1e2be04c7SGreg Kroah-Hartman /* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */ 2daedfb22SAlexei Starovoitov /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com 3daedfb22SAlexei Starovoitov * 4daedfb22SAlexei Starovoitov * This program is free software; you can redistribute it and/or 5daedfb22SAlexei Starovoitov * modify it under the terms of version 2 of the GNU General Public 6daedfb22SAlexei Starovoitov * License as published by the Free Software Foundation. 7daedfb22SAlexei Starovoitov */ 8daedfb22SAlexei Starovoitov #ifndef _UAPI__LINUX_BPF_H__ 9daedfb22SAlexei Starovoitov #define _UAPI__LINUX_BPF_H__ 10daedfb22SAlexei Starovoitov 11daedfb22SAlexei Starovoitov #include <linux/types.h> 12c15952dcSAlexei Starovoitov #include <linux/bpf_common.h> 13daedfb22SAlexei Starovoitov 14daedfb22SAlexei Starovoitov /* Extended instruction set based on top of classic BPF */ 15daedfb22SAlexei Starovoitov 16daedfb22SAlexei Starovoitov /* instruction classes */ 17d405c740SJiong Wang #define BPF_JMP32 0x06 /* jmp mode in word width */ 18daedfb22SAlexei Starovoitov #define BPF_ALU64 0x07 /* alu mode in double word width */ 19daedfb22SAlexei Starovoitov 20daedfb22SAlexei Starovoitov /* ld/ldx fields */ 21cb5f7334SJesper Dangaard Brouer #define BPF_DW 0x18 /* double word (64-bit) */ 22daedfb22SAlexei Starovoitov #define BPF_XADD 0xc0 /* exclusive add */ 23daedfb22SAlexei Starovoitov 24daedfb22SAlexei Starovoitov /* alu/jmp fields */ 25daedfb22SAlexei Starovoitov #define BPF_MOV 0xb0 /* mov reg to reg */ 26daedfb22SAlexei Starovoitov #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 27daedfb22SAlexei Starovoitov 28daedfb22SAlexei Starovoitov /* change endianness of a register */ 29daedfb22SAlexei Starovoitov #define BPF_END 0xd0 /* flags for endianness conversion: */ 30daedfb22SAlexei Starovoitov #define BPF_TO_LE 0x00 /* convert to little-endian */ 31daedfb22SAlexei Starovoitov #define BPF_TO_BE 0x08 /* convert to big-endian */ 32daedfb22SAlexei Starovoitov #define BPF_FROM_LE BPF_TO_LE 33daedfb22SAlexei Starovoitov #define BPF_FROM_BE BPF_TO_BE 34daedfb22SAlexei Starovoitov 3592b31a9aSDaniel Borkmann /* jmp encodings */ 36daedfb22SAlexei Starovoitov #define BPF_JNE 0x50 /* jump != */ 3792b31a9aSDaniel Borkmann #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 3892b31a9aSDaniel Borkmann #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 39daedfb22SAlexei Starovoitov #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 40daedfb22SAlexei Starovoitov #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 4192b31a9aSDaniel Borkmann #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 4292b31a9aSDaniel Borkmann #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 43daedfb22SAlexei Starovoitov #define BPF_CALL 0x80 /* function call */ 44daedfb22SAlexei Starovoitov #define BPF_EXIT 0x90 /* function return */ 45daedfb22SAlexei Starovoitov 46daedfb22SAlexei Starovoitov /* Register numbers */ 47daedfb22SAlexei Starovoitov enum { 48daedfb22SAlexei Starovoitov BPF_REG_0 = 0, 49daedfb22SAlexei Starovoitov BPF_REG_1, 50daedfb22SAlexei Starovoitov BPF_REG_2, 51daedfb22SAlexei Starovoitov BPF_REG_3, 52daedfb22SAlexei Starovoitov BPF_REG_4, 53daedfb22SAlexei Starovoitov BPF_REG_5, 54daedfb22SAlexei Starovoitov BPF_REG_6, 55daedfb22SAlexei Starovoitov BPF_REG_7, 56daedfb22SAlexei Starovoitov BPF_REG_8, 57daedfb22SAlexei Starovoitov BPF_REG_9, 58daedfb22SAlexei Starovoitov BPF_REG_10, 59daedfb22SAlexei Starovoitov __MAX_BPF_REG, 60daedfb22SAlexei Starovoitov }; 61daedfb22SAlexei Starovoitov 62daedfb22SAlexei Starovoitov /* BPF has 10 general purpose 64-bit registers and stack frame. */ 63daedfb22SAlexei Starovoitov #define MAX_BPF_REG __MAX_BPF_REG 64daedfb22SAlexei Starovoitov 65daedfb22SAlexei Starovoitov struct bpf_insn { 66daedfb22SAlexei Starovoitov __u8 code; /* opcode */ 67daedfb22SAlexei Starovoitov __u8 dst_reg:4; /* dest register */ 68daedfb22SAlexei Starovoitov __u8 src_reg:4; /* source register */ 69daedfb22SAlexei Starovoitov __s16 off; /* signed offset */ 70daedfb22SAlexei Starovoitov __s32 imm; /* signed immediate constant */ 71daedfb22SAlexei Starovoitov }; 72daedfb22SAlexei Starovoitov 73b95a5c4dSDaniel Mack /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 74b95a5c4dSDaniel Mack struct bpf_lpm_trie_key { 75b95a5c4dSDaniel Mack __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 761e6e9d0fSGustavo A. R. Silva __u8 data[0]; /* Arbitrary size */ 77b95a5c4dSDaniel Mack }; 78b95a5c4dSDaniel Mack 79de9cbbaaSRoman Gushchin struct bpf_cgroup_storage_key { 80de9cbbaaSRoman Gushchin __u64 cgroup_inode_id; /* cgroup inode id */ 81de9cbbaaSRoman Gushchin __u32 attach_type; /* program attach type */ 82de9cbbaaSRoman Gushchin }; 83de9cbbaaSRoman Gushchin 845e7b3020SYonghong Song union bpf_iter_link_info { 855e7b3020SYonghong Song struct { 865e7b3020SYonghong Song __u32 map_fd; 875e7b3020SYonghong Song } map; 885e7b3020SYonghong Song }; 895e7b3020SYonghong Song 90b2197755SDaniel Borkmann /* BPF syscall commands, see bpf(2) man-page for details. */ 9199c55f7dSAlexei Starovoitov enum bpf_cmd { 9299c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 93db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 94db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 95db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 96db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 9709756af4SAlexei Starovoitov BPF_PROG_LOAD, 98b2197755SDaniel Borkmann BPF_OBJ_PIN, 99b2197755SDaniel Borkmann BPF_OBJ_GET, 100f4324551SDaniel Mack BPF_PROG_ATTACH, 101f4324551SDaniel Mack BPF_PROG_DETACH, 1021cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 10334ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 10434ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 105b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 106bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 1071e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 108468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 109c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 110f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 11178958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 11241bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 113bd513cd0SMauricio Vasquez B BPF_MAP_LOOKUP_AND_DELETE_ELEM, 11487df15deSDaniel Borkmann BPF_MAP_FREEZE, 1151b9ed84eSQuentin Monnet BPF_BTF_GET_NEXT_ID, 116cb4d03abSBrian Vazquez BPF_MAP_LOOKUP_BATCH, 11705799638SYonghong Song BPF_MAP_LOOKUP_AND_DELETE_BATCH, 118aa2e93b8SBrian Vazquez BPF_MAP_UPDATE_BATCH, 119aa2e93b8SBrian Vazquez BPF_MAP_DELETE_BATCH, 120af6eea57SAndrii Nakryiko BPF_LINK_CREATE, 1210c991ebcSAndrii Nakryiko BPF_LINK_UPDATE, 1222d602c8cSAndrii Nakryiko BPF_LINK_GET_FD_BY_ID, 1232d602c8cSAndrii Nakryiko BPF_LINK_GET_NEXT_ID, 124d46edd67SSong Liu BPF_ENABLE_STATS, 125ac51d99bSYonghong Song BPF_ITER_CREATE, 12673b11c2aSAndrii Nakryiko BPF_LINK_DETACH, 127ef15314aSYiFei Zhu BPF_PROG_BIND_MAP, 12899c55f7dSAlexei Starovoitov }; 12999c55f7dSAlexei Starovoitov 13099c55f7dSAlexei Starovoitov enum bpf_map_type { 13199c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 1320f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 13328fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 13404fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 135ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 136824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 137a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 138d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 1394ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 14029ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 1418f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 142b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 14356f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 144bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 145546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 146174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 1476710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 148fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 14981110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 150de9cbbaaSRoman Gushchin BPF_MAP_TYPE_CGROUP_STORAGE, 1515dc4c4b7SMartin KaFai Lau BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 152b741f163SRoman Gushchin BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 153f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_QUEUE, 154f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_STACK, 1556ac99e8fSMartin KaFai Lau BPF_MAP_TYPE_SK_STORAGE, 1566f9d451aSToke Høiland-Jørgensen BPF_MAP_TYPE_DEVMAP_HASH, 15785d33df3SMartin KaFai Lau BPF_MAP_TYPE_STRUCT_OPS, 158457f4436SAndrii Nakryiko BPF_MAP_TYPE_RINGBUF, 1598ea63684SKP Singh BPF_MAP_TYPE_INODE_STORAGE, 16099c55f7dSAlexei Starovoitov }; 16199c55f7dSAlexei Starovoitov 1626c4fc209SDaniel Borkmann /* Note that tracing related programs such as 1636c4fc209SDaniel Borkmann * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 1646c4fc209SDaniel Borkmann * are not subject to a stable API since kernel internal data 1656c4fc209SDaniel Borkmann * structures can change from release to release and may 1666c4fc209SDaniel Borkmann * therefore break existing tracing BPF programs. Tracing BPF 1676c4fc209SDaniel Borkmann * programs correspond to /a/ specific kernel which is to be 1686c4fc209SDaniel Borkmann * analyzed, and not /a/ specific kernel /and/ all future ones. 1696c4fc209SDaniel Borkmann */ 17009756af4SAlexei Starovoitov enum bpf_prog_type { 17109756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 172ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 1732541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 17496be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 17594caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 17698b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 1776a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 1780515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 1790e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 18061023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 1813a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 1823a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 1833a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 18440304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 185b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 186ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 1874f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 188c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 1894fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 190004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 191f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 1922dbb9b9eSMartin KaFai Lau BPF_PROG_TYPE_SK_REUSEPORT, 193d58e468bSPetar Penkov BPF_PROG_TYPE_FLOW_DISSECTOR, 1947b146cebSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SYSCTL, 1959df1c28bSMatt Mullins BPF_PROG_TYPE_RAW_TRACEPOINT_WRITABLE, 1960d01da6aSStanislav Fomichev BPF_PROG_TYPE_CGROUP_SOCKOPT, 197f1b9509cSAlexei Starovoitov BPF_PROG_TYPE_TRACING, 19827ae7997SMartin KaFai Lau BPF_PROG_TYPE_STRUCT_OPS, 199be8704ffSAlexei Starovoitov BPF_PROG_TYPE_EXT, 200fc611f47SKP Singh BPF_PROG_TYPE_LSM, 201e9ddbb77SJakub Sitnicki BPF_PROG_TYPE_SK_LOOKUP, 20209756af4SAlexei Starovoitov }; 20309756af4SAlexei Starovoitov 2040e33661dSDaniel Mack enum bpf_attach_type { 2050e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 2060e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 20761023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 20840304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 209464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 210464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 211ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 2124f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 2134fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 2144fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 215d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 216d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 217aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 218aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 2191cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 2201cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 221f4364dcfSSean Young BPF_LIRC_MODE2, 222d58e468bSPetar Penkov BPF_FLOW_DISSECTOR, 2237b146cebSAndrey Ignatov BPF_CGROUP_SYSCTL, 224983695faSDaniel Borkmann BPF_CGROUP_UDP4_RECVMSG, 225983695faSDaniel Borkmann BPF_CGROUP_UDP6_RECVMSG, 2260d01da6aSStanislav Fomichev BPF_CGROUP_GETSOCKOPT, 2270d01da6aSStanislav Fomichev BPF_CGROUP_SETSOCKOPT, 228f1b9509cSAlexei Starovoitov BPF_TRACE_RAW_TP, 229fec56f58SAlexei Starovoitov BPF_TRACE_FENTRY, 230fec56f58SAlexei Starovoitov BPF_TRACE_FEXIT, 231ae240823SKP Singh BPF_MODIFY_RETURN, 232fc611f47SKP Singh BPF_LSM_MAC, 23315d83c4dSYonghong Song BPF_TRACE_ITER, 2341b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETPEERNAME, 2351b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETPEERNAME, 2361b66d253SDaniel Borkmann BPF_CGROUP_INET4_GETSOCKNAME, 2371b66d253SDaniel Borkmann BPF_CGROUP_INET6_GETSOCKNAME, 238fbee97feSDavid Ahern BPF_XDP_DEVMAP, 239f5836749SStanislav Fomichev BPF_CGROUP_INET_SOCK_RELEASE, 24092164774SLorenzo Bianconi BPF_XDP_CPUMAP, 241e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP, 242aa8d3a71SAndrii Nakryiko BPF_XDP, 2430e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 2440e33661dSDaniel Mack }; 2450e33661dSDaniel Mack 2460e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 2470e33661dSDaniel Mack 248f2e10bffSAndrii Nakryiko enum bpf_link_type { 249f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_UNSPEC = 0, 250f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_RAW_TRACEPOINT = 1, 251f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_TRACING = 2, 252f2e10bffSAndrii Nakryiko BPF_LINK_TYPE_CGROUP = 3, 253de4e05caSYonghong Song BPF_LINK_TYPE_ITER = 4, 2547f045a49SJakub Sitnicki BPF_LINK_TYPE_NETNS = 5, 255aa8d3a71SAndrii Nakryiko BPF_LINK_TYPE_XDP = 6, 256f2e10bffSAndrii Nakryiko 257f2e10bffSAndrii Nakryiko MAX_BPF_LINK_TYPE, 258f2e10bffSAndrii Nakryiko }; 259f2e10bffSAndrii Nakryiko 260324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 261324bda9eSAlexei Starovoitov * 262324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 263324bda9eSAlexei Starovoitov * 264324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 265324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 266324bda9eSAlexei Starovoitov * 267324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 268324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 269324bda9eSAlexei Starovoitov * 270324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 271324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 272324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 273324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 274324bda9eSAlexei Starovoitov * 275324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 276324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 277324bda9eSAlexei Starovoitov * (those that were attached first, run first) 278324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 279324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 280324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 281324bda9eSAlexei Starovoitov * parent program has a chance to override it. 282324bda9eSAlexei Starovoitov * 2837dd68b32SAndrey Ignatov * With BPF_F_ALLOW_MULTI a new program is added to the end of the list of 2847dd68b32SAndrey Ignatov * programs for a cgroup. Though it's possible to replace an old program at 2857dd68b32SAndrey Ignatov * any position by also specifying BPF_F_REPLACE flag and position itself in 2867dd68b32SAndrey Ignatov * replace_bpf_fd attribute. Old program at this position will be released. 2877dd68b32SAndrey Ignatov * 288324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 289324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 290324bda9eSAlexei Starovoitov * Ex1: 291324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 292324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 293324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 294324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 295324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 296324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 297324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 298324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 299324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 300324bda9eSAlexei Starovoitov * 301324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 302324bda9eSAlexei Starovoitov * earlier programs. 3037f677633SAlexei Starovoitov */ 3047f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 305324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 3067dd68b32SAndrey Ignatov #define BPF_F_REPLACE (1U << 2) 3077f677633SAlexei Starovoitov 308e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 309e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 310e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 311e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 312e07b98d9SDavid S. Miller */ 313e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 314e07b98d9SDavid S. Miller 315e9ee9efcSDavid Miller /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 316e9ee9efcSDavid Miller * verifier will allow any alignment whatsoever. On platforms 317e9ee9efcSDavid Miller * with strict alignment requirements for loads ands stores (such 318e9ee9efcSDavid Miller * as sparc and mips) the verifier validates that all loads and 319e9ee9efcSDavid Miller * stores provably follow this requirement. This flag turns that 320e9ee9efcSDavid Miller * checking and enforcement off. 321e9ee9efcSDavid Miller * 322e9ee9efcSDavid Miller * It is mostly used for testing when we want to validate the 323e9ee9efcSDavid Miller * context and memory access aspects of the verifier, but because 324e9ee9efcSDavid Miller * of an unaligned access the alignment check would trigger before 325e9ee9efcSDavid Miller * the one we are interested in. 326e9ee9efcSDavid Miller */ 327e9ee9efcSDavid Miller #define BPF_F_ANY_ALIGNMENT (1U << 1) 328e9ee9efcSDavid Miller 329c240eff6SJiong Wang /* BPF_F_TEST_RND_HI32 is used in BPF_PROG_LOAD command for testing purpose. 330c240eff6SJiong Wang * Verifier does sub-register def/use analysis and identifies instructions whose 331c240eff6SJiong Wang * def only matters for low 32-bit, high 32-bit is never referenced later 332c240eff6SJiong Wang * through implicit zero extension. Therefore verifier notifies JIT back-ends 333c240eff6SJiong Wang * that it is safe to ignore clearing high 32-bit for these instructions. This 334c240eff6SJiong Wang * saves some back-ends a lot of code-gen. However such optimization is not 335c240eff6SJiong Wang * necessary on some arches, for example x86_64, arm64 etc, whose JIT back-ends 336c240eff6SJiong Wang * hence hasn't used verifier's analysis result. But, we really want to have a 337c240eff6SJiong Wang * way to be able to verify the correctness of the described optimization on 338c240eff6SJiong Wang * x86_64 on which testsuites are frequently exercised. 339c240eff6SJiong Wang * 340c240eff6SJiong Wang * So, this flag is introduced. Once it is set, verifier will randomize high 341c240eff6SJiong Wang * 32-bit for those instructions who has been identified as safe to ignore them. 342c240eff6SJiong Wang * Then, if verifier is not doing correct analysis, such randomization will 343c240eff6SJiong Wang * regress tests to expose bugs. 344c240eff6SJiong Wang */ 345c240eff6SJiong Wang #define BPF_F_TEST_RND_HI32 (1U << 2) 346c240eff6SJiong Wang 34710d274e8SAlexei Starovoitov /* The verifier internal test flag. Behavior is undefined */ 34810d274e8SAlexei Starovoitov #define BPF_F_TEST_STATE_FREQ (1U << 3) 34910d274e8SAlexei Starovoitov 3501e6c62a8SAlexei Starovoitov /* If BPF_F_SLEEPABLE is used in BPF_PROG_LOAD command, the verifier will 3511e6c62a8SAlexei Starovoitov * restrict map and helper usage for such programs. Sleepable BPF programs can 3521e6c62a8SAlexei Starovoitov * only be attached to hooks where kernel execution context allows sleeping. 3531e6c62a8SAlexei Starovoitov * Such programs are allowed to use helpers that may sleep like 3541e6c62a8SAlexei Starovoitov * bpf_copy_from_user(). 3551e6c62a8SAlexei Starovoitov */ 3561e6c62a8SAlexei Starovoitov #define BPF_F_SLEEPABLE (1U << 4) 3571e6c62a8SAlexei Starovoitov 358d8eca5bbSDaniel Borkmann /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 359d8eca5bbSDaniel Borkmann * two extensions: 360d8eca5bbSDaniel Borkmann * 361d8eca5bbSDaniel Borkmann * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 362d8eca5bbSDaniel Borkmann * insn[0].imm: map fd map fd 363d8eca5bbSDaniel Borkmann * insn[1].imm: 0 offset into value 364d8eca5bbSDaniel Borkmann * insn[0].off: 0 0 365d8eca5bbSDaniel Borkmann * insn[1].off: 0 0 366d8eca5bbSDaniel Borkmann * ldimm64 rewrite: address of map address of map[0]+offset 367d8eca5bbSDaniel Borkmann * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 368d8eca5bbSDaniel Borkmann */ 369f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 370d8eca5bbSDaniel Borkmann #define BPF_PSEUDO_MAP_VALUE 2 371f1a66f85SDaniel Borkmann 372cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 373cc8b0b92SAlexei Starovoitov * offset to another bpf function 374cc8b0b92SAlexei Starovoitov */ 375cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 376cc8b0b92SAlexei Starovoitov 3773274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 3781aae4bddSAndrii Nakryiko enum { 3791aae4bddSAndrii Nakryiko BPF_ANY = 0, /* create new element or update existing */ 3801aae4bddSAndrii Nakryiko BPF_NOEXIST = 1, /* create new element if it didn't exist */ 3811aae4bddSAndrii Nakryiko BPF_EXIST = 2, /* update existing element */ 3821aae4bddSAndrii Nakryiko BPF_F_LOCK = 4, /* spin_lock-ed map_lookup/map_update */ 3831aae4bddSAndrii Nakryiko }; 3843274f520SAlexei Starovoitov 38596eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 3861aae4bddSAndrii Nakryiko enum { 3871aae4bddSAndrii Nakryiko BPF_F_NO_PREALLOC = (1U << 0), 38829ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 3898f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 39029ba732aSMartin KaFai Lau * which can scale and perform better. 39129ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 39229ba732aSMartin KaFai Lau * across different LRU lists. 39329ba732aSMartin KaFai Lau */ 3941aae4bddSAndrii Nakryiko BPF_F_NO_COMMON_LRU = (1U << 1), 39596eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 3961aae4bddSAndrii Nakryiko BPF_F_NUMA_NODE = (1U << 2), 397cb4d2b3fSMartin KaFai Lau 398591fe988SDaniel Borkmann /* Flags for accessing BPF object from syscall side. */ 3991aae4bddSAndrii Nakryiko BPF_F_RDONLY = (1U << 3), 4001aae4bddSAndrii Nakryiko BPF_F_WRONLY = (1U << 4), 4016e71b04aSChenbo Feng 402615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 4031aae4bddSAndrii Nakryiko BPF_F_STACK_BUILD_ID = (1U << 5), 404615755a7SSong Liu 40596b3b6c9SLorenz Bauer /* Zero-initialize hash function seed. This should only be used for testing. */ 4061aae4bddSAndrii Nakryiko BPF_F_ZERO_SEED = (1U << 6), 40796b3b6c9SLorenz Bauer 408591fe988SDaniel Borkmann /* Flags for accessing BPF object from program side. */ 4091aae4bddSAndrii Nakryiko BPF_F_RDONLY_PROG = (1U << 7), 4101aae4bddSAndrii Nakryiko BPF_F_WRONLY_PROG = (1U << 8), 411591fe988SDaniel Borkmann 4128f51dfc7SStanislav Fomichev /* Clone map from listener for newly accepted socket */ 4131aae4bddSAndrii Nakryiko BPF_F_CLONE = (1U << 9), 4148f51dfc7SStanislav Fomichev 415fc970227SAndrii Nakryiko /* Enable memory-mapping BPF map */ 4161aae4bddSAndrii Nakryiko BPF_F_MMAPABLE = (1U << 10), 417*792cacccSSong Liu 418*792cacccSSong Liu /* Share perf_event among processes */ 419*792cacccSSong Liu BPF_F_PRESERVE_ELEMS = (1U << 11), 4201aae4bddSAndrii Nakryiko }; 421fc970227SAndrii Nakryiko 422f5bfcd95SAndrey Ignatov /* Flags for BPF_PROG_QUERY. */ 423f5bfcd95SAndrey Ignatov 424f5bfcd95SAndrey Ignatov /* Query effective (directly attached + inherited from ancestor cgroups) 425f5bfcd95SAndrey Ignatov * programs that will be executed for events within a cgroup. 426f5bfcd95SAndrey Ignatov * attach_flags with this flag are returned only for directly attached programs. 427f5bfcd95SAndrey Ignatov */ 4282f183360SLorenz Bauer #define BPF_F_QUERY_EFFECTIVE (1U << 0) 4292f183360SLorenz Bauer 4301b4d60ecSSong Liu /* Flags for BPF_PROG_TEST_RUN */ 4311b4d60ecSSong Liu 4321b4d60ecSSong Liu /* If set, run the test on the cpu specified by bpf_attr.test.cpu */ 4331b4d60ecSSong Liu #define BPF_F_TEST_RUN_ON_CPU (1U << 0) 4341b4d60ecSSong Liu 435d46edd67SSong Liu /* type for BPF_ENABLE_STATS */ 436d46edd67SSong Liu enum bpf_stats_type { 437d46edd67SSong Liu /* enabled run_time_ns and run_cnt */ 438d46edd67SSong Liu BPF_STATS_RUN_TIME = 0, 439d46edd67SSong Liu }; 440d46edd67SSong Liu 441615755a7SSong Liu enum bpf_stack_build_id_status { 442615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 443615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 444615755a7SSong Liu /* with valid build_id and offset */ 445615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 446615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 447615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 448615755a7SSong Liu }; 449615755a7SSong Liu 450615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 451615755a7SSong Liu struct bpf_stack_build_id { 452615755a7SSong Liu __s32 status; 453615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 454615755a7SSong Liu union { 455615755a7SSong Liu __u64 offset; 456615755a7SSong Liu __u64 ip; 457615755a7SSong Liu }; 458615755a7SSong Liu }; 459615755a7SSong Liu 4601aae4bddSAndrii Nakryiko #define BPF_OBJ_NAME_LEN 16U 4611aae4bddSAndrii Nakryiko 46299c55f7dSAlexei Starovoitov union bpf_attr { 46399c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 46499c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 46599c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 46699c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 46799c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 46896eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 46996eabe7aSMartin KaFai Lau * flags defined above. 47096eabe7aSMartin KaFai Lau */ 47156f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 47296eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 47396eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 47496eabe7aSMartin KaFai Lau */ 475067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 476a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 477a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 4789b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 4799b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 48085d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id;/* BTF type_id of a kernel- 48185d33df3SMartin KaFai Lau * struct stored as the 48285d33df3SMartin KaFai Lau * map value 48385d33df3SMartin KaFai Lau */ 48499c55f7dSAlexei Starovoitov }; 485db20fd2bSAlexei Starovoitov 486db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 487db20fd2bSAlexei Starovoitov __u32 map_fd; 488db20fd2bSAlexei Starovoitov __aligned_u64 key; 489db20fd2bSAlexei Starovoitov union { 490db20fd2bSAlexei Starovoitov __aligned_u64 value; 491db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 492db20fd2bSAlexei Starovoitov }; 4933274f520SAlexei Starovoitov __u64 flags; 494db20fd2bSAlexei Starovoitov }; 49509756af4SAlexei Starovoitov 496cb4d03abSBrian Vazquez struct { /* struct used by BPF_MAP_*_BATCH commands */ 497cb4d03abSBrian Vazquez __aligned_u64 in_batch; /* start batch, 498cb4d03abSBrian Vazquez * NULL to start from beginning 499cb4d03abSBrian Vazquez */ 500cb4d03abSBrian Vazquez __aligned_u64 out_batch; /* output: next start batch */ 501cb4d03abSBrian Vazquez __aligned_u64 keys; 502cb4d03abSBrian Vazquez __aligned_u64 values; 503cb4d03abSBrian Vazquez __u32 count; /* input/output: 504cb4d03abSBrian Vazquez * input: # of key/value 505cb4d03abSBrian Vazquez * elements 506cb4d03abSBrian Vazquez * output: # of filled elements 507cb4d03abSBrian Vazquez */ 508cb4d03abSBrian Vazquez __u32 map_fd; 509cb4d03abSBrian Vazquez __u64 elem_flags; 510cb4d03abSBrian Vazquez __u64 flags; 511cb4d03abSBrian Vazquez } batch; 512cb4d03abSBrian Vazquez 51309756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 51409756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 51509756af4SAlexei Starovoitov __u32 insn_cnt; 51609756af4SAlexei Starovoitov __aligned_u64 insns; 51709756af4SAlexei Starovoitov __aligned_u64 license; 518cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 519cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 520cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 5216c4fc209SDaniel Borkmann __u32 kern_version; /* not used */ 522e07b98d9SDavid S. Miller __u32 prog_flags; 523067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 5241f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 5255e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 5265e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 5275e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 5285e43f899SAndrey Ignatov */ 5295e43f899SAndrey Ignatov __u32 expected_attach_type; 530838e9690SYonghong Song __u32 prog_btf_fd; /* fd pointing to BTF type data */ 531838e9690SYonghong Song __u32 func_info_rec_size; /* userspace bpf_func_info size */ 532838e9690SYonghong Song __aligned_u64 func_info; /* func info */ 533838e9690SYonghong Song __u32 func_info_cnt; /* number of bpf_func_info records */ 534c454a46bSMartin KaFai Lau __u32 line_info_rec_size; /* userspace bpf_line_info size */ 535c454a46bSMartin KaFai Lau __aligned_u64 line_info; /* line info */ 536c454a46bSMartin KaFai Lau __u32 line_info_cnt; /* number of bpf_line_info records */ 537ccfe29ebSAlexei Starovoitov __u32 attach_btf_id; /* in-kernel BTF type id to attach to */ 5385b92a28aSAlexei Starovoitov __u32 attach_prog_fd; /* 0 to attach to vmlinux */ 53909756af4SAlexei Starovoitov }; 540b2197755SDaniel Borkmann 541b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 542b2197755SDaniel Borkmann __aligned_u64 pathname; 543b2197755SDaniel Borkmann __u32 bpf_fd; 5446e71b04aSChenbo Feng __u32 file_flags; 545b2197755SDaniel Borkmann }; 546f4324551SDaniel Mack 547f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 548f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 549f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 550f4324551SDaniel Mack __u32 attach_type; 5517f677633SAlexei Starovoitov __u32 attach_flags; 5527dd68b32SAndrey Ignatov __u32 replace_bpf_fd; /* previously attached eBPF 5537dd68b32SAndrey Ignatov * program to replace if 5547dd68b32SAndrey Ignatov * BPF_F_REPLACE is used 5557dd68b32SAndrey Ignatov */ 556f4324551SDaniel Mack }; 5571cf1cae9SAlexei Starovoitov 5581cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 5591cf1cae9SAlexei Starovoitov __u32 prog_fd; 5601cf1cae9SAlexei Starovoitov __u32 retval; 561b5a36b1eSLorenz Bauer __u32 data_size_in; /* input: len of data_in */ 562b5a36b1eSLorenz Bauer __u32 data_size_out; /* input/output: len of data_out 563b5a36b1eSLorenz Bauer * returns ENOSPC if data_out 564b5a36b1eSLorenz Bauer * is too small. 565b5a36b1eSLorenz Bauer */ 5661cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 5671cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 5681cf1cae9SAlexei Starovoitov __u32 repeat; 5691cf1cae9SAlexei Starovoitov __u32 duration; 570b0b9395dSStanislav Fomichev __u32 ctx_size_in; /* input: len of ctx_in */ 571b0b9395dSStanislav Fomichev __u32 ctx_size_out; /* input/output: len of ctx_out 572b0b9395dSStanislav Fomichev * returns ENOSPC if ctx_out 573b0b9395dSStanislav Fomichev * is too small. 