bpf.h 105 KB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420142114221423142414251426142714281429143014311432143314341435143614371438143914401441144214431444144514461447144814491450145114521453145414551456145714581459146014611462146314641465146614671468146914701471147214731474147514761477147814791480148114821483148414851486148714881489149014911492149314941495149614971498149915001501150215031504150515061507150815091510151115121513151415151516151715181519152015211522152315241525152615271528152915301531153215331534153515361537153815391540154115421543154415451546154715481549155015511552155315541555155615571558155915601561156215631564156515661567156815691570157115721573157415751576157715781579158015811582158315841585158615871588158915901591159215931594159515961597159815991600160116021603160416051606160716081609161016111612161316141615161616171618161916201621162216231624162516261627162816291630163116321633163416351636163716381639164016411642164316441645164616471648164916501651165216531654165516561657165816591660166116621663166416651666166716681669167016711672167316741675167616771678167916801681168216831684168516861687168816891690169116921693169416951696169716981699170017011702170317041705170617071708170917101711171217131714171517161717171817191720172117221723172417251726172717281729173017311732173317341735173617371738173917401741174217431744174517461747174817491750175117521753175417551756175717581759176017611762176317641765176617671768176917701771177217731774177517761777177817791780178117821783178417851786178717881789179017911792179317941795179617971798179918001801180218031804180518061807180818091810181118121813181418151816181718181819182018211822182318241825182618271828182918301831183218331834183518361837183818391840184118421843184418451846184718481849185018511852185318541855185618571858185918601861186218631864186518661867186818691870187118721873187418751876187718781879188018811882188318841885188618871888188918901891189218931894189518961897189818991900190119021903190419051906190719081909191019111912191319141915191619171918191919201921192219231924192519261927192819291930193119321933193419351936193719381939194019411942194319441945194619471948194919501951195219531954195519561957195819591960196119621963196419651966196719681969197019711972197319741975197619771978197919801981198219831984198519861987198819891990199119921993199419951996199719981999200020012002200320042005200620072008200920102011201220132014201520162017201820192020202120222023202420252026202720282029203020312032203320342035203620372038203920402041204220432044204520462047204820492050205120522053205420552056205720582059206020612062206320642065206620672068206920702071207220732074207520762077207820792080208120822083208420852086208720882089209020912092209320942095209620972098209921002101210221032104210521062107210821092110211121122113211421152116211721182119212021212122212321242125212621272128212921302131213221332134213521362137213821392140214121422143214421452146214721482149215021512152215321542155215621572158215921602161216221632164216521662167216821692170217121722173217421752176217721782179218021812182218321842185218621872188218921902191219221932194219521962197219821992200220122022203220422052206220722082209221022112212221322142215221622172218221922202221222222232224222522262227222822292230223122322233223422352236223722382239224022412242224322442245224622472248224922502251225222532254225522562257225822592260226122622263226422652266226722682269227022712272227322742275227622772278227922802281228222832284228522862287228822892290229122922293229422952296229722982299230023012302230323042305230623072308230923102311231223132314231523162317231823192320232123222323232423252326232723282329233023312332233323342335233623372338233923402341234223432344234523462347234823492350235123522353235423552356235723582359236023612362236323642365236623672368236923702371237223732374237523762377237823792380238123822383238423852386238723882389239023912392239323942395239623972398239924002401240224032404240524062407240824092410241124122413241424152416241724182419242024212422242324242425242624272428242924302431243224332434243524362437243824392440244124422443244424452446244724482449245024512452245324542455245624572458245924602461246224632464246524662467246824692470247124722473247424752476247724782479248024812482248324842485248624872488248924902491249224932494249524962497249824992500250125022503250425052506250725082509251025112512251325142515251625172518251925202521252225232524252525262527252825292530253125322533253425352536253725382539254025412542254325442545254625472548254925502551255225532554255525562557255825592560256125622563256425652566256725682569257025712572257325742575257625772578257925802581258225832584258525862587258825892590259125922593259425952596259725982599260026012602260326042605260626072608260926102611261226132614261526162617261826192620262126222623262426252626262726282629263026312632263326342635263626372638263926402641264226432644264526462647264826492650265126522653265426552656265726582659266026612662266326642665266626672668266926702671267226732674267526762677267826792680268126822683268426852686268726882689269026912692269326942695269626972698269927002701270227032704270527062707270827092710271127122713271427152716271727182719272027212722272327242725272627272728272927302731273227332734273527362737273827392740274127422743274427452746274727482749275027512752275327542755275627572758275927602761276227632764276527662767276827692770277127722773277427752776277727782779278027812782278327842785
  1. /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
  2. /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com
  3. *
  4. * This program is free software; you can redistribute it and/or
  5. * modify it under the terms of version 2 of the GNU General Public
  6. * License as published by the Free Software Foundation.
  7. */
  8. #ifndef _UAPI__LINUX_BPF_H__
  9. #define _UAPI__LINUX_BPF_H__
  10. #include <linux/types.h>
  11. #include <linux/bpf_common.h>
  12. /* Extended instruction set based on top of classic BPF */
  13. /* instruction classes */
  14. #define BPF_ALU64 0x07 /* alu mode in double word width */
  15. /* ld/ldx fields */
  16. #define BPF_DW 0x18 /* double word (64-bit) */
  17. #define BPF_XADD 0xc0 /* exclusive add */
  18. /* alu/jmp fields */
  19. #define BPF_MOV 0xb0 /* mov reg to reg */
  20. #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */
  21. /* change endianness of a register */
  22. #define BPF_END 0xd0 /* flags for endianness conversion: */
  23. #define BPF_TO_LE 0x00 /* convert to little-endian */
  24. #define BPF_TO_BE 0x08 /* convert to big-endian */
  25. #define BPF_FROM_LE BPF_TO_LE
  26. #define BPF_FROM_BE BPF_TO_BE
  27. /* jmp encodings */
  28. #define BPF_JNE 0x50 /* jump != */
  29. #define BPF_JLT 0xa0 /* LT is unsigned, '<' */
  30. #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */
  31. #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */
  32. #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */
  33. #define BPF_JSLT 0xc0 /* SLT is signed, '<' */
  34. #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */
  35. #define BPF_CALL 0x80 /* function call */
  36. #define BPF_EXIT 0x90 /* function return */
  37. /* Register numbers */
  38. enum {
  39. BPF_REG_0 = 0,
  40. BPF_REG_1,
  41. BPF_REG_2,
  42. BPF_REG_3,
  43. BPF_REG_4,
  44. BPF_REG_5,
  45. BPF_REG_6,
  46. BPF_REG_7,
  47. BPF_REG_8,
  48. BPF_REG_9,
  49. BPF_REG_10,
  50. __MAX_BPF_REG,
  51. };
  52. /* BPF has 10 general purpose 64-bit registers and stack frame. */
  53. #define MAX_BPF_REG __MAX_BPF_REG
  54. struct bpf_insn {
  55. __u8 code; /* opcode */
  56. __u8 dst_reg:4; /* dest register */
  57. __u8 src_reg:4; /* source register */
  58. __s16 off; /* signed offset */
  59. __s32 imm; /* signed immediate constant */
  60. };
  61. /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */
  62. struct bpf_lpm_trie_key {
  63. __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */
  64. __u8 data[0]; /* Arbitrary size */
  65. };
  66. struct bpf_cgroup_storage_key {
  67. __u64 cgroup_inode_id; /* cgroup inode id */
  68. __u32 attach_type; /* program attach type */
  69. };
  70. /* BPF syscall commands, see bpf(2) man-page for details. */
  71. enum bpf_cmd {
  72. BPF_MAP_CREATE,
  73. BPF_MAP_LOOKUP_ELEM,
  74. BPF_MAP_UPDATE_ELEM,
  75. BPF_MAP_DELETE_ELEM,
  76. BPF_MAP_GET_NEXT_KEY,
  77. BPF_PROG_LOAD,
  78. BPF_OBJ_PIN,
  79. BPF_OBJ_GET,
  80. BPF_PROG_ATTACH,
  81. BPF_PROG_DETACH,
  82. BPF_PROG_TEST_RUN,
  83. BPF_PROG_GET_NEXT_ID,
  84. BPF_MAP_GET_NEXT_ID,
  85. BPF_PROG_GET_FD_BY_ID,
  86. BPF_MAP_GET_FD_BY_ID,
  87. BPF_OBJ_GET_INFO_BY_FD,
  88. BPF_PROG_QUERY,
  89. BPF_RAW_TRACEPOINT_OPEN,
  90. BPF_BTF_LOAD,
  91. BPF_BTF_GET_FD_BY_ID,
  92. BPF_TASK_FD_QUERY,
  93. };
  94. enum bpf_map_type {
  95. BPF_MAP_TYPE_UNSPEC,
  96. BPF_MAP_TYPE_HASH,
  97. BPF_MAP_TYPE_ARRAY,
  98. BPF_MAP_TYPE_PROG_ARRAY,
  99. BPF_MAP_TYPE_PERF_EVENT_ARRAY,
  100. BPF_MAP_TYPE_PERCPU_HASH,
  101. BPF_MAP_TYPE_PERCPU_ARRAY,
  102. BPF_MAP_TYPE_STACK_TRACE,
  103. BPF_MAP_TYPE_CGROUP_ARRAY,
  104. BPF_MAP_TYPE_LRU_HASH,
  105. BPF_MAP_TYPE_LRU_PERCPU_HASH,
  106. BPF_MAP_TYPE_LPM_TRIE,
  107. BPF_MAP_TYPE_ARRAY_OF_MAPS,
  108. BPF_MAP_TYPE_HASH_OF_MAPS,
  109. BPF_MAP_TYPE_DEVMAP,
  110. BPF_MAP_TYPE_SOCKMAP,
  111. BPF_MAP_TYPE_CPUMAP,
  112. BPF_MAP_TYPE_XSKMAP,
  113. BPF_MAP_TYPE_SOCKHASH,
  114. BPF_MAP_TYPE_CGROUP_STORAGE,
  115. BPF_MAP_TYPE_REUSEPORT_SOCKARRAY,
  116. };
  117. enum bpf_prog_type {
  118. BPF_PROG_TYPE_UNSPEC,
  119. BPF_PROG_TYPE_SOCKET_FILTER,
  120. BPF_PROG_TYPE_KPROBE,
  121. BPF_PROG_TYPE_SCHED_CLS,
  122. BPF_PROG_TYPE_SCHED_ACT,
  123. BPF_PROG_TYPE_TRACEPOINT,
  124. BPF_PROG_TYPE_XDP,
  125. BPF_PROG_TYPE_PERF_EVENT,
  126. BPF_PROG_TYPE_CGROUP_SKB,
  127. BPF_PROG_TYPE_CGROUP_SOCK,
  128. BPF_PROG_TYPE_LWT_IN,
  129. BPF_PROG_TYPE_LWT_OUT,
  130. BPF_PROG_TYPE_LWT_XMIT,
  131. BPF_PROG_TYPE_SOCK_OPS,
  132. BPF_PROG_TYPE_SK_SKB,
  133. BPF_PROG_TYPE_CGROUP_DEVICE,
  134. BPF_PROG_TYPE_SK_MSG,
  135. BPF_PROG_TYPE_RAW_TRACEPOINT,
  136. BPF_PROG_TYPE_CGROUP_SOCK_ADDR,
  137. BPF_PROG_TYPE_LWT_SEG6LOCAL,
  138. BPF_PROG_TYPE_LIRC_MODE2,
  139. BPF_PROG_TYPE_SK_REUSEPORT,
  140. };
  141. enum bpf_attach_type {
  142. BPF_CGROUP_INET_INGRESS,
  143. BPF_CGROUP_INET_EGRESS,
  144. BPF_CGROUP_INET_SOCK_CREATE,
  145. BPF_CGROUP_SOCK_OPS,
  146. BPF_SK_SKB_STREAM_PARSER,
  147. BPF_SK_SKB_STREAM_VERDICT,
  148. BPF_CGROUP_DEVICE,
  149. BPF_SK_MSG_VERDICT,
  150. BPF_CGROUP_INET4_BIND,
  151. BPF_CGROUP_INET6_BIND,
  152. BPF_CGROUP_INET4_CONNECT,
  153. BPF_CGROUP_INET6_CONNECT,
  154. BPF_CGROUP_INET4_POST_BIND,
  155. BPF_CGROUP_INET6_POST_BIND,
  156. BPF_CGROUP_UDP4_SENDMSG,
  157. BPF_CGROUP_UDP6_SENDMSG,
  158. BPF_LIRC_MODE2,
  159. BPF_CGROUP_UDP4_RECVMSG = 19,
  160. BPF_CGROUP_UDP6_RECVMSG,
  161. __MAX_BPF_ATTACH_TYPE
  162. };
  163. #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE
  164. /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command
  165. *
  166. * NONE(default): No further bpf programs allowed in the subtree.
  167. *
  168. * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program,
  169. * the program in this cgroup yields to sub-cgroup program.
  170. *
  171. * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program,
  172. * that cgroup program gets run in addition to the program in this cgroup.
  173. *
  174. * Only one program is allowed to be attached to a cgroup with
  175. * NONE or BPF_F_ALLOW_OVERRIDE flag.
  176. * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will
  177. * release old program and attach the new one. Attach flags has to match.
  178. *
  179. * Multiple programs are allowed to be attached to a cgroup with
  180. * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order
  181. * (those that were attached first, run first)
  182. * The programs of sub-cgroup are executed first, then programs of
  183. * this cgroup and then programs of parent cgroup.
  184. * When children program makes decision (like picking TCP CA or sock bind)
  185. * parent program has a chance to override it.
  186. *
  187. * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups.
  188. * A cgroup with NONE doesn't allow any programs in sub-cgroups.
  189. * Ex1:
  190. * cgrp1 (MULTI progs A, B) ->
  191. * cgrp2 (OVERRIDE prog C) ->
  192. * cgrp3 (MULTI prog D) ->
  193. * cgrp4 (OVERRIDE prog E) ->
  194. * cgrp5 (NONE prog F)
  195. * the event in cgrp5 triggers execution of F,D,A,B in that order.
  196. * if prog F is detached, the execution is E,D,A,B
  197. * if prog F and D are detached, the execution is E,A,B
  198. * if prog F, E and D are detached, the execution is C,A,B
  199. *
  200. * All eligible programs are executed regardless of return code from
  201. * earlier programs.
  202. */
  203. #define BPF_F_ALLOW_OVERRIDE (1U << 0)
  204. #define BPF_F_ALLOW_MULTI (1U << 1)
  205. /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the
  206. * verifier will perform strict alignment checking as if the kernel
  207. * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set,
  208. * and NET_IP_ALIGN defined to 2.
  209. */
  210. #define BPF_F_STRICT_ALIGNMENT (1U << 0)
  211. /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */
  212. #define BPF_PSEUDO_MAP_FD 1
  213. /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative
  214. * offset to another bpf function
  215. */
  216. #define BPF_PSEUDO_CALL 1
  217. /* flags for BPF_MAP_UPDATE_ELEM command */
  218. #define BPF_ANY 0 /* create new element or update existing */
  219. #define BPF_NOEXIST 1 /* create new element if it didn't exist */
  220. #define BPF_EXIST 2 /* update existing element */
  221. /* flags for BPF_MAP_CREATE command */
  222. #define BPF_F_NO_PREALLOC (1U << 0)
  223. /* Instead of having one common LRU list in the
  224. * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list
  225. * which can scale and perform better.
