sock.h 61 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196
  1. /* Copyright (c) 2015 Samsung Electronics Co., Ltd. */
  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. *
  11. * Authors: Ross Biro
  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
  34. *
  35. *
  36. * This program is free software; you can redistribute it and/or
  37. * modify it under the terms of the GNU General Public License
  38. * as published by the Free Software Foundation; either version
  39. * 2 of the License, or (at your option) any later version.
  40. */
  41. /*
  42. * Changes:
  43. * KwnagHyun Kim <kh0304.kim@samsung.com> 2015/07/08
  44. * Baesung Park <baesung.park@samsung.com> 2015/07/08
  45. * Vignesh Saravanaperumal <vignesh1.s@samsung.com> 2015/07/08
  46. * Add codes to share UID/PID information
  47. *
  48. */
  49. #ifndef _SOCK_H
  50. #define _SOCK_H
  51. #include <linux/hardirq.h>
  52. #include <linux/kernel.h>
  53. #include <linux/list.h>
  54. #include <linux/list_nulls.h>
  55. #include <linux/timer.h>
  56. #include <linux/cache.h>
  57. #include <linux/lockdep.h>
  58. #include <linux/netdevice.h>
  59. #include <linux/skbuff.h> /* struct sk_buff */
  60. #include <linux/mm.h>
  61. #include <linux/security.h>
  62. #include <linux/slab.h>
  63. #include <linux/uaccess.h>
  64. #include <linux/memcontrol.h>
  65. #include <linux/res_counter.h>
  66. #include <linux/static_key.h>
  67. #include <linux/aio.h>
  68. #include <linux/sched.h>
  69. #include <linux/filter.h>
  70. #include <linux/rculist_nulls.h>
  71. #include <linux/poll.h>
  72. #include <linux/atomic.h>
  73. #include <net/dst.h>
  74. #include <net/checksum.h>
  75. #include <net/tcp_states.h>
  76. struct cgroup;
  77. struct cgroup_subsys;
  78. #ifdef CONFIG_NET
  79. int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss);
  80. void mem_cgroup_sockets_destroy(struct cgroup *cgrp);
  81. #else
  82. static inline
  83. int mem_cgroup_sockets_init(struct cgroup *cgrp, struct cgroup_subsys *ss)
  84. {
  85. return 0;
  86. }
  87. static inline
  88. void mem_cgroup_sockets_destroy(struct cgroup *cgrp)
  89. {
  90. }
  91. #endif
  92. /*
  93. * This structure really needs to be cleaned up.
  94. * Most of it is for TCP, and not used by any of
  95. * the other protocols.
  96. */
  97. /* Define this to get the SOCK_DBG debugging facility. */
  98. #define SOCK_DEBUGGING
  99. #ifdef SOCK_DEBUGGING
  100. #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
  101. printk(KERN_DEBUG msg); } while (0)
  102. #else
  103. /* Validate arguments and do nothing */
  104. static inline __printf(2, 3)
  105. void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
  106. {
  107. }
  108. #endif
  109. /* This is the per-socket lock. The spinlock provides a synchronization
  110. * between user contexts and software interrupt processing, whereas the
  111. * mini-semaphore synchronizes multiple users amongst themselves.
  112. */
  113. typedef struct {
  114. spinlock_t slock;
  115. int owned;
  116. wait_queue_head_t wq;
  117. /*
  118. * We express the mutex-alike socket_lock semantics
  119. * to the lock validator by explicitly managing
  120. * the slock as a lock variant (in addition to
  121. * the slock itself):
  122. */
  123. #ifdef CONFIG_DEBUG_LOCK_ALLOC
  124. struct lockdep_map dep_map;
  125. #endif
  126. } socket_lock_t;
  127. struct sock;
  128. struct proto;
  129. struct net;
  130. /**
  131. * struct sock_common - minimal network layer representation of sockets
  132. * @skc_daddr: Foreign IPv4 addr
  133. * @skc_rcv_saddr: Bound local IPv4 addr
  134. * @skc_hash: hash value used with various protocol lookup tables
  135. * @skc_u16hashes: two u16 hash values used by UDP lookup tables
  136. * @skc_family: network address family
  137. * @skc_state: Connection state
  138. * @skc_reuse: %SO_REUSEADDR setting
  139. * @skc_bound_dev_if: bound device index if != 0
  140. * @skc_bind_node: bind hash linkage for various protocol lookup tables
  141. * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
  142. * @skc_prot: protocol handlers inside a network family
  143. * @skc_net: reference to the network namespace of this socket
  144. * @skc_node: main hash linkage for various protocol lookup tables
  145. * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
  146. * @skc_tx_queue_mapping: tx queue number for this connection
  147. * @skc_refcnt: reference count
  148. *
  149. * This is the minimal network layer representation of sockets, the header
  150. * for struct sock and struct inet_timewait_sock.
  151. */
  152. struct sock_common {
  153. /* skc_daddr and skc_rcv_saddr must be grouped :
  154. * cf INET_MATCH() and INET_TW_MATCH()
  155. */
  156. __be32 skc_daddr;
  157. __be32 skc_rcv_saddr;
  158. union {
  159. unsigned int skc_hash;
  160. __u16 skc_u16hashes[2];
  161. };
  162. unsigned short skc_family;
  163. volatile unsigned char skc_state;
  164. unsigned char skc_reuse;
  165. int skc_bound_dev_if;
  166. int padding[2];
  167. union {
  168. struct hlist_node skc_bind_node;
  169. struct hlist_nulls_node skc_portaddr_node;
  170. };
  171. struct proto *skc_prot;
  172. #ifdef CONFIG_NET_NS
  173. struct net *skc_net;
  174. #endif
  175. /*
  176. * fields between dontcopy_begin/dontcopy_end
  177. * are not copied in sock_copy()
  178. */
  179. /* private: */
  180. int skc_dontcopy_begin[0];
  181. /* public: */
  182. union {
  183. struct hlist_node skc_node;
  184. struct hlist_nulls_node skc_nulls_node;
  185. };
  186. int skc_tx_queue_mapping;
  187. atomic_t skc_refcnt;
  188. /* private: */
  189. int skc_dontcopy_end[0];
  190. /* public: */
  191. };
  192. struct cg_proto;
  193. /**
  194. * struct sock - network layer representation of sockets
  195. * @__sk_common: shared layout with inet_timewait_sock
  196. * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
  197. * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
  198. * @sk_lock: synchronizer
  199. * @sk_rcvbuf: size of receive buffer in bytes
  200. * @sk_wq: sock wait queue and async head
  201. * @sk_dst_cache: destination cache
  202. * @sk_dst_lock: destination cache lock
  203. * @sk_policy: flow policy
  204. * @sk_receive_queue: incoming packets
  205. * @sk_wmem_alloc: transmit queue bytes committed
  206. * @sk_write_queue: Packet sending queue
  207. * @sk_async_wait_queue: DMA copied packets
  208. * @sk_omem_alloc: "o" is "option" or "other"
  209. * @sk_wmem_queued: persistent queue size
  210. * @sk_forward_alloc: space allocated forward
  211. * @sk_allocation: allocation mode
  212. * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
  213. * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
  214. * @sk_sndbuf: size of send buffer in bytes
  215. * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
  216. * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
  217. * @sk_no_check: %SO_NO_CHECK setting, wether or not checkup packets
  218. * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
  219. * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
  220. * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
  221. * @sk_gso_max_size: Maximum GSO segment size to build
  222. * @sk_gso_max_segs: Maximum number of GSO segments
  223. * @sk_lingertime: %SO_LINGER l_linger setting
  224. * @sk_backlog: always used with the per-socket spinlock held
  225. * @sk_callback_lock: used with the callbacks in the end of this struct
  226. * @sk_error_queue: rarely used
  227. * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
  228. * IPV6_ADDRFORM for instance)
  229. * @sk_err: last error
  230. * @sk_err_soft: errors that don't cause failure but are the cause of a
  231. * persistent failure not just 'timed out'
  232. * @sk_drops: raw/udp drops counter
  233. * @sk_ack_backlog: current listen backlog
  234. * @sk_max_ack_backlog: listen backlog set in listen()
  235. * @sk_priority: %SO_PRIORITY setting
  236. * @sk_cgrp_prioidx: socket group's priority map index
  237. * @sk_type: socket type (%SOCK_STREAM, etc)
  238. * @sk_protocol: which protocol this socket belongs in this network family
  239. * @sk_peer_pid: &struct pid for this socket's peer
  240. * @sk_peer_cred: %SO_PEERCRED setting
  241. * @sk_rcvlowat: %SO_RCVLOWAT setting
  242. * @sk_rcvtimeo: %SO_RCVTIMEO setting
  243. * @sk_sndtimeo: %SO_SNDTIMEO setting
  244. * @sk_rxhash: flow hash received from netif layer
  245. * @sk_filter: socket filtering instructions
  246. * @sk_protinfo: private area, net family specific, when not using slab
  247. * @sk_timer: sock cleanup timer
  248. * @sk_stamp: time stamp of last packet received
  249. * @sk_socket: Identd and reporting IO signals
  250. * @sk_user_data: RPC layer private data
  251. * @sk_sndmsg_page: cached page for sendmsg
  252. * @sk_sndmsg_off: cached offset for sendmsg
  253. * @sk_peek_off: current peek_offset value
  254. * @sk_send_head: front of stuff to transmit
  255. * @sk_security: used by security modules
  256. * @sk_mark: generic packet mark
  257. * @sk_classid: this socket's cgroup classid
  258. * @sk_cgrp: this socket's cgroup-specific proto data
  259. * @sk_write_pending: a write to stream socket waits to start
  260. * @sk_state_change: callback to indicate change in the state of the sock
  261. * @sk_data_ready: callback to indicate there is data to be processed
  262. * @sk_write_space: callback to indicate there is bf sending space available
  263. * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
  264. * @sk_backlog_rcv: callback to process the backlog
  265. * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
  266. */
  267. struct sock {
  268. /*
  269. * Now struct inet_timewait_sock also uses sock_common, so please just
  270. * don't add nothing before this first member (__sk_common) --acme
  271. */
  272. struct sock_common __sk_common;
  273. #define sk_node __sk_common.skc_node
  274. #define sk_nulls_node __sk_common.skc_nulls_node
  275. #define sk_refcnt __sk_common.skc_refcnt
  276. #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
  277. #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
  278. #define sk_dontcopy_end __sk_common.skc_dontcopy_end
  279. #define sk_hash __sk_common.skc_hash
  280. #define sk_family __sk_common.skc_family
  281. #define sk_state __sk_common.skc_state
  282. #define sk_reuse __sk_common.skc_reuse
  283. #define sk_bound_dev_if __sk_common.skc_bound_dev_if
  284. #define sk_bind_node __sk_common.skc_bind_node
  285. #define sk_prot __sk_common.skc_prot
  286. #define sk_net __sk_common.skc_net
  287. socket_lock_t sk_lock;
  288. struct sk_buff_head sk_receive_queue;
  289. /*
  290. * The backlog queue is special, it is always used with
  291. * the per-socket spinlock held and requires low latency
  292. * access. Therefore we special case it's implementation.