574b0b9395dSStanislav Fomichev */ 575b0b9395dSStanislav Fomichev __aligned_u64 ctx_in; 576b0b9395dSStanislav Fomichev __aligned_u64 ctx_out; 5771b4d60ecSSong Liu __u32 flags; 5781b4d60ecSSong Liu __u32 cpu; 5791cf1cae9SAlexei Starovoitov } test; 58034ad5580SMartin KaFai Lau 581b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 582b16d9aa4SMartin KaFai Lau union { 58334ad5580SMartin KaFai Lau __u32 start_id; 584b16d9aa4SMartin KaFai Lau __u32 prog_id; 585bd5f5f4eSMartin KaFai Lau __u32 map_id; 58678958fcaSMartin KaFai Lau __u32 btf_id; 587a3b80e10SAndrii Nakryiko __u32 link_id; 588b16d9aa4SMartin KaFai Lau }; 58934ad5580SMartin KaFai Lau __u32 next_id; 5906e71b04aSChenbo Feng __u32 open_flags; 59134ad5580SMartin KaFai Lau }; 5921e270976SMartin KaFai Lau 5931e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 5941e270976SMartin KaFai Lau __u32 bpf_fd; 5951e270976SMartin KaFai Lau __u32 info_len; 5961e270976SMartin KaFai Lau __aligned_u64 info; 5971e270976SMartin KaFai Lau } info; 598468e2f64SAlexei Starovoitov 599468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 600468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 601468e2f64SAlexei Starovoitov __u32 attach_type; 602468e2f64SAlexei Starovoitov __u32 query_flags; 603468e2f64SAlexei Starovoitov __u32 attach_flags; 604468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 605468e2f64SAlexei Starovoitov __u32 prog_cnt; 606468e2f64SAlexei Starovoitov } query; 607c4f6699dSAlexei Starovoitov 608af6eea57SAndrii Nakryiko struct { /* anonymous struct used by BPF_RAW_TRACEPOINT_OPEN command */ 609c4f6699dSAlexei Starovoitov __u64 name; 610c4f6699dSAlexei Starovoitov __u32 prog_fd; 611c4f6699dSAlexei Starovoitov } raw_tracepoint; 612f56a653cSMartin KaFai Lau 613f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 614f56a653cSMartin KaFai Lau __aligned_u64 btf; 615f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 616f56a653cSMartin KaFai Lau __u32 btf_size; 617f56a653cSMartin KaFai Lau __u32 btf_log_size; 618f56a653cSMartin KaFai Lau __u32 btf_log_level; 619f56a653cSMartin KaFai Lau }; 62041bdc4b4SYonghong Song 62141bdc4b4SYonghong Song struct { 62241bdc4b4SYonghong Song __u32 pid; /* input: pid */ 62341bdc4b4SYonghong Song __u32 fd; /* input: fd */ 62441bdc4b4SYonghong Song __u32 flags; /* input: flags */ 62541bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 62641bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 62741bdc4b4SYonghong Song * tp_name for tracepoint 62841bdc4b4SYonghong Song * symbol for kprobe 62941bdc4b4SYonghong Song * filename for uprobe 63041bdc4b4SYonghong Song */ 63141bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 63241bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 63341bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 63441bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 63541bdc4b4SYonghong Song } task_fd_query; 636af6eea57SAndrii Nakryiko 637af6eea57SAndrii Nakryiko struct { /* struct used by BPF_LINK_CREATE command */ 638af6eea57SAndrii Nakryiko __u32 prog_fd; /* eBPF program to attach */ 639aa8d3a71SAndrii Nakryiko union { 640af6eea57SAndrii Nakryiko __u32 target_fd; /* object to attach to */ 641aa8d3a71SAndrii Nakryiko __u32 target_ifindex; /* target ifindex */ 642aa8d3a71SAndrii Nakryiko }; 643af6eea57SAndrii Nakryiko __u32 attach_type; /* attach type */ 644af6eea57SAndrii Nakryiko __u32 flags; /* extra flags */ 6454a1e7c0cSToke Høiland-Jørgensen union { 6464a1e7c0cSToke Høiland-Jørgensen __u32 target_btf_id; /* btf_id of target to attach to */ 6474a1e7c0cSToke Høiland-Jørgensen struct { 6485e7b3020SYonghong Song __aligned_u64 iter_info; /* extra bpf_iter_link_info */ 6495e7b3020SYonghong Song __u32 iter_info_len; /* iter_info length */ 6504a1e7c0cSToke Høiland-Jørgensen }; 6514a1e7c0cSToke Høiland-Jørgensen }; 652af6eea57SAndrii Nakryiko } link_create; 6530c991ebcSAndrii Nakryiko 6540c991ebcSAndrii Nakryiko struct { /* struct used by BPF_LINK_UPDATE command */ 6550c991ebcSAndrii Nakryiko __u32 link_fd; /* link fd */ 6560c991ebcSAndrii Nakryiko /* new program fd to update link with */ 6570c991ebcSAndrii Nakryiko __u32 new_prog_fd; 6580c991ebcSAndrii Nakryiko __u32 flags; /* extra flags */ 6590c991ebcSAndrii Nakryiko /* expected link's program fd; is specified only if 6600c991ebcSAndrii Nakryiko * BPF_F_REPLACE flag is set in flags */ 6610c991ebcSAndrii Nakryiko __u32 old_prog_fd; 6620c991ebcSAndrii Nakryiko } link_update; 6630c991ebcSAndrii Nakryiko 66473b11c2aSAndrii Nakryiko struct { 66573b11c2aSAndrii Nakryiko __u32 link_fd; 66673b11c2aSAndrii Nakryiko } link_detach; 66773b11c2aSAndrii Nakryiko 668d46edd67SSong Liu struct { /* struct used by BPF_ENABLE_STATS command */ 669d46edd67SSong Liu __u32 type; 670d46edd67SSong Liu } enable_stats; 671d46edd67SSong Liu 672ac51d99bSYonghong Song struct { /* struct used by BPF_ITER_CREATE command */ 673ac51d99bSYonghong Song __u32 link_fd; 674ac51d99bSYonghong Song __u32 flags; 675ac51d99bSYonghong Song } iter_create; 676ac51d99bSYonghong Song 677ef15314aSYiFei Zhu struct { /* struct used by BPF_PROG_BIND_MAP command */ 678ef15314aSYiFei Zhu __u32 prog_fd; 679ef15314aSYiFei Zhu __u32 map_fd; 680ef15314aSYiFei Zhu __u32 flags; /* extra flags */ 681ef15314aSYiFei Zhu } prog_bind_map; 682ef15314aSYiFei Zhu 68399c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 68499c55f7dSAlexei Starovoitov 68556a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 68656a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 68756a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 68856a092c8SQuentin Monnet * and requires the rst2man utility: 689ebb676daSThomas Graf * 69056a092c8SQuentin Monnet * $ ./scripts/bpf_helpers_doc.py \ 69156a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 69256a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 69356a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 694ebb676daSThomas Graf * 69556a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 69656a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 69756a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 69856a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 699ebb676daSThomas Graf * 70056a092c8SQuentin Monnet * Start of BPF helper function descriptions: 701ad4a5223SQuentin Monnet * 702ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 703ad4a5223SQuentin Monnet * Description 704ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 705ad4a5223SQuentin Monnet * Return 706ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 707ad4a5223SQuentin Monnet * found. 708ad4a5223SQuentin Monnet * 709bdb7b79bSAndrii Nakryiko * long bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 710ad4a5223SQuentin Monnet * Description 711ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 712ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 713ad4a5223SQuentin Monnet * 714ad4a5223SQuentin Monnet * **BPF_NOEXIST** 715ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 716ad4a5223SQuentin Monnet * **BPF_EXIST** 717ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 718ad4a5223SQuentin Monnet * **BPF_ANY** 719ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 720ad4a5223SQuentin Monnet * 721ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 722ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 723ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 724ad4a5223SQuentin Monnet * Return 725ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 726ad4a5223SQuentin Monnet * 727bdb7b79bSAndrii Nakryiko * long bpf_map_delete_elem(struct bpf_map *map, const void *key) 728ad4a5223SQuentin Monnet * Description 729ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 730ad4a5223SQuentin Monnet * Return 731ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 732ad4a5223SQuentin Monnet * 733bdb7b79bSAndrii Nakryiko * long bpf_probe_read(void *dst, u32 size, const void *unsafe_ptr) 734ad4a5223SQuentin Monnet * Description 735ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 7366ae08ae3SDaniel Borkmann * kernel space address *unsafe_ptr* and store the data in *dst*. 7376ae08ae3SDaniel Borkmann * 738ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user**\ () or 739ab8d7809SQuentin Monnet * **bpf_probe_read_kernel**\ () instead. 740ad4a5223SQuentin Monnet * Return 741ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 742ad4a5223SQuentin Monnet * 743ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 744ad4a5223SQuentin Monnet * Description 745ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 74671d19214SMaciej Żenczykowski * Does not include time the system was suspended. 747ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_MONOTONIC**) 748ad4a5223SQuentin Monnet * Return 749ad4a5223SQuentin Monnet * Current *ktime*. 750ad4a5223SQuentin Monnet * 751bdb7b79bSAndrii Nakryiko * long bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 752ad4a5223SQuentin Monnet * Description 753ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 754ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 755ad4a5223SQuentin Monnet * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 756ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 757ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 758ad4a5223SQuentin Monnet * limited to five). 759ad4a5223SQuentin Monnet * 760ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 76155c33dfbSPeter Wu * Lines are discarded while *\/sys/kernel/debug/tracing/trace* is 76255c33dfbSPeter Wu * open, use *\/sys/kernel/debug/tracing/trace_pipe* to avoid this. 763ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 764ad4a5223SQuentin Monnet * one will get depends on the options set in 765ad4a5223SQuentin Monnet * *\/sys/kernel/debug/tracing/trace_options* (see also the 766ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 767ad4a5223SQuentin Monnet * defaults to something like: 768ad4a5223SQuentin Monnet * 769ad4a5223SQuentin Monnet * :: 770ad4a5223SQuentin Monnet * 771ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 772ad4a5223SQuentin Monnet * 773ad4a5223SQuentin Monnet * In the above: 774ad4a5223SQuentin Monnet * 775ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 776ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 777ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 778ad4a5223SQuentin Monnet * running. 779ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 780ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 781ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 782ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 783ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 784ad4a5223SQuentin Monnet * are set. 785ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 786ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 787ad4a5223SQuentin Monnet * instruction pointer register. 788ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 789ad4a5223SQuentin Monnet * *fmt*. 790ad4a5223SQuentin Monnet * 791ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 792ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 793ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 794ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 795ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 796ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 797ad4a5223SQuentin Monnet * encounters an unknown specifier. 798ad4a5223SQuentin Monnet * 799ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 800ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 801b16fc097STobias Klauser * block (spanning several lines) is printed to kernel logs and 802ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 803ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 804ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 805ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 806ad4a5223SQuentin Monnet * Return 807ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 808ad4a5223SQuentin Monnet * in case of failure. 809ad4a5223SQuentin Monnet * 8101fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 8111fdd08beSQuentin Monnet * Description 8121fdd08beSQuentin Monnet * Get a pseudo-random number. 8131fdd08beSQuentin Monnet * 8141fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 8151fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 8161fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 8171fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 8181fdd08beSQuentin Monnet * cryptographically secure. 8191fdd08beSQuentin Monnet * Return 8201fdd08beSQuentin Monnet * A random 32-bit unsigned value. 8211fdd08beSQuentin Monnet * 8221fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 8231fdd08beSQuentin Monnet * Description 8241fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 8251fdd08beSQuentin Monnet * all programs run with preemption disabled, which means that the 8261fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 8271fdd08beSQuentin Monnet * program. 8281fdd08beSQuentin Monnet * Return 8291fdd08beSQuentin Monnet * The SMP id of the processor running the program. 8301fdd08beSQuentin Monnet * 831bdb7b79bSAndrii Nakryiko * long bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 832ad4a5223SQuentin Monnet * Description 833ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 834ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 835ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 836ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 837ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 838ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 839ad4a5223SQuentin Monnet * 84032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 841ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 842ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 843ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 844ad4a5223SQuentin Monnet * direct packet access. 845ad4a5223SQuentin Monnet * Return 846ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 847ad4a5223SQuentin Monnet * 848bdb7b79bSAndrii Nakryiko * long bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 849ad4a5223SQuentin Monnet * Description 850ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 851ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 852ad4a5223SQuentin Monnet * must know the former value of the header field that was 853ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 854ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 855ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 856ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 857ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 858ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 859ad4a5223SQuentin Monnet * 860ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 861ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 862ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 863ad4a5223SQuentin Monnet * checksum to update. 864ad4a5223SQuentin Monnet * 86532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 866ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 867ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 868ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 869ad4a5223SQuentin Monnet * direct packet access. 870ad4a5223SQuentin Monnet * Return 871ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 872ad4a5223SQuentin Monnet * 873bdb7b79bSAndrii Nakryiko * long bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 874ad4a5223SQuentin Monnet * Description 875ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 876ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 877ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 878ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 879ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 880ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 881ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 882ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 883ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 884ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 885ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 886ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 887ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 888ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 889ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 890ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 891ad4a5223SQuentin Monnet * 892ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 893ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 894ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 895ad4a5223SQuentin Monnet * checksum to update. 896ad4a5223SQuentin Monnet * 89732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 898ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 899ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 900ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 901ad4a5223SQuentin Monnet * direct packet access. 902ad4a5223SQuentin Monnet * Return 903ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 904ad4a5223SQuentin Monnet * 905bdb7b79bSAndrii Nakryiko * long bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 906ad4a5223SQuentin Monnet * Description 907ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 908ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 909ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 910ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 911ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 912ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 913ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 914ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 915ad4a5223SQuentin Monnet * performed. 916ad4a5223SQuentin Monnet * 917ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 918ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 919ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 920ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 921ad4a5223SQuentin Monnet * 922ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 923ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 924ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 925ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 926ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 927ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 928ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 929ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 930ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 931ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 932ad4a5223SQuentin Monnet * which is currently set to 32. 933ad4a5223SQuentin Monnet * Return 934ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 935ad4a5223SQuentin Monnet * 936bdb7b79bSAndrii Nakryiko * long bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 937ad4a5223SQuentin Monnet * Description 938ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 939ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 940ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 941ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 942ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 943ad4a5223SQuentin Monnet * This is the only flag supported for now. 944ad4a5223SQuentin Monnet * 945ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 946ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 947ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 948ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 949ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 950ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 951ad4a5223SQuentin Monnet * 95232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 953ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 954ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 955ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 956ad4a5223SQuentin Monnet * direct packet access. 957ad4a5223SQuentin Monnet * Return 958ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 959c456dec4SQuentin Monnet * 960c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 961c456dec4SQuentin Monnet * Return 962c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 963c456dec4SQuentin Monnet * created as such: 964c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 965c456dec4SQuentin Monnet * *current_task*\ **->pid**. 966c456dec4SQuentin Monnet * 967c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 968c456dec4SQuentin Monnet * Return 969c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 970c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 971c456dec4SQuentin Monnet * 972bdb7b79bSAndrii Nakryiko * long bpf_get_current_comm(void *buf, u32 size_of_buf) 973c456dec4SQuentin Monnet * Description 974c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 975c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 976c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 977c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 978c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 979c456dec4SQuentin Monnet * it is filled with zeroes. 980c456dec4SQuentin Monnet * Return 981c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 982c456dec4SQuentin Monnet * 9831fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 9841fdd08beSQuentin Monnet * Description 9851fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 9861fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 9871fdd08beSQuentin Monnet * 9881fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 9891fdd08beSQuentin Monnet * 9901fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 9911fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 9921fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 9931fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 994da82c92fSMauro Carvalho Chehab * *Documentation/admin-guide/cgroup-v1/net_cls.rst*. 9951fdd08beSQuentin Monnet * 9961fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 9971fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 9981fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 9991fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 10001fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 10011fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 10021fdd08beSQuentin Monnet * 10031fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 10041fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 10051fdd08beSQuentin Monnet * "**y**" or to "**m**". 10061fdd08beSQuentin Monnet * Return 10071fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 10081fdd08beSQuentin Monnet * 1009bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 1010c456dec4SQuentin Monnet * Description 1011c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 1012c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 1013c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 1014c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 1015c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 1016c456dec4SQuentin Monnet * 101732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1018c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1019c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1020c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1021c456dec4SQuentin Monnet * direct packet access. 1022c456dec4SQuentin Monnet * Return 1023c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1024c456dec4SQuentin Monnet * 1025bdb7b79bSAndrii Nakryiko * long bpf_skb_vlan_pop(struct sk_buff *skb) 1026c456dec4SQuentin Monnet * Description 1027c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 1028c456dec4SQuentin Monnet * 102932e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1030c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1031c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 1032c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 1033c456dec4SQuentin Monnet * direct packet access. 1034c456dec4SQuentin Monnet * Return 1035c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1036c456dec4SQuentin Monnet * 1037bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1038c456dec4SQuentin Monnet * Description 1039c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 1040c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 1041c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 1042c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 1043c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 1044c456dec4SQuentin Monnet * IPv4. 1045c456dec4SQuentin Monnet * 1046c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 1047c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 1048c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 1049c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 1050c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 1051c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 1052c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 1053c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 1054c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 1055c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 1056c456dec4SQuentin Monnet * () helper. 1057c456dec4SQuentin Monnet * 1058c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 1059c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 1060c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 1061c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 1062c456dec4SQuentin Monnet * 1063c456dec4SQuentin Monnet * :: 1064c456dec4SQuentin Monnet * 1065c456dec4SQuentin Monnet * int ret; 1066c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 1067c456dec4SQuentin Monnet * 1068c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 1069c456dec4SQuentin Monnet * if (ret < 0) 1070c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1071c456dec4SQuentin Monnet * 1072c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 1073c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 1074c456dec4SQuentin Monnet * 1075c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 1076c456dec4SQuentin Monnet * 1077c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 1078c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 1079c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 1080c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 1081c456dec4SQuentin Monnet * configuration can be extracted from this helper. 1082c456dec4SQuentin Monnet * 1083c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 1084c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 1085c456dec4SQuentin Monnet * Return 1086c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1087c456dec4SQuentin Monnet * 1088bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 1089c456dec4SQuentin Monnet * Description 1090c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 1091c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 1092c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 1093c456dec4SQuentin Monnet * 1094c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 1095c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 1096c456dec4SQuentin Monnet * instead of IPv4. 1097c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 1098c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 1099c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 1100c456dec4SQuentin Monnet * and checksum set to zeroes. 1101c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 1102c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 1103c456dec4SQuentin Monnet * packet should not be fragmented. 1104c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 1105c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 1106c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 1107c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 1108c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 1109c456dec4SQuentin Monnet * as well in the future. 1110c456dec4SQuentin Monnet * 1111c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 1112c456dec4SQuentin Monnet * 1113c456dec4SQuentin Monnet * :: 1114c456dec4SQuentin Monnet * 1115c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 1116c456dec4SQuentin Monnet * populate key ... 1117c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 1118c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 1119c456dec4SQuentin Monnet * 1120c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 1121c456dec4SQuentin Monnet * helper for additional information. 1122c456dec4SQuentin Monnet * Return 1123c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1124c456dec4SQuentin Monnet * 1125c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 1126c6b5fb86SQuentin Monnet * Description 1127c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 1128c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 1129c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 1130c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 1131c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 1132c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 1133c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 1134c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 1135c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 1136c6b5fb86SQuentin Monnet * current CPU should be retrieved. 1137c6b5fb86SQuentin Monnet * 1138c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 1139c6b5fb86SQuentin Monnet * retrieved. 1140c6b5fb86SQuentin Monnet * 1141c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 1142c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 11433bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 1144c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 1145c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 11463bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 11473bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 11483bd5a09bSQuentin Monnet * () interface. Please refer to the description of 1149c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 1150c6b5fb86SQuentin Monnet * Return 1151c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 1152c6b5fb86SQuentin Monnet * negative error code in case of failure. 1153c6b5fb86SQuentin Monnet * 1154bdb7b79bSAndrii Nakryiko * long bpf_redirect(u32 ifindex, u64 flags) 1155c456dec4SQuentin Monnet * Description 1156c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 1157c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 1158c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 1159c456dec4SQuentin Monnet * increased performance. 1160c456dec4SQuentin Monnet * 1161c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 1162c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 1163c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 1164c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 1165c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 1166c456dec4SQuentin Monnet * flag at all. 1167c456dec4SQuentin Monnet * 1168f25975f4SToke Høiland-Jørgensen * The same effect can also be attained with the more generic 1169f25975f4SToke Høiland-Jørgensen * **bpf_redirect_map**\ (), which uses a BPF map to store the 1170f25975f4SToke Høiland-Jørgensen * redirect target instead of providing it directly to the helper. 1171c456dec4SQuentin Monnet * Return 1172c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 1173c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 1174c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 1175c456dec4SQuentin Monnet * error. 1176c456dec4SQuentin Monnet * 11771fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 11781fdd08beSQuentin Monnet * Description 11791fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 11801fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 1181b16fc097STobias Klauser * identifier retrieved is a user-provided tag, similar to the 11821fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 11831fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 11841fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 11851fdd08beSQuentin Monnet * 11861fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 11871fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 11881fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 11891fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 11901fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 11911fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 11921fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 11931fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 11941fdd08beSQuentin Monnet * 11951fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 11961fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 11971fdd08beSQuentin Monnet * Return 11981fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 11991fdd08beSQuentin Monnet * if none was found. 12001fdd08beSQuentin Monnet * 1201bdb7b79bSAndrii Nakryiko * long bpf_perf_event_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 1202c456dec4SQuentin Monnet * Description 1203c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 1204c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 1205c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 1206c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 1207c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 1208c456dec4SQuentin Monnet * 1209c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 1210c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 1211c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1212c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 1213c456dec4SQuentin Monnet * used. 1214c456dec4SQuentin Monnet * 1215c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 1216c456dec4SQuentin Monnet * pointed by *data*. 1217c456dec4SQuentin Monnet * 1218c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 1219c456dec4SQuentin Monnet * helper. 1220c456dec4SQuentin Monnet * 1221c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 1222c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 1223c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 1224c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 1225c456dec4SQuentin Monnet * into it. An example is available in file 1226c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 1227c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 1228c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 1229c456dec4SQuentin Monnet * 1230c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 1231c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 1232c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 1233c456dec4SQuentin Monnet * programs. 1234c456dec4SQuentin Monnet * 1235c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 1236c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 1237c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 1238c456dec4SQuentin Monnet * can be: 1239c456dec4SQuentin Monnet * 1240c456dec4SQuentin Monnet * * Only custom structs, 1241c456dec4SQuentin Monnet * * Only the packet payload, or 1242c456dec4SQuentin Monnet * * A combination of both. 1243c456dec4SQuentin Monnet * Return 1244c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1245c456dec4SQuentin Monnet * 1246bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes(const void *skb, u32 offset, void *to, u32 len) 12471fdd08beSQuentin Monnet * Description 12481fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 12491fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 12501fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 12511fdd08beSQuentin Monnet * *to*. 12521fdd08beSQuentin Monnet * 12531fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 12541fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 12551fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 12561fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 12571fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 12581fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 12591fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 12601fdd08beSQuentin Monnet * Return 12611fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 12621fdd08beSQuentin Monnet * 1263bdb7b79bSAndrii Nakryiko * long bpf_get_stackid(void *ctx, struct bpf_map *map, u64 flags) 1264c456dec4SQuentin Monnet * Description 1265c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 1266c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 1267c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 1268c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1269c456dec4SQuentin Monnet * 1270c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 1271c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 1272c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1273c456dec4SQuentin Monnet * a combination of the following flags: 1274c456dec4SQuentin Monnet * 1275c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 1276c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 1277c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 1278c456dec4SQuentin Monnet * Compare stacks by hash only. 1279c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 1280c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 1281c456dec4SQuentin Monnet * discard the old one. 1282c456dec4SQuentin Monnet * 1283c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 1284c456dec4SQuentin Monnet * can be further combined with other data (including other stack 1285c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 1286c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 1287c456dec4SQuentin Monnet * graphs). 1288c456dec4SQuentin Monnet * 1289c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 1290c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 1291c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 1292c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 1293c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1294c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 1295c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 1296c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 1297c456dec4SQuentin Monnet * 1298c456dec4SQuentin Monnet * :: 1299c456dec4SQuentin Monnet * 1300c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 1301c456dec4SQuentin Monnet * Return 1302c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 1303c456dec4SQuentin Monnet * in case of failure. 