include/linux/bpf.h

   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 __LINUX_BPF_H__
   8 #define __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
  66 /* BPF syscall commands, see bpf(2) man-page for details. */
  67 enum 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
  78 enum 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 };
  88
  89 enum bpf_prog_type {
  90         BPF_PROG_TYPE_UNSPEC,
  91         BPF_PROG_TYPE_SOCKET_FILTER,
  92         BPF_PROG_TYPE_KPROBE,
  93         BPF_PROG_TYPE_SCHED_CLS,
  94         BPF_PROG_TYPE_SCHED_ACT,
  95         BPF_PROG_TYPE_TRACEPOINT,
  96 };
  97
  98 #define BPF_PSEUDO_MAP_FD       1
  99
 100 /* flags for BPF_MAP_UPDATE_ELEM command */
 101 #define BPF_ANY         0 /* create new element or update existing */
 102 #define BPF_NOEXIST     1 /* create new element if it didn't exist */
 103 #define BPF_EXIST       2 /* update existing element */
 104
 105 #define BPF_F_NO_PREALLOC       (1U << 0)
 106
 107 union bpf_attr {
 108         struct { /* anonymous struct used by BPF_MAP_CREATE command */
 109                 __u32   map_type;       /* one of enum bpf_map_type */
 110                 __u32   key_size;       /* size of key in bytes */
 111                 __u32   value_size;     /* size of value in bytes */
 112                 __u32   max_entries;    /* max number of entries in a map */
 113                 __u32   map_flags;      /* prealloc or not */
 114         };
 115
 116         struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
 117                 __u32           map_fd;
 118                 __aligned_u64   key;
 119                 union {
 120                         __aligned_u64 value;
 121                         __aligned_u64 next_key;
 122                 };
 123                 __u64           flags;
 124         };
 125
 126         struct { /* anonymous struct used by BPF_PROG_LOAD command */
 127                 __u32           prog_type;      /* one of enum bpf_prog_type */
 128                 __u32           insn_cnt;
 129                 __aligned_u64   insns;
 130                 __aligned_u64   license;
 131                 __u32           log_level;      /* verbosity level of verifier */
 132                 __u32           log_size;       /* size of user buffer */
 133                 __aligned_u64   log_buf;        /* user supplied buffer */
 134                 __u32           kern_version;   /* checked when prog_type=kprobe */
 135         };
 136
 137         struct { /* anonymous struct used by BPF_OBJ_* commands */
 138                 __aligned_u64   pathname;
 139                 __u32           bpf_fd;
 140         };
 141 } __attribute__((aligned(8)));
 142
 143 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
 144  * function eBPF program intends to call
 145  */
 146 enum bpf_func_id {
 147         BPF_FUNC_unspec,
 148         BPF_FUNC_map_lookup_elem, /* void *map_lookup_elem(&map, &key) */
 149         BPF_FUNC_map_update_elem, /* int map_update_elem(&map, &key, &value, flags) */
 150         BPF_FUNC_map_delete_elem, /* int map_delete_elem(&map, &key) */
 151         BPF_FUNC_probe_read,      /* int bpf_probe_read(void *dst, int size, void *src) */
 152         BPF_FUNC_ktime_get_ns,    /* u64 bpf_ktime_get_ns(void) */
 153         BPF_FUNC_trace_printk,    /* int bpf_trace_printk(const char *fmt, int fmt_size, ...) */
 154         BPF_FUNC_get_prandom_u32, /* u32 prandom_u32(void) */
 155         BPF_FUNC_get_smp_processor_id, /* u32 raw_smp_processor_id(void) */
 156
 157         /**
 158          * skb_store_bytes(skb, offset, from, len, flags) - store bytes into packet
 159          * @skb: pointer to skb
 160          * @offset: offset within packet from skb->mac_header
 161          * @from: pointer where to copy bytes from
 162          * @len: number of bytes to store into packet
 163          * @flags: bit 0 - if true, recompute skb->csum
 164          *         other bits - reserved
 165          * Return: 0 on success
 166          */
 167         BPF_FUNC_skb_store_bytes,
 168
 169         /**
 170          * l3_csum_replace(skb, offset, from, to, flags) - recompute IP checksum
