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