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 */ 17daedfb22SAlexei Starovoitov #define BPF_ALU64 0x07 /* alu mode in double word width */ 18daedfb22SAlexei Starovoitov 19daedfb22SAlexei Starovoitov /* ld/ldx fields */ 20cb5f7334SJesper Dangaard Brouer #define BPF_DW 0x18 /* double word (64-bit) */ 21daedfb22SAlexei Starovoitov #define BPF_XADD 0xc0 /* exclusive add */ 22daedfb22SAlexei Starovoitov 23daedfb22SAlexei Starovoitov /* alu/jmp fields */ 24daedfb22SAlexei Starovoitov #define BPF_MOV 0xb0 /* mov reg to reg */ 25daedfb22SAlexei Starovoitov #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ 26daedfb22SAlexei Starovoitov 27daedfb22SAlexei Starovoitov /* change endianness of a register */ 28daedfb22SAlexei Starovoitov #define BPF_END 0xd0 /* flags for endianness conversion: */ 29daedfb22SAlexei Starovoitov #define BPF_TO_LE 0x00 /* convert to little-endian */ 30daedfb22SAlexei Starovoitov #define BPF_TO_BE 0x08 /* convert to big-endian */ 31daedfb22SAlexei Starovoitov #define BPF_FROM_LE BPF_TO_LE 32daedfb22SAlexei Starovoitov #define BPF_FROM_BE BPF_TO_BE 33daedfb22SAlexei Starovoitov 3492b31a9aSDaniel Borkmann /* jmp encodings */ 35daedfb22SAlexei Starovoitov #define BPF_JNE 0x50 /* jump != */ 3692b31a9aSDaniel Borkmann #define BPF_JLT 0xa0 /* LT is unsigned, '<' */ 3792b31a9aSDaniel Borkmann #define BPF_JLE 0xb0 /* LE is unsigned, '<=' */ 38daedfb22SAlexei Starovoitov #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ 39daedfb22SAlexei Starovoitov #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ 4092b31a9aSDaniel Borkmann #define BPF_JSLT 0xc0 /* SLT is signed, '<' */ 4192b31a9aSDaniel Borkmann #define BPF_JSLE 0xd0 /* SLE is signed, '<=' */ 42daedfb22SAlexei Starovoitov #define BPF_CALL 0x80 /* function call */ 43daedfb22SAlexei Starovoitov #define BPF_EXIT 0x90 /* function return */ 44daedfb22SAlexei Starovoitov 45daedfb22SAlexei Starovoitov /* Register numbers */ 46daedfb22SAlexei Starovoitov enum { 47daedfb22SAlexei Starovoitov BPF_REG_0 = 0, 48daedfb22SAlexei Starovoitov BPF_REG_1, 49daedfb22SAlexei Starovoitov BPF_REG_2, 50daedfb22SAlexei Starovoitov BPF_REG_3, 51daedfb22SAlexei Starovoitov BPF_REG_4, 52daedfb22SAlexei Starovoitov BPF_REG_5, 53daedfb22SAlexei Starovoitov BPF_REG_6, 54daedfb22SAlexei Starovoitov BPF_REG_7, 55daedfb22SAlexei Starovoitov BPF_REG_8, 56daedfb22SAlexei Starovoitov BPF_REG_9, 57daedfb22SAlexei Starovoitov BPF_REG_10, 58daedfb22SAlexei Starovoitov __MAX_BPF_REG, 59daedfb22SAlexei Starovoitov }; 60daedfb22SAlexei Starovoitov 61daedfb22SAlexei Starovoitov /* BPF has 10 general purpose 64-bit registers and stack frame. */ 62daedfb22SAlexei Starovoitov #define MAX_BPF_REG __MAX_BPF_REG 63daedfb22SAlexei Starovoitov 64daedfb22SAlexei Starovoitov struct bpf_insn { 65daedfb22SAlexei Starovoitov __u8 code; /* opcode */ 66daedfb22SAlexei Starovoitov __u8 dst_reg:4; /* dest register */ 67daedfb22SAlexei Starovoitov __u8 src_reg:4; /* source register */ 68daedfb22SAlexei Starovoitov __s16 off; /* signed offset */ 69daedfb22SAlexei Starovoitov __s32 imm; /* signed immediate constant */ 70daedfb22SAlexei Starovoitov }; 71daedfb22SAlexei Starovoitov 72b95a5c4dSDaniel Mack /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ 73b95a5c4dSDaniel Mack struct bpf_lpm_trie_key { 74b95a5c4dSDaniel Mack __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ 75b95a5c4dSDaniel Mack __u8 data[0]; /* Arbitrary size */ 76b95a5c4dSDaniel Mack }; 77b95a5c4dSDaniel Mack 78b2197755SDaniel Borkmann /* BPF syscall commands, see bpf(2) man-page for details. */ 7999c55f7dSAlexei Starovoitov enum bpf_cmd { 8099c55f7dSAlexei Starovoitov BPF_MAP_CREATE, 81db20fd2bSAlexei Starovoitov BPF_MAP_LOOKUP_ELEM, 82db20fd2bSAlexei Starovoitov BPF_MAP_UPDATE_ELEM, 83db20fd2bSAlexei Starovoitov BPF_MAP_DELETE_ELEM, 84db20fd2bSAlexei Starovoitov BPF_MAP_GET_NEXT_KEY, 8509756af4SAlexei Starovoitov BPF_PROG_LOAD, 86b2197755SDaniel Borkmann BPF_OBJ_PIN, 87b2197755SDaniel Borkmann BPF_OBJ_GET, 88f4324551SDaniel Mack BPF_PROG_ATTACH, 89f4324551SDaniel Mack BPF_PROG_DETACH, 901cf1cae9SAlexei Starovoitov BPF_PROG_TEST_RUN, 9134ad5580SMartin KaFai Lau BPF_PROG_GET_NEXT_ID, 9234ad5580SMartin KaFai Lau BPF_MAP_GET_NEXT_ID, 93b16d9aa4SMartin KaFai Lau BPF_PROG_GET_FD_BY_ID, 94bd5f5f4eSMartin KaFai Lau BPF_MAP_GET_FD_BY_ID, 951e270976SMartin KaFai Lau BPF_OBJ_GET_INFO_BY_FD, 96468e2f64SAlexei Starovoitov BPF_PROG_QUERY, 97c4f6699dSAlexei Starovoitov BPF_RAW_TRACEPOINT_OPEN, 98f56a653cSMartin KaFai Lau BPF_BTF_LOAD, 9978958fcaSMartin KaFai Lau BPF_BTF_GET_FD_BY_ID, 10041bdc4b4SYonghong Song BPF_TASK_FD_QUERY, 10199c55f7dSAlexei Starovoitov }; 10299c55f7dSAlexei Starovoitov 10399c55f7dSAlexei Starovoitov enum bpf_map_type { 10499c55f7dSAlexei Starovoitov BPF_MAP_TYPE_UNSPEC, 1050f8e4bd8SAlexei Starovoitov BPF_MAP_TYPE_HASH, 10628fbcfa0SAlexei Starovoitov BPF_MAP_TYPE_ARRAY, 10704fd61abSAlexei Starovoitov BPF_MAP_TYPE_PROG_ARRAY, 108ea317b26SKaixu Xia BPF_MAP_TYPE_PERF_EVENT_ARRAY, 109824bd0ceSAlexei Starovoitov BPF_MAP_TYPE_PERCPU_HASH, 110a10423b8SAlexei Starovoitov BPF_MAP_TYPE_PERCPU_ARRAY, 111d5a3b1f6SAlexei Starovoitov BPF_MAP_TYPE_STACK_TRACE, 1124ed8ec52SMartin KaFai Lau BPF_MAP_TYPE_CGROUP_ARRAY, 11329ba732aSMartin KaFai Lau BPF_MAP_TYPE_LRU_HASH, 1148f844938SMartin KaFai Lau BPF_MAP_TYPE_LRU_PERCPU_HASH, 115b95a5c4dSDaniel Mack BPF_MAP_TYPE_LPM_TRIE, 11656f668dfSMartin KaFai Lau BPF_MAP_TYPE_ARRAY_OF_MAPS, 117bcc6b1b7SMartin KaFai Lau BPF_MAP_TYPE_HASH_OF_MAPS, 118546ac1ffSJohn Fastabend BPF_MAP_TYPE_DEVMAP, 119174a79ffSJohn Fastabend BPF_MAP_TYPE_SOCKMAP, 1206710e112SJesper Dangaard Brouer BPF_MAP_TYPE_CPUMAP, 121fbfc504aSBjörn Töpel BPF_MAP_TYPE_XSKMAP, 12281110384SJohn Fastabend BPF_MAP_TYPE_SOCKHASH, 12399c55f7dSAlexei Starovoitov }; 12499c55f7dSAlexei Starovoitov 12509756af4SAlexei Starovoitov enum bpf_prog_type { 12609756af4SAlexei Starovoitov BPF_PROG_TYPE_UNSPEC, 127ddd872bcSAlexei Starovoitov BPF_PROG_TYPE_SOCKET_FILTER, 1282541517cSAlexei Starovoitov BPF_PROG_TYPE_KPROBE, 12996be4325SDaniel Borkmann BPF_PROG_TYPE_SCHED_CLS, 13094caee8cSDaniel Borkmann BPF_PROG_TYPE_SCHED_ACT, 13198b5c2c6SAlexei Starovoitov BPF_PROG_TYPE_TRACEPOINT, 1326a773a15SBrenden Blanco BPF_PROG_TYPE_XDP, 1330515e599SAlexei Starovoitov BPF_PROG_TYPE_PERF_EVENT, 1340e33661dSDaniel Mack BPF_PROG_TYPE_CGROUP_SKB, 13561023658SDavid Ahern BPF_PROG_TYPE_CGROUP_SOCK, 1363a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_IN, 1373a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_OUT, 1383a0af8fdSThomas Graf BPF_PROG_TYPE_LWT_XMIT, 13940304b2aSLawrence Brakmo BPF_PROG_TYPE_SOCK_OPS, 140b005fd18SJohn Fastabend BPF_PROG_TYPE_SK_SKB, 141ebc614f6SRoman Gushchin BPF_PROG_TYPE_CGROUP_DEVICE, 1424f738adbSJohn Fastabend BPF_PROG_TYPE_SK_MSG, 143c4f6699dSAlexei Starovoitov BPF_PROG_TYPE_RAW_TRACEPOINT, 1444fbac77dSAndrey Ignatov BPF_PROG_TYPE_CGROUP_SOCK_ADDR, 145004d4b27SMathieu Xhonneux BPF_PROG_TYPE_LWT_SEG6LOCAL, 146f4364dcfSSean Young BPF_PROG_TYPE_LIRC_MODE2, 14709756af4SAlexei Starovoitov }; 14809756af4SAlexei Starovoitov 1490e33661dSDaniel Mack enum bpf_attach_type { 1500e33661dSDaniel Mack BPF_CGROUP_INET_INGRESS, 1510e33661dSDaniel Mack BPF_CGROUP_INET_EGRESS, 15261023658SDavid Ahern BPF_CGROUP_INET_SOCK_CREATE, 15340304b2aSLawrence Brakmo BPF_CGROUP_SOCK_OPS, 154464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_PARSER, 155464bc0fdSJohn Fastabend BPF_SK_SKB_STREAM_VERDICT, 156ebc614f6SRoman Gushchin BPF_CGROUP_DEVICE, 1574f738adbSJohn Fastabend BPF_SK_MSG_VERDICT, 1584fbac77dSAndrey Ignatov BPF_CGROUP_INET4_BIND, 1594fbac77dSAndrey Ignatov BPF_CGROUP_INET6_BIND, 160d74bad4eSAndrey Ignatov BPF_CGROUP_INET4_CONNECT, 161d74bad4eSAndrey Ignatov BPF_CGROUP_INET6_CONNECT, 162aac3fc32SAndrey Ignatov BPF_CGROUP_INET4_POST_BIND, 163aac3fc32SAndrey Ignatov BPF_CGROUP_INET6_POST_BIND, 1641cedee13SAndrey Ignatov BPF_CGROUP_UDP4_SENDMSG, 1651cedee13SAndrey Ignatov BPF_CGROUP_UDP6_SENDMSG, 166f4364dcfSSean Young BPF_LIRC_MODE2, 1670e33661dSDaniel Mack __MAX_BPF_ATTACH_TYPE 1680e33661dSDaniel Mack }; 1690e33661dSDaniel Mack 1700e33661dSDaniel Mack #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE 1710e33661dSDaniel Mack 172324bda9eSAlexei Starovoitov /* cgroup-bpf attach flags used in BPF_PROG_ATTACH command 173324bda9eSAlexei Starovoitov * 174324bda9eSAlexei Starovoitov * NONE(default): No further bpf programs allowed in the subtree. 175324bda9eSAlexei Starovoitov * 176324bda9eSAlexei Starovoitov * BPF_F_ALLOW_OVERRIDE: If a sub-cgroup installs some bpf program, 177324bda9eSAlexei Starovoitov * the program in this cgroup yields to sub-cgroup program. 178324bda9eSAlexei Starovoitov * 179324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI: If a sub-cgroup installs some bpf program, 180324bda9eSAlexei Starovoitov * that cgroup program gets run in addition to the program in this cgroup. 181324bda9eSAlexei Starovoitov * 182324bda9eSAlexei Starovoitov * Only one program is allowed to be attached to a cgroup with 183324bda9eSAlexei Starovoitov * NONE or BPF_F_ALLOW_OVERRIDE flag. 184324bda9eSAlexei Starovoitov * Attaching another program on top of NONE or BPF_F_ALLOW_OVERRIDE will 185324bda9eSAlexei Starovoitov * release old program and attach the new one. Attach flags has to match. 186324bda9eSAlexei Starovoitov * 187324bda9eSAlexei Starovoitov * Multiple programs are allowed to be attached to a cgroup with 188324bda9eSAlexei Starovoitov * BPF_F_ALLOW_MULTI flag. They are executed in FIFO order 189324bda9eSAlexei Starovoitov * (those that were attached first, run first) 190324bda9eSAlexei Starovoitov * The programs of sub-cgroup are executed first, then programs of 191324bda9eSAlexei Starovoitov * this cgroup and then programs of parent cgroup. 192324bda9eSAlexei Starovoitov * When children program makes decision (like picking TCP CA or sock bind) 193324bda9eSAlexei Starovoitov * parent program has a chance to override it. 194324bda9eSAlexei Starovoitov * 195324bda9eSAlexei Starovoitov * A cgroup with MULTI or OVERRIDE flag allows any attach flags in sub-cgroups. 196324bda9eSAlexei Starovoitov * A cgroup with NONE doesn't allow any programs in sub-cgroups. 197324bda9eSAlexei Starovoitov * Ex1: 198324bda9eSAlexei Starovoitov * cgrp1 (MULTI progs A, B) -> 199324bda9eSAlexei Starovoitov * cgrp2 (OVERRIDE prog C) -> 200324bda9eSAlexei Starovoitov * cgrp3 (MULTI prog D) -> 201324bda9eSAlexei Starovoitov * cgrp4 (OVERRIDE prog E) -> 202324bda9eSAlexei Starovoitov * cgrp5 (NONE prog F) 203324bda9eSAlexei Starovoitov * the event in cgrp5 triggers execution of F,D,A,B in that order. 204324bda9eSAlexei Starovoitov * if prog F is detached, the execution is E,D,A,B 205324bda9eSAlexei Starovoitov * if prog F and D are detached, the execution is E,A,B 206324bda9eSAlexei Starovoitov * if prog F, E and D are detached, the execution is C,A,B 207324bda9eSAlexei Starovoitov * 208324bda9eSAlexei Starovoitov * All eligible programs are executed regardless of return code from 209324bda9eSAlexei Starovoitov * earlier programs. 2107f677633SAlexei Starovoitov */ 2117f677633SAlexei Starovoitov #define BPF_F_ALLOW_OVERRIDE (1U << 0) 212324bda9eSAlexei Starovoitov #define BPF_F_ALLOW_MULTI (1U << 1) 2137f677633SAlexei Starovoitov 214e07b98d9SDavid S. Miller /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the 215e07b98d9SDavid S. Miller * verifier will perform strict alignment checking as if the kernel 216e07b98d9SDavid S. Miller * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, 217e07b98d9SDavid S. Miller * and NET_IP_ALIGN defined to 2. 218e07b98d9SDavid S. Miller */ 219e07b98d9SDavid S. Miller #define BPF_F_STRICT_ALIGNMENT (1U << 0) 220e07b98d9SDavid S. Miller 221cc8b0b92SAlexei Starovoitov /* when bpf_ldimm64->src_reg == BPF_PSEUDO_MAP_FD, bpf_ldimm64->imm == fd */ 222f1a66f85SDaniel Borkmann #define BPF_PSEUDO_MAP_FD 1 223f1a66f85SDaniel Borkmann 224cc8b0b92SAlexei Starovoitov /* when bpf_call->src_reg == BPF_PSEUDO_CALL, bpf_call->imm == pc-relative 225cc8b0b92SAlexei Starovoitov * offset to another bpf function 226cc8b0b92SAlexei Starovoitov */ 227cc8b0b92SAlexei Starovoitov #define BPF_PSEUDO_CALL 1 228cc8b0b92SAlexei Starovoitov 2293274f520SAlexei Starovoitov /* flags for BPF_MAP_UPDATE_ELEM command */ 2303274f520SAlexei Starovoitov #define BPF_ANY 0 /* create new element or update existing */ 2313274f520SAlexei Starovoitov #define BPF_NOEXIST 1 /* create new element if it didn't exist */ 2323274f520SAlexei Starovoitov #define BPF_EXIST 2 /* update existing element */ 2333274f520SAlexei Starovoitov 23496eabe7aSMartin KaFai Lau /* flags for BPF_MAP_CREATE command */ 2356c905981SAlexei Starovoitov #define BPF_F_NO_PREALLOC (1U << 0) 23629ba732aSMartin KaFai Lau /* Instead of having one common LRU list in the 2378f844938SMartin KaFai Lau * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list 23829ba732aSMartin KaFai Lau * which can scale and perform better. 23929ba732aSMartin KaFai Lau * Note, the LRU nodes (including free nodes) cannot be moved 24029ba732aSMartin KaFai Lau * across different LRU lists. 24129ba732aSMartin KaFai Lau */ 24229ba732aSMartin KaFai Lau #define BPF_F_NO_COMMON_LRU (1U << 1) 24396eabe7aSMartin KaFai Lau /* Specify numa node during map creation */ 24496eabe7aSMartin KaFai Lau #define BPF_F_NUMA_NODE (1U << 2) 2456c905981SAlexei Starovoitov 246468e2f64SAlexei Starovoitov /* flags for BPF_PROG_QUERY */ 247468e2f64SAlexei Starovoitov #define BPF_F_QUERY_EFFECTIVE (1U << 0) 248468e2f64SAlexei Starovoitov 249cb4d2b3fSMartin KaFai Lau #define BPF_OBJ_NAME_LEN 16U 250cb4d2b3fSMartin KaFai Lau 2516e71b04aSChenbo Feng /* Flags for accessing BPF object */ 2526e71b04aSChenbo Feng #define BPF_F_RDONLY (1U << 3) 2536e71b04aSChenbo Feng #define BPF_F_WRONLY (1U << 4) 2546e71b04aSChenbo Feng 255615755a7SSong Liu /* Flag for stack_map, store build_id+offset instead of pointer */ 256615755a7SSong Liu #define BPF_F_STACK_BUILD_ID (1U << 5) 257615755a7SSong Liu 258615755a7SSong Liu enum bpf_stack_build_id_status { 259615755a7SSong Liu /* user space need an empty entry to identify end of a trace */ 260615755a7SSong Liu BPF_STACK_BUILD_ID_EMPTY = 0, 261615755a7SSong Liu /* with valid build_id and offset */ 262615755a7SSong Liu BPF_STACK_BUILD_ID_VALID = 1, 263615755a7SSong Liu /* couldn't get build_id, fallback to ip */ 264615755a7SSong Liu BPF_STACK_BUILD_ID_IP = 2, 265615755a7SSong Liu }; 266615755a7SSong Liu 267615755a7SSong Liu #define BPF_BUILD_ID_SIZE 20 268615755a7SSong Liu struct bpf_stack_build_id { 269615755a7SSong Liu __s32 status; 270615755a7SSong Liu unsigned char build_id[BPF_BUILD_ID_SIZE]; 271615755a7SSong Liu union { 272615755a7SSong Liu __u64 offset; 273615755a7SSong Liu __u64 ip; 274615755a7SSong Liu }; 275615755a7SSong Liu }; 276615755a7SSong Liu 27799c55f7dSAlexei Starovoitov union bpf_attr { 27899c55f7dSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_CREATE command */ 27999c55f7dSAlexei Starovoitov __u32 map_type; /* one of enum bpf_map_type */ 28099c55f7dSAlexei Starovoitov __u32 key_size; /* size of key in bytes */ 28199c55f7dSAlexei Starovoitov __u32 value_size; /* size of value in bytes */ 28299c55f7dSAlexei Starovoitov __u32 max_entries; /* max number of entries in a map */ 28396eabe7aSMartin KaFai Lau __u32 map_flags; /* BPF_MAP_CREATE related 28496eabe7aSMartin KaFai Lau * flags defined above. 28596eabe7aSMartin KaFai Lau */ 28656f668dfSMartin KaFai Lau __u32 inner_map_fd; /* fd pointing to the inner map */ 28796eabe7aSMartin KaFai Lau __u32 numa_node; /* numa node (effective only if 28896eabe7aSMartin KaFai Lau * BPF_F_NUMA_NODE is set). 28996eabe7aSMartin KaFai Lau */ 290067cae47SMartin KaFai Lau char map_name[BPF_OBJ_NAME_LEN]; 291a3884572SJakub Kicinski __u32 map_ifindex; /* ifindex of netdev to create on */ 292a26ca7c9SMartin KaFai Lau __u32 btf_fd; /* fd pointing to a BTF type data */ 2939b2cf328SMartin KaFai Lau __u32 btf_key_type_id; /* BTF type_id of the key */ 2949b2cf328SMartin KaFai Lau __u32 btf_value_type_id; /* BTF type_id of the value */ 29599c55f7dSAlexei Starovoitov }; 296db20fd2bSAlexei Starovoitov 297db20fd2bSAlexei Starovoitov struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 298db20fd2bSAlexei Starovoitov __u32 map_fd; 299db20fd2bSAlexei Starovoitov __aligned_u64 key; 300db20fd2bSAlexei Starovoitov union { 301db20fd2bSAlexei Starovoitov __aligned_u64 value; 302db20fd2bSAlexei Starovoitov __aligned_u64 next_key; 303db20fd2bSAlexei Starovoitov }; 3043274f520SAlexei Starovoitov __u64 flags; 305db20fd2bSAlexei Starovoitov }; 30609756af4SAlexei Starovoitov 30709756af4SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_LOAD command */ 30809756af4SAlexei Starovoitov __u32 prog_type; /* one of enum bpf_prog_type */ 30909756af4SAlexei Starovoitov __u32 insn_cnt; 31009756af4SAlexei Starovoitov __aligned_u64 insns; 31109756af4SAlexei Starovoitov __aligned_u64 license; 312cbd35700SAlexei Starovoitov __u32 log_level; /* verbosity level of verifier */ 313cbd35700SAlexei Starovoitov __u32 log_size; /* size of user buffer */ 314cbd35700SAlexei Starovoitov __aligned_u64 log_buf; /* user supplied buffer */ 3152541517cSAlexei Starovoitov __u32 kern_version; /* checked when prog_type=kprobe */ 316e07b98d9SDavid S. Miller __u32 prog_flags; 317067cae47SMartin KaFai Lau char prog_name[BPF_OBJ_NAME_LEN]; 3181f6f4cb7SJakub Kicinski __u32 prog_ifindex; /* ifindex of netdev to prep for */ 3195e43f899SAndrey Ignatov /* For some prog types expected attach type must be known at 3205e43f899SAndrey Ignatov * load time to verify attach type specific parts of prog 3215e43f899SAndrey Ignatov * (context accesses, allowed helpers, etc). 3225e43f899SAndrey Ignatov */ 3235e43f899SAndrey Ignatov __u32 expected_attach_type; 32409756af4SAlexei Starovoitov }; 325b2197755SDaniel Borkmann 326b2197755SDaniel Borkmann struct { /* anonymous struct used by BPF_OBJ_* commands */ 327b2197755SDaniel Borkmann __aligned_u64 pathname; 328b2197755SDaniel Borkmann __u32 bpf_fd; 3296e71b04aSChenbo Feng __u32 file_flags; 330b2197755SDaniel Borkmann }; 331f4324551SDaniel Mack 332f4324551SDaniel Mack struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ 333f4324551SDaniel Mack __u32 target_fd; /* container object to attach to */ 334f4324551SDaniel Mack __u32 attach_bpf_fd; /* eBPF program to attach */ 335f4324551SDaniel Mack __u32 attach_type; 3367f677633SAlexei Starovoitov __u32 attach_flags; 337f4324551SDaniel Mack }; 3381cf1cae9SAlexei Starovoitov 3391cf1cae9SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ 3401cf1cae9SAlexei Starovoitov __u32 prog_fd; 3411cf1cae9SAlexei Starovoitov __u32 retval; 3421cf1cae9SAlexei Starovoitov __u32 data_size_in; 3431cf1cae9SAlexei Starovoitov __u32 data_size_out; 3441cf1cae9SAlexei Starovoitov __aligned_u64 data_in; 3451cf1cae9SAlexei Starovoitov __aligned_u64 data_out; 3461cf1cae9SAlexei Starovoitov __u32 repeat; 3471cf1cae9SAlexei Starovoitov __u32 duration; 3481cf1cae9SAlexei Starovoitov } test; 34934ad5580SMartin KaFai Lau 350b16d9aa4SMartin KaFai Lau struct { /* anonymous struct used by BPF_*_GET_*_ID */ 351b16d9aa4SMartin KaFai Lau union { 35234ad5580SMartin KaFai Lau __u32 start_id; 353b16d9aa4SMartin KaFai Lau __u32 prog_id; 354bd5f5f4eSMartin KaFai Lau __u32 map_id; 35578958fcaSMartin KaFai Lau __u32 btf_id; 356b16d9aa4SMartin KaFai Lau }; 35734ad5580SMartin KaFai Lau __u32 next_id; 3586e71b04aSChenbo Feng __u32 open_flags; 35934ad5580SMartin KaFai Lau }; 3601e270976SMartin KaFai Lau 3611e270976SMartin KaFai Lau struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ 3621e270976SMartin KaFai Lau __u32 bpf_fd; 3631e270976SMartin KaFai Lau __u32 info_len; 3641e270976SMartin KaFai Lau __aligned_u64 info; 3651e270976SMartin KaFai Lau } info; 366468e2f64SAlexei Starovoitov 367468e2f64SAlexei Starovoitov struct { /* anonymous struct used by BPF_PROG_QUERY command */ 368468e2f64SAlexei Starovoitov __u32 target_fd; /* container object to query */ 369468e2f64SAlexei Starovoitov __u32 attach_type; 370468e2f64SAlexei Starovoitov __u32 query_flags; 371468e2f64SAlexei Starovoitov __u32 attach_flags; 372468e2f64SAlexei Starovoitov __aligned_u64 prog_ids; 373468e2f64SAlexei Starovoitov __u32 prog_cnt; 374468e2f64SAlexei Starovoitov } query; 375c4f6699dSAlexei Starovoitov 376c4f6699dSAlexei Starovoitov struct { 377c4f6699dSAlexei Starovoitov __u64 name; 378c4f6699dSAlexei Starovoitov __u32 prog_fd; 379c4f6699dSAlexei Starovoitov } raw_tracepoint; 380f56a653cSMartin KaFai Lau 381f56a653cSMartin KaFai Lau struct { /* anonymous struct for BPF_BTF_LOAD */ 382f56a653cSMartin KaFai Lau __aligned_u64 btf; 383f56a653cSMartin KaFai Lau __aligned_u64 btf_log_buf; 384f56a653cSMartin KaFai Lau __u32 btf_size; 385f56a653cSMartin KaFai Lau __u32 btf_log_size; 386f56a653cSMartin KaFai Lau __u32 btf_log_level; 387f56a653cSMartin KaFai Lau }; 38841bdc4b4SYonghong Song 38941bdc4b4SYonghong Song struct { 39041bdc4b4SYonghong Song __u32 pid; /* input: pid */ 39141bdc4b4SYonghong Song __u32 fd; /* input: fd */ 39241bdc4b4SYonghong Song __u32 flags; /* input: flags */ 39341bdc4b4SYonghong Song __u32 buf_len; /* input/output: buf len */ 39441bdc4b4SYonghong Song __aligned_u64 buf; /* input/output: 39541bdc4b4SYonghong Song * tp_name for tracepoint 39641bdc4b4SYonghong Song * symbol for kprobe 39741bdc4b4SYonghong Song * filename for uprobe 39841bdc4b4SYonghong Song */ 39941bdc4b4SYonghong Song __u32 prog_id; /* output: prod_id */ 40041bdc4b4SYonghong Song __u32 fd_type; /* output: BPF_FD_TYPE_* */ 40141bdc4b4SYonghong Song __u64 probe_offset; /* output: probe_offset */ 40241bdc4b4SYonghong Song __u64 probe_addr; /* output: probe_addr */ 40341bdc4b4SYonghong Song } task_fd_query; 40499c55f7dSAlexei Starovoitov } __attribute__((aligned(8))); 40599c55f7dSAlexei Starovoitov 40656a092c8SQuentin Monnet /* The description below is an attempt at providing documentation to eBPF 40756a092c8SQuentin Monnet * developers about the multiple available eBPF helper functions. It can be 40856a092c8SQuentin Monnet * parsed and used to produce a manual page. The workflow is the following, 40956a092c8SQuentin Monnet * and requires the rst2man utility: 410ebb676daSThomas Graf * 41156a092c8SQuentin Monnet * $ ./scripts/bpf_helpers_doc.py \ 41256a092c8SQuentin Monnet * --filename include/uapi/linux/bpf.h > /tmp/bpf-helpers.rst 41356a092c8SQuentin Monnet * $ rst2man /tmp/bpf-helpers.rst > /tmp/bpf-helpers.7 41456a092c8SQuentin Monnet * $ man /tmp/bpf-helpers.7 415ebb676daSThomas Graf * 41656a092c8SQuentin Monnet * Note that in order to produce this external documentation, some RST 41756a092c8SQuentin Monnet * formatting is used in the descriptions to get "bold" and "italics" in 41856a092c8SQuentin Monnet * manual pages. Also note that the few trailing white spaces are 41956a092c8SQuentin Monnet * intentional, removing them would break paragraphs for rst2man. 420ebb676daSThomas Graf * 42156a092c8SQuentin Monnet * Start of BPF helper function descriptions: 422ad4a5223SQuentin Monnet * 423ad4a5223SQuentin Monnet * void *bpf_map_lookup_elem(struct bpf_map *map, const void *key) 424ad4a5223SQuentin Monnet * Description 425ad4a5223SQuentin Monnet * Perform a lookup in *map* for an entry associated to *key*. 