xref: /openbmc/linux/tools/lib/bpf/btf_dump.c (revision 0e17c50f)
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 /* make sure libbpf doesn't use kernel-only integer typedefs */
22 #pragma GCC poison u8 u16 u32 u64 s8 s16 s32 s64
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 	/* whether type is referenced from any other type */
55 	__u8 referenced: 1;
56 };
57 
58 struct btf_dump {
59 	const struct btf *btf;
60 	const struct btf_ext *btf_ext;
61 	btf_dump_printf_fn_t printf_fn;
62 	struct btf_dump_opts opts;
63 
64 	/* per-type auxiliary state */
65 	struct btf_dump_type_aux_state *type_states;
66 	/* per-type optional cached unique name, must be freed, if present */
67 	const char **cached_names;
68 
69 	/* topo-sorted list of dependent type definitions */
70 	__u32 *emit_queue;
71 	int emit_queue_cap;
72 	int emit_queue_cnt;
73 
74 	/*
75 	 * stack of type declarations (e.g., chain of modifiers, arrays,
76 	 * funcs, etc)
77 	 */
78 	__u32 *decl_stack;
79 	int decl_stack_cap;
80 	int decl_stack_cnt;
81 
82 	/* maps struct/union/enum name to a number of name occurrences */
83 	struct hashmap *type_names;
84 	/*
85 	 * maps typedef identifiers and enum value names to a number of such
86 	 * name occurrences
87 	 */
88 	struct hashmap *ident_names;
89 };
90 
91 static size_t str_hash_fn(const void *key, void *ctx)
92 {
93 	const char *s = key;
94 	size_t h = 0;
95 
96 	while (*s) {
97 		h = h * 31 + *s;
98 		s++;
99 	}
100 	return h;
101 }
102 
103 static bool str_equal_fn(const void *a, const void *b, void *ctx)
104 {
105 	return strcmp(a, b) == 0;
106 }
107 
108 static const char *btf_name_of(const struct btf_dump *d, __u32 name_off)
109 {
110 	return btf__name_by_offset(d->btf, name_off);
111 }
112 
113 static void btf_dump_printf(const struct btf_dump *d, const char *fmt, ...)
114 {
115 	va_list args;
116 
117 	va_start(args, fmt);
118 	d->printf_fn(d->opts.ctx, fmt, args);
119 	va_end(args);
120 }
121 
122 static int btf_dump_mark_referenced(struct btf_dump *d);
123 
124 struct btf_dump *btf_dump__new(const struct btf *btf,
125 			       const struct btf_ext *btf_ext,
126 			       const struct btf_dump_opts *opts,
127 			       btf_dump_printf_fn_t printf_fn)
128 {
129 	struct btf_dump *d;
130 	int err;
131 
132 	d = calloc(1, sizeof(struct btf_dump));
133 	if (!d)
134 		return ERR_PTR(-ENOMEM);
135 
136 	d->btf = btf;
137 	d->btf_ext = btf_ext;
138 	d->printf_fn = printf_fn;
139 	d->opts.ctx = opts ? opts->ctx : NULL;
140 
141 	d->type_names = hashmap__new(str_hash_fn, str_equal_fn, NULL);
142 	if (IS_ERR(d->type_names)) {
143 		err = PTR_ERR(d->type_names);
144 		d->type_names = NULL;
145 		goto err;
146 	}
147 	d->ident_names = hashmap__new(str_hash_fn, str_equal_fn, NULL);
148 	if (IS_ERR(d->ident_names)) {
149 		err = PTR_ERR(d->ident_names);
150 		d->ident_names = NULL;
151 		goto err;
152 	}
153 	d->type_states = calloc(1 + btf__get_nr_types(d->btf),
154 				sizeof(d->type_states[0]));
155 	if (!d->type_states) {
156 		err = -ENOMEM;
157 		goto err;
158 	}
159 	d->cached_names = calloc(1 + btf__get_nr_types(d->btf),
160 				 sizeof(d->cached_names[0]));
161 	if (!d->cached_names) {
162 		err = -ENOMEM;
163 		goto err;
164 	}
165 
166 	/* VOID is special */
167 	d->type_states[0].order_state = ORDERED;
168 	d->type_states[0].emit_state = EMITTED;
169 
170 	/* eagerly determine referenced types for anon enums */
171 	err = btf_dump_mark_referenced(d);
172 	if (err)
173 		goto err;
174 
175 	return d;
176 err:
177 	btf_dump__free(d);
178 	return ERR_PTR(err);
179 }
180 
181 void btf_dump__free(struct btf_dump *d)
182 {
183 	int i, cnt;
184 
185 	if (!d)
186 		return;
187 
188 	free(d->type_states);
189 	if (d->cached_names) {
190 		/* any set cached name is owned by us and should be freed */
191 		for (i = 0, cnt = btf__get_nr_types(d->btf); i <= cnt; i++) {
192 			if (d->cached_names[i])
193 				free((void *)d->cached_names[i]);
194 		}
195 	}
196 	free(d->cached_names);
197 	free(d->emit_queue);
198 	free(d->decl_stack);
199 	hashmap__free(d->type_names);
200 	hashmap__free(d->ident_names);
201 
202 	free(d);
203 }
204 
205 static int btf_dump_order_type(struct btf_dump *d, __u32 id, bool through_ptr);
206 static void btf_dump_emit_type(struct btf_dump *d, __u32 id, __u32 cont_id);
207 
208 /*
209  * Dump BTF type in a compilable C syntax, including all the necessary
210  * dependent types, necessary for compilation. If some of the dependent types
211  * were already emitted as part of previous btf_dump__dump_type() invocation
212  * for another type, they won't be emitted again. This API allows callers to
213  * filter out BTF types according to user-defined criterias and emitted only
214  * minimal subset of types, necessary to compile everything. Full struct/union
215  * definitions will still be emitted, even if the only usage is through
216  * pointer and could be satisfied with just a forward declaration.
217  *
218  * Dumping is done in two high-level passes:
219  *   1. Topologically sort type definitions to satisfy C rules of compilation.
220  *   2. Emit type definitions in C syntax.
221  *
222  * Returns 0 on success; <0, otherwise.
