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 */ 76b95a5c4dSDaniel Mack __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 84b2197755SDaniel Borkmann /* BPF syscall commands, see bpf(2) man-page for details. */ 8599c55f7dSAlexei Starovoitov enum bpf_cmd { 8699c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 87db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 88db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 89db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 90db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 9109756af4SAlexei Starovoitov BPF_PROG_LOAD, 92b2197755SDaniel Borkmann BPF_OBJ_PIN, 93b2197755SDaniel Borkmann BPF_OBJ_GET, 94f4324551SDaniel Mack BPF_PROG_ATTACH, 95f4324551SDaniel Mack BPF_PROG_DETACH, 961cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 9734ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 9834ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 99b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 100bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 1011e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 102468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 103c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 104f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 10578958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 10641bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 107bd513cd0SMauricio Vasquez B BPF_MAP_LOOKUP_AND_DELETE_ELEM, 10887df15deSDaniel Borkmann BPF_MAP_FREEZE, 10999c55f7dSAlexei Starovoitov }; 11099c55f7dSAlexei Starovoitov 11199c55f7dSAlexei Starovoitov enum bpf_map_type { 11299c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 1130f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 11428fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 11504fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 116ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 117824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 118a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 119d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 1204ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 12129ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 1228f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 123b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 12456f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 125bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 126546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 127174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 1286710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 129fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 13081110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 131de9cbbaaSRoman Gushchin BPF_MAP_TYPE_CGROUP_STORAGE, 1325dc4c4b7SMartin KaFai Lau BPF_MAP_TYPE_REUSEPORT_SOCKARRAY, 133b741f163SRoman Gushchin BPF_MAP_TYPE_PERCPU_CGROUP_STORAGE, 134f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_QUEUE, 135f1a2e44aSMauricio Vasquez B BPF_MAP_TYPE_STACK, 13699c55f7dSAlexei Starovoitov }; 13799c55f7dSAlexei Starovoitov 1386c4fc209SDaniel Borkmann /* Note that tracing related programs such as 1396c4fc209SDaniel Borkmann * BPF_PROG_TYPE_{KPROBE,TRACEPOINT,PERF_EVENT,RAW_TRACEPOINT} 1406c4fc209SDaniel Borkmann * are not subject to a stable API since kernel internal data 1416c4fc209SDaniel Borkmann * structures can change from release to release and may 1426c4fc209SDaniel Borkmann * therefore break existing tracing BPF programs. Tracing BPF 1436c4fc209SDaniel Borkmann * programs correspond to /a/ specific kernel which is to be 1446c4fc209SDaniel Borkmann * analyzed, and not /a/ specific kernel /and/ all future ones. 1456c4fc209SDaniel Borkmann */ 14609756af4SAlexei Starovoitov enum bpf_prog_type { 14709756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 148ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 1492541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 15096be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 15194caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 15298b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 1536a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 1540515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 1550e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 15661023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 1573a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 1583a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 1593a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 16040304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 161b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 162ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 1634f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 164c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 1654fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 166004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 167f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 1682dbb9b9eSMartin KaFai Lau BPF_PROG_TYPE_SK_REUSEPORT, 169d58e468bSPetar Penkov BPF_PROG_TYPE_FLOW_DISSECTOR, 17009756af4SAlexei Starovoitov }; 17109756af4SAlexei Starovoitov 1720e33661dSDaniel Mack enum bpf_attach_type { 1730e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 1740e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 17561023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 17640304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 177464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 178464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 179ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 1804f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 1814fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 1824fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 183d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 184d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 185aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 186aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 1871cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 1881cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 189f4364dcfSSean Young BPF_LIRC_MODE2, 190d58e468bSPetar Penkov BPF_FLOW_DISSECTOR, 1910e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 1920e33661dSDaniel Mack }; 1930e33661dSDaniel Mack 1940e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1950e33661dSDaniel Mack 196324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 197324bda9eSAlexei Starovoitov * 198324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 199324bda9eSAlexei Starovoitov * 200324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 201324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 202324bda9eSAlexei Starovoitov * 203324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 204324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 205324bda9eSAlexei Starovoitov * 206324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 207324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 208324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 209324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 210324bda9eSAlexei Starovoitov * 211324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 212324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 213324bda9eSAlexei Starovoitov * (those that were attached first, run first) 214324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 215324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 216324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 217324bda9eSAlexei Starovoitov * parent program has a chance to override it. 218324bda9eSAlexei Starovoitov * 219324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 220324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 221324bda9eSAlexei Starovoitov * Ex1: 222324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 223324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 224324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 225324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 226324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 227324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 228324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 229324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 230324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 231324bda9eSAlexei Starovoitov * 232324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 233324bda9eSAlexei Starovoitov * earlier programs. 2347f677633SAlexei Starovoitov */ 2357f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 236324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 2377f677633SAlexei Starovoitov 238e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 239e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 240e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 241e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 242e07b98d9SDavid S. Miller */ 243e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 244e07b98d9SDavid S. Miller 245e9ee9efcSDavid Miller /* If BPF_F_ANY_ALIGNMENT is used in BPF_PROF_LOAD command, the 246e9ee9efcSDavid Miller * verifier will allow any alignment whatsoever. On platforms 247e9ee9efcSDavid Miller * with strict alignment requirements for loads ands stores (such 248e9ee9efcSDavid Miller * as sparc and mips) the verifier validates that all loads and 249e9ee9efcSDavid Miller * stores provably follow this requirement. This flag turns that 250e9ee9efcSDavid Miller * checking and enforcement off. 251e9ee9efcSDavid Miller * 252e9ee9efcSDavid Miller * It is mostly used for testing when we want to validate the 253e9ee9efcSDavid Miller * context and memory access aspects of the verifier, but because 254e9ee9efcSDavid Miller * of an unaligned access the alignment check would trigger before 255e9ee9efcSDavid Miller * the one we are interested in. 256e9ee9efcSDavid Miller */ 257e9ee9efcSDavid Miller #define BPF_F_ANY_ALIGNMENT (1U << 1) 258e9ee9efcSDavid Miller 259d8eca5bbSDaniel Borkmann /* When BPF ldimm64's insn[0].src_reg != 0 then this can have 260d8eca5bbSDaniel Borkmann * two extensions: 261d8eca5bbSDaniel Borkmann * 262d8eca5bbSDaniel Borkmann * insn[0].src_reg: BPF_PSEUDO_MAP_FD BPF_PSEUDO_MAP_VALUE 263d8eca5bbSDaniel Borkmann * insn[0].imm: map fd map fd 264d8eca5bbSDaniel Borkmann * insn[1].imm: 0 offset into value 265d8eca5bbSDaniel Borkmann * insn[0].off: 0 0 266d8eca5bbSDaniel Borkmann * insn[1].off: 0 0 267d8eca5bbSDaniel Borkmann * ldimm64 rewrite: address of map address of map[0]+offset 268d8eca5bbSDaniel Borkmann * verifier type: CONST_PTR_TO_MAP PTR_TO_MAP_VALUE 269d8eca5bbSDaniel Borkmann */ 270f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 271d8eca5bbSDaniel Borkmann #define BPF_PSEUDO_MAP_VALUE 2 272f1a66f85SDaniel Borkmann 273cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 274cc8b0b92SAlexei Starovoitov * offset to another bpf function 275cc8b0b92SAlexei Starovoitov */ 276cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 277cc8b0b92SAlexei Starovoitov 2783274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 2793274f520SAlexei Starovoitov #define BPF_ANY 0 /* create new element or update existing */ 2803274f520SAlexei Starovoitov #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 2813274f520SAlexei Starovoitov #define BPF_EXIST 2 /* update existing element */ 28296049f3aSAlexei Starovoitov #define BPF_F_LOCK 4 /* spin_lock-ed map_lookup/map_update */ 2833274f520SAlexei Starovoitov 28496eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 2856c905981SAlexei Starovoitov #define BPF_F_NO_PREALLOC (1U << 0) 28629ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 2878f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 28829ba732aSMartin KaFai Lau * which can scale and perform better. 28929ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 29029ba732aSMartin KaFai Lau * across different LRU lists. 29129ba732aSMartin KaFai Lau */ 29229ba732aSMartin KaFai Lau #define BPF_F_NO_COMMON_LRU (1U << 1) 29396eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 29496eabe7aSMartin KaFai Lau #define BPF_F_NUMA_NODE (1U << 2) 2956c905981SAlexei Starovoitov 296cb4d2b3fSMartin KaFai Lau #define BPF_OBJ_NAME_LEN 16U 297cb4d2b3fSMartin KaFai Lau 298591fe988SDaniel Borkmann /* Flags for accessing BPF object from syscall side. */ 2996e71b04aSChenbo Feng #define BPF_F_RDONLY (1U << 3) 3006e71b04aSChenbo Feng #define BPF_F_WRONLY (1U << 4) 3016e71b04aSChenbo Feng 302615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 303615755a7SSong Liu #define BPF_F_STACK_BUILD_ID (1U << 5) 304615755a7SSong Liu 30596b3b6c9SLorenz Bauer /* Zero-initialize hash function seed. This should only be used for testing. */ 30696b3b6c9SLorenz Bauer #define BPF_F_ZERO_SEED (1U << 6) 30796b3b6c9SLorenz Bauer 308591fe988SDaniel Borkmann /* Flags for accessing BPF object from program side. */ 309591fe988SDaniel Borkmann #define BPF_F_RDONLY_PROG (1U << 7) 310591fe988SDaniel Borkmann #define BPF_F_WRONLY_PROG (1U << 8) 311591fe988SDaniel Borkmann 3122f183360SLorenz Bauer /* flags for BPF_PROG_QUERY */ 3132f183360SLorenz Bauer #define BPF_F_QUERY_EFFECTIVE (1U << 0) 3142f183360SLorenz Bauer 315615755a7SSong Liu enum bpf_stack_build_id_status { 316615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 317615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 318615755a7SSong Liu /* with valid build_id and offset */ 319615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 320615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 321615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 322615755a7SSong Liu }; 323615755a7SSong Liu 324615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 325615755a7SSong Liu struct bpf_stack_build_id { 326615755a7SSong Liu __s32 status; 327615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 328615755a7SSong Liu union { 329615755a7SSong Liu __u64 offset; 330615755a7SSong Liu __u64 ip; 331615755a7SSong Liu }; 332615755a7SSong Liu }; 333615755a7SSong Liu 33499c55f7dSAlexei Starovoitov union bpf_attr { 33599c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 33699c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 33799c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 33899c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 33999c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 34096eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 34196eabe7aSMartin KaFai Lau * flags defined above. 34296eabe7aSMartin KaFai Lau */ 34356f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 34496eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 34596eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 34696eabe7aSMartin KaFai Lau */ 347067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 348a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 349a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 3509b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 3519b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 35299c55f7dSAlexei Starovoitov }; 353db20fd2bSAlexei Starovoitov 354db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 355db20fd2bSAlexei Starovoitov __u32 map_fd; 356db20fd2bSAlexei Starovoitov __aligned_u64 key; 357db20fd2bSAlexei Starovoitov union { 358db20fd2bSAlexei Starovoitov __aligned_u64 value; 359db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 360db20fd2bSAlexei Starovoitov }; 3613274f520SAlexei Starovoitov __u64 flags; 362db20fd2bSAlexei Starovoitov }; 36309756af4SAlexei Starovoitov 36409756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 36509756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 36609756af4SAlexei Starovoitov __u32 insn_cnt; 36709756af4SAlexei Starovoitov __aligned_u64 insns; 36809756af4SAlexei Starovoitov __aligned_u64 license; 369cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 370cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 371cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 3726c4fc209SDaniel Borkmann __u32 kern_version; /* not used */ 373e07b98d9SDavid S. Miller __u32 prog_flags; 374067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 3751f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 3765e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 3775e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 3785e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 3795e43f899SAndrey Ignatov */ 3805e43f899SAndrey Ignatov __u32 expected_attach_type; 381838e9690SYonghong Song __u32 prog_btf_fd; /* fd pointing to BTF type data */ 382838e9690SYonghong Song __u32 func_info_rec_size; /* userspace bpf_func_info size */ 383838e9690SYonghong Song __aligned_u64 func_info; /* func info */ 384838e9690SYonghong Song __u32 func_info_cnt; /* number of bpf_func_info records */ 385c454a46bSMartin KaFai Lau __u32 line_info_rec_size; /* userspace bpf_line_info size */ 386c454a46bSMartin KaFai Lau __aligned_u64 line_info; /* line info */ 387c454a46bSMartin KaFai Lau __u32 line_info_cnt; /* number of bpf_line_info records */ 38809756af4SAlexei Starovoitov }; 389b2197755SDaniel Borkmann 390b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 391b2197755SDaniel Borkmann __aligned_u64 pathname; 392b2197755SDaniel Borkmann __u32 bpf_fd; 3936e71b04aSChenbo Feng __u32 file_flags; 394b2197755SDaniel Borkmann }; 395f4324551SDaniel Mack 396f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 397f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 398f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 399f4324551SDaniel Mack __u32 attach_type; 4007f677633SAlexei Starovoitov __u32 attach_flags; 401f4324551SDaniel Mack }; 4021cf1cae9SAlexei Starovoitov 4031cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 4041cf1cae9SAlexei Starovoitov __u32 prog_fd; 4051cf1cae9SAlexei Starovoitov __u32 retval; 406b5a36b1eSLorenz Bauer __u32 data_size_in; /* input: len of data_in */ 407b5a36b1eSLorenz Bauer __u32 data_size_out; /* input/output: len of data_out 408b5a36b1eSLorenz Bauer * returns ENOSPC if data_out 409b5a36b1eSLorenz Bauer * is too small. 410b5a36b1eSLorenz Bauer */ 4111cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 4121cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 4131cf1cae9SAlexei Starovoitov __u32 repeat; 4141cf1cae9SAlexei Starovoitov __u32 duration; 415*b0b9395dSStanislav Fomichev __u32 ctx_size_in; /* input: len of ctx_in */ 416*b0b9395dSStanislav Fomichev __u32 ctx_size_out; /* input/output: len of ctx_out 417*b0b9395dSStanislav Fomichev * returns ENOSPC if ctx_out 418*b0b9395dSStanislav Fomichev * is too small. 419*b0b9395dSStanislav Fomichev */ 420*b0b9395dSStanislav Fomichev __aligned_u64 ctx_in; 421*b0b9395dSStanislav Fomichev __aligned_u64 ctx_out; 4221cf1cae9SAlexei Starovoitov } test; 42334ad5580SMartin KaFai Lau 424b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 425b16d9aa4SMartin KaFai Lau union { 42634ad5580SMartin KaFai Lau __u32 start_id; 427b16d9aa4SMartin KaFai Lau __u32 prog_id; 428bd5f5f4eSMartin KaFai Lau __u32 map_id; 42978958fcaSMartin KaFai Lau __u32 btf_id; 430b16d9aa4SMartin KaFai Lau }; 43134ad5580SMartin KaFai Lau __u32 next_id; 4326e71b04aSChenbo Feng __u32 open_flags; 43334ad5580SMartin KaFai Lau }; 4341e270976SMartin KaFai Lau 4351e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 4361e270976SMartin KaFai Lau __u32 bpf_fd; 4371e270976SMartin KaFai Lau __u32 info_len; 4381e270976SMartin KaFai Lau __aligned_u64 info; 4391e270976SMartin KaFai Lau } info; 440468e2f64SAlexei Starovoitov 441468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 442468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 443468e2f64SAlexei Starovoitov __u32 attach_type; 444468e2f64SAlexei Starovoitov __u32 query_flags; 445468e2f64SAlexei Starovoitov __u32 attach_flags; 446468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 447468e2f64SAlexei Starovoitov __u32 prog_cnt; 448468e2f64SAlexei Starovoitov } query; 449c4f6699dSAlexei Starovoitov 450c4f6699dSAlexei Starovoitov struct { 451c4f6699dSAlexei Starovoitov __u64 name; 452c4f6699dSAlexei Starovoitov __u32 prog_fd; 453c4f6699dSAlexei Starovoitov } raw_tracepoint; 454f56a653cSMartin KaFai Lau 455f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 456f56a653cSMartin KaFai Lau __aligned_u64 btf; 457f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 458f56a653cSMartin KaFai Lau __u32 btf_size; 459f56a653cSMartin KaFai Lau __u32 btf_log_size; 460f56a653cSMartin KaFai Lau __u32 btf_log_level; 461f56a653cSMartin KaFai Lau }; 46241bdc4b4SYonghong Song 46341bdc4b4SYonghong Song struct { 46441bdc4b4SYonghong Song __u32 pid; /* input: pid */ 46541bdc4b4SYonghong Song __u32 fd; /* input: fd */ 46641bdc4b4SYonghong Song __u32 flags; /* input: flags */ 46741bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 46841bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 46941bdc4b4SYonghong Song * tp_name for tracepoint 47041bdc4b4SYonghong Song * symbol for kprobe 47141bdc4b4SYonghong Song * filename for uprobe 47241bdc4b4SYonghong Song */ 47341bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 47441bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 47541bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 47641bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 47741bdc4b4SYonghong Song } task_fd_query; 47899c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 47999c55f7dSAlexei Starovoitov 48056a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 48156a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 48256a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 48356a092c8SQuentin Monnet * and requires the rst2man utility: 484ebb676daSThomas Graf * 48556a092c8SQuentin Monnet * $ ./scripts/bpf_helpers_doc.py \ 48656a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 48756a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 48856a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 489ebb676daSThomas Graf * 49056a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 49156a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 49256a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 49356a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 494ebb676daSThomas Graf * 49556a092c8SQuentin Monnet * Start of BPF helper function descriptions: 496ad4a5223SQuentin Monnet * 497ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 498ad4a5223SQuentin Monnet * Description 499ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 500ad4a5223SQuentin Monnet * Return 501ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 502ad4a5223SQuentin Monnet * found. 