Home | History | Annotate | Download | only in linux
      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 };
     85 
     86 enum bpf_prog_type {
     87 	BPF_PROG_TYPE_UNSPEC,
     88 	BPF_PROG_TYPE_SOCKET_FILTER,
     89 	BPF_PROG_TYPE_KPROBE,
     90 	BPF_PROG_TYPE_SCHED_CLS,
     91 	BPF_PROG_TYPE_SCHED_ACT,
     92 };
     93 
     94 #define BPF_PSEUDO_MAP_FD	1
     95 
     96 /* flags for BPF_MAP_UPDATE_ELEM command */
     97 #define BPF_ANY		0 /* create new element or update existing */
     98 #define BPF_NOEXIST	1 /* create new element if it didn't exist */
     99 #define BPF_EXIST	2 /* update existing element */
    100 
    101 union bpf_attr {
    102 	struct { /* anonymous struct used by BPF_MAP_CREATE command */
    103 		__u32	map_type;	/* one of enum bpf_map_type */
    104 		__u32	key_size;	/* size of key in bytes */
    105 		__u32	value_size;	/* size of value in bytes */
    106 		__u32	max_entries;	/* max number of entries in a map */
    107 	};
    108 
    109 	struct { /* anonymous struct used by BPF_MAP_*_ELEM commands */
    110 		__u32		map_fd;
    111 		__aligned_u64	key;
    112 		union {
    113 			__aligned_u64 value;
    114 			__aligned_u64 next_key;
    115 		};
    116 		__u64		flags;
    117 	};
    118 
    119 	struct { /* anonymous struct used by BPF_PROG_LOAD command */
    120 		__u32		prog_type;	/* one of enum bpf_prog_type */
    121 		__u32		insn_cnt;
    122 		__aligned_u64	insns;
    123 		__aligned_u64	license;
    124 		__u32		log_level;	/* verbosity level of verifier */
    125 		__u32		log_size;	/* size of user buffer */
    126 		__aligned_u64	log_buf;	/* user supplied buffer */
    127 		__u32		kern_version;	/* checked when prog_type=kprobe */
    128 	};
    129 
    130 	struct { /* anonymous struct used by BPF_OBJ_* commands */
    131 		__aligned_u64	pathname;
    132 		__u32		bpf_fd;
    133 	};
    134 } __attribute__((aligned(8)));
    135 
    136 /* integer value in 'imm' field of BPF_CALL instruction selects which helper
    137  * function eBPF program intends to call
    138  */
    139 enum bpf_func_id {
    140 	BPF_FUNC_unspec,
    141 	BPF_FUNC_map_lookup_elem, /* void *map_lookup_elem(&map, &key) */
    142 	BPF_FUNC_map_update_elem, /* int map_update_elem(&map, &key, &value, flags) */
    143 	BPF_FUNC_map_delete_elem, /* int map_delete_elem(&map, &key) */
    144 	BPF_FUNC_probe_read,      /* int bpf_probe_read(void *dst, int size, void *src) */
    145 	BPF_FUNC_ktime_get_ns,    /* u64 bpf_ktime_get_ns(void) */
    146 	BPF_FUNC_trace_printk,    /* int bpf_trace_printk(const char *fmt, int fmt_size, ...) */
    147 	BPF_FUNC_get_prandom_u32, /* u32 prandom_u32(void) */
    148 	BPF_FUNC_get_smp_processor_id, /* u32 raw_smp_processor_id(void) */
    149 
    150 	/**
    151 	 * skb_store_bytes(skb, offset, from, len, flags) - store bytes into packet
    152 	 * @skb: pointer to skb
    153 	 * @offset: offset within packet from skb->mac_header
    154 	 * @from: pointer where to copy bytes from
    155 	 * @len: number of bytes to store into packet
    156 	 * @flags: bit 0 - if true, recompute skb->csum
    157 	 *         other bits - reserved
    158 	 * Return: 0 on success
    159 	 */
    160 	BPF_FUNC_skb_store_bytes,
    161 
    162 	/**
    163 	 * l3_csum_replace(skb, offset, from, to, flags) - recompute IP checksum
    164 	 * @skb: pointer to skb
    165 	 * @offset: offset within packet where IP checksum is located
    166 	 * @from: old value of header field
    167 	 * @to: new value of header field
    168 	 * @flags: bits 0-3 - size of header field
    169 	 *         other bits - reserved
    170 	 * Return: 0 on success
    171 	 */
    172 	BPF_FUNC_l3_csum_replace,
    173 
    174 	/**
    175 	 * l4_csum_replace(skb, offset, from, to, flags) - recompute TCP/UDP checksum
    176 	 * @skb: pointer to skb
    177 	 * @offset: offset within packet where TCP/UDP checksum is located
    178 	 * @from: old value of header field
    179 	 * @to: new value of header field
    180 	 * @flags: bits 0-3 - size of header field
    181 	 *         bit 4 - is pseudo header
    182 	 *         other bits - reserved
    183 	 * Return: 0 on success
    184 	 */
    185 	BPF_FUNC_l4_csum_replace,
    186 
    187 	/**
    188 	 * bpf_tail_call(ctx, prog_array_map, index) - jump into another BPF program
    189 	 * @ctx: context pointer passed to next program
    190 	 * @prog_array_map: pointer to map which type is BPF_MAP_TYPE_PROG_ARRAY
