bpf.h 105 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783
  1. /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
  2. /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
  3. *
  4. * This program is free software; you can redistribute it and/or
  5. * modify it under the terms of version 2 of the GNU General Public
  6. * License as published by the Free Software Foundation.
  7. */
  8. #ifndef _UAPI__LINUX_BPF_H__
  9. #define _UAPI__LINUX_BPF_H__
  10. #include <linux/types.h>
  11. #include <linux/bpf_common.h>
  12. /* Extended instruction set based on top of classic BPF */
  13. /* instruction classes */
  14. #define BPF_ALU64 0x07 /* alu mode in double word width */
  15. /* ld/ldx fields */
  16. #define BPF_DW 0x18 /* double word (64-bit) */
  17. #define BPF_XADD 0xc0 /* exclusive add */
  18. /* alu/jmp fields */
  19. #define BPF_MOV 0xb0 /* mov reg to reg */
  20. #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
  21. /* change endianness of a register */
  22. #define BPF_END 0xd0 /* flags for endianness conversion: */
  23. #define BPF_TO_LE 0x00 /* convert to little-endian */
  24. #define BPF_TO_BE 0x08 /* convert to big-endian */
  25. #define BPF_FROM_LE BPF_TO_LE
  26. #define BPF_FROM_BE BPF_TO_BE
  27. /* jmp encodings */
  28. #define BPF_JNE 0x50 /* jump != */
  29. #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
  30. #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
  31. #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
  32. #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
  33. #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
  34. #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
  35. #define BPF_CALL 0x80 /* function call */
  36. #define BPF_EXIT 0x90 /* function return */
  37. /* Register numbers */
  38. enum {
  39. BPF_REG_0 = 0,
  40. BPF_REG_1,
  41. BPF_REG_2,
  42. BPF_REG_3,
  43. BPF_REG_4,
  44. BPF_REG_5,
  45. BPF_REG_6,
  46. BPF_REG_7,
  47. BPF_REG_8,
  48. BPF_REG_9,
  49. BPF_REG_10,
  50. __MAX_BPF_REG,
  51. };
  52. /* BPF has 10 general purpose 64-bit registers and stack frame. */
  53. #define MAX_BPF_REG __MAX_BPF_REG
  54. struct bpf_insn {
  55. __u8 code; /* opcode */
  56. __u8 dst_reg:4; /* dest register */
  57. __u8 src_reg:4; /* source register */
  58. __s16 off; /* signed offset */
  59. __s32 imm; /* signed immediate constant */
  60. };
  61. /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
  62. struct bpf_lpm_trie_key {
  63. __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
  64. __u8 data[0]; /* Arbitrary size */
  65. };
  66. struct bpf_cgroup_storage_key {
  67. __u64 cgroup_inode_id; /* cgroup inode id */
  68. __u32 attach_type; /* program attach type */
  69. };
  70. /* BPF syscall commands, see bpf(2) man-page for details. */
  71. enum bpf_cmd {
  72. BPF_MAP_CREATE,
  73. BPF_MAP_LOOKUP_ELEM,
  74. BPF_MAP_UPDATE_ELEM,
  75. BPF_MAP_DELETE_ELEM,
  76. BPF_MAP_GET_NEXT_KEY,
  77. BPF_PROG_LOAD,
  78. BPF_OBJ_PIN,
  79. BPF_OBJ_GET,
  80. BPF_PROG_ATTACH,
  81. BPF_PROG_DETACH,
  82. BPF_PROG_TEST_RUN,
  83. BPF_PROG_GET_NEXT_ID,
  84. BPF_MAP_GET_NEXT_ID,
  85. BPF_PROG_GET_FD_BY_ID,
  86. BPF_MAP_GET_FD_BY_ID,
  87. BPF_OBJ_GET_INFO_BY_FD,
  88. BPF_PROG_QUERY,
  89. BPF_RAW_TRACEPOINT_OPEN,
  90. BPF_BTF_LOAD,
  91. BPF_BTF_GET_FD_BY_ID,
  92. BPF_TASK_FD_QUERY,
  93. };
  94. enum bpf_map_type {
  95. BPF_MAP_TYPE_UNSPEC,
  96. BPF_MAP_TYPE_HASH,
  97. BPF_MAP_TYPE_ARRAY,
  98. BPF_MAP_TYPE_PROG_ARRAY,
  99. BPF_MAP_TYPE_PERF_EVENT_ARRAY,
  100. BPF_MAP_TYPE_PERCPU_HASH,
  101. BPF_MAP_TYPE_PERCPU_ARRAY,
  102. BPF_MAP_TYPE_STACK_TRACE,
  103. BPF_MAP_TYPE_CGROUP_ARRAY,
  104. BPF_MAP_TYPE_LRU_HASH,
  105. BPF_MAP_TYPE_LRU_PERCPU_HASH,
  106. BPF_MAP_TYPE_LPM_TRIE,
  107. BPF_MAP_TYPE_ARRAY_OF_MAPS,
  108. BPF_MAP_TYPE_HASH_OF_MAPS,
  109. BPF_MAP_TYPE_DEVMAP,
  110. BPF_MAP_TYPE_SOCKMAP,
  111. BPF_MAP_TYPE_CPUMAP,
  112. BPF_MAP_TYPE_XSKMAP,
  113. BPF_MAP_TYPE_SOCKHASH,
  114. BPF_MAP_TYPE_CGROUP_STORAGE,
  115. BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
  116. };
  117. enum bpf_prog_type {
  118. BPF_PROG_TYPE_UNSPEC,
  119. BPF_PROG_TYPE_SOCKET_FILTER,
  120. BPF_PROG_TYPE_KPROBE,
  121. BPF_PROG_TYPE_SCHED_CLS,
  122. BPF_PROG_TYPE_SCHED_ACT,
  123. BPF_PROG_TYPE_TRACEPOINT,
  124. BPF_PROG_TYPE_XDP,
  125. BPF_PROG_TYPE_PERF_EVENT,
  126. BPF_PROG_TYPE_CGROUP_SKB,
  127. BPF_PROG_TYPE_CGROUP_SOCK,
  128. BPF_PROG_TYPE_LWT_IN,
  129. BPF_PROG_TYPE_LWT_OUT,
  130. BPF_PROG_TYPE_LWT_XMIT,
  131. BPF_PROG_TYPE_SOCK_OPS,
  132. BPF_PROG_TYPE_SK_SKB,
  133. BPF_PROG_TYPE_CGROUP_DEVICE,
  134. BPF_PROG_TYPE_SK_MSG,
  135. BPF_PROG_TYPE_RAW_TRACEPOINT,
  136. BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
  137. BPF_PROG_TYPE_LWT_SEG6LOCAL,
  138. BPF_PROG_TYPE_LIRC_MODE2,
  139. BPF_PROG_TYPE_SK_REUSEPORT,
  140. };
  141. enum bpf_attach_type {
  142. BPF_CGROUP_INET_INGRESS,
  143. BPF_CGROUP_INET_EGRESS,
  144. BPF_CGROUP_INET_SOCK_CREATE,
  145. BPF_CGROUP_SOCK_OPS,
  146. BPF_SK_SKB_STREAM_PARSER,
  147. BPF_SK_SKB_STREAM_VERDICT,
  148. BPF_CGROUP_DEVICE,
  149. BPF_SK_MSG_VERDICT,
  150. BPF_CGROUP_INET4_BIND,
  151. BPF_CGROUP_INET6_BIND,
  152. BPF_CGROUP_INET4_CONNECT,
  153. BPF_CGROUP_INET6_CONNECT,
  154. BPF_CGROUP_INET4_POST_BIND,
  155. BPF_CGROUP_INET6_POST_BIND,
  156. BPF_CGROUP_UDP4_SENDMSG,
  157. BPF_CGROUP_UDP6_SENDMSG,
  158. BPF_LIRC_MODE2,
  159. __MAX_BPF_ATTACH_TYPE
  160. };
  161. #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
  162. /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
  163. *
  164. * NONE(default): No further bpf programs allowed in the subtree.
  165. *
  166. * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
  167. * the program in this cgroup yields to sub-cgroup program.
  168. *
  169. * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
  170. * that cgroup program gets run in addition to the program in this cgroup.
  171. *
  172. * Only one program is allowed to be attached to a cgroup with
  173. * NONE or BPF_F_ALLOW_OVERRIDE flag.
  174. * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
  175. * release old program and attach the new one. Attach flags has to match.
  176. *
  177. * Multiple programs are allowed to be attached to a cgroup with
  178. * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
  179. * (those that were attached first, run first)
  180. * The programs of sub-cgroup are executed first, then programs of
  181. * this cgroup and then programs of parent cgroup.
  182. * When children program makes decision (like picking TCP CA or sock bind)
  183. * parent program has a chance to override it.
  184. *
  185. * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
  186. * A cgroup with NONE doesn't allow any programs in sub-cgroups.
  187. * Ex1:
  188. * cgrp1 (MULTI progs A, B) ->
  189. * cgrp2 (OVERRIDE prog C) ->
  190. * cgrp3 (MULTI prog D) ->
  191. * cgrp4 (OVERRIDE prog E) ->
  192. * cgrp5 (NONE prog F)
  193. * the event in cgrp5 triggers execution of F,D,A,B in that order.
  194. * if prog F is detached, the execution is E,D,A,B
  195. * if prog F and D are detached, the execution is E,A,B
  196. * if prog F, E and D are detached, the execution is C,A,B
  197. *
  198. * All eligible programs are executed regardless of return code from
  199. * earlier programs.
  200. */
  201. #define BPF_F_ALLOW_OVERRIDE (1U << 0)
  202. #define BPF_F_ALLOW_MULTI (1U << 1)
  203. /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
  204. * verifier will perform strict alignment checking as if the kernel
  205. * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
  206. * and NET_IP_ALIGN defined to 2.
  207. */
  208. #define BPF_F_STRICT_ALIGNMENT (1U << 0)
  209. /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
  210. #define BPF_PSEUDO_MAP_FD 1
  211. /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
  212. * offset to another bpf function
  213. */
  214. #define BPF_PSEUDO_CALL 1
  215. /* flags for BPF_MAP_UPDATE_ELEM command */
  216. #define BPF_ANY 0 /* create new element or update existing */
  217. #define BPF_NOEXIST 1 /* create new element if it didn't exist */
  218. #define BPF_EXIST 2 /* update existing element */
  219. /* flags for BPF_MAP_CREATE command */
  220. #define BPF_F_NO_PREALLOC (1U << 0)
  221. /* Instead of having one common LRU list in the
  222. * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
  223. * which can scale and perform better.
  224. * Note, the LRU nodes (including free nodes) cannot be moved
  225. * across different LRU lists.
  226. */
  227. #define BPF_F_NO_COMMON_LRU (1U << 1)
  228. /* Specify numa node during map creation */
  229. #define BPF_F_NUMA_NODE (1U << 2)
  230. /* flags for BPF_PROG_QUERY */
  231. #define BPF_F_QUERY_EFFECTIVE (1U << 0)
  232. #define BPF_OBJ_NAME_LEN 16U
  233. /* Flags for accessing BPF object */
  234. #define BPF_F_RDONLY (1U << 3)
  235. #define BPF_F_WRONLY (1U << 4)
  236. /* Flag for stack_map, store build_id+offset instead of pointer */
  237. #define BPF_F_STACK_BUILD_ID (1U << 5)
  238. enum bpf_stack_build_id_status {
  239. /* user space need an empty entry to identify end of a trace */
  240. BPF_STACK_BUILD_ID_EMPTY = 0,
  241. /* with valid build_id and offset */
  242. BPF_STACK_BUILD_ID_VALID = 1,
  243. /* couldn't get build_id, fallback to ip */
  244. BPF_STACK_BUILD_ID_IP = 2,
  245. };
  246. #define BPF_BUILD_ID_SIZE 20
  247. struct bpf_stack_build_id {
  248. __s32 status;
  249. unsigned char build_id[BPF_BUILD_ID_SIZE];
  250. union {
  251. __u64 offset;
  252. __u64 ip;
  253. };
  254. };
  255. union bpf_attr {
  256. struct { /* anonymous struct used by BPF_MAP_CREATE command */
  257. __u32 map_type; /* one of enum bpf_map_type */
  258. __u32 key_size; /* size of key in bytes */
  259. __u32 value_size; /* size of value in bytes */
  260. __u32 max_entries; /* max number of entries in a map */
  261. __u32 map_flags; /* BPF_MAP_CREATE related
  262. * flags defined above.
  263. */
  264. __u32 inner_map_fd; /* fd pointing to the inner map */
  265. __u32 numa_node; /* numa node (effective only if
  266. * BPF_F_NUMA_NODE is set).
  267. */
  268. char map_name[BPF_OBJ_NAME_LEN];
  269. __u32 map_ifindex; /* ifindex of netdev to create on */
  270. __u32 btf_fd; /* fd pointing to a BTF type data */
  271. __u32 btf_key_type_id; /* BTF type_id of the key */
  272. __u32 btf_value_type_id; /* BTF type_id of the value */
  273. };
  274. struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
  275. __u32 map_fd;
  276. __aligned_u64 key;
  277. union {
  278. __aligned_u64 value;
  279. __aligned_u64 next_key;
  280. };
  281. __u64 flags;
  282. };
  283. struct { /* anonymous struct used by BPF_PROG_LOAD command */
  284. __u32 prog_type; /* one of enum bpf_prog_type */
  285. __u32 insn_cnt;
  286. __aligned_u64 insns;
  287. __aligned_u64 license;
  288. __u32 log_level; /* verbosity level of verifier */
  289. __u32 log_size; /* size of user buffer */
  290. __aligned_u64 log_buf; /* user supplied buffer */
  291. __u32 kern_version; /* checked when prog_type=kprobe */
  292. __u32 prog_flags;
  293. char prog_name[BPF_OBJ_NAME_LEN];
  294. __u32 prog_ifindex; /* ifindex of netdev to prep for */
  295. /* For some prog types expected attach type must be known at
  296. * load time to verify attach type specific parts of prog
  297. * (context accesses, allowed helpers, etc).