  293. * Note : rmem_alloc is in this structure to fill a hole
  294. * on 64bit arches, not because its logically part of
  295. * backlog.
  296. */
  297. struct {
  298. atomic_t rmem_alloc;
  299. int len;
  300. struct sk_buff *head;
  301. struct sk_buff *tail;
  302. } sk_backlog;
  303. #define sk_rmem_alloc sk_backlog.rmem_alloc
  304. int sk_forward_alloc;
  305. #ifdef CONFIG_RPS
  306. __u32 sk_rxhash;
  307. #endif
  308. atomic_t sk_drops;
  309. int sk_rcvbuf;
  310. struct sk_filter __rcu *sk_filter;
  311. struct socket_wq __rcu *sk_wq;
  312. #ifdef CONFIG_NET_DMA
  313. struct sk_buff_head sk_async_wait_queue;
  314. #endif
  315. #ifdef CONFIG_XFRM
  316. struct xfrm_policy *sk_policy[2];
  317. #endif
  318. unsigned long sk_flags;
  319. struct dst_entry *sk_dst_cache;
  320. spinlock_t sk_dst_lock;
  321. atomic_t sk_wmem_alloc;
  322. atomic_t sk_omem_alloc;
  323. int sk_sndbuf;
  324. struct sk_buff_head sk_write_queue;
  325. kmemcheck_bitfield_begin(flags);
  326. unsigned int sk_shutdown : 2,
  327. sk_no_check : 2,
  328. sk_userlocks : 4,
  329. sk_protocol : 8,
  330. sk_type : 16;
  331. #define SK_PROTOCOL_MAX ((u8)~0U)
  332. kmemcheck_bitfield_end(flags);
  333. int sk_wmem_queued;
  334. gfp_t sk_allocation;
  335. u32 sk_pacing_rate; /* bytes per second */
  336. u32 sk_max_pacing_rate;
  337. netdev_features_t sk_route_caps;
  338. netdev_features_t sk_route_nocaps;
  339. int sk_gso_type;
  340. unsigned int sk_gso_max_size;
  341. u16 sk_gso_max_segs;
  342. int sk_rcvlowat;
  343. unsigned long sk_lingertime;
  344. struct sk_buff_head sk_error_queue;
  345. struct proto *sk_prot_creator;
  346. rwlock_t sk_callback_lock;
  347. int sk_err,
  348. sk_err_soft;
  349. unsigned short sk_ack_backlog;
  350. unsigned short sk_max_ack_backlog;
  351. __u32 sk_priority;
  352. #ifdef CONFIG_CGROUPS
  353. __u32 sk_cgrp_prioidx;
  354. #endif
  355. struct pid *sk_peer_pid;
  356. const struct cred *sk_peer_cred;
  357. long sk_rcvtimeo;
  358. long sk_sndtimeo;
  359. void *sk_protinfo;
  360. struct timer_list sk_timer;
  361. ktime_t sk_stamp;
  362. struct socket *sk_socket;
  363. void *sk_user_data;
  364. struct page *sk_sndmsg_page;
  365. struct sk_buff *sk_send_head;
  366. __u32 sk_sndmsg_off;
  367. __s32 sk_peek_off;
  368. int sk_write_pending;
  369. #ifdef CONFIG_SECURITY
  370. void *sk_security;
  371. #endif
  372. __u32 sk_mark;
  373. kuid_t sk_uid;
  374. u32 sk_classid;
  375. struct cg_proto *sk_cgrp;
  376. void (*sk_state_change)(struct sock *sk);
  377. void (*sk_data_ready)(struct sock *sk, int bytes);
  378. void (*sk_write_space)(struct sock *sk);
  379. void (*sk_error_report)(struct sock *sk);
  380. int (*sk_backlog_rcv)(struct sock *sk,
  381. struct sk_buff *skb);
  382. void (*sk_destruct)(struct sock *sk);
  383. };
  384. #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
  385. #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
  386. #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
  387. static inline int sk_peek_offset(struct sock *sk, int flags)
  388. {
  389. if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
  390. return sk->sk_peek_off;
  391. else
  392. return 0;
  393. }
  394. static inline void sk_peek_offset_bwd(struct sock *sk, int val)
  395. {
  396. if (sk->sk_peek_off >= 0) {
  397. if (sk->sk_peek_off >= val)
  398. sk->sk_peek_off -= val;
  399. else
  400. sk->sk_peek_off = 0;
  401. }
  402. }
  403. static inline void sk_peek_offset_fwd(struct sock *sk, int val)
  404. {
  405. if (sk->sk_peek_off >= 0)
  406. sk->sk_peek_off += val;
  407. }
  408. /*
  409. * Hashed lists helper routines
  410. */
  411. static inline struct sock *sk_entry(const struct hlist_node *node)
  412. {
  413. return hlist_entry(node, struct sock, sk_node);
  414. }
  415. static inline struct sock *__sk_head(const struct hlist_head *head)
  416. {
  417. return hlist_entry(head->first, struct sock, sk_node);
  418. }
  419. static inline struct sock *sk_head(const struct hlist_head *head)
  420. {
  421. return hlist_empty(head) ? NULL : __sk_head(head);
  422. }
  423. static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
  424. {
  425. return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
  426. }
  427. static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
  428. {
  429. return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
  430. }
  431. static inline struct sock *sk_next(const struct sock *sk)
  432. {
  433. return sk->sk_node.next ?
  434. hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
  435. }
  436. static inline struct sock *sk_nulls_next(const struct sock *sk)
  437. {
  438. return (!is_a_nulls(sk->sk_nulls_node.next)) ?
  439. hlist_nulls_entry(sk->sk_nulls_node.next,
  440. struct sock, sk_nulls_node) :
  441. NULL;
  442. }
  443. static inline bool sk_unhashed(const struct sock *sk)
  444. {
  445. return hlist_unhashed(&sk->sk_node);
  446. }
  447. static inline bool sk_hashed(const struct sock *sk)
  448. {
  449. return !sk_unhashed(sk);
  450. }
  451. static inline void sk_node_init(struct hlist_node *node)
  452. {
  453. node->pprev = NULL;
  454. }
  455. static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
  456. {
  457. node->pprev = NULL;
  458. }
  459. static inline void __sk_del_node(struct sock *sk)
  460. {
  461. __hlist_del(&sk->sk_node);
  462. }
  463. /* NB: equivalent to hlist_del_init_rcu */
  464. static inline bool __sk_del_node_init(struct sock *sk)
  465. {
  466. if (sk_hashed(sk)) {
  467. __sk_del_node(sk);
  468. sk_node_init(&sk->sk_node);
  469. return true;
  470. }
  471. return false;
  472. }
  473. /* Grab socket reference count. This operation is valid only
  474. when sk is ALREADY grabbed f.e. it is found in hash table
  475. or a list and the lookup is made under lock preventing hash table
  476. modifications.
