bpf.h 111 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718171917201721172217231724172517261727172817291730173117321733173417351736173717381739174017411742174317441745174617471748174917501751175217531754175517561757175817591760176117621763176417651766176717681769177017711772177317741775177617771778177917801781178217831784178517861787178817891790179117921793179417951796179717981799180018011802180318041805180618071808180918101811181218131814181518161817181818191820182118221823182418251826182718281829183018311832183318341835183618371838183918401841184218431844184518461847184818491850185118521853185418551856185718581859186018611862186318641865186618671868186918701871187218731874187518761877187818791880188118821883188418851886188718881889189018911892189318941895189618971898189919001901190219031904190519061907190819091910191119121913191419151916191719181919192019211922192319241925192619271928192919301931193219331934193519361937193819391940194119421943194419451946194719481949195019511952195319541955195619571958195919601961196219631964196519661967196819691970197119721973197419751976197719781979198019811982198319841985198619871988198919901991199219931994199519961997199819992000200120022003200420052006200720082009201020112012201320142015201620172018201920202021202220232024202520262027202820292030203120322033203420352036203720382039204020412042204320442045204620472048204920502051205220532054205520562057205820592060206120622063206420652066206720682069207020712072207320742075207620772078207920802081208220832084208520862087208820892090209120922093209420952096209720982099210021012102210321042105210621072108210921102111211221132114211521162117211821192120212121222123212421252126212721282129213021312132213321342135213621372138213921402141214221432144214521462147214821492150215121522153215421552156215721582159216021612162216321642165216621672168216921702171217221732174217521762177217821792180218121822183218421852186218721882189219021912192219321942195219621972198219922002201220222032204220522062207220822092210221122122213221422152216221722182219222022212222222322242225222622272228222922302231223222332234223522362237223822392240224122422243224422452246224722482249225022512252225322542255225622572258225922602261226222632264226522662267226822692270227122722273227422752276227722782279228022812282228322842285228622872288228922902291229222932294229522962297229822992300230123022303230423052306230723082309231023112312231323142315231623172318231923202321232223232324232523262327232823292330233123322333233423352336233723382339234023412342234323442345234623472348234923502351235223532354235523562357235823592360236123622363236423652366236723682369237023712372237323742375237623772378237923802381238223832384238523862387238823892390239123922393239423952396239723982399240024012402240324042405240624072408240924102411241224132414241524162417241824192420242124222423242424252426242724282429243024312432243324342435243624372438243924402441244224432444244524462447244824492450245124522453245424552456245724582459246024612462246324642465246624672468246924702471247224732474247524762477247824792480248124822483248424852486248724882489249024912492249324942495249624972498249925002501250225032504250525062507250825092510251125122513251425152516251725182519252025212522252325242525252625272528252925302531253225332534253525362537253825392540254125422543254425452546254725482549255025512552255325542555255625572558255925602561256225632564256525662567256825692570257125722573257425752576257725782579258025812582258325842585258625872588258925902591259225932594259525962597259825992600260126022603260426052606260726082609261026112612261326142615261626172618261926202621262226232624262526262627262826292630263126322633263426352636263726382639264026412642264326442645264626472648264926502651265226532654265526562657265826592660266126622663266426652666266726682669267026712672267326742675267626772678267926802681268226832684268526862687268826892690269126922693269426952696269726982699270027012702270327042705270627072708270927102711271227132714271527162717271827192720272127222723272427252726272727282729273027312732273327342735273627372738273927402741274227432744274527462747274827492750275127522753275427552756275727582759276027612762276327642765276627672768276927702771277227732774277527762777277827792780278127822783278427852786278727882789279027912792279327942795279627972798279928002801280228032804280528062807280828092810281128122813281428152816281728182819282028212822282328242825282628272828282928302831283228332834283528362837283828392840284128422843284428452846284728482849285028512852285328542855285628572858285928602861286228632864286528662867286828692870287128722873287428752876287728782879288028812882288328842885288628872888288928902891289228932894289528962897289828992900290129022903290429052906290729082909291029112912291329142915291629172918291929202921292229232924292529262927292829292930293129322933293429352936293729382939294029412942294329442945294629472948294929502951295229532954295529562957295829592960296129622963
  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. BPF_MAP_LOOKUP_AND_DELETE_ELEM,
  94. };
  95. enum bpf_map_type {
  96. BPF_MAP_TYPE_UNSPEC,
  97. BPF_MAP_TYPE_HASH,
  98. BPF_MAP_TYPE_ARRAY,
  99. BPF_MAP_TYPE_PROG_ARRAY,
  100. BPF_MAP_TYPE_PERF_EVENT_ARRAY,
  101. BPF_MAP_TYPE_PERCPU_HASH,
  102. BPF_MAP_TYPE_PERCPU_ARRAY,
  103. BPF_MAP_TYPE_STACK_TRACE,
  104. BPF_MAP_TYPE_CGROUP_ARRAY,
  105. BPF_MAP_TYPE_LRU_HASH,
  106. BPF_MAP_TYPE_LRU_PERCPU_HASH,
  107. BPF_MAP_TYPE_LPM_TRIE,
  108. BPF_MAP_TYPE_ARRAY_OF_MAPS,
  109. BPF_MAP_TYPE_HASH_OF_MAPS,
  110. BPF_MAP_TYPE_DEVMAP,
  111. BPF_MAP_TYPE_SOCKMAP,
  112. BPF_MAP_TYPE_CPUMAP,
  113. BPF_MAP_TYPE_XSKMAP,
  114. BPF_MAP_TYPE_SOCKHASH,
  115. BPF_MAP_TYPE_CGROUP_STORAGE,
  116. BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
  117. BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE,
  118. BPF_MAP_TYPE_QUEUE,
  119. BPF_MAP_TYPE_STACK,
  120. };
  121. enum bpf_prog_type {
  122. BPF_PROG_TYPE_UNSPEC,
  123. BPF_PROG_TYPE_SOCKET_FILTER,
  124. BPF_PROG_TYPE_KPROBE,
  125. BPF_PROG_TYPE_SCHED_CLS,
  126. BPF_PROG_TYPE_SCHED_ACT,
  127. BPF_PROG_TYPE_TRACEPOINT,
  128. BPF_PROG_TYPE_XDP,
  129. BPF_PROG_TYPE_PERF_EVENT,
  130. BPF_PROG_TYPE_CGROUP_SKB,
  131. BPF_PROG_TYPE_CGROUP_SOCK,
  132. BPF_PROG_TYPE_LWT_IN,
  133. BPF_PROG_TYPE_LWT_OUT,
  134. BPF_PROG_TYPE_LWT_XMIT,
  135. BPF_PROG_TYPE_SOCK_OPS,
  136. BPF_PROG_TYPE_SK_SKB,
  137. BPF_PROG_TYPE_CGROUP_DEVICE,
  138. BPF_PROG_TYPE_SK_MSG,
  139. BPF_PROG_TYPE_RAW_TRACEPOINT,
  140. BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
  141. BPF_PROG_TYPE_LWT_SEG6LOCAL,
  142. BPF_PROG_TYPE_LIRC_MODE2,
  143. BPF_PROG_TYPE_SK_REUSEPORT,
  144. BPF_PROG_TYPE_FLOW_DISSECTOR,
  145. };
  146. enum bpf_attach_type {
  147. BPF_CGROUP_INET_INGRESS,
  148. BPF_CGROUP_INET_EGRESS,
  149. BPF_CGROUP_INET_SOCK_CREATE,
  150. BPF_CGROUP_SOCK_OPS,
  151. BPF_SK_SKB_STREAM_PARSER,
  152. BPF_SK_SKB_STREAM_VERDICT,
  153. BPF_CGROUP_DEVICE,
  154. BPF_SK_MSG_VERDICT,
  155. BPF_CGROUP_INET4_BIND,
  156. BPF_CGROUP_INET6_BIND,
  157. BPF_CGROUP_INET4_CONNECT,
  158. BPF_CGROUP_INET6_CONNECT,
  159. BPF_CGROUP_INET4_POST_BIND,
  160. BPF_CGROUP_INET6_POST_BIND,
  161. BPF_CGROUP_UDP4_SENDMSG,
  162. BPF_CGROUP_UDP6_SENDMSG,
  163. BPF_LIRC_MODE2,
  164. BPF_FLOW_DISSECTOR,
  165. __MAX_BPF_ATTACH_TYPE
  166. };
  167. #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
  168. /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
  169. *
  170. * NONE(default): No further bpf programs allowed in the subtree.
  171. *
  172. * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
  173. * the program in this cgroup yields to sub-cgroup program.
  174. *
  175. * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
  176. * that cgroup program gets run in addition to the program in this cgroup.
  177. *
  178. * Only one program is allowed to be attached to a cgroup with
  179. * NONE or BPF_F_ALLOW_OVERRIDE flag.
  180. * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
  181. * release old program and attach the new one. Attach flags has to match.
  182. *
  183. * Multiple programs are allowed to be attached to a cgroup with
  184. * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
  185. * (those that were attached first, run first)
  186. * The programs of sub-cgroup are executed first, then programs of
  187. * this cgroup and then programs of parent cgroup.
  188. * When children program makes decision (like picking TCP CA or sock bind)
  189. * parent program has a chance to override it.
  190. *
  191. * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
  192. * A cgroup with NONE doesn't allow any programs in sub-cgroups.
  193. * Ex1:
  194. * cgrp1 (MULTI progs A, B) ->
  195. * cgrp2 (OVERRIDE prog C) ->
  196. * cgrp3 (MULTI prog D) ->
  197. * cgrp4 (OVERRIDE prog E) ->
  198. * cgrp5 (NONE prog F)
  199. * the event in cgrp5 triggers execution of F,D,A,B in that order.
  200. * if prog F is detached, the execution is E,D,A,B
  201. * if prog F and D are detached, the execution is E,A,B
  202. * if prog F, E and D are detached, the execution is C,A,B
  203. *
  204. * All eligible programs are executed regardless of return code from
  205. * earlier programs.
  206. */
  207. #define BPF_F_ALLOW_OVERRIDE (1U << 0)
  208. #define BPF_F_ALLOW_MULTI (1U << 1)
  209. /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
  210. * verifier will perform strict alignment checking as if the kernel
  211. * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
  212. * and NET_IP_ALIGN defined to 2.
  213. */
  214. #define BPF_F_STRICT_ALIGNMENT (1U << 0)
  215. /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
  216. #define BPF_PSEUDO_MAP_FD 1
  217. /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
  218. * offset to another bpf function
  219. */
  220. #define BPF_PSEUDO_CALL 1
  221. /* flags for BPF_MAP_UPDATE_ELEM command */
  222. #define BPF_ANY 0 /* create new element or update existing */
  223. #define BPF_NOEXIST 1 /* create new element if it didn't exist */
  224. #define BPF_EXIST 2 /* update existing element */
  225. /* flags for BPF_MAP_CREATE command */
  226. #define BPF_F_NO_PREALLOC (1U << 0)
  227. /* Instead of having one common LRU list in the
  228. * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
  229. * which can scale and perform better.
  230. * Note, the LRU nodes (including free nodes) cannot be moved
  231. * across different LRU lists.
  232. */
  233. #define BPF_F_NO_COMMON_LRU (1U << 1)
  234. /* Specify numa node during map creation */
  235. #define BPF_F_NUMA_NODE (1U << 2)
  236. /* flags for BPF_PROG_QUERY */
  237. #define BPF_F_QUERY_EFFECTIVE (1U << 0)
  238. #define BPF_OBJ_NAME_LEN 16U
  239. /* Flags for accessing BPF object */
  240. #define BPF_F_RDONLY (1U << 3)
  241. #define BPF_F_WRONLY (1U << 4)
  242. /* Flag for stack_map, store build_id+offset instead of pointer */
  243. #define BPF_F_STACK_BUILD_ID (1U << 5)
  244. enum bpf_stack_build_id_status {
  245. /* user space need an empty entry to identify end of a trace */
  246. BPF_STACK_BUILD_ID_EMPTY = 0,
  247. /* with valid build_id and offset */
  248. BPF_STACK_BUILD_ID_VALID = 1,
  249. /* couldn't get build_id, fallback to ip */
  250. BPF_STACK_BUILD_ID_IP = 2,
  251. };
  252. #define BPF_BUILD_ID_SIZE 20
  253. struct bpf_stack_build_id {
  254. __s32 status;
  255. unsigned char build_id[BPF_BUILD_ID_SIZE];
  256. union {
  257. __u64 offset;
  258. __u64 ip;
  259. };
  260. };
  261. union bpf_attr {
  262. struct { /* anonymous struct used by BPF_MAP_CREATE command */
  263. __u32 map_type; /* one of enum bpf_map_type */
  264. __u32 key_size; /* size of key in bytes */
  265. __u32 value_size; /* size of value in bytes */
  266. __u32 max_entries; /* max number of entries in a map */
  267. __u32 map_flags; /* BPF_MAP_CREATE related
  268. * flags defined above.
  269. */
  270. __u32 inner_map_fd; /* fd pointing to the inner map */
  271. __u32 numa_node; /* numa node (effective only if
  272. * BPF_F_NUMA_NODE is set).
  273. */
  274. char map_name[BPF_OBJ_NAME_LEN];
  275. __u32 map_ifindex; /* ifindex of netdev to create on */
  276. __u32 btf_fd; /* fd pointing to a BTF type data */
  277. __u32 btf_key_type_id; /* BTF type_id of the key */
  278. __u32 btf_value_type_id; /* BTF type_id of the value */
  279. };
  280. struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
  281. __u32 map_fd;
  282. __aligned_u64 key;
  283. union {
  284. __aligned_u64 value;
  285. __aligned_u64 next_key;
  286. };
  287. __u64 flags;
  288. };
  289. struct { /* anonymous struct used by BPF_PROG_LOAD command */
  290. __u32 prog_type; /* one of enum bpf_prog_type */
  291. __u32 insn_cnt;
  292. __aligned_u64 insns;
  293. __aligned_u64 license;
  294. __u32 log_level; /* verbosity level of verifier */
  295. __u32 log_size; /* size of user buffer */
  296. __aligned_u64 log_buf; /* user supplied buffer */
  297. __u32 kern_version; /* checked when prog_type=kprobe */
  298. __u32 prog_flags;
  299. char prog_name[BPF_OBJ_NAME_LEN];
  300. __u32 prog_ifindex; /* ifindex of netdev to prep for */
  301. /* For some prog types expected attach type must be known at
  302. * load time to verify attach type specific parts of prog
  303. * (context accesses, allowed helpers, etc).
