xref: /linux/tools/lib/bpf/btf_dump.c (revision a6cdeeb16bff89c8486324f53577db058cbe81ba)
1 // SPDX-License-Identifier: (LGPL-2.1 OR BSD-2-Clause)
2 
3 /*
4  * BTF-to-C type converter.
5  *
6  * Copyright (c) 2019 Facebook
7  */
8 
9 #include <stdbool.h>
10 #include <stddef.h>
11 #include <stdlib.h>
12 #include <string.h>
13 #include <errno.h>
14 #include <linux/err.h>
15 #include <linux/btf.h>
16 #include "btf.h"
17 #include "hashmap.h"
18 #include "libbpf.h"
19 #include "libbpf_internal.h"
20 
21 #define min(x, y) ((x) < (y) ? (x) : (y))
22 #define max(x, y) ((x) < (y) ? (y) : (x))
23 
24 static const char PREFIXES[] = "\t\t\t\t\t\t\t\t\t\t\t\t\t";
25 static const size_t PREFIX_CNT = sizeof(PREFIXES) - 1;
26 
27 static const char *pfx(int lvl)
28 {
29 	return lvl >= PREFIX_CNT ? PREFIXES : &PREFIXES[PREFIX_CNT - lvl];
30 }
31 
32 enum btf_dump_type_order_state {
33 	NOT_ORDERED,
34 	ORDERING,
35 	ORDERED,
36 };
37 
38 enum btf_dump_type_emit_state {
39 	NOT_EMITTED,
40 	EMITTING,
41 	EMITTED,
42 };
43 
44 /* per-type auxiliary state */
45 struct btf_dump_type_aux_state {
46 	/* topological sorting state */
47 	enum btf_dump_type_order_state order_state: 2;
48 	/* emitting state used to determine the need for forward declaration */
49 	enum btf_dump_type_emit_state emit_state: 2;
50 	/* whether forward declaration was already emitted */
51 	__u8 fwd_emitted: 1;
52 	/* whether unique non-duplicate name was already assigned */
53 	__u8 name_resolved: 1;
54 };
55 
56 struct btf_dump {
57 	const struct btf *btf;
58 	const struct btf_ext *btf_ext;
59 	btf_dump_printf_fn_t printf_fn;
60 	struct btf_dump_opts opts;
61 
62 	/* per-type auxiliary state */
63 	struct btf_dump_type_aux_state *type_states;
64 	/* per-type optional cached unique name, must be freed, if present */
65 	const char **cached_names;
66 
67 	/* topo-sorted list of dependent type definitions */
68 	__u32 *emit_queue;
69 	int emit_queue_cap;
70 	int emit_queue_cnt;
71 
72 	/*
73 	 * stack of type declarations (e.g., chain of modifiers, arrays,
74 	 * funcs, etc)
75 	 */
76 	__u32 *decl_stack;
77 	int decl_stack_cap;
78 	int decl_stack_cnt;
79 
80 	/* maps struct/union/enum name to a number of name occurrences */
81 	struct hashmap *type_names;
82 	/*
83 	 * maps typedef identifiers and enum value names to a number of such
84 	 * name occurrences
85 	 */
86 	struct hashmap *ident_names;
87 };
88 
89 static size_t str_hash_fn(const void *key, void *ctx)
90 {
91 	const char *s = key;
92 	size_t h = 0;
93 
94 	while (*s) {
95 		h = h * 31 + *s;
96 		s++;
97 	}
98 	return h;
99 }
100 
101 static bool str_equal_fn(const void *a, const void *b, void *ctx)
102 {
103 	return strcmp(a, b) == 0;
104 }
105 
106 static __u16 btf_kind_of(const struct btf_type *t)
107 {
108 	return BTF_INFO_KIND(t->info);
109 }
110 
111 static __u16 btf_vlen_of(const struct btf_type *t)
112 {
113 	return BTF_INFO_VLEN(t->info);
114 }
115 
116 static bool btf_kflag_of(const struct btf_type *t)
117 {
118 	return BTF_INFO_KFLAG(t->info);
119 }
120 
121 static const char *btf_name_of(const struct btf_dump *d, __u32 name_off)
122 {
123 	return btf__name_by_offset(d->btf, name_off);
124 }
125 
126 static void btf_dump_printf(const struct btf_dump *d, const char *fmt, ...)
127 {
128 	va_list args;
129 
130 	va_start(args, fmt);
131 	d->printf_fn(d->opts.ctx, fmt, args);
132 	va_end(args);
133 }
134 
135 struct btf_dump *btf_dump__new(const struct btf *btf,
136 			       const struct btf_ext *btf_ext,
137 			       const struct btf_dump_opts *opts,
138 			       btf_dump_printf_fn_t printf_fn)
139 {
140 	struct btf_dump *d;
141 	int err;
142 
143 	d = calloc(1, sizeof(struct btf_dump));
144 	if (!d)
145 		return ERR_PTR(-ENOMEM);
146 
147 	d->btf = btf;
148 	d->btf_ext = btf_ext;
149 	d->printf_fn = printf_fn;
150 	d->opts.ctx = opts ? opts->ctx : NULL;
151 
152 	d->type_names = hashmap__new(str_hash_fn, str_equal_fn, NULL);
153 	if (IS_ERR(d->type_names)) {
154 		err = PTR_ERR(d->type_names);
155 		d->type_names = NULL;
156 		btf_dump__free(d);
157 		return ERR_PTR(err);
158 	}
159 	d->ident_names = hashmap__new(str_hash_fn, str_equal_fn, NULL);
160 	if (IS_ERR(d->ident_names)) {
161 		err = PTR_ERR(d->ident_names);
162 		d->ident_names = NULL;
163 		btf_dump__free(d);
164 		return ERR_PTR(err);
165 	}
166 
167 	return d;
168 }
169 
170 void btf_dump__free(struct btf_dump *d)
171 {
172 	int i, cnt;
173 
174 	if (!d)
175 		return;
176 
177 	free(d->type_states);
178 	if (d->cached_names) {
179 		/* any set cached name is owned by us and should be freed */
180 		for (i = 0, cnt = btf__get_nr_types(d->btf); i <= cnt; i++) {
181 			if (d->cached_names[i])
182 				free((void *)d->cached_names[i]);
183 		}
184 	}
185 	free(d->cached_names);
186 	free(d->emit_queue);
187 	free(d->decl_stack);
188 	hashmap__free(d->type_names);
189 	hashmap__free(d->ident_names);
190 
191 	free(d);
192 }
193 
194 static int btf_dump_order_type(struct btf_dump *d, __u32 id, bool through_ptr);
195 static void btf_dump_emit_type(struct btf_dump *d, __u32 id, __u32 cont_id);
196 
197 /*
198  * Dump BTF type in a compilable C syntax, including all the necessary
199  * dependent types, necessary for compilation. If some of the dependent types
200  * were already emitted as part of previous btf_dump__dump_type() invocation
201  * for another type, they won't be emitted again. This API allows callers to
202  * filter out BTF types according to user-defined criterias and emitted only
203  * minimal subset of types, necessary to compile everything. Full struct/union
204  * definitions will still be emitted, even if the only usage is through
205  * pointer and could be satisfied with just a forward declaration.