  298. */
  299. __u32 expected_attach_type;
  300. };
  301. struct { /* anonymous struct used by BPF_OBJ_* commands */
  302. __aligned_u64 pathname;
  303. __u32 bpf_fd;
  304. __u32 file_flags;
  305. };
  306. struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
  307. __u32 target_fd; /* container object to attach to */
  308. __u32 attach_bpf_fd; /* eBPF program to attach */
  309. __u32 attach_type;
  310. __u32 attach_flags;
  311. };
  312. struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
  313. __u32 prog_fd;
  314. __u32 retval;
  315. __u32 data_size_in;
  316. __u32 data_size_out;
  317. __aligned_u64 data_in;
  318. __aligned_u64 data_out;
  319. __u32 repeat;
  320. __u32 duration;
  321. } test;
  322. struct { /* anonymous struct used by BPF_*_GET_*_ID */
  323. union {
  324. __u32 start_id;
  325. __u32 prog_id;
  326. __u32 map_id;
  327. __u32 btf_id;
  328. };
  329. __u32 next_id;
  330. __u32 open_flags;
  331. };
  332. struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
  333. __u32 bpf_fd;
  334. __u32 info_len;
  335. __aligned_u64 info;
  336. } info;
  337. struct { /* anonymous struct used by BPF_PROG_QUERY command */
  338. __u32 target_fd; /* container object to query */
  339. __u32 attach_type;
  340. __u32 query_flags;
  341. __u32 attach_flags;
  342. __aligned_u64 prog_ids;
  343. __u32 prog_cnt;
  344. } query;
  345. struct {
  346. __u64 name;
  347. __u32 prog_fd;
  348. } raw_tracepoint;
  349. struct { /* anonymous struct for BPF_BTF_LOAD */
  350. __aligned_u64 btf;
  351. __aligned_u64 btf_log_buf;
  352. __u32 btf_size;
  353. __u32 btf_log_size;
  354. __u32 btf_log_level;
  355. };
  356. struct {
  357. __u32 pid; /* input: pid */
  358. __u32 fd; /* input: fd */
  359. __u32 flags; /* input: flags */
  360. __u32 buf_len; /* input/output: buf len */
  361. __aligned_u64 buf; /* input/output:
  362. * tp_name for tracepoint
  363. * symbol for kprobe
  364. * filename for uprobe
  365. */
  366. __u32 prog_id; /* output: prod_id */
  367. __u32 fd_type; /* output: BPF_FD_TYPE_* */
  368. __u64 probe_offset; /* output: probe_offset */
  369. __u64 probe_addr; /* output: probe_addr */
  370. } task_fd_query;
  371. } __attribute__((aligned(8)));
  372. /* The description below is an attempt at providing documentation to eBPF
  373. * developers about the multiple available eBPF helper functions. It can be
  374. * parsed and used to produce a manual page. The workflow is the following,
  375. * and requires the rst2man utility:
  376. *
  377. * $ ./scripts/bpf_helpers_doc.py \
  378. * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
  379. * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
  380. * $ man /tmp/bpf-helpers.7
  381. *
  382. * Note that in order to produce this external documentation, some RST
  383. * formatting is used in the descriptions to get "bold" and "italics" in
  384. * manual pages. Also note that the few trailing white spaces are
  385. * intentional, removing them would break paragraphs for rst2man.
  386. *
  387. * Start of BPF helper function descriptions:
  388. *
  389. * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
  390. * Description
  391. * Perform a lookup in *map* for an entry associated to *key*.
  392. * Return
  393. * Map value associated to *key*, or **NULL** if no entry was
  394. * found.
  395. *
  396. * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
  397. * Description
  398. * Add or update the value of the entry associated to *key* in
  399. * *map* with *value*. *flags* is one of:
  400. *
  401. * **BPF_NOEXIST**
  402. * The entry for *key* must not exist in the map.
  403. * **BPF_EXIST**
  404. * The entry for *key* must already exist in the map.
  405. * **BPF_ANY**
  406. * No condition on the existence of the entry for *key*.
  407. *
  408. * Flag value **BPF_NOEXIST** cannot be used for maps of types
  409. * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
  410. * elements always exist), the helper would return an error.
  411. * Return
  412. * 0 on success, or a negative error in case of failure.
  413. *
  414. * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
  415. * Description
  416. * Delete entry with *key* from *map*.
  417. * Return
  418. * 0 on success, or a negative error in case of failure.
  419. *
  420. * int bpf_probe_read(void *dst, u32 size, const void *src)
  421. * Description
  422. * For tracing programs, safely attempt to read *size* bytes from
  423. * address *src* and store the data in *dst*.
  424. * Return
  425. * 0 on success, or a negative error in case of failure.
  426. *
  427. * u64 bpf_ktime_get_ns(void)
  428. * Description
  429. * Return the time elapsed since system boot, in nanoseconds.
  430. * Return
  431. * Current *ktime*.
  432. *
  433. * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
  434. * Description
  435. * This helper is a "printk()-like" facility for debugging. It
  436. * prints a message defined by format *fmt* (of size *fmt_size*)
  437. * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
  438. * available. It can take up to three additional **u64**
  439. * arguments (as an eBPF helpers, the total number of arguments is
  440. * limited to five).
  441. *
  442. * Each time the helper is called, it appends a line to the trace.
  443. * The format of the trace is customizable, and the exact output
  444. * one will get depends on the options set in
  445. * *\/sys/kernel/debug/tracing/trace_options* (see also the
  446. * *README* file under the same directory). However, it usually
  447. * defaults to something like:
  448. *
  449. * ::
  450. *
  451. * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
  452. *
  453. * In the above:
  454. *
  455. * * ``telnet`` is the name of the current task.
  456. * * ``470`` is the PID of the current task.
  457. * * ``001`` is the CPU number on which the task is
  458. * running.
  459. * * In ``.N..``, each character refers to a set of
  460. * options (whether irqs are enabled, scheduling
  461. * options, whether hard/softirqs are running, level of
  462. * preempt_disabled respectively). **N** means that
  463. * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
  464. * are set.
  465. * * ``419421.045894`` is a timestamp.
  466. * * ``0x00000001`` is a fake value used by BPF for the
  467. * instruction pointer register.
  468. * * ``<formatted msg>`` is the message formatted with
  469. * *fmt*.
  470. *
  471. * The conversion specifiers supported by *fmt* are similar, but
  472. * more limited than for printk(). They are **%d**, **%i**,
  473. * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
  474. * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
  475. * of field, padding with zeroes, etc.) is available, and the
  476. * helper will return **-EINVAL** (but print nothing) if it
  477. * encounters an unknown specifier.
  478. *
  479. * Also, note that **bpf_trace_printk**\ () is slow, and should
  480. * only be used for debugging purposes. For this reason, a notice
  481. * bloc (spanning several lines) is printed to kernel logs and
  482. * states that the helper should not be used "for production use"
  483. * the first time this helper is used (or more precisely, when
  484. * **trace_printk**\ () buffers are allocated). For passing values
  485. * to user space, perf events should be preferred.
  486. * Return
  487. * The number of bytes written to the buffer, or a negative error
  488. * in case of failure.
  489. *
  490. * u32 bpf_get_prandom_u32(void)
  491. * Description
  492. * Get a pseudo-random number.
  493. *
  494. * From a security point of view, this helper uses its own
  495. * pseudo-random internal state, and cannot be used to infer the
  496. * seed of other random functions in the kernel. However, it is
  497. * essential to note that the generator used by the helper is not
  498. * cryptographically secure.
  499. * Return
  500. * A random 32-bit unsigned value.
  501. *
  502. * u32 bpf_get_smp_processor_id(void)
  503. * Description
  504. * Get the SMP (symmetric multiprocessing) processor id. Note that
  505. * all programs run with preemption disabled, which means that the
  506. * SMP processor id is stable during all the execution of the
  507. * program.
  508. * Return
  509. * The SMP id of the processor running the program.
  510. *
  511. * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
  512. * Description
  513. * Store *len* bytes from address *from* into the packet
  514. * associated to *skb*, at *offset*. *flags* are a combination of
  515. * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
  516. * checksum for the packet after storing the bytes) and
  517. * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
  518. * **->swhash** and *skb*\ **->l4hash** to 0).
  519. *
  520. * A call to this helper is susceptible to change the underlaying
  521. * packet buffer. Therefore, at load time, all checks on pointers
  522. * previously done by the verifier are invalidated and must be
  523. * performed again, if the helper is used in combination with
  524. * direct packet access.
  525. * Return
  526. * 0 on success, or a negative error in case of failure.
  527. *
  528. * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
  529. * Description
  530. * Recompute the layer 3 (e.g. IP) checksum for the packet
  531. * associated to *skb*. Computation is incremental, so the helper
  532. * must know the former value of the header field that was
  533. * modified (*from*), the new value of this field (*to*), and the
  534. * number of bytes (2 or 4) for this field, stored in *size*.
  535. * Alternatively, it is possible to store the difference between
  536. * the previous and the new values of the header field in *to*, by
  537. * setting *from* and *size* to 0. For both methods, *offset*
  538. * indicates the location of the IP checksum within the packet.
  539. *
  540. * This helper works in combination with **bpf_csum_diff**\ (),
  541. * which does not update the checksum in-place, but offers more
  542. * flexibility and can handle sizes larger than 2 or 4 for the
  543. * checksum to update.
  544. *
  545. * A call to this helper is susceptible to change the underlaying
  546. * packet buffer. Therefore, at load time, all checks on pointers
  547. * previously done by the verifier are invalidated and must be
  548. * performed again, if the helper is used in combination with
  549. * direct packet access.
  550. * Return
  551. * 0 on success, or a negative error in case of failure.
  552. *
  553. * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
  554. * Description
  555. * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
  556. * packet associated to *skb*. Computation is incremental, so the
  557. * helper must know the former value of the header field that was
  558. * modified (*from*), the new value of this field (*to*), and the
  559. * number of bytes (2 or 4) for this field, stored on the lowest
  560. * four bits of *flags*. Alternatively, it is possible to store
  561. * the difference between the previous and the new values of the
  562. * header field in *to*, by setting *from* and the four lowest
  563. * bits of *flags* to 0. For both methods, *offset* indicates the
  564. * location of the IP checksum within the packet. In addition to
  565. * the size of the field, *flags* can be added (bitwise OR) actual
  566. * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
  567. * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
  568. * for updates resulting in a null checksum the value is set to
  569. * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
  570. * the checksum is to be computed against a pseudo-header.
  571. *
  572. * This helper works in combination with **bpf_csum_diff**\ (),
  573. * which does not update the checksum in-place, but offers more
  574. * flexibility and can handle sizes larger than 2 or 4 for the
  575. * checksum to update.
  576. *
  577. * A call to this helper is susceptible to change the underlaying
  578. * packet buffer. Therefore, at load time, all checks on pointers
  579. * previously done by the verifier are invalidated and must be
  580. * performed again, if the helper is used in combination with
  581. * direct packet access.
  582. * Return
  583. * 0 on success, or a negative error in case of failure.
  584. *
  585. * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
  586. * Description
  587. * This special helper is used to trigger a "tail call", or in
  588. * other words, to jump into another eBPF program. The same stack
  589. * frame is used (but values on stack and in registers for the
  590. * caller are not accessible to the callee). This mechanism allows
  591. * for program chaining, either for raising the maximum number of
  592. * available eBPF instructions, or to execute given programs in
  593. * conditional blocks. For security reasons, there is an upper
  594. * limit to the number of successive tail calls that can be
  595. * performed.
  596. *
  597. * Upon call of this helper, the program attempts to jump into a
  598. * program referenced at index *index* in *prog_array_map*, a
  599. * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
  600. * *ctx*, a pointer to the context.
  601. *
  602. * If the call succeeds, the kernel immediately runs the first
  603. * instruction of the new program. This is not a function call,
  604. * and it never returns to the previous program. If the call
  605. * fails, then the helper has no effect, and the caller continues
  606. * to run its subsequent instructions. A call can fail if the
  607. * destination program for the jump does not exist (i.e. *index*
  608. * is superior to the number of entries in *prog_array_map*), or
  609. * if the maximum number of tail calls has been reached for this
  610. * chain of programs. This limit is defined in the kernel by the
  611. * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
  612. * which is currently set to 32.
  613. * Return
  614. * 0 on success, or a negative error in case of failure.
  615. *
  616. * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
  617. * Description
  618. * Clone and redirect the packet associated to *skb* to another
  619. * net device of index *ifindex*. Both ingress and egress
  620. * interfaces can be used for redirection. The **BPF_F_INGRESS**
  621. * value in *flags* is used to make the distinction (ingress path
  622. * is selected if the flag is present, egress path otherwise).
  623. * This is the only flag supported for now.
  624. *
  625. * In comparison with **bpf_redirect**\ () helper,
  626. * **bpf_clone_redirect**\ () has the associated cost of
  627. * duplicating the packet buffer, but this can be executed out of
  628. * the eBPF program. Conversely, **bpf_redirect**\ () is more
  629. * efficient, but it is handled through an action code where the
  630. * redirection happens only after the eBPF program has returned.
  631. *
  632. * A call to this helper is susceptible to change the underlaying
  633. * packet buffer. Therefore, at load time, all checks on pointers
  634. * previously done by the verifier are invalidated and must be
  635. * performed again, if the helper is used in combination with
  636. * direct packet access.
  637. * Return
  638. * 0 on success, or a negative error in case of failure.
  639. *
  640. * u64 bpf_get_current_pid_tgid(void)
  641. * Return
  642. * A 64-bit integer containing the current tgid and pid, and
  643. * created as such:
  644. * *current_task*\ **->tgid << 32 \|**
  645. * *current_task*\ **->pid**.
  646. *
  647. * u64 bpf_get_current_uid_gid(void)
  648. * Return
  649. * A 64-bit integer containing the current GID and UID, and
  650. * created as such: *current_gid* **<< 32 \|** *current_uid*.