223  */
224 int btf_dump__dump_type(struct btf_dump *d, __u32 id)
225 {
226 	int err, i;
227 
228 	if (id > btf__get_nr_types(d->btf))
229 		return -EINVAL;
230 
231 	d->emit_queue_cnt = 0;
232 	err = btf_dump_order_type(d, id, false);
233 	if (err < 0)
234 		return err;
235 
236 	for (i = 0; i < d->emit_queue_cnt; i++)
237 		btf_dump_emit_type(d, d->emit_queue[i], 0 /*top-level*/);
238 
239 	return 0;
240 }
241 
242 /*
243  * Mark all types that are referenced from any other type. This is used to
244  * determine top-level anonymous enums that need to be emitted as an
245  * independent type declarations.
246  * Anonymous enums come in two flavors: either embedded in a struct's field
247  * definition, in which case they have to be declared inline as part of field
248  * type declaration; or as a top-level anonymous enum, typically used for
249  * declaring global constants. It's impossible to distinguish between two
250  * without knowning whether given enum type was referenced from other type:
251  * top-level anonymous enum won't be referenced by anything, while embedded
252  * one will.
253  */
254 static int btf_dump_mark_referenced(struct btf_dump *d)
255 {
256 	int i, j, n = btf__get_nr_types(d->btf);
257 	const struct btf_type *t;
258 	__u16 vlen;
259 
260 	for (i = 1; i <= n; i++) {
261 		t = btf__type_by_id(d->btf, i);
262 		vlen = btf_vlen(t);
263 
264 		switch (btf_kind(t)) {
265 		case BTF_KIND_INT:
266 		case BTF_KIND_ENUM:
267 		case BTF_KIND_FWD:
268 			break;
269 
270 		case BTF_KIND_VOLATILE:
271 		case BTF_KIND_CONST:
272 		case BTF_KIND_RESTRICT:
273 		case BTF_KIND_PTR:
274 		case BTF_KIND_TYPEDEF:
275 		case BTF_KIND_FUNC:
276 		case BTF_KIND_VAR:
277 			d->type_states[t->type].referenced = 1;
278 			break;
279 
280 		case BTF_KIND_ARRAY: {
281 			const struct btf_array *a = btf_array(t);
282 
283 			d->type_states[a->index_type].referenced = 1;
284 			d->type_states[a->type].referenced = 1;
285 			break;
286 		}
287 		case BTF_KIND_STRUCT:
288 		case BTF_KIND_UNION: {
289 			const struct btf_member *m = btf_members(t);
290 
291 			for (j = 0; j < vlen; j++, m++)
292 				d->type_states[m->type].referenced = 1;
293 			break;
294 		}
295 		case BTF_KIND_FUNC_PROTO: {
296 			const struct btf_param *p = btf_params(t);
297 
298 			for (j = 0; j < vlen; j++, p++)
299 				d->type_states[p->type].referenced = 1;
300 			break;
301 		}
302 		case BTF_KIND_DATASEC: {
303 			const struct btf_var_secinfo *v = btf_var_secinfos(t);
304 
305 			for (j = 0; j < vlen; j++, v++)
306 				d->type_states[v->type].referenced = 1;
307 			break;
308 		}
309 		default:
310 			return -EINVAL;
311 		}
312 	}
313 	return 0;
314 }
315 static int btf_dump_add_emit_queue_id(struct btf_dump *d, __u32 id)
316 {
317 	__u32 *new_queue;
318 	size_t new_cap;
319 
320 	if (d->emit_queue_cnt >= d->emit_queue_cap) {
321 		new_cap = max(16, d->emit_queue_cap * 3 / 2);
322 		new_queue = realloc(d->emit_queue,
323 				    new_cap * sizeof(new_queue[0]));
324 		if (!new_queue)
325 			return -ENOMEM;
326 		d->emit_queue = new_queue;
327 		d->emit_queue_cap = new_cap;
328 	}
329 
330 	d->emit_queue[d->emit_queue_cnt++] = id;
331 	return 0;
332 }
333 
334 /*
335  * Determine order of emitting dependent types and specified type to satisfy
336  * C compilation rules.  This is done through topological sorting with an
337  * additional complication which comes from C rules. The main idea for C is
338  * that if some type is "embedded" into a struct/union, it's size needs to be
339  * known at the time of definition of containing type. E.g., for:
340  *
341  *	struct A {};
342  *	struct B { struct A x; }
343  *
344  * struct A *HAS* to be defined before struct B, because it's "embedded",
345  * i.e., it is part of struct B layout. But in the following case:
346  *
347  *	struct A;
348  *	struct B { struct A *x; }
349  *	struct A {};
350  *
351  * it's enough to just have a forward declaration of struct A at the time of
352  * struct B definition, as struct B has a pointer to struct A, so the size of
353  * field x is known without knowing struct A size: it's sizeof(void *).
354  *
355  * Unfortunately, there are some trickier cases we need to handle, e.g.:
356  *
357  *	struct A {}; // if this was forward-declaration: compilation error
358  *	struct B {
359  *		struct { // anonymous struct
360  *			struct A y;
361  *		} *x;
362  *	};
363  *
364  * In this case, struct B's field x is a pointer, so it's size is known
365  * regardless of the size of (anonymous) struct it points to. But because this
366  * struct is anonymous and thus defined inline inside struct B, *and* it
367  * embeds struct A, compiler requires full definition of struct A to be known
368  * before struct B can be defined. This creates a transitive dependency
369  * between struct A and struct B. If struct A was forward-declared before
370  * struct B definition and fully defined after struct B definition, that would
371  * trigger compilation error.
372  *
373  * All this means that while we are doing topological sorting on BTF type
374  * graph, we need to determine relationships between different types (graph
375  * nodes):
376  *   - weak link (relationship) between X and Y, if Y *CAN* be
377  *   forward-declared at the point of X definition;
378  *   - strong link, if Y *HAS* to be fully-defined before X can be defined.
379  *
380  * The rule is as follows. Given a chain of BTF types from X to Y, if there is
381  * BTF_KIND_PTR type in the chain and at least one non-anonymous type
382  * Z (excluding X, including Y), then link is weak. Otherwise, it's strong.
383  * Weak/strong relationship is determined recursively during DFS traversal and
384  * is returned as a result from btf_dump_order_type().
385  *
386  * btf_dump_order_type() is trying to avoid unnecessary forward declarations,
387  * but it is not guaranteeing that no extraneous forward declarations will be
388  * emitted.
