791 | return sk->sk_family == AF_UNIX && sk->sk_type == SOCK_STREAM; | |
2792 | } | |
2793 | ||
1da177e4 LT |
2794 | /** |
2795 | * sk_eat_skb - Release a skb if it is no longer needed | |
4dc3b16b PP |
2796 | * @sk: socket to eat this skb from |
2797 | * @skb: socket buffer to eat | |
1da177e4 LT |
2798 | * |
2799 | * This routine must be called with interrupts disabled or with the socket | |
2800 | * locked so that the sk_buff queue operation is ok. | |
2801 | */ | |
7bced397 | 2802 | static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb) |
1da177e4 LT |
2803 | { |
2804 | __skb_unlink(skb, &sk->sk_receive_queue); | |
2805 | __kfree_skb(skb); | |
2806 | } | |
2807 | ||
cf7fbe66 JS |
2808 | static inline bool |
2809 | skb_sk_is_prefetched(struct sk_buff *skb) | |
2810 | { | |
2811 | #ifdef CONFIG_INET | |
2812 | return skb->destructor == sock_pfree; | |
2813 | #else | |
2814 | return false; | |
2815 | #endif /* CONFIG_INET */ | |
2816 | } | |
2817 | ||
7ae215d2 JS |
2818 | /* This helper checks if a socket is a full socket, |
2819 | * ie _not_ a timewait or request socket. | |
2820 | */ | |
2821 | static inline bool sk_fullsock(const struct sock *sk) | |
2822 | { | |
2823 | return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT | TCPF_NEW_SYN_RECV); | |
2824 | } | |
2825 | ||
2826 | static inline bool | |
2827 | sk_is_refcounted(struct sock *sk) | |
2828 | { | |
2829 | /* Only full sockets have sk->sk_flags. */ | |
2830 | return !sk_fullsock(sk) || !sock_flag(sk, SOCK_RCU_FREE); | |
2831 | } | |
2832 | ||
71489e21 | 2833 | /** |
045065f0 LR |
2834 | * skb_steal_sock - steal a socket from an sk_buff |
2835 | * @skb: sk_buff to steal the socket from | |
2836 | * @refcounted: is set to true if the socket is reference-counted | |
9c02bec9 | 2837 | * @prefetched: is set to true if the socket was assigned from bpf |
71489e21 JS |
2838 | */ |
2839 | static inline struct sock * | |
9c02bec9 | 2840 | skb_steal_sock(struct sk_buff *skb, bool *refcounted, bool *prefetched) |
23542618 | 2841 | { |
efc27f8c | 2842 | if (skb->sk) { |
23542618 KK |
2843 | struct sock *sk = skb->sk; |
2844 | ||
71489e21 | 2845 | *refcounted = true; |
9c02bec9 LB |
2846 | *prefetched = skb_sk_is_prefetched(skb); |
2847 | if (*prefetched) | |
7ae215d2 | 2848 | *refcounted = sk_is_refcounted(sk); |
23542618 KK |
2849 | skb->destructor = NULL; |
2850 | skb->sk = NULL; | |
2851 | return sk; | |
2852 | } | |
9c02bec9 | 2853 | *prefetched = false; |
71489e21 | 2854 | *refcounted = false; |
23542618 KK |
2855 | return NULL; |
2856 | } | |
2857 | ||
ebf4e808 IL |
2858 | /* Checks if this SKB belongs to an HW offloaded socket |
2859 | * and whether any SW fallbacks are required based on dev. | |
41477662 | 2860 | * Check decrypted mark in case skb_orphan() cleared socket. |
ebf4e808 IL |
2861 | */ |
2862 | static inline struct sk_buff *sk_validate_xmit_skb(struct sk_buff *skb, | |
2863 | struct net_device *dev) | |
2864 | { | |
2865 | #ifdef CONFIG_SOCK_VALIDATE_XMIT | |
2866 | struct sock *sk = skb->sk; | |
2867 | ||
41477662 | 2868 | if (sk && sk_fullsock(sk) && sk->sk_validate_xmit_skb) { |
ebf4e808 | 2869 | skb = sk->sk_validate_xmit_skb(sk, dev, skb); |
