Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 1 | /* Copyright (c) 2011-2014 PLUMgrid, http://plumgrid.com |
| 2 | * |
| 3 | * This program is free software; you can redistribute it and/or |
| 4 | * modify it under the terms of version 2 of the GNU General Public |
| 5 | * License as published by the Free Software Foundation. |
| 6 | */ |
| 7 | #ifndef _UAPI__LINUX_BPF_H__ |
| 8 | #define _UAPI__LINUX_BPF_H__ |
| 9 | |
| 10 | #include <linux/types.h> |
| 11 | #include <linux/bpf_common.h> |
| 12 | |
| 13 | /* Extended instruction set based on top of classic BPF */ |
| 14 | |
| 15 | /* instruction classes */ |
| 16 | #define BPF_ALU64 0x07 /* alu mode in double word width */ |
| 17 | |
| 18 | /* ld/ldx fields */ |
| 19 | #define BPF_DW 0x18 /* double word */ |
| 20 | #define BPF_XADD 0xc0 /* exclusive add */ |
| 21 | |
| 22 | /* alu/jmp fields */ |
| 23 | #define BPF_MOV 0xb0 /* mov reg to reg */ |
| 24 | #define BPF_ARSH 0xc0 /* sign extending arithmetic shift right */ |
| 25 | |
| 26 | /* change endianness of a register */ |
| 27 | #define BPF_END 0xd0 /* flags for endianness conversion: */ |
| 28 | #define BPF_TO_LE 0x00 /* convert to little-endian */ |
| 29 | #define BPF_TO_BE 0x08 /* convert to big-endian */ |
| 30 | #define BPF_FROM_LE BPF_TO_LE |
| 31 | #define BPF_FROM_BE BPF_TO_BE |
| 32 | |
| 33 | #define BPF_JNE 0x50 /* jump != */ |
| 34 | #define BPF_JSGT 0x60 /* SGT is signed '>', GT in x86 */ |
| 35 | #define BPF_JSGE 0x70 /* SGE is signed '>=', GE in x86 */ |
| 36 | #define BPF_CALL 0x80 /* function call */ |
| 37 | #define BPF_EXIT 0x90 /* function return */ |
| 38 | |
| 39 | /* Register numbers */ |
| 40 | enum { |
| 41 | BPF_REG_0 = 0, |
| 42 | BPF_REG_1, |
| 43 | BPF_REG_2, |
| 44 | BPF_REG_3, |
| 45 | BPF_REG_4, |
| 46 | BPF_REG_5, |
| 47 | BPF_REG_6, |
| 48 | BPF_REG_7, |
| 49 | BPF_REG_8, |
| 50 | BPF_REG_9, |
| 51 | BPF_REG_10, |
| 52 | __MAX_BPF_REG, |
| 53 | }; |
| 54 | |
| 55 | /* BPF has 10 general purpose 64-bit registers and stack frame. */ |
| 56 | #define MAX_BPF_REG __MAX_BPF_REG |
| 57 | |
| 58 | struct bpf_insn { |
| 59 | __u8 code; /* opcode */ |
| 60 | __u8 dst_reg:4; /* dest register */ |
| 61 | __u8 src_reg:4; /* source register */ |
| 62 | __s16 off; /* signed offset */ |
| 63 | __s32 imm; /* signed immediate constant */ |
| 64 | }; |
| 65 | |
Mickaël Salaün | 9a73826 | 2017-02-10 00:21:35 +0100 | [diff] [blame] | 66 | /* Key of an a BPF_MAP_TYPE_LPM_TRIE entry */ |
| 67 | struct bpf_lpm_trie_key { |
| 68 | __u32 prefixlen; /* up to 32 for AF_INET, 128 for AF_INET6 */ |
| 69 | __u8 data[0]; /* Arbitrary size */ |
| 70 | }; |
| 71 | |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 72 | /* BPF syscall commands, see bpf(2) man-page for details. */ |
| 73 | enum bpf_cmd { |
| 74 | BPF_MAP_CREATE, |
| 75 | BPF_MAP_LOOKUP_ELEM, |
| 76 | BPF_MAP_UPDATE_ELEM, |
| 77 | BPF_MAP_DELETE_ELEM, |
| 78 | BPF_MAP_GET_NEXT_KEY, |
| 79 | BPF_PROG_LOAD, |
| 80 | BPF_OBJ_PIN, |
| 81 | BPF_OBJ_GET, |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 82 | BPF_PROG_ATTACH, |
| 83 | BPF_PROG_DETACH, |
Alexei Starovoitov | 3084887 | 2017-03-30 21:45:39 -0700 | [diff] [blame] | 84 | BPF_PROG_TEST_RUN, |
Martin KaFai Lau | 95b9afd | 2017-06-05 12:15:53 -0700 | [diff] [blame] | 85 | BPF_PROG_GET_NEXT_ID, |
| 86 | BPF_MAP_GET_NEXT_ID, |
| 87 | BPF_PROG_GET_FD_BY_ID, |
| 88 | BPF_MAP_GET_FD_BY_ID, |
| 89 | BPF_OBJ_GET_INFO_BY_FD, |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 90 | }; |
| 91 | |
| 92 | enum bpf_map_type { |
| 93 | BPF_MAP_TYPE_UNSPEC, |
| 94 | BPF_MAP_TYPE_HASH, |
| 95 | BPF_MAP_TYPE_ARRAY, |
| 96 | BPF_MAP_TYPE_PROG_ARRAY, |
| 97 | BPF_MAP_TYPE_PERF_EVENT_ARRAY, |
| 98 | BPF_MAP_TYPE_PERCPU_HASH, |
| 99 | BPF_MAP_TYPE_PERCPU_ARRAY, |
| 100 | BPF_MAP_TYPE_STACK_TRACE, |
Arnaldo Carvalho de Melo | 791cceb | 2016-08-09 11:48:07 -0300 | [diff] [blame] | 101 | BPF_MAP_TYPE_CGROUP_ARRAY, |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 102 | BPF_MAP_TYPE_LRU_HASH, |
| 103 | BPF_MAP_TYPE_LRU_PERCPU_HASH, |
Mickaël Salaün | 9a73826 | 2017-02-10 00:21:35 +0100 | [diff] [blame] | 104 | BPF_MAP_TYPE_LPM_TRIE, |
Martin KaFai Lau | fb30d4b | 2017-03-22 10:00:35 -0700 | [diff] [blame] | 105 | BPF_MAP_TYPE_ARRAY_OF_MAPS, |
