sock.h 61 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192
  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. static inline int sk_peek_offset(struct sock *sk, int flags)
  385. {
  386. if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
  387. return sk->sk_peek_off;
  388. else
  389. return 0;
  390. }
  391. static inline void sk_peek_offset_bwd(struct sock *sk, int val)
  392. {
  393. if (sk->sk_peek_off >= 0) {
  394. if (sk->sk_peek_off >= val)
  395. sk->sk_peek_off -= val;
  396. else
  397. sk->sk_peek_off = 0;
  398. }
  399. }
  400. static inline void sk_peek_offset_fwd(struct sock *sk, int val)
  401. {
  402. if (sk->sk_peek_off >= 0)
  403. sk->sk_peek_off += val;
  404. }
  405. /*
  406. * Hashed lists helper routines
  407. */
  408. static inline struct sock *sk_entry(const struct hlist_node *node)
  409. {
  410. return hlist_entry(node, struct sock, sk_node);
  411. }
  412. static inline struct sock *__sk_head(const struct hlist_head *head)
  413. {
  414. return hlist_entry(head->first, struct sock, sk_node);
  415. }
  416. static inline struct sock *sk_head(const struct hlist_head *head)
  417. {
  418. return hlist_empty(head) ? NULL : __sk_head(head);
  419. }
  420. static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
  421. {
  422. return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
  423. }
  424. static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
  425. {
  426. return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
  427. }
  428. static inline struct sock *sk_next(const struct sock *sk)
  429. {
  430. return sk->sk_node.next ?
  431. hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
  432. }
  433. static inline struct sock *sk_nulls_next(const struct sock *sk)
  434. {
  435. return (!is_a_nulls(sk->sk_nulls_node.next)) ?
  436. hlist_nulls_entry(sk->sk_nulls_node.next,
  437. struct sock, sk_nulls_node) :
  438. NULL;
  439. }
  440. static inline bool sk_unhashed(const struct sock *sk)
  441. {
  442. return hlist_unhashed(&sk->sk_node);
  443. }
  444. static inline bool sk_hashed(const struct sock *sk)
  445. {
  446. return !sk_unhashed(sk);
  447. }
  448. static inline void sk_node_init(struct hlist_node *node)
  449. {
  450. node->pprev = NULL;
  451. }
  452. static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
  453. {
  454. node->pprev = NULL;
  455. }
  456. static inline void __sk_del_node(struct sock *sk)
  457. {
  458. __hlist_del(&sk->sk_node);
  459. }
  460. /* NB: equivalent to hlist_del_init_rcu */
  461. static inline bool __sk_del_node_init(struct sock *sk)
  462. {
  463. if (sk_hashed(sk)) {
  464. __sk_del_node(sk);
  465. sk_node_init(&sk->sk_node);
  466. return true;
  467. }
  468. return false;
  469. }
  470. /* Grab socket reference count. This operation is valid only
  471. when sk is ALREADY grabbed f.e. it is found in hash table
  472. or a list and the lookup is made under lock preventing hash table
  473. modifications.
  474. */
  475. static inline void sock_hold(struct sock *sk)
  476. {
  477. atomic_inc(&sk->sk_refcnt);
  478. }
  479. /* Ungrab socket in the context, which assumes that socket refcnt
  480. cannot hit zero, f.e. it is true in context of any socketcall.
  481. */
  482. static inline void __sock_put(struct sock *sk)
  483. {
  484. atomic_dec(&sk->sk_refcnt);
  485. }
  486. static inline bool sk_del_node_init(struct sock *sk)
  487. {
  488. bool rc = __sk_del_node_init(sk);
  489. if (rc) {
  490. /* paranoid for a while -acme */
  491. WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
  492. __sock_put(sk);
  493. }
  494. return rc;
  495. }
  496. #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
  497. static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
  498. {
  499. if (sk_hashed(sk)) {
  500. hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
  501. return true;
  502. }
  503. return false;
  504. }
  505. static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
  506. {
  507. bool rc = __sk_nulls_del_node_init_rcu(sk);
  508. if (rc) {
  509. /* paranoid for a while -acme */
  510. WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
  511. __sock_put(sk);
  512. }
  513. return rc;
  514. }
  515. static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
  516. {
  517. hlist_add_head(&sk->sk_node, list);
  518. }
  519. static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
  520. {
  521. sock_hold(sk);
  522. __sk_add_node(sk, list);
  523. }
  524. static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
  525. {
  526. sock_hold(sk);
  527. hlist_add_head_rcu(&sk->sk_node, list);
  528. }
  529. static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
  530. {
  531. hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
  532. }
  533. static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
  534. {
  535. sock_hold(sk);
  536. __sk_nulls_add_node_rcu(sk, list);
  537. }
  538. static inline void __sk_del_bind_node(struct sock *sk)
  539. {
  540. __hlist_del(&sk->sk_bind_node);
  541. }
  542. static inline void sk_add_bind_node(struct sock *sk,
  543. struct hlist_head *list)
  544. {
  545. hlist_add_head(&sk->sk_bind_node, list);
  546. }
  547. #define sk_for_each(__sk, node, list) \
  548. hlist_for_each_entry(__sk, node, list, sk_node)
  549. #define sk_for_each_rcu(__sk, node, list) \
  550. hlist_for_each_entry_rcu(__sk, node, list, sk_node)
  551. #define sk_nulls_for_each(__sk, node, list) \
  552. hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
  553. #define sk_nulls_for_each_rcu(__sk, node, list) \
  554. hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
  555. #define sk_for_each_from(__sk, node) \
  556. if (__sk && ({ node = &(__sk)->sk_node; 1; })) \
  557. hlist_for_each_entry_from(__sk, node, sk_node)
  558. #define sk_nulls_for_each_from(__sk, node) \
  559. if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
  560. hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
  561. #define sk_for_each_safe(__sk, node, tmp, list) \
  562. hlist_for_each_entry_safe(__sk, node, tmp, list, sk_node)
  563. #define sk_for_each_bound(__sk, node, list) \
  564. hlist_for_each_entry(__sk, node, list, sk_bind_node)
  565. /* Sock flags */
  566. enum sock_flags {
  567. SOCK_DEAD,
  568. SOCK_DONE,
  569. SOCK_URGINLINE,
  570. SOCK_KEEPOPEN,
  571. SOCK_LINGER,
  572. SOCK_DESTROY,
  573. SOCK_BROADCAST,
  574. SOCK_TIMESTAMP,
  575. SOCK_ZAPPED,
  576. SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
  577. SOCK_DBG, /* %SO_DEBUG setting */
  578. SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
  579. SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
  580. SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
  581. SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
  582. SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
  583. SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
  584. SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
  585. SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
  586. SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
  587. SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
  588. SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
  589. SOCK_FASYNC, /* fasync() active */
  590. SOCK_RXQ_OVFL,
  591. SOCK_ZEROCOPY, /* buffers from userspace */
  592. SOCK_WIFI_STATUS, /* push wifi status to userspace */
  593. SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
  594. * Will use last 4 bytes of packet sent from
  595. * user-space instead.
