1 // SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
2 /* Copyright (c) 2018 Facebook */
12 #include <sys/utsname.h>
13 #include <sys/param.h>
15 #include <linux/kernel.h>
16 #include <linux/err.h>
17 #include <linux/btf.h>
22 #include "libbpf_internal.h"
26 #define BTF_MAX_NR_TYPES 0x7fffffffU
27 #define BTF_MAX_STR_OFFSET 0x7fffffffU
29 static struct btf_type btf_void;
32 /* raw BTF data in native endianness */
34 /* raw BTF data in non-native endianness */
35 void *raw_data_swapped;
37 /* whether target endianness differs from the native one */
41 * When BTF is loaded from an ELF or raw memory it is stored
42 * in a contiguous memory block. The hdr, type_data, and, strs_data
43 * point inside that memory region to their respective parts of BTF
46 * +--------------------------------+
47 * | Header | Types | Strings |
48 * +--------------------------------+
53 * strs_data------------+
55 * If BTF data is later modified, e.g., due to types added or
56 * removed, BTF deduplication performed, etc, this contiguous
57 * representation is broken up into three independently allocated
58 * memory regions to be able to modify them independently.
59 * raw_data is nulled out at that point, but can be later allocated
60 * and cached again if user calls btf__raw_data(), at which point
61 * raw_data will contain a contiguous copy of header, types, and
64 * +----------+ +---------+ +-----------+
65 * | Header | | Types | | Strings |
66 * +----------+ +---------+ +-----------+
71 * strset__data(strs_set)-----+
73 * +----------+---------+-----------+
74 * | Header | Types | Strings |
75 * raw_data----->+----------+---------+-----------+
77 struct btf_header *hdr;
80 size_t types_data_cap; /* used size stored in hdr->type_len */
82 /* type ID to `struct btf_type *` lookup index
83 * type_offs[0] corresponds to the first non-VOID type:
84 * - for base BTF it's type [1];
85 * - for split BTF it's the first non-base BTF type.
89 /* number of types in this BTF instance:
90 * - doesn't include special [0] void type;
91 * - for split BTF counts number of types added on top of base BTF.
94 /* if not NULL, points to the base BTF on top of which the current
98 /* BTF type ID of the first type in this BTF instance:
99 * - for base BTF it's equal to 1;
100 * - for split BTF it's equal to biggest type ID of base BTF plus 1.
103 /* logical string offset of this BTF instance:
104 * - for base BTF it's equal to 0;
105 * - for split BTF it's equal to total size of base BTF's string section size.
109 /* only one of strs_data or strs_set can be non-NULL, depending on
110 * whether BTF is in a modifiable state (strs_set is used) or not
111 * (strs_data points inside raw_data)
114 /* a set of unique strings */
115 struct strset *strs_set;
116 /* whether strings are already deduplicated */
119 /* BTF object FD, if loaded into kernel */
122 /* Pointer size (in bytes) for a target architecture of this BTF */
126 static inline __u64 ptr_to_u64(const void *ptr)
128 return (__u64) (unsigned long) ptr;
131 /* Ensure given dynamically allocated memory region pointed to by *data* with
132 * capacity of *cap_cnt* elements each taking *elem_sz* bytes has enough
133 * memory to accommodate *add_cnt* new elements, assuming *cur_cnt* elements
134 * are already used. At most *max_cnt* elements can be ever allocated.
135 * If necessary, memory is reallocated and all existing data is copied over,
136 * new pointer to the memory region is stored at *data, new memory region
137 * capacity (in number of elements) is stored in *cap.
138 * On success, memory pointer to the beginning of unused memory is returned.
139 * On error, NULL is returned.
141 void *libbpf_add_mem(void **data, size_t *cap_cnt, size_t elem_sz,
142 size_t cur_cnt, size_t max_cnt, size_t add_cnt)
147 if (cur_cnt + add_cnt <= *cap_cnt)
148 return *data + cur_cnt * elem_sz;
150 /* requested more than the set limit */
151 if (cur_cnt + add_cnt > max_cnt)
155 new_cnt += new_cnt / 4; /* expand by 25% */
156 if (new_cnt < 16) /* but at least 16 elements */
158 if (new_cnt > max_cnt) /* but not exceeding a set limit */
160 if (new_cnt < cur_cnt + add_cnt) /* also ensure we have enough memory */
161 new_cnt = cur_cnt + add_cnt;
163 new_data = libbpf_reallocarray(*data, new_cnt, elem_sz);
167 /* zero out newly allocated portion of memory */
168 memset(new_data + (*cap_cnt) * elem_sz, 0, (new_cnt - *cap_cnt) * elem_sz);
172 return new_data + cur_cnt * elem_sz;
175 /* Ensure given dynamically allocated memory region has enough allocated space
176 * to accommodate *need_cnt* elements of size *elem_sz* bytes each
178 int libbpf_ensure_mem(void **data, size_t *cap_cnt, size_t elem_sz, size_t need_cnt)
182 if (need_cnt <= *cap_cnt)
185 p = libbpf_add_mem(data, cap_cnt, elem_sz, *cap_cnt, SIZE_MAX, need_cnt - *cap_cnt);
192 static void *btf_add_type_offs_mem(struct btf *btf, size_t add_cnt)
194 return libbpf_add_mem((void **)&btf->type_offs, &btf->type_offs_cap, sizeof(__u32),
195 btf->nr_types, BTF_MAX_NR_TYPES, add_cnt);
198 static int btf_add_type_idx_entry(struct btf *btf, __u32 type_off)
202 p = btf_add_type_offs_mem(btf, 1);
210 static void btf_bswap_hdr(struct btf_header *h)
212 h->magic = bswap_16(h->magic);
213 h->hdr_len = bswap_32(h->hdr_len);
214 h->type_off = bswap_32(h->type_off);
215 h->type_len = bswap_32(h->type_len);
216 h->str_off = bswap_32(h->str_off);
217 h->str_len = bswap_32(h->str_len);
220 static int btf_parse_hdr(struct btf *btf)
222 struct btf_header *hdr = btf->hdr;
225 if (btf->raw_size < sizeof(struct btf_header)) {
226 pr_debug("BTF header not found\n");
230 if (hdr->magic == bswap_16(BTF_MAGIC)) {
231 btf->swapped_endian = true;
232 if (bswap_32(hdr->hdr_len) != sizeof(struct btf_header)) {
233 pr_warn("Can't load BTF with non-native endianness due to unsupported header length %u\n",
234 bswap_32(hdr->hdr_len));
238 } else if (hdr->magic != BTF_MAGIC) {
239 pr_debug("Invalid BTF magic: %x\n", hdr->magic);
243 if (btf->raw_size < hdr->hdr_len) {
244 pr_debug("BTF header len %u larger than data size %u\n",
245 hdr->hdr_len, btf->raw_size);
249 meta_left = btf->raw_size - hdr->hdr_len;
250 if (meta_left < (long long)hdr->str_off + hdr->str_len) {
251 pr_debug("Invalid BTF total size: %u\n", btf->raw_size);
255 if ((long long)hdr->type_off + hdr->type_len > hdr->str_off) {
256 pr_debug("Invalid BTF data sections layout: type data at %u + %u, strings data at %u + %u\n",
257 hdr->type_off, hdr->type_len, hdr->str_off, hdr->str_len);
261 if (hdr->type_off % 4) {
262 pr_debug("BTF type section is not aligned to 4 bytes\n");
269 static int btf_parse_str_sec(struct btf *btf)
271 const struct btf_header *hdr = btf->hdr;
272 const char *start = btf->strs_data;
273 const char *end = start + btf->hdr->str_len;
275 if (btf->base_btf && hdr->str_len == 0)
277 if (!hdr->str_len || hdr->str_len - 1 > BTF_MAX_STR_OFFSET || end[-1]) {
278 pr_debug("Invalid BTF string section\n");
281 if (!btf->base_btf && start[0]) {
282 pr_debug("Invalid BTF string section\n");
288 static int btf_type_size(const struct btf_type *t)
290 const int base_size = sizeof(struct btf_type);
291 __u16 vlen = btf_vlen(t);
293 switch (btf_kind(t)) {
296 case BTF_KIND_VOLATILE:
297 case BTF_KIND_RESTRICT:
299 case BTF_KIND_TYPEDEF:
302 case BTF_KIND_TYPE_TAG:
305 return base_size + sizeof(__u32);
307 return base_size + vlen * sizeof(struct btf_enum);
308 case BTF_KIND_ENUM64:
309 return base_size + vlen * sizeof(struct btf_enum64);
311 return base_size + sizeof(struct btf_array);
312 case BTF_KIND_STRUCT:
314 return base_size + vlen * sizeof(struct btf_member);
315 case BTF_KIND_FUNC_PROTO:
316 return base_size + vlen * sizeof(struct btf_param);
318 return base_size + sizeof(struct btf_var);
319 case BTF_KIND_DATASEC:
320 return base_size + vlen * sizeof(struct btf_var_secinfo);
321 case BTF_KIND_DECL_TAG:
322 return base_size + sizeof(struct btf_decl_tag);
324 pr_debug("Unsupported BTF_KIND:%u\n", btf_kind(t));
329 static void btf_bswap_type_base(struct btf_type *t)
331 t->name_off = bswap_32(t->name_off);
332 t->info = bswap_32(t->info);
333 t->type = bswap_32(t->type);
336 static int btf_bswap_type_rest(struct btf_type *t)
338 struct btf_var_secinfo *v;
339 struct btf_enum64 *e64;
340 struct btf_member *m;
344 __u16 vlen = btf_vlen(t);
347 switch (btf_kind(t)) {
350 case BTF_KIND_VOLATILE:
351 case BTF_KIND_RESTRICT:
353 case BTF_KIND_TYPEDEF:
356 case BTF_KIND_TYPE_TAG:
359 *(__u32 *)(t + 1) = bswap_32(*(__u32 *)(t + 1));
362 for (i = 0, e = btf_enum(t); i < vlen; i++, e++) {
363 e->name_off = bswap_32(e->name_off);
364 e->val = bswap_32(e->val);
367 case BTF_KIND_ENUM64:
368 for (i = 0, e64 = btf_enum64(t); i < vlen; i++, e64++) {
369 e64->name_off = bswap_32(e64->name_off);
370 e64->val_lo32 = bswap_32(e64->val_lo32);
371 e64->val_hi32 = bswap_32(e64->val_hi32);
376 a->type = bswap_32(a->type);
377 a->index_type = bswap_32(a->index_type);
378 a->nelems = bswap_32(a->nelems);
380 case BTF_KIND_STRUCT:
382 for (i = 0, m = btf_members(t); i < vlen; i++, m++) {
383 m->name_off = bswap_32(m->name_off);
384 m->type = bswap_32(m->type);
385 m->offset = bswap_32(m->offset);
388 case BTF_KIND_FUNC_PROTO:
389 for (i = 0, p = btf_params(t); i < vlen; i++, p++) {
390 p->name_off = bswap_32(p->name_off);
391 p->type = bswap_32(p->type);
395 btf_var(t)->linkage = bswap_32(btf_var(t)->linkage);
397 case BTF_KIND_DATASEC:
398 for (i = 0, v = btf_var_secinfos(t); i < vlen; i++, v++) {
399 v->type = bswap_32(v->type);
400 v->offset = bswap_32(v->offset);
401 v->size = bswap_32(v->size);
404 case BTF_KIND_DECL_TAG:
405 btf_decl_tag(t)->component_idx = bswap_32(btf_decl_tag(t)->component_idx);
408 pr_debug("Unsupported BTF_KIND:%u\n", btf_kind(t));
413 static int btf_parse_type_sec(struct btf *btf)
415 struct btf_header *hdr = btf->hdr;
416 void *next_type = btf->types_data;
417 void *end_type = next_type + hdr->type_len;
420 while (next_type + sizeof(struct btf_type) <= end_type) {
421 if (btf->swapped_endian)
422 btf_bswap_type_base(next_type);
424 type_size = btf_type_size(next_type);
427 if (next_type + type_size > end_type) {
428 pr_warn("BTF type [%d] is malformed\n", btf->start_id + btf->nr_types);
432 if (btf->swapped_endian && btf_bswap_type_rest(next_type))
435 err = btf_add_type_idx_entry(btf, next_type - btf->types_data);
439 next_type += type_size;
443 if (next_type != end_type) {
444 pr_warn("BTF types data is malformed\n");
451 __u32 btf__type_cnt(const struct btf *btf)
453 return btf->start_id + btf->nr_types;
456 const struct btf *btf__base_btf(const struct btf *btf)
458 return btf->base_btf;
461 /* internal helper returning non-const pointer to a type */
462 struct btf_type *btf_type_by_id(const struct btf *btf, __u32 type_id)
466 if (type_id < btf->start_id)
467 return btf_type_by_id(btf->base_btf, type_id);
468 return btf->types_data + btf->type_offs[type_id - btf->start_id];
471 const struct btf_type *btf__type_by_id(const struct btf *btf, __u32 type_id)
473 if (type_id >= btf->start_id + btf->nr_types)
474 return errno = EINVAL, NULL;
475 return btf_type_by_id((struct btf *)btf, type_id);
478 static int determine_ptr_size(const struct btf *btf)
480 static const char * const long_aliases[] = {
493 const struct btf_type *t;
497 if (btf->base_btf && btf->base_btf->ptr_sz > 0)
498 return btf->base_btf->ptr_sz;
500 n = btf__type_cnt(btf);
501 for (i = 1; i < n; i++) {
502 t = btf__type_by_id(btf, i);
506 if (t->size != 4 && t->size != 8)
509 name = btf__name_by_offset(btf, t->name_off);
513 for (j = 0; j < ARRAY_SIZE(long_aliases); j++) {
514 if (strcmp(name, long_aliases[j]) == 0)
522 static size_t btf_ptr_sz(const struct btf *btf)
525 ((struct btf *)btf)->ptr_sz = determine_ptr_size(btf);
526 return btf->ptr_sz < 0 ? sizeof(void *) : btf->ptr_sz;
529 /* Return pointer size this BTF instance assumes. The size is heuristically
530 * determined by looking for 'long' or 'unsigned long' integer type and
531 * recording its size in bytes. If BTF type information doesn't have any such
532 * type, this function returns 0. In the latter case, native architecture's
533 * pointer size is assumed, so will be either 4 or 8, depending on
534 * architecture that libbpf was compiled for. It's possible to override
535 * guessed value by using btf__set_pointer_size() API.