389  *
390  * To avoid extra work, algorithm marks some of BTF types as ORDERED, when
391  * it's done with them, but not for all (e.g., VOLATILE, CONST, RESTRICT,
392  * ARRAY, FUNC_PROTO), as weak/strong semantics for those depends on the
393  * entire graph path, so depending where from one came to that BTF type, it
394  * might cause weak or strong ordering. For types like STRUCT/UNION/INT/ENUM,
395  * once they are processed, there is no need to do it again, so they are
396  * marked as ORDERED. We can mark PTR as ORDERED as well, as it semi-forces
397  * weak link, unless subsequent referenced STRUCT/UNION/ENUM is anonymous. But
398  * in any case, once those are processed, no need to do it again, as the
399  * result won't change.
400  *
401  * Returns:
402  *   - 1, if type is part of strong link (so there is strong topological
403  *   ordering requirements);
404  *   - 0, if type is part of weak link (so can be satisfied through forward
405  *   declaration);
406  *   - <0, on error (e.g., unsatisfiable type loop detected).
407  */
408 static int btf_dump_order_type(struct btf_dump *d, __u32 id, bool through_ptr)
409 {
410 	/*
411 	 * Order state is used to detect strong link cycles, but only for BTF
412 	 * kinds that are or could be an independent definition (i.e.,
413 	 * stand-alone fwd decl, enum, typedef, struct, union). Ptrs, arrays,
414 	 * func_protos, modifiers are just means to get to these definitions.
415 	 * Int/void don't need definitions, they are assumed to be always
416 	 * properly defined.  We also ignore datasec, var, and funcs for now.
417 	 * So for all non-defining kinds, we never even set ordering state,
418 	 * for defining kinds we set ORDERING and subsequently ORDERED if it
419 	 * forms a strong link.
420 	 */
421 	struct btf_dump_type_aux_state *tstate = &d->type_states[id];
422 	const struct btf_type *t;
423 	__u16 vlen;
424 	int err, i;
425 
426 	/* return true, letting typedefs know that it's ok to be emitted */
427 	if (tstate->order_state == ORDERED)
428 		return 1;
429 
430 	t = btf__type_by_id(d->btf, id);
431 
432 	if (tstate->order_state == ORDERING) {
433 		/* type loop, but resolvable through fwd declaration */
434 		if (btf_is_composite(t) && through_ptr && t->name_off != 0)
435 			return 0;
436 		pr_warn("unsatisfiable type cycle, id:[%u]\n", id);
437 		return -ELOOP;
438 	}
439 
440 	switch (btf_kind(t)) {
441 	case BTF_KIND_INT:
442 		tstate->order_state = ORDERED;
443 		return 0;
444 
445 	case BTF_KIND_PTR:
446 		err = btf_dump_order_type(d, t->type, true);
447 		tstate->order_state = ORDERED;
448 		return err;
449 
450 	case BTF_KIND_ARRAY:
451 		return btf_dump_order_type(d, btf_array(t)->type, through_ptr);
452 
453 	case BTF_KIND_STRUCT:
454 	case BTF_KIND_UNION: {
455 		const struct btf_member *m = btf_members(t);
456 		/*
457 		 * struct/union is part of strong link, only if it's embedded
458 		 * (so no ptr in a path) or it's anonymous (so has to be
459 		 * defined inline, even if declared through ptr)
460 		 */
461 		if (through_ptr && t->name_off != 0)
462 			return 0;
463 
464 		tstate->order_state = ORDERING;
465 
466 		vlen = btf_vlen(t);
467 		for (i = 0; i < vlen; i++, m++) {
468 			err = btf_dump_order_type(d, m->type, false);
469 			if (err < 0)
470 				return err;
471 		}
472 
473 		if (t->name_off != 0) {
474 			err = btf_dump_add_emit_queue_id(d, id);
475 			if (err < 0)
476 				return err;
477 		}
478 
479 		tstate->order_state = ORDERED;
480 		return 1;
481 	}
482 	case BTF_KIND_ENUM:
483 	case BTF_KIND_FWD:
484 		/*
485 		 * non-anonymous or non-referenced enums are top-level
486 		 * declarations and should be emitted. Same logic can be
487 		 * applied to FWDs, it won't hurt anyways.
488 		 */
489 		if (t->name_off != 0 || !tstate->referenced) {
490 			err = btf_dump_add_emit_queue_id(d, id);
491 			if (err)
492 				return err;
493 		}
494 		tstate->order_state = ORDERED;
495 		return 1;
496 
497 	case BTF_KIND_TYPEDEF: {
498 		int is_strong;
499 
500 		is_strong = btf_dump_order_type(d, t->type, through_ptr);
501 		if (is_strong < 0)
502 			return is_strong;
503 
504 		/* typedef is similar to struct/union w.r.t. fwd-decls */
505 		if (through_ptr && !is_strong)
506 			return 0;
507 
508 		/* typedef is always a named definition */
509 		err = btf_dump_add_emit_queue_id(d, id);
510 		if (err)
511 			return err;
512 
513 		d->type_states[id].order_state = ORDERED;
514 		return 1;
515 	}
516 	case BTF_KIND_VOLATILE:
517 	case BTF_KIND_CONST:
518 	case BTF_KIND_RESTRICT:
519 		return btf_dump_order_type(d, t->type, through_ptr);
520 
521 	case BTF_KIND_FUNC_PROTO: {
522 		const struct btf_param *p = btf_params(t);
523 		bool is_strong;
524 
525 		err = btf_dump_order_type(d, t->type, through_ptr);
526 		if (err < 0)
527 			return err;
528 		is_strong = err > 0;
529 
530 		vlen = btf_vlen(t);
531 		for (i = 0; i < vlen; i++, p++) {
532 			err = btf_dump_order_type(d, p->type, through_ptr);
533 			if (err < 0)
534 				return err;
535 			if (err > 0)
536 				is_strong = true;
537 		}
538 		return is_strong;
539 	}
540 	case BTF_KIND_FUNC:
541 	case BTF_KIND_VAR:
542 	case BTF_KIND_DATASEC:
543 		d->type_states[id].order_state = ORDERED;
544 		return 0;
545 
546 	default:
547 		return -EINVAL;
548 	}
549 }
550 
551 static void btf_dump_emit_struct_fwd(struct btf_dump *d, __u32 id,
552 				     const struct btf_type *t);
553 static void btf_dump_emit_struct_def(struct btf_dump *d, __u32 id,
554 				     const struct btf_type *t, int lvl);
555 
556 static void btf_dump_emit_enum_fwd(struct btf_dump *d, __u32 id,
557 				   const struct btf_type *t);
558 static void btf_dump_emit_enum_def(struct btf_dump *d, __u32 id,
559 				   const struct btf_type *t, int lvl);
560 
561 static void btf_dump_emit_fwd_def(struct btf_dump *d, __u32 id,
562 				  const struct btf_type *t);
563 
564 static void btf_dump_emit_typedef_def(struct btf_dump *d, __u32 id,
565 				      const struct btf_type *t, int lvl);
566 
567 /* a local view into a shared stack */
568 struct id_stack {
569 	const __u32 *ids;
570 	int cnt;
571 };
572 
573 static void btf_dump_emit_type_decl(struct btf_dump *d, __u32 id,
574 				    const char *fname, int lvl);
575 static void btf_dump_emit_type_chain(struct btf_dump *d,
576 				     struct id_stack *decl_stack,
577 				     const char *fname, int lvl);
578 
579 static const char *btf_dump_type_name(struct btf_dump *d, __u32 id);
580 static const char *btf_dump_ident_name(struct btf_dump *d, __u32 id);
581 static size_t btf_dump_name_dups(struct btf_dump *d, struct hashmap *name_map,
582 				 const char *orig_name);
583 
584 static bool btf_dump_is_blacklisted(struct btf_dump *d, __u32 id)
585 {
586 	const struct btf_type *t = btf__type_by_id(d->btf, id);
587 
588 	/* __builtin_va_list is a compiler built-in, which causes compilation
589 	 * errors, when compiling w/ different compiler, then used to compile
590 	 * original code (e.g., GCC to compile kernel, Clang to use generated
591 	 * C header from BTF). As it is built-in, it should be already defined
592 	 * properly internally in compiler.