  304. */
  305. __u32 expected_attach_type;
  306. };
  307. struct { /* anonymous struct used by BPF_OBJ_* commands */
  308. __aligned_u64 pathname;
  309. __u32 bpf_fd;
  310. __u32 file_flags;
  311. };
  312. struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
  313. __u32 target_fd; /* container object to attach to */
  314. __u32 attach_bpf_fd; /* eBPF program to attach */
  315. __u32 attach_type;
  316. __u32 attach_flags;
  317. };
  318. struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
  319. __u32 prog_fd;
  320. __u32 retval;
  321. __u32 data_size_in;
  322. __u32 data_size_out;
  323. __aligned_u64 data_in;
  324. __aligned_u64 data_out;
  325. __u32 repeat;
  326. __u32 duration;
  327. } test;
  328. struct { /* anonymous struct used by BPF_*_GET_*_ID */
  329. union {
  330. __u32 start_id;
  331. __u32 prog_id;
  332. __u32 map_id;
  333. __u32 btf_id;
  334. };
  335. __u32 next_id;
  336. __u32 open_flags;
  337. };
  338. struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
  339. __u32 bpf_fd;
  340. __u32 info_len;
  341. __aligned_u64 info;
  342. } info;
  343. struct { /* anonymous struct used by BPF_PROG_QUERY command */
  344. __u32 target_fd; /* container object to query */
  345. __u32 attach_type;
  346. __u32 query_flags;
  347. __u32 attach_flags;
  348. __aligned_u64 prog_ids;
  349. __u32 prog_cnt;
  350. } query;
  351. struct {
  352. __u64 name;
  353. __u32 prog_fd;
  354. } raw_tracepoint;
  355. struct { /* anonymous struct for BPF_BTF_LOAD */
  356. __aligned_u64 btf;
  357. __aligned_u64 btf_log_buf;
  358. __u32 btf_size;
  359. __u32 btf_log_size;
  360. __u32 btf_log_level;
  361. };
  362. struct {
  363. __u32 pid; /* input: pid */
  364. __u32 fd; /* input: fd */
  365. __u32 flags; /* input: flags */
  366. __u32 buf_len; /* input/output: buf len */
  367. __aligned_u64 buf; /* input/output:
  368. * tp_name for tracepoint
  369. * symbol for kprobe
  370. * filename for uprobe
  371. */
  372. __u32 prog_id; /* output: prod_id */
  373. __u32 fd_type; /* output: BPF_FD_TYPE_* */
  374. __u64 probe_offset; /* output: probe_offset */
  375. __u64 probe_addr; /* output: probe_addr */
  376. } task_fd_query;
  377. } __attribute__((aligned(8)));
  378. /* The description below is an attempt at providing documentation to eBPF
  379. * developers about the multiple available eBPF helper functions. It can be
  380. * parsed and used to produce a manual page. The workflow is the following,
  381. * and requires the rst2man utility:
  382. *
  383. * $ ./scripts/bpf_helpers_doc.py \
  384. * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
  385. * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
  386. * $ man /tmp/bpf-helpers.7
  387. *
  388. * Note that in order to produce this external documentation, some RST
  389. * formatting is used in the descriptions to get "bold" and "italics" in
  390. * manual pages. Also note that the few trailing white spaces are
  391. * intentional, removing them would break paragraphs for rst2man.
  392. *
  393. * Start of BPF helper function descriptions:
  394. *
  395. * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
  396. * Description
  397. * Perform a lookup in *map* for an entry associated to *key*.
  398. * Return
  399. * Map value associated to *key*, or **NULL** if no entry was
  400. * found.
  401. *
  402. * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
  403. * Description
  404. * Add or update the value of the entry associated to *key* in
  405. * *map* with *value*. *flags* is one of:
  406. *
  407. * **BPF_NOEXIST**
  408. * The entry for *key* must not exist in the map.
  409. * **BPF_EXIST**
  410. * The entry for *key* must already exist in the map.
  411. * **BPF_ANY**
  412. * No condition on the existence of the entry for *key*.
  413. *
  414. * Flag value **BPF_NOEXIST** cannot be used for maps of types
  415. * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
  416. * elements always exist), the helper would return an error.
  417. * Return
  418. * 0 on success, or a negative error in case of failure.
  419. *
  420. * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
  421. * Description
  422. * Delete entry with *key* from *map*.
  423. * Return
  424. * 0 on success, or a negative error in case of failure.
  425. *
  426. * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags)
  427. * Description
  428. * Push an element *value* in *map*. *flags* is one of:
  429. *
  430. * **BPF_EXIST**
  431. * If the queue/stack is full, the oldest element is removed to
  432. * make room for this.
  433. * Return
  434. * 0 on success, or a negative error in case of failure.
  435. *
  436. * int bpf_map_pop_elem(struct bpf_map *map, void *value)
  437. * Description
  438. * Pop an element from *map*.
  439. * Return
  440. * 0 on success, or a negative error in case of failure.
  441. *
  442. * int bpf_map_peek_elem(struct bpf_map *map, void *value)
  443. * Description
  444. * Get an element from *map* without removing it.
  445. * Return
  446. * 0 on success, or a negative error in case of failure.
  447. *
  448. * int bpf_probe_read(void *dst, u32 size, const void *src)
  449. * Description
  450. * For tracing programs, safely attempt to read *size* bytes from
  451. * address *src* and store the data in *dst*.
  452. * Return
  453. * 0 on success, or a negative error in case of failure.
  454. *
  455. * u64 bpf_ktime_get_ns(void)
  456. * Description
  457. * Return the time elapsed since system boot, in nanoseconds.
  458. * Return
  459. * Current *ktime*.
  460. *
  461. * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
  462. * Description
  463. * This helper is a "printk()-like" facility for debugging. It
  464. * prints a message defined by format *fmt* (of size *fmt_size*)
  465. * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
  466. * available. It can take up to three additional **u64**
  467. * arguments (as an eBPF helpers, the total number of arguments is
  468. * limited to five).
  469. *
  470. * Each time the helper is called, it appends a line to the trace.
  471. * The format of the trace is customizable, and the exact output
  472. * one will get depends on the options set in
  473. * *\/sys/kernel/debug/tracing/trace_options* (see also the
  474. * *README* file under the same directory). However, it usually
  475. * defaults to something like:
  476. *
  477. * ::
  478. *
  479. * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
  480. *
  481. * In the above:
  482. *
  483. * * ``telnet`` is the name of the current task.
  484. * * ``470`` is the PID of the current task.
  485. * * ``001`` is the CPU number on which the task is
  486. * running.
  487. * * In ``.N..``, each character refers to a set of
  488. * options (whether irqs are enabled, scheduling
  489. * options, whether hard/softirqs are running, level of
  490. * preempt_disabled respectively). **N** means that
  491. * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
  492. * are set.
  493. * * ``419421.045894`` is a timestamp.
  494. * * ``0x00000001`` is a fake value used by BPF for the
  495. * instruction pointer register.
  496. * * ``<formatted msg>`` is the message formatted with
  497. * *fmt*.
  498. *
  499. * The conversion specifiers supported by *fmt* are similar, but
  500. * more limited than for printk(). They are **%d**, **%i**,
  501. * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
  502. * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
  503. * of field, padding with zeroes, etc.) is available, and the
  504. * helper will return **-EINVAL** (but print nothing) if it
  505. * encounters an unknown specifier.
  506. *
  507. * Also, note that **bpf_trace_printk**\ () is slow, and should
  508. * only be used for debugging purposes. For this reason, a notice
  509. * bloc (spanning several lines) is printed to kernel logs and
  510. * states that the helper should not be used "for production use"
  511. * the first time this helper is used (or more precisely, when
  512. * **trace_printk**\ () buffers are allocated). For passing values
  513. * to user space, perf events should be preferred.
  514. * Return
  515. * The number of bytes written to the buffer, or a negative error
  516. * in case of failure.
  517. *
  518. * u32 bpf_get_prandom_u32(void)
  519. * Description
  520. * Get a pseudo-random number.
  521. *
  522. * From a security point of view, this helper uses its own
  523. * pseudo-random internal state, and cannot be used to infer the
  524. * seed of other random functions in the kernel. However, it is
  525. * essential to note that the generator used by the helper is not
  526. * cryptographically secure.
  527. * Return
  528. * A random 32-bit unsigned value.
  529. *
  530. * u32 bpf_get_smp_processor_id(void)
  531. * Description
  532. * Get the SMP (symmetric multiprocessing) processor id. Note that
  533. * all programs run with preemption disabled, which means that the
  534. * SMP processor id is stable during all the execution of the
  535. * program.
  536. * Return
  537. * The SMP id of the processor running the program.
  538. *
  539. * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
  540. * Description
  541. * Store *len* bytes from address *from* into the packet
  542. * associated to *skb*, at *offset*. *flags* are a combination of
  543. * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
  544. * checksum for the packet after storing the bytes) and
  545. * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
  546. * **->swhash** and *skb*\ **->l4hash** to 0).
  547. *
  548. * A call to this helper is susceptible to change the underlaying
  549. * packet buffer. Therefore, at load time, all checks on pointers
  550. * previously done by the verifier are invalidated and must be
  551. * performed again, if the helper is used in combination with
  552. * direct packet access.
  553. * Return
  554. * 0 on success, or a negative error in case of failure.
  555. *
  556. * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
  557. * Description
  558. * Recompute the layer 3 (e.g. IP) checksum for the packet
  559. * associated to *skb*. Computation is incremental, so the helper
  560. * must know the former value of the header field that was
  561. * modified (*from*), the new value of this field (*to*), and the
  562. * number of bytes (2 or 4) for this field, stored in *size*.
  563. * Alternatively, it is possible to store the difference between
  564. * the previous and the new values of the header field in *to*, by
  565. * setting *from* and *size* to 0. For both methods, *offset*
  566. * indicates the location of the IP checksum within the packet.
  567. *
  568. * This helper works in combination with **bpf_csum_diff**\ (),
  569. * which does not update the checksum in-place, but offers more
  570. * flexibility and can handle sizes larger than 2 or 4 for the
  571. * checksum to update.
  572. *
  573. * A call to this helper is susceptible to change the underlaying
  574. * packet buffer. Therefore, at load time, all checks on pointers
  575. * previously done by the verifier are invalidated and must be
  576. * performed again, if the helper is used in combination with
  577. * direct packet access.
  578. * Return
  579. * 0 on success, or a negative error in case of failure.
  580. *
  581. * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
  582. * Description
  583. * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
  584. * packet associated to *skb*. Computation is incremental, so the
  585. * helper must know the former value of the header field that was
  586. * modified (*from*), the new value of this field (*to*), and the
  587. * number of bytes (2 or 4) for this field, stored on the lowest
  588. * four bits of *flags*. Alternatively, it is possible to store
  589. * the difference between the previous and the new values of the
  590. * header field in *to*, by setting *from* and the four lowest
  591. * bits of *flags* to 0. For both methods, *offset* indicates the
  592. * location of the IP checksum within the packet. In addition to
  593. * the size of the field, *flags* can be added (bitwise OR) actual
  594. * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
  595. * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
  596. * for updates resulting in a null checksum the value is set to
  597. * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
  598. * the checksum is to be computed against a pseudo-header.
  599. *
  600. * This helper works in combination with **bpf_csum_diff**\ (),
  601. * which does not update the checksum in-place, but offers more
  602. * flexibility and can handle sizes larger than 2 or 4 for the
  603. * checksum to update.
  604. *
  605. * A call to this helper is susceptible to change the underlaying
  606. * packet buffer. Therefore, at load time, all checks on pointers
  607. * previously done by the verifier are invalidated and must be
  608. * performed again, if the helper is used in combination with
  609. * direct packet access.
  610. * Return
  611. * 0 on success, or a negative error in case of failure.
  612. *
  613. * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
  614. * Description
  615. * This special helper is used to trigger a "tail call", or in
  616. * other words, to jump into another eBPF program. The same stack
  617. * frame is used (but values on stack and in registers for the
  618. * caller are not accessible to the callee). This mechanism allows
  619. * for program chaining, either for raising the maximum number of
  620. * available eBPF instructions, or to execute given programs in
  621. * conditional blocks. For security reasons, there is an upper
  622. * limit to the number of successive tail calls that can be
  623. * performed.
  624. *
  625. * Upon call of this helper, the program attempts to jump into a
  626. * program referenced at index *index* in *prog_array_map*, a
  627. * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
  628. * *ctx*, a pointer to the context.
  629. *
  630. * If the call succeeds, the kernel immediately runs the first
  631. * instruction of the new program. This is not a function call,
  632. * and it never returns to the previous program. If the call
  633. * fails, then the helper has no effect, and the caller continues
  634. * to run its subsequent instructions. A call can fail if the
  635. * destination program for the jump does not exist (i.e. *index*
  636. * is superior to the number of entries in *prog_array_map*), or
  637. * if the maximum number of tail calls has been reached for this
  638. * chain of programs. This limit is defined in the kernel by the
  639. * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
  640. * which is currently set to 32.
  641. * Return
  642. * 0 on success, or a negative error in case of failure.
  643. *
  644. * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
  645. * Description
  646. * Clone and redirect the packet associated to *skb* to another
  647. * net device of index *ifindex*. Both ingress and egress
  648. * interfaces can be used for redirection. The **BPF_F_INGRESS**
  649. * value in *flags* is used to make the distinction (ingress path
  650. * is selected if the flag is present, egress path otherwise).
  651. * This is the only flag supported for now.
  652. *
  653. * In comparison with **bpf_redirect**\ () helper,
  654. * **bpf_clone_redirect**\ () has the associated cost of
  655. * duplicating the packet buffer, but this can be executed out of
  656. * the eBPF program. Conversely, **bpf_redirect**\ () is more
  657. * efficient, but it is handled through an action code where the
  658. * redirection happens only after the eBPF program has returned.
  659. *
  660. * A call to this helper is susceptible to change the underlaying
  661. * packet buffer. Therefore, at load time, all checks on pointers
  662. * previously done by the verifier are invalidated and must be
  663. * performed again, if the helper is used in combination with
  664. * direct packet access.
  665. * Return
  666. * 0 on success, or a negative error in case of failure.
  667. *
  668. * u64 bpf_get_current_pid_tgid(void)
  669. * Return
  670. * A 64-bit integer containing the current tgid and pid, and
  671. * created as such:
  672. * *current_task*\ **->tgid << 32 \|**
  673. * *current_task*\ **->pid**.
  674. *
  675. * u64 bpf_get_current_uid_gid(void)
  676. * Return
  677. * A 64-bit integer containing the current GID and UID, and
  678. * created as such: *current_gid* **<< 32 \|** *current_uid*.
  679. *
  680. * int bpf_get_current_comm(char *buf, u32 size_of_buf)
  681. * Description
  682. * Copy the **comm** attribute of the current task into *buf* of
  683. * *size_of_buf*. The **comm** attribute contains the name of
  684. * the executable (excluding the path) for the current task. The
  685. * *size_of_buf* must be strictly positive. On success, the
  686. * helper makes sure that the *buf* is NUL-terminated. On failure,
  687. * it is filled with zeroes.
  688. * Return
  689. * 0 on success, or a negative error in case of failure.
  690. *
  691. * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
  692. * Description
  693. * Retrieve the classid for the current task, i.e. for the net_cls
  694. * cgroup to which *skb* belongs.
  695. *
  696. * This helper can be used on TC egress path, but not on ingress.