  651. *
  652. * int bpf_get_current_comm(char *buf, u32 size_of_buf)
  653. * Description
  654. * Copy the **comm** attribute of the current task into *buf* of
  655. * *size_of_buf*. The **comm** attribute contains the name of
  656. * the executable (excluding the path) for the current task. The
  657. * *size_of_buf* must be strictly positive. On success, the
  658. * helper makes sure that the *buf* is NUL-terminated. On failure,
  659. * it is filled with zeroes.
  660. * Return
  661. * 0 on success, or a negative error in case of failure.
  662. *
  663. * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
  664. * Description
  665. * Retrieve the classid for the current task, i.e. for the net_cls
  666. * cgroup to which *skb* belongs.
  667. *
  668. * This helper can be used on TC egress path, but not on ingress.
  669. *
  670. * The net_cls cgroup provides an interface to tag network packets
  671. * based on a user-provided identifier for all traffic coming from
  672. * the tasks belonging to the related cgroup. See also the related
  673. * kernel documentation, available from the Linux sources in file
  674. * *Documentation/cgroup-v1/net_cls.txt*.
  675. *
  676. * The Linux kernel has two versions for cgroups: there are
  677. * cgroups v1 and cgroups v2. Both are available to users, who can
  678. * use a mixture of them, but note that the net_cls cgroup is for
  679. * cgroup v1 only. This makes it incompatible with BPF programs
  680. * run on cgroups, which is a cgroup-v2-only feature (a socket can
  681. * only hold data for one version of cgroups at a time).
  682. *
  683. * This helper is only available is the kernel was compiled with
  684. * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
  685. * "**y**" or to "**m**".
  686. * Return
  687. * The classid, or 0 for the default unconfigured classid.
  688. *
  689. * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
  690. * Description
  691. * Push a *vlan_tci* (VLAN tag control information) of protocol
  692. * *vlan_proto* to the packet associated to *skb*, then update
  693. * the checksum. Note that if *vlan_proto* is different from
  694. * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
  695. * be **ETH_P_8021Q**.
  696. *
  697. * A call to this helper is susceptible to change the underlaying
  698. * packet buffer. Therefore, at load time, all checks on pointers
  699. * previously done by the verifier are invalidated and must be
  700. * performed again, if the helper is used in combination with
  701. * direct packet access.
  702. * Return
  703. * 0 on success, or a negative error in case of failure.
  704. *
  705. * int bpf_skb_vlan_pop(struct sk_buff *skb)
  706. * Description
  707. * Pop a VLAN header from the packet associated to *skb*.
  708. *
  709. * A call to this helper is susceptible to change the underlaying
  710. * packet buffer. Therefore, at load time, all checks on pointers
  711. * previously done by the verifier are invalidated and must be
  712. * performed again, if the helper is used in combination with
  713. * direct packet access.
  714. * Return
  715. * 0 on success, or a negative error in case of failure.
  716. *
  717. * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  718. * Description
  719. * Get tunnel metadata. This helper takes a pointer *key* to an
  720. * empty **struct bpf_tunnel_key** of **size**, that will be
  721. * filled with tunnel metadata for the packet associated to *skb*.
  722. * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
  723. * indicates that the tunnel is based on IPv6 protocol instead of
  724. * IPv4.
  725. *
  726. * The **struct bpf_tunnel_key** is an object that generalizes the
  727. * principal parameters used by various tunneling protocols into a
  728. * single struct. This way, it can be used to easily make a
  729. * decision based on the contents of the encapsulation header,
  730. * "summarized" in this struct. In particular, it holds the IP
  731. * address of the remote end (IPv4 or IPv6, depending on the case)
  732. * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
  733. * this struct exposes the *key*\ **->tunnel_id**, which is
  734. * generally mapped to a VNI (Virtual Network Identifier), making
  735. * it programmable together with the **bpf_skb_set_tunnel_key**\
  736. * () helper.
  737. *
  738. * Let's imagine that the following code is part of a program
  739. * attached to the TC ingress interface, on one end of a GRE
  740. * tunnel, and is supposed to filter out all messages coming from
  741. * remote ends with IPv4 address other than 10.0.0.1:
  742. *
  743. * ::
  744. *
  745. * int ret;
  746. * struct bpf_tunnel_key key = {};
  747. *
  748. * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
  749. * if (ret < 0)
  750. * return TC_ACT_SHOT; // drop packet
  751. *
  752. * if (key.remote_ipv4 != 0x0a000001)
  753. * return TC_ACT_SHOT; // drop packet
  754. *
  755. * return TC_ACT_OK; // accept packet
  756. *
  757. * This interface can also be used with all encapsulation devices
  758. * that can operate in "collect metadata" mode: instead of having
  759. * one network device per specific configuration, the "collect
  760. * metadata" mode only requires a single device where the
  761. * configuration can be extracted from this helper.
  762. *
  763. * This can be used together with various tunnels such as VXLan,
  764. * Geneve, GRE or IP in IP (IPIP).
  765. * Return
  766. * 0 on success, or a negative error in case of failure.
  767. *
  768. * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  769. * Description
  770. * Populate tunnel metadata for packet associated to *skb.* The
  771. * tunnel metadata is set to the contents of *key*, of *size*. The
  772. * *flags* can be set to a combination of the following values:
  773. *
  774. * **BPF_F_TUNINFO_IPV6**
  775. * Indicate that the tunnel is based on IPv6 protocol
  776. * instead of IPv4.
  777. * **BPF_F_ZERO_CSUM_TX**
  778. * For IPv4 packets, add a flag to tunnel metadata
  779. * indicating that checksum computation should be skipped
  780. * and checksum set to zeroes.
  781. * **BPF_F_DONT_FRAGMENT**
  782. * Add a flag to tunnel metadata indicating that the
  783. * packet should not be fragmented.
  784. * **BPF_F_SEQ_NUMBER**
  785. * Add a flag to tunnel metadata indicating that a
  786. * sequence number should be added to tunnel header before
  787. * sending the packet. This flag was added for GRE
  788. * encapsulation, but might be used with other protocols
  789. * as well in the future.
  790. *
  791. * Here is a typical usage on the transmit path:
  792. *
  793. * ::
  794. *
  795. * struct bpf_tunnel_key key;
  796. * populate key ...
  797. * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
  798. * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
  799. *
  800. * See also the description of the **bpf_skb_get_tunnel_key**\ ()
  801. * helper for additional information.
  802. * Return
  803. * 0 on success, or a negative error in case of failure.
  804. *
  805. * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
  806. * Description
  807. * Read the value of a perf event counter. This helper relies on a
  808. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
  809. * the perf event counter is selected when *map* is updated with
  810. * perf event file descriptors. The *map* is an array whose size
  811. * is the number of available CPUs, and each cell contains a value
  812. * relative to one CPU. The value to retrieve is indicated by
  813. * *flags*, that contains the index of the CPU to look up, masked
  814. * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  815. * **BPF_F_CURRENT_CPU** to indicate that the value for the
  816. * current CPU should be retrieved.
  817. *
  818. * Note that before Linux 4.13, only hardware perf event can be
  819. * retrieved.
  820. *
  821. * Also, be aware that the newer helper
  822. * **bpf_perf_event_read_value**\ () is recommended over
  823. * **bpf_perf_event_read**\ () in general. The latter has some ABI
  824. * quirks where error and counter value are used as a return code
  825. * (which is wrong to do since ranges may overlap). This issue is
  826. * fixed with **bpf_perf_event_read_value**\ (), which at the same
  827. * time provides more features over the **bpf_perf_event_read**\
  828. * () interface. Please refer to the description of
  829. * **bpf_perf_event_read_value**\ () for details.
  830. * Return
  831. * The value of the perf event counter read from the map, or a
  832. * negative error code in case of failure.
  833. *
  834. * int bpf_redirect(u32 ifindex, u64 flags)
  835. * Description
  836. * Redirect the packet to another net device of index *ifindex*.
  837. * This helper is somewhat similar to **bpf_clone_redirect**\
  838. * (), except that the packet is not cloned, which provides
  839. * increased performance.
  840. *
  841. * Except for XDP, both ingress and egress interfaces can be used
  842. * for redirection. The **BPF_F_INGRESS** value in *flags* is used
  843. * to make the distinction (ingress path is selected if the flag
  844. * is present, egress path otherwise). Currently, XDP only
  845. * supports redirection to the egress interface, and accepts no
  846. * flag at all.
  847. *
  848. * The same effect can be attained with the more generic
  849. * **bpf_redirect_map**\ (), which requires specific maps to be
  850. * used but offers better performance.
  851. * Return
  852. * For XDP, the helper returns **XDP_REDIRECT** on success or
  853. * **XDP_ABORTED** on error. For other program types, the values
  854. * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
  855. * error.
  856. *
  857. * u32 bpf_get_route_realm(struct sk_buff *skb)
  858. * Description
  859. * Retrieve the realm or the route, that is to say the
  860. * **tclassid** field of the destination for the *skb*. The
  861. * indentifier retrieved is a user-provided tag, similar to the
  862. * one used with the net_cls cgroup (see description for
  863. * **bpf_get_cgroup_classid**\ () helper), but here this tag is
  864. * held by a route (a destination entry), not by a task.
  865. *
  866. * Retrieving this identifier works with the clsact TC egress hook
  867. * (see also **tc-bpf(8)**), or alternatively on conventional
  868. * classful egress qdiscs, but not on TC ingress path. In case of
  869. * clsact TC egress hook, this has the advantage that, internally,
  870. * the destination entry has not been dropped yet in the transmit
  871. * path. Therefore, the destination entry does not need to be
  872. * artificially held via **netif_keep_dst**\ () for a classful
  873. * qdisc until the *skb* is freed.
  874. *
  875. * This helper is available only if the kernel was compiled with
  876. * **CONFIG_IP_ROUTE_CLASSID** configuration option.
  877. * Return
  878. * The realm of the route for the packet associated to *skb*, or 0
  879. * if none was found.
  880. *
  881. * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  882. * Description
  883. * Write raw *data* blob into a special BPF perf event held by
  884. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  885. * event must have the following attributes: **PERF_SAMPLE_RAW**
  886. * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  887. * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  888. *
  889. * The *flags* are used to indicate the index in *map* for which
  890. * the value must be put, masked with **BPF_F_INDEX_MASK**.
  891. * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  892. * to indicate that the index of the current CPU core should be
  893. * used.
  894. *
  895. * The value to write, of *size*, is passed through eBPF stack and
  896. * pointed by *data*.
  897. *
  898. * The context of the program *ctx* needs also be passed to the
  899. * helper.
  900. *
  901. * On user space, a program willing to read the values needs to
  902. * call **perf_event_open**\ () on the perf event (either for
  903. * one or for all CPUs) and to store the file descriptor into the
  904. * *map*. This must be done before the eBPF program can send data
  905. * into it. An example is available in file
  906. * *samples/bpf/trace_output_user.c* in the Linux kernel source
  907. * tree (the eBPF program counterpart is in
  908. * *samples/bpf/trace_output_kern.c*).
  909. *
  910. * **bpf_perf_event_output**\ () achieves better performance
  911. * than **bpf_trace_printk**\ () for sharing data with user
  912. * space, and is much better suitable for streaming data from eBPF
  913. * programs.
  914. *
  915. * Note that this helper is not restricted to tracing use cases
  916. * and can be used with programs attached to TC or XDP as well,
  917. * where it allows for passing data to user space listeners. Data
  918. * can be:
  919. *
  920. * * Only custom structs,
  921. * * Only the packet payload, or
  922. * * A combination of both.
  923. * Return
  924. * 0 on success, or a negative error in case of failure.
  925. *
  926. * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
  927. * Description
  928. * This helper was provided as an easy way to load data from a
  929. * packet. It can be used to load *len* bytes from *offset* from
  930. * the packet associated to *skb*, into the buffer pointed by
  931. * *to*.
  932. *
  933. * Since Linux 4.7, usage of this helper has mostly been replaced
  934. * by "direct packet access", enabling packet data to be
  935. * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
  936. * pointing respectively to the first byte of packet data and to
  937. * the byte after the last byte of packet data. However, it
  938. * remains useful if one wishes to read large quantities of data
  939. * at once from a packet into the eBPF stack.
  940. * Return
  941. * 0 on success, or a negative error in case of failure.
  942. *
  943. * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
  944. * Description
  945. * Walk a user or a kernel stack and return its id. To achieve
  946. * this, the helper needs *ctx*, which is a pointer to the context
  947. * on which the tracing program is executed, and a pointer to a
  948. * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
  949. *
  950. * The last argument, *flags*, holds the number of stack frames to
  951. * skip (from 0 to 255), masked with
  952. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  953. * a combination of the following flags:
  954. *
  955. * **BPF_F_USER_STACK**
  956. * Collect a user space stack instead of a kernel stack.
  957. * **BPF_F_FAST_STACK_CMP**
  958. * Compare stacks by hash only.
  959. * **BPF_F_REUSE_STACKID**
  960. * If two different stacks hash into the same *stackid*,
  961. * discard the old one.
  962. *
  963. * The stack id retrieved is a 32 bit long integer handle which
  964. * can be further combined with other data (including other stack
  965. * ids) and used as a key into maps. This can be useful for
  966. * generating a variety of graphs (such as flame graphs or off-cpu
  967. * graphs).
  968. *
  969. * For walking a stack, this helper is an improvement over
  970. * **bpf_probe_read**\ (), which can be used with unrolled loops
  971. * but is not efficient and consumes a lot of eBPF instructions.
  972. * Instead, **bpf_get_stackid**\ () can collect up to
  973. * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
  974. * this limit can be controlled with the **sysctl** program, and
  975. * that it should be manually increased in order to profile long
  976. * user stacks (such as stacks for Java programs). To do so, use:
  977. *
  978. * ::
  979. *
  980. * # sysctl kernel.perf_event_max_stack=<new value>
  981. * Return
  982. * The positive or null stack id on success, or a negative error
  983. * in case of failure.
  984. *
  985. * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
  986. * Description
  987. * Compute a checksum difference, from the raw buffer pointed by
  988. * *from*, of length *from_size* (that must be a multiple of 4),
  989. * towards the raw buffer pointed by *to*, of size *to_size*
  990. * (same remark). An optional *seed* can be added to the value
  991. * (this can be cascaded, the seed may come from a previous call
  992. * to the helper).
