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 1
   9
  10 #include <uapi/linux/bpf.h>
  11
  12 #include <linux/workqueue.h>
  13 #include <linux/file.h>
  14 #include <linux/percpu.h>
  15 #include <linux/err.h>
  16 #include <linux/rbtree_latch.h>
  17 #include <linux/numa.h>
  18
  19 struct perf_event;
  20 struct bpf_prog;
  21 struct bpf_map;
  22
  23 /* map is generic key/value storage optionally accesible by eBPF programs */
  24 struct bpf_map_ops {
  25         /* funcs callable from userspace (via syscall) */
  26         struct bpf_map *(*map_alloc)(union bpf_attr *attr);
  27         void (*map_release)(struct bpf_map *map, struct file *map_file);
  28         void (*map_free)(struct bpf_map *map);
  29         int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
  30
  31         /* funcs callable from userspace and from eBPF programs */
  32         void *(*map_lookup_elem)(struct bpf_map *map, void *key);
  33         int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
  34         int (*map_delete_elem)(struct bpf_map *map, void *key);
  35
  36         /* funcs called by prog_array and perf_event_array map */
  37         void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
  38                                 int fd);
  39         void (*map_fd_put_ptr)(void *ptr);
  40         u32 (*map_gen_lookup)(struct bpf_map *map, struct bpf_insn *insn_buf);
  41         u32 (*map_fd_sys_lookup_elem)(void *ptr);
  42         int (*map_attach)(struct bpf_map *map,
  43                           struct bpf_prog *p1, struct bpf_prog *p2);
  44 };
  45
  46 struct bpf_map {
  47         atomic_t refcnt;
  48         enum bpf_map_type map_type;
  49         u32 key_size;
  50         u32 value_size;
  51         u32 max_entries;
  52         u32 map_flags;
  53         u32 pages;
  54         u32 id;
  55         int numa_node;
  56         struct user_struct *user;
  57         const struct bpf_map_ops *ops;
  58         struct work_struct work;
  59         atomic_t usercnt;
  60         struct bpf_map *inner_map_meta;
  61 };
  62
  63 /* function argument constraints */
  64 enum bpf_arg_type {
  65         ARG_DONTCARE = 0,       /* unused argument in helper function */
  66
  67         /* the following constraints used to prototype
  68          * bpf_map_lookup/update/delete_elem() functions
  69          */
  70         ARG_CONST_MAP_PTR,      /* const argument used as pointer to bpf_map */
  71         ARG_PTR_TO_MAP_KEY,     /* pointer to stack used as map key */
  72         ARG_PTR_TO_MAP_VALUE,   /* pointer to stack used as map value */
  73
  74         /* the following constraints used to prototype bpf_memcmp() and other
  75          * functions that access data on eBPF program stack
  76          */
  77         ARG_PTR_TO_MEM,         /* pointer to valid memory (stack, packet, map value) */
  78         ARG_PTR_TO_UNINIT_MEM,  /* pointer to memory does not need to be initialized,
  79                                  * helper function must fill all bytes or clear
  80                                  * them in error case.
  81                                  */
  82
  83         ARG_CONST_SIZE,         /* number of bytes accessed from memory */
  84         ARG_CONST_SIZE_OR_ZERO, /* number of bytes accessed from memory or 0 */
  85
  86         ARG_PTR_TO_CTX,         /* pointer to context */
  87         ARG_ANYTHING,           /* any (initialized) argument is ok */
  88 };
  89
  90 /* type of values returned from helper functions */
  91 enum bpf_return_type {
  92         RET_INTEGER,                    /* function returns integer */
  93         RET_VOID,                       /* function doesn't return anything */
  94         RET_PTR_TO_MAP_VALUE_OR_NULL,   /* returns a pointer to map elem value or NULL */
  95 };
  96
  97 /* eBPF function prototype used by verifier to allow BPF_CALLs from eBPF programs
  98  * to in-kernel helper functions and for adjusting imm32 field in BPF_CALL
  99  * instructions after verifying
 100  */
 101 struct bpf_func_proto {
 102         u64 (*func)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
 103         bool gpl_only;
 104         bool pkt_access;
 105         enum bpf_return_type ret_type;
 106         enum bpf_arg_type arg1_type;
 107         enum bpf_arg_type arg2_type;
 108         enum bpf_arg_type arg3_type;
 109         enum bpf_arg_type arg4_type;
 110         enum bpf_arg_type arg5_type;
 111 };
 112
 113 /* bpf_context is intentionally undefined structure. Pointer to bpf_context is
 114  * the first argument to eBPF programs.