537 size_t btf__pointer_size(const struct btf *btf)
540 ((struct btf *)btf)->ptr_sz = determine_ptr_size(btf);
543 /* not enough BTF type info to guess */
549 /* Override or set pointer size in bytes. Only values of 4 and 8 are
552 int btf__set_pointer_size(struct btf *btf, size_t ptr_sz)
554 if (ptr_sz != 4 && ptr_sz != 8)
555 return libbpf_err(-EINVAL);
556 btf->ptr_sz = ptr_sz;
560 static bool is_host_big_endian(void)
562 #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
564 #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
567 # error "Unrecognized __BYTE_ORDER__"
571 enum btf_endianness btf__endianness(const struct btf *btf)
573 if (is_host_big_endian())
574 return btf->swapped_endian ? BTF_LITTLE_ENDIAN : BTF_BIG_ENDIAN;
576 return btf->swapped_endian ? BTF_BIG_ENDIAN : BTF_LITTLE_ENDIAN;
579 int btf__set_endianness(struct btf *btf, enum btf_endianness endian)
581 if (endian != BTF_LITTLE_ENDIAN && endian != BTF_BIG_ENDIAN)
582 return libbpf_err(-EINVAL);
584 btf->swapped_endian = is_host_big_endian() != (endian == BTF_BIG_ENDIAN);
585 if (!btf->swapped_endian) {
586 free(btf->raw_data_swapped);
587 btf->raw_data_swapped = NULL;
592 static bool btf_type_is_void(const struct btf_type *t)
594 return t == &btf_void || btf_is_fwd(t);
597 static bool btf_type_is_void_or_null(const struct btf_type *t)
599 return !t || btf_type_is_void(t);
602 #define MAX_RESOLVE_DEPTH 32
604 __s64 btf__resolve_size(const struct btf *btf, __u32 type_id)
606 const struct btf_array *array;
607 const struct btf_type *t;
612 t = btf__type_by_id(btf, type_id);
613 for (i = 0; i < MAX_RESOLVE_DEPTH && !btf_type_is_void_or_null(t); i++) {
614 switch (btf_kind(t)) {
616 case BTF_KIND_STRUCT:
619 case BTF_KIND_ENUM64:
620 case BTF_KIND_DATASEC:
625 size = btf_ptr_sz(btf);
627 case BTF_KIND_TYPEDEF:
628 case BTF_KIND_VOLATILE:
630 case BTF_KIND_RESTRICT:
632 case BTF_KIND_DECL_TAG:
633 case BTF_KIND_TYPE_TAG:
637 array = btf_array(t);
638 if (nelems && array->nelems > UINT32_MAX / nelems)
639 return libbpf_err(-E2BIG);
640 nelems *= array->nelems;
641 type_id = array->type;
644 return libbpf_err(-EINVAL);
647 t = btf__type_by_id(btf, type_id);
652 return libbpf_err(-EINVAL);
653 if (nelems && size > UINT32_MAX / nelems)
654 return libbpf_err(-E2BIG);
656 return nelems * size;
659 int btf__align_of(const struct btf *btf, __u32 id)
661 const struct btf_type *t = btf__type_by_id(btf, id);
662 __u16 kind = btf_kind(t);
667 case BTF_KIND_ENUM64:
669 return min(btf_ptr_sz(btf), (size_t)t->size);
671 return btf_ptr_sz(btf);
672 case BTF_KIND_TYPEDEF:
673 case BTF_KIND_VOLATILE:
675 case BTF_KIND_RESTRICT:
676 case BTF_KIND_TYPE_TAG:
677 return btf__align_of(btf, t->type);
679 return btf__align_of(btf, btf_array(t)->type);
680 case BTF_KIND_STRUCT:
681 case BTF_KIND_UNION: {
682 const struct btf_member *m = btf_members(t);
683 __u16 vlen = btf_vlen(t);
684 int i, max_align = 1, align;
686 for (i = 0; i < vlen; i++, m++) {
687 align = btf__align_of(btf, m->type);
689 return libbpf_err(align);
690 max_align = max(max_align, align);
692 /* if field offset isn't aligned according to field
693 * type's alignment, then struct must be packed
695 if (btf_member_bitfield_size(t, i) == 0 &&
696 (m->offset % (8 * align)) != 0)
700 /* if struct/union size isn't a multiple of its alignment,
701 * then struct must be packed
703 if ((t->size % max_align) != 0)
709 pr_warn("unsupported BTF_KIND:%u\n", btf_kind(t));
710 return errno = EINVAL, 0;
714 int btf__resolve_type(const struct btf *btf, __u32 type_id)
716 const struct btf_type *t;
719 t = btf__type_by_id(btf, type_id);
720 while (depth < MAX_RESOLVE_DEPTH &&
721 !btf_type_is_void_or_null(t) &&
722 (btf_is_mod(t) || btf_is_typedef(t) || btf_is_var(t))) {
724 t = btf__type_by_id(btf, type_id);
728 if (depth == MAX_RESOLVE_DEPTH || btf_type_is_void_or_null(t))
729 return libbpf_err(-EINVAL);
734 __s32 btf__find_by_name(const struct btf *btf, const char *type_name)
736 __u32 i, nr_types = btf__type_cnt(btf);
738 if (!strcmp(type_name, "void"))
741 for (i = 1; i < nr_types; i++) {
742 const struct btf_type *t = btf__type_by_id(btf, i);
743 const char *name = btf__name_by_offset(btf, t->name_off);
745 if (name && !strcmp(type_name, name))
749 return libbpf_err(-ENOENT);
752 static __s32 btf_find_by_name_kind(const struct btf *btf, int start_id,
753 const char *type_name, __u32 kind)
755 __u32 i, nr_types = btf__type_cnt(btf);
757 if (kind == BTF_KIND_UNKN || !strcmp(type_name, "void"))
760 for (i = start_id; i < nr_types; i++) {
761 const struct btf_type *t = btf__type_by_id(btf, i);
764 if (btf_kind(t) != kind)
766 name = btf__name_by_offset(btf, t->name_off);
767 if (name && !strcmp(type_name, name))
771 return libbpf_err(-ENOENT);
774 __s32 btf__find_by_name_kind_own(const struct btf *btf, const char *type_name,
777 return btf_find_by_name_kind(btf, btf->start_id, type_name, kind);
780 __s32 btf__find_by_name_kind(const struct btf *btf, const char *type_name,
783 return btf_find_by_name_kind(btf, 1, type_name, kind);
786 static bool btf_is_modifiable(const struct btf *btf)
788 return (void *)btf->hdr != btf->raw_data;
791 void btf__free(struct btf *btf)
793 if (IS_ERR_OR_NULL(btf))
799 if (btf_is_modifiable(btf)) {
800 /* if BTF was modified after loading, it will have a split
801 * in-memory representation for header, types, and strings
802 * sections, so we need to free all of them individually. It
803 * might still have a cached contiguous raw data present,
804 * which will be unconditionally freed below.
807 free(btf->types_data);
808 strset__free(btf->strs_set);
811 free(btf->raw_data_swapped);
812 free(btf->type_offs);
816 static struct btf *btf_new_empty(struct btf *base_btf)
820 btf = calloc(1, sizeof(*btf));
822 return ERR_PTR(-ENOMEM);
826 btf->start_str_off = 0;
828 btf->ptr_sz = sizeof(void *);
829 btf->swapped_endian = false;
832 btf->base_btf = base_btf;
833 btf->start_id = btf__type_cnt(base_btf);
834 btf->start_str_off = base_btf->hdr->str_len;
837 /* +1 for empty string at offset 0 */
838 btf->raw_size = sizeof(struct btf_header) + (base_btf ? 0 : 1);
839 btf->raw_data = calloc(1, btf->raw_size);
840 if (!btf->raw_data) {
842 return ERR_PTR(-ENOMEM);
845 btf->hdr = btf->raw_data;
846 btf->hdr->hdr_len = sizeof(struct btf_header);
847 btf->hdr->magic = BTF_MAGIC;
848 btf->hdr->version = BTF_VERSION;
850 btf->types_data = btf->raw_data + btf->hdr->hdr_len;
851 btf->strs_data = btf->raw_data + btf->hdr->hdr_len;
852 btf->hdr->str_len = base_btf ? 0 : 1; /* empty string at offset 0 */
857 struct btf *btf__new_empty(void)
859 return libbpf_ptr(btf_new_empty(NULL));
862 struct btf *btf__new_empty_split(struct btf *base_btf)
864 return libbpf_ptr(btf_new_empty(base_btf));
867 static struct btf *btf_new(const void *data, __u32 size, struct btf *base_btf)
872 btf = calloc(1, sizeof(struct btf));
874 return ERR_PTR(-ENOMEM);
878 btf->start_str_off = 0;
882 btf->base_btf = base_btf;
883 btf->start_id = btf__type_cnt(base_btf);
884 btf->start_str_off = base_btf->hdr->str_len;
887 btf->raw_data = malloc(size);
888 if (!btf->raw_data) {
892 memcpy(btf->raw_data, data, size);
893 btf->raw_size = size;
895 btf->hdr = btf->raw_data;
896 err = btf_parse_hdr(btf);
900 btf->strs_data = btf->raw_data + btf->hdr->hdr_len + btf->hdr->str_off;
901 btf->types_data = btf->raw_data + btf->hdr->hdr_len + btf->hdr->type_off;
903 err = btf_parse_str_sec(btf);
904 err = err ?: btf_parse_type_sec(btf);
917 struct btf *btf__new(const void *data, __u32 size)
919 return libbpf_ptr(btf_new(data, size, NULL));
922 static struct btf *btf_parse_elf(const char *path, struct btf *base_btf,
923 struct btf_ext **btf_ext)
925 Elf_Data *btf_data = NULL, *btf_ext_data = NULL;
926 int err = 0, fd = -1, idx = 0;
927 struct btf *btf = NULL;
933 if (elf_version(EV_CURRENT) == EV_NONE) {
934 pr_warn("failed to init libelf for %s\n", path);
935 return ERR_PTR(-LIBBPF_ERRNO__LIBELF);
938 fd = open(path, O_RDONLY | O_CLOEXEC);
941 pr_warn("failed to open %s: %s\n", path, strerror(errno));
945 err = -LIBBPF_ERRNO__FORMAT;
947 elf = elf_begin(fd, ELF_C_READ, NULL);
949 pr_warn("failed to open %s as ELF file\n", path);
952 if (!gelf_getehdr(elf, &ehdr)) {
953 pr_warn("failed to get EHDR from %s\n", path);
957 if (elf_getshdrstrndx(elf, &shstrndx)) {
958 pr_warn("failed to get section names section index for %s\n",
963 if (!elf_rawdata(elf_getscn(elf, shstrndx), NULL)) {
964 pr_warn("failed to get e_shstrndx from %s\n", path);
968 while ((scn = elf_nextscn(elf, scn)) != NULL) {
973 if (gelf_getshdr(scn, &sh) != &sh) {
974 pr_warn("failed to get section(%d) header from %s\n",
978 name = elf_strptr(elf, shstrndx, sh.sh_name);
980 pr_warn("failed to get section(%d) name from %s\n",
984 if (strcmp(name, BTF_ELF_SEC) == 0) {
985 btf_data = elf_getdata(scn, 0);
987 pr_warn("failed to get section(%d, %s) data from %s\n",
992 } else if (btf_ext && strcmp(name, BTF_EXT_ELF_SEC) == 0) {
993 btf_ext_data = elf_getdata(scn, 0);
995 pr_warn("failed to get section(%d, %s) data from %s\n",
1004 pr_warn("failed to find '%s' ELF section in %s\n", BTF_ELF_SEC, path);
1008 btf = btf_new(btf_data->d_buf, btf_data->d_size, base_btf);
1009 err = libbpf_get_error(btf);
1013 switch (gelf_getclass(elf)) {
1015 btf__set_pointer_size(btf, 4);
1018 btf__set_pointer_size(btf, 8);
1021 pr_warn("failed to get ELF class (bitness) for %s\n", path);
1025 if (btf_ext && btf_ext_data) {
1026 *btf_ext = btf_ext__new(btf_ext_data->d_buf, btf_ext_data->d_size);
1027 err = libbpf_get_error(*btf_ext);
1030 } else if (btf_ext) {
1042 btf_ext__free(*btf_ext);
1045 return ERR_PTR(err);
1048 struct btf *btf__parse_elf(const char *path, struct btf_ext **btf_ext)
1050 return libbpf_ptr(btf_parse_elf(path, NULL, btf_ext));
1053 struct btf *btf__parse_elf_split(const char *path, struct btf *base_btf)
1055 return libbpf_ptr(btf_parse_elf(path, base_btf, NULL));
1058 static struct btf *btf_parse_raw(const char *path, struct btf *base_btf)
1060 struct btf *btf = NULL;
1067 f = fopen(path, "rbe");
1073 /* check BTF magic */
1074 if (fread(&magic, 1, sizeof(magic), f) < sizeof(magic)) {
1078 if (magic != BTF_MAGIC && magic != bswap_16(BTF_MAGIC)) {
1079 /* definitely not a raw BTF */
1085 if (fseek(f, 0, SEEK_END)) {
1094 /* rewind to the start */
1095 if (fseek(f, 0, SEEK_SET)) {
1100 /* pre-alloc memory and read all of BTF data */
1106 if (fread(data, 1, sz, f) < sz) {
1111 /* finally parse BTF data */
1112 btf = btf_new(data, sz, base_btf);
1118 return err ? ERR_PTR(err) : btf;
1121 struct btf *btf__parse_raw(const char *path)
1123 return libbpf_ptr(btf_parse_raw(path, NULL));
1126 struct btf *btf__parse_raw_split(const char *path, struct btf *base_btf)
1128 return libbpf_ptr(btf_parse_raw(path, base_btf));
1131 static struct btf *btf_parse(const char *path, struct btf *base_btf, struct btf_ext **btf_ext)
1139 btf = btf_parse_raw(path, base_btf);
1140 err = libbpf_get_error(btf);
1144 return ERR_PTR(err);
1145 return btf_parse_elf(path, base_btf, btf_ext);
1148 struct btf *btf__parse(const char *path, struct btf_ext **btf_ext)
1150 return libbpf_ptr(btf_parse(path, NULL, btf_ext));
1153 struct btf *btf__parse_split(const char *path, struct btf *base_btf)
1155 return libbpf_ptr(btf_parse(path, base_btf, NULL));
1158 static void *btf_get_raw_data(const struct btf *btf, __u32 *size, bool swap_endian);
1160 int btf_load_into_kernel(struct btf *btf, char *log_buf, size_t log_sz, __u32 log_level)
1162 LIBBPF_OPTS(bpf_btf_load_opts, opts);
1163 __u32 buf_sz = 0, raw_size;
1164 char *buf = NULL, *tmp;
1169 return libbpf_err(-EEXIST);
1170 if (log_sz && !log_buf)
1171 return libbpf_err(-EINVAL);
1173 /* cache native raw data representation */
1174 raw_data = btf_get_raw_data(btf, &raw_size, false);
1179 btf->raw_size = raw_size;
1180 btf->raw_data = raw_data;
1183 /* if log_level is 0, we won't provide log_buf/log_size to the kernel,
1184 * initially. Only if BTF loading fails, we bump log_level to 1 and
1185 * retry, using either auto-allocated or custom log_buf. This way
1186 * non-NULL custom log_buf provides a buffer just in case, but hopes
1187 * for successful load and no need for log_buf.
1190 /* if caller didn't provide custom log_buf, we'll keep
1191 * allocating our own progressively bigger buffers for BTF
1195 buf_sz = max((__u32)BPF_LOG_BUF_SIZE, buf_sz * 2);
1196 tmp = realloc(buf, buf_sz);
1205 opts.log_buf = log_buf ? log_buf : buf;
1206 opts.log_size = log_buf ? log_sz : buf_sz;
1207 opts.log_level = log_level;
1210 btf->fd = bpf_btf_load(raw_data, raw_size, &opts);
1212 /* time to turn on verbose mode and try again */
1213 if (log_level == 0) {
1217 /* only retry if caller didn't provide custom log_buf, but
1218 * make sure we can never overflow buf_sz
1220 if (!log_buf && errno == ENOSPC && buf_sz <= UINT_MAX / 2)
1224 pr_warn("BTF loading error: %d\n", err);
1225 /* don't print out contents of custom log_buf */
1226 if (!log_buf && buf[0])
1227 pr_warn("-- BEGIN BTF LOAD LOG ---\n%s\n-- END BTF LOAD LOG --\n", buf);
1232 return libbpf_err(err);
1235 int btf__load_into_kernel(struct btf *btf)
1237 return btf_load_into_kernel(btf, NULL, 0, 0);
1240 int btf__fd(const struct btf *btf)
1245 void btf__set_fd(struct btf *btf, int fd)
1250 static const void *btf_strs_data(const struct btf *btf)
1252 return btf->strs_data ? btf->strs_data : strset__data(btf->strs_set);
1255 static void *btf_get_raw_data(const struct btf *btf, __u32 *size, bool swap_endian)
1257 struct btf_header *hdr = btf->hdr;
1263 data = swap_endian ? btf->raw_data_swapped : btf->raw_data;
1265 *size = btf->raw_size;
1269 data_sz = hdr->hdr_len + hdr->type_len + hdr->str_len;
1270 data = calloc(1, data_sz);
1275 memcpy(p, hdr, hdr->hdr_len);
1280 memcpy(p, btf->types_data, hdr->type_len);
1282 for (i = 0; i < btf->nr_types; i++) {
1283 t = p + btf->type_offs[i];
1284 /* btf_bswap_type_rest() relies on native t->info, so
1285 * we swap base type info after we swapped all the
1286 * additional information
1288 if (btf_bswap_type_rest(t))
1290 btf_bswap_type_base(t);
1295 memcpy(p, btf_strs_data(btf), hdr->str_len);
1305 const void *btf__raw_data(const struct btf *btf_ro, __u32 *size)
1307 struct btf *btf = (struct btf *)btf_ro;
1311 data = btf_get_raw_data(btf, &data_sz, btf->swapped_endian);
1313 return errno = ENOMEM, NULL;
1315 btf->raw_size = data_sz;
1316 if (btf->swapped_endian)
1317 btf->raw_data_swapped = data;
1319 btf->raw_data = data;
1324 __attribute__((alias("btf__raw_data")))
1325 const void *btf__get_raw_data(const struct btf *btf, __u32 *size);
1327 const char *btf__str_by_offset(const struct btf *btf, __u32 offset)
1329 if (offset < btf->start_str_off)
1330 return btf__str_by_offset(btf->base_btf, offset);
1331 else if (offset - btf->start_str_off < btf->hdr->str_len)
1332 return btf_strs_data(btf) + (offset - btf->start_str_off);
1334 return errno = EINVAL, NULL;
1337 const char *btf__name_by_offset(const struct btf *btf, __u32 offset)
1339 return btf__str_by_offset(btf, offset);
1342 struct btf *btf_get_from_fd(int btf_fd, struct btf *base_btf)
1344 struct bpf_btf_info btf_info;
1345 __u32 len = sizeof(btf_info);
1351 /* we won't know btf_size until we call bpf_btf_get_info_by_fd(). so
1352 * let's start with a sane default - 4KiB here - and resize it only if
1353 * bpf_btf_get_info_by_fd() needs a bigger buffer.
1356 ptr = malloc(last_size);
1358 return ERR_PTR(-ENOMEM);
1360 memset(&btf_info, 0, sizeof(btf_info));
1361 btf_info.btf = ptr_to_u64(ptr);
1362 btf_info.btf_size = last_size;
1363 err = bpf_btf_get_info_by_fd(btf_fd, &btf_info, &len);
1365 if (!err && btf_info.btf_size > last_size) {
1368 last_size = btf_info.btf_size;
1369 temp_ptr = realloc(ptr, last_size);
1371 btf = ERR_PTR(-ENOMEM);
1376 len = sizeof(btf_info);
1377 memset(&btf_info, 0, sizeof(btf_info));
1378 btf_info.btf = ptr_to_u64(ptr);
1379 btf_info.btf_size = last_size;
1381 err = bpf_btf_get_info_by_fd(btf_fd, &btf_info, &len);
1384 if (err || btf_info.btf_size > last_size) {
1385 btf = err ? ERR_PTR(-errno) : ERR_PTR(-E2BIG);
1389 btf = btf_new(ptr, btf_info.btf_size, base_btf);
1396 struct btf *btf__load_from_kernel_by_id_split(__u32 id, struct btf *base_btf)
1401 btf_fd = bpf_btf_get_fd_by_id(id);
1403 return libbpf_err_ptr(-errno);
1405 btf = btf_get_from_fd(btf_fd, base_btf);
1408 return libbpf_ptr(btf);
1411 struct btf *btf__load_from_kernel_by_id(__u32 id)
1413 return btf__load_from_kernel_by_id_split(id, NULL);
1416 static void btf_invalidate_raw_data(struct btf *btf)
1418 if (btf->raw_data) {
1419 free(btf->raw_data);
1420 btf->raw_data = NULL;
1422 if (btf->raw_data_swapped) {
1423 free(btf->raw_data_swapped);
1424 btf->raw_data_swapped = NULL;
1428 /* Ensure BTF is ready to be modified (by splitting into a three memory
1429 * regions for header, types, and strings). Also invalidate cached
1432 static int btf_ensure_modifiable(struct btf *btf)
1435 struct strset *set = NULL;
1438 if (btf_is_modifiable(btf)) {
1439 /* any BTF modification invalidates raw_data */
1440 btf_invalidate_raw_data(btf);
1444 /* split raw data into three memory regions */
1445 hdr = malloc(btf->hdr->hdr_len);
1446 types = malloc(btf->hdr->type_len);
1450 memcpy(hdr, btf->hdr, btf->hdr->hdr_len);
1451 memcpy(types, btf->types_data, btf->hdr->type_len);
1453 /* build lookup index for all strings */
1454 set = strset__new(BTF_MAX_STR_OFFSET, btf->strs_data, btf->hdr->str_len);
1460 /* only when everything was successful, update internal state */
1462 btf->types_data = types;
1463 btf->types_data_cap = btf->hdr->type_len;
1464 btf->strs_data = NULL;
1465 btf->strs_set = set;
1466 /* if BTF was created from scratch, all strings are guaranteed to be
1467 * unique and deduplicated
1469 if (btf->hdr->str_len == 0)
1470 btf->strs_deduped = true;
1471 if (!btf->base_btf && btf->hdr->str_len == 1)
1472 btf->strs_deduped = true;
1474 /* invalidate raw_data representation */
1475 btf_invalidate_raw_data(btf);
1486 /* Find an offset in BTF string section that corresponds to a given string *s*.