  226. * Note, the LRU nodes (including free nodes) cannot be moved
  227. * across different LRU lists.
  228. */
  229. #define BPF_F_NO_COMMON_LRU (1U << 1)
  230. /* Specify numa node during map creation */
  231. #define BPF_F_NUMA_NODE (1U << 2)
  232. /* flags for BPF_PROG_QUERY */
  233. #define BPF_F_QUERY_EFFECTIVE (1U << 0)
  234. #define BPF_OBJ_NAME_LEN 16U
  235. /* Flags for accessing BPF object */
  236. #define BPF_F_RDONLY (1U << 3)
  237. #define BPF_F_WRONLY (1U << 4)
  238. /* Flag for stack_map, store build_id+offset instead of pointer */
  239. #define BPF_F_STACK_BUILD_ID (1U << 5)
  240. enum bpf_stack_build_id_status {
  241. /* user space need an empty entry to identify end of a trace */
  242. BPF_STACK_BUILD_ID_EMPTY = 0,
  243. /* with valid build_id and offset */
  244. BPF_STACK_BUILD_ID_VALID = 1,
  245. /* couldn't get build_id, fallback to ip */
  246. BPF_STACK_BUILD_ID_IP = 2,
  247. };
  248. #define BPF_BUILD_ID_SIZE 20
  249. struct bpf_stack_build_id {
  250. __s32 status;
  251. unsigned char build_id[BPF_BUILD_ID_SIZE];
  252. union {
  253. __u64 offset;
  254. __u64 ip;
  255. };
  256. };
  257. union bpf_attr {
  258. struct { /* anonymous struct used by BPF_MAP_CREATE command */
  259. __u32 map_type; /* one of enum bpf_map_type */
  260. __u32 key_size; /* size of key in bytes */
  261. __u32 value_size; /* size of value in bytes */
  262. __u32 max_entries; /* max number of entries in a map */
  263. __u32 map_flags; /* BPF_MAP_CREATE related
  264. * flags defined above.
  265. */
  266. __u32 inner_map_fd; /* fd pointing to the inner map */
  267. __u32 numa_node; /* numa node (effective only if
  268. * BPF_F_NUMA_NODE is set).
  269. */
  270. char map_name[BPF_OBJ_NAME_LEN];
  271. __u32 map_ifindex; /* ifindex of netdev to create on */
  272. __u32 btf_fd; /* fd pointing to a BTF type data */
  273. __u32 btf_key_type_id; /* BTF type_id of the key */
  274. __u32 btf_value_type_id; /* BTF type_id of the value */
  275. };
  276. struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
  277. __u32 map_fd;
  278. __aligned_u64 key;
  279. union {
  280. __aligned_u64 value;
  281. __aligned_u64 next_key;
  282. };
  283. __u64 flags;
  284. };
  285. struct { /* anonymous struct used by BPF_PROG_LOAD command */
  286. __u32 prog_type; /* one of enum bpf_prog_type */
  287. __u32 insn_cnt;
  288. __aligned_u64 insns;
  289. __aligned_u64 license;
  290. __u32 log_level; /* verbosity level of verifier */
  291. __u32 log_size; /* size of user buffer */
  292. __aligned_u64 log_buf; /* user supplied buffer */
  293. __u32 kern_version; /* checked when prog_type=kprobe */
  294. __u32 prog_flags;
  295. char prog_name[BPF_OBJ_NAME_LEN];
  296. __u32 prog_ifindex; /* ifindex of netdev to prep for */
  297. /* For some prog types expected attach type must be known at
  298. * load time to verify attach type specific parts of prog
  299. * (context accesses, allowed helpers, etc).
  300. */
  301. __u32 expected_attach_type;
  302. };
  303. struct { /* anonymous struct used by BPF_OBJ_* commands */
  304. __aligned_u64 pathname;
  305. __u32 bpf_fd;
  306. __u32 file_flags;
  307. };
  308. struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */
  309. __u32 target_fd; /* container object to attach to */
  310. __u32 attach_bpf_fd; /* eBPF program to attach */
  311. __u32 attach_type;
  312. __u32 attach_flags;
  313. };
  314. struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */
  315. __u32 prog_fd;
  316. __u32 retval;
  317. __u32 data_size_in;
  318. __u32 data_size_out;
  319. __aligned_u64 data_in;
  320. __aligned_u64 data_out;
  321. __u32 repeat;
  322. __u32 duration;
  323. } test;
  324. struct { /* anonymous struct used by BPF_*_GET_*_ID */
  325. union {
  326. __u32 start_id;
  327. __u32 prog_id;
  328. __u32 map_id;
  329. __u32 btf_id;
  330. };
  331. __u32 next_id;
  332. __u32 open_flags;
  333. };
  334. struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */
  335. __u32 bpf_fd;
  336. __u32 info_len;
  337. __aligned_u64 info;
  338. } info;
  339. struct { /* anonymous struct used by BPF_PROG_QUERY command */
  340. __u32 target_fd; /* container object to query */
  341. __u32 attach_type;
  342. __u32 query_flags;
  343. __u32 attach_flags;
  344. __aligned_u64 prog_ids;
  345. __u32 prog_cnt;
  346. } query;
  347. struct {
  348. __u64 name;
  349. __u32 prog_fd;
  350. } raw_tracepoint;
  351. struct { /* anonymous struct for BPF_BTF_LOAD */
  352. __aligned_u64 btf;
  353. __aligned_u64 btf_log_buf;
  354. __u32 btf_size;
  355. __u32 btf_log_size;
  356. __u32 btf_log_level;
  357. };
  358. struct {
  359. __u32 pid; /* input: pid */
  360. __u32 fd; /* input: fd */
  361. __u32 flags; /* input: flags */
  362. __u32 buf_len; /* input/output: buf len */
  363. __aligned_u64 buf; /* input/output:
  364. * tp_name for tracepoint
  365. * symbol for kprobe
  366. * filename for uprobe
  367. */
  368. __u32 prog_id; /* output: prod_id */
  369. __u32 fd_type; /* output: BPF_FD_TYPE_* */
  370. __u64 probe_offset; /* output: probe_offset */
  371. __u64 probe_addr; /* output: probe_addr */
  372. } task_fd_query;
  373. } __attribute__((aligned(8)));
  374. /* The description below is an attempt at providing documentation to eBPF
  375. * developers about the multiple available eBPF helper functions. It can be
  376. * parsed and used to produce a manual page. The workflow is the following,
  377. * and requires the rst2man utility:
  378. *
  379. * $ ./scripts/bpf_helpers_doc.py \
  380. * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst
  381. * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7
  382. * $ man /tmp/bpf-helpers.7
  383. *
  384. * Note that in order to produce this external documentation, some RST
  385. * formatting is used in the descriptions to get "bold" and "italics" in
  386. * manual pages. Also note that the few trailing white spaces are
  387. * intentional, removing them would break paragraphs for rst2man.
  388. *
  389. * Start of BPF helper function descriptions:
  390. *
  391. * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key)
  392. * Description
  393. * Perform a lookup in *map* for an entry associated to *key*.
  394. * Return
  395. * Map value associated to *key*, or **NULL** if no entry was
  396. * found.
  397. *
  398. * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags)
  399. * Description
  400. * Add or update the value of the entry associated to *key* in
  401. * *map* with *value*. *flags* is one of:
  402. *
  403. * **BPF_NOEXIST**
  404. * The entry for *key* must not exist in the map.
  405. * **BPF_EXIST**
  406. * The entry for *key* must already exist in the map.
  407. * **BPF_ANY**
  408. * No condition on the existence of the entry for *key*.
  409. *
  410. * Flag value **BPF_NOEXIST** cannot be used for maps of types
  411. * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all
  412. * elements always exist), the helper would return an error.
  413. * Return
  414. * 0 on success, or a negative error in case of failure.
  415. *
  416. * int bpf_map_delete_elem(struct bpf_map *map, const void *key)
  417. * Description
  418. * Delete entry with *key* from *map*.
  419. * Return
  420. * 0 on success, or a negative error in case of failure.
  421. *
  422. * int bpf_probe_read(void *dst, u32 size, const void *src)
  423. * Description
  424. * For tracing programs, safely attempt to read *size* bytes from
  425. * address *src* and store the data in *dst*.
  426. * Return
  427. * 0 on success, or a negative error in case of failure.
  428. *
  429. * u64 bpf_ktime_get_ns(void)
  430. * Description
  431. * Return the time elapsed since system boot, in nanoseconds.
  432. * Return
  433. * Current *ktime*.
  434. *
  435. * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...)
  436. * Description
  437. * This helper is a "printk()-like" facility for debugging. It
  438. * prints a message defined by format *fmt* (of size *fmt_size*)
  439. * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if
  440. * available. It can take up to three additional **u64**
  441. * arguments (as an eBPF helpers, the total number of arguments is
  442. * limited to five).
  443. *
  444. * Each time the helper is called, it appends a line to the trace.
  445. * The format of the trace is customizable, and the exact output
  446. * one will get depends on the options set in
  447. * *\/sys/kernel/debug/tracing/trace_options* (see also the
  448. * *README* file under the same directory). However, it usually
  449. * defaults to something like:
  450. *
  451. * ::
  452. *
  453. * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg>
  454. *
  455. * In the above:
  456. *
  457. * * ``telnet`` is the name of the current task.
  458. * * ``470`` is the PID of the current task.
  459. * * ``001`` is the CPU number on which the task is
  460. * running.
  461. * * In ``.N..``, each character refers to a set of
  462. * options (whether irqs are enabled, scheduling
  463. * options, whether hard/softirqs are running, level of
  464. * preempt_disabled respectively). **N** means that
  465. * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED**
  466. * are set.
  467. * * ``419421.045894`` is a timestamp.
  468. * * ``0x00000001`` is a fake value used by BPF for the
  469. * instruction pointer register.
  470. * * ``<formatted msg>`` is the message formatted with
  471. * *fmt*.
  472. *
  473. * The conversion specifiers supported by *fmt* are similar, but
  474. * more limited than for printk(). They are **%d**, **%i**,
  475. * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**,
  476. * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size
  477. * of field, padding with zeroes, etc.) is available, and the
  478. * helper will return **-EINVAL** (but print nothing) if it
  479. * encounters an unknown specifier.
  480. *
  481. * Also, note that **bpf_trace_printk**\ () is slow, and should
  482. * only be used for debugging purposes. For this reason, a notice
  483. * bloc (spanning several lines) is printed to kernel logs and
  484. * states that the helper should not be used "for production use"
  485. * the first time this helper is used (or more precisely, when
  486. * **trace_printk**\ () buffers are allocated). For passing values
  487. * to user space, perf events should be preferred.
  488. * Return
  489. * The number of bytes written to the buffer, or a negative error
  490. * in case of failure.
  491. *
  492. * u32 bpf_get_prandom_u32(void)
  493. * Description
  494. * Get a pseudo-random number.
  495. *
  496. * From a security point of view, this helper uses its own
  497. * pseudo-random internal state, and cannot be used to infer the
  498. * seed of other random functions in the kernel. However, it is
  499. * essential to note that the generator used by the helper is not
  500. * cryptographically secure.
  501. * Return
  502. * A random 32-bit unsigned value.
  503. *
  504. * u32 bpf_get_smp_processor_id(void)
  505. * Description
  506. * Get the SMP (symmetric multiprocessing) processor id. Note that
  507. * all programs run with preemption disabled, which means that the
  508. * SMP processor id is stable during all the execution of the
  509. * program.
  510. * Return
  511. * The SMP id of the processor running the program.
  512. *
  513. * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags)
  514. * Description
  515. * Store *len* bytes from address *from* into the packet
  516. * associated to *skb*, at *offset*. *flags* are a combination of
  517. * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the
  518. * checksum for the packet after storing the bytes) and
  519. * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\
  520. * **->swhash** and *skb*\ **->l4hash** to 0).
  521. *
  522. * A call to this helper is susceptible to change the underlaying
  523. * packet buffer. Therefore, at load time, all checks on pointers
  524. * previously done by the verifier are invalidated and must be
  525. * performed again, if the helper is used in combination with
  526. * direct packet access.
  527. * Return
  528. * 0 on success, or a negative error in case of failure.
  529. *
  530. * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size)
  531. * Description
  532. * Recompute the layer 3 (e.g. IP) checksum for the packet
  533. * associated to *skb*. Computation is incremental, so the helper
  534. * must know the former value of the header field that was
  535. * modified (*from*), the new value of this field (*to*), and the
  536. * number of bytes (2 or 4) for this field, stored in *size*.
  537. * Alternatively, it is possible to store the difference between
  538. * the previous and the new values of the header field in *to*, by
  539. * setting *from* and *size* to 0. For both methods, *offset*
  540. * indicates the location of the IP checksum within the packet.
  541. *
  542. * This helper works in combination with **bpf_csum_diff**\ (),
  543. * which does not update the checksum in-place, but offers more
  544. * flexibility and can handle sizes larger than 2 or 4 for the
  545. * checksum to update.
  546. *
  547. * A call to this helper is susceptible to change the underlaying
  548. * packet buffer. Therefore, at load time, all checks on pointers
  549. * previously done by the verifier are invalidated and must be
  550. * performed again, if the helper is used in combination with
  551. * direct packet access.
  552. * Return
  553. * 0 on success, or a negative error in case of failure.
  554. *
  555. * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags)
  556. * Description
  557. * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the
  558. * packet associated to *skb*. Computation is incremental, so the
  559. * helper must know the former value of the header field that was
  560. * modified (*from*), the new value of this field (*to*), and the
  561. * number of bytes (2 or 4) for this field, stored on the lowest
  562. * four bits of *flags*. Alternatively, it is possible to store
  563. * the difference between the previous and the new values of the
  564. * header field in *to*, by setting *from* and the four lowest
  565. * bits of *flags* to 0. For both methods, *offset* indicates the
  566. * location of the IP checksum within the packet. In addition to
  567. * the size of the field, *flags* can be added (bitwise OR) actual
  568. * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left
  569. * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and
  570. * for updates resulting in a null checksum the value is set to
  571. * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates
  572. * the checksum is to be computed against a pseudo-header.
  573. *
  574. * This helper works in combination with **bpf_csum_diff**\ (),
  575. * which does not update the checksum in-place, but offers more
  576. * flexibility and can handle sizes larger than 2 or 4 for the
  577. * checksum to update.
  578. *
  579. * A call to this helper is susceptible to change the underlaying
  580. * packet buffer. Therefore, at load time, all checks on pointers
  581. * previously done by the verifier are invalidated and must be
  582. * performed again, if the helper is used in combination with
  583. * direct packet access.
  584. * Return
  585. * 0 on success, or a negative error in case of failure.
