sock.h 64 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287
  1. /*
  2. * INET An implementation of the TCP/IP protocol suite for the LINUX
  3. * operating system. INET is implemented using the BSD Socket
  4. * interface as the means of communication with the user level.
  5. *
  6. * Definitions for the AF_INET socket handler.
  7. *
  8. * Version: @(#)sock.h 1.0.4 05/13/93
  9. *
  10. * Authors: Ross Biro
  11. * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
  12. * Corey Minyard <wf-rch!minyard@relay.EU.net>
  13. * Florian La Roche <flla@stud.uni-sb.de>
  14. *
  15. * Fixes:
  16. * Alan Cox : Volatiles in skbuff pointers. See
  17. * skbuff comments. May be overdone,
  18. * better to prove they can be removed
  19. * than the reverse.
  20. * Alan Cox : Added a zapped field for tcp to note
  21. * a socket is reset and must stay shut up
  22. * Alan Cox : New fields for options
  23. * Pauline Middelink : identd support
  24. * Alan Cox : Eliminate low level recv/recvfrom
  25. * David S. Miller : New socket lookup architecture.
  26. * Steve Whitehouse: Default routines for sock_ops
  27. * Arnaldo C. Melo : removed net_pinfo, tp_pinfo and made
  28. * protinfo be just a void pointer, as the
  29. * protocol specific parts were moved to
  30. * respective headers and ipv4/v6, etc now
  31. * use private slabcaches for its socks
  32. * Pedro Hortas : New flags field for socket options
  33. *
  34. *
  35. * This program is free software; you can redistribute it and/or
  36. * modify it under the terms of the GNU General Public License
  37. * as published by the Free Software Foundation; either version
  38. * 2 of the License, or (at your option) any later version.
  39. */
  40. #ifndef _SOCK_H
  41. #define _SOCK_H
  42. #include <linux/hardirq.h>
  43. #include <linux/kernel.h>
  44. #include <linux/list.h>
  45. #include <linux/list_nulls.h>
  46. #include <linux/timer.h>
  47. #include <linux/cache.h>
  48. #include <linux/bitops.h>
  49. #include <linux/lockdep.h>
  50. #include <linux/netdevice.h>
  51. #include <linux/skbuff.h> /* struct sk_buff */
  52. #include <linux/mm.h>
  53. #include <linux/security.h>
  54. #include <linux/slab.h>
  55. #include <linux/uaccess.h>
  56. #include <linux/memcontrol.h>
  57. #include <linux/res_counter.h>
  58. #include <linux/static_key.h>
  59. #include <linux/aio.h>
  60. #include <linux/sched.h>
  61. #include <linux/filter.h>
  62. #include <linux/rculist_nulls.h>
  63. #include <linux/poll.h>
  64. #include <linux/atomic.h>
  65. #include <net/dst.h>
  66. #include <net/checksum.h>
  67. struct cgroup;
  68. struct cgroup_subsys;
  69. #ifdef CONFIG_NET
  70. int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss);
  71. void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg);
  72. #else
  73. static inline
  74. int mem_cgroup_sockets_init(struct mem_cgroup *memcg, struct cgroup_subsys *ss)
  75. {
  76. return 0;
  77. }
  78. static inline
  79. void mem_cgroup_sockets_destroy(struct mem_cgroup *memcg)
  80. {
  81. }
  82. #endif
  83. /*
  84. * This structure really needs to be cleaned up.
  85. * Most of it is for TCP, and not used by any of
  86. * the other protocols.
  87. */
  88. /* Define this to get the SOCK_DBG debugging facility. */
  89. #define SOCK_DEBUGGING
  90. #ifdef SOCK_DEBUGGING
  91. #define SOCK_DEBUG(sk, msg...) do { if ((sk) && sock_flag((sk), SOCK_DBG)) \
  92. printk(KERN_DEBUG msg); } while (0)
  93. #else
  94. /* Validate arguments and do nothing */
  95. static inline __printf(2, 3)
  96. void SOCK_DEBUG(const struct sock *sk, const char *msg, ...)
  97. {
  98. }
  99. #endif
  100. /* This is the per-socket lock. The spinlock provides a synchronization
  101. * between user contexts and software interrupt processing, whereas the
  102. * mini-semaphore synchronizes multiple users amongst themselves.
  103. */
  104. typedef struct {
  105. spinlock_t slock;
  106. int owned;
  107. wait_queue_head_t wq;
  108. /*
  109. * We express the mutex-alike socket_lock semantics
  110. * to the lock validator by explicitly managing
  111. * the slock as a lock variant (in addition to
  112. * the slock itself):
  113. */
  114. #ifdef CONFIG_DEBUG_LOCK_ALLOC
  115. struct lockdep_map dep_map;
  116. #endif
  117. } socket_lock_t;
  118. struct sock;
  119. struct proto;
  120. struct net;
  121. typedef __u32 __bitwise __portpair;
  122. typedef __u64 __bitwise __addrpair;
  123. /**
  124. * struct sock_common - minimal network layer representation of sockets
  125. * @skc_daddr: Foreign IPv4 addr
  126. * @skc_rcv_saddr: Bound local IPv4 addr
  127. * @skc_hash: hash value used with various protocol lookup tables
  128. * @skc_u16hashes: two u16 hash values used by UDP lookup tables
  129. * @skc_dport: placeholder for inet_dport/tw_dport
  130. * @skc_num: placeholder for inet_num/tw_num
  131. * @skc_family: network address family
  132. * @skc_state: Connection state
  133. * @skc_reuse: %SO_REUSEADDR setting
  134. * @skc_reuseport: %SO_REUSEPORT setting
  135. * @skc_bound_dev_if: bound device index if != 0
  136. * @skc_bind_node: bind hash linkage for various protocol lookup tables
  137. * @skc_portaddr_node: second hash linkage for UDP/UDP-Lite protocol
  138. * @skc_prot: protocol handlers inside a network family
  139. * @skc_net: reference to the network namespace of this socket
  140. * @skc_node: main hash linkage for various protocol lookup tables
  141. * @skc_nulls_node: main hash linkage for TCP/UDP/UDP-Lite protocol
  142. * @skc_tx_queue_mapping: tx queue number for this connection
  143. * @skc_refcnt: reference count
  144. *
  145. * This is the minimal network layer representation of sockets, the header
  146. * for struct sock and struct inet_timewait_sock.
  147. */
  148. struct sock_common {
  149. /* skc_daddr and skc_rcv_saddr must be grouped on a 8 bytes aligned
  150. * address on 64bit arches : cf INET_MATCH()
  151. */
  152. union {
  153. __addrpair skc_addrpair;
  154. struct {
  155. __be32 skc_daddr;
  156. __be32 skc_rcv_saddr;
  157. };
  158. };
  159. union {
  160. unsigned int skc_hash;
  161. __u16 skc_u16hashes[2];
  162. };
  163. /* skc_dport && skc_num must be grouped as well */
  164. union {
  165. __portpair skc_portpair;
  166. struct {
  167. __be16 skc_dport;
  168. __u16 skc_num;
  169. };
  170. };
  171. unsigned short skc_family;
  172. volatile unsigned char skc_state;
  173. unsigned char skc_reuse:4;
  174. unsigned char skc_reuseport:4;
  175. int skc_bound_dev_if;
  176. union {
  177. struct hlist_node skc_bind_node;
  178. struct hlist_nulls_node skc_portaddr_node;
  179. };
  180. struct proto *skc_prot;
  181. #ifdef CONFIG_NET_NS
  182. struct net *skc_net;
  183. #endif
  184. #if IS_ENABLED(CONFIG_IPV6)
  185. struct in6_addr skc_v6_daddr;
  186. struct in6_addr skc_v6_rcv_saddr;
  187. #endif
  188. /*
  189. * fields between dontcopy_begin/dontcopy_end
  190. * are not copied in sock_copy()
  191. */
  192. /* private: */
  193. int skc_dontcopy_begin[0];
  194. /* public: */
  195. union {
  196. struct hlist_node skc_node;
  197. struct hlist_nulls_node skc_nulls_node;
  198. };
  199. int skc_tx_queue_mapping;
  200. atomic_t skc_refcnt;
  201. /* private: */
  202. int skc_dontcopy_end[0];
  203. /* public: */
  204. };
  205. struct cg_proto;
  206. /**
  207. * struct sock - network layer representation of sockets
  208. * @__sk_common: shared layout with inet_timewait_sock
  209. * @sk_shutdown: mask of %SEND_SHUTDOWN and/or %RCV_SHUTDOWN
  210. * @sk_userlocks: %SO_SNDBUF and %SO_RCVBUF settings
  211. * @sk_lock: synchronizer
  212. * @sk_rcvbuf: size of receive buffer in bytes
  213. * @sk_wq: sock wait queue and async head
  214. * @sk_rx_dst: receive input route used by early demux
  215. * @sk_dst_cache: destination cache
  216. * @sk_dst_lock: destination cache lock
  217. * @sk_policy: flow policy
  218. * @sk_receive_queue: incoming packets
  219. * @sk_wmem_alloc: transmit queue bytes committed
  220. * @sk_write_queue: Packet sending queue
  221. * @sk_async_wait_queue: DMA copied packets
  222. * @sk_omem_alloc: "o" is "option" or "other"
  223. * @sk_wmem_queued: persistent queue size
  224. * @sk_forward_alloc: space allocated forward
  225. * @sk_napi_id: id of the last napi context to receive data for sk
  226. * @sk_ll_usec: usecs to busypoll when there is no data
  227. * @sk_allocation: allocation mode
  228. * @sk_pacing_rate: Pacing rate (if supported by transport/packet scheduler)
  229. * @sk_max_pacing_rate: Maximum pacing rate (%SO_MAX_PACING_RATE)
  230. * @sk_sndbuf: size of send buffer in bytes
  231. * @sk_flags: %SO_LINGER (l_onoff), %SO_BROADCAST, %SO_KEEPALIVE,
  232. * %SO_OOBINLINE settings, %SO_TIMESTAMPING settings
  233. * @sk_no_check: %SO_NO_CHECK setting, whether or not checkup packets
  234. * @sk_route_caps: route capabilities (e.g. %NETIF_F_TSO)
  235. * @sk_route_nocaps: forbidden route capabilities (e.g NETIF_F_GSO_MASK)
  236. * @sk_gso_type: GSO type (e.g. %SKB_GSO_TCPV4)
  237. * @sk_gso_max_size: Maximum GSO segment size to build
  238. * @sk_gso_max_segs: Maximum number of GSO segments
  239. * @sk_lingertime: %SO_LINGER l_linger setting
  240. * @sk_backlog: always used with the per-socket spinlock held
  241. * @sk_callback_lock: used with the callbacks in the end of this struct
  242. * @sk_error_queue: rarely used
  243. * @sk_prot_creator: sk_prot of original sock creator (see ipv6_setsockopt,
  244. * IPV6_ADDRFORM for instance)
  245. * @sk_err: last error
  246. * @sk_err_soft: errors that don't cause failure but are the cause of a
  247. * persistent failure not just 'timed out'
  248. * @sk_drops: raw/udp drops counter
  249. * @sk_ack_backlog: current listen backlog
  250. * @sk_max_ack_backlog: listen backlog set in listen()
  251. * @sk_priority: %SO_PRIORITY setting
