1 #ifndef __BPF_EXPERIMENTAL__ 2 #define __BPF_EXPERIMENTAL__ 3 4 #include <vmlinux.h> 5 #include <bpf/bpf_tracing.h> 6 #include <bpf/bpf_helpers.h> 7 #include <bpf/bpf_core_read.h> 8 9 #define __contains(name, node) __attribute__((btf_decl_tag("contains:" #name ":" #node))) 10 11 /* Description 12 * Allocates an object of the type represented by 'local_type_id' in 13 * program BTF. User may use the bpf_core_type_id_local macro to pass the 14 * type ID of a struct in program BTF. 15 * 16 * The 'local_type_id' parameter must be a known constant. 17 * The 'meta' parameter is rewritten by the verifier, no need for BPF 18 * program to set it. 19 * Returns 20 * A pointer to an object of the type corresponding to the passed in 21 * 'local_type_id', or NULL on failure. 22 */ 23 extern void *bpf_obj_new_impl(__u64 local_type_id, void *meta) __ksym; 24 25 /* Convenience macro to wrap over bpf_obj_new_impl */ 26 #define bpf_obj_new(type) ((type *)bpf_obj_new_impl(bpf_core_type_id_local(type), NULL)) 27 28 /* Description 29 * Free an allocated object. All fields of the object that require 30 * destruction will be destructed before the storage is freed. 31 * 32 * The 'meta' parameter is rewritten by the verifier, no need for BPF 33 * program to set it. 34 * Returns 35 * Void. 36 */ 37 extern void bpf_obj_drop_impl(void *kptr, void *meta) __ksym; 38 39 /* Convenience macro to wrap over bpf_obj_drop_impl */ 40 #define bpf_obj_drop(kptr) bpf_obj_drop_impl(kptr, NULL) 41 42 /* Description 43 * Increment the refcount on a refcounted local kptr, turning the 44 * non-owning reference input into an owning reference in the process. 45 * 46 * The 'meta' parameter is rewritten by the verifier, no need for BPF 47 * program to set it. 48 * Returns 49 * An owning reference to the object pointed to by 'kptr' 50 */ 51 extern void *bpf_refcount_acquire_impl(void *kptr, void *meta) __ksym; 52 53 /* Convenience macro to wrap over bpf_refcount_acquire_impl */ 54 #define bpf_refcount_acquire(kptr) bpf_refcount_acquire_impl(kptr, NULL) 55 56 /* Description 57 * Add a new entry to the beginning of the BPF linked list. 58 * 59 * The 'meta' and 'off' parameters are rewritten by the verifier, no need 60 * for BPF programs to set them 61 * Returns 62 * 0 if the node was successfully added 63 * -EINVAL if the node wasn't added because it's already in a list 64 */ 65 extern int bpf_list_push_front_impl(struct bpf_list_head *head, 66 struct bpf_list_node *node, 67 void *meta, __u64 off) __ksym; 68 69 /* Convenience macro to wrap over bpf_list_push_front_impl */ 70 #define bpf_list_push_front(head, node) bpf_list_push_front_impl(head, node, NULL, 0) 71 72 /* Description 73 * Add a new entry to the end of the BPF linked list. 74 * 75 * The 'meta' and 'off' parameters are rewritten by the verifier, no need 76 * for BPF programs to set them 77 * Returns 78 * 0 if the node was successfully added 79 * -EINVAL if the node wasn't added because it's already in a list 80 */ 81 extern int bpf_list_push_back_impl(struct bpf_list_head *head, 82 struct bpf_list_node *node, 83 void *meta, __u64 off) __ksym; 84 85 /* Convenience macro to wrap over bpf_list_push_back_impl */ 86 #define bpf_list_push_back(head, node) bpf_list_push_back_impl(head, node, NULL, 0) 87 88 /* Description 89 * Remove the entry at the beginning of the BPF linked list. 90 * Returns 91 * Pointer to bpf_list_node of deleted entry, or NULL if list is empty. 