| 106 | BPF_MAP_TYPE_HASH_OF_MAPS, |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 107 | }; |
| 108 | |
| 109 | enum bpf_prog_type { |
| 110 | BPF_PROG_TYPE_UNSPEC, |
| 111 | BPF_PROG_TYPE_SOCKET_FILTER, |
| 112 | BPF_PROG_TYPE_KPROBE, |
| 113 | BPF_PROG_TYPE_SCHED_CLS, |
| 114 | BPF_PROG_TYPE_SCHED_ACT, |
| 115 | BPF_PROG_TYPE_TRACEPOINT, |
Arnaldo Carvalho de Melo | 791cceb | 2016-08-09 11:48:07 -0300 | [diff] [blame] | 116 | BPF_PROG_TYPE_XDP, |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 117 | BPF_PROG_TYPE_PERF_EVENT, |
| 118 | BPF_PROG_TYPE_CGROUP_SKB, |
| 119 | BPF_PROG_TYPE_CGROUP_SOCK, |
| 120 | BPF_PROG_TYPE_LWT_IN, |
| 121 | BPF_PROG_TYPE_LWT_OUT, |
| 122 | BPF_PROG_TYPE_LWT_XMIT, |
Lawrence Brakmo | 04df41e | 2017-06-30 20:02:55 -0700 | [diff] [blame] | 123 | BPF_PROG_TYPE_SOCK_OPS, |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 124 | }; |
| 125 | |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 126 | enum bpf_attach_type { |
| 127 | BPF_CGROUP_INET_INGRESS, |
| 128 | BPF_CGROUP_INET_EGRESS, |
| 129 | BPF_CGROUP_INET_SOCK_CREATE, |
Lawrence Brakmo | 04df41e | 2017-06-30 20:02:55 -0700 | [diff] [blame] | 130 | BPF_CGROUP_SOCK_OPS, |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 131 | __MAX_BPF_ATTACH_TYPE |
| 132 | }; |
| 133 | |
| 134 | #define MAX_BPF_ATTACH_TYPE __MAX_BPF_ATTACH_TYPE |
| 135 | |
Stephen Rothwell | 5463b3d | 2017-02-14 08:22:20 +1100 | [diff] [blame] | 136 | /* If BPF_F_ALLOW_OVERRIDE flag is used in BPF_PROG_ATTACH command |
| 137 | * to the given target_fd cgroup the descendent cgroup will be able to |
| 138 | * override effective bpf program that was inherited from this cgroup |
| 139 | */ |
| 140 | #define BPF_F_ALLOW_OVERRIDE (1U << 0) |
| 141 | |
David S. Miller | e07b98d | 2017-05-10 11:38:07 -0700 | [diff] [blame] | 142 | /* If BPF_F_STRICT_ALIGNMENT is used in BPF_PROG_LOAD command, the |
| 143 | * verifier will perform strict alignment checking as if the kernel |
| 144 | * has been built with CONFIG_EFFICIENT_UNALIGNED_ACCESS not set, |
| 145 | * and NET_IP_ALIGN defined to 2. |
| 146 | */ |
| 147 | #define BPF_F_STRICT_ALIGNMENT (1U << 0) |
| 148 | |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 149 | #define BPF_PSEUDO_MAP_FD 1 |
| 150 | |
| 151 | /* flags for BPF_MAP_UPDATE_ELEM command */ |
| 152 | #define BPF_ANY 0 /* create new element or update existing */ |
| 153 | #define BPF_NOEXIST 1 /* create new element if it didn't exist */ |
| 154 | #define BPF_EXIST 2 /* update existing element */ |
| 155 | |
| 156 | #define BPF_F_NO_PREALLOC (1U << 0) |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 157 | /* Instead of having one common LRU list in the |
| 158 | * BPF_MAP_TYPE_LRU_[PERCPU_]HASH map, use a percpu LRU list |
| 159 | * which can scale and perform better. |
| 160 | * Note, the LRU nodes (including free nodes) cannot be moved |
| 161 | * across different LRU lists. |
| 162 | */ |
| 163 | #define BPF_F_NO_COMMON_LRU (1U << 1) |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 164 | |
| 165 | union bpf_attr { |
| 166 | struct { /* anonymous struct used by BPF_MAP_CREATE command */ |
| 167 | __u32 map_type; /* one of enum bpf_map_type */ |
| 168 | __u32 key_size; /* size of key in bytes */ |
| 169 | __u32 value_size; /* size of value in bytes */ |
| 170 | __u32 max_entries; /* max number of entries in a map */ |
| 171 | __u32 map_flags; /* prealloc or not */ |
Martin KaFai Lau | fb30d4b | 2017-03-22 10:00:35 -0700 | [diff] [blame] | 172 | __u32 inner_map_fd; /* fd pointing to the inner map */ |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 173 | }; |
| 174 | |
| 175 | struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ |
| 176 | __u32 map_fd; |
| 177 | __aligned_u64 key; |
| 178 | union { |
| 179 | __aligned_u64 value; |
| 180 | __aligned_u64 next_key; |
| 181 | }; |
| 182 | __u64 flags; |
| 183 | }; |
| 184 | |
| 185 | struct { /* anonymous struct used by BPF_PROG_LOAD command */ |
| 186 | __u32 prog_type; /* one of enum bpf_prog_type */ |
| 187 | __u32 insn_cnt; |
| 188 | __aligned_u64 insns; |
| 189 | __aligned_u64 license; |
| 190 | __u32 log_level; /* verbosity level of verifier */ |
| 191 | __u32 log_size; /* size of user buffer */ |
| 192 | __aligned_u64 log_buf; /* user supplied buffer */ |