  252. * @sk_cgrp_prioidx: socket group's priority map index
  253. * @sk_type: socket type (%SOCK_STREAM, etc)
  254. * @sk_protocol: which protocol this socket belongs in this network family
  255. * @sk_peer_pid: &struct pid for this socket's peer
  256. * @sk_peer_cred: %SO_PEERCRED setting
  257. * @sk_rcvlowat: %SO_RCVLOWAT setting
  258. * @sk_rcvtimeo: %SO_RCVTIMEO setting
  259. * @sk_sndtimeo: %SO_SNDTIMEO setting
  260. * @sk_rxhash: flow hash received from netif layer
  261. * @sk_filter: socket filtering instructions
  262. * @sk_protinfo: private area, net family specific, when not using slab
  263. * @sk_timer: sock cleanup timer
  264. * @sk_stamp: time stamp of last packet received
  265. * @sk_socket: Identd and reporting IO signals
  266. * @sk_user_data: RPC layer private data
  267. * @sk_frag: cached page frag
  268. * @sk_peek_off: current peek_offset value
  269. * @sk_send_head: front of stuff to transmit
  270. * @sk_security: used by security modules
  271. * @sk_mark: generic packet mark
  272. * @sk_classid: this socket's cgroup classid
  273. * @sk_cgrp: this socket's cgroup-specific proto data
  274. * @sk_write_pending: a write to stream socket waits to start
  275. * @sk_state_change: callback to indicate change in the state of the sock
  276. * @sk_data_ready: callback to indicate there is data to be processed
  277. * @sk_write_space: callback to indicate there is bf sending space available
  278. * @sk_error_report: callback to indicate errors (e.g. %MSG_ERRQUEUE)
  279. * @sk_backlog_rcv: callback to process the backlog
  280. * @sk_destruct: called at sock freeing time, i.e. when all refcnt == 0
  281. */
  282. struct sock {
  283. /*
  284. * Now struct inet_timewait_sock also uses sock_common, so please just
  285. * don't add nothing before this first member (__sk_common) --acme
  286. */
  287. struct sock_common __sk_common;
  288. #define sk_node __sk_common.skc_node
  289. #define sk_nulls_node __sk_common.skc_nulls_node
  290. #define sk_refcnt __sk_common.skc_refcnt
  291. #define sk_tx_queue_mapping __sk_common.skc_tx_queue_mapping
  292. #define sk_dontcopy_begin __sk_common.skc_dontcopy_begin
  293. #define sk_dontcopy_end __sk_common.skc_dontcopy_end
  294. #define sk_hash __sk_common.skc_hash
  295. #define sk_portpair __sk_common.skc_portpair
  296. #define sk_num __sk_common.skc_num
  297. #define sk_dport __sk_common.skc_dport
  298. #define sk_addrpair __sk_common.skc_addrpair
  299. #define sk_daddr __sk_common.skc_daddr
  300. #define sk_rcv_saddr __sk_common.skc_rcv_saddr
  301. #define sk_family __sk_common.skc_family
  302. #define sk_state __sk_common.skc_state
  303. #define sk_reuse __sk_common.skc_reuse
  304. #define sk_reuseport __sk_common.skc_reuseport
  305. #define sk_bound_dev_if __sk_common.skc_bound_dev_if
  306. #define sk_bind_node __sk_common.skc_bind_node
  307. #define sk_prot __sk_common.skc_prot
  308. #define sk_net __sk_common.skc_net
  309. #define sk_v6_daddr __sk_common.skc_v6_daddr
  310. #define sk_v6_rcv_saddr __sk_common.skc_v6_rcv_saddr
  311. socket_lock_t sk_lock;
  312. struct sk_buff_head sk_receive_queue;
  313. /*
  314. * The backlog queue is special, it is always used with
  315. * the per-socket spinlock held and requires low latency
  316. * access. Therefore we special case it's implementation.
  317. * Note : rmem_alloc is in this structure to fill a hole
  318. * on 64bit arches, not because its logically part of
  319. * backlog.
  320. */
  321. struct {
  322. atomic_t rmem_alloc;
  323. int len;
  324. struct sk_buff *head;
  325. struct sk_buff *tail;
  326. } sk_backlog;
  327. #define sk_rmem_alloc sk_backlog.rmem_alloc
  328. int sk_forward_alloc;
  329. #ifdef CONFIG_RPS
  330. __u32 sk_rxhash;
  331. #endif
  332. #ifdef CONFIG_NET_RX_BUSY_POLL
  333. unsigned int sk_napi_id;
  334. unsigned int sk_ll_usec;
  335. #endif
  336. atomic_t sk_drops;
  337. int sk_rcvbuf;
  338. struct sk_filter __rcu *sk_filter;
  339. struct socket_wq __rcu *sk_wq;
  340. #ifdef CONFIG_NET_DMA
  341. struct sk_buff_head sk_async_wait_queue;
  342. #endif
  343. #ifdef CONFIG_XFRM
  344. struct xfrm_policy *sk_policy[2];
  345. #endif
  346. unsigned long sk_flags;
  347. struct dst_entry *sk_rx_dst;
  348. struct dst_entry __rcu *sk_dst_cache;
  349. spinlock_t sk_dst_lock;
  350. atomic_t sk_wmem_alloc;
  351. atomic_t sk_omem_alloc;
  352. int sk_sndbuf;
  353. struct sk_buff_head sk_write_queue;
  354. kmemcheck_bitfield_begin(flags);
  355. unsigned int sk_shutdown : 2,
  356. sk_no_check : 2,
  357. sk_userlocks : 4,
  358. sk_protocol : 8,
  359. sk_type : 16;
  360. kmemcheck_bitfield_end(flags);
  361. int sk_wmem_queued;
  362. gfp_t sk_allocation;
  363. u32 sk_pacing_rate; /* bytes per second */
  364. u32 sk_max_pacing_rate;
  365. netdev_features_t sk_route_caps;
  366. netdev_features_t sk_route_nocaps;
  367. int sk_gso_type;
  368. unsigned int sk_gso_max_size;
  369. u16 sk_gso_max_segs;
  370. int sk_rcvlowat;
  371. unsigned long sk_lingertime;
  372. struct sk_buff_head sk_error_queue;
  373. struct proto *sk_prot_creator;
  374. rwlock_t sk_callback_lock;
  375. int sk_err,
  376. sk_err_soft;
  377. unsigned short sk_ack_backlog;
  378. unsigned short sk_max_ack_backlog;
  379. __u32 sk_priority;
  380. #if IS_ENABLED(CONFIG_NETPRIO_CGROUP)
  381. __u32 sk_cgrp_prioidx;
  382. #endif
  383. struct pid *sk_peer_pid;
  384. const struct cred *sk_peer_cred;
  385. long sk_rcvtimeo;
  386. long sk_sndtimeo;
  387. void *sk_protinfo;
  388. struct timer_list sk_timer;
  389. ktime_t sk_stamp;
  390. struct socket *sk_socket;
  391. void *sk_user_data;
  392. struct page_frag sk_frag;
  393. struct sk_buff *sk_send_head;
  394. __s32 sk_peek_off;
  395. int sk_write_pending;
  396. #ifdef CONFIG_SECURITY
  397. void *sk_security;
  398. #endif
  399. __u32 sk_mark;
  400. u32 sk_classid;
  401. struct cg_proto *sk_cgrp;
  402. void (*sk_state_change)(struct sock *sk);
  403. void (*sk_data_ready)(struct sock *sk, int bytes);
  404. void (*sk_write_space)(struct sock *sk);
  405. void (*sk_error_report)(struct sock *sk);
  406. int (*sk_backlog_rcv)(struct sock *sk,
  407. struct sk_buff *skb);
  408. void (*sk_destruct)(struct sock *sk);
  409. };
  410. #define __sk_user_data(sk) ((*((void __rcu **)&(sk)->sk_user_data)))
  411. #define rcu_dereference_sk_user_data(sk) rcu_dereference(__sk_user_data((sk)))
  412. #define rcu_assign_sk_user_data(sk, ptr) rcu_assign_pointer(__sk_user_data((sk)), ptr)
  413. /*
  414. * SK_CAN_REUSE and SK_NO_REUSE on a socket mean that the socket is OK
  415. * or not whether his port will be reused by someone else. SK_FORCE_REUSE
  416. * on a socket means that the socket will reuse everybody else's port
  417. * without looking at the other's sk_reuse value.
  418. */
  419. #define SK_NO_REUSE 0
  420. #define SK_CAN_REUSE 1
  421. #define SK_FORCE_REUSE 2
  422. static inline int sk_peek_offset(struct sock *sk, int flags)
  423. {
  424. if ((flags & MSG_PEEK) && (sk->sk_peek_off >= 0))
  425. return sk->sk_peek_off;
  426. else
  427. return 0;
  428. }
  429. static inline void sk_peek_offset_bwd(struct sock *sk, int val)
  430. {
  431. if (sk->sk_peek_off >= 0) {
  432. if (sk->sk_peek_off >= val)
  433. sk->sk_peek_off -= val;
  434. else
  435. sk->sk_peek_off = 0;
  436. }
  437. }
  438. static inline void sk_peek_offset_fwd(struct sock *sk, int val)
  439. {
  440. if (sk->sk_peek_off >= 0)
  441. sk->sk_peek_off += val;
  442. }
  443. /*
  444. * Hashed lists helper routines
  445. */
  446. static inline struct sock *sk_entry(const struct hlist_node *node)
  447. {
  448. return hlist_entry(node, struct sock, sk_node);
  449. }
  450. static inline struct sock *__sk_head(const struct hlist_head *head)
  451. {
  452. return hlist_entry(head->first, struct sock, sk_node);
  453. }
  454. static inline struct sock *sk_head(const struct hlist_head *head)
  455. {
  456. return hlist_empty(head) ? NULL : __sk_head(head);
  457. }
  458. static inline struct sock *__sk_nulls_head(const struct hlist_nulls_head *head)
  459. {
  460. return hlist_nulls_entry(head->first, struct sock, sk_nulls_node);
  461. }
  462. static inline struct sock *sk_nulls_head(const struct hlist_nulls_head *head)
  463. {
  464. return hlist_nulls_empty(head) ? NULL : __sk_nulls_head(head);
  465. }
  466. static inline struct sock *sk_next(const struct sock *sk)
  467. {
  468. return sk->sk_node.next ?
  469. hlist_entry(sk->sk_node.next, struct sock, sk_node) : NULL;
  470. }
  471. static inline struct sock *sk_nulls_next(const struct sock *sk)
  472. {
  473. return (!is_a_nulls(sk->sk_nulls_node.next)) ?
  474. hlist_nulls_entry(sk->sk_nulls_node.next,
  475. struct sock, sk_nulls_node) :
  476. NULL;
  477. }
  478. static inline bool sk_unhashed(const struct sock *sk)
  479. {
  480. return hlist_unhashed(&sk->sk_node);
  481. }
  482. static inline bool sk_hashed(const struct sock *sk)
  483. {
  484. return !sk_unhashed(sk);
  485. }
  486. static inline void sk_node_init(struct hlist_node *node)
  487. {
  488. node->pprev = NULL;
  489. }
  490. static inline void sk_nulls_node_init(struct hlist_nulls_node *node)
  491. {
  492. node->pprev = NULL;
  493. }
  494. static inline void __sk_del_node(struct sock *sk)
  495. {
  496. __hlist_del(&sk->sk_node);
  497. }
  498. /* NB: equivalent to hlist_del_init_rcu */
  499. static inline bool __sk_del_node_init(struct sock *sk)
  500. {
  501. if (sk_hashed(sk)) {
  502. __sk_del_node(sk);
  503. sk_node_init(&sk->sk_node);
  504. return true;
  505. }
  506. return false;
  507. }
  508. /* Grab socket reference count. This operation is valid only
  509. when sk is ALREADY grabbed f.e. it is found in hash table
  510. or a list and the lookup is made under lock preventing hash table
  511. modifications.