  993. *
  994. * This is flexible enough to be used in several ways:
  995. *
  996. * * With *from_size* == 0, *to_size* > 0 and *seed* set to
  997. * checksum, it can be used when pushing new data.
  998. * * With *from_size* > 0, *to_size* == 0 and *seed* set to
  999. * checksum, it can be used when removing data from a packet.
  1000. * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
  1001. * can be used to compute a diff. Note that *from_size* and
  1002. * *to_size* do not need to be equal.
  1003. *
  1004. * This helper can be used in combination with
  1005. * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
  1006. * which one can feed in the difference computed with
  1007. * **bpf_csum_diff**\ ().
  1008. * Return
  1009. * The checksum result, or a negative error code in case of
  1010. * failure.
  1011. *
  1012. * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
  1013. * Description
  1014. * Retrieve tunnel options metadata for the packet associated to
  1015. * *skb*, and store the raw tunnel option data to the buffer *opt*
  1016. * of *size*.
  1017. *
  1018. * This helper can be used with encapsulation devices that can
  1019. * operate in "collect metadata" mode (please refer to the related
  1020. * note in the description of **bpf_skb_get_tunnel_key**\ () for
  1021. * more details). A particular example where this can be used is
  1022. * in combination with the Geneve encapsulation protocol, where it
  1023. * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
  1024. * and retrieving arbitrary TLVs (Type-Length-Value headers) from
  1025. * the eBPF program. This allows for full customization of these
  1026. * headers.
  1027. * Return
  1028. * The size of the option data retrieved.
  1029. *
  1030. * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
  1031. * Description
  1032. * Set tunnel options metadata for the packet associated to *skb*
  1033. * to the option data contained in the raw buffer *opt* of *size*.
  1034. *
  1035. * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
  1036. * helper for additional information.
  1037. * Return
  1038. * 0 on success, or a negative error in case of failure.
  1039. *
  1040. * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
  1041. * Description
  1042. * Change the protocol of the *skb* to *proto*. Currently
  1043. * supported are transition from IPv4 to IPv6, and from IPv6 to
  1044. * IPv4. The helper takes care of the groundwork for the
  1045. * transition, including resizing the socket buffer. The eBPF
  1046. * program is expected to fill the new headers, if any, via
  1047. * **skb_store_bytes**\ () and to recompute the checksums with
  1048. * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
  1049. * (). The main case for this helper is to perform NAT64
  1050. * operations out of an eBPF program.
  1051. *
  1052. * Internally, the GSO type is marked as dodgy so that headers are
  1053. * checked and segments are recalculated by the GSO/GRO engine.
  1054. * The size for GSO target is adapted as well.
  1055. *
  1056. * All values for *flags* are reserved for future usage, and must
  1057. * be left at zero.
  1058. *
  1059. * A call to this helper is susceptible to change the underlaying
  1060. * packet buffer. Therefore, at load time, all checks on pointers
  1061. * previously done by the verifier are invalidated and must be
  1062. * performed again, if the helper is used in combination with
  1063. * direct packet access.
  1064. * Return
  1065. * 0 on success, or a negative error in case of failure.
  1066. *
  1067. * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
  1068. * Description
  1069. * Change the packet type for the packet associated to *skb*. This
  1070. * comes down to setting *skb*\ **->pkt_type** to *type*, except
  1071. * the eBPF program does not have a write access to *skb*\
  1072. * **->pkt_type** beside this helper. Using a helper here allows
  1073. * for graceful handling of errors.
  1074. *
  1075. * The major use case is to change incoming *skb*s to
  1076. * **PACKET_HOST** in a programmatic way instead of having to
  1077. * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
  1078. * example.
  1079. *
  1080. * Note that *type* only allows certain values. At this time, they
  1081. * are:
  1082. *
  1083. * **PACKET_HOST**
  1084. * Packet is for us.
  1085. * **PACKET_BROADCAST**
  1086. * Send packet to all.
  1087. * **PACKET_MULTICAST**
  1088. * Send packet to group.
  1089. * **PACKET_OTHERHOST**
  1090. * Send packet to someone else.
  1091. * Return
  1092. * 0 on success, or a negative error in case of failure.
  1093. *
  1094. * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
  1095. * Description
  1096. * Check whether *skb* is a descendant of the cgroup2 held by
  1097. * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  1098. * Return
  1099. * The return value depends on the result of the test, and can be:
  1100. *
  1101. * * 0, if the *skb* failed the cgroup2 descendant test.
  1102. * * 1, if the *skb* succeeded the cgroup2 descendant test.
  1103. * * A negative error code, if an error occurred.
  1104. *
  1105. * u32 bpf_get_hash_recalc(struct sk_buff *skb)
  1106. * Description
  1107. * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
  1108. * not set, in particular if the hash was cleared due to mangling,
  1109. * recompute this hash. Later accesses to the hash can be done
  1110. * directly with *skb*\ **->hash**.
  1111. *
  1112. * Calling **bpf_set_hash_invalid**\ (), changing a packet
  1113. * prototype with **bpf_skb_change_proto**\ (), or calling
  1114. * **bpf_skb_store_bytes**\ () with the
  1115. * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
  1116. * the hash and to trigger a new computation for the next call to
  1117. * **bpf_get_hash_recalc**\ ().
  1118. * Return
  1119. * The 32-bit hash.
  1120. *
  1121. * u64 bpf_get_current_task(void)
  1122. * Return
  1123. * A pointer to the current task struct.
  1124. *
  1125. * int bpf_probe_write_user(void *dst, const void *src, u32 len)
  1126. * Description
  1127. * Attempt in a safe way to write *len* bytes from the buffer
  1128. * *src* to *dst* in memory. It only works for threads that are in
  1129. * user context, and *dst* must be a valid user space address.
  1130. *
  1131. * This helper should not be used to implement any kind of
  1132. * security mechanism because of TOC-TOU attacks, but rather to
  1133. * debug, divert, and manipulate execution of semi-cooperative
  1134. * processes.
  1135. *
  1136. * Keep in mind that this feature is meant for experiments, and it
  1137. * has a risk of crashing the system and running programs.
  1138. * Therefore, when an eBPF program using this helper is attached,
  1139. * a warning including PID and process name is printed to kernel
  1140. * logs.
  1141. * Return
  1142. * 0 on success, or a negative error in case of failure.
  1143. *
  1144. * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
  1145. * Description
  1146. * Check whether the probe is being run is the context of a given
  1147. * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
  1148. * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  1149. * Return
  1150. * The return value depends on the result of the test, and can be:
  1151. *
  1152. * * 0, if the *skb* task belongs to the cgroup2.
  1153. * * 1, if the *skb* task does not belong to the cgroup2.
  1154. * * A negative error code, if an error occurred.
  1155. *
  1156. * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
  1157. * Description
  1158. * Resize (trim or grow) the packet associated to *skb* to the
  1159. * new *len*. The *flags* are reserved for future usage, and must
  1160. * be left at zero.
  1161. *
  1162. * The basic idea is that the helper performs the needed work to
  1163. * change the size of the packet, then the eBPF program rewrites
  1164. * the rest via helpers like **bpf_skb_store_bytes**\ (),
  1165. * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
  1166. * and others. This helper is a slow path utility intended for
  1167. * replies with control messages. And because it is targeted for
  1168. * slow path, the helper itself can afford to be slow: it
  1169. * implicitly linearizes, unclones and drops offloads from the
  1170. * *skb*.
  1171. *
  1172. * A call to this helper is susceptible to change the underlaying
  1173. * packet buffer. Therefore, at load time, all checks on pointers
  1174. * previously done by the verifier are invalidated and must be
  1175. * performed again, if the helper is used in combination with
  1176. * direct packet access.
  1177. * Return
  1178. * 0 on success, or a negative error in case of failure.
  1179. *
  1180. * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
  1181. * Description
  1182. * Pull in non-linear data in case the *skb* is non-linear and not
  1183. * all of *len* are part of the linear section. Make *len* bytes
  1184. * from *skb* readable and writable. If a zero value is passed for
  1185. * *len*, then the whole length of the *skb* is pulled.
  1186. *
  1187. * This helper is only needed for reading and writing with direct
  1188. * packet access.
  1189. *
  1190. * For direct packet access, testing that offsets to access
  1191. * are within packet boundaries (test on *skb*\ **->data_end**) is
  1192. * susceptible to fail if offsets are invalid, or if the requested
  1193. * data is in non-linear parts of the *skb*. On failure the
  1194. * program can just bail out, or in the case of a non-linear
  1195. * buffer, use a helper to make the data available. The
  1196. * **bpf_skb_load_bytes**\ () helper is a first solution to access
  1197. * the data. Another one consists in using **bpf_skb_pull_data**
  1198. * to pull in once the non-linear parts, then retesting and
  1199. * eventually access the data.
  1200. *
  1201. * At the same time, this also makes sure the *skb* is uncloned,
  1202. * which is a necessary condition for direct write. As this needs
  1203. * to be an invariant for the write part only, the verifier
  1204. * detects writes and adds a prologue that is calling
  1205. * **bpf_skb_pull_data()** to effectively unclone the *skb* from
  1206. * the very beginning in case it is indeed cloned.
  1207. *
  1208. * A call to this helper is susceptible to change the underlaying
  1209. * packet buffer. Therefore, at load time, all checks on pointers
  1210. * previously done by the verifier are invalidated and must be
  1211. * performed again, if the helper is used in combination with
  1212. * direct packet access.
  1213. * Return
  1214. * 0 on success, or a negative error in case of failure.
  1215. *
  1216. * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
  1217. * Description
  1218. * Add the checksum *csum* into *skb*\ **->csum** in case the
  1219. * driver has supplied a checksum for the entire packet into that
  1220. * field. Return an error otherwise. This helper is intended to be
  1221. * used in combination with **bpf_csum_diff**\ (), in particular
  1222. * when the checksum needs to be updated after data has been
  1223. * written into the packet through direct packet access.
  1224. * Return
  1225. * The checksum on success, or a negative error code in case of
  1226. * failure.
  1227. *
  1228. * void bpf_set_hash_invalid(struct sk_buff *skb)
  1229. * Description
  1230. * Invalidate the current *skb*\ **->hash**. It can be used after
  1231. * mangling on headers through direct packet access, in order to
  1232. * indicate that the hash is outdated and to trigger a
  1233. * recalculation the next time the kernel tries to access this
  1234. * hash or when the **bpf_get_hash_recalc**\ () helper is called.
  1235. *
  1236. * int bpf_get_numa_node_id(void)
  1237. * Description
  1238. * Return the id of the current NUMA node. The primary use case
  1239. * for this helper is the selection of sockets for the local NUMA
  1240. * node, when the program is attached to sockets using the
  1241. * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
  1242. * but the helper is also available to other eBPF program types,
  1243. * similarly to **bpf_get_smp_processor_id**\ ().
  1244. * Return
  1245. * The id of current NUMA node.
  1246. *
  1247. * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
  1248. * Description
  1249. * Grows headroom of packet associated to *skb* and adjusts the
  1250. * offset of the MAC header accordingly, adding *len* bytes of
  1251. * space. It automatically extends and reallocates memory as
  1252. * required.
  1253. *
  1254. * This helper can be used on a layer 3 *skb* to push a MAC header
  1255. * for redirection into a layer 2 device.
  1256. *
  1257. * All values for *flags* are reserved for future usage, and must
  1258. * be left at zero.
  1259. *
  1260. * A call to this helper is susceptible to change the underlaying
  1261. * packet buffer. Therefore, at load time, all checks on pointers
  1262. * previously done by the verifier are invalidated and must be
  1263. * performed again, if the helper is used in combination with
  1264. * direct packet access.
  1265. * Return
  1266. * 0 on success, or a negative error in case of failure.
  1267. *
  1268. * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
  1269. * Description
  1270. * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
  1271. * it is possible to use a negative value for *delta*. This helper
  1272. * can be used to prepare the packet for pushing or popping
  1273. * headers.
  1274. *
  1275. * A call to this helper is susceptible to change the underlaying
  1276. * packet buffer. Therefore, at load time, all checks on pointers
  1277. * previously done by the verifier are invalidated and must be
  1278. * performed again, if the helper is used in combination with
  1279. * direct packet access.
  1280. * Return
  1281. * 0 on success, or a negative error in case of failure.
  1282. *
  1283. * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
  1284. * Description
  1285. * Copy a NUL terminated string from an unsafe address
  1286. * *unsafe_ptr* to *dst*. The *size* should include the
  1287. * terminating NUL byte. In case the string length is smaller than
  1288. * *size*, the target is not padded with further NUL bytes. If the
  1289. * string length is larger than *size*, just *size*-1 bytes are
  1290. * copied and the last byte is set to NUL.
  1291. *
  1292. * On success, the length of the copied string is returned. This
  1293. * makes this helper useful in tracing programs for reading
  1294. * strings, and more importantly to get its length at runtime. See
  1295. * the following snippet:
  1296. *
  1297. * ::
  1298. *
  1299. * SEC("kprobe/sys_open")
  1300. * void bpf_sys_open(struct pt_regs *ctx)
  1301. * {
  1302. * char buf[PATHLEN]; // PATHLEN is defined to 256
  1303. * int res = bpf_probe_read_str(buf, sizeof(buf),
  1304. * ctx->di);
  1305. *
  1306. * // Consume buf, for example push it to
  1307. * // userspace via bpf_perf_event_output(); we
  1308. * // can use res (the string length) as event
  1309. * // size, after checking its boundaries.
  1310. * }
  1311. *
  1312. * In comparison, using **bpf_probe_read()** helper here instead
  1313. * to read the string would require to estimate the length at
  1314. * compile time, and would often result in copying more memory
  1315. * than necessary.
  1316. *
  1317. * Another useful use case is when parsing individual process
  1318. * arguments or individual environment variables navigating
  1319. * *current*\ **->mm->arg_start** and *current*\
  1320. * **->mm->env_start**: using this helper and the return value,
  1321. * one can quickly iterate at the right offset of the memory area.