426ad4a5223SQuentin Monnet * Return 427ad4a5223SQuentin Monnet * Map value associated to *key*, or **NULL** if no entry was 428ad4a5223SQuentin Monnet * found. 429ad4a5223SQuentin Monnet * 430ad4a5223SQuentin Monnet * int bpf_map_update_elem(struct bpf_map *map, const void *key, const void *value, u64 flags) 431ad4a5223SQuentin Monnet * Description 432ad4a5223SQuentin Monnet * Add or update the value of the entry associated to *key* in 433ad4a5223SQuentin Monnet * *map* with *value*. *flags* is one of: 434ad4a5223SQuentin Monnet * 435ad4a5223SQuentin Monnet * **BPF_NOEXIST** 436ad4a5223SQuentin Monnet * The entry for *key* must not exist in the map. 437ad4a5223SQuentin Monnet * **BPF_EXIST** 438ad4a5223SQuentin Monnet * The entry for *key* must already exist in the map. 439ad4a5223SQuentin Monnet * **BPF_ANY** 440ad4a5223SQuentin Monnet * No condition on the existence of the entry for *key*. 441ad4a5223SQuentin Monnet * 442ad4a5223SQuentin Monnet * Flag value **BPF_NOEXIST** cannot be used for maps of types 443ad4a5223SQuentin Monnet * **BPF_MAP_TYPE_ARRAY** or **BPF_MAP_TYPE_PERCPU_ARRAY** (all 444ad4a5223SQuentin Monnet * elements always exist), the helper would return an error. 445ad4a5223SQuentin Monnet * Return 446ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 447ad4a5223SQuentin Monnet * 448ad4a5223SQuentin Monnet * int bpf_map_delete_elem(struct bpf_map *map, const void *key) 449ad4a5223SQuentin Monnet * Description 450ad4a5223SQuentin Monnet * Delete entry with *key* from *map*. 451ad4a5223SQuentin Monnet * Return 452ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 453ad4a5223SQuentin Monnet * 454ad4a5223SQuentin Monnet * int bpf_probe_read(void *dst, u32 size, const void *src) 455ad4a5223SQuentin Monnet * Description 456ad4a5223SQuentin Monnet * For tracing programs, safely attempt to read *size* bytes from 457ad4a5223SQuentin Monnet * address *src* and store the data in *dst*. 458ad4a5223SQuentin Monnet * Return 459ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 460ad4a5223SQuentin Monnet * 461ad4a5223SQuentin Monnet * u64 bpf_ktime_get_ns(void) 462ad4a5223SQuentin Monnet * Description 463ad4a5223SQuentin Monnet * Return the time elapsed since system boot, in nanoseconds. 464ad4a5223SQuentin Monnet * Return 465ad4a5223SQuentin Monnet * Current *ktime*. 466ad4a5223SQuentin Monnet * 467ad4a5223SQuentin Monnet * int bpf_trace_printk(const char *fmt, u32 fmt_size, ...) 468ad4a5223SQuentin Monnet * Description 469ad4a5223SQuentin Monnet * This helper is a "printk()-like" facility for debugging. It 470ad4a5223SQuentin Monnet * prints a message defined by format *fmt* (of size *fmt_size*) 471ad4a5223SQuentin Monnet * to file *\/sys/kernel/debug/tracing/trace* from DebugFS, if 472ad4a5223SQuentin Monnet * available. It can take up to three additional **u64** 473ad4a5223SQuentin Monnet * arguments (as an eBPF helpers, the total number of arguments is 474ad4a5223SQuentin Monnet * limited to five). 475ad4a5223SQuentin Monnet * 476ad4a5223SQuentin Monnet * Each time the helper is called, it appends a line to the trace. 477ad4a5223SQuentin Monnet * The format of the trace is customizable, and the exact output 478ad4a5223SQuentin Monnet * one will get depends on the options set in 479ad4a5223SQuentin Monnet * *\/sys/kernel/debug/tracing/trace_options* (see also the 480ad4a5223SQuentin Monnet * *README* file under the same directory). However, it usually 481ad4a5223SQuentin Monnet * defaults to something like: 482ad4a5223SQuentin Monnet * 483ad4a5223SQuentin Monnet * :: 484ad4a5223SQuentin Monnet * 485ad4a5223SQuentin Monnet * telnet-470 [001] .N.. 419421.045894: 0x00000001: <formatted msg> 486ad4a5223SQuentin Monnet * 487ad4a5223SQuentin Monnet * In the above: 488ad4a5223SQuentin Monnet * 489ad4a5223SQuentin Monnet * * ``telnet`` is the name of the current task. 490ad4a5223SQuentin Monnet * * ``470`` is the PID of the current task. 491ad4a5223SQuentin Monnet * * ``001`` is the CPU number on which the task is 492ad4a5223SQuentin Monnet * running. 493ad4a5223SQuentin Monnet * * In ``.N..``, each character refers to a set of 494ad4a5223SQuentin Monnet * options (whether irqs are enabled, scheduling 495ad4a5223SQuentin Monnet * options, whether hard/softirqs are running, level of 496ad4a5223SQuentin Monnet * preempt_disabled respectively). **N** means that 497ad4a5223SQuentin Monnet * **TIF_NEED_RESCHED** and **PREEMPT_NEED_RESCHED** 498ad4a5223SQuentin Monnet * are set. 499ad4a5223SQuentin Monnet * * ``419421.045894`` is a timestamp. 500ad4a5223SQuentin Monnet * * ``0x00000001`` is a fake value used by BPF for the 501ad4a5223SQuentin Monnet * instruction pointer register. 502ad4a5223SQuentin Monnet * * ``<formatted msg>`` is the message formatted with 503ad4a5223SQuentin Monnet * *fmt*. 504ad4a5223SQuentin Monnet * 505ad4a5223SQuentin Monnet * The conversion specifiers supported by *fmt* are similar, but 506ad4a5223SQuentin Monnet * more limited than for printk(). They are **%d**, **%i**, 507ad4a5223SQuentin Monnet * **%u**, **%x**, **%ld**, **%li**, **%lu**, **%lx**, **%lld**, 508ad4a5223SQuentin Monnet * **%lli**, **%llu**, **%llx**, **%p**, **%s**. No modifier (size 509ad4a5223SQuentin Monnet * of field, padding with zeroes, etc.) is available, and the 510ad4a5223SQuentin Monnet * helper will return **-EINVAL** (but print nothing) if it 511ad4a5223SQuentin Monnet * encounters an unknown specifier. 512ad4a5223SQuentin Monnet * 513ad4a5223SQuentin Monnet * Also, note that **bpf_trace_printk**\ () is slow, and should 514ad4a5223SQuentin Monnet * only be used for debugging purposes. For this reason, a notice 515ad4a5223SQuentin Monnet * bloc (spanning several lines) is printed to kernel logs and 516ad4a5223SQuentin Monnet * states that the helper should not be used "for production use" 517ad4a5223SQuentin Monnet * the first time this helper is used (or more precisely, when 518ad4a5223SQuentin Monnet * **trace_printk**\ () buffers are allocated). For passing values 519ad4a5223SQuentin Monnet * to user space, perf events should be preferred. 520ad4a5223SQuentin Monnet * Return 521ad4a5223SQuentin Monnet * The number of bytes written to the buffer, or a negative error 522ad4a5223SQuentin Monnet * in case of failure. 523ad4a5223SQuentin Monnet * 5241fdd08beSQuentin Monnet * u32 bpf_get_prandom_u32(void) 5251fdd08beSQuentin Monnet * Description 5261fdd08beSQuentin Monnet * Get a pseudo-random number. 5271fdd08beSQuentin Monnet * 5281fdd08beSQuentin Monnet * From a security point of view, this helper uses its own 5291fdd08beSQuentin Monnet * pseudo-random internal state, and cannot be used to infer the 5301fdd08beSQuentin Monnet * seed of other random functions in the kernel. However, it is 5311fdd08beSQuentin Monnet * essential to note that the generator used by the helper is not 5321fdd08beSQuentin Monnet * cryptographically secure. 5331fdd08beSQuentin Monnet * Return 5341fdd08beSQuentin Monnet * A random 32-bit unsigned value. 5351fdd08beSQuentin Monnet * 5361fdd08beSQuentin Monnet * u32 bpf_get_smp_processor_id(void) 5371fdd08beSQuentin Monnet * Description 5381fdd08beSQuentin Monnet * Get the SMP (symmetric multiprocessing) processor id. Note that 5391fdd08beSQuentin Monnet * all programs run with preemption disabled, which means that the 5401fdd08beSQuentin Monnet * SMP processor id is stable during all the execution of the 5411fdd08beSQuentin Monnet * program. 5421fdd08beSQuentin Monnet * Return 5431fdd08beSQuentin Monnet * The SMP id of the processor running the program. 5441fdd08beSQuentin Monnet * 545ad4a5223SQuentin Monnet * int bpf_skb_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len, u64 flags) 546ad4a5223SQuentin Monnet * Description 547ad4a5223SQuentin Monnet * Store *len* bytes from address *from* into the packet 548ad4a5223SQuentin Monnet * associated to *skb*, at *offset*. *flags* are a combination of 549ad4a5223SQuentin Monnet * **BPF_F_RECOMPUTE_CSUM** (automatically recompute the 550ad4a5223SQuentin Monnet * checksum for the packet after storing the bytes) and 551ad4a5223SQuentin Monnet * **BPF_F_INVALIDATE_HASH** (set *skb*\ **->hash**, *skb*\ 552ad4a5223SQuentin Monnet * **->swhash** and *skb*\ **->l4hash** to 0). 553ad4a5223SQuentin Monnet * 554ad4a5223SQuentin Monnet * A call to this helper is susceptible to change the underlaying 555ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 556ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 557ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 558ad4a5223SQuentin Monnet * direct packet access. 559ad4a5223SQuentin Monnet * Return 560ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 561ad4a5223SQuentin Monnet * 562ad4a5223SQuentin Monnet * int bpf_l3_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 size) 563ad4a5223SQuentin Monnet * Description 564ad4a5223SQuentin Monnet * Recompute the layer 3 (e.g. IP) checksum for the packet 565ad4a5223SQuentin Monnet * associated to *skb*. Computation is incremental, so the helper 566ad4a5223SQuentin Monnet * must know the former value of the header field that was 567ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 568ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored in *size*. 569ad4a5223SQuentin Monnet * Alternatively, it is possible to store the difference between 570ad4a5223SQuentin Monnet * the previous and the new values of the header field in *to*, by 571ad4a5223SQuentin Monnet * setting *from* and *size* to 0. For both methods, *offset* 572ad4a5223SQuentin Monnet * indicates the location of the IP checksum within the packet. 573ad4a5223SQuentin Monnet * 574ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 575ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 576ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 577ad4a5223SQuentin Monnet * checksum to update. 578ad4a5223SQuentin Monnet * 579ad4a5223SQuentin Monnet * A call to this helper is susceptible to change the underlaying 580ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 581ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 582ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 583ad4a5223SQuentin Monnet * direct packet access. 584ad4a5223SQuentin Monnet * Return 585ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 586ad4a5223SQuentin Monnet * 587ad4a5223SQuentin Monnet * int bpf_l4_csum_replace(struct sk_buff *skb, u32 offset, u64 from, u64 to, u64 flags) 588ad4a5223SQuentin Monnet * Description 589ad4a5223SQuentin Monnet * Recompute the layer 4 (e.g. TCP, UDP or ICMP) checksum for the 590ad4a5223SQuentin Monnet * packet associated to *skb*. Computation is incremental, so the 591ad4a5223SQuentin Monnet * helper must know the former value of the header field that was 592ad4a5223SQuentin Monnet * modified (*from*), the new value of this field (*to*), and the 593ad4a5223SQuentin Monnet * number of bytes (2 or 4) for this field, stored on the lowest 594ad4a5223SQuentin Monnet * four bits of *flags*. Alternatively, it is possible to store 595ad4a5223SQuentin Monnet * the difference between the previous and the new values of the 596ad4a5223SQuentin Monnet * header field in *to*, by setting *from* and the four lowest 597ad4a5223SQuentin Monnet * bits of *flags* to 0. For both methods, *offset* indicates the 598ad4a5223SQuentin Monnet * location of the IP checksum within the packet. In addition to 599ad4a5223SQuentin Monnet * the size of the field, *flags* can be added (bitwise OR) actual 600ad4a5223SQuentin Monnet * flags. With **BPF_F_MARK_MANGLED_0**, a null checksum is left 601ad4a5223SQuentin Monnet * untouched (unless **BPF_F_MARK_ENFORCE** is added as well), and 602ad4a5223SQuentin Monnet * for updates resulting in a null checksum the value is set to 603ad4a5223SQuentin Monnet * **CSUM_MANGLED_0** instead. Flag **BPF_F_PSEUDO_HDR** indicates 604ad4a5223SQuentin Monnet * the checksum is to be computed against a pseudo-header. 605ad4a5223SQuentin Monnet * 606ad4a5223SQuentin Monnet * This helper works in combination with **bpf_csum_diff**\ (), 607ad4a5223SQuentin Monnet * which does not update the checksum in-place, but offers more 608ad4a5223SQuentin Monnet * flexibility and can handle sizes larger than 2 or 4 for the 609ad4a5223SQuentin Monnet * checksum to update. 610ad4a5223SQuentin Monnet * 611ad4a5223SQuentin Monnet * A call to this helper is susceptible to change the underlaying 612ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 613ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 614ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 615ad4a5223SQuentin Monnet * direct packet access. 616ad4a5223SQuentin Monnet * Return 617ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 618ad4a5223SQuentin Monnet * 619ad4a5223SQuentin Monnet * int bpf_tail_call(void *ctx, struct bpf_map *prog_array_map, u32 index) 620ad4a5223SQuentin Monnet * Description 621ad4a5223SQuentin Monnet * This special helper is used to trigger a "tail call", or in 622ad4a5223SQuentin Monnet * other words, to jump into another eBPF program. The same stack 623ad4a5223SQuentin Monnet * frame is used (but values on stack and in registers for the 624ad4a5223SQuentin Monnet * caller are not accessible to the callee). This mechanism allows 625ad4a5223SQuentin Monnet * for program chaining, either for raising the maximum number of 626ad4a5223SQuentin Monnet * available eBPF instructions, or to execute given programs in 627ad4a5223SQuentin Monnet * conditional blocks. For security reasons, there is an upper 628ad4a5223SQuentin Monnet * limit to the number of successive tail calls that can be 629ad4a5223SQuentin Monnet * performed. 630ad4a5223SQuentin Monnet * 631ad4a5223SQuentin Monnet * Upon call of this helper, the program attempts to jump into a 632ad4a5223SQuentin Monnet * program referenced at index *index* in *prog_array_map*, a 633ad4a5223SQuentin Monnet * special map of type **BPF_MAP_TYPE_PROG_ARRAY**, and passes 634ad4a5223SQuentin Monnet * *ctx*, a pointer to the context. 635ad4a5223SQuentin Monnet * 636ad4a5223SQuentin Monnet * If the call succeeds, the kernel immediately runs the first 637ad4a5223SQuentin Monnet * instruction of the new program. This is not a function call, 638ad4a5223SQuentin Monnet * and it never returns to the previous program. If the call 639ad4a5223SQuentin Monnet * fails, then the helper has no effect, and the caller continues 640ad4a5223SQuentin Monnet * to run its subsequent instructions. A call can fail if the 641ad4a5223SQuentin Monnet * destination program for the jump does not exist (i.e. *index* 642ad4a5223SQuentin Monnet * is superior to the number of entries in *prog_array_map*), or 643ad4a5223SQuentin Monnet * if the maximum number of tail calls has been reached for this 644ad4a5223SQuentin Monnet * chain of programs. This limit is defined in the kernel by the 645ad4a5223SQuentin Monnet * macro **MAX_TAIL_CALL_CNT** (not accessible to user space), 646ad4a5223SQuentin Monnet * which is currently set to 32. 647ad4a5223SQuentin Monnet * Return 648ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 649ad4a5223SQuentin Monnet * 650ad4a5223SQuentin Monnet * int bpf_clone_redirect(struct sk_buff *skb, u32 ifindex, u64 flags) 651ad4a5223SQuentin Monnet * Description 652ad4a5223SQuentin Monnet * Clone and redirect the packet associated to *skb* to another 653ad4a5223SQuentin Monnet * net device of index *ifindex*. Both ingress and egress 654ad4a5223SQuentin Monnet * interfaces can be used for redirection. The **BPF_F_INGRESS** 655ad4a5223SQuentin Monnet * value in *flags* is used to make the distinction (ingress path 656ad4a5223SQuentin Monnet * is selected if the flag is present, egress path otherwise). 657ad4a5223SQuentin Monnet * This is the only flag supported for now. 658ad4a5223SQuentin Monnet * 659ad4a5223SQuentin Monnet * In comparison with **bpf_redirect**\ () helper, 660ad4a5223SQuentin Monnet * **bpf_clone_redirect**\ () has the associated cost of 661ad4a5223SQuentin Monnet * duplicating the packet buffer, but this can be executed out of 662ad4a5223SQuentin Monnet * the eBPF program. Conversely, **bpf_redirect**\ () is more 663ad4a5223SQuentin Monnet * efficient, but it is handled through an action code where the 664ad4a5223SQuentin Monnet * redirection happens only after the eBPF program has returned. 665ad4a5223SQuentin Monnet * 666ad4a5223SQuentin Monnet * A call to this helper is susceptible to change the underlaying 667ad4a5223SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 668ad4a5223SQuentin Monnet * previously done by the verifier are invalidated and must be 669ad4a5223SQuentin Monnet * performed again, if the helper is used in combination with 670ad4a5223SQuentin Monnet * direct packet access. 671ad4a5223SQuentin Monnet * Return 672ad4a5223SQuentin Monnet * 0 on success, or a negative error in case of failure. 673c456dec4SQuentin Monnet * 674c456dec4SQuentin Monnet * u64 bpf_get_current_pid_tgid(void) 675c456dec4SQuentin Monnet * Return 676c456dec4SQuentin Monnet * A 64-bit integer containing the current tgid and pid, and 677c456dec4SQuentin Monnet * created as such: 678c456dec4SQuentin Monnet * *current_task*\ **->tgid << 32 \|** 679c456dec4SQuentin Monnet * *current_task*\ **->pid**. 680c456dec4SQuentin Monnet * 681c456dec4SQuentin Monnet * u64 bpf_get_current_uid_gid(void) 682c456dec4SQuentin Monnet * Return 683c456dec4SQuentin Monnet * A 64-bit integer containing the current GID and UID, and 684c456dec4SQuentin Monnet * created as such: *current_gid* **<< 32 \|** *current_uid*. 685c456dec4SQuentin Monnet * 686c456dec4SQuentin Monnet * int bpf_get_current_comm(char *buf, u32 size_of_buf) 687c456dec4SQuentin Monnet * Description 688c456dec4SQuentin Monnet * Copy the **comm** attribute of the current task into *buf* of 689c456dec4SQuentin Monnet * *size_of_buf*. The **comm** attribute contains the name of 690c456dec4SQuentin Monnet * the executable (excluding the path) for the current task. The 691c456dec4SQuentin Monnet * *size_of_buf* must be strictly positive. On success, the 692c456dec4SQuentin Monnet * helper makes sure that the *buf* is NUL-terminated. On failure, 693c456dec4SQuentin Monnet * it is filled with zeroes. 694c456dec4SQuentin Monnet * Return 695c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 696c456dec4SQuentin Monnet * 6971fdd08beSQuentin Monnet * u32 bpf_get_cgroup_classid(struct sk_buff *skb) 6981fdd08beSQuentin Monnet * Description 6991fdd08beSQuentin Monnet * Retrieve the classid for the current task, i.e. for the net_cls 7001fdd08beSQuentin Monnet * cgroup to which *skb* belongs. 7011fdd08beSQuentin Monnet * 7021fdd08beSQuentin Monnet * This helper can be used on TC egress path, but not on ingress. 7031fdd08beSQuentin Monnet * 7041fdd08beSQuentin Monnet * The net_cls cgroup provides an interface to tag network packets 7051fdd08beSQuentin Monnet * based on a user-provided identifier for all traffic coming from 7061fdd08beSQuentin Monnet * the tasks belonging to the related cgroup. See also the related 7071fdd08beSQuentin Monnet * kernel documentation, available from the Linux sources in file 7081fdd08beSQuentin Monnet * *Documentation/cgroup-v1/net_cls.txt*. 