593 	 */
594 	if (t->name_off == 0)
595 		return false;
596 	return strcmp(btf_name_of(d, t->name_off), "__builtin_va_list") == 0;
597 }
598 
599 /*
600  * Emit C-syntax definitions of types from chains of BTF types.
601  *
602  * High-level handling of determining necessary forward declarations are handled
603  * by btf_dump_emit_type() itself, but all nitty-gritty details of emitting type
604  * declarations/definitions in C syntax  are handled by a combo of
605  * btf_dump_emit_type_decl()/btf_dump_emit_type_chain() w/ delegation to
606  * corresponding btf_dump_emit_*_{def,fwd}() functions.
607  *
608  * We also keep track of "containing struct/union type ID" to determine when
609  * we reference it from inside and thus can avoid emitting unnecessary forward
610  * declaration.
611  *
612  * This algorithm is designed in such a way, that even if some error occurs
613  * (either technical, e.g., out of memory, or logical, i.e., malformed BTF
614  * that doesn't comply to C rules completely), algorithm will try to proceed
615  * and produce as much meaningful output as possible.
616  */
617 static void btf_dump_emit_type(struct btf_dump *d, __u32 id, __u32 cont_id)
618 {
619 	struct btf_dump_type_aux_state *tstate = &d->type_states[id];
620 	bool top_level_def = cont_id == 0;
621 	const struct btf_type *t;
622 	__u16 kind;
623 
624 	if (tstate->emit_state == EMITTED)
625 		return;
626 
627 	t = btf__type_by_id(d->btf, id);
628 	kind = btf_kind(t);
629 
630 	if (tstate->emit_state == EMITTING) {
631 		if (tstate->fwd_emitted)
632 			return;
633 
634 		switch (kind) {
635 		case BTF_KIND_STRUCT:
636 		case BTF_KIND_UNION:
637 			/*
638 			 * if we are referencing a struct/union that we are
639 			 * part of - then no need for fwd declaration
640 			 */
641 			if (id == cont_id)
642 				return;
643 			if (t->name_off == 0) {
644 				pr_warn("anonymous struct/union loop, id:[%u]\n",
645 					id);
646 				return;
647 			}
648 			btf_dump_emit_struct_fwd(d, id, t);
649 			btf_dump_printf(d, ";\n\n");
650 			tstate->fwd_emitted = 1;
651 			break;
652 		case BTF_KIND_TYPEDEF:
653 			/*
654 			 * for typedef fwd_emitted means typedef definition
655 			 * was emitted, but it can be used only for "weak"
656 			 * references through pointer only, not for embedding
657 			 */
658 			if (!btf_dump_is_blacklisted(d, id)) {
659 				btf_dump_emit_typedef_def(d, id, t, 0);
660 				btf_dump_printf(d, ";\n\n");
661 			};
662 			tstate->fwd_emitted = 1;
663 			break;
664 		default:
665 			break;
666 		}
667 
668 		return;
669 	}
670 
671 	switch (kind) {
672 	case BTF_KIND_INT:
673 		tstate->emit_state = EMITTED;
674 		break;
675 	case BTF_KIND_ENUM:
676 		if (top_level_def) {
677 			btf_dump_emit_enum_def(d, id, t, 0);
678 			btf_dump_printf(d, ";\n\n");
679 		}
680 		tstate->emit_state = EMITTED;
681 		break;
682 	case BTF_KIND_PTR:
683 	case BTF_KIND_VOLATILE:
684 	case BTF_KIND_CONST:
685 	case BTF_KIND_RESTRICT:
686 		btf_dump_emit_type(d, t->type, cont_id);
687 		break;
688 	case BTF_KIND_ARRAY:
689 		btf_dump_emit_type(d, btf_array(t)->type, cont_id);
690 		break;
691 	case BTF_KIND_FWD:
692 		btf_dump_emit_fwd_def(d, id, t);
693 		btf_dump_printf(d, ";\n\n");
694 		tstate->emit_state = EMITTED;
695 		break;
696 	case BTF_KIND_TYPEDEF:
697 		tstate->emit_state = EMITTING;
698 		btf_dump_emit_type(d, t->type, id);
699 		/*
700 		 * typedef can server as both definition and forward
701 		 * declaration; at this stage someone depends on
702 		 * typedef as a forward declaration (refers to it
703 		 * through pointer), so unless we already did it,
704 		 * emit typedef as a forward declaration
705 		 */
706 		if (!tstate->fwd_emitted && !btf_dump_is_blacklisted(d, id)) {
707 			btf_dump_emit_typedef_def(d, id, t, 0);
708 			btf_dump_printf(d, ";\n\n");
709 		}
710 		tstate->emit_state = EMITTED;
711 		break;
712 	case BTF_KIND_STRUCT:
713 	case BTF_KIND_UNION:
714 		tstate->emit_state = EMITTING;
715 		/* if it's a top-level struct/union definition or struct/union
716 		 * is anonymous, then in C we'll be emitting all fields and
717 		 * their types (as opposed to just `struct X`), so we need to
718 		 * make sure that all types, referenced from struct/union
719 		 * members have necessary forward-declarations, where
720 		 * applicable
721 		 */
722 		if (top_level_def || t->name_off == 0) {
723 			const struct btf_member *m = btf_members(t);
724 			__u16 vlen = btf_vlen(t);
725 			int i, new_cont_id;
726 
727 			new_cont_id = t->name_off == 0 ? cont_id : id;
728 			for (i = 0; i < vlen; i++, m++)
729 				btf_dump_emit_type(d, m->type, new_cont_id);
730 		} else if (!tstate->fwd_emitted && id != cont_id) {
731 			btf_dump_emit_struct_fwd(d, id, t);