  586. *
  587. * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index)
  588. * Description
  589. * This special helper is used to trigger a "tail call", or in
  590. * other words, to jump into another eBPF program. The same stack
  591. * frame is used (but values on stack and in registers for the
  592. * caller are not accessible to the callee). This mechanism allows
  593. * for program chaining, either for raising the maximum number of
  594. * available eBPF instructions, or to execute given programs in
  595. * conditional blocks. For security reasons, there is an upper
  596. * limit to the number of successive tail calls that can be
  597. * performed.
  598. *
  599. * Upon call of this helper, the program attempts to jump into a
  600. * program referenced at index *index* in *prog_array_map*, a
  601. * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes
  602. * *ctx*, a pointer to the context.
  603. *
  604. * If the call succeeds, the kernel immediately runs the first
  605. * instruction of the new program. This is not a function call,
  606. * and it never returns to the previous program. If the call
  607. * fails, then the helper has no effect, and the caller continues
  608. * to run its subsequent instructions. A call can fail if the
  609. * destination program for the jump does not exist (i.e. *index*
  610. * is superior to the number of entries in *prog_array_map*), or
  611. * if the maximum number of tail calls has been reached for this
  612. * chain of programs. This limit is defined in the kernel by the
  613. * macro **MAX_TAIL_CALL_CNT** (not accessible to user space),
  614. * which is currently set to 32.
  615. * Return
  616. * 0 on success, or a negative error in case of failure.
  617. *
  618. * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags)
  619. * Description
  620. * Clone and redirect the packet associated to *skb* to another
  621. * net device of index *ifindex*. Both ingress and egress
  622. * interfaces can be used for redirection. The **BPF_F_INGRESS**
  623. * value in *flags* is used to make the distinction (ingress path
  624. * is selected if the flag is present, egress path otherwise).
  625. * This is the only flag supported for now.
  626. *
  627. * In comparison with **bpf_redirect**\ () helper,
  628. * **bpf_clone_redirect**\ () has the associated cost of
  629. * duplicating the packet buffer, but this can be executed out of
  630. * the eBPF program. Conversely, **bpf_redirect**\ () is more
  631. * efficient, but it is handled through an action code where the
  632. * redirection happens only after the eBPF program has returned.
  633. *
  634. * A call to this helper is susceptible to change the underlaying
  635. * packet buffer. Therefore, at load time, all checks on pointers
  636. * previously done by the verifier are invalidated and must be
  637. * performed again, if the helper is used in combination with
  638. * direct packet access.
  639. * Return
  640. * 0 on success, or a negative error in case of failure.
  641. *
  642. * u64 bpf_get_current_pid_tgid(void)
  643. * Return
  644. * A 64-bit integer containing the current tgid and pid, and
  645. * created as such:
  646. * *current_task*\ **->tgid << 32 \|**
  647. * *current_task*\ **->pid**.
  648. *
  649. * u64 bpf_get_current_uid_gid(void)
  650. * Return
  651. * A 64-bit integer containing the current GID and UID, and
  652. * created as such: *current_gid* **<< 32 \|** *current_uid*.
  653. *
  654. * int bpf_get_current_comm(char *buf, u32 size_of_buf)
  655. * Description
  656. * Copy the **comm** attribute of the current task into *buf* of
  657. * *size_of_buf*. The **comm** attribute contains the name of
  658. * the executable (excluding the path) for the current task. The
  659. * *size_of_buf* must be strictly positive. On success, the
  660. * helper makes sure that the *buf* is NUL-terminated. On failure,
  661. * it is filled with zeroes.
  662. * Return
  663. * 0 on success, or a negative error in case of failure.
  664. *
  665. * u32 bpf_get_cgroup_classid(struct sk_buff *skb)
  666. * Description
  667. * Retrieve the classid for the current task, i.e. for the net_cls
  668. * cgroup to which *skb* belongs.
  669. *
  670. * This helper can be used on TC egress path, but not on ingress.
  671. *
  672. * The net_cls cgroup provides an interface to tag network packets
  673. * based on a user-provided identifier for all traffic coming from
  674. * the tasks belonging to the related cgroup. See also the related
  675. * kernel documentation, available from the Linux sources in file
  676. * *Documentation/cgroup-v1/net_cls.txt*.
  677. *
  678. * The Linux kernel has two versions for cgroups: there are
  679. * cgroups v1 and cgroups v2. Both are available to users, who can
  680. * use a mixture of them, but note that the net_cls cgroup is for
  681. * cgroup v1 only. This makes it incompatible with BPF programs
  682. * run on cgroups, which is a cgroup-v2-only feature (a socket can
  683. * only hold data for one version of cgroups at a time).
  684. *
  685. * This helper is only available is the kernel was compiled with
  686. * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to
  687. * "**y**" or to "**m**".
  688. * Return
  689. * The classid, or 0 for the default unconfigured classid.
  690. *
  691. * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci)
  692. * Description
  693. * Push a *vlan_tci* (VLAN tag control information) of protocol
  694. * *vlan_proto* to the packet associated to *skb*, then update
  695. * the checksum. Note that if *vlan_proto* is different from
  696. * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to
  697. * be **ETH_P_8021Q**.
  698. *
  699. * A call to this helper is susceptible to change the underlaying
  700. * packet buffer. Therefore, at load time, all checks on pointers
  701. * previously done by the verifier are invalidated and must be
  702. * performed again, if the helper is used in combination with
  703. * direct packet access.
  704. * Return
  705. * 0 on success, or a negative error in case of failure.
  706. *
  707. * int bpf_skb_vlan_pop(struct sk_buff *skb)
  708. * Description
  709. * Pop a VLAN header from the packet associated to *skb*.
  710. *
  711. * A call to this helper is susceptible to change the underlaying
  712. * packet buffer. Therefore, at load time, all checks on pointers
  713. * previously done by the verifier are invalidated and must be
  714. * performed again, if the helper is used in combination with
  715. * direct packet access.
  716. * Return
  717. * 0 on success, or a negative error in case of failure.
  718. *
  719. * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  720. * Description
  721. * Get tunnel metadata. This helper takes a pointer *key* to an
  722. * empty **struct bpf_tunnel_key** of **size**, that will be
  723. * filled with tunnel metadata for the packet associated to *skb*.
  724. * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which
  725. * indicates that the tunnel is based on IPv6 protocol instead of
  726. * IPv4.
  727. *
  728. * The **struct bpf_tunnel_key** is an object that generalizes the
  729. * principal parameters used by various tunneling protocols into a
  730. * single struct. This way, it can be used to easily make a
  731. * decision based on the contents of the encapsulation header,
  732. * "summarized" in this struct. In particular, it holds the IP
  733. * address of the remote end (IPv4 or IPv6, depending on the case)
  734. * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also,
  735. * this struct exposes the *key*\ **->tunnel_id**, which is
  736. * generally mapped to a VNI (Virtual Network Identifier), making
  737. * it programmable together with the **bpf_skb_set_tunnel_key**\
  738. * () helper.
  739. *
  740. * Let's imagine that the following code is part of a program
  741. * attached to the TC ingress interface, on one end of a GRE
  742. * tunnel, and is supposed to filter out all messages coming from
  743. * remote ends with IPv4 address other than 10.0.0.1:
  744. *
  745. * ::
  746. *
  747. * int ret;
  748. * struct bpf_tunnel_key key = {};
  749. *
  750. * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0);
  751. * if (ret < 0)
  752. * return TC_ACT_SHOT; // drop packet
  753. *
  754. * if (key.remote_ipv4 != 0x0a000001)
  755. * return TC_ACT_SHOT; // drop packet
  756. *
  757. * return TC_ACT_OK; // accept packet
  758. *
  759. * This interface can also be used with all encapsulation devices
  760. * that can operate in "collect metadata" mode: instead of having
  761. * one network device per specific configuration, the "collect
  762. * metadata" mode only requires a single device where the
  763. * configuration can be extracted from this helper.
  764. *
  765. * This can be used together with various tunnels such as VXLan,
  766. * Geneve, GRE or IP in IP (IPIP).
  767. * Return
  768. * 0 on success, or a negative error in case of failure.
  769. *
  770. * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags)
  771. * Description
  772. * Populate tunnel metadata for packet associated to *skb.* The
  773. * tunnel metadata is set to the contents of *key*, of *size*. The
  774. * *flags* can be set to a combination of the following values:
  775. *
  776. * **BPF_F_TUNINFO_IPV6**
  777. * Indicate that the tunnel is based on IPv6 protocol
  778. * instead of IPv4.
  779. * **BPF_F_ZERO_CSUM_TX**
  780. * For IPv4 packets, add a flag to tunnel metadata
  781. * indicating that checksum computation should be skipped
  782. * and checksum set to zeroes.
  783. * **BPF_F_DONT_FRAGMENT**
  784. * Add a flag to tunnel metadata indicating that the
  785. * packet should not be fragmented.
  786. * **BPF_F_SEQ_NUMBER**
  787. * Add a flag to tunnel metadata indicating that a
  788. * sequence number should be added to tunnel header before
  789. * sending the packet. This flag was added for GRE
  790. * encapsulation, but might be used with other protocols
  791. * as well in the future.
  792. *
  793. * Here is a typical usage on the transmit path:
  794. *
  795. * ::
  796. *
  797. * struct bpf_tunnel_key key;
  798. * populate key ...
  799. * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0);
  800. * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0);
  801. *
  802. * See also the description of the **bpf_skb_get_tunnel_key**\ ()
  803. * helper for additional information.
  804. * Return
  805. * 0 on success, or a negative error in case of failure.
  806. *
  807. * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags)
  808. * Description
  809. * Read the value of a perf event counter. This helper relies on a
  810. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of
  811. * the perf event counter is selected when *map* is updated with
  812. * perf event file descriptors. The *map* is an array whose size
  813. * is the number of available CPUs, and each cell contains a value
  814. * relative to one CPU. The value to retrieve is indicated by
  815. * *flags*, that contains the index of the CPU to look up, masked
  816. * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  817. * **BPF_F_CURRENT_CPU** to indicate that the value for the
  818. * current CPU should be retrieved.
  819. *
  820. * Note that before Linux 4.13, only hardware perf event can be
  821. * retrieved.
  822. *
  823. * Also, be aware that the newer helper
  824. * **bpf_perf_event_read_value**\ () is recommended over
  825. * **bpf_perf_event_read**\ () in general. The latter has some ABI
  826. * quirks where error and counter value are used as a return code
  827. * (which is wrong to do since ranges may overlap). This issue is
  828. * fixed with **bpf_perf_event_read_value**\ (), which at the same
  829. * time provides more features over the **bpf_perf_event_read**\
  830. * () interface. Please refer to the description of
  831. * **bpf_perf_event_read_value**\ () for details.
  832. * Return
  833. * The value of the perf event counter read from the map, or a
  834. * negative error code in case of failure.
  835. *
  836. * int bpf_redirect(u32 ifindex, u64 flags)
  837. * Description
  838. * Redirect the packet to another net device of index *ifindex*.
  839. * This helper is somewhat similar to **bpf_clone_redirect**\
  840. * (), except that the packet is not cloned, which provides
  841. * increased performance.
  842. *
  843. * Except for XDP, both ingress and egress interfaces can be used
  844. * for redirection. The **BPF_F_INGRESS** value in *flags* is used
  845. * to make the distinction (ingress path is selected if the flag
  846. * is present, egress path otherwise). Currently, XDP only
  847. * supports redirection to the egress interface, and accepts no
  848. * flag at all.
  849. *
  850. * The same effect can be attained with the more generic
  851. * **bpf_redirect_map**\ (), which requires specific maps to be
  852. * used but offers better performance.
  853. * Return
  854. * For XDP, the helper returns **XDP_REDIRECT** on success or
  855. * **XDP_ABORTED** on error. For other program types, the values
  856. * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on
  857. * error.
  858. *
  859. * u32 bpf_get_route_realm(struct sk_buff *skb)
  860. * Description
  861. * Retrieve the realm or the route, that is to say the
  862. * **tclassid** field of the destination for the *skb*. The
  863. * indentifier retrieved is a user-provided tag, similar to the
  864. * one used with the net_cls cgroup (see description for
  865. * **bpf_get_cgroup_classid**\ () helper), but here this tag is
  866. * held by a route (a destination entry), not by a task.
  867. *
  868. * Retrieving this identifier works with the clsact TC egress hook
  869. * (see also **tc-bpf(8)**), or alternatively on conventional
  870. * classful egress qdiscs, but not on TC ingress path. In case of
  871. * clsact TC egress hook, this has the advantage that, internally,
  872. * the destination entry has not been dropped yet in the transmit
  873. * path. Therefore, the destination entry does not need to be
  874. * artificially held via **netif_keep_dst**\ () for a classful
  875. * qdisc until the *skb* is freed.
  876. *
  877. * This helper is available only if the kernel was compiled with
  878. * **CONFIG_IP_ROUTE_CLASSID** configuration option.
  879. * Return
  880. * The realm of the route for the packet associated to *skb*, or 0
  881. * if none was found.
  882. *
  883. * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size)
  884. * Description
  885. * Write raw *data* blob into a special BPF perf event held by
  886. * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf
  887. * event must have the following attributes: **PERF_SAMPLE_RAW**
  888. * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and
  889. * **PERF_COUNT_SW_BPF_OUTPUT** as **config**.
  890. *
  891. * The *flags* are used to indicate the index in *map* for which
  892. * the value must be put, masked with **BPF_F_INDEX_MASK**.
  893. * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU**
  894. * to indicate that the index of the current CPU core should be
  895. * used.
  896. *
  897. * The value to write, of *size*, is passed through eBPF stack and
  898. * pointed by *data*.
  899. *
  900. * The context of the program *ctx* needs also be passed to the
  901. * helper.
  902. *
  903. * On user space, a program willing to read the values needs to
  904. * call **perf_event_open**\ () on the perf event (either for
  905. * one or for all CPUs) and to store the file descriptor into the
  906. * *map*. This must be done before the eBPF program can send data
  907. * into it. An example is available in file
  908. * *samples/bpf/trace_output_user.c* in the Linux kernel source
  909. * tree (the eBPF program counterpart is in
  910. * *samples/bpf/trace_output_kern.c*).
  911. *
  912. * **bpf_perf_event_output**\ () achieves better performance
  913. * than **bpf_trace_printk**\ () for sharing data with user
  914. * space, and is much better suitable for streaming data from eBPF
  915. * programs.
  916. *
  917. * Note that this helper is not restricted to tracing use cases
  918. * and can be used with programs attached to TC or XDP as well,
  919. * where it allows for passing data to user space listeners. Data
  920. * can be:
  921. *
  922. * * Only custom structs,
  923. * * Only the packet payload, or
  924. * * A combination of both.
  925. * Return
  926. * 0 on success, or a negative error in case of failure.
  927. *
  928. * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len)
  929. * Description
  930. * This helper was provided as an easy way to load data from a
  931. * packet. It can be used to load *len* bytes from *offset* from
  932. * the packet associated to *skb*, into the buffer pointed by
  933. * *to*.