  596. */
  597. };
  598. static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
  599. {
  600. nsk->sk_flags = osk->sk_flags;
  601. }
  602. static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
  603. {
  604. __set_bit(flag, &sk->sk_flags);
  605. }
  606. static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
  607. {
  608. __clear_bit(flag, &sk->sk_flags);
  609. }
  610. static inline int sock_flag(struct sock *sk, enum sock_flags flag)
  611. {
  612. return test_bit(flag, &sk->sk_flags);
  613. }
  614. static inline void sk_acceptq_removed(struct sock *sk)
  615. {
  616. sk->sk_ack_backlog--;
  617. }
  618. static inline void sk_acceptq_added(struct sock *sk)
  619. {
  620. sk->sk_ack_backlog++;
  621. }
  622. static inline bool sk_acceptq_is_full(const struct sock *sk)
  623. {
  624. return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
  625. }
  626. /*
  627. * Compute minimal free write space needed to queue new packets.
  628. */
  629. static inline int sk_stream_min_wspace(const struct sock *sk)
  630. {
  631. return sk->sk_wmem_queued >> 1;
  632. }
  633. static inline int sk_stream_wspace(const struct sock *sk)
  634. {
  635. return sk->sk_sndbuf - sk->sk_wmem_queued;
  636. }
  637. void sk_stream_write_space(struct sock *sk);
  638. static inline bool sk_stream_memory_free(const struct sock *sk)
  639. {
  640. return sk->sk_wmem_queued < sk->sk_sndbuf;
  641. }
  642. /* OOB backlog add */
  643. static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
  644. {
  645. /* dont let skb dst not refcounted, we are going to leave rcu lock */
  646. skb_dst_force(skb);
  647. if (!sk->sk_backlog.tail)
  648. sk->sk_backlog.head = skb;
  649. else
  650. sk->sk_backlog.tail->next = skb;
  651. sk->sk_backlog.tail = skb;
  652. skb->next = NULL;
  653. }
  654. /*
  655. * Take into account size of receive queue and backlog queue
  656. * Do not take into account this skb truesize,
  657. * to allow even a single big packet to come.
  658. */
  659. static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb)
  660. {
  661. unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
  662. return qsize > sk->sk_rcvbuf;
  663. }
  664. /* The per-socket spinlock must be held here. */
  665. static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb)
  666. {
  667. if (sk_rcvqueues_full(sk, skb))
  668. return -ENOBUFS;
  669. __sk_add_backlog(sk, skb);
  670. sk->sk_backlog.len += skb->truesize;
  671. return 0;
  672. }
  673. static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
  674. {
  675. return sk->sk_backlog_rcv(sk, skb);
  676. }
  677. static inline void sock_rps_record_flow(const struct sock *sk)
  678. {
  679. #ifdef CONFIG_RPS
  680. struct rps_sock_flow_table *sock_flow_table;
  681. rcu_read_lock();
  682. sock_flow_table = rcu_dereference(rps_sock_flow_table);
  683. rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
  684. rcu_read_unlock();
  685. #endif
  686. }
  687. static inline void sock_rps_reset_flow(const struct sock *sk)
  688. {
  689. #ifdef CONFIG_RPS
  690. struct rps_sock_flow_table *sock_flow_table;
  691. rcu_read_lock();
  692. sock_flow_table = rcu_dereference(rps_sock_flow_table);
  693. rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
  694. rcu_read_unlock();
  695. #endif
  696. }
  697. static inline void sock_rps_save_rxhash(struct sock *sk,
  698. const struct sk_buff *skb)
  699. {
  700. #ifdef CONFIG_RPS
  701. if (unlikely(sk->sk_rxhash != skb->rxhash)) {
  702. sock_rps_reset_flow(sk);
  703. sk->sk_rxhash = skb->rxhash;
  704. }
  705. #endif
  706. }
  707. static inline void sock_rps_reset_rxhash(struct sock *sk)
  708. {
  709. #ifdef CONFIG_RPS
  710. sock_rps_reset_flow(sk);
  711. sk->sk_rxhash = 0;
  712. #endif
  713. }
  714. #define sk_wait_event(__sk, __timeo, __condition) \
  715. ({ int __rc; \
  716. release_sock(__sk); \
  717. __rc = __condition; \
  718. if (!__rc) { \
  719. *(__timeo) = schedule_timeout(*(__timeo)); \
  720. } \
  721. lock_sock(__sk); \
  722. __rc = __condition; \
  723. __rc; \
  724. })
  725. int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
  726. int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
  727. void sk_stream_wait_close(struct sock *sk, long timeo_p);
  728. int sk_stream_error(struct sock *sk, int flags, int err);
  729. void sk_stream_kill_queues(struct sock *sk);
  730. int sk_wait_data(struct sock *sk, long *timeo, const struct sk_buff *skb);
  731. struct request_sock_ops;
  732. struct timewait_sock_ops;
  733. struct inet_hashinfo;
  734. struct raw_hashinfo;
  735. struct module;
  736. /*
  737. * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
  738. * un-modified. Special care is taken when initializing object to zero.
  739. */
  740. static inline void sk_prot_clear_nulls(struct sock *sk, int size)
  741. {
  742. if (offsetof(struct sock, sk_node.next) != 0)
  743. memset(sk, 0, offsetof(struct sock, sk_node.next));
  744. memset(&sk->sk_node.pprev, 0,
  745. size - offsetof(struct sock, sk_node.pprev));
  746. }
  747. /* Networking protocol blocks we attach to sockets.
  748. * socket layer -> transport layer interface
  749. * transport -> network interface is defined by struct inet_proto
  750. */
  751. struct proto {
  752. void (*close)(struct sock *sk,
  753. long timeout);
  754. int (*connect)(struct sock *sk,
  755. struct sockaddr *uaddr,
  756. int addr_len);
  757. int (*disconnect)(struct sock *sk, int flags);
  758. struct sock * (*accept)(struct sock *sk, int flags, int *err);
  759. int (*ioctl)(struct sock *sk, int cmd,
  760. unsigned long arg);
  761. int (*init)(struct sock *sk);
  762. void (*destroy)(struct sock *sk);
  763. void (*shutdown)(struct sock *sk, int how);
  764. int (*setsockopt)(struct sock *sk, int level,
  765. int optname, char __user *optval,
  766. unsigned int optlen);
  767. int (*getsockopt)(struct sock *sk, int level,
  768. int optname, char __user *optval,
  769. int __user *option);
  770. #ifdef CONFIG_COMPAT
  771. int (*compat_setsockopt)(struct sock *sk,
  772. int level,
  773. int optname, char __user *optval,
  774. unsigned int optlen);
  775. int (*compat_getsockopt)(struct sock *sk,
  776. int level,
  777. int optname, char __user *optval,
  778. int __user *option);
  779. int (*compat_ioctl)(struct sock *sk,
  780. unsigned int cmd, unsigned long arg);
  781. #endif
  782. int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
  783. struct msghdr *msg, size_t len);
  784. int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
  785. struct msghdr *msg,
  786. size_t len, int noblock, int flags,
  787. int *addr_len);
  788. int (*sendpage)(struct sock *sk, struct page *page,
  789. int offset, size_t size, int flags);
  790. int (*bind)(struct sock *sk,
  791. struct sockaddr *uaddr, int addr_len);
  792. int (*backlog_rcv) (struct sock *sk,
  793. struct sk_buff *skb);
  794. void (*release_cb)(struct sock *sk);
  795. /* Keeping track of sk's, looking them up, and port selection methods. */
  796. void (*hash)(struct sock *sk);
  797. void (*unhash)(struct sock *sk);
  798. void (*rehash)(struct sock *sk);
  799. int (*get_port)(struct sock *sk, unsigned short snum);
  800. void (*clear_sk)(struct sock *sk, int size);
  801. /* Keeping track of sockets in use */
  802. #ifdef CONFIG_PROC_FS
  803. unsigned int inuse_idx;
  804. #endif
  805. /* Memory pressure */
  806. void (*enter_memory_pressure)(struct sock *sk);
  807. atomic_long_t *memory_allocated; /* Current allocated memory. */
  808. struct percpu_counter *sockets_allocated; /* Current number of sockets. */
  809. /*
  810. * Pressure flag: try to collapse.