  512. */
  513. static inline void sock_hold(struct sock *sk)
  514. {
  515. atomic_inc(&sk->sk_refcnt);
  516. }
  517. /* Ungrab socket in the context, which assumes that socket refcnt
  518. cannot hit zero, f.e. it is true in context of any socketcall.
  519. */
  520. static inline void __sock_put(struct sock *sk)
  521. {
  522. atomic_dec(&sk->sk_refcnt);
  523. }
  524. static inline bool sk_del_node_init(struct sock *sk)
  525. {
  526. bool rc = __sk_del_node_init(sk);
  527. if (rc) {
  528. /* paranoid for a while -acme */
  529. WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
  530. __sock_put(sk);
  531. }
  532. return rc;
  533. }
  534. #define sk_del_node_init_rcu(sk) sk_del_node_init(sk)
  535. static inline bool __sk_nulls_del_node_init_rcu(struct sock *sk)
  536. {
  537. if (sk_hashed(sk)) {
  538. hlist_nulls_del_init_rcu(&sk->sk_nulls_node);
  539. return true;
  540. }
  541. return false;
  542. }
  543. static inline bool sk_nulls_del_node_init_rcu(struct sock *sk)
  544. {
  545. bool rc = __sk_nulls_del_node_init_rcu(sk);
  546. if (rc) {
  547. /* paranoid for a while -acme */
  548. WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
  549. __sock_put(sk);
  550. }
  551. return rc;
  552. }
  553. static inline void __sk_add_node(struct sock *sk, struct hlist_head *list)
  554. {
  555. hlist_add_head(&sk->sk_node, list);
  556. }
  557. static inline void sk_add_node(struct sock *sk, struct hlist_head *list)
  558. {
  559. sock_hold(sk);
  560. __sk_add_node(sk, list);
  561. }
  562. static inline void sk_add_node_rcu(struct sock *sk, struct hlist_head *list)
  563. {
  564. sock_hold(sk);
  565. hlist_add_head_rcu(&sk->sk_node, list);
  566. }
  567. static inline void __sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
  568. {
  569. hlist_nulls_add_head_rcu(&sk->sk_nulls_node, list);
  570. }
  571. static inline void sk_nulls_add_node_rcu(struct sock *sk, struct hlist_nulls_head *list)
  572. {
  573. sock_hold(sk);
  574. __sk_nulls_add_node_rcu(sk, list);
  575. }
  576. static inline void __sk_del_bind_node(struct sock *sk)
  577. {
  578. __hlist_del(&sk->sk_bind_node);
  579. }
  580. static inline void sk_add_bind_node(struct sock *sk,
  581. struct hlist_head *list)
  582. {
  583. hlist_add_head(&sk->sk_bind_node, list);
  584. }
  585. #define sk_for_each(__sk, list) \
  586. hlist_for_each_entry(__sk, list, sk_node)
  587. #define sk_for_each_rcu(__sk, list) \
  588. hlist_for_each_entry_rcu(__sk, list, sk_node)
  589. #define sk_nulls_for_each(__sk, node, list) \
  590. hlist_nulls_for_each_entry(__sk, node, list, sk_nulls_node)
  591. #define sk_nulls_for_each_rcu(__sk, node, list) \
  592. hlist_nulls_for_each_entry_rcu(__sk, node, list, sk_nulls_node)
  593. #define sk_for_each_from(__sk) \
  594. hlist_for_each_entry_from(__sk, sk_node)
  595. #define sk_nulls_for_each_from(__sk, node) \
  596. if (__sk && ({ node = &(__sk)->sk_nulls_node; 1; })) \
  597. hlist_nulls_for_each_entry_from(__sk, node, sk_nulls_node)
  598. #define sk_for_each_safe(__sk, tmp, list) \
  599. hlist_for_each_entry_safe(__sk, tmp, list, sk_node)
  600. #define sk_for_each_bound(__sk, list) \
  601. hlist_for_each_entry(__sk, list, sk_bind_node)
  602. static inline struct user_namespace *sk_user_ns(struct sock *sk)
  603. {
  604. /* Careful only use this in a context where these parameters
  605. * can not change and must all be valid, such as recvmsg from
  606. * userspace.
  607. */
  608. return sk->sk_socket->file->f_cred->user_ns;
  609. }
  610. /* Sock flags */
  611. enum sock_flags {
  612. SOCK_DEAD,
  613. SOCK_DONE,
  614. SOCK_URGINLINE,
  615. SOCK_KEEPOPEN,
  616. SOCK_LINGER,
  617. SOCK_DESTROY,
  618. SOCK_BROADCAST,
  619. SOCK_TIMESTAMP,
  620. SOCK_ZAPPED,
  621. SOCK_USE_WRITE_QUEUE, /* whether to call sk->sk_write_space in sock_wfree */
  622. SOCK_DBG, /* %SO_DEBUG setting */
  623. SOCK_RCVTSTAMP, /* %SO_TIMESTAMP setting */
  624. SOCK_RCVTSTAMPNS, /* %SO_TIMESTAMPNS setting */
  625. SOCK_LOCALROUTE, /* route locally only, %SO_DONTROUTE setting */
  626. SOCK_QUEUE_SHRUNK, /* write queue has been shrunk recently */
  627. SOCK_MEMALLOC, /* VM depends on this socket for swapping */
  628. SOCK_TIMESTAMPING_TX_HARDWARE, /* %SOF_TIMESTAMPING_TX_HARDWARE */
  629. SOCK_TIMESTAMPING_TX_SOFTWARE, /* %SOF_TIMESTAMPING_TX_SOFTWARE */
  630. SOCK_TIMESTAMPING_RX_HARDWARE, /* %SOF_TIMESTAMPING_RX_HARDWARE */
  631. SOCK_TIMESTAMPING_RX_SOFTWARE, /* %SOF_TIMESTAMPING_RX_SOFTWARE */
  632. SOCK_TIMESTAMPING_SOFTWARE, /* %SOF_TIMESTAMPING_SOFTWARE */
  633. SOCK_TIMESTAMPING_RAW_HARDWARE, /* %SOF_TIMESTAMPING_RAW_HARDWARE */
  634. SOCK_TIMESTAMPING_SYS_HARDWARE, /* %SOF_TIMESTAMPING_SYS_HARDWARE */
  635. SOCK_FASYNC, /* fasync() active */
  636. SOCK_RXQ_OVFL,
  637. SOCK_ZEROCOPY, /* buffers from userspace */
  638. SOCK_WIFI_STATUS, /* push wifi status to userspace */
  639. SOCK_NOFCS, /* Tell NIC not to do the Ethernet FCS.
  640. * Will use last 4 bytes of packet sent from
  641. * user-space instead.
  642. */
  643. SOCK_FILTER_LOCKED, /* Filter cannot be changed anymore */
  644. SOCK_SELECT_ERR_QUEUE, /* Wake select on error queue */
  645. };
  646. static inline void sock_copy_flags(struct sock *nsk, struct sock *osk)
  647. {
  648. nsk->sk_flags = osk->sk_flags;
  649. }
  650. static inline void sock_set_flag(struct sock *sk, enum sock_flags flag)
  651. {
  652. __set_bit(flag, &sk->sk_flags);
  653. }
  654. static inline void sock_reset_flag(struct sock *sk, enum sock_flags flag)
  655. {
  656. __clear_bit(flag, &sk->sk_flags);
  657. }
  658. static inline bool sock_flag(const struct sock *sk, enum sock_flags flag)
  659. {
  660. return test_bit(flag, &sk->sk_flags);
  661. }
  662. #ifdef CONFIG_NET
  663. extern struct static_key memalloc_socks;
  664. static inline int sk_memalloc_socks(void)
  665. {
  666. return static_key_false(&memalloc_socks);
  667. }
  668. #else
  669. static inline int sk_memalloc_socks(void)
  670. {
  671. return 0;
  672. }
  673. #endif
  674. static inline gfp_t sk_gfp_atomic(struct sock *sk, gfp_t gfp_mask)
  675. {
  676. return GFP_ATOMIC | (sk->sk_allocation & __GFP_MEMALLOC);
  677. }
  678. static inline void sk_acceptq_removed(struct sock *sk)
  679. {
  680. sk->sk_ack_backlog--;
  681. }
  682. static inline void sk_acceptq_added(struct sock *sk)
  683. {
  684. sk->sk_ack_backlog++;
  685. }
  686. static inline bool sk_acceptq_is_full(const struct sock *sk)
  687. {
  688. return sk->sk_ack_backlog > sk->sk_max_ack_backlog;
  689. }
  690. /*
  691. * Compute minimal free write space needed to queue new packets.
  692. */
  693. static inline int sk_stream_min_wspace(const struct sock *sk)
  694. {
  695. return sk->sk_wmem_queued >> 1;
  696. }
  697. static inline int sk_stream_wspace(const struct sock *sk)
  698. {
  699. return sk->sk_sndbuf - sk->sk_wmem_queued;
  700. }
  701. void sk_stream_write_space(struct sock *sk);
  702. /* OOB backlog add */
  703. static inline void __sk_add_backlog(struct sock *sk, struct sk_buff *skb)
  704. {
  705. /* dont let skb dst not refcounted, we are going to leave rcu lock */
  706. skb_dst_force(skb);
  707. if (!sk->sk_backlog.tail)
  708. sk->sk_backlog.head = skb;
  709. else
  710. sk->sk_backlog.tail->next = skb;
  711. sk->sk_backlog.tail = skb;
  712. skb->next = NULL;
  713. }
  714. /*
  715. * Take into account size of receive queue and backlog queue
  716. * Do not take into account this skb truesize,
  717. * to allow even a single big packet to come.