  697. *
  698. * The net_cls cgroup provides an interface to tag network packets
  699. * based on a user-provided identifier for all traffic coming from
  700. * the tasks belonging to the related cgroup. See also the related
  701. * kernel documentation, available from the Linux sources in file
  702. * *Documentation/cgroup-v1/net_cls.txt*.
  703. *
  704. * The Linux kernel has two versions for cgroups: there are
  705. * cgroups v1 and cgroups v2. Both are available to users, who can
  706. * use a mixture of them, but note that the net_cls cgroup is for
  707. * cgroup v1 only. This makes it incompatible with BPF programs
  708. * run on cgroups, which is a cgroup-v2-only feature (a socket can
  709. * only hold data for one version of cgroups at a time).
  710. *
  711. * This helper is only available is the kernel was compiled with
  712. * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
  713. * "**y**" or to "**m**".
  714. * Return
  715. * The classid, or 0 for the default unconfigured classid.
  716. *
  717. * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
  718. * Description
  719. * Push a *vlan_tci* (VLAN tag control information) of protocol
  720. * *vlan_proto* to the packet associated to *skb*, then update
  721. * the checksum. Note that if *vlan_proto* is different from
  722. * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
  723. * be **ETH_P_8021Q**.
  724. *
  725. * A call to this helper is susceptible to change the underlaying
  726. * packet buffer. Therefore, at load time, all checks on pointers
  727. * previously done by the verifier are invalidated and must be
  728. * performed again, if the helper is used in combination with
  729. * direct packet access.
  730. * Return
  731. * 0 on success, or a negative error in case of failure.
  732. *
  733. * int bpf_skb_vlan_pop(struct sk_buff *skb)
  734. * Description
  735. * Pop a VLAN header from the packet associated to *skb*.
  736. *
  737. * A call to this helper is susceptible to change the underlaying
  738. * packet buffer. Therefore, at load time, all checks on pointers
  739. * previously done by the verifier are invalidated and must be
  740. * performed again, if the helper is used in combination with
  741. * direct packet access.
  742. * Return
  743. * 0 on success, or a negative error in case of failure.
  744. *
  745. * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  746. * Description
  747. * Get tunnel metadata. This helper takes a pointer *key* to an
  748. * empty **struct bpf_tunnel_key** of **size**, that will be
  749. * filled with tunnel metadata for the packet associated to *skb*.
  750. * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
  751. * indicates that the tunnel is based on IPv6 protocol instead of
  752. * IPv4.
  753. *
  754. * The **struct bpf_tunnel_key** is an object that generalizes the
  755. * principal parameters used by various tunneling protocols into a
  756. * single struct. This way, it can be used to easily make a
  757. * decision based on the contents of the encapsulation header,
  758. * "summarized" in this struct. In particular, it holds the IP
  759. * address of the remote end (IPv4 or IPv6, depending on the case)
  760. * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
  761. * this struct exposes the *key*\ **->tunnel_id**, which is
  762. * generally mapped to a VNI (Virtual Network Identifier), making
  763. * it programmable together with the **bpf_skb_set_tunnel_key**\
  764. * () helper.
  765. *
  766. * Let's imagine that the following code is part of a program
  767. * attached to the TC ingress interface, on one end of a GRE
  768. * tunnel, and is supposed to filter out all messages coming from
  769. * remote ends with IPv4 address other than 10.0.0.1:
  770. *
  771. * ::
  772. *
  773. * int ret;
  774. * struct bpf_tunnel_key key = {};
  775. *
  776. * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
  777. * if (ret < 0)
  778. * return TC_ACT_SHOT; // drop packet
  779. *
  780. * if (key.remote_ipv4 != 0x0a000001)
  781. * return TC_ACT_SHOT; // drop packet
  782. *
  783. * return TC_ACT_OK; // accept packet
  784. *
  785. * This interface can also be used with all encapsulation devices
  786. * that can operate in "collect metadata" mode: instead of having
  787. * one network device per specific configuration, the "collect
  788. * metadata" mode only requires a single device where the
  789. * configuration can be extracted from this helper.
  790. *
  791. * This can be used together with various tunnels such as VXLan,
  792. * Geneve, GRE or IP in IP (IPIP).
  793. * Return
  794. * 0 on success, or a negative error in case of failure.
  795. *
  796. * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  797. * Description
  798. * Populate tunnel metadata for packet associated to *skb.* The
  799. * tunnel metadata is set to the contents of *key*, of *size*. The
  800. * *flags* can be set to a combination of the following values:
  801. *
  802. * **BPF_F_TUNINFO_IPV6**
  803. * Indicate that the tunnel is based on IPv6 protocol
  804. * instead of IPv4.
  805. * **BPF_F_ZERO_CSUM_TX**
  806. * For IPv4 packets, add a flag to tunnel metadata
  807. * indicating that checksum computation should be skipped
  808. * and checksum set to zeroes.
  809. * **BPF_F_DONT_FRAGMENT**
  810. * Add a flag to tunnel metadata indicating that the
  811. * packet should not be fragmented.
  812. * **BPF_F_SEQ_NUMBER**
  813. * Add a flag to tunnel metadata indicating that a
  814. * sequence number should be added to tunnel header before
  815. * sending the packet. This flag was added for GRE
  816. * encapsulation, but might be used with other protocols
  817. * as well in the future.
  818. *
  819. * Here is a typical usage on the transmit path:
  820. *
  821. * ::
  822. *
  823. * struct bpf_tunnel_key key;
  824. * populate key ...
  825. * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
  826. * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
  827. *
  828. * See also the description of the **bpf_skb_get_tunnel_key**\ ()
  829. * helper for additional information.
  830. * Return
  831. * 0 on success, or a negative error in case of failure.
  832. *
  833. * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
  834. * Description
  835. * Read the value of a perf event counter. This helper relies on a
  836. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
  837. * the perf event counter is selected when *map* is updated with
  838. * perf event file descriptors. The *map* is an array whose size
  839. * is the number of available CPUs, and each cell contains a value
  840. * relative to one CPU. The value to retrieve is indicated by
  841. * *flags*, that contains the index of the CPU to look up, masked
  842. * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  843. * **BPF_F_CURRENT_CPU** to indicate that the value for the
  844. * current CPU should be retrieved.
  845. *
  846. * Note that before Linux 4.13, only hardware perf event can be
  847. * retrieved.
  848. *
  849. * Also, be aware that the newer helper
  850. * **bpf_perf_event_read_value**\ () is recommended over
  851. * **bpf_perf_event_read**\ () in general. The latter has some ABI
  852. * quirks where error and counter value are used as a return code
  853. * (which is wrong to do since ranges may overlap). This issue is
  854. * fixed with **bpf_perf_event_read_value**\ (), which at the same
  855. * time provides more features over the **bpf_perf_event_read**\
  856. * () interface. Please refer to the description of
  857. * **bpf_perf_event_read_value**\ () for details.
  858. * Return
  859. * The value of the perf event counter read from the map, or a
  860. * negative error code in case of failure.
  861. *
  862. * int bpf_redirect(u32 ifindex, u64 flags)
  863. * Description
  864. * Redirect the packet to another net device of index *ifindex*.
  865. * This helper is somewhat similar to **bpf_clone_redirect**\
  866. * (), except that the packet is not cloned, which provides
  867. * increased performance.
  868. *
  869. * Except for XDP, both ingress and egress interfaces can be used
  870. * for redirection. The **BPF_F_INGRESS** value in *flags* is used
  871. * to make the distinction (ingress path is selected if the flag
  872. * is present, egress path otherwise). Currently, XDP only
  873. * supports redirection to the egress interface, and accepts no
  874. * flag at all.
  875. *
  876. * The same effect can be attained with the more generic
  877. * **bpf_redirect_map**\ (), which requires specific maps to be
  878. * used but offers better performance.
  879. * Return
  880. * For XDP, the helper returns **XDP_REDIRECT** on success or
  881. * **XDP_ABORTED** on error. For other program types, the values
  882. * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
  883. * error.
  884. *
  885. * u32 bpf_get_route_realm(struct sk_buff *skb)
  886. * Description
  887. * Retrieve the realm or the route, that is to say the
  888. * **tclassid** field of the destination for the *skb*. The
  889. * indentifier retrieved is a user-provided tag, similar to the
  890. * one used with the net_cls cgroup (see description for
  891. * **bpf_get_cgroup_classid**\ () helper), but here this tag is
  892. * held by a route (a destination entry), not by a task.
  893. *
  894. * Retrieving this identifier works with the clsact TC egress hook
  895. * (see also **tc-bpf(8)**), or alternatively on conventional
  896. * classful egress qdiscs, but not on TC ingress path. In case of
  897. * clsact TC egress hook, this has the advantage that, internally,
  898. * the destination entry has not been dropped yet in the transmit
  899. * path. Therefore, the destination entry does not need to be
  900. * artificially held via **netif_keep_dst**\ () for a classful
  901. * qdisc until the *skb* is freed.
  902. *
  903. * This helper is available only if the kernel was compiled with
  904. * **CONFIG_IP_ROUTE_CLASSID** configuration option.
  905. * Return
  906. * The realm of the route for the packet associated to *skb*, or 0
  907. * if none was found.
  908. *
  909. * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  910. * Description
  911. * Write raw *data* blob into a special BPF perf event held by
  912. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  913. * event must have the following attributes: **PERF_SAMPLE_RAW**
  914. * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  915. * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  916. *
  917. * The *flags* are used to indicate the index in *map* for which
  918. * the value must be put, masked with **BPF_F_INDEX_MASK**.
  919. * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  920. * to indicate that the index of the current CPU core should be
  921. * used.
  922. *
  923. * The value to write, of *size*, is passed through eBPF stack and
  924. * pointed by *data*.
  925. *
  926. * The context of the program *ctx* needs also be passed to the
  927. * helper.
  928. *
  929. * On user space, a program willing to read the values needs to
  930. * call **perf_event_open**\ () on the perf event (either for
  931. * one or for all CPUs) and to store the file descriptor into the
  932. * *map*. This must be done before the eBPF program can send data
  933. * into it. An example is available in file
  934. * *samples/bpf/trace_output_user.c* in the Linux kernel source
  935. * tree (the eBPF program counterpart is in
  936. * *samples/bpf/trace_output_kern.c*).
  937. *
  938. * **bpf_perf_event_output**\ () achieves better performance
  939. * than **bpf_trace_printk**\ () for sharing data with user
  940. * space, and is much better suitable for streaming data from eBPF
  941. * programs.
  942. *
  943. * Note that this helper is not restricted to tracing use cases
  944. * and can be used with programs attached to TC or XDP as well,
  945. * where it allows for passing data to user space listeners. Data
  946. * can be:
  947. *
  948. * * Only custom structs,
  949. * * Only the packet payload, or
  950. * * A combination of both.
  951. * Return
  952. * 0 on success, or a negative error in case of failure.
  953. *
  954. * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
  955. * Description
  956. * This helper was provided as an easy way to load data from a
  957. * packet. It can be used to load *len* bytes from *offset* from
  958. * the packet associated to *skb*, into the buffer pointed by
  959. * *to*.
  960. *
  961. * Since Linux 4.7, usage of this helper has mostly been replaced
  962. * by "direct packet access", enabling packet data to be
  963. * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
  964. * pointing respectively to the first byte of packet data and to
  965. * the byte after the last byte of packet data. However, it
  966. * remains useful if one wishes to read large quantities of data
  967. * at once from a packet into the eBPF stack.
  968. * Return
  969. * 0 on success, or a negative error in case of failure.
  970. *
  971. * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
  972. * Description
  973. * Walk a user or a kernel stack and return its id. To achieve
  974. * this, the helper needs *ctx*, which is a pointer to the context
  975. * on which the tracing program is executed, and a pointer to a
  976. * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
  977. *
  978. * The last argument, *flags*, holds the number of stack frames to
  979. * skip (from 0 to 255), masked with
  980. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  981. * a combination of the following flags:
  982. *
  983. * **BPF_F_USER_STACK**
  984. * Collect a user space stack instead of a kernel stack.
  985. * **BPF_F_FAST_STACK_CMP**
  986. * Compare stacks by hash only.
  987. * **BPF_F_REUSE_STACKID**
  988. * If two different stacks hash into the same *stackid*,
  989. * discard the old one.
  990. *
  991. * The stack id retrieved is a 32 bit long integer handle which
  992. * can be further combined with other data (including other stack
  993. * ids) and used as a key into maps. This can be useful for
  994. * generating a variety of graphs (such as flame graphs or off-cpu
  995. * graphs).
  996. *
  997. * For walking a stack, this helper is an improvement over
  998. * **bpf_probe_read**\ (), which can be used with unrolled loops
  999. * but is not efficient and consumes a lot of eBPF instructions.
  1000. * Instead, **bpf_get_stackid**\ () can collect up to
  1001. * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
  1002. * this limit can be controlled with the **sysctl** program, and
  1003. * that it should be manually increased in order to profile long
  1004. * user stacks (such as stacks for Java programs). To do so, use:
  1005. *
  1006. * ::
  1007. *
  1008. * # sysctl kernel.perf_event_max_stack=<new value>
  1009. * Return
  1010. * The positive or null stack id on success, or a negative error
  1011. * in case of failure.
  1012. *
  1013. * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
  1014. * Description
  1015. * Compute a checksum difference, from the raw buffer pointed by
  1016. * *from*, of length *from_size* (that must be a multiple of 4),
  1017. * towards the raw buffer pointed by *to*, of size *to_size*
  1018. * (same remark). An optional *seed* can be added to the value
  1019. * (this can be cascaded, the seed may come from a previous call
  1020. * to the helper).
  1021. *
  1022. * This is flexible enough to be used in several ways:
  1023. *
  1024. * * With *from_size* == 0, *to_size* > 0 and *seed* set to
  1025. * checksum, it can be used when pushing new data.
  1026. * * With *from_size* > 0, *to_size* == 0 and *seed* set to
  1027. * checksum, it can be used when removing data from a packet.
  1028. * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
  1029. * can be used to compute a diff. Note that *from_size* and
  1030. * *to_size* do not need to be equal.
  1031. *
  1032. * This helper can be used in combination with
  1033. * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
  1034. * which one can feed in the difference computed with
  1035. * **bpf_csum_diff**\ ().
  1036. * Return
  1037. * The checksum result, or a negative error code in case of
  1038. * failure.
  1039. *
  1040. * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
  1041. * Description
  1042. * Retrieve tunnel options metadata for the packet associated to
  1043. * *skb*, and store the raw tunnel option data to the buffer *opt*
  1044. * of *size*.
  1045. *
  1046. * This helper can be used with encapsulation devices that can
  1047. * operate in "collect metadata" mode (please refer to the related
  1048. * note in the description of **bpf_skb_get_tunnel_key**\ () for
  1049. * more details). A particular example where this can be used is
  1050. * in combination with the Geneve encapsulation protocol, where it
  1051. * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
  1052. * and retrieving arbitrary TLVs (Type-Length-Value headers) from
  1053. * the eBPF program. This allows for full customization of these
  1054. * headers.