  934. *
  935. * Since Linux 4.7, usage of this helper has mostly been replaced
  936. * by "direct packet access", enabling packet data to be
  937. * manipulated with *skb*\ **->data** and *skb*\ **->data_end**
  938. * pointing respectively to the first byte of packet data and to
  939. * the byte after the last byte of packet data. However, it
  940. * remains useful if one wishes to read large quantities of data
  941. * at once from a packet into the eBPF stack.
  942. * Return
  943. * 0 on success, or a negative error in case of failure.
  944. *
  945. * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags)
  946. * Description
  947. * Walk a user or a kernel stack and return its id. To achieve
  948. * this, the helper needs *ctx*, which is a pointer to the context
  949. * on which the tracing program is executed, and a pointer to a
  950. * *map* of type **BPF_MAP_TYPE_STACK_TRACE**.
  951. *
  952. * The last argument, *flags*, holds the number of stack frames to
  953. * skip (from 0 to 255), masked with
  954. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  955. * a combination of the following flags:
  956. *
  957. * **BPF_F_USER_STACK**
  958. * Collect a user space stack instead of a kernel stack.
  959. * **BPF_F_FAST_STACK_CMP**
  960. * Compare stacks by hash only.
  961. * **BPF_F_REUSE_STACKID**
  962. * If two different stacks hash into the same *stackid*,
  963. * discard the old one.
  964. *
  965. * The stack id retrieved is a 32 bit long integer handle which
  966. * can be further combined with other data (including other stack
  967. * ids) and used as a key into maps. This can be useful for
  968. * generating a variety of graphs (such as flame graphs or off-cpu
  969. * graphs).
  970. *
  971. * For walking a stack, this helper is an improvement over
  972. * **bpf_probe_read**\ (), which can be used with unrolled loops
  973. * but is not efficient and consumes a lot of eBPF instructions.
  974. * Instead, **bpf_get_stackid**\ () can collect up to
  975. * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that
  976. * this limit can be controlled with the **sysctl** program, and
  977. * that it should be manually increased in order to profile long
  978. * user stacks (such as stacks for Java programs). To do so, use:
  979. *
  980. * ::
  981. *
  982. * # sysctl kernel.perf_event_max_stack=<new value>
  983. * Return
  984. * The positive or null stack id on success, or a negative error
  985. * in case of failure.
  986. *
  987. * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed)
  988. * Description
  989. * Compute a checksum difference, from the raw buffer pointed by
  990. * *from*, of length *from_size* (that must be a multiple of 4),
  991. * towards the raw buffer pointed by *to*, of size *to_size*
  992. * (same remark). An optional *seed* can be added to the value
  993. * (this can be cascaded, the seed may come from a previous call
  994. * to the helper).
  995. *
  996. * This is flexible enough to be used in several ways:
  997. *
  998. * * With *from_size* == 0, *to_size* > 0 and *seed* set to
  999. * checksum, it can be used when pushing new data.
  1000. * * With *from_size* > 0, *to_size* == 0 and *seed* set to
  1001. * checksum, it can be used when removing data from a packet.
  1002. * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it
  1003. * can be used to compute a diff. Note that *from_size* and
  1004. * *to_size* do not need to be equal.
  1005. *
  1006. * This helper can be used in combination with
  1007. * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to
  1008. * which one can feed in the difference computed with
  1009. * **bpf_csum_diff**\ ().
  1010. * Return
  1011. * The checksum result, or a negative error code in case of
  1012. * failure.
  1013. *
  1014. * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
  1015. * Description
  1016. * Retrieve tunnel options metadata for the packet associated to
  1017. * *skb*, and store the raw tunnel option data to the buffer *opt*
  1018. * of *size*.
  1019. *
  1020. * This helper can be used with encapsulation devices that can
  1021. * operate in "collect metadata" mode (please refer to the related
  1022. * note in the description of **bpf_skb_get_tunnel_key**\ () for
  1023. * more details). A particular example where this can be used is
  1024. * in combination with the Geneve encapsulation protocol, where it
  1025. * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper)
  1026. * and retrieving arbitrary TLVs (Type-Length-Value headers) from
  1027. * the eBPF program. This allows for full customization of these
  1028. * headers.
  1029. * Return
  1030. * The size of the option data retrieved.
  1031. *
  1032. * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size)
  1033. * Description
  1034. * Set tunnel options metadata for the packet associated to *skb*
  1035. * to the option data contained in the raw buffer *opt* of *size*.
  1036. *
  1037. * See also the description of the **bpf_skb_get_tunnel_opt**\ ()
  1038. * helper for additional information.
  1039. * Return
  1040. * 0 on success, or a negative error in case of failure.
  1041. *
  1042. * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags)
  1043. * Description
  1044. * Change the protocol of the *skb* to *proto*. Currently
  1045. * supported are transition from IPv4 to IPv6, and from IPv6 to
  1046. * IPv4. The helper takes care of the groundwork for the
  1047. * transition, including resizing the socket buffer. The eBPF
  1048. * program is expected to fill the new headers, if any, via
  1049. * **skb_store_bytes**\ () and to recompute the checksums with
  1050. * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\
  1051. * (). The main case for this helper is to perform NAT64
  1052. * operations out of an eBPF program.
  1053. *
  1054. * Internally, the GSO type is marked as dodgy so that headers are
  1055. * checked and segments are recalculated by the GSO/GRO engine.
  1056. * The size for GSO target is adapted as well.
  1057. *
  1058. * All values for *flags* are reserved for future usage, and must
  1059. * be left at zero.
  1060. *
  1061. * A call to this helper is susceptible to change the underlaying
  1062. * packet buffer. Therefore, at load time, all checks on pointers
  1063. * previously done by the verifier are invalidated and must be
  1064. * performed again, if the helper is used in combination with
  1065. * direct packet access.
  1066. * Return
  1067. * 0 on success, or a negative error in case of failure.
  1068. *
  1069. * int bpf_skb_change_type(struct sk_buff *skb, u32 type)
  1070. * Description
  1071. * Change the packet type for the packet associated to *skb*. This
  1072. * comes down to setting *skb*\ **->pkt_type** to *type*, except
  1073. * the eBPF program does not have a write access to *skb*\
  1074. * **->pkt_type** beside this helper. Using a helper here allows
  1075. * for graceful handling of errors.
  1076. *
  1077. * The major use case is to change incoming *skb*s to
  1078. * **PACKET_HOST** in a programmatic way instead of having to
  1079. * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for
  1080. * example.
  1081. *
  1082. * Note that *type* only allows certain values. At this time, they
  1083. * are:
  1084. *
  1085. * **PACKET_HOST**
  1086. * Packet is for us.
  1087. * **PACKET_BROADCAST**
  1088. * Send packet to all.
  1089. * **PACKET_MULTICAST**
  1090. * Send packet to group.
  1091. * **PACKET_OTHERHOST**
  1092. * Send packet to someone else.
  1093. * Return
  1094. * 0 on success, or a negative error in case of failure.
  1095. *
  1096. * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index)
  1097. * Description
  1098. * Check whether *skb* is a descendant of the cgroup2 held by
  1099. * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  1100. * Return
  1101. * The return value depends on the result of the test, and can be:
  1102. *
  1103. * * 0, if the *skb* failed the cgroup2 descendant test.
  1104. * * 1, if the *skb* succeeded the cgroup2 descendant test.
  1105. * * A negative error code, if an error occurred.
  1106. *
  1107. * u32 bpf_get_hash_recalc(struct sk_buff *skb)
  1108. * Description
  1109. * Retrieve the hash of the packet, *skb*\ **->hash**. If it is
  1110. * not set, in particular if the hash was cleared due to mangling,
  1111. * recompute this hash. Later accesses to the hash can be done
  1112. * directly with *skb*\ **->hash**.
  1113. *
  1114. * Calling **bpf_set_hash_invalid**\ (), changing a packet
  1115. * prototype with **bpf_skb_change_proto**\ (), or calling
  1116. * **bpf_skb_store_bytes**\ () with the
  1117. * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear
  1118. * the hash and to trigger a new computation for the next call to
  1119. * **bpf_get_hash_recalc**\ ().
  1120. * Return
  1121. * The 32-bit hash.
  1122. *
  1123. * u64 bpf_get_current_task(void)
  1124. * Return
  1125. * A pointer to the current task struct.
  1126. *
  1127. * int bpf_probe_write_user(void *dst, const void *src, u32 len)
  1128. * Description
  1129. * Attempt in a safe way to write *len* bytes from the buffer
  1130. * *src* to *dst* in memory. It only works for threads that are in
  1131. * user context, and *dst* must be a valid user space address.
  1132. *
  1133. * This helper should not be used to implement any kind of
  1134. * security mechanism because of TOC-TOU attacks, but rather to
  1135. * debug, divert, and manipulate execution of semi-cooperative
  1136. * processes.
  1137. *
  1138. * Keep in mind that this feature is meant for experiments, and it
  1139. * has a risk of crashing the system and running programs.
  1140. * Therefore, when an eBPF program using this helper is attached,
  1141. * a warning including PID and process name is printed to kernel
  1142. * logs.
  1143. * Return
  1144. * 0 on success, or a negative error in case of failure.
  1145. *
  1146. * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index)
  1147. * Description
  1148. * Check whether the probe is being run is the context of a given
  1149. * subset of the cgroup2 hierarchy. The cgroup2 to test is held by
  1150. * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*.
  1151. * Return
  1152. * The return value depends on the result of the test, and can be:
  1153. *
  1154. * * 0, if the *skb* task belongs to the cgroup2.
  1155. * * 1, if the *skb* task does not belong to the cgroup2.
  1156. * * A negative error code, if an error occurred.
  1157. *
  1158. * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags)
  1159. * Description
  1160. * Resize (trim or grow) the packet associated to *skb* to the
  1161. * new *len*. The *flags* are reserved for future usage, and must
  1162. * be left at zero.
  1163. *
  1164. * The basic idea is that the helper performs the needed work to
  1165. * change the size of the packet, then the eBPF program rewrites
  1166. * the rest via helpers like **bpf_skb_store_bytes**\ (),
  1167. * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ ()
  1168. * and others. This helper is a slow path utility intended for
  1169. * replies with control messages. And because it is targeted for
  1170. * slow path, the helper itself can afford to be slow: it
  1171. * implicitly linearizes, unclones and drops offloads from the
  1172. * *skb*.
  1173. *
  1174. * A call to this helper is susceptible to change the underlaying
  1175. * packet buffer. Therefore, at load time, all checks on pointers
  1176. * previously done by the verifier are invalidated and must be
  1177. * performed again, if the helper is used in combination with
  1178. * direct packet access.
  1179. * Return
  1180. * 0 on success, or a negative error in case of failure.
  1181. *
  1182. * int bpf_skb_pull_data(struct sk_buff *skb, u32 len)
  1183. * Description
  1184. * Pull in non-linear data in case the *skb* is non-linear and not
  1185. * all of *len* are part of the linear section. Make *len* bytes
  1186. * from *skb* readable and writable. If a zero value is passed for
  1187. * *len*, then the whole length of the *skb* is pulled.
  1188. *
  1189. * This helper is only needed for reading and writing with direct
  1190. * packet access.
  1191. *
  1192. * For direct packet access, testing that offsets to access
  1193. * are within packet boundaries (test on *skb*\ **->data_end**) is
  1194. * susceptible to fail if offsets are invalid, or if the requested
  1195. * data is in non-linear parts of the *skb*. On failure the
  1196. * program can just bail out, or in the case of a non-linear
  1197. * buffer, use a helper to make the data available. The
  1198. * **bpf_skb_load_bytes**\ () helper is a first solution to access
  1199. * the data. Another one consists in using **bpf_skb_pull_data**
  1200. * to pull in once the non-linear parts, then retesting and
  1201. * eventually access the data.
  1202. *
  1203. * At the same time, this also makes sure the *skb* is uncloned,
  1204. * which is a necessary condition for direct write. As this needs
  1205. * to be an invariant for the write part only, the verifier
  1206. * detects writes and adds a prologue that is calling
  1207. * **bpf_skb_pull_data()** to effectively unclone the *skb* from
  1208. * the very beginning in case it is indeed cloned.
  1209. *
  1210. * A call to this helper is susceptible to change the underlaying
  1211. * packet buffer. Therefore, at load time, all checks on pointers
  1212. * previously done by the verifier are invalidated and must be
  1213. * performed again, if the helper is used in combination with
  1214. * direct packet access.
  1215. * Return
  1216. * 0 on success, or a negative error in case of failure.
  1217. *
  1218. * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum)
  1219. * Description
  1220. * Add the checksum *csum* into *skb*\ **->csum** in case the
  1221. * driver has supplied a checksum for the entire packet into that
  1222. * field. Return an error otherwise. This helper is intended to be
  1223. * used in combination with **bpf_csum_diff**\ (), in particular
  1224. * when the checksum needs to be updated after data has been
  1225. * written into the packet through direct packet access.
  1226. * Return
  1227. * The checksum on success, or a negative error code in case of
  1228. * failure.
  1229. *
  1230. * void bpf_set_hash_invalid(struct sk_buff *skb)
  1231. * Description
  1232. * Invalidate the current *skb*\ **->hash**. It can be used after
  1233. * mangling on headers through direct packet access, in order to
  1234. * indicate that the hash is outdated and to trigger a
  1235. * recalculation the next time the kernel tries to access this
  1236. * hash or when the **bpf_get_hash_recalc**\ () helper is called.
  1237. *
  1238. * int bpf_get_numa_node_id(void)
  1239. * Description
  1240. * Return the id of the current NUMA node. The primary use case
  1241. * for this helper is the selection of sockets for the local NUMA
  1242. * node, when the program is attached to sockets using the
  1243. * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**),
  1244. * but the helper is also available to other eBPF program types,
  1245. * similarly to **bpf_get_smp_processor_id**\ ().
  1246. * Return
  1247. * The id of current NUMA node.
  1248. *
  1249. * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags)
  1250. * Description
  1251. * Grows headroom of packet associated to *skb* and adjusts the
  1252. * offset of the MAC header accordingly, adding *len* bytes of
  1253. * space. It automatically extends and reallocates memory as
  1254. * required.
  1255. *
  1256. * This helper can be used on a layer 3 *skb* to push a MAC header
  1257. * for redirection into a layer 2 device.
  1258. *
  1259. * All values for *flags* are reserved for future usage, and must
  1260. * be left at zero.
  1261. *
  1262. * A call to this helper is susceptible to change the underlaying
  1263. * packet buffer. Therefore, at load time, all checks on pointers
  1264. * previously done by the verifier are invalidated and must be
  1265. * performed again, if the helper is used in combination with
  1266. * direct packet access.
  1267. * Return
  1268. * 0 on success, or a negative error in case of failure.
  1269. *
  1270. * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta)
  1271. * Description
  1272. * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that
  1273. * it is possible to use a negative value for *delta*. This helper
  1274. * can be used to prepare the packet for pushing or popping
  1275. * headers.
  1276. *
  1277. * A call to this helper is susceptible to change the underlaying
  1278. * packet buffer. Therefore, at load time, all checks on pointers
  1279. * previously done by the verifier are invalidated and must be
  1280. * performed again, if the helper is used in combination with
  1281. * direct packet access.
