1/* SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause) */ 2#ifndef __BPF_HELPERS__ 3#define __BPF_HELPERS__ 4 5/* 6 * Note that bpf programs need to include either 7 * vmlinux.h (auto-generated from BTF) or linux/types.h 8 * in advance since bpf_helper_defs.h uses such types 9 * as __u64. 10 */ 11#include "bpf_helper_defs.h" 12 13#define __uint(name, val) int (*name)[val] 14#define __type(name, val) typeof(val) *name 15#define __array(name, val) typeof(val) *name[] 16 17/* 18 * Helper macro to place programs, maps, license in 19 * different sections in elf_bpf file. Section names 20 * are interpreted by libbpf depending on the context (BPF programs, BPF maps, 21 * extern variables, etc). 22 * To allow use of SEC() with externs (e.g., for extern .maps declarations), 23 * make sure __attribute__((unused)) doesn't trigger compilation warning. 24 */ 25#if __GNUC__ && !__clang__ 26 27/* 28 * Pragma macros are broken on GCC 29 * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55578 30 * https://gcc.gnu.org/bugzilla/show_bug.cgi?id=90400 31 */ 32#define SEC(name) __attribute__((section(name), used)) 33 34#else 35 36#define SEC(name) \ 37 _Pragma("GCC diagnostic push") \ 38 _Pragma("GCC diagnostic ignored \"-Wignored-attributes\"") \ 39 __attribute__((section(name), used)) \ 40 _Pragma("GCC diagnostic pop") \ 41 42#endif 43 44/* Avoid 'linux/stddef.h' definition of '__always_inline'. */ 45#undef __always_inline 46#define __always_inline inline __attribute__((always_inline)) 47 48#ifndef __noinline 49#define __noinline __attribute__((noinline)) 50#endif 51#ifndef __weak 52#define __weak __attribute__((weak)) 53#endif 54 55/* 56 * Use __hidden attribute to mark a non-static BPF subprogram effectively 57 * static for BPF verifier's verification algorithm purposes, allowing more 58 * extensive and permissive BPF verification process, taking into account 59 * subprogram's caller context. 60 */ 61#define __hidden __attribute__((visibility("hidden"))) 62 63/* When utilizing vmlinux.h with BPF CO-RE, user BPF programs can't include 64 * any system-level headers (such as stddef.h, linux/version.h, etc), and 65 * commonly-used macros like NULL and KERNEL_VERSION aren't available through 66 * vmlinux.h. This just adds unnecessary hurdles and forces users to re-define 67 * them on their own. So as a convenience, provide such definitions here. 68 */ 69#ifndef NULL 70#define NULL ((void *)0) 71#endif 72 73#ifndef KERNEL_VERSION 74#define KERNEL_VERSION(a, b, c) (((a) << 16) + ((b) << 8) + ((c) > 255 ? 255 : (c))) 75#endif 76 77/* 78 * Helper macros to manipulate data structures 79 */ 80 81/* offsetof() definition that uses __builtin_offset() might not preserve field 82 * offset CO-RE relocation properly, so force-redefine offsetof() using 83 * old-school approach which works with CO-RE correctly 84 */ 85#undef offsetof 86#define offsetof(type, member) ((unsigned long)&((type *)0)->member) 87 88/* redefined container_of() to ensure we use the above offsetof() macro */ 89#undef container_of 90#define container_of(ptr, type, member) \ 91 ({ \ 92 void *__mptr = (void *)(ptr); \ 93 ((type *)(__mptr - offsetof(type, member))); \ 94 }) 95 96/* 97 * Compiler (optimization) barrier. 98 */ 99#ifndef barrier 100#define barrier() asm volatile("" ::: "memory") 101#endif 102 103/* Variable-specific compiler (optimization) barrier. It's a no-op which makes 104 * compiler believe that there is some black box modification of a given 105 * variable and thus prevents compiler from making extra assumption about its 106 * value and potential simplifications and optimizations on this variable. 107 * 108 * E.g., compiler might often delay or even omit 32-bit to 64-bit casting of 109 * a variable, making some code patterns unverifiable. Putting barrier_var() 110 * in place will ensure that cast is performed before the barrier_var() 111 * invocation, because compiler has to pessimistically assume that embedded 112 * asm section might perform some extra operations on that variable. 