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 */ 40enum { 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 58struct 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 66/* BPF syscall commands, see bpf(2) man-page for details. */ 67enum bpf_cmd { 68 BPF_MAP_CREATE, 69 BPF_MAP_LOOKUP_ELEM, 70 BPF_MAP_UPDATE_ELEM, 71 BPF_MAP_DELETE_ELEM, 72 BPF_MAP_GET_NEXT_KEY, 73 BPF_PROG_LOAD, 74 BPF_OBJ_PIN, 75 BPF_OBJ_GET, 76}; 77 78enum bpf_map_type { 79 BPF_MAP_TYPE_UNSPEC, 80 BPF_MAP_TYPE_HASH, 81 BPF_MAP_TYPE_ARRAY, 82 BPF_MAP_TYPE_PROG_ARRAY, 83 BPF_MAP_TYPE_PERF_EVENT_ARRAY, 84 BPF_MAP_TYPE_PERCPU_HASH, 85 BPF_MAP_TYPE_PERCPU_ARRAY, 86 BPF_MAP_TYPE_STACK_TRACE, 87 BPF_MAP_TYPE_CGROUP_ARRAY, 88}; 89 90enum bpf_prog_type { 91 BPF_PROG_TYPE_UNSPEC, 92 BPF_PROG_TYPE_SOCKET_FILTER, 93 BPF_PROG_TYPE_KPROBE, 94 BPF_PROG_TYPE_SCHED_CLS, 95 BPF_PROG_TYPE_SCHED_ACT, 96 BPF_PROG_TYPE_TRACEPOINT, 97 BPF_PROG_TYPE_XDP, 98}; 99 100#define BPF_PSEUDO_MAP_FD 1 101 102/* flags for BPF_MAP_UPDATE_ELEM command */ 103#define BPF_ANY 0 /* create new element or update existing */ 104#define BPF_NOEXIST 1 /* create new element if it didn't exist */ 105#define BPF_EXIST 2 /* update existing element */ 106 107#define BPF_F_NO_PREALLOC (1U << 0) 108 109union bpf_attr { 110 struct { /* anonymous struct used by BPF_MAP_CREATE command */ 111 __u32 map_type; /* one of enum bpf_map_type */ 112 __u32 key_size; /* size of key in bytes */ 113 __u32 value_size; /* size of value in bytes */ 114 __u32 max_entries; /* max number of entries in a map */ 115 __u32 map_flags; /* prealloc or not */ 116 }; 117 118 struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */ 119 __u32 map_fd; 120 __aligned_u64 key; 121 union { 122 __aligned_u64 value; 123 __aligned_u64 next_key; 124 }; 125 __u64 flags; 126 }; 127 128 struct { /* anonymous struct used by BPF_PROG_LOAD command */ 129 __u32 prog_type; /* one of enum bpf_prog_type */ 130 __u32 insn_cnt; 131 __aligned_u64 insns; 132 __aligned_u64 license; 133 __u32 log_level; /* verbosity level of verifier */ 134 __u32 log_size; /* size of user buffer */ 135 __aligned_u64 log_buf; /* user supplied buffer */ 136 __u32 kern_version; /* checked when prog_type=kprobe */ 137 }; 138 139 struct { /* anonymous struct used by BPF_OBJ_* commands */ 140 __aligned_u64 pathname; 141 __u32 bpf_fd; 142 }; 143} __attribute__((aligned(8))); 144 145/* integer value in 'imm' field of BPF_CALL instruction selects which helper 146 * function eBPF program intends to call 147 */ 148enum bpf_func_id { 149 BPF_FUNC_unspec, 150 BPF_FUNC_map_lookup_elem, /* void *map_lookup_elem(&map, &key) */ 151 BPF_FUNC_map_update_elem, /* int map_update_elem(&map, &key, &value, flags) */ 152 BPF_FUNC_map_delete_elem, /* int map_delete_elem(&map, &key) */ 153 BPF_FUNC_probe_read, /* int bpf_probe_read(void *dst, int size, void *src) */ 154 BPF_FUNC_ktime_get_ns, /* u64 bpf_ktime_get_ns(void) */ 155 BPF_FUNC_trace_printk, /* int bpf_trace_printk(const char *fmt, int fmt_size, ...) */ 156 BPF_FUNC_get_prandom_u32, /* u32 prandom_u32(void) */ 157 BPF_FUNC_get_smp_processor_id, /* u32 raw_smp_processor_id(void) */ 