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 #include <linux/workqueue.h>
  12 #include <linux/file.h>
  13 #include <linux/percpu.h>
  14 #include <linux/err.h>
  15
  16 struct perf_event;
  17 struct bpf_map;
  18
  19 /* map is generic key/value storage optionally accesible by eBPF programs */
  20 struct bpf_map_ops {
  21         /* funcs callable from userspace (via syscall) */
  22         struct bpf_map *(*map_alloc)(union bpf_attr *attr);
  23         void (*map_release)(struct bpf_map *map, struct file *map_file);
  24         void (*map_free)(struct bpf_map *map);
  25         int (*map_get_next_key)(struct bpf_map *map, void *key, void *next_key);
  26
  27         /* funcs callable from userspace and from eBPF programs */
  28         void *(*map_lookup_elem)(struct bpf_map *map, void *key);
  29         int (*map_update_elem)(struct bpf_map *map, void *key, void *value, u64 flags);
  30         int (*map_delete_elem)(struct bpf_map *map, void *key);
  31
  32         /* funcs called by prog_array and perf_event_array map */
  33         void *(*map_fd_get_ptr)(struct bpf_map *map, struct file *map_file,
  34                                 int fd);
  35         void (*map_fd_put_ptr)(void *ptr);
  36 };
  37
  38 struct bpf_map {
  39         atomic_t refcnt;
  40         enum bpf_map_type map_type;
  41         u32 key_size;
  42         u32 value_size;
  43         u32 max_entries;
  44         u32 map_flags;
  45         u32 pages;
  46         struct user_struct *user;
  47         const struct bpf_map_ops *ops;
  48         struct work_struct work;
  49         atomic_t usercnt;
  50 };
  51
  52 struct bpf_map_type_list {
  53         struct list_head list_node;
  54         const struct bpf_map_ops *ops;
  55         enum bpf_map_type type;
  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_STACK,       /* any pointer to eBPF program stack */
  73         ARG_PTR_TO_RAW_STACK,   /* any pointer to eBPF program stack, area does not
  74                                  * need to be initialized, helper function must fill
  75                                  * all bytes or clear them in error case.
  76                                  */
  77
  78         ARG_CONST_STACK_SIZE,   /* number of bytes accessed from stack */
  79         ARG_CONST_STACK_SIZE_OR_ZERO, /* number of bytes accessed from stack 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 struct bpf_verifier_ops {
 154         /* return eBPF function prototype for verification */
 155         const struct bpf_func_proto *(*get_func_proto)(enum bpf_func_id func_id);
 156
 157         /* return true if 'size' wide access at offset 'off' within bpf_context
 158          * with 'type' (read or write) is allowed
 159          */
 160         bool (*is_valid_access)(int off, int size, enum bpf_access_type type,
 161                                 enum bpf_reg_type *reg_type);
 162         int (*gen_prologue)(struct bpf_insn *insn, bool direct_write,
 163                             const struct bpf_prog *prog);
 164         u32 (*convert_ctx_access)(enum bpf_access_type type, int dst_reg,
 165                                   int src_reg, int ctx_off,
 166                                   struct bpf_insn *insn, struct bpf_prog *prog);
 167 };
 168
 169 struct bpf_prog_type_list {
 170         struct list_head list_node;
 171         const struct bpf_verifier_ops *ops;
 172         enum bpf_prog_type type;
 173 };
 174
 175 struct bpf_prog_aux {
 176         atomic_t refcnt;
 177         u32 used_map_cnt;
 178         u32 max_ctx_offset;
 179         const struct bpf_verifier_ops *ops;
 180         struct bpf_map **used_maps;
 181         struct bpf_prog *prog;
 182         struct user_struct *user;
 183         union {
 184                 struct work_struct work;
 185                 struct rcu_head rcu;
 186         };
 187 };
 188
 189 struct bpf_array {
 190         struct bpf_map map;
 191         u32 elem_size;
 192         /* 'ownership' of prog_array is claimed by the first program that
 193          * is going to use this map or by the first program which FD is stored
 194          * in the map to make sure that all callers and callees have the same
 195          * prog_type and JITed flag
 196          */
 197         enum bpf_prog_type owner_prog_type;
 198         bool owner_jited;
 199         union {
 200                 char value[0] __aligned(8);
 201                 void *ptrs[0] __aligned(8);
 202                 void __percpu *pptrs[0] __aligned(8);
 203         };
 204 };
 205
 206 #define MAX_TAIL_CALL_CNT 32
 207
 208 struct bpf_event_entry {
 209         struct perf_event *event;
 210         struct file *perf_file;
 211         struct file *map_file;
 212         struct rcu_head rcu;
 213 };
 214
 215 u64 bpf_tail_call(u64 ctx, u64 r2, u64 index, u64 r4, u64 r5);
 216 u64 bpf_get_stackid(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
 217
 218 bool bpf_prog_array_compatible(struct bpf_array *array, const struct bpf_prog *fp);
 219 void bpf_prog_calc_digest(struct bpf_prog *fp);
 220
 221 const struct bpf_func_proto *bpf_get_trace_printk_proto(void);
 222