1488 * - >0 offset into string section, if string is found;
1489 * - -ENOENT, if string is not in the string section;
1490 * - <0, on any other error.
1492 int btf__find_str(struct btf *btf, const char *s)
1496 if (btf->base_btf) {
1497 off = btf__find_str(btf->base_btf, s);
1502 /* BTF needs to be in a modifiable state to build string lookup index */
1503 if (btf_ensure_modifiable(btf))
1504 return libbpf_err(-ENOMEM);
1506 off = strset__find_str(btf->strs_set, s);
1508 return libbpf_err(off);
1510 return btf->start_str_off + off;
1513 /* Add a string s to the BTF string section.
1515 * - > 0 offset into string section, on success;
1518 int btf__add_str(struct btf *btf, const char *s)
1522 if (btf->base_btf) {
1523 off = btf__find_str(btf->base_btf, s);
1528 if (btf_ensure_modifiable(btf))
1529 return libbpf_err(-ENOMEM);
1531 off = strset__add_str(btf->strs_set, s);
1533 return libbpf_err(off);
1535 btf->hdr->str_len = strset__data_size(btf->strs_set);
1537 return btf->start_str_off + off;
1540 static void *btf_add_type_mem(struct btf *btf, size_t add_sz)
1542 return libbpf_add_mem(&btf->types_data, &btf->types_data_cap, 1,
1543 btf->hdr->type_len, UINT_MAX, add_sz);
1546 static void btf_type_inc_vlen(struct btf_type *t)
1548 t->info = btf_type_info(btf_kind(t), btf_vlen(t) + 1, btf_kflag(t));
1551 static int btf_commit_type(struct btf *btf, int data_sz)
1555 err = btf_add_type_idx_entry(btf, btf->hdr->type_len);
1557 return libbpf_err(err);
1559 btf->hdr->type_len += data_sz;
1560 btf->hdr->str_off += data_sz;
1562 return btf->start_id + btf->nr_types - 1;
1566 const struct btf *src;
1568 struct hashmap *str_off_map; /* map string offsets from src to dst */
1571 static int btf_rewrite_str(__u32 *str_off, void *ctx)
1573 struct btf_pipe *p = ctx;
1577 if (!*str_off) /* nothing to do for empty strings */
1580 if (p->str_off_map &&
1581 hashmap__find(p->str_off_map, *str_off, &mapped_off)) {
1582 *str_off = mapped_off;
1586 off = btf__add_str(p->dst, btf__str_by_offset(p->src, *str_off));
1590 /* Remember string mapping from src to dst. It avoids
1591 * performing expensive string comparisons.
1593 if (p->str_off_map) {
1594 err = hashmap__append(p->str_off_map, *str_off, off);
1603 int btf__add_type(struct btf *btf, const struct btf *src_btf, const struct btf_type *src_type)
1605 struct btf_pipe p = { .src = src_btf, .dst = btf };
1609 sz = btf_type_size(src_type);
1611 return libbpf_err(sz);
1613 /* deconstruct BTF, if necessary, and invalidate raw_data */
1614 if (btf_ensure_modifiable(btf))
1615 return libbpf_err(-ENOMEM);
1617 t = btf_add_type_mem(btf, sz);
1619 return libbpf_err(-ENOMEM);
1621 memcpy(t, src_type, sz);
1623 err = btf_type_visit_str_offs(t, btf_rewrite_str, &p);
1625 return libbpf_err(err);
1627 return btf_commit_type(btf, sz);
1630 static int btf_rewrite_type_ids(__u32 *type_id, void *ctx)
1632 struct btf *btf = ctx;
1634 if (!*type_id) /* nothing to do for VOID references */
1637 /* we haven't updated btf's type count yet, so
1638 * btf->start_id + btf->nr_types - 1 is the type ID offset we should
1639 * add to all newly added BTF types
1641 *type_id += btf->start_id + btf->nr_types - 1;
1645 static size_t btf_dedup_identity_hash_fn(long key, void *ctx);
1646 static bool btf_dedup_equal_fn(long k1, long k2, void *ctx);
1648 int btf__add_btf(struct btf *btf, const struct btf *src_btf)
1650 struct btf_pipe p = { .src = src_btf, .dst = btf };
1651 int data_sz, sz, cnt, i, err, old_strs_len;
1655 /* appending split BTF isn't supported yet */
1656 if (src_btf->base_btf)
1657 return libbpf_err(-ENOTSUP);
1659 /* deconstruct BTF, if necessary, and invalidate raw_data */
1660 if (btf_ensure_modifiable(btf))
1661 return libbpf_err(-ENOMEM);
1663 /* remember original strings section size if we have to roll back
1664 * partial strings section changes
1666 old_strs_len = btf->hdr->str_len;
1668 data_sz = src_btf->hdr->type_len;
1669 cnt = btf__type_cnt(src_btf) - 1;
1671 /* pre-allocate enough memory for new types */
1672 t = btf_add_type_mem(btf, data_sz);
1674 return libbpf_err(-ENOMEM);
1676 /* pre-allocate enough memory for type offset index for new types */
1677 off = btf_add_type_offs_mem(btf, cnt);
1679 return libbpf_err(-ENOMEM);
1681 /* Map the string offsets from src_btf to the offsets from btf to improve performance */
1682 p.str_off_map = hashmap__new(btf_dedup_identity_hash_fn, btf_dedup_equal_fn, NULL);
1683 if (IS_ERR(p.str_off_map))
1684 return libbpf_err(-ENOMEM);
1686 /* bulk copy types data for all types from src_btf */
1687 memcpy(t, src_btf->types_data, data_sz);
1689 for (i = 0; i < cnt; i++) {
1690 sz = btf_type_size(t);
1692 /* unlikely, has to be corrupted src_btf */
1697 /* fill out type ID to type offset mapping for lookups by type ID */
1698 *off = t - btf->types_data;
1700 /* add, dedup, and remap strings referenced by this BTF type */
1701 err = btf_type_visit_str_offs(t, btf_rewrite_str, &p);
1705 /* remap all type IDs referenced from this BTF type */
1706 err = btf_type_visit_type_ids(t, btf_rewrite_type_ids, btf);
1710 /* go to next type data and type offset index entry */
1715 /* Up until now any of the copied type data was effectively invisible,
1716 * so if we exited early before this point due to error, BTF would be
1717 * effectively unmodified. There would be extra internal memory
1718 * pre-allocated, but it would not be available for querying. But now
1719 * that we've copied and rewritten all the data successfully, we can
1720 * update type count and various internal offsets and sizes to
1721 * "commit" the changes and made them visible to the outside world.
1723 btf->hdr->type_len += data_sz;
1724 btf->hdr->str_off += data_sz;
1725 btf->nr_types += cnt;
1727 hashmap__free(p.str_off_map);
1729 /* return type ID of the first added BTF type */
1730 return btf->start_id + btf->nr_types - cnt;
1732 /* zero out preallocated memory as if it was just allocated with
1735 memset(btf->types_data + btf->hdr->type_len, 0, data_sz);
1736 memset(btf->strs_data + old_strs_len, 0, btf->hdr->str_len - old_strs_len);
1738 /* and now restore original strings section size; types data size
1739 * wasn't modified, so doesn't need restoring, see big comment above
1741 btf->hdr->str_len = old_strs_len;
1743 hashmap__free(p.str_off_map);
1745 return libbpf_err(err);
1749 * Append new BTF_KIND_INT type with:
1750 * - *name* - non-empty, non-NULL type name;
1751 * - *sz* - power-of-2 (1, 2, 4, ..) size of the type, in bytes;
1752 * - encoding is a combination of BTF_INT_SIGNED, BTF_INT_CHAR, BTF_INT_BOOL.
1754 * - >0, type ID of newly added BTF type;
1757 int btf__add_int(struct btf *btf, const char *name, size_t byte_sz, int encoding)
1762 /* non-empty name */
1763 if (!name || !name[0])
1764 return libbpf_err(-EINVAL);
1765 /* byte_sz must be power of 2 */
1766 if (!byte_sz || (byte_sz & (byte_sz - 1)) || byte_sz > 16)
1767 return libbpf_err(-EINVAL);
1768 if (encoding & ~(BTF_INT_SIGNED | BTF_INT_CHAR | BTF_INT_BOOL))
1769 return libbpf_err(-EINVAL);
1771 /* deconstruct BTF, if necessary, and invalidate raw_data */
1772 if (btf_ensure_modifiable(btf))
1773 return libbpf_err(-ENOMEM);
1775 sz = sizeof(struct btf_type) + sizeof(int);
1776 t = btf_add_type_mem(btf, sz);
1778 return libbpf_err(-ENOMEM);
1780 /* if something goes wrong later, we might end up with an extra string,
1781 * but that shouldn't be a problem, because BTF can't be constructed
1782 * completely anyway and will most probably be just discarded
1784 name_off = btf__add_str(btf, name);
1788 t->name_off = name_off;
1789 t->info = btf_type_info(BTF_KIND_INT, 0, 0);
1791 /* set INT info, we don't allow setting legacy bit offset/size */
1792 *(__u32 *)(t + 1) = (encoding << 24) | (byte_sz * 8);
1794 return btf_commit_type(btf, sz);
1798 * Append new BTF_KIND_FLOAT type with:
1799 * - *name* - non-empty, non-NULL type name;
1800 * - *sz* - size of the type, in bytes;
1802 * - >0, type ID of newly added BTF type;
1805 int btf__add_float(struct btf *btf, const char *name, size_t byte_sz)
1810 /* non-empty name */
1811 if (!name || !name[0])
1812 return libbpf_err(-EINVAL);
1814 /* byte_sz must be one of the explicitly allowed values */
1815 if (byte_sz != 2 && byte_sz != 4 && byte_sz != 8 && byte_sz != 12 &&
1817 return libbpf_err(-EINVAL);
1819 if (btf_ensure_modifiable(btf))
1820 return libbpf_err(-ENOMEM);
1822 sz = sizeof(struct btf_type);
1823 t = btf_add_type_mem(btf, sz);
1825 return libbpf_err(-ENOMEM);
1827 name_off = btf__add_str(btf, name);
1831 t->name_off = name_off;
1832 t->info = btf_type_info(BTF_KIND_FLOAT, 0, 0);
1835 return btf_commit_type(btf, sz);
1838 /* it's completely legal to append BTF types with type IDs pointing forward to
1839 * types that haven't been appended yet, so we only make sure that id looks
1840 * sane, we can't guarantee that ID will always be valid
1842 static int validate_type_id(int id)
1844 if (id < 0 || id > BTF_MAX_NR_TYPES)
1849 /* generic append function for PTR, TYPEDEF, CONST/VOLATILE/RESTRICT */
1850 static int btf_add_ref_kind(struct btf *btf, int kind, const char *name, int ref_type_id)
1853 int sz, name_off = 0;
1855 if (validate_type_id(ref_type_id))
1856 return libbpf_err(-EINVAL);
1858 if (btf_ensure_modifiable(btf))
1859 return libbpf_err(-ENOMEM);
1861 sz = sizeof(struct btf_type);
1862 t = btf_add_type_mem(btf, sz);
1864 return libbpf_err(-ENOMEM);
1866 if (name && name[0]) {
1867 name_off = btf__add_str(btf, name);
1872 t->name_off = name_off;
1873 t->info = btf_type_info(kind, 0, 0);
1874 t->type = ref_type_id;
1876 return btf_commit_type(btf, sz);
1880 * Append new BTF_KIND_PTR type with:
1881 * - *ref_type_id* - referenced type ID, it might not exist yet;
1883 * - >0, type ID of newly added BTF type;
1886 int btf__add_ptr(struct btf *btf, int ref_type_id)
1888 return btf_add_ref_kind(btf, BTF_KIND_PTR, NULL, ref_type_id);
1892 * Append new BTF_KIND_ARRAY type with:
1893 * - *index_type_id* - type ID of the type describing array index;
1894 * - *elem_type_id* - type ID of the type describing array element;
1895 * - *nr_elems* - the size of the array;
1897 * - >0, type ID of newly added BTF type;
1900 int btf__add_array(struct btf *btf, int index_type_id, int elem_type_id, __u32 nr_elems)
1903 struct btf_array *a;
1906 if (validate_type_id(index_type_id) || validate_type_id(elem_type_id))
1907 return libbpf_err(-EINVAL);
1909 if (btf_ensure_modifiable(btf))
1910 return libbpf_err(-ENOMEM);
1912 sz = sizeof(struct btf_type) + sizeof(struct btf_array);
1913 t = btf_add_type_mem(btf, sz);
1915 return libbpf_err(-ENOMEM);
1918 t->info = btf_type_info(BTF_KIND_ARRAY, 0, 0);
1922 a->type = elem_type_id;
1923 a->index_type = index_type_id;
1924 a->nelems = nr_elems;
1926 return btf_commit_type(btf, sz);
1929 /* generic STRUCT/UNION append function */
1930 static int btf_add_composite(struct btf *btf, int kind, const char *name, __u32 bytes_sz)
1933 int sz, name_off = 0;
1935 if (btf_ensure_modifiable(btf))
1936 return libbpf_err(-ENOMEM);
1938 sz = sizeof(struct btf_type);
1939 t = btf_add_type_mem(btf, sz);
1941 return libbpf_err(-ENOMEM);
1943 if (name && name[0]) {
1944 name_off = btf__add_str(btf, name);
1949 /* start out with vlen=0 and no kflag; this will be adjusted when
1950 * adding each member
1952 t->name_off = name_off;
1953 t->info = btf_type_info(kind, 0, 0);
1956 return btf_commit_type(btf, sz);
1960 * Append new BTF_KIND_STRUCT type with:
1961 * - *name* - name of the struct, can be NULL or empty for anonymous structs;
1962 * - *byte_sz* - size of the struct, in bytes;
1964 * Struct initially has no fields in it. Fields can be added by
1965 * btf__add_field() right after btf__add_struct() succeeds.
1968 * - >0, type ID of newly added BTF type;
1971 int btf__add_struct(struct btf *btf, const char *name, __u32 byte_sz)
1973 return btf_add_composite(btf, BTF_KIND_STRUCT, name, byte_sz);
1977 * Append new BTF_KIND_UNION type with:
1978 * - *name* - name of the union, can be NULL or empty for anonymous union;
1979 * - *byte_sz* - size of the union, in bytes;
1981 * Union initially has no fields in it. Fields can be added by
1982 * btf__add_field() right after btf__add_union() succeeds. All fields
1983 * should have *bit_offset* of 0.