92 */ 93 extern struct bpf_list_node *bpf_list_pop_front(struct bpf_list_head *head) __ksym; 94 95 /* Description 96 * Remove the entry at the end of the BPF linked list. 97 * Returns 98 * Pointer to bpf_list_node of deleted entry, or NULL if list is empty. 99 */ 100 extern struct bpf_list_node *bpf_list_pop_back(struct bpf_list_head *head) __ksym; 101 102 /* Description 103 * Remove 'node' from rbtree with root 'root' 104 * Returns 105 * Pointer to the removed node, or NULL if 'root' didn't contain 'node' 106 */ 107 extern struct bpf_rb_node *bpf_rbtree_remove(struct bpf_rb_root *root, 108 struct bpf_rb_node *node) __ksym; 109 110 /* Description 111 * Add 'node' to rbtree with root 'root' using comparator 'less' 112 * 113 * The 'meta' and 'off' parameters are rewritten by the verifier, no need 114 * for BPF programs to set them 115 * Returns 116 * 0 if the node was successfully added 117 * -EINVAL if the node wasn't added because it's already in a tree 118 */ 119 extern int bpf_rbtree_add_impl(struct bpf_rb_root *root, struct bpf_rb_node *node, 120 bool (less)(struct bpf_rb_node *a, const struct bpf_rb_node *b), 121 void *meta, __u64 off) __ksym; 122 123 /* Convenience macro to wrap over bpf_rbtree_add_impl */ 124 #define bpf_rbtree_add(head, node, less) bpf_rbtree_add_impl(head, node, less, NULL, 0) 125 126 /* Description 127 * Return the first (leftmost) node in input tree 128 * Returns 129 * Pointer to the node, which is _not_ removed from the tree. If the tree 130 * contains no nodes, returns NULL. 131 */ 132 extern struct bpf_rb_node *bpf_rbtree_first(struct bpf_rb_root *root) __ksym; 133 134 /* Description 135 * Allocates a percpu object of the type represented by 'local_type_id' in 136 * program BTF. User may use the bpf_core_type_id_local macro to pass the 137 * type ID of a struct in program BTF. 138 * 139 * The 'local_type_id' parameter must be a known constant. 140 * The 'meta' parameter is rewritten by the verifier, no need for BPF 141 * program to set it. 142 * Returns 143 * A pointer to a percpu object of the type corresponding to the passed in 144 * 'local_type_id', or NULL on failure. 145 */ 146 extern void *bpf_percpu_obj_new_impl(__u64 local_type_id, void *meta) __ksym; 147 148 /* Convenience macro to wrap over bpf_percpu_obj_new_impl */ 149 #define bpf_percpu_obj_new(type) ((type __percpu_kptr *)bpf_percpu_obj_new_impl(bpf_core_type_id_local(type), NULL)) 150 151 /* Description 152 * Free an allocated percpu object. All fields of the object that require 153 * destruction will be destructed before the storage is freed. 