  1055. * Return
  1056. * The size of the option data retrieved.
  1057. *
  1058. * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
  1059. * Description
  1060. * Set tunnel options metadata for the packet associated to *skb*
  1061. * to the option data contained in the raw buffer *opt* of *size*.
  1062. *
  1063. * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
  1064. * helper for additional information.
  1065. * Return
  1066. * 0 on success, or a negative error in case of failure.
  1067. *
  1068. * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
  1069. * Description
  1070. * Change the protocol of the *skb* to *proto*. Currently
  1071. * supported are transition from IPv4 to IPv6, and from IPv6 to
  1072. * IPv4. The helper takes care of the groundwork for the
  1073. * transition, including resizing the socket buffer. The eBPF
  1074. * program is expected to fill the new headers, if any, via
  1075. * **skb_store_bytes**\ () and to recompute the checksums with
  1076. * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
  1077. * (). The main case for this helper is to perform NAT64
  1078. * operations out of an eBPF program.
  1079. *
  1080. * Internally, the GSO type is marked as dodgy so that headers are
  1081. * checked and segments are recalculated by the GSO/GRO engine.
  1082. * The size for GSO target is adapted as well.
  1083. *
  1084. * All values for *flags* are reserved for future usage, and must
  1085. * be left at zero.
  1086. *
  1087. * A call to this helper is susceptible to change the underlaying
  1088. * packet buffer. Therefore, at load time, all checks on pointers
  1089. * previously done by the verifier are invalidated and must be
  1090. * performed again, if the helper is used in combination with
  1091. * direct packet access.
  1092. * Return
  1093. * 0 on success, or a negative error in case of failure.
  1094. *
  1095. * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
  1096. * Description
  1097. * Change the packet type for the packet associated to *skb*. This
  1098. * comes down to setting *skb*\ **->pkt_type** to *type*, except
  1099. * the eBPF program does not have a write access to *skb*\
  1100. * **->pkt_type** beside this helper. Using a helper here allows
  1101. * for graceful handling of errors.
  1102. *
  1103. * The major use case is to change incoming *skb*s to
  1104. * **PACKET_HOST** in a programmatic way instead of having to
  1105. * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
  1106. * example.
  1107. *
  1108. * Note that *type* only allows certain values. At this time, they
  1109. * are:
  1110. *
  1111. * **PACKET_HOST**
  1112. * Packet is for us.
  1113. * **PACKET_BROADCAST**
  1114. * Send packet to all.
  1115. * **PACKET_MULTICAST**
  1116. * Send packet to group.
  1117. * **PACKET_OTHERHOST**
  1118. * Send packet to someone else.
  1119. * Return
  1120. * 0 on success, or a negative error in case of failure.
  1121. *
  1122. * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
  1123. * Description
  1124. * Check whether *skb* is a descendant of the cgroup2 held by
  1125. * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  1126. * Return
  1127. * The return value depends on the result of the test, and can be:
  1128. *
  1129. * * 0, if the *skb* failed the cgroup2 descendant test.
  1130. * * 1, if the *skb* succeeded the cgroup2 descendant test.
  1131. * * A negative error code, if an error occurred.
  1132. *
  1133. * u32 bpf_get_hash_recalc(struct sk_buff *skb)
  1134. * Description
  1135. * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
  1136. * not set, in particular if the hash was cleared due to mangling,
  1137. * recompute this hash. Later accesses to the hash can be done
  1138. * directly with *skb*\ **->hash**.
  1139. *
  1140. * Calling **bpf_set_hash_invalid**\ (), changing a packet
  1141. * prototype with **bpf_skb_change_proto**\ (), or calling
  1142. * **bpf_skb_store_bytes**\ () with the
  1143. * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
  1144. * the hash and to trigger a new computation for the next call to
  1145. * **bpf_get_hash_recalc**\ ().
  1146. * Return
  1147. * The 32-bit hash.
  1148. *
  1149. * u64 bpf_get_current_task(void)
  1150. * Return
  1151. * A pointer to the current task struct.
  1152. *
  1153. * int bpf_probe_write_user(void *dst, const void *src, u32 len)
  1154. * Description
  1155. * Attempt in a safe way to write *len* bytes from the buffer
  1156. * *src* to *dst* in memory. It only works for threads that are in
  1157. * user context, and *dst* must be a valid user space address.
  1158. *
  1159. * This helper should not be used to implement any kind of
  1160. * security mechanism because of TOC-TOU attacks, but rather to
  1161. * debug, divert, and manipulate execution of semi-cooperative
  1162. * processes.
  1163. *
  1164. * Keep in mind that this feature is meant for experiments, and it
  1165. * has a risk of crashing the system and running programs.
  1166. * Therefore, when an eBPF program using this helper is attached,
  1167. * a warning including PID and process name is printed to kernel
  1168. * logs.
  1169. * Return
  1170. * 0 on success, or a negative error in case of failure.
  1171. *
  1172. * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
  1173. * Description
  1174. * Check whether the probe is being run is the context of a given
  1175. * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
  1176. * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  1177. * Return
  1178. * The return value depends on the result of the test, and can be:
  1179. *
  1180. * * 0, if the *skb* task belongs to the cgroup2.
  1181. * * 1, if the *skb* task does not belong to the cgroup2.
  1182. * * A negative error code, if an error occurred.
  1183. *
  1184. * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
  1185. * Description
  1186. * Resize (trim or grow) the packet associated to *skb* to the
  1187. * new *len*. The *flags* are reserved for future usage, and must
  1188. * be left at zero.
  1189. *
  1190. * The basic idea is that the helper performs the needed work to
  1191. * change the size of the packet, then the eBPF program rewrites
  1192. * the rest via helpers like **bpf_skb_store_bytes**\ (),
  1193. * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
  1194. * and others. This helper is a slow path utility intended for
  1195. * replies with control messages. And because it is targeted for
  1196. * slow path, the helper itself can afford to be slow: it
  1197. * implicitly linearizes, unclones and drops offloads from the
  1198. * *skb*.
  1199. *
  1200. * A call to this helper is susceptible to change the underlaying
  1201. * packet buffer. Therefore, at load time, all checks on pointers
  1202. * previously done by the verifier are invalidated and must be
  1203. * performed again, if the helper is used in combination with
  1204. * direct packet access.
  1205. * Return
  1206. * 0 on success, or a negative error in case of failure.
  1207. *
  1208. * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
  1209. * Description
  1210. * Pull in non-linear data in case the *skb* is non-linear and not
  1211. * all of *len* are part of the linear section. Make *len* bytes
  1212. * from *skb* readable and writable. If a zero value is passed for
  1213. * *len*, then the whole length of the *skb* is pulled.
  1214. *
  1215. * This helper is only needed for reading and writing with direct
  1216. * packet access.
  1217. *
  1218. * For direct packet access, testing that offsets to access
  1219. * are within packet boundaries (test on *skb*\ **->data_end**) is
  1220. * susceptible to fail if offsets are invalid, or if the requested
  1221. * data is in non-linear parts of the *skb*. On failure the
  1222. * program can just bail out, or in the case of a non-linear
  1223. * buffer, use a helper to make the data available. The
  1224. * **bpf_skb_load_bytes**\ () helper is a first solution to access
  1225. * the data. Another one consists in using **bpf_skb_pull_data**
  1226. * to pull in once the non-linear parts, then retesting and
  1227. * eventually access the data.
  1228. *
  1229. * At the same time, this also makes sure the *skb* is uncloned,
  1230. * which is a necessary condition for direct write. As this needs
  1231. * to be an invariant for the write part only, the verifier
  1232. * detects writes and adds a prologue that is calling
  1233. * **bpf_skb_pull_data()** to effectively unclone the *skb* from
  1234. * the very beginning in case it is indeed cloned.
  1235. *
  1236. * A call to this helper is susceptible to change the underlaying
  1237. * packet buffer. Therefore, at load time, all checks on pointers
  1238. * previously done by the verifier are invalidated and must be
  1239. * performed again, if the helper is used in combination with
  1240. * direct packet access.
  1241. * Return
  1242. * 0 on success, or a negative error in case of failure.
  1243. *
  1244. * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
  1245. * Description
  1246. * Add the checksum *csum* into *skb*\ **->csum** in case the
  1247. * driver has supplied a checksum for the entire packet into that
  1248. * field. Return an error otherwise. This helper is intended to be
  1249. * used in combination with **bpf_csum_diff**\ (), in particular
  1250. * when the checksum needs to be updated after data has been
  1251. * written into the packet through direct packet access.
  1252. * Return
  1253. * The checksum on success, or a negative error code in case of
  1254. * failure.
  1255. *
  1256. * void bpf_set_hash_invalid(struct sk_buff *skb)
  1257. * Description
  1258. * Invalidate the current *skb*\ **->hash**. It can be used after
  1259. * mangling on headers through direct packet access, in order to
  1260. * indicate that the hash is outdated and to trigger a
  1261. * recalculation the next time the kernel tries to access this
  1262. * hash or when the **bpf_get_hash_recalc**\ () helper is called.
  1263. *
  1264. * int bpf_get_numa_node_id(void)
  1265. * Description
  1266. * Return the id of the current NUMA node. The primary use case
  1267. * for this helper is the selection of sockets for the local NUMA
  1268. * node, when the program is attached to sockets using the
  1269. * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
  1270. * but the helper is also available to other eBPF program types,
  1271. * similarly to **bpf_get_smp_processor_id**\ ().
  1272. * Return
  1273. * The id of current NUMA node.
  1274. *
  1275. * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
  1276. * Description
  1277. * Grows headroom of packet associated to *skb* and adjusts the
  1278. * offset of the MAC header accordingly, adding *len* bytes of
  1279. * space. It automatically extends and reallocates memory as
  1280. * required.
  1281. *
  1282. * This helper can be used on a layer 3 *skb* to push a MAC header
  1283. * for redirection into a layer 2 device.
  1284. *
  1285. * All values for *flags* are reserved for future usage, and must
  1286. * be left at zero.
  1287. *
  1288. * A call to this helper is susceptible to change the underlaying
  1289. * packet buffer. Therefore, at load time, all checks on pointers
  1290. * previously done by the verifier are invalidated and must be
  1291. * performed again, if the helper is used in combination with
  1292. * direct packet access.
  1293. * Return
  1294. * 0 on success, or a negative error in case of failure.
  1295. *
  1296. * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
  1297. * Description
  1298. * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
  1299. * it is possible to use a negative value for *delta*. This helper
  1300. * can be used to prepare the packet for pushing or popping
  1301. * headers.
  1302. *
  1303. * A call to this helper is susceptible to change the underlaying
  1304. * packet buffer. Therefore, at load time, all checks on pointers
  1305. * previously done by the verifier are invalidated and must be
  1306. * performed again, if the helper is used in combination with
  1307. * direct packet access.
  1308. * Return
  1309. * 0 on success, or a negative error in case of failure.
  1310. *
  1311. * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
  1312. * Description
  1313. * Copy a NUL terminated string from an unsafe address
  1314. * *unsafe_ptr* to *dst*. The *size* should include the
  1315. * terminating NUL byte. In case the string length is smaller than
  1316. * *size*, the target is not padded with further NUL bytes. If the
  1317. * string length is larger than *size*, just *size*-1 bytes are
  1318. * copied and the last byte is set to NUL.
  1319. *
  1320. * On success, the length of the copied string is returned. This
  1321. * makes this helper useful in tracing programs for reading
  1322. * strings, and more importantly to get its length at runtime. See
  1323. * the following snippet:
  1324. *
  1325. * ::
  1326. *
  1327. * SEC("kprobe/sys_open")
  1328. * void bpf_sys_open(struct pt_regs *ctx)
  1329. * {
  1330. * char buf[PATHLEN]; // PATHLEN is defined to 256
  1331. * int res = bpf_probe_read_str(buf, sizeof(buf),
  1332. * ctx->di);
  1333. *
  1334. * // Consume buf, for example push it to
  1335. * // userspace via bpf_perf_event_output(); we
  1336. * // can use res (the string length) as event
  1337. * // size, after checking its boundaries.
  1338. * }
  1339. *
  1340. * In comparison, using **bpf_probe_read()** helper here instead
  1341. * to read the string would require to estimate the length at
  1342. * compile time, and would often result in copying more memory
  1343. * than necessary.
  1344. *
  1345. * Another useful use case is when parsing individual process
  1346. * arguments or individual environment variables navigating
  1347. * *current*\ **->mm->arg_start** and *current*\
  1348. * **->mm->env_start**: using this helper and the return value,
  1349. * one can quickly iterate at the right offset of the memory area.
  1350. * Return
  1351. * On success, the strictly positive length of the string,
  1352. * including the trailing NUL character. On error, a negative
  1353. * value.
  1354. *
  1355. * u64 bpf_get_socket_cookie(struct sk_buff *skb)
  1356. * Description
  1357. * If the **struct sk_buff** pointed by *skb* has a known socket,
  1358. * retrieve the cookie (generated by the kernel) of this socket.
  1359. * If no cookie has been set yet, generate a new cookie. Once
  1360. * generated, the socket cookie remains stable for the life of the
  1361. * socket. This helper can be useful for monitoring per socket
  1362. * networking traffic statistics as it provides a unique socket
  1363. * identifier per namespace.
  1364. * Return
  1365. * A 8-byte long non-decreasing number on success, or 0 if the
  1366. * socket field is missing inside *skb*.
  1367. *
  1368. * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
  1369. * Description
  1370. * Equivalent to bpf_get_socket_cookie() helper that accepts
  1371. * *skb*, but gets socket from **struct bpf_sock_addr** contex.
  1372. * Return
  1373. * A 8-byte long non-decreasing number.
  1374. *
  1375. * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
  1376. * Description
  1377. * Equivalent to bpf_get_socket_cookie() helper that accepts
  1378. * *skb*, but gets socket from **struct bpf_sock_ops** contex.
  1379. * Return
  1380. * A 8-byte long non-decreasing number.
  1381. *
  1382. * u32 bpf_get_socket_uid(struct sk_buff *skb)
  1383. * Return
  1384. * The owner UID of the socket associated to *skb*. If the socket
  1385. * is **NULL**, or if it is not a full socket (i.e. if it is a
  1386. * time-wait or a request socket instead), **overflowuid** value
  1387. * is returned (note that **overflowuid** might also be the actual
  1388. * UID value for the socket).
  1389. *
  1390. * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
  1391. * Description
  1392. * Set the full hash for *skb* (set the field *skb*\ **->hash**)
  1393. * to value *hash*.
  1394. * Return
  1395. * 0
  1396. *
  1397. * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
  1398. * Description
  1399. * Emulate a call to **setsockopt()** on the socket associated to
  1400. * *bpf_socket*, which must be a full socket. The *level* at
  1401. * which the option resides and the name *optname* of the option
  1402. * must be specified, see **setsockopt(2)** for more information.
  1403. * The option value of length *optlen* is pointed by *optval*.
  1404. *
  1405. * This helper actually implements a subset of **setsockopt()**.