  718. */
  719. static inline bool sk_rcvqueues_full(const struct sock *sk, const struct sk_buff *skb,
  720. unsigned int limit)
  721. {
  722. unsigned int qsize = sk->sk_backlog.len + atomic_read(&sk->sk_rmem_alloc);
  723. return qsize > limit;
  724. }
  725. /* The per-socket spinlock must be held here. */
  726. static inline __must_check int sk_add_backlog(struct sock *sk, struct sk_buff *skb,
  727. unsigned int limit)
  728. {
  729. if (sk_rcvqueues_full(sk, skb, limit))
  730. return -ENOBUFS;
  731. __sk_add_backlog(sk, skb);
  732. sk->sk_backlog.len += skb->truesize;
  733. return 0;
  734. }
  735. int __sk_backlog_rcv(struct sock *sk, struct sk_buff *skb);
  736. static inline int sk_backlog_rcv(struct sock *sk, struct sk_buff *skb)
  737. {
  738. if (sk_memalloc_socks() && skb_pfmemalloc(skb))
  739. return __sk_backlog_rcv(sk, skb);
  740. return sk->sk_backlog_rcv(sk, skb);
  741. }
  742. static inline void sock_rps_record_flow(const struct sock *sk)
  743. {
  744. #ifdef CONFIG_RPS
  745. struct rps_sock_flow_table *sock_flow_table;
  746. rcu_read_lock();
  747. sock_flow_table = rcu_dereference(rps_sock_flow_table);
  748. rps_record_sock_flow(sock_flow_table, sk->sk_rxhash);
  749. rcu_read_unlock();
  750. #endif
  751. }
  752. static inline void sock_rps_reset_flow(const struct sock *sk)
  753. {
  754. #ifdef CONFIG_RPS
  755. struct rps_sock_flow_table *sock_flow_table;
  756. rcu_read_lock();
  757. sock_flow_table = rcu_dereference(rps_sock_flow_table);
  758. rps_reset_sock_flow(sock_flow_table, sk->sk_rxhash);
  759. rcu_read_unlock();
  760. #endif
  761. }
  762. static inline void sock_rps_save_rxhash(struct sock *sk,
  763. const struct sk_buff *skb)
  764. {
  765. #ifdef CONFIG_RPS
  766. if (unlikely(sk->sk_rxhash != skb->rxhash)) {
  767. sock_rps_reset_flow(sk);
  768. sk->sk_rxhash = skb->rxhash;
  769. }
  770. #endif
  771. }
  772. static inline void sock_rps_reset_rxhash(struct sock *sk)
  773. {
  774. #ifdef CONFIG_RPS
  775. sock_rps_reset_flow(sk);
  776. sk->sk_rxhash = 0;
  777. #endif
  778. }
  779. #define sk_wait_event(__sk, __timeo, __condition) \
  780. ({ int __rc; \
  781. release_sock(__sk); \
  782. __rc = __condition; \
  783. if (!__rc) { \
  784. *(__timeo) = schedule_timeout(*(__timeo)); \
  785. } \
  786. lock_sock(__sk); \
  787. __rc = __condition; \
  788. __rc; \
  789. })
  790. int sk_stream_wait_connect(struct sock *sk, long *timeo_p);
  791. int sk_stream_wait_memory(struct sock *sk, long *timeo_p);
  792. void sk_stream_wait_close(struct sock *sk, long timeo_p);
  793. int sk_stream_error(struct sock *sk, int flags, int err);
  794. void sk_stream_kill_queues(struct sock *sk);
  795. void sk_set_memalloc(struct sock *sk);
  796. void sk_clear_memalloc(struct sock *sk);
  797. int sk_wait_data(struct sock *sk, long *timeo);
  798. struct request_sock_ops;
  799. struct timewait_sock_ops;
  800. struct inet_hashinfo;
  801. struct raw_hashinfo;
  802. struct module;
  803. /*
  804. * caches using SLAB_DESTROY_BY_RCU should let .next pointer from nulls nodes
  805. * un-modified. Special care is taken when initializing object to zero.
  806. */
  807. static inline void sk_prot_clear_nulls(struct sock *sk, int size)
  808. {
  809. if (offsetof(struct sock, sk_node.next) != 0)
  810. memset(sk, 0, offsetof(struct sock, sk_node.next));
  811. memset(&sk->sk_node.pprev, 0,
  812. size - offsetof(struct sock, sk_node.pprev));
  813. }
  814. /* Networking protocol blocks we attach to sockets.
  815. * socket layer -> transport layer interface
  816. * transport -> network interface is defined by struct inet_proto
  817. */
  818. struct proto {
  819. void (*close)(struct sock *sk,
  820. long timeout);
  821. int (*connect)(struct sock *sk,
  822. struct sockaddr *uaddr,
  823. int addr_len);
  824. int (*disconnect)(struct sock *sk, int flags);
  825. struct sock * (*accept)(struct sock *sk, int flags, int *err);
  826. int (*ioctl)(struct sock *sk, int cmd,
  827. unsigned long arg);
  828. int (*init)(struct sock *sk);
  829. void (*destroy)(struct sock *sk);
  830. void (*shutdown)(struct sock *sk, int how);
  831. int (*setsockopt)(struct sock *sk, int level,
  832. int optname, char __user *optval,
  833. unsigned int optlen);
  834. int (*getsockopt)(struct sock *sk, int level,
  835. int optname, char __user *optval,
  836. int __user *option);
  837. #ifdef CONFIG_COMPAT
  838. int (*compat_setsockopt)(struct sock *sk,
  839. int level,
  840. int optname, char __user *optval,
  841. unsigned int optlen);
  842. int (*compat_getsockopt)(struct sock *sk,
  843. int level,
  844. int optname, char __user *optval,
  845. int __user *option);
  846. int (*compat_ioctl)(struct sock *sk,
  847. unsigned int cmd, unsigned long arg);
  848. #endif
  849. int (*sendmsg)(struct kiocb *iocb, struct sock *sk,
  850. struct msghdr *msg, size_t len);
  851. int (*recvmsg)(struct kiocb *iocb, struct sock *sk,
  852. struct msghdr *msg,
  853. size_t len, int noblock, int flags,
  854. int *addr_len);
  855. int (*sendpage)(struct sock *sk, struct page *page,
  856. int offset, size_t size, int flags);
  857. int (*bind)(struct sock *sk,
  858. struct sockaddr *uaddr, int addr_len);
  859. int (*backlog_rcv) (struct sock *sk,
  860. struct sk_buff *skb);
  861. void (*release_cb)(struct sock *sk);
  862. void (*mtu_reduced)(struct sock *sk);
  863. /* Keeping track of sk's, looking them up, and port selection methods. */
  864. void (*hash)(struct sock *sk);
  865. void (*unhash)(struct sock *sk);
  866. void (*rehash)(struct sock *sk);
  867. int (*get_port)(struct sock *sk, unsigned short snum);
  868. void (*clear_sk)(struct sock *sk, int size);
  869. /* Keeping track of sockets in use */
  870. #ifdef CONFIG_PROC_FS
  871. unsigned int inuse_idx;
  872. #endif
  873. bool (*stream_memory_free)(const struct sock *sk);
  874. /* Memory pressure */
  875. void (*enter_memory_pressure)(struct sock *sk);
  876. atomic_long_t *memory_allocated; /* Current allocated memory. */
  877. struct percpu_counter *sockets_allocated; /* Current number of sockets. */
  878. /*
  879. * Pressure flag: try to collapse.
  880. * Technical note: it is used by multiple contexts non atomically.
  881. * All the __sk_mem_schedule() is of this nature: accounting
  882. * is strict, actions are advisory and have some latency.
  883. */
  884. int *memory_pressure;
  885. long *sysctl_mem;
  886. int *sysctl_wmem;
  887. int *sysctl_rmem;
  888. int max_header;
  889. bool no_autobind;
  890. struct kmem_cache *slab;
  891. unsigned int obj_size;
  892. int slab_flags;
  893. struct percpu_counter *orphan_count;
  894. struct request_sock_ops *rsk_prot;
  895. struct timewait_sock_ops *twsk_prot;
  896. union {
  897. struct inet_hashinfo *hashinfo;
  898. struct udp_table *udp_table;
  899. struct raw_hashinfo *raw_hash;
  900. } h;
  901. struct module *owner;
  902. char name[32];
  903. struct list_head node;
  904. #ifdef SOCK_REFCNT_DEBUG
  905. atomic_t socks;
  906. #endif
  907. #ifdef CONFIG_MEMCG_KMEM
  908. /*
  909. * cgroup specific init/deinit functions. Called once for all
  910. * protocols that implement it, from cgroups populate function.
  911. * This function has to setup any files the protocol want to
  912. * appear in the kmem cgroup filesystem.
  913. */
  914. int (*init_cgroup)(struct mem_cgroup *memcg,
  915. struct cgroup_subsys *ss);
  916. void (*destroy_cgroup)(struct mem_cgroup *memcg);
  917. struct cg_proto *(*proto_cgroup)(struct mem_cgroup *memcg);
  918. #endif
  919. };
  920. /*
  921. * Bits in struct cg_proto.flags
  922. */
  923. enum cg_proto_flags {
  924. /* Currently active and new sockets should be assigned to cgroups */
  925. MEMCG_SOCK_ACTIVE,
  926. /* It was ever activated; we must disarm static keys on destruction */
  927. MEMCG_SOCK_ACTIVATED,
  928. };
  929. struct cg_proto {
  930. void (*enter_memory_pressure)(struct sock *sk);
  931. struct res_counter *memory_allocated; /* Current allocated memory. */
  932. struct percpu_counter *sockets_allocated; /* Current number of sockets. */
  933. int *memory_pressure;
  934. long *sysctl_mem;
  935. unsigned long flags;
  936. /*
  937. * memcg field is used to find which memcg we belong directly
  938. * Each memcg struct can hold more than one cg_proto, so container_of
  939. * won't really cut.
  940. *
  941. * The elegant solution would be having an inverse function to
  942. * proto_cgroup in struct proto, but that means polluting the structure
  943. * for everybody, instead of just for memcg users.
  944. */
  945. struct mem_cgroup *memcg;
  946. };
  947. int proto_register(struct proto *prot, int alloc_slab);
  948. void proto_unregister(struct proto *prot);
  949. static inline bool memcg_proto_active(struct cg_proto *cg_proto)
  950. {
  951. return test_bit(MEMCG_SOCK_ACTIVE, &cg_proto->flags);
  952. }
  953. static inline bool memcg_proto_activated(struct cg_proto *cg_proto)
  954. {
  955. return test_bit(MEMCG_SOCK_ACTIVATED, &cg_proto->flags);
  956. }
  957. #ifdef SOCK_REFCNT_DEBUG
  958. static inline void sk_refcnt_debug_inc(struct sock *sk)
  959. {
  960. atomic_inc(&sk->sk_prot->socks);
  961. }
  962. static inline void sk_refcnt_debug_dec(struct sock *sk)
  963. {
  964. atomic_dec(&sk->sk_prot->socks);
  965. printk(KERN_DEBUG "%s socket %p released, %d are still alive\n",
  966. sk->sk_prot->name, sk, atomic_read(&sk->sk_prot->socks));
  967. }
  968. static inline void sk_refcnt_debug_release(const struct sock *sk)
  969. {
  970. if (atomic_read(&sk->sk_refcnt) != 1)
  971. printk(KERN_DEBUG "Destruction of the %s socket %p delayed, refcnt=%d\n",
  972. sk->sk_prot->name, sk, atomic_read(&sk->sk_refcnt));
  973. }
  974. #else /* SOCK_REFCNT_DEBUG */
  975. #define sk_refcnt_debug_inc(sk) do { } while (0)
  976. #define sk_refcnt_debug_dec(sk) do { } while (0)
  977. #define sk_refcnt_debug_release(sk) do { } while (0)
  978. #endif /* SOCK_REFCNT_DEBUG */
  979. #if defined(CONFIG_MEMCG_KMEM) && defined(CONFIG_NET)
  980. extern struct static_key memcg_socket_limit_enabled;
  981. static inline struct cg_proto *parent_cg_proto(struct proto *proto,
  982. struct cg_proto *cg_proto)
  983. {
  984. return proto->proto_cgroup(parent_mem_cgroup(cg_proto->memcg));
  985. }
  986. #define mem_cgroup_sockets_enabled static_key_false(&memcg_socket_limit_enabled)
  987. #else
  988. #define mem_cgroup_sockets_enabled 0
  989. static inline struct cg_proto *parent_cg_proto(struct proto *proto,
  990. struct cg_proto *cg_proto)
  991. {
  992. return NULL;
  993. }
  994. #endif
  995. static inline bool sk_stream_memory_free(const struct sock *sk)
  996. {
  997. if (sk->sk_wmem_queued >= sk->sk_sndbuf)
  998. return false;
  999. return sk->sk_prot->stream_memory_free ?