1304c456dec4SQuentin Monnet * 13051fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 13061fdd08beSQuentin Monnet * Description 13071fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 13081fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 13091fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 13101fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 13111fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 13121fdd08beSQuentin Monnet * to the helper). 13131fdd08beSQuentin Monnet * 13141fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 13151fdd08beSQuentin Monnet * 13161fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 13171fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 13181fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 13191fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 13201fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 13211fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 13221fdd08beSQuentin Monnet * *to_size* do not need to be equal. 13231fdd08beSQuentin Monnet * 13241fdd08beSQuentin Monnet * This helper can be used in combination with 13251fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 13261fdd08beSQuentin Monnet * which one can feed in the difference computed with 13271fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 13281fdd08beSQuentin Monnet * Return 13291fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 13301fdd08beSQuentin Monnet * failure. 13311fdd08beSQuentin Monnet * 1332bdb7b79bSAndrii Nakryiko * long bpf_skb_get_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 13331fdd08beSQuentin Monnet * Description 13341fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 13351fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 13361fdd08beSQuentin Monnet * of *size*. 13371fdd08beSQuentin Monnet * 13381fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 13391fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 13401fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 13411fdd08beSQuentin Monnet * more details). A particular example where this can be used is 13421fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 13431fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 13441fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 13451fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 13461fdd08beSQuentin Monnet * headers. 13471fdd08beSQuentin Monnet * Return 13481fdd08beSQuentin Monnet * The size of the option data retrieved. 13491fdd08beSQuentin Monnet * 1350bdb7b79bSAndrii Nakryiko * long bpf_skb_set_tunnel_opt(struct sk_buff *skb, void *opt, u32 size) 13511fdd08beSQuentin Monnet * Description 13521fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 13531fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 13541fdd08beSQuentin Monnet * 13551fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 13561fdd08beSQuentin Monnet * helper for additional information. 13571fdd08beSQuentin Monnet * Return 13581fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 13591fdd08beSQuentin Monnet * 1360bdb7b79bSAndrii Nakryiko * long bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 13611fdd08beSQuentin Monnet * Description 13621fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 13631fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 13641fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 13651fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 13661fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 13671fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 13681fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 13691fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 13701fdd08beSQuentin Monnet * operations out of an eBPF program. 13711fdd08beSQuentin Monnet * 13721fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 13731fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 13741fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 13751fdd08beSQuentin Monnet * 13761fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 13771fdd08beSQuentin Monnet * be left at zero. 13781fdd08beSQuentin Monnet * 137932e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 13801fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 13811fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 13821fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 13831fdd08beSQuentin Monnet * direct packet access. 13841fdd08beSQuentin Monnet * Return 13851fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 13861fdd08beSQuentin Monnet * 1387bdb7b79bSAndrii Nakryiko * long bpf_skb_change_type(struct sk_buff *skb, u32 type) 13881fdd08beSQuentin Monnet * Description 13891fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 13901fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 13911fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 13921fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 13931fdd08beSQuentin Monnet * for graceful handling of errors. 13941fdd08beSQuentin Monnet * 13951fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 13961fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 13971fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 13981fdd08beSQuentin Monnet * example. 13991fdd08beSQuentin Monnet * 14001fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 14011fdd08beSQuentin Monnet * are: 14021fdd08beSQuentin Monnet * 14031fdd08beSQuentin Monnet * **PACKET_HOST** 14041fdd08beSQuentin Monnet * Packet is for us. 14051fdd08beSQuentin Monnet * **PACKET_BROADCAST** 14061fdd08beSQuentin Monnet * Send packet to all. 14071fdd08beSQuentin Monnet * **PACKET_MULTICAST** 14081fdd08beSQuentin Monnet * Send packet to group. 14091fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 14101fdd08beSQuentin Monnet * Send packet to someone else. 14111fdd08beSQuentin Monnet * Return 14121fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 14131fdd08beSQuentin Monnet * 1414bdb7b79bSAndrii Nakryiko * long bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1415c6b5fb86SQuentin Monnet * Description 1416c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 1417c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1418c6b5fb86SQuentin Monnet * Return 1419c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 1420c6b5fb86SQuentin Monnet * 1421c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 1422c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 1423c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 1424c6b5fb86SQuentin Monnet * 1425fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1426fa15601aSQuentin Monnet * Description 1427fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1428fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 1429fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 1430fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 1431fa15601aSQuentin Monnet * 1432fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 1433fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 1434fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 1435fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1436fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 1437fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 1438fa15601aSQuentin Monnet * Return 1439fa15601aSQuentin Monnet * The 32-bit hash. 1440fa15601aSQuentin Monnet * 1441c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 1442c456dec4SQuentin Monnet * Return 1443c456dec4SQuentin Monnet * A pointer to the current task struct. 1444fa15601aSQuentin Monnet * 1445bdb7b79bSAndrii Nakryiko * long bpf_probe_write_user(void *dst, const void *src, u32 len) 1446c6b5fb86SQuentin Monnet * Description 1447c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 1448c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 1449c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 1450c6b5fb86SQuentin Monnet * 1451c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 1452c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 1453c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 1454c6b5fb86SQuentin Monnet * processes. 1455c6b5fb86SQuentin Monnet * 1456c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 1457c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 1458c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 1459c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 1460c6b5fb86SQuentin Monnet * logs. 1461c6b5fb86SQuentin Monnet * Return 1462c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1463c6b5fb86SQuentin Monnet * 1464bdb7b79bSAndrii Nakryiko * long bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1465c6b5fb86SQuentin Monnet * Description 1466c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 1467c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1468c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1469c6b5fb86SQuentin Monnet * Return 1470c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 1471c6b5fb86SQuentin Monnet * 14721aef5b43SSong Liu * * 0, if current task belongs to the cgroup2. 14731aef5b43SSong Liu * * 1, if current task does not belong to the cgroup2. 1474c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 1475c6b5fb86SQuentin Monnet * 1476bdb7b79bSAndrii Nakryiko * long bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1477fa15601aSQuentin Monnet * Description 1478fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 1479fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 1480fa15601aSQuentin Monnet * be left at zero. 1481fa15601aSQuentin Monnet * 1482fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 1483fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 1484fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 1485fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1486fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 1487fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 1488fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 1489fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 1490fa15601aSQuentin Monnet * *skb*. 1491fa15601aSQuentin Monnet * 149232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1493fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1494fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1495fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1496fa15601aSQuentin Monnet * direct packet access. 1497fa15601aSQuentin Monnet * Return 1498fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1499fa15601aSQuentin Monnet * 1500bdb7b79bSAndrii Nakryiko * long bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1501fa15601aSQuentin Monnet * Description 1502fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 1503fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 1504fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 1505fa15601aSQuentin Monnet * *len*, then the whole length of the *skb* is pulled. 1506fa15601aSQuentin Monnet * 1507fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 1508fa15601aSQuentin Monnet * packet access. 1509fa15601aSQuentin Monnet * 1510fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 1511fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 1512fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 1513fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 1514fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 1515fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 1516fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 1517fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 1518fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 1519fa15601aSQuentin Monnet * eventually access the data. 1520fa15601aSQuentin Monnet * 1521fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 1522fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 1523fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 1524fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 1525fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1526fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 1527fa15601aSQuentin Monnet * 152832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1529fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1530fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1531fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1532fa15601aSQuentin Monnet * direct packet access. 1533fa15601aSQuentin Monnet * Return 1534fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1535fa15601aSQuentin Monnet * 1536fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1537fa15601aSQuentin Monnet * Description 1538fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 1539fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 1540fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 1541fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 1542fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 1543fa15601aSQuentin Monnet * written into the packet through direct packet access. 1544fa15601aSQuentin Monnet * Return 1545fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 1546fa15601aSQuentin Monnet * failure. 1547fa15601aSQuentin Monnet * 1548fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 1549fa15601aSQuentin Monnet * Description 1550fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 1551fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 1552fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 1553fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 1554fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1555fa15601aSQuentin Monnet * 1556bdb7b79bSAndrii Nakryiko * long bpf_get_numa_node_id(void) 1557fa15601aSQuentin Monnet * Description 1558fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 1559fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 1560fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 1561fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1562fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 1563fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 1564fa15601aSQuentin Monnet * Return 1565fa15601aSQuentin Monnet * The id of current NUMA node. 1566fa15601aSQuentin Monnet * 1567bdb7b79bSAndrii Nakryiko * long bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1568c6b5fb86SQuentin Monnet * Description 1569c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 1570c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 1571c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 1572c6b5fb86SQuentin Monnet * required. 1573c6b5fb86SQuentin Monnet * 1574c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 1575c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 1576c6b5fb86SQuentin Monnet * 1577c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 1578c6b5fb86SQuentin Monnet * be left at zero. 1579c6b5fb86SQuentin Monnet * 158032e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1581c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1582c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 1583c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 1584c6b5fb86SQuentin Monnet * direct packet access. 1585c6b5fb86SQuentin Monnet * Return 1586c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1587c6b5fb86SQuentin Monnet * 1588bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1589c6b5fb86SQuentin Monnet * Description 1590c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1591c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 1592c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 1593c6b5fb86SQuentin Monnet * headers. 1594c6b5fb86SQuentin Monnet * 159532e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1596c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1597c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 1598c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 1599c6b5fb86SQuentin Monnet * direct packet access. 1600c6b5fb86SQuentin Monnet * Return 1601c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1602c6b5fb86SQuentin Monnet * 1603bdb7b79bSAndrii Nakryiko * long bpf_probe_read_str(void *dst, u32 size, const void *unsafe_ptr) 1604c6b5fb86SQuentin Monnet * Description 16056ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address 1606ab8d7809SQuentin Monnet * *unsafe_ptr* to *dst*. See **bpf_probe_read_kernel_str**\ () for 16076ae08ae3SDaniel Borkmann * more details. 1608c6b5fb86SQuentin Monnet * 1609ab8d7809SQuentin Monnet * Generally, use **bpf_probe_read_user_str**\ () or 1610ab8d7809SQuentin Monnet * **bpf_probe_read_kernel_str**\ () instead. 1611c6b5fb86SQuentin Monnet * Return 1612c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 1613c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 1614c6b5fb86SQuentin Monnet * value. 1615c6b5fb86SQuentin Monnet * 1616c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1617c6b5fb86SQuentin Monnet * Description 1618c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 1619c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 1620c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 1621c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 1622c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 1623cd48bddaSDaniel Borkmann * networking traffic statistics as it provides a global socket 1624cd48bddaSDaniel Borkmann * identifier that can be assumed unique. 1625c6b5fb86SQuentin Monnet * Return 1626c6b5fb86SQuentin Monnet * A 8-byte long non-decreasing number on success, or 0 if the 1627c6b5fb86SQuentin Monnet * socket field is missing inside *skb*. 1628c6b5fb86SQuentin Monnet * 1629d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1630d692f113SAndrey Ignatov * Description 1631d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 163262369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_addr** context. 1633d692f113SAndrey Ignatov * Return 1634d692f113SAndrey Ignatov * A 8-byte long non-decreasing number. 1635d692f113SAndrey Ignatov * 1636d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1637d692f113SAndrey Ignatov * Description 1638ab8d7809SQuentin Monnet * Equivalent to **bpf_get_socket_cookie**\ () helper that accepts 163962369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_ops** context. 1640d692f113SAndrey Ignatov * Return 1641d692f113SAndrey Ignatov * A 8-byte long non-decreasing number. 1642d692f113SAndrey Ignatov * 1643c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 1644c6b5fb86SQuentin Monnet * Return 1645c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 1646c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 1647c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 1648c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 1649c6b5fb86SQuentin Monnet * UID value for the socket). 1650c6b5fb86SQuentin Monnet * 1651bdb7b79bSAndrii Nakryiko * long bpf_set_hash(struct sk_buff *skb, u32 hash) 1652fa15601aSQuentin Monnet * Description 1653fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1654fa15601aSQuentin Monnet * to value *hash*. 1655fa15601aSQuentin Monnet * Return 1656fa15601aSQuentin Monnet * 0 1657fa15601aSQuentin Monnet * 1658bdb7b79bSAndrii Nakryiko * long bpf_setsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 16597aa79a86SQuentin Monnet * Description 16607aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 16617aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 16627aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 16637aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 16647aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 16657aa79a86SQuentin Monnet * 1666beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 1667ab8d7809SQuentin Monnet * 1668beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1669beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1670beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 1671beecf11bSStanislav Fomichev * 16727aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 16737aa79a86SQuentin Monnet * It supports the following *level*\ s: 16747aa79a86SQuentin Monnet * 16757aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 16767aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 1677f9bcf968SDmitry Yakunin * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**, 1678f9bcf968SDmitry Yakunin * **SO_BINDTODEVICE**, **SO_KEEPALIVE**. 16797aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 16807aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 1681f9bcf968SDmitry Yakunin * **TCP_BPF_SNDCWND_CLAMP**, **TCP_SAVE_SYN**, 1682f9bcf968SDmitry Yakunin * **TCP_KEEPIDLE**, **TCP_KEEPINTVL**, **TCP_KEEPCNT**, 1683f9bcf968SDmitry Yakunin * **TCP_SYNCNT**, **TCP_USER_TIMEOUT**. 16847aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 16857aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 16867aa79a86SQuentin Monnet * Return 16877aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 16887aa79a86SQuentin Monnet * 1689bdb7b79bSAndrii Nakryiko * long bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1690fa15601aSQuentin Monnet * Description 1691fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 1692fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 1693fa15601aSQuentin Monnet * 1694836e66c2SDaniel Borkmann * By default, the helper will reset any offloaded checksum 1695836e66c2SDaniel Borkmann * indicator of the skb to CHECKSUM_NONE. This can be avoided 1696836e66c2SDaniel Borkmann * by the following flag: 1697836e66c2SDaniel Borkmann * 1698836e66c2SDaniel Borkmann * * **BPF_F_ADJ_ROOM_NO_CSUM_RESET**: Do not reset offloaded 1699836e66c2SDaniel Borkmann * checksum data of the skb to CHECKSUM_NONE. 1700836e66c2SDaniel Borkmann * 170114aa3192SWillem de Bruijn * There are two supported modes at this time: 170214aa3192SWillem de Bruijn * 170314aa3192SWillem de Bruijn * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 170414aa3192SWillem de Bruijn * (room space is added or removed below the layer 2 header). 1705fa15601aSQuentin Monnet * 1706fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1707fa15601aSQuentin Monnet * (room space is added or removed below the layer 3 header). 1708fa15601aSQuentin Monnet * 1709868d5235SWillem de Bruijn * The following flags are supported at this time: 17102278f6ccSWillem de Bruijn * 17112278f6ccSWillem de Bruijn * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 17122278f6ccSWillem de Bruijn * Adjusting mss in this way is not allowed for datagrams. 1713fa15601aSQuentin Monnet * 171480867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4**, 171580867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6**: 1716868d5235SWillem de Bruijn * Any new space is reserved to hold a tunnel header. 1717868d5235SWillem de Bruijn * Configure skb offsets and other fields accordingly. 1718868d5235SWillem de Bruijn * 171980867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE**, 172080867c5eSQuentin Monnet * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP**: 1721868d5235SWillem de Bruijn * Use with ENCAP_L3 flags to further specify the tunnel type. 1722868d5235SWillem de Bruijn * 172380867c5eSQuentin Monnet * * **BPF_F_ADJ_ROOM_ENCAP_L2**\ (*len*): 172458dfc900SAlan Maguire * Use with ENCAP_L3/L4 flags to further specify the tunnel 172580867c5eSQuentin Monnet * type; *len* is the length of the inner MAC header. 172658dfc900SAlan Maguire * 172732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1728fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1729fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1730fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1731fa15601aSQuentin Monnet * direct packet access. 1732fa15601aSQuentin Monnet * Return 1733fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1734fa15601aSQuentin Monnet * 1735bdb7b79bSAndrii Nakryiko * long bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1736ab127040SQuentin Monnet * Description 1737ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 1738ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 1739ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 1740ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 1741ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 1742ab127040SQuentin Monnet * support) as of this writing). 1743ab127040SQuentin Monnet * 174443e74c02SToke Høiland-Jørgensen * The lower two bits of *flags* are used as the return code if 174543e74c02SToke Høiland-Jørgensen * the map lookup fails. This is so that the return value can be 1746ab8d7809SQuentin Monnet * one of the XDP program return codes up to **XDP_TX**, as chosen 1747ab8d7809SQuentin Monnet * by the caller. Any higher bits in the *flags* argument must be 174843e74c02SToke Høiland-Jørgensen * unset. 1749ab127040SQuentin Monnet * 1750ab8d7809SQuentin Monnet * See also **bpf_redirect**\ (), which only supports redirecting 1751ab8d7809SQuentin Monnet * to an ifindex, but doesn't require a map to do so. 1752ab127040SQuentin Monnet * Return 1753f25975f4SToke Høiland-Jørgensen * **XDP_REDIRECT** on success, or the value of the two lower bits 1754a33d3147SJakub Wilk * of the *flags* argument on error. 1755ab127040SQuentin Monnet * 1756bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_map(struct sk_buff *skb, struct bpf_map *map, u32 key, u64 flags) 1757ab127040SQuentin Monnet * Description 1758ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 1759ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1760ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 1761ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 1762ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 1763ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 1764ab127040SQuentin Monnet * Return 1765ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 1766ab127040SQuentin Monnet * 1767bdb7b79bSAndrii Nakryiko * long bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1768ab127040SQuentin Monnet * Description 1769ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 1770ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 1771ab127040SQuentin Monnet * *key*. *flags* is one of: 1772ab127040SQuentin Monnet * 1773ab127040SQuentin Monnet * **BPF_NOEXIST** 1774ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 1775ab127040SQuentin Monnet * **BPF_EXIST** 1776ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 1777ab127040SQuentin Monnet * **BPF_ANY** 1778ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 1779ab127040SQuentin Monnet * 1780ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 1781ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 1782ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 1783ab127040SQuentin Monnet * Return 1784ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 1785ab127040SQuentin Monnet * 1786bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1787fa15601aSQuentin Monnet * Description 1788fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1789fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 1790fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 1791fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 1792fa15601aSQuentin Monnet * called. 1793fa15601aSQuentin Monnet * 1794fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 1795fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 1796fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 1797fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 1798fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 1799fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 1800fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 1801fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 1802fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 1803fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 1804fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 1805fa15601aSQuentin Monnet * data they need. 1806fa15601aSQuentin Monnet * 180732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 1808fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1809fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1810fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1811fa15601aSQuentin Monnet * direct packet access. 1812fa15601aSQuentin Monnet * Return 1813fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 18147aa79a86SQuentin Monnet * 1815bdb7b79bSAndrii Nakryiko * long bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 18167aa79a86SQuentin Monnet * Description 18177aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 18187aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 18197aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 18207aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 18217aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 18227aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 18237aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 18247aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 18257aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 18267aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 18277aa79a86SQuentin Monnet * current CPU should be retrieved. 18287aa79a86SQuentin Monnet * 18297aa79a86SQuentin Monnet * This helper behaves in a way close to 18307aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 18317aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 18327aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 18337aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 18347aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 18357aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 18367aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 18377aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 18387aa79a86SQuentin Monnet * 18397aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 18407aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 18417aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 18427aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 18437aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 18447aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 18457aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 18467aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 18477aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 18487aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 18497aa79a86SQuentin Monnet * as follows. 18507aa79a86SQuentin Monnet * 18517aa79a86SQuentin Monnet * :: 18527aa79a86SQuentin Monnet * 18537aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 18547aa79a86SQuentin Monnet * 18557aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 18567aa79a86SQuentin Monnet * the time running for event since last normalization. The 18577aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 18587aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 1859ab8d7809SQuentin Monnet * eBPF program, users can use CPU id as the key (which is 18607aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 18617aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 18627aa79a86SQuentin Monnet * Return 18637aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 18647aa79a86SQuentin Monnet * 1865bdb7b79bSAndrii Nakryiko * long bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 18667aa79a86SQuentin Monnet * Description 18677aa79a86SQuentin Monnet * For en eBPF program attached to a perf event, retrieve the 18687aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 18697aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 18707aa79a86SQuentin Monnet * and running times are also stored in the structure (see 18717aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 18727aa79a86SQuentin Monnet * more details). 18737aa79a86SQuentin Monnet * Return 18747aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 18757aa79a86SQuentin Monnet * 1876bdb7b79bSAndrii Nakryiko * long bpf_getsockopt(void *bpf_socket, int level, int optname, void *optval, int optlen) 18777aa79a86SQuentin Monnet * Description 18787aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 18797aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 18807aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 18817aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 18827aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 18837aa79a86SQuentin Monnet * *opval* and of length *optlen*. 18847aa79a86SQuentin Monnet * 1885beecf11bSStanislav Fomichev * *bpf_socket* should be one of the following: 1886ab8d7809SQuentin Monnet * 1887beecf11bSStanislav Fomichev * * **struct bpf_sock_ops** for **BPF_PROG_TYPE_SOCK_OPS**. 1888beecf11bSStanislav Fomichev * * **struct bpf_sock_addr** for **BPF_CGROUP_INET4_CONNECT** 1889beecf11bSStanislav Fomichev * and **BPF_CGROUP_INET6_CONNECT**. 1890beecf11bSStanislav Fomichev * 18917aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 18927aa79a86SQuentin Monnet * It supports the following *level*\ s: 18937aa79a86SQuentin Monnet * 18947aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports *optname* 18957aa79a86SQuentin Monnet * **TCP_CONGESTION**. 18967aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 18977aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 18987aa79a86SQuentin Monnet * Return 18997aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 19007aa79a86SQuentin Monnet * 1901bdb7b79bSAndrii Nakryiko * long bpf_override_return(struct pt_regs *regs, u64 rc) 19027aa79a86SQuentin Monnet * Description 19037aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 19047aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 19057aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 19067aa79a86SQuentin Monnet * works. 19077aa79a86SQuentin Monnet * 1908ab8d7809SQuentin Monnet * This helper works by setting the PC (program counter) 19097aa79a86SQuentin Monnet * to an override function which is run in place of the original 19107aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 19117aa79a86SQuentin Monnet * all. The replacement function just returns with the required 19127aa79a86SQuentin Monnet * value. 19137aa79a86SQuentin Monnet * 19147aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 19157aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 19167aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 19177aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 19187aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 19197aa79a86SQuentin Monnet * 19207aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 19217aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 19227aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 19237aa79a86SQuentin Monnet * Return 19247aa79a86SQuentin Monnet * 0 19257aa79a86SQuentin Monnet * 1926bdb7b79bSAndrii Nakryiko * long bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 19277aa79a86SQuentin Monnet * Description 19287aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 19297aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 19307aa79a86SQuentin Monnet * *argval*. 