| 193 | __u32 kern_version; /* checked when prog_type=kprobe */ |
David S. Miller | e07b98d | 2017-05-10 11:38:07 -0700 | [diff] [blame] | 194 | __u32 prog_flags; |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 195 | }; |
| 196 | |
| 197 | struct { /* anonymous struct used by BPF_OBJ_* commands */ |
| 198 | __aligned_u64 pathname; |
| 199 | __u32 bpf_fd; |
| 200 | }; |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 201 | |
| 202 | struct { /* anonymous struct used by BPF_PROG_ATTACH/DETACH commands */ |
| 203 | __u32 target_fd; /* container object to attach to */ |
| 204 | __u32 attach_bpf_fd; /* eBPF program to attach */ |
| 205 | __u32 attach_type; |
Stephen Rothwell | 5463b3d | 2017-02-14 08:22:20 +1100 | [diff] [blame] | 206 | __u32 attach_flags; |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 207 | }; |
Alexei Starovoitov | 3084887 | 2017-03-30 21:45:39 -0700 | [diff] [blame] | 208 | |
| 209 | struct { /* anonymous struct used by BPF_PROG_TEST_RUN command */ |
| 210 | __u32 prog_fd; |
| 211 | __u32 retval; |
| 212 | __u32 data_size_in; |
| 213 | __u32 data_size_out; |
| 214 | __aligned_u64 data_in; |
| 215 | __aligned_u64 data_out; |
| 216 | __u32 repeat; |
| 217 | __u32 duration; |
| 218 | } test; |
Martin KaFai Lau | 95b9afd | 2017-06-05 12:15:53 -0700 | [diff] [blame] | 219 | |
| 220 | struct { /* anonymous struct used by BPF_*_GET_*_ID */ |
| 221 | union { |
| 222 | __u32 start_id; |
| 223 | __u32 prog_id; |
| 224 | __u32 map_id; |
| 225 | }; |
| 226 | __u32 next_id; |
| 227 | }; |
| 228 | |
| 229 | struct { /* anonymous struct used by BPF_OBJ_GET_INFO_BY_FD */ |
| 230 | __u32 bpf_fd; |
| 231 | __u32 info_len; |
| 232 | __aligned_u64 info; |
| 233 | } info; |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 234 | } __attribute__((aligned(8))); |
| 235 | |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 236 | /* BPF helper function descriptions: |
| 237 | * |
| 238 | * void *bpf_map_lookup_elem(&map, &key) |
| 239 | * Return: Map value or NULL |
| 240 | * |
| 241 | * int bpf_map_update_elem(&map, &key, &value, flags) |
| 242 | * Return: 0 on success or negative error |
| 243 | * |
| 244 | * int bpf_map_delete_elem(&map, &key) |
| 245 | * Return: 0 on success or negative error |
| 246 | * |
| 247 | * int bpf_probe_read(void *dst, int size, void *src) |
| 248 | * Return: 0 on success or negative error |
| 249 | * |
| 250 | * u64 bpf_ktime_get_ns(void) |
| 251 | * Return: current ktime |
| 252 | * |
| 253 | * int bpf_trace_printk(const char *fmt, int fmt_size, ...) |
| 254 | * Return: length of buffer written or negative error |
| 255 | * |
| 256 | * u32 bpf_prandom_u32(void) |
| 257 | * Return: random value |
| 258 | * |
| 259 | * u32 bpf_raw_smp_processor_id(void) |
| 260 | * Return: SMP processor ID |
| 261 | * |
| 262 | * int bpf_skb_store_bytes(skb, offset, from, len, flags) |
| 263 | * store bytes into packet |
| 264 | * @skb: pointer to skb |
| 265 | * @offset: offset within packet from skb->mac_header |
| 266 | * @from: pointer where to copy bytes from |
| 267 | * @len: number of bytes to store into packet |
| 268 | * @flags: bit 0 - if true, recompute skb->csum |
| 269 | * other bits - reserved |
| 270 | * Return: 0 on success or negative error |
| 271 | * |
| 272 | * int bpf_l3_csum_replace(skb, offset, from, to, flags) |
| 273 | * recompute IP checksum |
| 274 | * @skb: pointer to skb |
| 275 | * @offset: offset within packet where IP checksum is located |
| 276 | * @from: old value of header field |
| 277 | * @to: new value of header field |
| 278 | * @flags: bits 0-3 - size of header field |
| 279 | * other bits - reserved |
| 280 | * Return: 0 on success or negative error |
| 281 | * |
| 282 | * int bpf_l4_csum_replace(skb, offset, from, to, flags) |
| 283 | * recompute TCP/UDP checksum |
| 284 | * @skb: pointer to skb |
| 285 | * @offset: offset within packet where TCP/UDP checksum is located |
| 286 | * @from: old value of header field |
| 287 | * @to: new value of header field |
| 288 | * @flags: bits 0-3 - size of header field |
| 289 | * bit 4 - is pseudo header |
| 290 | * other bits - reserved |
| 291 | * Return: 0 on success or negative error |
| 292 | * |
| 293 | * int bpf_tail_call(ctx, prog_array_map, index) |