7091fdd08beSQuentin Monnet * 7101fdd08beSQuentin Monnet * The Linux kernel has two versions for cgroups: there are 7111fdd08beSQuentin Monnet * cgroups v1 and cgroups v2. Both are available to users, who can 7121fdd08beSQuentin Monnet * use a mixture of them, but note that the net_cls cgroup is for 7131fdd08beSQuentin Monnet * cgroup v1 only. This makes it incompatible with BPF programs 7141fdd08beSQuentin Monnet * run on cgroups, which is a cgroup-v2-only feature (a socket can 7151fdd08beSQuentin Monnet * only hold data for one version of cgroups at a time). 7161fdd08beSQuentin Monnet * 7171fdd08beSQuentin Monnet * This helper is only available is the kernel was compiled with 7181fdd08beSQuentin Monnet * the **CONFIG_CGROUP_NET_CLASSID** configuration option set to 7191fdd08beSQuentin Monnet * "**y**" or to "**m**". 7201fdd08beSQuentin Monnet * Return 7211fdd08beSQuentin Monnet * The classid, or 0 for the default unconfigured classid. 7221fdd08beSQuentin Monnet * 723c456dec4SQuentin Monnet * int bpf_skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci) 724c456dec4SQuentin Monnet * Description 725c456dec4SQuentin Monnet * Push a *vlan_tci* (VLAN tag control information) of protocol 726c456dec4SQuentin Monnet * *vlan_proto* to the packet associated to *skb*, then update 727c456dec4SQuentin Monnet * the checksum. Note that if *vlan_proto* is different from 728c456dec4SQuentin Monnet * **ETH_P_8021Q** and **ETH_P_8021AD**, it is considered to 729c456dec4SQuentin Monnet * be **ETH_P_8021Q**. 730c456dec4SQuentin Monnet * 731c456dec4SQuentin Monnet * A call to this helper is susceptible to change the underlaying 732c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 733c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 734c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 735c456dec4SQuentin Monnet * direct packet access. 736c456dec4SQuentin Monnet * Return 737c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 738c456dec4SQuentin Monnet * 739c456dec4SQuentin Monnet * int bpf_skb_vlan_pop(struct sk_buff *skb) 740c456dec4SQuentin Monnet * Description 741c456dec4SQuentin Monnet * Pop a VLAN header from the packet associated to *skb*. 742c456dec4SQuentin Monnet * 743c456dec4SQuentin Monnet * A call to this helper is susceptible to change the underlaying 744c456dec4SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 745c456dec4SQuentin Monnet * previously done by the verifier are invalidated and must be 746c456dec4SQuentin Monnet * performed again, if the helper is used in combination with 747c456dec4SQuentin Monnet * direct packet access. 748c456dec4SQuentin Monnet * Return 749c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 750c456dec4SQuentin Monnet * 751c456dec4SQuentin Monnet * int bpf_skb_get_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 752c456dec4SQuentin Monnet * Description 753c456dec4SQuentin Monnet * Get tunnel metadata. This helper takes a pointer *key* to an 754c456dec4SQuentin Monnet * empty **struct bpf_tunnel_key** of **size**, that will be 755c456dec4SQuentin Monnet * filled with tunnel metadata for the packet associated to *skb*. 756c456dec4SQuentin Monnet * The *flags* can be set to **BPF_F_TUNINFO_IPV6**, which 757c456dec4SQuentin Monnet * indicates that the tunnel is based on IPv6 protocol instead of 758c456dec4SQuentin Monnet * IPv4. 759c456dec4SQuentin Monnet * 760c456dec4SQuentin Monnet * The **struct bpf_tunnel_key** is an object that generalizes the 761c456dec4SQuentin Monnet * principal parameters used by various tunneling protocols into a 762c456dec4SQuentin Monnet * single struct. This way, it can be used to easily make a 763c456dec4SQuentin Monnet * decision based on the contents of the encapsulation header, 764c456dec4SQuentin Monnet * "summarized" in this struct. In particular, it holds the IP 765c456dec4SQuentin Monnet * address of the remote end (IPv4 or IPv6, depending on the case) 766c456dec4SQuentin Monnet * in *key*\ **->remote_ipv4** or *key*\ **->remote_ipv6**. Also, 767c456dec4SQuentin Monnet * this struct exposes the *key*\ **->tunnel_id**, which is 768c456dec4SQuentin Monnet * generally mapped to a VNI (Virtual Network Identifier), making 769c456dec4SQuentin Monnet * it programmable together with the **bpf_skb_set_tunnel_key**\ 770c456dec4SQuentin Monnet * () helper. 771c456dec4SQuentin Monnet * 772c456dec4SQuentin Monnet * Let's imagine that the following code is part of a program 773c456dec4SQuentin Monnet * attached to the TC ingress interface, on one end of a GRE 774c456dec4SQuentin Monnet * tunnel, and is supposed to filter out all messages coming from 775c456dec4SQuentin Monnet * remote ends with IPv4 address other than 10.0.0.1: 776c456dec4SQuentin Monnet * 777c456dec4SQuentin Monnet * :: 778c456dec4SQuentin Monnet * 779c456dec4SQuentin Monnet * int ret; 780c456dec4SQuentin Monnet * struct bpf_tunnel_key key = {}; 781c456dec4SQuentin Monnet * 782c456dec4SQuentin Monnet * ret = bpf_skb_get_tunnel_key(skb, &key, sizeof(key), 0); 783c456dec4SQuentin Monnet * if (ret < 0) 784c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 785c456dec4SQuentin Monnet * 786c456dec4SQuentin Monnet * if (key.remote_ipv4 != 0x0a000001) 787c456dec4SQuentin Monnet * return TC_ACT_SHOT; // drop packet 788c456dec4SQuentin Monnet * 789c456dec4SQuentin Monnet * return TC_ACT_OK; // accept packet 790c456dec4SQuentin Monnet * 791c456dec4SQuentin Monnet * This interface can also be used with all encapsulation devices 792c456dec4SQuentin Monnet * that can operate in "collect metadata" mode: instead of having 793c456dec4SQuentin Monnet * one network device per specific configuration, the "collect 794c456dec4SQuentin Monnet * metadata" mode only requires a single device where the 795c456dec4SQuentin Monnet * configuration can be extracted from this helper. 796c456dec4SQuentin Monnet * 797c456dec4SQuentin Monnet * This can be used together with various tunnels such as VXLan, 798c456dec4SQuentin Monnet * Geneve, GRE or IP in IP (IPIP). 799c456dec4SQuentin Monnet * Return 800c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 801c456dec4SQuentin Monnet * 802c456dec4SQuentin Monnet * int bpf_skb_set_tunnel_key(struct sk_buff *skb, struct bpf_tunnel_key *key, u32 size, u64 flags) 803c456dec4SQuentin Monnet * Description 804c456dec4SQuentin Monnet * Populate tunnel metadata for packet associated to *skb.* The 805c456dec4SQuentin Monnet * tunnel metadata is set to the contents of *key*, of *size*. The 806c456dec4SQuentin Monnet * *flags* can be set to a combination of the following values: 807c456dec4SQuentin Monnet * 808c456dec4SQuentin Monnet * **BPF_F_TUNINFO_IPV6** 809c456dec4SQuentin Monnet * Indicate that the tunnel is based on IPv6 protocol 810c456dec4SQuentin Monnet * instead of IPv4. 811c456dec4SQuentin Monnet * **BPF_F_ZERO_CSUM_TX** 812c456dec4SQuentin Monnet * For IPv4 packets, add a flag to tunnel metadata 813c456dec4SQuentin Monnet * indicating that checksum computation should be skipped 814c456dec4SQuentin Monnet * and checksum set to zeroes. 815c456dec4SQuentin Monnet * **BPF_F_DONT_FRAGMENT** 816c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that the 817c456dec4SQuentin Monnet * packet should not be fragmented. 818c456dec4SQuentin Monnet * **BPF_F_SEQ_NUMBER** 819c456dec4SQuentin Monnet * Add a flag to tunnel metadata indicating that a 820c456dec4SQuentin Monnet * sequence number should be added to tunnel header before 821c456dec4SQuentin Monnet * sending the packet. This flag was added for GRE 822c456dec4SQuentin Monnet * encapsulation, but might be used with other protocols 823c456dec4SQuentin Monnet * as well in the future. 824c456dec4SQuentin Monnet * 825c456dec4SQuentin Monnet * Here is a typical usage on the transmit path: 826c456dec4SQuentin Monnet * 827c456dec4SQuentin Monnet * :: 828c456dec4SQuentin Monnet * 829c456dec4SQuentin Monnet * struct bpf_tunnel_key key; 830c456dec4SQuentin Monnet * populate key ... 831c456dec4SQuentin Monnet * bpf_skb_set_tunnel_key(skb, &key, sizeof(key), 0); 832c456dec4SQuentin Monnet * bpf_clone_redirect(skb, vxlan_dev_ifindex, 0); 833c456dec4SQuentin Monnet * 834c456dec4SQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_key**\ () 835c456dec4SQuentin Monnet * helper for additional information. 836c456dec4SQuentin Monnet * Return 837c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 838c456dec4SQuentin Monnet * 839c6b5fb86SQuentin Monnet * u64 bpf_perf_event_read(struct bpf_map *map, u64 flags) 840c6b5fb86SQuentin Monnet * Description 841c6b5fb86SQuentin Monnet * Read the value of a perf event counter. This helper relies on a 842c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of 843c6b5fb86SQuentin Monnet * the perf event counter is selected when *map* is updated with 844c6b5fb86SQuentin Monnet * perf event file descriptors. The *map* is an array whose size 845c6b5fb86SQuentin Monnet * is the number of available CPUs, and each cell contains a value 846c6b5fb86SQuentin Monnet * relative to one CPU. The value to retrieve is indicated by 847c6b5fb86SQuentin Monnet * *flags*, that contains the index of the CPU to look up, masked 848c6b5fb86SQuentin Monnet * with **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 849c6b5fb86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 850c6b5fb86SQuentin Monnet * current CPU should be retrieved. 851c6b5fb86SQuentin Monnet * 852c6b5fb86SQuentin Monnet * Note that before Linux 4.13, only hardware perf event can be 853c6b5fb86SQuentin Monnet * retrieved. 854c6b5fb86SQuentin Monnet * 855c6b5fb86SQuentin Monnet * Also, be aware that the newer helper 856c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () is recommended over 8573bd5a09bSQuentin Monnet * **bpf_perf_event_read**\ () in general. The latter has some ABI 858c6b5fb86SQuentin Monnet * quirks where error and counter value are used as a return code 859c6b5fb86SQuentin Monnet * (which is wrong to do since ranges may overlap). This issue is 8603bd5a09bSQuentin Monnet * fixed with **bpf_perf_event_read_value**\ (), which at the same 8613bd5a09bSQuentin Monnet * time provides more features over the **bpf_perf_event_read**\ 8623bd5a09bSQuentin Monnet * () interface. Please refer to the description of 863c6b5fb86SQuentin Monnet * **bpf_perf_event_read_value**\ () for details. 864c6b5fb86SQuentin Monnet * Return 865c6b5fb86SQuentin Monnet * The value of the perf event counter read from the map, or a 866c6b5fb86SQuentin Monnet * negative error code in case of failure. 867c6b5fb86SQuentin Monnet * 868c456dec4SQuentin Monnet * int bpf_redirect(u32 ifindex, u64 flags) 869c456dec4SQuentin Monnet * Description 870c456dec4SQuentin Monnet * Redirect the packet to another net device of index *ifindex*. 871c456dec4SQuentin Monnet * This helper is somewhat similar to **bpf_clone_redirect**\ 872c456dec4SQuentin Monnet * (), except that the packet is not cloned, which provides 873c456dec4SQuentin Monnet * increased performance. 874c456dec4SQuentin Monnet * 875c456dec4SQuentin Monnet * Except for XDP, both ingress and egress interfaces can be used 876c456dec4SQuentin Monnet * for redirection. The **BPF_F_INGRESS** value in *flags* is used 877c456dec4SQuentin Monnet * to make the distinction (ingress path is selected if the flag 878c456dec4SQuentin Monnet * is present, egress path otherwise). Currently, XDP only 879c456dec4SQuentin Monnet * supports redirection to the egress interface, and accepts no 880c456dec4SQuentin Monnet * flag at all. 881c456dec4SQuentin Monnet * 882c456dec4SQuentin Monnet * The same effect can be attained with the more generic 883c456dec4SQuentin Monnet * **bpf_redirect_map**\ (), which requires specific maps to be 884c456dec4SQuentin Monnet * used but offers better performance. 885c456dec4SQuentin Monnet * Return 886c456dec4SQuentin Monnet * For XDP, the helper returns **XDP_REDIRECT** on success or 887c456dec4SQuentin Monnet * **XDP_ABORTED** on error. For other program types, the values 888c456dec4SQuentin Monnet * are **TC_ACT_REDIRECT** on success or **TC_ACT_SHOT** on 889c456dec4SQuentin Monnet * error. 890c456dec4SQuentin Monnet * 8911fdd08beSQuentin Monnet * u32 bpf_get_route_realm(struct sk_buff *skb) 8921fdd08beSQuentin Monnet * Description 8931fdd08beSQuentin Monnet * Retrieve the realm or the route, that is to say the 8941fdd08beSQuentin Monnet * **tclassid** field of the destination for the *skb*. The 8951fdd08beSQuentin Monnet * indentifier retrieved is a user-provided tag, similar to the 8961fdd08beSQuentin Monnet * one used with the net_cls cgroup (see description for 8971fdd08beSQuentin Monnet * **bpf_get_cgroup_classid**\ () helper), but here this tag is 8981fdd08beSQuentin Monnet * held by a route (a destination entry), not by a task. 8991fdd08beSQuentin Monnet * 9001fdd08beSQuentin Monnet * Retrieving this identifier works with the clsact TC egress hook 9011fdd08beSQuentin Monnet * (see also **tc-bpf(8)**), or alternatively on conventional 9021fdd08beSQuentin Monnet * classful egress qdiscs, but not on TC ingress path. In case of 9031fdd08beSQuentin Monnet * clsact TC egress hook, this has the advantage that, internally, 9041fdd08beSQuentin Monnet * the destination entry has not been dropped yet in the transmit 9051fdd08beSQuentin Monnet * path. Therefore, the destination entry does not need to be 9061fdd08beSQuentin Monnet * artificially held via **netif_keep_dst**\ () for a classful 9071fdd08beSQuentin Monnet * qdisc until the *skb* is freed. 9081fdd08beSQuentin Monnet * 9091fdd08beSQuentin Monnet * This helper is available only if the kernel was compiled with 9101fdd08beSQuentin Monnet * **CONFIG_IP_ROUTE_CLASSID** configuration option. 9111fdd08beSQuentin Monnet * Return 9121fdd08beSQuentin Monnet * The realm of the route for the packet associated to *skb*, or 0 9131fdd08beSQuentin Monnet * if none was found. 9141fdd08beSQuentin Monnet * 915c456dec4SQuentin Monnet * int bpf_perf_event_output(struct pt_reg *ctx, struct bpf_map *map, u64 flags, void *data, u64 size) 916c456dec4SQuentin Monnet * Description 917c456dec4SQuentin Monnet * Write raw *data* blob into a special BPF perf event held by 918c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. This perf 919c456dec4SQuentin Monnet * event must have the following attributes: **PERF_SAMPLE_RAW** 920c456dec4SQuentin Monnet * as **sample_type**, **PERF_TYPE_SOFTWARE** as **type**, and 921c456dec4SQuentin Monnet * **PERF_COUNT_SW_BPF_OUTPUT** as **config**. 922c456dec4SQuentin Monnet * 923c456dec4SQuentin Monnet * The *flags* are used to indicate the index in *map* for which 924c456dec4SQuentin Monnet * the value must be put, masked with **BPF_F_INDEX_MASK**. 925c456dec4SQuentin Monnet * Alternatively, *flags* can be set to **BPF_F_CURRENT_CPU** 926c456dec4SQuentin Monnet * to indicate that the index of the current CPU core should be 927c456dec4SQuentin Monnet * used. 928c456dec4SQuentin Monnet * 929c456dec4SQuentin Monnet * The value to write, of *size*, is passed through eBPF stack and 930c456dec4SQuentin Monnet * pointed by *data*. 931c456dec4SQuentin Monnet * 932c456dec4SQuentin Monnet * The context of the program *ctx* needs also be passed to the 933c456dec4SQuentin Monnet * helper. 934c456dec4SQuentin Monnet * 935c456dec4SQuentin Monnet * On user space, a program willing to read the values needs to 936c456dec4SQuentin Monnet * call **perf_event_open**\ () on the perf event (either for 937c456dec4SQuentin Monnet * one or for all CPUs) and to store the file descriptor into the 938c456dec4SQuentin Monnet * *map*. This must be done before the eBPF program can send data 939c456dec4SQuentin Monnet * into it. An example is available in file 940c456dec4SQuentin Monnet * *samples/bpf/trace_output_user.c* in the Linux kernel source 941c456dec4SQuentin Monnet * tree (the eBPF program counterpart is in 942c456dec4SQuentin Monnet * *samples/bpf/trace_output_kern.c*). 943c456dec4SQuentin Monnet * 944c456dec4SQuentin Monnet * **bpf_perf_event_output**\ () achieves better performance 945c456dec4SQuentin Monnet * than **bpf_trace_printk**\ () for sharing data with user 946c456dec4SQuentin Monnet * space, and is much better suitable for streaming data from eBPF 947c456dec4SQuentin Monnet * programs. 948c456dec4SQuentin Monnet * 949c456dec4SQuentin Monnet * Note that this helper is not restricted to tracing use cases 950c456dec4SQuentin Monnet * and can be used with programs attached to TC or XDP as well, 951c456dec4SQuentin Monnet * where it allows for passing data to user space listeners. Data 952c456dec4SQuentin Monnet * can be: 953c456dec4SQuentin Monnet * 954c456dec4SQuentin Monnet * * Only custom structs, 955c456dec4SQuentin Monnet * * Only the packet payload, or 956c456dec4SQuentin Monnet * * A combination of both. 957c456dec4SQuentin Monnet * Return 958c456dec4SQuentin Monnet * 0 on success, or a negative error in case of failure. 959c456dec4SQuentin Monnet * 9601fdd08beSQuentin Monnet * int bpf_skb_load_bytes(const struct sk_buff *skb, u32 offset, void *to, u32 len) 9611fdd08beSQuentin Monnet * Description 9621fdd08beSQuentin Monnet * This helper was provided as an easy way to load data from a 9631fdd08beSQuentin Monnet * packet. It can be used to load *len* bytes from *offset* from 9641fdd08beSQuentin Monnet * the packet associated to *skb*, into the buffer pointed by 9651fdd08beSQuentin Monnet * *to*. 9661fdd08beSQuentin Monnet * 9671fdd08beSQuentin Monnet * Since Linux 4.7, usage of this helper has mostly been replaced 9681fdd08beSQuentin Monnet * by "direct packet access", enabling packet data to be 9691fdd08beSQuentin Monnet * manipulated with *skb*\ **->data** and *skb*\ **->data_end** 9701fdd08beSQuentin Monnet * pointing respectively to the first byte of packet data and to 9711fdd08beSQuentin Monnet * the byte after the last byte of packet data. However, it 9721fdd08beSQuentin Monnet * remains useful if one wishes to read large quantities of data 9731fdd08beSQuentin Monnet * at once from a packet into the eBPF stack. 9741fdd08beSQuentin Monnet * Return 9751fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 9761fdd08beSQuentin Monnet * 977c456dec4SQuentin Monnet * int bpf_get_stackid(struct pt_reg *ctx, struct bpf_map *map, u64 flags) 978c456dec4SQuentin Monnet * Description 979c456dec4SQuentin Monnet * Walk a user or a kernel stack and return its id. To achieve 980c456dec4SQuentin Monnet * this, the helper needs *ctx*, which is a pointer to the context 981c456dec4SQuentin Monnet * on which the tracing program is executed, and a pointer to a 982c456dec4SQuentin Monnet * *map* of type **BPF_MAP_TYPE_STACK_TRACE**. 983c456dec4SQuentin Monnet * 984c456dec4SQuentin Monnet * The last argument, *flags*, holds the number of stack frames to 985c456dec4SQuentin Monnet * skip (from 0 to 255), masked with 986c456dec4SQuentin Monnet * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 987c456dec4SQuentin Monnet * a combination of the following flags: 988c456dec4SQuentin Monnet * 989c456dec4SQuentin Monnet * **BPF_F_USER_STACK** 990c456dec4SQuentin Monnet * Collect a user space stack instead of a kernel stack. 991c456dec4SQuentin Monnet * **BPF_F_FAST_STACK_CMP** 992c456dec4SQuentin Monnet * Compare stacks by hash only. 993c456dec4SQuentin Monnet * **BPF_F_REUSE_STACKID** 994c456dec4SQuentin Monnet * If two different stacks hash into the same *stackid*, 995c456dec4SQuentin Monnet * discard the old one. 996c456dec4SQuentin Monnet * 997c456dec4SQuentin Monnet * The stack id retrieved is a 32 bit long integer handle which 998c456dec4SQuentin Monnet * can be further combined with other data (including other stack 999c456dec4SQuentin Monnet * ids) and used as a key into maps. This can be useful for 1000c456dec4SQuentin Monnet * generating a variety of graphs (such as flame graphs or off-cpu 1001c456dec4SQuentin Monnet * graphs). 1002c456dec4SQuentin Monnet * 1003c456dec4SQuentin Monnet * For walking a stack, this helper is an improvement over 1004c456dec4SQuentin Monnet * **bpf_probe_read**\ (), which can be used with unrolled loops 1005c456dec4SQuentin Monnet * but is not efficient and consumes a lot of eBPF instructions. 1006c456dec4SQuentin Monnet * Instead, **bpf_get_stackid**\ () can collect up to 1007c456dec4SQuentin Monnet * **PERF_MAX_STACK_DEPTH** both kernel and user frames. Note that 1008c456dec4SQuentin Monnet * this limit can be controlled with the **sysctl** program, and 1009c456dec4SQuentin Monnet * that it should be manually increased in order to profile long 1010c456dec4SQuentin Monnet * user stacks (such as stacks for Java programs). To do so, use: 1011c456dec4SQuentin Monnet * 1012c456dec4SQuentin Monnet * :: 1013c456dec4SQuentin Monnet * 1014c456dec4SQuentin Monnet * # sysctl kernel.perf_event_max_stack=<new value> 1015c456dec4SQuentin Monnet * Return 1016c456dec4SQuentin Monnet * The positive or null stack id on success, or a negative error 1017c456dec4SQuentin Monnet * in case of failure. 