503ad4a5223SQuentin Monnet * 504ad4a5223SQuentin Monnet * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 505ad4a5223SQuentin Monnet * Description 506ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 507ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 508ad4a5223SQuentin Monnet * 509ad4a5223SQuentin Monnet * **BPF_NOEXIST** 510ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 511ad4a5223SQuentin Monnet * **BPF_EXIST** 512ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 513ad4a5223SQuentin Monnet * **BPF_ANY** 514ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 515ad4a5223SQuentin Monnet * 516ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 517ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 518ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 519ad4a5223SQuentin Monnet * Return 520ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 521ad4a5223SQuentin Monnet * 522ad4a5223SQuentin Monnet * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 523ad4a5223SQuentin Monnet * Description 524ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 525ad4a5223SQuentin Monnet * Return 526ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 527ad4a5223SQuentin Monnet * 528ad4a5223SQuentin Monnet * int bpf_probe_read(void *dst, u32 size, const void *src) 529ad4a5223SQuentin Monnet * Description 530ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 531ad4a5223SQuentin Monnet * address *src* and store the data in *dst*. 532ad4a5223SQuentin Monnet * Return 533ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 534ad4a5223SQuentin Monnet * 535ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 536ad4a5223SQuentin Monnet * Description 537ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 538ad4a5223SQuentin Monnet * Return 539ad4a5223SQuentin Monnet * Current *ktime*. 540ad4a5223SQuentin Monnet * 541ad4a5223SQuentin Monnet * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 542ad4a5223SQuentin Monnet * Description 543ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 544ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 545ad4a5223SQuentin Monnet * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 546ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 547ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 548ad4a5223SQuentin Monnet * limited to five). 549ad4a5223SQuentin Monnet * 550ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 551ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 552ad4a5223SQuentin Monnet * one will get depends on the options set in 553ad4a5223SQuentin Monnet * *\/sys/kernel/debug/tracing/trace_options* (see also the 554ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 555ad4a5223SQuentin Monnet * defaults to something like: 556ad4a5223SQuentin Monnet * 557ad4a5223SQuentin Monnet * :: 558ad4a5223SQuentin Monnet * 559ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 560ad4a5223SQuentin Monnet * 561ad4a5223SQuentin Monnet * In the above: 562ad4a5223SQuentin Monnet * 563ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 564ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 565ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 566ad4a5223SQuentin Monnet * running. 567ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 568ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 569ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 570ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 571ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 572ad4a5223SQuentin Monnet * are set. 573ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 574ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 575ad4a5223SQuentin Monnet * instruction pointer register. 576ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 577ad4a5223SQuentin Monnet * *fmt*. 578ad4a5223SQuentin Monnet * 579ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 580ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 581ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 582ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 583ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 584ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 585ad4a5223SQuentin Monnet * encounters an unknown specifier. 586ad4a5223SQuentin Monnet * 587ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 588ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 589ad4a5223SQuentin Monnet * bloc (spanning several lines) is printed to kernel logs and 590ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 591ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 592ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 593ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 594ad4a5223SQuentin Monnet * Return 595ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 596ad4a5223SQuentin Monnet * in case of failure. 597ad4a5223SQuentin Monnet * 5981fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 5991fdd08beSQuentin Monnet * Description 6001fdd08beSQuentin Monnet * Get a pseudo-random number. 6011fdd08beSQuentin Monnet * 6021fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 6031fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 6041fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 6051fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 6061fdd08beSQuentin Monnet * cryptographically secure. 6071fdd08beSQuentin Monnet * Return 6081fdd08beSQuentin Monnet * A random 32-bit unsigned value. 6091fdd08beSQuentin Monnet * 6101fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 6111fdd08beSQuentin Monnet * Description 6121fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 6131fdd08beSQuentin Monnet * all programs run with preemption disabled, which means that the 6141fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 6151fdd08beSQuentin Monnet * program. 6161fdd08beSQuentin Monnet * Return 6171fdd08beSQuentin Monnet * The SMP id of the processor running the program. 6181fdd08beSQuentin Monnet * 619ad4a5223SQuentin Monnet * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 620ad4a5223SQuentin Monnet * Description 621ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 622ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 623ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 624ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 625ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 626ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 627ad4a5223SQuentin Monnet * 628ad4a5223SQuentin Monnet * A call to this helper is susceptible to change the underlaying 629ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 630ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 631ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 632ad4a5223SQuentin Monnet * direct packet access. 633ad4a5223SQuentin Monnet * Return 634ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 635ad4a5223SQuentin Monnet * 636ad4a5223SQuentin Monnet * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 637ad4a5223SQuentin Monnet * Description 638ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 639ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 640ad4a5223SQuentin Monnet * must know the former value of the header field that was 641ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 642ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 643ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 644ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 645ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 646ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 647ad4a5223SQuentin Monnet * 648ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 649ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 650ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 651ad4a5223SQuentin Monnet * checksum to update. 652ad4a5223SQuentin Monnet * 653ad4a5223SQuentin Monnet * A call to this helper is susceptible to change the underlaying 654ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 655ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 656ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 657ad4a5223SQuentin Monnet * direct packet access. 658ad4a5223SQuentin Monnet * Return 659ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 660ad4a5223SQuentin Monnet * 661ad4a5223SQuentin Monnet * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 662ad4a5223SQuentin Monnet * Description 663ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 664ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 665ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 666ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 667ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 668ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 669ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 670ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 671ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 672ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 673ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 674ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 675ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 676ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 677ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 678ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 679ad4a5223SQuentin Monnet * 680ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 681ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 682ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 683ad4a5223SQuentin Monnet * checksum to update. 684ad4a5223SQuentin Monnet * 685ad4a5223SQuentin Monnet * A call to this helper is susceptible to change the underlaying 686ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 687ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 688ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 689ad4a5223SQuentin Monnet * direct packet access. 690ad4a5223SQuentin Monnet * Return 691ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 692ad4a5223SQuentin Monnet * 693ad4a5223SQuentin Monnet * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 694ad4a5223SQuentin Monnet * Description 695ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 696ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 697ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 698ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 699ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 700ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 701ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 702ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 703ad4a5223SQuentin Monnet * performed. 704ad4a5223SQuentin Monnet * 705ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 706ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 707ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 708ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 709ad4a5223SQuentin Monnet * 710ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 711ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 712ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 713ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 714ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 715ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 716ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 717ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 718ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 719ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 720ad4a5223SQuentin Monnet * which is currently set to 32. 721ad4a5223SQuentin Monnet * Return 722ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 723ad4a5223SQuentin Monnet * 724ad4a5223SQuentin Monnet * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 725ad4a5223SQuentin Monnet * Description 726ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 727ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 728ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 729ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 730ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 731ad4a5223SQuentin Monnet * This is the only flag supported for now. 732ad4a5223SQuentin Monnet * 733ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 734ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 735ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 736ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 737ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 738ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 739ad4a5223SQuentin Monnet * 740ad4a5223SQuentin Monnet * A call to this helper is susceptible to change the underlaying 741ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 742ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 743ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 744ad4a5223SQuentin Monnet * direct packet access. 745ad4a5223SQuentin Monnet * Return 746ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 747c456dec4SQuentin Monnet * 748c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 749c456dec4SQuentin Monnet * Return 750c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 751c456dec4SQuentin Monnet * created as such: 752c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 753c456dec4SQuentin Monnet * *current_task*\ **->pid**. 754c456dec4SQuentin Monnet * 755c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 756c456dec4SQuentin Monnet * Return 757c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 758c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 759c456dec4SQuentin Monnet * 760c456dec4SQuentin Monnet * int bpf_get_current_comm(char *buf, u32 size_of_buf) 761c456dec4SQuentin Monnet * Description 762c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 763c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 764c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 765c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 766c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 767c456dec4SQuentin Monnet * it is filled with zeroes. 768c456dec4SQuentin Monnet * Return 769c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 770c456dec4SQuentin Monnet * 7711fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 7721fdd08beSQuentin Monnet * Description 7731fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 7741fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 7751fdd08beSQuentin Monnet * 7761fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 7771fdd08beSQuentin Monnet * 7781fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 7791fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 7801fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 7811fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 7821fdd08beSQuentin Monnet * *Documentation/cgroup-v1/net_cls.txt*. 7831fdd08beSQuentin Monnet * 7841fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 7851fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 7861fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 7871fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 7881fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 7891fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 7901fdd08beSQuentin Monnet * 7911fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 7921fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 7931fdd08beSQuentin Monnet * "**y**" or to "**m**". 7941fdd08beSQuentin Monnet * Return 7951fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 7961fdd08beSQuentin Monnet * 797c456dec4SQuentin Monnet * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 798c456dec4SQuentin Monnet * Description 799c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 800c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 801c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 802c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 803c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 804c456dec4SQuentin Monnet * 805c456dec4SQuentin Monnet * A call to this helper is susceptible to change the underlaying 806c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 807c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 808c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 809c456dec4SQuentin Monnet * direct packet access. 810c456dec4SQuentin Monnet * Return 811c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 812c456dec4SQuentin Monnet * 813c456dec4SQuentin Monnet * int bpf_skb_vlan_pop(struct sk_buff *skb) 814c456dec4SQuentin Monnet * Description 815c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 816c456dec4SQuentin Monnet * 817c456dec4SQuentin Monnet * A call to this helper is susceptible to change the underlaying 818c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 819c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 820c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 821c456dec4SQuentin Monnet * direct packet access. 822c456dec4SQuentin Monnet * Return 823c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 824c456dec4SQuentin Monnet * 825c456dec4SQuentin Monnet * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 826c456dec4SQuentin Monnet * Description 827c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 828c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 829c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 830c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 831c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 832c456dec4SQuentin Monnet * IPv4. 833c456dec4SQuentin Monnet * 834c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 835c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 836c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 837c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 838c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 839c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 840c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 841c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 842c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 843c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 844c456dec4SQuentin Monnet * () helper. 845c456dec4SQuentin Monnet * 846c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 847c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 848c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 849c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 850c456dec4SQuentin Monnet * 851c456dec4SQuentin Monnet * :: 852c456dec4SQuentin Monnet * 853c456dec4SQuentin Monnet * int ret; 854c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 855c456dec4SQuentin Monnet * 856c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 857c456dec4SQuentin Monnet * if (ret < 0) 858c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 859c456dec4SQuentin Monnet * 860c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 861c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 862c456dec4SQuentin Monnet * 863c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 864c456dec4SQuentin Monnet * 865c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 866c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 867c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 868c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 869c456dec4SQuentin Monnet * configuration can be extracted from this helper. 870c456dec4SQuentin Monnet * 871c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 872c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 873c456dec4SQuentin Monnet * Return 874c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 875c456dec4SQuentin Monnet * 876c456dec4SQuentin Monnet * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 877c456dec4SQuentin Monnet * Description 878c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 879c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 880c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 881c456dec4SQuentin Monnet * 882c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 883c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 884c456dec4SQuentin Monnet * instead of IPv4. 885c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 886c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 887c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 888c456dec4SQuentin Monnet * and checksum set to zeroes. 889c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 890c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 891c456dec4SQuentin Monnet * packet should not be fragmented. 892c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 893c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 894c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 895c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 896c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 897c456dec4SQuentin Monnet * as well in the future. 898c456dec4SQuentin Monnet * 899c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 900c456dec4SQuentin Monnet * 901c456dec4SQuentin Monnet * :: 902c456dec4SQuentin Monnet * 903c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 904c456dec4SQuentin Monnet * populate key ... 905c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 906c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 907c456dec4SQuentin Monnet * 908c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 909c456dec4SQuentin Monnet * helper for additional information. 