206  *
207  * Dumping is done in two high-level passes:
208  *   1. Topologically sort type definitions to satisfy C rules of compilation.
209  *   2. Emit type definitions in C syntax.
210  *
211  * Returns 0 on success; <0, otherwise.
212  */
213 int btf_dump__dump_type(struct btf_dump *d, __u32 id)
214 {
215 	int err, i;
216 
217 	if (id > btf__get_nr_types(d->btf))
218 		return -EINVAL;
219 
220 	/* type states are lazily allocated, as they might not be needed */
221 	if (!d->type_states) {
222 		d->type_states = calloc(1 + btf__get_nr_types(d->btf),
223 					sizeof(d->type_states[0]));
224 		if (!d->type_states)
225 			return -ENOMEM;
226 		d->cached_names = calloc(1 + btf__get_nr_types(d->btf),
227 					 sizeof(d->cached_names[0]));
228 		if (!d->cached_names)
229 			return -ENOMEM;
230 
231 		/* VOID is special */
232 		d->type_states[0].order_state = ORDERED;
233 		d->type_states[0].emit_state = EMITTED;
234 	}
235 
236 	d->emit_queue_cnt = 0;
237 	err = btf_dump_order_type(d, id, false);
238 	if (err < 0)
239 		return err;
240 
241 	for (i = 0; i < d->emit_queue_cnt; i++)
242 		btf_dump_emit_type(d, d->emit_queue[i], 0 /*top-level*/);
243 
244 	return 0;
245 }
246 
247 static int btf_dump_add_emit_queue_id(struct btf_dump *d, __u32 id)
248 {
249 	__u32 *new_queue;
250 	size_t new_cap;
251 
252 	if (d->emit_queue_cnt >= d->emit_queue_cap) {
253 		new_cap = max(16, d->emit_queue_cap * 3 / 2);
254 		new_queue = realloc(d->emit_queue,
255 				    new_cap * sizeof(new_queue[0]));
256 		if (!new_queue)
257 			return -ENOMEM;
258 		d->emit_queue = new_queue;
259 		d->emit_queue_cap = new_cap;
260 	}
261 
262 	d->emit_queue[d->emit_queue_cnt++] = id;
263 	return 0;
264 }
265 
266 /*
267  * Determine order of emitting dependent types and specified type to satisfy
268  * C compilation rules.  This is done through topological sorting with an
269  * additional complication which comes from C rules. The main idea for C is
270  * that if some type is "embedded" into a struct/union, it's size needs to be
271  * known at the time of definition of containing type. E.g., for:
272  *
273  *	struct A {};
274  *	struct B { struct A x; }
275  *
276  * struct A *HAS* to be defined before struct B, because it's "embedded",
277  * i.e., it is part of struct B layout. But in the following case:
278  *
279  *	struct A;
280  *	struct B { struct A *x; }
281  *	struct A {};
282  *
283  * it's enough to just have a forward declaration of struct A at the time of
284  * struct B definition, as struct B has a pointer to struct A, so the size of
285  * field x is known without knowing struct A size: it's sizeof(void *).
286  *
287  * Unfortunately, there are some trickier cases we need to handle, e.g.:
288  *
289  *	struct A {}; // if this was forward-declaration: compilation error
290  *	struct B {
291  *		struct { // anonymous struct
292  *			struct A y;
293  *		} *x;
294  *	};
295  *
296  * In this case, struct B's field x is a pointer, so it's size is known
297  * regardless of the size of (anonymous) struct it points to. But because this
298  * struct is anonymous and thus defined inline inside struct B, *and* it
299  * embeds struct A, compiler requires full definition of struct A to be known
300  * before struct B can be defined. This creates a transitive dependency
301  * between struct A and struct B. If struct A was forward-declared before
302  * struct B definition and fully defined after struct B definition, that would
303  * trigger compilation error.
304  *
305  * All this means that while we are doing topological sorting on BTF type
306  * graph, we need to determine relationships between different types (graph
307  * nodes):
308  *   - weak link (relationship) between X and Y, if Y *CAN* be
309  *   forward-declared at the point of X definition;
310  *   - strong link, if Y *HAS* to be fully-defined before X can be defined.
311  *
312  * The rule is as follows. Given a chain of BTF types from X to Y, if there is
313  * BTF_KIND_PTR type in the chain and at least one non-anonymous type
314  * Z (excluding X, including Y), then link is weak. Otherwise, it's strong.
315  * Weak/strong relationship is determined recursively during DFS traversal and
316  * is returned as a result from btf_dump_order_type().
317  *
318  * btf_dump_order_type() is trying to avoid unnecessary forward declarations,
319  * but it is not guaranteeing that no extraneous forward declarations will be
320  * emitted.
321  *
322  * To avoid extra work, algorithm marks some of BTF types as ORDERED, when
323  * it's done with them, but not for all (e.g., VOLATILE, CONST, RESTRICT,
324  * ARRAY, FUNC_PROTO), as weak/strong semantics for those depends on the
325  * entire graph path, so depending where from one came to that BTF type, it
326  * might cause weak or strong ordering. For types like STRUCT/UNION/INT/ENUM,
327  * once they are processed, there is no need to do it again, so they are
328  * marked as ORDERED. We can mark PTR as ORDERED as well, as it semi-forces
329  * weak link, unless subsequent referenced STRUCT/UNION/ENUM is anonymous. But
330  * in any case, once those are processed, no need to do it again, as the
331  * result won't change.
332  *
333  * Returns:
334  *   - 1, if type is part of strong link (so there is strong topological
335  *   ordering requirements);
336  *   - 0, if type is part of weak link (so can be satisfied through forward
337  *   declaration);
338  *   - <0, on error (e.g., unsatisfiable type loop detected).
339  */
340 static int btf_dump_order_type(struct btf_dump *d, __u32 id, bool through_ptr)
341 {
342 	/*
343 	 * Order state is used to detect strong link cycles, but only for BTF
344 	 * kinds that are or could be an independent definition (i.e.,
345 	 * stand-alone fwd decl, enum, typedef, struct, union). Ptrs, arrays,
346 	 * func_protos, modifiers are just means to get to these definitions.