  811. * Technical note: it is used by multiple contexts non atomically.
  812. * All the __sk_mem_schedule() is of this nature: accounting
  813. * is strict, actions are advisory and have some latency.
  814. */
  815. int *memory_pressure;
  816. long *sysctl_mem;
  817. int *sysctl_wmem;
  818. int *sysctl_rmem;
  819. int max_header;
  820. bool no_autobind;
  821. struct kmem_cache *slab;
  822. unsigned int obj_size;
  823. int slab_flags;
  824. struct percpu_counter *orphan_count;
  825. struct request_sock_ops *rsk_prot;
  826. struct timewait_sock_ops *twsk_prot;
  827. union {
  828. struct inet_hashinfo *hashinfo;
  829. struct udp_table *udp_table;
  830. struct raw_hashinfo *raw_hash;
  831. } h;
  832. struct module *owner;
  833. char name[32];
  834. struct list_head node;
  835. #ifdef SOCK_REFCNT_DEBUG
  836. atomic_t socks;
  837. #endif
  838. #ifdef CONFIG_CGROUP_MEM_RES_CTLR_KMEM
  839. /*
  840. * cgroup specific init/deinit functions. Called once for all
  841. * protocols that implement it, from cgroups populate function.
  842. * This function has to setup any files the protocol want to
  843. * appear in the kmem cgroup filesystem.
  844. */
  845. int (*init_cgroup)(struct cgroup *cgrp,
  846. struct cgroup_subsys *ss);
  847. void (*destroy_cgroup)(struct cgroup *cgrp);
  848. struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
  849. #endif
  850. int (*diag_destroy)(struct sock *sk, int err);
  851. };
  852. struct cg_proto {
  853. void (*enter_memory_pressure)(struct sock *sk);
  854. struct res_counter *memory_allocated; /* Current allocated memory. */
  855. struct percpu_counter *sockets_allocated; /* Current number of sockets. */
  856. int *memory_pressure;
  857. long *sysctl_mem;
  858. /*
  859. * memcg field is used to find which memcg we belong directly
  860. * Each memcg struct can hold more than one cg_proto, so container_of
  861. * won't really cut.
  862. *
  863. * The elegant solution would be having an inverse function to
  864. * proto_cgroup in struct proto, but that means polluting the structure
  865. * for everybody, instead of just for memcg users.
  866. */
  867. struct mem_cgroup *memcg;
  868. };
  869. int proto_register(struct proto *prot, int alloc_slab);
  870. void proto_unregister(struct proto *prot);
  871. #ifdef SOCK_REFCNT_DEBUG
  872. static inline void sk_refcnt_debug_inc(struct sock *sk)
  873. {
  874. atomic_inc(&sk->sk_prot->socks);
  875. }
  876. static inline void sk_refcnt_debug_dec(struct sock *sk)
  877. {
  878. atomic_dec(&sk->sk_prot->socks);
  879. printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
  880. sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
  881. }
  882. static inline void sk_refcnt_debug_release(const struct sock *sk)
  883. {
  884. if (atomic_read(&sk->sk_refcnt) != 1)
  885. printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
  886. sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
  887. }
  888. #else /* SOCK_REFCNT_DEBUG */
  889. #define sk_refcnt_debug_inc(sk) do { } while (0)
  890. #define sk_refcnt_debug_dec(sk) do { } while (0)
  891. #define sk_refcnt_debug_release(sk) do { } while (0)
  892. #endif /* SOCK_REFCNT_DEBUG */
  893. #if defined(CONFIG_CGROUP_MEM_RES_CTLR_KMEM) && defined(CONFIG_NET)
  894. extern struct static_key memcg_socket_limit_enabled;
  895. static inline struct cg_proto *parent_cg_proto(struct proto *proto,
  896. struct cg_proto *cg_proto)
  897. {
  898. return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
  899. }
  900. #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
  901. #else
  902. #define mem_cgroup_sockets_enabled 0
  903. static inline struct cg_proto *parent_cg_proto(struct proto *proto,
  904. struct cg_proto *cg_proto)
  905. {
  906. return NULL;
  907. }
  908. #endif
  909. static inline bool sk_has_memory_pressure(const struct sock *sk)
  910. {
  911. return sk->sk_prot->memory_pressure != NULL;
  912. }
  913. static inline bool sk_under_memory_pressure(const struct sock *sk)
  914. {
  915. if (!sk->sk_prot->memory_pressure)
  916. return false;
  917. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  918. return !!*sk->sk_cgrp->memory_pressure;
  919. return !!*sk->sk_prot->memory_pressure;
  920. }
  921. static inline void sk_leave_memory_pressure(struct sock *sk)
  922. {
  923. int *memory_pressure = sk->sk_prot->memory_pressure;
  924. if (!memory_pressure)
  925. return;
  926. if (*memory_pressure)
  927. *memory_pressure = 0;
  928. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  929. struct cg_proto *cg_proto = sk->sk_cgrp;
  930. struct proto *prot = sk->sk_prot;
  931. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  932. if (*cg_proto->memory_pressure)
  933. *cg_proto->memory_pressure = 0;
  934. }
  935. }
  936. static inline void sk_enter_memory_pressure(struct sock *sk)
  937. {
  938. if (!sk->sk_prot->enter_memory_pressure)
  939. return;
  940. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  941. struct cg_proto *cg_proto = sk->sk_cgrp;
  942. struct proto *prot = sk->sk_prot;
  943. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  944. cg_proto->enter_memory_pressure(sk);
  945. }
  946. sk->sk_prot->enter_memory_pressure(sk);
  947. }
  948. static inline long sk_prot_mem_limits(const struct sock *sk, int index)
  949. {
  950. long *prot = sk->sk_prot->sysctl_mem;
  951. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  952. prot = sk->sk_cgrp->sysctl_mem;
  953. return prot[index];
  954. }
  955. static inline void memcg_memory_allocated_add(struct cg_proto *prot,
  956. unsigned long amt,
  957. int *parent_status)
  958. {
  959. struct res_counter *fail;
  960. int ret;
  961. ret = res_counter_charge_nofail(prot->memory_allocated,
  962. amt << PAGE_SHIFT, &fail);
  963. if (ret < 0)
  964. *parent_status = OVER_LIMIT;
  965. }
  966. static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
  967. unsigned long amt)
  968. {
  969. res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
  970. }
  971. static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
  972. {
  973. u64 ret;
  974. ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
  975. return ret >> PAGE_SHIFT;
  976. }
  977. static inline long
  978. sk_memory_allocated(const struct sock *sk)
  979. {
  980. struct proto *prot = sk->sk_prot;
  981. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  982. return memcg_memory_allocated_read(sk->sk_cgrp);
  983. return atomic_long_read(prot->memory_allocated);
  984. }
  985. static inline long
  986. sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
  987. {
  988. struct proto *prot = sk->sk_prot;
  989. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  990. memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
  991. /* update the root cgroup regardless */