910c456dec4SQuentin Monnet * Return 911c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 912c456dec4SQuentin Monnet * 913c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 914c6b5fb86SQuentin Monnet * Description 915c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 916c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 917c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 918c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 919c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 920c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 921c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 922c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 923c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 924c6b5fb86SQuentin Monnet * current CPU should be retrieved. 925c6b5fb86SQuentin Monnet * 926c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 927c6b5fb86SQuentin Monnet * retrieved. 928c6b5fb86SQuentin Monnet * 929c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 930c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 9313bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 932c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 933c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 9343bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 9353bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 9363bd5a09bSQuentin Monnet * () interface. Please refer to the description of 937c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 938c6b5fb86SQuentin Monnet * Return 939c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 940c6b5fb86SQuentin Monnet * negative error code in case of failure. 941c6b5fb86SQuentin Monnet * 942c456dec4SQuentin Monnet * int bpf_redirect(u32 ifindex, u64 flags) 943c456dec4SQuentin Monnet * Description 944c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 945c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 946c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 947c456dec4SQuentin Monnet * increased performance. 948c456dec4SQuentin Monnet * 949c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 950c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 951c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 952c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 953c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 954c456dec4SQuentin Monnet * flag at all. 955c456dec4SQuentin Monnet * 956c456dec4SQuentin Monnet * The same effect can be attained with the more generic 957c456dec4SQuentin Monnet * **bpf_redirect_map**\ (), which requires specific maps to be 958c456dec4SQuentin Monnet * used but offers better performance. 959c456dec4SQuentin Monnet * Return 960c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 961c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 962c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 963c456dec4SQuentin Monnet * error. 964c456dec4SQuentin Monnet * 9651fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 9661fdd08beSQuentin Monnet * Description 9671fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 9681fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 9691fdd08beSQuentin Monnet * indentifier retrieved is a user-provided tag, similar to the 9701fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 9711fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 9721fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 9731fdd08beSQuentin Monnet * 9741fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 9751fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 9761fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 9771fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 9781fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 9791fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 9801fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 9811fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 9821fdd08beSQuentin Monnet * 9831fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 9841fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 9851fdd08beSQuentin Monnet * Return 9861fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 9871fdd08beSQuentin Monnet * if none was found. 9881fdd08beSQuentin Monnet * 989c456dec4SQuentin Monnet * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 990c456dec4SQuentin Monnet * Description 991c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 992c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 993c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 994c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 995c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 996c456dec4SQuentin Monnet * 997c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 998c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 999c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 1000c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 1001c456dec4SQuentin Monnet * used. 1002c456dec4SQuentin Monnet * 1003c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 1004c456dec4SQuentin Monnet * pointed by *data*. 1005c456dec4SQuentin Monnet * 1006c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 1007c456dec4SQuentin Monnet * helper. 1008c456dec4SQuentin Monnet * 1009c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 1010c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 1011c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 1012c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 1013c456dec4SQuentin Monnet * into it. An example is available in file 1014c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 1015c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 1016c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 1017c456dec4SQuentin Monnet * 1018c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 1019c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 1020c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 1021c456dec4SQuentin Monnet * programs. 1022c456dec4SQuentin Monnet * 1023c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 1024c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 1025c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 1026c456dec4SQuentin Monnet * can be: 1027c456dec4SQuentin Monnet * 1028c456dec4SQuentin Monnet * * Only custom structs, 1029c456dec4SQuentin Monnet * * Only the packet payload, or 1030c456dec4SQuentin Monnet * * A combination of both. 1031c456dec4SQuentin Monnet * Return 1032c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 1033c456dec4SQuentin Monnet * 10341fdd08beSQuentin Monnet * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 10351fdd08beSQuentin Monnet * Description 10361fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 10371fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 10381fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 10391fdd08beSQuentin Monnet * *to*. 10401fdd08beSQuentin Monnet * 10411fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 10421fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 10431fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 10441fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 10451fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 10461fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 10471fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 10481fdd08beSQuentin Monnet * Return 10491fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 10501fdd08beSQuentin Monnet * 1051c456dec4SQuentin Monnet * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) 1052c456dec4SQuentin Monnet * Description 1053c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 1054c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 1055c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 1056c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 1057c456dec4SQuentin Monnet * 1058c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 1059c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 1060c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1061c456dec4SQuentin Monnet * a combination of the following flags: 1062c456dec4SQuentin Monnet * 1063c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 1064c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 1065c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 1066c456dec4SQuentin Monnet * Compare stacks by hash only. 1067c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 1068c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 1069c456dec4SQuentin Monnet * discard the old one. 1070c456dec4SQuentin Monnet * 1071c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 1072c456dec4SQuentin Monnet * can be further combined with other data (including other stack 1073c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 1074c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 1075c456dec4SQuentin Monnet * graphs). 1076c456dec4SQuentin Monnet * 1077c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 1078c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 1079c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 1080c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 1081c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1082c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 1083c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 1084c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 1085c456dec4SQuentin Monnet * 1086c456dec4SQuentin Monnet * :: 1087c456dec4SQuentin Monnet * 1088c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 1089c456dec4SQuentin Monnet * Return 1090c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 1091c456dec4SQuentin Monnet * in case of failure. 1092c456dec4SQuentin Monnet * 10931fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 10941fdd08beSQuentin Monnet * Description 10951fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 10961fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 10971fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 10981fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 10991fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 11001fdd08beSQuentin Monnet * to the helper). 11011fdd08beSQuentin Monnet * 11021fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 11031fdd08beSQuentin Monnet * 11041fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 11051fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 11061fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 11071fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 11081fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 11091fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 11101fdd08beSQuentin Monnet * *to_size* do not need to be equal. 11111fdd08beSQuentin Monnet * 11121fdd08beSQuentin Monnet * This helper can be used in combination with 11131fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 11141fdd08beSQuentin Monnet * which one can feed in the difference computed with 11151fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 11161fdd08beSQuentin Monnet * Return 11171fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 11181fdd08beSQuentin Monnet * failure. 11191fdd08beSQuentin Monnet * 11201fdd08beSQuentin Monnet * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 11211fdd08beSQuentin Monnet * Description 11221fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 11231fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 11241fdd08beSQuentin Monnet * of *size*. 11251fdd08beSQuentin Monnet * 11261fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 11271fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 11281fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 11291fdd08beSQuentin Monnet * more details). A particular example where this can be used is 11301fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 11311fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 11321fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 11331fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 11341fdd08beSQuentin Monnet * headers. 11351fdd08beSQuentin Monnet * Return 11361fdd08beSQuentin Monnet * The size of the option data retrieved. 11371fdd08beSQuentin Monnet * 11381fdd08beSQuentin Monnet * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 11391fdd08beSQuentin Monnet * Description 11401fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 11411fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 11421fdd08beSQuentin Monnet * 11431fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 11441fdd08beSQuentin Monnet * helper for additional information. 11451fdd08beSQuentin Monnet * Return 11461fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 11471fdd08beSQuentin Monnet * 11481fdd08beSQuentin Monnet * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 11491fdd08beSQuentin Monnet * Description 11501fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 11511fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 11521fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 11531fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 11541fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 11551fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 11561fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 11571fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 11581fdd08beSQuentin Monnet * operations out of an eBPF program. 11591fdd08beSQuentin Monnet * 11601fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 11611fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 11621fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 11631fdd08beSQuentin Monnet * 11641fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 11651fdd08beSQuentin Monnet * be left at zero. 11661fdd08beSQuentin Monnet * 11671fdd08beSQuentin Monnet * A call to this helper is susceptible to change the underlaying 11681fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 11691fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 11701fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 11711fdd08beSQuentin Monnet * direct packet access. 11721fdd08beSQuentin Monnet * Return 11731fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 11741fdd08beSQuentin Monnet * 11751fdd08beSQuentin Monnet * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 11761fdd08beSQuentin Monnet * Description 11771fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 11781fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 11791fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 11801fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 11811fdd08beSQuentin Monnet * for graceful handling of errors. 11821fdd08beSQuentin Monnet * 11831fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 11841fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 11851fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 11861fdd08beSQuentin Monnet * example. 11871fdd08beSQuentin Monnet * 11881fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 11891fdd08beSQuentin Monnet * are: 11901fdd08beSQuentin Monnet * 11911fdd08beSQuentin Monnet * **PACKET_HOST** 11921fdd08beSQuentin Monnet * Packet is for us. 11931fdd08beSQuentin Monnet * **PACKET_BROADCAST** 11941fdd08beSQuentin Monnet * Send packet to all. 11951fdd08beSQuentin Monnet * **PACKET_MULTICAST** 11961fdd08beSQuentin Monnet * Send packet to group. 11971fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 11981fdd08beSQuentin Monnet * Send packet to someone else. 11991fdd08beSQuentin Monnet * Return 12001fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 12011fdd08beSQuentin Monnet * 1202c6b5fb86SQuentin Monnet * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1203c6b5fb86SQuentin Monnet * Description 1204c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 1205c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1206c6b5fb86SQuentin Monnet * Return 1207c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 1208c6b5fb86SQuentin Monnet * 1209c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 1210c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 1211c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 1212c6b5fb86SQuentin Monnet * 1213fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1214fa15601aSQuentin Monnet * Description 1215fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1216fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 1217fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 1218fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 1219fa15601aSQuentin Monnet * 1220fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 1221fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 1222fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 1223fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1224fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 1225fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 1226fa15601aSQuentin Monnet * Return 1227fa15601aSQuentin Monnet * The 32-bit hash. 1228fa15601aSQuentin Monnet * 1229c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 1230c456dec4SQuentin Monnet * Return 1231c456dec4SQuentin Monnet * A pointer to the current task struct. 1232fa15601aSQuentin Monnet * 1233c6b5fb86SQuentin Monnet * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1234c6b5fb86SQuentin Monnet * Description 1235c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 1236c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 1237c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 1238c6b5fb86SQuentin Monnet * 1239c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 1240c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 1241c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 1242c6b5fb86SQuentin Monnet * processes. 1243c6b5fb86SQuentin Monnet * 1244c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 1245c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 1246c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 1247c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 1248c6b5fb86SQuentin Monnet * logs. 1249c6b5fb86SQuentin Monnet * Return 1250c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1251c6b5fb86SQuentin Monnet * 1252c6b5fb86SQuentin Monnet * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1253c6b5fb86SQuentin Monnet * Description 1254c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 1255c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1256c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1257c6b5fb86SQuentin Monnet * Return 1258c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 1259c6b5fb86SQuentin Monnet * 1260c6b5fb86SQuentin Monnet * * 0, if the *skb* task belongs to the cgroup2. 1261c6b5fb86SQuentin Monnet * * 1, if the *skb* task does not belong to the cgroup2. 1262c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 1263c6b5fb86SQuentin Monnet * 1264fa15601aSQuentin Monnet * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1265fa15601aSQuentin Monnet * Description 1266fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 1267fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 1268fa15601aSQuentin Monnet * be left at zero. 1269fa15601aSQuentin Monnet * 1270fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 1271fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 1272fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 1273fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1274fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 1275fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 1276fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 1277fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 1278fa15601aSQuentin Monnet * *skb*. 1279fa15601aSQuentin Monnet * 1280fa15601aSQuentin Monnet * A call to this helper is susceptible to change the underlaying 1281fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1282fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1283fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1284fa15601aSQuentin Monnet * direct packet access. 1285fa15601aSQuentin Monnet * Return 1286fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1287fa15601aSQuentin Monnet * 1288fa15601aSQuentin Monnet * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1289fa15601aSQuentin Monnet * Description 1290fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 1291fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 1292fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 1293fa15601aSQuentin Monnet * *len*, then the whole length of the *skb* is pulled. 1294fa15601aSQuentin Monnet * 1295fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 1296fa15601aSQuentin Monnet * packet access. 1297fa15601aSQuentin Monnet * 1298fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 1299fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 1300fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 1301fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 1302fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 1303fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 1304fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 1305fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 1306fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 1307fa15601aSQuentin Monnet * eventually access the data. 1308fa15601aSQuentin Monnet * 1309fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 1310fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 1311fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 1312fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 1313fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1314fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 1315fa15601aSQuentin Monnet * 1316fa15601aSQuentin Monnet * A call to this helper is susceptible to change the underlaying 1317fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1318fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1319fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1320fa15601aSQuentin Monnet * direct packet access. 