1018c456dec4SQuentin Monnet * 10191fdd08beSQuentin Monnet * s64 bpf_csum_diff(__be32 *from, u32 from_size, __be32 *to, u32 to_size, __wsum seed) 10201fdd08beSQuentin Monnet * Description 10211fdd08beSQuentin Monnet * Compute a checksum difference, from the raw buffer pointed by 10221fdd08beSQuentin Monnet * *from*, of length *from_size* (that must be a multiple of 4), 10231fdd08beSQuentin Monnet * towards the raw buffer pointed by *to*, of size *to_size* 10241fdd08beSQuentin Monnet * (same remark). An optional *seed* can be added to the value 10251fdd08beSQuentin Monnet * (this can be cascaded, the seed may come from a previous call 10261fdd08beSQuentin Monnet * to the helper). 10271fdd08beSQuentin Monnet * 10281fdd08beSQuentin Monnet * This is flexible enough to be used in several ways: 10291fdd08beSQuentin Monnet * 10301fdd08beSQuentin Monnet * * With *from_size* == 0, *to_size* > 0 and *seed* set to 10311fdd08beSQuentin Monnet * checksum, it can be used when pushing new data. 10321fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* == 0 and *seed* set to 10331fdd08beSQuentin Monnet * checksum, it can be used when removing data from a packet. 10341fdd08beSQuentin Monnet * * With *from_size* > 0, *to_size* > 0 and *seed* set to 0, it 10351fdd08beSQuentin Monnet * can be used to compute a diff. Note that *from_size* and 10361fdd08beSQuentin Monnet * *to_size* do not need to be equal. 10371fdd08beSQuentin Monnet * 10381fdd08beSQuentin Monnet * This helper can be used in combination with 10391fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ (), to 10401fdd08beSQuentin Monnet * which one can feed in the difference computed with 10411fdd08beSQuentin Monnet * **bpf_csum_diff**\ (). 10421fdd08beSQuentin Monnet * Return 10431fdd08beSQuentin Monnet * The checksum result, or a negative error code in case of 10441fdd08beSQuentin Monnet * failure. 10451fdd08beSQuentin Monnet * 10461fdd08beSQuentin Monnet * int bpf_skb_get_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 10471fdd08beSQuentin Monnet * Description 10481fdd08beSQuentin Monnet * Retrieve tunnel options metadata for the packet associated to 10491fdd08beSQuentin Monnet * *skb*, and store the raw tunnel option data to the buffer *opt* 10501fdd08beSQuentin Monnet * of *size*. 10511fdd08beSQuentin Monnet * 10521fdd08beSQuentin Monnet * This helper can be used with encapsulation devices that can 10531fdd08beSQuentin Monnet * operate in "collect metadata" mode (please refer to the related 10541fdd08beSQuentin Monnet * note in the description of **bpf_skb_get_tunnel_key**\ () for 10551fdd08beSQuentin Monnet * more details). A particular example where this can be used is 10561fdd08beSQuentin Monnet * in combination with the Geneve encapsulation protocol, where it 10571fdd08beSQuentin Monnet * allows for pushing (with **bpf_skb_get_tunnel_opt**\ () helper) 10581fdd08beSQuentin Monnet * and retrieving arbitrary TLVs (Type-Length-Value headers) from 10591fdd08beSQuentin Monnet * the eBPF program. This allows for full customization of these 10601fdd08beSQuentin Monnet * headers. 10611fdd08beSQuentin Monnet * Return 10621fdd08beSQuentin Monnet * The size of the option data retrieved. 10631fdd08beSQuentin Monnet * 10641fdd08beSQuentin Monnet * int bpf_skb_set_tunnel_opt(struct sk_buff *skb, u8 *opt, u32 size) 10651fdd08beSQuentin Monnet * Description 10661fdd08beSQuentin Monnet * Set tunnel options metadata for the packet associated to *skb* 10671fdd08beSQuentin Monnet * to the option data contained in the raw buffer *opt* of *size*. 10681fdd08beSQuentin Monnet * 10691fdd08beSQuentin Monnet * See also the description of the **bpf_skb_get_tunnel_opt**\ () 10701fdd08beSQuentin Monnet * helper for additional information. 10711fdd08beSQuentin Monnet * Return 10721fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 10731fdd08beSQuentin Monnet * 10741fdd08beSQuentin Monnet * int bpf_skb_change_proto(struct sk_buff *skb, __be16 proto, u64 flags) 10751fdd08beSQuentin Monnet * Description 10761fdd08beSQuentin Monnet * Change the protocol of the *skb* to *proto*. Currently 10771fdd08beSQuentin Monnet * supported are transition from IPv4 to IPv6, and from IPv6 to 10781fdd08beSQuentin Monnet * IPv4. The helper takes care of the groundwork for the 10791fdd08beSQuentin Monnet * transition, including resizing the socket buffer. The eBPF 10801fdd08beSQuentin Monnet * program is expected to fill the new headers, if any, via 10811fdd08beSQuentin Monnet * **skb_store_bytes**\ () and to recompute the checksums with 10821fdd08beSQuentin Monnet * **bpf_l3_csum_replace**\ () and **bpf_l4_csum_replace**\ 10831fdd08beSQuentin Monnet * (). The main case for this helper is to perform NAT64 10841fdd08beSQuentin Monnet * operations out of an eBPF program. 10851fdd08beSQuentin Monnet * 10861fdd08beSQuentin Monnet * Internally, the GSO type is marked as dodgy so that headers are 10871fdd08beSQuentin Monnet * checked and segments are recalculated by the GSO/GRO engine. 10881fdd08beSQuentin Monnet * The size for GSO target is adapted as well. 10891fdd08beSQuentin Monnet * 10901fdd08beSQuentin Monnet * All values for *flags* are reserved for future usage, and must 10911fdd08beSQuentin Monnet * be left at zero. 10921fdd08beSQuentin Monnet * 10931fdd08beSQuentin Monnet * A call to this helper is susceptible to change the underlaying 10941fdd08beSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 10951fdd08beSQuentin Monnet * previously done by the verifier are invalidated and must be 10961fdd08beSQuentin Monnet * performed again, if the helper is used in combination with 10971fdd08beSQuentin Monnet * direct packet access. 10981fdd08beSQuentin Monnet * Return 10991fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 11001fdd08beSQuentin Monnet * 11011fdd08beSQuentin Monnet * int bpf_skb_change_type(struct sk_buff *skb, u32 type) 11021fdd08beSQuentin Monnet * Description 11031fdd08beSQuentin Monnet * Change the packet type for the packet associated to *skb*. This 11041fdd08beSQuentin Monnet * comes down to setting *skb*\ **->pkt_type** to *type*, except 11051fdd08beSQuentin Monnet * the eBPF program does not have a write access to *skb*\ 11061fdd08beSQuentin Monnet * **->pkt_type** beside this helper. Using a helper here allows 11071fdd08beSQuentin Monnet * for graceful handling of errors. 11081fdd08beSQuentin Monnet * 11091fdd08beSQuentin Monnet * The major use case is to change incoming *skb*s to 11101fdd08beSQuentin Monnet * **PACKET_HOST** in a programmatic way instead of having to 11111fdd08beSQuentin Monnet * recirculate via **redirect**\ (..., **BPF_F_INGRESS**), for 11121fdd08beSQuentin Monnet * example. 11131fdd08beSQuentin Monnet * 11141fdd08beSQuentin Monnet * Note that *type* only allows certain values. At this time, they 11151fdd08beSQuentin Monnet * are: 11161fdd08beSQuentin Monnet * 11171fdd08beSQuentin Monnet * **PACKET_HOST** 11181fdd08beSQuentin Monnet * Packet is for us. 11191fdd08beSQuentin Monnet * **PACKET_BROADCAST** 11201fdd08beSQuentin Monnet * Send packet to all. 11211fdd08beSQuentin Monnet * **PACKET_MULTICAST** 11221fdd08beSQuentin Monnet * Send packet to group. 11231fdd08beSQuentin Monnet * **PACKET_OTHERHOST** 11241fdd08beSQuentin Monnet * Send packet to someone else. 11251fdd08beSQuentin Monnet * Return 11261fdd08beSQuentin Monnet * 0 on success, or a negative error in case of failure. 11271fdd08beSQuentin Monnet * 1128c6b5fb86SQuentin Monnet * int bpf_skb_under_cgroup(struct sk_buff *skb, struct bpf_map *map, u32 index) 1129c6b5fb86SQuentin Monnet * Description 1130c6b5fb86SQuentin Monnet * Check whether *skb* is a descendant of the cgroup2 held by 1131c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1132c6b5fb86SQuentin Monnet * Return 1133c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 1134c6b5fb86SQuentin Monnet * 1135c6b5fb86SQuentin Monnet * * 0, if the *skb* failed the cgroup2 descendant test. 1136c6b5fb86SQuentin Monnet * * 1, if the *skb* succeeded the cgroup2 descendant test. 1137c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 1138c6b5fb86SQuentin Monnet * 1139fa15601aSQuentin Monnet * u32 bpf_get_hash_recalc(struct sk_buff *skb) 1140fa15601aSQuentin Monnet * Description 1141fa15601aSQuentin Monnet * Retrieve the hash of the packet, *skb*\ **->hash**. If it is 1142fa15601aSQuentin Monnet * not set, in particular if the hash was cleared due to mangling, 1143fa15601aSQuentin Monnet * recompute this hash. Later accesses to the hash can be done 1144fa15601aSQuentin Monnet * directly with *skb*\ **->hash**. 1145fa15601aSQuentin Monnet * 1146fa15601aSQuentin Monnet * Calling **bpf_set_hash_invalid**\ (), changing a packet 1147fa15601aSQuentin Monnet * prototype with **bpf_skb_change_proto**\ (), or calling 1148fa15601aSQuentin Monnet * **bpf_skb_store_bytes**\ () with the 1149fa15601aSQuentin Monnet * **BPF_F_INVALIDATE_HASH** are actions susceptible to clear 1150fa15601aSQuentin Monnet * the hash and to trigger a new computation for the next call to 1151fa15601aSQuentin Monnet * **bpf_get_hash_recalc**\ (). 1152fa15601aSQuentin Monnet * Return 1153fa15601aSQuentin Monnet * The 32-bit hash. 1154fa15601aSQuentin Monnet * 1155c456dec4SQuentin Monnet * u64 bpf_get_current_task(void) 1156c456dec4SQuentin Monnet * Return 1157c456dec4SQuentin Monnet * A pointer to the current task struct. 1158fa15601aSQuentin Monnet * 1159c6b5fb86SQuentin Monnet * int bpf_probe_write_user(void *dst, const void *src, u32 len) 1160c6b5fb86SQuentin Monnet * Description 1161c6b5fb86SQuentin Monnet * Attempt in a safe way to write *len* bytes from the buffer 1162c6b5fb86SQuentin Monnet * *src* to *dst* in memory. It only works for threads that are in 1163c6b5fb86SQuentin Monnet * user context, and *dst* must be a valid user space address. 1164c6b5fb86SQuentin Monnet * 1165c6b5fb86SQuentin Monnet * This helper should not be used to implement any kind of 1166c6b5fb86SQuentin Monnet * security mechanism because of TOC-TOU attacks, but rather to 1167c6b5fb86SQuentin Monnet * debug, divert, and manipulate execution of semi-cooperative 1168c6b5fb86SQuentin Monnet * processes. 1169c6b5fb86SQuentin Monnet * 1170c6b5fb86SQuentin Monnet * Keep in mind that this feature is meant for experiments, and it 1171c6b5fb86SQuentin Monnet * has a risk of crashing the system and running programs. 1172c6b5fb86SQuentin Monnet * Therefore, when an eBPF program using this helper is attached, 1173c6b5fb86SQuentin Monnet * a warning including PID and process name is printed to kernel 1174c6b5fb86SQuentin Monnet * logs. 1175c6b5fb86SQuentin Monnet * Return 1176c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1177c6b5fb86SQuentin Monnet * 1178c6b5fb86SQuentin Monnet * int bpf_current_task_under_cgroup(struct bpf_map *map, u32 index) 1179c6b5fb86SQuentin Monnet * Description 1180c6b5fb86SQuentin Monnet * Check whether the probe is being run is the context of a given 1181c6b5fb86SQuentin Monnet * subset of the cgroup2 hierarchy. The cgroup2 to test is held by 1182c6b5fb86SQuentin Monnet * *map* of type **BPF_MAP_TYPE_CGROUP_ARRAY**, at *index*. 1183c6b5fb86SQuentin Monnet * Return 1184c6b5fb86SQuentin Monnet * The return value depends on the result of the test, and can be: 1185c6b5fb86SQuentin Monnet * 1186c6b5fb86SQuentin Monnet * * 0, if the *skb* task belongs to the cgroup2. 1187c6b5fb86SQuentin Monnet * * 1, if the *skb* task does not belong to the cgroup2. 1188c6b5fb86SQuentin Monnet * * A negative error code, if an error occurred. 1189c6b5fb86SQuentin Monnet * 1190fa15601aSQuentin Monnet * int bpf_skb_change_tail(struct sk_buff *skb, u32 len, u64 flags) 1191fa15601aSQuentin Monnet * Description 1192fa15601aSQuentin Monnet * Resize (trim or grow) the packet associated to *skb* to the 1193fa15601aSQuentin Monnet * new *len*. The *flags* are reserved for future usage, and must 1194fa15601aSQuentin Monnet * be left at zero. 1195fa15601aSQuentin Monnet * 1196fa15601aSQuentin Monnet * The basic idea is that the helper performs the needed work to 1197fa15601aSQuentin Monnet * change the size of the packet, then the eBPF program rewrites 1198fa15601aSQuentin Monnet * the rest via helpers like **bpf_skb_store_bytes**\ (), 1199fa15601aSQuentin Monnet * **bpf_l3_csum_replace**\ (), **bpf_l3_csum_replace**\ () 1200fa15601aSQuentin Monnet * and others. This helper is a slow path utility intended for 1201fa15601aSQuentin Monnet * replies with control messages. And because it is targeted for 1202fa15601aSQuentin Monnet * slow path, the helper itself can afford to be slow: it 1203fa15601aSQuentin Monnet * implicitly linearizes, unclones and drops offloads from the 1204fa15601aSQuentin Monnet * *skb*. 1205fa15601aSQuentin Monnet * 1206fa15601aSQuentin Monnet * A call to this helper is susceptible to change the underlaying 1207fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1208fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1209fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1210fa15601aSQuentin Monnet * direct packet access. 1211fa15601aSQuentin Monnet * Return 1212fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1213fa15601aSQuentin Monnet * 1214fa15601aSQuentin Monnet * int bpf_skb_pull_data(struct sk_buff *skb, u32 len) 1215fa15601aSQuentin Monnet * Description 1216fa15601aSQuentin Monnet * Pull in non-linear data in case the *skb* is non-linear and not 1217fa15601aSQuentin Monnet * all of *len* are part of the linear section. Make *len* bytes 1218fa15601aSQuentin Monnet * from *skb* readable and writable. If a zero value is passed for 1219fa15601aSQuentin Monnet * *len*, then the whole length of the *skb* is pulled. 1220fa15601aSQuentin Monnet * 1221fa15601aSQuentin Monnet * This helper is only needed for reading and writing with direct 1222fa15601aSQuentin Monnet * packet access. 1223fa15601aSQuentin Monnet * 1224fa15601aSQuentin Monnet * For direct packet access, testing that offsets to access 1225fa15601aSQuentin Monnet * are within packet boundaries (test on *skb*\ **->data_end**) is 1226fa15601aSQuentin Monnet * susceptible to fail if offsets are invalid, or if the requested 1227fa15601aSQuentin Monnet * data is in non-linear parts of the *skb*. On failure the 1228fa15601aSQuentin Monnet * program can just bail out, or in the case of a non-linear 1229fa15601aSQuentin Monnet * buffer, use a helper to make the data available. The 1230fa15601aSQuentin Monnet * **bpf_skb_load_bytes**\ () helper is a first solution to access 1231fa15601aSQuentin Monnet * the data. Another one consists in using **bpf_skb_pull_data** 1232fa15601aSQuentin Monnet * to pull in once the non-linear parts, then retesting and 1233fa15601aSQuentin Monnet * eventually access the data. 1234fa15601aSQuentin Monnet * 1235fa15601aSQuentin Monnet * At the same time, this also makes sure the *skb* is uncloned, 1236fa15601aSQuentin Monnet * which is a necessary condition for direct write. As this needs 1237fa15601aSQuentin Monnet * to be an invariant for the write part only, the verifier 1238fa15601aSQuentin Monnet * detects writes and adds a prologue that is calling 1239fa15601aSQuentin Monnet * **bpf_skb_pull_data()** to effectively unclone the *skb* from 1240fa15601aSQuentin Monnet * the very beginning in case it is indeed cloned. 1241fa15601aSQuentin Monnet * 1242fa15601aSQuentin Monnet * A call to this helper is susceptible to change the underlaying 1243fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1244fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1245fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1246fa15601aSQuentin Monnet * direct packet access. 1247fa15601aSQuentin Monnet * Return 1248fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1249fa15601aSQuentin Monnet * 1250fa15601aSQuentin Monnet * s64 bpf_csum_update(struct sk_buff *skb, __wsum csum) 1251fa15601aSQuentin Monnet * Description 1252fa15601aSQuentin Monnet * Add the checksum *csum* into *skb*\ **->csum** in case the 1253fa15601aSQuentin Monnet * driver has supplied a checksum for the entire packet into that 1254fa15601aSQuentin Monnet * field. Return an error otherwise. This helper is intended to be 1255fa15601aSQuentin Monnet * used in combination with **bpf_csum_diff**\ (), in particular 1256fa15601aSQuentin Monnet * when the checksum needs to be updated after data has been 1257fa15601aSQuentin Monnet * written into the packet through direct packet access. 1258fa15601aSQuentin Monnet * Return 1259fa15601aSQuentin Monnet * The checksum on success, or a negative error code in case of 1260fa15601aSQuentin Monnet * failure. 1261fa15601aSQuentin Monnet * 1262fa15601aSQuentin Monnet * void bpf_set_hash_invalid(struct sk_buff *skb) 1263fa15601aSQuentin Monnet * Description 1264fa15601aSQuentin Monnet * Invalidate the current *skb*\ **->hash**. It can be used after 1265fa15601aSQuentin Monnet * mangling on headers through direct packet access, in order to 1266fa15601aSQuentin Monnet * indicate that the hash is outdated and to trigger a 1267fa15601aSQuentin Monnet * recalculation the next time the kernel tries to access this 1268fa15601aSQuentin Monnet * hash or when the **bpf_get_hash_recalc**\ () helper is called. 1269fa15601aSQuentin Monnet * 1270fa15601aSQuentin Monnet * int bpf_get_numa_node_id(void) 1271fa15601aSQuentin Monnet * Description 1272fa15601aSQuentin Monnet * Return the id of the current NUMA node. The primary use case 1273fa15601aSQuentin Monnet * for this helper is the selection of sockets for the local NUMA 1274fa15601aSQuentin Monnet * node, when the program is attached to sockets using the 1275fa15601aSQuentin Monnet * **SO_ATTACH_REUSEPORT_EBPF** option (see also **socket(7)**), 1276fa15601aSQuentin Monnet * but the helper is also available to other eBPF program types, 1277fa15601aSQuentin Monnet * similarly to **bpf_get_smp_processor_id**\ (). 1278fa15601aSQuentin Monnet * Return 1279fa15601aSQuentin Monnet * The id of current NUMA node. 1280fa15601aSQuentin Monnet * 1281c6b5fb86SQuentin Monnet * int bpf_skb_change_head(struct sk_buff *skb, u32 len, u64 flags) 1282c6b5fb86SQuentin Monnet * Description 1283c6b5fb86SQuentin Monnet * Grows headroom of packet associated to *skb* and adjusts the 1284c6b5fb86SQuentin Monnet * offset of the MAC header accordingly, adding *len* bytes of 1285c6b5fb86SQuentin Monnet * space. It automatically extends and reallocates memory as 1286c6b5fb86SQuentin Monnet * required. 1287c6b5fb86SQuentin Monnet * 1288c6b5fb86SQuentin Monnet * This helper can be used on a layer 3 *skb* to push a MAC header 1289c6b5fb86SQuentin Monnet * for redirection into a layer 2 device. 1290c6b5fb86SQuentin Monnet * 1291c6b5fb86SQuentin Monnet * All values for *flags* are reserved for future usage, and must 1292c6b5fb86SQuentin Monnet * be left at zero. 1293c6b5fb86SQuentin Monnet * 1294c6b5fb86SQuentin Monnet * A call to this helper is susceptible to change the underlaying 1295c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1296c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 1297c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 1298c6b5fb86SQuentin Monnet * direct packet access. 1299c6b5fb86SQuentin Monnet * Return 1300c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1301c6b5fb86SQuentin Monnet * 1302c6b5fb86SQuentin Monnet * int bpf_xdp_adjust_head(struct xdp_buff *xdp_md, int delta) 1303c6b5fb86SQuentin Monnet * Description 1304c6b5fb86SQuentin Monnet * Adjust (move) *xdp_md*\ **->data** by *delta* bytes. Note that 1305c6b5fb86SQuentin Monnet * it is possible to use a negative value for *delta*. This helper 1306c6b5fb86SQuentin Monnet * can be used to prepare the packet for pushing or popping 1307c6b5fb86SQuentin Monnet * headers. 1308c6b5fb86SQuentin Monnet * 1309c6b5fb86SQuentin Monnet * A call to this helper is susceptible to change the underlaying 1310c6b5fb86SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1311c6b5fb86SQuentin Monnet * previously done by the verifier are invalidated and must be 1312c6b5fb86SQuentin Monnet * performed again, if the helper is used in combination with 1313c6b5fb86SQuentin Monnet * direct packet access. 1314c6b5fb86SQuentin Monnet * Return 1315c6b5fb86SQuentin Monnet * 0 on success, or a negative error in case of failure. 1316c6b5fb86SQuentin Monnet * 1317c6b5fb86SQuentin Monnet * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) 1318c6b5fb86SQuentin Monnet * Description 1319c6b5fb86SQuentin Monnet * Copy a NUL terminated string from an unsafe address 1320c6b5fb86SQuentin Monnet * *unsafe_ptr* to *dst*. The *size* should include the 1321c6b5fb86SQuentin Monnet * terminating NUL byte. In case the string length is smaller than 1322c6b5fb86SQuentin Monnet * *size*, the target is not padded with further NUL bytes. If the 1323c6b5fb86SQuentin Monnet * string length is larger than *size*, just *size*-1 bytes are 1324c6b5fb86SQuentin Monnet * copied and the last byte is set to NUL. 1325c6b5fb86SQuentin Monnet * 1326c6b5fb86SQuentin Monnet * On success, the length of the copied string is returned. This 1327c6b5fb86SQuentin Monnet * makes this helper useful in tracing programs for reading 1328c6b5fb86SQuentin Monnet * strings, and more importantly to get its length at runtime. See 1329c6b5fb86SQuentin Monnet * the following snippet: 1330c6b5fb86SQuentin Monnet * 1331c6b5fb86SQuentin Monnet * :: 1332c6b5fb86SQuentin Monnet * 1333c6b5fb86SQuentin Monnet * SEC("kprobe/sys_open") 1334c6b5fb86SQuentin Monnet * void bpf_sys_open(struct pt_regs *ctx) 1335c6b5fb86SQuentin Monnet * { 1336c6b5fb86SQuentin Monnet * char buf[PATHLEN]; // PATHLEN is defined to 256 1337c6b5fb86SQuentin Monnet * int res = bpf_probe_read_str(buf, sizeof(buf), 1338c6b5fb86SQuentin Monnet * ctx->di); 1339c6b5fb86SQuentin Monnet * 1340c6b5fb86SQuentin Monnet * // Consume buf, for example push it to 1341c6b5fb86SQuentin Monnet * // userspace via bpf_perf_event_output(); we 1342c6b5fb86SQuentin Monnet * // can use res (the string length) as event 1343c6b5fb86SQuentin Monnet * // size, after checking its boundaries. 