732 			btf_dump_printf(d, ";\n\n");
733 			tstate->fwd_emitted = 1;
734 		}
735 
736 		if (top_level_def) {
737 			btf_dump_emit_struct_def(d, id, t, 0);
738 			btf_dump_printf(d, ";\n\n");
739 			tstate->emit_state = EMITTED;
740 		} else {
741 			tstate->emit_state = NOT_EMITTED;
742 		}
743 		break;
744 	case BTF_KIND_FUNC_PROTO: {
745 		const struct btf_param *p = btf_params(t);
746 		__u16 vlen = btf_vlen(t);
747 		int i;
748 
749 		btf_dump_emit_type(d, t->type, cont_id);
750 		for (i = 0; i < vlen; i++, p++)
751 			btf_dump_emit_type(d, p->type, cont_id);
752 
753 		break;
754 	}
755 	default:
756 		break;
757 	}
758 }
759 
760 static bool btf_is_struct_packed(const struct btf *btf, __u32 id,
761 				 const struct btf_type *t)
762 {
763 	const struct btf_member *m;
764 	int align, i, bit_sz;
765 	__u16 vlen;
766 
767 	align = btf__align_of(btf, id);
768 	/* size of a non-packed struct has to be a multiple of its alignment*/
769 	if (align && t->size % align)
770 		return true;
771 
772 	m = btf_members(t);
773 	vlen = btf_vlen(t);
774 	/* all non-bitfield fields have to be naturally aligned */
775 	for (i = 0; i < vlen; i++, m++) {
776 		align = btf__align_of(btf, m->type);
777 		bit_sz = btf_member_bitfield_size(t, i);
778 		if (align && bit_sz == 0 && m->offset % (8 * align) != 0)
779 			return true;
780 	}
781 
782 	/*
783 	 * if original struct was marked as packed, but its layout is
784 	 * naturally aligned, we'll detect that it's not packed
785 	 */
786 	return false;
787 }
788 
789 static int chip_away_bits(int total, int at_most)
790 {
791 	return total % at_most ? : at_most;
792 }
793 
794 static void btf_dump_emit_bit_padding(const struct btf_dump *d,
795 				      int cur_off, int m_off, int m_bit_sz,
796 				      int align, int lvl)
797 {
798 	int off_diff = m_off - cur_off;
799 	int ptr_bits = sizeof(void *) * 8;
800 
801 	if (off_diff <= 0)
802 		/* no gap */
803 		return;
804 	if (m_bit_sz == 0 && off_diff < align * 8)
805 		/* natural padding will take care of a gap */
806 		return;
807 
808 	while (off_diff > 0) {
809 		const char *pad_type;
810 		int pad_bits;
811 
812 		if (ptr_bits > 32 && off_diff > 32) {
813 			pad_type = "long";
814 			pad_bits = chip_away_bits(off_diff, ptr_bits);
815 		} else if (off_diff > 16) {
816 			pad_type = "int";
817 			pad_bits = chip_away_bits(off_diff, 32);
818 		} else if (off_diff > 8) {
819 			pad_type = "short";
820 			pad_bits = chip_away_bits(off_diff, 16);
821 		} else {
822 			pad_type = "char";
823 			pad_bits = chip_away_bits(off_diff, 8);
824 		}
825 		btf_dump_printf(d, "\n%s%s: %d;", pfx(lvl), pad_type, pad_bits);
826 		off_diff -= pad_bits;
827 	}
828 }
829 
830 static void btf_dump_emit_struct_fwd(struct btf_dump *d, __u32 id,
831 				     const struct btf_type *t)
832 {
833 	btf_dump_printf(d, "%s %s",
834 			btf_is_struct(t) ? "struct" : "union",
835 			btf_dump_type_name(d, id));
836 }
837 
838 static void btf_dump_emit_struct_def(struct btf_dump *d,
839 				     __u32 id,
840 				     const struct btf_type *t,
841 				     int lvl)
842 {
843 	const struct btf_member *m = btf_members(t);
844 	bool is_struct = btf_is_struct(t);
845 	int align, i, packed, off = 0;
846 	__u16 vlen = btf_vlen(t);
847 
848 	packed = is_struct ? btf_is_struct_packed(d->btf, id, t) : 0;
849 
850 	btf_dump_printf(d, "%s%s%s {",
851 			is_struct ? "struct" : "union",
852 			t->name_off ? " " : "",
853 			btf_dump_type_name(d, id));
854 
855 	for (i = 0; i < vlen; i++, m++) {
856 		const char *fname;
857 		int m_off, m_sz;
858 
859 		fname = btf_name_of(d, m->name_off);
860 		m_sz = btf_member_bitfield_size(t, i);
861 		m_off = btf_member_bit_offset(t, i);
862 		align = packed ? 1 : btf__align_of(d->btf, m->type);
863 
864 		btf_dump_emit_bit_padding(d, off, m_off, m_sz, align, lvl + 1);
865 		btf_dump_printf(d, "\n%s", pfx(lvl + 1));
866 		btf_dump_emit_type_decl(d, m->type, fname, lvl + 1);
867 
868 		if (m_sz) {
869 			btf_dump_printf(d, ": %d", m_sz);
870 			off = m_off + m_sz;
871 		} else {
872 			m_sz = max(0, btf__resolve_size(d->btf, m->type));
873 			off = m_off + m_sz * 8;
874 		}
875 		btf_dump_printf(d, ";");
876 	}
877 
878 	/* pad at the end, if necessary */
879 	if (is_struct) {
880 		align = packed ? 1 : btf__align_of(d->btf, id);
881 		btf_dump_emit_bit_padding(d, off, t->size * 8, 0, align,
882 					  lvl + 1);
883 	}
884 
885 	if (vlen)
886 		btf_dump_printf(d, "\n");
887 	btf_dump_printf(d, "%s}", pfx(lvl));
888 	if (packed)
889 		btf_dump_printf(d, " __attribute__((packed))");
890 }
891 
892 static void btf_dump_emit_enum_fwd(struct btf_dump *d, __u32 id,
893 				   const struct btf_type *t)
894 {
895 	btf_dump_printf(d, "enum %s", btf_dump_type_name(d, id));
896 }