 115  * For socket filters: 'struct bpf_context *' == 'struct sk_buff *'
 116  */
 117 struct bpf_context;
 118
 119 enum bpf_access_type {
 120         BPF_READ = 1,
 121         BPF_WRITE = 2
 122 };
 123
 124 /* types of values stored in eBPF registers */
 125 /* Pointer types represent:
 126  * pointer
 127  * pointer + imm
 128  * pointer + (u16) var
 129  * pointer + (u16) var + imm
 130  * if (range > 0) then [ptr, ptr + range - off) is safe to access
 131  * if (id > 0) means that some 'var' was added
 132  * if (off > 0) means that 'imm' was added
 133  */
 134 enum bpf_reg_type {
 135         NOT_INIT = 0,            /* nothing was written into register */
 136         SCALAR_VALUE,            /* reg doesn't contain a valid pointer */
 137         PTR_TO_CTX,              /* reg points to bpf_context */
 138         CONST_PTR_TO_MAP,        /* reg points to struct bpf_map */
 139         PTR_TO_MAP_VALUE,        /* reg points to map element value */
 140         PTR_TO_MAP_VALUE_OR_NULL,/* points to map elem value or NULL */
 141         PTR_TO_STACK,            /* reg == frame_pointer + offset */
 142         PTR_TO_PACKET,           /* reg points to skb->data */
 143         PTR_TO_PACKET_END,       /* skb->data + headlen */
 144 };
 145
 146 /* The information passed from prog-specific *_is_valid_access
 147  * back to the verifier.
 148  */
 149 struct bpf_insn_access_aux {
 150         enum bpf_reg_type reg_type;
 151         int ctx_field_size;
 152 };
 153
 154 static inline void
 155 bpf_ctx_record_field_size(struct bpf_insn_access_aux *aux, u32 size)
 156 {
 157         aux->ctx_field_size = size;
 158 }
 159
 160 struct bpf_verifier_ops {
 161         /* return eBPF function prototype for verification */
 162         const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
 163
 164         /* return true if 'size' wide access at offset 'off' within bpf_context
 165          * with 'type' (read or write) is allowed
 166          */
 167         bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
 168                                 struct bpf_insn_access_aux *info);
 169         int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
 170                             const struct bpf_prog *prog);
 171         u32 (*convert_ctx_access)(enum bpf_access_type type,
 172                                   const struct bpf_insn *src,
 173                                   struct bpf_insn *dst,
 174                                   struct bpf_prog *prog, u32 *target_size);
 175         int (*test_run)(struct bpf_prog *prog, const union bpf_attr *kattr,
 176                         union bpf_attr __user *uattr);
 177 };
 178
 179 struct bpf_prog_aux {
 180         atomic_t refcnt;
 181         u32 used_map_cnt;
 182         u32 max_ctx_offset;
 183         u32 stack_depth;
 184         u32 id;
 185         struct latch_tree_node ksym_tnode;
 186         struct list_head ksym_lnode;
 187         const struct bpf_verifier_ops *ops;
 188         struct bpf_map **used_maps;
 189         struct bpf_prog *prog;
 190         struct user_struct *user;
 191         union {
 192                 struct work_struct work;
 193                 struct rcu_head rcu;
 194         };
 195 };
 196
 197 struct bpf_array {
 198         struct bpf_map map;
 199         u32 elem_size;
 200         /* 'ownership' of prog_array is claimed by the first program that
 201          * is going to use this map or by the first program which FD is stored
 202          * in the map to make sure that all callers and callees have the same
 203          * prog_type and JITed flag
 204          */
 205         enum bpf_prog_type owner_prog_type;
 206         bool owner_jited;
 207         union {
 208                 char value[0] __aligned(8);
 209                 void *ptrs[0] __aligned(8);