  1406. * It supports the following *level*\ s:
  1407. *
  1408. * * **SOL_SOCKET**, which supports the following *optname*\ s:
  1409. * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
  1410. * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
  1411. * * **IPPROTO_TCP**, which supports the following *optname*\ s:
  1412. * **TCP_CONGESTION**, **TCP_BPF_IW**,
  1413. * **TCP_BPF_SNDCWND_CLAMP**.
  1414. * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
  1415. * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
  1416. * Return
  1417. * 0 on success, or a negative error in case of failure.
  1418. *
  1419. * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags)
  1420. * Description
  1421. * Grow or shrink the room for data in the packet associated to
  1422. * *skb* by *len_diff*, and according to the selected *mode*.
  1423. *
  1424. * There is a single supported mode at this time:
  1425. *
  1426. * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  1427. * (room space is added or removed below the layer 3 header).
  1428. *
  1429. * All values for *flags* are reserved for future usage, and must
  1430. * be left at zero.
  1431. *
  1432. * A call to this helper is susceptible to change the underlaying
  1433. * packet buffer. Therefore, at load time, all checks on pointers
  1434. * previously done by the verifier are invalidated and must be
  1435. * performed again, if the helper is used in combination with
  1436. * direct packet access.
  1437. * Return
  1438. * 0 on success, or a negative error in case of failure.
  1439. *
  1440. * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
  1441. * Description
  1442. * Redirect the packet to the endpoint referenced by *map* at
  1443. * index *key*. Depending on its type, this *map* can contain
  1444. * references to net devices (for forwarding packets through other
  1445. * ports), or to CPUs (for redirecting XDP frames to another CPU;
  1446. * but this is only implemented for native XDP (with driver
  1447. * support) as of this writing).
  1448. *
  1449. * All values for *flags* are reserved for future usage, and must
  1450. * be left at zero.
  1451. *
  1452. * When used to redirect packets to net devices, this helper
  1453. * provides a high performance increase over **bpf_redirect**\ ().
  1454. * This is due to various implementation details of the underlying
  1455. * mechanisms, one of which is the fact that **bpf_redirect_map**\
  1456. * () tries to send packet as a "bulk" to the device.
  1457. * Return
  1458. * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
  1459. *
  1460. * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
  1461. * Description
  1462. * Redirect the packet to the socket referenced by *map* (of type
  1463. * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  1464. * egress interfaces can be used for redirection. The
  1465. * **BPF_F_INGRESS** value in *flags* is used to make the
  1466. * distinction (ingress path is selected if the flag is present,
  1467. * egress path otherwise). This is the only flag supported for now.
  1468. * Return
  1469. * **SK_PASS** on success, or **SK_DROP** on error.
  1470. *
  1471. * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
  1472. * Description
  1473. * Add an entry to, or update a *map* referencing sockets. The
  1474. * *skops* is used as a new value for the entry associated to
  1475. * *key*. *flags* is one of:
  1476. *
  1477. * **BPF_NOEXIST**
  1478. * The entry for *key* must not exist in the map.
  1479. * **BPF_EXIST**
  1480. * The entry for *key* must already exist in the map.
  1481. * **BPF_ANY**
  1482. * No condition on the existence of the entry for *key*.
  1483. *
  1484. * If the *map* has eBPF programs (parser and verdict), those will
  1485. * be inherited by the socket being added. If the socket is
  1486. * already attached to eBPF programs, this results in an error.
  1487. * Return
  1488. * 0 on success, or a negative error in case of failure.
  1489. *
  1490. * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
  1491. * Description
  1492. * Adjust the address pointed by *xdp_md*\ **->data_meta** by
  1493. * *delta* (which can be positive or negative). Note that this
  1494. * operation modifies the address stored in *xdp_md*\ **->data**,
  1495. * so the latter must be loaded only after the helper has been
  1496. * called.
  1497. *
  1498. * The use of *xdp_md*\ **->data_meta** is optional and programs
  1499. * are not required to use it. The rationale is that when the
  1500. * packet is processed with XDP (e.g. as DoS filter), it is
  1501. * possible to push further meta data along with it before passing
  1502. * to the stack, and to give the guarantee that an ingress eBPF
  1503. * program attached as a TC classifier on the same device can pick
  1504. * this up for further post-processing. Since TC works with socket
  1505. * buffers, it remains possible to set from XDP the **mark** or
  1506. * **priority** pointers, or other pointers for the socket buffer.
  1507. * Having this scratch space generic and programmable allows for
  1508. * more flexibility as the user is free to store whatever meta
  1509. * data they need.
  1510. *
  1511. * A call to this helper is susceptible to change the underlaying
  1512. * packet buffer. Therefore, at load time, all checks on pointers
  1513. * previously done by the verifier are invalidated and must be
  1514. * performed again, if the helper is used in combination with
  1515. * direct packet access.
  1516. * Return
  1517. * 0 on success, or a negative error in case of failure.
  1518. *
  1519. * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
  1520. * Description
  1521. * Read the value of a perf event counter, and store it into *buf*
  1522. * of size *buf_size*. This helper relies on a *map* of type
  1523. * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
  1524. * counter is selected when *map* is updated with perf event file
  1525. * descriptors. The *map* is an array whose size is the number of
  1526. * available CPUs, and each cell contains a value relative to one
  1527. * CPU. The value to retrieve is indicated by *flags*, that
  1528. * contains the index of the CPU to look up, masked with
  1529. * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  1530. * **BPF_F_CURRENT_CPU** to indicate that the value for the
  1531. * current CPU should be retrieved.
  1532. *
  1533. * This helper behaves in a way close to
  1534. * **bpf_perf_event_read**\ () helper, save that instead of
  1535. * just returning the value observed, it fills the *buf*
  1536. * structure. This allows for additional data to be retrieved: in
  1537. * particular, the enabled and running times (in *buf*\
  1538. * **->enabled** and *buf*\ **->running**, respectively) are
  1539. * copied. In general, **bpf_perf_event_read_value**\ () is
  1540. * recommended over **bpf_perf_event_read**\ (), which has some
  1541. * ABI issues and provides fewer functionalities.
  1542. *
  1543. * These values are interesting, because hardware PMU (Performance
  1544. * Monitoring Unit) counters are limited resources. When there are
  1545. * more PMU based perf events opened than available counters,
  1546. * kernel will multiplex these events so each event gets certain
  1547. * percentage (but not all) of the PMU time. In case that
  1548. * multiplexing happens, the number of samples or counter value
  1549. * will not reflect the case compared to when no multiplexing
  1550. * occurs. This makes comparison between different runs difficult.
  1551. * Typically, the counter value should be normalized before
  1552. * comparing to other experiments. The usual normalization is done
  1553. * as follows.
  1554. *
  1555. * ::
  1556. *
  1557. * normalized_counter = counter * t_enabled / t_running
  1558. *
  1559. * Where t_enabled is the time enabled for event and t_running is
  1560. * the time running for event since last normalization. The
  1561. * enabled and running times are accumulated since the perf event
  1562. * open. To achieve scaling factor between two invocations of an
  1563. * eBPF program, users can can use CPU id as the key (which is
  1564. * typical for perf array usage model) to remember the previous
  1565. * value and do the calculation inside the eBPF program.
  1566. * Return
  1567. * 0 on success, or a negative error in case of failure.
  1568. *
  1569. * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
  1570. * Description
  1571. * For en eBPF program attached to a perf event, retrieve the
  1572. * value of the event counter associated to *ctx* and store it in
  1573. * the structure pointed by *buf* and of size *buf_size*. Enabled
  1574. * and running times are also stored in the structure (see
  1575. * description of helper **bpf_perf_event_read_value**\ () for
  1576. * more details).
  1577. * Return
  1578. * 0 on success, or a negative error in case of failure.
  1579. *
  1580. * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
  1581. * Description
  1582. * Emulate a call to **getsockopt()** on the socket associated to
  1583. * *bpf_socket*, which must be a full socket. The *level* at
  1584. * which the option resides and the name *optname* of the option
  1585. * must be specified, see **getsockopt(2)** for more information.
  1586. * The retrieved value is stored in the structure pointed by
  1587. * *opval* and of length *optlen*.
  1588. *
  1589. * This helper actually implements a subset of **getsockopt()**.
  1590. * It supports the following *level*\ s:
  1591. *
  1592. * * **IPPROTO_TCP**, which supports *optname*
  1593. * **TCP_CONGESTION**.
  1594. * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
  1595. * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
  1596. * Return
  1597. * 0 on success, or a negative error in case of failure.
  1598. *
  1599. * int bpf_override_return(struct pt_reg *regs, u64 rc)
  1600. * Description
  1601. * Used for error injection, this helper uses kprobes to override
  1602. * the return value of the probed function, and to set it to *rc*.
  1603. * The first argument is the context *regs* on which the kprobe
  1604. * works.
  1605. *
  1606. * This helper works by setting setting the PC (program counter)
  1607. * to an override function which is run in place of the original
  1608. * probed function. This means the probed function is not run at
  1609. * all. The replacement function just returns with the required
  1610. * value.
  1611. *
  1612. * This helper has security implications, and thus is subject to
  1613. * restrictions. It is only available if the kernel was compiled
  1614. * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
  1615. * option, and in this case it only works on functions tagged with
  1616. * **ALLOW_ERROR_INJECTION** in the kernel code.
  1617. *
  1618. * Also, the helper is only available for the architectures having
  1619. * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
  1620. * x86 architecture is the only one to support this feature.
  1621. * Return
  1622. * 0
  1623. *
  1624. * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
  1625. * Description
  1626. * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
  1627. * for the full TCP socket associated to *bpf_sock_ops* to
  1628. * *argval*.
  1629. *
  1630. * The primary use of this field is to determine if there should
  1631. * be calls to eBPF programs of type
  1632. * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
  1633. * code. A program of the same type can change its value, per
  1634. * connection and as necessary, when the connection is
  1635. * established. This field is directly accessible for reading, but
  1636. * this helper must be used for updates in order to return an
  1637. * error if an eBPF program tries to set a callback that is not
  1638. * supported in the current kernel.
  1639. *
  1640. * The supported callback values that *argval* can combine are:
  1641. *
  1642. * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
  1643. * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
  1644. * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
  1645. *
  1646. * Here are some examples of where one could call such eBPF
  1647. * program:
  1648. *
  1649. * * When RTO fires.
  1650. * * When a packet is retransmitted.
  1651. * * When the connection terminates.
  1652. * * When a packet is sent.
  1653. * * When a packet is received.
  1654. * Return
  1655. * Code **-EINVAL** if the socket is not a full TCP socket;
  1656. * otherwise, a positive number containing the bits that could not
  1657. * be set is returned (which comes down to 0 if all bits were set
  1658. * as required).
  1659. *
  1660. * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
  1661. * Description
  1662. * This helper is used in programs implementing policies at the
  1663. * socket level. If the message *msg* is allowed to pass (i.e. if
  1664. * the verdict eBPF program returns **SK_PASS**), redirect it to
  1665. * the socket referenced by *map* (of type
  1666. * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  1667. * egress interfaces can be used for redirection. The
  1668. * **BPF_F_INGRESS** value in *flags* is used to make the
  1669. * distinction (ingress path is selected if the flag is present,
  1670. * egress path otherwise). This is the only flag supported for now.
  1671. * Return
  1672. * **SK_PASS** on success, or **SK_DROP** on error.
  1673. *
  1674. * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
  1675. * Description
  1676. * For socket policies, apply the verdict of the eBPF program to
  1677. * the next *bytes* (number of bytes) of message *msg*.
  1678. *
  1679. * For example, this helper can be used in the following cases:
  1680. *
  1681. * * A single **sendmsg**\ () or **sendfile**\ () system call
  1682. * contains multiple logical messages that the eBPF program is
  1683. * supposed to read and for which it should apply a verdict.
  1684. * * An eBPF program only cares to read the first *bytes* of a
  1685. * *msg*. If the message has a large payload, then setting up
  1686. * and calling the eBPF program repeatedly for all bytes, even
  1687. * though the verdict is already known, would create unnecessary
  1688. * overhead.
  1689. *
  1690. * When called from within an eBPF program, the helper sets a
  1691. * counter internal to the BPF infrastructure, that is used to
  1692. * apply the last verdict to the next *bytes*. If *bytes* is
  1693. * smaller than the current data being processed from a
  1694. * **sendmsg**\ () or **sendfile**\ () system call, the first
  1695. * *bytes* will be sent and the eBPF program will be re-run with
  1696. * the pointer for start of data pointing to byte number *bytes*
  1697. * **+ 1**. If *bytes* is larger than the current data being
  1698. * processed, then the eBPF verdict will be applied to multiple
  1699. * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
  1700. * consumed.
  1701. *
  1702. * Note that if a socket closes with the internal counter holding
  1703. * a non-zero value, this is not a problem because data is not
  1704. * being buffered for *bytes* and is sent as it is received.
  1705. * Return
  1706. * 0
  1707. *
  1708. * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
  1709. * Description
  1710. * For socket policies, prevent the execution of the verdict eBPF
  1711. * program for message *msg* until *bytes* (byte number) have been
  1712. * accumulated.
  1713. *
  1714. * This can be used when one needs a specific number of bytes
  1715. * before a verdict can be assigned, even if the data spans
  1716. * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
  1717. * case would be a user calling **sendmsg**\ () repeatedly with
  1718. * 1-byte long message segments. Obviously, this is bad for
  1719. * performance, but it is still valid. If the eBPF program needs
  1720. * *bytes* bytes to validate a header, this helper can be used to
  1721. * prevent the eBPF program to be called again until *bytes* have
  1722. * been accumulated.
  1723. * Return
  1724. * 0
  1725. *
  1726. * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
  1727. * Description
  1728. * For socket policies, pull in non-linear data from user space
  1729. * for *msg* and set pointers *msg*\ **->data** and *msg*\
  1730. * **->data_end** to *start* and *end* bytes offsets into *msg*,
  1731. * respectively.
  1732. *
  1733. * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
  1734. * *msg* it can only parse data that the (**data**, **data_end**)
  1735. * pointers have already consumed. For **sendmsg**\ () hooks this
  1736. * is likely the first scatterlist element. But for calls relying
  1737. * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
  1738. * be the range (**0**, **0**) because the data is shared with
  1739. * user space and by default the objective is to avoid allowing
  1740. * user space to modify data while (or after) eBPF verdict is
  1741. * being decided. This helper can be used to pull in data and to
  1742. * set the start and end pointer to given values. Data will be
  1743. * copied if necessary (i.e. if data was not linear and if start
  1744. * and end pointers do not point to the same chunk).
  1745. *
  1746. * A call to this helper is susceptible to change the underlaying
  1747. * packet buffer. Therefore, at load time, all checks on pointers
  1748. * previously done by the verifier are invalidated and must be
  1749. * performed again, if the helper is used in combination with
  1750. * direct packet access.
  1751. *
  1752. * All values for *flags* are reserved for future usage, and must
  1753. * be left at zero.
  1754. * Return
  1755. * 0 on success, or a negative error in case of failure.