  1000. sk->sk_prot->stream_memory_free(sk) : true;
  1001. }
  1002. static inline bool sk_stream_is_writeable(const struct sock *sk)
  1003. {
  1004. return sk_stream_wspace(sk) >= sk_stream_min_wspace(sk) &&
  1005. sk_stream_memory_free(sk);
  1006. }
  1007. static inline bool sk_has_memory_pressure(const struct sock *sk)
  1008. {
  1009. return sk->sk_prot->memory_pressure != NULL;
  1010. }
  1011. static inline bool sk_under_memory_pressure(const struct sock *sk)
  1012. {
  1013. if (!sk->sk_prot->memory_pressure)
  1014. return false;
  1015. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1016. return !!*sk->sk_cgrp->memory_pressure;
  1017. return !!*sk->sk_prot->memory_pressure;
  1018. }
  1019. static inline void sk_leave_memory_pressure(struct sock *sk)
  1020. {
  1021. int *memory_pressure = sk->sk_prot->memory_pressure;
  1022. if (!memory_pressure)
  1023. return;
  1024. if (*memory_pressure)
  1025. *memory_pressure = 0;
  1026. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1027. struct cg_proto *cg_proto = sk->sk_cgrp;
  1028. struct proto *prot = sk->sk_prot;
  1029. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  1030. if (*cg_proto->memory_pressure)
  1031. *cg_proto->memory_pressure = 0;
  1032. }
  1033. }
  1034. static inline void sk_enter_memory_pressure(struct sock *sk)
  1035. {
  1036. if (!sk->sk_prot->enter_memory_pressure)
  1037. return;
  1038. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1039. struct cg_proto *cg_proto = sk->sk_cgrp;
  1040. struct proto *prot = sk->sk_prot;
  1041. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  1042. cg_proto->enter_memory_pressure(sk);
  1043. }
  1044. sk->sk_prot->enter_memory_pressure(sk);
  1045. }
  1046. static inline long sk_prot_mem_limits(const struct sock *sk, int index)
  1047. {
  1048. long *prot = sk->sk_prot->sysctl_mem;
  1049. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1050. prot = sk->sk_cgrp->sysctl_mem;
  1051. return prot[index];
  1052. }
  1053. static inline void memcg_memory_allocated_add(struct cg_proto *prot,
  1054. unsigned long amt,
  1055. int *parent_status)
  1056. {
  1057. struct res_counter *fail;
  1058. int ret;
  1059. ret = res_counter_charge_nofail(prot->memory_allocated,
  1060. amt << PAGE_SHIFT, &fail);
  1061. if (ret < 0)
  1062. *parent_status = OVER_LIMIT;
  1063. }
  1064. static inline void memcg_memory_allocated_sub(struct cg_proto *prot,
  1065. unsigned long amt)
  1066. {
  1067. res_counter_uncharge(prot->memory_allocated, amt << PAGE_SHIFT);
  1068. }
  1069. static inline u64 memcg_memory_allocated_read(struct cg_proto *prot)
  1070. {
  1071. u64 ret;
  1072. ret = res_counter_read_u64(prot->memory_allocated, RES_USAGE);
  1073. return ret >> PAGE_SHIFT;
  1074. }
  1075. static inline long
  1076. sk_memory_allocated(const struct sock *sk)
  1077. {
  1078. struct proto *prot = sk->sk_prot;
  1079. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1080. return memcg_memory_allocated_read(sk->sk_cgrp);
  1081. return atomic_long_read(prot->memory_allocated);
  1082. }
  1083. static inline long
  1084. sk_memory_allocated_add(struct sock *sk, int amt, int *parent_status)
  1085. {
  1086. struct proto *prot = sk->sk_prot;
  1087. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1088. memcg_memory_allocated_add(sk->sk_cgrp, amt, parent_status);
  1089. /* update the root cgroup regardless */
  1090. atomic_long_add_return(amt, prot->memory_allocated);
  1091. return memcg_memory_allocated_read(sk->sk_cgrp);
  1092. }
  1093. return atomic_long_add_return(amt, prot->memory_allocated);
  1094. }
  1095. static inline void
  1096. sk_memory_allocated_sub(struct sock *sk, int amt)
  1097. {
  1098. struct proto *prot = sk->sk_prot;
  1099. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1100. memcg_memory_allocated_sub(sk->sk_cgrp, amt);
  1101. atomic_long_sub(amt, prot->memory_allocated);
  1102. }
  1103. static inline void sk_sockets_allocated_dec(struct sock *sk)
  1104. {
  1105. struct proto *prot = sk->sk_prot;
  1106. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1107. struct cg_proto *cg_proto = sk->sk_cgrp;
  1108. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  1109. percpu_counter_dec(cg_proto->sockets_allocated);
  1110. }
  1111. percpu_counter_dec(prot->sockets_allocated);
  1112. }
  1113. static inline void sk_sockets_allocated_inc(struct sock *sk)
  1114. {
  1115. struct proto *prot = sk->sk_prot;
  1116. if (mem_cgroup_sockets_enabled && sk->sk_cgrp) {
  1117. struct cg_proto *cg_proto = sk->sk_cgrp;
  1118. for (; cg_proto; cg_proto = parent_cg_proto(prot, cg_proto))
  1119. percpu_counter_inc(cg_proto->sockets_allocated);
  1120. }
  1121. percpu_counter_inc(prot->sockets_allocated);
  1122. }
  1123. static inline int
  1124. sk_sockets_allocated_read_positive(struct sock *sk)
  1125. {
  1126. struct proto *prot = sk->sk_prot;
  1127. if (mem_cgroup_sockets_enabled && sk->sk_cgrp)
  1128. return percpu_counter_read_positive(sk->sk_cgrp->sockets_allocated);
  1129. return percpu_counter_read_positive(prot->sockets_allocated);
  1130. }
  1131. static inline int
  1132. proto_sockets_allocated_sum_positive(struct proto *prot)
  1133. {
  1134. return percpu_counter_sum_positive(prot->sockets_allocated);
  1135. }
  1136. static inline long
  1137. proto_memory_allocated(struct proto *prot)
  1138. {
  1139. return atomic_long_read(prot->memory_allocated);
  1140. }
  1141. static inline bool
  1142. proto_memory_pressure(struct proto *prot)
  1143. {
  1144. if (!prot->memory_pressure)
  1145. return false;
  1146. return !!*prot->memory_pressure;
  1147. }
  1148. #ifdef CONFIG_PROC_FS
  1149. /* Called with local bh disabled */
  1150. void sock_prot_inuse_add(struct net *net, struct proto *prot, int inc);
  1151. int sock_prot_inuse_get(struct net *net, struct proto *proto);
  1152. #else
  1153. static inline void sock_prot_inuse_add(struct net *net, struct proto *prot,
  1154. int inc)
  1155. {
  1156. }
  1157. #endif
  1158. /* With per-bucket locks this operation is not-atomic, so that
  1159. * this version is not worse.
  1160. */
  1161. static inline void __sk_prot_rehash(struct sock *sk)
  1162. {
  1163. sk->sk_prot->unhash(sk);
  1164. sk->sk_prot->hash(sk);
  1165. }
  1166. void sk_prot_clear_portaddr_nulls(struct sock *sk, int size);
  1167. /* About 10 seconds */
  1168. #define SOCK_DESTROY_TIME (10*HZ)
  1169. /* Sockets 0-1023 can't be bound to unless you are superuser */
  1170. #define PROT_SOCK 1024
  1171. #define SHUTDOWN_MASK 3
  1172. #define RCV_SHUTDOWN 1
  1173. #define SEND_SHUTDOWN 2
  1174. #define SOCK_SNDBUF_LOCK 1
  1175. #define SOCK_RCVBUF_LOCK 2
  1176. #define SOCK_BINDADDR_LOCK 4
  1177. #define SOCK_BINDPORT_LOCK 8
  1178. /* sock_iocb: used to kick off async processing of socket ios */
  1179. struct sock_iocb {
  1180. struct list_head list;
  1181. int flags;
  1182. int size;
  1183. struct socket *sock;
  1184. struct sock *sk;
  1185. struct scm_cookie *scm;
  1186. struct msghdr *msg, async_msg;
  1187. struct kiocb *kiocb;
  1188. };
  1189. static inline struct sock_iocb *kiocb_to_siocb(struct kiocb *iocb)
  1190. {
  1191. return (struct sock_iocb *)iocb->private;
  1192. }
  1193. static inline struct kiocb *siocb_to_kiocb(struct sock_iocb *si)
  1194. {
  1195. return si->kiocb;
  1196. }
  1197. struct socket_alloc {
  1198. struct socket socket;
  1199. struct inode vfs_inode;
  1200. };
  1201. static inline struct socket *SOCKET_I(struct inode *inode)
  1202. {
  1203. return &container_of(inode, struct socket_alloc, vfs_inode)->socket;
  1204. }
  1205. static inline struct inode *SOCK_INODE(struct socket *socket)
  1206. {
  1207. return &container_of(socket, struct socket_alloc, socket)->vfs_inode;
  1208. }
  1209. /*
  1210. * Functions for memory accounting
  1211. */
  1212. int __sk_mem_schedule(struct sock *sk, int size, int kind);
  1213. void __sk_mem_reclaim(struct sock *sk);
  1214. #define SK_MEM_QUANTUM ((int)PAGE_SIZE)
  1215. #define SK_MEM_QUANTUM_SHIFT ilog2(SK_MEM_QUANTUM)
  1216. #define SK_MEM_SEND 0
  1217. #define SK_MEM_RECV 1
  1218. static inline int sk_mem_pages(int amt)
  1219. {
  1220. return (amt + SK_MEM_QUANTUM - 1) >> SK_MEM_QUANTUM_SHIFT;
  1221. }
  1222. static inline bool sk_has_account(struct sock *sk)
  1223. {
  1224. /* return true if protocol supports memory accounting */
  1225. return !!sk->sk_prot->memory_allocated;
  1226. }
  1227. static inline bool sk_wmem_schedule(struct sock *sk, int size)
  1228. {
  1229. if (!sk_has_account(sk))
  1230. return true;
  1231. return size <= sk->sk_forward_alloc ||
  1232. __sk_mem_schedule(sk, size, SK_MEM_SEND);
  1233. }
  1234. static inline bool
  1235. sk_rmem_schedule(struct sock *sk, struct sk_buff *skb, int size)
  1236. {
  1237. if (!sk_has_account(sk))
  1238. return true;
  1239. return size<= sk->sk_forward_alloc ||
  1240. __sk_mem_schedule(sk, size, SK_MEM_RECV) ||
  1241. skb_pfmemalloc(skb);
  1242. }
  1243. static inline void sk_mem_reclaim(struct sock *sk)
  1244. {
  1245. if (!sk_has_account(sk))
  1246. return;
  1247. if (sk->sk_forward_alloc >= SK_MEM_QUANTUM)
  1248. __sk_mem_reclaim(sk);
  1249. }
  1250. static inline void sk_mem_reclaim_partial(struct sock *sk)
  1251. {
  1252. if (!sk_has_account(sk))
  1253. return;
  1254. if (sk->sk_forward_alloc > SK_MEM_QUANTUM)
  1255. __sk_mem_reclaim(sk);
  1256. }
  1257. static inline void sk_mem_charge(struct sock *sk, int size)
  1258. {
  1259. if (!sk_has_account(sk))
  1260. return;
  1261. sk->sk_forward_alloc -= size;
  1262. }
  1263. static inline void sk_mem_uncharge(struct sock *sk, int size)
  1264. {
  1265. if (!sk_has_account(sk))
  1266. return;
  1267. sk->sk_forward_alloc += size;
  1268. }
  1269. static inline void sk_wmem_free_skb(struct sock *sk, struct sk_buff *skb)
  1270. {
  1271. sock_set_flag(sk, SOCK_QUEUE_SHRUNK);
  1272. sk->sk_wmem_queued -= skb->truesize;
  1273. sk_mem_uncharge(sk, skb->truesize);
  1274. __kfree_skb(skb);
  1275. }
  1276. /* Used by processes to "lock" a socket state, so that
  1277. * interrupts and bottom half handlers won't change it
  1278. * from under us. It essentially blocks any incoming
  1279. * packets, so that we won't get any new data or any
  1280. * packets that change the state of the socket.
