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