1344c6b5fb86SQuentin Monnet * } 1345c6b5fb86SQuentin Monnet * 1346c6b5fb86SQuentin Monnet * In comparison, using **bpf_probe_read()** helper here instead 1347c6b5fb86SQuentin Monnet * to read the string would require to estimate the length at 1348c6b5fb86SQuentin Monnet * compile time, and would often result in copying more memory 1349c6b5fb86SQuentin Monnet * than necessary. 1350c6b5fb86SQuentin Monnet * 1351c6b5fb86SQuentin Monnet * Another useful use case is when parsing individual process 1352c6b5fb86SQuentin Monnet * arguments or individual environment variables navigating 1353c6b5fb86SQuentin Monnet * *current*\ **->mm->arg_start** and *current*\ 1354c6b5fb86SQuentin Monnet * **->mm->env_start**: using this helper and the return value, 1355c6b5fb86SQuentin Monnet * one can quickly iterate at the right offset of the memory area. 1356c6b5fb86SQuentin Monnet * Return 1357c6b5fb86SQuentin Monnet * On success, the strictly positive length of the string, 1358c6b5fb86SQuentin Monnet * including the trailing NUL character. On error, a negative 1359c6b5fb86SQuentin Monnet * value. 1360c6b5fb86SQuentin Monnet * 1361c6b5fb86SQuentin Monnet * u64 bpf_get_socket_cookie(struct sk_buff *skb) 1362c6b5fb86SQuentin Monnet * Description 1363c6b5fb86SQuentin Monnet * If the **struct sk_buff** pointed by *skb* has a known socket, 1364c6b5fb86SQuentin Monnet * retrieve the cookie (generated by the kernel) of this socket. 1365c6b5fb86SQuentin Monnet * If no cookie has been set yet, generate a new cookie. Once 1366c6b5fb86SQuentin Monnet * generated, the socket cookie remains stable for the life of the 1367c6b5fb86SQuentin Monnet * socket. This helper can be useful for monitoring per socket 1368c6b5fb86SQuentin Monnet * networking traffic statistics as it provides a unique socket 1369c6b5fb86SQuentin Monnet * identifier per namespace. 1370c6b5fb86SQuentin Monnet * Return 1371c6b5fb86SQuentin Monnet * A 8-byte long non-decreasing number on success, or 0 if the 1372c6b5fb86SQuentin Monnet * socket field is missing inside *skb*. 1373c6b5fb86SQuentin Monnet * 1374c6b5fb86SQuentin Monnet * u32 bpf_get_socket_uid(struct sk_buff *skb) 1375c6b5fb86SQuentin Monnet * Return 1376c6b5fb86SQuentin Monnet * The owner UID of the socket associated to *skb*. If the socket 1377c6b5fb86SQuentin Monnet * is **NULL**, or if it is not a full socket (i.e. if it is a 1378c6b5fb86SQuentin Monnet * time-wait or a request socket instead), **overflowuid** value 1379c6b5fb86SQuentin Monnet * is returned (note that **overflowuid** might also be the actual 1380c6b5fb86SQuentin Monnet * UID value for the socket). 1381c6b5fb86SQuentin Monnet * 1382fa15601aSQuentin Monnet * u32 bpf_set_hash(struct sk_buff *skb, u32 hash) 1383fa15601aSQuentin Monnet * Description 1384fa15601aSQuentin Monnet * Set the full hash for *skb* (set the field *skb*\ **->hash**) 1385fa15601aSQuentin Monnet * to value *hash*. 1386fa15601aSQuentin Monnet * Return 1387fa15601aSQuentin Monnet * 0 1388fa15601aSQuentin Monnet * 1389a3ef8e9aSAndrey Ignatov * int bpf_setsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 13907aa79a86SQuentin Monnet * Description 13917aa79a86SQuentin Monnet * Emulate a call to **setsockopt()** on the socket associated to 13927aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 13937aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 13947aa79a86SQuentin Monnet * must be specified, see **setsockopt(2)** for more information. 13957aa79a86SQuentin Monnet * The option value of length *optlen* is pointed by *optval*. 13967aa79a86SQuentin Monnet * 13977aa79a86SQuentin Monnet * This helper actually implements a subset of **setsockopt()**. 13987aa79a86SQuentin Monnet * It supports the following *level*\ s: 13997aa79a86SQuentin Monnet * 14007aa79a86SQuentin Monnet * * **SOL_SOCKET**, which supports the following *optname*\ s: 14017aa79a86SQuentin Monnet * **SO_RCVBUF**, **SO_SNDBUF**, **SO_MAX_PACING_RATE**, 14027aa79a86SQuentin Monnet * **SO_PRIORITY**, **SO_RCVLOWAT**, **SO_MARK**. 14037aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports the following *optname*\ s: 14047aa79a86SQuentin Monnet * **TCP_CONGESTION**, **TCP_BPF_IW**, 14057aa79a86SQuentin Monnet * **TCP_BPF_SNDCWND_CLAMP**. 14067aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 14077aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 14087aa79a86SQuentin Monnet * Return 14097aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 14107aa79a86SQuentin Monnet * 1411fa15601aSQuentin Monnet * int bpf_skb_adjust_room(struct sk_buff *skb, u32 len_diff, u32 mode, u64 flags) 1412fa15601aSQuentin Monnet * Description 1413fa15601aSQuentin Monnet * Grow or shrink the room for data in the packet associated to 1414fa15601aSQuentin Monnet * *skb* by *len_diff*, and according to the selected *mode*. 1415fa15601aSQuentin Monnet * 1416fa15601aSQuentin Monnet * There is a single supported mode at this time: 1417fa15601aSQuentin Monnet * 1418fa15601aSQuentin Monnet * * **BPF_ADJ_ROOM_NET**: Adjust room at the network layer 1419fa15601aSQuentin Monnet * (room space is added or removed below the layer 3 header). 1420fa15601aSQuentin Monnet * 1421fa15601aSQuentin Monnet * All values for *flags* are reserved for future usage, and must 1422fa15601aSQuentin Monnet * be left at zero. 1423fa15601aSQuentin Monnet * 1424fa15601aSQuentin Monnet * A call to this helper is susceptible to change the underlaying 1425fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1426fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1427fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1428fa15601aSQuentin Monnet * direct packet access. 1429fa15601aSQuentin Monnet * Return 1430fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 1431fa15601aSQuentin Monnet * 1432ab127040SQuentin Monnet * int bpf_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1433ab127040SQuentin Monnet * Description 1434ab127040SQuentin Monnet * Redirect the packet to the endpoint referenced by *map* at 1435ab127040SQuentin Monnet * index *key*. Depending on its type, this *map* can contain 1436ab127040SQuentin Monnet * references to net devices (for forwarding packets through other 1437ab127040SQuentin Monnet * ports), or to CPUs (for redirecting XDP frames to another CPU; 1438ab127040SQuentin Monnet * but this is only implemented for native XDP (with driver 1439ab127040SQuentin Monnet * support) as of this writing). 1440ab127040SQuentin Monnet * 1441ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 1442ab127040SQuentin Monnet * be left at zero. 1443ab127040SQuentin Monnet * 1444ab127040SQuentin Monnet * When used to redirect packets to net devices, this helper 1445ab127040SQuentin Monnet * provides a high performance increase over **bpf_redirect**\ (). 1446ab127040SQuentin Monnet * This is due to various implementation details of the underlying 1447ab127040SQuentin Monnet * mechanisms, one of which is the fact that **bpf_redirect_map**\ 1448ab127040SQuentin Monnet * () tries to send packet as a "bulk" to the device. 1449ab127040SQuentin Monnet * Return 1450ab127040SQuentin Monnet * **XDP_REDIRECT** on success, or **XDP_ABORTED** on error. 1451ab127040SQuentin Monnet * 1452ab127040SQuentin Monnet * int bpf_sk_redirect_map(struct bpf_map *map, u32 key, u64 flags) 1453ab127040SQuentin Monnet * Description 1454ab127040SQuentin Monnet * Redirect the packet to the socket referenced by *map* (of type 1455ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1456ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 1457ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 1458ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 1459ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 1460ab127040SQuentin Monnet * Return 1461ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 1462ab127040SQuentin Monnet * 1463a3ef8e9aSAndrey Ignatov * int bpf_sock_map_update(struct bpf_sock_ops *skops, struct bpf_map *map, void *key, u64 flags) 1464ab127040SQuentin Monnet * Description 1465ab127040SQuentin Monnet * Add an entry to, or update a *map* referencing sockets. The 1466ab127040SQuentin Monnet * *skops* is used as a new value for the entry associated to 1467ab127040SQuentin Monnet * *key*. *flags* is one of: 1468ab127040SQuentin Monnet * 1469ab127040SQuentin Monnet * **BPF_NOEXIST** 1470ab127040SQuentin Monnet * The entry for *key* must not exist in the map. 1471ab127040SQuentin Monnet * **BPF_EXIST** 1472ab127040SQuentin Monnet * The entry for *key* must already exist in the map. 1473ab127040SQuentin Monnet * **BPF_ANY** 1474ab127040SQuentin Monnet * No condition on the existence of the entry for *key*. 1475ab127040SQuentin Monnet * 1476ab127040SQuentin Monnet * If the *map* has eBPF programs (parser and verdict), those will 1477ab127040SQuentin Monnet * be inherited by the socket being added. If the socket is 1478ab127040SQuentin Monnet * already attached to eBPF programs, this results in an error. 1479ab127040SQuentin Monnet * Return 1480ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 1481ab127040SQuentin Monnet * 1482fa15601aSQuentin Monnet * int bpf_xdp_adjust_meta(struct xdp_buff *xdp_md, int delta) 1483fa15601aSQuentin Monnet * Description 1484fa15601aSQuentin Monnet * Adjust the address pointed by *xdp_md*\ **->data_meta** by 1485fa15601aSQuentin Monnet * *delta* (which can be positive or negative). Note that this 1486fa15601aSQuentin Monnet * operation modifies the address stored in *xdp_md*\ **->data**, 1487fa15601aSQuentin Monnet * so the latter must be loaded only after the helper has been 1488fa15601aSQuentin Monnet * called. 1489fa15601aSQuentin Monnet * 1490fa15601aSQuentin Monnet * The use of *xdp_md*\ **->data_meta** is optional and programs 1491fa15601aSQuentin Monnet * are not required to use it. The rationale is that when the 1492fa15601aSQuentin Monnet * packet is processed with XDP (e.g. as DoS filter), it is 1493fa15601aSQuentin Monnet * possible to push further meta data along with it before passing 1494fa15601aSQuentin Monnet * to the stack, and to give the guarantee that an ingress eBPF 1495fa15601aSQuentin Monnet * program attached as a TC classifier on the same device can pick 1496fa15601aSQuentin Monnet * this up for further post-processing. Since TC works with socket 1497fa15601aSQuentin Monnet * buffers, it remains possible to set from XDP the **mark** or 1498fa15601aSQuentin Monnet * **priority** pointers, or other pointers for the socket buffer. 1499fa15601aSQuentin Monnet * Having this scratch space generic and programmable allows for 1500fa15601aSQuentin Monnet * more flexibility as the user is free to store whatever meta 1501fa15601aSQuentin Monnet * data they need. 1502fa15601aSQuentin Monnet * 1503fa15601aSQuentin Monnet * A call to this helper is susceptible to change the underlaying 1504fa15601aSQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1505fa15601aSQuentin Monnet * previously done by the verifier are invalidated and must be 1506fa15601aSQuentin Monnet * performed again, if the helper is used in combination with 1507fa15601aSQuentin Monnet * direct packet access. 1508fa15601aSQuentin Monnet * Return 1509fa15601aSQuentin Monnet * 0 on success, or a negative error in case of failure. 15107aa79a86SQuentin Monnet * 15117aa79a86SQuentin Monnet * int bpf_perf_event_read_value(struct bpf_map *map, u64 flags, struct bpf_perf_event_value *buf, u32 buf_size) 15127aa79a86SQuentin Monnet * Description 15137aa79a86SQuentin Monnet * Read the value of a perf event counter, and store it into *buf* 15147aa79a86SQuentin Monnet * of size *buf_size*. This helper relies on a *map* of type 15157aa79a86SQuentin Monnet * **BPF_MAP_TYPE_PERF_EVENT_ARRAY**. The nature of the perf event 15167aa79a86SQuentin Monnet * counter is selected when *map* is updated with perf event file 15177aa79a86SQuentin Monnet * descriptors. The *map* is an array whose size is the number of 15187aa79a86SQuentin Monnet * available CPUs, and each cell contains a value relative to one 15197aa79a86SQuentin Monnet * CPU. The value to retrieve is indicated by *flags*, that 15207aa79a86SQuentin Monnet * contains the index of the CPU to look up, masked with 15217aa79a86SQuentin Monnet * **BPF_F_INDEX_MASK**. Alternatively, *flags* can be set to 15227aa79a86SQuentin Monnet * **BPF_F_CURRENT_CPU** to indicate that the value for the 15237aa79a86SQuentin Monnet * current CPU should be retrieved. 15247aa79a86SQuentin Monnet * 15257aa79a86SQuentin Monnet * This helper behaves in a way close to 15267aa79a86SQuentin Monnet * **bpf_perf_event_read**\ () helper, save that instead of 15277aa79a86SQuentin Monnet * just returning the value observed, it fills the *buf* 15287aa79a86SQuentin Monnet * structure. This allows for additional data to be retrieved: in 15297aa79a86SQuentin Monnet * particular, the enabled and running times (in *buf*\ 15307aa79a86SQuentin Monnet * **->enabled** and *buf*\ **->running**, respectively) are 15317aa79a86SQuentin Monnet * copied. In general, **bpf_perf_event_read_value**\ () is 15327aa79a86SQuentin Monnet * recommended over **bpf_perf_event_read**\ (), which has some 15337aa79a86SQuentin Monnet * ABI issues and provides fewer functionalities. 15347aa79a86SQuentin Monnet * 15357aa79a86SQuentin Monnet * These values are interesting, because hardware PMU (Performance 15367aa79a86SQuentin Monnet * Monitoring Unit) counters are limited resources. When there are 15377aa79a86SQuentin Monnet * more PMU based perf events opened than available counters, 15387aa79a86SQuentin Monnet * kernel will multiplex these events so each event gets certain 15397aa79a86SQuentin Monnet * percentage (but not all) of the PMU time. In case that 15407aa79a86SQuentin Monnet * multiplexing happens, the number of samples or counter value 15417aa79a86SQuentin Monnet * will not reflect the case compared to when no multiplexing 15427aa79a86SQuentin Monnet * occurs. This makes comparison between different runs difficult. 15437aa79a86SQuentin Monnet * Typically, the counter value should be normalized before 15447aa79a86SQuentin Monnet * comparing to other experiments. The usual normalization is done 15457aa79a86SQuentin Monnet * as follows. 15467aa79a86SQuentin Monnet * 15477aa79a86SQuentin Monnet * :: 15487aa79a86SQuentin Monnet * 15497aa79a86SQuentin Monnet * normalized_counter = counter * t_enabled / t_running 15507aa79a86SQuentin Monnet * 15517aa79a86SQuentin Monnet * Where t_enabled is the time enabled for event and t_running is 15527aa79a86SQuentin Monnet * the time running for event since last normalization. The 15537aa79a86SQuentin Monnet * enabled and running times are accumulated since the perf event 15547aa79a86SQuentin Monnet * open. To achieve scaling factor between two invocations of an 15557aa79a86SQuentin Monnet * eBPF program, users can can use CPU id as the key (which is 15567aa79a86SQuentin Monnet * typical for perf array usage model) to remember the previous 15577aa79a86SQuentin Monnet * value and do the calculation inside the eBPF program. 15587aa79a86SQuentin Monnet * Return 15597aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 15607aa79a86SQuentin Monnet * 1561a3ef8e9aSAndrey Ignatov * int bpf_perf_prog_read_value(struct bpf_perf_event_data *ctx, struct bpf_perf_event_value *buf, u32 buf_size) 15627aa79a86SQuentin Monnet * Description 15637aa79a86SQuentin Monnet * For en eBPF program attached to a perf event, retrieve the 15647aa79a86SQuentin Monnet * value of the event counter associated to *ctx* and store it in 15657aa79a86SQuentin Monnet * the structure pointed by *buf* and of size *buf_size*. Enabled 15667aa79a86SQuentin Monnet * and running times are also stored in the structure (see 15677aa79a86SQuentin Monnet * description of helper **bpf_perf_event_read_value**\ () for 15687aa79a86SQuentin Monnet * more details). 15697aa79a86SQuentin Monnet * Return 15707aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 15717aa79a86SQuentin Monnet * 1572a3ef8e9aSAndrey Ignatov * int bpf_getsockopt(struct bpf_sock_ops *bpf_socket, int level, int optname, char *optval, int optlen) 15737aa79a86SQuentin Monnet * Description 15747aa79a86SQuentin Monnet * Emulate a call to **getsockopt()** on the socket associated to 15757aa79a86SQuentin Monnet * *bpf_socket*, which must be a full socket. The *level* at 15767aa79a86SQuentin Monnet * which the option resides and the name *optname* of the option 15777aa79a86SQuentin Monnet * must be specified, see **getsockopt(2)** for more information. 15787aa79a86SQuentin Monnet * The retrieved value is stored in the structure pointed by 15797aa79a86SQuentin Monnet * *opval* and of length *optlen*. 15807aa79a86SQuentin Monnet * 15817aa79a86SQuentin Monnet * This helper actually implements a subset of **getsockopt()**. 15827aa79a86SQuentin Monnet * It supports the following *level*\ s: 15837aa79a86SQuentin Monnet * 15847aa79a86SQuentin Monnet * * **IPPROTO_TCP**, which supports *optname* 15857aa79a86SQuentin Monnet * **TCP_CONGESTION**. 15867aa79a86SQuentin Monnet * * **IPPROTO_IP**, which supports *optname* **IP_TOS**. 15877aa79a86SQuentin Monnet * * **IPPROTO_IPV6**, which supports *optname* **IPV6_TCLASS**. 15887aa79a86SQuentin Monnet * Return 15897aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 15907aa79a86SQuentin Monnet * 15917aa79a86SQuentin Monnet * int bpf_override_return(struct pt_reg *regs, u64 rc) 15927aa79a86SQuentin Monnet * Description 15937aa79a86SQuentin Monnet * Used for error injection, this helper uses kprobes to override 15947aa79a86SQuentin Monnet * the return value of the probed function, and to set it to *rc*. 15957aa79a86SQuentin Monnet * The first argument is the context *regs* on which the kprobe 15967aa79a86SQuentin Monnet * works. 15977aa79a86SQuentin Monnet * 15987aa79a86SQuentin Monnet * This helper works by setting setting the PC (program counter) 15997aa79a86SQuentin Monnet * to an override function which is run in place of the original 16007aa79a86SQuentin Monnet * probed function. This means the probed function is not run at 16017aa79a86SQuentin Monnet * all. The replacement function just returns with the required 16027aa79a86SQuentin Monnet * value. 16037aa79a86SQuentin Monnet * 16047aa79a86SQuentin Monnet * This helper has security implications, and thus is subject to 16057aa79a86SQuentin Monnet * restrictions. It is only available if the kernel was compiled 16067aa79a86SQuentin Monnet * with the **CONFIG_BPF_KPROBE_OVERRIDE** configuration 16077aa79a86SQuentin Monnet * option, and in this case it only works on functions tagged with 16087aa79a86SQuentin Monnet * **ALLOW_ERROR_INJECTION** in the kernel code. 16097aa79a86SQuentin Monnet * 16107aa79a86SQuentin Monnet * Also, the helper is only available for the architectures having 16117aa79a86SQuentin Monnet * the CONFIG_FUNCTION_ERROR_INJECTION option. As of this writing, 16127aa79a86SQuentin Monnet * x86 architecture is the only one to support this feature. 16137aa79a86SQuentin Monnet * Return 16147aa79a86SQuentin Monnet * 0 16157aa79a86SQuentin Monnet * 1616a3ef8e9aSAndrey Ignatov * int bpf_sock_ops_cb_flags_set(struct bpf_sock_ops *bpf_sock, int argval) 16177aa79a86SQuentin Monnet * Description 16187aa79a86SQuentin Monnet * Attempt to set the value of the **bpf_sock_ops_cb_flags** field 16197aa79a86SQuentin Monnet * for the full TCP socket associated to *bpf_sock_ops* to 16207aa79a86SQuentin Monnet * *argval*. 16217aa79a86SQuentin Monnet * 16227aa79a86SQuentin Monnet * The primary use of this field is to determine if there should 16237aa79a86SQuentin Monnet * be calls to eBPF programs of type 16247aa79a86SQuentin Monnet * **BPF_PROG_TYPE_SOCK_OPS** at various points in the TCP 16257aa79a86SQuentin Monnet * code. A program of the same type can change its value, per 16267aa79a86SQuentin Monnet * connection and as necessary, when the connection is 16277aa79a86SQuentin Monnet * established. This field is directly accessible for reading, but 16287aa79a86SQuentin Monnet * this helper must be used for updates in order to return an 16297aa79a86SQuentin Monnet * error if an eBPF program tries to set a callback that is not 16307aa79a86SQuentin Monnet * supported in the current kernel. 