| 294 | * jump into another BPF program |
| 295 | * @ctx: context pointer passed to next program |
| 296 | * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY |
| 297 | * @index: index inside array that selects specific program to run |
| 298 | * Return: 0 on success or negative error |
| 299 | * |
| 300 | * int bpf_clone_redirect(skb, ifindex, flags) |
| 301 | * redirect to another netdev |
| 302 | * @skb: pointer to skb |
| 303 | * @ifindex: ifindex of the net device |
| 304 | * @flags: bit 0 - if set, redirect to ingress instead of egress |
| 305 | * other bits - reserved |
| 306 | * Return: 0 on success or negative error |
| 307 | * |
| 308 | * u64 bpf_get_current_pid_tgid(void) |
| 309 | * Return: current->tgid << 32 | current->pid |
| 310 | * |
| 311 | * u64 bpf_get_current_uid_gid(void) |
| 312 | * Return: current_gid << 32 | current_uid |
| 313 | * |
| 314 | * int bpf_get_current_comm(char *buf, int size_of_buf) |
| 315 | * stores current->comm into buf |
| 316 | * Return: 0 on success or negative error |
| 317 | * |
| 318 | * u32 bpf_get_cgroup_classid(skb) |
| 319 | * retrieve a proc's classid |
| 320 | * @skb: pointer to skb |
| 321 | * Return: classid if != 0 |
| 322 | * |
| 323 | * int bpf_skb_vlan_push(skb, vlan_proto, vlan_tci) |
| 324 | * Return: 0 on success or negative error |
| 325 | * |
| 326 | * int bpf_skb_vlan_pop(skb) |
| 327 | * Return: 0 on success or negative error |
| 328 | * |
| 329 | * int bpf_skb_get_tunnel_key(skb, key, size, flags) |
| 330 | * int bpf_skb_set_tunnel_key(skb, key, size, flags) |
| 331 | * retrieve or populate tunnel metadata |
| 332 | * @skb: pointer to skb |
| 333 | * @key: pointer to 'struct bpf_tunnel_key' |
| 334 | * @size: size of 'struct bpf_tunnel_key' |
| 335 | * @flags: room for future extensions |
| 336 | * Return: 0 on success or negative error |
| 337 | * |
Teng Qin | b7d3ed5 | 2017-06-02 21:03:54 -0700 | [diff] [blame] | 338 | * u64 bpf_perf_event_read(map, flags) |
| 339 | * read perf event counter value |
| 340 | * @map: pointer to perf_event_array map |
| 341 | * @flags: index of event in the map or bitmask flags |
| 342 | * Return: value of perf event counter read or error code |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 343 | * |
| 344 | * int bpf_redirect(ifindex, flags) |
| 345 | * redirect to another netdev |
| 346 | * @ifindex: ifindex of the net device |
| 347 | * @flags: bit 0 - if set, redirect to ingress instead of egress |
| 348 | * other bits - reserved |
| 349 | * Return: TC_ACT_REDIRECT |
| 350 | * |
| 351 | * u32 bpf_get_route_realm(skb) |
| 352 | * retrieve a dst's tclassid |
| 353 | * @skb: pointer to skb |
| 354 | * Return: realm if != 0 |
| 355 | * |
Teng Qin | b7d3ed5 | 2017-06-02 21:03:54 -0700 | [diff] [blame] | 356 | * int bpf_perf_event_output(ctx, map, flags, data, size) |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 357 | * output perf raw sample |
| 358 | * @ctx: struct pt_regs* |
| 359 | * @map: pointer to perf_event_array map |
Teng Qin | b7d3ed5 | 2017-06-02 21:03:54 -0700 | [diff] [blame] | 360 | * @flags: index of event in the map or bitmask flags |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 361 | * @data: data on stack to be output as raw data |
| 362 | * @size: size of data |
| 363 | * Return: 0 on success or negative error |
| 364 | * |
| 365 | * int bpf_get_stackid(ctx, map, flags) |
| 366 | * walk user or kernel stack and return id |
| 367 | * @ctx: struct pt_regs* |
| 368 | * @map: pointer to stack_trace map |
| 369 | * @flags: bits 0-7 - numer of stack frames to skip |
| 370 | * bit 8 - collect user stack instead of kernel |
| 371 | * bit 9 - compare stacks by hash only |
| 372 | * bit 10 - if two different stacks hash into the same stackid |
| 373 | * discard old |
| 374 | * other bits - reserved |
| 375 | * Return: >= 0 stackid on success or negative error |
| 376 | * |
| 377 | * s64 bpf_csum_diff(from, from_size, to, to_size, seed) |
| 378 | * calculate csum diff |
| 379 | * @from: raw from buffer |
| 380 | * @from_size: length of from buffer |
| 381 | * @to: raw to buffer |
| 382 | * @to_size: length of to buffer |
| 383 | * @seed: optional seed |
| 384 | * Return: csum result or negative error code |
| 385 | * |
| 386 | * int bpf_skb_get_tunnel_opt(skb, opt, size) |