  1281. *
  1282. * While locked, BH processing will add new packets to
  1283. * the backlog queue. This queue is processed by the
  1284. * owner of the socket lock right before it is released.
  1285. *
  1286. * Since ~2.3.5 it is also exclusive sleep lock serializing
  1287. * accesses from user process context.
  1288. */
  1289. #define sock_owned_by_user(sk) ((sk)->sk_lock.owned)
  1290. /*
  1291. * Macro so as to not evaluate some arguments when
  1292. * lockdep is not enabled.
  1293. *
  1294. * Mark both the sk_lock and the sk_lock.slock as a
  1295. * per-address-family lock class.
  1296. */
  1297. #define sock_lock_init_class_and_name(sk, sname, skey, name, key) \
  1298. do { \
  1299. sk->sk_lock.owned = 0; \
  1300. init_waitqueue_head(&sk->sk_lock.wq); \
  1301. spin_lock_init(&(sk)->sk_lock.slock); \
  1302. debug_check_no_locks_freed((void *)&(sk)->sk_lock, \
  1303. sizeof((sk)->sk_lock)); \
  1304. lockdep_set_class_and_name(&(sk)->sk_lock.slock, \
  1305. (skey), (sname)); \
  1306. lockdep_init_map(&(sk)->sk_lock.dep_map, (name), (key), 0); \
  1307. } while (0)
  1308. void lock_sock_nested(struct sock *sk, int subclass);
  1309. static inline void lock_sock(struct sock *sk)
  1310. {
  1311. lock_sock_nested(sk, 0);
  1312. }
  1313. void release_sock(struct sock *sk);
  1314. /* BH context may only use the following locking interface. */
  1315. #define bh_lock_sock(__sk) spin_lock(&((__sk)->sk_lock.slock))
  1316. #define bh_lock_sock_nested(__sk) \
  1317. spin_lock_nested(&((__sk)->sk_lock.slock), \
  1318. SINGLE_DEPTH_NESTING)
  1319. #define bh_unlock_sock(__sk) spin_unlock(&((__sk)->sk_lock.slock))
  1320. bool lock_sock_fast(struct sock *sk);
  1321. /**
  1322. * unlock_sock_fast - complement of lock_sock_fast
  1323. * @sk: socket
  1324. * @slow: slow mode
  1325. *
  1326. * fast unlock socket for user context.
  1327. * If slow mode is on, we call regular release_sock()
  1328. */
  1329. static inline void unlock_sock_fast(struct sock *sk, bool slow)
  1330. {
  1331. if (slow)
  1332. release_sock(sk);
  1333. else
  1334. spin_unlock_bh(&sk->sk_lock.slock);
  1335. }
  1336. struct sock *sk_alloc(struct net *net, int family, gfp_t priority,
  1337. struct proto *prot);
  1338. void sk_free(struct sock *sk);
  1339. void sk_release_kernel(struct sock *sk);
  1340. struct sock *sk_clone_lock(const struct sock *sk, const gfp_t priority);
  1341. struct sk_buff *sock_wmalloc(struct sock *sk, unsigned long size, int force,
  1342. gfp_t priority);
  1343. struct sk_buff *sock_rmalloc(struct sock *sk, unsigned long size, int force,
  1344. gfp_t priority);
  1345. void sock_wfree(struct sk_buff *skb);
  1346. void skb_orphan_partial(struct sk_buff *skb);
  1347. void sock_rfree(struct sk_buff *skb);
  1348. void sock_edemux(struct sk_buff *skb);
  1349. int sock_setsockopt(struct socket *sock, int level, int op,
  1350. char __user *optval, unsigned int optlen);
  1351. int sock_getsockopt(struct socket *sock, int level, int op,
  1352. char __user *optval, int __user *optlen);
  1353. struct sk_buff *sock_alloc_send_skb(struct sock *sk, unsigned long size,
  1354. int noblock, int *errcode);
  1355. struct sk_buff *sock_alloc_send_pskb(struct sock *sk, unsigned long header_len,
  1356. unsigned long data_len, int noblock,
  1357. int *errcode, int max_page_order);
  1358. void *sock_kmalloc(struct sock *sk, int size, gfp_t priority);
  1359. void sock_kfree_s(struct sock *sk, void *mem, int size);
  1360. void sk_send_sigurg(struct sock *sk);
  1361. /*
  1362. * Functions to fill in entries in struct proto_ops when a protocol
  1363. * does not implement a particular function.
  1364. */
  1365. int sock_no_bind(struct socket *, struct sockaddr *, int);
  1366. int sock_no_connect(struct socket *, struct sockaddr *, int, int);
  1367. int sock_no_socketpair(struct socket *, struct socket *);
  1368. int sock_no_accept(struct socket *, struct socket *, int);
  1369. int sock_no_getname(struct socket *, struct sockaddr *, int *, int);
  1370. unsigned int sock_no_poll(struct file *, struct socket *,
  1371. struct poll_table_struct *);
  1372. int sock_no_ioctl(struct socket *, unsigned int, unsigned long);
  1373. int sock_no_listen(struct socket *, int);
  1374. int sock_no_shutdown(struct socket *, int);
  1375. int sock_no_getsockopt(struct socket *, int , int, char __user *, int __user *);
  1376. int sock_no_setsockopt(struct socket *, int, int, char __user *, unsigned int);
  1377. int sock_no_sendmsg(struct kiocb *, struct socket *, struct msghdr *, size_t);
  1378. int sock_no_recvmsg(struct kiocb *, struct socket *, struct msghdr *, size_t,
  1379. int);
  1380. int sock_no_mmap(struct file *file, struct socket *sock,
  1381. struct vm_area_struct *vma);
  1382. ssize_t sock_no_sendpage(struct socket *sock, struct page *page, int offset,
  1383. size_t size, int flags);
  1384. /*
  1385. * Functions to fill in entries in struct proto_ops when a protocol
  1386. * uses the inet style.
  1387. */
  1388. int sock_common_getsockopt(struct socket *sock, int level, int optname,
  1389. char __user *optval, int __user *optlen);
  1390. int sock_common_recvmsg(struct kiocb *iocb, struct socket *sock,
  1391. struct msghdr *msg, size_t size, int flags);
  1392. int sock_common_setsockopt(struct socket *sock, int level, int optname,
  1393. char __user *optval, unsigned int optlen);
  1394. int compat_sock_common_getsockopt(struct socket *sock, int level,
  1395. int optname, char __user *optval, int __user *optlen);
  1396. int compat_sock_common_setsockopt(struct socket *sock, int level,
  1397. int optname, char __user *optval, unsigned int optlen);
  1398. void sk_common_release(struct sock *sk);
  1399. /*
  1400. * Default socket callbacks and setup code
  1401. */
  1402. /* Initialise core socket variables */
  1403. void sock_init_data(struct socket *sock, struct sock *sk);
  1404. void sk_filter_release_rcu(struct rcu_head *rcu);
  1405. /**
  1406. * sk_filter_release - release a socket filter
  1407. * @fp: filter to remove
  1408. *
  1409. * Remove a filter from a socket and release its resources.
  1410. */
  1411. static inline void sk_filter_release(struct sk_filter *fp)
  1412. {
  1413. if (atomic_dec_and_test(&fp->refcnt))
  1414. call_rcu(&fp->rcu, sk_filter_release_rcu);
  1415. }
  1416. static inline void sk_filter_uncharge(struct sock *sk, struct sk_filter *fp)
  1417. {
  1418. atomic_sub(sk_filter_size(fp->len), &sk->sk_omem_alloc);
  1419. sk_filter_release(fp);
  1420. }
  1421. static inline void sk_filter_charge(struct sock *sk, struct sk_filter *fp)
  1422. {
  1423. atomic_inc(&fp->refcnt);
  1424. atomic_add(sk_filter_size(fp->len), &sk->sk_omem_alloc);
  1425. }
  1426. /*
  1427. * Socket reference counting postulates.
  1428. *
  1429. * * Each user of socket SHOULD hold a reference count.
  1430. * * Each access point to socket (an hash table bucket, reference from a list,
  1431. * running timer, skb in flight MUST hold a reference count.
  1432. * * When reference count hits 0, it means it will never increase back.
  1433. * * When reference count hits 0, it means that no references from
  1434. * outside exist to this socket and current process on current CPU
  1435. * is last user and may/should destroy this socket.
  1436. * * sk_free is called from any context: process, BH, IRQ. When
  1437. * it is called, socket has no references from outside -> sk_free
  1438. * may release descendant resources allocated by the socket, but
  1439. * to the time when it is called, socket is NOT referenced by any
  1440. * hash tables, lists etc.
  1441. * * Packets, delivered from outside (from network or from another process)
  1442. * and enqueued on receive/error queues SHOULD NOT grab reference count,
  1443. * when they sit in queue. Otherwise, packets will leak to hole, when
  1444. * socket is looked up by one cpu and unhasing is made by another CPU.
  1445. * It is true for udp/raw, netlink (leak to receive and error queues), tcp
  1446. * (leak to backlog). Packet socket does all the processing inside
  1447. * BR_NETPROTO_LOCK, so that it has not this race condition. UNIX sockets
  1448. * use separate SMP lock, so that they are prone too.
  1449. */
  1450. /* Ungrab socket and destroy it, if it was the last reference. */
  1451. static inline void sock_put(struct sock *sk)
  1452. {
  1453. if (atomic_dec_and_test(&sk->sk_refcnt))
  1454. sk_free(sk);
  1455. }
  1456. /* Generic version of sock_put(), dealing with all sockets
  1457. * (TCP_TIMEWAIT, ESTABLISHED...)
  1458. */
  1459. void sock_gen_put(struct sock *sk);
  1460. int sk_receive_skb(struct sock *sk, struct sk_buff *skb, const int nested);
  1461. static inline void sk_tx_queue_set(struct sock *sk, int tx_queue)
  1462. {
  1463. sk->sk_tx_queue_mapping = tx_queue;
  1464. }
  1465. static inline void sk_tx_queue_clear(struct sock *sk)
  1466. {
  1467. sk->sk_tx_queue_mapping = -1;
  1468. }
  1469. static inline int sk_tx_queue_get(const struct sock *sk)
  1470. {
  1471. return sk ? sk->sk_tx_queue_mapping : -1;
  1472. }
  1473. static inline void sk_set_socket(struct sock *sk, struct socket *sock)
  1474. {
  1475. sk_tx_queue_clear(sk);
  1476. sk->sk_socket = sock;
  1477. }
  1478. static inline wait_queue_head_t *sk_sleep(struct sock *sk)
  1479. {
  1480. BUILD_BUG_ON(offsetof(struct socket_wq, wait) != 0);
  1481. return &rcu_dereference_raw(sk->sk_wq)->wait;
  1482. }
  1483. /* Detach socket from process context.
  1484. * Announce socket dead, detach it from wait queue and inode.
  1485. * Note that parent inode held reference count on this struct sock,
  1486. * we do not release it in this function, because protocol
  1487. * probably wants some additional cleanups or even continuing
  1488. * to work with this socket (TCP).