    191 	 * @index: index inside array that selects specific program to run
    192 	 * Return: 0 on success
    193 	 */
    194 	BPF_FUNC_tail_call,
    195 
    196 	/**
    197 	 * bpf_clone_redirect(skb, ifindex, flags) - redirect to another netdev
    198 	 * @skb: pointer to skb
    199 	 * @ifindex: ifindex of the net device
    200 	 * @flags: bit 0 - if set, redirect to ingress instead of egress
    201 	 *         other bits - reserved
    202 	 * Return: 0 on success
    203 	 */
    204 	BPF_FUNC_clone_redirect,
    205 
    206 	/**
    207 	 * u64 bpf_get_current_pid_tgid(void)
    208 	 * Return: current->tgid << 32 | current->pid
    209 	 */
    210 	BPF_FUNC_get_current_pid_tgid,
    211 
    212 	/**
    213 	 * u64 bpf_get_current_uid_gid(void)
    214 	 * Return: current_gid << 32 | current_uid
    215 	 */
    216 	BPF_FUNC_get_current_uid_gid,
    217 
    218 	/**
    219 	 * bpf_get_current_comm(char *buf, int size_of_buf)
    220 	 * stores current->comm into buf
    221 	 * Return: 0 on success
    222 	 */
    223 	BPF_FUNC_get_current_comm,
    224 
    225 	/**
    226 	 * bpf_get_cgroup_classid(skb) - retrieve a proc's classid
    227 	 * @skb: pointer to skb
    228 	 * Return: classid if != 0
    229 	 */
    230 	BPF_FUNC_get_cgroup_classid,
    231 	BPF_FUNC_skb_vlan_push, /* bpf_skb_vlan_push(skb, vlan_proto, vlan_tci) */
    232 	BPF_FUNC_skb_vlan_pop,  /* bpf_skb_vlan_pop(skb) */
    233 
    234 	/**
    235 	 * bpf_skb_[gs]et_tunnel_key(skb, key, size, flags)
    236 	 * retrieve or populate tunnel metadata
    237 	 * @skb: pointer to skb
    238 	 * @key: pointer to 'struct bpf_tunnel_key'
    239 	 * @size: size of 'struct bpf_tunnel_key'
    240 	 * @flags: room for future extensions
    241 	 * Retrun: 0 on success
    242 	 */
    243 	BPF_FUNC_skb_get_tunnel_key,
    244 	BPF_FUNC_skb_set_tunnel_key,
    245 	BPF_FUNC_perf_event_read,	/* u64 bpf_perf_event_read(&map, index) */
    246 	/**
    247 	 * bpf_redirect(ifindex, flags) - redirect to another netdev
    248 	 * @ifindex: ifindex of the net device
    249 	 * @flags: bit 0 - if set, redirect to ingress instead of egress
    250 	 *         other bits - reserved
    251 	 * Return: TC_ACT_REDIRECT
    252 	 */
    253 	BPF_FUNC_redirect,
    254 
    255 	/**
    256 	 * bpf_get_route_realm(skb) - retrieve a dst's tclassid
    257 	 * @skb: pointer to skb
    258 	 * Return: realm if != 0
    259 	 */
    260 	BPF_FUNC_get_route_realm,
    261 
    262 	/**
    263 	 * bpf_perf_event_output(ctx, map, index, data, size) - output perf raw sample
    264 	 * @ctx: struct pt_regs*
    265 	 * @map: pointer to perf_event_array map
    266 	 * @index: index of event in the map
    267 	 * @data: data on stack to be output as raw data
    268 	 * @size: size of data
    269 	 * Return: 0 on success
    270 	 */
    271 	BPF_FUNC_perf_event_output,
    272 	BPF_FUNC_skb_load_bytes,
    273 	__BPF_FUNC_MAX_ID,
    274 };
    275 
    276 /* All flags used by eBPF helper functions, placed here. */
    277 
    278 /* BPF_FUNC_skb_store_bytes flags. */
    279 #define BPF_F_RECOMPUTE_CSUM		(1ULL << 0)
    280 
    281 /* BPF_FUNC_l3_csum_replace and BPF_FUNC_l4_csum_replace flags.
    282  * First 4 bits are for passing the header field size.
    283  */
    284 #define BPF_F_HDR_FIELD_MASK		0xfULL
    285 
    286 /* BPF_FUNC_l4_csum_replace flags. */
    287 #define BPF_F_PSEUDO_HDR		(1ULL << 4)
    288 
    289 /* BPF_FUNC_clone_redirect and BPF_FUNC_redirect flags. */
    290 #define BPF_F_INGRESS			(1ULL << 0)
    291 
    292 /* BPF_FUNC_skb_set_tunnel_key and BPF_FUNC_skb_get_tunnel_key flags. */
    293 #define BPF_F_TUNINFO_IPV6		(1ULL << 0)
    294 
    295 /* user accessible mirror of in-kernel sk_buff.
    296  * new fields can only be added to the end of this structure
    297  */
    298 struct __sk_buff {
    299 	__u32 len;
    300 	__u32 pkt_type;
    301 	__u32 mark;
    302 	__u32 queue_mapping;
    303 	__u32 protocol;
    304 	__u32 vlan_present;
    305 	__u32 vlan_tci;
    306 	__u32 vlan_proto;
    307 	__u32 priority;
    308 	__u32 ingress_ifindex;
    309 	__u32 ifindex;
    310 	__u32 tc_index;
    311 	__u32 cb[5];
    312 	__u32 hash;
    313 	__u32 tc_classid;
    314 };
    315 
    316 struct bpf_tunnel_key {
    317 	__u32 tunnel_id;
    318 	union {
    319 		__u32 remote_ipv4;
    320 		__u32 remote_ipv6[4];
    321 	};
    322 	__u8 tunnel_tos;
    323 	__u8 tunnel_ttl;
    324 };
    325 
    326 #endif /* __LINUX_BPF_H__ */
    327