347 	 * Int/void don't need definitions, they are assumed to be always
348 	 * properly defined.  We also ignore datasec, var, and funcs for now.
349 	 * So for all non-defining kinds, we never even set ordering state,
350 	 * for defining kinds we set ORDERING and subsequently ORDERED if it
351 	 * forms a strong link.
352 	 */
353 	struct btf_dump_type_aux_state *tstate = &d->type_states[id];
354 	const struct btf_type *t;
355 	__u16 kind, vlen;
356 	int err, i;
357 
358 	/* return true, letting typedefs know that it's ok to be emitted */
359 	if (tstate->order_state == ORDERED)
360 		return 1;
361 
362 	t = btf__type_by_id(d->btf, id);
363 	kind = btf_kind_of(t);
364 
365 	if (tstate->order_state == ORDERING) {
366 		/* type loop, but resolvable through fwd declaration */
367 		if ((kind == BTF_KIND_STRUCT || kind == BTF_KIND_UNION) &&
368 		    through_ptr && t->name_off != 0)
369 			return 0;
370 		pr_warning("unsatisfiable type cycle, id:[%u]\n", id);
371 		return -ELOOP;
372 	}
373 
374 	switch (kind) {
375 	case BTF_KIND_INT:
376 		tstate->order_state = ORDERED;
377 		return 0;
378 
379 	case BTF_KIND_PTR:
380 		err = btf_dump_order_type(d, t->type, true);
381 		tstate->order_state = ORDERED;
382 		return err;
383 
384 	case BTF_KIND_ARRAY: {
385 		const struct btf_array *a = (void *)(t + 1);
386 
387 		return btf_dump_order_type(d, a->type, through_ptr);
388 	}
389 	case BTF_KIND_STRUCT:
390 	case BTF_KIND_UNION: {
391 		const struct btf_member *m = (void *)(t + 1);
392 		/*
393 		 * struct/union is part of strong link, only if it's embedded
394 		 * (so no ptr in a path) or it's anonymous (so has to be
395 		 * defined inline, even if declared through ptr)
396 		 */
397 		if (through_ptr && t->name_off != 0)
398 			return 0;
399 
400 		tstate->order_state = ORDERING;
401 
402 		vlen = btf_vlen_of(t);
403 		for (i = 0; i < vlen; i++, m++) {
404 			err = btf_dump_order_type(d, m->type, false);
405 			if (err < 0)
406 				return err;
407 		}
408 
409 		if (t->name_off != 0) {
410 			err = btf_dump_add_emit_queue_id(d, id);
411 			if (err < 0)
412 				return err;
413 		}
414 
415 		tstate->order_state = ORDERED;
416 		return 1;
417 	}
418 	case BTF_KIND_ENUM:
419 	case BTF_KIND_FWD:
420 		if (t->name_off != 0) {
421 			err = btf_dump_add_emit_queue_id(d, id);
422 			if (err)
423 				return err;
424 		}
425 		tstate->order_state = ORDERED;
426 		return 1;
427 
428 	case BTF_KIND_TYPEDEF: {
429 		int is_strong;
430 
431 		is_strong = btf_dump_order_type(d, t->type, through_ptr);
432 		if (is_strong < 0)
433 			return is_strong;
434 
435 		/* typedef is similar to struct/union w.r.t. fwd-decls */
436 		if (through_ptr && !is_strong)
437 			return 0;
438 
439 		/* typedef is always a named definition */
440 		err = btf_dump_add_emit_queue_id(d, id);
441 		if (err)
442 			return err;
443 
444 		d->type_states[id].order_state = ORDERED;
445 		return 1;
446 	}
447 	case BTF_KIND_VOLATILE:
448 	case BTF_KIND_CONST:
449 	case BTF_KIND_RESTRICT:
450 		return btf_dump_order_type(d, t->type, through_ptr);
451 
452 	case BTF_KIND_FUNC_PROTO: {
453 		const struct btf_param *p = (void *)(t + 1);
454 		bool is_strong;
455 
456 		err = btf_dump_order_type(d, t->type, through_ptr);
457 		if (err < 0)
458 			return err;
459 		is_strong = err > 0;
460 
461 		vlen = btf_vlen_of(t);
462 		for (i = 0; i < vlen; i++, p++) {
463 			err = btf_dump_order_type(d, p->type, through_ptr);
464 			if (err < 0)
465 				return err;
466 			if (err > 0)
467 				is_strong = true;
468 		}
469 		return is_strong;
470 	}
471 	case BTF_KIND_FUNC:
472 	case BTF_KIND_VAR:
473 	case BTF_KIND_DATASEC:
474 		d->type_states[id].order_state = ORDERED;
475 		return 0;
476 
477 	default:
478 		return -EINVAL;
479 	}
480 }
481 
482 static void btf_dump_emit_struct_fwd(struct btf_dump *d, __u32 id,
483 				     const struct btf_type *t);
484 static void btf_dump_emit_struct_def(struct btf_dump *d, __u32 id,
485 				     const struct btf_type *t, int lvl);
486 
487 static void btf_dump_emit_enum_fwd(struct btf_dump *d, __u32 id,
488 				   const struct btf_type *t);
489 static void btf_dump_emit_enum_def(struct btf_dump *d, __u32 id,
490 				   const struct btf_type *t, int lvl);
491 
492 static void btf_dump_emit_fwd_def(struct btf_dump *d, __u32 id,
493 				  const struct btf_type *t);
494 
495 static void btf_dump_emit_typedef_def(struct btf_dump *d, __u32 id,
496 				      const struct btf_type *t, int lvl);
497 
498 /* a local view into a shared stack */
499 struct id_stack {
500 	const __u32 *ids;
501 	int cnt;
502 };
503 
504 static void btf_dump_emit_type_decl(struct btf_dump *d, __u32 id,
505 				    const char *fname, int lvl);
506 static void btf_dump_emit_type_chain(struct btf_dump *d,
507 				     struct id_stack *decl_stack,
508 				     const char *fname, int lvl);
509 
510 static const char *btf_dump_type_name(struct btf_dump *d, __u32 id);
511 static const char *btf_dump_ident_name(struct btf_dump *d, __u32 id);
512 static size_t btf_dump_name_dups(struct btf_dump *d, struct hashmap *name_map,
513 				 const char *orig_name);
514 
515 static bool btf_dump_is_blacklisted(struct btf_dump *d, __u32 id)
516 {
517 	const struct btf_type *t = btf__type_by_id(d->btf, id);
518 
519 	/* __builtin_va_list is a compiler built-in, which causes compilation
520 	 * errors, when compiling w/ different compiler, then used to compile
521 	 * original code (e.g., GCC to compile kernel, Clang to use generated
522 	 * C header from BTF). As it is built-in, it should be already defined
523 	 * properly internally in compiler.