  477. */
  478. static inline void sock_hold(struct sock *sk)
  479. {
  480. atomic_inc(&sk->sk_refcnt);
  481. }
  482. /* Ungrab socket in the context, which assumes that socket refcnt
  483. cannot hit zero, f.e. it is true in context of any socketcall.
  484. */
  485. static inline void __sock_put(struct sock *sk)
  486. {
  487. atomic_dec(&sk->sk_refcnt);
  488. }
  489. static inline bool sk_del_node_init(struct sock *sk)
  490. {
  491. bool rc = __sk_del_node_init(sk);
  492. if (rc) {
  493. /* paranoid for a while -acme */
  494. WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
  495. __sock_put(sk);
  496. }
  497. return rc;
  498. }
  499. #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
  500. static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
  501. {
  502. if (sk_hashed(sk)) {
  503. hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
  504. return true;
  505. }
  506. return false;
  507. }
  508. static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
  509. {
  510. bool rc = __sk_nulls_del_node_init_rcu(sk);
  511. if (rc) {
  512. /* paranoid for a while -acme */
  513. WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
  514. __sock_put(sk);
  515. }
  516. return rc;
  517. }
  518. static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
  519. {
  520. hlist_add_head(&sk->sk_node, list);
  521. }
  522. static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
  523. {
  524. sock_hold(sk);
  525. __sk_add_node(sk, list);
  526. }
  527. static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
  528. {
  529. sock_hold(sk);
  530. hlist_add_head_rcu(&sk->sk_node, list);
  531. }
  532. static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
  533. {
  534. hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
  535. }
  536. static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
  537. {
  538. sock_hold(sk);
  539. __sk_nulls_add_node_rcu(sk, list);
  540. }
  541. static inline void __sk_del_bind_node(struct sock *sk)
  542. {
  543. __hlist_del(&sk->sk_bind_node);
  544. }
  545. static inline void sk_add_bind_node(struct sock *sk,
  546. struct hlist_head *list)
  547. {
  548. hlist_add_head(&sk->sk_bind_node, list);
  549. }
  550. #define sk_for_each(__sk, node, list) \
  551. hlist_for_each_entry(__sk, node, list, sk_node)
  552. #define sk_for_each_rcu(__sk, node, list) \
  553. hlist_for_each_entry_rcu(__sk, node, list, sk_node)
  554. #define sk_nulls_for_each(__sk, node, list) \
  555. hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
  556. #define sk_nulls_for_each_rcu(__sk, node, list) \
  557. hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
  558. #define sk_for_each_from(__sk, node) \
  559. if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
  560. hlist_for_each_entry_from(__sk, node, sk_node)
  561. #define sk_nulls_for_each_from(__sk, node) \
  562. if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
  563. hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
  564. #define sk_for_each_safe(__sk, node, tmp, list) \
  565. hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
  566. #define sk_for_each_bound(__sk, node, list) \
  567. hlist_for_each_entry(__sk, node, list, sk_bind_node)
  568. /* Sock flags */
  569. enum sock_flags {
  570. SOCK_DEAD,
  571. SOCK_DONE,
  572. SOCK_URGINLINE,
  573. SOCK_KEEPOPEN,
  574. SOCK_LINGER,
  575. SOCK_DESTROY,
  576. SOCK_BROADCAST,
  577. SOCK_TIMESTAMP,
  578. SOCK_ZAPPED,
  579. SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
  580. SOCK_DBG, /* %SO_DEBUG setting */
  581. SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
  582. SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
  583. SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
  584. SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
  585. SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
  586. SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
  587. SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
  588. SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
  589. SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
  590. SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
  591. SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
  592. SOCK_FASYNC, /* fasync() active */
  593. SOCK_RXQ_OVFL,
  594. SOCK_ZEROCOPY, /* buffers from userspace */
  595. SOCK_WIFI_STATUS, /* push wifi status to userspace */
  596. SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
  597. * Will use last 4 bytes of packet sent from
  598. * user-space instead.
  599. */
  600. };
  601. static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
  602. {
  603. nsk->sk_flags = osk->sk_flags;
  604. }
  605. static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
  606. {
  607. __set_bit(flag, &sk->sk_flags);
  608. }
  609. static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
  610. {
  611. __clear_bit(flag, &sk->sk_flags);
  612. }
  613. static inline int sock_flag(struct sock *sk, enum sock_flags flag)
  614. {
  615. return test_bit(flag, &sk->sk_flags);
  616. }
  617. static inline void sk_acceptq_removed(struct sock *sk)
  618. {
  619. sk->sk_ack_backlog--;
  620. }
  621. static inline void sk_acceptq_added(struct sock *sk)
  622. {
  623. sk->sk_ack_backlog++;
  624. }
  625. static inline bool sk_acceptq_is_full(const struct sock *sk)
  626. {
  627. return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
  628. }
  629. /*
  630. * Compute minimal free write space needed to queue new packets.
  631. */
  632. static inline int sk_stream_min_wspace(const struct sock *sk)
  633. {
  634. return sk->sk_wmem_queued >> 1;
  635. }
  636. static inline int sk_stream_wspace(const struct sock *sk)
  637. {
  638. return sk->sk_sndbuf - sk->sk_wmem_queued;
  639. }
  640. void sk_stream_write_space(struct sock *sk);
  641. static inline bool sk_stream_memory_free(const struct sock *sk)
  642. {
  643. return sk->sk_wmem_queued < sk->sk_sndbuf;
  644. }
  645. /* OOB backlog add */
  646. static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
  647. {
  648. /* dont let skb dst not refcounted, we are going to leave rcu lock */
  649. skb_dst_force(skb);
  650. if (!sk->sk_backlog.tail)
  651. sk->sk_backlog.head = skb;
  652. else
  653. sk->sk_backlog.tail->next = skb;
  654. sk->sk_backlog.tail = skb;
  655. skb->next = NULL;
  656. }
  657. /*
  658. * Take into account size of receive queue and backlog queue
  659. * Do not take into account this skb truesize,
  660. * to allow even a single big packet to come.
  661. */
  662. static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
  663. {
  664. unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
  665. return qsize > sk->sk_rcvbuf;
  666. }
  667. /* The per-socket spinlock must be held here. */
  668. static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
  669. {
  670. if (sk_rcvqueues_full(sk, skb))
  671. return -ENOBUFS;
  672. __sk_add_backlog(sk, skb);
  673. sk->sk_backlog.len += skb->truesize;
  674. return 0;
  675. }
  676. static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
  677. {
  678. return sk->sk_backlog_rcv(sk, skb);
  679. }
  680. static inline void sock_rps_record_flow(const struct sock *sk)
  681. {
  682. #ifdef CONFIG_RPS
  683. struct rps_sock_flow_table *sock_flow_table;
  684. rcu_read_lock();
  685. sock_flow_table = rcu_dereference(rps_sock_flow_table);
  686. rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
  687. rcu_read_unlock();
  688. #endif
  689. }
  690. static inline void sock_rps_reset_flow(const struct sock *sk)
  691. {
  692. #ifdef CONFIG_RPS
  693. struct rps_sock_flow_table *sock_flow_table;
  694. rcu_read_lock();
  695. sock_flow_table = rcu_dereference(rps_sock_flow_table);
  696. rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
  697. rcu_read_unlock();
  698. #endif
  699. }
  700. static inline void sock_rps_save_rxhash(struct sock *sk,
  701. const struct sk_buff *skb)
  702. {
  703. #ifdef CONFIG_RPS
  704. if (unlikely(sk->sk_rxhash != skb->rxhash)) {
  705. sock_rps_reset_flow(sk);
  706. sk->sk_rxhash = skb->rxhash;
  707. }
  708. #endif
  709. }
  710. static inline void sock_rps_reset_rxhash(struct sock *sk)
  711. {
  712. #ifdef CONFIG_RPS
  713. sock_rps_reset_flow(sk);
  714. sk->sk_rxhash = 0;
  715. #endif
  716. }
  717. #define sk_wait_event(__sk, __timeo, __condition) \
  718. ({ int __rc; \
  719. release_sock(__sk); \
  720. __rc = __condition; \
  721. if (!__rc) { \
  722. *(__timeo) = schedule_timeout(*(__timeo)); \
  723. } \
  724. lock_sock(__sk); \
  725. __rc = __condition; \
  726. __rc; \
  727. })
  728. int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
  729. int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
  730. void sk_stream_wait_close(struct sock *sk, long timeo_p);
  731. int sk_stream_error(struct sock *sk, int flags, int err);
  732. void sk_stream_kill_queues(struct sock *sk);
  733. int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
  734. struct request_sock_ops;
  735. struct timewait_sock_ops;
  736. struct inet_hashinfo;
  737. struct raw_hashinfo;
  738. struct module;
  739. /*
  740. * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
  741. * un-modified. Special care is taken when initializing object to zero.