154 * 155 * The 'meta' parameter is rewritten by the verifier, no need for BPF 156 * program to set it. 157 * Returns 158 * Void. 159 */ 160 extern void bpf_percpu_obj_drop_impl(void *kptr, void *meta) __ksym; 161 162 struct bpf_iter_task_vma; 163 164 extern int bpf_iter_task_vma_new(struct bpf_iter_task_vma *it, 165 struct task_struct *task, 166 __u64 addr) __ksym; 167 extern struct vm_area_struct *bpf_iter_task_vma_next(struct bpf_iter_task_vma *it) __ksym; 168 extern void bpf_iter_task_vma_destroy(struct bpf_iter_task_vma *it) __ksym; 169 170 /* Convenience macro to wrap over bpf_obj_drop_impl */ 171 #define bpf_percpu_obj_drop(kptr) bpf_percpu_obj_drop_impl(kptr, NULL) 172 173 /* Description 174 * Throw a BPF exception from the program, immediately terminating its 175 * execution and unwinding the stack. The supplied 'cookie' parameter 176 * will be the return value of the program when an exception is thrown, 177 * and the default exception callback is used. Otherwise, if an exception 178 * callback is set using the '__exception_cb(callback)' declaration tag 179 * on the main program, the 'cookie' parameter will be the callback's only 180 * input argument. 181 * 182 * Thus, in case of default exception callback, 'cookie' is subjected to 183 * constraints on the program's return value (as with R0 on exit). 184 * Otherwise, the return value of the marked exception callback will be 185 * subjected to the same checks. 186 * 187 * Note that throwing an exception with lingering resources (locks, 188 * references, etc.) will lead to a verification error. 189 * 190 * Note that callbacks *cannot* call this helper. 191 * Returns 192 * Never. 193 * Throws 194 * An exception with the specified 'cookie' value. 195 */ 196 extern void bpf_throw(u64 cookie) __ksym; 197 198 /* This macro must be used to mark the exception callback corresponding to the 199 * main program. For example: 200 * 201 * int exception_cb(u64 cookie) { 202 * return cookie; 203 * } 204 * 205 * SEC("tc") 206 * __exception_cb(exception_cb) 207 * int main_prog(struct __sk_buff *ctx) { 208 * ... 209 * return TC_ACT_OK; 210 * } 211 * 212 * Here, exception callback for the main program will be 'exception_cb'. Note 213 * that this attribute can only be used once, and multiple exception callbacks 214 * specified for the main program will lead to verification error. 215 */ 216 #define __exception_cb(name) __attribute__((btf_decl_tag("exception_callback:" #name))) 217 218 #define __bpf_assert_signed(x) _Generic((x), \ 219 unsigned long: 0, \ 220 unsigned long long: 0, \ 221 signed long: 1, \ 222 signed long long: 1 \ 223 ) 224 225 #define __bpf_assert_check(LHS, op, RHS) \ 226 _Static_assert(sizeof(&(LHS)), "1st argument must be an lvalue expression"); \ 227 _Static_assert(sizeof(LHS) == 8, "Only 8-byte integers are supported\n"); \ 228 _Static_assert(__builtin_constant_p(__bpf_assert_signed(LHS)), "internal static assert"); \ 229 _Static_assert(__builtin_constant_p((RHS)), "2nd argument must be a constant expression") 230 231 #define __bpf_assert(LHS, op, cons, RHS, VAL) \ 232 ({ \ 233 (void)bpf_throw; \ 234 asm volatile ("if %[lhs] " op " %[rhs] goto +2; r1 = %[value]; call bpf_throw" \ 235 : : [lhs] "r"(LHS), [rhs] cons(RHS), [value] "ri"(VAL) : ); \ 236 }) 237 238 #define __bpf_assert_op_sign(LHS, op, cons, RHS, VAL, supp_sign) \ 239 ({ \ 240 __bpf_assert_check(LHS, op, RHS); \ 241 if (__bpf_assert_signed(LHS) && !