897 
898 static void btf_dump_emit_enum_def(struct btf_dump *d, __u32 id,
899 				   const struct btf_type *t,
900 				   int lvl)
901 {
902 	const struct btf_enum *v = btf_enum(t);
903 	__u16 vlen = btf_vlen(t);
904 	const char *name;
905 	size_t dup_cnt;
906 	int i;
907 
908 	btf_dump_printf(d, "enum%s%s",
909 			t->name_off ? " " : "",
910 			btf_dump_type_name(d, id));
911 
912 	if (vlen) {
913 		btf_dump_printf(d, " {");
914 		for (i = 0; i < vlen; i++, v++) {
915 			name = btf_name_of(d, v->name_off);
916 			/* enumerators share namespace with typedef idents */
917 			dup_cnt = btf_dump_name_dups(d, d->ident_names, name);
918 			if (dup_cnt > 1) {
919 				btf_dump_printf(d, "\n%s%s___%zu = %d,",
920 						pfx(lvl + 1), name, dup_cnt,
921 						(__s32)v->val);
922 			} else {
923 				btf_dump_printf(d, "\n%s%s = %d,",
924 						pfx(lvl + 1), name,
925 						(__s32)v->val);
926 			}
927 		}
928 		btf_dump_printf(d, "\n%s}", pfx(lvl));
929 	}
930 }
931 
932 static void btf_dump_emit_fwd_def(struct btf_dump *d, __u32 id,
933 				  const struct btf_type *t)
934 {
935 	const char *name = btf_dump_type_name(d, id);
936 
937 	if (btf_kflag(t))
938 		btf_dump_printf(d, "union %s", name);
939 	else
940 		btf_dump_printf(d, "struct %s", name);
941 }
942 
943 static void btf_dump_emit_typedef_def(struct btf_dump *d, __u32 id,
944 				     const struct btf_type *t, int lvl)
945 {
946 	const char *name = btf_dump_ident_name(d, id);
947 
948 	/*
949 	 * Old GCC versions are emitting invalid typedef for __gnuc_va_list
950 	 * pointing to VOID. This generates warnings from btf_dump() and
951 	 * results in uncompilable header file, so we are fixing it up here
952 	 * with valid typedef into __builtin_va_list.
953 	 */
954 	if (t->type == 0 && strcmp(name, "__gnuc_va_list") == 0) {
955 		btf_dump_printf(d, "typedef __builtin_va_list __gnuc_va_list");
956 		return;
957 	}
958 
959 	btf_dump_printf(d, "typedef ");
960 	btf_dump_emit_type_decl(d, t->type, name, lvl);
961 }
962 
963 static int btf_dump_push_decl_stack_id(struct btf_dump *d, __u32 id)
964 {
965 	__u32 *new_stack;
966 	size_t new_cap;
967 
968 	if (d->decl_stack_cnt >= d->decl_stack_cap) {
969 		new_cap = max(16, d->decl_stack_cap * 3 / 2);
970 		new_stack = realloc(d->decl_stack,
971 				    new_cap * sizeof(new_stack[0]));
972 		if (!new_stack)
973 			return -ENOMEM;
974 		d->decl_stack = new_stack;
975 		d->decl_stack_cap = new_cap;
976 	}
977 
978 	d->decl_stack[d->decl_stack_cnt++] = id;
979 
980 	return 0;
981 }
982 
983 /*
984  * Emit type declaration (e.g., field type declaration in a struct or argument
985  * declaration in function prototype) in correct C syntax.
986  *
987  * For most types it's trivial, but there are few quirky type declaration
988  * cases worth mentioning:
989  *   - function prototypes (especially nesting of function prototypes);
990  *   - arrays;
991  *   - const/volatile/restrict for pointers vs other types.
992  *
993  * For a good discussion of *PARSING* C syntax (as a human), see
994  * Peter van der Linden's "Expert C Programming: Deep C Secrets",
995  * Ch.3 "Unscrambling Declarations in C".
996  *
997  * It won't help with BTF to C conversion much, though, as it's an opposite
998  * problem. So we came up with this algorithm in reverse to van der Linden's
999  * parsing algorithm. It goes from structured BTF representation of type
1000  * declaration to a valid compilable C syntax.
1001  *
1002  * For instance, consider this C typedef:
1003  *	typedef const int * const * arr[10] arr_t;
1004  * It will be represented in BTF with this chain of BTF types:
1005  *	[typedef] -> [array] -> [ptr] -> [const] -> [ptr] -> [const] -> [int]
1006  *
1007  * Notice how [const] modifier always goes before type it modifies in BTF type
1008  * graph, but in C syntax, const/volatile/restrict modifiers are written to
1009  * the right of pointers, but to the left of other types. There are also other
1010  * quirks, like function pointers, arrays of them, functions returning other
1011  * functions, etc.
1012  *
1013  * We handle that by pushing all the types to a stack, until we hit "terminal"
1014  * type (int/enum/struct/union/fwd). Then depending on the kind of a type on
1015  * top of a stack, modifiers are handled differently. Array/function pointers
1016  * have also wildly different syntax and how nesting of them are done. See
1017  * code for authoritative definition.
1018  *
1019  * To avoid allocating new stack for each independent chain of BTF types, we
1020  * share one bigger stack, with each chain working only on its own local view
1021  * of a stack frame. Some care is required to "pop" stack frames after
1022  * processing type declaration chain.