19317aa79a86SQuentin Monnet * 19327aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 19337aa79a86SQuentin Monnet * be calls to eBPF programs of type 19347aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 19357aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 19367aa79a86SQuentin Monnet * connection and as necessary, when the connection is 19377aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 19387aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 19397aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 19407aa79a86SQuentin Monnet * supported in the current kernel. 19417aa79a86SQuentin Monnet * 1942725721a6SViet Hoang Tran * *argval* is a flag array which can combine these flags: 19437aa79a86SQuentin Monnet * 19447aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 19457aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 19467aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 194723729ff2SStanislav Fomichev * * **BPF_SOCK_OPS_RTT_CB_FLAG** (every RTT) 19487aa79a86SQuentin Monnet * 1949725721a6SViet Hoang Tran * Therefore, this function can be used to clear a callback flag by 1950725721a6SViet Hoang Tran * setting the appropriate bit to zero. e.g. to disable the RTO 1951725721a6SViet Hoang Tran * callback: 1952725721a6SViet Hoang Tran * 1953725721a6SViet Hoang Tran * **bpf_sock_ops_cb_flags_set(bpf_sock,** 1954725721a6SViet Hoang Tran * **bpf_sock->bpf_sock_ops_cb_flags & ~BPF_SOCK_OPS_RTO_CB_FLAG)** 1955725721a6SViet Hoang Tran * 19567aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 19577aa79a86SQuentin Monnet * program: 19587aa79a86SQuentin Monnet * 19597aa79a86SQuentin Monnet * * When RTO fires. 19607aa79a86SQuentin Monnet * * When a packet is retransmitted. 19617aa79a86SQuentin Monnet * * When the connection terminates. 19627aa79a86SQuentin Monnet * * When a packet is sent. 19637aa79a86SQuentin Monnet * * When a packet is received. 19647aa79a86SQuentin Monnet * Return 19657aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 19667aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 19677aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 19687aa79a86SQuentin Monnet * as required). 19697aa79a86SQuentin Monnet * 1970bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1971ab127040SQuentin Monnet * Description 1972ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 1973ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 1974ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 1975ab127040SQuentin Monnet * the socket referenced by *map* (of type 1976ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1977ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 1978ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 1979ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 1980ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 1981ab127040SQuentin Monnet * Return 1982ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 1983ab127040SQuentin Monnet * 1984bdb7b79bSAndrii Nakryiko * long bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1985ab127040SQuentin Monnet * Description 1986ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 1987ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 1988ab127040SQuentin Monnet * 1989ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 1990ab127040SQuentin Monnet * 1991ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 1992ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 1993ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 1994ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 1995ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 1996ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 1997ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 1998ab127040SQuentin Monnet * overhead. 1999ab127040SQuentin Monnet * 2000ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 2001ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 2002ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 2003ab127040SQuentin Monnet * smaller than the current data being processed from a 2004ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 2005ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 2006ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 2007ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 2008ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 2009ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 2010ab127040SQuentin Monnet * consumed. 2011ab127040SQuentin Monnet * 2012ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 2013ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 2014ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 2015ab127040SQuentin Monnet * Return 2016ab127040SQuentin Monnet * 0 2017ab127040SQuentin Monnet * 2018bdb7b79bSAndrii Nakryiko * long bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 2019ab127040SQuentin Monnet * Description 2020ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 2021ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 2022ab127040SQuentin Monnet * accumulated. 2023ab127040SQuentin Monnet * 2024ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 2025ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 2026ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 2027ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 2028ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 2029ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 2030ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 2031ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 2032ab127040SQuentin Monnet * been accumulated. 2033ab127040SQuentin Monnet * Return 2034ab127040SQuentin Monnet * 0 2035ab127040SQuentin Monnet * 2036bdb7b79bSAndrii Nakryiko * long bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 2037ab127040SQuentin Monnet * Description 2038ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 2039ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 2040ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 2041ab127040SQuentin Monnet * respectively. 2042ab127040SQuentin Monnet * 2043ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 2044ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 2045ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 2046ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 2047ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 2048ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 2049ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 2050ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 2051ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 2052ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 2053ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 2054ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 2055ab127040SQuentin Monnet * 205632e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2057ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 2058ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 2059ab127040SQuentin Monnet * performed again, if the helper is used in combination with 2060ab127040SQuentin Monnet * direct packet access. 2061ab127040SQuentin Monnet * 2062ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 2063ab127040SQuentin Monnet * be left at zero. 2064ab127040SQuentin Monnet * Return 2065ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 2066ab127040SQuentin Monnet * 2067bdb7b79bSAndrii Nakryiko * long bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 20687aa79a86SQuentin Monnet * Description 20697aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 20707aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 20717aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 20727aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 20737aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 20747aa79a86SQuentin Monnet * 20757aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 20767aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 20778086fbafSStanislav Fomichev * **AF_INET6**). It's advised to pass zero port (**sin_port** 20788086fbafSStanislav Fomichev * or **sin6_port**) which triggers IP_BIND_ADDRESS_NO_PORT-like 20798086fbafSStanislav Fomichev * behavior and lets the kernel efficiently pick up an unused 20808086fbafSStanislav Fomichev * port as long as 4-tuple is unique. Passing non-zero port might 20818086fbafSStanislav Fomichev * lead to degraded performance. 20827aa79a86SQuentin Monnet * Return 20837aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 20842d020dd7SQuentin Monnet * 2085bdb7b79bSAndrii Nakryiko * long bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 20862d020dd7SQuentin Monnet * Description 20872d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 2088c8741e2bSJesper Dangaard Brouer * possible to both shrink and grow the packet tail. 2089c8741e2bSJesper Dangaard Brouer * Shrink done via *delta* being a negative integer. 20902d020dd7SQuentin Monnet * 209132e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 20922d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 20932d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 20942d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 20952d020dd7SQuentin Monnet * direct packet access. 20962d020dd7SQuentin Monnet * Return 20972d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 20982d020dd7SQuentin Monnet * 2099bdb7b79bSAndrii Nakryiko * long bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 21002d020dd7SQuentin Monnet * Description 21012d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 21022d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 21032d020dd7SQuentin Monnet * 21042d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 21052d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 21062d020dd7SQuentin Monnet * 21072d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 21082d020dd7SQuentin Monnet * be left at zero. 21092d020dd7SQuentin Monnet * 21102d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 21112d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 21122d020dd7SQuentin Monnet * Return 21132d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 2114c195651eSYonghong Song * 2115bdb7b79bSAndrii Nakryiko * long bpf_get_stack(void *ctx, void *buf, u32 size, u64 flags) 2116c195651eSYonghong Song * Description 2117c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 2118c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 2119c195651eSYonghong Song * to the context on which the tracing program is executed. 2120c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 2121c195651eSYonghong Song * a nonnegative *size*. 2122c195651eSYonghong Song * 2123c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 2124c195651eSYonghong Song * skip (from 0 to 255), masked with 2125c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 2126c195651eSYonghong Song * the following flags: 2127c195651eSYonghong Song * 2128c195651eSYonghong Song * **BPF_F_USER_STACK** 2129c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 2130c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 2131c195651eSYonghong Song * Collect buildid+offset instead of ips for user stack, 2132c195651eSYonghong Song * only valid if **BPF_F_USER_STACK** is also specified. 2133c195651eSYonghong Song * 2134c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 2135c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 2136c195651eSYonghong Song * to sufficient large buffer size. Note that 2137c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 2138c195651eSYonghong Song * that it should be manually increased in order to profile long 2139c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 2140c195651eSYonghong Song * 2141c195651eSYonghong Song * :: 2142c195651eSYonghong Song * 2143c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 2144c195651eSYonghong Song * Return 21457a279e93SQuentin Monnet * A non-negative value equal to or less than *size* on success, 21467a279e93SQuentin Monnet * or a negative error in case of failure. 21474e1ec56cSDaniel Borkmann * 2148bdb7b79bSAndrii Nakryiko * long bpf_skb_load_bytes_relative(const void *skb, u32 offset, void *to, u32 len, u32 start_header) 21494e1ec56cSDaniel Borkmann * Description 21504e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 21514e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 21524e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 21534e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 21544e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 21554e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 21564e1ec56cSDaniel Borkmann * 21574e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 21584e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 21594e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 21604e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 21614e1ec56cSDaniel Borkmann * 21624e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 21634e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 21644e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 21654e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 21664e1ec56cSDaniel Borkmann * is not available. 21674e1ec56cSDaniel Borkmann * Return 21684e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 21694e1ec56cSDaniel Borkmann * 2170bdb7b79bSAndrii Nakryiko * long bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 217187f5fc7eSDavid Ahern * Description 217287f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 217387f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 217487f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 217587f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 2176fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 2177fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 2178fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 21794c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 21804c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 218187f5fc7eSDavid Ahern * 218287f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 21837a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 21847a279e93SQuentin Monnet * following values: 218587f5fc7eSDavid Ahern * 21867a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 21877a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 21887a279e93SQuentin Monnet * rules. 21897a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 21907a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 21917a279e93SQuentin Monnet * ingress). 219287f5fc7eSDavid Ahern * 219387f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 219487f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 219587f5fc7eSDavid Ahern * Return 21964c79579bSDavid Ahern * * < 0 if any input argument is invalid 21974c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 21984c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 21992bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 220081110384SJohn Fastabend * 2201bdb7b79bSAndrii Nakryiko * long bpf_sock_hash_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 220281110384SJohn Fastabend * Description 220381110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 220481110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 220581110384SJohn Fastabend * *key*. *flags* is one of: 220681110384SJohn Fastabend * 220781110384SJohn Fastabend * **BPF_NOEXIST** 220881110384SJohn Fastabend * The entry for *key* must not exist in the map. 220981110384SJohn Fastabend * **BPF_EXIST** 221081110384SJohn Fastabend * The entry for *key* must already exist in the map. 221181110384SJohn Fastabend * **BPF_ANY** 221281110384SJohn Fastabend * No condition on the existence of the entry for *key*. 221381110384SJohn Fastabend * 221481110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 221581110384SJohn Fastabend * be inherited by the socket being added. If the socket is 221681110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 221781110384SJohn Fastabend * Return 221881110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 221981110384SJohn Fastabend * 2220bdb7b79bSAndrii Nakryiko * long bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 222181110384SJohn Fastabend * Description 222281110384SJohn Fastabend * This helper is used in programs implementing policies at the 222381110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 222481110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 222581110384SJohn Fastabend * the socket referenced by *map* (of type 222681110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 222781110384SJohn Fastabend * egress interfaces can be used for redirection. The 222881110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 222981110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 223081110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 223181110384SJohn Fastabend * Return 223281110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 223381110384SJohn Fastabend * 2234bdb7b79bSAndrii Nakryiko * long bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 223581110384SJohn Fastabend * Description 223681110384SJohn Fastabend * This helper is used in programs implementing policies at the 223781110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 223881110384SJohn Fastabend * if the verdeict eBPF program returns **SK_PASS**), redirect it 223981110384SJohn Fastabend * to the socket referenced by *map* (of type 224081110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 224181110384SJohn Fastabend * egress interfaces can be used for redirection. The 224281110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 224381110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 224481110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 224581110384SJohn Fastabend * Return 224681110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 2247fe94cc29SMathieu Xhonneux * 2248bdb7b79bSAndrii Nakryiko * long bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2249fe94cc29SMathieu Xhonneux * Description 2250fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 2251fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 2252fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 2253fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 2254fe94cc29SMathieu Xhonneux * 2255fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 2256fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 2257fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2258fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 2259fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 2260fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 2261fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2262fe94cc29SMathieu Xhonneux * the IPv6 header. 22633e0bd37cSPeter Oskolkov * **BPF_LWT_ENCAP_IP** 22643e0bd37cSPeter Oskolkov * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 22653e0bd37cSPeter Oskolkov * must be IPv4 or IPv6, followed by zero or more 226680867c5eSQuentin Monnet * additional headers, up to **LWT_BPF_MAX_HEADROOM** 226780867c5eSQuentin Monnet * total bytes in all prepended headers. Please note that 226880867c5eSQuentin Monnet * if **skb_is_gso**\ (*skb*) is true, no more than two 226980867c5eSQuentin Monnet * headers can be prepended, and the inner header, if 227080867c5eSQuentin Monnet * present, should be either GRE or UDP/GUE. 22713e0bd37cSPeter Oskolkov * 227280867c5eSQuentin Monnet * **BPF_LWT_ENCAP_SEG6**\ \* types can be called by BPF programs 227380867c5eSQuentin Monnet * of type **BPF_PROG_TYPE_LWT_IN**; **BPF_LWT_ENCAP_IP** type can 227480867c5eSQuentin Monnet * be called by bpf programs of types **BPF_PROG_TYPE_LWT_IN** and 227580867c5eSQuentin Monnet * **BPF_PROG_TYPE_LWT_XMIT**. 2276fe94cc29SMathieu Xhonneux * 227732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2278fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2279fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2280fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2281fe94cc29SMathieu Xhonneux * direct packet access. 2282fe94cc29SMathieu Xhonneux * Return 2283fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2284fe94cc29SMathieu Xhonneux * 2285bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2286fe94cc29SMathieu Xhonneux * Description 2287fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 2288fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2289fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 2290fe94cc29SMathieu Xhonneux * modified through this helper. 2291fe94cc29SMathieu Xhonneux * 229232e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2293fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2294fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2295fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2296fe94cc29SMathieu Xhonneux * direct packet access. 2297fe94cc29SMathieu Xhonneux * Return 2298fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2299fe94cc29SMathieu Xhonneux * 2300bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2301fe94cc29SMathieu Xhonneux * Description 2302fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 2303fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 2304fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 2305fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 2306fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 2307fe94cc29SMathieu Xhonneux * 230832e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2309fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2310fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2311fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2312fe94cc29SMathieu Xhonneux * direct packet access. 2313fe94cc29SMathieu Xhonneux * Return 2314fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2315fe94cc29SMathieu Xhonneux * 2316bdb7b79bSAndrii Nakryiko * long bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2317fe94cc29SMathieu Xhonneux * Description 2318fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 2319fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 2320fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 2321fe94cc29SMathieu Xhonneux * *action* can be one of: 2322fe94cc29SMathieu Xhonneux * 2323fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 2324fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 2325fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 2326fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 2327fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 2328fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 2329fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 2330fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 233180867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 2332fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2333fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 2334fe94cc29SMathieu Xhonneux * encapsulation policy. 233580867c5eSQuentin Monnet * Type of *param*: **struct ipv6_sr_hdr**. 2336fe94cc29SMathieu Xhonneux * 233732e7dc28SQuentin Monnet * A call to this helper is susceptible to change the underlying 2338fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2339fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2340fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2341fe94cc29SMathieu Xhonneux * direct packet access. 2342fe94cc29SMathieu Xhonneux * Return 2343fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2344f4364dcfSSean Young * 2345bdb7b79bSAndrii Nakryiko * long bpf_rc_repeat(void *ctx) 234662369db2SQuentin Monnet * Description 234762369db2SQuentin Monnet * This helper is used in programs implementing IR decoding, to 234862369db2SQuentin Monnet * report a successfully decoded repeat key message. This delays 234962369db2SQuentin Monnet * the generation of a key up event for previously generated 235062369db2SQuentin Monnet * key down event. 235162369db2SQuentin Monnet * 235262369db2SQuentin Monnet * Some IR protocols like NEC have a special IR message for 235362369db2SQuentin Monnet * repeating last button, for when a button is held down. 235462369db2SQuentin Monnet * 235562369db2SQuentin Monnet * The *ctx* should point to the lirc sample as passed into 235662369db2SQuentin Monnet * the program. 235762369db2SQuentin Monnet * 235862369db2SQuentin Monnet * This helper is only available is the kernel was compiled with 235962369db2SQuentin Monnet * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 236062369db2SQuentin Monnet * "**y**". 236162369db2SQuentin Monnet * Return 236262369db2SQuentin Monnet * 0 236362369db2SQuentin Monnet * 2364bdb7b79bSAndrii Nakryiko * long bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2365f4364dcfSSean Young * Description 2366f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 2367f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 2368f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 2369f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 2370f4364dcfSSean Young * an input key down event. After a period a key up event is 2371f4364dcfSSean Young * generated. This period can be extended by calling either 237290b1023fSQuentin Monnet * **bpf_rc_keydown**\ () again with the same values, or calling 237390b1023fSQuentin Monnet * **bpf_rc_repeat**\ (). 2374f4364dcfSSean Young * 2375f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 2376f4364dcfSSean Young * released and pressed again between consecutive scancodes. 2377f4364dcfSSean Young * 2378f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 2379f4364dcfSSean Young * the program. 2380f4364dcfSSean Young * 2381f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 2382f4364dcfSSean Young * **enum rc_proto** for some predefined values). 2383f4364dcfSSean Young * 2384f4364dcfSSean Young * This helper is only available is the kernel was compiled with 2385f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2386f4364dcfSSean Young * "**y**". 2387f4364dcfSSean Young * Return 2388f4364dcfSSean Young * 0 2389f4364dcfSSean Young * 239062369db2SQuentin Monnet * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2391cb20b08eSDaniel Borkmann * Description 2392cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 2393cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2394cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 2395cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 2396cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 2397cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 2398cb20b08eSDaniel Borkmann * to the same 64-bit id. 2399cb20b08eSDaniel Borkmann * 2400cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 2401cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 2402cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2403cb20b08eSDaniel Borkmann * Return 2404cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 2405bf6fa2c8SYonghong Song * 2406bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 2407bf6fa2c8SYonghong Song * Return 2408bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 2409bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 2410cd339431SRoman Gushchin * 241162369db2SQuentin Monnet * void *bpf_get_local_storage(void *map, u64 flags) 2412cd339431SRoman Gushchin * Description 2413cd339431SRoman Gushchin * Get the pointer to the local storage area. 2414cd339431SRoman Gushchin * The type and the size of the local storage is defined 2415cd339431SRoman Gushchin * by the *map* argument. 2416cd339431SRoman Gushchin * The *flags* meaning is specific for each map type, 2417cd339431SRoman Gushchin * and has to be 0 for cgroup local storage. 2418cd339431SRoman Gushchin * 241990b1023fSQuentin Monnet * Depending on the BPF program type, a local storage area 242090b1023fSQuentin Monnet * can be shared between multiple instances of the BPF program, 2421cd339431SRoman Gushchin * running simultaneously. 2422cd339431SRoman Gushchin * 2423cd339431SRoman Gushchin * A user should care about the synchronization by himself. 242490b1023fSQuentin Monnet * For example, by using the **BPF_STX_XADD** instruction to alter 2425cd339431SRoman Gushchin * the shared data. 2426cd339431SRoman Gushchin * Return 242790b1023fSQuentin Monnet * A pointer to the local storage area. 24282dbb9b9eSMartin KaFai Lau * 2429bdb7b79bSAndrii Nakryiko * long bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 24302dbb9b9eSMartin KaFai Lau * Description 243190b1023fSQuentin Monnet * Select a **SO_REUSEPORT** socket from a 243290b1023fSQuentin Monnet * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 243390b1023fSQuentin Monnet * It checks the selected socket is matching the incoming 243490b1023fSQuentin Monnet * request in the socket buffer. 24352dbb9b9eSMartin KaFai Lau * Return 24362dbb9b9eSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 24376acc9b43SJoe Stringer * 243862369db2SQuentin Monnet * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 243962369db2SQuentin Monnet * Description 244062369db2SQuentin Monnet * Return id of cgroup v2 that is ancestor of cgroup associated 244162369db2SQuentin Monnet * with the *skb* at the *ancestor_level*. The root cgroup is at 244262369db2SQuentin Monnet * *ancestor_level* zero and each step down the hierarchy 244362369db2SQuentin Monnet * increments the level. If *ancestor_level* == level of cgroup 244462369db2SQuentin Monnet * associated with *skb*, then return value will be same as that 244562369db2SQuentin Monnet * of **bpf_skb_cgroup_id**\ (). 244662369db2SQuentin Monnet * 244762369db2SQuentin Monnet * The helper is useful to implement policies based on cgroups 244862369db2SQuentin Monnet * that are upper in hierarchy than immediate cgroup associated 244962369db2SQuentin Monnet * with *skb*. 245062369db2SQuentin Monnet * 245162369db2SQuentin Monnet * The format of returned id and helper limitations are same as in 245262369db2SQuentin Monnet * **bpf_skb_cgroup_id**\ (). 245362369db2SQuentin Monnet * Return 245462369db2SQuentin Monnet * The id is returned or 0 in case the id could not be retrieved. 245562369db2SQuentin Monnet * 2456f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 24576acc9b43SJoe Stringer * Description 24586acc9b43SJoe Stringer * Look for TCP socket matching *tuple*, optionally in a child 24596acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 246090b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 24616acc9b43SJoe Stringer * 24626acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 24636acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 24646acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 24656acc9b43SJoe Stringer * 24666acc9b43SJoe Stringer * *tuple_size* must be one of: 24676acc9b43SJoe Stringer * 24686acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 24696acc9b43SJoe Stringer * Look for an IPv4 socket. 24706acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 24716acc9b43SJoe Stringer * Look for an IPv6 socket. 24726acc9b43SJoe Stringer * 2473f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 2474bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 2475f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 2476f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 2477f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 2478f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 2479f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 2480f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 24816acc9b43SJoe Stringer * 24826acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 24836acc9b43SJoe Stringer * be left at zero. 24846acc9b43SJoe Stringer * 24856acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 24866acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 24876acc9b43SJoe Stringer * Return 24880bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 24890bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 249080867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 249180867c5eSQuentin Monnet * tuple. 24926acc9b43SJoe Stringer * 2493f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 24946acc9b43SJoe Stringer * Description 24956acc9b43SJoe Stringer * Look for UDP socket matching *tuple*, optionally in a child 24966acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 249790b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 24986acc9b43SJoe Stringer * 24996acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 25006acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 25016acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 25026acc9b43SJoe Stringer * 25036acc9b43SJoe Stringer * *tuple_size* must be one of: 25046acc9b43SJoe Stringer * 25056acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 25066acc9b43SJoe Stringer * Look for an IPv4 socket. 25076acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 25086acc9b43SJoe Stringer * Look for an IPv6 socket. 25096acc9b43SJoe Stringer * 2510f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 2511bfdfa517SRandy Dunlap * socket lookup table in the netns associated with the *ctx* 2512f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 2513f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 2514f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 2515f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 2516f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 2517f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 25186acc9b43SJoe Stringer * 25196acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 25206acc9b43SJoe Stringer * be left at zero. 25216acc9b43SJoe Stringer * 25226acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 25236acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 25246acc9b43SJoe Stringer * Return 25250bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 25260bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 252780867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 252880867c5eSQuentin Monnet * tuple. 25296acc9b43SJoe Stringer * 2530a5fa25adSMartin KaFai Lau * long bpf_sk_release(void *sock) 25316acc9b43SJoe Stringer * Description 253290b1023fSQuentin Monnet * Release the reference held by *sock*. *sock* must be a 253390b1023fSQuentin Monnet * non-**NULL** pointer that was returned from 253490b1023fSQuentin Monnet * **bpf_sk_lookup_xxx**\ (). 253590b1023fSQuentin Monnet * Return 253690b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 253790b1023fSQuentin Monnet * 2538bdb7b79bSAndrii Nakryiko * long bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 253962369db2SQuentin Monnet * Description 254062369db2SQuentin Monnet * Push an element *value* in *map*. *flags* is one of: 254162369db2SQuentin Monnet * 254262369db2SQuentin Monnet * **BPF_EXIST** 254362369db2SQuentin Monnet * If the queue/stack is full, the oldest element is 254462369db2SQuentin Monnet * removed to make room for this. 254562369db2SQuentin Monnet * Return 254662369db2SQuentin Monnet * 0 on success, or a negative error in case of failure. 254762369db2SQuentin Monnet * 2548bdb7b79bSAndrii Nakryiko * long bpf_map_pop_elem(struct bpf_map *map, void *value) 254990b1023fSQuentin Monnet * Description 255090b1023fSQuentin Monnet * Pop an element from *map*. 