  992. atomic_long_add_return(amt, prot->memory_allocated);
  993. return memcg_memory_allocated_read(sk->sk_cgrp);
  994. }
  995. return atomic_long_add_return(amt, prot->memory_allocated);
  996. }
  997. static inline void
  998. sk_memory_allocated_sub(struct sock *sk, int amt)
  999. {
  1000. struct proto *prot = sk->sk_prot;
  1001. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1002. memcg_memory_allocated_sub(sk->sk_cgrp, amt);
  1003. atomic_long_sub(amt, prot->memory_allocated);
  1004. }
  1005. static inline void sk_sockets_allocated_dec(struct sock *sk)
  1006. {
  1007. struct proto *prot = sk->sk_prot;
  1008. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1009. struct cg_proto *cg_proto = sk->sk_cgrp;
  1010. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  1011. percpu_counter_dec(cg_proto->sockets_allocated);
  1012. }
  1013. percpu_counter_dec(prot->sockets_allocated);
  1014. }
  1015. static inline void sk_sockets_allocated_inc(struct sock *sk)
  1016. {
  1017. struct proto *prot = sk->sk_prot;
  1018. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1019. struct cg_proto *cg_proto = sk->sk_cgrp;
  1020. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  1021. percpu_counter_inc(cg_proto->sockets_allocated);
  1022. }
  1023. percpu_counter_inc(prot->sockets_allocated);
  1024. }
  1025. static inline int
  1026. sk_sockets_allocated_read_positive(struct sock *sk)
  1027. {
  1028. struct proto *prot = sk->sk_prot;
  1029. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1030. return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
  1031. return percpu_counter_read_positive(prot->sockets_allocated);
  1032. }
  1033. static inline int
  1034. proto_sockets_allocated_sum_positive(struct proto *prot)
  1035. {
  1036. return percpu_counter_sum_positive(prot->sockets_allocated);
  1037. }
  1038. static inline long
  1039. proto_memory_allocated(struct proto *prot)
  1040. {
  1041. return atomic_long_read(prot->memory_allocated);
  1042. }
  1043. static inline bool
  1044. proto_memory_pressure(struct proto *prot)
  1045. {
  1046. if (!prot->memory_pressure)
  1047. return false;
  1048. return !!*prot->memory_pressure;
  1049. }
  1050. #ifdef CONFIG_PROC_FS
  1051. /* Called with local bh disabled */
  1052. void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
  1053. int sock_prot_inuse_get(struct net *net, struct proto *proto);
  1054. #else
  1055. static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
  1056. int inc)
  1057. {
  1058. }
  1059. #endif
  1060. /* With per-bucket locks this operation is not-atomic, so that
  1061. * this version is not worse.
  1062. */
  1063. static inline void __sk_prot_rehash(struct sock *sk)
  1064. {
  1065. sk->sk_prot->unhash(sk);
  1066. sk->sk_prot->hash(sk);
  1067. }
  1068. void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
  1069. /* About 10 seconds */
  1070. #define SOCK_DESTROY_TIME (10*HZ)
  1071. /* Sockets 0-1023 can't be bound to unless you are superuser */
  1072. #define PROT_SOCK 1024
  1073. #define SHUTDOWN_MASK 3
  1074. #define RCV_SHUTDOWN 1
  1075. #define SEND_SHUTDOWN 2
  1076. #define SOCK_SNDBUF_LOCK 1
  1077. #define SOCK_RCVBUF_LOCK 2
  1078. #define SOCK_BINDADDR_LOCK 4
  1079. #define SOCK_BINDPORT_LOCK 8
  1080. /* sock_iocb: used to kick off async processing of socket ios */
  1081. struct sock_iocb {
  1082. struct list_head list;
  1083. int flags;
  1084. int size;
  1085. struct socket *sock;
  1086. struct sock *sk;
  1087. struct scm_cookie *scm;
  1088. struct msghdr *msg, async_msg;
  1089. struct kiocb *kiocb;
  1090. };
  1091. static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
  1092. {
  1093. return (struct sock_iocb *)iocb->private;
  1094. }
  1095. static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
  1096. {
  1097. return si->kiocb;
  1098. }
  1099. struct socket_alloc {
  1100. struct socket socket;
  1101. struct inode vfs_inode;
  1102. };
  1103. static inline struct socket *SOCKET_I(struct inode *inode)
  1104. {
  1105. return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
  1106. }
  1107. static inline struct inode *SOCK_INODE(struct socket *socket)
  1108. {
  1109. return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
  1110. }
  1111. /*
  1112. * Functions for memory accounting
  1113. */
  1114. int __sk_mem_schedule(struct sock *sk, int size, int kind);
  1115. void __sk_mem_reclaim(struct sock *sk);
  1116. #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
  1117. #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
  1118. #define SK_MEM_SEND 0
  1119. #define SK_MEM_RECV 1
  1120. static inline int sk_mem_pages(int amt)
  1121. {
  1122. return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
  1123. }
  1124. static inline bool sk_has_account(struct sock *sk)
  1125. {
  1126. /* return true if protocol supports memory accounting */
  1127. return !!sk->sk_prot->memory_allocated;
  1128. }
  1129. static inline bool sk_wmem_schedule(struct sock *sk, int size)
  1130. {
  1131. if (!sk_has_account(sk))
  1132. return true;
  1133. return size <= sk->sk_forward_alloc ||
  1134. __sk_mem_schedule(sk, size, SK_MEM_SEND);
  1135. }
  1136. static inline bool sk_rmem_schedule(struct sock *sk, int size)
  1137. {
  1138. if (!sk_has_account(sk))
  1139. return true;
  1140. return size <= sk->sk_forward_alloc ||
  1141. __sk_mem_schedule(sk, size, SK_MEM_RECV);
  1142. }
  1143. static inline void sk_mem_reclaim(struct sock *sk)
  1144. {
  1145. if (!sk_has_account(sk))
  1146. return;
  1147. if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
  1148. __sk_mem_reclaim(sk);
  1149. }
  1150. static inline void sk_mem_reclaim_partial(struct sock *sk)
  1151. {
  1152. if (!sk_has_account(sk))
  1153. return;
  1154. if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
  1155. __sk_mem_reclaim(sk);
  1156. }
  1157. static inline void sk_mem_charge(struct sock *sk, int size)
  1158. {
  1159. if (!sk_has_account(sk))
  1160. return;
  1161. sk->sk_forward_alloc -= size;
  1162. }
  1163. static inline void sk_mem_uncharge(struct sock *sk, int size)
  1164. {
  1165. if (!sk_has_account(sk))
  1166. return;
  1167. sk->sk_forward_alloc += size;
  1168. }
  1169. static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
  1170. {
  1171. sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
  1172. sk->sk_wmem_queued -= skb->truesize;
  1173. sk_mem_uncharge(sk, skb->truesize);
  1174. __kfree_skb(skb);
  1175. }
  1176. /* Used by processes to "lock" a socket state, so that
  1177. * interrupts and bottom half handlers won't change it
  1178. * from under us. It essentially blocks any incoming
  1179. * packets, so that we won't get any new data or any
  1180. * packets that change the state of the socket.