  742. */
  743. static inline void sk_prot_clear_nulls(struct sock *sk, int size)
  744. {
  745. if (offsetof(struct sock, sk_node.next) != 0)
  746. memset(sk, 0, offsetof(struct sock, sk_node.next));
  747. memset(&sk->sk_node.pprev, 0,
  748. size - offsetof(struct sock, sk_node.pprev));
  749. }
  750. /* Networking protocol blocks we attach to sockets.
  751. * socket layer -> transport layer interface
  752. * transport -> network interface is defined by struct inet_proto
  753. */
  754. struct proto {
  755. void (*close)(struct sock *sk,
  756. long timeout);
  757. int (*connect)(struct sock *sk,
  758. struct sockaddr *uaddr,
  759. int addr_len);
  760. int (*disconnect)(struct sock *sk, int flags);
  761. struct sock * (*accept)(struct sock *sk, int flags, int *err);
  762. int (*ioctl)(struct sock *sk, int cmd,
  763. unsigned long arg);
  764. int (*init)(struct sock *sk);
  765. void (*destroy)(struct sock *sk);
  766. void (*shutdown)(struct sock *sk, int how);
  767. int (*setsockopt)(struct sock *sk, int level,
  768. int optname, char __user *optval,
  769. unsigned int optlen);
  770. int (*getsockopt)(struct sock *sk, int level,
  771. int optname, char __user *optval,
  772. int __user *option);
  773. #ifdef CONFIG_COMPAT
  774. int (*compat_setsockopt)(struct sock *sk,
  775. int level,
  776. int optname, char __user *optval,
  777. unsigned int optlen);
  778. int (*compat_getsockopt)(struct sock *sk,
  779. int level,
  780. int optname, char __user *optval,
  781. int __user *option);
  782. int (*compat_ioctl)(struct sock *sk,
  783. unsigned int cmd, unsigned long arg);
  784. #endif
  785. int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
  786. struct msghdr *msg, size_t len);
  787. int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
  788. struct msghdr *msg,
  789. size_t len, int noblock, int flags,
  790. int *addr_len);
  791. int (*sendpage)(struct sock *sk, struct page *page,
  792. int offset, size_t size, int flags);
  793. int (*bind)(struct sock *sk,
  794. struct sockaddr *uaddr, int addr_len);
  795. int (*backlog_rcv) (struct sock *sk,
  796. struct sk_buff *skb);
  797. void (*release_cb)(struct sock *sk);
  798. /* Keeping track of sk's, looking them up, and port selection methods. */
  799. void (*hash)(struct sock *sk);
  800. void (*unhash)(struct sock *sk);
  801. void (*rehash)(struct sock *sk);
  802. int (*get_port)(struct sock *sk, unsigned short snum);
  803. void (*clear_sk)(struct sock *sk, int size);
  804. /* Keeping track of sockets in use */
  805. #ifdef CONFIG_PROC_FS
  806. unsigned int inuse_idx;
  807. #endif
  808. /* Memory pressure */
  809. void (*enter_memory_pressure)(struct sock *sk);
  810. atomic_long_t *memory_allocated; /* Current allocated memory. */
  811. struct percpu_counter *sockets_allocated; /* Current number of sockets. */
  812. /*
  813. * Pressure flag: try to collapse.
  814. * Technical note: it is used by multiple contexts non atomically.
  815. * All the __sk_mem_schedule() is of this nature: accounting
  816. * is strict, actions are advisory and have some latency.
  817. */
  818. int *memory_pressure;
  819. long *sysctl_mem;
  820. int *sysctl_wmem;
  821. int *sysctl_rmem;
  822. int max_header;
  823. bool no_autobind;
  824. struct kmem_cache *slab;
  825. unsigned int obj_size;
  826. int slab_flags;
  827. struct percpu_counter *orphan_count;
  828. struct request_sock_ops *rsk_prot;
  829. struct timewait_sock_ops *twsk_prot;
  830. union {
  831. struct inet_hashinfo *hashinfo;
  832. struct udp_table *udp_table;
  833. struct raw_hashinfo *raw_hash;
  834. } h;
  835. struct module *owner;
  836. char name[32];
  837. struct list_head node;
  838. #ifdef SOCK_REFCNT_DEBUG
  839. atomic_t socks;
  840. #endif
  841. #ifdef CONFIG_MEMCG_KMEM
  842. /*
  843. * cgroup specific init/deinit functions. Called once for all
  844. * protocols that implement it, from cgroups populate function.
  845. * This function has to setup any files the protocol want to
  846. * appear in the kmem cgroup filesystem.
  847. */
  848. int (*init_cgroup)(struct cgroup *cgrp,
  849. struct cgroup_subsys *ss);
  850. void (*destroy_cgroup)(struct cgroup *cgrp);
  851. struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
  852. #endif
  853. int (*diag_destroy)(struct sock *sk, int err);
  854. };
  855. struct cg_proto {
  856. void (*enter_memory_pressure)(struct sock *sk);
  857. struct res_counter *memory_allocated; /* Current allocated memory. */
  858. struct percpu_counter *sockets_allocated; /* Current number of sockets. */
  859. int *memory_pressure;
  860. long *sysctl_mem;
  861. /*
  862. * memcg field is used to find which memcg we belong directly
  863. * Each memcg struct can hold more than one cg_proto, so container_of
  864. * won't really cut.
  865. *
  866. * The elegant solution would be having an inverse function to
  867. * proto_cgroup in struct proto, but that means polluting the structure
  868. * for everybody, instead of just for memcg users.
  869. */
  870. struct mem_cgroup *memcg;
  871. };
  872. int proto_register(struct proto *prot, int alloc_slab);
  873. void proto_unregister(struct proto *prot);
  874. #ifdef SOCK_REFCNT_DEBUG
  875. static inline void sk_refcnt_debug_inc(struct sock *sk)
  876. {
  877. atomic_inc(&sk->sk_prot->socks);
  878. }
  879. static inline void sk_refcnt_debug_dec(struct sock *sk)
  880. {
  881. atomic_dec(&sk->sk_prot->socks);
  882. printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
  883. sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
  884. }
  885. static inline void sk_refcnt_debug_release(const struct sock *sk)
  886. {
  887. if (atomic_read(&sk->sk_refcnt) != 1)
  888. printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
  889. sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
  890. }
  891. #else /* SOCK_REFCNT_DEBUG */
  892. #define sk_refcnt_debug_inc(sk) do { } while (0)
  893. #define sk_refcnt_debug_dec(sk) do { } while (0)
  894. #define sk_refcnt_debug_release(sk) do { } while (0)
  895. #endif /* SOCK_REFCNT_DEBUG */
  896. #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
  897. extern struct static_key memcg_socket_limit_enabled;
  898. static inline struct cg_proto *parent_cg_proto(struct proto *proto,
  899. struct cg_proto *cg_proto)
  900. {
  901. return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
  902. }
  903. #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
  904. #else
  905. #define mem_cgroup_sockets_enabled 0
  906. static inline struct cg_proto *parent_cg_proto(struct proto *proto,
  907. struct cg_proto *cg_proto)
  908. {
  909. return NULL;
  910. }
  911. #endif
  912. static inline bool sk_has_memory_pressure(const struct sock *sk)
  913. {
  914. return sk->sk_prot->memory_pressure != NULL;
  915. }
  916. static inline bool sk_under_memory_pressure(const struct sock *sk)
  917. {
  918. if (!sk->sk_prot->memory_pressure)
  919. return false;
  920. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  921. return !!*sk->sk_cgrp->memory_pressure;
  922. return !!*sk->sk_prot->memory_pressure;
  923. }
  924. static inline void sk_leave_memory_pressure(struct sock *sk)
  925. {
  926. int *memory_pressure = sk->sk_prot->memory_pressure;
  927. if (!memory_pressure)
  928. return;
  929. if (*memory_pressure)
  930. *memory_pressure = 0;
  931. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  932. struct cg_proto *cg_proto = sk->sk_cgrp;
  933. struct proto *prot = sk->sk_prot;
  934. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  935. if (*cg_proto->memory_pressure)
  936. *cg_proto->memory_pressure = 0;
  937. }
  938. }
  939. static inline void sk_enter_memory_pressure(struct sock *sk)
  940. {
  941. if (!sk->sk_prot->enter_memory_pressure)
  942. return;
  943. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  944. struct cg_proto *cg_proto = sk->sk_cgrp;
  945. struct proto *prot = sk->sk_prot;
  946. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  947. cg_proto->enter_memory_pressure(sk);
  948. }
  949. sk->sk_prot->enter_memory_pressure(sk);
  950. }
  951. static inline long sk_prot_mem_limits(const struct sock *sk, int index)
  952. {
  953. long *prot = sk->sk_prot->sysctl_mem;
  954. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  955. prot = sk->sk_cgrp->sysctl_mem;
  956. return prot[index];
  957. }
  958. static inline void memcg_memory_allocated_add(struct cg_proto *prot,
  959. unsigned long amt,
  960. int *parent_status)
  961. {
  962. struct res_counter *fail;