  1282. * Return
  1283. * 0 on success, or a negative error in case of failure.
  1284. *
  1285. * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr)
  1286. * Description
  1287. * Copy a NUL terminated string from an unsafe address
  1288. * *unsafe_ptr* to *dst*. The *size* should include the
  1289. * terminating NUL byte. In case the string length is smaller than
  1290. * *size*, the target is not padded with further NUL bytes. If the
  1291. * string length is larger than *size*, just *size*-1 bytes are
  1292. * copied and the last byte is set to NUL.
  1293. *
  1294. * On success, the length of the copied string is returned. This
  1295. * makes this helper useful in tracing programs for reading
  1296. * strings, and more importantly to get its length at runtime. See
  1297. * the following snippet:
  1298. *
  1299. * ::
  1300. *
  1301. * SEC("kprobe/sys_open")
  1302. * void bpf_sys_open(struct pt_regs *ctx)
  1303. * {
  1304. * char buf[PATHLEN]; // PATHLEN is defined to 256
  1305. * int res = bpf_probe_read_str(buf, sizeof(buf),
  1306. * ctx->di);
  1307. *
  1308. * // Consume buf, for example push it to
  1309. * // userspace via bpf_perf_event_output(); we
  1310. * // can use res (the string length) as event
  1311. * // size, after checking its boundaries.
  1312. * }
  1313. *
  1314. * In comparison, using **bpf_probe_read()** helper here instead
  1315. * to read the string would require to estimate the length at
  1316. * compile time, and would often result in copying more memory
  1317. * than necessary.
  1318. *
  1319. * Another useful use case is when parsing individual process
  1320. * arguments or individual environment variables navigating
  1321. * *current*\ **->mm->arg_start** and *current*\
  1322. * **->mm->env_start**: using this helper and the return value,
  1323. * one can quickly iterate at the right offset of the memory area.
  1324. * Return
  1325. * On success, the strictly positive length of the string,
  1326. * including the trailing NUL character. On error, a negative
  1327. * value.
  1328. *
  1329. * u64 bpf_get_socket_cookie(struct sk_buff *skb)
  1330. * Description
  1331. * If the **struct sk_buff** pointed by *skb* has a known socket,
  1332. * retrieve the cookie (generated by the kernel) of this socket.
  1333. * If no cookie has been set yet, generate a new cookie. Once
  1334. * generated, the socket cookie remains stable for the life of the
  1335. * socket. This helper can be useful for monitoring per socket
  1336. * networking traffic statistics as it provides a unique socket
  1337. * identifier per namespace.
  1338. * Return
  1339. * A 8-byte long non-decreasing number on success, or 0 if the
  1340. * socket field is missing inside *skb*.
  1341. *
  1342. * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx)
  1343. * Description
  1344. * Equivalent to bpf_get_socket_cookie() helper that accepts
  1345. * *skb*, but gets socket from **struct bpf_sock_addr** contex.
  1346. * Return
  1347. * A 8-byte long non-decreasing number.
  1348. *
  1349. * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx)
  1350. * Description
  1351. * Equivalent to bpf_get_socket_cookie() helper that accepts
  1352. * *skb*, but gets socket from **struct bpf_sock_ops** contex.
  1353. * Return
  1354. * A 8-byte long non-decreasing number.
  1355. *
  1356. * u32 bpf_get_socket_uid(struct sk_buff *skb)
  1357. * Return
  1358. * The owner UID of the socket associated to *skb*. If the socket
  1359. * is **NULL**, or if it is not a full socket (i.e. if it is a
  1360. * time-wait or a request socket instead), **overflowuid** value
  1361. * is returned (note that **overflowuid** might also be the actual
  1362. * UID value for the socket).
  1363. *
  1364. * u32 bpf_set_hash(struct sk_buff *skb, u32 hash)
  1365. * Description
  1366. * Set the full hash for *skb* (set the field *skb*\ **->hash**)
  1367. * to value *hash*.
  1368. * Return
  1369. * 0
  1370. *
  1371. * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
  1372. * Description
  1373. * Emulate a call to **setsockopt()** on the socket associated to
  1374. * *bpf_socket*, which must be a full socket. The *level* at
  1375. * which the option resides and the name *optname* of the option
  1376. * must be specified, see **setsockopt(2)** for more information.
  1377. * The option value of length *optlen* is pointed by *optval*.
  1378. *
  1379. * This helper actually implements a subset of **setsockopt()**.
  1380. * It supports the following *level*\ s:
  1381. *
  1382. * * **SOL_SOCKET**, which supports the following *optname*\ s:
  1383. * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**,
  1384. * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**.
  1385. * * **IPPROTO_TCP**, which supports the following *optname*\ s:
  1386. * **TCP_CONGESTION**, **TCP_BPF_IW**,
  1387. * **TCP_BPF_SNDCWND_CLAMP**.
  1388. * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
  1389. * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
  1390. * Return
  1391. * 0 on success, or a negative error in case of failure.
  1392. *
  1393. * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags)
  1394. * Description
  1395. * Grow or shrink the room for data in the packet associated to
  1396. * *skb* by *len_diff*, and according to the selected *mode*.
  1397. *
  1398. * There is a single supported mode at this time:
  1399. *
  1400. * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer
  1401. * (room space is added or removed below the layer 3 header).
  1402. *
  1403. * All values for *flags* are reserved for future usage, and must
  1404. * be left at zero.
  1405. *
  1406. * A call to this helper is susceptible to change the underlaying
  1407. * packet buffer. Therefore, at load time, all checks on pointers
  1408. * previously done by the verifier are invalidated and must be
  1409. * performed again, if the helper is used in combination with
  1410. * direct packet access.
  1411. * Return
  1412. * 0 on success, or a negative error in case of failure.
  1413. *
  1414. * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags)
  1415. * Description
  1416. * Redirect the packet to the endpoint referenced by *map* at
  1417. * index *key*. Depending on its type, this *map* can contain
  1418. * references to net devices (for forwarding packets through other
  1419. * ports), or to CPUs (for redirecting XDP frames to another CPU;
  1420. * but this is only implemented for native XDP (with driver
  1421. * support) as of this writing).
  1422. *
  1423. * All values for *flags* are reserved for future usage, and must
  1424. * be left at zero.
  1425. *
  1426. * When used to redirect packets to net devices, this helper
  1427. * provides a high performance increase over **bpf_redirect**\ ().
  1428. * This is due to various implementation details of the underlying
  1429. * mechanisms, one of which is the fact that **bpf_redirect_map**\
  1430. * () tries to send packet as a "bulk" to the device.
  1431. * Return
  1432. * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error.
  1433. *
  1434. * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags)
  1435. * Description
  1436. * Redirect the packet to the socket referenced by *map* (of type
  1437. * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  1438. * egress interfaces can be used for redirection. The
  1439. * **BPF_F_INGRESS** value in *flags* is used to make the
  1440. * distinction (ingress path is selected if the flag is present,
  1441. * egress path otherwise). This is the only flag supported for now.
  1442. * Return
  1443. * **SK_PASS** on success, or **SK_DROP** on error.
  1444. *
  1445. * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags)
  1446. * Description
  1447. * Add an entry to, or update a *map* referencing sockets. The
  1448. * *skops* is used as a new value for the entry associated to
  1449. * *key*. *flags* is one of:
  1450. *
  1451. * **BPF_NOEXIST**
  1452. * The entry for *key* must not exist in the map.
  1453. * **BPF_EXIST**
  1454. * The entry for *key* must already exist in the map.
  1455. * **BPF_ANY**
  1456. * No condition on the existence of the entry for *key*.
  1457. *
  1458. * If the *map* has eBPF programs (parser and verdict), those will
  1459. * be inherited by the socket being added. If the socket is
  1460. * already attached to eBPF programs, this results in an error.
  1461. * Return
  1462. * 0 on success, or a negative error in case of failure.
  1463. *
  1464. * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta)
  1465. * Description
  1466. * Adjust the address pointed by *xdp_md*\ **->data_meta** by
  1467. * *delta* (which can be positive or negative). Note that this
  1468. * operation modifies the address stored in *xdp_md*\ **->data**,
  1469. * so the latter must be loaded only after the helper has been
  1470. * called.
  1471. *
  1472. * The use of *xdp_md*\ **->data_meta** is optional and programs
  1473. * are not required to use it. The rationale is that when the
  1474. * packet is processed with XDP (e.g. as DoS filter), it is
  1475. * possible to push further meta data along with it before passing
  1476. * to the stack, and to give the guarantee that an ingress eBPF
  1477. * program attached as a TC classifier on the same device can pick
  1478. * this up for further post-processing. Since TC works with socket
  1479. * buffers, it remains possible to set from XDP the **mark** or
  1480. * **priority** pointers, or other pointers for the socket buffer.
  1481. * Having this scratch space generic and programmable allows for
  1482. * more flexibility as the user is free to store whatever meta
  1483. * data they need.
  1484. *
  1485. * A call to this helper is susceptible to change the underlaying
  1486. * packet buffer. Therefore, at load time, all checks on pointers
  1487. * previously done by the verifier are invalidated and must be
  1488. * performed again, if the helper is used in combination with
  1489. * direct packet access.
  1490. * Return
  1491. * 0 on success, or a negative error in case of failure.
  1492. *
  1493. * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size)
  1494. * Description
  1495. * Read the value of a perf event counter, and store it into *buf*
  1496. * of size *buf_size*. This helper relies on a *map* of type
  1497. * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event
  1498. * counter is selected when *map* is updated with perf event file
  1499. * descriptors. The *map* is an array whose size is the number of
  1500. * available CPUs, and each cell contains a value relative to one
  1501. * CPU. The value to retrieve is indicated by *flags*, that
  1502. * contains the index of the CPU to look up, masked with
  1503. * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to
  1504. * **BPF_F_CURRENT_CPU** to indicate that the value for the
  1505. * current CPU should be retrieved.
  1506. *
  1507. * This helper behaves in a way close to
  1508. * **bpf_perf_event_read**\ () helper, save that instead of
  1509. * just returning the value observed, it fills the *buf*
  1510. * structure. This allows for additional data to be retrieved: in
  1511. * particular, the enabled and running times (in *buf*\
  1512. * **->enabled** and *buf*\ **->running**, respectively) are
  1513. * copied. In general, **bpf_perf_event_read_value**\ () is
  1514. * recommended over **bpf_perf_event_read**\ (), which has some
  1515. * ABI issues and provides fewer functionalities.
  1516. *
  1517. * These values are interesting, because hardware PMU (Performance
  1518. * Monitoring Unit) counters are limited resources. When there are
  1519. * more PMU based perf events opened than available counters,
  1520. * kernel will multiplex these events so each event gets certain
  1521. * percentage (but not all) of the PMU time. In case that
  1522. * multiplexing happens, the number of samples or counter value
  1523. * will not reflect the case compared to when no multiplexing
  1524. * occurs. This makes comparison between different runs difficult.
  1525. * Typically, the counter value should be normalized before
  1526. * comparing to other experiments. The usual normalization is done
  1527. * as follows.
  1528. *
  1529. * ::
  1530. *
  1531. * normalized_counter = counter * t_enabled / t_running
  1532. *
  1533. * Where t_enabled is the time enabled for event and t_running is
  1534. * the time running for event since last normalization. The
  1535. * enabled and running times are accumulated since the perf event
  1536. * open. To achieve scaling factor between two invocations of an
  1537. * eBPF program, users can can use CPU id as the key (which is
  1538. * typical for perf array usage model) to remember the previous
  1539. * value and do the calculation inside the eBPF program.
  1540. * Return
  1541. * 0 on success, or a negative error in case of failure.
  1542. *
  1543. * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size)
  1544. * Description
  1545. * For en eBPF program attached to a perf event, retrieve the
  1546. * value of the event counter associated to *ctx* and store it in
  1547. * the structure pointed by *buf* and of size *buf_size*. Enabled
  1548. * and running times are also stored in the structure (see
  1549. * description of helper **bpf_perf_event_read_value**\ () for
  1550. * more details).
  1551. * Return
  1552. * 0 on success, or a negative error in case of failure.
  1553. *
  1554. * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen)
  1555. * Description
  1556. * Emulate a call to **getsockopt()** on the socket associated to
  1557. * *bpf_socket*, which must be a full socket. The *level* at
  1558. * which the option resides and the name *optname* of the option
  1559. * must be specified, see **getsockopt(2)** for more information.
  1560. * The retrieved value is stored in the structure pointed by
  1561. * *opval* and of length *optlen*.
  1562. *
  1563. * This helper actually implements a subset of **getsockopt()**.
  1564. * It supports the following *level*\ s:
  1565. *
  1566. * * **IPPROTO_TCP**, which supports *optname*
  1567. * **TCP_CONGESTION**.
  1568. * * **IPPROTO_IP**, which supports *optname* **IP_TOS**.
  1569. * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**.
  1570. * Return
  1571. * 0 on success, or a negative error in case of failure.
  1572. *
  1573. * int bpf_override_return(struct pt_reg *regs, u64 rc)
  1574. * Description
  1575. * Used for error injection, this helper uses kprobes to override
  1576. * the return value of the probed function, and to set it to *rc*.
  1577. * The first argument is the context *regs* on which the kprobe
  1578. * works.
  1579. *
  1580. * This helper works by setting setting the PC (program counter)
  1581. * to an override function which is run in place of the original
  1582. * probed function. This means the probed function is not run at
  1583. * all. The replacement function just returns with the required
  1584. * value.
  1585. *
  1586. * This helper has security implications, and thus is subject to
  1587. * restrictions. It is only available if the kernel was compiled
  1588. * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration
  1589. * option, and in this case it only works on functions tagged with
  1590. * **ALLOW_ERROR_INJECTION** in the kernel code.
  1591. *
  1592. * Also, the helper is only available for the architectures having
  1593. * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing,
  1594. * x86 architecture is the only one to support this feature.
  1595. * Return
  1596. * 0
  1597. *
  1598. * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval)
  1599. * Description
  1600. * Attempt to set the value of the **bpf_sock_ops_cb_flags** field
  1601. * for the full TCP socket associated to *bpf_sock_ops* to
  1602. * *argval*.
  1603. *
  1604. * The primary use of this field is to determine if there should
  1605. * be calls to eBPF programs of type
  1606. * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP
  1607. * code. A program of the same type can change its value, per
  1608. * connection and as necessary, when the connection is
  1609. * established. This field is directly accessible for reading, but
  1610. * this helper must be used for updates in order to return an
  1611. * error if an eBPF program tries to set a callback that is not
  1612. * supported in the current kernel.
  1613. *
  1614. * The supported callback values that *argval* can combine are:
  1615. *
  1616. * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out)
  1617. * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission)
  1618. * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change)
  1619. *
  1620. * Here are some examples of where one could call such eBPF
  1621. * program:
  1622. *
  1623. * * When RTO fires.
  1624. * * When a packet is retransmitted.
  1625. * * When the connection terminates.
  1626. * * When a packet is sent.