 210                 void __percpu *pptrs[0] __aligned(8);
 211         };
 212 };
 213
 214 #define MAX_TAIL_CALL_CNT 32
 215
 216 struct bpf_event_entry {
 217         struct perf_event *event;
 218         struct file *perf_file;
 219         struct file *map_file;
 220         struct rcu_head rcu;
 221 };
 222
 223 u64 bpf_tail_call(u64 ctx, u64 r2, u64 index, u64 r4, u64 r5);
 224 u64 bpf_get_stackid(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
 225
 226 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
 227 int bpf_prog_calc_tag(struct bpf_prog *fp);
 228
 229 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
 230
 231 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
 232                                         unsigned long off, unsigned long len);
 233
 234 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
 235                      void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
 236
 237 int bpf_prog_test_run_xdp(struct bpf_prog *prog, const union bpf_attr *kattr,
 238                           union bpf_attr __user *uattr);
 239 int bpf_prog_test_run_skb(struct bpf_prog *prog, const union bpf_attr *kattr,
 240                           union bpf_attr __user *uattr);
 241
 242 #ifdef CONFIG_BPF_SYSCALL
 243 DECLARE_PER_CPU(int, bpf_prog_active);
 244
 245 #define BPF_PROG_TYPE(_id, _ops) \
 246         extern const struct bpf_verifier_ops _ops;
 247 #define BPF_MAP_TYPE(_id, _ops) \
 248         extern const struct bpf_map_ops _ops;
 249 #include <linux/bpf_types.h>
 250 #undef BPF_PROG_TYPE
 251 #undef BPF_MAP_TYPE
 252
 253 struct bpf_prog *bpf_prog_get(u32 ufd);
 254 struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type);
 255 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
 256 void bpf_prog_sub(struct bpf_prog *prog, int i);
 257 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
 258 struct bpf_prog * __must_check bpf_prog_inc_not_zero(struct bpf_prog *prog);
 259 void bpf_prog_put(struct bpf_prog *prog);
 260 int __bpf_prog_charge(struct user_struct *user, u32 pages);
 261 void __bpf_prog_uncharge(struct user_struct *user, u32 pages);
 262
 263 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
 264 struct bpf_map *__bpf_map_get(struct fd f);
 265 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
 266 void bpf_map_put_with_uref(struct bpf_map *map);
 267 void bpf_map_put(struct bpf_map *map);
 268 int bpf_map_precharge_memlock(u32 pages);
 269 void *bpf_map_area_alloc(size_t size, int numa_node);
 270 void bpf_map_area_free(void *base);
 271
 272 extern int sysctl_unprivileged_bpf_disabled;
 273
 274 int bpf_map_new_fd(struct bpf_map *map);
 275 int bpf_prog_new_fd(struct bpf_prog *prog);
 276
 277 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
 278 int bpf_obj_get_user(const char __user *pathname);
 279
 280 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
 281 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
 282 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
 283                            u64 flags);
 284 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
 285                             u64 flags);
 286
 287 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
 288
 289 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
 290                                  void *key, void *value, u64 map_flags);
 291 int bpf_fd_array_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
 292 void bpf_fd_array_map_clear(struct bpf_map *map);
 293 int bpf_fd_htab_map_update_elem(struct bpf_map *map, struct file *map_file,
 294                                 void *key, void *value, u64 map_flags);
 295 int bpf_fd_htab_map_lookup_elem(struct bpf_map *map, void *key, u32 *value);
 296
 297 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
 298  * forced to use 'long' read/writes to try to atomically copy long counters.