  1181. *
  1182. * While locked, BH processing will add new packets to
  1183. * the backlog queue. This queue is processed by the
  1184. * owner of the socket lock right before it is released.
  1185. *
  1186. * Since ~2.3.5 it is also exclusive sleep lock serializing
  1187. * accesses from user process context.
  1188. */
  1189. #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
  1190. /*
  1191. * Macro so as to not evaluate some arguments when
  1192. * lockdep is not enabled.
  1193. *
  1194. * Mark both the sk_lock and the sk_lock.slock as a
  1195. * per-address-family lock class.
  1196. */
  1197. #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
  1198. do { \
  1199. sk->sk_lock.owned = 0; \
  1200. init_waitqueue_head(&sk->sk_lock.wq); \
  1201. spin_lock_init(&(sk)->sk_lock.slock); \
  1202. debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
  1203. sizeof((sk)->sk_lock)); \
  1204. lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
  1205. (skey), (sname)); \
  1206. lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
  1207. } while (0)
  1208. void lock_sock_nested(struct sock *sk, int subclass);
  1209. static inline void lock_sock(struct sock *sk)
  1210. {
  1211. lock_sock_nested(sk, 0);
  1212. }
  1213. void release_sock(struct sock *sk);
  1214. /* BH context may only use the following locking interface. */
  1215. #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
  1216. #define bh_lock_sock_nested(__sk) \
  1217. spin_lock_nested(&((__sk)->sk_lock.slock), \
  1218. SINGLE_DEPTH_NESTING)
  1219. #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
  1220. bool lock_sock_fast(struct sock *sk);
  1221. /**
  1222. * unlock_sock_fast - complement of lock_sock_fast
  1223. * @sk: socket
  1224. * @slow: slow mode
  1225. *
  1226. * fast unlock socket for user context.
  1227. * If slow mode is on, we call regular release_sock()
  1228. */
  1229. static inline void unlock_sock_fast(struct sock *sk, bool slow)
  1230. {
  1231. if (slow)
  1232. release_sock(sk);
  1233. else
  1234. spin_unlock_bh(&sk->sk_lock.slock);
  1235. }
  1236. struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
  1237. struct proto *prot);
  1238. void sk_free(struct sock *sk);
  1239. void sk_release_kernel(struct sock *sk);
  1240. struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
  1241. struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
  1242. gfp_t priority);
  1243. struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
  1244. gfp_t priority);
  1245. void sock_wfree(struct sk_buff *skb);
  1246. void sock_rfree(struct sk_buff *skb);
  1247. int sock_setsockopt(struct socket *sock, int level, int op,
  1248. char __user *optval, unsigned int optlen);
  1249. int sock_getsockopt(struct socket *sock, int level, int op,
  1250. char __user *optval, int __user *optlen);
  1251. struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
  1252. int noblock, int *errcode);
  1253. struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
  1254. unsigned long data_len, int noblock,
  1255. int *errcode);
  1256. void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
  1257. void sock_kfree_s(struct sock *sk, void *mem, int size);
  1258. void sk_send_sigurg(struct sock *sk);
  1259. #ifdef CONFIG_CGROUPS
  1260. void sock_update_classid(struct sock *sk);
  1261. #else
  1262. static inline void sock_update_classid(struct sock *sk)
  1263. {
  1264. }
  1265. #endif
  1266. /*
  1267. * Functions to fill in entries in struct proto_ops when a protocol
  1268. * does not implement a particular function.
  1269. */
  1270. int sock_no_bind(struct socket *, struct sockaddr *, int);
  1271. int sock_no_connect(struct socket *, struct sockaddr *, int, int);
  1272. int sock_no_socketpair(struct socket *, struct socket *);
  1273. int sock_no_accept(struct socket *, struct socket *, int);
  1274. int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
  1275. unsigned int sock_no_poll(struct file *, struct socket *,
  1276. struct poll_table_struct *);
  1277. int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
  1278. int sock_no_listen(struct socket *, int);
  1279. int sock_no_shutdown(struct socket *, int);
  1280. int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
  1281. int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
  1282. int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
  1283. int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
  1284. int);
  1285. int sock_no_mmap(struct file *file, struct socket *sock,
  1286. struct vm_area_struct *vma);
  1287. ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
  1288. size_t size, int flags);
  1289. /*
  1290. * Functions to fill in entries in struct proto_ops when a protocol
  1291. * uses the inet style.
  1292. */
  1293. int sock_common_getsockopt(struct socket *sock, int level, int optname,
  1294. char __user *optval, int __user *optlen);
  1295. int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
  1296. struct msghdr *msg, size_t size, int flags);
  1297. int sock_common_setsockopt(struct socket *sock, int level, int optname,
  1298. char __user *optval, unsigned int optlen);
  1299. int compat_sock_common_getsockopt(struct socket *sock, int level,
  1300. int optname, char __user *optval, int __user *optlen);
  1301. int compat_sock_common_setsockopt(struct socket *sock, int level,
  1302. int optname, char __user *optval, unsigned int optlen);
  1303. void sk_common_release(struct sock *sk);
  1304. /*
  1305. * Default socket callbacks and setup code
  1306. */
  1307. /* Initialise core socket variables */
  1308. void sock_init_data(struct socket *sock, struct sock *sk);
  1309. void sk_filter_release_rcu(struct rcu_head *rcu);
  1310. /**
  1311. * sk_filter_release - release a socket filter
  1312. * @fp: filter to remove
  1313. *
  1314. * Remove a filter from a socket and release its resources.
  1315. */
  1316. static inline void sk_filter_release(struct sk_filter *fp)
  1317. {
  1318. if (atomic_dec_and_test(&fp->refcnt))
  1319. call_rcu(&fp->rcu, sk_filter_release_rcu);
  1320. }
  1321. static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
  1322. {
  1323. unsigned int size = sk_filter_len(fp);
  1324. atomic_sub(size, &sk->sk_omem_alloc);
  1325. sk_filter_release(fp);
  1326. }
  1327. static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
  1328. {
  1329. atomic_inc(&fp->refcnt);
  1330. atomic_add(sk_filter_len(fp), &sk->sk_omem_alloc);
  1331. }
  1332. /*
  1333. * Socket reference counting postulates.
  1334. *
  1335. * * Each user of socket SHOULD hold a reference count.
  1336. * * Each access point to socket (an hash table bucket, reference from a list,
  1337. * running timer, skb in flight MUST hold a reference count.
  1338. * * When reference count hits 0, it means it will never increase back.
  1339. * * When reference count hits 0, it means that no references from
  1340. * outside exist to this socket and current process on current CPU
  1341. * is last user and may/should destroy this socket.