  1322. * Return
  1323. * On success, the strictly positive length of the string,
  1324. * including the trailing NUL character. On error, a negative
  1325. * value.
  1326. *
  1327. * u64 bpf_get_socket_cookie(struct sk_buff *skb)
  1328. * Description
  1329. * If the **struct sk_buff** pointed by *skb* has a known socket,
  1330. * retrieve the cookie (generated by the kernel) of this socket.
  1331. * If no cookie has been set yet, generate a new cookie. Once
  1332. * generated, the socket cookie remains stable for the life of the
  1333. * socket. This helper can be useful for monitoring per socket
  1334. * networking traffic statistics as it provides a unique socket
  1335. * identifier per namespace.
  1336. * Return
  1337. * A 8-byte long non-decreasing number on success, or 0 if the
  1338. * socket field is missing inside *skb*.
  1339. *
  1340. * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
  1341. * Description
  1342. * Equivalent to bpf_get_socket_cookie() helper that accepts
  1343. * *skb*, but gets socket from **struct bpf_sock_addr** contex.
  1344. * Return
  1345. * A 8-byte long non-decreasing number.
  1346. *
  1347. * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
  1348. * Description
  1349. * Equivalent to bpf_get_socket_cookie() helper that accepts
  1350. * *skb*, but gets socket from **struct bpf_sock_ops** contex.
  1351. * Return
  1352. * A 8-byte long non-decreasing number.
  1353. *
  1354. * u32 bpf_get_socket_uid(struct sk_buff *skb)
  1355. * Return
  1356. * The owner UID of the socket associated to *skb*. If the socket
  1357. * is **NULL**, or if it is not a full socket (i.e. if it is a
  1358. * time-wait or a request socket instead), **overflowuid** value
  1359. * is returned (note that **overflowuid** might also be the actual
  1360. * UID value for the socket).
  1361. *
  1362. * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
  1363. * Description
  1364. * Set the full hash for *skb* (set the field *skb*\ **->hash**)
  1365. * to value *hash*.
  1366. * Return
  1367. * 0
  1368. *
  1369. * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
  1370. * Description
  1371. * Emulate a call to **setsockopt()** on the socket associated to
  1372. * *bpf_socket*, which must be a full socket. The *level* at
  1373. * which the option resides and the name *optname* of the option
  1374. * must be specified, see **setsockopt(2)** for more information.
  1375. * The option value of length *optlen* is pointed by *optval*.
  1376. *
  1377. * This helper actually implements a subset of **setsockopt()**.
  1378. * It supports the following *level*\ s:
  1379. *
  1380. * * **SOL_SOCKET**, which supports the following *optname*\ s:
  1381. * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
  1382. * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
  1383. * * **IPPROTO_TCP**, which supports the following *optname*\ s:
  1384. * **TCP_CONGESTION**, **TCP_BPF_IW**,
  1385. * **TCP_BPF_SNDCWND_CLAMP**.
  1386. * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
  1387. * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
  1388. * Return
  1389. * 0 on success, or a negative error in case of failure.
  1390. *
  1391. * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags)
  1392. * Description
  1393. * Grow or shrink the room for data in the packet associated to
  1394. * *skb* by *len_diff*, and according to the selected *mode*.
  1395. *
  1396. * There is a single supported mode at this time:
  1397. *
  1398. * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  1399. * (room space is added or removed below the layer 3 header).
  1400. *
  1401. * All values for *flags* are reserved for future usage, and must
  1402. * be left at zero.
  1403. *
  1404. * A call to this helper is susceptible to change the underlaying
  1405. * packet buffer. Therefore, at load time, all checks on pointers
  1406. * previously done by the verifier are invalidated and must be
  1407. * performed again, if the helper is used in combination with
  1408. * direct packet access.
  1409. * Return
  1410. * 0 on success, or a negative error in case of failure.
  1411. *
  1412. * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
  1413. * Description
  1414. * Redirect the packet to the endpoint referenced by *map* at
  1415. * index *key*. Depending on its type, this *map* can contain
  1416. * references to net devices (for forwarding packets through other
  1417. * ports), or to CPUs (for redirecting XDP frames to another CPU;
  1418. * but this is only implemented for native XDP (with driver
  1419. * support) as of this writing).
  1420. *
  1421. * All values for *flags* are reserved for future usage, and must
  1422. * be left at zero.
  1423. *
  1424. * When used to redirect packets to net devices, this helper
  1425. * provides a high performance increase over **bpf_redirect**\ ().
  1426. * This is due to various implementation details of the underlying
  1427. * mechanisms, one of which is the fact that **bpf_redirect_map**\
  1428. * () tries to send packet as a "bulk" to the device.
  1429. * Return
  1430. * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
  1431. *
  1432. * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
  1433. * Description
  1434. * Redirect the packet to the socket referenced by *map* (of type
  1435. * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  1436. * egress interfaces can be used for redirection. The
  1437. * **BPF_F_INGRESS** value in *flags* is used to make the
  1438. * distinction (ingress path is selected if the flag is present,
  1439. * egress path otherwise). This is the only flag supported for now.
  1440. * Return
  1441. * **SK_PASS** on success, or **SK_DROP** on error.
  1442. *
  1443. * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
  1444. * Description
  1445. * Add an entry to, or update a *map* referencing sockets. The
  1446. * *skops* is used as a new value for the entry associated to
  1447. * *key*. *flags* is one of:
  1448. *
  1449. * **BPF_NOEXIST**
  1450. * The entry for *key* must not exist in the map.
  1451. * **BPF_EXIST**
  1452. * The entry for *key* must already exist in the map.
  1453. * **BPF_ANY**
  1454. * No condition on the existence of the entry for *key*.
  1455. *
  1456. * If the *map* has eBPF programs (parser and verdict), those will
  1457. * be inherited by the socket being added. If the socket is
  1458. * already attached to eBPF programs, this results in an error.
  1459. * Return
  1460. * 0 on success, or a negative error in case of failure.
  1461. *
  1462. * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
  1463. * Description
  1464. * Adjust the address pointed by *xdp_md*\ **->data_meta** by
  1465. * *delta* (which can be positive or negative). Note that this
  1466. * operation modifies the address stored in *xdp_md*\ **->data**,
  1467. * so the latter must be loaded only after the helper has been
  1468. * called.
  1469. *
  1470. * The use of *xdp_md*\ **->data_meta** is optional and programs
  1471. * are not required to use it. The rationale is that when the
  1472. * packet is processed with XDP (e.g. as DoS filter), it is
  1473. * possible to push further meta data along with it before passing
  1474. * to the stack, and to give the guarantee that an ingress eBPF
  1475. * program attached as a TC classifier on the same device can pick
  1476. * this up for further post-processing. Since TC works with socket
  1477. * buffers, it remains possible to set from XDP the **mark** or
  1478. * **priority** pointers, or other pointers for the socket buffer.
  1479. * Having this scratch space generic and programmable allows for
  1480. * more flexibility as the user is free to store whatever meta
  1481. * data they need.
  1482. *
  1483. * A call to this helper is susceptible to change the underlaying
  1484. * packet buffer. Therefore, at load time, all checks on pointers
  1485. * previously done by the verifier are invalidated and must be
  1486. * performed again, if the helper is used in combination with
  1487. * direct packet access.
  1488. * Return
  1489. * 0 on success, or a negative error in case of failure.
  1490. *
  1491. * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
  1492. * Description
  1493. * Read the value of a perf event counter, and store it into *buf*
  1494. * of size *buf_size*. This helper relies on a *map* of type
  1495. * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
  1496. * counter is selected when *map* is updated with perf event file
  1497. * descriptors. The *map* is an array whose size is the number of
  1498. * available CPUs, and each cell contains a value relative to one
  1499. * CPU. The value to retrieve is indicated by *flags*, that
  1500. * contains the index of the CPU to look up, masked with
  1501. * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  1502. * **BPF_F_CURRENT_CPU** to indicate that the value for the
  1503. * current CPU should be retrieved.
  1504. *
  1505. * This helper behaves in a way close to
  1506. * **bpf_perf_event_read**\ () helper, save that instead of
  1507. * just returning the value observed, it fills the *buf*
  1508. * structure. This allows for additional data to be retrieved: in
  1509. * particular, the enabled and running times (in *buf*\
  1510. * **->enabled** and *buf*\ **->running**, respectively) are
  1511. * copied. In general, **bpf_perf_event_read_value**\ () is
  1512. * recommended over **bpf_perf_event_read**\ (), which has some
  1513. * ABI issues and provides fewer functionalities.
  1514. *
  1515. * These values are interesting, because hardware PMU (Performance
  1516. * Monitoring Unit) counters are limited resources. When there are
  1517. * more PMU based perf events opened than available counters,
  1518. * kernel will multiplex these events so each event gets certain
  1519. * percentage (but not all) of the PMU time. In case that
  1520. * multiplexing happens, the number of samples or counter value
  1521. * will not reflect the case compared to when no multiplexing
  1522. * occurs. This makes comparison between different runs difficult.
  1523. * Typically, the counter value should be normalized before
  1524. * comparing to other experiments. The usual normalization is done
  1525. * as follows.
  1526. *
  1527. * ::
  1528. *
  1529. * normalized_counter = counter * t_enabled / t_running
  1530. *
  1531. * Where t_enabled is the time enabled for event and t_running is
  1532. * the time running for event since last normalization. The
  1533. * enabled and running times are accumulated since the perf event
  1534. * open. To achieve scaling factor between two invocations of an
  1535. * eBPF program, users can can use CPU id as the key (which is
  1536. * typical for perf array usage model) to remember the previous
  1537. * value and do the calculation inside the eBPF program.
  1538. * Return
  1539. * 0 on success, or a negative error in case of failure.
  1540. *
  1541. * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
  1542. * Description
  1543. * For en eBPF program attached to a perf event, retrieve the
  1544. * value of the event counter associated to *ctx* and store it in
  1545. * the structure pointed by *buf* and of size *buf_size*. Enabled
  1546. * and running times are also stored in the structure (see
  1547. * description of helper **bpf_perf_event_read_value**\ () for
  1548. * more details).
  1549. * Return
  1550. * 0 on success, or a negative error in case of failure.
  1551. *
  1552. * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
  1553. * Description
  1554. * Emulate a call to **getsockopt()** on the socket associated to
  1555. * *bpf_socket*, which must be a full socket. The *level* at
  1556. * which the option resides and the name *optname* of the option
  1557. * must be specified, see **getsockopt(2)** for more information.
  1558. * The retrieved value is stored in the structure pointed by
  1559. * *opval* and of length *optlen*.
  1560. *
  1561. * This helper actually implements a subset of **getsockopt()**.
  1562. * It supports the following *level*\ s:
  1563. *
  1564. * * **IPPROTO_TCP**, which supports *optname*
  1565. * **TCP_CONGESTION**.
  1566. * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
  1567. * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
  1568. * Return
  1569. * 0 on success, or a negative error in case of failure.
  1570. *
  1571. * int bpf_override_return(struct pt_reg *regs, u64 rc)
  1572. * Description
  1573. * Used for error injection, this helper uses kprobes to override
  1574. * the return value of the probed function, and to set it to *rc*.
  1575. * The first argument is the context *regs* on which the kprobe
  1576. * works.
  1577. *
  1578. * This helper works by setting setting the PC (program counter)
  1579. * to an override function which is run in place of the original
  1580. * probed function. This means the probed function is not run at
  1581. * all. The replacement function just returns with the required
  1582. * value.
  1583. *
  1584. * This helper has security implications, and thus is subject to
  1585. * restrictions. It is only available if the kernel was compiled
  1586. * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
  1587. * option, and in this case it only works on functions tagged with
  1588. * **ALLOW_ERROR_INJECTION** in the kernel code.
  1589. *
  1590. * Also, the helper is only available for the architectures having
  1591. * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
  1592. * x86 architecture is the only one to support this feature.
  1593. * Return
  1594. * 0
  1595. *
  1596. * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
  1597. * Description
  1598. * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
  1599. * for the full TCP socket associated to *bpf_sock_ops* to
  1600. * *argval*.
  1601. *
  1602. * The primary use of this field is to determine if there should
  1603. * be calls to eBPF programs of type
  1604. * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
  1605. * code. A program of the same type can change its value, per
  1606. * connection and as necessary, when the connection is
  1607. * established. This field is directly accessible for reading, but
  1608. * this helper must be used for updates in order to return an
  1609. * error if an eBPF program tries to set a callback that is not
  1610. * supported in the current kernel.
  1611. *
  1612. * The supported callback values that *argval* can combine are:
  1613. *
  1614. * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
  1615. * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
  1616. * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
  1617. *
  1618. * Here are some examples of where one could call such eBPF
  1619. * program:
  1620. *
  1621. * * When RTO fires.
  1622. * * When a packet is retransmitted.
  1623. * * When the connection terminates.
  1624. * * When a packet is sent.
  1625. * * When a packet is received.
  1626. * Return
  1627. * Code **-EINVAL** if the socket is not a full TCP socket;
  1628. * otherwise, a positive number containing the bits that could not
  1629. * be set is returned (which comes down to 0 if all bits were set
  1630. * as required).
  1631. *
  1632. * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
  1633. * Description
  1634. * This helper is used in programs implementing policies at the
  1635. * socket level. If the message *msg* is allowed to pass (i.e. if
  1636. * the verdict eBPF program returns **SK_PASS**), redirect it to
  1637. * the socket referenced by *map* (of type
  1638. * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  1639. * egress interfaces can be used for redirection. The
  1640. * **BPF_F_INGRESS** value in *flags* is used to make the
  1641. * distinction (ingress path is selected if the flag is present,
  1642. * egress path otherwise). This is the only flag supported for now.
  1643. * Return
  1644. * **SK_PASS** on success, or **SK_DROP** on error.
  1645. *
  1646. * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
  1647. * Description
  1648. * For socket policies, apply the verdict of the eBPF program to
  1649. * the next *bytes* (number of bytes) of message *msg*.