255190b1023fSQuentin Monnet * Return 255290b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 255390b1023fSQuentin Monnet * 2554bdb7b79bSAndrii Nakryiko * long bpf_map_peek_elem(struct bpf_map *map, void *value) 255590b1023fSQuentin Monnet * Description 255690b1023fSQuentin Monnet * Get an element from *map* without removing it. 25576acc9b43SJoe Stringer * Return 25586acc9b43SJoe Stringer * 0 on success, or a negative error in case of failure. 25596fff607eSJohn Fastabend * 2560bdb7b79bSAndrii Nakryiko * long bpf_msg_push_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 25616fff607eSJohn Fastabend * Description 256290b1023fSQuentin Monnet * For socket policies, insert *len* bytes into *msg* at offset 25636fff607eSJohn Fastabend * *start*. 25646fff607eSJohn Fastabend * 25656fff607eSJohn Fastabend * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 256690b1023fSQuentin Monnet * *msg* it may want to insert metadata or options into the *msg*. 25676fff607eSJohn Fastabend * This can later be read and used by any of the lower layer BPF 25686fff607eSJohn Fastabend * hooks. 25696fff607eSJohn Fastabend * 25706fff607eSJohn Fastabend * This helper may fail if under memory pressure (a malloc 25716fff607eSJohn Fastabend * fails) in these cases BPF programs will get an appropriate 25726fff607eSJohn Fastabend * error and BPF programs will need to handle them. 25736fff607eSJohn Fastabend * Return 25746fff607eSJohn Fastabend * 0 on success, or a negative error in case of failure. 25757246d8edSJohn Fastabend * 2576bdb7b79bSAndrii Nakryiko * long bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 len, u64 flags) 25777246d8edSJohn Fastabend * Description 25785f0e5412SAndrii Nakryiko * Will remove *len* bytes from a *msg* starting at byte *start*. 25797246d8edSJohn Fastabend * This may result in **ENOMEM** errors under certain situations if 25807246d8edSJohn Fastabend * an allocation and copy are required due to a full ring buffer. 25817246d8edSJohn Fastabend * However, the helper will try to avoid doing the allocation 25827246d8edSJohn Fastabend * if possible. Other errors can occur if input parameters are 258390b1023fSQuentin Monnet * invalid either due to *start* byte not being valid part of *msg* 25847246d8edSJohn Fastabend * payload and/or *pop* value being to large. 25857246d8edSJohn Fastabend * Return 258690b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 258701d3240aSSean Young * 2588bdb7b79bSAndrii Nakryiko * long bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 258901d3240aSSean Young * Description 259001d3240aSSean Young * This helper is used in programs implementing IR decoding, to 259101d3240aSSean Young * report a successfully decoded pointer movement. 259201d3240aSSean Young * 259301d3240aSSean Young * The *ctx* should point to the lirc sample as passed into 259401d3240aSSean Young * the program. 259501d3240aSSean Young * 259601d3240aSSean Young * This helper is only available is the kernel was compiled with 259701d3240aSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 259801d3240aSSean Young * "**y**". 259901d3240aSSean Young * Return 260001d3240aSSean Young * 0 260146f8bc92SMartin KaFai Lau * 2602bdb7b79bSAndrii Nakryiko * long bpf_spin_lock(struct bpf_spin_lock *lock) 26030eb09785SQuentin Monnet * Description 26040eb09785SQuentin Monnet * Acquire a spinlock represented by the pointer *lock*, which is 26050eb09785SQuentin Monnet * stored as part of a value of a map. Taking the lock allows to 26060eb09785SQuentin Monnet * safely update the rest of the fields in that value. The 26070eb09785SQuentin Monnet * spinlock can (and must) later be released with a call to 26080eb09785SQuentin Monnet * **bpf_spin_unlock**\ (\ *lock*\ ). 26090eb09785SQuentin Monnet * 26100eb09785SQuentin Monnet * Spinlocks in BPF programs come with a number of restrictions 26110eb09785SQuentin Monnet * and constraints: 26120eb09785SQuentin Monnet * 26130eb09785SQuentin Monnet * * **bpf_spin_lock** objects are only allowed inside maps of 26140eb09785SQuentin Monnet * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 26150eb09785SQuentin Monnet * list could be extended in the future). 26160eb09785SQuentin Monnet * * BTF description of the map is mandatory. 26170eb09785SQuentin Monnet * * The BPF program can take ONE lock at a time, since taking two 26180eb09785SQuentin Monnet * or more could cause dead locks. 26190eb09785SQuentin Monnet * * Only one **struct bpf_spin_lock** is allowed per map element. 26200eb09785SQuentin Monnet * * When the lock is taken, calls (either BPF to BPF or helpers) 26210eb09785SQuentin Monnet * are not allowed. 26220eb09785SQuentin Monnet * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 26230eb09785SQuentin Monnet * allowed inside a spinlock-ed region. 26240eb09785SQuentin Monnet * * The BPF program MUST call **bpf_spin_unlock**\ () to release 26250eb09785SQuentin Monnet * the lock, on all execution paths, before it returns. 26260eb09785SQuentin Monnet * * The BPF program can access **struct bpf_spin_lock** only via 26270eb09785SQuentin Monnet * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 26280eb09785SQuentin Monnet * helpers. Loading or storing data into the **struct 26290eb09785SQuentin Monnet * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 26300eb09785SQuentin Monnet * * To use the **bpf_spin_lock**\ () helper, the BTF description 26310eb09785SQuentin Monnet * of the map value must be a struct and have **struct 26320eb09785SQuentin Monnet * bpf_spin_lock** *anyname*\ **;** field at the top level. 26330eb09785SQuentin Monnet * Nested lock inside another struct is not allowed. 26340eb09785SQuentin Monnet * * The **struct bpf_spin_lock** *lock* field in a map value must 26350eb09785SQuentin Monnet * be aligned on a multiple of 4 bytes in that value. 26360eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 26370eb09785SQuentin Monnet * the **bpf_spin_lock** field to user space. 26380eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 26390eb09785SQuentin Monnet * a BPF program, do not update the **bpf_spin_lock** field. 26400eb09785SQuentin Monnet * * **bpf_spin_lock** cannot be on the stack or inside a 26410eb09785SQuentin Monnet * networking packet (it can only be inside of a map values). 26420eb09785SQuentin Monnet * * **bpf_spin_lock** is available to root only. 26430eb09785SQuentin Monnet * * Tracing programs and socket filter programs cannot use 26440eb09785SQuentin Monnet * **bpf_spin_lock**\ () due to insufficient preemption checks 26450eb09785SQuentin Monnet * (but this may change in the future). 26460eb09785SQuentin Monnet * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 26470eb09785SQuentin Monnet * Return 26480eb09785SQuentin Monnet * 0 26490eb09785SQuentin Monnet * 2650bdb7b79bSAndrii Nakryiko * long bpf_spin_unlock(struct bpf_spin_lock *lock) 26510eb09785SQuentin Monnet * Description 26520eb09785SQuentin Monnet * Release the *lock* previously locked by a call to 26530eb09785SQuentin Monnet * **bpf_spin_lock**\ (\ *lock*\ ). 26540eb09785SQuentin Monnet * Return 26550eb09785SQuentin Monnet * 0 26560eb09785SQuentin Monnet * 265746f8bc92SMartin KaFai Lau * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 265846f8bc92SMartin KaFai Lau * Description 265946f8bc92SMartin KaFai Lau * This helper gets a **struct bpf_sock** pointer such 266062369db2SQuentin Monnet * that all the fields in this **bpf_sock** can be accessed. 266146f8bc92SMartin KaFai Lau * Return 266262369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 266346f8bc92SMartin KaFai Lau * case of failure. 2664655a51e5SMartin KaFai Lau * 2665655a51e5SMartin KaFai Lau * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2666655a51e5SMartin KaFai Lau * Description 2667655a51e5SMartin KaFai Lau * This helper gets a **struct bpf_tcp_sock** pointer from a 2668655a51e5SMartin KaFai Lau * **struct bpf_sock** pointer. 2669655a51e5SMartin KaFai Lau * Return 267062369db2SQuentin Monnet * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2671655a51e5SMartin KaFai Lau * case of failure. 2672f7c917baSbrakmo * 2673bdb7b79bSAndrii Nakryiko * long bpf_skb_ecn_set_ce(struct sk_buff *skb) 2674f7c917baSbrakmo * Description 267562369db2SQuentin Monnet * Set ECN (Explicit Congestion Notification) field of IP header 267662369db2SQuentin Monnet * to **CE** (Congestion Encountered) if current value is **ECT** 267762369db2SQuentin Monnet * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 267862369db2SQuentin Monnet * and IPv4. 2679f7c917baSbrakmo * Return 268062369db2SQuentin Monnet * 1 if the **CE** flag is set (either by the current helper call 268162369db2SQuentin Monnet * or because it was already present), 0 if it is not set. 2682dbafd7ddSMartin KaFai Lau * 2683dbafd7ddSMartin KaFai Lau * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2684dbafd7ddSMartin KaFai Lau * Description 268562369db2SQuentin Monnet * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 268662369db2SQuentin Monnet * **bpf_sk_release**\ () is unnecessary and not allowed. 2687dbafd7ddSMartin KaFai Lau * Return 268862369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 2689dbafd7ddSMartin KaFai Lau * case of failure. 2690edbf8c01SLorenz Bauer * 2691edbf8c01SLorenz Bauer * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2692edbf8c01SLorenz Bauer * Description 2693edbf8c01SLorenz Bauer * Look for TCP socket matching *tuple*, optionally in a child 2694edbf8c01SLorenz Bauer * network namespace *netns*. The return value must be checked, 2695edbf8c01SLorenz Bauer * and if non-**NULL**, released via **bpf_sk_release**\ (). 2696edbf8c01SLorenz Bauer * 269780867c5eSQuentin Monnet * This function is identical to **bpf_sk_lookup_tcp**\ (), except 269880867c5eSQuentin Monnet * that it also returns timewait or request sockets. Use 269980867c5eSQuentin Monnet * **bpf_sk_fullsock**\ () or **bpf_tcp_sock**\ () to access the 270080867c5eSQuentin Monnet * full structure. 2701edbf8c01SLorenz Bauer * 2702edbf8c01SLorenz Bauer * This helper is available only if the kernel was compiled with 2703edbf8c01SLorenz Bauer * **CONFIG_NET** configuration option. 2704edbf8c01SLorenz Bauer * Return 2705edbf8c01SLorenz Bauer * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2706edbf8c01SLorenz Bauer * For sockets with reuseport option, the **struct bpf_sock** 270780867c5eSQuentin Monnet * result is from *reuse*\ **->socks**\ [] using the hash of the 270880867c5eSQuentin Monnet * tuple. 270939904084SLorenz Bauer * 2710c0df236eSMartin KaFai Lau * long bpf_tcp_check_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 271139904084SLorenz Bauer * Description 271280867c5eSQuentin Monnet * Check whether *iph* and *th* contain a valid SYN cookie ACK for 271380867c5eSQuentin Monnet * the listening socket in *sk*. 271439904084SLorenz Bauer * 271580867c5eSQuentin Monnet * *iph* points to the start of the IPv4 or IPv6 header, while 271680867c5eSQuentin Monnet * *iph_len* contains **sizeof**\ (**struct iphdr**) or 271780867c5eSQuentin Monnet * **sizeof**\ (**struct ip6hdr**). 271839904084SLorenz Bauer * 271980867c5eSQuentin Monnet * *th* points to the start of the TCP header, while *th_len* 272080867c5eSQuentin Monnet * contains **sizeof**\ (**struct tcphdr**). 272139904084SLorenz Bauer * Return 272280867c5eSQuentin Monnet * 0 if *iph* and *th* are a valid SYN cookie ACK, or a negative 272380867c5eSQuentin Monnet * error otherwise. 2724808649fbSAndrey Ignatov * 2725bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_name(struct bpf_sysctl *ctx, char *buf, size_t buf_len, u64 flags) 2726808649fbSAndrey Ignatov * Description 2727808649fbSAndrey Ignatov * Get name of sysctl in /proc/sys/ and copy it into provided by 2728808649fbSAndrey Ignatov * program buffer *buf* of size *buf_len*. 2729808649fbSAndrey Ignatov * 2730808649fbSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 2731808649fbSAndrey Ignatov * 2732808649fbSAndrey Ignatov * If *flags* is zero, full name (e.g. "net/ipv4/tcp_mem") is 2733808649fbSAndrey Ignatov * copied. Use **BPF_F_SYSCTL_BASE_NAME** flag to copy base name 2734808649fbSAndrey Ignatov * only (e.g. "tcp_mem"). 2735808649fbSAndrey Ignatov * Return 2736808649fbSAndrey Ignatov * Number of character copied (not including the trailing NUL). 2737808649fbSAndrey Ignatov * 2738808649fbSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 2739808649fbSAndrey Ignatov * truncated name in this case). 27401d11b301SAndrey Ignatov * 2741bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_current_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 27421d11b301SAndrey Ignatov * Description 27431d11b301SAndrey Ignatov * Get current value of sysctl as it is presented in /proc/sys 27441d11b301SAndrey Ignatov * (incl. newline, etc), and copy it as a string into provided 27451d11b301SAndrey Ignatov * by program buffer *buf* of size *buf_len*. 27461d11b301SAndrey Ignatov * 27471d11b301SAndrey Ignatov * The whole value is copied, no matter what file position user 27481d11b301SAndrey Ignatov * space issued e.g. sys_read at. 27491d11b301SAndrey Ignatov * 27501d11b301SAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 27511d11b301SAndrey Ignatov * Return 27521d11b301SAndrey Ignatov * Number of character copied (not including the trailing NUL). 27531d11b301SAndrey Ignatov * 27541d11b301SAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 27551d11b301SAndrey Ignatov * truncated name in this case). 27561d11b301SAndrey Ignatov * 27571d11b301SAndrey Ignatov * **-EINVAL** if current value was unavailable, e.g. because 27581d11b301SAndrey Ignatov * sysctl is uninitialized and read returns -EIO for it. 27594e63acdfSAndrey Ignatov * 2760bdb7b79bSAndrii Nakryiko * long bpf_sysctl_get_new_value(struct bpf_sysctl *ctx, char *buf, size_t buf_len) 27614e63acdfSAndrey Ignatov * Description 27624e63acdfSAndrey Ignatov * Get new value being written by user space to sysctl (before 27634e63acdfSAndrey Ignatov * the actual write happens) and copy it as a string into 27644e63acdfSAndrey Ignatov * provided by program buffer *buf* of size *buf_len*. 27654e63acdfSAndrey Ignatov * 27664e63acdfSAndrey Ignatov * User space may write new value at file position > 0. 27674e63acdfSAndrey Ignatov * 27684e63acdfSAndrey Ignatov * The buffer is always NUL terminated, unless it's zero-sized. 27694e63acdfSAndrey Ignatov * Return 27704e63acdfSAndrey Ignatov * Number of character copied (not including the trailing NUL). 27714e63acdfSAndrey Ignatov * 27724e63acdfSAndrey Ignatov * **-E2BIG** if the buffer wasn't big enough (*buf* will contain 27734e63acdfSAndrey Ignatov * truncated name in this case). 27744e63acdfSAndrey Ignatov * 27754e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 27764e63acdfSAndrey Ignatov * 2777bdb7b79bSAndrii Nakryiko * long bpf_sysctl_set_new_value(struct bpf_sysctl *ctx, const char *buf, size_t buf_len) 27784e63acdfSAndrey Ignatov * Description 27794e63acdfSAndrey Ignatov * Override new value being written by user space to sysctl with 27804e63acdfSAndrey Ignatov * value provided by program in buffer *buf* of size *buf_len*. 27814e63acdfSAndrey Ignatov * 27824e63acdfSAndrey Ignatov * *buf* should contain a string in same form as provided by user 27834e63acdfSAndrey Ignatov * space on sysctl write. 27844e63acdfSAndrey Ignatov * 27854e63acdfSAndrey Ignatov * User space may write new value at file position > 0. To override 27864e63acdfSAndrey Ignatov * the whole sysctl value file position should be set to zero. 27874e63acdfSAndrey Ignatov * Return 27884e63acdfSAndrey Ignatov * 0 on success. 27894e63acdfSAndrey Ignatov * 27904e63acdfSAndrey Ignatov * **-E2BIG** if the *buf_len* is too big. 27914e63acdfSAndrey Ignatov * 27924e63acdfSAndrey Ignatov * **-EINVAL** if sysctl is being read. 2793d7a4cb9bSAndrey Ignatov * 2794bdb7b79bSAndrii Nakryiko * long bpf_strtol(const char *buf, size_t buf_len, u64 flags, long *res) 2795d7a4cb9bSAndrey Ignatov * Description 2796d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 2797d7a4cb9bSAndrey Ignatov * size *buf_len* to a long integer according to the given base 2798d7a4cb9bSAndrey Ignatov * and save the result in *res*. 2799d7a4cb9bSAndrey Ignatov * 2800d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 280180867c5eSQuentin Monnet * (as determined by **isspace**\ (3)) followed by a single 280280867c5eSQuentin Monnet * optional '**-**' sign. 2803d7a4cb9bSAndrey Ignatov * 2804d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 2805d7a4cb9bSAndrey Ignatov * are currently unused. 2806d7a4cb9bSAndrey Ignatov * 2807d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 280880867c5eSQuentin Monnet * similar to user space **strtol**\ (3). 2809d7a4cb9bSAndrey Ignatov * Return 2810d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 281180867c5eSQuentin Monnet * no more than *buf_len*. 2812d7a4cb9bSAndrey Ignatov * 2813d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 2814d7a4cb9bSAndrey Ignatov * was provided. 2815d7a4cb9bSAndrey Ignatov * 2816d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 2817d7a4cb9bSAndrey Ignatov * 2818bdb7b79bSAndrii Nakryiko * long bpf_strtoul(const char *buf, size_t buf_len, u64 flags, unsigned long *res) 2819d7a4cb9bSAndrey Ignatov * Description 2820d7a4cb9bSAndrey Ignatov * Convert the initial part of the string from buffer *buf* of 2821d7a4cb9bSAndrey Ignatov * size *buf_len* to an unsigned long integer according to the 2822d7a4cb9bSAndrey Ignatov * given base and save the result in *res*. 2823d7a4cb9bSAndrey Ignatov * 2824d7a4cb9bSAndrey Ignatov * The string may begin with an arbitrary amount of white space 282580867c5eSQuentin Monnet * (as determined by **isspace**\ (3)). 2826d7a4cb9bSAndrey Ignatov * 2827d7a4cb9bSAndrey Ignatov * Five least significant bits of *flags* encode base, other bits 2828d7a4cb9bSAndrey Ignatov * are currently unused. 2829d7a4cb9bSAndrey Ignatov * 2830d7a4cb9bSAndrey Ignatov * Base must be either 8, 10, 16 or 0 to detect it automatically 283180867c5eSQuentin Monnet * similar to user space **strtoul**\ (3). 2832d7a4cb9bSAndrey Ignatov * Return 2833d7a4cb9bSAndrey Ignatov * Number of characters consumed on success. Must be positive but 283480867c5eSQuentin Monnet * no more than *buf_len*. 2835d7a4cb9bSAndrey Ignatov * 2836d7a4cb9bSAndrey Ignatov * **-EINVAL** if no valid digits were found or unsupported base 2837d7a4cb9bSAndrey Ignatov * was provided. 2838d7a4cb9bSAndrey Ignatov * 2839d7a4cb9bSAndrey Ignatov * **-ERANGE** if resulting value was out of range. 28406ac99e8fSMartin KaFai Lau * 284130897832SKP Singh * void *bpf_sk_storage_get(struct bpf_map *map, void *sk, void *value, u64 flags) 28426ac99e8fSMartin KaFai Lau * Description 284380867c5eSQuentin Monnet * Get a bpf-local-storage from a *sk*. 28446ac99e8fSMartin KaFai Lau * 28456ac99e8fSMartin KaFai Lau * Logically, it could be thought of getting the value from 28466ac99e8fSMartin KaFai Lau * a *map* with *sk* as the **key**. From this 28476ac99e8fSMartin KaFai Lau * perspective, the usage is not much different from 284880867c5eSQuentin Monnet * **bpf_map_lookup_elem**\ (*map*, **&**\ *sk*) except this 284980867c5eSQuentin Monnet * helper enforces the key must be a full socket and the map must 285080867c5eSQuentin Monnet * be a **BPF_MAP_TYPE_SK_STORAGE** also. 28516ac99e8fSMartin KaFai Lau * 28526ac99e8fSMartin KaFai Lau * Underneath, the value is stored locally at *sk* instead of 285380867c5eSQuentin Monnet * the *map*. The *map* is used as the bpf-local-storage 285480867c5eSQuentin Monnet * "type". The bpf-local-storage "type" (i.e. the *map*) is 285580867c5eSQuentin Monnet * searched against all bpf-local-storages residing at *sk*. 28566ac99e8fSMartin KaFai Lau * 285730897832SKP Singh * *sk* is a kernel **struct sock** pointer for LSM program. 285830897832SKP Singh * *sk* is a **struct bpf_sock** pointer for other program types. 285930897832SKP Singh * 286080867c5eSQuentin Monnet * An optional *flags* (**BPF_SK_STORAGE_GET_F_CREATE**) can be 28616ac99e8fSMartin KaFai Lau * used such that a new bpf-local-storage will be 28626ac99e8fSMartin KaFai Lau * created if one does not exist. *value* can be used 286380867c5eSQuentin Monnet * together with **BPF_SK_STORAGE_GET_F_CREATE** to specify 28646ac99e8fSMartin KaFai Lau * the initial value of a bpf-local-storage. If *value* is 286580867c5eSQuentin Monnet * **NULL**, the new bpf-local-storage will be zero initialized. 28666ac99e8fSMartin KaFai Lau * Return 28676ac99e8fSMartin KaFai Lau * A bpf-local-storage pointer is returned on success. 28686ac99e8fSMartin KaFai Lau * 28696ac99e8fSMartin KaFai Lau * **NULL** if not found or there was an error in adding 28706ac99e8fSMartin KaFai Lau * a new bpf-local-storage. 28716ac99e8fSMartin KaFai Lau * 287230897832SKP Singh * long bpf_sk_storage_delete(struct bpf_map *map, void *sk) 28736ac99e8fSMartin KaFai Lau * Description 287480867c5eSQuentin Monnet * Delete a bpf-local-storage from a *sk*. 28756ac99e8fSMartin KaFai Lau * Return 28766ac99e8fSMartin KaFai Lau * 0 on success. 28776ac99e8fSMartin KaFai Lau * 28786ac99e8fSMartin KaFai Lau * **-ENOENT** if the bpf-local-storage cannot be found. 2879592a3498SMartin KaFai Lau * **-EINVAL** if sk is not a fullsock (e.g. a request_sock). 28808b401f9eSYonghong Song * 2881bdb7b79bSAndrii Nakryiko * long bpf_send_signal(u32 sig) 28828b401f9eSYonghong Song * Description 28838482941fSYonghong Song * Send signal *sig* to the process of the current task. 28848482941fSYonghong Song * The signal may be delivered to any of this process's threads. 28858b401f9eSYonghong Song * Return 28868b401f9eSYonghong Song * 0 on success or successfully queued. 28878b401f9eSYonghong Song * 28888b401f9eSYonghong Song * **-EBUSY** if work queue under nmi is full. 28898b401f9eSYonghong Song * 28908b401f9eSYonghong Song * **-EINVAL** if *sig* is invalid. 28918b401f9eSYonghong Song * 28928b401f9eSYonghong Song * **-EPERM** if no permission to send the *sig*. 28938b401f9eSYonghong Song * 28948b401f9eSYonghong Song * **-EAGAIN** if bpf program can try again. 289570d66244SPetar Penkov * 2896c0df236eSMartin KaFai Lau * s64 bpf_tcp_gen_syncookie(void *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 289770d66244SPetar Penkov * Description 289870d66244SPetar Penkov * Try to issue a SYN cookie for the packet with corresponding 289970d66244SPetar Penkov * IP/TCP headers, *iph* and *th*, on the listening socket in *sk*. 290070d66244SPetar Penkov * 290170d66244SPetar Penkov * *iph* points to the start of the IPv4 or IPv6 header, while 290270d66244SPetar Penkov * *iph_len* contains **sizeof**\ (**struct iphdr**) or 290370d66244SPetar Penkov * **sizeof**\ (**struct ip6hdr**). 290470d66244SPetar Penkov * 290570d66244SPetar Penkov * *th* points to the start of the TCP header, while *th_len* 290670d66244SPetar Penkov * contains the length of the TCP header. 290770d66244SPetar Penkov * Return 290870d66244SPetar Penkov * On success, lower 32 bits hold the generated SYN cookie in 290970d66244SPetar Penkov * followed by 16 bits which hold the MSS value for that cookie, 291070d66244SPetar Penkov * and the top 16 bits are unused. 291170d66244SPetar Penkov * 291270d66244SPetar Penkov * On failure, the returned value is one of the following: 291370d66244SPetar Penkov * 291470d66244SPetar Penkov * **-EINVAL** SYN cookie cannot be issued due to error 291570d66244SPetar Penkov * 291670d66244SPetar Penkov * **-ENOENT** SYN cookie should not be issued (no SYN flood) 291770d66244SPetar Penkov * 291870d66244SPetar Penkov * **-EOPNOTSUPP** kernel configuration does not enable SYN cookies 291970d66244SPetar Penkov * 292070d66244SPetar Penkov * **-EPROTONOSUPPORT** IP packet version is not 4 or 6 2921a7658e1aSAlexei Starovoitov * 2922bdb7b79bSAndrii Nakryiko * long bpf_skb_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 2923a7658e1aSAlexei Starovoitov * Description 2924a7658e1aSAlexei Starovoitov * Write raw *data* blob into a special BPF perf event held by 2925a7658e1aSAlexei Starovoitov * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 2926a7658e1aSAlexei Starovoitov * event must have the following attributes: **PERF_SAMPLE_RAW** 2927a7658e1aSAlexei Starovoitov * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 2928a7658e1aSAlexei Starovoitov * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 2929a7658e1aSAlexei Starovoitov * 2930a7658e1aSAlexei Starovoitov * The *flags* are used to indicate the index in *map* for which 2931a7658e1aSAlexei Starovoitov * the value must be put, masked with **BPF_F_INDEX_MASK**. 2932a7658e1aSAlexei Starovoitov * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 2933a7658e1aSAlexei Starovoitov * to indicate that the index of the current CPU core should be 2934a7658e1aSAlexei Starovoitov * used. 2935a7658e1aSAlexei Starovoitov * 2936a7658e1aSAlexei Starovoitov * The value to write, of *size*, is passed through eBPF stack and 2937a7658e1aSAlexei Starovoitov * pointed by *data*. 2938a7658e1aSAlexei Starovoitov * 2939a7658e1aSAlexei Starovoitov * *ctx* is a pointer to in-kernel struct sk_buff. 2940a7658e1aSAlexei Starovoitov * 2941a7658e1aSAlexei Starovoitov * This helper is similar to **bpf_perf_event_output**\ () but 2942a7658e1aSAlexei Starovoitov * restricted to raw_tracepoint bpf programs. 2943a7658e1aSAlexei Starovoitov * Return 2944a7658e1aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 29456ae08ae3SDaniel Borkmann * 2946bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user(void *dst, u32 size, const void *unsafe_ptr) 29476ae08ae3SDaniel Borkmann * Description 29486ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from user space address 29496ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 29506ae08ae3SDaniel Borkmann * Return 29516ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 29526ae08ae3SDaniel Borkmann * 2953bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel(void *dst, u32 size, const void *unsafe_ptr) 29546ae08ae3SDaniel Borkmann * Description 29556ae08ae3SDaniel Borkmann * Safely attempt to read *size* bytes from kernel space address 29566ae08ae3SDaniel Borkmann * *unsafe_ptr* and store the data in *dst*. 29576ae08ae3SDaniel Borkmann * Return 29586ae08ae3SDaniel Borkmann * 0 on success, or a negative error in case of failure. 29596ae08ae3SDaniel Borkmann * 2960bdb7b79bSAndrii Nakryiko * long bpf_probe_read_user_str(void *dst, u32 size, const void *unsafe_ptr) 29616ae08ae3SDaniel Borkmann * Description 29626ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe user address 29636ae08ae3SDaniel Borkmann * *unsafe_ptr* to *dst*. The *size* should include the 29646ae08ae3SDaniel Borkmann * terminating NUL byte. In case the string length is smaller than 29656ae08ae3SDaniel Borkmann * *size*, the target is not padded with further NUL bytes. If the 29666ae08ae3SDaniel Borkmann * string length is larger than *size*, just *size*-1 bytes are 29676ae08ae3SDaniel Borkmann * copied and the last byte is set to NUL. 29686ae08ae3SDaniel Borkmann * 29696ae08ae3SDaniel Borkmann * On success, the length of the copied string is returned. This 29706ae08ae3SDaniel Borkmann * makes this helper useful in tracing programs for reading 29716ae08ae3SDaniel Borkmann * strings, and more importantly to get its length at runtime. See 29726ae08ae3SDaniel Borkmann * the following snippet: 29736ae08ae3SDaniel Borkmann * 29746ae08ae3SDaniel Borkmann * :: 29756ae08ae3SDaniel Borkmann * 29766ae08ae3SDaniel Borkmann * SEC("kprobe/sys_open") 29776ae08ae3SDaniel Borkmann * void bpf_sys_open(struct pt_regs *ctx) 29786ae08ae3SDaniel Borkmann * { 29796ae08ae3SDaniel Borkmann * char buf[PATHLEN]; // PATHLEN is defined to 256 29806ae08ae3SDaniel Borkmann * int res = bpf_probe_read_user_str(buf, sizeof(buf), 29816ae08ae3SDaniel Borkmann * ctx->di); 29826ae08ae3SDaniel Borkmann * 29836ae08ae3SDaniel Borkmann * // Consume buf, for example push it to 29846ae08ae3SDaniel Borkmann * // userspace via bpf_perf_event_output(); we 29856ae08ae3SDaniel Borkmann * // can use res (the string length) as event 29866ae08ae3SDaniel Borkmann * // size, after checking its boundaries. 29876ae08ae3SDaniel Borkmann * } 29886ae08ae3SDaniel Borkmann * 2989ab8d7809SQuentin Monnet * In comparison, using **bpf_probe_read_user**\ () helper here 29906ae08ae3SDaniel Borkmann * instead to read the string would require to estimate the length 29916ae08ae3SDaniel Borkmann * at compile time, and would often result in copying more memory 29926ae08ae3SDaniel Borkmann * than necessary. 29936ae08ae3SDaniel Borkmann * 29946ae08ae3SDaniel Borkmann * Another useful use case is when parsing individual process 29956ae08ae3SDaniel Borkmann * arguments or individual environment variables navigating 29966ae08ae3SDaniel Borkmann * *current*\ **->mm->arg_start** and *current*\ 29976ae08ae3SDaniel Borkmann * **->mm->env_start**: using this helper and the return value, 29986ae08ae3SDaniel Borkmann * one can quickly iterate at the right offset of the memory area. 29996ae08ae3SDaniel Borkmann * Return 30006ae08ae3SDaniel Borkmann * On success, the strictly positive length of the string, 30016ae08ae3SDaniel Borkmann * including the trailing NUL character. On error, a negative 30026ae08ae3SDaniel Borkmann * value. 30036ae08ae3SDaniel Borkmann * 3004bdb7b79bSAndrii Nakryiko * long bpf_probe_read_kernel_str(void *dst, u32 size, const void *unsafe_ptr) 30056ae08ae3SDaniel Borkmann * Description 30066ae08ae3SDaniel Borkmann * Copy a NUL terminated string from an unsafe kernel address *unsafe_ptr* 3007ab8d7809SQuentin Monnet * to *dst*. Same semantics as with **bpf_probe_read_user_str**\ () apply. 30086ae08ae3SDaniel Borkmann * Return 30096ae08ae3SDaniel Borkmann * On success, the strictly positive length of the string, including 30106ae08ae3SDaniel Borkmann * the trailing NUL character. On error, a negative value. 3011206057feSMartin KaFai Lau * 3012bdb7b79bSAndrii Nakryiko * long bpf_tcp_send_ack(void *tp, u32 rcv_nxt) 3013206057feSMartin KaFai Lau * Description 3014ab8d7809SQuentin Monnet * Send out a tcp-ack. *tp* is the in-kernel struct **tcp_sock**. 3015206057feSMartin KaFai Lau * *rcv_nxt* is the ack_seq to be sent out. 3016206057feSMartin KaFai Lau * Return 3017206057feSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 3018206057feSMartin KaFai Lau * 3019bdb7b79bSAndrii Nakryiko * long bpf_send_signal_thread(u32 sig) 30208482941fSYonghong Song * Description 30218482941fSYonghong Song * Send signal *sig* to the thread corresponding to the current task. 30228482941fSYonghong Song * Return 30238482941fSYonghong Song * 0 on success or successfully queued. 30248482941fSYonghong Song * 30258482941fSYonghong Song * **-EBUSY** if work queue under nmi is full. 30268482941fSYonghong Song * 30278482941fSYonghong Song * **-EINVAL** if *sig* is invalid. 30288482941fSYonghong Song * 30298482941fSYonghong Song * **-EPERM** if no permission to send the *sig*. 30308482941fSYonghong Song * 30318482941fSYonghong Song * **-EAGAIN** if bpf program can try again. 30325576b991SMartin KaFai Lau * 30335576b991SMartin KaFai Lau * u64 bpf_jiffies64(void) 30345576b991SMartin KaFai Lau * Description 30355576b991SMartin KaFai Lau * Obtain the 64bit jiffies 30365576b991SMartin KaFai Lau * Return 30375576b991SMartin KaFai Lau * The 64 bit jiffies 3038fff7b643SDaniel Xu * 3039bdb7b79bSAndrii Nakryiko * long bpf_read_branch_records(struct bpf_perf_event_data *ctx, void *buf, u32 size, u64 flags) 3040fff7b643SDaniel Xu * Description 3041fff7b643SDaniel Xu * For an eBPF program attached to a perf event, retrieve the 3042ab8d7809SQuentin Monnet * branch records (**struct perf_branch_entry**) associated to *ctx* 3043fff7b643SDaniel Xu * and store it in the buffer pointed by *buf* up to size 3044fff7b643SDaniel Xu * *size* bytes. 3045fff7b643SDaniel Xu * Return 3046fff7b643SDaniel Xu * On success, number of bytes written to *buf*. On error, a 3047fff7b643SDaniel Xu * negative value. 3048fff7b643SDaniel Xu * 3049fff7b643SDaniel Xu * The *flags* can be set to **BPF_F_GET_BRANCH_RECORDS_SIZE** to 3050fff7b643SDaniel Xu * instead return the number of bytes required to store all the 3051fff7b643SDaniel Xu * branch entries. If this flag is set, *buf* may be NULL. 3052fff7b643SDaniel Xu * 3053fff7b643SDaniel Xu * **-EINVAL** if arguments invalid or **size** not a multiple 3054ab8d7809SQuentin Monnet * of **sizeof**\ (**struct perf_branch_entry**\ ). 3055fff7b643SDaniel Xu * 3056fff7b643SDaniel Xu * **-ENOENT** if architecture does not support branch records. 3057b4490c5cSCarlos Neira * 3058bdb7b79bSAndrii Nakryiko * long bpf_get_ns_current_pid_tgid(u64 dev, u64 ino, struct bpf_pidns_info *nsdata, u32 size) 3059b4490c5cSCarlos Neira * Description 3060b4490c5cSCarlos Neira * Returns 0 on success, values for *pid* and *tgid* as seen from the current 3061b4490c5cSCarlos Neira * *namespace* will be returned in *nsdata*. 3062ab8d7809SQuentin Monnet * Return 3063ab8d7809SQuentin Monnet * 0 on success, or one of the following in case of failure: 3064b4490c5cSCarlos Neira * 3065b4490c5cSCarlos Neira * **-EINVAL** if dev and inum supplied don't match dev_t and inode number 3066b4490c5cSCarlos Neira * with nsfs of current task, or if dev conversion to dev_t lost high bits. 3067b4490c5cSCarlos Neira * 3068b4490c5cSCarlos Neira * **-ENOENT** if pidns does not exists for the current task. 3069b4490c5cSCarlos Neira * 3070bdb7b79bSAndrii Nakryiko * long bpf_xdp_output(void *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 3071d831ee84SEelco Chaudron * Description 3072d831ee84SEelco Chaudron * Write raw *data* blob into a special BPF perf event held by 3073d831ee84SEelco Chaudron * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 3074d831ee84SEelco Chaudron * event must have the following attributes: **PERF_SAMPLE_RAW** 3075d831ee84SEelco Chaudron * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 3076d831ee84SEelco Chaudron * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 3077d831ee84SEelco Chaudron * 3078d831ee84SEelco Chaudron * The *flags* are used to indicate the index in *map* for which 3079d831ee84SEelco Chaudron * the value must be put, masked with **BPF_F_INDEX_MASK**. 3080d831ee84SEelco Chaudron * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 3081d831ee84SEelco Chaudron * to indicate that the index of the current CPU core should be 3082d831ee84SEelco Chaudron * used. 3083d831ee84SEelco Chaudron * 3084d831ee84SEelco Chaudron * The value to write, of *size*, is passed through eBPF stack and 3085d831ee84SEelco Chaudron * pointed by *data*. 3086d831ee84SEelco Chaudron * 3087d831ee84SEelco Chaudron * *ctx* is a pointer to in-kernel struct xdp_buff. 3088d831ee84SEelco Chaudron * 3089d831ee84SEelco Chaudron * This helper is similar to **bpf_perf_eventoutput**\ () but 3090d831ee84SEelco Chaudron * restricted to raw_tracepoint bpf programs. 3091d831ee84SEelco Chaudron * Return 3092d831ee84SEelco Chaudron * 0 on success, or a negative error in case of failure. 