 299  * Best-effort only.  No barriers here, since it _will_ race with concurrent
 300  * updates from BPF programs. Called from bpf syscall and mostly used with
 301  * size 8 or 16 bytes, so ask compiler to inline it.
 302  */
 303 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
 304 {
 305         const long *lsrc = src;
 306         long *ldst = dst;
 307
 308         size /= sizeof(long);
 309         while (size--)
 310                 *ldst++ = *lsrc++;
 311 }
 312
 313 /* verify correctness of eBPF program */
 314 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
 315
 316 /* Map specifics */
 317 struct net_device  *__dev_map_lookup_elem(struct bpf_map *map, u32 key);
 318 void __dev_map_insert_ctx(struct bpf_map *map, u32 index);
 319 void __dev_map_flush(struct bpf_map *map);
 320
 321 /* Return map's numa specified by userspace */
 322 static inline int bpf_map_attr_numa_node(const union bpf_attr *attr)
 323 {
 324         return (attr->map_flags & BPF_F_NUMA_NODE) ?
 325                 attr->numa_node : NUMA_NO_NODE;
 326 }
 327
 328 #else
 329 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
 330 {
 331         return ERR_PTR(-EOPNOTSUPP);
 332 }
 333
 334 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
 335                                                  enum bpf_prog_type type)
 336 {
 337         return ERR_PTR(-EOPNOTSUPP);
 338 }
 339 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
 340                                                           int i)
 341 {
 342         return ERR_PTR(-EOPNOTSUPP);
 343 }
 344
 345 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
 346 {
 347 }
 348
 349 static inline void bpf_prog_put(struct bpf_prog *prog)
 350 {
 351 }
 352
 353 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
 354 {
 355         return ERR_PTR(-EOPNOTSUPP);
 356 }
 357
 358 static inline struct bpf_prog *__must_check
 359 bpf_prog_inc_not_zero(struct bpf_prog *prog)
 360 {
 361         return ERR_PTR(-EOPNOTSUPP);
 362 }
 363
 364 static inline int __bpf_prog_charge(struct user_struct *user, u32 pages)
 365 {
 366         return 0;
 367 }
 368
 369 static inline void __bpf_prog_uncharge(struct user_struct *user, u32 pages)
 370 {
 371 }
 372
 373 static inline struct net_device  *__dev_map_lookup_elem(struct bpf_map *map,
 374                                                        u32 key)
 375 {
 376         return NULL;
 377 }
 378
 379 static inline void __dev_map_insert_ctx(struct bpf_map *map, u32 index)
 380 {
 381 }
 382
 383 static inline void __dev_map_flush(struct bpf_map *map)
 384 {
 385 }
 386 #endif /* CONFIG_BPF_SYSCALL */
 387
 388 #if defined(CONFIG_STREAM_PARSER) && defined(CONFIG_BPF_SYSCALL)
 389 struct sock  *__sock_map_lookup_elem(struct bpf_map *map, u32 key);
 390 #else
 391 static inline struct sock  *__sock_map_lookup_elem(struct bpf_map *map, u32 key)
 392 {
 393         return NULL;
 394 }
 395 #endif
 396
 397 /* verifier prototypes for helper functions called from eBPF programs */
 398 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
 399 extern const struct bpf_func_proto bpf_map_update_elem_proto;
 400 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
 401
 402 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
 403 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
 404 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
 405 extern const struct bpf_func_proto bpf_tail_call_proto;
 406 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
 407 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
 408 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
 409 extern const struct bpf_func_proto bpf_get_current_comm_proto;
 410 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
 411 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
 412 extern const struct bpf_func_proto bpf_get_stackid_proto;
 413 extern const struct bpf_func_proto bpf_sock_map_update_proto;
 414
 415 /* Shared helpers among cBPF and eBPF. */
 416 void bpf_user_rnd_init_once(void);
 417 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
 418
 419 #endif /* _LINUX_BPF_H */