  1650. *
  1651. * For example, this helper can be used in the following cases:
  1652. *
  1653. * * A single **sendmsg**\ () or **sendfile**\ () system call
  1654. * contains multiple logical messages that the eBPF program is
  1655. * supposed to read and for which it should apply a verdict.
  1656. * * An eBPF program only cares to read the first *bytes* of a
  1657. * *msg*. If the message has a large payload, then setting up
  1658. * and calling the eBPF program repeatedly for all bytes, even
  1659. * though the verdict is already known, would create unnecessary
  1660. * overhead.
  1661. *
  1662. * When called from within an eBPF program, the helper sets a
  1663. * counter internal to the BPF infrastructure, that is used to
  1664. * apply the last verdict to the next *bytes*. If *bytes* is
  1665. * smaller than the current data being processed from a
  1666. * **sendmsg**\ () or **sendfile**\ () system call, the first
  1667. * *bytes* will be sent and the eBPF program will be re-run with
  1668. * the pointer for start of data pointing to byte number *bytes*
  1669. * **+ 1**. If *bytes* is larger than the current data being
  1670. * processed, then the eBPF verdict will be applied to multiple
  1671. * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
  1672. * consumed.
  1673. *
  1674. * Note that if a socket closes with the internal counter holding
  1675. * a non-zero value, this is not a problem because data is not
  1676. * being buffered for *bytes* and is sent as it is received.
  1677. * Return
  1678. * 0
  1679. *
  1680. * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
  1681. * Description
  1682. * For socket policies, prevent the execution of the verdict eBPF
  1683. * program for message *msg* until *bytes* (byte number) have been
  1684. * accumulated.
  1685. *
  1686. * This can be used when one needs a specific number of bytes
  1687. * before a verdict can be assigned, even if the data spans
  1688. * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
  1689. * case would be a user calling **sendmsg**\ () repeatedly with
  1690. * 1-byte long message segments. Obviously, this is bad for
  1691. * performance, but it is still valid. If the eBPF program needs
  1692. * *bytes* bytes to validate a header, this helper can be used to
  1693. * prevent the eBPF program to be called again until *bytes* have
  1694. * been accumulated.
  1695. * Return
  1696. * 0
  1697. *
  1698. * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
  1699. * Description
  1700. * For socket policies, pull in non-linear data from user space
  1701. * for *msg* and set pointers *msg*\ **->data** and *msg*\
  1702. * **->data_end** to *start* and *end* bytes offsets into *msg*,
  1703. * respectively.
  1704. *
  1705. * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
  1706. * *msg* it can only parse data that the (**data**, **data_end**)
  1707. * pointers have already consumed. For **sendmsg**\ () hooks this
  1708. * is likely the first scatterlist element. But for calls relying
  1709. * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
  1710. * be the range (**0**, **0**) because the data is shared with
  1711. * user space and by default the objective is to avoid allowing
  1712. * user space to modify data while (or after) eBPF verdict is
  1713. * being decided. This helper can be used to pull in data and to
  1714. * set the start and end pointer to given values. Data will be
  1715. * copied if necessary (i.e. if data was not linear and if start
  1716. * and end pointers do not point to the same chunk).
  1717. *
  1718. * A call to this helper is susceptible to change the underlaying
  1719. * packet buffer. Therefore, at load time, all checks on pointers
  1720. * previously done by the verifier are invalidated and must be
  1721. * performed again, if the helper is used in combination with
  1722. * direct packet access.
  1723. *
  1724. * All values for *flags* are reserved for future usage, and must
  1725. * be left at zero.
  1726. * Return
  1727. * 0 on success, or a negative error in case of failure.
  1728. *
  1729. * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
  1730. * Description
  1731. * Bind the socket associated to *ctx* to the address pointed by
  1732. * *addr*, of length *addr_len*. This allows for making outgoing
  1733. * connection from the desired IP address, which can be useful for
  1734. * example when all processes inside a cgroup should use one
  1735. * single IP address on a host that has multiple IP configured.
  1736. *
  1737. * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
  1738. * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
  1739. * **AF_INET6**). Looking for a free port to bind to can be
  1740. * expensive, therefore binding to port is not permitted by the
  1741. * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
  1742. * must be set to zero.
  1743. * Return
  1744. * 0 on success, or a negative error in case of failure.
  1745. *
  1746. * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
  1747. * Description
  1748. * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
  1749. * only possible to shrink the packet as of this writing,
  1750. * therefore *delta* must be a negative integer.
  1751. *
  1752. * A call to this helper is susceptible to change the underlaying
  1753. * packet buffer. Therefore, at load time, all checks on pointers
  1754. * previously done by the verifier are invalidated and must be
  1755. * performed again, if the helper is used in combination with
  1756. * direct packet access.
  1757. * Return
  1758. * 0 on success, or a negative error in case of failure.
  1759. *
  1760. * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
  1761. * Description
  1762. * Retrieve the XFRM state (IP transform framework, see also
  1763. * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
  1764. *
  1765. * The retrieved value is stored in the **struct bpf_xfrm_state**
  1766. * pointed by *xfrm_state* and of length *size*.
  1767. *
  1768. * All values for *flags* are reserved for future usage, and must
  1769. * be left at zero.
  1770. *
  1771. * This helper is available only if the kernel was compiled with
  1772. * **CONFIG_XFRM** configuration option.
  1773. * Return
  1774. * 0 on success, or a negative error in case of failure.
  1775. *
  1776. * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
  1777. * Description
  1778. * Return a user or a kernel stack in bpf program provided buffer.
  1779. * To achieve this, the helper needs *ctx*, which is a pointer
  1780. * to the context on which the tracing program is executed.
  1781. * To store the stacktrace, the bpf program provides *buf* with
  1782. * a nonnegative *size*.
  1783. *
  1784. * The last argument, *flags*, holds the number of stack frames to
  1785. * skip (from 0 to 255), masked with
  1786. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  1787. * the following flags:
  1788. *
  1789. * **BPF_F_USER_STACK**
  1790. * Collect a user space stack instead of a kernel stack.
  1791. * **BPF_F_USER_BUILD_ID**
  1792. * Collect buildid+offset instead of ips for user stack,
  1793. * only valid if **BPF_F_USER_STACK** is also specified.
  1794. *
  1795. * **bpf_get_stack**\ () can collect up to
  1796. * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
  1797. * to sufficient large buffer size. Note that
  1798. * this limit can be controlled with the **sysctl** program, and
  1799. * that it should be manually increased in order to profile long
  1800. * user stacks (such as stacks for Java programs). To do so, use:
  1801. *
  1802. * ::
  1803. *
  1804. * # sysctl kernel.perf_event_max_stack=<new value>
  1805. * Return
  1806. * A non-negative value equal to or less than *size* on success,
  1807. * or a negative error in case of failure.
  1808. *
  1809. * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
  1810. * Description
  1811. * This helper is similar to **bpf_skb_load_bytes**\ () in that
  1812. * it provides an easy way to load *len* bytes from *offset*
  1813. * from the packet associated to *skb*, into the buffer pointed
  1814. * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
  1815. * a fifth argument *start_header* exists in order to select a
  1816. * base offset to start from. *start_header* can be one of:
  1817. *
  1818. * **BPF_HDR_START_MAC**
  1819. * Base offset to load data from is *skb*'s mac header.
  1820. * **BPF_HDR_START_NET**
  1821. * Base offset to load data from is *skb*'s network header.
  1822. *
  1823. * In general, "direct packet access" is the preferred method to
  1824. * access packet data, however, this helper is in particular useful
  1825. * in socket filters where *skb*\ **->data** does not always point
  1826. * to the start of the mac header and where "direct packet access"
  1827. * is not available.
  1828. * Return
  1829. * 0 on success, or a negative error in case of failure.
  1830. *
  1831. * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
  1832. * Description
  1833. * Do FIB lookup in kernel tables using parameters in *params*.
  1834. * If lookup is successful and result shows packet is to be
  1835. * forwarded, the neighbor tables are searched for the nexthop.
  1836. * If successful (ie., FIB lookup shows forwarding and nexthop
  1837. * is resolved), the nexthop address is returned in ipv4_dst
  1838. * or ipv6_dst based on family, smac is set to mac address of
  1839. * egress device, dmac is set to nexthop mac address, rt_metric
  1840. * is set to metric from route (IPv4/IPv6 only), and ifindex
  1841. * is set to the device index of the nexthop from the FIB lookup.
  1842. *
  1843. * *plen* argument is the size of the passed in struct.
  1844. * *flags* argument can be a combination of one or more of the
  1845. * following values:
  1846. *
  1847. * **BPF_FIB_LOOKUP_DIRECT**
  1848. * Do a direct table lookup vs full lookup using FIB
  1849. * rules.
  1850. * **BPF_FIB_LOOKUP_OUTPUT**
  1851. * Perform lookup from an egress perspective (default is
  1852. * ingress).
  1853. *
  1854. * *ctx* is either **struct xdp_md** for XDP programs or
  1855. * **struct sk_buff** tc cls_act programs.
  1856. * Return
  1857. * * < 0 if any input argument is invalid
  1858. * * 0 on success (packet is forwarded, nexthop neighbor exists)
  1859. * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
  1860. * packet is not forwarded or needs assist from full stack
  1861. *
  1862. * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
  1863. * Description
  1864. * Add an entry to, or update a sockhash *map* referencing sockets.
  1865. * The *skops* is used as a new value for the entry associated to
  1866. * *key*. *flags* is one of:
  1867. *
  1868. * **BPF_NOEXIST**
  1869. * The entry for *key* must not exist in the map.
  1870. * **BPF_EXIST**
  1871. * The entry for *key* must already exist in the map.
  1872. * **BPF_ANY**
  1873. * No condition on the existence of the entry for *key*.
  1874. *
  1875. * If the *map* has eBPF programs (parser and verdict), those will
  1876. * be inherited by the socket being added. If the socket is
  1877. * already attached to eBPF programs, this results in an error.
  1878. * Return
  1879. * 0 on success, or a negative error in case of failure.
  1880. *
  1881. * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
  1882. * Description
  1883. * This helper is used in programs implementing policies at the
  1884. * socket level. If the message *msg* is allowed to pass (i.e. if
  1885. * the verdict eBPF program returns **SK_PASS**), redirect it to
  1886. * the socket referenced by *map* (of type
  1887. * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  1888. * egress interfaces can be used for redirection. The
  1889. * **BPF_F_INGRESS** value in *flags* is used to make the
  1890. * distinction (ingress path is selected if the flag is present,
  1891. * egress path otherwise). This is the only flag supported for now.
  1892. * Return
  1893. * **SK_PASS** on success, or **SK_DROP** on error.
  1894. *
  1895. * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
  1896. * Description
  1897. * This helper is used in programs implementing policies at the
  1898. * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
  1899. * if the verdeict eBPF program returns **SK_PASS**), redirect it
  1900. * to the socket referenced by *map* (of type
  1901. * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  1902. * egress interfaces can be used for redirection. The
  1903. * **BPF_F_INGRESS** value in *flags* is used to make the
  1904. * distinction (ingress path is selected if the flag is present,
  1905. * egress otherwise). This is the only flag supported for now.
  1906. * Return
  1907. * **SK_PASS** on success, or **SK_DROP** on error.
  1908. *
  1909. * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
  1910. * Description
  1911. * Encapsulate the packet associated to *skb* within a Layer 3
  1912. * protocol header. This header is provided in the buffer at
  1913. * address *hdr*, with *len* its size in bytes. *type* indicates
  1914. * the protocol of the header and can be one of:
  1915. *
  1916. * **BPF_LWT_ENCAP_SEG6**
  1917. * IPv6 encapsulation with Segment Routing Header
  1918. * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
  1919. * the IPv6 header is computed by the kernel.
  1920. * **BPF_LWT_ENCAP_SEG6_INLINE**
  1921. * Only works if *skb* contains an IPv6 packet. Insert a
  1922. * Segment Routing Header (**struct ipv6_sr_hdr**) inside
  1923. * the IPv6 header.
  1924. *
  1925. * A call to this helper is susceptible to change the underlaying
  1926. * packet buffer. Therefore, at load time, all checks on pointers
  1927. * previously done by the verifier are invalidated and must be
  1928. * performed again, if the helper is used in combination with
  1929. * direct packet access.
  1930. * Return
  1931. * 0 on success, or a negative error in case of failure.
  1932. *
  1933. * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
  1934. * Description
  1935. * Store *len* bytes from address *from* into the packet
  1936. * associated to *skb*, at *offset*. Only the flags, tag and TLVs
  1937. * inside the outermost IPv6 Segment Routing Header can be
  1938. * modified through this helper.
  1939. *
  1940. * A call to this helper is susceptible to change the underlaying
  1941. * packet buffer. Therefore, at load time, all checks on pointers
  1942. * previously done by the verifier are invalidated and must be
  1943. * performed again, if the helper is used in combination with
  1944. * direct packet access.
  1945. * Return
  1946. * 0 on success, or a negative error in case of failure.
  1947. *
  1948. * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
  1949. * Description
  1950. * Adjust the size allocated to TLVs in the outermost IPv6
  1951. * Segment Routing Header contained in the packet associated to
  1952. * *skb*, at position *offset* by *delta* bytes. Only offsets
  1953. * after the segments are accepted. *delta* can be as well
  1954. * positive (growing) as negative (shrinking).
  1955. *
  1956. * A call to this helper is susceptible to change the underlaying
  1957. * packet buffer. Therefore, at load time, all checks on pointers
  1958. * previously done by the verifier are invalidated and must be
  1959. * performed again, if the helper is used in combination with
  1960. * direct packet access.
  1961. * Return
  1962. * 0 on success, or a negative error in case of failure.
  1963. *
  1964. * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
  1965. * Description
  1966. * Apply an IPv6 Segment Routing action of type *action* to the
  1967. * packet associated to *skb*. Each action takes a parameter
  1968. * contained at address *param*, and of length *param_len* bytes.
  1969. * *action* can be one of:
  1970. *
  1971. * **SEG6_LOCAL_ACTION_END_X**
  1972. * End.X action: Endpoint with Layer-3 cross-connect.