(supp_sign)) \ 242 __bpf_assert(LHS, "s" #op, cons, RHS, VAL); \ 243 else \ 244 __bpf_assert(LHS, #op, cons, RHS, VAL); \ 245 }) 246 247 #define __bpf_assert_op(LHS, op, RHS, VAL, supp_sign) \ 248 ({ \ 249 if (sizeof(typeof(RHS)) == 8) { \ 250 const typeof(RHS) rhs_var = (RHS); \ 251 __bpf_assert_op_sign(LHS, op, "r", rhs_var, VAL, supp_sign); \ 252 } else { \ 253 __bpf_assert_op_sign(LHS, op, "i", RHS, VAL, supp_sign); \ 254 } \ 255 }) 256 257 #define __cmp_cannot_be_signed(x) \ 258 __builtin_strcmp(#x, "==") == 0 || __builtin_strcmp(#x, "!=") == 0 || \ 259 __builtin_strcmp(#x, "&") == 0 260 261 #define __is_signed_type(type) (((type)(-1)) < (type)1) 262 263 #define __bpf_cmp(LHS, OP, PRED, RHS, DEFAULT) \ 264 ({ \ 265 __label__ l_true; \ 266 bool ret = DEFAULT; \ 267 asm volatile goto("if %[lhs] " OP " %[rhs] goto %l[l_true]" \ 268 :: [lhs] "r"((short)LHS), [rhs] PRED (RHS) :: l_true); \ 269 ret = !DEFAULT; \ 270 l_true: \ 271 ret; \ 272 }) 273 274 /* C type conversions coupled with comparison operator are tricky. 275 * Make sure BPF program is compiled with -Wsign-compare then 276 * __lhs OP __rhs below will catch the mistake. 277 * Be aware that we check only __lhs to figure out the sign of compare. 278 */ 279 #define _bpf_cmp(LHS, OP, RHS, UNLIKELY) \ 280 ({ \ 281 typeof(LHS) __lhs = (LHS); \ 282 typeof(RHS) __rhs = (RHS); \ 283 bool ret; \ 284 _Static_assert(sizeof(&(LHS)), "1st argument must be an lvalue expression"); \ 285 (void)(__lhs OP __rhs); \ 286 if (__cmp_cannot_be_signed(OP) || !__is_signed_type(typeof(__lhs))) { \ 287 if (sizeof(__rhs) == 8) \ 288 /* "i" will truncate 64-bit constant into s32, \ 289 * so we have to use extra register via "r". \ 290 */ \ 291 ret = __bpf_cmp(__lhs, #OP, "r", __rhs, UNLIKELY); \ 292 else \ 293 ret = __bpf_cmp(__lhs, #OP, "ri", __rhs, UNLIKELY); \ 294 } else { \ 295 if (sizeof(__rhs) == 8) \ 296 ret = __bpf_cmp(__lhs, "s"#OP, "r", __rhs, UNLIKELY); \ 297 else \ 298 ret = __bpf_cmp(__lhs, "s"#OP, "ri", __rhs, UNLIKELY); \ 299 } \ 300 ret; \ 301 }) 302 303 #ifndef bpf_cmp_unlikely 304 #define bpf_cmp_unlikely(LHS, OP, RHS) _bpf_cmp(LHS, OP, RHS, true) 305 #endif 306 307 #ifndef bpf_cmp_likely 308 #define bpf_cmp_likely(LHS, OP, RHS) \ 309 ({ \ 310 bool ret = 0; \ 311 if (__builtin_strcmp(#OP, "==") == 0) \ 312 ret = _bpf_cmp(LHS, !=, RHS, false); \ 313 else if (__builtin_strcmp(#OP, "!=") == 0) \ 314 ret = _bpf_cmp(LHS, ==, RHS, false); \ 315 else if (__builtin_strcmp(#OP, "<=") == 0) \ 316 ret = _bpf_cmp(LHS, >, RHS, false); \ 317 else if (__builtin_strcmp(#OP, "<") == 0) \ 318 ret = _bpf_cmp(LHS, >=, RHS, false); \ 319 else if (__builtin_strcmp(#OP, ">") == 0) \ 320 ret = _bpf_cmp(LHS, <=, RHS, false); \ 321 else if (__builtin_strcmp(#OP, ">=") == 0) \ 322 ret = _bpf_cmp(LHS, <, RHS, false); \ 323 else \ 324 asm volatile("r0 " #OP " invalid compare"); \ 325 ret; \ 326 }) 327 #endif 328 329 /* 330 * Note that cond_break can only be portably used in the body of a breakable 331 * construct, whereas can_loop can be used anywhere. 