  1627. * * When a packet is received.
  1628. * Return
  1629. * Code **-EINVAL** if the socket is not a full TCP socket;
  1630. * otherwise, a positive number containing the bits that could not
  1631. * be set is returned (which comes down to 0 if all bits were set
  1632. * as required).
  1633. *
  1634. * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags)
  1635. * Description
  1636. * This helper is used in programs implementing policies at the
  1637. * socket level. If the message *msg* is allowed to pass (i.e. if
  1638. * the verdict eBPF program returns **SK_PASS**), redirect it to
  1639. * the socket referenced by *map* (of type
  1640. * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and
  1641. * egress interfaces can be used for redirection. The
  1642. * **BPF_F_INGRESS** value in *flags* is used to make the
  1643. * distinction (ingress path is selected if the flag is present,
  1644. * egress path otherwise). This is the only flag supported for now.
  1645. * Return
  1646. * **SK_PASS** on success, or **SK_DROP** on error.
  1647. *
  1648. * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes)
  1649. * Description
  1650. * For socket policies, apply the verdict of the eBPF program to
  1651. * the next *bytes* (number of bytes) of message *msg*.
  1652. *
  1653. * For example, this helper can be used in the following cases:
  1654. *
  1655. * * A single **sendmsg**\ () or **sendfile**\ () system call
  1656. * contains multiple logical messages that the eBPF program is
  1657. * supposed to read and for which it should apply a verdict.
  1658. * * An eBPF program only cares to read the first *bytes* of a
  1659. * *msg*. If the message has a large payload, then setting up
  1660. * and calling the eBPF program repeatedly for all bytes, even
  1661. * though the verdict is already known, would create unnecessary
  1662. * overhead.
  1663. *
  1664. * When called from within an eBPF program, the helper sets a
  1665. * counter internal to the BPF infrastructure, that is used to
  1666. * apply the last verdict to the next *bytes*. If *bytes* is
  1667. * smaller than the current data being processed from a
  1668. * **sendmsg**\ () or **sendfile**\ () system call, the first
  1669. * *bytes* will be sent and the eBPF program will be re-run with
  1670. * the pointer for start of data pointing to byte number *bytes*
  1671. * **+ 1**. If *bytes* is larger than the current data being
  1672. * processed, then the eBPF verdict will be applied to multiple
  1673. * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are
  1674. * consumed.
  1675. *
  1676. * Note that if a socket closes with the internal counter holding
  1677. * a non-zero value, this is not a problem because data is not
  1678. * being buffered for *bytes* and is sent as it is received.
  1679. * Return
  1680. * 0
  1681. *
  1682. * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes)
  1683. * Description
  1684. * For socket policies, prevent the execution of the verdict eBPF
  1685. * program for message *msg* until *bytes* (byte number) have been
  1686. * accumulated.
  1687. *
  1688. * This can be used when one needs a specific number of bytes
  1689. * before a verdict can be assigned, even if the data spans
  1690. * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme
  1691. * case would be a user calling **sendmsg**\ () repeatedly with
  1692. * 1-byte long message segments. Obviously, this is bad for
  1693. * performance, but it is still valid. If the eBPF program needs
  1694. * *bytes* bytes to validate a header, this helper can be used to
  1695. * prevent the eBPF program to be called again until *bytes* have
  1696. * been accumulated.
  1697. * Return
  1698. * 0
  1699. *
  1700. * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags)
  1701. * Description
  1702. * For socket policies, pull in non-linear data from user space
  1703. * for *msg* and set pointers *msg*\ **->data** and *msg*\
  1704. * **->data_end** to *start* and *end* bytes offsets into *msg*,
  1705. * respectively.
  1706. *
  1707. * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a
  1708. * *msg* it can only parse data that the (**data**, **data_end**)
  1709. * pointers have already consumed. For **sendmsg**\ () hooks this
  1710. * is likely the first scatterlist element. But for calls relying
  1711. * on the **sendpage** handler (e.g. **sendfile**\ ()) this will
  1712. * be the range (**0**, **0**) because the data is shared with
  1713. * user space and by default the objective is to avoid allowing
  1714. * user space to modify data while (or after) eBPF verdict is
  1715. * being decided. This helper can be used to pull in data and to
  1716. * set the start and end pointer to given values. Data will be
  1717. * copied if necessary (i.e. if data was not linear and if start
  1718. * and end pointers do not point to the same chunk).
  1719. *
  1720. * A call to this helper is susceptible to change the underlaying
  1721. * packet buffer. Therefore, at load time, all checks on pointers
  1722. * previously done by the verifier are invalidated and must be
  1723. * performed again, if the helper is used in combination with
  1724. * direct packet access.
  1725. *
  1726. * All values for *flags* are reserved for future usage, and must
  1727. * be left at zero.
  1728. * Return
  1729. * 0 on success, or a negative error in case of failure.
  1730. *
  1731. * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len)
  1732. * Description
  1733. * Bind the socket associated to *ctx* to the address pointed by
  1734. * *addr*, of length *addr_len*. This allows for making outgoing
  1735. * connection from the desired IP address, which can be useful for
  1736. * example when all processes inside a cgroup should use one
  1737. * single IP address on a host that has multiple IP configured.
  1738. *
  1739. * This helper works for IPv4 and IPv6, TCP and UDP sockets. The
  1740. * domain (*addr*\ **->sa_family**) must be **AF_INET** (or
  1741. * **AF_INET6**). Looking for a free port to bind to can be
  1742. * expensive, therefore binding to port is not permitted by the
  1743. * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively)
  1744. * must be set to zero.
  1745. * Return
  1746. * 0 on success, or a negative error in case of failure.
  1747. *
  1748. * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta)
  1749. * Description
  1750. * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is
  1751. * only possible to shrink the packet as of this writing,
  1752. * therefore *delta* must be a negative integer.
  1753. *
  1754. * A call to this helper is susceptible to change the underlaying
  1755. * packet buffer. Therefore, at load time, all checks on pointers
  1756. * previously done by the verifier are invalidated and must be
  1757. * performed again, if the helper is used in combination with
  1758. * direct packet access.
  1759. * Return
  1760. * 0 on success, or a negative error in case of failure.
  1761. *
  1762. * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags)
  1763. * Description
  1764. * Retrieve the XFRM state (IP transform framework, see also
  1765. * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*.
  1766. *
  1767. * The retrieved value is stored in the **struct bpf_xfrm_state**
  1768. * pointed by *xfrm_state* and of length *size*.
  1769. *
  1770. * All values for *flags* are reserved for future usage, and must
  1771. * be left at zero.
  1772. *
  1773. * This helper is available only if the kernel was compiled with
  1774. * **CONFIG_XFRM** configuration option.
  1775. * Return
  1776. * 0 on success, or a negative error in case of failure.
  1777. *
  1778. * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags)
  1779. * Description
  1780. * Return a user or a kernel stack in bpf program provided buffer.
  1781. * To achieve this, the helper needs *ctx*, which is a pointer
  1782. * to the context on which the tracing program is executed.
  1783. * To store the stacktrace, the bpf program provides *buf* with
  1784. * a nonnegative *size*.
  1785. *
  1786. * The last argument, *flags*, holds the number of stack frames to
  1787. * skip (from 0 to 255), masked with
  1788. * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set
  1789. * the following flags:
  1790. *
  1791. * **BPF_F_USER_STACK**
  1792. * Collect a user space stack instead of a kernel stack.
  1793. * **BPF_F_USER_BUILD_ID**
  1794. * Collect buildid+offset instead of ips for user stack,
  1795. * only valid if **BPF_F_USER_STACK** is also specified.
  1796. *
  1797. * **bpf_get_stack**\ () can collect up to
  1798. * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject
  1799. * to sufficient large buffer size. Note that
  1800. * this limit can be controlled with the **sysctl** program, and
  1801. * that it should be manually increased in order to profile long
  1802. * user stacks (such as stacks for Java programs). To do so, use:
  1803. *
  1804. * ::
  1805. *
  1806. * # sysctl kernel.perf_event_max_stack=<new value>
  1807. * Return
  1808. * A non-negative value equal to or less than *size* on success,
  1809. * or a negative error in case of failure.
  1810. *
  1811. * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header)
  1812. * Description
  1813. * This helper is similar to **bpf_skb_load_bytes**\ () in that
  1814. * it provides an easy way to load *len* bytes from *offset*
  1815. * from the packet associated to *skb*, into the buffer pointed
  1816. * by *to*. The difference to **bpf_skb_load_bytes**\ () is that
  1817. * a fifth argument *start_header* exists in order to select a
  1818. * base offset to start from. *start_header* can be one of:
  1819. *
  1820. * **BPF_HDR_START_MAC**
  1821. * Base offset to load data from is *skb*'s mac header.
  1822. * **BPF_HDR_START_NET**
  1823. * Base offset to load data from is *skb*'s network header.
  1824. *
  1825. * In general, "direct packet access" is the preferred method to
  1826. * access packet data, however, this helper is in particular useful
  1827. * in socket filters where *skb*\ **->data** does not always point
  1828. * to the start of the mac header and where "direct packet access"
  1829. * is not available.
  1830. * Return
  1831. * 0 on success, or a negative error in case of failure.
  1832. *
  1833. * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags)
  1834. * Description
  1835. * Do FIB lookup in kernel tables using parameters in *params*.
  1836. * If lookup is successful and result shows packet is to be
  1837. * forwarded, the neighbor tables are searched for the nexthop.
  1838. * If successful (ie., FIB lookup shows forwarding and nexthop
  1839. * is resolved), the nexthop address is returned in ipv4_dst
  1840. * or ipv6_dst based on family, smac is set to mac address of
  1841. * egress device, dmac is set to nexthop mac address, rt_metric
  1842. * is set to metric from route (IPv4/IPv6 only), and ifindex
  1843. * is set to the device index of the nexthop from the FIB lookup.
  1844. *
  1845. * *plen* argument is the size of the passed in struct.
  1846. * *flags* argument can be a combination of one or more of the
  1847. * following values:
  1848. *
  1849. * **BPF_FIB_LOOKUP_DIRECT**
  1850. * Do a direct table lookup vs full lookup using FIB
  1851. * rules.
  1852. * **BPF_FIB_LOOKUP_OUTPUT**
  1853. * Perform lookup from an egress perspective (default is
  1854. * ingress).
  1855. *
  1856. * *ctx* is either **struct xdp_md** for XDP programs or
  1857. * **struct sk_buff** tc cls_act programs.
  1858. * Return
  1859. * * < 0 if any input argument is invalid
  1860. * * 0 on success (packet is forwarded, nexthop neighbor exists)
  1861. * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the
  1862. * packet is not forwarded or needs assist from full stack
  1863. *
  1864. * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags)
  1865. * Description
  1866. * Add an entry to, or update a sockhash *map* referencing sockets.
  1867. * The *skops* is used as a new value for the entry associated to
  1868. * *key*. *flags* is one of:
  1869. *
  1870. * **BPF_NOEXIST**
  1871. * The entry for *key* must not exist in the map.
  1872. * **BPF_EXIST**
  1873. * The entry for *key* must already exist in the map.
  1874. * **BPF_ANY**
  1875. * No condition on the existence of the entry for *key*.
  1876. *
  1877. * If the *map* has eBPF programs (parser and verdict), those will
  1878. * be inherited by the socket being added. If the socket is
  1879. * already attached to eBPF programs, this results in an error.
  1880. * Return
  1881. * 0 on success, or a negative error in case of failure.
  1882. *
  1883. * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags)
  1884. * Description
  1885. * This helper is used in programs implementing policies at the
  1886. * socket level. If the message *msg* is allowed to pass (i.e. if
  1887. * the verdict eBPF program returns **SK_PASS**), redirect it to
  1888. * the socket referenced by *map* (of type
  1889. * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  1890. * egress interfaces can be used for redirection. The
  1891. * **BPF_F_INGRESS** value in *flags* is used to make the
  1892. * distinction (ingress path is selected if the flag is present,
  1893. * egress path otherwise). This is the only flag supported for now.
  1894. * Return
  1895. * **SK_PASS** on success, or **SK_DROP** on error.
  1896. *
  1897. * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags)
  1898. * Description
  1899. * This helper is used in programs implementing policies at the
  1900. * skb socket level. If the sk_buff *skb* is allowed to pass (i.e.
  1901. * if the verdeict eBPF program returns **SK_PASS**), redirect it
  1902. * to the socket referenced by *map* (of type
  1903. * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and
  1904. * egress interfaces can be used for redirection. The
  1905. * **BPF_F_INGRESS** value in *flags* is used to make the
  1906. * distinction (ingress path is selected if the flag is present,
  1907. * egress otherwise). This is the only flag supported for now.
  1908. * Return
  1909. * **SK_PASS** on success, or **SK_DROP** on error.
  1910. *
  1911. * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len)
  1912. * Description
  1913. * Encapsulate the packet associated to *skb* within a Layer 3
  1914. * protocol header. This header is provided in the buffer at
  1915. * address *hdr*, with *len* its size in bytes. *type* indicates
  1916. * the protocol of the header and can be one of:
  1917. *
  1918. * **BPF_LWT_ENCAP_SEG6**
  1919. * IPv6 encapsulation with Segment Routing Header
  1920. * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH,
  1921. * the IPv6 header is computed by the kernel.
  1922. * **BPF_LWT_ENCAP_SEG6_INLINE**
  1923. * Only works if *skb* contains an IPv6 packet. Insert a
  1924. * Segment Routing Header (**struct ipv6_sr_hdr**) inside
  1925. * the IPv6 header.
  1926. *
  1927. * A call to this helper is susceptible to change the underlaying
  1928. * packet buffer. Therefore, at load time, all checks on pointers
  1929. * previously done by the verifier are invalidated and must be
  1930. * performed again, if the helper is used in combination with
  1931. * direct packet access.
  1932. * Return
  1933. * 0 on success, or a negative error in case of failure.
  1934. *
  1935. * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len)
  1936. * Description
  1937. * Store *len* bytes from address *from* into the packet
  1938. * associated to *skb*, at *offset*. Only the flags, tag and TLVs
  1939. * inside the outermost IPv6 Segment Routing Header can be
  1940. * modified through this helper.
  1941. *
  1942. * A call to this helper is susceptible to change the underlaying
  1943. * packet buffer. Therefore, at load time, all checks on pointers
  1944. * previously done by the verifier are invalidated and must be
  1945. * performed again, if the helper is used in combination with
  1946. * direct packet access.
  1947. * Return
  1948. * 0 on success, or a negative error in case of failure.
  1949. *
  1950. * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta)
  1951. * Description
  1952. * Adjust the size allocated to TLVs in the outermost IPv6
  1953. * Segment Routing Header contained in the packet associated to
  1954. * *skb*, at position *offset* by *delta* bytes. Only offsets
  1955. * after the segments are accepted. *delta* can be as well
  1956. * positive (growing) as negative (shrinking).
  1957. *
  1958. * A call to this helper is susceptible to change the underlaying
  1959. * packet buffer. Therefore, at load time, all checks on pointers
  1960. * previously done by the verifier are invalidated and must be
  1961. * performed again, if the helper is used in combination with
  1962. * direct packet access.