  1342. * * sk_free is called from any context: process, BH, IRQ. When
  1343. * it is called, socket has no references from outside -> sk_free
  1344. * may release descendant resources allocated by the socket, but
  1345. * to the time when it is called, socket is NOT referenced by any
  1346. * hash tables, lists etc.
  1347. * * Packets, delivered from outside (from network or from another process)
  1348. * and enqueued on receive/error queues SHOULD NOT grab reference count,
  1349. * when they sit in queue. Otherwise, packets will leak to hole, when
  1350. * socket is looked up by one cpu and unhasing is made by another CPU.
  1351. * It is true for udp/raw, netlink (leak to receive and error queues), tcp
  1352. * (leak to backlog). Packet socket does all the processing inside
  1353. * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
  1354. * use separate SMP lock, so that they are prone too.
  1355. */
  1356. /* Ungrab socket and destroy it, if it was the last reference. */
  1357. static inline void sock_put(struct sock *sk)
  1358. {
  1359. if (atomic_dec_and_test(&sk->sk_refcnt))
  1360. sk_free(sk);
  1361. }
  1362. int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
  1363. static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
  1364. {
  1365. sk->sk_tx_queue_mapping = tx_queue;
  1366. }
  1367. static inline void sk_tx_queue_clear(struct sock *sk)
  1368. {
  1369. sk->sk_tx_queue_mapping = -1;
  1370. }
  1371. static inline int sk_tx_queue_get(const struct sock *sk)
  1372. {
  1373. return sk ? sk->sk_tx_queue_mapping : -1;
  1374. }
  1375. static inline void sk_set_socket(struct sock *sk, struct socket *sock)
  1376. {
  1377. sk_tx_queue_clear(sk);
  1378. sk->sk_socket = sock;
  1379. }
  1380. static inline wait_queue_head_t *sk_sleep(struct sock *sk)
  1381. {
  1382. BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
  1383. return &rcu_dereference_raw(sk->sk_wq)->wait;
  1384. }
  1385. /* Detach socket from process context.
  1386. * Announce socket dead, detach it from wait queue and inode.
  1387. * Note that parent inode held reference count on this struct sock,
  1388. * we do not release it in this function, because protocol
  1389. * probably wants some additional cleanups or even continuing
  1390. * to work with this socket (TCP).
  1391. */
  1392. static inline void sock_orphan(struct sock *sk)
  1393. {
  1394. write_lock_bh(&sk->sk_callback_lock);
  1395. sock_set_flag(sk, SOCK_DEAD);
  1396. sk_set_socket(sk, NULL);
  1397. sk->sk_wq = NULL;
  1398. write_unlock_bh(&sk->sk_callback_lock);
  1399. }
  1400. static inline void sock_graft(struct sock *sk, struct socket *parent)
  1401. {
  1402. write_lock_bh(&sk->sk_callback_lock);
  1403. sk->sk_wq = parent->wq;
  1404. parent->sk = sk;
  1405. sk_set_socket(sk, parent);
  1406. sk->sk_uid = SOCK_INODE(parent)->i_uid;
  1407. security_sock_graft(sk, parent);
  1408. write_unlock_bh(&sk->sk_callback_lock);
  1409. }
  1410. int sock_i_uid(struct sock *sk);
  1411. unsigned long sock_i_ino(struct sock *sk);
  1412. static inline kuid_t sock_net_uid(const struct net *net, const struct sock *sk)
  1413. {
  1414. return sk ? sk->sk_uid : make_kuid(net->user_ns, 0);
  1415. }
  1416. static inline struct dst_entry *
  1417. __sk_dst_get(struct sock *sk)
  1418. {
  1419. return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
  1420. lockdep_is_held(&sk->sk_lock.slock));
  1421. }
  1422. static inline struct dst_entry *
  1423. sk_dst_get(struct sock *sk)
  1424. {
  1425. struct dst_entry *dst;
  1426. rcu_read_lock();
  1427. dst = rcu_dereference(sk->sk_dst_cache);
  1428. if (dst)
  1429. dst_hold(dst);
  1430. rcu_read_unlock();
  1431. return dst;
  1432. }
  1433. void sk_reset_txq(struct sock *sk);
  1434. static inline void dst_negative_advice(struct sock *sk)
  1435. {
  1436. struct dst_entry *ndst, *dst = __sk_dst_get(sk);
  1437. if (dst && dst->ops->negative_advice) {
  1438. ndst = dst->ops->negative_advice(dst);
  1439. if (ndst != dst) {
  1440. rcu_assign_pointer(sk->sk_dst_cache, ndst);
  1441. sk_reset_txq(sk);
  1442. }
  1443. }
  1444. }
  1445. static inline void
  1446. __sk_dst_set(struct sock *sk, struct dst_entry *dst)
  1447. {
  1448. struct dst_entry *old_dst;
  1449. sk_tx_queue_clear(sk);
  1450. /*
  1451. * This can be called while sk is owned by the caller only,
  1452. * with no state that can be checked in a rcu_dereference_check() cond
  1453. */
  1454. old_dst = rcu_dereference_raw(sk->sk_dst_cache);
  1455. rcu_assign_pointer(sk->sk_dst_cache, dst);
  1456. dst_release(old_dst);
  1457. }
  1458. static inline void
  1459. sk_dst_set(struct sock *sk, struct dst_entry *dst)
  1460. {
  1461. spin_lock(&sk->sk_dst_lock);
  1462. __sk_dst_set(sk, dst);
  1463. spin_unlock(&sk->sk_dst_lock);
  1464. }
  1465. static inline void
  1466. __sk_dst_reset(struct sock *sk)
  1467. {
  1468. __sk_dst_set(sk, NULL);
  1469. }
  1470. static inline void
  1471. sk_dst_reset(struct sock *sk)
  1472. {
  1473. spin_lock(&sk->sk_dst_lock);
  1474. __sk_dst_reset(sk);
  1475. spin_unlock(&sk->sk_dst_lock);
  1476. }
  1477. struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
  1478. struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
  1479. static inline bool sk_can_gso(const struct sock *sk)
  1480. {
  1481. return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
  1482. }
  1483. void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
  1484. static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
  1485. {
  1486. sk->sk_route_nocaps |= flags;
  1487. sk->sk_route_caps &= ~flags;
  1488. }
  1489. static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
  1490. char __user *from, char *to,
  1491. int copy, int offset)
  1492. {
  1493. if (skb->ip_summed == CHECKSUM_NONE) {
  1494. int err = 0;
  1495. __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
  1496. if (err)
  1497. return err;
  1498. skb->csum = csum_block_add(skb->csum, csum, offset);
  1499. } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
  1500. if (!access_ok(VERIFY_READ, from, copy) ||
  1501. __copy_from_user_nocache(to, from, copy))
  1502. return -EFAULT;
  1503. } else if (copy_from_user(to, from, copy))
  1504. return -EFAULT;
  1505. return 0;
  1506. }
  1507. static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
  1508. char __user *from, int copy)
  1509. {
  1510. int err, offset = skb->len;
  1511. err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
  1512. copy, offset);
  1513. if (err)
  1514. __skb_trim(skb, offset);
  1515. return err;
  1516. }
  1517. static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
  1518. struct sk_buff *skb,
  1519. struct page *page,
  1520. int off, int copy)
  1521. {
  1522. int err;
  1523. err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
  1524. copy, skb->len);
  1525. if (err)
  1526. return err;
  1527. skb->len += copy;
  1528. skb->data_len += copy;
  1529. skb->truesize += copy;
  1530. sk->sk_wmem_queued += copy;
  1531. sk_mem_charge(sk, copy);
  1532. return 0;
  1533. }
  1534. static inline int skb_copy_to_page(struct sock *sk, char __user *from,
  1535. struct sk_buff *skb, struct page *page,
  1536. int off, int copy)
  1537. {
  1538. if (skb->ip_summed == CHECKSUM_NONE) {
  1539. int err = 0;
  1540. __wsum csum = csum_and_copy_from_user(from,