3093f318903cSDaniel Borkmann * 3094f318903cSDaniel Borkmann * u64 bpf_get_netns_cookie(void *ctx) 3095f318903cSDaniel Borkmann * Description 3096f318903cSDaniel Borkmann * Retrieve the cookie (generated by the kernel) of the network 3097f318903cSDaniel Borkmann * namespace the input *ctx* is associated with. The network 3098f318903cSDaniel Borkmann * namespace cookie remains stable for its lifetime and provides 3099f318903cSDaniel Borkmann * a global identifier that can be assumed unique. If *ctx* is 3100f318903cSDaniel Borkmann * NULL, then the helper returns the cookie for the initial 3101f318903cSDaniel Borkmann * network namespace. The cookie itself is very similar to that 3102ab8d7809SQuentin Monnet * of **bpf_get_socket_cookie**\ () helper, but for network 3103ab8d7809SQuentin Monnet * namespaces instead of sockets. 3104f318903cSDaniel Borkmann * Return 3105f318903cSDaniel Borkmann * A 8-byte long opaque number. 31060f09abd1SDaniel Borkmann * 31070f09abd1SDaniel Borkmann * u64 bpf_get_current_ancestor_cgroup_id(int ancestor_level) 31080f09abd1SDaniel Borkmann * Description 31090f09abd1SDaniel Borkmann * Return id of cgroup v2 that is ancestor of the cgroup associated 31100f09abd1SDaniel Borkmann * with the current task at the *ancestor_level*. The root cgroup 31110f09abd1SDaniel Borkmann * is at *ancestor_level* zero and each step down the hierarchy 31120f09abd1SDaniel Borkmann * increments the level. If *ancestor_level* == level of cgroup 31130f09abd1SDaniel Borkmann * associated with the current task, then return value will be the 31140f09abd1SDaniel Borkmann * same as that of **bpf_get_current_cgroup_id**\ (). 31150f09abd1SDaniel Borkmann * 31160f09abd1SDaniel Borkmann * The helper is useful to implement policies based on cgroups 31170f09abd1SDaniel Borkmann * that are upper in hierarchy than immediate cgroup associated 31180f09abd1SDaniel Borkmann * with the current task. 31190f09abd1SDaniel Borkmann * 31200f09abd1SDaniel Borkmann * The format of returned id and helper limitations are same as in 31210f09abd1SDaniel Borkmann * **bpf_get_current_cgroup_id**\ (). 31220f09abd1SDaniel Borkmann * Return 31230f09abd1SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3124cf7fbe66SJoe Stringer * 312527e5203bSMartin KaFai Lau * long bpf_sk_assign(struct sk_buff *skb, void *sk, u64 flags) 3126cf7fbe66SJoe Stringer * Description 3127e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 3128e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SCHED_CLS** and 3129e9ddbb77SJakub Sitnicki * **BPF_PROG_TYPE_SCHED_ACT** programs. 3130e9ddbb77SJakub Sitnicki * 3131cf7fbe66SJoe Stringer * Assign the *sk* to the *skb*. When combined with appropriate 3132cf7fbe66SJoe Stringer * routing configuration to receive the packet towards the socket, 3133cf7fbe66SJoe Stringer * will cause *skb* to be delivered to the specified socket. 3134cf7fbe66SJoe Stringer * Subsequent redirection of *skb* via **bpf_redirect**\ (), 3135cf7fbe66SJoe Stringer * **bpf_clone_redirect**\ () or other methods outside of BPF may 3136cf7fbe66SJoe Stringer * interfere with successful delivery to the socket. 3137cf7fbe66SJoe Stringer * 3138cf7fbe66SJoe Stringer * This operation is only valid from TC ingress path. 3139cf7fbe66SJoe Stringer * 3140cf7fbe66SJoe Stringer * The *flags* argument must be zero. 3141cf7fbe66SJoe Stringer * Return 3142ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 3143cf7fbe66SJoe Stringer * 3144ab8d7809SQuentin Monnet * **-EINVAL** if specified *flags* are not supported. 3145ab8d7809SQuentin Monnet * 3146ab8d7809SQuentin Monnet * **-ENOENT** if the socket is unavailable for assignment. 3147ab8d7809SQuentin Monnet * 3148ab8d7809SQuentin Monnet * **-ENETUNREACH** if the socket is unreachable (wrong netns). 3149ab8d7809SQuentin Monnet * 3150ab8d7809SQuentin Monnet * **-EOPNOTSUPP** if the operation is not supported, for example 3151ab8d7809SQuentin Monnet * a call from outside of TC ingress. 3152ab8d7809SQuentin Monnet * 3153ab8d7809SQuentin Monnet * **-ESOCKTNOSUPPORT** if the socket type is not supported 3154ab8d7809SQuentin Monnet * (reuseport). 315571d19214SMaciej Żenczykowski * 3156e9ddbb77SJakub Sitnicki * long bpf_sk_assign(struct bpf_sk_lookup *ctx, struct bpf_sock *sk, u64 flags) 3157e9ddbb77SJakub Sitnicki * Description 3158e9ddbb77SJakub Sitnicki * Helper is overloaded depending on BPF program type. This 3159e9ddbb77SJakub Sitnicki * description applies to **BPF_PROG_TYPE_SK_LOOKUP** programs. 3160e9ddbb77SJakub Sitnicki * 3161e9ddbb77SJakub Sitnicki * Select the *sk* as a result of a socket lookup. 3162e9ddbb77SJakub Sitnicki * 3163e9ddbb77SJakub Sitnicki * For the operation to succeed passed socket must be compatible 3164e9ddbb77SJakub Sitnicki * with the packet description provided by the *ctx* object. 3165e9ddbb77SJakub Sitnicki * 3166e9ddbb77SJakub Sitnicki * L4 protocol (**IPPROTO_TCP** or **IPPROTO_UDP**) must 3167e9ddbb77SJakub Sitnicki * be an exact match. While IP family (**AF_INET** or 3168e9ddbb77SJakub Sitnicki * **AF_INET6**) must be compatible, that is IPv6 sockets 3169e9ddbb77SJakub Sitnicki * that are not v6-only can be selected for IPv4 packets. 3170e9ddbb77SJakub Sitnicki * 3171e9ddbb77SJakub Sitnicki * Only TCP listeners and UDP unconnected sockets can be 3172e9ddbb77SJakub Sitnicki * selected. *sk* can also be NULL to reset any previous 3173e9ddbb77SJakub Sitnicki * selection. 3174e9ddbb77SJakub Sitnicki * 3175e9ddbb77SJakub Sitnicki * *flags* argument can combination of following values: 3176e9ddbb77SJakub Sitnicki * 3177e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_REPLACE** to override the previous 3178e9ddbb77SJakub Sitnicki * socket selection, potentially done by a BPF program 3179e9ddbb77SJakub Sitnicki * that ran before us. 3180e9ddbb77SJakub Sitnicki * 3181e9ddbb77SJakub Sitnicki * * **BPF_SK_LOOKUP_F_NO_REUSEPORT** to skip 3182e9ddbb77SJakub Sitnicki * load-balancing within reuseport group for the socket 3183e9ddbb77SJakub Sitnicki * being selected. 3184e9ddbb77SJakub Sitnicki * 3185e9ddbb77SJakub Sitnicki * On success *ctx->sk* will point to the selected socket. 3186e9ddbb77SJakub Sitnicki * 3187e9ddbb77SJakub Sitnicki * Return 3188e9ddbb77SJakub Sitnicki * 0 on success, or a negative errno in case of failure. 3189e9ddbb77SJakub Sitnicki * 3190e9ddbb77SJakub Sitnicki * * **-EAFNOSUPPORT** if socket family (*sk->family*) is 3191e9ddbb77SJakub Sitnicki * not compatible with packet family (*ctx->family*). 3192e9ddbb77SJakub Sitnicki * 3193e9ddbb77SJakub Sitnicki * * **-EEXIST** if socket has been already selected, 3194e9ddbb77SJakub Sitnicki * potentially by another program, and 3195e9ddbb77SJakub Sitnicki * **BPF_SK_LOOKUP_F_REPLACE** flag was not specified. 3196e9ddbb77SJakub Sitnicki * 3197e9ddbb77SJakub Sitnicki * * **-EINVAL** if unsupported flags were specified. 3198e9ddbb77SJakub Sitnicki * 3199e9ddbb77SJakub Sitnicki * * **-EPROTOTYPE** if socket L4 protocol 3200e9ddbb77SJakub Sitnicki * (*sk->protocol*) doesn't match packet protocol 3201e9ddbb77SJakub Sitnicki * (*ctx->protocol*). 3202e9ddbb77SJakub Sitnicki * 3203e9ddbb77SJakub Sitnicki * * **-ESOCKTNOSUPPORT** if socket is not in allowed 3204e9ddbb77SJakub Sitnicki * state (TCP listening or UDP unconnected). 3205e9ddbb77SJakub Sitnicki * 320671d19214SMaciej Żenczykowski * u64 bpf_ktime_get_boot_ns(void) 320771d19214SMaciej Żenczykowski * Description 320871d19214SMaciej Żenczykowski * Return the time elapsed since system boot, in nanoseconds. 320971d19214SMaciej Żenczykowski * Does include the time the system was suspended. 3210ab8d7809SQuentin Monnet * See: **clock_gettime**\ (**CLOCK_BOOTTIME**) 321171d19214SMaciej Żenczykowski * Return 321271d19214SMaciej Żenczykowski * Current *ktime*. 3213492e639fSYonghong Song * 3214bdb7b79bSAndrii Nakryiko * long bpf_seq_printf(struct seq_file *m, const char *fmt, u32 fmt_size, const void *data, u32 data_len) 3215492e639fSYonghong Song * Description 3216ab8d7809SQuentin Monnet * **bpf_seq_printf**\ () uses seq_file **seq_printf**\ () to print 3217ab8d7809SQuentin Monnet * out the format string. 3218492e639fSYonghong Song * The *m* represents the seq_file. The *fmt* and *fmt_size* are for 3219492e639fSYonghong Song * the format string itself. The *data* and *data_len* are format string 3220ab8d7809SQuentin Monnet * arguments. The *data* are a **u64** array and corresponding format string 3221492e639fSYonghong Song * values are stored in the array. For strings and pointers where pointees 3222492e639fSYonghong Song * are accessed, only the pointer values are stored in the *data* array. 3223ab8d7809SQuentin Monnet * The *data_len* is the size of *data* in bytes. 3224492e639fSYonghong Song * 3225492e639fSYonghong Song * Formats **%s**, **%p{i,I}{4,6}** requires to read kernel memory. 3226492e639fSYonghong Song * Reading kernel memory may fail due to either invalid address or 3227492e639fSYonghong Song * valid address but requiring a major memory fault. If reading kernel memory 3228492e639fSYonghong Song * fails, the string for **%s** will be an empty string, and the ip 3229492e639fSYonghong Song * address for **%p{i,I}{4,6}** will be 0. Not returning error to 3230ab8d7809SQuentin Monnet * bpf program is consistent with what **bpf_trace_printk**\ () does for now. 3231492e639fSYonghong Song * Return 3232ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 3233492e639fSYonghong Song * 3234ab8d7809SQuentin Monnet * **-EBUSY** if per-CPU memory copy buffer is busy, can try again 3235492e639fSYonghong Song * by returning 1 from bpf program. 3236ab8d7809SQuentin Monnet * 3237ab8d7809SQuentin Monnet * **-EINVAL** if arguments are invalid, or if *fmt* is invalid/unsupported. 3238ab8d7809SQuentin Monnet * 3239ab8d7809SQuentin Monnet * **-E2BIG** if *fmt* contains too many format specifiers. 3240ab8d7809SQuentin Monnet * 3241ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3242492e639fSYonghong Song * 3243bdb7b79bSAndrii Nakryiko * long bpf_seq_write(struct seq_file *m, const void *data, u32 len) 3244492e639fSYonghong Song * Description 3245ab8d7809SQuentin Monnet * **bpf_seq_write**\ () uses seq_file **seq_write**\ () to write the data. 3246492e639fSYonghong Song * The *m* represents the seq_file. The *data* and *len* represent the 3247492e639fSYonghong Song * data to write in bytes. 3248492e639fSYonghong Song * Return 3249ab8d7809SQuentin Monnet * 0 on success, or a negative error in case of failure: 3250492e639fSYonghong Song * 3251ab8d7809SQuentin Monnet * **-EOVERFLOW** if an overflow happened: The same object will be tried again. 3252f307fa2cSAndrey Ignatov * 3253a5fa25adSMartin KaFai Lau * u64 bpf_sk_cgroup_id(void *sk) 3254f307fa2cSAndrey Ignatov * Description 3255f307fa2cSAndrey Ignatov * Return the cgroup v2 id of the socket *sk*. 3256f307fa2cSAndrey Ignatov * 3257a5fa25adSMartin KaFai Lau * *sk* must be a non-**NULL** pointer to a socket, e.g. one 3258f307fa2cSAndrey Ignatov * returned from **bpf_sk_lookup_xxx**\ (), 3259f307fa2cSAndrey Ignatov * **bpf_sk_fullsock**\ (), etc. The format of returned id is 3260f307fa2cSAndrey Ignatov * same as in **bpf_skb_cgroup_id**\ (). 3261f307fa2cSAndrey Ignatov * 3262f307fa2cSAndrey Ignatov * This helper is available only if the kernel was compiled with 3263f307fa2cSAndrey Ignatov * the **CONFIG_SOCK_CGROUP_DATA** configuration option. 3264f307fa2cSAndrey Ignatov * Return 3265f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 3266f307fa2cSAndrey Ignatov * 3267a5fa25adSMartin KaFai Lau * u64 bpf_sk_ancestor_cgroup_id(void *sk, int ancestor_level) 3268f307fa2cSAndrey Ignatov * Description 3269f307fa2cSAndrey Ignatov * Return id of cgroup v2 that is ancestor of cgroup associated 3270f307fa2cSAndrey Ignatov * with the *sk* at the *ancestor_level*. The root cgroup is at 3271f307fa2cSAndrey Ignatov * *ancestor_level* zero and each step down the hierarchy 3272f307fa2cSAndrey Ignatov * increments the level. If *ancestor_level* == level of cgroup 3273f307fa2cSAndrey Ignatov * associated with *sk*, then return value will be same as that 3274f307fa2cSAndrey Ignatov * of **bpf_sk_cgroup_id**\ (). 3275f307fa2cSAndrey Ignatov * 3276f307fa2cSAndrey Ignatov * The helper is useful to implement policies based on cgroups 3277f307fa2cSAndrey Ignatov * that are upper in hierarchy than immediate cgroup associated 3278f307fa2cSAndrey Ignatov * with *sk*. 3279f307fa2cSAndrey Ignatov * 3280f307fa2cSAndrey Ignatov * The format of returned id and helper limitations are same as in 3281f307fa2cSAndrey Ignatov * **bpf_sk_cgroup_id**\ (). 3282f307fa2cSAndrey Ignatov * Return 3283f307fa2cSAndrey Ignatov * The id is returned or 0 in case the id could not be retrieved. 3284457f4436SAndrii Nakryiko * 3285e1613b57SAndrii Nakryiko * long bpf_ringbuf_output(void *ringbuf, void *data, u64 size, u64 flags) 3286457f4436SAndrii Nakryiko * Description 3287457f4436SAndrii Nakryiko * Copy *size* bytes from *data* into a ring buffer *ringbuf*. 3288bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3289bcc7f554SQuentin Monnet * of new data availability is sent. 3290bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3291bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 3292457f4436SAndrii Nakryiko * Return 3293bcc7f554SQuentin Monnet * 0 on success, or a negative error in case of failure. 3294457f4436SAndrii Nakryiko * 3295457f4436SAndrii Nakryiko * void *bpf_ringbuf_reserve(void *ringbuf, u64 size, u64 flags) 3296457f4436SAndrii Nakryiko * Description 3297457f4436SAndrii Nakryiko * Reserve *size* bytes of payload in a ring buffer *ringbuf*. 3298457f4436SAndrii Nakryiko * Return 3299457f4436SAndrii Nakryiko * Valid pointer with *size* bytes of memory available; NULL, 3300457f4436SAndrii Nakryiko * otherwise. 3301457f4436SAndrii Nakryiko * 3302457f4436SAndrii Nakryiko * void bpf_ringbuf_submit(void *data, u64 flags) 3303457f4436SAndrii Nakryiko * Description 3304457f4436SAndrii Nakryiko * Submit reserved ring buffer sample, pointed to by *data*. 3305bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3306bcc7f554SQuentin Monnet * of new data availability is sent. 3307bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3308bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 3309457f4436SAndrii Nakryiko * Return 3310457f4436SAndrii Nakryiko * Nothing. Always succeeds. 3311457f4436SAndrii Nakryiko * 3312457f4436SAndrii Nakryiko * void bpf_ringbuf_discard(void *data, u64 flags) 3313457f4436SAndrii Nakryiko * Description 3314457f4436SAndrii Nakryiko * Discard reserved ring buffer sample, pointed to by *data*. 3315bcc7f554SQuentin Monnet * If **BPF_RB_NO_WAKEUP** is specified in *flags*, no notification 3316bcc7f554SQuentin Monnet * of new data availability is sent. 3317bcc7f554SQuentin Monnet * If **BPF_RB_FORCE_WAKEUP** is specified in *flags*, notification 3318bcc7f554SQuentin Monnet * of new data availability is sent unconditionally. 3319457f4436SAndrii Nakryiko * Return 3320457f4436SAndrii Nakryiko * Nothing. Always succeeds. 3321457f4436SAndrii Nakryiko * 3322457f4436SAndrii Nakryiko * u64 bpf_ringbuf_query(void *ringbuf, u64 flags) 3323457f4436SAndrii Nakryiko * Description 3324457f4436SAndrii Nakryiko * Query various characteristics of provided ring buffer. What 3325457f4436SAndrii Nakryiko * exactly is queries is determined by *flags*: 3326bcc7f554SQuentin Monnet * 3327bcc7f554SQuentin Monnet * * **BPF_RB_AVAIL_DATA**: Amount of data not yet consumed. 3328bcc7f554SQuentin Monnet * * **BPF_RB_RING_SIZE**: The size of ring buffer. 3329bcc7f554SQuentin Monnet * * **BPF_RB_CONS_POS**: Consumer position (can wrap around). 3330bcc7f554SQuentin Monnet * * **BPF_RB_PROD_POS**: Producer(s) position (can wrap around). 3331bcc7f554SQuentin Monnet * 3332bcc7f554SQuentin Monnet * Data returned is just a momentary snapshot of actual values 3333457f4436SAndrii Nakryiko * and could be inaccurate, so this facility should be used to 3334457f4436SAndrii Nakryiko * power heuristics and for reporting, not to make 100% correct 3335457f4436SAndrii Nakryiko * calculation. 3336457f4436SAndrii Nakryiko * Return 3337bcc7f554SQuentin Monnet * Requested value, or 0, if *flags* are not recognized. 33387cdec54fSDaniel Borkmann * 3339bdb7b79bSAndrii Nakryiko * long bpf_csum_level(struct sk_buff *skb, u64 level) 33407cdec54fSDaniel Borkmann * Description 33417cdec54fSDaniel Borkmann * Change the skbs checksum level by one layer up or down, or 33427cdec54fSDaniel Borkmann * reset it entirely to none in order to have the stack perform 33437cdec54fSDaniel Borkmann * checksum validation. The level is applicable to the following 33447cdec54fSDaniel Borkmann * protocols: TCP, UDP, GRE, SCTP, FCOE. For example, a decap of 33457cdec54fSDaniel Borkmann * | ETH | IP | UDP | GUE | IP | TCP | into | ETH | IP | TCP | 33467cdec54fSDaniel Borkmann * through **bpf_skb_adjust_room**\ () helper with passing in 33477cdec54fSDaniel Borkmann * **BPF_F_ADJ_ROOM_NO_CSUM_RESET** flag would require one call 33487cdec54fSDaniel Borkmann * to **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_DEC** since 33497cdec54fSDaniel Borkmann * the UDP header is removed. Similarly, an encap of the latter 33507cdec54fSDaniel Borkmann * into the former could be accompanied by a helper call to 33517cdec54fSDaniel Borkmann * **bpf_csum_level**\ () with **BPF_CSUM_LEVEL_INC** if the 33527cdec54fSDaniel Borkmann * skb is still intended to be processed in higher layers of the 33537cdec54fSDaniel Borkmann * stack instead of just egressing at tc. 33547cdec54fSDaniel Borkmann * 33557cdec54fSDaniel Borkmann * There are three supported level settings at this time: 33567cdec54fSDaniel Borkmann * 33577cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_INC**: Increases skb->csum_level for skbs 33587cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 33597cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_DEC**: Decreases skb->csum_level for skbs 33607cdec54fSDaniel Borkmann * with CHECKSUM_UNNECESSARY. 33617cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_RESET**: Resets skb->csum_level to 0 and 33627cdec54fSDaniel Borkmann * sets CHECKSUM_NONE to force checksum validation by the stack. 33637cdec54fSDaniel Borkmann * * **BPF_CSUM_LEVEL_QUERY**: No-op, returns the current 33647cdec54fSDaniel Borkmann * skb->csum_level. 33657cdec54fSDaniel Borkmann * Return 33667cdec54fSDaniel Borkmann * 0 on success, or a negative error in case of failure. In the 33677cdec54fSDaniel Borkmann * case of **BPF_CSUM_LEVEL_QUERY**, the current skb->csum_level 33687cdec54fSDaniel Borkmann * is returned or the error code -EACCES in case the skb is not 33697cdec54fSDaniel Borkmann * subject to CHECKSUM_UNNECESSARY. 3370af7ec138SYonghong Song * 3371af7ec138SYonghong Song * struct tcp6_sock *bpf_skc_to_tcp6_sock(void *sk) 3372af7ec138SYonghong Song * Description 3373af7ec138SYonghong Song * Dynamically cast a *sk* pointer to a *tcp6_sock* pointer. 3374af7ec138SYonghong Song * Return 3375938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 3376478cfbdfSYonghong Song * 3377478cfbdfSYonghong Song * struct tcp_sock *bpf_skc_to_tcp_sock(void *sk) 3378478cfbdfSYonghong Song * Description 3379478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_sock* pointer. 3380478cfbdfSYonghong Song * Return 3381938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 3382478cfbdfSYonghong Song * 3383478cfbdfSYonghong Song * struct tcp_timewait_sock *bpf_skc_to_tcp_timewait_sock(void *sk) 3384478cfbdfSYonghong Song * Description 3385478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_timewait_sock* pointer. 3386478cfbdfSYonghong Song * Return 3387938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 3388478cfbdfSYonghong Song * 3389478cfbdfSYonghong Song * struct tcp_request_sock *bpf_skc_to_tcp_request_sock(void *sk) 3390478cfbdfSYonghong Song * Description 3391478cfbdfSYonghong Song * Dynamically cast a *sk* pointer to a *tcp_request_sock* pointer. 3392478cfbdfSYonghong Song * Return 3393938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 33940d4fad3eSYonghong Song * 33950d4fad3eSYonghong Song * struct udp6_sock *bpf_skc_to_udp6_sock(void *sk) 33960d4fad3eSYonghong Song * Description 33970d4fad3eSYonghong Song * Dynamically cast a *sk* pointer to a *udp6_sock* pointer. 33980d4fad3eSYonghong Song * Return 3399938c3efdSQuentin Monnet * *sk* if casting is valid, or **NULL** otherwise. 3400fa28dcb8SSong Liu * 3401fa28dcb8SSong Liu * long bpf_get_task_stack(struct task_struct *task, void *buf, u32 size, u64 flags) 3402fa28dcb8SSong Liu * Description 3403fa28dcb8SSong Liu * Return a user or a kernel stack in bpf program provided buffer. 3404fa28dcb8SSong Liu * To achieve this, the helper needs *task*, which is a valid 3405938c3efdSQuentin Monnet * pointer to **struct task_struct**. To store the stacktrace, the 3406fa28dcb8SSong Liu * bpf program provides *buf* with a nonnegative *size*. 3407fa28dcb8SSong Liu * 3408fa28dcb8SSong Liu * The last argument, *flags*, holds the number of stack frames to 3409fa28dcb8SSong Liu * skip (from 0 to 255), masked with 3410fa28dcb8SSong Liu * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 3411fa28dcb8SSong Liu * the following flags: 3412fa28dcb8SSong Liu * 3413fa28dcb8SSong Liu * **BPF_F_USER_STACK** 3414fa28dcb8SSong Liu * Collect a user space stack instead of a kernel stack. 3415fa28dcb8SSong Liu * **BPF_F_USER_BUILD_ID** 3416fa28dcb8SSong Liu * Collect buildid+offset instead of ips for user stack, 3417fa28dcb8SSong Liu * only valid if **BPF_F_USER_STACK** is also specified. 3418fa28dcb8SSong Liu * 3419fa28dcb8SSong Liu * **bpf_get_task_stack**\ () can collect up to 3420fa28dcb8SSong Liu * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 3421fa28dcb8SSong Liu * to sufficient large buffer size. Note that 3422fa28dcb8SSong Liu * this limit can be controlled with the **sysctl** program, and 3423fa28dcb8SSong Liu * that it should be manually increased in order to profile long 3424fa28dcb8SSong Liu * user stacks (such as stacks for Java programs). To do so, use: 3425fa28dcb8SSong Liu * 3426fa28dcb8SSong Liu * :: 3427fa28dcb8SSong Liu * 3428fa28dcb8SSong Liu * # sysctl kernel.perf_event_max_stack=<new value> 3429fa28dcb8SSong Liu * Return 3430fa28dcb8SSong Liu * A non-negative value equal to or less than *size* on success, 3431fa28dcb8SSong Liu * or a negative error in case of failure. 3432fa28dcb8SSong Liu * 34330813a841SMartin KaFai Lau * long bpf_load_hdr_opt(struct bpf_sock_ops *skops, void *searchby_res, u32 len, u64 flags) 34340813a841SMartin KaFai Lau * Description 34350813a841SMartin KaFai Lau * Load header option. Support reading a particular TCP header 3436938c3efdSQuentin Monnet * option for bpf program (**BPF_PROG_TYPE_SOCK_OPS**). 34370813a841SMartin KaFai Lau * 34380813a841SMartin KaFai Lau * If *flags* is 0, it will search the option from the 3439938c3efdSQuentin Monnet * *skops*\ **->skb_data**. The comment in **struct bpf_sock_ops** 34400813a841SMartin KaFai Lau * has details on what skb_data contains under different 3441938c3efdSQuentin Monnet * *skops*\ **->op**. 34420813a841SMartin KaFai Lau * 34430813a841SMartin KaFai Lau * The first byte of the *searchby_res* specifies the 34440813a841SMartin KaFai Lau * kind that it wants to search. 34450813a841SMartin KaFai Lau * 34460813a841SMartin KaFai Lau * If the searching kind is an experimental kind 34470813a841SMartin KaFai Lau * (i.e. 253 or 254 according to RFC6994). It also 34480813a841SMartin KaFai Lau * needs to specify the "magic" which is either 34490813a841SMartin KaFai Lau * 2 bytes or 4 bytes. It then also needs to 34500813a841SMartin KaFai Lau * specify the size of the magic by using 34510813a841SMartin KaFai Lau * the 2nd byte which is "kind-length" of a TCP 34520813a841SMartin KaFai Lau * header option and the "kind-length" also 34530813a841SMartin KaFai Lau * includes the first 2 bytes "kind" and "kind-length" 34540813a841SMartin KaFai Lau * itself as a normal TCP header option also does. 34550813a841SMartin KaFai Lau * 34560813a841SMartin KaFai Lau * For example, to search experimental kind 254 with 34570813a841SMartin KaFai Lau * 2 byte magic 0xeB9F, the searchby_res should be 34580813a841SMartin KaFai Lau * [ 254, 4, 0xeB, 0x9F, 0, 0, .... 0 ]. 34590813a841SMartin KaFai Lau * 34600813a841SMartin KaFai Lau * To search for the standard window scale option (3), 3461938c3efdSQuentin Monnet * the *searchby_res* should be [ 3, 0, 0, .... 0 ]. 34620813a841SMartin KaFai Lau * Note, kind-length must be 0 for regular option. 34630813a841SMartin KaFai Lau * 34640813a841SMartin KaFai Lau * Searching for No-Op (0) and End-of-Option-List (1) are 34650813a841SMartin KaFai Lau * not supported. 34660813a841SMartin KaFai Lau * 34670813a841SMartin KaFai Lau * *len* must be at least 2 bytes which is the minimal size 34680813a841SMartin KaFai Lau * of a header option. 34690813a841SMartin KaFai Lau * 34700813a841SMartin KaFai Lau * Supported flags: 3471938c3efdSQuentin Monnet * 34720813a841SMartin KaFai Lau * * **BPF_LOAD_HDR_OPT_TCP_SYN** to search from the 34730813a841SMartin KaFai Lau * saved_syn packet or the just-received syn packet. 34740813a841SMartin KaFai Lau * 34750813a841SMartin KaFai Lau * Return 34760813a841SMartin KaFai Lau * > 0 when found, the header option is copied to *searchby_res*. 3477938c3efdSQuentin Monnet * The return value is the total length copied. On failure, a 3478938c3efdSQuentin Monnet * negative error code is returned: 34790813a841SMartin KaFai Lau * 3480938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 34810813a841SMartin KaFai Lau * 3482938c3efdSQuentin Monnet * **-ENOMSG** if the option is not found. 34830813a841SMartin KaFai Lau * 3484938c3efdSQuentin Monnet * **-ENOENT** if no syn packet is available when 3485938c3efdSQuentin Monnet * **BPF_LOAD_HDR_OPT_TCP_SYN** is used. 34860813a841SMartin KaFai Lau * 3487938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space. Only *len* number of 34880813a841SMartin KaFai Lau * bytes are copied. 34890813a841SMartin KaFai Lau * 3490938c3efdSQuentin Monnet * **-EFAULT** on failure to parse the header options in the 3491938c3efdSQuentin Monnet * packet. 34920813a841SMartin KaFai Lau * 3493938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 3494938c3efdSQuentin Monnet * *skops*\ **->op**. 34950813a841SMartin KaFai Lau * 34960813a841SMartin KaFai Lau * long bpf_store_hdr_opt(struct bpf_sock_ops *skops, const void *from, u32 len, u64 flags) 34970813a841SMartin KaFai Lau * Description 34980813a841SMartin KaFai Lau * Store header option. The data will be copied 34990813a841SMartin KaFai Lau * from buffer *from* with length *len* to the TCP header. 35000813a841SMartin KaFai Lau * 35010813a841SMartin KaFai Lau * The buffer *from* should have the whole option that 35020813a841SMartin KaFai Lau * includes the kind, kind-length, and the actual 35030813a841SMartin KaFai Lau * option data. The *len* must be at least kind-length 35040813a841SMartin KaFai Lau * long. The kind-length does not have to be 4 byte 35050813a841SMartin KaFai Lau * aligned. The kernel will take care of the padding 35060813a841SMartin KaFai Lau * and setting the 4 bytes aligned value to th->doff. 35070813a841SMartin KaFai Lau * 35080813a841SMartin KaFai Lau * This helper will check for duplicated option 35090813a841SMartin KaFai Lau * by searching the same option in the outgoing skb. 35100813a841SMartin KaFai Lau * 35110813a841SMartin KaFai Lau * This helper can only be called during 3512938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 35130813a841SMartin KaFai Lau * 35140813a841SMartin KaFai Lau * Return 35150813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 35160813a841SMartin KaFai Lau * 3517938c3efdSQuentin Monnet * **-EINVAL** If param is invalid. 35180813a841SMartin KaFai Lau * 3519938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 35200813a841SMartin KaFai Lau * Nothing has been written 35210813a841SMartin KaFai Lau * 3522938c3efdSQuentin Monnet * **-EEXIST** if the option already exists. 35230813a841SMartin KaFai Lau * 3524938c3efdSQuentin Monnet * **-EFAULT** on failrue to parse the existing header options. 35250813a841SMartin KaFai Lau * 3526938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 3527938c3efdSQuentin Monnet * *skops*\ **->op**. 35280813a841SMartin KaFai Lau * 35290813a841SMartin KaFai Lau * long bpf_reserve_hdr_opt(struct bpf_sock_ops *skops, u32 len, u64 flags) 35300813a841SMartin KaFai Lau * Description 35310813a841SMartin KaFai Lau * Reserve *len* bytes for the bpf header option. The 3532938c3efdSQuentin Monnet * space will be used by **bpf_store_hdr_opt**\ () later in 3533938c3efdSQuentin Monnet * **BPF_SOCK_OPS_WRITE_HDR_OPT_CB**. 35340813a841SMartin KaFai Lau * 3535938c3efdSQuentin Monnet * If **bpf_reserve_hdr_opt**\ () is called multiple times, 35360813a841SMartin KaFai Lau * the total number of bytes will be reserved. 35370813a841SMartin KaFai Lau * 35380813a841SMartin KaFai Lau * This helper can only be called during 3539938c3efdSQuentin Monnet * **BPF_SOCK_OPS_HDR_OPT_LEN_CB**. 35400813a841SMartin KaFai Lau * 35410813a841SMartin KaFai Lau * Return 35420813a841SMartin KaFai Lau * 0 on success, or negative error in case of failure: 35430813a841SMartin KaFai Lau * 3544938c3efdSQuentin Monnet * **-EINVAL** if a parameter is invalid. 35450813a841SMartin KaFai Lau * 3546938c3efdSQuentin Monnet * **-ENOSPC** if there is not enough space in the header. 35470813a841SMartin KaFai Lau * 3548938c3efdSQuentin Monnet * **-EPERM** if the helper cannot be used under the current 3549938c3efdSQuentin Monnet * *skops*\ **->op**. 35506e22ab9dSJiri Olsa * 35518ea63684SKP Singh * void *bpf_inode_storage_get(struct bpf_map *map, void *inode, void *value, u64 flags) 35528ea63684SKP Singh * Description 35538ea63684SKP Singh * Get a bpf_local_storage from an *inode*. 35548ea63684SKP Singh * 35558ea63684SKP Singh * Logically, it could be thought of as getting the value from 35568ea63684SKP Singh * a *map* with *inode* as the **key**. From this 35578ea63684SKP Singh * perspective, the usage is not much different from 35588ea63684SKP Singh * **bpf_map_lookup_elem**\ (*map*, **&**\ *inode*) except this 35598ea63684SKP Singh * helper enforces the key must be an inode and the map must also 35608ea63684SKP Singh * be a **BPF_MAP_TYPE_INODE_STORAGE**. 35618ea63684SKP Singh * 35628ea63684SKP Singh * Underneath, the value is stored locally at *inode* instead of 35638ea63684SKP Singh * the *map*. The *map* is used as the bpf-local-storage 35648ea63684SKP Singh * "type". The bpf-local-storage "type" (i.e. the *map*) is 35658ea63684SKP Singh * searched against all bpf_local_storage residing at *inode*. 35668ea63684SKP Singh * 35678ea63684SKP Singh * An optional *flags* (**BPF_LOCAL_STORAGE_GET_F_CREATE**) can be 35688ea63684SKP Singh * used such that a new bpf_local_storage will be 35698ea63684SKP Singh * created if one does not exist. *value* can be used 35708ea63684SKP Singh * together with **BPF_LOCAL_STORAGE_GET_F_CREATE** to specify 35718ea63684SKP Singh * the initial value of a bpf_local_storage. If *value* is 35728ea63684SKP Singh * **NULL**, the new bpf_local_storage will be zero initialized. 35738ea63684SKP Singh * Return 35748ea63684SKP Singh * A bpf_local_storage pointer is returned on success. 35758ea63684SKP Singh * 35768ea63684SKP Singh * **NULL** if not found or there was an error in adding 35778ea63684SKP Singh * a new bpf_local_storage. 35788ea63684SKP Singh * 35798ea63684SKP Singh * int bpf_inode_storage_delete(struct bpf_map *map, void *inode) 35808ea63684SKP Singh * Description 35818ea63684SKP Singh * Delete a bpf_local_storage from an *inode*. 35828ea63684SKP Singh * Return 35838ea63684SKP Singh * 0 on success. 35848ea63684SKP Singh * 35858ea63684SKP Singh * **-ENOENT** if the bpf_local_storage cannot be found. 35866e22ab9dSJiri Olsa * 35876e22ab9dSJiri Olsa * long bpf_d_path(struct path *path, char *buf, u32 sz) 35886e22ab9dSJiri Olsa * Description 3589938c3efdSQuentin Monnet * Return full path for given **struct path** object, which 3590938c3efdSQuentin Monnet * needs to be the kernel BTF *path* object. The path is 3591938c3efdSQuentin Monnet * returned in the provided buffer *buf* of size *sz* and 35926e22ab9dSJiri Olsa * is zero terminated. 35936e22ab9dSJiri Olsa * 35946e22ab9dSJiri Olsa * Return 35956e22ab9dSJiri Olsa * On success, the strictly positive length of the string, 35966e22ab9dSJiri Olsa * including the trailing NUL character. On error, a negative 35976e22ab9dSJiri Olsa * value. 359807be4c4aSAlexei Starovoitov * 359907be4c4aSAlexei Starovoitov * long bpf_copy_from_user(void *dst, u32 size, const void *user_ptr) 360007be4c4aSAlexei Starovoitov * Description 360107be4c4aSAlexei Starovoitov * Read *size* bytes from user space address *user_ptr* and store 3602938c3efdSQuentin Monnet * the data in *dst*. This is a wrapper of **copy_from_user**\ (). 360307be4c4aSAlexei Starovoitov * Return 360407be4c4aSAlexei Starovoitov * 0 on success, or a negative error in case of failure. 3605c4d0bfb4SAlan Maguire * 3606c4d0bfb4SAlan Maguire * long bpf_snprintf_btf(char *str, u32 str_size, struct btf_ptr *ptr, u32 btf_ptr_size, u64 flags) 3607c4d0bfb4SAlan Maguire * Description 3608c4d0bfb4SAlan Maguire * Use BTF to store a string representation of *ptr*->ptr in *str*, 3609c4d0bfb4SAlan Maguire * using *ptr*->type_id. This value should specify the type 3610c4d0bfb4SAlan Maguire * that *ptr*->ptr points to. LLVM __builtin_btf_type_id(type, 1) 3611c4d0bfb4SAlan Maguire * can be used to look up vmlinux BTF type ids. Traversing the 3612c4d0bfb4SAlan Maguire * data structure using BTF, the type information and values are 3613c4d0bfb4SAlan Maguire * stored in the first *str_size* - 1 bytes of *str*. Safe copy of 3614c4d0bfb4SAlan Maguire * the pointer data is carried out to avoid kernel crashes during 3615c4d0bfb4SAlan Maguire * operation. Smaller types can use string space on the stack; 3616c4d0bfb4SAlan Maguire * larger programs can use map data to store the string 3617c4d0bfb4SAlan Maguire * representation. 3618c4d0bfb4SAlan Maguire * 3619c4d0bfb4SAlan Maguire * The string can be subsequently shared with userspace via 3620c4d0bfb4SAlan Maguire * bpf_perf_event_output() or ring buffer interfaces. 3621c4d0bfb4SAlan Maguire * bpf_trace_printk() is to be avoided as it places too small 3622c4d0bfb4SAlan Maguire * a limit on string size to be useful. 3623c4d0bfb4SAlan Maguire * 3624c4d0bfb4SAlan Maguire * *flags* is a combination of 3625c4d0bfb4SAlan Maguire * 3626c4d0bfb4SAlan Maguire * **BTF_F_COMPACT** 3627c4d0bfb4SAlan Maguire * no formatting around type information 3628c4d0bfb4SAlan Maguire * **BTF_F_NONAME** 3629c4d0bfb4SAlan Maguire * no struct/union member names/types 3630c4d0bfb4SAlan Maguire * **BTF_F_PTR_RAW** 3631c4d0bfb4SAlan Maguire * show raw (unobfuscated) pointer values; 3632c4d0bfb4SAlan Maguire * equivalent to printk specifier %px. 3633c4d0bfb4SAlan Maguire * **BTF_F_ZERO** 3634c4d0bfb4SAlan Maguire * show zero-valued struct/union members; they 3635c4d0bfb4SAlan Maguire * are not displayed by default 3636c4d0bfb4SAlan Maguire * 3637c4d0bfb4SAlan Maguire * Return 3638c4d0bfb4SAlan Maguire * The number of bytes that were written (or would have been 3639c4d0bfb4SAlan Maguire * written if output had to be truncated due to string size), 3640c4d0bfb4SAlan Maguire * or a negative error in cases of failure. 