158 159 /** 160 * skb_store_bytes(skb, offset, from, len, flags) - store bytes into packet 161 * @skb: pointer to skb 162 * @offset: offset within packet from skb->mac_header 163 * @from: pointer where to copy bytes from 164 * @len: number of bytes to store into packet 165 * @flags: bit 0 - if true, recompute skb->csum 166 * other bits - reserved 167 * Return: 0 on success 168 */ 169 BPF_FUNC_skb_store_bytes, 170 171 /** 172 * l3_csum_replace(skb, offset, from, to, flags) - recompute IP checksum 173 * @skb: pointer to skb 174 * @offset: offset within packet where IP checksum is located 175 * @from: old value of header field 176 * @to: new value of header field 177 * @flags: bits 0-3 - size of header field 178 * other bits - reserved 179 * Return: 0 on success 180 */ 181 BPF_FUNC_l3_csum_replace, 182 183 /** 184 * l4_csum_replace(skb, offset, from, to, flags) - recompute TCP/UDP checksum 185 * @skb: pointer to skb 186 * @offset: offset within packet where TCP/UDP checksum is located 187 * @from: old value of header field 188 * @to: new value of header field 189 * @flags: bits 0-3 - size of header field 190 * bit 4 - is pseudo header 191 * other bits - reserved 192 * Return: 0 on success 193 */ 194 BPF_FUNC_l4_csum_replace, 195 196 /** 197 * bpf_tail_call(ctx, prog_array_map, index) - jump into another BPF program 198 * @ctx: context pointer passed to next program 199 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY 200 * @index: index inside array that selects specific program to run 201 * Return: 0 on success 202 */ 203 BPF_FUNC_tail_call, 204 205 /** 206 * bpf_clone_redirect(skb, ifindex, flags) - redirect to another netdev 207 * @skb: pointer to skb 208 * @ifindex: ifindex of the net device 209 * @flags: bit 0 - if set, redirect to ingress instead of egress 210 * other bits - reserved 211 * Return: 0 on success 212 */ 213 BPF_FUNC_clone_redirect, 214 215 /** 216 * u64 bpf_get_current_pid_tgid(void) 217 * Return: current->tgid << 32 | current->pid 218 */ 219 BPF_FUNC_get_current_pid_tgid, 220 221 /** 222 * u64 bpf_get_current_uid_gid(void) 223 * Return: current_gid << 32 | current_uid 224 */ 225 BPF_FUNC_get_current_uid_gid, 226 227 /** 228 * bpf_get_current_comm(char *buf, int size_of_buf) 229 * stores current->comm into buf 230 * Return: 0 on success 231 */ 232 BPF_FUNC_get_current_comm, 233 234 /** 235 * bpf_get_cgroup_classid(skb) - retrieve a proc's classid 236 * @skb: pointer to skb 237 * Return: classid if != 0 238 */ 239 BPF_FUNC_get_cgroup_classid, 240 BPF_FUNC_skb_vlan_push, /* bpf_skb_vlan_push(skb, vlan_proto, vlan_tci) */ 241 BPF_FUNC_skb_vlan_pop, /* bpf_skb_vlan_pop(skb) */ 242 243 /** 244 * bpf_skb_[gs]et_tunnel_key(skb, key, size, flags) 245 * retrieve or populate tunnel metadata 246 * @skb: pointer to skb 247 * @key: pointer to 'struct bpf_tunnel_key' 248 * @size: size of 'struct bpf_tunnel_key' 249 * @flags: room for future extensions 250 * Retrun: 0 on success 251 */ 252 BPF_FUNC_skb_get_tunnel_key, 253 BPF_FUNC_skb_set_tunnel_key, 254 BPF_FUNC_perf_event_read, /* u64 bpf_perf_event_read(&map, index) */ 255 /** 256 * bpf_redirect(ifindex, flags) - redirect