  1973. * Type of *param*: **struct in6_addr**.
  1974. * **SEG6_LOCAL_ACTION_END_T**
  1975. * End.T action: Endpoint with specific IPv6 table lookup.
  1976. * Type of *param*: **int**.
  1977. * **SEG6_LOCAL_ACTION_END_B6**
  1978. * End.B6 action: Endpoint bound to an SRv6 policy.
  1979. * Type of param: **struct ipv6_sr_hdr**.
  1980. * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
  1981. * End.B6.Encap action: Endpoint bound to an SRv6
  1982. * encapsulation policy.
  1983. * Type of param: **struct ipv6_sr_hdr**.
  1984. *
  1985. * A call to this helper is susceptible to change the underlaying
  1986. * packet buffer. Therefore, at load time, all checks on pointers
  1987. * previously done by the verifier are invalidated and must be
  1988. * performed again, if the helper is used in combination with
  1989. * direct packet access.
  1990. * Return
  1991. * 0 on success, or a negative error in case of failure.
  1992. *
  1993. * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
  1994. * Description
  1995. * This helper is used in programs implementing IR decoding, to
  1996. * report a successfully decoded key press with *scancode*,
  1997. * *toggle* value in the given *protocol*. The scancode will be
  1998. * translated to a keycode using the rc keymap, and reported as
  1999. * an input key down event. After a period a key up event is
  2000. * generated. This period can be extended by calling either
  2001. * **bpf_rc_keydown** () again with the same values, or calling
  2002. * **bpf_rc_repeat** ().
  2003. *
  2004. * Some protocols include a toggle bit, in case the button was
  2005. * released and pressed again between consecutive scancodes.
  2006. *
  2007. * The *ctx* should point to the lirc sample as passed into
  2008. * the program.
  2009. *
  2010. * The *protocol* is the decoded protocol number (see
  2011. * **enum rc_proto** for some predefined values).
  2012. *
  2013. * This helper is only available is the kernel was compiled with
  2014. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  2015. * "**y**".
  2016. * Return
  2017. * 0
  2018. *
  2019. * int bpf_rc_repeat(void *ctx)
  2020. * Description
  2021. * This helper is used in programs implementing IR decoding, to
  2022. * report a successfully decoded repeat key message. This delays
  2023. * the generation of a key up event for previously generated
  2024. * key down event.
  2025. *
  2026. * Some IR protocols like NEC have a special IR message for
  2027. * repeating last button, for when a button is held down.
  2028. *
  2029. * The *ctx* should point to the lirc sample as passed into
  2030. * the program.
  2031. *
  2032. * This helper is only available is the kernel was compiled with
  2033. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  2034. * "**y**".
  2035. * Return
  2036. * 0
  2037. *
  2038. * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
  2039. * Description
  2040. * Return the cgroup v2 id of the socket associated with the *skb*.
  2041. * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
  2042. * helper for cgroup v1 by providing a tag resp. identifier that
  2043. * can be matched on or used for map lookups e.g. to implement
  2044. * policy. The cgroup v2 id of a given path in the hierarchy is
  2045. * exposed in user space through the f_handle API in order to get
  2046. * to the same 64-bit id.
  2047. *
  2048. * This helper can be used on TC egress path, but not on ingress,
  2049. * and is available only if the kernel was compiled with the
  2050. * **CONFIG_SOCK_CGROUP_DATA** configuration option.
  2051. * Return
  2052. * The id is returned or 0 in case the id could not be retrieved.
  2053. *
  2054. * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
  2055. * Description
  2056. * Return id of cgroup v2 that is ancestor of cgroup associated
  2057. * with the *skb* at the *ancestor_level*. The root cgroup is at
  2058. * *ancestor_level* zero and each step down the hierarchy
  2059. * increments the level. If *ancestor_level* == level of cgroup
  2060. * associated with *skb*, then return value will be same as that
  2061. * of **bpf_skb_cgroup_id**\ ().
  2062. *
  2063. * The helper is useful to implement policies based on cgroups
  2064. * that are upper in hierarchy than immediate cgroup associated
  2065. * with *skb*.
  2066. *
  2067. * The format of returned id and helper limitations are same as in
  2068. * **bpf_skb_cgroup_id**\ ().
  2069. * Return
  2070. * The id is returned or 0 in case the id could not be retrieved.
  2071. *
  2072. * u64 bpf_get_current_cgroup_id(void)
  2073. * Return
  2074. * A 64-bit integer containing the current cgroup id based
  2075. * on the cgroup within which the current task is running.
  2076. *
  2077. * void* get_local_storage(void *map, u64 flags)
  2078. * Description
  2079. * Get the pointer to the local storage area.
  2080. * The type and the size of the local storage is defined
  2081. * by the *map* argument.
  2082. * The *flags* meaning is specific for each map type,
  2083. * and has to be 0 for cgroup local storage.
  2084. *
  2085. * Depending on the bpf program type, a local storage area
  2086. * can be shared between multiple instances of the bpf program,
  2087. * running simultaneously.
  2088. *
  2089. * A user should care about the synchronization by himself.
  2090. * For example, by using the BPF_STX_XADD instruction to alter
  2091. * the shared data.
  2092. * Return
  2093. * Pointer to the local storage area.
  2094. *
  2095. * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
  2096. * Description
  2097. * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map
  2098. * It checks the selected sk is matching the incoming
  2099. * request in the skb.
  2100. * Return
  2101. * 0 on success, or a negative error in case of failure.
  2102. */
  2103. #define __BPF_FUNC_MAPPER(FN) \
  2104. FN(unspec), \
  2105. FN(map_lookup_elem), \
  2106. FN(map_update_elem), \
  2107. FN(map_delete_elem), \
  2108. FN(probe_read), \
  2109. FN(ktime_get_ns), \
  2110. FN(trace_printk), \
  2111. FN(get_prandom_u32), \
  2112. FN(get_smp_processor_id), \
  2113. FN(skb_store_bytes), \
  2114. FN(l3_csum_replace), \
  2115. FN(l4_csum_replace), \
  2116. FN(tail_call), \
  2117. FN(clone_redirect), \
  2118. FN(get_current_pid_tgid), \
  2119. FN(get_current_uid_gid), \
  2120. FN(get_current_comm), \
  2121. FN(get_cgroup_classid), \
  2122. FN(skb_vlan_push), \
  2123. FN(skb_vlan_pop), \
  2124. FN(skb_get_tunnel_key), \
  2125. FN(skb_set_tunnel_key), \
  2126. FN(perf_event_read), \
  2127. FN(redirect), \
  2128. FN(get_route_realm), \
  2129. FN(perf_event_output), \
  2130. FN(skb_load_bytes), \
  2131. FN(get_stackid), \
  2132. FN(csum_diff), \
  2133. FN(skb_get_tunnel_opt), \
  2134. FN(skb_set_tunnel_opt), \
  2135. FN(skb_change_proto), \
  2136. FN(skb_change_type), \
  2137. FN(skb_under_cgroup), \
  2138. FN(get_hash_recalc), \
  2139. FN(get_current_task), \
  2140. FN(probe_write_user), \
  2141. FN(current_task_under_cgroup), \
  2142. FN(skb_change_tail), \
  2143. FN(skb_pull_data), \
  2144. FN(csum_update), \
  2145. FN(set_hash_invalid), \
  2146. FN(get_numa_node_id), \
  2147. FN(skb_change_head), \
  2148. FN(xdp_adjust_head), \
  2149. FN(probe_read_str), \
  2150. FN(get_socket_cookie), \
  2151. FN(get_socket_uid), \
  2152. FN(set_hash), \
  2153. FN(setsockopt), \
  2154. FN(skb_adjust_room), \
  2155. FN(redirect_map), \
  2156. FN(sk_redirect_map), \
  2157. FN(sock_map_update), \
  2158. FN(xdp_adjust_meta), \
  2159. FN(perf_event_read_value), \
  2160. FN(perf_prog_read_value), \
  2161. FN(getsockopt), \
  2162. FN(override_return), \
  2163. FN(sock_ops_cb_flags_set), \
  2164. FN(msg_redirect_map), \
  2165. FN(msg_apply_bytes), \
  2166. FN(msg_cork_bytes), \
  2167. FN(msg_pull_data), \
  2168. FN(bind), \
  2169. FN(xdp_adjust_tail), \
  2170. FN(skb_get_xfrm_state), \
  2171. FN(get_stack), \
  2172. FN(skb_load_bytes_relative), \
  2173. FN(fib_lookup), \
  2174. FN(sock_hash_update), \
  2175. FN(msg_redirect_hash), \
  2176. FN(sk_redirect_hash), \
  2177. FN(lwt_push_encap), \
  2178. FN(lwt_seg6_store_bytes), \
  2179. FN(lwt_seg6_adjust_srh), \
  2180. FN(lwt_seg6_action), \
  2181. FN(rc_repeat), \
  2182. FN(rc_keydown), \
  2183. FN(skb_cgroup_id), \
  2184. FN(get_current_cgroup_id), \
  2185. FN(get_local_storage), \
  2186. FN(sk_select_reuseport), \
  2187. FN(skb_ancestor_cgroup_id),
  2188. /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  2189. * function eBPF program intends to call
  2190. */
  2191. #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
  2192. enum bpf_func_id {
  2193. __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
  2194. __BPF_FUNC_MAX_ID,
  2195. };
  2196. #undef __BPF_ENUM_FN
  2197. /* All flags used by eBPF helper functions, placed here. */
  2198. /* BPF_FUNC_skb_store_bytes flags. */
  2199. #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
  2200. #define BPF_F_INVALIDATE_HASH (1ULL << 1)
  2201. /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
  2202. * First 4 bits are for passing the header field size.
  2203. */
  2204. #define BPF_F_HDR_FIELD_MASK 0xfULL
  2205. /* BPF_FUNC_l4_csum_replace flags. */
  2206. #define BPF_F_PSEUDO_HDR (1ULL << 4)
  2207. #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
  2208. #define BPF_F_MARK_ENFORCE (1ULL << 6)
  2209. /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
  2210. #define BPF_F_INGRESS (1ULL << 0)
  2211. /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
  2212. #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
  2213. /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
  2214. #define BPF_F_SKIP_FIELD_MASK 0xffULL
  2215. #define BPF_F_USER_STACK (1ULL << 8)
  2216. /* flags used by BPF_FUNC_get_stackid only. */
  2217. #define BPF_F_FAST_STACK_CMP (1ULL << 9)
  2218. #define BPF_F_REUSE_STACKID (1ULL << 10)
  2219. /* flags used by BPF_FUNC_get_stack only. */
  2220. #define BPF_F_USER_BUILD_ID (1ULL << 11)
  2221. /* BPF_FUNC_skb_set_tunnel_key flags. */
  2222. #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
  2223. #define BPF_F_DONT_FRAGMENT (1ULL << 2)
  2224. #define BPF_F_SEQ_NUMBER (1ULL << 3)
  2225. /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
  2226. * BPF_FUNC_perf_event_read_value flags.
  2227. */
  2228. #define BPF_F_INDEX_MASK 0xffffffffULL
  2229. #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
  2230. /* BPF_FUNC_perf_event_output for sk_buff input context. */
  2231. #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
  2232. /* Mode for BPF_FUNC_skb_adjust_room helper. */
  2233. enum bpf_adj_room_mode {
  2234. BPF_ADJ_ROOM_NET,
  2235. };
  2236. /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
  2237. enum bpf_hdr_start_off {
  2238. BPF_HDR_START_MAC,
  2239. BPF_HDR_START_NET,
  2240. };
  2241. /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
  2242. enum bpf_lwt_encap_mode {
  2243. BPF_LWT_ENCAP_SEG6,
  2244. BPF_LWT_ENCAP_SEG6_INLINE
  2245. };
  2246. /* user accessible mirror of in-kernel sk_buff.
  2247. * new fields can only be added to the end of this structure
  2248. */
  2249. struct __sk_buff {
  2250. __u32 len;
  2251. __u32 pkt_type;
  2252. __u32 mark;
  2253. __u32 queue_mapping;
  2254. __u32 protocol;
  2255. __u32 vlan_present;
  2256. __u32 vlan_tci;
  2257. __u32 vlan_proto;
  2258. __u32 priority;
  2259. __u32 ingress_ifindex;
  2260. __u32 ifindex;
  2261. __u32 tc_index;
  2262. __u32 cb[5];
  2263. __u32 hash;
  2264. __u32 tc_classid;
  2265. __u32 data;
  2266. __u32 data_end;
  2267. __u32 napi_id;
  2268. /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
  2269. __u32 family;
  2270. __u32 remote_ip4; /* Stored in network byte order */
  2271. __u32 local_ip4; /* Stored in network byte order */
  2272. __u32 remote_ip6[4]; /* Stored in network byte order */
  2273. __u32 local_ip6[4]; /* Stored in network byte order */
  2274. __u32 remote_port; /* Stored in network byte order */
  2275. __u32 local_port; /* stored in host byte order */
  2276. /* ... here. */
  2277. __u32 data_meta;
  2278. };
  2279. struct bpf_tunnel_key {
  2280. __u32 tunnel_id;
  2281. union {
  2282. __u32 remote_ipv4;
  2283. __u32 remote_ipv6[4];
  2284. };
  2285. __u8 tunnel_tos;
  2286. __u8 tunnel_ttl;
  2287. __u16 tunnel_ext; /* Padding, future use. */
  2288. __u32 tunnel_label;
  2289. };
  2290. /* user accessible mirror of in-kernel xfrm_state.
  2291. * new fields can only be added to the end of this structure
  2292. */
  2293. struct bpf_xfrm_state {
  2294. __u32 reqid;
  2295. __u32 spi; /* Stored in network byte order */
  2296. __u16 family;
  2297. __u16 ext; /* Padding, future use. */
  2298. union {
  2299. __u32 remote_ipv4; /* Stored in network byte order */
  2300. __u32 remote_ipv6[4]; /* Stored in network byte order */
  2301. };
  2302. };
  2303. /* Generic BPF return codes which all BPF program types may support.