3641eb411377SAlan Maguire * 3642eb411377SAlan Maguire * long bpf_seq_printf_btf(struct seq_file *m, struct btf_ptr *ptr, u32 ptr_size, u64 flags) 3643eb411377SAlan Maguire * Description 3644eb411377SAlan Maguire * Use BTF to write to seq_write a string representation of 3645eb411377SAlan Maguire * *ptr*->ptr, using *ptr*->type_id as per bpf_snprintf_btf(). 3646eb411377SAlan Maguire * *flags* are identical to those used for bpf_snprintf_btf. 3647eb411377SAlan Maguire * Return 3648eb411377SAlan Maguire * 0 on success or a negative error in case of failure. 3649b426ce83SDaniel Borkmann * 3650b426ce83SDaniel Borkmann * u64 bpf_skb_cgroup_classid(struct sk_buff *skb) 3651b426ce83SDaniel Borkmann * Description 3652b426ce83SDaniel Borkmann * See **bpf_get_cgroup_classid**\ () for the main description. 3653b426ce83SDaniel Borkmann * This helper differs from **bpf_get_cgroup_classid**\ () in that 3654b426ce83SDaniel Borkmann * the cgroup v1 net_cls class is retrieved only from the *skb*'s 3655b426ce83SDaniel Borkmann * associated socket instead of the current process. 3656b426ce83SDaniel Borkmann * Return 3657b426ce83SDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 3658b4ab3141SDaniel Borkmann * 3659b4ab3141SDaniel Borkmann * long bpf_redirect_neigh(u32 ifindex, u64 flags) 3660b4ab3141SDaniel Borkmann * Description 3661b4ab3141SDaniel Borkmann * Redirect the packet to another net device of index *ifindex* 3662b4ab3141SDaniel Borkmann * and fill in L2 addresses from neighboring subsystem. This helper 3663b4ab3141SDaniel Borkmann * is somewhat similar to **bpf_redirect**\ (), except that it 3664b4ab3141SDaniel Borkmann * fills in e.g. MAC addresses based on the L3 information from 3665b4ab3141SDaniel Borkmann * the packet. This helper is supported for IPv4 and IPv6 protocols. 3666b4ab3141SDaniel Borkmann * The *flags* argument is reserved and must be 0. The helper is 3667b4ab3141SDaniel Borkmann * currently only supported for tc BPF program types. 3668b4ab3141SDaniel Borkmann * Return 3669b4ab3141SDaniel Borkmann * The helper returns **TC_ACT_REDIRECT** on success or 3670b4ab3141SDaniel Borkmann * **TC_ACT_SHOT** on error. 36717a4b28c6SDaniel Borkmann */ 3672ebb676daSThomas Graf #define __BPF_FUNC_MAPPER(FN) \ 3673ebb676daSThomas Graf FN(unspec), \ 3674ebb676daSThomas Graf FN(map_lookup_elem), \ 3675ebb676daSThomas Graf FN(map_update_elem), \ 3676ebb676daSThomas Graf FN(map_delete_elem), \ 3677ebb676daSThomas Graf FN(probe_read), \ 3678ebb676daSThomas Graf FN(ktime_get_ns), \ 3679ebb676daSThomas Graf FN(trace_printk), \ 3680ebb676daSThomas Graf FN(get_prandom_u32), \ 3681ebb676daSThomas Graf FN(get_smp_processor_id), \ 3682ebb676daSThomas Graf FN(skb_store_bytes), \ 3683ebb676daSThomas Graf FN(l3_csum_replace), \ 3684ebb676daSThomas Graf FN(l4_csum_replace), \ 3685ebb676daSThomas Graf FN(tail_call), \ 3686ebb676daSThomas Graf FN(clone_redirect), \ 3687ebb676daSThomas Graf FN(get_current_pid_tgid), \ 3688ebb676daSThomas Graf FN(get_current_uid_gid), \ 3689ebb676daSThomas Graf FN(get_current_comm), \ 3690ebb676daSThomas Graf FN(get_cgroup_classid), \ 3691ebb676daSThomas Graf FN(skb_vlan_push), \ 3692ebb676daSThomas Graf FN(skb_vlan_pop), \ 3693ebb676daSThomas Graf FN(skb_get_tunnel_key), \ 3694ebb676daSThomas Graf FN(skb_set_tunnel_key), \ 3695ebb676daSThomas Graf FN(perf_event_read), \ 3696ebb676daSThomas Graf FN(redirect), \ 3697ebb676daSThomas Graf FN(get_route_realm), \ 3698ebb676daSThomas Graf FN(perf_event_output), \ 3699ebb676daSThomas Graf FN(skb_load_bytes), \ 3700ebb676daSThomas Graf FN(get_stackid), \ 3701ebb676daSThomas Graf FN(csum_diff), \ 3702ebb676daSThomas Graf FN(skb_get_tunnel_opt), \ 3703ebb676daSThomas Graf FN(skb_set_tunnel_opt), \ 3704ebb676daSThomas Graf FN(skb_change_proto), \ 3705ebb676daSThomas Graf FN(skb_change_type), \ 3706ebb676daSThomas Graf FN(skb_under_cgroup), \ 3707ebb676daSThomas Graf FN(get_hash_recalc), \ 3708ebb676daSThomas Graf FN(get_current_task), \ 3709ebb676daSThomas Graf FN(probe_write_user), \ 3710ebb676daSThomas Graf FN(current_task_under_cgroup), \ 3711ebb676daSThomas Graf FN(skb_change_tail), \ 3712ebb676daSThomas Graf FN(skb_pull_data), \ 3713ebb676daSThomas Graf FN(csum_update), \ 3714ebb676daSThomas Graf FN(set_hash_invalid), \ 37153a0af8fdSThomas Graf FN(get_numa_node_id), \ 371617bedab2SMartin KaFai Lau FN(skb_change_head), \ 3717a5e8c070SGianluca Borello FN(xdp_adjust_head), \ 371891b8270fSChenbo Feng FN(probe_read_str), \ 37196acc5c29SChenbo Feng FN(get_socket_cookie), \ 3720ded092cdSDaniel Borkmann FN(get_socket_uid), \ 37218c4b4c7eSLawrence Brakmo FN(set_hash), \ 37222be7e212SDaniel Borkmann FN(setsockopt), \ 372397f91a7cSJohn Fastabend FN(skb_adjust_room), \ 3724174a79ffSJohn Fastabend FN(redirect_map), \ 3725174a79ffSJohn Fastabend FN(sk_redirect_map), \ 3726174a79ffSJohn Fastabend FN(sock_map_update), \ 3727908432caSYonghong Song FN(xdp_adjust_meta), \ 37284bebdc7aSYonghong Song FN(perf_event_read_value), \ 3729cd86d1fdSLawrence Brakmo FN(perf_prog_read_value), \ 37309802d865SJosef Bacik FN(getsockopt), \ 3731b13d8807SLawrence Brakmo FN(override_return), \ 37324f738adbSJohn Fastabend FN(sock_ops_cb_flags_set), \ 37332a100317SJohn Fastabend FN(msg_redirect_map), \ 373491843d54SJohn Fastabend FN(msg_apply_bytes), \ 3735015632bbSJohn Fastabend FN(msg_cork_bytes), \ 3736d74bad4eSAndrey Ignatov FN(msg_pull_data), \ 3737b32cc5b9SNikita V. Shirokov FN(bind), \ 373812bed760SEyal Birger FN(xdp_adjust_tail), \ 3739c195651eSYonghong Song FN(skb_get_xfrm_state), \ 37404e1ec56cSDaniel Borkmann FN(get_stack), \ 374187f5fc7eSDavid Ahern FN(skb_load_bytes_relative), \ 374281110384SJohn Fastabend FN(fib_lookup), \ 374381110384SJohn Fastabend FN(sock_hash_update), \ 374481110384SJohn Fastabend FN(msg_redirect_hash), \ 3745fe94cc29SMathieu Xhonneux FN(sk_redirect_hash), \ 3746fe94cc29SMathieu Xhonneux FN(lwt_push_encap), \ 3747fe94cc29SMathieu Xhonneux FN(lwt_seg6_store_bytes), \ 3748fe94cc29SMathieu Xhonneux FN(lwt_seg6_adjust_srh), \ 3749f4364dcfSSean Young FN(lwt_seg6_action), \ 3750f4364dcfSSean Young FN(rc_repeat), \ 3751cb20b08eSDaniel Borkmann FN(rc_keydown), \ 3752bf6fa2c8SYonghong Song FN(skb_cgroup_id), \ 3753cd339431SRoman Gushchin FN(get_current_cgroup_id), \ 37542dbb9b9eSMartin KaFai Lau FN(get_local_storage), \ 375577236281SAndrey Ignatov FN(sk_select_reuseport), \ 37566acc9b43SJoe Stringer FN(skb_ancestor_cgroup_id), \ 37576acc9b43SJoe Stringer FN(sk_lookup_tcp), \ 37586acc9b43SJoe Stringer FN(sk_lookup_udp), \ 3759f1a2e44aSMauricio Vasquez B FN(sk_release), \ 3760f1a2e44aSMauricio Vasquez B FN(map_push_elem), \ 3761f1a2e44aSMauricio Vasquez B FN(map_pop_elem), \ 37626fff607eSJohn Fastabend FN(map_peek_elem), \ 37637246d8edSJohn Fastabend FN(msg_push_data), \ 376401d3240aSSean Young FN(msg_pop_data), \ 3765d83525caSAlexei Starovoitov FN(rc_pointer_rel), \ 3766d83525caSAlexei Starovoitov FN(spin_lock), \ 376746f8bc92SMartin KaFai Lau FN(spin_unlock), \ 3768655a51e5SMartin KaFai Lau FN(sk_fullsock), \ 3769f7c917baSbrakmo FN(tcp_sock), \ 3770dbafd7ddSMartin KaFai Lau FN(skb_ecn_set_ce), \ 3771edbf8c01SLorenz Bauer FN(get_listener_sock), \ 377239904084SLorenz Bauer FN(skc_lookup_tcp), \ 3773808649fbSAndrey Ignatov FN(tcp_check_syncookie), \ 37741d11b301SAndrey Ignatov FN(sysctl_get_name), \ 37754e63acdfSAndrey Ignatov FN(sysctl_get_current_value), \ 37764e63acdfSAndrey Ignatov FN(sysctl_get_new_value), \ 3777d7a4cb9bSAndrey Ignatov FN(sysctl_set_new_value), \ 3778d7a4cb9bSAndrey Ignatov FN(strtol), \ 37796ac99e8fSMartin KaFai Lau FN(strtoul), \ 37806ac99e8fSMartin KaFai Lau FN(sk_storage_get), \ 37818b401f9eSYonghong Song FN(sk_storage_delete), \ 378270d66244SPetar Penkov FN(send_signal), \ 3783a7658e1aSAlexei Starovoitov FN(tcp_gen_syncookie), \ 37846ae08ae3SDaniel Borkmann FN(skb_output), \ 37856ae08ae3SDaniel Borkmann FN(probe_read_user), \ 37866ae08ae3SDaniel Borkmann FN(probe_read_kernel), \ 37876ae08ae3SDaniel Borkmann FN(probe_read_user_str), \ 3788206057feSMartin KaFai Lau FN(probe_read_kernel_str), \ 37898482941fSYonghong Song FN(tcp_send_ack), \ 37905576b991SMartin KaFai Lau FN(send_signal_thread), \ 3791fff7b643SDaniel Xu FN(jiffies64), \ 3792b4490c5cSCarlos Neira FN(read_branch_records), \ 3793d831ee84SEelco Chaudron FN(get_ns_current_pid_tgid), \ 3794f318903cSDaniel Borkmann FN(xdp_output), \ 37950f09abd1SDaniel Borkmann FN(get_netns_cookie), \ 3796cf7fbe66SJoe Stringer FN(get_current_ancestor_cgroup_id), \ 379771d19214SMaciej Żenczykowski FN(sk_assign), \ 3798492e639fSYonghong Song FN(ktime_get_boot_ns), \ 3799492e639fSYonghong Song FN(seq_printf), \ 3800f307fa2cSAndrey Ignatov FN(seq_write), \ 3801f307fa2cSAndrey Ignatov FN(sk_cgroup_id), \ 3802457f4436SAndrii Nakryiko FN(sk_ancestor_cgroup_id), \ 3803457f4436SAndrii Nakryiko FN(ringbuf_output), \ 3804457f4436SAndrii Nakryiko FN(ringbuf_reserve), \ 3805457f4436SAndrii Nakryiko FN(ringbuf_submit), \ 3806457f4436SAndrii Nakryiko FN(ringbuf_discard), \ 38077cdec54fSDaniel Borkmann FN(ringbuf_query), \ 3808af7ec138SYonghong Song FN(csum_level), \ 3809478cfbdfSYonghong Song FN(skc_to_tcp6_sock), \ 3810478cfbdfSYonghong Song FN(skc_to_tcp_sock), \ 3811478cfbdfSYonghong Song FN(skc_to_tcp_timewait_sock), \ 38120d4fad3eSYonghong Song FN(skc_to_tcp_request_sock), \ 3813fa28dcb8SSong Liu FN(skc_to_udp6_sock), \ 3814fa28dcb8SSong Liu FN(get_task_stack), \ 38150813a841SMartin KaFai Lau FN(load_hdr_opt), \ 38160813a841SMartin KaFai Lau FN(store_hdr_opt), \ 38178ea63684SKP Singh FN(reserve_hdr_opt), \ 38188ea63684SKP Singh FN(inode_storage_get), \ 38198ea63684SKP Singh FN(inode_storage_delete), \ 38206e22ab9dSJiri Olsa FN(d_path), \ 382107be4c4aSAlexei Starovoitov FN(copy_from_user), \ 3822c4d0bfb4SAlan Maguire FN(snprintf_btf), \ 3823eb411377SAlan Maguire FN(seq_printf_btf), \ 3824b426ce83SDaniel Borkmann FN(skb_cgroup_classid), \ 3825b4ab3141SDaniel Borkmann FN(redirect_neigh), \ 3826fa28dcb8SSong Liu /* */ 38277a4b28c6SDaniel Borkmann 3828ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 3829ebb676daSThomas Graf * function eBPF program intends to call 38302d0e30c3SDaniel Borkmann */ 3831ebb676daSThomas Graf #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 3832ebb676daSThomas Graf enum bpf_func_id { 3833ebb676daSThomas Graf __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 383409756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 383509756af4SAlexei Starovoitov }; 3836ebb676daSThomas Graf #undef __BPF_ENUM_FN 383709756af4SAlexei Starovoitov 3838781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 3839781c53bcSDaniel Borkmann 3840781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 38411aae4bddSAndrii Nakryiko enum { 38421aae4bddSAndrii Nakryiko BPF_F_RECOMPUTE_CSUM = (1ULL << 0), 38431aae4bddSAndrii Nakryiko BPF_F_INVALIDATE_HASH = (1ULL << 1), 38441aae4bddSAndrii Nakryiko }; 3845781c53bcSDaniel Borkmann 3846781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 3847781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 3848781c53bcSDaniel Borkmann */ 38491aae4bddSAndrii Nakryiko enum { 38501aae4bddSAndrii Nakryiko BPF_F_HDR_FIELD_MASK = 0xfULL, 38511aae4bddSAndrii Nakryiko }; 3852781c53bcSDaniel Borkmann 3853781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 38541aae4bddSAndrii Nakryiko enum { 38551aae4bddSAndrii Nakryiko BPF_F_PSEUDO_HDR = (1ULL << 4), 38561aae4bddSAndrii Nakryiko BPF_F_MARK_MANGLED_0 = (1ULL << 5), 38571aae4bddSAndrii Nakryiko BPF_F_MARK_ENFORCE = (1ULL << 6), 38581aae4bddSAndrii Nakryiko }; 3859781c53bcSDaniel Borkmann 3860781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 38611aae4bddSAndrii Nakryiko enum { 38621aae4bddSAndrii Nakryiko BPF_F_INGRESS = (1ULL << 0), 38631aae4bddSAndrii Nakryiko }; 3864781c53bcSDaniel Borkmann 3865c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 38661aae4bddSAndrii Nakryiko enum { 38671aae4bddSAndrii Nakryiko BPF_F_TUNINFO_IPV6 = (1ULL << 0), 38681aae4bddSAndrii Nakryiko }; 3869c6c33454SDaniel Borkmann 3870c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 38711aae4bddSAndrii Nakryiko enum { 38721aae4bddSAndrii Nakryiko BPF_F_SKIP_FIELD_MASK = 0xffULL, 38731aae4bddSAndrii Nakryiko BPF_F_USER_STACK = (1ULL << 8), 3874c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 38751aae4bddSAndrii Nakryiko BPF_F_FAST_STACK_CMP = (1ULL << 9), 38761aae4bddSAndrii Nakryiko BPF_F_REUSE_STACKID = (1ULL << 10), 3877c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 38781aae4bddSAndrii Nakryiko BPF_F_USER_BUILD_ID = (1ULL << 11), 38791aae4bddSAndrii Nakryiko }; 3880d5a3b1f6SAlexei Starovoitov 38812da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 38821aae4bddSAndrii Nakryiko enum { 38831aae4bddSAndrii Nakryiko BPF_F_ZERO_CSUM_TX = (1ULL << 1), 38841aae4bddSAndrii Nakryiko BPF_F_DONT_FRAGMENT = (1ULL << 2), 38851aae4bddSAndrii Nakryiko BPF_F_SEQ_NUMBER = (1ULL << 3), 38861aae4bddSAndrii Nakryiko }; 38872da897e5SDaniel Borkmann 3888908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 3889908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 3890908432caSYonghong Song */ 38911aae4bddSAndrii Nakryiko enum { 38921aae4bddSAndrii Nakryiko BPF_F_INDEX_MASK = 0xffffffffULL, 38931aae4bddSAndrii Nakryiko BPF_F_CURRENT_CPU = BPF_F_INDEX_MASK, 3894555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 38951aae4bddSAndrii Nakryiko BPF_F_CTXLEN_MASK = (0xfffffULL << 32), 38961aae4bddSAndrii Nakryiko }; 38971e33759cSDaniel Borkmann 3898f71c6143SJoe Stringer /* Current network namespace */ 38991aae4bddSAndrii Nakryiko enum { 39001aae4bddSAndrii Nakryiko BPF_F_CURRENT_NETNS = (-1L), 39011aae4bddSAndrii Nakryiko }; 3902f71c6143SJoe Stringer 39037cdec54fSDaniel Borkmann /* BPF_FUNC_csum_level level values. */ 39047cdec54fSDaniel Borkmann enum { 39057cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_QUERY, 39067cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_INC, 39077cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_DEC, 39087cdec54fSDaniel Borkmann BPF_CSUM_LEVEL_RESET, 39097cdec54fSDaniel Borkmann }; 39107cdec54fSDaniel Borkmann 39112278f6ccSWillem de Bruijn /* BPF_FUNC_skb_adjust_room flags. */ 39121aae4bddSAndrii Nakryiko enum { 39131aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_FIXED_GSO = (1ULL << 0), 39141aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 = (1ULL << 1), 39151aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 = (1ULL << 2), 39161aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_GRE = (1ULL << 3), 39171aae4bddSAndrii Nakryiko BPF_F_ADJ_ROOM_ENCAP_L4_UDP = (1ULL << 4), 3918836e66c2SDaniel Borkmann BPF_F_ADJ_ROOM_NO_CSUM_RESET = (1ULL << 5), 39191aae4bddSAndrii Nakryiko }; 39202278f6ccSWillem de Bruijn 39211aae4bddSAndrii Nakryiko enum { 39221aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_MASK = 0xff, 39231aae4bddSAndrii Nakryiko BPF_ADJ_ROOM_ENCAP_L2_SHIFT = 56, 39241aae4bddSAndrii Nakryiko }; 392558dfc900SAlan Maguire 392658dfc900SAlan Maguire #define BPF_F_ADJ_ROOM_ENCAP_L2(len) (((__u64)len & \ 392758dfc900SAlan Maguire BPF_ADJ_ROOM_ENCAP_L2_MASK) \ 392858dfc900SAlan Maguire << BPF_ADJ_ROOM_ENCAP_L2_SHIFT) 3929868d5235SWillem de Bruijn 3930808649fbSAndrey Ignatov /* BPF_FUNC_sysctl_get_name flags. */ 39311aae4bddSAndrii Nakryiko enum { 39321aae4bddSAndrii Nakryiko BPF_F_SYSCTL_BASE_NAME = (1ULL << 0), 39331aae4bddSAndrii Nakryiko }; 3934808649fbSAndrey Ignatov 3935f836a56eSKP Singh /* BPF_FUNC_<kernel_obj>_storage_get flags */ 39361aae4bddSAndrii Nakryiko enum { 3937f836a56eSKP Singh BPF_LOCAL_STORAGE_GET_F_CREATE = (1ULL << 0), 3938f836a56eSKP Singh /* BPF_SK_STORAGE_GET_F_CREATE is only kept for backward compatibility 3939f836a56eSKP Singh * and BPF_LOCAL_STORAGE_GET_F_CREATE must be used instead. 3940f836a56eSKP Singh */ 3941f836a56eSKP Singh BPF_SK_STORAGE_GET_F_CREATE = BPF_LOCAL_STORAGE_GET_F_CREATE, 39421aae4bddSAndrii Nakryiko }; 39436ac99e8fSMartin KaFai Lau 3944fff7b643SDaniel Xu /* BPF_FUNC_read_branch_records flags. */ 39451aae4bddSAndrii Nakryiko enum { 39461aae4bddSAndrii Nakryiko BPF_F_GET_BRANCH_RECORDS_SIZE = (1ULL << 0), 39471aae4bddSAndrii Nakryiko }; 3948fff7b643SDaniel Xu 3949457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_commit, BPF_FUNC_bpf_ringbuf_discard, and 3950457f4436SAndrii Nakryiko * BPF_FUNC_bpf_ringbuf_output flags. 3951457f4436SAndrii Nakryiko */ 3952457f4436SAndrii Nakryiko enum { 3953457f4436SAndrii Nakryiko BPF_RB_NO_WAKEUP = (1ULL << 0), 3954457f4436SAndrii Nakryiko BPF_RB_FORCE_WAKEUP = (1ULL << 1), 3955457f4436SAndrii Nakryiko }; 3956457f4436SAndrii Nakryiko 3957457f4436SAndrii Nakryiko /* BPF_FUNC_bpf_ringbuf_query flags */ 3958457f4436SAndrii Nakryiko enum { 3959457f4436SAndrii Nakryiko BPF_RB_AVAIL_DATA = 0, 3960457f4436SAndrii Nakryiko BPF_RB_RING_SIZE = 1, 3961457f4436SAndrii Nakryiko BPF_RB_CONS_POS = 2, 3962457f4436SAndrii Nakryiko BPF_RB_PROD_POS = 3, 3963457f4436SAndrii Nakryiko }; 3964457f4436SAndrii Nakryiko 3965457f4436SAndrii Nakryiko /* BPF ring buffer constants */ 3966457f4436SAndrii Nakryiko enum { 3967457f4436SAndrii Nakryiko BPF_RINGBUF_BUSY_BIT = (1U << 31), 3968457f4436SAndrii Nakryiko BPF_RINGBUF_DISCARD_BIT = (1U << 30), 3969457f4436SAndrii Nakryiko BPF_RINGBUF_HDR_SZ = 8, 3970457f4436SAndrii Nakryiko }; 3971457f4436SAndrii Nakryiko 3972e9ddbb77SJakub Sitnicki /* BPF_FUNC_sk_assign flags in bpf_sk_lookup context. */ 3973e9ddbb77SJakub Sitnicki enum { 3974e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_REPLACE = (1ULL << 0), 3975e9ddbb77SJakub Sitnicki BPF_SK_LOOKUP_F_NO_REUSEPORT = (1ULL << 1), 3976e9ddbb77SJakub Sitnicki }; 3977e9ddbb77SJakub Sitnicki 39782be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 39792be7e212SDaniel Borkmann enum bpf_adj_room_mode { 39802be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 398114aa3192SWillem de Bruijn BPF_ADJ_ROOM_MAC, 39822be7e212SDaniel Borkmann }; 39832be7e212SDaniel Borkmann 39844e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 39854e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 39864e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 39874e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 39884e1ec56cSDaniel Borkmann }; 39894e1ec56cSDaniel Borkmann 3990fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 3991fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 3992fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 39933e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_SEG6_INLINE, 39943e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_IP, 3995fe94cc29SMathieu Xhonneux }; 3996fe94cc29SMathieu Xhonneux 3997b7df9adaSDaniel Borkmann #define __bpf_md_ptr(type, name) \ 3998b7df9adaSDaniel Borkmann union { \ 3999b7df9adaSDaniel Borkmann type name; \ 4000b7df9adaSDaniel Borkmann __u64 :64; \ 4001b7df9adaSDaniel Borkmann } __attribute__((aligned(8))) 4002b7df9adaSDaniel Borkmann 40039bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 40049bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 40059bac3d6dSAlexei Starovoitov */ 40069bac3d6dSAlexei Starovoitov struct __sk_buff { 40079bac3d6dSAlexei Starovoitov __u32 len; 40089bac3d6dSAlexei Starovoitov __u32 pkt_type; 40099bac3d6dSAlexei Starovoitov __u32 mark; 40109bac3d6dSAlexei Starovoitov __u32 queue_mapping; 4011c2497395SAlexei Starovoitov __u32 protocol; 4012c2497395SAlexei Starovoitov __u32 vlan_present; 4013c2497395SAlexei Starovoitov __u32 vlan_tci; 401427cd5452SMichal Sekletar __u32 vlan_proto; 4015bcad5718SDaniel Borkmann __u32 priority; 401637e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 401737e82c2fSAlexei Starovoitov __u32 ifindex; 4018d691f9e8SAlexei Starovoitov __u32 tc_index; 4019d691f9e8SAlexei Starovoitov __u32 cb[5]; 4020ba7591d8SDaniel Borkmann __u32 hash; 4021045efa82SDaniel Borkmann __u32 tc_classid; 4022969bf05eSAlexei Starovoitov __u32 data; 4023969bf05eSAlexei Starovoitov __u32 data_end; 4024b1d9fc41SDaniel Borkmann __u32 napi_id; 40258a31db56SJohn Fastabend 4026de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 40278a31db56SJohn Fastabend __u32 family; 40288a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 40298a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 40308a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 40318a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 40328a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 40338a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 4034de8f3a83SDaniel Borkmann /* ... here. */ 4035de8f3a83SDaniel Borkmann 4036de8f3a83SDaniel Borkmann __u32 data_meta; 4037b7df9adaSDaniel Borkmann __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 4038f11216b2SVlad Dumitrescu __u64 tstamp; 4039e3da08d0SPetar Penkov __u32 wire_len; 4040d9ff286aSEric Dumazet __u32 gso_segs; 404146f8bc92SMartin KaFai Lau __bpf_md_ptr(struct bpf_sock *, sk); 4042cf62089bSWillem de Bruijn __u32 gso_size; 40439bac3d6dSAlexei Starovoitov }; 40449bac3d6dSAlexei Starovoitov 4045d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 4046d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 4047c6c33454SDaniel Borkmann union { 4048d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 4049c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 4050c6c33454SDaniel Borkmann }; 4051c6c33454SDaniel Borkmann __u8 tunnel_tos; 4052c6c33454SDaniel Borkmann __u8 tunnel_ttl; 40531fbc2e0cSDaniel Borkmann __u16 tunnel_ext; /* Padding, future use. */ 40544018ab18SDaniel Borkmann __u32 tunnel_label; 4055d3aa45ceSAlexei Starovoitov }; 4056d3aa45ceSAlexei Starovoitov 405712bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 405812bed760SEyal Birger * new fields can only be added to the end of this structure 405912bed760SEyal Birger */ 406012bed760SEyal Birger struct bpf_xfrm_state { 406112bed760SEyal Birger __u32 reqid; 406212bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 406312bed760SEyal Birger __u16 family; 40641fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 406512bed760SEyal Birger union { 406612bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 406712bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 406812bed760SEyal Birger }; 406912bed760SEyal Birger }; 407012bed760SEyal Birger 40713a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 40723a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 40733a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 40743a0af8fdSThomas Graf * programs. 40753a0af8fdSThomas Graf * 40763a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 40773a0af8fdSThomas Graf */ 40783a0af8fdSThomas Graf enum bpf_ret_code { 40793a0af8fdSThomas Graf BPF_OK = 0, 40803a0af8fdSThomas Graf /* 1 reserved */ 40813a0af8fdSThomas Graf BPF_DROP = 2, 40823a0af8fdSThomas Graf /* 3-6 reserved */ 40833a0af8fdSThomas Graf BPF_REDIRECT = 7, 40843e0bd37cSPeter Oskolkov /* >127 are reserved for prog type specific return codes. 40853e0bd37cSPeter Oskolkov * 40863e0bd37cSPeter Oskolkov * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 40873e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 40883e0bd37cSPeter Oskolkov * changed and should be routed based on its new L3 header. 40893e0bd37cSPeter Oskolkov * (This is an L3 redirect, as opposed to L2 redirect 40903e0bd37cSPeter Oskolkov * represented by BPF_REDIRECT above). 40913e0bd37cSPeter Oskolkov */ 40923e0bd37cSPeter Oskolkov BPF_LWT_REROUTE = 128, 40933a0af8fdSThomas Graf }; 40943a0af8fdSThomas Graf 409561023658SDavid Ahern struct bpf_sock { 409661023658SDavid Ahern __u32 bound_dev_if; 4097aa4c1037SDavid Ahern __u32 family; 4098aa4c1037SDavid Ahern __u32 type; 4099aa4c1037SDavid Ahern __u32 protocol; 4100482dca93SDavid Ahern __u32 mark; 4101482dca93SDavid Ahern __u32 priority; 4102aa65d696SMartin KaFai Lau /* IP address also allows 1 and 2 bytes access */ 4103aa65d696SMartin KaFai Lau __u32 src_ip4; 4104aa65d696SMartin KaFai Lau __u32 src_ip6[4]; 4105aa65d696SMartin KaFai Lau __u32 src_port; /* host byte order */ 4106aa65d696SMartin KaFai Lau __u32 dst_port; /* network byte order */ 4107aa65d696SMartin KaFai Lau __u32 dst_ip4; 4108aa65d696SMartin KaFai Lau __u32 dst_ip6[4]; 4109aa65d696SMartin KaFai Lau __u32 state; 4110c3c16f2eSAmritha Nambiar __s32 rx_queue_mapping; 411161023658SDavid Ahern }; 411261023658SDavid Ahern 4113655a51e5SMartin KaFai Lau struct bpf_tcp_sock { 4114655a51e5SMartin KaFai Lau __u32 snd_cwnd; /* Sending congestion window */ 4115655a51e5SMartin KaFai Lau __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 4116655a51e5SMartin KaFai Lau __u32 rtt_min; 4117655a51e5SMartin KaFai Lau __u32 snd_ssthresh; /* Slow start size threshold */ 4118655a51e5SMartin KaFai Lau __u32 rcv_nxt; /* What we want to receive next */ 4119655a51e5SMartin KaFai Lau __u32 snd_nxt; /* Next sequence we send */ 4120655a51e5SMartin KaFai Lau __u32 snd_una; /* First byte we want an ack for */ 4121655a51e5SMartin KaFai Lau __u32 mss_cache; /* Cached effective mss, not including SACKS */ 4122655a51e5SMartin KaFai Lau __u32 ecn_flags; /* ECN status bits. */ 4123655a51e5SMartin KaFai Lau __u32 rate_delivered; /* saved rate sample: packets delivered */ 4124655a51e5SMartin KaFai Lau __u32 rate_interval_us; /* saved rate sample: time elapsed */ 4125655a51e5SMartin KaFai Lau __u32 packets_out; /* Packets which are "in flight" */ 4126655a51e5SMartin KaFai Lau __u32 retrans_out; /* Retransmitted packets out */ 4127655a51e5SMartin KaFai Lau __u32 total_retrans; /* Total retransmits for entire connection */ 4128655a51e5SMartin KaFai Lau __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 4129655a51e5SMartin KaFai Lau * total number of segments in. 4130655a51e5SMartin KaFai Lau */ 4131655a51e5SMartin KaFai Lau __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 4132655a51e5SMartin KaFai Lau * total number of data segments in. 4133655a51e5SMartin KaFai Lau */ 4134655a51e5SMartin KaFai Lau __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 4135655a51e5SMartin KaFai Lau * The total number of segments sent. 4136655a51e5SMartin KaFai Lau */ 4137655a51e5SMartin KaFai Lau __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 4138655a51e5SMartin KaFai Lau * total number of data segments sent. 4139655a51e5SMartin KaFai Lau */ 4140655a51e5SMartin KaFai Lau __u32 lost_out; /* Lost packets */ 4141655a51e5SMartin KaFai Lau __u32 sacked_out; /* SACK'd packets */ 4142655a51e5SMartin KaFai Lau __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 4143655a51e5SMartin KaFai Lau * sum(delta(rcv_nxt)), or how many bytes 4144655a51e5SMartin KaFai Lau * were acked. 4145655a51e5SMartin KaFai Lau */ 4146655a51e5SMartin KaFai Lau __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 4147655a51e5SMartin KaFai Lau * sum(delta(snd_una)), or how many bytes 4148655a51e5SMartin KaFai Lau * were acked. 4149655a51e5SMartin KaFai Lau */ 41500357746dSStanislav Fomichev __u32 dsack_dups; /* RFC4898 tcpEStatsStackDSACKDups 41510357746dSStanislav Fomichev * total number of DSACK blocks received 41520357746dSStanislav Fomichev */ 41530357746dSStanislav Fomichev __u32 delivered; /* Total data packets delivered incl. rexmits */ 41540357746dSStanislav Fomichev __u32 delivered_ce; /* Like the above but only ECE marked packets */ 4155c2cb5e82SStanislav Fomichev __u32 icsk_retransmits; /* Number of unrecovered [RTO] timeouts */ 4156655a51e5SMartin KaFai Lau }; 4157655a51e5SMartin KaFai Lau 41586acc9b43SJoe Stringer struct bpf_sock_tuple { 41596acc9b43SJoe Stringer union { 41606acc9b43SJoe Stringer struct { 41616acc9b43SJoe Stringer __be32 saddr; 41626acc9b43SJoe Stringer __be32 daddr; 41636acc9b43SJoe Stringer __be16 sport; 41646acc9b43SJoe Stringer __be16 dport; 41656acc9b43SJoe Stringer } ipv4; 41666acc9b43SJoe Stringer struct { 41676acc9b43SJoe Stringer __be32 saddr[4]; 41686acc9b43SJoe Stringer __be32 daddr[4]; 41696acc9b43SJoe Stringer __be16 sport; 41706acc9b43SJoe Stringer __be16 dport; 41716acc9b43SJoe Stringer } ipv6; 41726acc9b43SJoe Stringer }; 41736acc9b43SJoe Stringer }; 41746acc9b43SJoe Stringer 4175fada7fdcSJonathan Lemon struct bpf_xdp_sock { 4176fada7fdcSJonathan Lemon __u32 queue_id; 4177fada7fdcSJonathan Lemon }; 4178fada7fdcSJonathan Lemon 417917bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 418017bedab2SMartin KaFai Lau 41816a773a15SBrenden Blanco /* User return codes for XDP prog type. 41826a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 41839beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 41849beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 41856a773a15SBrenden Blanco */ 41866a773a15SBrenden Blanco enum xdp_action { 41876a773a15SBrenden Blanco XDP_ABORTED = 0, 41886a773a15SBrenden Blanco XDP_DROP, 41896a773a15SBrenden Blanco XDP_PASS, 41906ce96ca3SBrenden Blanco XDP_TX, 4191814abfabSJohn Fastabend XDP_REDIRECT, 41926a773a15SBrenden Blanco }; 41936a773a15SBrenden Blanco 41946a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 41956a773a15SBrenden Blanco * new fields must be added to the end of this structure 41966a773a15SBrenden Blanco */ 41976a773a15SBrenden Blanco struct xdp_md { 41986a773a15SBrenden Blanco __u32 data; 41996a773a15SBrenden Blanco __u32 data_end; 4200de8f3a83SDaniel Borkmann __u32 data_meta; 4201daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 420202dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 420302dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 420464b59025SDavid Ahern 420564b59025SDavid Ahern __u32 egress_ifindex; /* txq->dev->ifindex */ 42066a773a15SBrenden Blanco }; 42076a773a15SBrenden Blanco 4208281920b7SJesper Dangaard Brouer /* DEVMAP map-value layout 4209281920b7SJesper Dangaard Brouer * 4210281920b7SJesper Dangaard Brouer * The struct data-layout of map-value is a configuration interface. 4211281920b7SJesper Dangaard Brouer * New members can only be added to the end of this structure. 4212281920b7SJesper Dangaard Brouer */ 4213281920b7SJesper Dangaard Brouer struct bpf_devmap_val { 4214281920b7SJesper Dangaard Brouer __u32 ifindex; /* device index */ 4215281920b7SJesper Dangaard Brouer union { 4216281920b7SJesper Dangaard Brouer int fd; /* prog fd on map write */ 4217281920b7SJesper Dangaard Brouer __u32 id; /* prog id on map read */ 4218281920b7SJesper Dangaard Brouer } bpf_prog; 4219281920b7SJesper Dangaard Brouer }; 4220281920b7SJesper Dangaard Brouer 4221644bfe51SLorenzo Bianconi /* CPUMAP map-value layout 4222644bfe51SLorenzo Bianconi * 4223644bfe51SLorenzo Bianconi * The struct data-layout of map-value is a configuration interface. 4224644bfe51SLorenzo Bianconi * New members can only be added to the end of this structure. 4225644bfe51SLorenzo Bianconi */ 4226644bfe51SLorenzo Bianconi struct bpf_cpumap_val { 4227644bfe51SLorenzo Bianconi __u32 qsize; /* queue size to remote target CPU */ 422892164774SLorenzo Bianconi union { 422992164774SLorenzo Bianconi int fd; /* prog fd on map write */ 423092164774SLorenzo Bianconi __u32 id; /* prog id on map read */ 423192164774SLorenzo Bianconi } bpf_prog; 4232644bfe51SLorenzo Bianconi }; 4233644bfe51SLorenzo Bianconi 4234174a79ffSJohn Fastabend enum sk_action { 4235bfa64075SJohn Fastabend SK_DROP = 0, 4236bfa64075SJohn Fastabend SK_PASS, 4237174a79ffSJohn Fastabend }; 4238174a79ffSJohn Fastabend 42394f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 42404f738adbSJohn Fastabend * be added to the end of this structure 42414f738adbSJohn Fastabend */ 42424f738adbSJohn Fastabend struct sk_msg_md { 4243b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 4244b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 4245303def35SJohn Fastabend 4246303def35SJohn Fastabend __u32 family; 4247303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 4248303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 4249303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 4250303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 4251303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 4252303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 42533bdbd022SJohn Fastabend __u32 size; /* Total size of sk_msg */ 425413d70f5aSJohn Fastabend 425513d70f5aSJohn Fastabend __bpf_md_ptr(struct bpf_sock *, sk); /* current socket */ 42564f738adbSJohn Fastabend }; 42574f738adbSJohn Fastabend 42582dbb9b9eSMartin KaFai Lau struct sk_reuseport_md { 42592dbb9b9eSMartin KaFai Lau /* 42602dbb9b9eSMartin KaFai Lau * Start of directly accessible data. It begins from 42612dbb9b9eSMartin KaFai Lau * the tcp/udp header. 42622dbb9b9eSMartin KaFai Lau */ 4263b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 4264b7df9adaSDaniel Borkmann /* End of directly accessible data */ 4265b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 42662dbb9b9eSMartin KaFai Lau /* 42672dbb9b9eSMartin KaFai Lau * Total length of packet (starting from the tcp/udp header). 42682dbb9b9eSMartin KaFai Lau * Note that the directly accessible bytes (data_end - data) 42692dbb9b9eSMartin KaFai Lau * could be less than this "len". Those bytes could be 42702dbb9b9eSMartin KaFai Lau * indirectly read by a helper "bpf_skb_load_bytes()". 