1023  */
1024 int btf_dump__emit_type_decl(struct btf_dump *d, __u32 id,
1025 			     const struct btf_dump_emit_type_decl_opts *opts)
1026 {
1027 	const char *fname;
1028 	int lvl;
1029 
1030 	if (!OPTS_VALID(opts, btf_dump_emit_type_decl_opts))
1031 		return -EINVAL;
1032 
1033 	fname = OPTS_GET(opts, field_name, NULL);
1034 	lvl = OPTS_GET(opts, indent_level, 0);
1035 	btf_dump_emit_type_decl(d, id, fname, lvl);
1036 	return 0;
1037 }
1038 
1039 static void btf_dump_emit_type_decl(struct btf_dump *d, __u32 id,
1040 				    const char *fname, int lvl)
1041 {
1042 	struct id_stack decl_stack;
1043 	const struct btf_type *t;
1044 	int err, stack_start;
1045 
1046 	stack_start = d->decl_stack_cnt;
1047 	for (;;) {
1048 		err = btf_dump_push_decl_stack_id(d, id);
1049 		if (err < 0) {
1050 			/*
1051 			 * if we don't have enough memory for entire type decl
1052 			 * chain, restore stack, emit warning, and try to
1053 			 * proceed nevertheless
1054 			 */
1055 			pr_warn("not enough memory for decl stack:%d", err);
1056 			d->decl_stack_cnt = stack_start;
1057 			return;
1058 		}
1059 
1060 		/* VOID */
1061 		if (id == 0)
1062 			break;
1063 
1064 		t = btf__type_by_id(d->btf, id);
1065 		switch (btf_kind(t)) {
1066 		case BTF_KIND_PTR:
1067 		case BTF_KIND_VOLATILE:
1068 		case BTF_KIND_CONST:
1069 		case BTF_KIND_RESTRICT:
1070 		case BTF_KIND_FUNC_PROTO:
1071 			id = t->type;
1072 			break;
1073 		case BTF_KIND_ARRAY:
1074 			id = btf_array(t)->type;
1075 			break;
1076 		case BTF_KIND_INT:
1077 		case BTF_KIND_ENUM:
1078 		case BTF_KIND_FWD:
1079 		case BTF_KIND_STRUCT:
1080 		case BTF_KIND_UNION:
1081 		case BTF_KIND_TYPEDEF:
1082 			goto done;
1083 		default:
1084 			pr_warn("unexpected type in decl chain, kind:%u, id:[%u]\n",
1085 				btf_kind(t), id);
1086 			goto done;
1087 		}
1088 	}
1089 done:
1090 	/*
1091 	 * We might be inside a chain of declarations (e.g., array of function
1092 	 * pointers returning anonymous (so inlined) structs, having another
1093 	 * array field). Each of those needs its own "stack frame" to handle
1094 	 * emitting of declarations. Those stack frames are non-overlapping
1095 	 * portions of shared btf_dump->decl_stack. To make it a bit nicer to
1096 	 * handle this set of nested stacks, we create a view corresponding to
1097 	 * our own "stack frame" and work with it as an independent stack.
1098 	 * We'll need to clean up after emit_type_chain() returns, though.
1099 	 */
1100 	decl_stack.ids = d->decl_stack + stack_start;
1101 	decl_stack.cnt = d->decl_stack_cnt - stack_start;
1102 	btf_dump_emit_type_chain(d, &decl_stack, fname, lvl);
1103 	/*
1104 	 * emit_type_chain() guarantees that it will pop its entire decl_stack
1105 	 * frame before returning. But it works with a read-only view into
1106 	 * decl_stack, so it doesn't actually pop anything from the
1107 	 * perspective of shared btf_dump->decl_stack, per se. We need to
1108 	 * reset decl_stack state to how it was before us to avoid it growing
1109 	 * all the time.
1110 	 */
1111 	d->decl_stack_cnt = stack_start;
1112 }
1113 
1114 static void btf_dump_emit_mods(struct btf_dump *d, struct id_stack *decl_stack)
1115 {
1116 	const struct btf_type *t;
1117 	__u32 id;
1118 
1119 	while (decl_stack->cnt) {
1120 		id = decl_stack->ids[decl_stack->cnt - 1];
1121 		t = btf__type_by_id(d->btf, id);
1122 
1123 		switch (btf_kind(t)) {
1124 		case BTF_KIND_VOLATILE:
1125 			btf_dump_printf(d, "volatile ");
1126 			break;
1127 		case BTF_KIND_CONST:
1128 			btf_dump_printf(d, "const ");
1129 			break;
1130 		case BTF_KIND_RESTRICT:
1131 			btf_dump_printf(d, "restrict ");
1132 			break;
1133 		default:
1134 			return;
1135 		}
1136 		decl_stack->cnt--;
1137 	}
1138 }
1139 
1140 static void btf_dump_emit_name(const struct btf_dump *d,
1141 			       const char *name, bool last_was_ptr)
1142 {
1143 	bool separate = name[0] && !last_was_ptr;
1144 
1145 	btf_dump_printf(d, "%s%s", separate ? " " : "", name);
1146 }
1147 
1148 static void btf_dump_emit_type_chain(struct btf_dump *d,
1149 				     struct id_stack *decls,
1150 				     const char *fname, int lvl)
1151 {
1152 	/*
1153 	 * last_was_ptr is used to determine if we need to separate pointer
1154 	 * asterisk (*) from previous part of type signature with space, so
1155 	 * that we get `int ***`, instead of `int * * *`. We default to true
1156 	 * for cases where we have single pointer in a chain. E.g., in ptr ->
1157 	 * func_proto case. func_proto will start a new emit_type_chain call
1158 	 * with just ptr, which should be emitted as (*) or (*<fname>), so we
1159 	 * don't want to prepend space for that last pointer.