16317aa79a86SQuentin Monnet * 16327aa79a86SQuentin Monnet * The supported callback values that *argval* can combine are: 16337aa79a86SQuentin Monnet * 16347aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RTO_CB_FLAG** (retransmission time out) 16357aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_RETRANS_CB_FLAG** (retransmission) 16367aa79a86SQuentin Monnet * * **BPF_SOCK_OPS_STATE_CB_FLAG** (TCP state change) 16377aa79a86SQuentin Monnet * 16387aa79a86SQuentin Monnet * Here are some examples of where one could call such eBPF 16397aa79a86SQuentin Monnet * program: 16407aa79a86SQuentin Monnet * 16417aa79a86SQuentin Monnet * * When RTO fires. 16427aa79a86SQuentin Monnet * * When a packet is retransmitted. 16437aa79a86SQuentin Monnet * * When the connection terminates. 16447aa79a86SQuentin Monnet * * When a packet is sent. 16457aa79a86SQuentin Monnet * * When a packet is received. 16467aa79a86SQuentin Monnet * Return 16477aa79a86SQuentin Monnet * Code **-EINVAL** if the socket is not a full TCP socket; 16487aa79a86SQuentin Monnet * otherwise, a positive number containing the bits that could not 16497aa79a86SQuentin Monnet * be set is returned (which comes down to 0 if all bits were set 16507aa79a86SQuentin Monnet * as required). 16517aa79a86SQuentin Monnet * 1652ab127040SQuentin Monnet * int bpf_msg_redirect_map(struct sk_msg_buff *msg, struct bpf_map *map, u32 key, u64 flags) 1653ab127040SQuentin Monnet * Description 1654ab127040SQuentin Monnet * This helper is used in programs implementing policies at the 1655ab127040SQuentin Monnet * socket level. If the message *msg* is allowed to pass (i.e. if 1656ab127040SQuentin Monnet * the verdict eBPF program returns **SK_PASS**), redirect it to 1657ab127040SQuentin Monnet * the socket referenced by *map* (of type 1658ab127040SQuentin Monnet * **BPF_MAP_TYPE_SOCKMAP**) at index *key*. Both ingress and 1659ab127040SQuentin Monnet * egress interfaces can be used for redirection. The 1660ab127040SQuentin Monnet * **BPF_F_INGRESS** value in *flags* is used to make the 1661ab127040SQuentin Monnet * distinction (ingress path is selected if the flag is present, 1662ab127040SQuentin Monnet * egress path otherwise). This is the only flag supported for now. 1663ab127040SQuentin Monnet * Return 1664ab127040SQuentin Monnet * **SK_PASS** on success, or **SK_DROP** on error. 1665ab127040SQuentin Monnet * 1666ab127040SQuentin Monnet * int bpf_msg_apply_bytes(struct sk_msg_buff *msg, u32 bytes) 1667ab127040SQuentin Monnet * Description 1668ab127040SQuentin Monnet * For socket policies, apply the verdict of the eBPF program to 1669ab127040SQuentin Monnet * the next *bytes* (number of bytes) of message *msg*. 1670ab127040SQuentin Monnet * 1671ab127040SQuentin Monnet * For example, this helper can be used in the following cases: 1672ab127040SQuentin Monnet * 1673ab127040SQuentin Monnet * * A single **sendmsg**\ () or **sendfile**\ () system call 1674ab127040SQuentin Monnet * contains multiple logical messages that the eBPF program is 1675ab127040SQuentin Monnet * supposed to read and for which it should apply a verdict. 1676ab127040SQuentin Monnet * * An eBPF program only cares to read the first *bytes* of a 1677ab127040SQuentin Monnet * *msg*. If the message has a large payload, then setting up 1678ab127040SQuentin Monnet * and calling the eBPF program repeatedly for all bytes, even 1679ab127040SQuentin Monnet * though the verdict is already known, would create unnecessary 1680ab127040SQuentin Monnet * overhead. 1681ab127040SQuentin Monnet * 1682ab127040SQuentin Monnet * When called from within an eBPF program, the helper sets a 1683ab127040SQuentin Monnet * counter internal to the BPF infrastructure, that is used to 1684ab127040SQuentin Monnet * apply the last verdict to the next *bytes*. If *bytes* is 1685ab127040SQuentin Monnet * smaller than the current data being processed from a 1686ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () system call, the first 1687ab127040SQuentin Monnet * *bytes* will be sent and the eBPF program will be re-run with 1688ab127040SQuentin Monnet * the pointer for start of data pointing to byte number *bytes* 1689ab127040SQuentin Monnet * **+ 1**. If *bytes* is larger than the current data being 1690ab127040SQuentin Monnet * processed, then the eBPF verdict will be applied to multiple 1691ab127040SQuentin Monnet * **sendmsg**\ () or **sendfile**\ () calls until *bytes* are 1692ab127040SQuentin Monnet * consumed. 1693ab127040SQuentin Monnet * 1694ab127040SQuentin Monnet * Note that if a socket closes with the internal counter holding 1695ab127040SQuentin Monnet * a non-zero value, this is not a problem because data is not 1696ab127040SQuentin Monnet * being buffered for *bytes* and is sent as it is received. 1697ab127040SQuentin Monnet * Return 1698ab127040SQuentin Monnet * 0 1699ab127040SQuentin Monnet * 1700ab127040SQuentin Monnet * int bpf_msg_cork_bytes(struct sk_msg_buff *msg, u32 bytes) 1701ab127040SQuentin Monnet * Description 1702ab127040SQuentin Monnet * For socket policies, prevent the execution of the verdict eBPF 1703ab127040SQuentin Monnet * program for message *msg* until *bytes* (byte number) have been 1704ab127040SQuentin Monnet * accumulated. 1705ab127040SQuentin Monnet * 1706ab127040SQuentin Monnet * This can be used when one needs a specific number of bytes 1707ab127040SQuentin Monnet * before a verdict can be assigned, even if the data spans 1708ab127040SQuentin Monnet * multiple **sendmsg**\ () or **sendfile**\ () calls. The extreme 1709ab127040SQuentin Monnet * case would be a user calling **sendmsg**\ () repeatedly with 1710ab127040SQuentin Monnet * 1-byte long message segments. Obviously, this is bad for 1711ab127040SQuentin Monnet * performance, but it is still valid. If the eBPF program needs 1712ab127040SQuentin Monnet * *bytes* bytes to validate a header, this helper can be used to 1713ab127040SQuentin Monnet * prevent the eBPF program to be called again until *bytes* have 1714ab127040SQuentin Monnet * been accumulated. 1715ab127040SQuentin Monnet * Return 1716ab127040SQuentin Monnet * 0 1717ab127040SQuentin Monnet * 1718ab127040SQuentin Monnet * int bpf_msg_pull_data(struct sk_msg_buff *msg, u32 start, u32 end, u64 flags) 1719ab127040SQuentin Monnet * Description 1720ab127040SQuentin Monnet * For socket policies, pull in non-linear data from user space 1721ab127040SQuentin Monnet * for *msg* and set pointers *msg*\ **->data** and *msg*\ 1722ab127040SQuentin Monnet * **->data_end** to *start* and *end* bytes offsets into *msg*, 1723ab127040SQuentin Monnet * respectively. 1724ab127040SQuentin Monnet * 1725ab127040SQuentin Monnet * If a program of type **BPF_PROG_TYPE_SK_MSG** is run on a 1726ab127040SQuentin Monnet * *msg* it can only parse data that the (**data**, **data_end**) 1727ab127040SQuentin Monnet * pointers have already consumed. For **sendmsg**\ () hooks this 1728ab127040SQuentin Monnet * is likely the first scatterlist element. But for calls relying 1729ab127040SQuentin Monnet * on the **sendpage** handler (e.g. **sendfile**\ ()) this will 1730ab127040SQuentin Monnet * be the range (**0**, **0**) because the data is shared with 1731ab127040SQuentin Monnet * user space and by default the objective is to avoid allowing 1732ab127040SQuentin Monnet * user space to modify data while (or after) eBPF verdict is 1733ab127040SQuentin Monnet * being decided. This helper can be used to pull in data and to 1734ab127040SQuentin Monnet * set the start and end pointer to given values. Data will be 1735ab127040SQuentin Monnet * copied if necessary (i.e. if data was not linear and if start 1736ab127040SQuentin Monnet * and end pointers do not point to the same chunk). 1737ab127040SQuentin Monnet * 1738ab127040SQuentin Monnet * A call to this helper is susceptible to change the underlaying 1739ab127040SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 1740ab127040SQuentin Monnet * previously done by the verifier are invalidated and must be 1741ab127040SQuentin Monnet * performed again, if the helper is used in combination with 1742ab127040SQuentin Monnet * direct packet access. 1743ab127040SQuentin Monnet * 1744ab127040SQuentin Monnet * All values for *flags* are reserved for future usage, and must 1745ab127040SQuentin Monnet * be left at zero. 1746ab127040SQuentin Monnet * Return 1747ab127040SQuentin Monnet * 0 on success, or a negative error in case of failure. 1748ab127040SQuentin Monnet * 1749a3ef8e9aSAndrey Ignatov * int bpf_bind(struct bpf_sock_addr *ctx, struct sockaddr *addr, int addr_len) 17507aa79a86SQuentin Monnet * Description 17517aa79a86SQuentin Monnet * Bind the socket associated to *ctx* to the address pointed by 17527aa79a86SQuentin Monnet * *addr*, of length *addr_len*. This allows for making outgoing 17537aa79a86SQuentin Monnet * connection from the desired IP address, which can be useful for 17547aa79a86SQuentin Monnet * example when all processes inside a cgroup should use one 17557aa79a86SQuentin Monnet * single IP address on a host that has multiple IP configured. 17567aa79a86SQuentin Monnet * 17577aa79a86SQuentin Monnet * This helper works for IPv4 and IPv6, TCP and UDP sockets. The 17587aa79a86SQuentin Monnet * domain (*addr*\ **->sa_family**) must be **AF_INET** (or 17597aa79a86SQuentin Monnet * **AF_INET6**). Looking for a free port to bind to can be 17607aa79a86SQuentin Monnet * expensive, therefore binding to port is not permitted by the 17617aa79a86SQuentin Monnet * helper: *addr*\ **->sin_port** (or **sin6_port**, respectively) 17627aa79a86SQuentin Monnet * must be set to zero. 17637aa79a86SQuentin Monnet * Return 17647aa79a86SQuentin Monnet * 0 on success, or a negative error in case of failure. 17652d020dd7SQuentin Monnet * 17662d020dd7SQuentin Monnet * int bpf_xdp_adjust_tail(struct xdp_buff *xdp_md, int delta) 17672d020dd7SQuentin Monnet * Description 17682d020dd7SQuentin Monnet * Adjust (move) *xdp_md*\ **->data_end** by *delta* bytes. It is 17692d020dd7SQuentin Monnet * only possible to shrink the packet as of this writing, 17702d020dd7SQuentin Monnet * therefore *delta* must be a negative integer. 17712d020dd7SQuentin Monnet * 17722d020dd7SQuentin Monnet * A call to this helper is susceptible to change the underlaying 17732d020dd7SQuentin Monnet * packet buffer. Therefore, at load time, all checks on pointers 17742d020dd7SQuentin Monnet * previously done by the verifier are invalidated and must be 17752d020dd7SQuentin Monnet * performed again, if the helper is used in combination with 17762d020dd7SQuentin Monnet * direct packet access. 17772d020dd7SQuentin Monnet * Return 17782d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 17792d020dd7SQuentin Monnet * 17802d020dd7SQuentin Monnet * int bpf_skb_get_xfrm_state(struct sk_buff *skb, u32 index, struct bpf_xfrm_state *xfrm_state, u32 size, u64 flags) 17812d020dd7SQuentin Monnet * Description 17822d020dd7SQuentin Monnet * Retrieve the XFRM state (IP transform framework, see also 17832d020dd7SQuentin Monnet * **ip-xfrm(8)**) at *index* in XFRM "security path" for *skb*. 17842d020dd7SQuentin Monnet * 17852d020dd7SQuentin Monnet * The retrieved value is stored in the **struct bpf_xfrm_state** 17862d020dd7SQuentin Monnet * pointed by *xfrm_state* and of length *size*. 17872d020dd7SQuentin Monnet * 17882d020dd7SQuentin Monnet * All values for *flags* are reserved for future usage, and must 17892d020dd7SQuentin Monnet * be left at zero. 17902d020dd7SQuentin Monnet * 17912d020dd7SQuentin Monnet * This helper is available only if the kernel was compiled with 17922d020dd7SQuentin Monnet * **CONFIG_XFRM** configuration option. 17932d020dd7SQuentin Monnet * Return 17942d020dd7SQuentin Monnet * 0 on success, or a negative error in case of failure. 1795c195651eSYonghong Song * 1796c195651eSYonghong Song * int bpf_get_stack(struct pt_regs *regs, void *buf, u32 size, u64 flags) 1797c195651eSYonghong Song * Description 1798c195651eSYonghong Song * Return a user or a kernel stack in bpf program provided buffer. 1799c195651eSYonghong Song * To achieve this, the helper needs *ctx*, which is a pointer 1800c195651eSYonghong Song * to the context on which the tracing program is executed. 1801c195651eSYonghong Song * To store the stacktrace, the bpf program provides *buf* with 1802c195651eSYonghong Song * a nonnegative *size*. 1803c195651eSYonghong Song * 1804c195651eSYonghong Song * The last argument, *flags*, holds the number of stack frames to 1805c195651eSYonghong Song * skip (from 0 to 255), masked with 1806c195651eSYonghong Song * **BPF_F_SKIP_FIELD_MASK**. The next bits can be used to set 1807c195651eSYonghong Song * the following flags: 1808c195651eSYonghong Song * 1809c195651eSYonghong Song * **BPF_F_USER_STACK** 1810c195651eSYonghong Song * Collect a user space stack instead of a kernel stack. 1811c195651eSYonghong Song * **BPF_F_USER_BUILD_ID** 1812c195651eSYonghong Song * Collect buildid+offset instead of ips for user stack, 1813c195651eSYonghong Song * only valid if **BPF_F_USER_STACK** is also specified. 1814c195651eSYonghong Song * 1815c195651eSYonghong Song * **bpf_get_stack**\ () can collect up to 1816c195651eSYonghong Song * **PERF_MAX_STACK_DEPTH** both kernel and user frames, subject 1817c195651eSYonghong Song * to sufficient large buffer size. Note that 1818c195651eSYonghong Song * this limit can be controlled with the **sysctl** program, and 1819c195651eSYonghong Song * that it should be manually increased in order to profile long 1820c195651eSYonghong Song * user stacks (such as stacks for Java programs). To do so, use: 1821c195651eSYonghong Song * 1822c195651eSYonghong Song * :: 1823c195651eSYonghong Song * 1824c195651eSYonghong Song * # sysctl kernel.perf_event_max_stack=<new value> 1825c195651eSYonghong Song * Return 18267a279e93SQuentin Monnet * A non-negative value equal to or less than *size* on success, 18277a279e93SQuentin Monnet * or a negative error in case of failure. 18284e1ec56cSDaniel Borkmann * 18292bae79d2SQuentin Monnet * int bpf_skb_load_bytes_relative(const struct sk_buff *skb, u32 offset, void *to, u32 len, u32 start_header) 18304e1ec56cSDaniel Borkmann * Description 18314e1ec56cSDaniel Borkmann * This helper is similar to **bpf_skb_load_bytes**\ () in that 18324e1ec56cSDaniel Borkmann * it provides an easy way to load *len* bytes from *offset* 18334e1ec56cSDaniel Borkmann * from the packet associated to *skb*, into the buffer pointed 18344e1ec56cSDaniel Borkmann * by *to*. The difference to **bpf_skb_load_bytes**\ () is that 18354e1ec56cSDaniel Borkmann * a fifth argument *start_header* exists in order to select a 18364e1ec56cSDaniel Borkmann * base offset to start from. *start_header* can be one of: 18374e1ec56cSDaniel Borkmann * 18384e1ec56cSDaniel Borkmann * **BPF_HDR_START_MAC** 18394e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s mac header. 18404e1ec56cSDaniel Borkmann * **BPF_HDR_START_NET** 18414e1ec56cSDaniel Borkmann * Base offset to load data from is *skb*'s network header. 18424e1ec56cSDaniel Borkmann * 18434e1ec56cSDaniel Borkmann * In general, "direct packet access" is the preferred method to 18444e1ec56cSDaniel Borkmann * access packet data, however, this helper is in particular useful 18454e1ec56cSDaniel Borkmann * in socket filters where *skb*\ **->data** does not always point 18464e1ec56cSDaniel Borkmann * to the start of the mac header and where "direct packet access" 18474e1ec56cSDaniel Borkmann * is not available. 18484e1ec56cSDaniel Borkmann * Return 18494e1ec56cSDaniel Borkmann * 0 on success, or a negative error in case of failure. 18504e1ec56cSDaniel Borkmann * 185187f5fc7eSDavid Ahern * int bpf_fib_lookup(void *ctx, struct bpf_fib_lookup *params, int plen, u32 flags) 185287f5fc7eSDavid Ahern * Description 185387f5fc7eSDavid Ahern * Do FIB lookup in kernel tables using parameters in *params*. 185487f5fc7eSDavid Ahern * If lookup is successful and result shows packet is to be 185587f5fc7eSDavid Ahern * forwarded, the neighbor tables are searched for the nexthop. 185687f5fc7eSDavid Ahern * If successful (ie., FIB lookup shows forwarding and nexthop 1857fa898d76SDavid Ahern * is resolved), the nexthop address is returned in ipv4_dst 1858fa898d76SDavid Ahern * or ipv6_dst based on family, smac is set to mac address of 1859fa898d76SDavid Ahern * egress device, dmac is set to nexthop mac address, rt_metric 18604c79579bSDavid Ahern * is set to metric from route (IPv4/IPv6 only), and ifindex 18614c79579bSDavid Ahern * is set to the device index of the nexthop from the FIB lookup. 186287f5fc7eSDavid Ahern * 186387f5fc7eSDavid Ahern * *plen* argument is the size of the passed in struct. 18647a279e93SQuentin Monnet * *flags* argument can be a combination of one or more of the 18657a279e93SQuentin Monnet * following values: 186687f5fc7eSDavid Ahern * 18677a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_DIRECT** 18687a279e93SQuentin Monnet * Do a direct table lookup vs full lookup using FIB 18697a279e93SQuentin Monnet * rules. 18707a279e93SQuentin Monnet * **BPF_FIB_LOOKUP_OUTPUT** 18717a279e93SQuentin Monnet * Perform lookup from an egress perspective (default is 18727a279e93SQuentin Monnet * ingress). 187387f5fc7eSDavid Ahern * 187487f5fc7eSDavid Ahern * *ctx* is either **struct xdp_md** for XDP programs or 187587f5fc7eSDavid Ahern * **struct sk_buff** tc cls_act programs. 187687f5fc7eSDavid Ahern * Return 18774c79579bSDavid Ahern * * < 0 if any input argument is invalid 18784c79579bSDavid Ahern * * 0 on success (packet is forwarded, nexthop neighbor exists) 18794c79579bSDavid Ahern * * > 0 one of **BPF_FIB_LKUP_RET_** codes explaining why the 18802bae79d2SQuentin Monnet * packet is not forwarded or needs assist from full stack 188181110384SJohn Fastabend * 188281110384SJohn Fastabend * int bpf_sock_hash_update(struct bpf_sock_ops_kern *skops, struct bpf_map *map, void *key, u64 flags) 188381110384SJohn Fastabend * Description 188481110384SJohn Fastabend * Add an entry to, or update a sockhash *map* referencing sockets. 188581110384SJohn Fastabend * The *skops* is used as a new value for the entry associated to 188681110384SJohn Fastabend * *key*. *flags* is one of: 188781110384SJohn Fastabend * 188881110384SJohn Fastabend * **BPF_NOEXIST** 188981110384SJohn Fastabend * The entry for *key* must not exist in the map. 189081110384SJohn Fastabend * **BPF_EXIST** 189181110384SJohn Fastabend * The entry for *key* must already exist in the map. 189281110384SJohn Fastabend * **BPF_ANY** 189381110384SJohn Fastabend * No condition on the existence of the entry for *key*. 189481110384SJohn Fastabend * 189581110384SJohn Fastabend * If the *map* has eBPF programs (parser and verdict), those will 189681110384SJohn Fastabend * be inherited by the socket being added. If the socket is 189781110384SJohn Fastabend * already attached to eBPF programs, this results in an error. 189881110384SJohn Fastabend * Return 189981110384SJohn Fastabend * 0 on success, or a negative error in case of failure. 190081110384SJohn Fastabend * 190181110384SJohn Fastabend * int bpf_msg_redirect_hash(struct sk_msg_buff *msg, struct bpf_map *map, void *key, u64 flags) 190281110384SJohn Fastabend * Description 190381110384SJohn Fastabend * This helper is used in programs implementing policies at the 190481110384SJohn Fastabend * socket level. If the message *msg* is allowed to pass (i.e. if 190581110384SJohn Fastabend * the verdict eBPF program returns **SK_PASS**), redirect it to 190681110384SJohn Fastabend * the socket referenced by *map* (of type 190781110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 190881110384SJohn Fastabend * egress interfaces can be used for redirection. The 190981110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 191081110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 191181110384SJohn Fastabend * egress path otherwise). This is the only flag supported for now. 191281110384SJohn Fastabend * Return 191381110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 191481110384SJohn Fastabend * 191581110384SJohn Fastabend * int bpf_sk_redirect_hash(struct sk_buff *skb, struct bpf_map *map, void *key, u64 flags) 191681110384SJohn Fastabend * Description 191781110384SJohn Fastabend * This helper is used in programs implementing policies at the 191881110384SJohn Fastabend * skb socket level. If the sk_buff *skb* is allowed to pass (i.e. 191981110384SJohn Fastabend * if the verdeict eBPF program returns **SK_PASS**), redirect it 192081110384SJohn Fastabend * to the socket referenced by *map* (of type 192181110384SJohn Fastabend * **BPF_MAP_TYPE_SOCKHASH**) using hash *key*. Both ingress and 192281110384SJohn Fastabend * egress interfaces can be used for redirection. The 192381110384SJohn Fastabend * **BPF_F_INGRESS** value in *flags* is used to make the 192481110384SJohn Fastabend * distinction (ingress path is selected if the flag is present, 192581110384SJohn Fastabend * egress otherwise). This is the only flag supported for now. 192681110384SJohn Fastabend * Return 192781110384SJohn Fastabend * **SK_PASS** on success, or **SK_DROP** on error. 