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