42712dbb9b9eSMartin KaFai Lau */ 42722dbb9b9eSMartin KaFai Lau __u32 len; 42732dbb9b9eSMartin KaFai Lau /* 42742dbb9b9eSMartin KaFai Lau * Eth protocol in the mac header (network byte order). e.g. 42752dbb9b9eSMartin KaFai Lau * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 42762dbb9b9eSMartin KaFai Lau */ 42772dbb9b9eSMartin KaFai Lau __u32 eth_protocol; 42782dbb9b9eSMartin KaFai Lau __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 42792dbb9b9eSMartin KaFai Lau __u32 bind_inany; /* Is sock bound to an INANY address? */ 42802dbb9b9eSMartin KaFai Lau __u32 hash; /* A hash of the packet 4 tuples */ 42812dbb9b9eSMartin KaFai Lau }; 42822dbb9b9eSMartin KaFai Lau 42831e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 42841e270976SMartin KaFai Lau 42851e270976SMartin KaFai Lau struct bpf_prog_info { 42861e270976SMartin KaFai Lau __u32 type; 42871e270976SMartin KaFai Lau __u32 id; 42881e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 42891e270976SMartin KaFai Lau __u32 jited_prog_len; 42901e270976SMartin KaFai Lau __u32 xlated_prog_len; 42911e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 42921e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 4293cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 4294cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 4295cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 4296cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 4297067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 4298675fc275SJakub Kicinski __u32 ifindex; 4299b85fab0eSJiri Olsa __u32 gpl_compatible:1; 43000472301aSBaruch Siach __u32 :31; /* alignment pad */ 4301675fc275SJakub Kicinski __u64 netns_dev; 4302675fc275SJakub Kicinski __u64 netns_ino; 4303dbecd738SSandipan Das __u32 nr_jited_ksyms; 4304815581c1SSandipan Das __u32 nr_jited_func_lens; 4305dbecd738SSandipan Das __aligned_u64 jited_ksyms; 4306815581c1SSandipan Das __aligned_u64 jited_func_lens; 4307838e9690SYonghong Song __u32 btf_id; 4308838e9690SYonghong Song __u32 func_info_rec_size; 4309838e9690SYonghong Song __aligned_u64 func_info; 431011d8b82dSYonghong Song __u32 nr_func_info; 431111d8b82dSYonghong Song __u32 nr_line_info; 4312c454a46bSMartin KaFai Lau __aligned_u64 line_info; 4313c454a46bSMartin KaFai Lau __aligned_u64 jited_line_info; 431411d8b82dSYonghong Song __u32 nr_jited_line_info; 4315c454a46bSMartin KaFai Lau __u32 line_info_rec_size; 4316c454a46bSMartin KaFai Lau __u32 jited_line_info_rec_size; 4317c872bdb3SSong Liu __u32 nr_prog_tags; 4318c872bdb3SSong Liu __aligned_u64 prog_tags; 43195f8f8b93SAlexei Starovoitov __u64 run_time_ns; 43205f8f8b93SAlexei Starovoitov __u64 run_cnt; 43211e270976SMartin KaFai Lau } __attribute__((aligned(8))); 43221e270976SMartin KaFai Lau 43231e270976SMartin KaFai Lau struct bpf_map_info { 43241e270976SMartin KaFai Lau __u32 type; 43251e270976SMartin KaFai Lau __u32 id; 43261e270976SMartin KaFai Lau __u32 key_size; 43271e270976SMartin KaFai Lau __u32 value_size; 43281e270976SMartin KaFai Lau __u32 max_entries; 43291e270976SMartin KaFai Lau __u32 map_flags; 4330067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 433152775b33SJakub Kicinski __u32 ifindex; 433285d33df3SMartin KaFai Lau __u32 btf_vmlinux_value_type_id; 433352775b33SJakub Kicinski __u64 netns_dev; 433452775b33SJakub Kicinski __u64 netns_ino; 433578958fcaSMartin KaFai Lau __u32 btf_id; 43369b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 43379b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 43381e270976SMartin KaFai Lau } __attribute__((aligned(8))); 43391e270976SMartin KaFai Lau 434062dab84cSMartin KaFai Lau struct bpf_btf_info { 434162dab84cSMartin KaFai Lau __aligned_u64 btf; 434262dab84cSMartin KaFai Lau __u32 btf_size; 434362dab84cSMartin KaFai Lau __u32 id; 434462dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 434562dab84cSMartin KaFai Lau 4346f2e10bffSAndrii Nakryiko struct bpf_link_info { 4347f2e10bffSAndrii Nakryiko __u32 type; 4348f2e10bffSAndrii Nakryiko __u32 id; 4349f2e10bffSAndrii Nakryiko __u32 prog_id; 4350f2e10bffSAndrii Nakryiko union { 4351f2e10bffSAndrii Nakryiko struct { 4352f2e10bffSAndrii Nakryiko __aligned_u64 tp_name; /* in/out: tp_name buffer ptr */ 4353f2e10bffSAndrii Nakryiko __u32 tp_name_len; /* in/out: tp_name buffer len */ 4354f2e10bffSAndrii Nakryiko } raw_tracepoint; 4355f2e10bffSAndrii Nakryiko struct { 4356f2e10bffSAndrii Nakryiko __u32 attach_type; 4357f2e10bffSAndrii Nakryiko } tracing; 4358f2e10bffSAndrii Nakryiko struct { 4359f2e10bffSAndrii Nakryiko __u64 cgroup_id; 4360f2e10bffSAndrii Nakryiko __u32 attach_type; 4361f2e10bffSAndrii Nakryiko } cgroup; 43627f045a49SJakub Sitnicki struct { 43636b0a249aSYonghong Song __aligned_u64 target_name; /* in/out: target_name buffer ptr */ 43646b0a249aSYonghong Song __u32 target_name_len; /* in/out: target_name buffer len */ 43656b0a249aSYonghong Song union { 4366b0c9eb37SYonghong Song struct { 43676b0a249aSYonghong Song __u32 map_id; 43686b0a249aSYonghong Song } map; 4369b0c9eb37SYonghong Song }; 43706b0a249aSYonghong Song } iter; 43716b0a249aSYonghong Song struct { 43727f045a49SJakub Sitnicki __u32 netns_ino; 43737f045a49SJakub Sitnicki __u32 attach_type; 43747f045a49SJakub Sitnicki } netns; 4375c1931c97SAndrii Nakryiko struct { 4376c1931c97SAndrii Nakryiko __u32 ifindex; 4377c1931c97SAndrii Nakryiko } xdp; 4378f2e10bffSAndrii Nakryiko }; 4379f2e10bffSAndrii Nakryiko } __attribute__((aligned(8))); 4380f2e10bffSAndrii Nakryiko 43814fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 43824fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 4383bfdfa517SRandy Dunlap * attach type). 43844fbac77dSAndrey Ignatov */ 43854fbac77dSAndrey Ignatov struct bpf_sock_addr { 43864fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 43874fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 43884fbac77dSAndrey Ignatov * Stored in network byte order. 43894fbac77dSAndrey Ignatov */ 4390d4ecfeb1SStanislav Fomichev __u32 user_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 43914fbac77dSAndrey Ignatov * Stored in network byte order. 43924fbac77dSAndrey Ignatov */ 43937aebfa1bSAndrey Ignatov __u32 user_port; /* Allows 1,2,4-byte read and 4-byte write. 43944fbac77dSAndrey Ignatov * Stored in network byte order 43954fbac77dSAndrey Ignatov */ 43964fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 43974fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 43984fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 4399600c70baSStanislav Fomichev __u32 msg_src_ip4; /* Allows 1,2,4-byte read and 4-byte write. 44001cedee13SAndrey Ignatov * Stored in network byte order. 44011cedee13SAndrey Ignatov */ 4402d4ecfeb1SStanislav Fomichev __u32 msg_src_ip6[4]; /* Allows 1,2,4,8-byte read and 4,8-byte write. 44031cedee13SAndrey Ignatov * Stored in network byte order. 44041cedee13SAndrey Ignatov */ 4405fb85c4a7SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 44064fbac77dSAndrey Ignatov }; 44074fbac77dSAndrey Ignatov 440840304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 440940304b2aSLawrence Brakmo * and their replies. 441040304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 441140304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 441240304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 441340304b2aSLawrence Brakmo */ 441440304b2aSLawrence Brakmo struct bpf_sock_ops { 441540304b2aSLawrence Brakmo __u32 op; 441640304b2aSLawrence Brakmo union { 4417de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 4418de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 4419de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 442040304b2aSLawrence Brakmo }; 442140304b2aSLawrence Brakmo __u32 family; 442240304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 442340304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 442440304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 442540304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 442640304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 442740304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 4428f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 4429f19397a5SLawrence Brakmo * there is a full socket. If not, the 4430f19397a5SLawrence Brakmo * fields read as zero. 4431f19397a5SLawrence Brakmo */ 4432f19397a5SLawrence Brakmo __u32 snd_cwnd; 4433f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 4434b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 443544f0e430SLawrence Brakmo __u32 state; 443644f0e430SLawrence Brakmo __u32 rtt_min; 443744f0e430SLawrence Brakmo __u32 snd_ssthresh; 443844f0e430SLawrence Brakmo __u32 rcv_nxt; 443944f0e430SLawrence Brakmo __u32 snd_nxt; 444044f0e430SLawrence Brakmo __u32 snd_una; 444144f0e430SLawrence Brakmo __u32 mss_cache; 444244f0e430SLawrence Brakmo __u32 ecn_flags; 444344f0e430SLawrence Brakmo __u32 rate_delivered; 444444f0e430SLawrence Brakmo __u32 rate_interval_us; 444544f0e430SLawrence Brakmo __u32 packets_out; 444644f0e430SLawrence Brakmo __u32 retrans_out; 444744f0e430SLawrence Brakmo __u32 total_retrans; 444844f0e430SLawrence Brakmo __u32 segs_in; 444944f0e430SLawrence Brakmo __u32 data_segs_in; 445044f0e430SLawrence Brakmo __u32 segs_out; 445144f0e430SLawrence Brakmo __u32 data_segs_out; 445244f0e430SLawrence Brakmo __u32 lost_out; 445344f0e430SLawrence Brakmo __u32 sacked_out; 445444f0e430SLawrence Brakmo __u32 sk_txhash; 445544f0e430SLawrence Brakmo __u64 bytes_received; 445644f0e430SLawrence Brakmo __u64 bytes_acked; 44571314ef56SStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 44580813a841SMartin KaFai Lau /* [skb_data, skb_data_end) covers the whole TCP header. 44590813a841SMartin KaFai Lau * 44600813a841SMartin KaFai Lau * BPF_SOCK_OPS_PARSE_HDR_OPT_CB: The packet received 44610813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN_CB: Not useful because the 44620813a841SMartin KaFai Lau * header has not been written. 44630813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB: The header and options have 44640813a841SMartin KaFai Lau * been written so far. 44650813a841SMartin KaFai Lau * BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB: The SYNACK that concludes 44660813a841SMartin KaFai Lau * the 3WHS. 44670813a841SMartin KaFai Lau * BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB: The ACK that concludes 44680813a841SMartin KaFai Lau * the 3WHS. 44690813a841SMartin KaFai Lau * 44700813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also be used to read a particular option. 44710813a841SMartin KaFai Lau */ 44720813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data); 44730813a841SMartin KaFai Lau __bpf_md_ptr(void *, skb_data_end); 44740813a841SMartin KaFai Lau __u32 skb_len; /* The total length of a packet. 44750813a841SMartin KaFai Lau * It includes the header, options, 44760813a841SMartin KaFai Lau * and payload. 44770813a841SMartin KaFai Lau */ 44780813a841SMartin KaFai Lau __u32 skb_tcp_flags; /* tcp_flags of the header. It provides 44790813a841SMartin KaFai Lau * an easy way to check for tcp_flags 44800813a841SMartin KaFai Lau * without parsing skb_data. 44810813a841SMartin KaFai Lau * 44820813a841SMartin KaFai Lau * In particular, the skb_tcp_flags 44830813a841SMartin KaFai Lau * will still be available in 44840813a841SMartin KaFai Lau * BPF_SOCK_OPS_HDR_OPT_LEN even though 44850813a841SMartin KaFai Lau * the outgoing header has not 44860813a841SMartin KaFai Lau * been written yet. 44870813a841SMartin KaFai Lau */ 448840304b2aSLawrence Brakmo }; 448940304b2aSLawrence Brakmo 4490b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 44911aae4bddSAndrii Nakryiko enum { 44921aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTO_CB_FLAG = (1<<0), 44931aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RETRANS_CB_FLAG = (1<<1), 44941aae4bddSAndrii Nakryiko BPF_SOCK_OPS_STATE_CB_FLAG = (1<<2), 44951aae4bddSAndrii Nakryiko BPF_SOCK_OPS_RTT_CB_FLAG = (1<<3), 44960813a841SMartin KaFai Lau /* Call bpf for all received TCP headers. The bpf prog will be 44970813a841SMartin KaFai Lau * called under sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB 44980813a841SMartin KaFai Lau * 44990813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 45000813a841SMartin KaFai Lau * for the header option related helpers that will be useful 45010813a841SMartin KaFai Lau * to the bpf programs. 45020813a841SMartin KaFai Lau * 45030813a841SMartin KaFai Lau * It could be used at the client/active side (i.e. connect() side) 45040813a841SMartin KaFai Lau * when the server told it that the server was in syncookie 45050813a841SMartin KaFai Lau * mode and required the active side to resend the bpf-written 45060813a841SMartin KaFai Lau * options. The active side can keep writing the bpf-options until 45070813a841SMartin KaFai Lau * it received a valid packet from the server side to confirm 45080813a841SMartin KaFai Lau * the earlier packet (and options) has been received. The later 45090813a841SMartin KaFai Lau * example patch is using it like this at the active side when the 45100813a841SMartin KaFai Lau * server is in syncookie mode. 45110813a841SMartin KaFai Lau * 45120813a841SMartin KaFai Lau * The bpf prog will usually turn this off in the common cases. 45130813a841SMartin KaFai Lau */ 451400d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_ALL_HDR_OPT_CB_FLAG = (1<<4), 45150813a841SMartin KaFai Lau /* Call bpf when kernel has received a header option that 45160813a841SMartin KaFai Lau * the kernel cannot handle. The bpf prog will be called under 45170813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_PARSE_HDR_OPT_CB. 45180813a841SMartin KaFai Lau * 45190813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_PARSE_HDR_OPT_CB 45200813a841SMartin KaFai Lau * for the header option related helpers that will be useful 45210813a841SMartin KaFai Lau * to the bpf programs. 45220813a841SMartin KaFai Lau */ 452300d211a4SMartin KaFai Lau BPF_SOCK_OPS_PARSE_UNKNOWN_HDR_OPT_CB_FLAG = (1<<5), 45240813a841SMartin KaFai Lau /* Call bpf when the kernel is writing header options for the 45250813a841SMartin KaFai Lau * outgoing packet. The bpf prog will first be called 45260813a841SMartin KaFai Lau * to reserve space in a skb under 45270813a841SMartin KaFai Lau * sock_ops->op == BPF_SOCK_OPS_HDR_OPT_LEN_CB. Then 45280813a841SMartin KaFai Lau * the bpf prog will be called to write the header option(s) 45290813a841SMartin KaFai Lau * under sock_ops->op == BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 45300813a841SMartin KaFai Lau * 45310813a841SMartin KaFai Lau * Please refer to the comment in BPF_SOCK_OPS_HDR_OPT_LEN_CB 45320813a841SMartin KaFai Lau * and BPF_SOCK_OPS_WRITE_HDR_OPT_CB for the header option 45330813a841SMartin KaFai Lau * related helpers that will be useful to the bpf programs. 45340813a841SMartin KaFai Lau * 45350813a841SMartin KaFai Lau * The kernel gets its chance to reserve space and write 45360813a841SMartin KaFai Lau * options first before the BPF program does. 45370813a841SMartin KaFai Lau */ 4538331fca43SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB_FLAG = (1<<6), 45391aae4bddSAndrii Nakryiko /* Mask of all currently supported cb flags */ 4540331fca43SMartin KaFai Lau BPF_SOCK_OPS_ALL_CB_FLAGS = 0x7F, 45411aae4bddSAndrii Nakryiko }; 4542b13d8807SLawrence Brakmo 454340304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 454440304b2aSLawrence Brakmo * New entries can only be added at the end 454540304b2aSLawrence Brakmo */ 454640304b2aSLawrence Brakmo enum { 454740304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 45488550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 45498550f328SLawrence Brakmo * -1 if default value should be used 45508550f328SLawrence Brakmo */ 455113d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 455213d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 455313d3b1ebSLawrence Brakmo * value should be used 455413d3b1ebSLawrence Brakmo */ 45559872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 45569872a4bdSLawrence Brakmo * active connection is initialized 45579872a4bdSLawrence Brakmo */ 45589872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 45599872a4bdSLawrence Brakmo * active connection is 45609872a4bdSLawrence Brakmo * established 45619872a4bdSLawrence Brakmo */ 45629872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 45639872a4bdSLawrence Brakmo * passive connection is 45649872a4bdSLawrence Brakmo * established 45659872a4bdSLawrence Brakmo */ 456691b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 456791b5b21cSLawrence Brakmo * needs ECN 456891b5b21cSLawrence Brakmo */ 4569e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 4570e6546ef6SLawrence Brakmo * based on the path and may be 4571e6546ef6SLawrence Brakmo * dependent on the congestion control 4572e6546ef6SLawrence Brakmo * algorithm. In general it indicates 4573e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 4574e6546ef6SLawrence Brakmo * this indicate congestion 4575e6546ef6SLawrence Brakmo */ 4576f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 4577f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 4578f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 4579f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 4580f89013f6SLawrence Brakmo */ 4581a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 4582a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 4583a31ad29eSLawrence Brakmo * Arg2: # segments 4584a31ad29eSLawrence Brakmo * Arg3: return value of 4585a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 4586a31ad29eSLawrence Brakmo */ 4587d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 4588d4487491SLawrence Brakmo * Arg1: old_state 4589d4487491SLawrence Brakmo * Arg2: new_state 4590d4487491SLawrence Brakmo */ 4591f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 4592f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 4593f333ee0cSAndrey Ignatov */ 459423729ff2SStanislav Fomichev BPF_SOCK_OPS_RTT_CB, /* Called on every RTT. 459523729ff2SStanislav Fomichev */ 45960813a841SMartin KaFai Lau BPF_SOCK_OPS_PARSE_HDR_OPT_CB, /* Parse the header option. 45970813a841SMartin KaFai Lau * It will be called to handle 45980813a841SMartin KaFai Lau * the packets received at 45990813a841SMartin KaFai Lau * an already established 46000813a841SMartin KaFai Lau * connection. 46010813a841SMartin KaFai Lau * 46020813a841SMartin KaFai Lau * sock_ops->skb_data: 46030813a841SMartin KaFai Lau * Referring to the received skb. 46040813a841SMartin KaFai Lau * It covers the TCP header only. 46050813a841SMartin KaFai Lau * 46060813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 46070813a841SMartin KaFai Lau * be used to search for a 46080813a841SMartin KaFai Lau * particular option. 46090813a841SMartin KaFai Lau */ 46100813a841SMartin KaFai Lau BPF_SOCK_OPS_HDR_OPT_LEN_CB, /* Reserve space for writing the 46110813a841SMartin KaFai Lau * header option later in 46120813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 46130813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 46140813a841SMartin KaFai Lau * writing SYNACK only) 46150813a841SMartin KaFai Lau * 46160813a841SMartin KaFai Lau * sock_ops->skb_data: 46170813a841SMartin KaFai Lau * Not available because no header has 46180813a841SMartin KaFai Lau * been written yet. 46190813a841SMartin KaFai Lau * 46200813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 46210813a841SMartin KaFai Lau * The tcp_flags of the 46220813a841SMartin KaFai Lau * outgoing skb. (e.g. SYN, ACK, FIN). 46230813a841SMartin KaFai Lau * 46240813a841SMartin KaFai Lau * bpf_reserve_hdr_opt() should 46250813a841SMartin KaFai Lau * be used to reserve space. 46260813a841SMartin KaFai Lau */ 46270813a841SMartin KaFai Lau BPF_SOCK_OPS_WRITE_HDR_OPT_CB, /* Write the header options 46280813a841SMartin KaFai Lau * Arg1: bool want_cookie. (in 46290813a841SMartin KaFai Lau * writing SYNACK only) 46300813a841SMartin KaFai Lau * 46310813a841SMartin KaFai Lau * sock_ops->skb_data: 46320813a841SMartin KaFai Lau * Referring to the outgoing skb. 46330813a841SMartin KaFai Lau * It covers the TCP header 46340813a841SMartin KaFai Lau * that has already been written 46350813a841SMartin KaFai Lau * by the kernel and the 46360813a841SMartin KaFai Lau * earlier bpf-progs. 46370813a841SMartin KaFai Lau * 46380813a841SMartin KaFai Lau * sock_ops->skb_tcp_flags: 46390813a841SMartin KaFai Lau * The tcp_flags of the outgoing 46400813a841SMartin KaFai Lau * skb. (e.g. SYN, ACK, FIN). 46410813a841SMartin KaFai Lau * 46420813a841SMartin KaFai Lau * bpf_store_hdr_opt() should 46430813a841SMartin KaFai Lau * be used to write the 46440813a841SMartin KaFai Lau * option. 46450813a841SMartin KaFai Lau * 46460813a841SMartin KaFai Lau * bpf_load_hdr_opt() can also 46470813a841SMartin KaFai Lau * be used to search for a 46480813a841SMartin KaFai Lau * particular option that 46490813a841SMartin KaFai Lau * has already been written 46500813a841SMartin KaFai Lau * by the kernel or the 46510813a841SMartin KaFai Lau * earlier bpf-progs. 46520813a841SMartin KaFai Lau */ 4653d4487491SLawrence Brakmo }; 4654d4487491SLawrence Brakmo 4655d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 4656d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 4657d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 4658d4487491SLawrence Brakmo * the BPF sock_ops function. 4659d4487491SLawrence Brakmo */ 4660d4487491SLawrence Brakmo enum { 4661d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 4662d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 4663d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 4664d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 4665d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 4666d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 4667d4487491SLawrence Brakmo BPF_TCP_CLOSE, 4668d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 4669d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 4670d4487491SLawrence Brakmo BPF_TCP_LISTEN, 4671d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 4672d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 4673d4487491SLawrence Brakmo 4674d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 467540304b2aSLawrence Brakmo }; 467640304b2aSLawrence Brakmo 46771aae4bddSAndrii Nakryiko enum { 46781aae4bddSAndrii Nakryiko TCP_BPF_IW = 1001, /* Set TCP initial congestion window */ 46791aae4bddSAndrii Nakryiko TCP_BPF_SNDCWND_CLAMP = 1002, /* Set sndcwnd_clamp */ 46802b8ee4f0SMartin KaFai Lau TCP_BPF_DELACK_MAX = 1003, /* Max delay ack in usecs */ 4681ca584ba0SMartin KaFai Lau TCP_BPF_RTO_MIN = 1004, /* Min delay ack in usecs */ 46820813a841SMartin KaFai Lau /* Copy the SYN pkt to optval 46830813a841SMartin KaFai Lau * 46840813a841SMartin KaFai Lau * BPF_PROG_TYPE_SOCK_OPS only. It is similar to the 46850813a841SMartin KaFai Lau * bpf_getsockopt(TCP_SAVED_SYN) but it does not limit 46860813a841SMartin KaFai Lau * to only getting from the saved_syn. It can either get the 46870813a841SMartin KaFai Lau * syn packet from: 46880813a841SMartin KaFai Lau * 46890813a841SMartin KaFai Lau * 1. the just-received SYN packet (only available when writing the 46900813a841SMartin KaFai Lau * SYNACK). It will be useful when it is not necessary to 46910813a841SMartin KaFai Lau * save the SYN packet for latter use. It is also the only way 46920813a841SMartin KaFai Lau * to get the SYN during syncookie mode because the syn 46930813a841SMartin KaFai Lau * packet cannot be saved during syncookie. 46940813a841SMartin KaFai Lau * 46950813a841SMartin KaFai Lau * OR 46960813a841SMartin KaFai Lau * 46970813a841SMartin KaFai Lau * 2. the earlier saved syn which was done by 46980813a841SMartin KaFai Lau * bpf_setsockopt(TCP_SAVE_SYN). 46990813a841SMartin KaFai Lau * 47000813a841SMartin KaFai Lau * The bpf_getsockopt(TCP_BPF_SYN*) option will hide where the 47010813a841SMartin KaFai Lau * SYN packet is obtained. 47020813a841SMartin KaFai Lau * 47030813a841SMartin KaFai Lau * If the bpf-prog does not need the IP[46] header, the 47040813a841SMartin KaFai Lau * bpf-prog can avoid parsing the IP header by using 47050813a841SMartin KaFai Lau * TCP_BPF_SYN. Otherwise, the bpf-prog can get both 47060813a841SMartin KaFai Lau * IP[46] and TCP header by using TCP_BPF_SYN_IP. 47070813a841SMartin KaFai Lau * 47080813a841SMartin KaFai Lau * >0: Total number of bytes copied 47090813a841SMartin KaFai Lau * -ENOSPC: Not enough space in optval. Only optlen number of 47100813a841SMartin KaFai Lau * bytes is copied. 47110813a841SMartin KaFai Lau * -ENOENT: The SYN skb is not available now and the earlier SYN pkt 47120813a841SMartin KaFai Lau * is not saved by setsockopt(TCP_SAVE_SYN). 47130813a841SMartin KaFai Lau */ 47140813a841SMartin KaFai Lau TCP_BPF_SYN = 1005, /* Copy the TCP header */ 47150813a841SMartin KaFai Lau TCP_BPF_SYN_IP = 1006, /* Copy the IP[46] and TCP header */ 4716267cf9faSMartin KaFai Lau TCP_BPF_SYN_MAC = 1007, /* Copy the MAC, IP[46], and TCP header */ 47170813a841SMartin KaFai Lau }; 47180813a841SMartin KaFai Lau 47190813a841SMartin KaFai Lau enum { 47200813a841SMartin KaFai Lau BPF_LOAD_HDR_OPT_TCP_SYN = (1ULL << 0), 47210813a841SMartin KaFai Lau }; 47220813a841SMartin KaFai Lau 47230813a841SMartin KaFai Lau /* args[0] value during BPF_SOCK_OPS_HDR_OPT_LEN_CB and 47240813a841SMartin KaFai Lau * BPF_SOCK_OPS_WRITE_HDR_OPT_CB. 47250813a841SMartin KaFai Lau */ 47260813a841SMartin KaFai Lau enum { 47270813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_CURRENT_MSS = 1, /* Kernel is finding the 47280813a841SMartin KaFai Lau * total option spaces 47290813a841SMartin KaFai Lau * required for an established 47300813a841SMartin KaFai Lau * sk in order to calculate the 47310813a841SMartin KaFai Lau * MSS. No skb is actually 47320813a841SMartin KaFai Lau * sent. 47330813a841SMartin KaFai Lau */ 47340813a841SMartin KaFai Lau BPF_WRITE_HDR_TCP_SYNACK_COOKIE = 2, /* Kernel is in syncookie mode 47350813a841SMartin KaFai Lau * when sending a SYN. 47360813a841SMartin KaFai Lau */ 47371aae4bddSAndrii Nakryiko }; 4738fc747810SLawrence Brakmo 4739908432caSYonghong Song struct bpf_perf_event_value { 4740908432caSYonghong Song __u64 counter; 4741908432caSYonghong Song __u64 enabled; 4742908432caSYonghong Song __u64 running; 4743908432caSYonghong Song }; 4744908432caSYonghong Song 47451aae4bddSAndrii Nakryiko enum { 47461aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_MKNOD = (1ULL << 0), 47471aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_READ = (1ULL << 1), 47481aae4bddSAndrii Nakryiko BPF_DEVCG_ACC_WRITE = (1ULL << 2), 47491aae4bddSAndrii Nakryiko }; 4750ebc614f6SRoman Gushchin 47511aae4bddSAndrii Nakryiko enum { 47521aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_BLOCK = (1ULL << 0), 47531aae4bddSAndrii Nakryiko BPF_DEVCG_DEV_CHAR = (1ULL << 1), 47541aae4bddSAndrii Nakryiko }; 4755ebc614f6SRoman Gushchin 4756ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 475706ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 475806ef0ccbSYonghong Song __u32 access_type; 4759ebc614f6SRoman Gushchin __u32 major; 4760ebc614f6SRoman Gushchin __u32 minor; 4761ebc614f6SRoman Gushchin }; 4762ebc614f6SRoman Gushchin 4763c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 4764c4f6699dSAlexei Starovoitov __u64 args[0]; 4765c4f6699dSAlexei Starovoitov }; 4766c4f6699dSAlexei Starovoitov 476787f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 476887f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 476987f5fc7eSDavid Ahern */ 47701aae4bddSAndrii Nakryiko enum { 47711aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_DIRECT = (1U << 0), 47721aae4bddSAndrii Nakryiko BPF_FIB_LOOKUP_OUTPUT = (1U << 1), 47731aae4bddSAndrii Nakryiko }; 477487f5fc7eSDavid Ahern 47754c79579bSDavid Ahern enum { 47764c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 47774c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 47784c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 47794c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 47804c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 47814c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 47824c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 47834c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 47844c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 47854c79579bSDavid Ahern }; 47864c79579bSDavid Ahern 478787f5fc7eSDavid Ahern struct bpf_fib_lookup { 4788fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 4789fa898d76SDavid Ahern * output: network family of egress nexthop 4790fa898d76SDavid Ahern */ 4791fa898d76SDavid Ahern __u8 family; 479287f5fc7eSDavid Ahern 479387f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 479487f5fc7eSDavid Ahern __u8 l4_protocol; 479587f5fc7eSDavid Ahern __be16 sport; 479687f5fc7eSDavid Ahern __be16 dport; 479787f5fc7eSDavid Ahern 479887f5fc7eSDavid Ahern /* total length of packet from network header - used for MTU check */ 479987f5fc7eSDavid Ahern __u16 tot_len; 48004c79579bSDavid Ahern 48014c79579bSDavid Ahern /* input: L3 device index for lookup 48024c79579bSDavid Ahern * output: device index from FIB lookup 48034c79579bSDavid Ahern */ 48044c79579bSDavid Ahern __u32 ifindex; 480587f5fc7eSDavid Ahern 480687f5fc7eSDavid Ahern union { 480787f5fc7eSDavid Ahern /* inputs to lookup */ 480887f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 4809bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 481087f5fc7eSDavid Ahern 4811fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 481287f5fc7eSDavid Ahern __u32 rt_metric; 481387f5fc7eSDavid Ahern }; 481487f5fc7eSDavid Ahern 481587f5fc7eSDavid Ahern union { 481687f5fc7eSDavid Ahern __be32 ipv4_src; 481787f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 481887f5fc7eSDavid Ahern }; 481987f5fc7eSDavid Ahern 4820fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 4821fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 4822fa898d76SDavid Ahern * if FIB lookup returns gateway route 482387f5fc7eSDavid Ahern */ 482487f5fc7eSDavid Ahern union { 482587f5fc7eSDavid Ahern __be32 ipv4_dst; 482687f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 482787f5fc7eSDavid Ahern }; 482887f5fc7eSDavid Ahern 482987f5fc7eSDavid Ahern /* output */ 483087f5fc7eSDavid Ahern __be16 h_vlan_proto; 483187f5fc7eSDavid Ahern __be16 h_vlan_TCI; 483287f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 483387f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 483487f5fc7eSDavid Ahern }; 483587f5fc7eSDavid Ahern 483641bdc4b4SYonghong Song enum bpf_task_fd_type { 483741bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 483841bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 483941bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 484041bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 484141bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 484241bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 484341bdc4b4SYonghong Song }; 484441bdc4b4SYonghong Song 48451aae4bddSAndrii Nakryiko enum { 48461aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_PARSE_1ST_FRAG = (1U << 0), 48471aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_FLOW_LABEL = (1U << 1), 48481aae4bddSAndrii Nakryiko BPF_FLOW_DISSECTOR_F_STOP_AT_ENCAP = (1U << 2), 48491aae4bddSAndrii Nakryiko }; 4850086f9568SStanislav Fomichev 4851d58e468bSPetar Penkov struct bpf_flow_keys { 4852d58e468bSPetar Penkov __u16 nhoff; 4853d58e468bSPetar Penkov __u16 thoff; 4854d58e468bSPetar Penkov __u16 addr_proto; /* ETH_P_* of valid addrs */ 4855d58e468bSPetar Penkov __u8 is_frag; 4856d58e468bSPetar Penkov __u8 is_first_frag; 4857d58e468bSPetar Penkov __u8 is_encap; 4858d58e468bSPetar Penkov __u8 ip_proto; 4859d58e468bSPetar Penkov __be16 n_proto; 4860d58e468bSPetar Penkov __be16 sport; 4861d58e468bSPetar Penkov __be16 dport; 4862d58e468bSPetar Penkov union { 4863d58e468bSPetar Penkov struct { 4864d58e468bSPetar Penkov __be32 ipv4_src; 4865d58e468bSPetar Penkov __be32 ipv4_dst; 4866d58e468bSPetar Penkov }; 4867d58e468bSPetar Penkov struct { 4868d58e468bSPetar Penkov __u32 ipv6_src[4]; /* in6_addr; network order */ 4869d58e468bSPetar Penkov __u32 ipv6_dst[4]; /* in6_addr; network order */ 4870d58e468bSPetar Penkov }; 4871d58e468bSPetar Penkov }; 4872086f9568SStanislav Fomichev __u32 flags; 487371c99e32SStanislav Fomichev __be32 flow_label; 4874d58e468bSPetar Penkov }; 4875d58e468bSPetar Penkov 4876838e9690SYonghong Song struct bpf_func_info { 4877d30d42e0SMartin KaFai Lau __u32 insn_off; 4878838e9690SYonghong Song __u32 type_id; 4879838e9690SYonghong Song }; 4880838e9690SYonghong Song 4881c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 4882c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 4883c454a46bSMartin KaFai Lau 4884c454a46bSMartin KaFai Lau struct bpf_line_info { 4885c454a46bSMartin KaFai Lau __u32 insn_off; 4886c454a46bSMartin KaFai Lau __u32 file_name_off; 4887c454a46bSMartin KaFai Lau __u32 line_off; 4888c454a46bSMartin KaFai Lau __u32 line_col; 4889c454a46bSMartin KaFai Lau }; 4890c454a46bSMartin KaFai Lau 4891d83525caSAlexei Starovoitov struct bpf_spin_lock { 4892d83525caSAlexei Starovoitov __u32 val; 4893d83525caSAlexei Starovoitov }; 48947b146cebSAndrey Ignatov 48957b146cebSAndrey Ignatov struct bpf_sysctl { 48967b146cebSAndrey Ignatov __u32 write; /* Sysctl is being read (= 0) or written (= 1). 48977b146cebSAndrey Ignatov * Allows 1,2,4-byte read, but no write. 48987b146cebSAndrey Ignatov */ 4899e1550bfeSAndrey Ignatov __u32 file_pos; /* Sysctl file position to read from, write to. 4900e1550bfeSAndrey Ignatov * Allows 1,2,4-byte read an 4-byte write. 4901e1550bfeSAndrey Ignatov */ 49027b146cebSAndrey Ignatov }; 49037b146cebSAndrey Ignatov 49040d01da6aSStanislav Fomichev struct bpf_sockopt { 49050d01da6aSStanislav Fomichev __bpf_md_ptr(struct bpf_sock *, sk); 49060d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval); 49070d01da6aSStanislav Fomichev __bpf_md_ptr(void *, optval_end); 49080d01da6aSStanislav Fomichev 49090d01da6aSStanislav Fomichev __s32 level; 49100d01da6aSStanislav Fomichev __s32 optname; 49110d01da6aSStanislav Fomichev __s32 optlen; 49120d01da6aSStanislav Fomichev __s32 retval; 49130d01da6aSStanislav Fomichev }; 49140d01da6aSStanislav Fomichev 4915b4490c5cSCarlos Neira struct bpf_pidns_info { 4916b4490c5cSCarlos Neira __u32 pid; 4917b4490c5cSCarlos Neira __u32 tgid; 4918b4490c5cSCarlos Neira }; 4919e9ddbb77SJakub Sitnicki 4920e9ddbb77SJakub Sitnicki /* User accessible data for SK_LOOKUP programs. Add new fields at the end. */ 4921e9ddbb77SJakub Sitnicki struct bpf_sk_lookup { 4922e9ddbb77SJakub Sitnicki __bpf_md_ptr(struct bpf_sock *, sk); /* Selected socket */ 4923e9ddbb77SJakub Sitnicki 4924e9ddbb77SJakub Sitnicki __u32 family; /* Protocol family (AF_INET, AF_INET6) */ 4925e9ddbb77SJakub Sitnicki __u32 protocol; /* IP protocol (IPPROTO_TCP, IPPROTO_UDP) */ 4926e9ddbb77SJakub Sitnicki __u32 remote_ip4; /* Network byte order */ 4927e9ddbb77SJakub Sitnicki __u32 remote_ip6[4]; /* Network byte order */ 4928e9ddbb77SJakub Sitnicki __u32 remote_port; /* Network byte order */ 4929e9ddbb77SJakub Sitnicki __u32 local_ip4; /* Network byte order */ 4930e9ddbb77SJakub Sitnicki __u32 local_ip6[4]; /* Network byte order */ 4931e9ddbb77SJakub Sitnicki __u32 local_port; /* Host byte order */ 4932e9ddbb77SJakub Sitnicki }; 4933e9ddbb77SJakub Sitnicki 4934c4d0bfb4SAlan Maguire /* 4935c4d0bfb4SAlan Maguire * struct btf_ptr is used for typed pointer representation; the 4936c4d0bfb4SAlan Maguire * type id is used to render the pointer data as the appropriate type 4937c4d0bfb4SAlan Maguire * via the bpf_snprintf_btf() helper described above. A flags field - 4938c4d0bfb4SAlan Maguire * potentially to specify additional details about the BTF pointer 4939c4d0bfb4SAlan Maguire * (rather than its mode of display) - is included for future use. 4940c4d0bfb4SAlan Maguire * Display flags - BTF_F_* - are passed to bpf_snprintf_btf separately. 4941c4d0bfb4SAlan Maguire */ 4942c4d0bfb4SAlan Maguire struct btf_ptr { 4943c4d0bfb4SAlan Maguire void *ptr; 4944c4d0bfb4SAlan Maguire __u32 type_id; 4945c4d0bfb4SAlan Maguire __u32 flags; /* BTF ptr flags; unused at present. */ 4946c4d0bfb4SAlan Maguire }; 4947c4d0bfb4SAlan Maguire 4948c4d0bfb4SAlan Maguire /* 4949c4d0bfb4SAlan Maguire * Flags to control bpf_snprintf_btf() behaviour. 4950c4d0bfb4SAlan Maguire * - BTF_F_COMPACT: no formatting around type information 4951c4d0bfb4SAlan Maguire * - BTF_F_NONAME: no struct/union member names/types 4952c4d0bfb4SAlan Maguire * - BTF_F_PTR_RAW: show raw (unobfuscated) pointer values; 4953c4d0bfb4SAlan Maguire * equivalent to %px. 4954c4d0bfb4SAlan Maguire * - BTF_F_ZERO: show zero-valued struct/union members; they 4955c4d0bfb4SAlan Maguire * are not displayed by default 4956c4d0bfb4SAlan Maguire */ 4957c4d0bfb4SAlan Maguire enum { 4958c4d0bfb4SAlan Maguire BTF_F_COMPACT = (1ULL << 0), 4959c4d0bfb4SAlan Maguire BTF_F_NONAME = (1ULL << 1), 4960c4d0bfb4SAlan Maguire BTF_F_PTR_RAW = (1ULL << 2), 4961c4d0bfb4SAlan Maguire BTF_F_ZERO = (1ULL << 3), 4962c4d0bfb4SAlan Maguire }; 4963c4d0bfb4SAlan Maguire 4964daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 4965