1928fe94cc29SMathieu Xhonneux * 1929fe94cc29SMathieu Xhonneux * int bpf_lwt_push_encap(struct sk_buff *skb, u32 type, void *hdr, u32 len) 1930fe94cc29SMathieu Xhonneux * Description 1931fe94cc29SMathieu Xhonneux * Encapsulate the packet associated to *skb* within a Layer 3 1932fe94cc29SMathieu Xhonneux * protocol header. This header is provided in the buffer at 1933fe94cc29SMathieu Xhonneux * address *hdr*, with *len* its size in bytes. *type* indicates 1934fe94cc29SMathieu Xhonneux * the protocol of the header and can be one of: 1935fe94cc29SMathieu Xhonneux * 1936fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6** 1937fe94cc29SMathieu Xhonneux * IPv6 encapsulation with Segment Routing Header 1938fe94cc29SMathieu Xhonneux * (**struct ipv6_sr_hdr**). *hdr* only contains the SRH, 1939fe94cc29SMathieu Xhonneux * the IPv6 header is computed by the kernel. 1940fe94cc29SMathieu Xhonneux * **BPF_LWT_ENCAP_SEG6_INLINE** 1941fe94cc29SMathieu Xhonneux * Only works if *skb* contains an IPv6 packet. Insert a 1942fe94cc29SMathieu Xhonneux * Segment Routing Header (**struct ipv6_sr_hdr**) inside 1943fe94cc29SMathieu Xhonneux * the IPv6 header. 1944fe94cc29SMathieu Xhonneux * 1945fe94cc29SMathieu Xhonneux * A call to this helper is susceptible to change the underlaying 1946fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 1947fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 1948fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 1949fe94cc29SMathieu Xhonneux * direct packet access. 1950fe94cc29SMathieu Xhonneux * Return 1951fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 1952fe94cc29SMathieu Xhonneux * 1953fe94cc29SMathieu Xhonneux * int bpf_lwt_seg6_store_bytes(struct sk_buff *skb, u32 offset, const void *from, u32 len) 1954fe94cc29SMathieu Xhonneux * Description 1955fe94cc29SMathieu Xhonneux * Store *len* bytes from address *from* into the packet 1956fe94cc29SMathieu Xhonneux * associated to *skb*, at *offset*. Only the flags, tag and TLVs 1957fe94cc29SMathieu Xhonneux * inside the outermost IPv6 Segment Routing Header can be 1958fe94cc29SMathieu Xhonneux * modified through this helper. 1959fe94cc29SMathieu Xhonneux * 1960fe94cc29SMathieu Xhonneux * A call to this helper is susceptible to change the underlaying 1961fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 1962fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 1963fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 1964fe94cc29SMathieu Xhonneux * direct packet access. 1965fe94cc29SMathieu Xhonneux * Return 1966fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 1967fe94cc29SMathieu Xhonneux * 1968fe94cc29SMathieu Xhonneux * int bpf_lwt_seg6_adjust_srh(struct sk_buff *skb, u32 offset, s32 delta) 1969fe94cc29SMathieu Xhonneux * Description 1970fe94cc29SMathieu Xhonneux * Adjust the size allocated to TLVs in the outermost IPv6 1971fe94cc29SMathieu Xhonneux * Segment Routing Header contained in the packet associated to 1972fe94cc29SMathieu Xhonneux * *skb*, at position *offset* by *delta* bytes. Only offsets 1973fe94cc29SMathieu Xhonneux * after the segments are accepted. *delta* can be as well 1974fe94cc29SMathieu Xhonneux * positive (growing) as negative (shrinking). 1975fe94cc29SMathieu Xhonneux * 1976fe94cc29SMathieu Xhonneux * A call to this helper is susceptible to change the underlaying 1977fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 1978fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 1979fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 1980fe94cc29SMathieu Xhonneux * direct packet access. 1981fe94cc29SMathieu Xhonneux * Return 1982fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 1983fe94cc29SMathieu Xhonneux * 1984fe94cc29SMathieu Xhonneux * int bpf_lwt_seg6_action(struct sk_buff *skb, u32 action, void *param, u32 param_len) 1985fe94cc29SMathieu Xhonneux * Description 1986fe94cc29SMathieu Xhonneux * Apply an IPv6 Segment Routing action of type *action* to the 1987fe94cc29SMathieu Xhonneux * packet associated to *skb*. Each action takes a parameter 1988fe94cc29SMathieu Xhonneux * contained at address *param*, and of length *param_len* bytes. 1989fe94cc29SMathieu Xhonneux * *action* can be one of: 1990fe94cc29SMathieu Xhonneux * 1991fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_X** 1992fe94cc29SMathieu Xhonneux * End.X action: Endpoint with Layer-3 cross-connect. 1993fe94cc29SMathieu Xhonneux * Type of *param*: **struct in6_addr**. 1994fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_T** 1995fe94cc29SMathieu Xhonneux * End.T action: Endpoint with specific IPv6 table lookup. 1996fe94cc29SMathieu Xhonneux * Type of *param*: **int**. 1997fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6** 1998fe94cc29SMathieu Xhonneux * End.B6 action: Endpoint bound to an SRv6 policy. 1999fe94cc29SMathieu Xhonneux * Type of param: **struct ipv6_sr_hdr**. 2000fe94cc29SMathieu Xhonneux * **SEG6_LOCAL_ACTION_END_B6_ENCAP** 2001fe94cc29SMathieu Xhonneux * End.B6.Encap action: Endpoint bound to an SRv6 2002fe94cc29SMathieu Xhonneux * encapsulation policy. 2003fe94cc29SMathieu Xhonneux * Type of param: **struct ipv6_sr_hdr**. 2004fe94cc29SMathieu Xhonneux * 2005fe94cc29SMathieu Xhonneux * A call to this helper is susceptible to change the underlaying 2006fe94cc29SMathieu Xhonneux * packet buffer. Therefore, at load time, all checks on pointers 2007fe94cc29SMathieu Xhonneux * previously done by the verifier are invalidated and must be 2008fe94cc29SMathieu Xhonneux * performed again, if the helper is used in combination with 2009fe94cc29SMathieu Xhonneux * direct packet access. 2010fe94cc29SMathieu Xhonneux * Return 2011fe94cc29SMathieu Xhonneux * 0 on success, or a negative error in case of failure. 2012f4364dcfSSean Young * 2013f4364dcfSSean Young * int bpf_rc_keydown(void *ctx, u32 protocol, u64 scancode, u32 toggle) 2014f4364dcfSSean Young * Description 2015f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 2016f4364dcfSSean Young * report a successfully decoded key press with *scancode*, 2017f4364dcfSSean Young * *toggle* value in the given *protocol*. The scancode will be 2018f4364dcfSSean Young * translated to a keycode using the rc keymap, and reported as 2019f4364dcfSSean Young * an input key down event. After a period a key up event is 2020f4364dcfSSean Young * generated. This period can be extended by calling either 2021f4364dcfSSean Young * **bpf_rc_keydown** () again with the same values, or calling 2022f4364dcfSSean Young * **bpf_rc_repeat** (). 2023f4364dcfSSean Young * 2024f4364dcfSSean Young * Some protocols include a toggle bit, in case the button was 2025f4364dcfSSean Young * released and pressed again between consecutive scancodes. 2026f4364dcfSSean Young * 2027f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 2028f4364dcfSSean Young * the program. 2029f4364dcfSSean Young * 2030f4364dcfSSean Young * The *protocol* is the decoded protocol number (see 2031f4364dcfSSean Young * **enum rc_proto** for some predefined values). 2032f4364dcfSSean Young * 2033f4364dcfSSean Young * This helper is only available is the kernel was compiled with 2034f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2035f4364dcfSSean Young * "**y**". 2036f4364dcfSSean Young * Return 2037f4364dcfSSean Young * 0 2038f4364dcfSSean Young * 2039f4364dcfSSean Young * int bpf_rc_repeat(void *ctx) 2040f4364dcfSSean Young * Description 2041f4364dcfSSean Young * This helper is used in programs implementing IR decoding, to 2042f4364dcfSSean Young * report a successfully decoded repeat key message. This delays 2043f4364dcfSSean Young * the generation of a key up event for previously generated 2044f4364dcfSSean Young * key down event. 2045f4364dcfSSean Young * 2046f4364dcfSSean Young * Some IR protocols like NEC have a special IR message for 2047f4364dcfSSean Young * repeating last button, for when a button is held down. 2048f4364dcfSSean Young * 2049f4364dcfSSean Young * The *ctx* should point to the lirc sample as passed into 2050f4364dcfSSean Young * the program. 2051f4364dcfSSean Young * 2052f4364dcfSSean Young * This helper is only available is the kernel was compiled with 2053f4364dcfSSean Young * the **CONFIG_BPF_LIRC_MODE2** configuration option set to 2054f4364dcfSSean Young * "**y**". 2055f4364dcfSSean Young * Return 2056f4364dcfSSean Young * 0 2057cb20b08eSDaniel Borkmann * 2058cb20b08eSDaniel Borkmann * uint64_t bpf_skb_cgroup_id(struct sk_buff *skb) 2059cb20b08eSDaniel Borkmann * Description 2060cb20b08eSDaniel Borkmann * Return the cgroup v2 id of the socket associated with the *skb*. 2061cb20b08eSDaniel Borkmann * This is roughly similar to the **bpf_get_cgroup_classid**\ () 2062cb20b08eSDaniel Borkmann * helper for cgroup v1 by providing a tag resp. identifier that 2063cb20b08eSDaniel Borkmann * can be matched on or used for map lookups e.g. to implement 2064cb20b08eSDaniel Borkmann * policy. The cgroup v2 id of a given path in the hierarchy is 2065cb20b08eSDaniel Borkmann * exposed in user space through the f_handle API in order to get 2066cb20b08eSDaniel Borkmann * to the same 64-bit id. 2067cb20b08eSDaniel Borkmann * 2068cb20b08eSDaniel Borkmann * This helper can be used on TC egress path, but not on ingress, 2069cb20b08eSDaniel Borkmann * and is available only if the kernel was compiled with the 2070cb20b08eSDaniel Borkmann * **CONFIG_SOCK_CGROUP_DATA** configuration option. 2071cb20b08eSDaniel Borkmann * Return 2072cb20b08eSDaniel Borkmann * The id is returned or 0 in case the id could not be retrieved. 2073bf6fa2c8SYonghong Song * 2074bf6fa2c8SYonghong Song * u64 bpf_get_current_cgroup_id(void) 2075bf6fa2c8SYonghong Song * Return 2076bf6fa2c8SYonghong Song * A 64-bit integer containing the current cgroup id based 2077bf6fa2c8SYonghong Song * on the cgroup within which the current task is running. 20787a4b28c6SDaniel Borkmann */ 2079ebb676daSThomas Graf #define __BPF_FUNC_MAPPER(FN) \ 2080ebb676daSThomas Graf FN(unspec), \ 2081ebb676daSThomas Graf FN(map_lookup_elem), \ 2082ebb676daSThomas Graf FN(map_update_elem), \ 2083ebb676daSThomas Graf FN(map_delete_elem), \ 2084ebb676daSThomas Graf FN(probe_read), \ 2085ebb676daSThomas Graf FN(ktime_get_ns), \ 2086ebb676daSThomas Graf FN(trace_printk), \ 2087ebb676daSThomas Graf FN(get_prandom_u32), \ 2088ebb676daSThomas Graf FN(get_smp_processor_id), \ 2089ebb676daSThomas Graf FN(skb_store_bytes), \ 2090ebb676daSThomas Graf FN(l3_csum_replace), \ 2091ebb676daSThomas Graf FN(l4_csum_replace), \ 2092ebb676daSThomas Graf FN(tail_call), \ 2093ebb676daSThomas Graf FN(clone_redirect), \ 2094ebb676daSThomas Graf FN(get_current_pid_tgid), \ 2095ebb676daSThomas Graf FN(get_current_uid_gid), \ 2096ebb676daSThomas Graf FN(get_current_comm), \ 2097ebb676daSThomas Graf FN(get_cgroup_classid), \ 2098ebb676daSThomas Graf FN(skb_vlan_push), \ 2099ebb676daSThomas Graf FN(skb_vlan_pop), \ 2100ebb676daSThomas Graf FN(skb_get_tunnel_key), \ 2101ebb676daSThomas Graf FN(skb_set_tunnel_key), \ 2102ebb676daSThomas Graf FN(perf_event_read), \ 2103ebb676daSThomas Graf FN(redirect), \ 2104ebb676daSThomas Graf FN(get_route_realm), \ 2105ebb676daSThomas Graf FN(perf_event_output), \ 2106ebb676daSThomas Graf FN(skb_load_bytes), \ 2107ebb676daSThomas Graf FN(get_stackid), \ 2108ebb676daSThomas Graf FN(csum_diff), \ 2109ebb676daSThomas Graf FN(skb_get_tunnel_opt), \ 2110ebb676daSThomas Graf FN(skb_set_tunnel_opt), \ 2111ebb676daSThomas Graf FN(skb_change_proto), \ 2112ebb676daSThomas Graf FN(skb_change_type), \ 2113ebb676daSThomas Graf FN(skb_under_cgroup), \ 2114ebb676daSThomas Graf FN(get_hash_recalc), \ 2115ebb676daSThomas Graf FN(get_current_task), \ 2116ebb676daSThomas Graf FN(probe_write_user), \ 2117ebb676daSThomas Graf FN(current_task_under_cgroup), \ 2118ebb676daSThomas Graf FN(skb_change_tail), \ 2119ebb676daSThomas Graf FN(skb_pull_data), \ 2120ebb676daSThomas Graf FN(csum_update), \ 2121ebb676daSThomas Graf FN(set_hash_invalid), \ 21223a0af8fdSThomas Graf FN(get_numa_node_id), \ 212317bedab2SMartin KaFai Lau FN(skb_change_head), \ 2124a5e8c070SGianluca Borello FN(xdp_adjust_head), \ 212591b8270fSChenbo Feng FN(probe_read_str), \ 21266acc5c29SChenbo Feng FN(get_socket_cookie), \ 2127ded092cdSDaniel Borkmann FN(get_socket_uid), \ 21288c4b4c7eSLawrence Brakmo FN(set_hash), \ 21292be7e212SDaniel Borkmann FN(setsockopt), \ 213097f91a7cSJohn Fastabend FN(skb_adjust_room), \ 2131174a79ffSJohn Fastabend FN(redirect_map), \ 2132174a79ffSJohn Fastabend FN(sk_redirect_map), \ 2133174a79ffSJohn Fastabend FN(sock_map_update), \ 2134908432caSYonghong Song FN(xdp_adjust_meta), \ 21354bebdc7aSYonghong Song FN(perf_event_read_value), \ 2136cd86d1fdSLawrence Brakmo FN(perf_prog_read_value), \ 21379802d865SJosef Bacik FN(getsockopt), \ 2138b13d8807SLawrence Brakmo FN(override_return), \ 21394f738adbSJohn Fastabend FN(sock_ops_cb_flags_set), \ 21402a100317SJohn Fastabend FN(msg_redirect_map), \ 214191843d54SJohn Fastabend FN(msg_apply_bytes), \ 2142015632bbSJohn Fastabend FN(msg_cork_bytes), \ 2143d74bad4eSAndrey Ignatov FN(msg_pull_data), \ 2144b32cc5b9SNikita V. Shirokov FN(bind), \ 214512bed760SEyal Birger FN(xdp_adjust_tail), \ 2146c195651eSYonghong Song FN(skb_get_xfrm_state), \ 21474e1ec56cSDaniel Borkmann FN(get_stack), \ 214887f5fc7eSDavid Ahern FN(skb_load_bytes_relative), \ 214981110384SJohn Fastabend FN(fib_lookup), \ 215081110384SJohn Fastabend FN(sock_hash_update), \ 215181110384SJohn Fastabend FN(msg_redirect_hash), \ 2152fe94cc29SMathieu Xhonneux FN(sk_redirect_hash), \ 2153fe94cc29SMathieu Xhonneux FN(lwt_push_encap), \ 2154fe94cc29SMathieu Xhonneux FN(lwt_seg6_store_bytes), \ 2155fe94cc29SMathieu Xhonneux FN(lwt_seg6_adjust_srh), \ 2156f4364dcfSSean Young FN(lwt_seg6_action), \ 2157f4364dcfSSean Young FN(rc_repeat), \ 2158cb20b08eSDaniel Borkmann FN(rc_keydown), \ 2159bf6fa2c8SYonghong Song FN(skb_cgroup_id), \ 2160bf6fa2c8SYonghong Song FN(get_current_cgroup_id), 21617a4b28c6SDaniel Borkmann 2162ebb676daSThomas Graf /* integer value in 'imm' field of BPF_CALL instruction selects which helper 2163ebb676daSThomas Graf * function eBPF program intends to call 21642d0e30c3SDaniel Borkmann */ 2165ebb676daSThomas Graf #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x 2166ebb676daSThomas Graf enum bpf_func_id { 2167ebb676daSThomas Graf __BPF_FUNC_MAPPER(__BPF_ENUM_FN) 216809756af4SAlexei Starovoitov __BPF_FUNC_MAX_ID, 216909756af4SAlexei Starovoitov }; 2170ebb676daSThomas Graf #undef __BPF_ENUM_FN 217109756af4SAlexei Starovoitov 2172781c53bcSDaniel Borkmann /* All flags used by eBPF helper functions, placed here. */ 2173781c53bcSDaniel Borkmann 2174781c53bcSDaniel Borkmann /* BPF_FUNC_skb_store_bytes flags. */ 2175781c53bcSDaniel Borkmann #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 21768afd54c8SDaniel Borkmann #define BPF_F_INVALIDATE_HASH (1ULL << 1) 2177781c53bcSDaniel Borkmann 2178781c53bcSDaniel Borkmann /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 2179781c53bcSDaniel Borkmann * First 4 bits are for passing the header field size. 2180781c53bcSDaniel Borkmann */ 2181781c53bcSDaniel Borkmann #define BPF_F_HDR_FIELD_MASK 0xfULL 2182781c53bcSDaniel Borkmann 2183781c53bcSDaniel Borkmann /* BPF_FUNC_l4_csum_replace flags. */ 2184781c53bcSDaniel Borkmann #define BPF_F_PSEUDO_HDR (1ULL << 4) 21852f72959aSDaniel Borkmann #define BPF_F_MARK_MANGLED_0 (1ULL << 5) 2186d1b662adSDaniel Borkmann #define BPF_F_MARK_ENFORCE (1ULL << 6) 2187781c53bcSDaniel Borkmann 2188781c53bcSDaniel Borkmann /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 2189781c53bcSDaniel Borkmann #define BPF_F_INGRESS (1ULL << 0) 2190781c53bcSDaniel Borkmann 2191c6c33454SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 2192c6c33454SDaniel Borkmann #define BPF_F_TUNINFO_IPV6 (1ULL << 0) 2193c6c33454SDaniel Borkmann 2194c195651eSYonghong Song /* flags for both BPF_FUNC_get_stackid and BPF_FUNC_get_stack. */ 2195d5a3b1f6SAlexei Starovoitov #define BPF_F_SKIP_FIELD_MASK 0xffULL 2196d5a3b1f6SAlexei Starovoitov #define BPF_F_USER_STACK (1ULL << 8) 2197c195651eSYonghong Song /* flags used by BPF_FUNC_get_stackid only. */ 2198d5a3b1f6SAlexei Starovoitov #define BPF_F_FAST_STACK_CMP (1ULL << 9) 2199d5a3b1f6SAlexei Starovoitov #define BPF_F_REUSE_STACKID (1ULL << 10) 2200c195651eSYonghong Song /* flags used by BPF_FUNC_get_stack only. */ 2201c195651eSYonghong Song #define BPF_F_USER_BUILD_ID (1ULL << 11) 2202d5a3b1f6SAlexei Starovoitov 22032da897e5SDaniel Borkmann /* BPF_FUNC_skb_set_tunnel_key flags. */ 22042da897e5SDaniel Borkmann #define BPF_F_ZERO_CSUM_TX (1ULL << 1) 220522080870SDaniel Borkmann #define BPF_F_DONT_FRAGMENT (1ULL << 2) 220677a5196aSWilliam Tu #define BPF_F_SEQ_NUMBER (1ULL << 3) 22072da897e5SDaniel Borkmann 2208908432caSYonghong Song /* BPF_FUNC_perf_event_output, BPF_FUNC_perf_event_read and 2209908432caSYonghong Song * BPF_FUNC_perf_event_read_value flags. 2210908432caSYonghong Song */ 22111e33759cSDaniel Borkmann #define BPF_F_INDEX_MASK 0xffffffffULL 22121e33759cSDaniel Borkmann #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 2213555c8a86SDaniel Borkmann /* BPF_FUNC_perf_event_output for sk_buff input context. */ 2214555c8a86SDaniel Borkmann #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 22151e33759cSDaniel Borkmann 22162be7e212SDaniel Borkmann /* Mode for BPF_FUNC_skb_adjust_room helper. */ 22172be7e212SDaniel Borkmann enum bpf_adj_room_mode { 22182be7e212SDaniel Borkmann BPF_ADJ_ROOM_NET, 22192be7e212SDaniel Borkmann }; 22202be7e212SDaniel Borkmann 22214e1ec56cSDaniel Borkmann /* Mode for BPF_FUNC_skb_load_bytes_relative helper. */ 22224e1ec56cSDaniel Borkmann enum bpf_hdr_start_off { 22234e1ec56cSDaniel Borkmann BPF_HDR_START_MAC, 22244e1ec56cSDaniel Borkmann BPF_HDR_START_NET, 22254e1ec56cSDaniel Borkmann }; 22264e1ec56cSDaniel Borkmann 2227fe94cc29SMathieu Xhonneux /* Encapsulation type for BPF_FUNC_lwt_push_encap helper. */ 2228fe94cc29SMathieu Xhonneux enum bpf_lwt_encap_mode { 2229fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6, 2230fe94cc29SMathieu Xhonneux BPF_LWT_ENCAP_SEG6_INLINE 2231fe94cc29SMathieu Xhonneux }; 2232fe94cc29SMathieu Xhonneux 22339bac3d6dSAlexei Starovoitov /* user accessible mirror of in-kernel sk_buff. 22349bac3d6dSAlexei Starovoitov * new fields can only be added to the end of this structure 22359bac3d6dSAlexei Starovoitov */ 22369bac3d6dSAlexei Starovoitov struct __sk_buff { 22379bac3d6dSAlexei Starovoitov __u32 len; 22389bac3d6dSAlexei Starovoitov __u32 pkt_type; 22399bac3d6dSAlexei Starovoitov __u32 mark; 22409bac3d6dSAlexei Starovoitov __u32 queue_mapping; 2241c2497395SAlexei Starovoitov __u32 protocol; 2242c2497395SAlexei Starovoitov __u32 vlan_present; 2243c2497395SAlexei Starovoitov __u32 vlan_tci; 224427cd5452SMichal Sekletar __u32 vlan_proto; 2245bcad5718SDaniel Borkmann __u32 priority; 224637e82c2fSAlexei Starovoitov __u32 ingress_ifindex; 224737e82c2fSAlexei Starovoitov __u32 ifindex; 2248d691f9e8SAlexei Starovoitov __u32 tc_index; 2249d691f9e8SAlexei Starovoitov __u32 cb[5]; 2250ba7591d8SDaniel Borkmann __u32 hash; 2251045efa82SDaniel Borkmann __u32 tc_classid; 2252969bf05eSAlexei Starovoitov __u32 data; 2253969bf05eSAlexei Starovoitov __u32 data_end; 2254b1d9fc41SDaniel Borkmann __u32 napi_id; 22558a31db56SJohn Fastabend 2256de8f3a83SDaniel Borkmann /* Accessed by BPF_PROG_TYPE_sk_skb types from here to ... */ 22578a31db56SJohn Fastabend __u32 family; 22588a31db56SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 22598a31db56SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 22608a31db56SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 22618a31db56SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 22628a31db56SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 22638a31db56SJohn Fastabend __u32 local_port; /* stored in host byte order */ 2264de8f3a83SDaniel Borkmann /* ... here. */ 2265de8f3a83SDaniel Borkmann 2266de8f3a83SDaniel Borkmann __u32 data_meta; 22679bac3d6dSAlexei Starovoitov }; 22689bac3d6dSAlexei Starovoitov 2269d3aa45ceSAlexei Starovoitov struct bpf_tunnel_key { 2270d3aa45ceSAlexei Starovoitov __u32 tunnel_id; 2271c6c33454SDaniel Borkmann union { 2272d3aa45ceSAlexei Starovoitov __u32 remote_ipv4; 2273c6c33454SDaniel Borkmann __u32 remote_ipv6[4]; 2274c6c33454SDaniel Borkmann }; 2275c6c33454SDaniel Borkmann __u8 tunnel_tos; 2276c6c33454SDaniel Borkmann __u8 tunnel_ttl; 22771fbc2e0cSDaniel Borkmann __u16 tunnel_ext; /* Padding, future use. */ 22784018ab18SDaniel Borkmann __u32 tunnel_label; 2279d3aa45ceSAlexei Starovoitov }; 2280d3aa45ceSAlexei Starovoitov 228112bed760SEyal Birger /* user accessible mirror of in-kernel xfrm_state. 228212bed760SEyal Birger * new fields can only be added to the end of this structure 228312bed760SEyal Birger */ 228412bed760SEyal Birger struct bpf_xfrm_state { 228512bed760SEyal Birger __u32 reqid; 228612bed760SEyal Birger __u32 spi; /* Stored in network byte order */ 228712bed760SEyal Birger __u16 family; 22881fbc2e0cSDaniel Borkmann __u16 ext; /* Padding, future use. */ 228912bed760SEyal Birger union { 229012bed760SEyal Birger __u32 remote_ipv4; /* Stored in network byte order */ 229112bed760SEyal Birger __u32 remote_ipv6[4]; /* Stored in network byte order */ 229212bed760SEyal Birger }; 229312bed760SEyal Birger }; 229412bed760SEyal Birger 22953a0af8fdSThomas Graf /* Generic BPF return codes which all BPF program types may support. 22963a0af8fdSThomas Graf * The values are binary compatible with their TC_ACT_* counter-part to 22973a0af8fdSThomas Graf * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT 22983a0af8fdSThomas Graf * programs. 