| 387 | * retrieve tunnel options metadata |
| 388 | * @skb: pointer to skb |
| 389 | * @opt: pointer to raw tunnel option data |
| 390 | * @size: size of @opt |
| 391 | * Return: option size |
| 392 | * |
| 393 | * int bpf_skb_set_tunnel_opt(skb, opt, size) |
| 394 | * populate tunnel options metadata |
| 395 | * @skb: pointer to skb |
| 396 | * @opt: pointer to raw tunnel option data |
| 397 | * @size: size of @opt |
| 398 | * Return: 0 on success or negative error |
| 399 | * |
| 400 | * int bpf_skb_change_proto(skb, proto, flags) |
| 401 | * Change protocol of the skb. Currently supported is v4 -> v6, |
| 402 | * v6 -> v4 transitions. The helper will also resize the skb. eBPF |
| 403 | * program is expected to fill the new headers via skb_store_bytes |
| 404 | * and lX_csum_replace. |
| 405 | * @skb: pointer to skb |
| 406 | * @proto: new skb->protocol type |
| 407 | * @flags: reserved |
| 408 | * Return: 0 on success or negative error |
| 409 | * |
| 410 | * int bpf_skb_change_type(skb, type) |
| 411 | * Change packet type of skb. |
| 412 | * @skb: pointer to skb |
| 413 | * @type: new skb->pkt_type type |
| 414 | * Return: 0 on success or negative error |
| 415 | * |
| 416 | * int bpf_skb_under_cgroup(skb, map, index) |
| 417 | * Check cgroup2 membership of skb |
| 418 | * @skb: pointer to skb |
| 419 | * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type |
| 420 | * @index: index of the cgroup in the bpf_map |
| 421 | * Return: |
| 422 | * == 0 skb failed the cgroup2 descendant test |
| 423 | * == 1 skb succeeded the cgroup2 descendant test |
| 424 | * < 0 error |
| 425 | * |
| 426 | * u32 bpf_get_hash_recalc(skb) |
| 427 | * Retrieve and possibly recalculate skb->hash. |
| 428 | * @skb: pointer to skb |
| 429 | * Return: hash |
| 430 | * |
| 431 | * u64 bpf_get_current_task(void) |
| 432 | * Returns current task_struct |
| 433 | * Return: current |
| 434 | * |
| 435 | * int bpf_probe_write_user(void *dst, void *src, int len) |
| 436 | * safely attempt to write to a location |
| 437 | * @dst: destination address in userspace |
| 438 | * @src: source address on stack |
| 439 | * @len: number of bytes to copy |
| 440 | * Return: 0 on success or negative error |
| 441 | * |
| 442 | * int bpf_current_task_under_cgroup(map, index) |
| 443 | * Check cgroup2 membership of current task |
| 444 | * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type |
| 445 | * @index: index of the cgroup in the bpf_map |
| 446 | * Return: |
| 447 | * == 0 current failed the cgroup2 descendant test |
| 448 | * == 1 current succeeded the cgroup2 descendant test |
| 449 | * < 0 error |
| 450 | * |
| 451 | * int bpf_skb_change_tail(skb, len, flags) |
| 452 | * The helper will resize the skb to the given new size, to be used f.e. |
| 453 | * with control messages. |
| 454 | * @skb: pointer to skb |
| 455 | * @len: new skb length |
| 456 | * @flags: reserved |
| 457 | * Return: 0 on success or negative error |
| 458 | * |
| 459 | * int bpf_skb_pull_data(skb, len) |
| 460 | * The helper will pull in non-linear data in case the skb is non-linear |
| 461 | * and not all of len are part of the linear section. Only needed for |
| 462 | * read/write with direct packet access. |
| 463 | * @skb: pointer to skb |
| 464 | * @len: len to make read/writeable |
| 465 | * Return: 0 on success or negative error |
| 466 | * |
| 467 | * s64 bpf_csum_update(skb, csum) |
| 468 | * Adds csum into skb->csum in case of CHECKSUM_COMPLETE. |
| 469 | * @skb: pointer to skb |
| 470 | * @csum: csum to add |
| 471 | * Return: csum on success or negative error |
| 472 | * |
| 473 | * void bpf_set_hash_invalid(skb) |
| 474 | * Invalidate current skb->hash. |
| 475 | * @skb: pointer to skb |
| 476 | * |
| 477 | * int bpf_get_numa_node_id() |
| 478 | * Return: Id of current NUMA node. |
| 479 | * |
| 480 | * int bpf_skb_change_head() |
| 481 | * Grows headroom of skb and adjusts MAC header offset accordingly. |
| 482 | * Will extends/reallocae as required automatically. |
| 483 | * May change skb data pointer and will thus invalidate any check |
| 484 | * performed for direct packet access. |
| 485 | * @skb: pointer to skb |
| 486 | * @len: length of header to be pushed in front |
| 487 | * @flags: Flags (unused for now) |