1321fa15601aSQuentin Monnet * Return 1322fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1323fa15601aSQuentin Monnet * 1324fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1325fa15601aSQuentin Monnet * Description 1326fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 1327fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 1328fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 1329fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 1330fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 1331fa15601aSQuentin Monnet * written into the packet through direct packet access. 1332fa15601aSQuentin Monnet * Return 1333fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 1334fa15601aSQuentin Monnet * failure. 1335fa15601aSQuentin Monnet * 1336fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 1337fa15601aSQuentin Monnet * Description 1338fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 1339fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 1340fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 1341fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 1342fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1343fa15601aSQuentin Monnet * 1344fa15601aSQuentin Monnet * int bpf_get_numa_node_id(void) 1345fa15601aSQuentin Monnet * Description 1346fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 1347fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 1348fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 1349fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1350fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 1351fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 1352fa15601aSQuentin Monnet * Return 1353fa15601aSQuentin Monnet * The id of current NUMA node. 1354fa15601aSQuentin Monnet * 1355c6b5fb86SQuentin Monnet * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1356c6b5fb86SQuentin Monnet * Description 1357c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 1358c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 1359c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 1360c6b5fb86SQuentin Monnet * required. 1361c6b5fb86SQuentin Monnet * 1362c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 1363c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 1364c6b5fb86SQuentin Monnet * 1365c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 1366c6b5fb86SQuentin Monnet * be left at zero. 1367c6b5fb86SQuentin Monnet * 1368c6b5fb86SQuentin Monnet * A call to this helper is susceptible to change the underlaying 1369c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1370c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 1371c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 1372c6b5fb86SQuentin Monnet * direct packet access. 1373c6b5fb86SQuentin Monnet * Return 1374c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1375c6b5fb86SQuentin Monnet * 1376c6b5fb86SQuentin Monnet * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1377c6b5fb86SQuentin Monnet * Description 1378c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1379c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 1380c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 1381c6b5fb86SQuentin Monnet * headers. 1382c6b5fb86SQuentin Monnet * 1383c6b5fb86SQuentin Monnet * A call to this helper is susceptible to change the underlaying 1384c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1385c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 1386c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 1387c6b5fb86SQuentin Monnet * direct packet access. 1388c6b5fb86SQuentin Monnet * Return 1389c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1390c6b5fb86SQuentin Monnet * 1391c6b5fb86SQuentin Monnet * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1392c6b5fb86SQuentin Monnet * Description 1393c6b5fb86SQuentin Monnet * Copy a NUL terminated string from an unsafe address 1394c6b5fb86SQuentin Monnet * *unsafe_ptr* to *dst*. The *size* should include the 1395c6b5fb86SQuentin Monnet * terminating NUL byte. In case the string length is smaller than 1396c6b5fb86SQuentin Monnet * *size*, the target is not padded with further NUL bytes. If the 1397c6b5fb86SQuentin Monnet * string length is larger than *size*, just *size*-1 bytes are 1398c6b5fb86SQuentin Monnet * copied and the last byte is set to NUL. 1399c6b5fb86SQuentin Monnet * 1400c6b5fb86SQuentin Monnet * On success, the length of the copied string is returned. This 1401c6b5fb86SQuentin Monnet * makes this helper useful in tracing programs for reading 1402c6b5fb86SQuentin Monnet * strings, and more importantly to get its length at runtime. See 1403c6b5fb86SQuentin Monnet * the following snippet: 1404c6b5fb86SQuentin Monnet * 1405c6b5fb86SQuentin Monnet * :: 1406c6b5fb86SQuentin Monnet * 1407c6b5fb86SQuentin Monnet * SEC("kprobe/sys_open") 1408c6b5fb86SQuentin Monnet * void bpf_sys_open(struct pt_regs *ctx) 1409c6b5fb86SQuentin Monnet * { 1410c6b5fb86SQuentin Monnet * char buf[PATHLEN]; // PATHLEN is defined to 256 1411c6b5fb86SQuentin Monnet * int res = bpf_probe_read_str(buf, sizeof(buf), 1412c6b5fb86SQuentin Monnet * ctx->di); 1413c6b5fb86SQuentin Monnet * 1414c6b5fb86SQuentin Monnet * // Consume buf, for example push it to 1415c6b5fb86SQuentin Monnet * // userspace via bpf_perf_event_output(); we 1416c6b5fb86SQuentin Monnet * // can use res (the string length) as event 1417c6b5fb86SQuentin Monnet * // size, after checking its boundaries. 1418c6b5fb86SQuentin Monnet * } 1419c6b5fb86SQuentin Monnet * 1420c6b5fb86SQuentin Monnet * In comparison, using **bpf_probe_read()** helper here instead 1421c6b5fb86SQuentin Monnet * to read the string would require to estimate the length at 1422c6b5fb86SQuentin Monnet * compile time, and would often result in copying more memory 1423c6b5fb86SQuentin Monnet * than necessary. 1424c6b5fb86SQuentin Monnet * 1425c6b5fb86SQuentin Monnet * Another useful use case is when parsing individual process 1426c6b5fb86SQuentin Monnet * arguments or individual environment variables navigating 1427c6b5fb86SQuentin Monnet * *current*\ **->mm->arg_start** and *current*\ 1428c6b5fb86SQuentin Monnet * **->mm->env_start**: using this helper and the return value, 1429c6b5fb86SQuentin Monnet * one can quickly iterate at the right offset of the memory area. 1430c6b5fb86SQuentin Monnet * Return 1431c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 1432c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 1433c6b5fb86SQuentin Monnet * value. 1434c6b5fb86SQuentin Monnet * 1435c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1436c6b5fb86SQuentin Monnet * Description 1437c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 1438c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 1439c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 1440c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 1441c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 1442c6b5fb86SQuentin Monnet * networking traffic statistics as it provides a unique socket 1443c6b5fb86SQuentin Monnet * identifier per namespace. 1444c6b5fb86SQuentin Monnet * Return 1445c6b5fb86SQuentin Monnet * A 8-byte long non-decreasing number on success, or 0 if the 1446c6b5fb86SQuentin Monnet * socket field is missing inside *skb*. 1447c6b5fb86SQuentin Monnet * 1448d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_addr *ctx) 1449d692f113SAndrey Ignatov * Description 1450d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 145162369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_addr** context. 1452d692f113SAndrey Ignatov * Return 1453d692f113SAndrey Ignatov * A 8-byte long non-decreasing number. 1454d692f113SAndrey Ignatov * 1455d692f113SAndrey Ignatov * u64 bpf_get_socket_cookie(struct bpf_sock_ops *ctx) 1456d692f113SAndrey Ignatov * Description 1457d692f113SAndrey Ignatov * Equivalent to bpf_get_socket_cookie() helper that accepts 145862369db2SQuentin Monnet * *skb*, but gets socket from **struct bpf_sock_ops** context. 1459d692f113SAndrey Ignatov * Return 1460d692f113SAndrey Ignatov * A 8-byte long non-decreasing number. 1461d692f113SAndrey Ignatov * 1462c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 1463c6b5fb86SQuentin Monnet * Return 1464c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 1465c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 1466c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 1467c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 1468c6b5fb86SQuentin Monnet * UID value for the socket). 1469c6b5fb86SQuentin Monnet * 1470fa15601aSQuentin Monnet * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1471fa15601aSQuentin Monnet * Description 1472fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1473fa15601aSQuentin Monnet * to value *hash*. 1474fa15601aSQuentin Monnet * Return 1475fa15601aSQuentin Monnet * 0 1476fa15601aSQuentin Monnet * 1477a3ef8e9aSAndrey Ignatov * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 14787aa79a86SQuentin Monnet * Description 14797aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 14807aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 14817aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 14827aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 14837aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 14847aa79a86SQuentin Monnet * 14857aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 14867aa79a86SQuentin Monnet * It supports the following *level*\ s: 14877aa79a86SQuentin Monnet * 14887aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 14897aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 14907aa79a86SQuentin Monnet * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 14917aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 14927aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 14937aa79a86SQuentin Monnet * **TCP_BPF_SNDCWND_CLAMP**. 14947aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 14957aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 14967aa79a86SQuentin Monnet * Return 14977aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 14987aa79a86SQuentin Monnet * 1499b55cbc8dSNicolas Dichtel * int bpf_skb_adjust_room(struct sk_buff *skb, s32 len_diff, u32 mode, u64 flags) 1500fa15601aSQuentin Monnet * Description 1501fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 1502fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 1503fa15601aSQuentin Monnet * 150414aa3192SWillem de Bruijn * There are two supported modes at this time: 150514aa3192SWillem de Bruijn * 150614aa3192SWillem de Bruijn * * **BPF_ADJ_ROOM_MAC**: Adjust room at the mac layer 150714aa3192SWillem de Bruijn * (room space is added or removed below the layer 2 header). 1508fa15601aSQuentin Monnet * 1509fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1510fa15601aSQuentin Monnet * (room space is added or removed below the layer 3 header). 1511fa15601aSQuentin Monnet * 1512868d5235SWillem de Bruijn * The following flags are supported at this time: 15132278f6ccSWillem de Bruijn * 15142278f6ccSWillem de Bruijn * * **BPF_F_ADJ_ROOM_FIXED_GSO**: Do not adjust gso_size. 15152278f6ccSWillem de Bruijn * Adjusting mss in this way is not allowed for datagrams. 1516fa15601aSQuentin Monnet * 1517868d5235SWillem de Bruijn * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 **: 1518868d5235SWillem de Bruijn * * **BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 **: 1519868d5235SWillem de Bruijn * Any new space is reserved to hold a tunnel header. 1520868d5235SWillem de Bruijn * Configure skb offsets and other fields accordingly. 1521868d5235SWillem de Bruijn * 1522868d5235SWillem de Bruijn * * **BPF_F_ADJ_ROOM_ENCAP_L4_GRE **: 1523868d5235SWillem de Bruijn * * **BPF_F_ADJ_ROOM_ENCAP_L4_UDP **: 1524868d5235SWillem de Bruijn * Use with ENCAP_L3 flags to further specify the tunnel type. 1525868d5235SWillem de Bruijn * 1526fa15601aSQuentin Monnet * A call to this helper is susceptible to change the underlaying 1527fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1528fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1529fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1530fa15601aSQuentin Monnet * direct packet access. 1531fa15601aSQuentin Monnet * Return 1532fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1533fa15601aSQuentin Monnet * 1534ab127040SQuentin Monnet * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1535ab127040SQuentin Monnet * Description 1536ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 1537ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 1538ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 1539ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 1540ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 1541ab127040SQuentin Monnet * support) as of this writing). 1542ab127040SQuentin Monnet * 1543ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 1544ab127040SQuentin Monnet * be left at zero. 1545ab127040SQuentin Monnet * 1546ab127040SQuentin Monnet * When used to redirect packets to net devices, this helper 1547ab127040SQuentin Monnet * provides a high performance increase over **bpf_redirect**\ (). 1548ab127040SQuentin Monnet * This is due to various implementation details of the underlying 1549ab127040SQuentin Monnet * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1550ab127040SQuentin Monnet * () tries to send packet as a "bulk" to the device. 1551ab127040SQuentin Monnet * Return 1552ab127040SQuentin Monnet * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1553ab127040SQuentin Monnet * 1554ab127040SQuentin Monnet * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1555ab127040SQuentin Monnet * Description 1556ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 1557ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1558ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 1559ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 1560ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 1561ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 1562ab127040SQuentin Monnet * Return 1563ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 1564ab127040SQuentin Monnet * 1565a3ef8e9aSAndrey Ignatov * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1566ab127040SQuentin Monnet * Description 1567ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 1568ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 1569ab127040SQuentin Monnet * *key*. *flags* is one of: 1570ab127040SQuentin Monnet * 1571ab127040SQuentin Monnet * **BPF_NOEXIST** 1572ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 1573ab127040SQuentin Monnet * **BPF_EXIST** 1574ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 1575ab127040SQuentin Monnet * **BPF_ANY** 1576ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 1577ab127040SQuentin Monnet * 1578ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 1579ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 1580ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 1581ab127040SQuentin Monnet * Return 1582ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 1583ab127040SQuentin Monnet * 1584fa15601aSQuentin Monnet * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1585fa15601aSQuentin Monnet * Description 1586fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1587fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 1588fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 1589fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 1590fa15601aSQuentin Monnet * called. 1591fa15601aSQuentin Monnet * 1592fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 1593fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 1594fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 1595fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 1596fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 1597fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 1598fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 1599fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 1600fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 1601fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 1602fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 1603fa15601aSQuentin Monnet * data they need. 1604fa15601aSQuentin Monnet * 1605fa15601aSQuentin Monnet * A call to this helper is susceptible to change the underlaying 1606fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1607fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1608fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1609fa15601aSQuentin Monnet * direct packet access. 1610fa15601aSQuentin Monnet * Return 1611fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 16127aa79a86SQuentin Monnet * 16137aa79a86SQuentin Monnet * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 16147aa79a86SQuentin Monnet * Description 16157aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 16167aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 16177aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 16187aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 16197aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 16207aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 16217aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 16227aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 16237aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 16247aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 16257aa79a86SQuentin Monnet * current CPU should be retrieved. 16267aa79a86SQuentin Monnet * 16277aa79a86SQuentin Monnet * This helper behaves in a way close to 16287aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 16297aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 16307aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 16317aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 16327aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 16337aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 16347aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 16357aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 16367aa79a86SQuentin Monnet * 16377aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 16387aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 16397aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 16407aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 16417aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 16427aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 16437aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 16447aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 16457aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 16467aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 16477aa79a86SQuentin Monnet * as follows. 16487aa79a86SQuentin Monnet * 16497aa79a86SQuentin Monnet * :: 16507aa79a86SQuentin Monnet * 16517aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 16527aa79a86SQuentin Monnet * 16537aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 16547aa79a86SQuentin Monnet * the time running for event since last normalization. The 16557aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 16567aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 16577aa79a86SQuentin Monnet * eBPF program, users can can use CPU id as the key (which is 16587aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 16597aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 16607aa79a86SQuentin Monnet * Return 16617aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 16627aa79a86SQuentin Monnet * 1663a3ef8e9aSAndrey Ignatov * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 16647aa79a86SQuentin Monnet * Description 16657aa79a86SQuentin Monnet * For en eBPF program attached to a perf event, retrieve the 16667aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 16677aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 16687aa79a86SQuentin Monnet * and running times are also stored in the structure (see 16697aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 16707aa79a86SQuentin Monnet * more details). 16717aa79a86SQuentin Monnet * Return 16727aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 16737aa79a86SQuentin Monnet * 1674a3ef8e9aSAndrey Ignatov * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 16757aa79a86SQuentin Monnet * Description 16767aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 16777aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 16787aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 16797aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 16807aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 16817aa79a86SQuentin Monnet * *opval* and of length *optlen*. 16827aa79a86SQuentin Monnet * 16837aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 16847aa79a86SQuentin Monnet * It supports the following *level*\ s: 16857aa79a86SQuentin Monnet * 16867aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports *optname* 16877aa79a86SQuentin Monnet * **TCP_CONGESTION**. 16887aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 16897aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 16907aa79a86SQuentin Monnet * Return 16917aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 16927aa79a86SQuentin Monnet * 16937aa79a86SQuentin Monnet * int bpf_override_return(struct pt_reg *regs, u64 rc) 16947aa79a86SQuentin Monnet * Description 16957aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 16967aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 16977aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 16987aa79a86SQuentin Monnet * works. 16997aa79a86SQuentin Monnet * 17007aa79a86SQuentin Monnet * This helper works by setting setting the PC (program counter) 17017aa79a86SQuentin Monnet * to an override function which is run in place of the original 17027aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 17037aa79a86SQuentin Monnet * all. The replacement function just returns with the required 17047aa79a86SQuentin Monnet * value. 