22993a0af8fdSThomas Graf * 23003a0af8fdSThomas Graf * XDP is handled seprately, see XDP_*. 23013a0af8fdSThomas Graf */ 23023a0af8fdSThomas Graf enum bpf_ret_code { 23033a0af8fdSThomas Graf BPF_OK = 0, 23043a0af8fdSThomas Graf /* 1 reserved */ 23053a0af8fdSThomas Graf BPF_DROP = 2, 23063a0af8fdSThomas Graf /* 3-6 reserved */ 23073a0af8fdSThomas Graf BPF_REDIRECT = 7, 23083a0af8fdSThomas Graf /* >127 are reserved for prog type specific return codes */ 23093a0af8fdSThomas Graf }; 23103a0af8fdSThomas Graf 231161023658SDavid Ahern struct bpf_sock { 231261023658SDavid Ahern __u32 bound_dev_if; 2313aa4c1037SDavid Ahern __u32 family; 2314aa4c1037SDavid Ahern __u32 type; 2315aa4c1037SDavid Ahern __u32 protocol; 2316482dca93SDavid Ahern __u32 mark; 2317482dca93SDavid Ahern __u32 priority; 2318aac3fc32SAndrey Ignatov __u32 src_ip4; /* Allows 1,2,4-byte read. 2319aac3fc32SAndrey Ignatov * Stored in network byte order. 2320aac3fc32SAndrey Ignatov */ 2321aac3fc32SAndrey Ignatov __u32 src_ip6[4]; /* Allows 1,2,4-byte read. 2322aac3fc32SAndrey Ignatov * Stored in network byte order. 2323aac3fc32SAndrey Ignatov */ 2324aac3fc32SAndrey Ignatov __u32 src_port; /* Allows 4-byte read. 2325aac3fc32SAndrey Ignatov * Stored in host byte order 2326aac3fc32SAndrey Ignatov */ 232761023658SDavid Ahern }; 232861023658SDavid Ahern 232917bedab2SMartin KaFai Lau #define XDP_PACKET_HEADROOM 256 233017bedab2SMartin KaFai Lau 23316a773a15SBrenden Blanco /* User return codes for XDP prog type. 23326a773a15SBrenden Blanco * A valid XDP program must return one of these defined values. All other 23339beb8bedSDaniel Borkmann * return codes are reserved for future use. Unknown return codes will 23349beb8bedSDaniel Borkmann * result in packet drops and a warning via bpf_warn_invalid_xdp_action(). 23356a773a15SBrenden Blanco */ 23366a773a15SBrenden Blanco enum xdp_action { 23376a773a15SBrenden Blanco XDP_ABORTED = 0, 23386a773a15SBrenden Blanco XDP_DROP, 23396a773a15SBrenden Blanco XDP_PASS, 23406ce96ca3SBrenden Blanco XDP_TX, 2341814abfabSJohn Fastabend XDP_REDIRECT, 23426a773a15SBrenden Blanco }; 23436a773a15SBrenden Blanco 23446a773a15SBrenden Blanco /* user accessible metadata for XDP packet hook 23456a773a15SBrenden Blanco * new fields must be added to the end of this structure 23466a773a15SBrenden Blanco */ 23476a773a15SBrenden Blanco struct xdp_md { 23486a773a15SBrenden Blanco __u32 data; 23496a773a15SBrenden Blanco __u32 data_end; 2350de8f3a83SDaniel Borkmann __u32 data_meta; 2351daaf24c6SJesper Dangaard Brouer /* Below access go through struct xdp_rxq_info */ 235202dd3291SJesper Dangaard Brouer __u32 ingress_ifindex; /* rxq->dev->ifindex */ 235302dd3291SJesper Dangaard Brouer __u32 rx_queue_index; /* rxq->queue_index */ 23546a773a15SBrenden Blanco }; 23556a773a15SBrenden Blanco 2356174a79ffSJohn Fastabend enum sk_action { 2357bfa64075SJohn Fastabend SK_DROP = 0, 2358bfa64075SJohn Fastabend SK_PASS, 2359174a79ffSJohn Fastabend }; 2360174a79ffSJohn Fastabend 23614f738adbSJohn Fastabend /* user accessible metadata for SK_MSG packet hook, new fields must 23624f738adbSJohn Fastabend * be added to the end of this structure 23634f738adbSJohn Fastabend */ 23644f738adbSJohn Fastabend struct sk_msg_md { 23654f738adbSJohn Fastabend void *data; 23664f738adbSJohn Fastabend void *data_end; 2367303def35SJohn Fastabend 2368303def35SJohn Fastabend __u32 family; 2369303def35SJohn Fastabend __u32 remote_ip4; /* Stored in network byte order */ 2370303def35SJohn Fastabend __u32 local_ip4; /* Stored in network byte order */ 2371303def35SJohn Fastabend __u32 remote_ip6[4]; /* Stored in network byte order */ 2372303def35SJohn Fastabend __u32 local_ip6[4]; /* Stored in network byte order */ 2373303def35SJohn Fastabend __u32 remote_port; /* Stored in network byte order */ 2374303def35SJohn Fastabend __u32 local_port; /* stored in host byte order */ 23754f738adbSJohn Fastabend }; 23764f738adbSJohn Fastabend 23771e270976SMartin KaFai Lau #define BPF_TAG_SIZE 8 23781e270976SMartin KaFai Lau 23791e270976SMartin KaFai Lau struct bpf_prog_info { 23801e270976SMartin KaFai Lau __u32 type; 23811e270976SMartin KaFai Lau __u32 id; 23821e270976SMartin KaFai Lau __u8 tag[BPF_TAG_SIZE]; 23831e270976SMartin KaFai Lau __u32 jited_prog_len; 23841e270976SMartin KaFai Lau __u32 xlated_prog_len; 23851e270976SMartin KaFai Lau __aligned_u64 jited_prog_insns; 23861e270976SMartin KaFai Lau __aligned_u64 xlated_prog_insns; 2387cb4d2b3fSMartin KaFai Lau __u64 load_time; /* ns since boottime */ 2388cb4d2b3fSMartin KaFai Lau __u32 created_by_uid; 2389cb4d2b3fSMartin KaFai Lau __u32 nr_map_ids; 2390cb4d2b3fSMartin KaFai Lau __aligned_u64 map_ids; 2391067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 2392675fc275SJakub Kicinski __u32 ifindex; 2393b85fab0eSJiri Olsa __u32 gpl_compatible:1; 2394675fc275SJakub Kicinski __u64 netns_dev; 2395675fc275SJakub Kicinski __u64 netns_ino; 2396dbecd738SSandipan Das __u32 nr_jited_ksyms; 2397815581c1SSandipan Das __u32 nr_jited_func_lens; 2398dbecd738SSandipan Das __aligned_u64 jited_ksyms; 2399815581c1SSandipan Das __aligned_u64 jited_func_lens; 24001e270976SMartin KaFai Lau } __attribute__((aligned(8))); 24011e270976SMartin KaFai Lau 24021e270976SMartin KaFai Lau struct bpf_map_info { 24031e270976SMartin KaFai Lau __u32 type; 24041e270976SMartin KaFai Lau __u32 id; 24051e270976SMartin KaFai Lau __u32 key_size; 24061e270976SMartin KaFai Lau __u32 value_size; 24071e270976SMartin KaFai Lau __u32 max_entries; 24081e270976SMartin KaFai Lau __u32 map_flags; 2409067cae47SMartin KaFai Lau char name[BPF_OBJ_NAME_LEN]; 241052775b33SJakub Kicinski __u32 ifindex; 241136f9814aSDaniel Borkmann __u32 :32; 241252775b33SJakub Kicinski __u64 netns_dev; 241352775b33SJakub Kicinski __u64 netns_ino; 241478958fcaSMartin KaFai Lau __u32 btf_id; 24159b2cf328SMartin KaFai Lau __u32 btf_key_type_id; 24169b2cf328SMartin KaFai Lau __u32 btf_value_type_id; 24171e270976SMartin KaFai Lau } __attribute__((aligned(8))); 24181e270976SMartin KaFai Lau 241962dab84cSMartin KaFai Lau struct bpf_btf_info { 242062dab84cSMartin KaFai Lau __aligned_u64 btf; 242162dab84cSMartin KaFai Lau __u32 btf_size; 242262dab84cSMartin KaFai Lau __u32 id; 242362dab84cSMartin KaFai Lau } __attribute__((aligned(8))); 242462dab84cSMartin KaFai Lau 24254fbac77dSAndrey Ignatov /* User bpf_sock_addr struct to access socket fields and sockaddr struct passed 24264fbac77dSAndrey Ignatov * by user and intended to be used by socket (e.g. to bind to, depends on 24274fbac77dSAndrey Ignatov * attach attach type). 24284fbac77dSAndrey Ignatov */ 24294fbac77dSAndrey Ignatov struct bpf_sock_addr { 24304fbac77dSAndrey Ignatov __u32 user_family; /* Allows 4-byte read, but no write. */ 24314fbac77dSAndrey Ignatov __u32 user_ip4; /* Allows 1,2,4-byte read and 4-byte write. 24324fbac77dSAndrey Ignatov * Stored in network byte order. 24334fbac77dSAndrey Ignatov */ 24344fbac77dSAndrey Ignatov __u32 user_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 24354fbac77dSAndrey Ignatov * Stored in network byte order. 24364fbac77dSAndrey Ignatov */ 24374fbac77dSAndrey Ignatov __u32 user_port; /* Allows 4-byte read and write. 24384fbac77dSAndrey Ignatov * Stored in network byte order 24394fbac77dSAndrey Ignatov */ 24404fbac77dSAndrey Ignatov __u32 family; /* Allows 4-byte read, but no write */ 24414fbac77dSAndrey Ignatov __u32 type; /* Allows 4-byte read, but no write */ 24424fbac77dSAndrey Ignatov __u32 protocol; /* Allows 4-byte read, but no write */ 24431cedee13SAndrey Ignatov __u32 msg_src_ip4; /* Allows 1,2,4-byte read an 4-byte write. 24441cedee13SAndrey Ignatov * Stored in network byte order. 24451cedee13SAndrey Ignatov */ 24461cedee13SAndrey Ignatov __u32 msg_src_ip6[4]; /* Allows 1,2,4-byte read an 4-byte write. 24471cedee13SAndrey Ignatov * Stored in network byte order. 24481cedee13SAndrey Ignatov */ 24494fbac77dSAndrey Ignatov }; 24504fbac77dSAndrey Ignatov 245140304b2aSLawrence Brakmo /* User bpf_sock_ops struct to access socket values and specify request ops 245240304b2aSLawrence Brakmo * and their replies. 245340304b2aSLawrence Brakmo * Some of this fields are in network (bigendian) byte order and may need 245440304b2aSLawrence Brakmo * to be converted before use (bpf_ntohl() defined in samples/bpf/bpf_endian.h). 245540304b2aSLawrence Brakmo * New fields can only be added at the end of this structure 245640304b2aSLawrence Brakmo */ 245740304b2aSLawrence Brakmo struct bpf_sock_ops { 245840304b2aSLawrence Brakmo __u32 op; 245940304b2aSLawrence Brakmo union { 2460de525be2SLawrence Brakmo __u32 args[4]; /* Optionally passed to bpf program */ 2461de525be2SLawrence Brakmo __u32 reply; /* Returned by bpf program */ 2462de525be2SLawrence Brakmo __u32 replylong[4]; /* Optionally returned by bpf prog */ 246340304b2aSLawrence Brakmo }; 246440304b2aSLawrence Brakmo __u32 family; 246540304b2aSLawrence Brakmo __u32 remote_ip4; /* Stored in network byte order */ 246640304b2aSLawrence Brakmo __u32 local_ip4; /* Stored in network byte order */ 246740304b2aSLawrence Brakmo __u32 remote_ip6[4]; /* Stored in network byte order */ 246840304b2aSLawrence Brakmo __u32 local_ip6[4]; /* Stored in network byte order */ 246940304b2aSLawrence Brakmo __u32 remote_port; /* Stored in network byte order */ 247040304b2aSLawrence Brakmo __u32 local_port; /* stored in host byte order */ 2471f19397a5SLawrence Brakmo __u32 is_fullsock; /* Some TCP fields are only valid if 2472f19397a5SLawrence Brakmo * there is a full socket. If not, the 2473f19397a5SLawrence Brakmo * fields read as zero. 2474f19397a5SLawrence Brakmo */ 2475f19397a5SLawrence Brakmo __u32 snd_cwnd; 2476f19397a5SLawrence Brakmo __u32 srtt_us; /* Averaged RTT << 3 in usecs */ 2477b13d8807SLawrence Brakmo __u32 bpf_sock_ops_cb_flags; /* flags defined in uapi/linux/tcp.h */ 247844f0e430SLawrence Brakmo __u32 state; 247944f0e430SLawrence Brakmo __u32 rtt_min; 248044f0e430SLawrence Brakmo __u32 snd_ssthresh; 248144f0e430SLawrence Brakmo __u32 rcv_nxt; 248244f0e430SLawrence Brakmo __u32 snd_nxt; 248344f0e430SLawrence Brakmo __u32 snd_una; 248444f0e430SLawrence Brakmo __u32 mss_cache; 248544f0e430SLawrence Brakmo __u32 ecn_flags; 248644f0e430SLawrence Brakmo __u32 rate_delivered; 248744f0e430SLawrence Brakmo __u32 rate_interval_us; 248844f0e430SLawrence Brakmo __u32 packets_out; 248944f0e430SLawrence Brakmo __u32 retrans_out; 249044f0e430SLawrence Brakmo __u32 total_retrans; 249144f0e430SLawrence Brakmo __u32 segs_in; 249244f0e430SLawrence Brakmo __u32 data_segs_in; 249344f0e430SLawrence Brakmo __u32 segs_out; 249444f0e430SLawrence Brakmo __u32 data_segs_out; 249544f0e430SLawrence Brakmo __u32 lost_out; 249644f0e430SLawrence Brakmo __u32 sacked_out; 249744f0e430SLawrence Brakmo __u32 sk_txhash; 249844f0e430SLawrence Brakmo __u64 bytes_received; 249944f0e430SLawrence Brakmo __u64 bytes_acked; 250040304b2aSLawrence Brakmo }; 250140304b2aSLawrence Brakmo 2502b13d8807SLawrence Brakmo /* Definitions for bpf_sock_ops_cb_flags */ 2503f89013f6SLawrence Brakmo #define BPF_SOCK_OPS_RTO_CB_FLAG (1<<0) 2504a31ad29eSLawrence Brakmo #define BPF_SOCK_OPS_RETRANS_CB_FLAG (1<<1) 2505d4487491SLawrence Brakmo #define BPF_SOCK_OPS_STATE_CB_FLAG (1<<2) 2506d4487491SLawrence Brakmo #define BPF_SOCK_OPS_ALL_CB_FLAGS 0x7 /* Mask of all currently 2507b13d8807SLawrence Brakmo * supported cb flags 2508b13d8807SLawrence Brakmo */ 2509b13d8807SLawrence Brakmo 251040304b2aSLawrence Brakmo /* List of known BPF sock_ops operators. 251140304b2aSLawrence Brakmo * New entries can only be added at the end 251240304b2aSLawrence Brakmo */ 251340304b2aSLawrence Brakmo enum { 251440304b2aSLawrence Brakmo BPF_SOCK_OPS_VOID, 25158550f328SLawrence Brakmo BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or 25168550f328SLawrence Brakmo * -1 if default value should be used 25178550f328SLawrence Brakmo */ 251813d3b1ebSLawrence Brakmo BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized 251913d3b1ebSLawrence Brakmo * window (in packets) or -1 if default 252013d3b1ebSLawrence Brakmo * value should be used 252113d3b1ebSLawrence Brakmo */ 25229872a4bdSLawrence Brakmo BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an 25239872a4bdSLawrence Brakmo * active connection is initialized 25249872a4bdSLawrence Brakmo */ 25259872a4bdSLawrence Brakmo BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an 25269872a4bdSLawrence Brakmo * active connection is 25279872a4bdSLawrence Brakmo * established 25289872a4bdSLawrence Brakmo */ 25299872a4bdSLawrence Brakmo BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a 25309872a4bdSLawrence Brakmo * passive connection is 25319872a4bdSLawrence Brakmo * established 25329872a4bdSLawrence Brakmo */ 253391b5b21cSLawrence Brakmo BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control 253491b5b21cSLawrence Brakmo * needs ECN 253591b5b21cSLawrence Brakmo */ 2536e6546ef6SLawrence Brakmo BPF_SOCK_OPS_BASE_RTT, /* Get base RTT. The correct value is 2537e6546ef6SLawrence Brakmo * based on the path and may be 2538e6546ef6SLawrence Brakmo * dependent on the congestion control 2539e6546ef6SLawrence Brakmo * algorithm. In general it indicates 2540e6546ef6SLawrence Brakmo * a congestion threshold. RTTs above 2541e6546ef6SLawrence Brakmo * this indicate congestion 2542e6546ef6SLawrence Brakmo */ 2543f89013f6SLawrence Brakmo BPF_SOCK_OPS_RTO_CB, /* Called when an RTO has triggered. 2544f89013f6SLawrence Brakmo * Arg1: value of icsk_retransmits 2545f89013f6SLawrence Brakmo * Arg2: value of icsk_rto 2546f89013f6SLawrence Brakmo * Arg3: whether RTO has expired 2547f89013f6SLawrence Brakmo */ 2548a31ad29eSLawrence Brakmo BPF_SOCK_OPS_RETRANS_CB, /* Called when skb is retransmitted. 2549a31ad29eSLawrence Brakmo * Arg1: sequence number of 1st byte 2550a31ad29eSLawrence Brakmo * Arg2: # segments 2551a31ad29eSLawrence Brakmo * Arg3: return value of 2552a31ad29eSLawrence Brakmo * tcp_transmit_skb (0 => success) 2553a31ad29eSLawrence Brakmo */ 2554d4487491SLawrence Brakmo BPF_SOCK_OPS_STATE_CB, /* Called when TCP changes state. 2555d4487491SLawrence Brakmo * Arg1: old_state 2556d4487491SLawrence Brakmo * Arg2: new_state 2557d4487491SLawrence Brakmo */ 2558*f333ee0cSAndrey Ignatov BPF_SOCK_OPS_TCP_LISTEN_CB, /* Called on listen(2), right after 2559*f333ee0cSAndrey Ignatov * socket transition to LISTEN state. 2560*f333ee0cSAndrey Ignatov */ 2561d4487491SLawrence Brakmo }; 2562d4487491SLawrence Brakmo 2563d4487491SLawrence Brakmo /* List of TCP states. There is a build check in net/ipv4/tcp.c to detect 2564d4487491SLawrence Brakmo * changes between the TCP and BPF versions. Ideally this should never happen. 2565d4487491SLawrence Brakmo * If it does, we need to add code to convert them before calling 2566d4487491SLawrence Brakmo * the BPF sock_ops function. 2567d4487491SLawrence Brakmo */ 2568d4487491SLawrence Brakmo enum { 2569d4487491SLawrence Brakmo BPF_TCP_ESTABLISHED = 1, 2570d4487491SLawrence Brakmo BPF_TCP_SYN_SENT, 2571d4487491SLawrence Brakmo BPF_TCP_SYN_RECV, 2572d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT1, 2573d4487491SLawrence Brakmo BPF_TCP_FIN_WAIT2, 2574d4487491SLawrence Brakmo BPF_TCP_TIME_WAIT, 2575d4487491SLawrence Brakmo BPF_TCP_CLOSE, 2576d4487491SLawrence Brakmo BPF_TCP_CLOSE_WAIT, 2577d4487491SLawrence Brakmo BPF_TCP_LAST_ACK, 2578d4487491SLawrence Brakmo BPF_TCP_LISTEN, 2579d4487491SLawrence Brakmo BPF_TCP_CLOSING, /* Now a valid state */ 2580d4487491SLawrence Brakmo BPF_TCP_NEW_SYN_RECV, 2581d4487491SLawrence Brakmo 2582d4487491SLawrence Brakmo BPF_TCP_MAX_STATES /* Leave at the end! */ 258340304b2aSLawrence Brakmo }; 258440304b2aSLawrence Brakmo 2585fc747810SLawrence Brakmo #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ 258613bf9641SLawrence Brakmo #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ 2587fc747810SLawrence Brakmo 2588908432caSYonghong Song struct bpf_perf_event_value { 2589908432caSYonghong Song __u64 counter; 2590908432caSYonghong Song __u64 enabled; 2591908432caSYonghong Song __u64 running; 2592908432caSYonghong Song }; 2593908432caSYonghong Song 2594ebc614f6SRoman Gushchin #define BPF_DEVCG_ACC_MKNOD (1ULL << 0) 2595ebc614f6SRoman Gushchin #define BPF_DEVCG_ACC_READ (1ULL << 1) 2596ebc614f6SRoman Gushchin #define BPF_DEVCG_ACC_WRITE (1ULL << 2) 2597ebc614f6SRoman Gushchin 2598ebc614f6SRoman Gushchin #define BPF_DEVCG_DEV_BLOCK (1ULL << 0) 2599ebc614f6SRoman Gushchin #define BPF_DEVCG_DEV_CHAR (1ULL << 1) 2600ebc614f6SRoman Gushchin 2601ebc614f6SRoman Gushchin struct bpf_cgroup_dev_ctx { 260206ef0ccbSYonghong Song /* access_type encoded as (BPF_DEVCG_ACC_* << 16) | BPF_DEVCG_DEV_* */ 260306ef0ccbSYonghong Song __u32 access_type; 2604ebc614f6SRoman Gushchin __u32 major; 2605ebc614f6SRoman Gushchin __u32 minor; 2606ebc614f6SRoman Gushchin }; 2607ebc614f6SRoman Gushchin 2608c4f6699dSAlexei Starovoitov struct bpf_raw_tracepoint_args { 2609c4f6699dSAlexei Starovoitov __u64 args[0]; 2610c4f6699dSAlexei Starovoitov }; 2611c4f6699dSAlexei Starovoitov 261287f5fc7eSDavid Ahern /* DIRECT: Skip the FIB rules and go to FIB table associated with device 261387f5fc7eSDavid Ahern * OUTPUT: Do lookup from egress perspective; default is ingress 261487f5fc7eSDavid Ahern */ 261587f5fc7eSDavid Ahern #define BPF_FIB_LOOKUP_DIRECT BIT(0) 261687f5fc7eSDavid Ahern #define BPF_FIB_LOOKUP_OUTPUT BIT(1) 261787f5fc7eSDavid Ahern 26184c79579bSDavid Ahern enum { 26194c79579bSDavid Ahern BPF_FIB_LKUP_RET_SUCCESS, /* lookup successful */ 26204c79579bSDavid Ahern BPF_FIB_LKUP_RET_BLACKHOLE, /* dest is blackholed; can be dropped */ 26214c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNREACHABLE, /* dest is unreachable; can be dropped */ 26224c79579bSDavid Ahern BPF_FIB_LKUP_RET_PROHIBIT, /* dest not allowed; can be dropped */ 26234c79579bSDavid Ahern BPF_FIB_LKUP_RET_NOT_FWDED, /* packet is not forwarded */ 26244c79579bSDavid Ahern BPF_FIB_LKUP_RET_FWD_DISABLED, /* fwding is not enabled on ingress */ 26254c79579bSDavid Ahern BPF_FIB_LKUP_RET_UNSUPP_LWT, /* fwd requires encapsulation */ 26264c79579bSDavid Ahern BPF_FIB_LKUP_RET_NO_NEIGH, /* no neighbor entry for nh */ 26274c79579bSDavid Ahern BPF_FIB_LKUP_RET_FRAG_NEEDED, /* fragmentation required to fwd */ 26284c79579bSDavid Ahern }; 26294c79579bSDavid Ahern 263087f5fc7eSDavid Ahern struct bpf_fib_lookup { 2631fa898d76SDavid Ahern /* input: network family for lookup (AF_INET, AF_INET6) 2632fa898d76SDavid Ahern * output: network family of egress nexthop 2633fa898d76SDavid Ahern */ 2634fa898d76SDavid Ahern __u8 family; 263587f5fc7eSDavid Ahern 263687f5fc7eSDavid Ahern /* set if lookup is to consider L4 data - e.g., FIB rules */ 263787f5fc7eSDavid Ahern __u8 l4_protocol; 263887f5fc7eSDavid Ahern __be16 sport; 263987f5fc7eSDavid Ahern __be16 dport; 264087f5fc7eSDavid Ahern 264187f5fc7eSDavid Ahern /* total length of packet from network header - used for MTU check */ 264287f5fc7eSDavid Ahern __u16 tot_len; 26434c79579bSDavid Ahern 26444c79579bSDavid Ahern /* input: L3 device index for lookup 26454c79579bSDavid Ahern * output: device index from FIB lookup 26464c79579bSDavid Ahern */ 26474c79579bSDavid Ahern __u32 ifindex; 264887f5fc7eSDavid Ahern 264987f5fc7eSDavid Ahern union { 265087f5fc7eSDavid Ahern /* inputs to lookup */ 265187f5fc7eSDavid Ahern __u8 tos; /* AF_INET */ 2652bd3a08aaSDavid Ahern __be32 flowinfo; /* AF_INET6, flow_label + priority */ 265387f5fc7eSDavid Ahern 2654fa898d76SDavid Ahern /* output: metric of fib result (IPv4/IPv6 only) */ 265587f5fc7eSDavid Ahern __u32 rt_metric; 265687f5fc7eSDavid Ahern }; 265787f5fc7eSDavid Ahern 265887f5fc7eSDavid Ahern union { 265987f5fc7eSDavid Ahern __be32 ipv4_src; 266087f5fc7eSDavid Ahern __u32 ipv6_src[4]; /* in6_addr; network order */ 266187f5fc7eSDavid Ahern }; 266287f5fc7eSDavid Ahern 2663fa898d76SDavid Ahern /* input to bpf_fib_lookup, ipv{4,6}_dst is destination address in 2664fa898d76SDavid Ahern * network header. output: bpf_fib_lookup sets to gateway address 2665fa898d76SDavid Ahern * if FIB lookup returns gateway route 266687f5fc7eSDavid Ahern */ 266787f5fc7eSDavid Ahern union { 266887f5fc7eSDavid Ahern __be32 ipv4_dst; 266987f5fc7eSDavid Ahern __u32 ipv6_dst[4]; /* in6_addr; network order */ 267087f5fc7eSDavid Ahern }; 267187f5fc7eSDavid Ahern 267287f5fc7eSDavid Ahern /* output */ 267387f5fc7eSDavid Ahern __be16 h_vlan_proto; 267487f5fc7eSDavid Ahern __be16 h_vlan_TCI; 267587f5fc7eSDavid Ahern __u8 smac[6]; /* ETH_ALEN */ 267687f5fc7eSDavid Ahern __u8 dmac[6]; /* ETH_ALEN */ 267787f5fc7eSDavid Ahern }; 267887f5fc7eSDavid Ahern 267941bdc4b4SYonghong Song enum bpf_task_fd_type { 268041bdc4b4SYonghong Song BPF_FD_TYPE_RAW_TRACEPOINT, /* tp name */ 268141bdc4b4SYonghong Song BPF_FD_TYPE_TRACEPOINT, /* tp name */ 268241bdc4b4SYonghong Song BPF_FD_TYPE_KPROBE, /* (symbol + offset) or addr */ 268341bdc4b4SYonghong Song BPF_FD_TYPE_KRETPROBE, /* (symbol + offset) or addr */ 268441bdc4b4SYonghong Song BPF_FD_TYPE_UPROBE, /* filename + offset */ 268541bdc4b4SYonghong Song BPF_FD_TYPE_URETPROBE, /* filename + offset */ 268641bdc4b4SYonghong Song }; 268741bdc4b4SYonghong Song 2688daedfb22SAlexei Starovoitov #endif /* _UAPI__LINUX_BPF_H__ */ 2689