  1963. * Return
  1964. * 0 on success, or a negative error in case of failure.
  1965. *
  1966. * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len)
  1967. * Description
  1968. * Apply an IPv6 Segment Routing action of type *action* to the
  1969. * packet associated to *skb*. Each action takes a parameter
  1970. * contained at address *param*, and of length *param_len* bytes.
  1971. * *action* can be one of:
  1972. *
  1973. * **SEG6_LOCAL_ACTION_END_X**
  1974. * End.X action: Endpoint with Layer-3 cross-connect.
  1975. * Type of *param*: **struct in6_addr**.
  1976. * **SEG6_LOCAL_ACTION_END_T**
  1977. * End.T action: Endpoint with specific IPv6 table lookup.
  1978. * Type of *param*: **int**.
  1979. * **SEG6_LOCAL_ACTION_END_B6**
  1980. * End.B6 action: Endpoint bound to an SRv6 policy.
  1981. * Type of param: **struct ipv6_sr_hdr**.
  1982. * **SEG6_LOCAL_ACTION_END_B6_ENCAP**
  1983. * End.B6.Encap action: Endpoint bound to an SRv6
  1984. * encapsulation policy.
  1985. * Type of param: **struct ipv6_sr_hdr**.
  1986. *
  1987. * A call to this helper is susceptible to change the underlaying
  1988. * packet buffer. Therefore, at load time, all checks on pointers
  1989. * previously done by the verifier are invalidated and must be
  1990. * performed again, if the helper is used in combination with
  1991. * direct packet access.
  1992. * Return
  1993. * 0 on success, or a negative error in case of failure.
  1994. *
  1995. * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle)
  1996. * Description
  1997. * This helper is used in programs implementing IR decoding, to
  1998. * report a successfully decoded key press with *scancode*,
  1999. * *toggle* value in the given *protocol*. The scancode will be
  2000. * translated to a keycode using the rc keymap, and reported as
  2001. * an input key down event. After a period a key up event is
  2002. * generated. This period can be extended by calling either
  2003. * **bpf_rc_keydown** () again with the same values, or calling
  2004. * **bpf_rc_repeat** ().
  2005. *
  2006. * Some protocols include a toggle bit, in case the button was
  2007. * released and pressed again between consecutive scancodes.
  2008. *
  2009. * The *ctx* should point to the lirc sample as passed into
  2010. * the program.
  2011. *
  2012. * The *protocol* is the decoded protocol number (see
  2013. * **enum rc_proto** for some predefined values).
  2014. *
  2015. * This helper is only available is the kernel was compiled with
  2016. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  2017. * "**y**".
  2018. * Return
  2019. * 0
  2020. *
  2021. * int bpf_rc_repeat(void *ctx)
  2022. * Description
  2023. * This helper is used in programs implementing IR decoding, to
  2024. * report a successfully decoded repeat key message. This delays
  2025. * the generation of a key up event for previously generated
  2026. * key down event.
  2027. *
  2028. * Some IR protocols like NEC have a special IR message for
  2029. * repeating last button, for when a button is held down.
  2030. *
  2031. * The *ctx* should point to the lirc sample as passed into
  2032. * the program.
  2033. *
  2034. * This helper is only available is the kernel was compiled with
  2035. * the **CONFIG_BPF_LIRC_MODE2** configuration option set to
  2036. * "**y**".
  2037. * Return
  2038. * 0
  2039. *
  2040. * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb)
  2041. * Description
  2042. * Return the cgroup v2 id of the socket associated with the *skb*.
  2043. * This is roughly similar to the **bpf_get_cgroup_classid**\ ()
  2044. * helper for cgroup v1 by providing a tag resp. identifier that
  2045. * can be matched on or used for map lookups e.g. to implement
  2046. * policy. The cgroup v2 id of a given path in the hierarchy is
  2047. * exposed in user space through the f_handle API in order to get
  2048. * to the same 64-bit id.
  2049. *
  2050. * This helper can be used on TC egress path, but not on ingress,
  2051. * and is available only if the kernel was compiled with the
  2052. * **CONFIG_SOCK_CGROUP_DATA** configuration option.
  2053. * Return
  2054. * The id is returned or 0 in case the id could not be retrieved.
  2055. *
  2056. * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level)
  2057. * Description
  2058. * Return id of cgroup v2 that is ancestor of cgroup associated
  2059. * with the *skb* at the *ancestor_level*. The root cgroup is at
  2060. * *ancestor_level* zero and each step down the hierarchy
  2061. * increments the level. If *ancestor_level* == level of cgroup
  2062. * associated with *skb*, then return value will be same as that
  2063. * of **bpf_skb_cgroup_id**\ ().
  2064. *
  2065. * The helper is useful to implement policies based on cgroups
  2066. * that are upper in hierarchy than immediate cgroup associated
  2067. * with *skb*.
  2068. *
  2069. * The format of returned id and helper limitations are same as in
  2070. * **bpf_skb_cgroup_id**\ ().
  2071. * Return
  2072. * The id is returned or 0 in case the id could not be retrieved.
  2073. *
  2074. * u64 bpf_get_current_cgroup_id(void)
  2075. * Return
  2076. * A 64-bit integer containing the current cgroup id based
  2077. * on the cgroup within which the current task is running.
  2078. *
  2079. * void* get_local_storage(void *map, u64 flags)
  2080. * Description
  2081. * Get the pointer to the local storage area.
  2082. * The type and the size of the local storage is defined
  2083. * by the *map* argument.
  2084. * The *flags* meaning is specific for each map type,
  2085. * and has to be 0 for cgroup local storage.
  2086. *
  2087. * Depending on the bpf program type, a local storage area
  2088. * can be shared between multiple instances of the bpf program,
  2089. * running simultaneously.
  2090. *
  2091. * A user should care about the synchronization by himself.
  2092. * For example, by using the BPF_STX_XADD instruction to alter
  2093. * the shared data.
  2094. * Return
  2095. * Pointer to the local storage area.
  2096. *
  2097. * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags)
  2098. * Description
  2099. * Select a SO_REUSEPORT sk from a BPF_MAP_TYPE_REUSEPORT_ARRAY map
  2100. * It checks the selected sk is matching the incoming
  2101. * request in the skb.
  2102. * Return
  2103. * 0 on success, or a negative error in case of failure.
  2104. */
  2105. #define __BPF_FUNC_MAPPER(FN) \
  2106. FN(unspec), \
  2107. FN(map_lookup_elem), \
  2108. FN(map_update_elem), \
  2109. FN(map_delete_elem), \
  2110. FN(probe_read), \
  2111. FN(ktime_get_ns), \
  2112. FN(trace_printk), \
  2113. FN(get_prandom_u32), \
  2114. FN(get_smp_processor_id), \
  2115. FN(skb_store_bytes), \
  2116. FN(l3_csum_replace), \
  2117. FN(l4_csum_replace), \
  2118. FN(tail_call), \
  2119. FN(clone_redirect), \
  2120. FN(get_current_pid_tgid), \
  2121. FN(get_current_uid_gid), \
  2122. FN(get_current_comm), \
  2123. FN(get_cgroup_classid), \
  2124. FN(skb_vlan_push), \
  2125. FN(skb_vlan_pop), \
  2126. FN(skb_get_tunnel_key), \
  2127. FN(skb_set_tunnel_key), \
  2128. FN(perf_event_read), \
  2129. FN(redirect), \
  2130. FN(get_route_realm), \
  2131. FN(perf_event_output), \
  2132. FN(skb_load_bytes), \
  2133. FN(get_stackid), \
  2134. FN(csum_diff), \
  2135. FN(skb_get_tunnel_opt), \
  2136. FN(skb_set_tunnel_opt), \
  2137. FN(skb_change_proto), \
  2138. FN(skb_change_type), \
  2139. FN(skb_under_cgroup), \
  2140. FN(get_hash_recalc), \
  2141. FN(get_current_task), \
  2142. FN(probe_write_user), \
  2143. FN(current_task_under_cgroup), \
  2144. FN(skb_change_tail), \
  2145. FN(skb_pull_data), \
  2146. FN(csum_update), \
  2147. FN(set_hash_invalid), \
  2148. FN(get_numa_node_id), \
  2149. FN(skb_change_head), \
  2150. FN(xdp_adjust_head), \
  2151. FN(probe_read_str), \
  2152. FN(get_socket_cookie), \
  2153. FN(get_socket_uid), \
  2154. FN(set_hash), \
  2155. FN(setsockopt), \
  2156. FN(skb_adjust_room), \
  2157. FN(redirect_map), \
  2158. FN(sk_redirect_map), \
  2159. FN(sock_map_update), \
  2160. FN(xdp_adjust_meta), \
  2161. FN(perf_event_read_value), \
  2162. FN(perf_prog_read_value), \
  2163. FN(getsockopt), \
  2164. FN(override_return), \
  2165. FN(sock_ops_cb_flags_set), \
  2166. FN(msg_redirect_map), \
  2167. FN(msg_apply_bytes), \
  2168. FN(msg_cork_bytes), \
  2169. FN(msg_pull_data), \
  2170. FN(bind), \
  2171. FN(xdp_adjust_tail), \
  2172. FN(skb_get_xfrm_state), \
  2173. FN(get_stack), \
  2174. FN(skb_load_bytes_relative), \
  2175. FN(fib_lookup), \
  2176. FN(sock_hash_update), \
  2177. FN(msg_redirect_hash), \
  2178. FN(sk_redirect_hash), \
  2179. FN(lwt_push_encap), \
  2180. FN(lwt_seg6_store_bytes), \
  2181. FN(lwt_seg6_adjust_srh), \
  2182. FN(lwt_seg6_action), \
  2183. FN(rc_repeat), \
  2184. FN(rc_keydown), \
  2185. FN(skb_cgroup_id), \
  2186. FN(get_current_cgroup_id), \
  2187. FN(get_local_storage), \
  2188. FN(sk_select_reuseport), \
  2189. FN(skb_ancestor_cgroup_id),
  2190. /* integer value in 'imm' field of BPF_CALL instruction selects which helper
  2191. * function eBPF program intends to call
  2192. */
  2193. #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x
  2194. enum bpf_func_id {
  2195. __BPF_FUNC_MAPPER(__BPF_ENUM_FN)
  2196. __BPF_FUNC_MAX_ID,
  2197. };
  2198. #undef __BPF_ENUM_FN
  2199. /* All flags used by eBPF helper functions, placed here. */
  2200. /* BPF_FUNC_skb_store_bytes flags. */
  2201. #define BPF_F_RECOMPUTE_CSUM (1ULL << 0)
  2202. #define BPF_F_INVALIDATE_HASH (1ULL << 1)
  2203. /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
  2204. * First 4 bits are for passing the header field size.
  2205. */
  2206. #define BPF_F_HDR_FIELD_MASK 0xfULL
  2207. /* BPF_FUNC_l4_csum_replace flags. */
  2208. #define BPF_F_PSEUDO_HDR (1ULL << 4)
  2209. #define BPF_F_MARK_MANGLED_0 (1ULL << 5)
  2210. #define BPF_F_MARK_ENFORCE (1ULL << 6)
  2211. /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
  2212. #define BPF_F_INGRESS (1ULL << 0)
  2213. /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
  2214. #define BPF_F_TUNINFO_IPV6 (1ULL << 0)
  2215. /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */
  2216. #define BPF_F_SKIP_FIELD_MASK 0xffULL
  2217. #define BPF_F_USER_STACK (1ULL << 8)
  2218. /* flags used by BPF_FUNC_get_stackid only. */
  2219. #define BPF_F_FAST_STACK_CMP (1ULL << 9)
  2220. #define BPF_F_REUSE_STACKID (1ULL << 10)
  2221. /* flags used by BPF_FUNC_get_stack only. */
  2222. #define BPF_F_USER_BUILD_ID (1ULL << 11)
  2223. /* BPF_FUNC_skb_set_tunnel_key flags. */
  2224. #define BPF_F_ZERO_CSUM_TX (1ULL << 1)
  2225. #define BPF_F_DONT_FRAGMENT (1ULL << 2)
  2226. #define BPF_F_SEQ_NUMBER (1ULL << 3)
  2227. /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and
  2228. * BPF_FUNC_perf_event_read_value flags.
  2229. */
  2230. #define BPF_F_INDEX_MASK 0xffffffffULL
  2231. #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK
  2232. /* BPF_FUNC_perf_event_output for sk_buff input context. */
  2233. #define BPF_F_CTXLEN_MASK (0xfffffULL << 32)
  2234. /* Mode for BPF_FUNC_skb_adjust_room helper. */
  2235. enum bpf_adj_room_mode {
  2236. BPF_ADJ_ROOM_NET,
  2237. };
  2238. /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */
  2239. enum bpf_hdr_start_off {
  2240. BPF_HDR_START_MAC,
  2241. BPF_HDR_START_NET,
  2242. };
  2243. /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */
  2244. enum bpf_lwt_encap_mode {
  2245. BPF_LWT_ENCAP_SEG6,
  2246. BPF_LWT_ENCAP_SEG6_INLINE
  2247. };
  2248. /* user accessible mirror of in-kernel sk_buff.
  2249. * new fields can only be added to the end of this structure
  2250. */
  2251. struct __sk_buff {
  2252. __u32 len;
  2253. __u32 pkt_type;
  2254. __u32 mark;
  2255. __u32 queue_mapping;
  2256. __u32 protocol;
  2257. __u32 vlan_present;
  2258. __u32 vlan_tci;
  2259. __u32 vlan_proto;
  2260. __u32 priority;
  2261. __u32 ingress_ifindex;
  2262. __u32 ifindex;
  2263. __u32 tc_index;
  2264. __u32 cb[5];
  2265. __u32 hash;
  2266. __u32 tc_classid;
  2267. __u32 data;
  2268. __u32 data_end;
  2269. __u32 napi_id;
  2270. /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */
  2271. __u32 family;
  2272. __u32 remote_ip4; /* Stored in network byte order */
  2273. __u32 local_ip4; /* Stored in network byte order */
  2274. __u32 remote_ip6[4]; /* Stored in network byte order */
  2275. __u32 local_ip6[4]; /* Stored in network byte order */
  2276. __u32 remote_port; /* Stored in network byte order */
  2277. __u32 local_port; /* stored in host byte order */
  2278. /* ... here. */
  2279. __u32 data_meta;
  2280. };
  2281. struct bpf_tunnel_key {
  2282. __u32 tunnel_id;
  2283. union {
  2284. __u32 remote_ipv4;
  2285. __u32 remote_ipv6[4];
  2286. };
  2287. __u8 tunnel_tos;
  2288. __u8 tunnel_ttl;
  2289. __u16 tunnel_ext; /* Padding, future use. */
  2290. __u32 tunnel_label;
  2291. };
  2292. /* user accessible mirror of in-kernel xfrm_state.