  1489. */
  1490. static inline void sock_orphan(struct sock *sk)
  1491. {
  1492. write_lock_bh(&sk->sk_callback_lock);
  1493. sock_set_flag(sk, SOCK_DEAD);
  1494. sk_set_socket(sk, NULL);
  1495. sk->sk_wq = NULL;
  1496. write_unlock_bh(&sk->sk_callback_lock);
  1497. }
  1498. static inline void sock_graft(struct sock *sk, struct socket *parent)
  1499. {
  1500. write_lock_bh(&sk->sk_callback_lock);
  1501. sk->sk_wq = parent->wq;
  1502. parent->sk = sk;
  1503. sk_set_socket(sk, parent);
  1504. security_sock_graft(sk, parent);
  1505. write_unlock_bh(&sk->sk_callback_lock);
  1506. }
  1507. kuid_t sock_i_uid(struct sock *sk);
  1508. unsigned long sock_i_ino(struct sock *sk);
  1509. static inline struct dst_entry *
  1510. __sk_dst_get(struct sock *sk)
  1511. {
  1512. return rcu_dereference_check(sk->sk_dst_cache, sock_owned_by_user(sk) ||
  1513. lockdep_is_held(&sk->sk_lock.slock));
  1514. }
  1515. static inline struct dst_entry *
  1516. sk_dst_get(struct sock *sk)
  1517. {
  1518. struct dst_entry *dst;
  1519. rcu_read_lock();
  1520. dst = rcu_dereference(sk->sk_dst_cache);
  1521. if (dst)
  1522. dst_hold(dst);
  1523. rcu_read_unlock();
  1524. return dst;
  1525. }
  1526. void sk_reset_txq(struct sock *sk);
  1527. static inline void dst_negative_advice(struct sock *sk)
  1528. {
  1529. struct dst_entry *ndst, *dst = __sk_dst_get(sk);
  1530. if (dst && dst->ops->negative_advice) {
  1531. ndst = dst->ops->negative_advice(dst);
  1532. if (ndst != dst) {
  1533. rcu_assign_pointer(sk->sk_dst_cache, ndst);
  1534. sk_reset_txq(sk);
  1535. }
  1536. }
  1537. }
  1538. static inline void
  1539. __sk_dst_set(struct sock *sk, struct dst_entry *dst)
  1540. {
  1541. struct dst_entry *old_dst;
  1542. sk_tx_queue_clear(sk);
  1543. /*
  1544. * This can be called while sk is owned by the caller only,
  1545. * with no state that can be checked in a rcu_dereference_check() cond
  1546. */
  1547. old_dst = rcu_dereference_raw(sk->sk_dst_cache);
  1548. rcu_assign_pointer(sk->sk_dst_cache, dst);
  1549. dst_release(old_dst);
  1550. }
  1551. static inline void
  1552. sk_dst_set(struct sock *sk, struct dst_entry *dst)
  1553. {
  1554. spin_lock(&sk->sk_dst_lock);
  1555. __sk_dst_set(sk, dst);
  1556. spin_unlock(&sk->sk_dst_lock);
  1557. }
  1558. static inline void
  1559. __sk_dst_reset(struct sock *sk)
  1560. {
  1561. __sk_dst_set(sk, NULL);
  1562. }
  1563. static inline void
  1564. sk_dst_reset(struct sock *sk)
  1565. {
  1566. spin_lock(&sk->sk_dst_lock);
  1567. __sk_dst_reset(sk);
  1568. spin_unlock(&sk->sk_dst_lock);
  1569. }
  1570. struct dst_entry *__sk_dst_check(struct sock *sk, u32 cookie);
  1571. struct dst_entry *sk_dst_check(struct sock *sk, u32 cookie);
  1572. static inline bool sk_can_gso(const struct sock *sk)
  1573. {
  1574. return net_gso_ok(sk->sk_route_caps, sk->sk_gso_type);
  1575. }
  1576. void sk_setup_caps(struct sock *sk, struct dst_entry *dst);
  1577. static inline void sk_nocaps_add(struct sock *sk, netdev_features_t flags)
  1578. {
  1579. sk->sk_route_nocaps |= flags;
  1580. sk->sk_route_caps &= ~flags;
  1581. }
  1582. static inline int skb_do_copy_data_nocache(struct sock *sk, struct sk_buff *skb,
  1583. char __user *from, char *to,
  1584. int copy, int offset)
  1585. {
  1586. if (skb->ip_summed == CHECKSUM_NONE) {
  1587. int err = 0;
  1588. __wsum csum = csum_and_copy_from_user(from, to, copy, 0, &err);
  1589. if (err)
  1590. return err;
  1591. skb->csum = csum_block_add(skb->csum, csum, offset);
  1592. } else if (sk->sk_route_caps & NETIF_F_NOCACHE_COPY) {
  1593. if (!access_ok(VERIFY_READ, from, copy) ||
  1594. __copy_from_user_nocache(to, from, copy))
  1595. return -EFAULT;
  1596. } else if (copy_from_user(to, from, copy))
  1597. return -EFAULT;
  1598. return 0;
  1599. }
  1600. static inline int skb_add_data_nocache(struct sock *sk, struct sk_buff *skb,
  1601. char __user *from, int copy)
  1602. {
  1603. int err, offset = skb->len;
  1604. err = skb_do_copy_data_nocache(sk, skb, from, skb_put(skb, copy),
  1605. copy, offset);
  1606. if (err)
  1607. __skb_trim(skb, offset);
  1608. return err;
  1609. }
  1610. static inline int skb_copy_to_page_nocache(struct sock *sk, char __user *from,
  1611. struct sk_buff *skb,
  1612. struct page *page,
  1613. int off, int copy)
  1614. {
  1615. int err;
  1616. err = skb_do_copy_data_nocache(sk, skb, from, page_address(page) + off,
  1617. copy, skb->len);
  1618. if (err)
  1619. return err;
  1620. skb->len += copy;
  1621. skb->data_len += copy;
  1622. skb->truesize += copy;
  1623. sk->sk_wmem_queued += copy;
  1624. sk_mem_charge(sk, copy);
  1625. return 0;
  1626. }
  1627. static inline int skb_copy_to_page(struct sock *sk, char __user *from,
  1628. struct sk_buff *skb, struct page *page,
  1629. int off, int copy)
  1630. {
  1631. if (skb->ip_summed == CHECKSUM_NONE) {
  1632. int err = 0;
  1633. __wsum csum = csum_and_copy_from_user(from,
  1634. page_address(page) + off,
  1635. copy, 0, &err);
  1636. if (err)
  1637. return err;
  1638. skb->csum = csum_block_add(skb->csum, csum, skb->len);
  1639. } else if (copy_from_user(page_address(page) + off, from, copy))
  1640. return -EFAULT;
  1641. skb->len += copy;
  1642. skb->data_len += copy;
  1643. skb->truesize += copy;
  1644. sk->sk_wmem_queued += copy;
  1645. sk_mem_charge(sk, copy);
  1646. return 0;
  1647. }
  1648. /**
  1649. * sk_wmem_alloc_get - returns write allocations
  1650. * @sk: socket
  1651. *
  1652. * Returns sk_wmem_alloc minus initial offset of one
  1653. */
  1654. static inline int sk_wmem_alloc_get(const struct sock *sk)
  1655. {
  1656. return atomic_read(&sk->sk_wmem_alloc) - 1;
  1657. }
  1658. /**
  1659. * sk_rmem_alloc_get - returns read allocations
  1660. * @sk: socket
  1661. *
  1662. * Returns sk_rmem_alloc
  1663. */
  1664. static inline int sk_rmem_alloc_get(const struct sock *sk)
  1665. {
  1666. return atomic_read(&sk->sk_rmem_alloc);
  1667. }
  1668. /**
  1669. * sk_has_allocations - check if allocations are outstanding
  1670. * @sk: socket
  1671. *
  1672. * Returns true if socket has write or read allocations
  1673. */
  1674. static inline bool sk_has_allocations(const struct sock *sk)
  1675. {
  1676. return sk_wmem_alloc_get(sk) || sk_rmem_alloc_get(sk);
  1677. }
  1678. /**
  1679. * wq_has_sleeper - check if there are any waiting processes
  1680. * @wq: struct socket_wq
  1681. *
  1682. * Returns true if socket_wq has waiting processes
  1683. *
  1684. * The purpose of the wq_has_sleeper and sock_poll_wait is to wrap the memory
  1685. * barrier call. They were added due to the race found within the tcp code.
  1686. *
  1687. * Consider following tcp code paths:
  1688. *
  1689. * CPU1 CPU2
  1690. *
  1691. * sys_select receive packet
  1692. * ... ...
  1693. * __add_wait_queue update tp->rcv_nxt
  1694. * ... ...
  1695. * tp->rcv_nxt check sock_def_readable
  1696. * ... {
  1697. * schedule rcu_read_lock();
  1698. * wq = rcu_dereference(sk->sk_wq);
  1699. * if (wq && waitqueue_active(&wq->wait))
  1700. * wake_up_interruptible(&wq->wait)
  1701. * ...
  1702. * }
  1703. *
  1704. * The race for tcp fires when the __add_wait_queue changes done by CPU1 stay
  1705. * in its cache, and so does the tp->rcv_nxt update on CPU2 side. The CPU1
  1706. * could then endup calling schedule and sleep forever if there are no more
  1707. * data on the socket.
  1708. *
  1709. */
  1710. static inline bool wq_has_sleeper(struct socket_wq *wq)
  1711. {
  1712. /* We need to be sure we are in sync with the
  1713. * add_wait_queue modifications to the wait queue.
  1714. *
  1715. * This memory barrier is paired in the sock_poll_wait.
  1716. */
  1717. smp_mb();
  1718. return wq && waitqueue_active(&wq->wait);
  1719. }
  1720. /**
  1721. * sock_poll_wait - place memory barrier behind the poll_wait call.
  1722. * @filp: file
  1723. * @wait_address: socket wait queue
  1724. * @p: poll_table
  1725. *
  1726. * See the comments in the wq_has_sleeper function.
  1727. */
  1728. static inline void sock_poll_wait(struct file *filp,
  1729. wait_queue_head_t *wait_address, poll_table *p)
  1730. {
  1731. if (!poll_does_not_wait(p) && wait_address) {
  1732. poll_wait(filp, wait_address, p);
  1733. /* We need to be sure we are in sync with the
  1734. * socket flags modification.
  1735. *
  1736. * This memory barrier is paired in the wq_has_sleeper.
  1737. */
  1738. smp_mb();
  1739. }
  1740. }
  1741. /*
  1742. * Queue a received datagram if it will fit. Stream and sequenced
  1743. * protocols can't normally use this as they need to fit buffers in
  1744. * and play with them.
  1745. *
  1746. * Inlined as it's very short and called for pretty much every
  1747. * packet ever received.