17057aa79a86SQuentin Monnet * 17067aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 17077aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 17087aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 17097aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 17107aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 17117aa79a86SQuentin Monnet * 17127aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 17137aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 17147aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 17157aa79a86SQuentin Monnet * Return 17167aa79a86SQuentin Monnet * 0 17177aa79a86SQuentin Monnet * 1718a3ef8e9aSAndrey Ignatov * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 17197aa79a86SQuentin Monnet * Description 17207aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 17217aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 17227aa79a86SQuentin Monnet * *argval*. 17237aa79a86SQuentin Monnet * 17247aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 17257aa79a86SQuentin Monnet * be calls to eBPF programs of type 17267aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 17277aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 17287aa79a86SQuentin Monnet * connection and as necessary, when the connection is 17297aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 17307aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 17317aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 17327aa79a86SQuentin Monnet * supported in the current kernel. 17337aa79a86SQuentin Monnet * 17347aa79a86SQuentin Monnet * The supported callback values that *argval* can combine are: 17357aa79a86SQuentin Monnet * 17367aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 17377aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 17387aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 17397aa79a86SQuentin Monnet * 17407aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 17417aa79a86SQuentin Monnet * program: 17427aa79a86SQuentin Monnet * 17437aa79a86SQuentin Monnet * * When RTO fires. 17447aa79a86SQuentin Monnet * * When a packet is retransmitted. 17457aa79a86SQuentin Monnet * * When the connection terminates. 17467aa79a86SQuentin Monnet * * When a packet is sent. 17477aa79a86SQuentin Monnet * * When a packet is received. 17487aa79a86SQuentin Monnet * Return 17497aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 17507aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 17517aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 17527aa79a86SQuentin Monnet * as required). 17537aa79a86SQuentin Monnet * 1754ab127040SQuentin Monnet * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1755ab127040SQuentin Monnet * Description 1756ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 1757ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 1758ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 1759ab127040SQuentin Monnet * the socket referenced by *map* (of type 1760ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1761ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 1762ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 1763ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 1764ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 1765ab127040SQuentin Monnet * Return 1766ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 1767ab127040SQuentin Monnet * 1768ab127040SQuentin Monnet * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1769ab127040SQuentin Monnet * Description 1770ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 1771ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 1772ab127040SQuentin Monnet * 1773ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 1774ab127040SQuentin Monnet * 1775ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 1776ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 1777ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 1778ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 1779ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 1780ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 1781ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 1782ab127040SQuentin Monnet * overhead. 1783ab127040SQuentin Monnet * 1784ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 1785ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 1786ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 1787ab127040SQuentin Monnet * smaller than the current data being processed from a 1788ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 1789ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 1790ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 1791ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 1792ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 1793ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1794ab127040SQuentin Monnet * consumed. 1795ab127040SQuentin Monnet * 1796ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 1797ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 1798ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 1799ab127040SQuentin Monnet * Return 1800ab127040SQuentin Monnet * 0 1801ab127040SQuentin Monnet * 1802ab127040SQuentin Monnet * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1803ab127040SQuentin Monnet * Description 1804ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 1805ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 1806ab127040SQuentin Monnet * accumulated. 1807ab127040SQuentin Monnet * 1808ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 1809ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 1810ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1811ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 1812ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 1813ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 1814ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 1815ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 1816ab127040SQuentin Monnet * been accumulated. 1817ab127040SQuentin Monnet * Return 1818ab127040SQuentin Monnet * 0 1819ab127040SQuentin Monnet * 1820ab127040SQuentin Monnet * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1821ab127040SQuentin Monnet * Description 1822ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 1823ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1824ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 1825ab127040SQuentin Monnet * respectively. 1826ab127040SQuentin Monnet * 1827ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1828ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 1829ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 1830ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 1831ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1832ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 1833ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 1834ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 1835ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 1836ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 1837ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 1838ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 1839ab127040SQuentin Monnet * 1840ab127040SQuentin Monnet * A call to this helper is susceptible to change the underlaying 1841ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1842ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 1843ab127040SQuentin Monnet * performed again, if the helper is used in combination with 1844ab127040SQuentin Monnet * direct packet access. 1845ab127040SQuentin Monnet * 1846ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 1847ab127040SQuentin Monnet * be left at zero. 1848ab127040SQuentin Monnet * Return 1849ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 1850ab127040SQuentin Monnet * 1851a3ef8e9aSAndrey Ignatov * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 18527aa79a86SQuentin Monnet * Description 18537aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 18547aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 18557aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 18567aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 18577aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 18587aa79a86SQuentin Monnet * 18597aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 18607aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 18617aa79a86SQuentin Monnet * **AF_INET6**). Looking for a free port to bind to can be 18627aa79a86SQuentin Monnet * expensive, therefore binding to port is not permitted by the 18637aa79a86SQuentin Monnet * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 18647aa79a86SQuentin Monnet * must be set to zero. 18657aa79a86SQuentin Monnet * Return 18667aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 18672d020dd7SQuentin Monnet * 18682d020dd7SQuentin Monnet * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 18692d020dd7SQuentin Monnet * Description 18702d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 18712d020dd7SQuentin Monnet * only possible to shrink the packet as of this writing, 18722d020dd7SQuentin Monnet * therefore *delta* must be a negative integer. 18732d020dd7SQuentin Monnet * 18742d020dd7SQuentin Monnet * A call to this helper is susceptible to change the underlaying 18752d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 18762d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 18772d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 18782d020dd7SQuentin Monnet * direct packet access. 18792d020dd7SQuentin Monnet * Return 18802d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 18812d020dd7SQuentin Monnet * 18822d020dd7SQuentin Monnet * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 18832d020dd7SQuentin Monnet * Description 18842d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 18852d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 18862d020dd7SQuentin Monnet * 18872d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 18882d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 18892d020dd7SQuentin Monnet * 18902d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 18912d020dd7SQuentin Monnet * be left at zero. 18922d020dd7SQuentin Monnet * 18932d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 18942d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 18952d020dd7SQuentin Monnet * Return 18962d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 1897c195651eSYonghong Song * 1898c195651eSYonghong Song * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1899c195651eSYonghong Song * Description 1900c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 1901c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 1902c195651eSYonghong Song * to the context on which the tracing program is executed. 1903c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 1904c195651eSYonghong Song * a nonnegative *size*. 1905c195651eSYonghong Song * 1906c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 1907c195651eSYonghong Song * skip (from 0 to 255), masked with 1908c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1909c195651eSYonghong Song * the following flags: 1910c195651eSYonghong Song * 1911c195651eSYonghong Song * **BPF_F_USER_STACK** 1912c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 1913c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 1914c195651eSYonghong Song * Collect buildid+offset instead of ips for user stack, 1915c195651eSYonghong Song * only valid if **BPF_F_USER_STACK** is also specified. 1916c195651eSYonghong Song * 1917c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 1918c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1919c195651eSYonghong Song * to sufficient large buffer size. Note that 1920c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 1921c195651eSYonghong Song * that it should be manually increased in order to profile long 1922c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 1923c195651eSYonghong Song * 1924c195651eSYonghong Song * :: 1925c195651eSYonghong Song * 1926c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 1927c195651eSYonghong Song * Return 19287a279e93SQuentin Monnet * A non-negative value equal to or less than *size* on success, 19297a279e93SQuentin Monnet * or a negative error in case of failure. 19304e1ec56cSDaniel Borkmann * 19312bae79d2SQuentin Monnet * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 19324e1ec56cSDaniel Borkmann * Description 19334e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 19344e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 19354e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 19364e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 19374e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 19384e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 19394e1ec56cSDaniel Borkmann * 19404e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 19414e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 19424e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 19434e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 19444e1ec56cSDaniel Borkmann * 19454e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 19464e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 19474e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 19484e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 19494e1ec56cSDaniel Borkmann * is not available. 19504e1ec56cSDaniel Borkmann * Return 19514e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 19524e1ec56cSDaniel Borkmann * 195387f5fc7eSDavid Ahern * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 195487f5fc7eSDavid Ahern * Description 195587f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 195687f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 195787f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 195887f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 1959fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 1960fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 1961fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 19624c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 19634c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 196487f5fc7eSDavid Ahern * 196587f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 19667a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 19677a279e93SQuentin Monnet * following values: 196887f5fc7eSDavid Ahern * 19697a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 19707a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 19717a279e93SQuentin Monnet * rules. 19727a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 19737a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 19747a279e93SQuentin Monnet * ingress). 197587f5fc7eSDavid Ahern * 197687f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 197787f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 197887f5fc7eSDavid Ahern * Return 19794c79579bSDavid Ahern * * < 0 if any input argument is invalid 19804c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 19814c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 19822bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 198381110384SJohn Fastabend * 198481110384SJohn Fastabend * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) 198581110384SJohn Fastabend * Description 198681110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 198781110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 198881110384SJohn Fastabend * *key*. *flags* is one of: 198981110384SJohn Fastabend * 199081110384SJohn Fastabend * **BPF_NOEXIST** 199181110384SJohn Fastabend * The entry for *key* must not exist in the map. 199281110384SJohn Fastabend * **BPF_EXIST** 199381110384SJohn Fastabend * The entry for *key* must already exist in the map. 199481110384SJohn Fastabend * **BPF_ANY** 199581110384SJohn Fastabend * No condition on the existence of the entry for *key*. 199681110384SJohn Fastabend * 199781110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 199881110384SJohn Fastabend * be inherited by the socket being added. If the socket is 199981110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 200081110384SJohn Fastabend * Return 200181110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 200281110384SJohn Fastabend * 200381110384SJohn Fastabend * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 200481110384SJohn Fastabend * Description 200581110384SJohn Fastabend * This helper is used in programs implementing policies at the 200681110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 200781110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 200881110384SJohn Fastabend * the socket referenced by *map* (of type 200981110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 201081110384SJohn Fastabend * egress interfaces can be used for redirection. The 201181110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 201281110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 201381110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 201481110384SJohn Fastabend * Return 201581110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 201681110384SJohn Fastabend * 201781110384SJohn Fastabend * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 201881110384SJohn Fastabend * Description 201981110384SJohn Fastabend * This helper is used in programs implementing policies at the 202081110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 202181110384SJohn Fastabend * if the verdeict eBPF program returns **SK_PASS**), redirect it 202281110384SJohn Fastabend * to the socket referenced by *map* (of type 202381110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 202481110384SJohn Fastabend * egress interfaces can be used for redirection. The 202581110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 202681110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 202781110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 202881110384SJohn Fastabend * Return 202981110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 2030fe94cc29SMathieu Xhonneux * 2031fe94cc29SMathieu Xhonneux * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 2032fe94cc29SMathieu Xhonneux * Description 2033fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 2034fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 2035fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 2036fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 2037fe94cc29SMathieu Xhonneux * 2038fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 2039fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 2040fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 2041fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 2042fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 2043fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 2044fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 2045fe94cc29SMathieu Xhonneux * the IPv6 header. 20463e0bd37cSPeter Oskolkov * **BPF_LWT_ENCAP_IP** 20473e0bd37cSPeter Oskolkov * IP encapsulation (GRE/GUE/IPIP/etc). The outer header 20483e0bd37cSPeter Oskolkov * must be IPv4 or IPv6, followed by zero or more 20493e0bd37cSPeter Oskolkov * additional headers, up to LWT_BPF_MAX_HEADROOM total 20503e0bd37cSPeter Oskolkov * bytes in all prepended headers. Please note that 20513e0bd37cSPeter Oskolkov * if skb_is_gso(skb) is true, no more than two headers 20523e0bd37cSPeter Oskolkov * can be prepended, and the inner header, if present, 20533e0bd37cSPeter Oskolkov * should be either GRE or UDP/GUE. 20543e0bd37cSPeter Oskolkov * 20553e0bd37cSPeter Oskolkov * BPF_LWT_ENCAP_SEG6*** types can be called by bpf programs of 20563e0bd37cSPeter Oskolkov * type BPF_PROG_TYPE_LWT_IN; BPF_LWT_ENCAP_IP type can be called 20573e0bd37cSPeter Oskolkov * by bpf programs of types BPF_PROG_TYPE_LWT_IN and 20583e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT. 2059fe94cc29SMathieu Xhonneux * 2060fe94cc29SMathieu Xhonneux * A call to this helper is susceptible to change the underlaying 2061fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2062fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2063fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2064fe94cc29SMathieu Xhonneux * direct packet access. 2065fe94cc29SMathieu Xhonneux * Return 2066fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2067fe94cc29SMathieu Xhonneux * 2068fe94cc29SMathieu Xhonneux * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 2069fe94cc29SMathieu Xhonneux * Description 2070fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 2071fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 2072fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 2073fe94cc29SMathieu Xhonneux * modified through this helper. 2074fe94cc29SMathieu Xhonneux * 2075fe94cc29SMathieu Xhonneux * A call to this helper is susceptible to change the underlaying 2076fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2077fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2078fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2079fe94cc29SMathieu Xhonneux * direct packet access. 2080fe94cc29SMathieu Xhonneux * Return 2081fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2082fe94cc29SMathieu Xhonneux * 2083fe94cc29SMathieu Xhonneux * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 2084fe94cc29SMathieu Xhonneux * Description 2085fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 2086fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 2087fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 2088fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 2089fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 2090fe94cc29SMathieu Xhonneux * 2091fe94cc29SMathieu Xhonneux * A call to this helper is susceptible to change the underlaying 2092fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2093fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2094fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2095fe94cc29SMathieu Xhonneux * direct packet access. 