to another netdev 257 * @ifindex: ifindex of the net device 258 * @flags: bit 0 - if set, redirect to ingress instead of egress 259 * other bits - reserved 260 * Return: TC_ACT_REDIRECT 261 */ 262 BPF_FUNC_redirect, 263 264 /** 265 * bpf_get_route_realm(skb) - retrieve a dst's tclassid 266 * @skb: pointer to skb 267 * Return: realm if != 0 268 */ 269 BPF_FUNC_get_route_realm, 270 271 /** 272 * bpf_perf_event_output(ctx, map, index, data, size) - output perf raw sample 273 * @ctx: struct pt_regs* 274 * @map: pointer to perf_event_array map 275 * @index: index of event in the map 276 * @data: data on stack to be output as raw data 277 * @size: size of data 278 * Return: 0 on success 279 */ 280 BPF_FUNC_perf_event_output, 281 BPF_FUNC_skb_load_bytes, 282 283 /** 284 * bpf_get_stackid(ctx, map, flags) - walk user or kernel stack and return id 285 * @ctx: struct pt_regs* 286 * @map: pointer to stack_trace map 287 * @flags: bits 0-7 - numer of stack frames to skip 288 * bit 8 - collect user stack instead of kernel 289 * bit 9 - compare stacks by hash only 290 * bit 10 - if two different stacks hash into the same stackid 291 * discard old 292 * other bits - reserved 293 * Return: >= 0 stackid on success or negative error 294 */ 295 BPF_FUNC_get_stackid, 296 297 /** 298 * bpf_csum_diff(from, from_size, to, to_size, seed) - calculate csum diff 299 * @from: raw from buffer 300 * @from_size: length of from buffer 301 * @to: raw to buffer 302 * @to_size: length of to buffer 303 * @seed: optional seed 304 * Return: csum result 305 */ 306 BPF_FUNC_csum_diff, 307 308 /** 309 * bpf_skb_[gs]et_tunnel_opt(skb, opt, size) 310 * retrieve or populate tunnel options metadata 311 * @skb: pointer to skb 312 * @opt: pointer to raw tunnel option data 313 * @size: size of @opt 314 * Return: 0 on success for set, option size for get 315 */ 316 BPF_FUNC_skb_get_tunnel_opt, 317 BPF_FUNC_skb_set_tunnel_opt, 318 319 /** 320 * bpf_skb_change_proto(skb, proto, flags) 321 * Change protocol of the skb. Currently supported is 322 * v4 -> v6, v6 -> v4 transitions. The helper will also 323 * resize the skb. eBPF program is expected to fill the 324 * new headers via skb_store_bytes and lX_csum_replace. 325 * @skb: pointer to skb 326 * @proto: new skb->protocol type 327 * @flags: reserved 328 * Return: 0 on success or negative error 329 */ 330 BPF_FUNC_skb_change_proto, 331 332 /** 333 * bpf_skb_change_type(skb, type) 334 * Change packet type of skb. 335 * @skb: pointer to skb 336 * @type: new skb->pkt_type type 337 * Return: 0 on success or negative error 338 */ 339 BPF_FUNC_skb_change_type, 340 341 /** 342 * bpf_skb_under_cgroup(skb, map, index) - Check cgroup2 membership of skb 343 * @skb: pointer to skb 344 * @map: pointer to bpf_map in BPF_MAP_TYPE_CGROUP_ARRAY type 345 * @index: index of the cgroup in the bpf_map 346 * Return: 347 * == 0 skb failed the cgroup2 descendant test 348 * == 1 skb succeeded the cgroup2 descendant test 349 * < 0 error 350 */ 351 BPF_FUNC_skb_under_cgroup, 352 353 /** 354 * bpf_get_hash_recalc(skb) 355 * Retrieve and possibly recalculate skb->hash. 356 * @skb: pointer to skb 