 223 typedef unsigned long (*bpf_ctx_copy_t)(void *dst, const void *src,
 224                                         unsigned long off, unsigned long len);
 225
 226 u64 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
 227                      void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy);
 228
 229 #ifdef CONFIG_BPF_SYSCALL
 230 DECLARE_PER_CPU(int, bpf_prog_active);
 231
 232 void bpf_register_prog_type(struct bpf_prog_type_list *tl);
 233 void bpf_register_map_type(struct bpf_map_type_list *tl);
 234
 235 struct bpf_prog *bpf_prog_get(u32 ufd);
 236 struct bpf_prog *bpf_prog_get_type(u32 ufd, enum bpf_prog_type type);
 237 struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog, int i);
 238 void bpf_prog_sub(struct bpf_prog *prog, int i);
 239 struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog);
 240 void bpf_prog_put(struct bpf_prog *prog);
 241
 242 struct bpf_map *bpf_map_get_with_uref(u32 ufd);
 243 struct bpf_map *__bpf_map_get(struct fd f);
 244 struct bpf_map * __must_check bpf_map_inc(struct bpf_map *map, bool uref);
 245 void bpf_map_put_with_uref(struct bpf_map *map);
 246 void bpf_map_put(struct bpf_map *map);
 247 int bpf_map_precharge_memlock(u32 pages);
 248
 249 extern int sysctl_unprivileged_bpf_disabled;
 250
 251 int bpf_map_new_fd(struct bpf_map *map);
 252 int bpf_prog_new_fd(struct bpf_prog *prog);
 253
 254 int bpf_obj_pin_user(u32 ufd, const char __user *pathname);
 255 int bpf_obj_get_user(const char __user *pathname);
 256
 257 int bpf_percpu_hash_copy(struct bpf_map *map, void *key, void *value);
 258 int bpf_percpu_array_copy(struct bpf_map *map, void *key, void *value);
 259 int bpf_percpu_hash_update(struct bpf_map *map, void *key, void *value,
 260                            u64 flags);
 261 int bpf_percpu_array_update(struct bpf_map *map, void *key, void *value,
 262                             u64 flags);
 263
 264 int bpf_stackmap_copy(struct bpf_map *map, void *key, void *value);
 265
 266 int bpf_fd_array_map_update_elem(struct bpf_map *map, struct file *map_file,
 267                                  void *key, void *value, u64 map_flags);
 268 void bpf_fd_array_map_clear(struct bpf_map *map);
 269
 270 /* memcpy that is used with 8-byte aligned pointers, power-of-8 size and
 271  * forced to use 'long' read/writes to try to atomically copy long counters.
 272  * Best-effort only.  No barriers here, since it _will_ race with concurrent
 273  * updates from BPF programs. Called from bpf syscall and mostly used with
 274  * size 8 or 16 bytes, so ask compiler to inline it.
 275  */
 276 static inline void bpf_long_memcpy(void *dst, const void *src, u32 size)
 277 {
 278         const long *lsrc = src;
 279         long *ldst = dst;
 280
 281         size /= sizeof(long);
 282         while (size--)
 283                 *ldst++ = *lsrc++;
 284 }
 285
 286 /* verify correctness of eBPF program */
 287 int bpf_check(struct bpf_prog **fp, union bpf_attr *attr);
 288 #else
 289 static inline void bpf_register_prog_type(struct bpf_prog_type_list *tl)
 290 {
 291 }
 292
 293 static inline struct bpf_prog *bpf_prog_get(u32 ufd)
 294 {
 295         return ERR_PTR(-EOPNOTSUPP);
 296 }
 297
 298 static inline struct bpf_prog *bpf_prog_get_type(u32 ufd,
 299                                                  enum bpf_prog_type type)
 300 {
 301         return ERR_PTR(-EOPNOTSUPP);
 302 }
 303 static inline struct bpf_prog * __must_check bpf_prog_add(struct bpf_prog *prog,
 304                                                           int i)
 305 {
 306         return ERR_PTR(-EOPNOTSUPP);
 307 }
 308
 309 static inline void bpf_prog_sub(struct bpf_prog *prog, int i)
 310 {
 311 }
 312
 313 static inline void bpf_prog_put(struct bpf_prog *prog)
 314 {
 315 }
 316
 317 static inline struct bpf_prog * __must_check bpf_prog_inc(struct bpf_prog *prog)
 318 {
 319         return ERR_PTR(-EOPNOTSUPP);
 320 }
 321 #endif /* CONFIG_BPF_SYSCALL */
 322
 323 /* verifier prototypes for helper functions called from eBPF programs */
 324 extern const struct bpf_func_proto bpf_map_lookup_elem_proto;
 325 extern const struct bpf_func_proto bpf_map_update_elem_proto;
 326 extern const struct bpf_func_proto bpf_map_delete_elem_proto;
 327
 328 extern const struct bpf_func_proto bpf_get_prandom_u32_proto;
 329 extern const struct bpf_func_proto bpf_get_smp_processor_id_proto;
 330 extern const struct bpf_func_proto bpf_get_numa_node_id_proto;
 331 extern const struct bpf_func_proto bpf_tail_call_proto;
 332 extern const struct bpf_func_proto bpf_ktime_get_ns_proto;
 333 extern const struct bpf_func_proto bpf_get_current_pid_tgid_proto;
 334 extern const struct bpf_func_proto bpf_get_current_uid_gid_proto;
 335 extern const struct bpf_func_proto bpf_get_current_comm_proto;
 336 extern const struct bpf_func_proto bpf_skb_vlan_push_proto;
 337 extern const struct bpf_func_proto bpf_skb_vlan_pop_proto;
 338 extern const struct bpf_func_proto bpf_get_stackid_proto;
 339
 340 /* Shared helpers among cBPF and eBPF. */
 341 void bpf_user_rnd_init_once(void);
 342 u64 bpf_user_rnd_u32(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5);
 343
 344 #endif /* _LINUX_BPF_H */