  1748. */
  1749. static inline void skb_set_owner_w(struct sk_buff *skb, struct sock *sk)
  1750. {
  1751. skb_orphan(skb);
  1752. skb->sk = sk;
  1753. skb->destructor = sock_wfree;
  1754. /*
  1755. * We used to take a refcount on sk, but following operation
  1756. * is enough to guarantee sk_free() wont free this sock until
  1757. * all in-flight packets are completed
  1758. */
  1759. atomic_add(skb->truesize, &sk->sk_wmem_alloc);
  1760. }
  1761. static inline void skb_set_owner_r(struct sk_buff *skb, struct sock *sk)
  1762. {
  1763. skb_orphan(skb);
  1764. skb->sk = sk;
  1765. skb->destructor = sock_rfree;
  1766. atomic_add(skb->truesize, &sk->sk_rmem_alloc);
  1767. sk_mem_charge(sk, skb->truesize);
  1768. }
  1769. void sk_reset_timer(struct sock *sk, struct timer_list *timer,
  1770. unsigned long expires);
  1771. void sk_stop_timer(struct sock *sk, struct timer_list *timer);
  1772. int sock_queue_rcv_skb(struct sock *sk, struct sk_buff *skb);
  1773. int sock_queue_err_skb(struct sock *sk, struct sk_buff *skb);
  1774. /*
  1775. * Recover an error report and clear atomically
  1776. */
  1777. static inline int sock_error(struct sock *sk)
  1778. {
  1779. int err;
  1780. if (likely(!sk->sk_err))
  1781. return 0;
  1782. err = xchg(&sk->sk_err, 0);
  1783. return -err;
  1784. }
  1785. static inline unsigned long sock_wspace(struct sock *sk)
  1786. {
  1787. int amt = 0;
  1788. if (!(sk->sk_shutdown & SEND_SHUTDOWN)) {
  1789. amt = sk->sk_sndbuf - atomic_read(&sk->sk_wmem_alloc);
  1790. if (amt < 0)
  1791. amt = 0;
  1792. }
  1793. return amt;
  1794. }
  1795. static inline void sk_wake_async(struct sock *sk, int how, int band)
  1796. {
  1797. if (sock_flag(sk, SOCK_FASYNC))
  1798. sock_wake_async(sk->sk_socket, how, band);
  1799. }
  1800. /* Since sk_{r,w}mem_alloc sums skb->truesize, even a small frame might
  1801. * need sizeof(sk_buff) + MTU + padding, unless net driver perform copybreak.
  1802. * Note: for send buffers, TCP works better if we can build two skbs at
  1803. * minimum.
  1804. */
  1805. #define TCP_SKB_MIN_TRUESIZE (2048 + SKB_DATA_ALIGN(sizeof(struct sk_buff)))
  1806. #define SOCK_MIN_SNDBUF (TCP_SKB_MIN_TRUESIZE * 2)
  1807. #define SOCK_MIN_RCVBUF TCP_SKB_MIN_TRUESIZE
  1808. static inline void sk_stream_moderate_sndbuf(struct sock *sk)
  1809. {
  1810. if (!(sk->sk_userlocks & SOCK_SNDBUF_LOCK)) {
  1811. sk->sk_sndbuf = min(sk->sk_sndbuf, sk->sk_wmem_queued >> 1);
  1812. sk->sk_sndbuf = max_t(u32, sk->sk_sndbuf, SOCK_MIN_SNDBUF);
  1813. }
  1814. }
  1815. struct sk_buff *sk_stream_alloc_skb(struct sock *sk, int size, gfp_t gfp);
  1816. /**
  1817. * sk_page_frag - return an appropriate page_frag
  1818. * @sk: socket
  1819. *
  1820. * If socket allocation mode allows current thread to sleep, it means its
  1821. * safe to use the per task page_frag instead of the per socket one.
  1822. */
  1823. static inline struct page_frag *sk_page_frag(struct sock *sk)
  1824. {
  1825. if (sk->sk_allocation & __GFP_WAIT)
  1826. return &current->task_frag;
  1827. return &sk->sk_frag;
  1828. }
  1829. bool sk_page_frag_refill(struct sock *sk, struct page_frag *pfrag);
  1830. /*
  1831. * Default write policy as shown to user space via poll/select/SIGIO
  1832. */
  1833. static inline bool sock_writeable(const struct sock *sk)
  1834. {
  1835. return atomic_read(&sk->sk_wmem_alloc) < (sk->sk_sndbuf >> 1);
  1836. }
  1837. static inline gfp_t gfp_any(void)
  1838. {
  1839. return in_softirq() ? GFP_ATOMIC : GFP_KERNEL;
  1840. }
  1841. static inline long sock_rcvtimeo(const struct sock *sk, bool noblock)
  1842. {
  1843. return noblock ? 0 : sk->sk_rcvtimeo;
  1844. }
  1845. static inline long sock_sndtimeo(const struct sock *sk, bool noblock)
  1846. {
  1847. return noblock ? 0 : sk->sk_sndtimeo;
  1848. }
  1849. static inline int sock_rcvlowat(const struct sock *sk, int waitall, int len)
  1850. {
  1851. return (waitall ? len : min_t(int, sk->sk_rcvlowat, len)) ? : 1;
  1852. }
  1853. /* Alas, with timeout socket operations are not restartable.
  1854. * Compare this to poll().
  1855. */
  1856. static inline int sock_intr_errno(long timeo)
  1857. {
  1858. return timeo == MAX_SCHEDULE_TIMEOUT ? -ERESTARTSYS : -EINTR;
  1859. }
  1860. void __sock_recv_timestamp(struct msghdr *msg, struct sock *sk,
  1861. struct sk_buff *skb);
  1862. void __sock_recv_wifi_status(struct msghdr *msg, struct sock *sk,
  1863. struct sk_buff *skb);
  1864. static inline void
  1865. sock_recv_timestamp(struct msghdr *msg, struct sock *sk, struct sk_buff *skb)
  1866. {
  1867. ktime_t kt = skb->tstamp;
  1868. struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
  1869. /*
  1870. * generate control messages if
  1871. * - receive time stamping in software requested (SOCK_RCVTSTAMP
  1872. * or SOCK_TIMESTAMPING_RX_SOFTWARE)
  1873. * - software time stamp available and wanted
  1874. * (SOCK_TIMESTAMPING_SOFTWARE)
  1875. * - hardware time stamps available and wanted
  1876. * (SOCK_TIMESTAMPING_SYS_HARDWARE or
  1877. * SOCK_TIMESTAMPING_RAW_HARDWARE)
  1878. */
  1879. if (sock_flag(sk, SOCK_RCVTSTAMP) ||
  1880. sock_flag(sk, SOCK_TIMESTAMPING_RX_SOFTWARE) ||
  1881. (kt.tv64 && sock_flag(sk, SOCK_TIMESTAMPING_SOFTWARE)) ||
  1882. (hwtstamps->hwtstamp.tv64 &&
  1883. sock_flag(sk, SOCK_TIMESTAMPING_RAW_HARDWARE)) ||
  1884. (hwtstamps->syststamp.tv64 &&
  1885. sock_flag(sk, SOCK_TIMESTAMPING_SYS_HARDWARE)))
  1886. __sock_recv_timestamp(msg, sk, skb);
  1887. else
  1888. sk->sk_stamp = kt;
  1889. if (sock_flag(sk, SOCK_WIFI_STATUS) && skb->wifi_acked_valid)
  1890. __sock_recv_wifi_status(msg, sk, skb);
  1891. }
  1892. void __sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
  1893. struct sk_buff *skb);
  1894. static inline void sock_recv_ts_and_drops(struct msghdr *msg, struct sock *sk,
  1895. struct sk_buff *skb)
  1896. {
  1897. #define FLAGS_TS_OR_DROPS ((1UL << SOCK_RXQ_OVFL) | \
  1898. (1UL << SOCK_RCVTSTAMP) | \
  1899. (1UL << SOCK_TIMESTAMPING_RX_SOFTWARE) | \
  1900. (1UL << SOCK_TIMESTAMPING_SOFTWARE) | \
  1901. (1UL << SOCK_TIMESTAMPING_RAW_HARDWARE) | \
  1902. (1UL << SOCK_TIMESTAMPING_SYS_HARDWARE))
  1903. if (sk->sk_flags & FLAGS_TS_OR_DROPS)
  1904. __sock_recv_ts_and_drops(msg, sk, skb);
  1905. else
  1906. sk->sk_stamp = skb->tstamp;
  1907. }
  1908. /**
  1909. * sock_tx_timestamp - checks whether the outgoing packet is to be time stamped
  1910. * @sk: socket sending this packet
  1911. * @tx_flags: filled with instructions for time stamping
  1912. *
  1913. * Currently only depends on SOCK_TIMESTAMPING* flags.
  1914. */
  1915. void sock_tx_timestamp(struct sock *sk, __u8 *tx_flags);
  1916. /**
  1917. * sk_eat_skb - Release a skb if it is no longer needed
  1918. * @sk: socket to eat this skb from
  1919. * @skb: socket buffer to eat
  1920. * @copied_early: flag indicating whether DMA operations copied this data early
  1921. *
  1922. * This routine must be called with interrupts disabled or with the socket
  1923. * locked so that the sk_buff queue operation is ok.
  1924. */
  1925. #ifdef CONFIG_NET_DMA
  1926. static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
  1927. {
  1928. __skb_unlink(skb, &sk->sk_receive_queue);
  1929. if (!copied_early)
  1930. __kfree_skb(skb);
  1931. else
  1932. __skb_queue_tail(&sk->sk_async_wait_queue, skb);
  1933. }
  1934. #else
  1935. static inline void sk_eat_skb(struct sock *sk, struct sk_buff *skb, bool copied_early)
  1936. {
  1937. __skb_unlink(skb, &sk->sk_receive_queue);
  1938. __kfree_skb(skb);
  1939. }
  1940. #endif
  1941. static inline
  1942. struct net *sock_net(const struct sock *sk)
  1943. {
  1944. return read_pnet(&sk->sk_net);
  1945. }
  1946. static inline
  1947. void sock_net_set(struct sock *sk, struct net *net)
  1948. {
  1949. write_pnet(&sk->sk_net, net);
  1950. }
  1951. /*
  1952. * Kernel sockets, f.e. rtnl or icmp_socket, are a part of a namespace.
  1953. * They should not hold a reference to a namespace in order to allow
  1954. * to stop it.
  1955. * Sockets after sk_change_net should be released using sk_release_kernel
  1956. */
  1957. static inline void sk_change_net(struct sock *sk, struct net *net)
  1958. {
  1959. put_net(sock_net(sk));
  1960. sock_net_set(sk, hold_net(net));
  1961. }
  1962. static inline struct sock *skb_steal_sock(struct sk_buff *skb)
  1963. {
  1964. if (skb->sk) {
  1965. struct sock *sk = skb->sk;
  1966. skb->destructor = NULL;
  1967. skb->sk = NULL;
  1968. return sk;
  1969. }
  1970. return NULL;
  1971. }
  1972. void sock_enable_timestamp(struct sock *sk, int flag);
  1973. int sock_get_timestamp(struct sock *, struct timeval __user *);
  1974. int sock_get_timestampns(struct sock *, struct timespec __user *);
  1975. int sock_recv_errqueue(struct sock *sk, struct msghdr *msg, int len, int level,
  1976. int type);
  1977. /*
  1978. * Enable debug/info messages
  1979. */
  1980. extern int net_msg_warn;
  1981. #define NETDEBUG(fmt, args...) \
  1982. do { if (net_msg_warn) printk(fmt,##args); } while (0)
  1983. #define LIMIT_NETDEBUG(fmt, args...) \
  1984. do { if (net_msg_warn && net_ratelimit()) printk(fmt,##args); } while(0)
  1985. extern __u32 sysctl_wmem_max;
  1986. extern __u32 sysctl_rmem_max;
  1987. extern int sysctl_optmem_max;
  1988. extern __u32 sysctl_wmem_default;
  1989. extern __u32 sysctl_rmem_default;
  1990. #endif /* _SOCK_H */