357 * Return: hash 358 */ 359 BPF_FUNC_get_hash_recalc, 360 361 /** 362 * u64 bpf_get_current_task(void) 363 * Returns current task_struct 364 * Return: current 365 */ 366 BPF_FUNC_get_current_task, 367 368 /** 369 * bpf_probe_write_user(void *dst, void *src, int len) 370 * safely attempt to write to a location 371 * @dst: destination address in userspace 372 * @src: source address on stack 373 * @len: number of bytes to copy 374 * Return: 0 on success or negative error 375 */ 376 BPF_FUNC_probe_write_user, 377 378 __BPF_FUNC_MAX_ID, 379}; 380 381/* All flags used by eBPF helper functions, placed here. */ 382 383/* BPF_FUNC_skb_store_bytes flags. */ 384#define BPF_F_RECOMPUTE_CSUM (1ULL << 0) 385#define BPF_F_INVALIDATE_HASH (1ULL << 1) 386 387/* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags. 388 * First 4 bits are for passing the header field size. 389 */ 390#define BPF_F_HDR_FIELD_MASK 0xfULL 391 392/* BPF_FUNC_l4_csum_replace flags. */ 393#define BPF_F_PSEUDO_HDR (1ULL << 4) 394#define BPF_F_MARK_MANGLED_0 (1ULL << 5) 395 396/* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */ 397#define BPF_F_INGRESS (1ULL << 0) 398 399/* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */ 400#define BPF_F_TUNINFO_IPV6 (1ULL << 0) 401 402/* BPF_FUNC_get_stackid flags. */ 403#define BPF_F_SKIP_FIELD_MASK 0xffULL 404#define BPF_F_USER_STACK (1ULL << 8) 405#define BPF_F_FAST_STACK_CMP (1ULL << 9) 406#define BPF_F_REUSE_STACKID (1ULL << 10) 407 408/* BPF_FUNC_skb_set_tunnel_key flags. */ 409#define BPF_F_ZERO_CSUM_TX (1ULL << 1) 410#define BPF_F_DONT_FRAGMENT (1ULL << 2) 411 412/* BPF_FUNC_perf_event_output and BPF_FUNC_perf_event_read flags. */ 413#define BPF_F_INDEX_MASK 0xffffffffULL 414#define BPF_F_CURRENT_CPU BPF_F_INDEX_MASK 415/* BPF_FUNC_perf_event_output for sk_buff input context. */ 416#define BPF_F_CTXLEN_MASK (0xfffffULL << 32) 417 418/* user accessible mirror of in-kernel sk_buff. 419 * new fields can only be added to the end of this structure 420 */ 421struct __sk_buff { 422 __u32 len; 423 __u32 pkt_type; 424 __u32 mark; 425 __u32 queue_mapping; 426 __u32 protocol; 427 __u32 vlan_present; 428 __u32 vlan_tci; 429 __u32 vlan_proto; 430 __u32 priority; 431 __u32 ingress_ifindex; 432 __u32 ifindex; 433 __u32 tc_index; 434 __u32 cb[5]; 435 __u32 hash; 436 __u32 tc_classid; 437 __u32 data; 438 __u32 data_end; 439}; 440 441struct bpf_tunnel_key { 442 __u32 tunnel_id; 443 union { 444 __u32 remote_ipv4; 445 __u32 remote_ipv6[4]; 446 }; 447 __u8 tunnel_tos; 448 __u8 tunnel_ttl; 449 __u16 tunnel_ext; 450 __u32 tunnel_label; 451}; 452 453/* User return codes for XDP prog type. 454 * A valid XDP program must return one of these defined values. All other 455 * return codes are reserved for future use. Unknown return codes will result 456 * in packet drop. 457 */ 458enum xdp_action { 459 XDP_ABORTED = 0, 460 XDP_DROP, 461 XDP_PASS, 462 XDP_TX, 463}; 464 465/* user accessible metadata for XDP packet hook 466 * new fields must be added to the end of this structure 467 */ 468struct xdp_md { 469 __u32 data; 470 __u32 data_end; 471}; 472 473#endif /* _UAPI__LINUX_BPF_H__ */ 474