1986 * - >0, type ID of newly added BTF type;
1989 int btf__add_union(struct btf *btf, const char *name, __u32 byte_sz)
1991 return btf_add_composite(btf, BTF_KIND_UNION, name, byte_sz);
1994 static struct btf_type *btf_last_type(struct btf *btf)
1996 return btf_type_by_id(btf, btf__type_cnt(btf) - 1);
2000 * Append new field for the current STRUCT/UNION type with:
2001 * - *name* - name of the field, can be NULL or empty for anonymous field;
2002 * - *type_id* - type ID for the type describing field type;
2003 * - *bit_offset* - bit offset of the start of the field within struct/union;
2004 * - *bit_size* - bit size of a bitfield, 0 for non-bitfield fields;
2009 int btf__add_field(struct btf *btf, const char *name, int type_id,
2010 __u32 bit_offset, __u32 bit_size)
2013 struct btf_member *m;
2015 int sz, name_off = 0;
2017 /* last type should be union/struct */
2018 if (btf->nr_types == 0)
2019 return libbpf_err(-EINVAL);
2020 t = btf_last_type(btf);
2021 if (!btf_is_composite(t))
2022 return libbpf_err(-EINVAL);
2024 if (validate_type_id(type_id))
2025 return libbpf_err(-EINVAL);
2026 /* best-effort bit field offset/size enforcement */
2027 is_bitfield = bit_size || (bit_offset % 8 != 0);
2028 if (is_bitfield && (bit_size == 0 || bit_size > 255 || bit_offset > 0xffffff))
2029 return libbpf_err(-EINVAL);
2031 /* only offset 0 is allowed for unions */
2032 if (btf_is_union(t) && bit_offset)
2033 return libbpf_err(-EINVAL);
2035 /* decompose and invalidate raw data */
2036 if (btf_ensure_modifiable(btf))
2037 return libbpf_err(-ENOMEM);
2039 sz = sizeof(struct btf_member);
2040 m = btf_add_type_mem(btf, sz);
2042 return libbpf_err(-ENOMEM);
2044 if (name && name[0]) {
2045 name_off = btf__add_str(btf, name);
2050 m->name_off = name_off;
2052 m->offset = bit_offset | (bit_size << 24);
2054 /* btf_add_type_mem can invalidate t pointer */
2055 t = btf_last_type(btf);
2056 /* update parent type's vlen and kflag */
2057 t->info = btf_type_info(btf_kind(t), btf_vlen(t) + 1, is_bitfield || btf_kflag(t));
2059 btf->hdr->type_len += sz;
2060 btf->hdr->str_off += sz;
2064 static int btf_add_enum_common(struct btf *btf, const char *name, __u32 byte_sz,
2065 bool is_signed, __u8 kind)
2068 int sz, name_off = 0;
2070 /* byte_sz must be power of 2 */
2071 if (!byte_sz || (byte_sz & (byte_sz - 1)) || byte_sz > 8)
2072 return libbpf_err(-EINVAL);
2074 if (btf_ensure_modifiable(btf))
2075 return libbpf_err(-ENOMEM);
2077 sz = sizeof(struct btf_type);
2078 t = btf_add_type_mem(btf, sz);
2080 return libbpf_err(-ENOMEM);
2082 if (name && name[0]) {
2083 name_off = btf__add_str(btf, name);
2088 /* start out with vlen=0; it will be adjusted when adding enum values */
2089 t->name_off = name_off;
2090 t->info = btf_type_info(kind, 0, is_signed);
2093 return btf_commit_type(btf, sz);
2097 * Append new BTF_KIND_ENUM type with:
2098 * - *name* - name of the enum, can be NULL or empty for anonymous enums;
2099 * - *byte_sz* - size of the enum, in bytes.
2101 * Enum initially has no enum values in it (and corresponds to enum forward
2102 * declaration). Enumerator values can be added by btf__add_enum_value()
2103 * immediately after btf__add_enum() succeeds.
2106 * - >0, type ID of newly added BTF type;
2109 int btf__add_enum(struct btf *btf, const char *name, __u32 byte_sz)
2112 * set the signedness to be unsigned, it will change to signed
2113 * if any later enumerator is negative.
2115 return btf_add_enum_common(btf, name, byte_sz, false, BTF_KIND_ENUM);
2119 * Append new enum value for the current ENUM type with:
2120 * - *name* - name of the enumerator value, can't be NULL or empty;
2121 * - *value* - integer value corresponding to enum value *name*;
2126 int btf__add_enum_value(struct btf *btf, const char *name, __s64 value)
2132 /* last type should be BTF_KIND_ENUM */
2133 if (btf->nr_types == 0)
2134 return libbpf_err(-EINVAL);
2135 t = btf_last_type(btf);
2136 if (!btf_is_enum(t))
2137 return libbpf_err(-EINVAL);
2139 /* non-empty name */
2140 if (!name || !name[0])
2141 return libbpf_err(-EINVAL);
2142 if (value < INT_MIN || value > UINT_MAX)
2143 return libbpf_err(-E2BIG);
2145 /* decompose and invalidate raw data */
2146 if (btf_ensure_modifiable(btf))
2147 return libbpf_err(-ENOMEM);
2149 sz = sizeof(struct btf_enum);
2150 v = btf_add_type_mem(btf, sz);
2152 return libbpf_err(-ENOMEM);
2154 name_off = btf__add_str(btf, name);
2158 v->name_off = name_off;
2161 /* update parent type's vlen */
2162 t = btf_last_type(btf);
2163 btf_type_inc_vlen(t);
2165 /* if negative value, set signedness to signed */
2167 t->info = btf_type_info(btf_kind(t), btf_vlen(t), true);
2169 btf->hdr->type_len += sz;
2170 btf->hdr->str_off += sz;
2175 * Append new BTF_KIND_ENUM64 type with:
2176 * - *name* - name of the enum, can be NULL or empty for anonymous enums;
2177 * - *byte_sz* - size of the enum, in bytes.
2178 * - *is_signed* - whether the enum values are signed or not;
2180 * Enum initially has no enum values in it (and corresponds to enum forward
2181 * declaration). Enumerator values can be added by btf__add_enum64_value()
2182 * immediately after btf__add_enum64() succeeds.
2185 * - >0, type ID of newly added BTF type;
2188 int btf__add_enum64(struct btf *btf, const char *name, __u32 byte_sz,
2191 return btf_add_enum_common(btf, name, byte_sz, is_signed,
2196 * Append new enum value for the current ENUM64 type with:
2197 * - *name* - name of the enumerator value, can't be NULL or empty;
2198 * - *value* - integer value corresponding to enum value *name*;
2203 int btf__add_enum64_value(struct btf *btf, const char *name, __u64 value)
2205 struct btf_enum64 *v;
2209 /* last type should be BTF_KIND_ENUM64 */
2210 if (btf->nr_types == 0)
2211 return libbpf_err(-EINVAL);
2212 t = btf_last_type(btf);
2213 if (!btf_is_enum64(t))
2214 return libbpf_err(-EINVAL);
2216 /* non-empty name */
2217 if (!name || !name[0])
2218 return libbpf_err(-EINVAL);
2220 /* decompose and invalidate raw data */
2221 if (btf_ensure_modifiable(btf))
2222 return libbpf_err(-ENOMEM);
2224 sz = sizeof(struct btf_enum64);
2225 v = btf_add_type_mem(btf, sz);
2227 return libbpf_err(-ENOMEM);
2229 name_off = btf__add_str(btf, name);
2233 v->name_off = name_off;
2234 v->val_lo32 = (__u32)value;
2235 v->val_hi32 = value >> 32;
2237 /* update parent type's vlen */
2238 t = btf_last_type(btf);
2239 btf_type_inc_vlen(t);
2241 btf->hdr->type_len += sz;
2242 btf->hdr->str_off += sz;
2247 * Append new BTF_KIND_FWD type with:
2248 * - *name*, non-empty/non-NULL name;
2249 * - *fwd_kind*, kind of forward declaration, one of BTF_FWD_STRUCT,
2250 * BTF_FWD_UNION, or BTF_FWD_ENUM;
2252 * - >0, type ID of newly added BTF type;
2255 int btf__add_fwd(struct btf *btf, const char *name, enum btf_fwd_kind fwd_kind)
2257 if (!name || !name[0])
2258 return libbpf_err(-EINVAL);
2261 case BTF_FWD_STRUCT:
2262 case BTF_FWD_UNION: {
2266 id = btf_add_ref_kind(btf, BTF_KIND_FWD, name, 0);
2269 t = btf_type_by_id(btf, id);
2270 t->info = btf_type_info(BTF_KIND_FWD, 0, fwd_kind == BTF_FWD_UNION);
2274 /* enum forward in BTF currently is just an enum with no enum
2275 * values; we also assume a standard 4-byte size for it
2277 return btf__add_enum(btf, name, sizeof(int));
2279 return libbpf_err(-EINVAL);
2284 * Append new BTF_KING_TYPEDEF type with:
2285 * - *name*, non-empty/non-NULL name;
2286 * - *ref_type_id* - referenced type ID, it might not exist yet;
2288 * - >0, type ID of newly added BTF type;
2291 int btf__add_typedef(struct btf *btf, const char *name, int ref_type_id)
2293 if (!name || !name[0])
2294 return libbpf_err(-EINVAL);
2296 return btf_add_ref_kind(btf, BTF_KIND_TYPEDEF, name, ref_type_id);
2300 * Append new BTF_KIND_VOLATILE type with:
2301 * - *ref_type_id* - referenced type ID, it might not exist yet;
2303 * - >0, type ID of newly added BTF type;
2306 int btf__add_volatile(struct btf *btf, int ref_type_id)
2308 return btf_add_ref_kind(btf, BTF_KIND_VOLATILE, NULL, ref_type_id);
2312 * Append new BTF_KIND_CONST type with:
2313 * - *ref_type_id* - referenced type ID, it might not exist yet;
2315 * - >0, type ID of newly added BTF type;
2318 int btf__add_const(struct btf *btf, int ref_type_id)
2320 return btf_add_ref_kind(btf, BTF_KIND_CONST, NULL, ref_type_id);
2324 * Append new BTF_KIND_RESTRICT type with:
2325 * - *ref_type_id* - referenced type ID, it might not exist yet;
2327 * - >0, type ID of newly added BTF type;
2330 int btf__add_restrict(struct btf *btf, int ref_type_id)
2332 return btf_add_ref_kind(btf, BTF_KIND_RESTRICT, NULL, ref_type_id);
2336 * Append new BTF_KIND_TYPE_TAG type with:
2337 * - *value*, non-empty/non-NULL tag value;
2338 * - *ref_type_id* - referenced type ID, it might not exist yet;
2340 * - >0, type ID of newly added BTF type;
2343 int btf__add_type_tag(struct btf *btf, const char *value, int ref_type_id)
2345 if (!value || !value[0])
2346 return libbpf_err(-EINVAL);
2348 return btf_add_ref_kind(btf, BTF_KIND_TYPE_TAG, value, ref_type_id);
2352 * Append new BTF_KIND_FUNC type with:
2353 * - *name*, non-empty/non-NULL name;
2354 * - *proto_type_id* - FUNC_PROTO's type ID, it might not exist yet;
2356 * - >0, type ID of newly added BTF type;
2359 int btf__add_func(struct btf *btf, const char *name,
2360 enum btf_func_linkage linkage, int proto_type_id)
2364 if (!name || !name[0])
2365 return libbpf_err(-EINVAL);
2366 if (linkage != BTF_FUNC_STATIC && linkage != BTF_FUNC_GLOBAL &&
2367 linkage != BTF_FUNC_EXTERN)
2368 return libbpf_err(-EINVAL);
2370 id = btf_add_ref_kind(btf, BTF_KIND_FUNC, name, proto_type_id);
2372 struct btf_type *t = btf_type_by_id(btf, id);
2374 t->info = btf_type_info(BTF_KIND_FUNC, linkage, 0);
2376 return libbpf_err(id);
2380 * Append new BTF_KIND_FUNC_PROTO with:
2381 * - *ret_type_id* - type ID for return result of a function.
2383 * Function prototype initially has no arguments, but they can be added by
2384 * btf__add_func_param() one by one, immediately after
2385 * btf__add_func_proto() succeeded.
2388 * - >0, type ID of newly added BTF type;
2391 int btf__add_func_proto(struct btf *btf, int ret_type_id)
2396 if (validate_type_id(ret_type_id))
2397 return libbpf_err(-EINVAL);
2399 if (btf_ensure_modifiable(btf))
2400 return libbpf_err(-ENOMEM);
2402 sz = sizeof(struct btf_type);
2403 t = btf_add_type_mem(btf, sz);
2405 return libbpf_err(-ENOMEM);
2407 /* start out with vlen=0; this will be adjusted when adding enum
2408 * values, if necessary
2411 t->info = btf_type_info(BTF_KIND_FUNC_PROTO, 0, 0);
2412 t->type = ret_type_id;
2414 return btf_commit_type(btf, sz);
2418 * Append new function parameter for current FUNC_PROTO type with:
2419 * - *name* - parameter name, can be NULL or empty;
2420 * - *type_id* - type ID describing the type of the parameter.
2425 int btf__add_func_param(struct btf *btf, const char *name, int type_id)
2428 struct btf_param *p;
2429 int sz, name_off = 0;
2431 if (validate_type_id(type_id))
2432 return libbpf_err(-EINVAL);
2434 /* last type should be BTF_KIND_FUNC_PROTO */
2435 if (btf->nr_types == 0)
2436 return libbpf_err(-EINVAL);
2437 t = btf_last_type(btf);
2438 if (!btf_is_func_proto(t))
2439 return libbpf_err(-EINVAL);
2441 /* decompose and invalidate raw data */
2442 if (btf_ensure_modifiable(btf))
2443 return libbpf_err(-ENOMEM);
2445 sz = sizeof(struct btf_param);
2446 p = btf_add_type_mem(btf, sz);
2448 return libbpf_err(-ENOMEM);
2450 if (name && name[0]) {
2451 name_off = btf__add_str(btf, name);
2456 p->name_off = name_off;
2459 /* update parent type's vlen */
2460 t = btf_last_type(btf);
2461 btf_type_inc_vlen(t);
2463 btf->hdr->type_len += sz;
2464 btf->hdr->str_off += sz;
2469 * Append new BTF_KIND_VAR type with:
2470 * - *name* - non-empty/non-NULL name;
2471 * - *linkage* - variable linkage, one of BTF_VAR_STATIC,
2472 * BTF_VAR_GLOBAL_ALLOCATED, or BTF_VAR_GLOBAL_EXTERN;
2473 * - *type_id* - type ID of the type describing the type of the variable.
2475 * - >0, type ID of newly added BTF type;
2478 int btf__add_var(struct btf *btf, const char *name, int linkage, int type_id)
2484 /* non-empty name */
2485 if (!name || !name[0])
2486 return libbpf_err(-EINVAL);
2487 if (linkage != BTF_VAR_STATIC && linkage != BTF_VAR_GLOBAL_ALLOCATED &&
2488 linkage != BTF_VAR_GLOBAL_EXTERN)
2489 return libbpf_err(-EINVAL);
2490 if (validate_type_id(type_id))
2491 return libbpf_err(-EINVAL);
2493 /* deconstruct BTF, if necessary, and invalidate raw_data */
2494 if (btf_ensure_modifiable(btf))
2495 return libbpf_err(-ENOMEM);
2497 sz = sizeof(struct btf_type) + sizeof(struct btf_var);
2498 t = btf_add_type_mem(btf, sz);
2500 return libbpf_err(-ENOMEM);
2502 name_off = btf__add_str(btf, name);
2506 t->name_off = name_off;
2507 t->info = btf_type_info(BTF_KIND_VAR, 0, 0);
2511 v->linkage = linkage;
2513 return btf_commit_type(btf, sz);
2517 * Append new BTF_KIND_DATASEC type with:
2518 * - *name* - non-empty/non-NULL name;
2519 * - *byte_sz* - data section size, in bytes.
2521 * Data section is initially empty. Variables info can be added with
2522 * btf__add_datasec_var_info() calls, after btf__add_datasec() succeeds.
2525 * - >0, type ID of newly added BTF type;
2528 int btf__add_datasec(struct btf *btf, const char *name, __u32 byte_sz)
2533 /* non-empty name */
2534 if (!name || !name[0])
2535 return libbpf_err(-EINVAL);
2537 if (btf_ensure_modifiable(btf))
2538 return libbpf_err(-ENOMEM);
2540 sz = sizeof(struct btf_type);
2541 t = btf_add_type_mem(btf, sz);
2543 return libbpf_err(-ENOMEM);
2545 name_off = btf__add_str(btf, name);
2549 /* start with vlen=0, which will be update as var_secinfos are added */
2550 t->name_off = name_off;
2551 t->info = btf_type_info(BTF_KIND_DATASEC, 0, 0);
2554 return btf_commit_type(btf, sz);
2558 * Append new data section variable information entry for current DATASEC type:
2559 * - *var_type_id* - type ID, describing type of the variable;
2560 * - *offset* - variable offset within data section, in bytes;
2561 * - *byte_sz* - variable size, in bytes.