  1756. *
  1757. * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
  1758. * Description
  1759. * Bind the socket associated to *ctx* to the address pointed by
  1760. * *addr*, of length *addr_len*. This allows for making outgoing
  1761. * connection from the desired IP address, which can be useful for
  1762. * example when all processes inside a cgroup should use one
  1763. * single IP address on a host that has multiple IP configured.
  1764. *
  1765. * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
  1766. * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
  1767. * **AF_INET6**). Looking for a free port to bind to can be
  1768. * expensive, therefore binding to port is not permitted by the
  1769. * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
  1770. * must be set to zero.
  1771. * Return
  1772. * 0 on success, or a negative error in case of failure.
  1773. *
  1774. * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
  1775. * Description
  1776. * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
  1777. * only possible to shrink the packet as of this writing,
  1778. * therefore *delta* must be a negative integer.
  1779. *
  1780. * A call to this helper is susceptible to change the underlaying
  1781. * packet buffer. Therefore, at load time, all checks on pointers
  1782. * previously done by the verifier are invalidated and must be
  1783. * performed again, if the helper is used in combination with
  1784. * direct packet access.
  1785. * Return
  1786. * 0 on success, or a negative error in case of failure.
  1787. *
  1788. * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
  1789. * Description
  1790. * Retrieve the XFRM state (IP transform framework, see also
  1791. * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
  1792. *
  1793. * The retrieved value is stored in the **struct bpf_xfrm_state**
  1794. * pointed by *xfrm_state* and of length *size*.
  1795. *
  1796. * All values for *flags* are reserved for future usage, and must
  1797. * be left at zero.
  1798. *
  1799. * This helper is available only if the kernel was compiled with
  1800. * **CONFIG_XFRM** configuration option.
  1801. * Return
  1802. * 0 on success, or a negative error in case of failure.
  1803. *
  1804. * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
  1805. * Description
  1806. * Return a user or a kernel stack in bpf program provided buffer.
  1807. * To achieve this, the helper needs *ctx*, which is a pointer
  1808. * to the context on which the tracing program is executed.
  1809. * To store the stacktrace, the bpf program provides *buf* with
  1810. * a nonnegative *size*.
  1811. *
  1812. * The last argument, *flags*, holds the number of stack frames to
  1813. * skip (from 0 to 255), masked with
  1814. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  1815. * the following flags:
  1816. *
  1817. * **BPF_F_USER_STACK**
  1818. * Collect a user space stack instead of a kernel stack.
  1819. * **BPF_F_USER_BUILD_ID**
  1820. * Collect buildid+offset instead of ips for user stack,
  1821. * only valid if **BPF_F_USER_STACK** is also specified.
  1822. *
  1823. * **bpf_get_stack**\ () can collect up to
  1824. * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
  1825. * to sufficient large buffer size. Note that
  1826. * this limit can be controlled with the **sysctl** program, and
  1827. * that it should be manually increased in order to profile long
  1828. * user stacks (such as stacks for Java programs). To do so, use:
  1829. *
  1830. * ::
  1831. *
  1832. * # sysctl kernel.perf_event_max_stack=<new value>
  1833. * Return
  1834. * A non-negative value equal to or less than *size* on success,
  1835. * or a negative error in case of failure.
  1836. *
  1837. * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
  1838. * Description
  1839. * This helper is similar to **bpf_skb_load_bytes**\ () in that
  1840. * it provides an easy way to load *len* bytes from *offset*
  1841. * from the packet associated to *skb*, into the buffer pointed
  1842. * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
  1843. * a fifth argument *start_header* exists in order to select a
  1844. * base offset to start from. *start_header* can be one of:
  1845. *
  1846. * **BPF_HDR_START_MAC**
  1847. * Base offset to load data from is *skb*'s mac header.
  1848. * **BPF_HDR_START_NET**
  1849. * Base offset to load data from is *skb*'s network header.
  1850. *
  1851. * In general, "direct packet access" is the preferred method to
  1852. * access packet data, however, this helper is in particular useful
  1853. * in socket filters where *skb*\ **->data** does not always point
  1854. * to the start of the mac header and where "direct packet access"
  1855. * is not available.
  1856. * Return
  1857. * 0 on success, or a negative error in case of failure.
  1858. *
  1859. * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
  1860. * Description
  1861. * Do FIB lookup in kernel tables using parameters in *params*.
  1862. * If lookup is successful and result shows packet is to be
  1863. * forwarded, the neighbor tables are searched for the nexthop.
  1864. * If successful (ie., FIB lookup shows forwarding and nexthop
  1865. * is resolved), the nexthop address is returned in ipv4_dst
  1866. * or ipv6_dst based on family, smac is set to mac address of
  1867. * egress device, dmac is set to nexthop mac address, rt_metric
  1868. * is set to metric from route (IPv4/IPv6 only), and ifindex
  1869. * is set to the device index of the nexthop from the FIB lookup.
  1870. *
  1871. * *plen* argument is the size of the passed in struct.
  1872. * *flags* argument can be a combination of one or more of the
  1873. * following values:
  1874. *
  1875. * **BPF_FIB_LOOKUP_DIRECT**
  1876. * Do a direct table lookup vs full lookup using FIB
  1877. * rules.
  1878. * **BPF_FIB_LOOKUP_OUTPUT**
  1879. * Perform lookup from an egress perspective (default is
  1880. * ingress).
  1881. *
  1882. * *ctx* is either **struct xdp_md** for XDP programs or
  1883. * **struct sk_buff** tc cls_act programs.
  1884. * Return
  1885. * * < 0 if any input argument is invalid
  1886. * * 0 on success (packet is forwarded, nexthop neighbor exists)
  1887. * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
  1888. * packet is not forwarded or needs assist from full stack
  1889. *
  1890. * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
  1891. * Description
  1892. * Add an entry to, or update a sockhash *map* referencing sockets.
  1893. * The *skops* is used as a new value for the entry associated to
  1894. * *key*. *flags* is one of:
  1895. *
  1896. * **BPF_NOEXIST**
  1897. * The entry for *key* must not exist in the map.
  1898. * **BPF_EXIST**
  1899. * The entry for *key* must already exist in the map.
  1900. * **BPF_ANY**
  1901. * No condition on the existence of the entry for *key*.
  1902. *
  1903. * If the *map* has eBPF programs (parser and verdict), those will
  1904. * be inherited by the socket being added. If the socket is
  1905. * already attached to eBPF programs, this results in an error.
  1906. * Return
  1907. * 0 on success, or a negative error in case of failure.
  1908. *
  1909. * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
  1910. * Description
  1911. * This helper is used in programs implementing policies at the
  1912. * socket level. If the message *msg* is allowed to pass (i.e. if
  1913. * the verdict eBPF program returns **SK_PASS**), redirect it to
  1914. * the socket referenced by *map* (of type
  1915. * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  1916. * egress interfaces can be used for redirection. The
  1917. * **BPF_F_INGRESS** value in *flags* is used to make the
  1918. * distinction (ingress path is selected if the flag is present,
  1919. * egress path otherwise). This is the only flag supported for now.
  1920. * Return
  1921. * **SK_PASS** on success, or **SK_DROP** on error.
  1922. *
  1923. * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
  1924. * Description
  1925. * This helper is used in programs implementing policies at the
  1926. * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
  1927. * if the verdeict eBPF program returns **SK_PASS**), redirect it
  1928. * to the socket referenced by *map* (of type
  1929. * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  1930. * egress interfaces can be used for redirection. The
  1931. * **BPF_F_INGRESS** value in *flags* is used to make the
  1932. * distinction (ingress path is selected if the flag is present,
  1933. * egress otherwise). This is the only flag supported for now.
  1934. * Return
  1935. * **SK_PASS** on success, or **SK_DROP** on error.
  1936. *
  1937. * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
  1938. * Description
  1939. * Encapsulate the packet associated to *skb* within a Layer 3
  1940. * protocol header. This header is provided in the buffer at
  1941. * address *hdr*, with *len* its size in bytes. *type* indicates
  1942. * the protocol of the header and can be one of:
  1943. *
  1944. * **BPF_LWT_ENCAP_SEG6**
  1945. * IPv6 encapsulation with Segment Routing Header
  1946. * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
  1947. * the IPv6 header is computed by the kernel.
  1948. * **BPF_LWT_ENCAP_SEG6_INLINE**
  1949. * Only works if *skb* contains an IPv6 packet. Insert a
  1950. * Segment Routing Header (**struct ipv6_sr_hdr**) inside
  1951. * the IPv6 header.
  1952. *
  1953. * A call to this helper is susceptible to change the underlaying
  1954. * packet buffer. Therefore, at load time, all checks on pointers
  1955. * previously done by the verifier are invalidated and must be
  1956. * performed again, if the helper is used in combination with
  1957. * direct packet access.
  1958. * Return
  1959. * 0 on success, or a negative error in case of failure.
  1960. *
  1961. * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
  1962. * Description
  1963. * Store *len* bytes from address *from* into the packet
  1964. * associated to *skb*, at *offset*. Only the flags, tag and TLVs
  1965. * inside the outermost IPv6 Segment Routing Header can be
  1966. * modified through this helper.
  1967. *
  1968. * A call to this helper is susceptible to change the underlaying
  1969. * packet buffer. Therefore, at load time, all checks on pointers
  1970. * previously done by the verifier are invalidated and must be
  1971. * performed again, if the helper is used in combination with
  1972. * direct packet access.
  1973. * Return
  1974. * 0 on success, or a negative error in case of failure.
  1975. *
  1976. * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
  1977. * Description
  1978. * Adjust the size allocated to TLVs in the outermost IPv6
  1979. * Segment Routing Header contained in the packet associated to
  1980. * *skb*, at position *offset* by *delta* bytes. Only offsets
  1981. * after the segments are accepted. *delta* can be as well
  1982. * positive (growing) as negative (shrinking).
  1983. *
  1984. * A call to this helper is susceptible to change the underlaying
  1985. * packet buffer. Therefore, at load time, all checks on pointers
  1986. * previously done by the verifier are invalidated and must be
  1987. * performed again, if the helper is used in combination with
  1988. * direct packet access.
  1989. * Return
  1990. * 0 on success, or a negative error in case of failure.
  1991. *
  1992. * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
  1993. * Description
  1994. * Apply an IPv6 Segment Routing action of type *action* to the
  1995. * packet associated to *skb*. Each action takes a parameter
  1996. * contained at address *param*, and of length *param_len* bytes.
  1997. * *action* can be one of:
  1998. *
  1999. * **SEG6_LOCAL_ACTION_END_X**
  2000. * End.X action: Endpoint with Layer-3 cross-connect.
  2001. * Type of *param*: **struct in6_addr**.
  2002. * **SEG6_LOCAL_ACTION_END_T**
  2003. * End.T action: Endpoint with specific IPv6 table lookup.
  2004. * Type of *param*: **int**.
  2005. * **SEG6_LOCAL_ACTION_END_B6**
  2006. * End.B6 action: Endpoint bound to an SRv6 policy.
  2007. * Type of param: **struct ipv6_sr_hdr**.
  2008. * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
  2009. * End.B6.Encap action: Endpoint bound to an SRv6
  2010. * encapsulation policy.
  2011. * Type of param: **struct ipv6_sr_hdr**.
  2012. *
  2013. * A call to this helper is susceptible to change the underlaying
  2014. * packet buffer. Therefore, at load time, all checks on pointers
  2015. * previously done by the verifier are invalidated and must be
  2016. * performed again, if the helper is used in combination with
  2017. * direct packet access.
  2018. * Return
  2019. * 0 on success, or a negative error in case of failure.
  2020. *
  2021. * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
  2022. * Description
  2023. * This helper is used in programs implementing IR decoding, to
  2024. * report a successfully decoded key press with *scancode*,
  2025. * *toggle* value in the given *protocol*. The scancode will be
  2026. * translated to a keycode using the rc keymap, and reported as
  2027. * an input key down event. After a period a key up event is
  2028. * generated. This period can be extended by calling either
  2029. * **bpf_rc_keydown** () again with the same values, or calling
  2030. * **bpf_rc_repeat** ().
  2031. *
  2032. * Some protocols include a toggle bit, in case the button was
  2033. * released and pressed again between consecutive scancodes.
  2034. *
  2035. * The *ctx* should point to the lirc sample as passed into
  2036. * the program.
  2037. *
  2038. * The *protocol* is the decoded protocol number (see
  2039. * **enum rc_proto** for some predefined values).
  2040. *
  2041. * This helper is only available is the kernel was compiled with
  2042. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  2043. * "**y**".
  2044. * Return
  2045. * 0
  2046. *
  2047. * int bpf_rc_repeat(void *ctx)
  2048. * Description
  2049. * This helper is used in programs implementing IR decoding, to
  2050. * report a successfully decoded repeat key message. This delays
  2051. * the generation of a key up event for previously generated
  2052. * key down event.
  2053. *
  2054. * Some IR protocols like NEC have a special IR message for
  2055. * repeating last button, for when a button is held down.
  2056. *
  2057. * The *ctx* should point to the lirc sample as passed into
  2058. * the program.
  2059. *
  2060. * This helper is only available is the kernel was compiled with
  2061. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  2062. * "**y**".
  2063. * Return
  2064. * 0
  2065. *
  2066. * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
  2067. * Description
  2068. * Return the cgroup v2 id of the socket associated with the *skb*.
  2069. * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
  2070. * helper for cgroup v1 by providing a tag resp. identifier that
  2071. * can be matched on or used for map lookups e.g. to implement
  2072. * policy. The cgroup v2 id of a given path in the hierarchy is
  2073. * exposed in user space through the f_handle API in order to get
  2074. * to the same 64-bit id.
  2075. *
  2076. * This helper can be used on TC egress path, but not on ingress,
  2077. * and is available only if the kernel was compiled with the
  2078. * **CONFIG_SOCK_CGROUP_DATA** configuration option.
  2079. * Return
  2080. * The id is returned or 0 in case the id could not be retrieved.
  2081. *
  2082. * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
  2083. * Description
  2084. * Return id of cgroup v2 that is ancestor of cgroup associated
  2085. * with the *skb* at the *ancestor_level*. The root cgroup is at
  2086. * *ancestor_level* zero and each step down the hierarchy
  2087. * increments the level. If *ancestor_level* == level of cgroup
  2088. * associated with *skb*, then return value will be same as that
  2089. * of **bpf_skb_cgroup_id**\ ().
  2090. *
  2091. * The helper is useful to implement policies based on cgroups
  2092. * that are upper in hierarchy than immediate cgroup associated
  2093. * with *skb*.
  2094. *
  2095. * The format of returned id and helper limitations are same as in
  2096. * **bpf_skb_cgroup_id**\ ().
  2097. * Return
  2098. * The id is returned or 0 in case the id could not be retrieved.
  2099. *
  2100. * u64 bpf_get_current_cgroup_id(void)
  2101. * Return
  2102. * A 64-bit integer containing the current cgroup id based
  2103. * on the cgroup within which the current task is running.
  2104. *
  2105. * void* get_local_storage(void *map, u64 flags)
  2106. * Description
  2107. * Get the pointer to the local storage area.