332 */ 333 #ifdef __BPF_FEATURE_MAY_GOTO 334 #define can_loop \ 335 ({ __label__ l_break, l_continue; \ 336 bool ret = true; \ 337 asm volatile goto("may_goto %l[l_break]" \ 338 :::: l_break); \ 339 goto l_continue; \ 340 l_break: ret = false; \ 341 l_continue:; \ 342 ret; \ 343 }) 344 345 #define cond_break \ 346 ({ __label__ l_break, l_continue; \ 347 asm volatile goto("may_goto %l[l_break]" \ 348 :::: l_break); \ 349 goto l_continue; \ 350 l_break: break; \ 351 l_continue:; \ 352 }) 353 #else 354 #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ 355 #define can_loop \ 356 ({ __label__ l_break, l_continue; \ 357 bool ret = true; \ 358 asm volatile goto("1:.byte 0xe5; \ 359 .byte 0; \ 360 .long ((%l[l_break] - 1b - 8) / 8) & 0xffff; \ 361 .short 0" \ 362 :::: l_break); \ 363 goto l_continue; \ 364 l_break: ret = false; \ 365 l_continue:; \ 366 ret; \ 367 }) 368 369 #define cond_break \ 370 ({ __label__ l_break, l_continue; \ 371 asm volatile goto("1:.byte 0xe5; \ 372 .byte 0; \ 373 .long ((%l[l_break] - 1b - 8) / 8) & 0xffff; \ 374 .short 0" \ 375 :::: l_break); \ 376 goto l_continue; \ 377 l_break: break; \ 378 l_continue:; \ 379 }) 380 #else 381 #define can_loop \ 382 ({ __label__ l_break, l_continue; \ 383 bool ret = true; \ 384 asm volatile goto("1:.byte 0xe5; \ 385 .byte 0; \ 386 .long (((%l[l_break] - 1b - 8) / 8) & 0xffff) << 16; \ 387 .short 0" \ 388 :::: l_break); \ 389 goto l_continue; \ 390 l_break: ret = false; \ 391 l_continue:; \ 392 ret; \ 393 }) 394 395 #define cond_break \ 396 ({ __label__ l_break, l_continue; \ 397 asm volatile goto("1:.byte 0xe5; \ 398 .byte 0; \ 399 .long (((%l[l_break] - 1b - 8) / 8) & 0xffff) << 16; \ 400 .short 0" \ 401 :::: l_break); \ 402 goto l_continue; \ 403 l_break: break; \ 404 l_continue:; \ 405 }) 406 #endif 407 #endif 408 409 #ifndef bpf_nop_mov 410 #define bpf_nop_mov(var) \ 411 asm volatile("%[reg]=%[reg]"::[reg]"r"((short)var)) 412 #endif 413 414 /* emit instruction: 415 * rX = rX .off = BPF_ADDR_SPACE_CAST .imm32 = (dst_as << 16) | src_as 416 */ 417 #ifndef bpf_addr_space_cast 418 #define bpf_addr_space_cast(var, dst_as, src_as)\ 419 asm volatile(".byte 0xBF; \ 420 .ifc %[reg], r0; \ 421 .byte 0x00; \ 422 .endif; \ 423 .ifc %[reg], r1; \ 424 .byte 0x11; \ 425 .endif; \ 426 .ifc %[reg], r2; \ 427 .byte 0x22; \ 428 .endif; \ 429 .ifc %[reg], r3; \ 430 .byte 0x33; \ 431 .endif; \ 432 .ifc %[reg], r4; \ 433 .byte 0x44; \ 434 .endif; \ 435 .ifc %[reg], r5; \ 436 .byte 0x55; \ 437 .endif; \ 438 .ifc %[reg], r6; \ 439 .byte 0x66; \ 440 .endif; \ 441 .ifc %[reg], r7; \ 442 .byte 0x77; \ 443 .endif; \ 444 .ifc %[reg], r8; \ 445 .byte 0x88; \ 446 .endif; \ 447 .ifc %[reg], r9; \ 448 .byte 0x99; \ 449 .endif; \ 450 .short %[off]; \ 451 .long %[as]" \ 452 : [reg]"+r"(var) \ 453 : [off]"i"(BPF_ADDR_SPACE_CAST) \ 454 , [as]"i"((dst_as << 16) | src_as)); 455 #endif 456 457 void bpf_preempt_disable(void) __weak __ksym; 458 void bpf_preempt_enable(void) __weak __ksym; 459 460 typedef struct { 461 } __bpf_preempt_t; 462 463 static inline __bpf_preempt_t __bpf_preempt_constructor(void) 464 { 465 __bpf_preempt_t ret = {}; 466 467 bpf_preempt_disable(); 468 return ret; 469 } 470 static inline void __bpf_preempt_destructor(__bpf_preempt_t *t) 471 { 472 bpf_preempt_enable(); 473 } 474 #define bpf_guard_preempt() \ 475 __bpf_preempt_t ___bpf_apply(preempt, __COUNTER__) \ 476 __attribute__((__unused__, __cleanup__(__bpf_preempt_destructor))) = \ 477 __bpf_preempt_constructor() 478 479 /* Description 480 * Assert that a conditional expression is true. 481 * Returns 482 * Void. 483 * Throws 484 * An exception with the value zero when the assertion fails. 