2567 int btf__add_datasec_var_info(struct btf *btf, int var_type_id, __u32 offset, __u32 byte_sz)
2570 struct btf_var_secinfo *v;
2573 /* last type should be BTF_KIND_DATASEC */
2574 if (btf->nr_types == 0)
2575 return libbpf_err(-EINVAL);
2576 t = btf_last_type(btf);
2577 if (!btf_is_datasec(t))
2578 return libbpf_err(-EINVAL);
2580 if (validate_type_id(var_type_id))
2581 return libbpf_err(-EINVAL);
2583 /* decompose and invalidate raw data */
2584 if (btf_ensure_modifiable(btf))
2585 return libbpf_err(-ENOMEM);
2587 sz = sizeof(struct btf_var_secinfo);
2588 v = btf_add_type_mem(btf, sz);
2590 return libbpf_err(-ENOMEM);
2592 v->type = var_type_id;
2596 /* update parent type's vlen */
2597 t = btf_last_type(btf);
2598 btf_type_inc_vlen(t);
2600 btf->hdr->type_len += sz;
2601 btf->hdr->str_off += sz;
2606 * Append new BTF_KIND_DECL_TAG type with:
2607 * - *value* - non-empty/non-NULL string;
2608 * - *ref_type_id* - referenced type ID, it might not exist yet;
2609 * - *component_idx* - -1 for tagging reference type, otherwise struct/union
2610 * member or function argument index;
2612 * - >0, type ID of newly added BTF type;
2615 int btf__add_decl_tag(struct btf *btf, const char *value, int ref_type_id,
2621 if (!value || !value[0] || component_idx < -1)
2622 return libbpf_err(-EINVAL);
2624 if (validate_type_id(ref_type_id))
2625 return libbpf_err(-EINVAL);
2627 if (btf_ensure_modifiable(btf))
2628 return libbpf_err(-ENOMEM);
2630 sz = sizeof(struct btf_type) + sizeof(struct btf_decl_tag);
2631 t = btf_add_type_mem(btf, sz);
2633 return libbpf_err(-ENOMEM);
2635 value_off = btf__add_str(btf, value);
2639 t->name_off = value_off;
2640 t->info = btf_type_info(BTF_KIND_DECL_TAG, 0, false);
2641 t->type = ref_type_id;
2642 btf_decl_tag(t)->component_idx = component_idx;
2644 return btf_commit_type(btf, sz);
2647 struct btf_ext_sec_setup_param {
2651 struct btf_ext_info *ext_info;
2655 static int btf_ext_setup_info(struct btf_ext *btf_ext,
2656 struct btf_ext_sec_setup_param *ext_sec)
2658 const struct btf_ext_info_sec *sinfo;
2659 struct btf_ext_info *ext_info;
2660 __u32 info_left, record_size;
2662 /* The start of the info sec (including the __u32 record_size). */
2665 if (ext_sec->len == 0)
2668 if (ext_sec->off & 0x03) {
2669 pr_debug(".BTF.ext %s section is not aligned to 4 bytes\n",
2674 info = btf_ext->data + btf_ext->hdr->hdr_len + ext_sec->off;
2675 info_left = ext_sec->len;
2677 if (btf_ext->data + btf_ext->data_size < info + ext_sec->len) {
2678 pr_debug("%s section (off:%u len:%u) is beyond the end of the ELF section .BTF.ext\n",
2679 ext_sec->desc, ext_sec->off, ext_sec->len);
2683 /* At least a record size */
2684 if (info_left < sizeof(__u32)) {
2685 pr_debug(".BTF.ext %s record size not found\n", ext_sec->desc);
2689 /* The record size needs to meet the minimum standard */
2690 record_size = *(__u32 *)info;
2691 if (record_size < ext_sec->min_rec_size ||
2692 record_size & 0x03) {
2693 pr_debug("%s section in .BTF.ext has invalid record size %u\n",
2694 ext_sec->desc, record_size);
2698 sinfo = info + sizeof(__u32);
2699 info_left -= sizeof(__u32);
2701 /* If no records, return failure now so .BTF.ext won't be used. */
2703 pr_debug("%s section in .BTF.ext has no records", ext_sec->desc);
2708 unsigned int sec_hdrlen = sizeof(struct btf_ext_info_sec);
2709 __u64 total_record_size;
2712 if (info_left < sec_hdrlen) {
2713 pr_debug("%s section header is not found in .BTF.ext\n",
2718 num_records = sinfo->num_info;
2719 if (num_records == 0) {
2720 pr_debug("%s section has incorrect num_records in .BTF.ext\n",
2725 total_record_size = sec_hdrlen + (__u64)num_records * record_size;
2726 if (info_left < total_record_size) {
2727 pr_debug("%s section has incorrect num_records in .BTF.ext\n",
2732 info_left -= total_record_size;
2733 sinfo = (void *)sinfo + total_record_size;
2737 ext_info = ext_sec->ext_info;
2738 ext_info->len = ext_sec->len - sizeof(__u32);
2739 ext_info->rec_size = record_size;
2740 ext_info->info = info + sizeof(__u32);
2741 ext_info->sec_cnt = sec_cnt;
2746 static int btf_ext_setup_func_info(struct btf_ext *btf_ext)
2748 struct btf_ext_sec_setup_param param = {
2749 .off = btf_ext->hdr->func_info_off,
2750 .len = btf_ext->hdr->func_info_len,
2751 .min_rec_size = sizeof(struct bpf_func_info_min),
2752 .ext_info = &btf_ext->func_info,
2756 return btf_ext_setup_info(btf_ext, ¶m);
2759 static int btf_ext_setup_line_info(struct btf_ext *btf_ext)
2761 struct btf_ext_sec_setup_param param = {
2762 .off = btf_ext->hdr->line_info_off,
2763 .len = btf_ext->hdr->line_info_len,
2764 .min_rec_size = sizeof(struct bpf_line_info_min),
2765 .ext_info = &btf_ext->line_info,
2766 .desc = "line_info",
2769 return btf_ext_setup_info(btf_ext, ¶m);
2772 static int btf_ext_setup_core_relos(struct btf_ext *btf_ext)
2774 struct btf_ext_sec_setup_param param = {
2775 .off = btf_ext->hdr->core_relo_off,
2776 .len = btf_ext->hdr->core_relo_len,
2777 .min_rec_size = sizeof(struct bpf_core_relo),
2778 .ext_info = &btf_ext->core_relo_info,
2779 .desc = "core_relo",
2782 return btf_ext_setup_info(btf_ext, ¶m);
2785 static int btf_ext_parse_hdr(__u8 *data, __u32 data_size)
2787 const struct btf_ext_header *hdr = (struct btf_ext_header *)data;
2789 if (data_size < offsetofend(struct btf_ext_header, hdr_len) ||
2790 data_size < hdr->hdr_len) {
2791 pr_debug("BTF.ext header not found");
2795 if (hdr->magic == bswap_16(BTF_MAGIC)) {
2796 pr_warn("BTF.ext in non-native endianness is not supported\n");
2798 } else if (hdr->magic != BTF_MAGIC) {
2799 pr_debug("Invalid BTF.ext magic:%x\n", hdr->magic);
2803 if (hdr->version != BTF_VERSION) {
2804 pr_debug("Unsupported BTF.ext version:%u\n", hdr->version);
2809 pr_debug("Unsupported BTF.ext flags:%x\n", hdr->flags);
2813 if (data_size == hdr->hdr_len) {
2814 pr_debug("BTF.ext has no data\n");
2821 void btf_ext__free(struct btf_ext *btf_ext)
2823 if (IS_ERR_OR_NULL(btf_ext))
2825 free(btf_ext->func_info.sec_idxs);
2826 free(btf_ext->line_info.sec_idxs);
2827 free(btf_ext->core_relo_info.sec_idxs);
2828 free(btf_ext->data);
2832 struct btf_ext *btf_ext__new(const __u8 *data, __u32 size)
2834 struct btf_ext *btf_ext;
2837 btf_ext = calloc(1, sizeof(struct btf_ext));
2839 return libbpf_err_ptr(-ENOMEM);
2841 btf_ext->data_size = size;
2842 btf_ext->data = malloc(size);
2843 if (!btf_ext->data) {
2847 memcpy(btf_ext->data, data, size);
2849 err = btf_ext_parse_hdr(btf_ext->data, size);
2853 if (btf_ext->hdr->hdr_len < offsetofend(struct btf_ext_header, line_info_len)) {
2858 err = btf_ext_setup_func_info(btf_ext);
2862 err = btf_ext_setup_line_info(btf_ext);
2866 if (btf_ext->hdr->hdr_len < offsetofend(struct btf_ext_header, core_relo_len))
2867 goto done; /* skip core relos parsing */
2869 err = btf_ext_setup_core_relos(btf_ext);
2875 btf_ext__free(btf_ext);
2876 return libbpf_err_ptr(err);
2882 const void *btf_ext__get_raw_data(const struct btf_ext *btf_ext, __u32 *size)
2884 *size = btf_ext->data_size;
2885 return btf_ext->data;
2890 static struct btf_dedup *btf_dedup_new(struct btf *btf, const struct btf_dedup_opts *opts);
2891 static void btf_dedup_free(struct btf_dedup *d);
2892 static int btf_dedup_prep(struct btf_dedup *d);
2893 static int btf_dedup_strings(struct btf_dedup *d);
2894 static int btf_dedup_prim_types(struct btf_dedup *d);
2895 static int btf_dedup_struct_types(struct btf_dedup *d);
2896 static int btf_dedup_ref_types(struct btf_dedup *d);
2897 static int btf_dedup_resolve_fwds(struct btf_dedup *d);
2898 static int btf_dedup_compact_types(struct btf_dedup *d);
2899 static int btf_dedup_remap_types(struct btf_dedup *d);
2902 * Deduplicate BTF types and strings.
2904 * BTF dedup algorithm takes as an input `struct btf` representing `.BTF` ELF
2905 * section with all BTF type descriptors and string data. It overwrites that
2906 * memory in-place with deduplicated types and strings without any loss of
2907 * information. If optional `struct btf_ext` representing '.BTF.ext' ELF section
2908 * is provided, all the strings referenced from .BTF.ext section are honored
2909 * and updated to point to the right offsets after deduplication.
2911 * If function returns with error, type/string data might be garbled and should
2914 * More verbose and detailed description of both problem btf_dedup is solving,
2915 * as well as solution could be found at:
2916 * https://facebookmicrosites.github.io/bpf/blog/2018/11/14/btf-enhancement.html
2918 * Problem description and justification
2919 * =====================================
2921 * BTF type information is typically emitted either as a result of conversion
2922 * from DWARF to BTF or directly by compiler. In both cases, each compilation
2923 * unit contains information about a subset of all the types that are used
2924 * in an application. These subsets are frequently overlapping and contain a lot
2925 * of duplicated information when later concatenated together into a single
2926 * binary. This algorithm ensures that each unique type is represented by single
2927 * BTF type descriptor, greatly reducing resulting size of BTF data.
2929 * Compilation unit isolation and subsequent duplication of data is not the only
2930 * problem. The same type hierarchy (e.g., struct and all the type that struct
2931 * references) in different compilation units can be represented in BTF to
2932 * various degrees of completeness (or, rather, incompleteness) due to
2933 * struct/union forward declarations.
2935 * Let's take a look at an example, that we'll use to better understand the
2936 * problem (and solution). Suppose we have two compilation units, each using
2937 * same `struct S`, but each of them having incomplete type information about
2966 * In case of CU #1, BTF data will know only that `struct B` exist (but no
2967 * more), but will know the complete type information about `struct A`. While
2968 * for CU #2, it will know full type information about `struct B`, but will
2969 * only know about forward declaration of `struct A` (in BTF terms, it will
2970 * have `BTF_KIND_FWD` type descriptor with name `B`).
2972 * This compilation unit isolation means that it's possible that there is no
2973 * single CU with complete type information describing structs `S`, `A`, and
2974 * `B`. Also, we might get tons of duplicated and redundant type information.
2976 * Additional complication we need to keep in mind comes from the fact that
2977 * types, in general, can form graphs containing cycles, not just DAGs.
2979 * While algorithm does deduplication, it also merges and resolves type
2980 * information (unless disabled throught `struct btf_opts`), whenever possible.
2981 * E.g., in the example above with two compilation units having partial type
2982 * information for structs `A` and `B`, the output of algorithm will emit
2983 * a single copy of each BTF type that describes structs `A`, `B`, and `S`
2984 * (as well as type information for `int` and pointers), as if they were defined
2985 * in a single compilation unit as:
3005 * Algorithm completes its work in 7 separate passes:
3007 * 1. Strings deduplication.
3008 * 2. Primitive types deduplication (int, enum, fwd).
3009 * 3. Struct/union types deduplication.
3010 * 4. Resolve unambiguous forward declarations.
3011 * 5. Reference types deduplication (pointers, typedefs, arrays, funcs, func
3012 * protos, and const/volatile/restrict modifiers).
3013 * 6. Types compaction.
3014 * 7. Types remapping.
3016 * Algorithm determines canonical type descriptor, which is a single
3017 * representative type for each truly unique type. This canonical type is the
3018 * one that will go into final deduplicated BTF type information. For
3019 * struct/unions, it is also the type that algorithm will merge additional type
3020 * information into (while resolving FWDs), as it discovers it from data in
3021 * other CUs. Each input BTF type eventually gets either mapped to itself, if
3022 * that type is canonical, or to some other type, if that type is equivalent
3023 * and was chosen as canonical representative. This mapping is stored in
3024 * `btf_dedup->map` array. This map is also used to record STRUCT/UNION that
3025 * FWD type got resolved to.
3027 * To facilitate fast discovery of canonical types, we also maintain canonical
3028 * index (`btf_dedup->dedup_table`), which maps type descriptor's signature hash
3029 * (i.e., hashed kind, name, size, fields, etc) into a list of canonical types
3030 * that match that signature. With sufficiently good choice of type signature
3031 * hashing function, we can limit number of canonical types for each unique type
3032 * signature to a very small number, allowing to find canonical type for any
3033 * duplicated type very quickly.
3035 * Struct/union deduplication is the most critical part and algorithm for
3036 * deduplicating structs/unions is described in greater details in comments for
3037 * `btf_dedup_is_equiv` function.
3039 int btf__dedup(struct btf *btf, const struct btf_dedup_opts *opts)
3041 struct btf_dedup *d;
3044 if (!OPTS_VALID(opts, btf_dedup_opts))
3045 return libbpf_err(-EINVAL);
3047 d = btf_dedup_new(btf, opts);
3049 pr_debug("btf_dedup_new failed: %ld", PTR_ERR(d));
3050 return libbpf_err(-EINVAL);
3053 if (btf_ensure_modifiable(btf)) {
3058 err = btf_dedup_prep(d);
3060 pr_debug("btf_dedup_prep failed:%d\n", err);
3063 err = btf_dedup_strings(d);
3065 pr_debug("btf_dedup_strings failed:%d\n", err);
3068 err = btf_dedup_prim_types(d);
3070 pr_debug("btf_dedup_prim_types failed:%d\n", err);
3073 err = btf_dedup_struct_types(d);
3075 pr_debug("btf_dedup_struct_types failed:%d\n", err);
3078 err = btf_dedup_resolve_fwds(d);
3080 pr_debug("btf_dedup_resolve_fwds failed:%d\n", err);
3083 err = btf_dedup_ref_types(d);
3085 pr_debug("btf_dedup_ref_types failed:%d\n", err);
3088 err = btf_dedup_compact_types(d);
3090 pr_debug("btf_dedup_compact_types failed:%d\n", err);
3093 err = btf_dedup_remap_types(d);
3095 pr_debug("btf_dedup_remap_types failed:%d\n", err);
3101 return libbpf_err(err);
3104 #define BTF_UNPROCESSED_ID ((__u32)-1)
3105 #define BTF_IN_PROGRESS_ID ((__u32)-2)
3108 /* .BTF section to be deduped in-place */
3111 * Optional .BTF.ext section. When provided, any strings referenced
3112 * from it will be taken into account when deduping strings
3114 struct btf_ext *btf_ext;
3116 * This is a map from any type's signature hash to a list of possible
3117 * canonical representative type candidates. Hash collisions are
3118 * ignored, so even types of various kinds can share same list of
3119 * candidates, which is fine because we rely on subsequent
3120 * btf_xxx_equal() checks to authoritatively verify type equality.
3122 struct hashmap *dedup_table;
3123 /* Canonical types map */
3125 /* Hypothetical mapping, used during type graph equivalence checks */
3130 /* Whether hypothetical mapping, if successful, would need to adjust
3131 * already canonicalized types (due to a new forward declaration to
3132 * concrete type resolution). In such case, during split BTF dedup
3133 * candidate type would still be considered as different, because base
3134 * BTF is considered to be immutable.
3136 bool hypot_adjust_canon;
3137 /* Various option modifying behavior of algorithm */
3138 struct btf_dedup_opts opts;
3139 /* temporary strings deduplication state */
3140 struct strset *strs_set;
3143 static long hash_combine(long h, long value)
3145 return h * 31 + value;
3148 #define for_each_dedup_cand(d, node, hash) \
3149 hashmap__for_each_key_entry(d->dedup_table, node, hash)
3151 static int btf_dedup_table_add(struct btf_dedup *d, long hash, __u32 type_id)
3153 return hashmap__append(d->dedup_table, hash, type_id);
3156 static int btf_dedup_hypot_map_add(struct btf_dedup *d,
3157 __u32 from_id, __u32 to_id)
3159 if (d->hypot_cnt == d->hypot_cap) {
3162 d->hypot_cap += max((size_t)16, d->hypot_cap / 2);
3163 new_list = libbpf_reallocarray(d->hypot_list, d->hypot_cap, sizeof(__u32));
3166 d->hypot_list = new_list;
3168 d->hypot_list[d->hypot_cnt++] = from_id;
3169 d->hypot_map[from_id] = to_id;
3173 static void btf_dedup_clear_hypot_map(struct btf_dedup *d)
3177 for (i = 0; i < d->hypot_cnt; i++)
3178 d->hypot_map[d->hypot_list[i]] = BTF_UNPROCESSED_ID;
3180 d->hypot_adjust_canon = false;
3183 static void btf_dedup_free(struct btf_dedup *d)
3185 hashmap__free(d->dedup_table);
3186 d->dedup_table = NULL;
3192 d->hypot_map = NULL;
3194 free(d->hypot_list);
3195 d->hypot_list = NULL;
3200 static size_t btf_dedup_identity_hash_fn(long key, void *ctx)
3205 static size_t btf_dedup_collision_hash_fn(long key, void *ctx)
3210 static bool btf_dedup_equal_fn(long k1, long k2, void *ctx)
3215 static struct btf_dedup *btf_dedup_new(struct btf *btf, const struct btf_dedup_opts *opts)
3217 struct btf_dedup *d = calloc(1, sizeof(struct btf_dedup));
3218 hashmap_hash_fn hash_fn = btf_dedup_identity_hash_fn;
3219 int i, err = 0, type_cnt;
3222 return ERR_PTR(-ENOMEM);
3224 if (OPTS_GET(opts, force_collisions, false))
3225 hash_fn = btf_dedup_collision_hash_fn;
3228 d->btf_ext = OPTS_GET(opts, btf_ext, NULL);
3230 d->dedup_table = hashmap__new(hash_fn, btf_dedup_equal_fn, NULL);
3231 if (IS_ERR(d->dedup_table)) {
3232 err = PTR_ERR(d->dedup_table);
3233 d->dedup_table = NULL;
3237 type_cnt = btf__type_cnt(btf);
3238 d->map = malloc(sizeof(__u32) * type_cnt);
3243 /* special BTF "void" type is made canonical immediately */
3245 for (i = 1; i < type_cnt; i++) {
3246 struct btf_type *t = btf_type_by_id(d->btf, i);
3248 /* VAR and DATASEC are never deduped and are self-canonical */
3249 if (btf_is_var(t) || btf_is_datasec(t))
3252 d->map[i] = BTF_UNPROCESSED_ID;
3255 d->hypot_map = malloc(sizeof(__u32) * type_cnt);
3256 if (!d->hypot_map) {
3260 for (i = 0; i < type_cnt; i++)
3261 d->hypot_map[i] = BTF_UNPROCESSED_ID;
3266 return ERR_PTR(err);
3273 * Iterate over all possible places in .BTF and .BTF.ext that can reference
3274 * string and pass pointer to it to a provided callback `fn`.