524 	 */
525 	if (t->name_off == 0)
526 		return false;
527 	return strcmp(btf_name_of(d, t->name_off), "__builtin_va_list") == 0;
528 }
529 
530 /*
531  * Emit C-syntax definitions of types from chains of BTF types.
532  *
533  * High-level handling of determining necessary forward declarations are handled
534  * by btf_dump_emit_type() itself, but all nitty-gritty details of emitting type
535  * declarations/definitions in C syntax  are handled by a combo of
536  * btf_dump_emit_type_decl()/btf_dump_emit_type_chain() w/ delegation to
537  * corresponding btf_dump_emit_*_{def,fwd}() functions.
538  *
539  * We also keep track of "containing struct/union type ID" to determine when
540  * we reference it from inside and thus can avoid emitting unnecessary forward
541  * declaration.
542  *
543  * This algorithm is designed in such a way, that even if some error occurs
544  * (either technical, e.g., out of memory, or logical, i.e., malformed BTF
545  * that doesn't comply to C rules completely), algorithm will try to proceed
546  * and produce as much meaningful output as possible.
547  */
548 static void btf_dump_emit_type(struct btf_dump *d, __u32 id, __u32 cont_id)
549 {
550 	struct btf_dump_type_aux_state *tstate = &d->type_states[id];
551 	bool top_level_def = cont_id == 0;
552 	const struct btf_type *t;
553 	__u16 kind;
554 
555 	if (tstate->emit_state == EMITTED)
556 		return;
557 
558 	t = btf__type_by_id(d->btf, id);
559 	kind = btf_kind_of(t);
560 
561 	if (top_level_def && t->name_off == 0) {
562 		pr_warning("unexpected nameless definition, id:[%u]\n", id);
563 		return;
564 	}
565 
566 	if (tstate->emit_state == EMITTING) {
567 		if (tstate->fwd_emitted)
568 			return;
569 
570 		switch (kind) {
571 		case BTF_KIND_STRUCT:
572 		case BTF_KIND_UNION:
573 			/*
574 			 * if we are referencing a struct/union that we are
575 			 * part of - then no need for fwd declaration
576 			 */
577 			if (id == cont_id)
578 				return;
579 			if (t->name_off == 0) {
580 				pr_warning("anonymous struct/union loop, id:[%u]\n",
581 					   id);
582 				return;
583 			}
584 			btf_dump_emit_struct_fwd(d, id, t);
585 			btf_dump_printf(d, ";\n\n");
586 			tstate->fwd_emitted = 1;
587 			break;
588 		case BTF_KIND_TYPEDEF:
589 			/*
590 			 * for typedef fwd_emitted means typedef definition
591 			 * was emitted, but it can be used only for "weak"
592 			 * references through pointer only, not for embedding
593 			 */
594 			if (!btf_dump_is_blacklisted(d, id)) {
595 				btf_dump_emit_typedef_def(d, id, t, 0);
596 				btf_dump_printf(d, ";\n\n");
597 			};
598 			tstate->fwd_emitted = 1;
599 			break;
600 		default:
601 			break;
602 		}
603 
604 		return;
605 	}
606 
607 	switch (kind) {
608 	case BTF_KIND_INT:
609 		tstate->emit_state = EMITTED;
610 		break;
611 	case BTF_KIND_ENUM:
612 		if (top_level_def) {
613 			btf_dump_emit_enum_def(d, id, t, 0);
614 			btf_dump_printf(d, ";\n\n");
615 		}
616 		tstate->emit_state = EMITTED;
617 		break;
618 	case BTF_KIND_PTR:
619 	case BTF_KIND_VOLATILE:
620 	case BTF_KIND_CONST:
621 	case BTF_KIND_RESTRICT:
622 		btf_dump_emit_type(d, t->type, cont_id);
623 		break;
624 	case BTF_KIND_ARRAY: {
625 		const struct btf_array *a = (void *)(t + 1);
626 
627 		btf_dump_emit_type(d, a->type, cont_id);
628 		break;
629 	}
630 	case BTF_KIND_FWD:
631 		btf_dump_emit_fwd_def(d, id, t);
632 		btf_dump_printf(d, ";\n\n");
633 		tstate->emit_state = EMITTED;
634 		break;
635 	case BTF_KIND_TYPEDEF:
636 		tstate->emit_state = EMITTING;
637 		btf_dump_emit_type(d, t->type, id);
638 		/*
639 		 * typedef can server as both definition and forward
640 		 * declaration; at this stage someone depends on
641 		 * typedef as a forward declaration (refers to it
642 		 * through pointer), so unless we already did it,
643 		 * emit typedef as a forward declaration
644 		 */
645 		if (!tstate->fwd_emitted && !btf_dump_is_blacklisted(d, id)) {
646 			btf_dump_emit_typedef_def(d, id, t, 0);
647 			btf_dump_printf(d, ";\n\n");
648 		}
649 		tstate->emit_state = EMITTED;
650 		break;
651 	case BTF_KIND_STRUCT:
652 	case BTF_KIND_UNION:
653 		tstate->emit_state = EMITTING;
654 		/* if it's a top-level struct/union definition or struct/union
655 		 * is anonymous, then in C we'll be emitting all fields and
656 		 * their types (as opposed to just `struct X`), so we need to
657 		 * make sure that all types, referenced from struct/union
658 		 * members have necessary forward-declarations, where
659 		 * applicable
660 		 */
661 		if (top_level_def || t->name_off == 0) {
662 			const struct btf_member *m = (void *)(t + 1);
663 			__u16 vlen = btf_vlen_of(t);
664 			int i, new_cont_id;
665 
666 			new_cont_id = t->name_off == 0 ? cont_id : id;
667 			for (i = 0; i < vlen; i++, m++)
668 				btf_dump_emit_type(d, m->type, new_cont_id);
669 		} else if (!tstate->fwd_emitted && id != cont_id) {
670 			btf_dump_emit_struct_fwd(d, id, t);
671 			btf_dump_printf(d, ";\n\n");
672 			tstate->fwd_emitted = 1;