113 * 114 * This is a variable-specific variant of more global barrier(). 115 */ 116#ifndef barrier_var 117#define barrier_var(var) asm volatile("" : "+r"(var)) 118#endif 119 120/* 121 * Helper macro to throw a compilation error if __bpf_unreachable() gets 122 * built into the resulting code. This works given BPF back end does not 123 * implement __builtin_trap(). This is useful to assert that certain paths 124 * of the program code are never used and hence eliminated by the compiler. 125 * 126 * For example, consider a switch statement that covers known cases used by 127 * the program. __bpf_unreachable() can then reside in the default case. If 128 * the program gets extended such that a case is not covered in the switch 129 * statement, then it will throw a build error due to the default case not 130 * being compiled out. 131 */ 132#ifndef __bpf_unreachable 133# define __bpf_unreachable() __builtin_trap() 134#endif 135 136/* 137 * Helper function to perform a tail call with a constant/immediate map slot. 138 */ 139#if __clang_major__ >= 8 && defined(__bpf__) 140static __always_inline void 141bpf_tail_call_static(void *ctx, const void *map, const __u32 slot) 142{ 143 if (!__builtin_constant_p(slot)) 144 __bpf_unreachable(); 145 146 /* 147 * Provide a hard guarantee that LLVM won't optimize setting r2 (map 148 * pointer) and r3 (constant map index) from _different paths_ ending 149 * up at the _same_ call insn as otherwise we won't be able to use the 150 * jmpq/nopl retpoline-free patching by the x86-64 JIT in the kernel 151 * given they mismatch. See also d2e4c1e6c294 ("bpf: Constant map key 152 * tracking for prog array pokes") for details on verifier tracking. 153 * 154 * Note on clobber list: we need to stay in-line with BPF calling 155 * convention, so even if we don't end up using r0, r4, r5, we need 156 * to mark them as clobber so that LLVM doesn't end up using them 157 * before / after the call. 158 */ 159 asm volatile("r1 = %[ctx]\n\t" 160 "r2 = %[map]\n\t" 161 "r3 = %[slot]\n\t" 162 "call 12" 163 :: [ctx]"r"(ctx), [map]"r"(map), [slot]"i"(slot) 164 : "r0", "r1", "r2", "r3", "r4", "r5"); 165} 166#endif 167 168enum libbpf_pin_type { 169 LIBBPF_PIN_NONE, 170 /* PIN_BY_NAME: pin maps by name (in /sys/fs/bpf by default) */ 171 LIBBPF_PIN_BY_NAME, 172}; 173 174enum libbpf_tristate { 175 TRI_NO = 0, 176 TRI_YES = 1, 177 TRI_MODULE = 2, 178}; 179 180#define __kconfig __attribute__((section(".kconfig"))) 181#define __ksym __attribute__((section(".ksyms"))) 182#define __kptr_untrusted __attribute__((btf_type_tag("kptr_untrusted"))) 183#define __kptr __attribute__((btf_type_tag("kptr"))) 184#define __percpu_kptr __attribute__((btf_type_tag("percpu_kptr"))) 185 186#define bpf_ksym_exists(sym) ({ \ 187 _Static_assert(!__builtin_constant_p(!!sym), #sym " should be marked as __weak"); \ 188 !!sym; \ 189}) 190 191#ifndef ___bpf_concat 192#define ___bpf_concat(a, b) a ## b 193#endif 194#ifndef ___bpf_apply 195#define ___bpf_apply(fn, n) ___bpf_concat(fn, n) 196#endif 197#ifndef ___bpf_nth 198#define ___bpf_nth(_, _1, _2, _3, _4, _5, _6, _7, _8, _9, _a, _b, _c, N, ...) N 199#endif 200#ifndef ___bpf_narg 201#define ___bpf_narg(...) \ 202 ___bpf_nth(_, ##__VA_ARGS__, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0) 203#endif 204 205#define ___bpf_fill0(arr, p, x) do {} while (0) 206#define ___bpf_fill1(arr, p, x) arr[p] = x 207#define ___bpf_fill2(arr, p, x, args...) arr[p] = x; ___bpf_fill1(arr, p + 1, args) 208#define ___bpf_fill3(arr, p, x, args...) arr[p] = x; ___bpf_fill2(arr, p + 1, args) 209#define ___bpf_fill4(arr, p, x, args...) arr[p] = x; ___bpf_fill3(arr, p + 1, args) 210#define ___bpf_fill5(arr, p, x, args...) arr[p] = x; ___bpf_fill4(arr, p + 1, args) 211#define ___bpf_fill6(arr, p, x, args...) arr[p] = x; ___bpf_fill5(arr, p + 1, args) 212#define ___bpf_fill7(arr, p, x, args...) arr[p] = x; ___bpf_fill6(arr, p + 1, args) 213#define ___bpf_fill8(arr, p, x, args...) arr[p] = x; ___bpf_fill7(arr, p + 1, args) 214#define ___bpf_fill9(arr, p, x, args...) arr[p] = x; ___bpf_fill8(arr, p + 1, args) 215#define ___bpf_fill10(arr, p, x, args...) arr[p] = x; ___bpf_fill9(arr, p + 1, args) 216#define ___bpf_fill11(arr, p, x, args...) arr[p] = x; ___bpf_fill10(arr, p + 1, args) 217#define ___bpf_fill12(arr, p, x, args...) arr[p] = x; ___bpf_fill11(arr, p + 1, args) 218#define ___bpf_fill(arr, args...) \ 219 ___bpf_apply(___bpf_fill, ___bpf_narg(args))(arr, 0, args) 220 221/* 222 * BPF_SEQ_PRINTF to wrap bpf_seq_printf to-be-printed values 223 * in a structure. 224 */ 225#define BPF_SEQ_PRINTF(seq, fmt, args...) \ 226({ \ 227 static const char ___fmt[] = fmt; \ 228 unsigned long long ___param[___bpf_narg(args)]; \ 229 \ 230 _Pragma("GCC diagnostic push") \ 231 _Pragma("GCC diagnostic ignored \"-Wint-conversion\"") \ 232 ___bpf_fill(___param, args); \ 233 _Pragma("GCC diagnostic pop") \ 234 \ 235 bpf_seq_printf(seq, ___fmt, sizeof(___fmt), \ 236 ___param, sizeof(___param)); \ 237}) 238 239/* 240 * BPF_SNPRINTF wraps the bpf_snprintf helper with variadic arguments instead of 241 * an array of u64. 242 */ 243#define BPF_SNPRINTF(out, out_size, fmt, args...) \ 244({ \ 245 static const char ___fmt[] = fmt; \ 246 unsigned long long ___param[___bpf_narg(args)]; \ 247 \ 248 _Pragma("GCC diagnostic push") \ 249 _Pragma("GCC diagnostic ignored \"-Wint-conversion\"") \ 250 ___bpf_fill(___param, args); \ 251 _Pragma("GCC diagnostic pop") \ 252 \ 253 bpf_snprintf(out, out_size, ___fmt, \ 254 ___param, sizeof(___param)); \ 255}) 256 257#ifdef BPF_NO_GLOBAL_DATA 258#define BPF_PRINTK_FMT_MOD 259#else 260#define BPF_PRINTK_FMT_MOD static const 261#endif 262 263#define __bpf_printk(fmt, ...) \ 264({ \ 265 BPF_PRINTK_FMT_MOD char ____fmt[] = fmt; \ 266 bpf_trace_printk(____fmt, sizeof(____fmt), \ 267 ##__VA_ARGS__); \ 268}) 269 270/* 271 * __bpf_vprintk wraps the bpf_trace_vprintk helper with variadic arguments 272 * instead of an array of u64. 273 */ 274#define __bpf_vprintk(fmt, args...) \ 275({ \ 276 static const char ___fmt[] = fmt; \ 277 unsigned long long ___param[___bpf_narg(args)]; \ 278 \ 279 _Pragma("GCC diagnostic push") \ 280 _Pragma("GCC diagnostic ignored \"-Wint-conversion\"") \ 281 ___bpf_fill(___param, args); \ 282 _Pragma("GCC diagnostic pop") \ 283 \ 284 bpf_trace_vprintk(___fmt, sizeof(___fmt), \ 285 ___param, sizeof(___param)); \ 286}) 287 288/* Use __bpf_printk when bpf_printk call has 3 or fewer fmt args 289 * Otherwise use __bpf_vprintk 290 */ 291#define ___bpf_pick_printk(...) \ 292 ___bpf_nth(_, ##__VA_ARGS__, __bpf_vprintk, __bpf_vprintk, __bpf_vprintk, \ 293 __bpf_vprintk, __bpf_vprintk, __bpf_vprintk, __bpf_vprintk, \ 294 __bpf_vprintk, __bpf_vprintk, __bpf_printk /*3*/, __bpf_printk /*2*/,\ 295 __bpf_printk /*1*/, __bpf_printk /*0*/) 296 297/* Helper macro to print out debug messages */ 298#define bpf_printk(fmt, args...) ___bpf_pick_printk(args)(fmt, ##args) 299 300struct bpf_iter_num; 301 302extern int bpf_iter_num_new(struct bpf_iter_num *it, int start, int end) __weak __ksym; 303extern int *bpf_iter_num_next(struct bpf_iter_num *it) __weak __ksym; 304extern void bpf_iter_num_destroy(struct bpf_iter_num *it) __weak __ksym; 305 306#ifndef bpf_for_each 307/* bpf_for_each(iter_type, cur_elem, args...) provides generic construct for 308 * using BPF open-coded iterators without having to write mundane explicit 309 * low-level loop logic. Instead, it provides for()-like generic construct 310 * that can be used pretty naturally. E.g., for some hypothetical cgroup 311 * iterator, you'd write: 312 * 313 * struct cgroup *cg, *parent_cg = <...>; 314 * 315 * bpf_for_each(cgroup, cg, parent_cg, CG_ITER_CHILDREN) { 316 * bpf_printk("Child cgroup id = %d", cg->cgroup_id); 317 * if (cg->cgroup_id == 123) 318 * break; 319 * } 320 * 321 * I.e., it looks almost like high-level for each loop in other languages, 322 * supports continue/break, and is verifiable by BPF verifier. 