| 488 | * Return: 0 on success or negative error |
| 489 | * |
| 490 | * int bpf_xdp_adjust_head(xdp_md, delta) |
| 491 | * Adjust the xdp_md.data by delta |
| 492 | * @xdp_md: pointer to xdp_md |
| 493 | * @delta: An positive/negative integer to be added to xdp_md.data |
| 494 | * Return: 0 on success or negative on error |
Mickaël Salaün | 9a73826 | 2017-02-10 00:21:35 +0100 | [diff] [blame] | 495 | * |
| 496 | * int bpf_probe_read_str(void *dst, int size, const void *unsafe_ptr) |
| 497 | * Copy a NUL terminated string from unsafe address. In case the string |
| 498 | * length is smaller than size, the target is not padded with further NUL |
| 499 | * bytes. In case the string length is larger than size, just count-1 |
| 500 | * bytes are copied and the last byte is set to NUL. |
| 501 | * @dst: destination address |
| 502 | * @size: maximum number of bytes to copy, including the trailing NUL |
| 503 | * @unsafe_ptr: unsafe address |
| 504 | * Return: |
| 505 | * > 0 length of the string including the trailing NUL on success |
| 506 | * < 0 error |
Alexei Starovoitov | 3084887 | 2017-03-30 21:45:39 -0700 | [diff] [blame] | 507 | * |
Alexander Alemayhu | 3c60a53 | 2017-04-08 22:08:10 +0200 | [diff] [blame] | 508 | * u64 bpf_get_socket_cookie(skb) |
Alexei Starovoitov | 3084887 | 2017-03-30 21:45:39 -0700 | [diff] [blame] | 509 | * Get the cookie for the socket stored inside sk_buff. |
| 510 | * @skb: pointer to skb |
| 511 | * Return: 8 Bytes non-decreasing number on success or 0 if the socket |
| 512 | * field is missing inside sk_buff |
| 513 | * |
| 514 | * u32 bpf_get_socket_uid(skb) |
| 515 | * Get the owner uid of the socket stored inside sk_buff. |
| 516 | * @skb: pointer to skb |
David S. Miller | e07b98d | 2017-05-10 11:38:07 -0700 | [diff] [blame] | 517 | * Return: uid of the socket owner on success or overflowuid if failed. |
Daniel Borkmann | ded092c | 2017-06-11 00:50:47 +0200 | [diff] [blame] | 518 | * |
| 519 | * u32 bpf_set_hash(skb, hash) |
| 520 | * Set full skb->hash. |
| 521 | * @skb: pointer to skb |
| 522 | * @hash: hash to set |
Lawrence Brakmo | 04df41e | 2017-06-30 20:02:55 -0700 | [diff] [blame] | 523 | * |
| 524 | * int bpf_setsockopt(bpf_socket, level, optname, optval, optlen) |
| 525 | * Calls setsockopt. Not all opts are available, only those with |
| 526 | * integer optvals plus TCP_CONGESTION. |
| 527 | * Supported levels: SOL_SOCKET and IPROTO_TCP |
| 528 | * @bpf_socket: pointer to bpf_socket |
| 529 | * @level: SOL_SOCKET or IPROTO_TCP |
| 530 | * @optname: option name |
| 531 | * @optval: pointer to option value |
| 532 | * @optlen: length of optval in byes |
| 533 | * Return: 0 or negative error |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 534 | */ |
| 535 | #define __BPF_FUNC_MAPPER(FN) \ |
| 536 | FN(unspec), \ |
| 537 | FN(map_lookup_elem), \ |
| 538 | FN(map_update_elem), \ |
| 539 | FN(map_delete_elem), \ |
| 540 | FN(probe_read), \ |
| 541 | FN(ktime_get_ns), \ |
| 542 | FN(trace_printk), \ |
| 543 | FN(get_prandom_u32), \ |
| 544 | FN(get_smp_processor_id), \ |
| 545 | FN(skb_store_bytes), \ |
| 546 | FN(l3_csum_replace), \ |
| 547 | FN(l4_csum_replace), \ |
| 548 | FN(tail_call), \ |
| 549 | FN(clone_redirect), \ |
| 550 | FN(get_current_pid_tgid), \ |
| 551 | FN(get_current_uid_gid), \ |
| 552 | FN(get_current_comm), \ |
| 553 | FN(get_cgroup_classid), \ |
| 554 | FN(skb_vlan_push), \ |
| 555 | FN(skb_vlan_pop), \ |
| 556 | FN(skb_get_tunnel_key), \ |
| 557 | FN(skb_set_tunnel_key), \ |
| 558 | FN(perf_event_read), \ |
| 559 | FN(redirect), \ |
| 560 | FN(get_route_realm), \ |
| 561 | FN(perf_event_output), \ |
| 562 | FN(skb_load_bytes), \ |
| 563 | FN(get_stackid), \ |
| 564 | FN(csum_diff), \ |
| 565 | FN(skb_get_tunnel_opt), \ |
| 566 | FN(skb_set_tunnel_opt), \ |
| 567 | FN(skb_change_proto), \ |
| 568 | FN(skb_change_type), \ |
| 569 | FN(skb_under_cgroup), \ |
| 570 | FN(get_hash_recalc), \ |
| 571 | FN(get_current_task), \ |
| 572 | FN(probe_write_user), \ |
| 573 | FN(current_task_under_cgroup), \ |
| 574 | FN(skb_change_tail), \ |
| 575 | FN(skb_pull_data), \ |
| 576 | FN(csum_update), \ |
| 577 | FN(set_hash_invalid), \ |
| 578 | FN(get_numa_node_id), \ |
| 579 | FN(skb_change_head), \ |