  963. int ret;
  964. ret = res_counter_charge_nofail(prot->memory_allocated,
  965. amt << PAGE_SHIFT, &fail);
  966. if (ret < 0)
  967. *parent_status = OVER_LIMIT;
  968. }
  969. static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
  970. unsigned long amt)
  971. {
  972. res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
  973. }
  974. static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
  975. {
  976. u64 ret;
  977. ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
  978. return ret >> PAGE_SHIFT;
  979. }
  980. static inline long
  981. sk_memory_allocated(const struct sock *sk)
  982. {
  983. struct proto *prot = sk->sk_prot;
  984. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  985. return memcg_memory_allocated_read(sk->sk_cgrp);
  986. return atomic_long_read(prot->memory_allocated);
  987. }
  988. static inline long
  989. sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
  990. {
  991. struct proto *prot = sk->sk_prot;
  992. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  993. memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
  994. /* update the root cgroup regardless */
  995. atomic_long_add_return(amt, prot->memory_allocated);
  996. return memcg_memory_allocated_read(sk->sk_cgrp);
  997. }
  998. return atomic_long_add_return(amt, prot->memory_allocated);
  999. }
  1000. static inline void
  1001. sk_memory_allocated_sub(struct sock *sk, int amt)
  1002. {
  1003. struct proto *prot = sk->sk_prot;
  1004. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1005. memcg_memory_allocated_sub(sk->sk_cgrp, amt);
  1006. atomic_long_sub(amt, prot->memory_allocated);
  1007. }
  1008. static inline void sk_sockets_allocated_dec(struct sock *sk)
  1009. {
  1010. struct proto *prot = sk->sk_prot;
  1011. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1012. struct cg_proto *cg_proto = sk->sk_cgrp;
  1013. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  1014. percpu_counter_dec(cg_proto->sockets_allocated);
  1015. }
  1016. percpu_counter_dec(prot->sockets_allocated);
  1017. }
  1018. static inline void sk_sockets_allocated_inc(struct sock *sk)
  1019. {
  1020. struct proto *prot = sk->sk_prot;
  1021. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1022. struct cg_proto *cg_proto = sk->sk_cgrp;
  1023. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  1024. percpu_counter_inc(cg_proto->sockets_allocated);
  1025. }
  1026. percpu_counter_inc(prot->sockets_allocated);
  1027. }
  1028. static inline u64
  1029. sk_sockets_allocated_read_positive(struct sock *sk)
  1030. {
  1031. struct proto *prot = sk->sk_prot;
  1032. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1033. return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
  1034. return percpu_counter_read_positive(prot->sockets_allocated);
  1035. }
  1036. static inline int
  1037. proto_sockets_allocated_sum_positive(struct proto *prot)
  1038. {
  1039. return percpu_counter_sum_positive(prot->sockets_allocated);
  1040. }
  1041. static inline long
  1042. proto_memory_allocated(struct proto *prot)
  1043. {
  1044. return atomic_long_read(prot->memory_allocated);
  1045. }
  1046. static inline bool
  1047. proto_memory_pressure(struct proto *prot)
  1048. {
  1049. if (!prot->memory_pressure)
  1050. return false;
  1051. return !!*prot->memory_pressure;
  1052. }
  1053. #ifdef CONFIG_PROC_FS
  1054. /* Called with local bh disabled */
  1055. void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
  1056. int sock_prot_inuse_get(struct net *net, struct proto *proto);
  1057. #else
  1058. static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
  1059. int inc)
  1060. {
  1061. }
  1062. #endif
  1063. /* With per-bucket locks this operation is not-atomic, so that
  1064. * this version is not worse.
  1065. */
  1066. static inline void __sk_prot_rehash(struct sock *sk)
  1067. {
  1068. sk->sk_prot->unhash(sk);
  1069. sk->sk_prot->hash(sk);
  1070. }
  1071. void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
  1072. /* About 10 seconds */
  1073. #define SOCK_DESTROY_TIME (10*HZ)
  1074. /* Sockets 0-1023 can't be bound to unless you are superuser */
  1075. #define PROT_SOCK 1024
  1076. #define SHUTDOWN_MASK 3
  1077. #define RCV_SHUTDOWN 1
  1078. #define SEND_SHUTDOWN 2
  1079. #define SOCK_SNDBUF_LOCK 1
  1080. #define SOCK_RCVBUF_LOCK 2
  1081. #define SOCK_BINDADDR_LOCK 4
  1082. #define SOCK_BINDPORT_LOCK 8
  1083. /* sock_iocb: used to kick off async processing of socket ios */
  1084. struct sock_iocb {
  1085. struct list_head list;
  1086. int flags;
  1087. int size;
  1088. struct socket *sock;
  1089. struct sock *sk;
  1090. struct scm_cookie *scm;
  1091. struct msghdr *msg, async_msg;
  1092. struct kiocb *kiocb;
  1093. };
  1094. static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
  1095. {
  1096. return (struct sock_iocb *)iocb->private;
  1097. }
  1098. static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
  1099. {
  1100. return si->kiocb;
  1101. }
  1102. struct socket_alloc {
  1103. struct socket socket;
  1104. struct inode vfs_inode;
  1105. };
  1106. static inline struct socket *SOCKET_I(struct inode *inode)
  1107. {
  1108. return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
  1109. }
  1110. static inline struct inode *SOCK_INODE(struct socket *socket)
  1111. {
  1112. return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
  1113. }
  1114. /*
  1115. * Functions for memory accounting
  1116. */
  1117. int __sk_mem_schedule(struct sock *sk, int size, int kind);
  1118. void __sk_mem_reclaim(struct sock *sk);
  1119. #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
  1120. #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
  1121. #define SK_MEM_SEND 0
  1122. #define SK_MEM_RECV 1
  1123. static inline int sk_mem_pages(int amt)
  1124. {
  1125. return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
  1126. }
  1127. static inline bool sk_has_account(struct sock *sk)
  1128. {
  1129. /* return true if protocol supports memory accounting */
  1130. return !!sk->sk_prot->memory_allocated;
  1131. }
  1132. static inline bool sk_wmem_schedule(struct sock *sk, int size)
  1133. {
  1134. if (!sk_has_account(sk))
  1135. return true;
  1136. return size <= sk->sk_forward_alloc ||
  1137. __sk_mem_schedule(sk, size, SK_MEM_SEND);
  1138. }
  1139. static inline bool sk_rmem_schedule(struct sock *sk, int size)
  1140. {
  1141. if (!sk_has_account(sk))
  1142. return true;
  1143. return size <= sk->sk_forward_alloc ||
  1144. __sk_mem_schedule(sk, size, SK_MEM_RECV);
  1145. }
  1146. static inline void sk_mem_reclaim(struct sock *sk)
  1147. {
  1148. if (!sk_has_account(sk))
  1149. return;
  1150. if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
  1151. __sk_mem_reclaim(sk);
  1152. }
  1153. static inline void sk_mem_reclaim_partial(struct sock *sk)
  1154. {
  1155. if (!sk_has_account(sk))
  1156. return;
  1157. if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
  1158. __sk_mem_reclaim(sk);
  1159. }
  1160. static inline void sk_mem_charge(struct sock *sk, int size)
  1161. {
  1162. if (!sk_has_account(sk))
  1163. return;
  1164. sk->sk_forward_alloc -= size;
  1165. }
  1166. static inline void sk_mem_uncharge(struct sock *sk, int size)
  1167. {
  1168. if (!sk_has_account(sk))
  1169. return;
  1170. sk->sk_forward_alloc += size;
  1171. }
  1172. static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
  1173. {
  1174. sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
  1175. sk->sk_wmem_queued -= skb->truesize;
  1176. sk_mem_uncharge(sk, skb->truesize);
  1177. __kfree_skb(skb);
  1178. }
  1179. /* Used by processes to "lock" a socket state, so that
  1180. * interrupts and bottom half handlers won't change it
  1181. * from under us. It essentially blocks any incoming
  1182. * packets, so that we won't get any new data or any
  1183. * packets that change the state of the socket.
  1184. *
  1185. * While locked, BH processing will add new packets to
  1186. * the backlog queue. This queue is processed by the
  1187. * owner of the socket lock right before it is released.
  1188. *
  1189. * Since ~2.3.5 it is also exclusive sleep lock serializing
  1190. * accesses from user process context.
  1191. */
  1192. #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
  1193. /*
  1194. * Macro so as to not evaluate some arguments when
  1195. * lockdep is not enabled.
  1196. *
  1197. * Mark both the sk_lock and the sk_lock.slock as a
  1198. * per-address-family lock class.