323 * 324 * For iterating integers, the difference betwen bpf_for_each(num, i, N, M) 325 * and bpf_for(i, N, M) is in that bpf_for() provides additional proof to 326 * verifier that i is in [N, M) range, and in bpf_for_each() case i is `int 327 * *`, not just `int`. So for integers bpf_for() is more convenient. 328 * 329 * Note: this macro relies on C99 feature of allowing to declare variables 330 * inside for() loop, bound to for() loop lifetime. It also utilizes GCC 331 * extension: __attribute__((cleanup(<func>))), supported by both GCC and 332 * Clang. 333 */ 334#define bpf_for_each(type, cur, args...) for ( \ 335 /* initialize and define destructor */ \ 336 struct bpf_iter_##type ___it __attribute__((aligned(8), /* enforce, just in case */, \ 337 cleanup(bpf_iter_##type##_destroy))), \ 338 /* ___p pointer is just to call bpf_iter_##type##_new() *once* to init ___it */ \ 339 *___p __attribute__((unused)) = ( \ 340 bpf_iter_##type##_new(&___it, ##args), \ 341 /* this is a workaround for Clang bug: it currently doesn't emit BTF */ \ 342 /* for bpf_iter_##type##_destroy() when used from cleanup() attribute */ \ 343 (void)bpf_iter_##type##_destroy, (void *)0); \ 344 /* iteration and termination check */ \ 345 (((cur) = bpf_iter_##type##_next(&___it))); \ 346) 347#endif /* bpf_for_each */ 348 349#ifndef bpf_for 350/* bpf_for(i, start, end) implements a for()-like looping construct that sets 351 * provided integer variable *i* to values starting from *start* through, 352 * but not including, *end*. It also proves to BPF verifier that *i* belongs 353 * to range [start, end), so this can be used for accessing arrays without 354 * extra checks. 355 * 356 * Note: *start* and *end* are assumed to be expressions with no side effects 357 * and whose values do not change throughout bpf_for() loop execution. They do 358 * not have to be statically known or constant, though. 359 * 360 * Note: similarly to bpf_for_each(), it relies on C99 feature of declaring for() 361 * loop bound variables and cleanup attribute, supported by GCC and Clang. 362 */ 363#define bpf_for(i, start, end) for ( \ 364 /* initialize and define destructor */ \ 365 struct bpf_iter_num ___it __attribute__((aligned(8), /* enforce, just in case */ \ 366 cleanup(bpf_iter_num_destroy))), \ 367 /* ___p pointer is necessary to call bpf_iter_num_new() *once* to init ___it */ \ 368 *___p __attribute__((unused)) = ( \ 369 bpf_iter_num_new(&___it, (start), (end)), \ 370 /* this is a workaround for Clang bug: it currently doesn't emit BTF */ \ 371 /* for bpf_iter_num_destroy() when used from cleanup() attribute */ \ 372 (void)bpf_iter_num_destroy, (void *)0); \ 373 ({ \ 374 /* iteration step */ \ 375 int *___t = bpf_iter_num_next(&___it); \ 376 /* termination and bounds check */ \ 377 (___t && ((i) = *___t, (i) >= (start) && (i) < (end))); \ 378 }); \ 379) 380#endif /* bpf_for */ 381 382#ifndef bpf_repeat 383/* bpf_repeat(N) performs N iterations without exposing iteration number 384 * 385 * Note: similarly to bpf_for_each(), it relies on C99 feature of declaring for() 386 * loop bound variables and cleanup attribute, supported by GCC and Clang. 387 */ 388#define bpf_repeat(N) for ( \ 389 /* initialize and define destructor */ \ 390 struct bpf_iter_num ___it __attribute__((aligned(8), /* enforce, just in case */ \ 391 cleanup(bpf_iter_num_destroy))), \ 392 /* ___p pointer is necessary to call bpf_iter_num_new() *once* to init ___it */ \ 393 *___p __attribute__((unused)) = ( \ 394 bpf_iter_num_new(&___it, 0, (N)), \ 395 /* this is a workaround for Clang bug: it currently doesn't emit BTF */ \ 396 /* for bpf_iter_num_destroy() when used from cleanup() attribute */ \ 397 (void)bpf_iter_num_destroy, (void *)0); \ 398 bpf_iter_num_next(&___it); \ 399 /* nothing here */ \ 400) 401#endif /* bpf_repeat */ 402 403#endif 404