2096fe94cc29SMathieu Xhonneux * Return 2097fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2098fe94cc29SMathieu Xhonneux * 2099fe94cc29SMathieu Xhonneux * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 2100fe94cc29SMathieu Xhonneux * Description 2101fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 2102fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 2103fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 2104fe94cc29SMathieu Xhonneux * *action* can be one of: 2105fe94cc29SMathieu Xhonneux * 2106fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 2107fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 2108fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 2109fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 2110fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 2111fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 2112fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 2113fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 2114fe94cc29SMathieu Xhonneux * Type of param: **struct ipv6_sr_hdr**. 2115fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2116fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 2117fe94cc29SMathieu Xhonneux * encapsulation policy. 2118fe94cc29SMathieu Xhonneux * Type of param: **struct ipv6_sr_hdr**. 2119fe94cc29SMathieu Xhonneux * 2120fe94cc29SMathieu Xhonneux * A call to this helper is susceptible to change the underlaying 2121fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2122fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2123fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2124fe94cc29SMathieu Xhonneux * direct packet access. 2125fe94cc29SMathieu Xhonneux * Return 2126fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2127f4364dcfSSean Young * 212862369db2SQuentin Monnet * int bpf_rc_repeat(void *ctx) 212962369db2SQuentin Monnet * Description 213062369db2SQuentin Monnet * This helper is used in programs implementing IR decoding, to 213162369db2SQuentin Monnet * report a successfully decoded repeat key message. This delays 213262369db2SQuentin Monnet * the generation of a key up event for previously generated 213362369db2SQuentin Monnet * key down event. 213462369db2SQuentin Monnet * 213562369db2SQuentin Monnet * Some IR protocols like NEC have a special IR message for 213662369db2SQuentin Monnet * repeating last button, for when a button is held down. 213762369db2SQuentin Monnet * 213862369db2SQuentin Monnet * The *ctx* should point to the lirc sample as passed into 213962369db2SQuentin Monnet * the program. 214062369db2SQuentin Monnet * 214162369db2SQuentin Monnet * This helper is only available is the kernel was compiled with 214262369db2SQuentin Monnet * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 214362369db2SQuentin Monnet * "**y**". 214462369db2SQuentin Monnet * Return 214562369db2SQuentin Monnet * 0 214662369db2SQuentin Monnet * 2147f4364dcfSSean Young * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2148f4364dcfSSean Young * Description 2149f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 2150f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 2151f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 2152f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 2153f4364dcfSSean Young * an input key down event. After a period a key up event is 2154f4364dcfSSean Young * generated. This period can be extended by calling either 215590b1023fSQuentin Monnet * **bpf_rc_keydown**\ () again with the same values, or calling 215690b1023fSQuentin Monnet * **bpf_rc_repeat**\ (). 2157f4364dcfSSean Young * 2158f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 2159f4364dcfSSean Young * released and pressed again between consecutive scancodes. 2160f4364dcfSSean Young * 2161f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 2162f4364dcfSSean Young * the program. 2163f4364dcfSSean Young * 2164f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 2165f4364dcfSSean Young * **enum rc_proto** for some predefined values). 2166f4364dcfSSean Young * 2167f4364dcfSSean Young * This helper is only available is the kernel was compiled with 2168f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2169f4364dcfSSean Young * "**y**". 2170f4364dcfSSean Young * Return 2171f4364dcfSSean Young * 0 2172f4364dcfSSean Young * 217362369db2SQuentin Monnet * u64 bpf_skb_cgroup_id(struct sk_buff *skb) 2174cb20b08eSDaniel Borkmann * Description 2175cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 2176cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2177cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 2178cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 2179cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 2180cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 2181cb20b08eSDaniel Borkmann * to the same 64-bit id. 2182cb20b08eSDaniel Borkmann * 2183cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 2184cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 2185cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2186cb20b08eSDaniel Borkmann * Return 2187cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 2188bf6fa2c8SYonghong Song * 2189bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 2190bf6fa2c8SYonghong Song * Return 2191bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 2192bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 2193cd339431SRoman Gushchin * 219462369db2SQuentin Monnet * void *bpf_get_local_storage(void *map, u64 flags) 2195cd339431SRoman Gushchin * Description 2196cd339431SRoman Gushchin * Get the pointer to the local storage area. 2197cd339431SRoman Gushchin * The type and the size of the local storage is defined 2198cd339431SRoman Gushchin * by the *map* argument. 2199cd339431SRoman Gushchin * The *flags* meaning is specific for each map type, 2200cd339431SRoman Gushchin * and has to be 0 for cgroup local storage. 2201cd339431SRoman Gushchin * 220290b1023fSQuentin Monnet * Depending on the BPF program type, a local storage area 220390b1023fSQuentin Monnet * can be shared between multiple instances of the BPF program, 2204cd339431SRoman Gushchin * running simultaneously. 2205cd339431SRoman Gushchin * 2206cd339431SRoman Gushchin * A user should care about the synchronization by himself. 220790b1023fSQuentin Monnet * For example, by using the **BPF_STX_XADD** instruction to alter 2208cd339431SRoman Gushchin * the shared data. 2209cd339431SRoman Gushchin * Return 221090b1023fSQuentin Monnet * A pointer to the local storage area. 22112dbb9b9eSMartin KaFai Lau * 22122dbb9b9eSMartin KaFai Lau * int bpf_sk_select_reuseport(struct sk_reuseport_md *reuse, struct bpf_map *map, void *key, u64 flags) 22132dbb9b9eSMartin KaFai Lau * Description 221490b1023fSQuentin Monnet * Select a **SO_REUSEPORT** socket from a 221590b1023fSQuentin Monnet * **BPF_MAP_TYPE_REUSEPORT_ARRAY** *map*. 221690b1023fSQuentin Monnet * It checks the selected socket is matching the incoming 221790b1023fSQuentin Monnet * request in the socket buffer. 22182dbb9b9eSMartin KaFai Lau * Return 22192dbb9b9eSMartin KaFai Lau * 0 on success, or a negative error in case of failure. 22206acc9b43SJoe Stringer * 222162369db2SQuentin Monnet * u64 bpf_skb_ancestor_cgroup_id(struct sk_buff *skb, int ancestor_level) 222262369db2SQuentin Monnet * Description 222362369db2SQuentin Monnet * Return id of cgroup v2 that is ancestor of cgroup associated 222462369db2SQuentin Monnet * with the *skb* at the *ancestor_level*. The root cgroup is at 222562369db2SQuentin Monnet * *ancestor_level* zero and each step down the hierarchy 222662369db2SQuentin Monnet * increments the level. If *ancestor_level* == level of cgroup 222762369db2SQuentin Monnet * associated with *skb*, then return value will be same as that 222862369db2SQuentin Monnet * of **bpf_skb_cgroup_id**\ (). 222962369db2SQuentin Monnet * 223062369db2SQuentin Monnet * The helper is useful to implement policies based on cgroups 223162369db2SQuentin Monnet * that are upper in hierarchy than immediate cgroup associated 223262369db2SQuentin Monnet * with *skb*. 223362369db2SQuentin Monnet * 223462369db2SQuentin Monnet * The format of returned id and helper limitations are same as in 223562369db2SQuentin Monnet * **bpf_skb_cgroup_id**\ (). 223662369db2SQuentin Monnet * Return 223762369db2SQuentin Monnet * The id is returned or 0 in case the id could not be retrieved. 223862369db2SQuentin Monnet * 2239f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 22406acc9b43SJoe Stringer * Description 22416acc9b43SJoe Stringer * Look for TCP socket matching *tuple*, optionally in a child 22426acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 224390b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 22446acc9b43SJoe Stringer * 22456acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 22466acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 22476acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 22486acc9b43SJoe Stringer * 22496acc9b43SJoe Stringer * *tuple_size* must be one of: 22506acc9b43SJoe Stringer * 22516acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 22526acc9b43SJoe Stringer * Look for an IPv4 socket. 22536acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 22546acc9b43SJoe Stringer * Look for an IPv6 socket. 22556acc9b43SJoe Stringer * 2256f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 2257f71c6143SJoe Stringer * socket lookup table in the netns associated with the *ctx* will 2258f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 2259f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 2260f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 2261f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 2262f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 2263f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 22646acc9b43SJoe Stringer * 22656acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 22666acc9b43SJoe Stringer * be left at zero. 22676acc9b43SJoe Stringer * 22686acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 22696acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 22706acc9b43SJoe Stringer * Return 22710bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 22720bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 22730bd72117SDaniel Borkmann * result is from **reuse->socks**\ [] using the hash of the tuple. 22746acc9b43SJoe Stringer * 2275f71c6143SJoe Stringer * struct bpf_sock *bpf_sk_lookup_udp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 22766acc9b43SJoe Stringer * Description 22776acc9b43SJoe Stringer * Look for UDP socket matching *tuple*, optionally in a child 22786acc9b43SJoe Stringer * network namespace *netns*. The return value must be checked, 227990b1023fSQuentin Monnet * and if non-**NULL**, released via **bpf_sk_release**\ (). 22806acc9b43SJoe Stringer * 22816acc9b43SJoe Stringer * The *ctx* should point to the context of the program, such as 22826acc9b43SJoe Stringer * the skb or socket (depending on the hook in use). This is used 22836acc9b43SJoe Stringer * to determine the base network namespace for the lookup. 22846acc9b43SJoe Stringer * 22856acc9b43SJoe Stringer * *tuple_size* must be one of: 22866acc9b43SJoe Stringer * 22876acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv4**) 22886acc9b43SJoe Stringer * Look for an IPv4 socket. 22896acc9b43SJoe Stringer * **sizeof**\ (*tuple*\ **->ipv6**) 22906acc9b43SJoe Stringer * Look for an IPv6 socket. 22916acc9b43SJoe Stringer * 2292f71c6143SJoe Stringer * If the *netns* is a negative signed 32-bit integer, then the 2293f71c6143SJoe Stringer * socket lookup table in the netns associated with the *ctx* will 2294f71c6143SJoe Stringer * will be used. For the TC hooks, this is the netns of the device 2295f71c6143SJoe Stringer * in the skb. For socket hooks, this is the netns of the socket. 2296f71c6143SJoe Stringer * If *netns* is any other signed 32-bit value greater than or 2297f71c6143SJoe Stringer * equal to zero then it specifies the ID of the netns relative to 2298f71c6143SJoe Stringer * the netns associated with the *ctx*. *netns* values beyond the 2299f71c6143SJoe Stringer * range of 32-bit integers are reserved for future use. 23006acc9b43SJoe Stringer * 23016acc9b43SJoe Stringer * All values for *flags* are reserved for future usage, and must 23026acc9b43SJoe Stringer * be left at zero. 23036acc9b43SJoe Stringer * 23046acc9b43SJoe Stringer * This helper is available only if the kernel was compiled with 23056acc9b43SJoe Stringer * **CONFIG_NET** configuration option. 23066acc9b43SJoe Stringer * Return 23070bd72117SDaniel Borkmann * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 23080bd72117SDaniel Borkmann * For sockets with reuseport option, the **struct bpf_sock** 23090bd72117SDaniel Borkmann * result is from **reuse->socks**\ [] using the hash of the tuple. 23106acc9b43SJoe Stringer * 231190b1023fSQuentin Monnet * int bpf_sk_release(struct bpf_sock *sock) 23126acc9b43SJoe Stringer * Description 231390b1023fSQuentin Monnet * Release the reference held by *sock*. *sock* must be a 231490b1023fSQuentin Monnet * non-**NULL** pointer that was returned from 231590b1023fSQuentin Monnet * **bpf_sk_lookup_xxx**\ (). 231690b1023fSQuentin Monnet * Return 231790b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 231890b1023fSQuentin Monnet * 231962369db2SQuentin Monnet * int bpf_map_push_elem(struct bpf_map *map, const void *value, u64 flags) 232062369db2SQuentin Monnet * Description 232162369db2SQuentin Monnet * Push an element *value* in *map*. *flags* is one of: 232262369db2SQuentin Monnet * 232362369db2SQuentin Monnet * **BPF_EXIST** 232462369db2SQuentin Monnet * If the queue/stack is full, the oldest element is 232562369db2SQuentin Monnet * removed to make room for this. 232662369db2SQuentin Monnet * Return 232762369db2SQuentin Monnet * 0 on success, or a negative error in case of failure. 232862369db2SQuentin Monnet * 232990b1023fSQuentin Monnet * int bpf_map_pop_elem(struct bpf_map *map, void *value) 233090b1023fSQuentin Monnet * Description 233190b1023fSQuentin Monnet * Pop an element from *map*. 233290b1023fSQuentin Monnet * Return 233390b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 233490b1023fSQuentin Monnet * 233590b1023fSQuentin Monnet * int bpf_map_peek_elem(struct bpf_map *map, void *value) 233690b1023fSQuentin Monnet * Description 233790b1023fSQuentin Monnet * Get an element from *map* without removing it. 23386acc9b43SJoe Stringer * Return 23396acc9b43SJoe Stringer * 0 on success, or a negative error in case of failure. 23406fff607eSJohn Fastabend * 23416fff607eSJohn Fastabend * int bpf_msg_push_data(struct sk_buff *skb, u32 start, u32 len, u64 flags) 23426fff607eSJohn Fastabend * Description 234390b1023fSQuentin Monnet * For socket policies, insert *len* bytes into *msg* at offset 23446fff607eSJohn Fastabend * *start*. 23456fff607eSJohn Fastabend * 23466fff607eSJohn Fastabend * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 234790b1023fSQuentin Monnet * *msg* it may want to insert metadata or options into the *msg*. 23486fff607eSJohn Fastabend * This can later be read and used by any of the lower layer BPF 23496fff607eSJohn Fastabend * hooks. 23506fff607eSJohn Fastabend * 23516fff607eSJohn Fastabend * This helper may fail if under memory pressure (a malloc 23526fff607eSJohn Fastabend * fails) in these cases BPF programs will get an appropriate 23536fff607eSJohn Fastabend * error and BPF programs will need to handle them. 23546fff607eSJohn Fastabend * Return 23556fff607eSJohn Fastabend * 0 on success, or a negative error in case of failure. 23567246d8edSJohn Fastabend * 23577246d8edSJohn Fastabend * int bpf_msg_pop_data(struct sk_msg_buff *msg, u32 start, u32 pop, u64 flags) 23587246d8edSJohn Fastabend * Description 23597246d8edSJohn Fastabend * Will remove *pop* bytes from a *msg* starting at byte *start*. 23607246d8edSJohn Fastabend * This may result in **ENOMEM** errors under certain situations if 23617246d8edSJohn Fastabend * an allocation and copy are required due to a full ring buffer. 23627246d8edSJohn Fastabend * However, the helper will try to avoid doing the allocation 23637246d8edSJohn Fastabend * if possible. Other errors can occur if input parameters are 236490b1023fSQuentin Monnet * invalid either due to *start* byte not being valid part of *msg* 23657246d8edSJohn Fastabend * payload and/or *pop* value being to large. 23667246d8edSJohn Fastabend * Return 236790b1023fSQuentin Monnet * 0 on success, or a negative error in case of failure. 236801d3240aSSean Young * 236901d3240aSSean Young * int bpf_rc_pointer_rel(void *ctx, s32 rel_x, s32 rel_y) 237001d3240aSSean Young * Description 237101d3240aSSean Young * This helper is used in programs implementing IR decoding, to 237201d3240aSSean Young * report a successfully decoded pointer movement. 237301d3240aSSean Young * 237401d3240aSSean Young * The *ctx* should point to the lirc sample as passed into 237501d3240aSSean Young * the program. 237601d3240aSSean Young * 237701d3240aSSean Young * This helper is only available is the kernel was compiled with 237801d3240aSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 237901d3240aSSean Young * "**y**". 238001d3240aSSean Young * Return 238101d3240aSSean Young * 0 238246f8bc92SMartin KaFai Lau * 23830eb09785SQuentin Monnet * int bpf_spin_lock(struct bpf_spin_lock *lock) 23840eb09785SQuentin Monnet * Description 23850eb09785SQuentin Monnet * Acquire a spinlock represented by the pointer *lock*, which is 23860eb09785SQuentin Monnet * stored as part of a value of a map. Taking the lock allows to 23870eb09785SQuentin Monnet * safely update the rest of the fields in that value. The 23880eb09785SQuentin Monnet * spinlock can (and must) later be released with a call to 23890eb09785SQuentin Monnet * **bpf_spin_unlock**\ (\ *lock*\ ). 23900eb09785SQuentin Monnet * 23910eb09785SQuentin Monnet * Spinlocks in BPF programs come with a number of restrictions 23920eb09785SQuentin Monnet * and constraints: 23930eb09785SQuentin Monnet * 23940eb09785SQuentin Monnet * * **bpf_spin_lock** objects are only allowed inside maps of 23950eb09785SQuentin Monnet * types **BPF_MAP_TYPE_HASH** and **BPF_MAP_TYPE_ARRAY** (this 23960eb09785SQuentin Monnet * list could be extended in the future). 23970eb09785SQuentin Monnet * * BTF description of the map is mandatory. 23980eb09785SQuentin Monnet * * The BPF program can take ONE lock at a time, since taking two 23990eb09785SQuentin Monnet * or more could cause dead locks. 24000eb09785SQuentin Monnet * * Only one **struct bpf_spin_lock** is allowed per map element. 24010eb09785SQuentin Monnet * * When the lock is taken, calls (either BPF to BPF or helpers) 24020eb09785SQuentin Monnet * are not allowed. 24030eb09785SQuentin Monnet * * The **BPF_LD_ABS** and **BPF_LD_IND** instructions are not 24040eb09785SQuentin Monnet * allowed inside a spinlock-ed region. 24050eb09785SQuentin Monnet * * The BPF program MUST call **bpf_spin_unlock**\ () to release 24060eb09785SQuentin Monnet * the lock, on all execution paths, before it returns. 24070eb09785SQuentin Monnet * * The BPF program can access **struct bpf_spin_lock** only via 24080eb09785SQuentin Monnet * the **bpf_spin_lock**\ () and **bpf_spin_unlock**\ () 24090eb09785SQuentin Monnet * helpers. Loading or storing data into the **struct 24100eb09785SQuentin Monnet * bpf_spin_lock** *lock*\ **;** field of a map is not allowed. 24110eb09785SQuentin Monnet * * To use the **bpf_spin_lock**\ () helper, the BTF description 24120eb09785SQuentin Monnet * of the map value must be a struct and have **struct 24130eb09785SQuentin Monnet * bpf_spin_lock** *anyname*\ **;** field at the top level. 24140eb09785SQuentin Monnet * Nested lock inside another struct is not allowed. 24150eb09785SQuentin Monnet * * The **struct bpf_spin_lock** *lock* field in a map value must 24160eb09785SQuentin Monnet * be aligned on a multiple of 4 bytes in that value. 24170eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_LOOKUP_ELEM** does not copy 24180eb09785SQuentin Monnet * the **bpf_spin_lock** field to user space. 24190eb09785SQuentin Monnet * * Syscall with command **BPF_MAP_UPDATE_ELEM**, or update from 24200eb09785SQuentin Monnet * a BPF program, do not update the **bpf_spin_lock** field. 24210eb09785SQuentin Monnet * * **bpf_spin_lock** cannot be on the stack or inside a 24220eb09785SQuentin Monnet * networking packet (it can only be inside of a map values). 24230eb09785SQuentin Monnet * * **bpf_spin_lock** is available to root only. 24240eb09785SQuentin Monnet * * Tracing programs and socket filter programs cannot use 24250eb09785SQuentin Monnet * **bpf_spin_lock**\ () due to insufficient preemption checks 24260eb09785SQuentin Monnet * (but this may change in the future). 24270eb09785SQuentin Monnet * * **bpf_spin_lock** is not allowed in inner maps of map-in-map. 24280eb09785SQuentin Monnet * Return 24290eb09785SQuentin Monnet * 0 24300eb09785SQuentin Monnet * 24310eb09785SQuentin Monnet * int bpf_spin_unlock(struct bpf_spin_lock *lock) 24320eb09785SQuentin Monnet * Description 24330eb09785SQuentin Monnet * Release the *lock* previously locked by a call to 24340eb09785SQuentin Monnet * **bpf_spin_lock**\ (\ *lock*\ ). 24350eb09785SQuentin Monnet * Return 24360eb09785SQuentin Monnet * 0 24370eb09785SQuentin Monnet * 243846f8bc92SMartin KaFai Lau * struct bpf_sock *bpf_sk_fullsock(struct bpf_sock *sk) 243946f8bc92SMartin KaFai Lau * Description 244046f8bc92SMartin KaFai Lau * This helper gets a **struct bpf_sock** pointer such 244162369db2SQuentin Monnet * that all the fields in this **bpf_sock** can be accessed. 244246f8bc92SMartin KaFai Lau * Return 244362369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 244446f8bc92SMartin KaFai Lau * case of failure. 2445655a51e5SMartin KaFai Lau * 2446655a51e5SMartin KaFai Lau * struct bpf_tcp_sock *bpf_tcp_sock(struct bpf_sock *sk) 2447655a51e5SMartin KaFai Lau * Description 2448655a51e5SMartin KaFai Lau * This helper gets a **struct bpf_tcp_sock** pointer from a 2449655a51e5SMartin KaFai Lau * **struct bpf_sock** pointer. 