  1541. page_address(page) + off,
  1542. copy, 0, &err);
  1543. if (err)
  1544. return err;
  1545. skb->csum = csum_block_add(skb->csum, csum, skb->len);
  1546. } else if (copy_from_user(page_address(page) + off, from, copy))
  1547. return -EFAULT;
  1548. skb->len += copy;
  1549. skb->data_len += copy;
  1550. skb->truesize += copy;
  1551. sk->sk_wmem_queued += copy;
  1552. sk_mem_charge(sk, copy);
  1553. return 0;
  1554. }
  1555. /**
  1556. * sk_wmem_alloc_get - returns write allocations
  1557. * @sk: socket
  1558. *
  1559. * Returns sk_wmem_alloc minus initial offset of one
  1560. */
  1561. static inline int sk_wmem_alloc_get(const struct sock *sk)
  1562. {
  1563. return atomic_read(&sk->sk_wmem_alloc) - 1;
  1564. }
  1565. /**
  1566. * sk_rmem_alloc_get - returns read allocations
  1567. * @sk: socket
  1568. *
  1569. * Returns sk_rmem_alloc
  1570. */
  1571. static inline int sk_rmem_alloc_get(const struct sock *sk)
  1572. {
  1573. return atomic_read(&sk->sk_rmem_alloc);
  1574. }
  1575. /**
  1576. * sk_has_allocations - check if allocations are outstanding
  1577. * @sk: socket
  1578. *
  1579. * Returns true if socket has write or read allocations
  1580. */
  1581. static inline bool sk_has_allocations(const struct sock *sk)
  1582. {
  1583. return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
  1584. }
  1585. /**
  1586. * wq_has_sleeper - check if there are any waiting processes
  1587. * @wq: struct socket_wq
  1588. *
  1589. * Returns true if socket_wq has waiting processes
  1590. *
  1591. * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
  1592. * barrier call. They were added due to the race found within the tcp code.
  1593. *
  1594. * Consider following tcp code paths:
  1595. *
  1596. * CPU1 CPU2
  1597. *
  1598. * sys_select receive packet
  1599. * ... ...
  1600. * __add_wait_queue update tp->rcv_nxt
  1601. * ... ...
  1602. * tp->rcv_nxt check sock_def_readable
  1603. * ... {
  1604. * schedule rcu_read_lock();
  1605. * wq = rcu_dereference(sk->sk_wq);
  1606. * if (wq && waitqueue_active(&wq->wait))
  1607. * wake_up_interruptible(&wq->wait)
  1608. * ...
  1609. * }
  1610. *
  1611. * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
  1612. * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
  1613. * could then endup calling schedule and sleep forever if there are no more
  1614. * data on the socket.
  1615. *
  1616. */
  1617. static inline bool wq_has_sleeper(struct socket_wq *wq)
  1618. {
  1619. /* We need to be sure we are in sync with the
  1620. * add_wait_queue modifications to the wait queue.
  1621. *
  1622. * This memory barrier is paired in the sock_poll_wait.
  1623. */
  1624. smp_mb();
  1625. return wq && waitqueue_active(&wq->wait);
  1626. }
  1627. /**
  1628. * sock_poll_wait - place memory barrier behind the poll_wait call.
  1629. * @filp: file
  1630. * @wait_address: socket wait queue
  1631. * @p: poll_table
  1632. *
  1633. * See the comments in the wq_has_sleeper function.
  1634. */
  1635. static inline void sock_poll_wait(struct file *filp,
  1636. wait_queue_head_t *wait_address, poll_table *p)
  1637. {
  1638. if (!poll_does_not_wait(p) && wait_address) {
  1639. poll_wait(filp, wait_address, p);
  1640. /* We need to be sure we are in sync with the
  1641. * socket flags modification.
  1642. *
  1643. * This memory barrier is paired in the wq_has_sleeper.
  1644. */
  1645. smp_mb();
  1646. }
  1647. }
  1648. /*
  1649. * Queue a received datagram if it will fit. Stream and sequenced
  1650. * protocols can't normally use this as they need to fit buffers in
  1651. * and play with them.
  1652. *
  1653. * Inlined as it's very short and called for pretty much every
  1654. * packet ever received.
  1655. */
  1656. static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
  1657. {
  1658. skb_orphan(skb);
  1659. skb->sk = sk;
  1660. skb->destructor = sock_wfree;
  1661. /*
  1662. * We used to take a refcount on sk, but following operation
  1663. * is enough to guarantee sk_free() wont free this sock until
  1664. * all in-flight packets are completed
  1665. */
  1666. atomic_add(skb->truesize, &sk->sk_wmem_alloc);
  1667. }
  1668. static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
  1669. {
  1670. skb_orphan(skb);
  1671. skb->sk = sk;
  1672. skb->destructor = sock_rfree;
  1673. atomic_add(skb->truesize, &sk->sk_rmem_alloc);
  1674. sk_mem_charge(sk, skb->truesize);
  1675. }
  1676. void sk_reset_timer(struct sock *sk, struct timer_list *timer,
  1677. unsigned long expires);
  1678. void sk_stop_timer(struct sock *sk, struct timer_list *timer);
  1679. int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
  1680. int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
  1681. /*
  1682. * Recover an error report and clear atomically
  1683. */
  1684. static inline int sock_error(struct sock *sk)
  1685. {
  1686. int err;
  1687. if (likely(!sk->sk_err))
  1688. return 0;
  1689. err = xchg(&sk->sk_err, 0);
  1690. return -err;
  1691. }
  1692. static inline unsigned long sock_wspace(struct sock *sk)
  1693. {
  1694. int amt = 0;
  1695. if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
  1696. amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
  1697. if (amt < 0)
  1698. amt = 0;
  1699. }
  1700. return amt;
  1701. }
  1702. static inline void sk_wake_async(struct sock *sk, int how, int band)
  1703. {
  1704. if (sock_flag(sk, SOCK_FASYNC))
  1705. sock_wake_async(sk->sk_socket, how, band);
  1706. }
  1707. #define SOCK_MIN_SNDBUF 2048
  1708. /*
  1709. * Since sk_rmem_alloc sums skb->truesize, even a small frame might need
  1710. * sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak
  1711. */
  1712. #define SOCK_MIN_RCVBUF (2048 + sizeof(struct sk_buff))
  1713. static inline void sk_stream_moderate_sndbuf(struct sock *sk)
  1714. {
  1715. if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
  1716. sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
  1717. sk->sk_sndbuf = max(sk->sk_sndbuf, SOCK_MIN_SNDBUF);
  1718. }
  1719. }
  1720. struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
  1721. static inline struct page *sk_stream_alloc_page(struct sock *sk)
  1722. {
  1723. struct page *page = NULL;
  1724. page = alloc_pages(sk->sk_allocation, 0);
  1725. if (!page) {
  1726. sk_enter_memory_pressure(sk);
  1727. sk_stream_moderate_sndbuf(sk);
  1728. }
  1729. return page;
  1730. }
  1731. /*
  1732. * Default write policy as shown to user space via poll/select/SIGIO
  1733. */
  1734. static inline bool sock_writeable(const struct sock *sk)
  1735. {
  1736. return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
  1737. }
  1738. static inline gfp_t gfp_any(void)
  1739. {
  1740. return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
  1741. }
  1742. static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
  1743. {
  1744. return noblock ? 0 : sk->sk_rcvtimeo;
  1745. }
  1746. static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
  1747. {
  1748. return noblock ? 0 : sk->sk_sndtimeo;
  1749. }
  1750. static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
  1751. {
  1752. return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
  1753. }
  1754. /* Alas, with timeout socket operations are not restartable.