  2293. * new fields can only be added to the end of this structure
  2294. */
  2295. struct bpf_xfrm_state {
  2296. __u32 reqid;
  2297. __u32 spi; /* Stored in network byte order */
  2298. __u16 family;
  2299. __u16 ext; /* Padding, future use. */
  2300. union {
  2301. __u32 remote_ipv4; /* Stored in network byte order */
  2302. __u32 remote_ipv6[4]; /* Stored in network byte order */
  2303. };
  2304. };
  2305. /* Generic BPF return codes which all BPF program types may support.
  2306. * The values are binary compatible with their TC_ACT_* counter-part to
  2307. * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT
  2308. * programs.
  2309. *
  2310. * XDP is handled seprately, see XDP_*.
  2311. */
  2312. enum bpf_ret_code {
  2313. BPF_OK = 0,
  2314. /* 1 reserved */
  2315. BPF_DROP = 2,
  2316. /* 3-6 reserved */
  2317. BPF_REDIRECT = 7,
  2318. /* >127 are reserved for prog type specific return codes */
  2319. };
  2320. struct bpf_sock {
  2321. __u32 bound_dev_if;
  2322. __u32 family;
  2323. __u32 type;
  2324. __u32 protocol;
  2325. __u32 mark;
  2326. __u32 priority;
  2327. __u32 src_ip4; /* Allows 1,2,4-byte read.
  2328. * Stored in network byte order.
  2329. */
  2330. __u32 src_ip6[4]; /* Allows 1,2,4-byte read.
  2331. * Stored in network byte order.
  2332. */
  2333. __u32 src_port; /* Allows 4-byte read.
  2334. * Stored in host byte order
  2335. */
  2336. };
  2337. #define XDP_PACKET_HEADROOM 256
  2338. /* User return codes for XDP prog type.
  2339. * A valid XDP program must return one of these defined values. All other
  2340. * return codes are reserved for future use. Unknown return codes will
  2341. * result in packet drops and a warning via bpf_warn_invalid_xdp_action().
  2342. */
  2343. enum xdp_action {
  2344. XDP_ABORTED = 0,
  2345. XDP_DROP,
  2346. XDP_PASS,
  2347. XDP_TX,
  2348. XDP_REDIRECT,
  2349. };
  2350. /* user accessible metadata for XDP packet hook
  2351. * new fields must be added to the end of this structure
  2352. */
  2353. struct xdp_md {
  2354. __u32 data;
  2355. __u32 data_end;
  2356. __u32 data_meta;
  2357. /* Below access go through struct xdp_rxq_info */
  2358. __u32 ingress_ifindex; /* rxq->dev->ifindex */
  2359. __u32 rx_queue_index; /* rxq->queue_index */
  2360. };
  2361. enum sk_action {
  2362. SK_DROP = 0,
  2363. SK_PASS,
  2364. };
  2365. /* user accessible metadata for SK_MSG packet hook, new fields must
  2366. * be added to the end of this structure
  2367. */
  2368. struct sk_msg_md {
  2369. void *data;
  2370. void *data_end;
  2371. __u32 family;
  2372. __u32 remote_ip4; /* Stored in network byte order */
  2373. __u32 local_ip4; /* Stored in network byte order */
  2374. __u32 remote_ip6[4]; /* Stored in network byte order */
  2375. __u32 local_ip6[4]; /* Stored in network byte order */
  2376. __u32 remote_port; /* Stored in network byte order */
  2377. __u32 local_port; /* stored in host byte order */
  2378. };
  2379. struct sk_reuseport_md {
  2380. /*
  2381. * Start of directly accessible data. It begins from
  2382. * the tcp/udp header.
  2383. */
  2384. void *data;
  2385. void *data_end; /* End of directly accessible data */
  2386. /*
  2387. * Total length of packet (starting from the tcp/udp header).
  2388. * Note that the directly accessible bytes (data_end - data)
  2389. * could be less than this "len". Those bytes could be
  2390. * indirectly read by a helper "bpf_skb_load_bytes()".
  2391. */
  2392. __u32 len;
  2393. /*
  2394. * Eth protocol in the mac header (network byte order). e.g.
  2395. * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD)
  2396. */
  2397. __u32 eth_protocol;
  2398. __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */
  2399. __u32 bind_inany; /* Is sock bound to an INANY address? */
  2400. __u32 hash; /* A hash of the packet 4 tuples */
  2401. };
  2402. #define BPF_TAG_SIZE 8
  2403. struct bpf_prog_info {
  2404. __u32 type;
  2405. __u32 id;
  2406. __u8 tag[BPF_TAG_SIZE];
  2407. __u32 jited_prog_len;
  2408. __u32 xlated_prog_len;
  2409. __aligned_u64 jited_prog_insns;
  2410. __aligned_u64 xlated_prog_insns;
  2411. __u64 load_time; /* ns since boottime */
  2412. __u32 created_by_uid;
  2413. __u32 nr_map_ids;
  2414. __aligned_u64 map_ids;
  2415. char name[BPF_OBJ_NAME_LEN];
  2416. __u32 ifindex;
  2417. __u32 gpl_compatible:1;
  2418. __u32 :31; /* alignment pad */
  2419. __u64 netns_dev;
  2420. __u64 netns_ino;
  2421. __u32 nr_jited_ksyms;
  2422. __u32 nr_jited_func_lens;
  2423. __aligned_u64 jited_ksyms;
  2424. __aligned_u64 jited_func_lens;
  2425. } __attribute__((aligned(8)));
  2426. struct bpf_map_info {
  2427. __u32 type;
  2428. __u32 id;
  2429. __u32 key_size;
  2430. __u32 value_size;
  2431. __u32 max_entries;
  2432. __u32 map_flags;
  2433. char name[BPF_OBJ_NAME_LEN];
  2434. __u32 ifindex;
  2435. __u32 :32;
  2436. __u64 netns_dev;
  2437. __u64 netns_ino;
  2438. __u32 btf_id;
  2439. __u32 btf_key_type_id;
  2440. __u32 btf_value_type_id;
  2441. } __attribute__((aligned(8)));
  2442. struct bpf_btf_info {
  2443. __aligned_u64 btf;
  2444. __u32 btf_size;
  2445. __u32 id;
  2446. } __attribute__((aligned(8)));
  2447. /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed
  2448. * by user and intended to be used by socket (e.g. to bind to, depends on
  2449. * attach attach type).
  2450. */
  2451. struct bpf_sock_addr {
  2452. __u32 user_family; /* Allows 4-byte read, but no write. */
  2453. __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write.
  2454. * Stored in network byte order.
  2455. */
  2456. __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
  2457. * Stored in network byte order.
  2458. */
  2459. __u32 user_port; /* Allows 4-byte read and write.
  2460. * Stored in network byte order
  2461. */
  2462. __u32 family; /* Allows 4-byte read, but no write */
  2463. __u32 type; /* Allows 4-byte read, but no write */
  2464. __u32 protocol; /* Allows 4-byte read, but no write */
  2465. __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write.
  2466. * Stored in network byte order.
  2467. */
  2468. __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write.
  2469. * Stored in network byte order.
  2470. */
  2471. };
  2472. /* User bpf_sock_ops struct to access socket values and specify request ops
  2473. * and their replies.
  2474. * Some of this fields are in network (bigendian) byte order and may need
  2475. * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h).
  2476. * New fields can only be added at the end of this structure
  2477. */
  2478. struct bpf_sock_ops {
  2479. __u32 op;
  2480. union {
  2481. __u32 args[4]; /* Optionally passed to bpf program */
  2482. __u32 reply; /* Returned by bpf program */
  2483. __u32 replylong[4]; /* Optionally returned by bpf prog */
  2484. };
  2485. __u32 family;
  2486. __u32 remote_ip4; /* Stored in network byte order */
  2487. __u32 local_ip4; /* Stored in network byte order */
  2488. __u32 remote_ip6[4]; /* Stored in network byte order */
  2489. __u32 local_ip6[4]; /* Stored in network byte order */
  2490. __u32 remote_port; /* Stored in network byte order */
  2491. __u32 local_port; /* stored in host byte order */
  2492. __u32 is_fullsock; /* Some TCP fields are only valid if
  2493. * there is a full socket. If not, the
  2494. * fields read as zero.
  2495. */
  2496. __u32 snd_cwnd;
  2497. __u32 srtt_us; /* Averaged RTT << 3 in usecs */
  2498. __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */
  2499. __u32 state;
  2500. __u32 rtt_min;
  2501. __u32 snd_ssthresh;
  2502. __u32 rcv_nxt;
  2503. __u32 snd_nxt;
  2504. __u32 snd_una;
  2505. __u32 mss_cache;
  2506. __u32 ecn_flags;
  2507. __u32 rate_delivered;
  2508. __u32 rate_interval_us;
  2509. __u32 packets_out;
  2510. __u32 retrans_out;
  2511. __u32 total_retrans;
  2512. __u32 segs_in;
  2513. __u32 data_segs_in;
  2514. __u32 segs_out;
  2515. __u32 data_segs_out;
  2516. __u32 lost_out;
  2517. __u32 sacked_out;
  2518. __u32 sk_txhash;
  2519. __u64 bytes_received;
  2520. __u64 bytes_acked;
  2521. };
  2522. /* Definitions for bpf_sock_ops_cb_flags */
  2523. #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0)
  2524. #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1)
  2525. #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2)
  2526. #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently
  2527. * supported cb flags
  2528. */
  2529. /* List of known BPF sock_ops operators.
  2530. * New entries can only be added at the end
  2531. */
  2532. enum {
  2533. BPF_SOCK_OPS_VOID,
  2534. BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or
  2535. * -1 if default value should be used
  2536. */
  2537. BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized
  2538. * window (in packets) or -1 if default
  2539. * value should be used
  2540. */
  2541. BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an
  2542. * active connection is initialized
  2543. */
  2544. BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an
  2545. * active connection is
  2546. * established
  2547. */
  2548. BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a
  2549. * passive connection is
  2550. * established
  2551. */
  2552. BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control
  2553. * needs ECN
  2554. */
  2555. BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is
  2556. * based on the path and may be
  2557. * dependent on the congestion control
  2558. * algorithm. In general it indicates
  2559. * a congestion threshold. RTTs above
  2560. * this indicate congestion
  2561. */
  2562. BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered.
  2563. * Arg1: value of icsk_retransmits
  2564. * Arg2: value of icsk_rto
  2565. * Arg3: whether RTO has expired
  2566. */
  2567. BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted.
  2568. * Arg1: sequence number of 1st byte
  2569. * Arg2: # segments
  2570. * Arg3: return value of
  2571. * tcp_transmit_skb (0 => success)
  2572. */
  2573. BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state.
  2574. * Arg1: old_state
  2575. * Arg2: new_state
  2576. */
  2577. BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after
  2578. * socket transition to LISTEN state.
  2579. */
  2580. };
  2581. /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect
  2582. * changes between the TCP and BPF versions. Ideally this should never happen.
  2583. * If it does, we need to add code to convert them before calling
  2584. * the BPF sock_ops function.
  2585. */
  2586. enum {
  2587. BPF_TCP_ESTABLISHED = 1,
  2588. BPF_TCP_SYN_SENT,
  2589. BPF_TCP_SYN_RECV,
  2590. BPF_TCP_FIN_WAIT1,
  2591. BPF_TCP_FIN_WAIT2,
  2592. BPF_TCP_TIME_WAIT,
  2593. BPF_TCP_CLOSE,
  2594. BPF_TCP_CLOSE_WAIT,
  2595. BPF_TCP_LAST_ACK,
  2596. BPF_TCP_LISTEN,
  2597. BPF_TCP_CLOSING, /* Now a valid state */
  2598. BPF_TCP_NEW_SYN_RECV,
  2599. BPF_TCP_MAX_STATES /* Leave at the end! */
  2600. };
  2601. #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */
  2602. #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */
  2603. struct bpf_perf_event_value {
  2604. __u64 counter;
  2605. __u64 enabled;
  2606. __u64 running;
  2607. };
  2608. #define BPF_DEVCG_ACC_MKNOD (1ULL << 0)
  2609. #define BPF_DEVCG_ACC_READ (1ULL << 1)
  2610. #define BPF_DEVCG_ACC_WRITE (1ULL << 2)
  2611. #define BPF_DEVCG_DEV_BLOCK (1ULL << 0)
  2612. #define BPF_DEVCG_DEV_CHAR (1ULL << 1)
  2613. struct bpf_cgroup_dev_ctx {
  2614. /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */
  2615. __u32 access_type;
  2616. __u32 major;
  2617. __u32 minor;
  2618. };
  2619. struct bpf_raw_tracepoint_args {
  2620. __u64 args[0];
  2621. };
  2622. /* DIRECT: Skip the FIB rules and go to FIB table associated with device
  2623. * OUTPUT: Do lookup from egress perspective; default is ingress
  2624. */
  2625. #define BPF_FIB_LOOKUP_DIRECT (1U << 0)
  2626. #define BPF_FIB_LOOKUP_OUTPUT (1U << 1)
  2627. enum {
  2628. BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */
  2629. BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */
  2630. BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */
  2631. BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */
  2632. BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */
  2633. BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */
  2634. BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */
  2635. BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */
  2636. BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */
  2637. };
  2638. struct bpf_fib_lookup {
  2639. /* input: network family for lookup (AF_INET, AF_INET6)
  2640. * output: network family of egress nexthop
  2641. */
  2642. __u8 family;
  2643. /* set if lookup is to consider L4 data - e.g., FIB rules */
  2644. __u8 l4_protocol;
  2645. __be16 sport;
  2646. __be16 dport;
  2647. /* total length of packet from network header - used for MTU check */
  2648. __u16 tot_len;
  2649. /* input: L3 device index for lookup
  2650. * output: device index from FIB lookup
  2651. */
  2652. __u32 ifindex;
  2653. union {
  2654. /* inputs to lookup */
  2655. __u8 tos; /* AF_INET */
  2656. __be32 flowinfo; /* AF_INET6, flow_label + priority */
  2657. /* output: metric of fib result (IPv4/IPv6 only) */
  2658. __u32 rt_metric;
  2659. };
  2660. union {
  2661. __be32 ipv4_src;
  2662. __u32 ipv6_src[4]; /* in6_addr; network order */
  2663. };
  2664. /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in
  2665. * network header. output: bpf_fib_lookup sets to gateway address
  2666. * if FIB lookup returns gateway route
  2667. */
  2668. union {
  2669. __be32 ipv4_dst;
  2670. __u32 ipv6_dst[4]; /* in6_addr; network order */
  2671. };
  2672. /* output */
  2673. __be16 h_vlan_proto;
  2674. __be16 h_vlan_TCI;
  2675. __u8 smac[6]; /* ETH_ALEN */
  2676. __u8 dmac[6]; /* ETH_ALEN */
  2677. };
  2678. enum bpf_task_fd_type {
  2679. BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */
  2680. BPF_FD_TYPE_TRACEPOINT, /* tp name */
  2681. BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */
  2682. BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */
  2683. BPF_FD_TYPE_UPROBE, /* filename + offset */
  2684. BPF_FD_TYPE_URETPROBE, /* filename + offset */
  2685. };
  2686. #endif /* _UAPI__LINUX_BPF_H__ */