2450655a51e5SMartin KaFai Lau * Return 245162369db2SQuentin Monnet * A **struct bpf_tcp_sock** pointer on success, or **NULL** in 2452655a51e5SMartin KaFai Lau * case of failure. 2453f7c917baSbrakmo * 2454f7c917baSbrakmo * int bpf_skb_ecn_set_ce(struct sk_buf *skb) 2455f7c917baSbrakmo * Description 245662369db2SQuentin Monnet * Set ECN (Explicit Congestion Notification) field of IP header 245762369db2SQuentin Monnet * to **CE** (Congestion Encountered) if current value is **ECT** 245862369db2SQuentin Monnet * (ECN Capable Transport). Otherwise, do nothing. Works with IPv6 245962369db2SQuentin Monnet * and IPv4. 2460f7c917baSbrakmo * Return 246162369db2SQuentin Monnet * 1 if the **CE** flag is set (either by the current helper call 246262369db2SQuentin Monnet * or because it was already present), 0 if it is not set. 2463dbafd7ddSMartin KaFai Lau * 2464dbafd7ddSMartin KaFai Lau * struct bpf_sock *bpf_get_listener_sock(struct bpf_sock *sk) 2465dbafd7ddSMartin KaFai Lau * Description 246662369db2SQuentin Monnet * Return a **struct bpf_sock** pointer in **TCP_LISTEN** state. 246762369db2SQuentin Monnet * **bpf_sk_release**\ () is unnecessary and not allowed. 2468dbafd7ddSMartin KaFai Lau * Return 246962369db2SQuentin Monnet * A **struct bpf_sock** pointer on success, or **NULL** in 2470dbafd7ddSMartin KaFai Lau * case of failure. 2471edbf8c01SLorenz Bauer * 2472edbf8c01SLorenz Bauer * struct bpf_sock *bpf_skc_lookup_tcp(void *ctx, struct bpf_sock_tuple *tuple, u32 tuple_size, u64 netns, u64 flags) 2473edbf8c01SLorenz Bauer * Description 2474edbf8c01SLorenz Bauer * Look for TCP socket matching *tuple*, optionally in a child 2475edbf8c01SLorenz Bauer * network namespace *netns*. The return value must be checked, 2476edbf8c01SLorenz Bauer * and if non-**NULL**, released via **bpf_sk_release**\ (). 2477edbf8c01SLorenz Bauer * 2478edbf8c01SLorenz Bauer * This function is identical to bpf_sk_lookup_tcp, except that it 2479edbf8c01SLorenz Bauer * also returns timewait or request sockets. Use bpf_sk_fullsock 2480edbf8c01SLorenz Bauer * or bpf_tcp_socket to access the full structure. 2481edbf8c01SLorenz Bauer * 2482edbf8c01SLorenz Bauer * This helper is available only if the kernel was compiled with 2483edbf8c01SLorenz Bauer * **CONFIG_NET** configuration option. 2484edbf8c01SLorenz Bauer * Return 2485edbf8c01SLorenz Bauer * Pointer to **struct bpf_sock**, or **NULL** in case of failure. 2486edbf8c01SLorenz Bauer * For sockets with reuseport option, the **struct bpf_sock** 2487edbf8c01SLorenz Bauer * result is from **reuse->socks**\ [] using the hash of the tuple. 248839904084SLorenz Bauer * 248939904084SLorenz Bauer * int bpf_tcp_check_syncookie(struct bpf_sock *sk, void *iph, u32 iph_len, struct tcphdr *th, u32 th_len) 249039904084SLorenz Bauer * Description 249139904084SLorenz Bauer * Check whether iph and th contain a valid SYN cookie ACK for 249239904084SLorenz Bauer * the listening socket in sk. 249339904084SLorenz Bauer * 249439904084SLorenz Bauer * iph points to the start of the IPv4 or IPv6 header, while 249539904084SLorenz Bauer * iph_len contains sizeof(struct iphdr) or sizeof(struct ip6hdr). 249639904084SLorenz Bauer * 249739904084SLorenz Bauer * th points to the start of the TCP header, while th_len contains 249839904084SLorenz Bauer * sizeof(struct tcphdr). 249939904084SLorenz Bauer * 250039904084SLorenz Bauer * Return 250139904084SLorenz Bauer * 0 if iph and th are a valid SYN cookie ACK, or a negative error 250239904084SLorenz Bauer * otherwise. 25037a4b28c6SDaniel Borkmann */ 2504ebb676daSThomas Graf #define __BPF_FUNC_MAPPER(FN) \ 2505ebb676daSThomas Graf FN(unspec), \ 2506ebb676daSThomas Graf FN(map_lookup_elem), \ 2507ebb676daSThomas Graf FN(map_update_elem), \ 2508ebb676daSThomas Graf FN(map_delete_elem), \ 2509ebb676daSThomas Graf FN(probe_read), \ 2510ebb676daSThomas Graf FN(ktime_get_ns), \ 2511ebb676daSThomas Graf FN(trace_printk), \ 2512ebb676daSThomas Graf FN(get_prandom_u32), \ 2513ebb676daSThomas Graf FN(get_smp_processor_id), \ 2514ebb676daSThomas Graf FN(skb_store_bytes), \ 2515ebb676daSThomas Graf FN(l3_csum_replace), \ 2516ebb676daSThomas Graf FN(l4_csum_replace), \ 2517ebb676daSThomas Graf FN(tail_call), \ 2518ebb676daSThomas Graf FN(clone_redirect), \ 2519ebb676daSThomas Graf FN(get_current_pid_tgid), \ 2520ebb676daSThomas Graf FN(get_current_uid_gid), \ 2521ebb676daSThomas Graf FN(get_current_comm), \ 2522ebb676daSThomas Graf FN(get_cgroup_classid), \ 2523ebb676daSThomas Graf FN(skb_vlan_push), \ 2524ebb676daSThomas Graf FN(skb_vlan_pop), \ 2525ebb676daSThomas Graf FN(skb_get_tunnel_key), \ 2526ebb676daSThomas Graf FN(skb_set_tunnel_key), \ 2527ebb676daSThomas Graf FN(perf_event_read), \ 2528ebb676daSThomas Graf FN(redirect), \ 2529ebb676daSThomas Graf FN(get_route_realm), \ 2530ebb676daSThomas Graf FN(perf_event_output), \ 2531ebb676daSThomas Graf FN(skb_load_bytes), \ 2532ebb676daSThomas Graf FN(get_stackid), \ 2533ebb676daSThomas Graf FN(csum_diff), \ 2534ebb676daSThomas Graf FN(skb_get_tunnel_opt), \ 2535ebb676daSThomas Graf FN(skb_set_tunnel_opt), \ 2536ebb676daSThomas Graf FN(skb_change_proto), \ 2537ebb676daSThomas Graf FN(skb_change_type), \ 2538ebb676daSThomas Graf FN(skb_under_cgroup), \ 2539ebb676daSThomas Graf FN(get_hash_recalc), \ 2540ebb676daSThomas Graf FN(get_current_task), \ 2541ebb676daSThomas Graf FN(probe_write_user), \ 2542ebb676daSThomas Graf FN(current_task_under_cgroup), \ 2543ebb676daSThomas Graf FN(skb_change_tail), \ 2544ebb676daSThomas Graf FN(skb_pull_data), \ 2545ebb676daSThomas Graf FN(csum_update), \ 2546ebb676daSThomas Graf FN(set_hash_invalid), \ 25473a0af8fdSThomas Graf FN(get_numa_node_id), \ 254817bedab2SMartin KaFai Lau FN(skb_change_head), \ 2549a5e8c070SGianluca Borello FN(xdp_adjust_head), \ 255091b8270fSChenbo Feng FN(probe_read_str), \ 25516acc5c29SChenbo Feng FN(get_socket_cookie), \ 2552ded092cdSDaniel Borkmann FN(get_socket_uid), \ 25538c4b4c7eSLawrence Brakmo FN(set_hash), \ 25542be7e212SDaniel Borkmann FN(setsockopt), \ 255597f91a7cSJohn Fastabend FN(skb_adjust_room), \ 2556174a79ffSJohn Fastabend FN(redirect_map), \ 2557174a79ffSJohn Fastabend FN(sk_redirect_map), \ 2558174a79ffSJohn Fastabend FN(sock_map_update), \ 2559908432caSYonghong Song FN(xdp_adjust_meta), \ 25604bebdc7aSYonghong Song FN(perf_event_read_value), \ 2561cd86d1fdSLawrence Brakmo FN(perf_prog_read_value), \ 25629802d865SJosef Bacik FN(getsockopt), \ 2563b13d8807SLawrence Brakmo FN(override_return), \ 25644f738adbSJohn Fastabend FN(sock_ops_cb_flags_set), \ 25652a100317SJohn Fastabend FN(msg_redirect_map), \ 256691843d54SJohn Fastabend FN(msg_apply_bytes), \ 2567015632bbSJohn Fastabend FN(msg_cork_bytes), \ 2568d74bad4eSAndrey Ignatov FN(msg_pull_data), \ 2569b32cc5b9SNikita V. Shirokov FN(bind), \ 257012bed760SEyal Birger FN(xdp_adjust_tail), \ 2571c195651eSYonghong Song FN(skb_get_xfrm_state), \ 25724e1ec56cSDaniel Borkmann FN(get_stack), \ 257387f5fc7eSDavid Ahern FN(skb_load_bytes_relative), \ 257481110384SJohn Fastabend FN(fib_lookup), \ 257581110384SJohn Fastabend FN(sock_hash_update), \ 257681110384SJohn Fastabend FN(msg_redirect_hash), \ 2577fe94cc29SMathieu Xhonneux FN(sk_redirect_hash), \ 2578fe94cc29SMathieu Xhonneux FN(lwt_push_encap), \ 2579fe94cc29SMathieu Xhonneux FN(lwt_seg6_store_bytes), \ 2580fe94cc29SMathieu Xhonneux FN(lwt_seg6_adjust_srh), \ 2581f4364dcfSSean Young FN(lwt_seg6_action), \ 2582f4364dcfSSean Young FN(rc_repeat), \ 2583cb20b08eSDaniel Borkmann FN(rc_keydown), \ 2584bf6fa2c8SYonghong Song FN(skb_cgroup_id), \ 2585cd339431SRoman Gushchin FN(get_current_cgroup_id), \ 25862dbb9b9eSMartin KaFai Lau FN(get_local_storage), \ 258777236281SAndrey Ignatov FN(sk_select_reuseport), \ 25886acc9b43SJoe Stringer FN(skb_ancestor_cgroup_id), \ 25896acc9b43SJoe Stringer FN(sk_lookup_tcp), \ 25906acc9b43SJoe Stringer FN(sk_lookup_udp), \ 2591f1a2e44aSMauricio Vasquez B FN(sk_release), \ 2592f1a2e44aSMauricio Vasquez B FN(map_push_elem), \ 2593f1a2e44aSMauricio Vasquez B FN(map_pop_elem), \ 25946fff607eSJohn Fastabend FN(map_peek_elem), \ 25957246d8edSJohn Fastabend FN(msg_push_data), \ 259601d3240aSSean Young FN(msg_pop_data), \ 2597d83525caSAlexei Starovoitov FN(rc_pointer_rel), \ 2598d83525caSAlexei Starovoitov FN(spin_lock), \ 259946f8bc92SMartin KaFai Lau FN(spin_unlock), \ 2600655a51e5SMartin KaFai Lau FN(sk_fullsock), \ 2601f7c917baSbrakmo FN(tcp_sock), \ 2602dbafd7ddSMartin KaFai Lau FN(skb_ecn_set_ce), \ 2603edbf8c01SLorenz Bauer FN(get_listener_sock), \ 260439904084SLorenz Bauer FN(skc_lookup_tcp), \ 260539904084SLorenz Bauer FN(tcp_check_syncookie), 26067a4b28c6SDaniel Borkmann 2607ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2608ebb676daSThomas Graf * function eBPF program intends to call 26092d0e30c3SDaniel Borkmann */ 2610ebb676daSThomas Graf #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2611ebb676daSThomas Graf enum bpf_func_id { 2612ebb676daSThomas Graf __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 261309756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 261409756af4SAlexei Starovoitov }; 2615ebb676daSThomas Graf #undef __BPF_ENUM_FN 261609756af4SAlexei Starovoitov 2617781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 2618781c53bcSDaniel Borkmann 2619781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 2620781c53bcSDaniel Borkmann #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 26218afd54c8SDaniel Borkmann #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2622781c53bcSDaniel Borkmann 2623781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2624781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 2625781c53bcSDaniel Borkmann */ 2626781c53bcSDaniel Borkmann #define BPF_F_HDR_FIELD_MASK 0xfULL 2627781c53bcSDaniel Borkmann 2628781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 2629781c53bcSDaniel Borkmann #define BPF_F_PSEUDO_HDR (1ULL << 4) 26302f72959aSDaniel Borkmann #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2631d1b662adSDaniel Borkmann #define BPF_F_MARK_ENFORCE (1ULL << 6) 2632781c53bcSDaniel Borkmann 2633781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2634781c53bcSDaniel Borkmann #define BPF_F_INGRESS (1ULL << 0) 2635781c53bcSDaniel Borkmann 2636c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2637c6c33454SDaniel Borkmann #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2638c6c33454SDaniel Borkmann 2639c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2640d5a3b1f6SAlexei Starovoitov #define BPF_F_SKIP_FIELD_MASK 0xffULL 2641d5a3b1f6SAlexei Starovoitov #define BPF_F_USER_STACK (1ULL << 8) 2642c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 2643d5a3b1f6SAlexei Starovoitov #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2644d5a3b1f6SAlexei Starovoitov #define BPF_F_REUSE_STACKID (1ULL << 10) 2645c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 2646c195651eSYonghong Song #define BPF_F_USER_BUILD_ID (1ULL << 11) 2647d5a3b1f6SAlexei Starovoitov 26482da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 26492da897e5SDaniel Borkmann #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 265022080870SDaniel Borkmann #define BPF_F_DONT_FRAGMENT (1ULL << 2) 265177a5196aSWilliam Tu #define BPF_F_SEQ_NUMBER (1ULL << 3) 26522da897e5SDaniel Borkmann 2653908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2654908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 2655908432caSYonghong Song */ 26561e33759cSDaniel Borkmann #define BPF_F_INDEX_MASK 0xffffffffULL 26571e33759cSDaniel Borkmann #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2658555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2659555c8a86SDaniel Borkmann #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 26601e33759cSDaniel Borkmann 2661f71c6143SJoe Stringer /* Current network namespace */ 2662f71c6143SJoe Stringer #define BPF_F_CURRENT_NETNS (-1L) 2663f71c6143SJoe Stringer 26642278f6ccSWillem de Bruijn /* BPF_FUNC_skb_adjust_room flags. */ 26652278f6ccSWillem de Bruijn #define BPF_F_ADJ_ROOM_FIXED_GSO (1ULL << 0) 26662278f6ccSWillem de Bruijn 2667868d5235SWillem de Bruijn #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV4 (1ULL << 1) 2668868d5235SWillem de Bruijn #define BPF_F_ADJ_ROOM_ENCAP_L3_IPV6 (1ULL << 2) 2669868d5235SWillem de Bruijn #define BPF_F_ADJ_ROOM_ENCAP_L4_GRE (1ULL << 3) 2670868d5235SWillem de Bruijn #define BPF_F_ADJ_ROOM_ENCAP_L4_UDP (1ULL << 4) 2671868d5235SWillem de Bruijn 26722be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 26732be7e212SDaniel Borkmann enum bpf_adj_room_mode { 26742be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 267514aa3192SWillem de Bruijn BPF_ADJ_ROOM_MAC, 26762be7e212SDaniel Borkmann }; 26772be7e212SDaniel Borkmann 26784e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 26794e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 26804e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 26814e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 26824e1ec56cSDaniel Borkmann }; 26834e1ec56cSDaniel Borkmann 2684fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 2685fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 2686fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 26873e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_SEG6_INLINE, 26883e0bd37cSPeter Oskolkov BPF_LWT_ENCAP_IP, 2689fe94cc29SMathieu Xhonneux }; 2690fe94cc29SMathieu Xhonneux 2691b7df9adaSDaniel Borkmann #define __bpf_md_ptr(type, name) \ 2692b7df9adaSDaniel Borkmann union { \ 2693b7df9adaSDaniel Borkmann type name; \ 2694b7df9adaSDaniel Borkmann __u64 :64; \ 2695b7df9adaSDaniel Borkmann } __attribute__((aligned(8))) 2696b7df9adaSDaniel Borkmann 26979bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 26989bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 26999bac3d6dSAlexei Starovoitov */ 27009bac3d6dSAlexei Starovoitov struct __sk_buff { 27019bac3d6dSAlexei Starovoitov __u32 len; 27029bac3d6dSAlexei Starovoitov __u32 pkt_type; 27039bac3d6dSAlexei Starovoitov __u32 mark; 27049bac3d6dSAlexei Starovoitov __u32 queue_mapping; 2705c2497395SAlexei Starovoitov __u32 protocol; 2706c2497395SAlexei Starovoitov __u32 vlan_present; 2707c2497395SAlexei Starovoitov __u32 vlan_tci; 270827cd5452SMichal Sekletar __u32 vlan_proto; 2709bcad5718SDaniel Borkmann __u32 priority; 271037e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 271137e82c2fSAlexei Starovoitov __u32 ifindex; 2712d691f9e8SAlexei Starovoitov __u32 tc_index; 2713d691f9e8SAlexei Starovoitov __u32 cb[5]; 2714ba7591d8SDaniel Borkmann __u32 hash; 2715045efa82SDaniel Borkmann __u32 tc_classid; 2716969bf05eSAlexei Starovoitov __u32 data; 2717969bf05eSAlexei Starovoitov __u32 data_end; 2718b1d9fc41SDaniel Borkmann __u32 napi_id; 27198a31db56SJohn Fastabend 2720de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 27218a31db56SJohn Fastabend __u32 family; 27228a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 27238a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 27248a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 27258a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 27268a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 27278a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 2728de8f3a83SDaniel Borkmann /* ... here. */ 2729de8f3a83SDaniel Borkmann 2730de8f3a83SDaniel Borkmann __u32 data_meta; 2731b7df9adaSDaniel Borkmann __bpf_md_ptr(struct bpf_flow_keys *, flow_keys); 2732f11216b2SVlad Dumitrescu __u64 tstamp; 2733e3da08d0SPetar Penkov __u32 wire_len; 2734d9ff286aSEric Dumazet __u32 gso_segs; 273546f8bc92SMartin KaFai Lau __bpf_md_ptr(struct bpf_sock *, sk); 27369bac3d6dSAlexei Starovoitov }; 27379bac3d6dSAlexei Starovoitov 2738d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 2739d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 2740c6c33454SDaniel Borkmann union { 2741d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 2742c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 2743c6c33454SDaniel Borkmann }; 2744c6c33454SDaniel Borkmann __u8 tunnel_tos; 2745c6c33454SDaniel Borkmann __u8 tunnel_ttl; 27461fbc2e0cSDaniel Borkmann __u16 tunnel_ext; /* Padding, future use. */ 27474018ab18SDaniel Borkmann __u32 tunnel_label; 2748d3aa45ceSAlexei Starovoitov }; 2749d3aa45ceSAlexei Starovoitov 275012bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 275112bed760SEyal Birger * new fields can only be added to the end of this structure 275212bed760SEyal Birger */ 275312bed760SEyal Birger struct bpf_xfrm_state { 275412bed760SEyal Birger __u32 reqid; 275512bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 275612bed760SEyal Birger __u16 family; 27571fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 275812bed760SEyal Birger union { 275912bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 276012bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 276112bed760SEyal Birger }; 276212bed760SEyal Birger }; 276312bed760SEyal Birger 27643a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 27653a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 27663a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 27673a0af8fdSThomas Graf * programs. 27683a0af8fdSThomas Graf * 27693a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 27703a0af8fdSThomas Graf */ 27713a0af8fdSThomas Graf enum bpf_ret_code { 27723a0af8fdSThomas Graf BPF_OK = 0, 27733a0af8fdSThomas Graf /* 1 reserved */ 27743a0af8fdSThomas Graf BPF_DROP = 2, 27753a0af8fdSThomas Graf /* 3-6 reserved */ 27763a0af8fdSThomas Graf BPF_REDIRECT = 7, 27773e0bd37cSPeter Oskolkov /* >127 are reserved for prog type specific return codes. 27783e0bd37cSPeter Oskolkov * 27793e0bd37cSPeter Oskolkov * BPF_LWT_REROUTE: used by BPF_PROG_TYPE_LWT_IN and 27803e0bd37cSPeter Oskolkov * BPF_PROG_TYPE_LWT_XMIT to indicate that skb had been 27813e0bd37cSPeter Oskolkov * changed and should be routed based on its new L3 header. 27823e0bd37cSPeter Oskolkov * (This is an L3 redirect, as opposed to L2 redirect 27833e0bd37cSPeter Oskolkov * represented by BPF_REDIRECT above). 27843e0bd37cSPeter Oskolkov */ 27853e0bd37cSPeter Oskolkov BPF_LWT_REROUTE = 128, 27863a0af8fdSThomas Graf }; 27873a0af8fdSThomas Graf 278861023658SDavid Ahern struct bpf_sock { 278961023658SDavid Ahern __u32 bound_dev_if; 2790aa4c1037SDavid Ahern __u32 family; 2791aa4c1037SDavid Ahern __u32 type; 2792aa4c1037SDavid Ahern __u32 protocol; 2793482dca93SDavid Ahern __u32 mark; 2794482dca93SDavid Ahern __u32 priority; 2795aa65d696SMartin KaFai Lau /* IP address also allows 1 and 2 bytes access */ 2796aa65d696SMartin KaFai Lau __u32 src_ip4; 2797aa65d696SMartin KaFai Lau __u32 src_ip6[4]; 2798aa65d696SMartin KaFai Lau __u32 src_port; /* host byte order */ 2799aa65d696SMartin KaFai Lau __u32 dst_port; /* network byte order */ 2800aa65d696SMartin KaFai Lau __u32 dst_ip4; 2801aa65d696SMartin KaFai Lau __u32 dst_ip6[4]; 2802aa65d696SMartin KaFai Lau __u32 state; 280361023658SDavid Ahern }; 280461023658SDavid Ahern 2805655a51e5SMartin KaFai Lau struct bpf_tcp_sock { 2806655a51e5SMartin KaFai Lau __u32 snd_cwnd; /* Sending congestion window */ 2807655a51e5SMartin KaFai Lau __u32 srtt_us; /* smoothed round trip time << 3 in usecs */ 2808655a51e5SMartin KaFai Lau __u32 rtt_min; 2809655a51e5SMartin KaFai Lau __u32 snd_ssthresh; /* Slow start size threshold */ 2810655a51e5SMartin KaFai Lau __u32 rcv_nxt; /* What we want to receive next */ 2811655a51e5SMartin KaFai Lau __u32 snd_nxt; /* Next sequence we send */ 2812655a51e5SMartin KaFai Lau __u32 snd_una; /* First byte we want an ack for */ 2813655a51e5SMartin KaFai Lau __u32 mss_cache; /* Cached effective mss, not including SACKS */ 2814655a51e5SMartin KaFai Lau __u32 ecn_flags; /* ECN status bits. */ 2815655a51e5SMartin KaFai Lau __u32 rate_delivered; /* saved rate sample: packets delivered */ 2816655a51e5SMartin KaFai Lau __u32 rate_interval_us; /* saved rate sample: time elapsed */ 2817655a51e5SMartin KaFai Lau __u32 packets_out; /* Packets which are "in flight" */ 2818655a51e5SMartin KaFai Lau __u32 retrans_out; /* Retransmitted packets out */ 2819655a51e5SMartin KaFai Lau __u32 total_retrans; /* Total retransmits for entire connection */ 2820655a51e5SMartin KaFai Lau __u32 segs_in; /* RFC4898 tcpEStatsPerfSegsIn 2821655a51e5SMartin KaFai Lau * total number of segments in. 2822655a51e5SMartin KaFai Lau */ 2823655a51e5SMartin KaFai Lau __u32 data_segs_in; /* RFC4898 tcpEStatsPerfDataSegsIn 2824655a51e5SMartin KaFai Lau * total number of data segments in. 2825655a51e5SMartin KaFai Lau */ 2826655a51e5SMartin KaFai Lau __u32 segs_out; /* RFC4898 tcpEStatsPerfSegsOut 2827655a51e5SMartin KaFai Lau * The total number of segments sent. 2828655a51e5SMartin KaFai Lau */ 2829655a51e5SMartin KaFai Lau __u32 data_segs_out; /* RFC4898 tcpEStatsPerfDataSegsOut 2830655a51e5SMartin KaFai Lau * total number of data segments sent. 2831655a51e5SMartin KaFai Lau */ 2832655a51e5SMartin KaFai Lau __u32 lost_out; /* Lost packets */ 2833655a51e5SMartin KaFai Lau __u32 sacked_out; /* SACK'd packets */ 2834655a51e5SMartin KaFai Lau __u64 bytes_received; /* RFC4898 tcpEStatsAppHCThruOctetsReceived 2835655a51e5SMartin KaFai Lau * sum(delta(rcv_nxt)), or how many bytes 2836655a51e5SMartin KaFai Lau * were acked. 2837655a51e5SMartin KaFai Lau */ 2838655a51e5SMartin KaFai Lau __u64 bytes_acked; /* RFC4898 tcpEStatsAppHCThruOctetsAcked 2839655a51e5SMartin KaFai Lau * sum(delta(snd_una)), or how many bytes 2840655a51e5SMartin KaFai Lau * were acked. 