  1755. * Compare this to poll().
  1756. */
  1757. static inline int sock_intr_errno(long timeo)
  1758. {
  1759. return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
  1760. }
  1761. void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
  1762. struct sk_buff *skb);
  1763. void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
  1764. struct sk_buff *skb);
  1765. static inline void
  1766. sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
  1767. {
  1768. ktime_t kt = skb->tstamp;
  1769. struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
  1770. /*
  1771. * generate control messages if
  1772. * - receive time stamping in software requested (SOCK_RCVTSTAMP
  1773. * or SOCK_TIMESTAMPING_RX_SOFTWARE)
  1774. * - software time stamp available and wanted
  1775. * (SOCK_TIMESTAMPING_SOFTWARE)
  1776. * - hardware time stamps available and wanted
  1777. * (SOCK_TIMESTAMPING_SYS_HARDWARE or
  1778. * SOCK_TIMESTAMPING_RAW_HARDWARE)
  1779. */
  1780. if (sock_flag(sk, SOCK_RCVTSTAMP) ||
  1781. sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
  1782. (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
  1783. (hwtstamps->hwtstamp.tv64 &&
  1784. sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
  1785. (hwtstamps->syststamp.tv64 &&
  1786. sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
  1787. __sock_recv_timestamp(msg, sk, skb);
  1788. else
  1789. sk->sk_stamp = kt;
  1790. if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
  1791. __sock_recv_wifi_status(msg, sk, skb);
  1792. }
  1793. void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
  1794. struct sk_buff *skb);
  1795. static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
  1796. struct sk_buff *skb)
  1797. {
  1798. #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
  1799. (1UL << SOCK_RCVTSTAMP) | \
  1800. (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
  1801. (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
  1802. (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
  1803. (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
  1804. if (sk->sk_flags & FLAGS_TS_OR_DROPS)
  1805. __sock_recv_ts_and_drops(msg, sk, skb);
  1806. else
  1807. sk->sk_stamp = skb->tstamp;
  1808. }
  1809. /**
  1810. * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
  1811. * @sk: socket sending this packet
  1812. * @tx_flags: filled with instructions for time stamping
  1813. *
  1814. * Currently only depends on SOCK_TIMESTAMPING* flags.
  1815. */
  1816. void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
  1817. /**
  1818. * sk_eat_skb - Release a skb if it is no longer needed
  1819. * @sk: socket to eat this skb from
  1820. * @skb: socket buffer to eat
  1821. * @copied_early: flag indicating whether DMA operations copied this data early
  1822. *
  1823. * This routine must be called with interrupts disabled or with the socket
  1824. * locked so that the sk_buff queue operation is ok.
  1825. */
  1826. #ifdef CONFIG_NET_DMA
  1827. static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
  1828. {
  1829. __skb_unlink(skb, &sk->sk_receive_queue);
  1830. if (!copied_early)
  1831. __kfree_skb(skb);
  1832. else
  1833. __skb_queue_tail(&sk->sk_async_wait_queue, skb);
  1834. }
  1835. #else
  1836. static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
  1837. {
  1838. __skb_unlink(skb, &sk->sk_receive_queue);
  1839. __kfree_skb(skb);
  1840. }
  1841. #endif
  1842. static inline
  1843. struct net *sock_net(const struct sock *sk)
  1844. {
  1845. return read_pnet(&sk->sk_net);
  1846. }
  1847. static inline
  1848. void sock_net_set(struct sock *sk, struct net *net)
  1849. {
  1850. write_pnet(&sk->sk_net, net);
  1851. }
  1852. /*
  1853. * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
  1854. * They should not hold a reference to a namespace in order to allow
  1855. * to stop it.
  1856. * Sockets after sk_change_net should be released using sk_release_kernel
  1857. */
  1858. static inline void sk_change_net(struct sock *sk, struct net *net)
  1859. {
  1860. put_net(sock_net(sk));
  1861. sock_net_set(sk, hold_net(net));
  1862. }
  1863. static inline struct sock *skb_steal_sock(struct sk_buff *skb)
  1864. {
  1865. if (unlikely(skb->sk)) {
  1866. struct sock *sk = skb->sk;
  1867. skb->destructor = NULL;
  1868. skb->sk = NULL;
  1869. return sk;
  1870. }
  1871. return NULL;
  1872. }
  1873. /* This helper checks if a socket is a full socket,
  1874. * ie _not_ a timewait or request socket.
  1875. * TODO: Check for TCPF_NEW_SYN_RECV when that starts to exist.
  1876. */
  1877. static inline bool sk_fullsock(const struct sock *sk)
  1878. {
  1879. return (1 << sk->sk_state) & ~(TCPF_TIME_WAIT);
  1880. }
  1881. void sock_enable_timestamp(struct sock *sk, int flag);
  1882. int sock_get_timestamp(struct sock *, struct timeval __user *);
  1883. int sock_get_timestampns(struct sock *, struct timespec __user *);
  1884. /*
  1885. * Enable debug/info messages
  1886. */
  1887. extern int net_msg_warn;
  1888. #define NETDEBUG(fmt, args...) \
  1889. do { if (net_msg_warn) printk(fmt,##args); } while (0)
  1890. #define LIMIT_NETDEBUG(fmt, args...) \
  1891. do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
  1892. extern __u32 sysctl_wmem_max;
  1893. extern __u32 sysctl_rmem_max;
  1894. void sk_init(void);
  1895. extern int sysctl_optmem_max;
  1896. extern __u32 sysctl_wmem_default;
  1897. extern __u32 sysctl_rmem_default;
  1898. /* SOCKEV Notifier Events */
  1899. #define SOCKEV_SOCKET 0x00
  1900. #define SOCKEV_BIND 0x01
  1901. #define SOCKEV_LISTEN 0x02
  1902. #define SOCKEV_ACCEPT 0x03
  1903. #define SOCKEV_CONNECT 0x04
  1904. #define SOCKEV_SHUTDOWN 0x05
  1905. int sockev_register_notify(struct notifier_block *nb);
  1906. int sockev_unregister_notify(struct notifier_block *nb);
  1907. #endif /* _SOCK_H */