  1199. */
  1200. #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
  1201. do { \
  1202. sk->sk_lock.owned = 0; \
  1203. init_waitqueue_head(&sk->sk_lock.wq); \
  1204. spin_lock_init(&(sk)->sk_lock.slock); \
  1205. debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
  1206. sizeof((sk)->sk_lock)); \
  1207. lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
  1208. (skey), (sname)); \
  1209. lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
  1210. } while (0)
  1211. void lock_sock_nested(struct sock *sk, int subclass);
  1212. static inline void lock_sock(struct sock *sk)
  1213. {
  1214. lock_sock_nested(sk, 0);
  1215. }
  1216. void release_sock(struct sock *sk);
  1217. /* BH context may only use the following locking interface. */
  1218. #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
  1219. #define bh_lock_sock_nested(__sk) \
  1220. spin_lock_nested(&((__sk)->sk_lock.slock), \
  1221. SINGLE_DEPTH_NESTING)
  1222. #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
  1223. bool lock_sock_fast(struct sock *sk);
  1224. /**
  1225. * unlock_sock_fast - complement of lock_sock_fast
  1226. * @sk: socket
  1227. * @slow: slow mode
  1228. *
  1229. * fast unlock socket for user context.
  1230. * If slow mode is on, we call regular release_sock()
  1231. */
  1232. static inline void unlock_sock_fast(struct sock *sk, bool slow)
  1233. {
  1234. if (slow)
  1235. release_sock(sk);
  1236. else
  1237. spin_unlock_bh(&sk->sk_lock.slock);
  1238. }
  1239. struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
  1240. struct proto *prot);
  1241. void sk_free(struct sock *sk);
  1242. void sk_release_kernel(struct sock *sk);
  1243. struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
  1244. struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
  1245. gfp_t priority);
  1246. struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
  1247. gfp_t priority);
  1248. void sock_wfree(struct sk_buff *skb);
  1249. void sock_rfree(struct sk_buff *skb);
  1250. int sock_setsockopt(struct socket *sock, int level, int op,
  1251. char __user *optval, unsigned int optlen);
  1252. int sock_getsockopt(struct socket *sock, int level, int op,
  1253. char __user *optval, int __user *optlen);
  1254. struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
  1255. int noblock, int *errcode);
  1256. struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
  1257. unsigned long data_len, int noblock,
  1258. int *errcode);
  1259. void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
  1260. void sock_kfree_s(struct sock *sk, void *mem, int size);
  1261. void sk_send_sigurg(struct sock *sk);
  1262. #ifdef CONFIG_CGROUPS
  1263. void sock_update_classid(struct sock *sk);
  1264. #else
  1265. static inline void sock_update_classid(struct sock *sk)
  1266. {
  1267. }
  1268. #endif
  1269. /*
  1270. * Functions to fill in entries in struct proto_ops when a protocol
  1271. * does not implement a particular function.
  1272. */
  1273. int sock_no_bind(struct socket *, struct sockaddr *, int);
  1274. int sock_no_connect(struct socket *, struct sockaddr *, int, int);
  1275. int sock_no_socketpair(struct socket *, struct socket *);
  1276. int sock_no_accept(struct socket *, struct socket *, int);
  1277. int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
  1278. unsigned int sock_no_poll(struct file *, struct socket *,
  1279. struct poll_table_struct *);
  1280. int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
  1281. int sock_no_listen(struct socket *, int);
  1282. int sock_no_shutdown(struct socket *, int);
  1283. int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
  1284. int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
  1285. int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
  1286. int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
  1287. int);
  1288. int sock_no_mmap(struct file *file, struct socket *sock,
  1289. struct vm_area_struct *vma);
  1290. ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
  1291. size_t size, int flags);
  1292. /*
  1293. * Functions to fill in entries in struct proto_ops when a protocol
  1294. * uses the inet style.
  1295. */
  1296. int sock_common_getsockopt(struct socket *sock, int level, int optname,
  1297. char __user *optval, int __user *optlen);
  1298. int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
  1299. struct msghdr *msg, size_t size, int flags);
  1300. int sock_common_setsockopt(struct socket *sock, int level, int optname,
  1301. char __user *optval, unsigned int optlen);
  1302. int compat_sock_common_getsockopt(struct socket *sock, int level,
  1303. int optname, char __user *optval, int __user *optlen);
  1304. int compat_sock_common_setsockopt(struct socket *sock, int level,
  1305. int optname, char __user *optval, unsigned int optlen);
  1306. void sk_common_release(struct sock *sk);
  1307. /*
  1308. * Default socket callbacks and setup code
  1309. */
  1310. /* Initialise core socket variables */
  1311. void sock_init_data(struct socket *sock, struct sock *sk);
  1312. void sk_filter_release_rcu(struct rcu_head *rcu);
  1313. /**
  1314. * sk_filter_release - release a socket filter
  1315. * @fp: filter to remove
  1316. *
  1317. * Remove a filter from a socket and release its resources.
  1318. */
  1319. static inline void sk_filter_release(struct sk_filter *fp)
  1320. {
  1321. if (atomic_dec_and_test(&fp->refcnt))
  1322. call_rcu(&fp->rcu, sk_filter_release_rcu);
  1323. }
  1324. static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
  1325. {
  1326. unsigned int size = sk_filter_len(fp);
  1327. atomic_sub(size, &sk->sk_omem_alloc);
  1328. sk_filter_release(fp);
  1329. }
  1330. static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
  1331. {
  1332. atomic_inc(&fp->refcnt);
  1333. atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
  1334. }
  1335. /*
  1336. * Socket reference counting postulates.
  1337. *
  1338. * * Each user of socket SHOULD hold a reference count.
  1339. * * Each access point to socket (an hash table bucket, reference from a list,
  1340. * running timer, skb in flight MUST hold a reference count.
  1341. * * When reference count hits 0, it means it will never increase back.
  1342. * * When reference count hits 0, it means that no references from
  1343. * outside exist to this socket and current process on current CPU
  1344. * is last user and may/should destroy this socket.
  1345. * * sk_free is called from any context: process, BH, IRQ. When
  1346. * it is called, socket has no references from outside -> sk_free
  1347. * may release descendant resources allocated by the socket, but
  1348. * to the time when it is called, socket is NOT referenced by any
  1349. * hash tables, lists etc.
  1350. * * Packets, delivered from outside (from network or from another process)
  1351. * and enqueued on receive/error queues SHOULD NOT grab reference count,
  1352. * when they sit in queue. Otherwise, packets will leak to hole, when
  1353. * socket is looked up by one cpu and unhasing is made by another CPU.
  1354. * It is true for udp/raw, netlink (leak to receive and error queues), tcp
  1355. * (leak to backlog). Packet socket does all the processing inside
  1356. * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
  1357. * use separate SMP lock, so that they are prone too.
  1358. */
  1359. /* Ungrab socket and destroy it, if it was the last reference. */
  1360. static inline void sock_put(struct sock *sk)
  1361. {
  1362. if (atomic_dec_and_test(&sk->sk_refcnt))
  1363. sk_free(sk);
  1364. }
  1365. int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
  1366. static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
  1367. {
  1368. sk->sk_tx_queue_mapping = tx_queue;
  1369. }
  1370. static inline void sk_tx_queue_clear(struct sock *sk)
  1371. {
  1372. sk->sk_tx_queue_mapping = -1;
  1373. }
  1374. static inline int sk_tx_queue_get(const struct sock *sk)
  1375. {
  1376. return sk ? sk->sk_tx_queue_mapping : -1;
  1377. }
  1378. static inline void sk_set_socket(struct sock *sk, struct socket *sock)
  1379. {
  1380. sk->sk_socket = sock;
  1381. }
  1382. static inline wait_queue_head_t *sk_sleep(struct sock *sk)
  1383. {
  1384. BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
  1385. return &rcu_dereference_raw(sk->sk_wq)->wait;
  1386. }
  1387. /* Detach socket from process context.
  1388. * Announce socket dead, detach it from wait queue and inode.
  1389. * Note that parent inode held reference count on this struct sock,
  1390. * we do not release it in this function, because protocol
  1391. * probably wants some additional cleanups or even continuing
  1392. * to work with this socket (TCP).