 171          * @skb: pointer to skb
 172          * @offset: offset within packet where IP checksum is located
 173          * @from: old value of header field
 174          * @to: new value of header field
 175          * @flags: bits 0-3 - size of header field
 176          *         other bits - reserved
 177          * Return: 0 on success
 178          */
 179         BPF_FUNC_l3_csum_replace,
 180
 181         /**
 182          * l4_csum_replace(skb, offset, from, to, flags) - recompute TCP/UDP checksum
 183          * @skb: pointer to skb
 184          * @offset: offset within packet where TCP/UDP checksum is located
 185          * @from: old value of header field
 186          * @to: new value of header field
 187          * @flags: bits 0-3 - size of header field
 188          *         bit 4 - is pseudo header
 189          *         other bits - reserved
 190          * Return: 0 on success
 191          */
 192         BPF_FUNC_l4_csum_replace,
 193
 194         /**
 195          * bpf_tail_call(ctx, prog_array_map, index) - jump into another BPF program
 196          * @ctx: context pointer passed to next program
 197          * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
 198          * @index: index inside array that selects specific program to run
 199          * Return: 0 on success
 200          */
 201         BPF_FUNC_tail_call,
 202
 203         /**
 204          * bpf_clone_redirect(skb, ifindex, flags) - redirect to another netdev
 205          * @skb: pointer to skb
 206          * @ifindex: ifindex of the net device
 207          * @flags: bit 0 - if set, redirect to ingress instead of egress
 208          *         other bits - reserved
 209          * Return: 0 on success
 210          */
 211         BPF_FUNC_clone_redirect,
 212
 213         /**
 214          * u64 bpf_get_current_pid_tgid(void)
 215          * Return: current->tgid << 32 | current->pid
 216          */
 217         BPF_FUNC_get_current_pid_tgid,
 218
 219         /**
 220          * u64 bpf_get_current_uid_gid(void)
 221          * Return: current_gid << 32 | current_uid
 222          */
 223         BPF_FUNC_get_current_uid_gid,
 224
 225         /**
 226          * bpf_get_current_comm(char *buf, int size_of_buf)
 227          * stores current->comm into buf
 228          * Return: 0 on success
 229          */
 230         BPF_FUNC_get_current_comm,
 231
 232         /**
 233          * bpf_get_cgroup_classid(skb) - retrieve a proc's classid
 234          * @skb: pointer to skb
 235          * Return: classid if != 0
 236          */
 237         BPF_FUNC_get_cgroup_classid,
 238         BPF_FUNC_skb_vlan_push, /* bpf_skb_vlan_push(skb, vlan_proto, vlan_tci) */
 239         BPF_FUNC_skb_vlan_pop,  /* bpf_skb_vlan_pop(skb) */
 240
 241         /**
 242          * bpf_skb_[gs]et_tunnel_key(skb, key, size, flags)
 243          * retrieve or populate tunnel metadata
 244          * @skb: pointer to skb
 245          * @key: pointer to 'struct bpf_tunnel_key'
 246          * @size: size of 'struct bpf_tunnel_key'
 247          * @flags: room for future extensions
 248          * Retrun: 0 on success
 249          */
 250         BPF_FUNC_skb_get_tunnel_key,
 251         BPF_FUNC_skb_set_tunnel_key,
 252         BPF_FUNC_perf_event_read,       /* u64 bpf_perf_event_read(&map, index) */
 253         /**
 254          * bpf_redirect(ifindex, flags) - redirect to another netdev
 255          * @ifindex: ifindex of the net device
 256          * @flags: bit 0 - if set, redirect to ingress instead of egress
 257          *         other bits - reserved
 258          * Return: TC_ACT_REDIRECT
 259          */
 260         BPF_FUNC_redirect,
 261
 262         /**
 263          * bpf_get_route_realm(skb) - retrieve a dst's tclassid
 264          * @skb: pointer to skb
 265          * Return: realm if != 0
 266          */
 267         BPF_FUNC_get_route_realm,
 268
 269         /**
 270          * bpf_perf_event_output(ctx, map, index, data, size) - output perf raw sample
 271          * @ctx: struct pt_regs*
 272          * @map: pointer to perf_event_array map
 273          * @index: index of event in the map