673 		}
674 
675 		if (top_level_def) {
676 			btf_dump_emit_struct_def(d, id, t, 0);
677 			btf_dump_printf(d, ";\n\n");
678 			tstate->emit_state = EMITTED;
679 		} else {
680 			tstate->emit_state = NOT_EMITTED;
681 		}
682 		break;
683 	case BTF_KIND_FUNC_PROTO: {
684 		const struct btf_param *p = (void *)(t + 1);
685 		__u16 vlen = btf_vlen_of(t);
686 		int i;
687 
688 		btf_dump_emit_type(d, t->type, cont_id);
689 		for (i = 0; i < vlen; i++, p++)
690 			btf_dump_emit_type(d, p->type, cont_id);
691 
692 		break;
693 	}
694 	default:
695 		break;
696 	}
697 }
698 
699 static int btf_align_of(const struct btf *btf, __u32 id)
700 {
701 	const struct btf_type *t = btf__type_by_id(btf, id);
702 	__u16 kind = btf_kind_of(t);
703 
704 	switch (kind) {
705 	case BTF_KIND_INT:
706 	case BTF_KIND_ENUM:
707 		return min(sizeof(void *), t->size);
708 	case BTF_KIND_PTR:
709 		return sizeof(void *);
710 	case BTF_KIND_TYPEDEF:
711 	case BTF_KIND_VOLATILE:
712 	case BTF_KIND_CONST:
713 	case BTF_KIND_RESTRICT:
714 		return btf_align_of(btf, t->type);
715 	case BTF_KIND_ARRAY: {
716 		const struct btf_array *a = (void *)(t + 1);
717 
718 		return btf_align_of(btf, a->type);
719 	}
720 	case BTF_KIND_STRUCT:
721 	case BTF_KIND_UNION: {
722 		const struct btf_member *m = (void *)(t + 1);
723 		__u16 vlen = btf_vlen_of(t);
724 		int i, align = 1;
725 
726 		for (i = 0; i < vlen; i++, m++)
727 			align = max(align, btf_align_of(btf, m->type));
728 
729 		return align;
730 	}
731 	default:
732 		pr_warning("unsupported BTF_KIND:%u\n", btf_kind_of(t));
733 		return 1;
734 	}
735 }
736 
737 static bool btf_is_struct_packed(const struct btf *btf, __u32 id,
738 				 const struct btf_type *t)
739 {
740 	const struct btf_member *m;
741 	int align, i, bit_sz;
742 	__u16 vlen;
743 	bool kflag;
744 
745 	align = btf_align_of(btf, id);
746 	/* size of a non-packed struct has to be a multiple of its alignment*/
747 	if (t->size % align)
748 		return true;
749 
750 	m = (void *)(t + 1);
751 	kflag = btf_kflag_of(t);
752 	vlen = btf_vlen_of(t);
753 	/* all non-bitfield fields have to be naturally aligned */
754 	for (i = 0; i < vlen; i++, m++) {
755 		align = btf_align_of(btf, m->type);
756 		bit_sz = kflag ? BTF_MEMBER_BITFIELD_SIZE(m->offset) : 0;
757 		if (bit_sz == 0 && m->offset % (8 * align) != 0)
758 			return true;
759 	}
760 
761 	/*
762 	 * if original struct was marked as packed, but its layout is
763 	 * naturally aligned, we'll detect that it's not packed
764 	 */
765 	return false;
766 }
767 
768 static int chip_away_bits(int total, int at_most)
769 {
770 	return total % at_most ? : at_most;
771 }
772 
773 static void btf_dump_emit_bit_padding(const struct btf_dump *d,
774 				      int cur_off, int m_off, int m_bit_sz,
775 				      int align, int lvl)
776 {
777 	int off_diff = m_off - cur_off;
778 	int ptr_bits = sizeof(void *) * 8;
779 
780 	if (off_diff <= 0)
781 		/* no gap */
782 		return;
783 	if (m_bit_sz == 0 && off_diff < align * 8)
784 		/* natural padding will take care of a gap */
785 		return;
786 
787 	while (off_diff > 0) {
788 		const char *pad_type;
789 		int pad_bits;
790 
791 		if (ptr_bits > 32 && off_diff > 32) {
792 			pad_type = "long";
793 			pad_bits = chip_away_bits(off_diff, ptr_bits);
794 		} else if (off_diff > 16) {
795 			pad_type = "int";
796 			pad_bits = chip_away_bits(off_diff, 32);
797 		} else if (off_diff > 8) {
798 			pad_type = "short";
799 			pad_bits = chip_away_bits(off_diff, 16);
800 		} else {
801 			pad_type = "char";
802 			pad_bits = chip_away_bits(off_diff, 8);
803 		}
804 		btf_dump_printf(d, "\n%s%s: %d;", pfx(lvl), pad_type, pad_bits);
805 		off_diff -= pad_bits;
806 	}
807 }
808 
809 static void btf_dump_emit_struct_fwd(struct btf_dump *d, __u32 id,
810 				     const struct btf_type *t)
811 {
812 	btf_dump_printf(d, "%s %s",
813 			btf_kind_of(t) == BTF_KIND_STRUCT ? "struct" : "union",
814 			btf_dump_type_name(d, id));
815 }
816 
817 static void btf_dump_emit_struct_def(struct btf_dump *d,
818 				     __u32 id,
819 				     const struct btf_type *t,
820 				     int lvl)
821 {
822 	const struct btf_member *m = (void *)(t + 1);
823 	bool kflag = btf_kflag_of(t), is_struct;
824 	int align, i, packed, off = 0;
825 	__u16 vlen = btf_vlen_of(t);
826 
827 	is_struct = btf_kind_of(t) == BTF_KIND_STRUCT;
828 	packed = is_struct ? btf_is_struct_packed(d->btf, id, t) : 0;
829 	align = packed ? 1 : btf_align_of(d->btf, id);
830 
831 	btf_dump_printf(d, "%s%s%s {",
832 			is_struct ? "struct" : "union",
833 			t->name_off ? " " : "",
834 			btf_dump_type_name(d, id));
835 
836 	for (i = 0; i < vlen; i++, m++) {
837 		const char *fname;
838 		int m_off, m_sz;
839 
840 		fname = btf_name_of(d, m->name_off);
841 		m_sz = kflag ? BTF_MEMBER_BITFIELD_SIZE(m->offset) : 0;
842 		m_off = kflag ? BTF_MEMBER_BIT_OFFSET(m->offset) : m->offset;
843 		align = packed ? 1 : btf_align_of(d->btf, m->type);
844 
845 		btf_dump_emit_bit_padding(d, off, m_off, m_sz, align, lvl + 1);