  2108. * The type and the size of the local storage is defined
  2109. * by the *map* argument.
  2110. * The *flags* meaning is specific for each map type,
  2111. * and has to be 0 for cgroup local storage.
  2112. *
  2113. * Depending on the bpf program type, a local storage area
  2114. * can be shared between multiple instances of the bpf program,
  2115. * running simultaneously.
  2116. *
  2117. * A user should care about the synchronization by himself.
  2118. * For example, by using the BPF_STX_XADD instruction to alter
  2119. * the shared data.
  2120. * Return
  2121. * Pointer to the local storage area.
  2122. *
  2123. * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
  2124. * Description
  2125. * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map
  2126. * It checks the selected sk is matching the incoming
  2127. * request in the skb.
  2128. * Return
  2129. * 0 on success, or a negative error in case of failure.
  2130. *
  2131. * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
  2132. * Description
  2133. * Look for TCP socket matching *tuple*, optionally in a child
  2134. * network namespace *netns*. The return value must be checked,
  2135. * and if non-NULL, released via **bpf_sk_release**\ ().
  2136. *
  2137. * The *ctx* should point to the context of the program, such as
  2138. * the skb or socket (depending on the hook in use). This is used
  2139. * to determine the base network namespace for the lookup.
  2140. *
  2141. * *tuple_size* must be one of:
  2142. *
  2143. * **sizeof**\ (*tuple*\ **->ipv4**)
  2144. * Look for an IPv4 socket.
  2145. * **sizeof**\ (*tuple*\ **->ipv6**)
  2146. * Look for an IPv6 socket.
  2147. *
  2148. * If the *netns* is a negative signed 32-bit integer, then the
  2149. * socket lookup table in the netns associated with the *ctx* will
  2150. * will be used. For the TC hooks, this is the netns of the device
  2151. * in the skb. For socket hooks, this is the netns of the socket.
  2152. * If *netns* is any other signed 32-bit value greater than or
  2153. * equal to zero then it specifies the ID of the netns relative to
  2154. * the netns associated with the *ctx*. *netns* values beyond the
  2155. * range of 32-bit integers are reserved for future use.
  2156. *
  2157. * All values for *flags* are reserved for future usage, and must
  2158. * be left at zero.
  2159. *
  2160. * This helper is available only if the kernel was compiled with
  2161. * **CONFIG_NET** configuration option.
  2162. * Return
  2163. * Pointer to *struct bpf_sock*, or NULL in case of failure.
  2164. * For sockets with reuseport option, the *struct bpf_sock*
  2165. * result is from reuse->socks[] using the hash of the tuple.
  2166. *
  2167. * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags)
  2168. * Description
  2169. * Look for UDP socket matching *tuple*, optionally in a child
  2170. * network namespace *netns*. The return value must be checked,
  2171. * and if non-NULL, released via **bpf_sk_release**\ ().
  2172. *
  2173. * The *ctx* should point to the context of the program, such as
  2174. * the skb or socket (depending on the hook in use). This is used
  2175. * to determine the base network namespace for the lookup.
  2176. *
  2177. * *tuple_size* must be one of:
  2178. *
  2179. * **sizeof**\ (*tuple*\ **->ipv4**)
  2180. * Look for an IPv4 socket.
  2181. * **sizeof**\ (*tuple*\ **->ipv6**)
  2182. * Look for an IPv6 socket.
  2183. *
  2184. * If the *netns* is a negative signed 32-bit integer, then the
  2185. * socket lookup table in the netns associated with the *ctx* will
  2186. * will be used. For the TC hooks, this is the netns of the device
  2187. * in the skb. For socket hooks, this is the netns of the socket.
  2188. * If *netns* is any other signed 32-bit value greater than or
  2189. * equal to zero then it specifies the ID of the netns relative to
  2190. * the netns associated with the *ctx*. *netns* values beyond the
  2191. * range of 32-bit integers are reserved for future use.
  2192. *
  2193. * All values for *flags* are reserved for future usage, and must
  2194. * be left at zero.
  2195. *
  2196. * This helper is available only if the kernel was compiled with
  2197. * **CONFIG_NET** configuration option.
  2198. * Return
  2199. * Pointer to *struct bpf_sock*, or NULL in case of failure.
  2200. * For sockets with reuseport option, the *struct bpf_sock*
  2201. * result is from reuse->socks[] using the hash of the tuple.
  2202. *
  2203. * int bpf_sk_release(struct bpf_sock *sk)
  2204. * Description
  2205. * Release the reference held by *sock*. *sock* must be a non-NULL
  2206. * pointer that was returned from bpf_sk_lookup_xxx\ ().
  2207. * Return
  2208. * 0 on success, or a negative error in case of failure.
  2209. *
  2210. * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags)
  2211. * Description
  2212. * For socket policies, insert *len* bytes into msg at offset
  2213. * *start*.
  2214. *
  2215. * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
  2216. * *msg* it may want to insert metadata or options into the msg.
  2217. * This can later be read and used by any of the lower layer BPF
  2218. * hooks.
  2219. *
  2220. * This helper may fail if under memory pressure (a malloc
  2221. * fails) in these cases BPF programs will get an appropriate
  2222. * error and BPF programs will need to handle them.
  2223. *
  2224. * Return
  2225. * 0 on success, or a negative error in case of failure.
  2226. */
  2227. #define __BPF_FUNC_MAPPER(FN) \
  2228. FN(unspec), \
  2229. FN(map_lookup_elem), \
  2230. FN(map_update_elem), \
  2231. FN(map_delete_elem), \
  2232. FN(probe_read), \
  2233. FN(ktime_get_ns), \
  2234. FN(trace_printk), \
  2235. FN(get_prandom_u32), \
  2236. FN(get_smp_processor_id), \
  2237. FN(skb_store_bytes), \
  2238. FN(l3_csum_replace), \
  2239. FN(l4_csum_replace), \
  2240. FN(tail_call), \
  2241. FN(clone_redirect), \
  2242. FN(get_current_pid_tgid), \
  2243. FN(get_current_uid_gid), \
  2244. FN(get_current_comm), \
  2245. FN(get_cgroup_classid), \
  2246. FN(skb_vlan_push), \
  2247. FN(skb_vlan_pop), \
  2248. FN(skb_get_tunnel_key), \
  2249. FN(skb_set_tunnel_key), \
  2250. FN(perf_event_read), \
  2251. FN(redirect), \
  2252. FN(get_route_realm), \
  2253. FN(perf_event_output), \
  2254. FN(skb_load_bytes), \
  2255. FN(get_stackid), \
  2256. FN(csum_diff), \
  2257. FN(skb_get_tunnel_opt), \
  2258. FN(skb_set_tunnel_opt), \
  2259. FN(skb_change_proto), \
  2260. FN(skb_change_type), \
  2261. FN(skb_under_cgroup), \
  2262. FN(get_hash_recalc), \
  2263. FN(get_current_task), \
  2264. FN(probe_write_user), \
  2265. FN(current_task_under_cgroup), \
  2266. FN(skb_change_tail), \
  2267. FN(skb_pull_data), \
  2268. FN(csum_update), \
  2269. FN(set_hash_invalid), \
  2270. FN(get_numa_node_id), \
  2271. FN(skb_change_head), \
  2272. FN(xdp_adjust_head), \
  2273. FN(probe_read_str), \
  2274. FN(get_socket_cookie), \
  2275. FN(get_socket_uid), \
  2276. FN(set_hash), \
  2277. FN(setsockopt), \
  2278. FN(skb_adjust_room), \
  2279. FN(redirect_map), \
  2280. FN(sk_redirect_map), \
  2281. FN(sock_map_update), \
  2282. FN(xdp_adjust_meta), \
  2283. FN(perf_event_read_value), \
  2284. FN(perf_prog_read_value), \
  2285. FN(getsockopt), \
  2286. FN(override_return), \
  2287. FN(sock_ops_cb_flags_set), \
  2288. FN(msg_redirect_map), \
  2289. FN(msg_apply_bytes), \
  2290. FN(msg_cork_bytes), \
  2291. FN(msg_pull_data), \
  2292. FN(bind), \
  2293. FN(xdp_adjust_tail), \
  2294. FN(skb_get_xfrm_state), \
  2295. FN(get_stack), \
  2296. FN(skb_load_bytes_relative), \
  2297. FN(fib_lookup), \
  2298. FN(sock_hash_update), \
  2299. FN(msg_redirect_hash), \
  2300. FN(sk_redirect_hash), \
  2301. FN(lwt_push_encap), \
  2302. FN(lwt_seg6_store_bytes), \
  2303. FN(lwt_seg6_adjust_srh), \
  2304. FN(lwt_seg6_action), \
  2305. FN(rc_repeat), \
  2306. FN(rc_keydown), \
  2307. FN(skb_cgroup_id), \
  2308. FN(get_current_cgroup_id), \
  2309. FN(get_local_storage), \
  2310. FN(sk_select_reuseport), \
  2311. FN(skb_ancestor_cgroup_id), \
  2312. FN(sk_lookup_tcp), \
  2313. FN(sk_lookup_udp), \
  2314. FN(sk_release), \
  2315. FN(map_push_elem), \
  2316. FN(map_pop_elem), \
  2317. FN(map_peek_elem), \
  2318. FN(msg_push_data),
  2319. /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  2320. * function eBPF program intends to call
  2321. */
  2322. #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
  2323. enum bpf_func_id {
  2324. __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
  2325. __BPF_FUNC_MAX_ID,
  2326. };
  2327. #undef __BPF_ENUM_FN
  2328. /* All flags used by eBPF helper functions, placed here. */
  2329. /* BPF_FUNC_skb_store_bytes flags. */
  2330. #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
  2331. #define BPF_F_INVALIDATE_HASH (1ULL << 1)
  2332. /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
  2333. * First 4 bits are for passing the header field size.
  2334. */
  2335. #define BPF_F_HDR_FIELD_MASK 0xfULL
  2336. /* BPF_FUNC_l4_csum_replace flags. */
  2337. #define BPF_F_PSEUDO_HDR (1ULL << 4)
  2338. #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
  2339. #define BPF_F_MARK_ENFORCE (1ULL << 6)
  2340. /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
  2341. #define BPF_F_INGRESS (1ULL << 0)
  2342. /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
  2343. #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
  2344. /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
  2345. #define BPF_F_SKIP_FIELD_MASK 0xffULL
  2346. #define BPF_F_USER_STACK (1ULL << 8)
  2347. /* flags used by BPF_FUNC_get_stackid only. */
  2348. #define BPF_F_FAST_STACK_CMP (1ULL << 9)
  2349. #define BPF_F_REUSE_STACKID (1ULL << 10)
  2350. /* flags used by BPF_FUNC_get_stack only. */
  2351. #define BPF_F_USER_BUILD_ID (1ULL << 11)
  2352. /* BPF_FUNC_skb_set_tunnel_key flags. */
  2353. #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
  2354. #define BPF_F_DONT_FRAGMENT (1ULL << 2)
  2355. #define BPF_F_SEQ_NUMBER (1ULL << 3)
  2356. /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
  2357. * BPF_FUNC_perf_event_read_value flags.
  2358. */
  2359. #define BPF_F_INDEX_MASK 0xffffffffULL
  2360. #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
  2361. /* BPF_FUNC_perf_event_output for sk_buff input context. */
  2362. #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
  2363. /* Current network namespace */
  2364. #define BPF_F_CURRENT_NETNS (-1L)
  2365. /* Mode for BPF_FUNC_skb_adjust_room helper. */
  2366. enum bpf_adj_room_mode {
  2367. BPF_ADJ_ROOM_NET,
  2368. };
  2369. /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
  2370. enum bpf_hdr_start_off {
  2371. BPF_HDR_START_MAC,
  2372. BPF_HDR_START_NET,
  2373. };
  2374. /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
  2375. enum bpf_lwt_encap_mode {
  2376. BPF_LWT_ENCAP_SEG6,
  2377. BPF_LWT_ENCAP_SEG6_INLINE
  2378. };
  2379. #define __bpf_md_ptr(type, name) \
  2380. union { \
  2381. type name; \
  2382. __u64 :64; \
  2383. } __attribute__((aligned(8)))
  2384. /* user accessible mirror of in-kernel sk_buff.
  2385. * new fields can only be added to the end of this structure
  2386. */
  2387. struct __sk_buff {
  2388. __u32 len;
  2389. __u32 pkt_type;
  2390. __u32 mark;
  2391. __u32 queue_mapping;
  2392. __u32 protocol;
  2393. __u32 vlan_present;
  2394. __u32 vlan_tci;
  2395. __u32 vlan_proto;
  2396. __u32 priority;
  2397. __u32 ingress_ifindex;
  2398. __u32 ifindex;
  2399. __u32 tc_index;
  2400. __u32 cb[5];
  2401. __u32 hash;
  2402. __u32 tc_classid;
  2403. __u32 data;
  2404. __u32 data_end;
  2405. __u32 napi_id;
  2406. /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
  2407. __u32 family;
  2408. __u32 remote_ip4; /* Stored in network byte order */
  2409. __u32 local_ip4; /* Stored in network byte order */
  2410. __u32 remote_ip6[4]; /* Stored in network byte order */
  2411. __u32 local_ip6[4]; /* Stored in network byte order */
  2412. __u32 remote_port; /* Stored in network byte order */
  2413. __u32 local_port; /* stored in host byte order */
  2414. /* ... here. */
  2415. __u32 data_meta;
  2416. __bpf_md_ptr(struct bpf_flow_keys *, flow_keys);
  2417. };
  2418. struct bpf_tunnel_key {
  2419. __u32 tunnel_id;
  2420. union {
  2421. __u32 remote_ipv4;
  2422. __u32 remote_ipv6[4];
  2423. };
  2424. __u8 tunnel_tos;
  2425. __u8 tunnel_ttl;
  2426. __u16 tunnel_ext; /* Padding, future use. */
  2427. __u32 tunnel_label;
  2428. };
  2429. /* user accessible mirror of in-kernel xfrm_state.
  2430. * new fields can only be added to the end of this structure
  2431. */
  2432. struct bpf_xfrm_state {
  2433. __u32 reqid;
  2434. __u32 spi; /* Stored in network byte order */
  2435. __u16 family;
  2436. __u16 ext; /* Padding, future use. */
  2437. union {
  2438. __u32 remote_ipv4; /* Stored in network byte order */
  2439. __u32 remote_ipv6[4]; /* Stored in network byte order */
  2440. };
  2441. };
  2442. /* Generic BPF return codes which all BPF program types may support.
  2443. * The values are binary compatible with their TC_ACT_* counter-part to
  2444. * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
  2445. * programs.
  2446. *
  2447. * XDP is handled seprately, see XDP_*.