 274          * @data: data on stack to be output as raw data
 275          * @size: size of data
 276          * Return: 0 on success
 277          */
 278         BPF_FUNC_perf_event_output,
 279         BPF_FUNC_skb_load_bytes,
 280
 281         /**
 282          * bpf_get_stackid(ctx, map, flags) - walk user or kernel stack and return id
 283          * @ctx: struct pt_regs*
 284          * @map: pointer to stack_trace map
 285          * @flags: bits 0-7 - numer of stack frames to skip
 286          *         bit 8 - collect user stack instead of kernel
 287          *         bit 9 - compare stacks by hash only
 288          *         bit 10 - if two different stacks hash into the same stackid
 289          *                  discard old
 290          *         other bits - reserved
 291          * Return: >= 0 stackid on success or negative error
 292          */
 293         BPF_FUNC_get_stackid,
 294
 295         /**
 296          * bpf_csum_diff(from, from_size, to, to_size, seed) - calculate csum diff
 297          * @from: raw from buffer
 298          * @from_size: length of from buffer
 299          * @to: raw to buffer
 300          * @to_size: length of to buffer
 301          * @seed: optional seed
 302          * Return: csum result
 303          */
 304         BPF_FUNC_csum_diff,
 305
 306         /**
 307          * bpf_skb_[gs]et_tunnel_opt(skb, opt, size)
 308          * retrieve or populate tunnel options metadata
 309          * @skb: pointer to skb
 310          * @opt: pointer to raw tunnel option data
 311          * @size: size of @opt
 312          * Return: 0 on success for set, option size for get
 313          */
 314         BPF_FUNC_skb_get_tunnel_opt,
 315         BPF_FUNC_skb_set_tunnel_opt,
 316         __BPF_FUNC_MAX_ID,
 317 };
 318
 319 /* All flags used by eBPF helper functions, placed here. */
 320
 321 /* BPF_FUNC_skb_store_bytes flags. */
 322 #define BPF_F_RECOMPUTE_CSUM            (1ULL << 0)
 323 #define BPF_F_INVALIDATE_HASH           (1ULL << 1)
 324
 325 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
 326  * First 4 bits are for passing the header field size.
 327  */
 328 #define BPF_F_HDR_FIELD_MASK            0xfULL
 329
 330 /* BPF_FUNC_l4_csum_replace flags. */
 331 #define BPF_F_PSEUDO_HDR                (1ULL << 4)
 332 #define BPF_F_MARK_MANGLED_0            (1ULL << 5)
 333
 334 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
 335 #define BPF_F_INGRESS                   (1ULL << 0)
 336
 337 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
 338 #define BPF_F_TUNINFO_IPV6              (1ULL << 0)
 339
 340 /* BPF_FUNC_get_stackid flags. */
 341 #define BPF_F_SKIP_FIELD_MASK           0xffULL
 342 #define BPF_F_USER_STACK                (1ULL << 8)
 343 #define BPF_F_FAST_STACK_CMP            (1ULL << 9)
 344 #define BPF_F_REUSE_STACKID             (1ULL << 10)
 345
 346 /* BPF_FUNC_skb_set_tunnel_key flags. */
 347 #define BPF_F_ZERO_CSUM_TX              (1ULL << 1)
 348 #define BPF_F_DONT_FRAGMENT             (1ULL << 2)
 349
 350 /* BPF_FUNC_perf_event_output flags. */
 351 #define BPF_F_INDEX_MASK                0xffffffffULL
 352 #define BPF_F_CURRENT_CPU               BPF_F_INDEX_MASK
 353
 354 /* user accessible mirror of in-kernel sk_buff.
 355  * new fields can only be added to the end of this structure
 356  */
 357 struct __sk_buff {
 358         __u32 len;
 359         __u32 pkt_type;
 360         __u32 mark;
 361         __u32 queue_mapping;
 362         __u32 protocol;
 363         __u32 vlan_present;
 364         __u32 vlan_tci;
 365         __u32 vlan_proto;
 366         __u32 priority;
 367         __u32 ingress_ifindex;
 368         __u32 ifindex;
 369         __u32 tc_index;
 370         __u32 cb[5];
 371         __u32 hash;
 372         __u32 tc_classid;
 373         __u32 data;
 374         __u32 data_end;
 375 };
 376
 377 struct bpf_tunnel_key {
 378         __u32 tunnel_id;
 379         union {
 380                 __u32 remote_ipv4;
 381                 __u32 remote_ipv6[4];
 382         };
 383         __u8 tunnel_tos;
 384         __u8 tunnel_ttl;
 385         __u16 tunnel_ext;
 386         __u32 tunnel_label;
 387 };
 388
 389 #endif /* __LINUX_BPF_H__ */