846 		btf_dump_printf(d, "\n%s", pfx(lvl + 1));
847 		btf_dump_emit_type_decl(d, m->type, fname, lvl + 1);
848 
849 		if (m_sz) {
850 			btf_dump_printf(d, ": %d", m_sz);
851 			off = m_off + m_sz;
852 		} else {
853 			m_sz = max(0, btf__resolve_size(d->btf, m->type));
854 			off = m_off + m_sz * 8;
855 		}
856 		btf_dump_printf(d, ";");
857 	}
858 
859 	if (vlen)
860 		btf_dump_printf(d, "\n");
861 	btf_dump_printf(d, "%s}", pfx(lvl));
862 	if (packed)
863 		btf_dump_printf(d, " __attribute__((packed))");
864 }
865 
866 static void btf_dump_emit_enum_fwd(struct btf_dump *d, __u32 id,
867 				   const struct btf_type *t)
868 {
869 	btf_dump_printf(d, "enum %s", btf_dump_type_name(d, id));
870 }
871 
872 static void btf_dump_emit_enum_def(struct btf_dump *d, __u32 id,
873 				   const struct btf_type *t,
874 				   int lvl)
875 {
876 	const struct btf_enum *v = (void *)(t+1);
877 	__u16 vlen = btf_vlen_of(t);
878 	const char *name;
879 	size_t dup_cnt;
880 	int i;
881 
882 	btf_dump_printf(d, "enum%s%s",
883 			t->name_off ? " " : "",
884 			btf_dump_type_name(d, id));
885 
886 	if (vlen) {
887 		btf_dump_printf(d, " {");
888 		for (i = 0; i < vlen; i++, v++) {
889 			name = btf_name_of(d, v->name_off);
890 			/* enumerators share namespace with typedef idents */
891 			dup_cnt = btf_dump_name_dups(d, d->ident_names, name);
892 			if (dup_cnt > 1) {
893 				btf_dump_printf(d, "\n%s%s___%zu = %d,",
894 						pfx(lvl + 1), name, dup_cnt,
895 						(__s32)v->val);
896 			} else {
897 				btf_dump_printf(d, "\n%s%s = %d,",
898 						pfx(lvl + 1), name,
899 						(__s32)v->val);
900 			}
901 		}
902 		btf_dump_printf(d, "\n%s}", pfx(lvl));
903 	}
904 }
905 
906 static void btf_dump_emit_fwd_def(struct btf_dump *d, __u32 id,
907 				  const struct btf_type *t)
908 {
909 	const char *name = btf_dump_type_name(d, id);
910 
911 	if (btf_kflag_of(t))
912 		btf_dump_printf(d, "union %s", name);
913 	else
914 		btf_dump_printf(d, "struct %s", name);
915 }
916 
917 static void btf_dump_emit_typedef_def(struct btf_dump *d, __u32 id,
918 				     const struct btf_type *t, int lvl)
919 {
920 	const char *name = btf_dump_ident_name(d, id);
921 
922 	btf_dump_printf(d, "typedef ");
923 	btf_dump_emit_type_decl(d, t->type, name, lvl);
924 }
925 
926 static int btf_dump_push_decl_stack_id(struct btf_dump *d, __u32 id)
927 {
928 	__u32 *new_stack;
929 	size_t new_cap;
930 
931 	if (d->decl_stack_cnt >= d->decl_stack_cap) {
932 		new_cap = max(16, d->decl_stack_cap * 3 / 2);
933 		new_stack = realloc(d->decl_stack,
934 				    new_cap * sizeof(new_stack[0]));
935 		if (!new_stack)
936 			return -ENOMEM;
937 		d->decl_stack = new_stack;
938 		d->decl_stack_cap = new_cap;
939 	}
940 
941 	d->decl_stack[d->decl_stack_cnt++] = id;
942 
943 	return 0;
944 }
945 
946 /*
947  * Emit type declaration (e.g., field type declaration in a struct or argument
948  * declaration in function prototype) in correct C syntax.
949  *
950  * For most types it's trivial, but there are few quirky type declaration
951  * cases worth mentioning:
952  *   - function prototypes (especially nesting of function prototypes);
953  *   - arrays;
954  *   - const/volatile/restrict for pointers vs other types.
955  *
956  * For a good discussion of *PARSING* C syntax (as a human), see
957  * Peter van der Linden's "Expert C Programming: Deep C Secrets",
958  * Ch.3 "Unscrambling Declarations in C".
959  *
960  * It won't help with BTF to C conversion much, though, as it's an opposite
961  * problem. So we came up with this algorithm in reverse to van der Linden's
962  * parsing algorithm. It goes from structured BTF representation of type
963  * declaration to a valid compilable C syntax.
964  *
965  * For instance, consider this C typedef:
966  *	typedef const int * const * arr[10] arr_t;
967  * It will be represented in BTF with this chain of BTF types:
968  *	[typedef] -> [array] -> [ptr] -> [const] -> [ptr] -> [const] -> [int]
969  *
970  * Notice how [const] modifier always goes before type it modifies in BTF type
971  * graph, but in C syntax, const/volatile/restrict modifiers are written to
972  * the right of pointers, but to the left of other types. There are also other
973  * quirks, like function pointers, arrays of them, functions returning other
974  * functions, etc.
975  *
976  * We handle that by pushing all the types to a stack, until we hit "terminal"
977  * type (int/enum/struct/union/fwd). Then depending on the kind of a type on
978  * top of a stack, modifiers are handled differently. Array/function pointers
979  * have also wildly different syntax and how nesting of them are done. See
980  * code for authoritative definition.
981  *
982  * To avoid allocating new stack for each independent chain of BTF types, we
983  * share one bigger stack, with each chain working only on its own local view
984  * of a stack frame. Some care is required to "pop" stack frames after
985  * processing type declaration chain.