41477662 JK |
2870 | #ifdef CONFIG_TLS_DEVICE |
2871 | } else if (unlikely(skb->decrypted)) { | |
2872 | pr_warn_ratelimited("unencrypted skb with no associated socket - dropping\n"); | |
2873 | kfree_skb(skb); | |
2874 | skb = NULL; | |
2875 | #endif | |
2876 | } | |
ebf4e808 IL |
2877 | #endif |
2878 | ||
2879 | return skb; | |
2880 | } | |
2881 | ||
e446f9df ED |
2882 | /* This helper checks if a socket is a LISTEN or NEW_SYN_RECV |
2883 | * SYNACK messages can be attached to either ones (depending on SYNCOOKIE) | |
2884 | */ | |
2885 | static inline bool sk_listener(const struct sock *sk) | |
2886 | { | |
2887 | return (1 << sk->sk_state) & (TCPF_LISTEN | TCPF_NEW_SYN_RECV); | |
2888 | } | |
2889 | ||
193d357d | 2890 | void sock_enable_timestamp(struct sock *sk, enum sock_flags flag); |
69336bd2 JP |
2891 | int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level, |
2892 | int type); | |
1da177e4 | 2893 | |
a3b299da EB |
2894 | bool sk_ns_capable(const struct sock *sk, |
2895 | struct user_namespace *user_ns, int cap); | |
2896 | bool sk_capable(const struct sock *sk, int cap); | |
2897 | bool sk_net_capable(const struct sock *sk, int cap); | |
2898 | ||
a2d133b1 JH |
2899 | void sk_get_meminfo(const struct sock *sk, u32 *meminfo); |
2900 | ||
eaa72dc4 ED |
2901 | /* Take into consideration the size of the struct sk_buff overhead in the |
2902 | * determination of these values, since that is non-constant across | |
2903 | * platforms. This makes socket queueing behavior and performance | |
2904 | * not depend upon such differences. | |
2905 | */ | |
2906 | #define _SK_MEM_PACKETS 256 | |
2907 | #define _SK_MEM_OVERHEAD SKB_TRUESIZE(256) | |
2908 | #define SK_WMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
2909 | #define SK_RMEM_MAX (_SK_MEM_OVERHEAD * _SK_MEM_PACKETS) | |
2910 | ||
1da177e4 LT |
2911 | extern __u32 sysctl_wmem_max; |
2912 | extern __u32 sysctl_rmem_max; | |
2913 | ||
b245be1f | 2914 | extern int sysctl_tstamp_allow_data; |
6baf1f41 | 2915 | |
20380731 ACM |
2916 | extern __u32 sysctl_wmem_default; |
2917 | extern __u32 sysctl_rmem_default; | |
20380731 | 2918 | |
723783d0 | 2919 | #define SKB_FRAG_PAGE_ORDER get_order(32768) |
ce27ec60 ED |
2920 | DECLARE_STATIC_KEY_FALSE(net_high_order_alloc_disable_key); |
2921 | ||
a3dcaf17 ED |
2922 | static inline int sk_get_wmem0(const struct sock *sk, const struct proto *proto) |
2923 | { | |
2924 | /* Does this proto have per netns sysctl_wmem ? */ | |
2925 | if (proto->sysctl_wmem_offset) | |
02739545 | 2926 | return READ_ONCE(*(int *)((void *)sock_net(sk) + proto->sysctl_wmem_offset)); |
a3dcaf17 | 2927 | |
02739545 | 2928 | return READ_ONCE(*proto->sysctl_wmem); |
a3dcaf17 ED |
2929 | } |
2930 | ||
2931 | static inline int sk_get_rmem0(const struct sock *sk, const struct proto *proto) | |
2932 | { | |
2933 | /* Does this proto have per netns sysctl_rmem ? */ | |
2934 | if (proto->sysctl_rmem_offset) | |
02739545 | 2935 | return READ_ONCE(*(int *)((void *)sock_net(sk) + proto->sysctl_rmem_offset)); |
a3dcaf17 | 2936 | |
02739545 | 2937 | return READ_ONCE(*proto->sysctl_rmem); |
a3dcaf17 ED |
2938 | } |
2939 | ||