3276 static int btf_for_each_str_off(struct btf_dedup *d, str_off_visit_fn fn, void *ctx)
3280 for (i = 0; i < d->btf->nr_types; i++) {
3281 struct btf_type *t = btf_type_by_id(d->btf, d->btf->start_id + i);
3283 r = btf_type_visit_str_offs(t, fn, ctx);
3291 r = btf_ext_visit_str_offs(d->btf_ext, fn, ctx);
3298 static int strs_dedup_remap_str_off(__u32 *str_off_ptr, void *ctx)
3300 struct btf_dedup *d = ctx;
3301 __u32 str_off = *str_off_ptr;
3305 /* don't touch empty string or string in main BTF */
3306 if (str_off == 0 || str_off < d->btf->start_str_off)
3309 s = btf__str_by_offset(d->btf, str_off);
3310 if (d->btf->base_btf) {
3311 err = btf__find_str(d->btf->base_btf, s);
3320 off = strset__add_str(d->strs_set, s);
3324 *str_off_ptr = d->btf->start_str_off + off;
3329 * Dedup string and filter out those that are not referenced from either .BTF
3330 * or .BTF.ext (if provided) sections.
3332 * This is done by building index of all strings in BTF's string section,
3333 * then iterating over all entities that can reference strings (e.g., type
3334 * names, struct field names, .BTF.ext line info, etc) and marking corresponding
3335 * strings as used. After that all used strings are deduped and compacted into
3336 * sequential blob of memory and new offsets are calculated. Then all the string
3337 * references are iterated again and rewritten using new offsets.
3339 static int btf_dedup_strings(struct btf_dedup *d)
3343 if (d->btf->strs_deduped)
3346 d->strs_set = strset__new(BTF_MAX_STR_OFFSET, NULL, 0);
3347 if (IS_ERR(d->strs_set)) {
3348 err = PTR_ERR(d->strs_set);
3352 if (!d->btf->base_btf) {
3353 /* insert empty string; we won't be looking it up during strings
3354 * dedup, but it's good to have it for generic BTF string lookups
3356 err = strset__add_str(d->strs_set, "");
3361 /* remap string offsets */
3362 err = btf_for_each_str_off(d, strs_dedup_remap_str_off, d);
3366 /* replace BTF string data and hash with deduped ones */
3367 strset__free(d->btf->strs_set);
3368 d->btf->hdr->str_len = strset__data_size(d->strs_set);
3369 d->btf->strs_set = d->strs_set;
3371 d->btf->strs_deduped = true;
3375 strset__free(d->strs_set);
3381 static long btf_hash_common(struct btf_type *t)
3385 h = hash_combine(0, t->name_off);
3386 h = hash_combine(h, t->info);
3387 h = hash_combine(h, t->size);
3391 static bool btf_equal_common(struct btf_type *t1, struct btf_type *t2)
3393 return t1->name_off == t2->name_off &&
3394 t1->info == t2->info &&
3395 t1->size == t2->size;
3398 /* Calculate type signature hash of INT or TAG. */
3399 static long btf_hash_int_decl_tag(struct btf_type *t)
3401 __u32 info = *(__u32 *)(t + 1);
3404 h = btf_hash_common(t);
3405 h = hash_combine(h, info);
3409 /* Check structural equality of two INTs or TAGs. */
3410 static bool btf_equal_int_tag(struct btf_type *t1, struct btf_type *t2)
3414 if (!btf_equal_common(t1, t2))
3416 info1 = *(__u32 *)(t1 + 1);
3417 info2 = *(__u32 *)(t2 + 1);
3418 return info1 == info2;
3421 /* Calculate type signature hash of ENUM/ENUM64. */
3422 static long btf_hash_enum(struct btf_type *t)
3426 /* don't hash vlen, enum members and size to support enum fwd resolving */
3427 h = hash_combine(0, t->name_off);
3431 static bool btf_equal_enum_members(struct btf_type *t1, struct btf_type *t2)
3433 const struct btf_enum *m1, *m2;
3437 vlen = btf_vlen(t1);
3440 for (i = 0; i < vlen; i++) {
3441 if (m1->name_off != m2->name_off || m1->val != m2->val)
3449 static bool btf_equal_enum64_members(struct btf_type *t1, struct btf_type *t2)
3451 const struct btf_enum64 *m1, *m2;
3455 vlen = btf_vlen(t1);
3456 m1 = btf_enum64(t1);
3457 m2 = btf_enum64(t2);
3458 for (i = 0; i < vlen; i++) {
3459 if (m1->name_off != m2->name_off || m1->val_lo32 != m2->val_lo32 ||
3460 m1->val_hi32 != m2->val_hi32)
3468 /* Check structural equality of two ENUMs or ENUM64s. */
3469 static bool btf_equal_enum(struct btf_type *t1, struct btf_type *t2)
3471 if (!btf_equal_common(t1, t2))
3474 /* t1 & t2 kinds are identical because of btf_equal_common */
3475 if (btf_kind(t1) == BTF_KIND_ENUM)
3476 return btf_equal_enum_members(t1, t2);
3478 return btf_equal_enum64_members(t1, t2);
3481 static inline bool btf_is_enum_fwd(struct btf_type *t)
3483 return btf_is_any_enum(t) && btf_vlen(t) == 0;
3486 static bool btf_compat_enum(struct btf_type *t1, struct btf_type *t2)
3488 if (!btf_is_enum_fwd(t1) && !btf_is_enum_fwd(t2))
3489 return btf_equal_enum(t1, t2);
3490 /* At this point either t1 or t2 or both are forward declarations, thus:
3491 * - skip comparing vlen because it is zero for forward declarations;
3492 * - skip comparing size to allow enum forward declarations
3493 * to be compatible with enum64 full declarations;
3494 * - skip comparing kind for the same reason.
3496 return t1->name_off == t2->name_off &&
3497 btf_is_any_enum(t1) && btf_is_any_enum(t2);
3501 * Calculate type signature hash of STRUCT/UNION, ignoring referenced type IDs,
3502 * as referenced type IDs equivalence is established separately during type
3503 * graph equivalence check algorithm.
3505 static long btf_hash_struct(struct btf_type *t)
3507 const struct btf_member *member = btf_members(t);
3508 __u32 vlen = btf_vlen(t);
3509 long h = btf_hash_common(t);
3512 for (i = 0; i < vlen; i++) {
3513 h = hash_combine(h, member->name_off);
3514 h = hash_combine(h, member->offset);
3515 /* no hashing of referenced type ID, it can be unresolved yet */
3522 * Check structural compatibility of two STRUCTs/UNIONs, ignoring referenced
3523 * type IDs. This check is performed during type graph equivalence check and
3524 * referenced types equivalence is checked separately.
3526 static bool btf_shallow_equal_struct(struct btf_type *t1, struct btf_type *t2)
3528 const struct btf_member *m1, *m2;
3532 if (!btf_equal_common(t1, t2))
3535 vlen = btf_vlen(t1);
3536 m1 = btf_members(t1);
3537 m2 = btf_members(t2);
3538 for (i = 0; i < vlen; i++) {
3539 if (m1->name_off != m2->name_off || m1->offset != m2->offset)
3548 * Calculate type signature hash of ARRAY, including referenced type IDs,
3549 * under assumption that they were already resolved to canonical type IDs and
3550 * are not going to change.
3552 static long btf_hash_array(struct btf_type *t)
3554 const struct btf_array *info = btf_array(t);
3555 long h = btf_hash_common(t);
3557 h = hash_combine(h, info->type);
3558 h = hash_combine(h, info->index_type);
3559 h = hash_combine(h, info->nelems);
3564 * Check exact equality of two ARRAYs, taking into account referenced
3565 * type IDs, under assumption that they were already resolved to canonical
3566 * type IDs and are not going to change.
3567 * This function is called during reference types deduplication to compare
3568 * ARRAY to potential canonical representative.
3570 static bool btf_equal_array(struct btf_type *t1, struct btf_type *t2)
3572 const struct btf_array *info1, *info2;
3574 if (!btf_equal_common(t1, t2))
3577 info1 = btf_array(t1);
3578 info2 = btf_array(t2);
3579 return info1->type == info2->type &&
3580 info1->index_type == info2->index_type &&
3581 info1->nelems == info2->nelems;
3585 * Check structural compatibility of two ARRAYs, ignoring referenced type
3586 * IDs. This check is performed during type graph equivalence check and
3587 * referenced types equivalence is checked separately.
3589 static bool btf_compat_array(struct btf_type *t1, struct btf_type *t2)
3591 if (!btf_equal_common(t1, t2))
3594 return btf_array(t1)->nelems == btf_array(t2)->nelems;
3598 * Calculate type signature hash of FUNC_PROTO, including referenced type IDs,
3599 * under assumption that they were already resolved to canonical type IDs and
3600 * are not going to change.
3602 static long btf_hash_fnproto(struct btf_type *t)
3604 const struct btf_param *member = btf_params(t);
3605 __u16 vlen = btf_vlen(t);
3606 long h = btf_hash_common(t);
3609 for (i = 0; i < vlen; i++) {
3610 h = hash_combine(h, member->name_off);
3611 h = hash_combine(h, member->type);
3618 * Check exact equality of two FUNC_PROTOs, taking into account referenced
3619 * type IDs, under assumption that they were already resolved to canonical
3620 * type IDs and are not going to change.
3621 * This function is called during reference types deduplication to compare
3622 * FUNC_PROTO to potential canonical representative.
3624 static bool btf_equal_fnproto(struct btf_type *t1, struct btf_type *t2)
3626 const struct btf_param *m1, *m2;
3630 if (!btf_equal_common(t1, t2))
3633 vlen = btf_vlen(t1);
3634 m1 = btf_params(t1);
3635 m2 = btf_params(t2);
3636 for (i = 0; i < vlen; i++) {
3637 if (m1->name_off != m2->name_off || m1->type != m2->type)
3646 * Check structural compatibility of two FUNC_PROTOs, ignoring referenced type
3647 * IDs. This check is performed during type graph equivalence check and
3648 * referenced types equivalence is checked separately.
3650 static bool btf_compat_fnproto(struct btf_type *t1, struct btf_type *t2)
3652 const struct btf_param *m1, *m2;
3656 /* skip return type ID */
3657 if (t1->name_off != t2->name_off || t1->info != t2->info)
3660 vlen = btf_vlen(t1);
3661 m1 = btf_params(t1);
3662 m2 = btf_params(t2);
3663 for (i = 0; i < vlen; i++) {
3664 if (m1->name_off != m2->name_off)
3672 /* Prepare split BTF for deduplication by calculating hashes of base BTF's
3673 * types and initializing the rest of the state (canonical type mapping) for
3674 * the fixed base BTF part.
3676 static int btf_dedup_prep(struct btf_dedup *d)
3682 if (!d->btf->base_btf)
3685 for (type_id = 1; type_id < d->btf->start_id; type_id++) {
3686 t = btf_type_by_id(d->btf, type_id);
3688 /* all base BTF types are self-canonical by definition */
3689 d->map[type_id] = type_id;
3691 switch (btf_kind(t)) {
3693 case BTF_KIND_DATASEC:
3694 /* VAR and DATASEC are never hash/deduplicated */
3696 case BTF_KIND_CONST:
3697 case BTF_KIND_VOLATILE:
3698 case BTF_KIND_RESTRICT:
3701 case BTF_KIND_TYPEDEF:
3703 case BTF_KIND_FLOAT:
3704 case BTF_KIND_TYPE_TAG:
3705 h = btf_hash_common(t);
3708 case BTF_KIND_DECL_TAG:
3709 h = btf_hash_int_decl_tag(t);
3712 case BTF_KIND_ENUM64:
3713 h = btf_hash_enum(t);
3715 case BTF_KIND_STRUCT:
3716 case BTF_KIND_UNION:
3717 h = btf_hash_struct(t);
3719 case BTF_KIND_ARRAY:
3720 h = btf_hash_array(t);
3722 case BTF_KIND_FUNC_PROTO:
3723 h = btf_hash_fnproto(t);
3726 pr_debug("unknown kind %d for type [%d]\n", btf_kind(t), type_id);
3729 if (btf_dedup_table_add(d, h, type_id))
3737 * Deduplicate primitive types, that can't reference other types, by calculating
3738 * their type signature hash and comparing them with any possible canonical
3739 * candidate. If no canonical candidate matches, type itself is marked as
3740 * canonical and is added into `btf_dedup->dedup_table` as another candidate.
3742 static int btf_dedup_prim_type(struct btf_dedup *d, __u32 type_id)
3744 struct btf_type *t = btf_type_by_id(d->btf, type_id);
3745 struct hashmap_entry *hash_entry;
3746 struct btf_type *cand;
3747 /* if we don't find equivalent type, then we are canonical */
3748 __u32 new_id = type_id;
3752 switch (btf_kind(t)) {
3753 case BTF_KIND_CONST:
3754 case BTF_KIND_VOLATILE:
3755 case BTF_KIND_RESTRICT:
3757 case BTF_KIND_TYPEDEF:
3758 case BTF_KIND_ARRAY:
3759 case BTF_KIND_STRUCT:
3760 case BTF_KIND_UNION:
3762 case BTF_KIND_FUNC_PROTO:
3764 case BTF_KIND_DATASEC:
3765 case BTF_KIND_DECL_TAG:
3766 case BTF_KIND_TYPE_TAG:
3770 h = btf_hash_int_decl_tag(t);
3771 for_each_dedup_cand(d, hash_entry, h) {
3772 cand_id = hash_entry->value;
3773 cand = btf_type_by_id(d->btf, cand_id);
3774 if (btf_equal_int_tag(t, cand)) {
3782 case BTF_KIND_ENUM64:
3783 h = btf_hash_enum(t);
3784 for_each_dedup_cand(d, hash_entry, h) {
3785 cand_id = hash_entry->value;
3786 cand = btf_type_by_id(d->btf, cand_id);
3787 if (btf_equal_enum(t, cand)) {
3791 if (btf_compat_enum(t, cand)) {
3792 if (btf_is_enum_fwd(t)) {
3793 /* resolve fwd to full enum */
3797 /* resolve canonical enum fwd to full enum */
3798 d->map[cand_id] = type_id;
3804 case BTF_KIND_FLOAT:
3805 h = btf_hash_common(t);
3806 for_each_dedup_cand(d, hash_entry, h) {
3807 cand_id = hash_entry->value;
3808 cand = btf_type_by_id(d->btf, cand_id);
3809 if (btf_equal_common(t, cand)) {
3820 d->map[type_id] = new_id;
3821 if (type_id == new_id && btf_dedup_table_add(d, h, type_id))
3827 static int btf_dedup_prim_types(struct btf_dedup *d)
3831 for (i = 0; i < d->btf->nr_types; i++) {
3832 err = btf_dedup_prim_type(d, d->btf->start_id + i);
3840 * Check whether type is already mapped into canonical one (could be to itself).
3842 static inline bool is_type_mapped(struct btf_dedup *d, uint32_t type_id)
3844 return d->map[type_id] <= BTF_MAX_NR_TYPES;
3848 * Resolve type ID into its canonical type ID, if any; otherwise return original
3849 * type ID. If type is FWD and is resolved into STRUCT/UNION already, follow
3850 * STRUCT/UNION link and resolve it into canonical type ID as well.
3852 static inline __u32 resolve_type_id(struct btf_dedup *d, __u32 type_id)
3854 while (is_type_mapped(d, type_id) && d->map[type_id] != type_id)
3855 type_id = d->map[type_id];
3860 * Resolve FWD to underlying STRUCT/UNION, if any; otherwise return original
3863 static uint32_t resolve_fwd_id(struct btf_dedup *d, uint32_t type_id)
3865 __u32 orig_type_id = type_id;
3867 if (!btf_is_fwd(btf__type_by_id(d->btf, type_id)))
3870 while (is_type_mapped(d, type_id) && d->map[type_id] != type_id)
3871 type_id = d->map[type_id];
3873 if (!btf_is_fwd(btf__type_by_id(d->btf, type_id)))
3876 return orig_type_id;
3880 static inline __u16 btf_fwd_kind(struct btf_type *t)
3882 return btf_kflag(t) ? BTF_KIND_UNION : BTF_KIND_STRUCT;
3885 /* Check if given two types are identical ARRAY definitions */
3886 static bool btf_dedup_identical_arrays(struct btf_dedup *d, __u32 id1, __u32 id2)
3888 struct btf_type *t1, *t2;
3890 t1 = btf_type_by_id(d->btf, id1);
3891 t2 = btf_type_by_id(d->btf, id2);
3892 if (!btf_is_array(t1) || !btf_is_array(t2))
3895 return btf_equal_array(t1, t2);
3898 /* Check if given two types are identical STRUCT/UNION definitions */
3899 static bool btf_dedup_identical_structs(struct btf_dedup *d, __u32 id1, __u32 id2)
3901 const struct btf_member *m1, *m2;
3902 struct btf_type *t1, *t2;
3905 t1 = btf_type_by_id(d->btf, id1);
3906 t2 = btf_type_by_id(d->btf, id2);
3908 if (!btf_is_composite(t1) || btf_kind(t1) != btf_kind(t2))
3911 if (!btf_shallow_equal_struct(t1, t2))
3914 m1 = btf_members(t1);
3915 m2 = btf_members(t2);
3916 for (i = 0, n = btf_vlen(t1); i < n; i++, m1++, m2++) {
3917 if (m1->type != m2->type &&
3918 !btf_dedup_identical_arrays(d, m1->type, m2->type) &&
3919 !btf_dedup_identical_structs(d, m1->type, m2->type))
3926 * Check equivalence of BTF type graph formed by candidate struct/union (we'll
3927 * call it "candidate graph" in this description for brevity) to a type graph
3928 * formed by (potential) canonical struct/union ("canonical graph" for brevity
3929 * here, though keep in mind that not all types in canonical graph are
3930 * necessarily canonical representatives themselves, some of them might be
3931 * duplicates or its uniqueness might not have been established yet).