Mickaël Salaün | 9a73826 | 2017-02-10 00:21:35 +0100 | [diff] [blame] | 580 | FN(xdp_adjust_head), \ |
Chenbo Feng | 91b8270 | 2017-03-22 17:27:34 -0700 | [diff] [blame] | 581 | FN(probe_read_str), \ |
Chenbo Feng | 6acc5c2 | 2017-03-22 17:27:35 -0700 | [diff] [blame] | 582 | FN(get_socket_cookie), \ |
Daniel Borkmann | ded092c | 2017-06-11 00:50:47 +0200 | [diff] [blame] | 583 | FN(get_socket_uid), \ |
Lawrence Brakmo | 04df41e | 2017-06-30 20:02:55 -0700 | [diff] [blame] | 584 | FN(set_hash), \ |
| 585 | FN(setsockopt), |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 586 | |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 587 | /* integer value in 'imm' field of BPF_CALL instruction selects which helper |
| 588 | * function eBPF program intends to call |
| 589 | */ |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 590 | #define __BPF_ENUM_FN(x) BPF_FUNC_ ## x |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 591 | enum bpf_func_id { |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 592 | __BPF_FUNC_MAPPER(__BPF_ENUM_FN) |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 593 | __BPF_FUNC_MAX_ID, |
| 594 | }; |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 595 | #undef __BPF_ENUM_FN |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 596 | |
| 597 | /* All flags used by eBPF helper functions, placed here. */ |
| 598 | |
| 599 | /* BPF_FUNC_skb_store_bytes flags. */ |
| 600 | #define BPF_F_RECOMPUTE_CSUM (1ULL << 0) |
| 601 | #define BPF_F_INVALIDATE_HASH (1ULL << 1) |
| 602 | |
| 603 | /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. |
| 604 | * First 4 bits are for passing the header field size. |
| 605 | */ |
| 606 | #define BPF_F_HDR_FIELD_MASK 0xfULL |
| 607 | |
| 608 | /* BPF_FUNC_l4_csum_replace flags. */ |
| 609 | #define BPF_F_PSEUDO_HDR (1ULL << 4) |
| 610 | #define BPF_F_MARK_MANGLED_0 (1ULL << 5) |
Mickaël Salaün | 9a73826 | 2017-02-10 00:21:35 +0100 | [diff] [blame] | 611 | #define BPF_F_MARK_ENFORCE (1ULL << 6) |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 612 | |
| 613 | /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ |
| 614 | #define BPF_F_INGRESS (1ULL << 0) |
| 615 | |
| 616 | /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ |
| 617 | #define BPF_F_TUNINFO_IPV6 (1ULL << 0) |
| 618 | |
| 619 | /* BPF_FUNC_get_stackid flags. */ |
| 620 | #define BPF_F_SKIP_FIELD_MASK 0xffULL |
| 621 | #define BPF_F_USER_STACK (1ULL << 8) |
| 622 | #define BPF_F_FAST_STACK_CMP (1ULL << 9) |
| 623 | #define BPF_F_REUSE_STACKID (1ULL << 10) |
| 624 | |
| 625 | /* BPF_FUNC_skb_set_tunnel_key flags. */ |
| 626 | #define BPF_F_ZERO_CSUM_TX (1ULL << 1) |
| 627 | #define BPF_F_DONT_FRAGMENT (1ULL << 2) |
| 628 | |
Arnaldo Carvalho de Melo | 791cceb | 2016-08-09 11:48:07 -0300 | [diff] [blame] | 629 | /* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */ |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 630 | #define BPF_F_INDEX_MASK 0xffffffffULL |
| 631 | #define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK |
Arnaldo Carvalho de Melo | 791cceb | 2016-08-09 11:48:07 -0300 | [diff] [blame] | 632 | /* BPF_FUNC_perf_event_output for sk_buff input context. */ |
| 633 | #define BPF_F_CTXLEN_MASK (0xfffffULL << 32) |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 634 | |
| 635 | /* user accessible mirror of in-kernel sk_buff. |
| 636 | * new fields can only be added to the end of this structure |
| 637 | */ |
| 638 | struct __sk_buff { |
| 639 | __u32 len; |
| 640 | __u32 pkt_type; |
| 641 | __u32 mark; |
| 642 | __u32 queue_mapping; |
| 643 | __u32 protocol; |
| 644 | __u32 vlan_present; |
| 645 | __u32 vlan_tci; |
| 646 | __u32 vlan_proto; |
| 647 | __u32 priority; |
| 648 | __u32 ingress_ifindex; |
| 649 | __u32 ifindex; |
| 650 | __u32 tc_index; |
| 651 | __u32 cb[5]; |
| 652 | __u32 hash; |
| 653 | __u32 tc_classid; |
| 654 | __u32 data; |
| 655 | __u32 data_end; |
Daniel Borkmann | b1d9fc4 | 2017-04-19 23:01:17 +0200 | [diff] [blame] | 656 | __u32 napi_id; |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 657 | }; |
| 658 | |
| 659 | struct bpf_tunnel_key { |
| 660 | __u32 tunnel_id; |
| 661 | union { |
| 662 | __u32 remote_ipv4; |
| 663 | __u32 remote_ipv6[4]; |
| 664 | }; |
| 665 | __u8 tunnel_tos; |