986  */
987 static void btf_dump_emit_type_decl(struct btf_dump *d, __u32 id,
988 				    const char *fname, int lvl)
989 {
990 	struct id_stack decl_stack;
991 	const struct btf_type *t;
992 	int err, stack_start;
993 	__u16 kind;
994 
995 	stack_start = d->decl_stack_cnt;
996 	for (;;) {
997 		err = btf_dump_push_decl_stack_id(d, id);
998 		if (err < 0) {
999 			/*
1000 			 * if we don't have enough memory for entire type decl
1001 			 * chain, restore stack, emit warning, and try to
1002 			 * proceed nevertheless
1003 			 */
1004 			pr_warning("not enough memory for decl stack:%d", err);
1005 			d->decl_stack_cnt = stack_start;
1006 			return;
1007 		}
1008 
1009 		/* VOID */
1010 		if (id == 0)
1011 			break;
1012 
1013 		t = btf__type_by_id(d->btf, id);
1014 		kind = btf_kind_of(t);
1015 		switch (kind) {
1016 		case BTF_KIND_PTR:
1017 		case BTF_KIND_VOLATILE:
1018 		case BTF_KIND_CONST:
1019 		case BTF_KIND_RESTRICT:
1020 		case BTF_KIND_FUNC_PROTO:
1021 			id = t->type;
1022 			break;
1023 		case BTF_KIND_ARRAY: {
1024 			const struct btf_array *a = (void *)(t + 1);
1025 
1026 			id = a->type;
1027 			break;
1028 		}
1029 		case BTF_KIND_INT:
1030 		case BTF_KIND_ENUM:
1031 		case BTF_KIND_FWD:
1032 		case BTF_KIND_STRUCT:
1033 		case BTF_KIND_UNION:
1034 		case BTF_KIND_TYPEDEF:
1035 			goto done;
1036 		default:
1037 			pr_warning("unexpected type in decl chain, kind:%u, id:[%u]\n",
1038 				   kind, id);
1039 			goto done;
1040 		}
1041 	}
1042 done:
1043 	/*
1044 	 * We might be inside a chain of declarations (e.g., array of function
1045 	 * pointers returning anonymous (so inlined) structs, having another
1046 	 * array field). Each of those needs its own "stack frame" to handle
1047 	 * emitting of declarations. Those stack frames are non-overlapping
1048 	 * portions of shared btf_dump->decl_stack. To make it a bit nicer to
1049 	 * handle this set of nested stacks, we create a view corresponding to
1050 	 * our own "stack frame" and work with it as an independent stack.
1051 	 * We'll need to clean up after emit_type_chain() returns, though.
1052 	 */
1053 	decl_stack.ids = d->decl_stack + stack_start;
1054 	decl_stack.cnt = d->decl_stack_cnt - stack_start;
1055 	btf_dump_emit_type_chain(d, &decl_stack, fname, lvl);
1056 	/*
1057 	 * emit_type_chain() guarantees that it will pop its entire decl_stack
1058 	 * frame before returning. But it works with a read-only view into
1059 	 * decl_stack, so it doesn't actually pop anything from the
1060 	 * perspective of shared btf_dump->decl_stack, per se. We need to
1061 	 * reset decl_stack state to how it was before us to avoid it growing
1062 	 * all the time.
1063 	 */
1064 	d->decl_stack_cnt = stack_start;
1065 }
1066 
1067 static void btf_dump_emit_mods(struct btf_dump *d, struct id_stack *decl_stack)
1068 {
1069 	const struct btf_type *t;
1070 	__u32 id;
1071 
1072 	while (decl_stack->cnt) {
1073 		id = decl_stack->ids[decl_stack->cnt - 1];
1074 		t = btf__type_by_id(d->btf, id);
1075 
1076 		switch (btf_kind_of(t)) {
1077 		case BTF_KIND_VOLATILE:
1078 			btf_dump_printf(d, "volatile ");
1079 			break;
1080 		case BTF_KIND_CONST:
1081 			btf_dump_printf(d, "const ");
1082 			break;
1083 		case BTF_KIND_RESTRICT:
1084 			btf_dump_printf(d, "restrict ");
1085 			break;
1086 		default:
1087 			return;
1088 		}
1089 		decl_stack->cnt--;
1090 	}
1091 }
1092 
1093 static bool btf_is_mod_kind(const struct btf *btf, __u32 id)
1094 {
1095 	const struct btf_type *t = btf__type_by_id(btf, id);
1096 
1097 	switch (btf_kind_of(t)) {
1098 	case BTF_KIND_VOLATILE:
1099 	case BTF_KIND_CONST:
1100 	case BTF_KIND_RESTRICT:
1101 		return true;
1102 	default:
1103 		return false;
1104 	}
1105 }
1106 
1107 static void btf_dump_emit_name(const struct btf_dump *d,
1108 			       const char *name, bool last_was_ptr)
1109 {
1110 	bool separate = name[0] && !last_was_ptr;
1111 
1112 	btf_dump_printf(d, "%s%s", separate ? " " : "", name);
1113 }
1114 
1115 static void btf_dump_emit_type_chain(struct btf_dump *d,
1116 				     struct id_stack *decls,
1117 				     const char *fname, int lvl)
1118 {
1119 	/*
1120 	 * last_was_ptr is used to determine if we need to separate pointer
1121 	 * asterisk (*) from previous part of type signature with space, so
1122 	 * that we get `int ***`, instead of `int * * *`. We default to true
1123 	 * for cases where we have single pointer in a chain. E.g., in ptr ->
1124 	 * func_proto case. func_proto will start a new emit_type_chain call
1125 	 * with just ptr, which should be emitted as (*) or (*<fname>), so we
1126 	 * don't want to prepend space for that last pointer.
1127 	 */
1128 	bool last_was_ptr = true;
1129 	const struct btf_type *t;
1130 	const char *name;
1131 	__u16 kind;
1132 	__u32 id;
1133 
1134 	while (decls->cnt) {
1135 		id = decls->ids[--decls->cnt];
1136 		if (id == 0) {
1137 			/* VOID is a special snowflake */
1138 			btf_dump_emit_mods(d, decls);
1139 			btf_dump_printf(d, "void");
1140 			last_was_ptr = false;
1141 			continue;
1142 		}
1143 
1144 		t = btf__type_by_id(d->btf, id);
1145 		kind = btf_kind_of(t);
1146 
1147 		switch (kind) {
1148 		case BTF_KIND_INT:
1149 			btf_dump_emit_mods(d, decls);
1150 			name = btf_name_of(d, t->name_off);
1151 			btf_dump_printf(d, "%s", name);
1152 			break;
1153 		case BTF_KIND_STRUCT:
1154 		case BTF_KIND_UNION:
1155 			btf_dump_emit_mods(d, decls);
1156 			/* inline anonymous struct/union */
1157 			if (t->name_off == 0)
1158 				btf_dump_emit_struct_def(d, id, t, lvl);
1159 			else
1160 				btf_dump_emit_struct_fwd(d, id, t);
1161 			break;
1162 		case BTF_KIND_ENUM:
1163 			btf_dump_emit_mods(d, decls);
1164 			/* inline anonymous enum */
1165 			if (t->name_off == 0)
1166 				btf_dump_emit_enum_def(d, id, t, lvl);
1167 			else
1168 				btf_dump_emit_enum_fwd(d, id, t);
1169 			break;
1170 		case BTF_KIND_FWD:
1171 			btf_dump_emit_mods(d, decls);
1172 			btf_dump_emit_fwd_def(d, id, t);
1173 			break;
1174 		case BTF_KIND_TYPEDEF:
1175 			btf_dump_emit_mods(d, decls);
1176 			btf_dump_printf(d, "%s", btf_dump_ident_name(d, id));
1177 			break;
1178 		case BTF_KIND_PTR:
1179 			btf_dump_printf(d, "%s", last_was_ptr ? "*" : " *");
1180 			break;
1181 		case BTF_KIND_VOLATILE:
1182 			btf_dump_printf(d, " volatile");
1183 			break;
1184 		case BTF_KIND_CONST:
1185 			btf_dump_printf(d, " const");
1186 			break;
1187 		case BTF_KIND_RESTRICT:
1188 			btf_dump_printf(d, " restrict");
1189 			break;
1190 		case BTF_KIND_ARRAY: {
1191 			const struct btf_array *a = (void *)(t + 1);
1192 			const struct btf_type *next_t;
1193 			__u32 next_id;
1194 			bool multidim;
1195 			/*
1196 			 * GCC has a bug
1197 			 * (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=8354)
1198 			 * which causes it to emit extra const/volatile
1199 			 * modifiers for an array, if array's element type has
1200 			 * const/volatile modifiers. Clang doesn't do that.