3933 * - >0, if type graphs are equivalent;
3934 * - 0, if not equivalent;
3937 * Algorithm performs side-by-side DFS traversal of both type graphs and checks
3938 * equivalence of BTF types at each step. If at any point BTF types in candidate
3939 * and canonical graphs are not compatible structurally, whole graphs are
3940 * incompatible. If types are structurally equivalent (i.e., all information
3941 * except referenced type IDs is exactly the same), a mapping from `canon_id` to
3942 * a `cand_id` is recored in hypothetical mapping (`btf_dedup->hypot_map`).
3943 * If a type references other types, then those referenced types are checked
3944 * for equivalence recursively.
3946 * During DFS traversal, if we find that for current `canon_id` type we
3947 * already have some mapping in hypothetical map, we check for two possible
3949 * - `canon_id` is mapped to exactly the same type as `cand_id`. This will
3950 * happen when type graphs have cycles. In this case we assume those two
3951 * types are equivalent.
3952 * - `canon_id` is mapped to different type. This is contradiction in our
3953 * hypothetical mapping, because same graph in canonical graph corresponds
3954 * to two different types in candidate graph, which for equivalent type
3955 * graphs shouldn't happen. This condition terminates equivalence check
3956 * with negative result.
3958 * If type graphs traversal exhausts types to check and find no contradiction,
3959 * then type graphs are equivalent.
3961 * When checking types for equivalence, there is one special case: FWD types.
3962 * If FWD type resolution is allowed and one of the types (either from canonical
3963 * or candidate graph) is FWD and other is STRUCT/UNION (depending on FWD's kind
3964 * flag) and their names match, hypothetical mapping is updated to point from
3965 * FWD to STRUCT/UNION. If graphs will be determined as equivalent successfully,
3966 * this mapping will be used to record FWD -> STRUCT/UNION mapping permanently.
3968 * Technically, this could lead to incorrect FWD to STRUCT/UNION resolution,
3969 * if there are two exactly named (or anonymous) structs/unions that are
3970 * compatible structurally, one of which has FWD field, while other is concrete
3971 * STRUCT/UNION, but according to C sources they are different structs/unions
3972 * that are referencing different types with the same name. This is extremely
3973 * unlikely to happen, but btf_dedup API allows to disable FWD resolution if
3974 * this logic is causing problems.
3976 * Doing FWD resolution means that both candidate and/or canonical graphs can
3977 * consists of portions of the graph that come from multiple compilation units.
3978 * This is due to the fact that types within single compilation unit are always
3979 * deduplicated and FWDs are already resolved, if referenced struct/union
3980 * definiton is available. So, if we had unresolved FWD and found corresponding
3981 * STRUCT/UNION, they will be from different compilation units. This
3982 * consequently means that when we "link" FWD to corresponding STRUCT/UNION,
3983 * type graph will likely have at least two different BTF types that describe
3984 * same type (e.g., most probably there will be two different BTF types for the
3985 * same 'int' primitive type) and could even have "overlapping" parts of type
3986 * graph that describe same subset of types.
3988 * This in turn means that our assumption that each type in canonical graph
3989 * must correspond to exactly one type in candidate graph might not hold
3990 * anymore and will make it harder to detect contradictions using hypothetical
3991 * map. To handle this problem, we allow to follow FWD -> STRUCT/UNION
3992 * resolution only in canonical graph. FWDs in candidate graphs are never
3993 * resolved. To see why it's OK, let's check all possible situations w.r.t. FWDs
3995 * - Both types in canonical and candidate graphs are FWDs. If they are
3996 * structurally equivalent, then they can either be both resolved to the
3997 * same STRUCT/UNION or not resolved at all. In both cases they are
3998 * equivalent and there is no need to resolve FWD on candidate side.
3999 * - Both types in canonical and candidate graphs are concrete STRUCT/UNION,
4000 * so nothing to resolve as well, algorithm will check equivalence anyway.
4001 * - Type in canonical graph is FWD, while type in candidate is concrete
4002 * STRUCT/UNION. In this case candidate graph comes from single compilation
4003 * unit, so there is exactly one BTF type for each unique C type. After
4004 * resolving FWD into STRUCT/UNION, there might be more than one BTF type
4005 * in canonical graph mapping to single BTF type in candidate graph, but
4006 * because hypothetical mapping maps from canonical to candidate types, it's
4007 * alright, and we still maintain the property of having single `canon_id`
4008 * mapping to single `cand_id` (there could be two different `canon_id`
4009 * mapped to the same `cand_id`, but it's not contradictory).
4010 * - Type in canonical graph is concrete STRUCT/UNION, while type in candidate
4011 * graph is FWD. In this case we are just going to check compatibility of
4012 * STRUCT/UNION and corresponding FWD, and if they are compatible, we'll
4013 * assume that whatever STRUCT/UNION FWD resolves to must be equivalent to
4014 * a concrete STRUCT/UNION from canonical graph. If the rest of type graphs
4015 * turn out equivalent, we'll re-resolve FWD to concrete STRUCT/UNION from
4018 static int btf_dedup_is_equiv(struct btf_dedup *d, __u32 cand_id,
4021 struct btf_type *cand_type;
4022 struct btf_type *canon_type;
4023 __u32 hypot_type_id;
4028 /* if both resolve to the same canonical, they must be equivalent */
4029 if (resolve_type_id(d, cand_id) == resolve_type_id(d, canon_id))
4032 canon_id = resolve_fwd_id(d, canon_id);
4034 hypot_type_id = d->hypot_map[canon_id];
4035 if (hypot_type_id <= BTF_MAX_NR_TYPES) {
4036 if (hypot_type_id == cand_id)
4038 /* In some cases compiler will generate different DWARF types
4039 * for *identical* array type definitions and use them for
4040 * different fields within the *same* struct. This breaks type
4041 * equivalence check, which makes an assumption that candidate
4042 * types sub-graph has a consistent and deduped-by-compiler
4043 * types within a single CU. So work around that by explicitly
4044 * allowing identical array types here.
4046 if (btf_dedup_identical_arrays(d, hypot_type_id, cand_id))
4048 /* It turns out that similar situation can happen with
4049 * struct/union sometimes, sigh... Handle the case where
4050 * structs/unions are exactly the same, down to the referenced
4051 * type IDs. Anything more complicated (e.g., if referenced
4052 * types are different, but equivalent) is *way more*
4053 * complicated and requires a many-to-many equivalence mapping.
4055 if (btf_dedup_identical_structs(d, hypot_type_id, cand_id))
4060 if (btf_dedup_hypot_map_add(d, canon_id, cand_id))
4063 cand_type = btf_type_by_id(d->btf, cand_id);
4064 canon_type = btf_type_by_id(d->btf, canon_id);
4065 cand_kind = btf_kind(cand_type);
4066 canon_kind = btf_kind(canon_type);
4068 if (cand_type->name_off != canon_type->name_off)
4071 /* FWD <--> STRUCT/UNION equivalence check, if enabled */
4072 if ((cand_kind == BTF_KIND_FWD || canon_kind == BTF_KIND_FWD)
4073 && cand_kind != canon_kind) {
4077 if (cand_kind == BTF_KIND_FWD) {
4078 real_kind = canon_kind;
4079 fwd_kind = btf_fwd_kind(cand_type);
4081 real_kind = cand_kind;
4082 fwd_kind = btf_fwd_kind(canon_type);
4083 /* we'd need to resolve base FWD to STRUCT/UNION */
4084 if (fwd_kind == real_kind && canon_id < d->btf->start_id)
4085 d->hypot_adjust_canon = true;
4087 return fwd_kind == real_kind;
4090 if (cand_kind != canon_kind)
4093 switch (cand_kind) {
4095 return btf_equal_int_tag(cand_type, canon_type);
4098 case BTF_KIND_ENUM64:
4099 return btf_compat_enum(cand_type, canon_type);
4102 case BTF_KIND_FLOAT:
4103 return btf_equal_common(cand_type, canon_type);
4105 case BTF_KIND_CONST:
4106 case BTF_KIND_VOLATILE:
4107 case BTF_KIND_RESTRICT:
4109 case BTF_KIND_TYPEDEF:
4111 case BTF_KIND_TYPE_TAG:
4112 if (cand_type->info != canon_type->info)
4114 return btf_dedup_is_equiv(d, cand_type->type, canon_type->type);
4116 case BTF_KIND_ARRAY: {
4117 const struct btf_array *cand_arr, *canon_arr;
4119 if (!btf_compat_array(cand_type, canon_type))
4121 cand_arr = btf_array(cand_type);
4122 canon_arr = btf_array(canon_type);
4123 eq = btf_dedup_is_equiv(d, cand_arr->index_type, canon_arr->index_type);
4126 return btf_dedup_is_equiv(d, cand_arr->type, canon_arr->type);
4129 case BTF_KIND_STRUCT:
4130 case BTF_KIND_UNION: {
4131 const struct btf_member *cand_m, *canon_m;
4134 if (!btf_shallow_equal_struct(cand_type, canon_type))
4136 vlen = btf_vlen(cand_type);
4137 cand_m = btf_members(cand_type);
4138 canon_m = btf_members(canon_type);
4139 for (i = 0; i < vlen; i++) {
4140 eq = btf_dedup_is_equiv(d, cand_m->type, canon_m->type);
4150 case BTF_KIND_FUNC_PROTO: {
4151 const struct btf_param *cand_p, *canon_p;
4154 if (!btf_compat_fnproto(cand_type, canon_type))
4156 eq = btf_dedup_is_equiv(d, cand_type->type, canon_type->type);
4159 vlen = btf_vlen(cand_type);
4160 cand_p = btf_params(cand_type);
4161 canon_p = btf_params(canon_type);
4162 for (i = 0; i < vlen; i++) {
4163 eq = btf_dedup_is_equiv(d, cand_p->type, canon_p->type);
4179 * Use hypothetical mapping, produced by successful type graph equivalence
4180 * check, to augment existing struct/union canonical mapping, where possible.
4182 * If BTF_KIND_FWD resolution is allowed, this mapping is also used to record
4183 * FWD -> STRUCT/UNION correspondence as well. FWD resolution is bidirectional:
4184 * it doesn't matter if FWD type was part of canonical graph or candidate one,
4185 * we are recording the mapping anyway. As opposed to carefulness required
4186 * for struct/union correspondence mapping (described below), for FWD resolution
4187 * it's not important, as by the time that FWD type (reference type) will be
4188 * deduplicated all structs/unions will be deduped already anyway.
4190 * Recording STRUCT/UNION mapping is purely a performance optimization and is
4191 * not required for correctness. It needs to be done carefully to ensure that
4192 * struct/union from candidate's type graph is not mapped into corresponding
4193 * struct/union from canonical type graph that itself hasn't been resolved into
4194 * canonical representative. The only guarantee we have is that canonical
4195 * struct/union was determined as canonical and that won't change. But any
4196 * types referenced through that struct/union fields could have been not yet
4197 * resolved, so in case like that it's too early to establish any kind of
4198 * correspondence between structs/unions.
4200 * No canonical correspondence is derived for primitive types (they are already
4201 * deduplicated completely already anyway) or reference types (they rely on
4202 * stability of struct/union canonical relationship for equivalence checks).
4204 static void btf_dedup_merge_hypot_map(struct btf_dedup *d)
4206 __u32 canon_type_id, targ_type_id;
4207 __u16 t_kind, c_kind;
4211 for (i = 0; i < d->hypot_cnt; i++) {
4212 canon_type_id = d->hypot_list[i];
4213 targ_type_id = d->hypot_map[canon_type_id];
4214 t_id = resolve_type_id(d, targ_type_id);
4215 c_id = resolve_type_id(d, canon_type_id);
4216 t_kind = btf_kind(btf__type_by_id(d->btf, t_id));
4217 c_kind = btf_kind(btf__type_by_id(d->btf, c_id));
4219 * Resolve FWD into STRUCT/UNION.
4220 * It's ok to resolve FWD into STRUCT/UNION that's not yet
4221 * mapped to canonical representative (as opposed to
4222 * STRUCT/UNION <--> STRUCT/UNION mapping logic below), because
4223 * eventually that struct is going to be mapped and all resolved
4224 * FWDs will automatically resolve to correct canonical
4225 * representative. This will happen before ref type deduping,
4226 * which critically depends on stability of these mapping. This
4227 * stability is not a requirement for STRUCT/UNION equivalence
4231 /* if it's the split BTF case, we still need to point base FWD
4232 * to STRUCT/UNION in a split BTF, because FWDs from split BTF
4233 * will be resolved against base FWD. If we don't point base
4234 * canonical FWD to the resolved STRUCT/UNION, then all the
4235 * FWDs in split BTF won't be correctly resolved to a proper
4238 if (t_kind != BTF_KIND_FWD && c_kind == BTF_KIND_FWD)
4239 d->map[c_id] = t_id;
4241 /* if graph equivalence determined that we'd need to adjust
4242 * base canonical types, then we need to only point base FWDs
4243 * to STRUCTs/UNIONs and do no more modifications. For all
4244 * other purposes the type graphs were not equivalent.
4246 if (d->hypot_adjust_canon)
4249 if (t_kind == BTF_KIND_FWD && c_kind != BTF_KIND_FWD)
4250 d->map[t_id] = c_id;
4252 if ((t_kind == BTF_KIND_STRUCT || t_kind == BTF_KIND_UNION) &&
4253 c_kind != BTF_KIND_FWD &&
4254 is_type_mapped(d, c_id) &&
4255 !is_type_mapped(d, t_id)) {
4257 * as a perf optimization, we can map struct/union
4258 * that's part of type graph we just verified for
4259 * equivalence. We can do that for struct/union that has
4260 * canonical representative only, though.
4262 d->map[t_id] = c_id;
4268 * Deduplicate struct/union types.
4270 * For each struct/union type its type signature hash is calculated, taking
4271 * into account type's name, size, number, order and names of fields, but
4272 * ignoring type ID's referenced from fields, because they might not be deduped
4273 * completely until after reference types deduplication phase. This type hash
4274 * is used to iterate over all potential canonical types, sharing same hash.
4275 * For each canonical candidate we check whether type graphs that they form
4276 * (through referenced types in fields and so on) are equivalent using algorithm
4277 * implemented in `btf_dedup_is_equiv`. If such equivalence is found and
4278 * BTF_KIND_FWD resolution is allowed, then hypothetical mapping
4279 * (btf_dedup->hypot_map) produced by aforementioned type graph equivalence
4280 * algorithm is used to record FWD -> STRUCT/UNION mapping. It's also used to
4281 * potentially map other structs/unions to their canonical representatives,
4282 * if such relationship hasn't yet been established. This speeds up algorithm
4283 * by eliminating some of the duplicate work.
4285 * If no matching canonical representative was found, struct/union is marked
4286 * as canonical for itself and is added into btf_dedup->dedup_table hash map
4287 * for further look ups.
4289 static int btf_dedup_struct_type(struct btf_dedup *d, __u32 type_id)
4291 struct btf_type *cand_type, *t;
4292 struct hashmap_entry *hash_entry;
4293 /* if we don't find equivalent type, then we are canonical */
4294 __u32 new_id = type_id;
4298 /* already deduped or is in process of deduping (loop detected) */
4299 if (d->map[type_id] <= BTF_MAX_NR_TYPES)
4302 t = btf_type_by_id(d->btf, type_id);
4305 if (kind != BTF_KIND_STRUCT && kind != BTF_KIND_UNION)
4308 h = btf_hash_struct(t);
4309 for_each_dedup_cand(d, hash_entry, h) {
4310 __u32 cand_id = hash_entry->value;
4314 * Even though btf_dedup_is_equiv() checks for
4315 * btf_shallow_equal_struct() internally when checking two
4316 * structs (unions) for equivalence, we need to guard here
4317 * from picking matching FWD type as a dedup candidate.
4318 * This can happen due to hash collision. In such case just
4319 * relying on btf_dedup_is_equiv() would lead to potentially
4320 * creating a loop (FWD -> STRUCT and STRUCT -> FWD), because
4321 * FWD and compatible STRUCT/UNION are considered equivalent.
4323 cand_type = btf_type_by_id(d->btf, cand_id);
4324 if (!btf_shallow_equal_struct(t, cand_type))
4327 btf_dedup_clear_hypot_map(d);
4328 eq = btf_dedup_is_equiv(d, type_id, cand_id);
4333 btf_dedup_merge_hypot_map(d);
4334 if (d->hypot_adjust_canon) /* not really equivalent */
4340 d->map[type_id] = new_id;
4341 if (type_id == new_id && btf_dedup_table_add(d, h, type_id))
4347 static int btf_dedup_struct_types(struct btf_dedup *d)
4351 for (i = 0; i < d->btf->nr_types; i++) {
4352 err = btf_dedup_struct_type(d, d->btf->start_id + i);
4360 * Deduplicate reference type.
4362 * Once all primitive and struct/union types got deduplicated, we can easily
4363 * deduplicate all other (reference) BTF types. This is done in two steps:
4365 * 1. Resolve all referenced type IDs into their canonical type IDs. This
4366 * resolution can be done either immediately for primitive or struct/union types
4367 * (because they were deduped in previous two phases) or recursively for
4368 * reference types. Recursion will always terminate at either primitive or
4369 * struct/union type, at which point we can "unwind" chain of reference types
4370 * one by one. There is no danger of encountering cycles because in C type
4371 * system the only way to form type cycle is through struct/union, so any chain
4372 * of reference types, even those taking part in a type cycle, will inevitably
4373 * reach struct/union at some point.