485 */ 486 #define bpf_assert(cond) if (!(cond)) bpf_throw(0); 487 488 /* Description 489 * Assert that a conditional expression is true. 490 * Returns 491 * Void. 492 * Throws 493 * An exception with the specified value when the assertion fails. 494 */ 495 #define bpf_assert_with(cond, value) if (!(cond)) bpf_throw(value); 496 497 /* Description 498 * Assert that LHS is in the range [BEG, END] (inclusive of both). This 499 * statement updates the known bounds of LHS during verification. Note 500 * that both BEG and END must be constant values, and must fit within the 501 * data type of LHS. 502 * Returns 503 * Void. 504 * Throws 505 * An exception with the value zero when the assertion fails. 506 */ 507 #define bpf_assert_range(LHS, BEG, END) \ 508 ({ \ 509 _Static_assert(BEG <= END, "BEG must be <= END"); \ 510 barrier_var(LHS); \ 511 __bpf_assert_op(LHS, >=, BEG, 0, false); \ 512 __bpf_assert_op(LHS, <=, END, 0, false); \ 513 }) 514 515 /* Description 516 * Assert that LHS is in the range [BEG, END] (inclusive of both). This 517 * statement updates the known bounds of LHS during verification. Note 518 * that both BEG and END must be constant values, and must fit within the 519 * data type of LHS. 520 * Returns 521 * Void. 522 * Throws 523 * An exception with the specified value when the assertion fails. 524 */ 525 #define bpf_assert_range_with(LHS, BEG, END, value) \ 526 ({ \ 527 _Static_assert(BEG <= END, "BEG must be <= END"); \ 528 barrier_var(LHS); \ 529 __bpf_assert_op(LHS, >=, BEG, value, false); \ 530 __bpf_assert_op(LHS, <=, END, value, false); \ 531 }) 532 533 struct bpf_iter_css_task; 534 struct cgroup_subsys_state; 535 extern int bpf_iter_css_task_new(struct bpf_iter_css_task *it, 536 struct cgroup_subsys_state *css, unsigned int flags) __weak __ksym; 537 extern struct task_struct *bpf_iter_css_task_next(struct bpf_iter_css_task *it) __weak __ksym; 538 extern void bpf_iter_css_task_destroy(struct bpf_iter_css_task *it) __weak __ksym; 539 540 struct bpf_iter_task; 541 extern int bpf_iter_task_new(struct bpf_iter_task *it, 542 struct task_struct *task, unsigned int flags) __weak __ksym; 543 extern struct task_struct *bpf_iter_task_next(struct bpf_iter_task *it) __weak __ksym; 544 extern void bpf_iter_task_destroy(struct bpf_iter_task *it) __weak __ksym; 545 546 struct bpf_iter_css; 547 extern int bpf_iter_css_new(struct bpf_iter_css *it, 548 struct cgroup_subsys_state *start, unsigned int flags) __weak __ksym; 549 extern struct cgroup_subsys_state *bpf_iter_css_next(struct bpf_iter_css *it) __weak __ksym; 550 extern void bpf_iter_css_destroy(struct bpf_iter_css *it) __weak __ksym; 551 552 extern int bpf_wq_init(struct bpf_wq *wq, void *p__map, unsigned int flags) __weak __ksym; 553 extern int bpf_wq_start(struct bpf_wq *wq, unsigned int flags) __weak __ksym; 554 extern int bpf_wq_set_callback_impl(struct bpf_wq *wq, 555 int (callback_fn)(void *map, int *key, void *value), 556 unsigned int flags__k, void *aux__ign) __ksym; 557 #define bpf_wq_set_callback(timer, cb, flags) \ 558 bpf_wq_set_callback_impl(timer, cb, flags, NULL) 559 #endif 560