  2304. * The values are binary compatible with their TC_ACT_* counter-part to
  2305. * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
  2306. * programs.
  2307. *
  2308. * XDP is handled seprately, see XDP_*.
  2309. */
  2310. enum bpf_ret_code {
  2311. BPF_OK = 0,
  2312. /* 1 reserved */
  2313. BPF_DROP = 2,
  2314. /* 3-6 reserved */
  2315. BPF_REDIRECT = 7,
  2316. /* >127 are reserved for prog type specific return codes */
  2317. };
  2318. struct bpf_sock {
  2319. __u32 bound_dev_if;
  2320. __u32 family;
  2321. __u32 type;
  2322. __u32 protocol;
  2323. __u32 mark;
  2324. __u32 priority;
  2325. __u32 src_ip4; /* Allows 1,2,4-byte read.
  2326. * Stored in network byte order.
  2327. */
  2328. __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
  2329. * Stored in network byte order.
  2330. */
  2331. __u32 src_port; /* Allows 4-byte read.
  2332. * Stored in host byte order
  2333. */
  2334. };
  2335. #define XDP_PACKET_HEADROOM 256
  2336. /* User return codes for XDP prog type.
  2337. * A valid XDP program must return one of these defined values. All other
  2338. * return codes are reserved for future use. Unknown return codes will
  2339. * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
  2340. */
  2341. enum xdp_action {
  2342. XDP_ABORTED = 0,
  2343. XDP_DROP,
  2344. XDP_PASS,
  2345. XDP_TX,
  2346. XDP_REDIRECT,
  2347. };
  2348. /* user accessible metadata for XDP packet hook
  2349. * new fields must be added to the end of this structure
  2350. */
  2351. struct xdp_md {
  2352. __u32 data;
  2353. __u32 data_end;
  2354. __u32 data_meta;
  2355. /* Below access go through struct xdp_rxq_info */
  2356. __u32 ingress_ifindex; /* rxq->dev->ifindex */
  2357. __u32 rx_queue_index; /* rxq->queue_index */
  2358. };
  2359. enum sk_action {
  2360. SK_DROP = 0,
  2361. SK_PASS,
  2362. };
  2363. /* user accessible metadata for SK_MSG packet hook, new fields must
  2364. * be added to the end of this structure
  2365. */
  2366. struct sk_msg_md {
  2367. void *data;
  2368. void *data_end;
  2369. __u32 family;
  2370. __u32 remote_ip4; /* Stored in network byte order */
  2371. __u32 local_ip4; /* Stored in network byte order */
  2372. __u32 remote_ip6[4]; /* Stored in network byte order */
  2373. __u32 local_ip6[4]; /* Stored in network byte order */
  2374. __u32 remote_port; /* Stored in network byte order */
  2375. __u32 local_port; /* stored in host byte order */
  2376. };
  2377. struct sk_reuseport_md {
  2378. /*
  2379. * Start of directly accessible data. It begins from
  2380. * the tcp/udp header.
  2381. */
  2382. void *data;
  2383. void *data_end; /* End of directly accessible data */
  2384. /*
  2385. * Total length of packet (starting from the tcp/udp header).
  2386. * Note that the directly accessible bytes (data_end - data)
  2387. * could be less than this "len". Those bytes could be
  2388. * indirectly read by a helper "bpf_skb_load_bytes()".
  2389. */
  2390. __u32 len;
  2391. /*
  2392. * Eth protocol in the mac header (network byte order). e.g.
  2393. * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
  2394. */
  2395. __u32 eth_protocol;
  2396. __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
  2397. __u32 bind_inany; /* Is sock bound to an INANY address? */
  2398. __u32 hash; /* A hash of the packet 4 tuples */
  2399. };
  2400. #define BPF_TAG_SIZE 8
  2401. struct bpf_prog_info {
  2402. __u32 type;
  2403. __u32 id;
  2404. __u8 tag[BPF_TAG_SIZE];
  2405. __u32 jited_prog_len;
  2406. __u32 xlated_prog_len;
  2407. __aligned_u64 jited_prog_insns;
  2408. __aligned_u64 xlated_prog_insns;
  2409. __u64 load_time; /* ns since boottime */
  2410. __u32 created_by_uid;
  2411. __u32 nr_map_ids;
  2412. __aligned_u64 map_ids;
  2413. char name[BPF_OBJ_NAME_LEN];
  2414. __u32 ifindex;
  2415. __u32 gpl_compatible:1;
  2416. __u32 :31; /* alignment pad */
  2417. __u64 netns_dev;
  2418. __u64 netns_ino;
  2419. __u32 nr_jited_ksyms;
  2420. __u32 nr_jited_func_lens;
  2421. __aligned_u64 jited_ksyms;
  2422. __aligned_u64 jited_func_lens;
  2423. } __attribute__((aligned(8)));
  2424. struct bpf_map_info {
  2425. __u32 type;
  2426. __u32 id;
  2427. __u32 key_size;
  2428. __u32 value_size;
  2429. __u32 max_entries;
  2430. __u32 map_flags;
  2431. char name[BPF_OBJ_NAME_LEN];
  2432. __u32 ifindex;
  2433. __u32 :32;
  2434. __u64 netns_dev;
  2435. __u64 netns_ino;
  2436. __u32 btf_id;
  2437. __u32 btf_key_type_id;
  2438. __u32 btf_value_type_id;
  2439. } __attribute__((aligned(8)));
  2440. struct bpf_btf_info {
  2441. __aligned_u64 btf;
  2442. __u32 btf_size;
  2443. __u32 id;
  2444. } __attribute__((aligned(8)));
  2445. /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
  2446. * by user and intended to be used by socket (e.g. to bind to, depends on
  2447. * attach attach type).
  2448. */
  2449. struct bpf_sock_addr {
  2450. __u32 user_family; /* Allows 4-byte read, but no write. */
  2451. __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
  2452. * Stored in network byte order.
  2453. */
  2454. __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
  2455. * Stored in network byte order.
  2456. */
  2457. __u32 user_port; /* Allows 4-byte read and write.
  2458. * Stored in network byte order
  2459. */
  2460. __u32 family; /* Allows 4-byte read, but no write */
  2461. __u32 type; /* Allows 4-byte read, but no write */
  2462. __u32 protocol; /* Allows 4-byte read, but no write */
  2463. __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
  2464. * Stored in network byte order.
  2465. */
  2466. __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
  2467. * Stored in network byte order.
  2468. */
  2469. };
  2470. /* User bpf_sock_ops struct to access socket values and specify request ops
  2471. * and their replies.
  2472. * Some of this fields are in network (bigendian) byte order and may need
  2473. * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
  2474. * New fields can only be added at the end of this structure
  2475. */
  2476. struct bpf_sock_ops {
  2477. __u32 op;
  2478. union {
  2479. __u32 args[4]; /* Optionally passed to bpf program */
  2480. __u32 reply; /* Returned by bpf program */
  2481. __u32 replylong[4]; /* Optionally returned by bpf prog */
  2482. };
  2483. __u32 family;
  2484. __u32 remote_ip4; /* Stored in network byte order */
  2485. __u32 local_ip4; /* Stored in network byte order */
  2486. __u32 remote_ip6[4]; /* Stored in network byte order */
  2487. __u32 local_ip6[4]; /* Stored in network byte order */
  2488. __u32 remote_port; /* Stored in network byte order */
  2489. __u32 local_port; /* stored in host byte order */
  2490. __u32 is_fullsock; /* Some TCP fields are only valid if
  2491. * there is a full socket. If not, the
  2492. * fields read as zero.
  2493. */
  2494. __u32 snd_cwnd;
  2495. __u32 srtt_us; /* Averaged RTT << 3 in usecs */
  2496. __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
  2497. __u32 state;
  2498. __u32 rtt_min;
  2499. __u32 snd_ssthresh;
  2500. __u32 rcv_nxt;
  2501. __u32 snd_nxt;
  2502. __u32 snd_una;
  2503. __u32 mss_cache;
  2504. __u32 ecn_flags;
  2505. __u32 rate_delivered;
  2506. __u32 rate_interval_us;
  2507. __u32 packets_out;
  2508. __u32 retrans_out;
  2509. __u32 total_retrans;
  2510. __u32 segs_in;
  2511. __u32 data_segs_in;
  2512. __u32 segs_out;
  2513. __u32 data_segs_out;
  2514. __u32 lost_out;
  2515. __u32 sacked_out;
  2516. __u32 sk_txhash;
  2517. __u64 bytes_received;
  2518. __u64 bytes_acked;
  2519. };
  2520. /* Definitions for bpf_sock_ops_cb_flags */
  2521. #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
  2522. #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
  2523. #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
  2524. #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
  2525. * supported cb flags
  2526. */
  2527. /* List of known BPF sock_ops operators.
  2528. * New entries can only be added at the end
  2529. */
  2530. enum {
  2531. BPF_SOCK_OPS_VOID,
  2532. BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
  2533. * -1 if default value should be used
  2534. */
  2535. BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
  2536. * window (in packets) or -1 if default
  2537. * value should be used
  2538. */
  2539. BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
  2540. * active connection is initialized
  2541. */
  2542. BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
  2543. * active connection is
  2544. * established
  2545. */
  2546. BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
  2547. * passive connection is
  2548. * established
  2549. */
  2550. BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
  2551. * needs ECN
  2552. */
  2553. BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
  2554. * based on the path and may be
  2555. * dependent on the congestion control
  2556. * algorithm. In general it indicates
  2557. * a congestion threshold. RTTs above
  2558. * this indicate congestion
  2559. */
  2560. BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
  2561. * Arg1: value of icsk_retransmits
  2562. * Arg2: value of icsk_rto
  2563. * Arg3: whether RTO has expired
  2564. */
  2565. BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
  2566. * Arg1: sequence number of 1st byte
  2567. * Arg2: # segments
  2568. * Arg3: return value of
  2569. * tcp_transmit_skb (0 => success)
  2570. */
  2571. BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
  2572. * Arg1: old_state
  2573. * Arg2: new_state
  2574. */
  2575. BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
  2576. * socket transition to LISTEN state.
  2577. */
  2578. };
  2579. /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
  2580. * changes between the TCP and BPF versions. Ideally this should never happen.
  2581. * If it does, we need to add code to convert them before calling
  2582. * the BPF sock_ops function.
  2583. */
  2584. enum {
  2585. BPF_TCP_ESTABLISHED = 1,
  2586. BPF_TCP_SYN_SENT,
  2587. BPF_TCP_SYN_RECV,
  2588. BPF_TCP_FIN_WAIT1,
  2589. BPF_TCP_FIN_WAIT2,
  2590. BPF_TCP_TIME_WAIT,
  2591. BPF_TCP_CLOSE,
  2592. BPF_TCP_CLOSE_WAIT,
  2593. BPF_TCP_LAST_ACK,
  2594. BPF_TCP_LISTEN,
  2595. BPF_TCP_CLOSING, /* Now a valid state */
  2596. BPF_TCP_NEW_SYN_RECV,
  2597. BPF_TCP_MAX_STATES /* Leave at the end! */
  2598. };
  2599. #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
  2600. #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
  2601. struct bpf_perf_event_value {
  2602. __u64 counter;
  2603. __u64 enabled;
  2604. __u64 running;
  2605. };
  2606. #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
  2607. #define BPF_DEVCG_ACC_READ (1ULL << 1)
  2608. #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
  2609. #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
  2610. #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
  2611. struct bpf_cgroup_dev_ctx {
  2612. /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
  2613. __u32 access_type;
  2614. __u32 major;
  2615. __u32 minor;
  2616. };
  2617. struct bpf_raw_tracepoint_args {
  2618. __u64 args[0];
  2619. };
  2620. /* DIRECT: Skip the FIB rules and go to FIB table associated with device
  2621. * OUTPUT: Do lookup from egress perspective; default is ingress
  2622. */
  2623. #define BPF_FIB_LOOKUP_DIRECT BIT(0)
  2624. #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
  2625. enum {
  2626. BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
  2627. BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
  2628. BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
  2629. BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
  2630. BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
  2631. BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
  2632. BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
  2633. BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
  2634. BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
  2635. };
  2636. struct bpf_fib_lookup {
  2637. /* input: network family for lookup (AF_INET, AF_INET6)
  2638. * output: network family of egress nexthop
  2639. */
  2640. __u8 family;
  2641. /* set if lookup is to consider L4 data - e.g., FIB rules */
  2642. __u8 l4_protocol;
  2643. __be16 sport;
  2644. __be16 dport;
  2645. /* total length of packet from network header - used for MTU check */
  2646. __u16 tot_len;
  2647. /* input: L3 device index for lookup
  2648. * output: device index from FIB lookup
  2649. */
  2650. __u32 ifindex;
  2651. union {
  2652. /* inputs to lookup */
  2653. __u8 tos; /* AF_INET */
  2654. __be32 flowinfo; /* AF_INET6, flow_label + priority */
  2655. /* output: metric of fib result (IPv4/IPv6 only) */
  2656. __u32 rt_metric;
  2657. };
  2658. union {
  2659. __be32 ipv4_src;
  2660. __u32 ipv6_src[4]; /* in6_addr; network order */
  2661. };
  2662. /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
  2663. * network header. output: bpf_fib_lookup sets to gateway address
  2664. * if FIB lookup returns gateway route
  2665. */
  2666. union {
  2667. __be32 ipv4_dst;
  2668. __u32 ipv6_dst[4]; /* in6_addr; network order */
  2669. };
  2670. /* output */
  2671. __be16 h_vlan_proto;
  2672. __be16 h_vlan_TCI;
  2673. __u8 smac[6]; /* ETH_ALEN */
  2674. __u8 dmac[6]; /* ETH_ALEN */
  2675. };
  2676. enum bpf_task_fd_type {
  2677. BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
  2678. BPF_FD_TYPE_TRACEPOINT, /* tp name */
  2679. BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
  2680. BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
  2681. BPF_FD_TYPE_UPROBE, /* filename + offset */
  2682. BPF_FD_TYPE_URETPROBE, /* filename + offset */
  2683. };
  2684. #endif /* _UAPI__LINUX_BPF_H__ */