4375 * 2. Once all referenced type IDs are resolved into canonical ones, BTF type
4376 * becomes "stable", in the sense that no further deduplication will cause
4377 * any changes to it. With that, it's now possible to calculate type's signature
4378 * hash (this time taking into account referenced type IDs) and loop over all
4379 * potential canonical representatives. If no match was found, current type
4380 * will become canonical representative of itself and will be added into
4381 * btf_dedup->dedup_table as another possible canonical representative.
4383 static int btf_dedup_ref_type(struct btf_dedup *d, __u32 type_id)
4385 struct hashmap_entry *hash_entry;
4386 __u32 new_id = type_id, cand_id;
4387 struct btf_type *t, *cand;
4388 /* if we don't find equivalent type, then we are representative type */
4392 if (d->map[type_id] == BTF_IN_PROGRESS_ID)
4394 if (d->map[type_id] <= BTF_MAX_NR_TYPES)
4395 return resolve_type_id(d, type_id);
4397 t = btf_type_by_id(d->btf, type_id);
4398 d->map[type_id] = BTF_IN_PROGRESS_ID;
4400 switch (btf_kind(t)) {
4401 case BTF_KIND_CONST:
4402 case BTF_KIND_VOLATILE:
4403 case BTF_KIND_RESTRICT:
4405 case BTF_KIND_TYPEDEF:
4407 case BTF_KIND_TYPE_TAG:
4408 ref_type_id = btf_dedup_ref_type(d, t->type);
4409 if (ref_type_id < 0)
4411 t->type = ref_type_id;
4413 h = btf_hash_common(t);
4414 for_each_dedup_cand(d, hash_entry, h) {
4415 cand_id = hash_entry->value;
4416 cand = btf_type_by_id(d->btf, cand_id);
4417 if (btf_equal_common(t, cand)) {
4424 case BTF_KIND_DECL_TAG:
4425 ref_type_id = btf_dedup_ref_type(d, t->type);
4426 if (ref_type_id < 0)
4428 t->type = ref_type_id;
4430 h = btf_hash_int_decl_tag(t);
4431 for_each_dedup_cand(d, hash_entry, h) {
4432 cand_id = hash_entry->value;
4433 cand = btf_type_by_id(d->btf, cand_id);
4434 if (btf_equal_int_tag(t, cand)) {
4441 case BTF_KIND_ARRAY: {
4442 struct btf_array *info = btf_array(t);
4444 ref_type_id = btf_dedup_ref_type(d, info->type);
4445 if (ref_type_id < 0)
4447 info->type = ref_type_id;
4449 ref_type_id = btf_dedup_ref_type(d, info->index_type);
4450 if (ref_type_id < 0)
4452 info->index_type = ref_type_id;
4454 h = btf_hash_array(t);
4455 for_each_dedup_cand(d, hash_entry, h) {
4456 cand_id = hash_entry->value;
4457 cand = btf_type_by_id(d->btf, cand_id);
4458 if (btf_equal_array(t, cand)) {
4466 case BTF_KIND_FUNC_PROTO: {
4467 struct btf_param *param;
4471 ref_type_id = btf_dedup_ref_type(d, t->type);
4472 if (ref_type_id < 0)
4474 t->type = ref_type_id;
4477 param = btf_params(t);
4478 for (i = 0; i < vlen; i++) {
4479 ref_type_id = btf_dedup_ref_type(d, param->type);
4480 if (ref_type_id < 0)
4482 param->type = ref_type_id;
4486 h = btf_hash_fnproto(t);
4487 for_each_dedup_cand(d, hash_entry, h) {
4488 cand_id = hash_entry->value;
4489 cand = btf_type_by_id(d->btf, cand_id);
4490 if (btf_equal_fnproto(t, cand)) {
4502 d->map[type_id] = new_id;
4503 if (type_id == new_id && btf_dedup_table_add(d, h, type_id))
4509 static int btf_dedup_ref_types(struct btf_dedup *d)
4513 for (i = 0; i < d->btf->nr_types; i++) {
4514 err = btf_dedup_ref_type(d, d->btf->start_id + i);
4518 /* we won't need d->dedup_table anymore */
4519 hashmap__free(d->dedup_table);
4520 d->dedup_table = NULL;
4525 * Collect a map from type names to type ids for all canonical structs
4526 * and unions. If the same name is shared by several canonical types
4527 * use a special value 0 to indicate this fact.
4529 static int btf_dedup_fill_unique_names_map(struct btf_dedup *d, struct hashmap *names_map)
4531 __u32 nr_types = btf__type_cnt(d->btf);
4538 * Iterate over base and split module ids in order to get all
4539 * available structs in the map.
4541 for (type_id = 1; type_id < nr_types; ++type_id) {
4542 t = btf_type_by_id(d->btf, type_id);
4545 if (kind != BTF_KIND_STRUCT && kind != BTF_KIND_UNION)
4548 /* Skip non-canonical types */
4549 if (type_id != d->map[type_id])
4552 err = hashmap__add(names_map, t->name_off, type_id);
4554 err = hashmap__set(names_map, t->name_off, 0, NULL, NULL);
4563 static int btf_dedup_resolve_fwd(struct btf_dedup *d, struct hashmap *names_map, __u32 type_id)
4565 struct btf_type *t = btf_type_by_id(d->btf, type_id);
4566 enum btf_fwd_kind fwd_kind = btf_kflag(t);
4567 __u16 cand_kind, kind = btf_kind(t);
4568 struct btf_type *cand_t;
4571 if (kind != BTF_KIND_FWD)
4574 /* Skip if this FWD already has a mapping */
4575 if (type_id != d->map[type_id])
4578 if (!hashmap__find(names_map, t->name_off, &cand_id))
4581 /* Zero is a special value indicating that name is not unique */
4585 cand_t = btf_type_by_id(d->btf, cand_id);
4586 cand_kind = btf_kind(cand_t);
4587 if ((cand_kind == BTF_KIND_STRUCT && fwd_kind != BTF_FWD_STRUCT) ||
4588 (cand_kind == BTF_KIND_UNION && fwd_kind != BTF_FWD_UNION))
4591 d->map[type_id] = cand_id;
4597 * Resolve unambiguous forward declarations.
4599 * The lion's share of all FWD declarations is resolved during
4600 * `btf_dedup_struct_types` phase when different type graphs are
4601 * compared against each other. However, if in some compilation unit a
4602 * FWD declaration is not a part of a type graph compared against
4603 * another type graph that declaration's canonical type would not be
4609 * struct foo *some_global;
4613 * struct foo { int u; };
4614 * struct foo *another_global;
4616 * After `btf_dedup_struct_types` the BTF looks as follows:
4618 * [1] STRUCT 'foo' size=4 vlen=1 ...
4619 * [2] INT 'int' size=4 ...
4620 * [3] PTR '(anon)' type_id=1
4621 * [4] FWD 'foo' fwd_kind=struct
4622 * [5] PTR '(anon)' type_id=4
4624 * This pass assumes that such FWD declarations should be mapped to
4625 * structs or unions with identical name in case if the name is not
4628 static int btf_dedup_resolve_fwds(struct btf_dedup *d)
4631 struct hashmap *names_map;
4633 names_map = hashmap__new(btf_dedup_identity_hash_fn, btf_dedup_equal_fn, NULL);
4634 if (IS_ERR(names_map))
4635 return PTR_ERR(names_map);
4637 err = btf_dedup_fill_unique_names_map(d, names_map);
4641 for (i = 0; i < d->btf->nr_types; i++) {
4642 err = btf_dedup_resolve_fwd(d, names_map, d->btf->start_id + i);
4648 hashmap__free(names_map);
4655 * After we established for each type its corresponding canonical representative
4656 * type, we now can eliminate types that are not canonical and leave only
4657 * canonical ones layed out sequentially in memory by copying them over
4658 * duplicates. During compaction btf_dedup->hypot_map array is reused to store
4659 * a map from original type ID to a new compacted type ID, which will be used
4660 * during next phase to "fix up" type IDs, referenced from struct/union and
4663 static int btf_dedup_compact_types(struct btf_dedup *d)
4666 __u32 next_type_id = d->btf->start_id;
4667 const struct btf_type *t;
4671 /* we are going to reuse hypot_map to store compaction remapping */
4672 d->hypot_map[0] = 0;
4673 /* base BTF types are not renumbered */
4674 for (id = 1; id < d->btf->start_id; id++)
4675 d->hypot_map[id] = id;
4676 for (i = 0, id = d->btf->start_id; i < d->btf->nr_types; i++, id++)
4677 d->hypot_map[id] = BTF_UNPROCESSED_ID;
4679 p = d->btf->types_data;
4681 for (i = 0, id = d->btf->start_id; i < d->btf->nr_types; i++, id++) {
4682 if (d->map[id] != id)
4685 t = btf__type_by_id(d->btf, id);
4686 len = btf_type_size(t);
4691 d->hypot_map[id] = next_type_id;
4692 d->btf->type_offs[next_type_id - d->btf->start_id] = p - d->btf->types_data;
4697 /* shrink struct btf's internal types index and update btf_header */
4698 d->btf->nr_types = next_type_id - d->btf->start_id;
4699 d->btf->type_offs_cap = d->btf->nr_types;
4700 d->btf->hdr->type_len = p - d->btf->types_data;
4701 new_offs = libbpf_reallocarray(d->btf->type_offs, d->btf->type_offs_cap,
4703 if (d->btf->type_offs_cap && !new_offs)
4705 d->btf->type_offs = new_offs;
4706 d->btf->hdr->str_off = d->btf->hdr->type_len;
4707 d->btf->raw_size = d->btf->hdr->hdr_len + d->btf->hdr->type_len + d->btf->hdr->str_len;
4712 * Figure out final (deduplicated and compacted) type ID for provided original
4713 * `type_id` by first resolving it into corresponding canonical type ID and
4714 * then mapping it to a deduplicated type ID, stored in btf_dedup->hypot_map,
4715 * which is populated during compaction phase.
4717 static int btf_dedup_remap_type_id(__u32 *type_id, void *ctx)
4719 struct btf_dedup *d = ctx;
4720 __u32 resolved_type_id, new_type_id;
4722 resolved_type_id = resolve_type_id(d, *type_id);
4723 new_type_id = d->hypot_map[resolved_type_id];
4724 if (new_type_id > BTF_MAX_NR_TYPES)
4727 *type_id = new_type_id;
4732 * Remap referenced type IDs into deduped type IDs.
4734 * After BTF types are deduplicated and compacted, their final type IDs may
4735 * differ from original ones. The map from original to a corresponding
4736 * deduped type ID is stored in btf_dedup->hypot_map and is populated during
4737 * compaction phase. During remapping phase we are rewriting all type IDs
4738 * referenced from any BTF type (e.g., struct fields, func proto args, etc) to
4739 * their final deduped type IDs.
4741 static int btf_dedup_remap_types(struct btf_dedup *d)
4745 for (i = 0; i < d->btf->nr_types; i++) {
4746 struct btf_type *t = btf_type_by_id(d->btf, d->btf->start_id + i);
4748 r = btf_type_visit_type_ids(t, btf_dedup_remap_type_id, d);
4756 r = btf_ext_visit_type_ids(d->btf_ext, btf_dedup_remap_type_id, d);
4764 * Probe few well-known locations for vmlinux kernel image and try to load BTF
4765 * data out of it to use for target BTF.
4767 struct btf *btf__load_vmlinux_btf(void)
4769 const char *locations[] = {
4770 /* try canonical vmlinux BTF through sysfs first */
4771 "/sys/kernel/btf/vmlinux",
4772 /* fall back to trying to find vmlinux on disk otherwise */
4773 "/boot/vmlinux-%1$s",
4774 "/lib/modules/%1$s/vmlinux-%1$s",
4775 "/lib/modules/%1$s/build/vmlinux",
4776 "/usr/lib/modules/%1$s/kernel/vmlinux",
4777 "/usr/lib/debug/boot/vmlinux-%1$s",
4778 "/usr/lib/debug/boot/vmlinux-%1$s.debug",
4779 "/usr/lib/debug/lib/modules/%1$s/vmlinux",
4781 char path[PATH_MAX + 1];
4788 for (i = 0; i < ARRAY_SIZE(locations); i++) {
4789 snprintf(path, PATH_MAX, locations[i], buf.release);
4791 if (faccessat(AT_FDCWD, path, R_OK, AT_EACCESS))
4794 btf = btf__parse(path, NULL);
4795 err = libbpf_get_error(btf);
4796 pr_debug("loading kernel BTF '%s': %d\n", path, err);
4803 pr_warn("failed to find valid kernel BTF\n");
4804 return libbpf_err_ptr(-ESRCH);
4807 struct btf *libbpf_find_kernel_btf(void) __attribute__((alias("btf__load_vmlinux_btf")));
4809 struct btf *btf__load_module_btf(const char *module_name, struct btf *vmlinux_btf)
4813 snprintf(path, sizeof(path), "/sys/kernel/btf/%s", module_name);
4814 return btf__parse_split(path, vmlinux_btf);
4817 int btf_type_visit_type_ids(struct btf_type *t, type_id_visit_fn visit, void *ctx)
4821 switch (btf_kind(t)) {
4823 case BTF_KIND_FLOAT:
4825 case BTF_KIND_ENUM64:
4829 case BTF_KIND_CONST:
4830 case BTF_KIND_VOLATILE:
4831 case BTF_KIND_RESTRICT:
4833 case BTF_KIND_TYPEDEF:
4836 case BTF_KIND_DECL_TAG:
4837 case BTF_KIND_TYPE_TAG:
4838 return visit(&t->type, ctx);
4840 case BTF_KIND_ARRAY: {
4841 struct btf_array *a = btf_array(t);
4843 err = visit(&a->type, ctx);
4844 err = err ?: visit(&a->index_type, ctx);
4848 case BTF_KIND_STRUCT:
4849 case BTF_KIND_UNION: {
4850 struct btf_member *m = btf_members(t);
4852 for (i = 0, n = btf_vlen(t); i < n; i++, m++) {
4853 err = visit(&m->type, ctx);
4860 case BTF_KIND_FUNC_PROTO: {
4861 struct btf_param *m = btf_params(t);
4863 err = visit(&t->type, ctx);
4866 for (i = 0, n = btf_vlen(t); i < n; i++, m++) {
4867 err = visit(&m->type, ctx);
4874 case BTF_KIND_DATASEC: {
4875 struct btf_var_secinfo *m = btf_var_secinfos(t);
4877 for (i = 0, n = btf_vlen(t); i < n; i++, m++) {
4878 err = visit(&m->type, ctx);
4890 int btf_type_visit_str_offs(struct btf_type *t, str_off_visit_fn visit, void *ctx)
4894 err = visit(&t->name_off, ctx);
4898 switch (btf_kind(t)) {
4899 case BTF_KIND_STRUCT:
4900 case BTF_KIND_UNION: {
4901 struct btf_member *m = btf_members(t);
4903 for (i = 0, n = btf_vlen(t); i < n; i++, m++) {
4904 err = visit(&m->name_off, ctx);
4910 case BTF_KIND_ENUM: {
4911 struct btf_enum *m = btf_enum(t);
4913 for (i = 0, n = btf_vlen(t); i < n; i++, m++) {
4914 err = visit(&m->name_off, ctx);
4920 case BTF_KIND_ENUM64: {
4921 struct btf_enum64 *m = btf_enum64(t);
4923 for (i = 0, n = btf_vlen(t); i < n; i++, m++) {
4924 err = visit(&m->name_off, ctx);
4930 case BTF_KIND_FUNC_PROTO: {
4931 struct btf_param *m = btf_params(t);
4933 for (i = 0, n = btf_vlen(t); i < n; i++, m++) {
4934 err = visit(&m->name_off, ctx);
4947 int btf_ext_visit_type_ids(struct btf_ext *btf_ext, type_id_visit_fn visit, void *ctx)
4949 const struct btf_ext_info *seg;
4950 struct btf_ext_info_sec *sec;
4953 seg = &btf_ext->func_info;
4954 for_each_btf_ext_sec(seg, sec) {
4955 struct bpf_func_info_min *rec;
4957 for_each_btf_ext_rec(seg, sec, i, rec) {
4958 err = visit(&rec->type_id, ctx);
4964 seg = &btf_ext->core_relo_info;
4965 for_each_btf_ext_sec(seg, sec) {
4966 struct bpf_core_relo *rec;
4968 for_each_btf_ext_rec(seg, sec, i, rec) {
4969 err = visit(&rec->type_id, ctx);
4978 int btf_ext_visit_str_offs(struct btf_ext *btf_ext, str_off_visit_fn visit, void *ctx)
4980 const struct btf_ext_info *seg;
4981 struct btf_ext_info_sec *sec;
4984 seg = &btf_ext->func_info;
4985 for_each_btf_ext_sec(seg, sec) {
4986 err = visit(&sec->sec_name_off, ctx);
4991 seg = &btf_ext->line_info;
4992 for_each_btf_ext_sec(seg, sec) {
4993 struct bpf_line_info_min *rec;
4995 err = visit(&sec->sec_name_off, ctx);
4999 for_each_btf_ext_rec(seg, sec, i, rec) {
5000 err = visit(&rec->file_name_off, ctx);
5003 err = visit(&rec->line_off, ctx);
5009 seg = &btf_ext->core_relo_info;
5010 for_each_btf_ext_sec(seg, sec) {
5011 struct bpf_core_relo *rec;
5013 err = visit(&sec->sec_name_off, ctx);
5017 for_each_btf_ext_rec(seg, sec, i, rec) {
5018 err = visit(&rec->access_str_off, ctx);