1201 			 * In general, it doesn't seem very meaningful to have
1202 			 * a const/volatile modifier for array, so we are
1203 			 * going to silently skip them here.
1204 			 */
1205 			while (decls->cnt) {
1206 				next_id = decls->ids[decls->cnt - 1];
1207 				if (btf_is_mod_kind(d->btf, next_id))
1208 					decls->cnt--;
1209 				else
1210 					break;
1211 			}
1212 
1213 			if (decls->cnt == 0) {
1214 				btf_dump_emit_name(d, fname, last_was_ptr);
1215 				btf_dump_printf(d, "[%u]", a->nelems);
1216 				return;
1217 			}
1218 
1219 			next_t = btf__type_by_id(d->btf, next_id);
1220 			multidim = btf_kind_of(next_t) == BTF_KIND_ARRAY;
1221 			/* we need space if we have named non-pointer */
1222 			if (fname[0] && !last_was_ptr)
1223 				btf_dump_printf(d, " ");
1224 			/* no parentheses for multi-dimensional array */
1225 			if (!multidim)
1226 				btf_dump_printf(d, "(");
1227 			btf_dump_emit_type_chain(d, decls, fname, lvl);
1228 			if (!multidim)
1229 				btf_dump_printf(d, ")");
1230 			btf_dump_printf(d, "[%u]", a->nelems);
1231 			return;
1232 		}
1233 		case BTF_KIND_FUNC_PROTO: {
1234 			const struct btf_param *p = (void *)(t + 1);
1235 			__u16 vlen = btf_vlen_of(t);
1236 			int i;
1237 
1238 			btf_dump_emit_mods(d, decls);
1239 			if (decls->cnt) {
1240 				btf_dump_printf(d, " (");
1241 				btf_dump_emit_type_chain(d, decls, fname, lvl);
1242 				btf_dump_printf(d, ")");
1243 			} else {
1244 				btf_dump_emit_name(d, fname, last_was_ptr);
1245 			}
1246 			btf_dump_printf(d, "(");
1247 			/*
1248 			 * Clang for BPF target generates func_proto with no
1249 			 * args as a func_proto with a single void arg (e.g.,
1250 			 * `int (*f)(void)` vs just `int (*f)()`). We are
1251 			 * going to pretend there are no args for such case.
1252 			 */
1253 			if (vlen == 1 && p->type == 0) {
1254 				btf_dump_printf(d, ")");
1255 				return;
1256 			}
1257 
1258 			for (i = 0; i < vlen; i++, p++) {
1259 				if (i > 0)
1260 					btf_dump_printf(d, ", ");
1261 
1262 				/* last arg of type void is vararg */
1263 				if (i == vlen - 1 && p->type == 0) {
1264 					btf_dump_printf(d, "...");
1265 					break;
1266 				}
1267 
1268 				name = btf_name_of(d, p->name_off);
1269 				btf_dump_emit_type_decl(d, p->type, name, lvl);
1270 			}
1271 
1272 			btf_dump_printf(d, ")");
1273 			return;
1274 		}
1275 		default:
1276 			pr_warning("unexpected type in decl chain, kind:%u, id:[%u]\n",
1277 				   kind, id);
1278 			return;
1279 		}
1280 
1281 		last_was_ptr = kind == BTF_KIND_PTR;
1282 	}
1283 
1284 	btf_dump_emit_name(d, fname, last_was_ptr);
1285 }
1286 
1287 /* return number of duplicates (occurrences) of a given name */
1288 static size_t btf_dump_name_dups(struct btf_dump *d, struct hashmap *name_map,
1289 				 const char *orig_name)
1290 {
1291 	size_t dup_cnt = 0;
1292 
1293 	hashmap__find(name_map, orig_name, (void **)&dup_cnt);
1294 	dup_cnt++;
1295 	hashmap__set(name_map, orig_name, (void *)dup_cnt, NULL, NULL);
1296 
1297 	return dup_cnt;
1298 }
1299 
1300 static const char *btf_dump_resolve_name(struct btf_dump *d, __u32 id,
1301 					 struct hashmap *name_map)
1302 {
1303 	struct btf_dump_type_aux_state *s = &d->type_states[id];
1304 	const struct btf_type *t = btf__type_by_id(d->btf, id);
1305 	const char *orig_name = btf_name_of(d, t->name_off);
1306 	const char **cached_name = &d->cached_names[id];
1307 	size_t dup_cnt;
1308 
1309 	if (t->name_off == 0)
1310 		return "";
1311 
1312 	if (s->name_resolved)
1313 		return *cached_name ? *cached_name : orig_name;
1314 
1315 	dup_cnt = btf_dump_name_dups(d, name_map, orig_name);
1316 	if (dup_cnt > 1) {
1317 		const size_t max_len = 256;
1318 		char new_name[max_len];
1319 
1320 		snprintf(new_name, max_len, "%s___%zu", orig_name, dup_cnt);
1321 		*cached_name = strdup(new_name);
1322 	}
1323 
1324 	s->name_resolved = 1;
1325 	return *cached_name ? *cached_name : orig_name;
1326 }
1327 
1328 static const char *btf_dump_type_name(struct btf_dump *d, __u32 id)
1329 {
1330 	return btf_dump_resolve_name(d, id, d->type_names);
1331 }
1332 
1333 static const char *btf_dump_ident_name(struct btf_dump *d, __u32 id)
1334 {
1335 	return btf_dump_resolve_name(d, id, d->ident_names);
1336 }
1337