  1393. */
  1394. static inline void sock_orphan(struct sock *sk)
  1395. {
  1396. write_lock_bh(&sk->sk_callback_lock);
  1397. sock_set_flag(sk, SOCK_DEAD);
  1398. sk_set_socket(sk, NULL);
  1399. sk->sk_wq = NULL;
  1400. write_unlock_bh(&sk->sk_callback_lock);
  1401. }
  1402. static inline void sock_graft(struct sock *sk, struct socket *parent)
  1403. {
  1404. write_lock_bh(&sk->sk_callback_lock);
  1405. sk->sk_wq = parent->wq;
  1406. parent->sk = sk;
  1407. sk_set_socket(sk, parent);
  1408. sk->sk_uid = SOCK_INODE(parent)->i_uid;
  1409. security_sock_graft(sk, parent);
  1410. write_unlock_bh(&sk->sk_callback_lock);
  1411. }
  1412. int sock_i_uid(struct sock *sk);
  1413. unsigned long sock_i_ino(struct sock *sk);
  1414. static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
  1415. {
  1416. return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
  1417. }
  1418. static inline struct dst_entry *
  1419. __sk_dst_get(struct sock *sk)
  1420. {
  1421. return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
  1422. lockdep_is_held(&sk->sk_lock.slock));
  1423. }
  1424. static inline struct dst_entry *
  1425. sk_dst_get(struct sock *sk)
  1426. {
  1427. struct dst_entry *dst;
  1428. rcu_read_lock();
  1429. dst = rcu_dereference(sk->sk_dst_cache);
  1430. if (dst)
  1431. dst_hold(dst);
  1432. rcu_read_unlock();
  1433. return dst;
  1434. }
  1435. void sk_reset_txq(struct sock *sk);
  1436. static inline void dst_negative_advice(struct sock *sk)
  1437. {
  1438. struct dst_entry *ndst, *dst = __sk_dst_get(sk);
  1439. if (dst && dst->ops->negative_advice) {
  1440. ndst = dst->ops->negative_advice(dst);
  1441. if (ndst != dst) {
  1442. rcu_assign_pointer(sk->sk_dst_cache, ndst);
  1443. sk_reset_txq(sk);
  1444. }
  1445. }
  1446. }
  1447. static inline void
  1448. __sk_dst_set(struct sock *sk, struct dst_entry *dst)
  1449. {
  1450. struct dst_entry *old_dst;
  1451. sk_tx_queue_clear(sk);
  1452. /*
  1453. * This can be called while sk is owned by the caller only,
  1454. * with no state that can be checked in a rcu_dereference_check() cond
  1455. */
  1456. old_dst = rcu_dereference_raw(sk->sk_dst_cache);
  1457. rcu_assign_pointer(sk->sk_dst_cache, dst);
  1458. dst_release(old_dst);
  1459. }
  1460. static inline void
  1461. sk_dst_set(struct sock *sk, struct dst_entry *dst)
  1462. {
  1463. spin_lock(&sk->sk_dst_lock);
  1464. __sk_dst_set(sk, dst);
  1465. spin_unlock(&sk->sk_dst_lock);
  1466. }
  1467. static inline void
  1468. __sk_dst_reset(struct sock *sk)
  1469. {
  1470. __sk_dst_set(sk, NULL);
  1471. }
  1472. static inline void
  1473. sk_dst_reset(struct sock *sk)
  1474. {
  1475. spin_lock(&sk->sk_dst_lock);
  1476. __sk_dst_reset(sk);
  1477. spin_unlock(&sk->sk_dst_lock);
  1478. }
  1479. struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
  1480. struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
  1481. static inline bool sk_can_gso(const struct sock *sk)
  1482. {
  1483. return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
  1484. }
  1485. void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
  1486. static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
  1487. {
  1488. sk->sk_route_nocaps |= flags;
  1489. sk->sk_route_caps &= ~flags;
  1490. }
  1491. static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
  1492. char __user *from, char *to,
  1493. int copy, int offset)
  1494. {
  1495. if (skb->ip_summed == CHECKSUM_NONE) {
  1496. int err = 0;
  1497. __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
  1498. if (err)
  1499. return err;
  1500. skb->csum = csum_block_add(skb->csum, csum, offset);
  1501. } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
  1502. if (!access_ok(VERIFY_READ, from, copy) ||
  1503. __copy_from_user_nocache(to, from, copy))
  1504. return -EFAULT;
  1505. } else if (copy_from_user(to, from, copy))
  1506. return -EFAULT;
  1507. return 0;
  1508. }
  1509. static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
  1510. char __user *from, int copy)
  1511. {
  1512. int err, offset = skb->len;
  1513. err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
  1514. copy, offset);
  1515. if (err)
  1516. __skb_trim(skb, offset);
  1517. return err;
  1518. }
  1519. static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
  1520. struct sk_buff *skb,
  1521. struct page *page,
  1522. int off, int copy)
  1523. {
  1524. int err;
  1525. err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
  1526. copy, skb->len);
  1527. if (err)
  1528. return err;
  1529. skb->len += copy;
  1530. skb->data_len += copy;
  1531. skb->truesize += copy;
  1532. sk->sk_wmem_queued += copy;
  1533. sk_mem_charge(sk, copy);
  1534. return 0;
  1535. }
  1536. static inline int skb_copy_to_page(struct sock *sk, char __user *from,
  1537. struct sk_buff *skb, struct page *page,
  1538. int off, int copy)
  1539. {
  1540. if (skb->ip_summed == CHECKSUM_NONE) {
  1541. int err = 0;
  1542. __wsum csum = csum_and_copy_from_user(from,
  1543. page_address(page) + off,
  1544. copy, 0, &err);
  1545. if (err)
  1546. return err;
  1547. skb->csum = csum_block_add(skb->csum, csum, skb->len);
  1548. } else if (copy_from_user(page_address(page) + off, from, copy))
  1549. return -EFAULT;
  1550. skb->len += copy;
  1551. skb->data_len += copy;
  1552. skb->truesize += copy;
  1553. sk->sk_wmem_queued += copy;
  1554. sk_mem_charge(sk, copy);
  1555. return 0;
  1556. }
  1557. /**
  1558. * sk_wmem_alloc_get - returns write allocations
  1559. * @sk: socket
  1560. *
  1561. * Returns sk_wmem_alloc minus initial offset of one
  1562. */
  1563. static inline int sk_wmem_alloc_get(const struct sock *sk)
  1564. {
  1565. return atomic_read(&sk->sk_wmem_alloc) - 1;
  1566. }
  1567. /**
  1568. * sk_rmem_alloc_get - returns read allocations
  1569. * @sk: socket
  1570. *
  1571. * Returns sk_rmem_alloc
  1572. */
  1573. static inline int sk_rmem_alloc_get(const struct sock *sk)
  1574. {
  1575. return atomic_read(&sk->sk_rmem_alloc);
  1576. }
  1577. /**
  1578. * sk_has_allocations - check if allocations are outstanding
  1579. * @sk: socket
  1580. *
  1581. * Returns true if socket has write or read allocations
  1582. */
  1583. static inline bool sk_has_allocations(const struct sock *sk)
  1584. {
  1585. return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
  1586. }
  1587. /**
  1588. * wq_has_sleeper - check if there are any waiting processes
  1589. * @wq: struct socket_wq
  1590. *
  1591. * Returns true if socket_wq has waiting processes
  1592. *
  1593. * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
  1594. * barrier call. They were added due to the race found within the tcp code.
  1595. *
  1596. * Consider following tcp code paths:
  1597. *
  1598. * CPU1 CPU2
  1599. *
  1600. * sys_select receive packet
  1601. * ... ...
  1602. * __add_wait_queue update tp->rcv_nxt
  1603. * ... ...
  1604. * tp->rcv_nxt check sock_def_readable
  1605. * ... {
  1606. * schedule rcu_read_lock();
  1607. * wq = rcu_dereference(sk->sk_wq);
  1608. * if (wq && waitqueue_active(&wq->wait))
  1609. * wake_up_interruptible(&wq->wait)
  1610. * ...
  1611. * }
  1612. *
  1613. * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
  1614. * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
  1615. * could then endup calling schedule and sleep forever if there are no more
  1616. * data on the socket.
  1617. *
  1618. */
  1619. static inline bool wq_has_sleeper(struct socket_wq *wq)
  1620. {
  1621. /* We need to be sure we are in sync with the
  1622. * add_wait_queue modifications to the wait queue.
  1623. *
  1624. * This memory barrier is paired in the sock_poll_wait.
  1625. */
  1626. smp_mb();
  1627. return wq && waitqueue_active(&wq->wait);
  1628. }
  1629. /**
  1630. * sock_poll_wait - place memory barrier behind the poll_wait call.
  1631. * @filp: file
  1632. * @wait_address: socket wait queue
  1633. * @p: poll_table
  1634. *
  1635. * See the comments in the wq_has_sleeper function.
  1636. */
  1637. static inline void sock_poll_wait(struct file *filp,
  1638. wait_queue_head_t *wait_address, poll_table *p)
  1639. {
  1640. if (!poll_does_not_wait(p) && wait_address) {
  1641. poll_wait(filp, wait_address, p);
  1642. /* We need to be sure we are in sync with the
  1643. * socket flags modification.
  1644. *
  1645. * This memory barrier is paired in the wq_has_sleeper.
  1646. */
  1647. smp_mb();
  1648. }
  1649. }
  1650. /*
  1651. * Queue a received datagram if it will fit. Stream and sequenced
  1652. * protocols can't normally use this as they need to fit buffers in
  1653. * and play with them.
  1654. *
  1655. * Inlined as it's very short and called for pretty much every
  1656. * packet ever received.