c9f1f58d ED |
2940 | /* Default TCP Small queue budget is ~1 ms of data (1sec >> 10) |
2941 | * Some wifi drivers need to tweak it to get more chunks. | |
2942 | * They can use this helper from their ndo_start_xmit() | |
2943 | */ | |
2944 | static inline void sk_pacing_shift_update(struct sock *sk, int val) | |
2945 | { | |
7c68fa2b | 2946 | if (!sk || !sk_fullsock(sk) || READ_ONCE(sk->sk_pacing_shift) == val) |
c9f1f58d | 2947 | return; |
7c68fa2b | 2948 | WRITE_ONCE(sk->sk_pacing_shift, val); |
c9f1f58d ED |
2949 | } |
2950 | ||
54dc3e33 DA |
2951 | /* if a socket is bound to a device, check that the given device |
2952 | * index is either the same or that the socket is bound to an L3 | |
2953 | * master device and the given device index is also enslaved to | |
2954 | * that L3 master | |
2955 | */ | |
2956 | static inline bool sk_dev_equal_l3scope(struct sock *sk, int dif) | |
2957 | { | |
4c971d2f | 2958 | int bound_dev_if = READ_ONCE(sk->sk_bound_dev_if); |
54dc3e33 DA |
2959 | int mdif; |
2960 | ||
4c971d2f | 2961 | if (!bound_dev_if || bound_dev_if == dif) |
54dc3e33 DA |
2962 | return true; |
2963 | ||
2964 | mdif = l3mdev_master_ifindex_by_index(sock_net(sk), dif); | |
4c971d2f | 2965 | if (mdif && mdif == bound_dev_if) |
54dc3e33 DA |
2966 | return true; |
2967 | ||
2968 | return false; | |
2969 | } | |
2970 | ||
43a825af BT |
2971 | void sock_def_readable(struct sock *sk); |
2972 | ||
8ea204c2 | 2973 | int sock_bindtoindex(struct sock *sk, int ifindex, bool lock_sk); |
371087aa | 2974 | void sock_set_timestamp(struct sock *sk, int optname, bool valbool); |
d463126e YL |
2975 | int sock_set_timestamping(struct sock *sk, int optname, |
2976 | struct so_timestamping timestamping); | |
ced122d9 | 2977 | |
783da70e | 2978 | void sock_enable_timestamps(struct sock *sk); |
c433594c | 2979 | void sock_no_linger(struct sock *sk); |
ce3d9544 | 2980 | void sock_set_keepalive(struct sock *sk); |
6e434967 | 2981 | void sock_set_priority(struct sock *sk, u32 priority); |
26cfabf9 | 2982 | void sock_set_rcvbuf(struct sock *sk, int val); |
84d1c617 | 2983 | void sock_set_mark(struct sock *sk, u32 val); |
b58f0e8f | 2984 | void sock_set_reuseaddr(struct sock *sk); |
fe31a326 | 2985 | void sock_set_reuseport(struct sock *sk); |
76ee0785 | 2986 | void sock_set_sndtimeo(struct sock *sk, s64 secs); |
b58f0e8f | 2987 | |
c0425a42 CH |
2988 | int sock_bind_add(struct sock *sk, struct sockaddr *addr, int addr_len); |
2989 | ||
4c1e34c0 RP |
2990 | int sock_get_timeout(long timeo, void *optval, bool old_timeval); |
2991 | int sock_copy_user_timeval(struct __kernel_sock_timeval *tv, | |
2992 | sockptr_t optval, int optlen, bool old_timeval); | |
2993 | ||
e1d001fa BL |
2994 | int sock_ioctl_inout(struct sock *sk, unsigned int cmd, |
2995 | void __user *arg, void *karg, size_t size); | |
2996 | int sk_ioctl(struct sock *sk, unsigned int cmd, void __user *arg); | |
7b50ecfc CW |
2997 | static inline bool sk_is_readable(struct sock *sk) |
2998 | { | |
2999 | if (sk->sk_prot->sock_is_readable) | |
3000 | return sk->sk_prot->sock_is_readable(sk); | |
3001 | return false; | |
3002 | } | |
1da177e4 | 3003 | #endif /* _SOCK_H */ |