| 666 | __u8 tunnel_ttl; |
| 667 | __u16 tunnel_ext; |
| 668 | __u32 tunnel_label; |
| 669 | }; |
| 670 | |
Joe Stringer | 0cb34dc | 2016-12-08 18:46:14 -0800 | [diff] [blame] | 671 | /* Generic BPF return codes which all BPF program types may support. |
| 672 | * The values are binary compatible with their TC_ACT_* counter-part to |
| 673 | * provide backwards compatibility with existing SCHED_CLS and SCHED_ACT |
| 674 | * programs. |
| 675 | * |
| 676 | * XDP is handled seprately, see XDP_*. |
| 677 | */ |
| 678 | enum bpf_ret_code { |
| 679 | BPF_OK = 0, |
| 680 | /* 1 reserved */ |
| 681 | BPF_DROP = 2, |
| 682 | /* 3-6 reserved */ |
| 683 | BPF_REDIRECT = 7, |
| 684 | /* >127 are reserved for prog type specific return codes */ |
| 685 | }; |
| 686 | |
| 687 | struct bpf_sock { |
| 688 | __u32 bound_dev_if; |
| 689 | __u32 family; |
| 690 | __u32 type; |
| 691 | __u32 protocol; |
| 692 | }; |
| 693 | |
| 694 | #define XDP_PACKET_HEADROOM 256 |
| 695 | |
Arnaldo Carvalho de Melo | 791cceb | 2016-08-09 11:48:07 -0300 | [diff] [blame] | 696 | /* User return codes for XDP prog type. |
| 697 | * A valid XDP program must return one of these defined values. All other |
| 698 | * return codes are reserved for future use. Unknown return codes will result |
| 699 | * in packet drop. |
| 700 | */ |
| 701 | enum xdp_action { |
| 702 | XDP_ABORTED = 0, |
| 703 | XDP_DROP, |
| 704 | XDP_PASS, |
| 705 | XDP_TX, |
| 706 | }; |
| 707 | |
| 708 | /* user accessible metadata for XDP packet hook |
| 709 | * new fields must be added to the end of this structure |
| 710 | */ |
| 711 | struct xdp_md { |
| 712 | __u32 data; |
| 713 | __u32 data_end; |
| 714 | }; |
| 715 | |
Martin KaFai Lau | 95b9afd | 2017-06-05 12:15:53 -0700 | [diff] [blame] | 716 | #define BPF_TAG_SIZE 8 |
| 717 | |
| 718 | struct bpf_prog_info { |
| 719 | __u32 type; |
| 720 | __u32 id; |
| 721 | __u8 tag[BPF_TAG_SIZE]; |
| 722 | __u32 jited_prog_len; |
| 723 | __u32 xlated_prog_len; |
| 724 | __aligned_u64 jited_prog_insns; |
| 725 | __aligned_u64 xlated_prog_insns; |
| 726 | } __attribute__((aligned(8))); |
| 727 | |
| 728 | struct bpf_map_info { |
| 729 | __u32 type; |
| 730 | __u32 id; |
| 731 | __u32 key_size; |
| 732 | __u32 value_size; |
| 733 | __u32 max_entries; |
| 734 | __u32 map_flags; |
| 735 | } __attribute__((aligned(8))); |
| 736 | |
Lawrence Brakmo | 04df41e | 2017-06-30 20:02:55 -0700 | [diff] [blame] | 737 | /* User bpf_sock_ops struct to access socket values and specify request ops |
| 738 | * and their replies. |
| 739 | * New fields can only be added at the end of this structure |
| 740 | */ |
| 741 | struct bpf_sock_ops { |
| 742 | __u32 op; |
| 743 | union { |
| 744 | __u32 reply; |
| 745 | __u32 replylong[4]; |
| 746 | }; |
| 747 | __u32 family; |
| 748 | __u32 remote_ip4; |
| 749 | __u32 local_ip4; |
| 750 | __u32 remote_ip6[4]; |
| 751 | __u32 local_ip6[4]; |
| 752 | __u32 remote_port; |
| 753 | __u32 local_port; |
| 754 | }; |
| 755 | |
| 756 | /* List of known BPF sock_ops operators. |
| 757 | * New entries can only be added at the end |
| 758 | */ |
| 759 | enum { |
| 760 | BPF_SOCK_OPS_VOID, |
| 761 | BPF_SOCK_OPS_TIMEOUT_INIT, /* Should return SYN-RTO value to use or |
| 762 | * -1 if default value should be used |
| 763 | */ |
| 764 | BPF_SOCK_OPS_RWND_INIT, /* Should return initial advertized |
| 765 | * window (in packets) or -1 if default |
| 766 | * value should be used |
| 767 | */ |
| 768 | BPF_SOCK_OPS_TCP_CONNECT_CB, /* Calls BPF program right before an |
| 769 | * active connection is initialized |
| 770 | */ |
| 771 | BPF_SOCK_OPS_ACTIVE_ESTABLISHED_CB, /* Calls BPF program when an |
| 772 | * active connection is |
| 773 | * established |
| 774 | */ |
| 775 | BPF_SOCK_OPS_PASSIVE_ESTABLISHED_CB, /* Calls BPF program when a |
| 776 | * passive connection is |
| 777 | * established |
| 778 | */ |
| 779 | BPF_SOCK_OPS_NEEDS_ECN, /* If connection's congestion control |
| 780 | * needs ECN |
| 781 | */ |
| 782 | }; |
| 783 | |
| 784 | #define TCP_BPF_IW 1001 /* Set TCP initial congestion window */ |
| 785 | #define TCP_BPF_SNDCWND_CLAMP 1002 /* Set sndcwnd_clamp */ |
| 786 | |
Arnaldo Carvalho de Melo | 971e827 | 2016-07-11 16:38:05 -0300 | [diff] [blame] | 787 | #endif /* _UAPI__LINUX_BPF_H__ */ |