  1657. */
  1658. static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
  1659. {
  1660. skb_orphan(skb);
  1661. skb->sk = sk;
  1662. skb->destructor = sock_wfree;
  1663. /*
  1664. * We used to take a refcount on sk, but following operation
  1665. * is enough to guarantee sk_free() wont free this sock until
  1666. * all in-flight packets are completed
  1667. */
  1668. atomic_add(skb->truesize, &sk->sk_wmem_alloc);
  1669. }
  1670. static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
  1671. {
  1672. skb_orphan(skb);
  1673. skb->sk = sk;
  1674. skb->destructor = sock_rfree;
  1675. atomic_add(skb->truesize, &sk->sk_rmem_alloc);
  1676. sk_mem_charge(sk, skb->truesize);
  1677. }
  1678. void sk_reset_timer(struct sock *sk, struct timer_list *timer,
  1679. unsigned long expires);
  1680. void sk_stop_timer(struct sock *sk, struct timer_list *timer);
  1681. int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
  1682. int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
  1683. /*
  1684. * Recover an error report and clear atomically
  1685. */
  1686. static inline int sock_error(struct sock *sk)
  1687. {
  1688. int err;
  1689. if (likely(!sk->sk_err))
  1690. return 0;
  1691. err = xchg(&sk->sk_err, 0);
  1692. return -err;
  1693. }
  1694. static inline unsigned long sock_wspace(struct sock *sk)
  1695. {
  1696. int amt = 0;
  1697. if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
  1698. amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
  1699. if (amt < 0)
  1700. amt = 0;
  1701. }
  1702. return amt;
  1703. }
  1704. static inline void sk_wake_async(struct sock *sk, int how, int band)
  1705. {
  1706. if (sock_flag(sk, SOCK_FASYNC))
  1707. sock_wake_async(sk->sk_socket, how, band);
  1708. }
  1709. #define SOCK_MIN_SNDBUF 2048
  1710. /*
  1711. * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
  1712. * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
  1713. */
  1714. #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
  1715. static inline void sk_stream_moderate_sndbuf(struct sock *sk)
  1716. {
  1717. if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
  1718. sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
  1719. sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
  1720. }
  1721. }
  1722. struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
  1723. static inline struct page *sk_stream_alloc_page(struct sock *sk)
  1724. {
  1725. struct page *page = NULL;
  1726. page = alloc_pages(sk->sk_allocation, 0);
  1727. if (!page) {
  1728. sk_enter_memory_pressure(sk);
  1729. sk_stream_moderate_sndbuf(sk);
  1730. }
  1731. return page;
  1732. }
  1733. /*
  1734. * Default write policy as shown to user space via poll/select/SIGIO
  1735. */
  1736. static inline bool sock_writeable(const struct sock *sk)
  1737. {
  1738. return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
  1739. }
  1740. static inline gfp_t gfp_any(void)
  1741. {
  1742. return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
  1743. }
  1744. static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
  1745. {
  1746. return noblock ? 0 : sk->sk_rcvtimeo;
  1747. }
  1748. static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
  1749. {
  1750. return noblock ? 0 : sk->sk_sndtimeo;
  1751. }
  1752. static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
  1753. {
  1754. return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
  1755. }
  1756. /* Alas, with timeout socket operations are not restartable.
  1757. * Compare this to poll().
  1758. */
  1759. static inline int sock_intr_errno(long timeo)
  1760. {
  1761. return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
  1762. }
  1763. void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
  1764. struct sk_buff *skb);
  1765. void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
  1766. struct sk_buff *skb);
  1767. static inline void
  1768. sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
  1769. {
  1770. ktime_t kt = skb->tstamp;
  1771. struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
  1772. /*
  1773. * generate control messages if
  1774. * - receive time stamping in software requested (SOCK_RCVTSTAMP
  1775. * or SOCK_TIMESTAMPING_RX_SOFTWARE)
  1776. * - software time stamp available and wanted
  1777. * (SOCK_TIMESTAMPING_SOFTWARE)
  1778. * - hardware time stamps available and wanted
  1779. * (SOCK_TIMESTAMPING_SYS_HARDWARE or
  1780. * SOCK_TIMESTAMPING_RAW_HARDWARE)
  1781. */
  1782. if (sock_flag(sk, SOCK_RCVTSTAMP) ||
  1783. sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
  1784. (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
  1785. (hwtstamps->hwtstamp.tv64 &&
  1786. sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
  1787. (hwtstamps->syststamp.tv64 &&
  1788. sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
  1789. __sock_recv_timestamp(msg, sk, skb);
  1790. else
  1791. sk->sk_stamp = kt;
  1792. if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
  1793. __sock_recv_wifi_status(msg, sk, skb);
  1794. }
  1795. void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
  1796. struct sk_buff *skb);
  1797. static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
  1798. struct sk_buff *skb)
  1799. {
  1800. #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
  1801. (1UL << SOCK_RCVTSTAMP) | \
  1802. (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
  1803. (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
  1804. (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
  1805. (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
  1806. if (sk->sk_flags & FLAGS_TS_OR_DROPS)
  1807. __sock_recv_ts_and_drops(msg, sk, skb);
  1808. else
  1809. sk->sk_stamp = skb->tstamp;
  1810. }
  1811. /**
  1812. * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
  1813. * @sk: socket sending this packet
  1814. * @tx_flags: filled with instructions for time stamping
  1815. *
  1816. * Currently only depends on SOCK_TIMESTAMPING* flags.
  1817. */
  1818. void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
  1819. /**
  1820. * sk_eat_skb - Release a skb if it is no longer needed
  1821. * @sk: socket to eat this skb from
  1822. * @skb: socket buffer to eat
  1823. * @copied_early: flag indicating whether DMA operations copied this data early
  1824. *
  1825. * This routine must be called with interrupts disabled or with the socket
  1826. * locked so that the sk_buff queue operation is ok.
  1827. */
  1828. #ifdef CONFIG_NET_DMA
  1829. static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
  1830. {
  1831. __skb_unlink(skb, &sk->sk_receive_queue);
  1832. if (!copied_early)
  1833. __kfree_skb(skb);
  1834. else
  1835. __skb_queue_tail(&sk->sk_async_wait_queue, skb);
  1836. }
  1837. #else
  1838. static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
  1839. {
  1840. __skb_unlink(skb, &sk->sk_receive_queue);
  1841. __kfree_skb(skb);
  1842. }
  1843. #endif
  1844. static inline
  1845. struct net *sock_net(const struct sock *sk)
  1846. {
  1847. return read_pnet(&sk->sk_net);
  1848. }
  1849. static inline
  1850. void sock_net_set(struct sock *sk, struct net *net)
  1851. {
  1852. write_pnet(&sk->sk_net, net);
  1853. }
  1854. /*
  1855. * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
  1856. * They should not hold a reference to a namespace in order to allow
  1857. * to stop it.
  1858. * Sockets after sk_change_net should be released using sk_release_kernel
  1859. */
  1860. static inline void sk_change_net(struct sock *sk, struct net *net)
  1861. {
  1862. put_net(sock_net(sk));
  1863. sock_net_set(sk, hold_net(net));
  1864. }
  1865. static inline struct sock *skb_steal_sock(struct sk_buff *skb)
  1866. {
  1867. if (unlikely(skb->sk)) {
  1868. struct sock *sk = skb->sk;
  1869. skb->destructor = NULL;
  1870. skb->sk = NULL;
  1871. return sk;
  1872. }
  1873. return NULL;
  1874. }
  1875. /* This helper checks if a socket is a full socket,
  1876. * ie _not_ a timewait or request socket.
  1877. * TODO: Check for TCPF_NEW_SYN_RECV when that starts to exist.
  1878. */
  1879. static inline bool sk_fullsock(const struct sock *sk)
  1880. {
  1881. return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT);
  1882. }
  1883. void sock_enable_timestamp(struct sock *sk, int flag);
  1884. int sock_get_timestamp(struct sock *, struct timeval __user *);
  1885. int sock_get_timestampns(struct sock *, struct timespec __user *);
  1886. /*
  1887. * Enable debug/info messages
  1888. */
  1889. extern int net_msg_warn;
  1890. #define NETDEBUG(fmt, args...) \
  1891. do { if (net_msg_warn) printk(fmt,##args); } while (0)
  1892. #define LIMIT_NETDEBUG(fmt, args...) \
  1893. do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
  1894. extern __u32 sysctl_wmem_max;
  1895. extern __u32 sysctl_rmem_max;
  1896. void sk_init(void);
  1897. extern int sysctl_optmem_max;
  1898. extern __u32 sysctl_wmem_default;
  1899. extern __u32 sysctl_rmem_default;
  1900. /* SOCKEV Notifier Events */
  1901. #define SOCKEV_SOCKET 0x00
  1902. #define SOCKEV_BIND 0x01
  1903. #define SOCKEV_LISTEN 0x02
  1904. #define SOCKEV_ACCEPT 0x03
  1905. #define SOCKEV_CONNECT 0x04
  1906. #define SOCKEV_SHUTDOWN 0x05
  1907. int sockev_register_notify(struct notifier_block *nb);
  1908. int sockev_unregister_notify(struct notifier_block *nb);
  1909. #endif /* _SOCK_H */