  2448. */
  2449. enum bpf_ret_code {
  2450. BPF_OK = 0,
  2451. /* 1 reserved */
  2452. BPF_DROP = 2,
  2453. /* 3-6 reserved */
  2454. BPF_REDIRECT = 7,
  2455. /* >127 are reserved for prog type specific return codes */
  2456. };
  2457. struct bpf_sock {
  2458. __u32 bound_dev_if;
  2459. __u32 family;
  2460. __u32 type;
  2461. __u32 protocol;
  2462. __u32 mark;
  2463. __u32 priority;
  2464. __u32 src_ip4; /* Allows 1,2,4-byte read.
  2465. * Stored in network byte order.
  2466. */
  2467. __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
  2468. * Stored in network byte order.
  2469. */
  2470. __u32 src_port; /* Allows 4-byte read.
  2471. * Stored in host byte order
  2472. */
  2473. };
  2474. struct bpf_sock_tuple {
  2475. union {
  2476. struct {
  2477. __be32 saddr;
  2478. __be32 daddr;
  2479. __be16 sport;
  2480. __be16 dport;
  2481. } ipv4;
  2482. struct {
  2483. __be32 saddr[4];
  2484. __be32 daddr[4];
  2485. __be16 sport;
  2486. __be16 dport;
  2487. } ipv6;
  2488. };
  2489. };
  2490. #define XDP_PACKET_HEADROOM 256
  2491. /* User return codes for XDP prog type.
  2492. * A valid XDP program must return one of these defined values. All other
  2493. * return codes are reserved for future use. Unknown return codes will
  2494. * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
  2495. */
  2496. enum xdp_action {
  2497. XDP_ABORTED = 0,
  2498. XDP_DROP,
  2499. XDP_PASS,
  2500. XDP_TX,
  2501. XDP_REDIRECT,
  2502. };
  2503. /* user accessible metadata for XDP packet hook
  2504. * new fields must be added to the end of this structure
  2505. */
  2506. struct xdp_md {
  2507. __u32 data;
  2508. __u32 data_end;
  2509. __u32 data_meta;
  2510. /* Below access go through struct xdp_rxq_info */
  2511. __u32 ingress_ifindex; /* rxq->dev->ifindex */
  2512. __u32 rx_queue_index; /* rxq->queue_index */
  2513. };
  2514. enum sk_action {
  2515. SK_DROP = 0,
  2516. SK_PASS,
  2517. };
  2518. /* user accessible metadata for SK_MSG packet hook, new fields must
  2519. * be added to the end of this structure
  2520. */
  2521. struct sk_msg_md {
  2522. __bpf_md_ptr(void *, data);
  2523. __bpf_md_ptr(void *, data_end);
  2524. __u32 family;
  2525. __u32 remote_ip4; /* Stored in network byte order */
  2526. __u32 local_ip4; /* Stored in network byte order */
  2527. __u32 remote_ip6[4]; /* Stored in network byte order */
  2528. __u32 local_ip6[4]; /* Stored in network byte order */
  2529. __u32 remote_port; /* Stored in network byte order */
  2530. __u32 local_port; /* stored in host byte order */
  2531. };
  2532. struct sk_reuseport_md {
  2533. /*
  2534. * Start of directly accessible data. It begins from
  2535. * the tcp/udp header.
  2536. */
  2537. __bpf_md_ptr(void *, data);
  2538. /* End of directly accessible data */
  2539. __bpf_md_ptr(void *, data_end);
  2540. /*
  2541. * Total length of packet (starting from the tcp/udp header).
  2542. * Note that the directly accessible bytes (data_end - data)
  2543. * could be less than this "len". Those bytes could be
  2544. * indirectly read by a helper "bpf_skb_load_bytes()".
  2545. */
  2546. __u32 len;
  2547. /*
  2548. * Eth protocol in the mac header (network byte order). e.g.
  2549. * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
  2550. */
  2551. __u32 eth_protocol;
  2552. __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
  2553. __u32 bind_inany; /* Is sock bound to an INANY address? */
  2554. __u32 hash; /* A hash of the packet 4 tuples */
  2555. };
  2556. #define BPF_TAG_SIZE 8
  2557. struct bpf_prog_info {
  2558. __u32 type;
  2559. __u32 id;
  2560. __u8 tag[BPF_TAG_SIZE];
  2561. __u32 jited_prog_len;
  2562. __u32 xlated_prog_len;
  2563. __aligned_u64 jited_prog_insns;
  2564. __aligned_u64 xlated_prog_insns;
  2565. __u64 load_time; /* ns since boottime */
  2566. __u32 created_by_uid;
  2567. __u32 nr_map_ids;
  2568. __aligned_u64 map_ids;
  2569. char name[BPF_OBJ_NAME_LEN];
  2570. __u32 ifindex;
  2571. __u32 gpl_compatible:1;
  2572. __u64 netns_dev;
  2573. __u64 netns_ino;
  2574. __u32 nr_jited_ksyms;
  2575. __u32 nr_jited_func_lens;
  2576. __aligned_u64 jited_ksyms;
  2577. __aligned_u64 jited_func_lens;
  2578. } __attribute__((aligned(8)));
  2579. struct bpf_map_info {
  2580. __u32 type;
  2581. __u32 id;
  2582. __u32 key_size;
  2583. __u32 value_size;
  2584. __u32 max_entries;
  2585. __u32 map_flags;
  2586. char name[BPF_OBJ_NAME_LEN];
  2587. __u32 ifindex;
  2588. __u32 :32;
  2589. __u64 netns_dev;
  2590. __u64 netns_ino;
  2591. __u32 btf_id;
  2592. __u32 btf_key_type_id;
  2593. __u32 btf_value_type_id;
  2594. } __attribute__((aligned(8)));
  2595. struct bpf_btf_info {
  2596. __aligned_u64 btf;
  2597. __u32 btf_size;
  2598. __u32 id;
  2599. } __attribute__((aligned(8)));
  2600. /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
  2601. * by user and intended to be used by socket (e.g. to bind to, depends on
  2602. * attach attach type).
  2603. */
  2604. struct bpf_sock_addr {
  2605. __u32 user_family; /* Allows 4-byte read, but no write. */
  2606. __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
  2607. * Stored in network byte order.
  2608. */
  2609. __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
  2610. * Stored in network byte order.
  2611. */
  2612. __u32 user_port; /* Allows 4-byte read and write.
  2613. * Stored in network byte order
  2614. */
  2615. __u32 family; /* Allows 4-byte read, but no write */
  2616. __u32 type; /* Allows 4-byte read, but no write */
  2617. __u32 protocol; /* Allows 4-byte read, but no write */
  2618. __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
  2619. * Stored in network byte order.
  2620. */
  2621. __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
  2622. * Stored in network byte order.
  2623. */
  2624. };
  2625. /* User bpf_sock_ops struct to access socket values and specify request ops
  2626. * and their replies.
  2627. * Some of this fields are in network (bigendian) byte order and may need
  2628. * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
  2629. * New fields can only be added at the end of this structure
  2630. */
  2631. struct bpf_sock_ops {
  2632. __u32 op;
  2633. union {
  2634. __u32 args[4]; /* Optionally passed to bpf program */
  2635. __u32 reply; /* Returned by bpf program */
  2636. __u32 replylong[4]; /* Optionally returned by bpf prog */
  2637. };
  2638. __u32 family;
  2639. __u32 remote_ip4; /* Stored in network byte order */
  2640. __u32 local_ip4; /* Stored in network byte order */
  2641. __u32 remote_ip6[4]; /* Stored in network byte order */
  2642. __u32 local_ip6[4]; /* Stored in network byte order */
  2643. __u32 remote_port; /* Stored in network byte order */
  2644. __u32 local_port; /* stored in host byte order */
  2645. __u32 is_fullsock; /* Some TCP fields are only valid if
  2646. * there is a full socket. If not, the
  2647. * fields read as zero.
  2648. */
  2649. __u32 snd_cwnd;
  2650. __u32 srtt_us; /* Averaged RTT << 3 in usecs */
  2651. __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
  2652. __u32 state;
  2653. __u32 rtt_min;
  2654. __u32 snd_ssthresh;
  2655. __u32 rcv_nxt;
  2656. __u32 snd_nxt;
  2657. __u32 snd_una;
  2658. __u32 mss_cache;
  2659. __u32 ecn_flags;
  2660. __u32 rate_delivered;
  2661. __u32 rate_interval_us;
  2662. __u32 packets_out;
  2663. __u32 retrans_out;
  2664. __u32 total_retrans;
  2665. __u32 segs_in;
  2666. __u32 data_segs_in;
  2667. __u32 segs_out;
  2668. __u32 data_segs_out;
  2669. __u32 lost_out;
  2670. __u32 sacked_out;
  2671. __u32 sk_txhash;
  2672. __u64 bytes_received;
  2673. __u64 bytes_acked;
  2674. };
  2675. /* Definitions for bpf_sock_ops_cb_flags */
  2676. #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
  2677. #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
  2678. #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
  2679. #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
  2680. * supported cb flags
  2681. */
  2682. /* List of known BPF sock_ops operators.
  2683. * New entries can only be added at the end
  2684. */
  2685. enum {
  2686. BPF_SOCK_OPS_VOID,
  2687. BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
  2688. * -1 if default value should be used
  2689. */
  2690. BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
  2691. * window (in packets) or -1 if default
  2692. * value should be used
  2693. */
  2694. BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
  2695. * active connection is initialized
  2696. */
  2697. BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
  2698. * active connection is
  2699. * established
  2700. */
  2701. BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
  2702. * passive connection is
  2703. * established
  2704. */
  2705. BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
  2706. * needs ECN
  2707. */
  2708. BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
  2709. * based on the path and may be
  2710. * dependent on the congestion control
  2711. * algorithm. In general it indicates
  2712. * a congestion threshold. RTTs above
  2713. * this indicate congestion
  2714. */
  2715. BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
  2716. * Arg1: value of icsk_retransmits
  2717. * Arg2: value of icsk_rto
  2718. * Arg3: whether RTO has expired
  2719. */
  2720. BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
  2721. * Arg1: sequence number of 1st byte
  2722. * Arg2: # segments
  2723. * Arg3: return value of
  2724. * tcp_transmit_skb (0 => success)
  2725. */
  2726. BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
  2727. * Arg1: old_state
  2728. * Arg2: new_state
  2729. */
  2730. BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
  2731. * socket transition to LISTEN state.
  2732. */
  2733. };
  2734. /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
  2735. * changes between the TCP and BPF versions. Ideally this should never happen.
  2736. * If it does, we need to add code to convert them before calling
  2737. * the BPF sock_ops function.
  2738. */
  2739. enum {
  2740. BPF_TCP_ESTABLISHED = 1,
  2741. BPF_TCP_SYN_SENT,
  2742. BPF_TCP_SYN_RECV,
  2743. BPF_TCP_FIN_WAIT1,
  2744. BPF_TCP_FIN_WAIT2,
  2745. BPF_TCP_TIME_WAIT,
  2746. BPF_TCP_CLOSE,
  2747. BPF_TCP_CLOSE_WAIT,
  2748. BPF_TCP_LAST_ACK,
  2749. BPF_TCP_LISTEN,
  2750. BPF_TCP_CLOSING, /* Now a valid state */
  2751. BPF_TCP_NEW_SYN_RECV,
  2752. BPF_TCP_MAX_STATES /* Leave at the end! */
  2753. };
  2754. #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
  2755. #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
  2756. struct bpf_perf_event_value {
  2757. __u64 counter;
  2758. __u64 enabled;
  2759. __u64 running;
  2760. };
  2761. #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
  2762. #define BPF_DEVCG_ACC_READ (1ULL << 1)
  2763. #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
  2764. #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
  2765. #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
  2766. struct bpf_cgroup_dev_ctx {
  2767. /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
  2768. __u32 access_type;
  2769. __u32 major;
  2770. __u32 minor;
  2771. };
  2772. struct bpf_raw_tracepoint_args {
  2773. __u64 args[0];
  2774. };
  2775. /* DIRECT: Skip the FIB rules and go to FIB table associated with device
  2776. * OUTPUT: Do lookup from egress perspective; default is ingress
  2777. */
  2778. #define BPF_FIB_LOOKUP_DIRECT BIT(0)
  2779. #define BPF_FIB_LOOKUP_OUTPUT BIT(1)
  2780. enum {
  2781. BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
  2782. BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
  2783. BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
  2784. BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
  2785. BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
  2786. BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
  2787. BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
  2788. BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
  2789. BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
  2790. };
  2791. struct bpf_fib_lookup {
  2792. /* input: network family for lookup (AF_INET, AF_INET6)
  2793. * output: network family of egress nexthop
  2794. */
  2795. __u8 family;
  2796. /* set if lookup is to consider L4 data - e.g., FIB rules */
  2797. __u8 l4_protocol;
  2798. __be16 sport;
  2799. __be16 dport;
  2800. /* total length of packet from network header - used for MTU check */
  2801. __u16 tot_len;
  2802. /* input: L3 device index for lookup
  2803. * output: device index from FIB lookup
  2804. */
  2805. __u32 ifindex;
  2806. union {
  2807. /* inputs to lookup */
  2808. __u8 tos; /* AF_INET */
  2809. __be32 flowinfo; /* AF_INET6, flow_label + priority */
  2810. /* output: metric of fib result (IPv4/IPv6 only) */
  2811. __u32 rt_metric;
  2812. };
  2813. union {
  2814. __be32 ipv4_src;
  2815. __u32 ipv6_src[4]; /* in6_addr; network order */
  2816. };
  2817. /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
  2818. * network header. output: bpf_fib_lookup sets to gateway address
  2819. * if FIB lookup returns gateway route
  2820. */
  2821. union {
  2822. __be32 ipv4_dst;
  2823. __u32 ipv6_dst[4]; /* in6_addr; network order */
  2824. };
  2825. /* output */
  2826. __be16 h_vlan_proto;
  2827. __be16 h_vlan_TCI;
  2828. __u8 smac[6]; /* ETH_ALEN */
  2829. __u8 dmac[6]; /* ETH_ALEN */
  2830. };
  2831. enum bpf_task_fd_type {
  2832. BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
  2833. BPF_FD_TYPE_TRACEPOINT, /* tp name */
  2834. BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
  2835. BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
  2836. BPF_FD_TYPE_UPROBE, /* filename + offset */
  2837. BPF_FD_TYPE_URETPROBE, /* filename + offset */
  2838. };
  2839. struct bpf_flow_keys {
  2840. __u16 nhoff;
  2841. __u16 thoff;
  2842. __u16 addr_proto; /* ETH_P_* of valid addrs */
  2843. __u8 is_frag;
  2844. __u8 is_first_frag;
  2845. __u8 is_encap;
  2846. __u8 ip_proto;
  2847. __be16 n_proto;
  2848. __be16 sport;
  2849. __be16 dport;
  2850. union {
  2851. struct {
  2852. __be32 ipv4_src;
  2853. __be32 ipv4_dst;
  2854. };
  2855. struct {
  2856. __u32 ipv6_src[4]; /* in6_addr; network order */
  2857. __u32 ipv6_dst[4]; /* in6_addr; network order */
  2858. };
  2859. };
  2860. };
  2861. #endif /* _UAPI__LINUX_BPF_H__ */