2841655a51e5SMartin KaFai Lau */ 2842655a51e5SMartin KaFai Lau }; 2843655a51e5SMartin KaFai Lau 28446acc9b43SJoe Stringer struct bpf_sock_tuple { 28456acc9b43SJoe Stringer union { 28466acc9b43SJoe Stringer struct { 28476acc9b43SJoe Stringer __be32 saddr; 28486acc9b43SJoe Stringer __be32 daddr; 28496acc9b43SJoe Stringer __be16 sport; 28506acc9b43SJoe Stringer __be16 dport; 28516acc9b43SJoe Stringer } ipv4; 28526acc9b43SJoe Stringer struct { 28536acc9b43SJoe Stringer __be32 saddr[4]; 28546acc9b43SJoe Stringer __be32 daddr[4]; 28556acc9b43SJoe Stringer __be16 sport; 28566acc9b43SJoe Stringer __be16 dport; 28576acc9b43SJoe Stringer } ipv6; 28586acc9b43SJoe Stringer }; 28596acc9b43SJoe Stringer }; 28606acc9b43SJoe Stringer 286117bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 286217bedab2SMartin KaFai Lau 28636a773a15SBrenden Blanco /* User return codes for XDP prog type. 28646a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 28659beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 28669beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 28676a773a15SBrenden Blanco */ 28686a773a15SBrenden Blanco enum xdp_action { 28696a773a15SBrenden Blanco XDP_ABORTED = 0, 28706a773a15SBrenden Blanco XDP_DROP, 28716a773a15SBrenden Blanco XDP_PASS, 28726ce96ca3SBrenden Blanco XDP_TX, 2873814abfabSJohn Fastabend XDP_REDIRECT, 28746a773a15SBrenden Blanco }; 28756a773a15SBrenden Blanco 28766a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 28776a773a15SBrenden Blanco * new fields must be added to the end of this structure 28786a773a15SBrenden Blanco */ 28796a773a15SBrenden Blanco struct xdp_md { 28806a773a15SBrenden Blanco __u32 data; 28816a773a15SBrenden Blanco __u32 data_end; 2882de8f3a83SDaniel Borkmann __u32 data_meta; 2883daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 288402dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 288502dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 28866a773a15SBrenden Blanco }; 28876a773a15SBrenden Blanco 2888174a79ffSJohn Fastabend enum sk_action { 2889bfa64075SJohn Fastabend SK_DROP = 0, 2890bfa64075SJohn Fastabend SK_PASS, 2891174a79ffSJohn Fastabend }; 2892174a79ffSJohn Fastabend 28934f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 28944f738adbSJohn Fastabend * be added to the end of this structure 28954f738adbSJohn Fastabend */ 28964f738adbSJohn Fastabend struct sk_msg_md { 2897b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 2898b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 2899303def35SJohn Fastabend 2900303def35SJohn Fastabend __u32 family; 2901303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 2902303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 2903303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 2904303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 2905303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 2906303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 29073bdbd022SJohn Fastabend __u32 size; /* Total size of sk_msg */ 29084f738adbSJohn Fastabend }; 29094f738adbSJohn Fastabend 29102dbb9b9eSMartin KaFai Lau struct sk_reuseport_md { 29112dbb9b9eSMartin KaFai Lau /* 29122dbb9b9eSMartin KaFai Lau * Start of directly accessible data. It begins from 29132dbb9b9eSMartin KaFai Lau * the tcp/udp header. 29142dbb9b9eSMartin KaFai Lau */ 2915b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data); 2916b7df9adaSDaniel Borkmann /* End of directly accessible data */ 2917b7df9adaSDaniel Borkmann __bpf_md_ptr(void *, data_end); 29182dbb9b9eSMartin KaFai Lau /* 29192dbb9b9eSMartin KaFai Lau * Total length of packet (starting from the tcp/udp header). 29202dbb9b9eSMartin KaFai Lau * Note that the directly accessible bytes (data_end - data) 29212dbb9b9eSMartin KaFai Lau * could be less than this "len". Those bytes could be 29222dbb9b9eSMartin KaFai Lau * indirectly read by a helper "bpf_skb_load_bytes()". 29232dbb9b9eSMartin KaFai Lau */ 29242dbb9b9eSMartin KaFai Lau __u32 len; 29252dbb9b9eSMartin KaFai Lau /* 29262dbb9b9eSMartin KaFai Lau * Eth protocol in the mac header (network byte order). e.g. 29272dbb9b9eSMartin KaFai Lau * ETH_P_IP(0x0800) and ETH_P_IPV6(0x86DD) 29282dbb9b9eSMartin KaFai Lau */ 29292dbb9b9eSMartin KaFai Lau __u32 eth_protocol; 29302dbb9b9eSMartin KaFai Lau __u32 ip_protocol; /* IP protocol. e.g. IPPROTO_TCP, IPPROTO_UDP */ 29312dbb9b9eSMartin KaFai Lau __u32 bind_inany; /* Is sock bound to an INANY address? */ 29322dbb9b9eSMartin KaFai Lau __u32 hash; /* A hash of the packet 4 tuples */ 29332dbb9b9eSMartin KaFai Lau }; 29342dbb9b9eSMartin KaFai Lau 29351e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 29361e270976SMartin KaFai Lau 29371e270976SMartin KaFai Lau struct bpf_prog_info { 29381e270976SMartin KaFai Lau __u32 type; 29391e270976SMartin KaFai Lau __u32 id; 29401e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 29411e270976SMartin KaFai Lau __u32 jited_prog_len; 29421e270976SMartin KaFai Lau __u32 xlated_prog_len; 29431e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 29441e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 2945cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 2946cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 2947cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 2948cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 2949067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 2950675fc275SJakub Kicinski __u32 ifindex; 2951b85fab0eSJiri Olsa __u32 gpl_compatible:1; 2952675fc275SJakub Kicinski __u64 netns_dev; 2953675fc275SJakub Kicinski __u64 netns_ino; 2954dbecd738SSandipan Das __u32 nr_jited_ksyms; 2955815581c1SSandipan Das __u32 nr_jited_func_lens; 2956dbecd738SSandipan Das __aligned_u64 jited_ksyms; 2957815581c1SSandipan Das __aligned_u64 jited_func_lens; 2958838e9690SYonghong Song __u32 btf_id; 2959838e9690SYonghong Song __u32 func_info_rec_size; 2960838e9690SYonghong Song __aligned_u64 func_info; 296111d8b82dSYonghong Song __u32 nr_func_info; 296211d8b82dSYonghong Song __u32 nr_line_info; 2963c454a46bSMartin KaFai Lau __aligned_u64 line_info; 2964c454a46bSMartin KaFai Lau __aligned_u64 jited_line_info; 296511d8b82dSYonghong Song __u32 nr_jited_line_info; 2966c454a46bSMartin KaFai Lau __u32 line_info_rec_size; 2967c454a46bSMartin KaFai Lau __u32 jited_line_info_rec_size; 2968c872bdb3SSong Liu __u32 nr_prog_tags; 2969c872bdb3SSong Liu __aligned_u64 prog_tags; 29705f8f8b93SAlexei Starovoitov __u64 run_time_ns; 29715f8f8b93SAlexei Starovoitov __u64 run_cnt; 29721e270976SMartin KaFai Lau } __attribute__((aligned(8))); 29731e270976SMartin KaFai Lau 29741e270976SMartin KaFai Lau struct bpf_map_info { 29751e270976SMartin KaFai Lau __u32 type; 29761e270976SMartin KaFai Lau __u32 id; 29771e270976SMartin KaFai Lau __u32 key_size; 29781e270976SMartin KaFai Lau __u32 value_size; 29791e270976SMartin KaFai Lau __u32 max_entries; 29801e270976SMartin KaFai Lau __u32 map_flags; 2981067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 298252775b33SJakub Kicinski __u32 ifindex; 298336f9814aSDaniel Borkmann __u32 :32; 298452775b33SJakub Kicinski __u64 netns_dev; 298552775b33SJakub Kicinski __u64 netns_ino; 298678958fcaSMartin KaFai Lau __u32 btf_id; 29879b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 29889b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 29891e270976SMartin KaFai Lau } __attribute__((aligned(8))); 29901e270976SMartin KaFai Lau 299162dab84cSMartin KaFai Lau struct bpf_btf_info { 299262dab84cSMartin KaFai Lau __aligned_u64 btf; 299362dab84cSMartin KaFai Lau __u32 btf_size; 299462dab84cSMartin KaFai Lau __u32 id; 299562dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 299662dab84cSMartin KaFai Lau 29974fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 29984fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 29994fbac77dSAndrey Ignatov * attach attach type). 30004fbac77dSAndrey Ignatov */ 30014fbac77dSAndrey Ignatov struct bpf_sock_addr { 30024fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 30034fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 30044fbac77dSAndrey Ignatov * Stored in network byte order. 30054fbac77dSAndrey Ignatov */ 30064fbac77dSAndrey Ignatov __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 30074fbac77dSAndrey Ignatov * Stored in network byte order. 30084fbac77dSAndrey Ignatov */ 30094fbac77dSAndrey Ignatov __u32 user_port; /* Allows 4-byte read and write. 30104fbac77dSAndrey Ignatov * Stored in network byte order 30114fbac77dSAndrey Ignatov */ 30124fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 30134fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 30144fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 30151cedee13SAndrey Ignatov __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write. 30161cedee13SAndrey Ignatov * Stored in network byte order. 30171cedee13SAndrey Ignatov */ 30181cedee13SAndrey Ignatov __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 30191cedee13SAndrey Ignatov * Stored in network byte order. 30201cedee13SAndrey Ignatov */ 30214fbac77dSAndrey Ignatov }; 30224fbac77dSAndrey Ignatov 302340304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 302440304b2aSLawrence Brakmo * and their replies. 302540304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 302640304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 302740304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 302840304b2aSLawrence Brakmo */ 302940304b2aSLawrence Brakmo struct bpf_sock_ops { 303040304b2aSLawrence Brakmo __u32 op; 303140304b2aSLawrence Brakmo union { 3032de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 3033de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 3034de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 303540304b2aSLawrence Brakmo }; 303640304b2aSLawrence Brakmo __u32 family; 303740304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 303840304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 303940304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 304040304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 304140304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 304240304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 3043f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 3044f19397a5SLawrence Brakmo * there is a full socket. If not, the 3045f19397a5SLawrence Brakmo * fields read as zero. 3046f19397a5SLawrence Brakmo */ 3047f19397a5SLawrence Brakmo __u32 snd_cwnd; 3048f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 3049b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 305044f0e430SLawrence Brakmo __u32 state; 305144f0e430SLawrence Brakmo __u32 rtt_min; 305244f0e430SLawrence Brakmo __u32 snd_ssthresh; 305344f0e430SLawrence Brakmo __u32 rcv_nxt; 305444f0e430SLawrence Brakmo __u32 snd_nxt; 305544f0e430SLawrence Brakmo __u32 snd_una; 305644f0e430SLawrence Brakmo __u32 mss_cache; 305744f0e430SLawrence Brakmo __u32 ecn_flags; 305844f0e430SLawrence Brakmo __u32 rate_delivered; 305944f0e430SLawrence Brakmo __u32 rate_interval_us; 306044f0e430SLawrence Brakmo __u32 packets_out; 306144f0e430SLawrence Brakmo __u32 retrans_out; 306244f0e430SLawrence Brakmo __u32 total_retrans; 306344f0e430SLawrence Brakmo __u32 segs_in; 306444f0e430SLawrence Brakmo __u32 data_segs_in; 306544f0e430SLawrence Brakmo __u32 segs_out; 306644f0e430SLawrence Brakmo __u32 data_segs_out; 306744f0e430SLawrence Brakmo __u32 lost_out; 306844f0e430SLawrence Brakmo __u32 sacked_out; 306944f0e430SLawrence Brakmo __u32 sk_txhash; 307044f0e430SLawrence Brakmo __u64 bytes_received; 307144f0e430SLawrence Brakmo __u64 bytes_acked; 307240304b2aSLawrence Brakmo }; 307340304b2aSLawrence Brakmo 3074b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 3075f89013f6SLawrence Brakmo #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 3076a31ad29eSLawrence Brakmo #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 3077d4487491SLawrence Brakmo #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 3078d4487491SLawrence Brakmo #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently 3079b13d8807SLawrence Brakmo * supported cb flags 3080b13d8807SLawrence Brakmo */ 3081b13d8807SLawrence Brakmo 308240304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 308340304b2aSLawrence Brakmo * New entries can only be added at the end 308440304b2aSLawrence Brakmo */ 308540304b2aSLawrence Brakmo enum { 308640304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 30878550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 30888550f328SLawrence Brakmo * -1 if default value should be used 30898550f328SLawrence Brakmo */ 309013d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 309113d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 309213d3b1ebSLawrence Brakmo * value should be used 309313d3b1ebSLawrence Brakmo */ 30949872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 30959872a4bdSLawrence Brakmo * active connection is initialized 30969872a4bdSLawrence Brakmo */ 30979872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 30989872a4bdSLawrence Brakmo * active connection is 30999872a4bdSLawrence Brakmo * established 31009872a4bdSLawrence Brakmo */ 31019872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 31029872a4bdSLawrence Brakmo * passive connection is 31039872a4bdSLawrence Brakmo * established 31049872a4bdSLawrence Brakmo */ 310591b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 310691b5b21cSLawrence Brakmo * needs ECN 310791b5b21cSLawrence Brakmo */ 3108e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 3109e6546ef6SLawrence Brakmo * based on the path and may be 3110e6546ef6SLawrence Brakmo * dependent on the congestion control 3111e6546ef6SLawrence Brakmo * algorithm. In general it indicates 3112e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 3113e6546ef6SLawrence Brakmo * this indicate congestion 3114e6546ef6SLawrence Brakmo */ 3115f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 3116f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 3117f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 3118f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 3119f89013f6SLawrence Brakmo */ 3120a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 3121a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 3122a31ad29eSLawrence Brakmo * Arg2: # segments 3123a31ad29eSLawrence Brakmo * Arg3: return value of 3124a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 3125a31ad29eSLawrence Brakmo */ 3126d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 3127d4487491SLawrence Brakmo * Arg1: old_state 3128d4487491SLawrence Brakmo * Arg2: new_state 3129d4487491SLawrence Brakmo */ 3130f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 3131f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 3132f333ee0cSAndrey Ignatov */ 3133d4487491SLawrence Brakmo }; 3134d4487491SLawrence Brakmo 3135d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 3136d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 3137d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 3138d4487491SLawrence Brakmo * the BPF sock_ops function. 3139d4487491SLawrence Brakmo */ 3140d4487491SLawrence Brakmo enum { 3141d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 3142d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 3143d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 3144d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 3145d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 3146d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 3147d4487491SLawrence Brakmo BPF_TCP_CLOSE, 3148d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 3149d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 3150d4487491SLawrence Brakmo BPF_TCP_LISTEN, 3151d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 3152d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 3153d4487491SLawrence Brakmo 3154d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 315540304b2aSLawrence Brakmo }; 315640304b2aSLawrence Brakmo 3157fc747810SLawrence Brakmo #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 315813bf9641SLawrence Brakmo #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 3159fc747810SLawrence Brakmo 3160908432caSYonghong Song struct bpf_perf_event_value { 3161908432caSYonghong Song __u64 counter; 3162908432caSYonghong Song __u64 enabled; 3163908432caSYonghong Song __u64 running; 3164908432caSYonghong Song }; 3165908432caSYonghong Song 3166ebc614f6SRoman Gushchin #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 3167ebc614f6SRoman Gushchin #define BPF_DEVCG_ACC_READ (1ULL << 1) 3168ebc614f6SRoman Gushchin #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 3169ebc614f6SRoman Gushchin 3170ebc614f6SRoman Gushchin #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 3171ebc614f6SRoman Gushchin #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 3172ebc614f6SRoman Gushchin 3173ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 317406ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 317506ef0ccbSYonghong Song __u32 access_type; 3176ebc614f6SRoman Gushchin __u32 major; 3177ebc614f6SRoman Gushchin __u32 minor; 3178ebc614f6SRoman Gushchin }; 3179ebc614f6SRoman Gushchin 3180c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 3181c4f6699dSAlexei Starovoitov __u64 args[0]; 3182c4f6699dSAlexei Starovoitov }; 3183c4f6699dSAlexei Starovoitov 318487f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 318587f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 318687f5fc7eSDavid Ahern */ 318787f5fc7eSDavid Ahern #define BPF_FIB_LOOKUP_DIRECT BIT(0) 318887f5fc7eSDavid Ahern #define BPF_FIB_LOOKUP_OUTPUT BIT(1) 318987f5fc7eSDavid Ahern 31904c79579bSDavid Ahern enum { 31914c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 31924c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 31934c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 31944c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 31954c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 31964c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 31974c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 31984c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 31994c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 32004c79579bSDavid Ahern }; 32014c79579bSDavid Ahern 320287f5fc7eSDavid Ahern struct bpf_fib_lookup { 3203fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 3204fa898d76SDavid Ahern * output: network family of egress nexthop 3205fa898d76SDavid Ahern */ 3206fa898d76SDavid Ahern __u8 family; 320787f5fc7eSDavid Ahern 320887f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 320987f5fc7eSDavid Ahern __u8 l4_protocol; 321087f5fc7eSDavid Ahern __be16 sport; 321187f5fc7eSDavid Ahern __be16 dport; 321287f5fc7eSDavid Ahern 321387f5fc7eSDavid Ahern /* total length of packet from network header - used for MTU check */ 321487f5fc7eSDavid Ahern __u16 tot_len; 32154c79579bSDavid Ahern 32164c79579bSDavid Ahern /* input: L3 device index for lookup 32174c79579bSDavid Ahern * output: device index from FIB lookup 32184c79579bSDavid Ahern */ 32194c79579bSDavid Ahern __u32 ifindex; 322087f5fc7eSDavid Ahern 322187f5fc7eSDavid Ahern union { 322287f5fc7eSDavid Ahern /* inputs to lookup */ 322387f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 3224bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 322587f5fc7eSDavid Ahern 3226fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 322787f5fc7eSDavid Ahern __u32 rt_metric; 322887f5fc7eSDavid Ahern }; 322987f5fc7eSDavid Ahern 323087f5fc7eSDavid Ahern union { 323187f5fc7eSDavid Ahern __be32 ipv4_src; 323287f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 323387f5fc7eSDavid Ahern }; 323487f5fc7eSDavid Ahern 3235fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 3236fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 3237fa898d76SDavid Ahern * if FIB lookup returns gateway route 323887f5fc7eSDavid Ahern */ 323987f5fc7eSDavid Ahern union { 324087f5fc7eSDavid Ahern __be32 ipv4_dst; 324187f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 324287f5fc7eSDavid Ahern }; 324387f5fc7eSDavid Ahern 324487f5fc7eSDavid Ahern /* output */ 324587f5fc7eSDavid Ahern __be16 h_vlan_proto; 324687f5fc7eSDavid Ahern __be16 h_vlan_TCI; 324787f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 324887f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 324987f5fc7eSDavid Ahern }; 325087f5fc7eSDavid Ahern 325141bdc4b4SYonghong Song enum bpf_task_fd_type { 325241bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 325341bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 325441bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 325541bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 325641bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 325741bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 325841bdc4b4SYonghong Song }; 325941bdc4b4SYonghong Song 3260d58e468bSPetar Penkov struct bpf_flow_keys { 3261d58e468bSPetar Penkov __u16 nhoff; 3262d58e468bSPetar Penkov __u16 thoff; 3263d58e468bSPetar Penkov __u16 addr_proto; /* ETH_P_* of valid addrs */ 3264d58e468bSPetar Penkov __u8 is_frag; 3265d58e468bSPetar Penkov __u8 is_first_frag; 3266d58e468bSPetar Penkov __u8 is_encap; 3267d58e468bSPetar Penkov __u8 ip_proto; 3268d58e468bSPetar Penkov __be16 n_proto; 3269d58e468bSPetar Penkov __be16 sport; 3270d58e468bSPetar Penkov __be16 dport; 3271d58e468bSPetar Penkov union { 3272d58e468bSPetar Penkov struct { 3273d58e468bSPetar Penkov __be32 ipv4_src; 3274d58e468bSPetar Penkov __be32 ipv4_dst; 3275d58e468bSPetar Penkov }; 3276d58e468bSPetar Penkov struct { 3277d58e468bSPetar Penkov __u32 ipv6_src[4]; /* in6_addr; network order */ 3278d58e468bSPetar Penkov __u32 ipv6_dst[4]; /* in6_addr; network order */ 3279d58e468bSPetar Penkov }; 3280d58e468bSPetar Penkov }; 3281d58e468bSPetar Penkov }; 3282d58e468bSPetar Penkov 3283838e9690SYonghong Song struct bpf_func_info { 3284d30d42e0SMartin KaFai Lau __u32 insn_off; 3285838e9690SYonghong Song __u32 type_id; 3286838e9690SYonghong Song }; 3287838e9690SYonghong Song 3288c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_NUM(line_col) ((line_col) >> 10) 3289c454a46bSMartin KaFai Lau #define BPF_LINE_INFO_LINE_COL(line_col) ((line_col) & 0x3ff) 3290c454a46bSMartin KaFai Lau 3291c454a46bSMartin KaFai Lau struct bpf_line_info { 3292c454a46bSMartin KaFai Lau __u32 insn_off; 3293c454a46bSMartin KaFai Lau __u32 file_name_off; 3294c454a46bSMartin KaFai Lau __u32 line_off; 3295c454a46bSMartin KaFai Lau __u32 line_col; 3296c454a46bSMartin KaFai Lau }; 3297c454a46bSMartin KaFai Lau 3298d83525caSAlexei Starovoitov struct bpf_spin_lock { 3299d83525caSAlexei Starovoitov __u32 val; 3300d83525caSAlexei Starovoitov }; 3301daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 3302