1160 	 */
1161 	bool last_was_ptr = true;
1162 	const struct btf_type *t;
1163 	const char *name;
1164 	__u16 kind;
1165 	__u32 id;
1166 
1167 	while (decls->cnt) {
1168 		id = decls->ids[--decls->cnt];
1169 		if (id == 0) {
1170 			/* VOID is a special snowflake */
1171 			btf_dump_emit_mods(d, decls);
1172 			btf_dump_printf(d, "void");
1173 			last_was_ptr = false;
1174 			continue;
1175 		}
1176 
1177 		t = btf__type_by_id(d->btf, id);
1178 		kind = btf_kind(t);
1179 
1180 		switch (kind) {
1181 		case BTF_KIND_INT:
1182 			btf_dump_emit_mods(d, decls);
1183 			name = btf_name_of(d, t->name_off);
1184 			btf_dump_printf(d, "%s", name);
1185 			break;
1186 		case BTF_KIND_STRUCT:
1187 		case BTF_KIND_UNION:
1188 			btf_dump_emit_mods(d, decls);
1189 			/* inline anonymous struct/union */
1190 			if (t->name_off == 0)
1191 				btf_dump_emit_struct_def(d, id, t, lvl);
1192 			else
1193 				btf_dump_emit_struct_fwd(d, id, t);
1194 			break;
1195 		case BTF_KIND_ENUM:
1196 			btf_dump_emit_mods(d, decls);
1197 			/* inline anonymous enum */
1198 			if (t->name_off == 0)
1199 				btf_dump_emit_enum_def(d, id, t, lvl);
1200 			else
1201 				btf_dump_emit_enum_fwd(d, id, t);
1202 			break;
1203 		case BTF_KIND_FWD:
1204 			btf_dump_emit_mods(d, decls);
1205 			btf_dump_emit_fwd_def(d, id, t);
1206 			break;
1207 		case BTF_KIND_TYPEDEF:
1208 			btf_dump_emit_mods(d, decls);
1209 			btf_dump_printf(d, "%s", btf_dump_ident_name(d, id));
1210 			break;
1211 		case BTF_KIND_PTR:
1212 			btf_dump_printf(d, "%s", last_was_ptr ? "*" : " *");
1213 			break;
1214 		case BTF_KIND_VOLATILE:
1215 			btf_dump_printf(d, " volatile");
1216 			break;
1217 		case BTF_KIND_CONST:
1218 			btf_dump_printf(d, " const");
1219 			break;
1220 		case BTF_KIND_RESTRICT:
1221 			btf_dump_printf(d, " restrict");
1222 			break;
1223 		case BTF_KIND_ARRAY: {
1224 			const struct btf_array *a = btf_array(t);
1225 			const struct btf_type *next_t;
1226 			__u32 next_id;
1227 			bool multidim;
1228 			/*
1229 			 * GCC has a bug
1230 			 * (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=8354)
1231 			 * which causes it to emit extra const/volatile
1232 			 * modifiers for an array, if array's element type has
1233 			 * const/volatile modifiers. Clang doesn't do that.
1234 			 * In general, it doesn't seem very meaningful to have
1235 			 * a const/volatile modifier for array, so we are
1236 			 * going to silently skip them here.
1237 			 */
1238 			while (decls->cnt) {
1239 				next_id = decls->ids[decls->cnt - 1];
1240 				next_t = btf__type_by_id(d->btf, next_id);
1241 				if (btf_is_mod(next_t))
1242 					decls->cnt--;
1243 				else
1244 					break;
1245 			}
1246 
1247 			if (decls->cnt == 0) {
1248 				btf_dump_emit_name(d, fname, last_was_ptr);
1249 				btf_dump_printf(d, "[%u]", a->nelems);
1250 				return;
1251 			}
1252 
1253 			next_id = decls->ids[decls->cnt - 1];
1254 			next_t = btf__type_by_id(d->btf, next_id);
1255 			multidim = btf_is_array(next_t);
1256 			/* we need space if we have named non-pointer */
1257 			if (fname[0] && !last_was_ptr)
1258 				btf_dump_printf(d, " ");
1259 			/* no parentheses for multi-dimensional array */
1260 			if (!multidim)
1261 				btf_dump_printf(d, "(");
1262 			btf_dump_emit_type_chain(d, decls, fname, lvl);
1263 			if (!multidim)
1264 				btf_dump_printf(d, ")");
1265 			btf_dump_printf(d, "[%u]", a->nelems);
1266 			return;
1267 		}
1268 		case BTF_KIND_FUNC_PROTO: {
1269 			const struct btf_param *p = btf_params(t);
1270 			__u16 vlen = btf_vlen(t);
1271 			int i;
1272 
1273 			btf_dump_emit_mods(d, decls);
1274 			if (decls->cnt) {
1275 				btf_dump_printf(d, " (");
1276 				btf_dump_emit_type_chain(d, decls, fname, lvl);
1277 				btf_dump_printf(d, ")");
1278 			} else {
1279 				btf_dump_emit_name(d, fname, last_was_ptr);
1280 			}
1281 			btf_dump_printf(d, "(");
1282 			/*
1283 			 * Clang for BPF target generates func_proto with no
1284 			 * args as a func_proto with a single void arg (e.g.,
1285 			 * `int (*f)(void)` vs just `int (*f)()`). We are
1286 			 * going to pretend there are no args for such case.
1287 			 */
1288 			if (vlen == 1 && p->type == 0) {
1289 				btf_dump_printf(d, ")");
1290 				return;
1291 			}
1292 
1293 			for (i = 0; i < vlen; i++, p++) {
1294 				if (i > 0)
1295 					btf_dump_printf(d, ", ");
1296 
1297 				/* last arg of type void is vararg */
1298 				if (i == vlen - 1 && p->type == 0) {
1299 					btf_dump_printf(d, "...");
1300 					break;
1301 				}
1302 
1303 				name = btf_name_of(d, p->name_off);
1304 				btf_dump_emit_type_decl(d, p->type, name, lvl);
1305 			}
1306 
1307 			btf_dump_printf(d, ")");
1308 			return;
1309 		}
1310 		default:
1311 			pr_warn("unexpected type in decl chain, kind:%u, id:[%u]\n",
1312 				kind, id);
1313 			return;
1314 		}
1315 
1316 		last_was_ptr = kind == BTF_KIND_PTR;
1317 	}
1318 
1319 	btf_dump_emit_name(d, fname, last_was_ptr);
1320 }
1321 
1322 /* return number of duplicates (occurrences) of a given name */
1323 static size_t btf_dump_name_dups(struct btf_dump *d, struct hashmap *name_map,
1324 				 const char *orig_name)
1325 {
1326 	size_t dup_cnt = 0;
1327 
1328 	hashmap__find(name_map, orig_name, (void **)&dup_cnt);
1329 	dup_cnt++;
1330 	hashmap__set(name_map, orig_name, (void *)dup_cnt, NULL, NULL);
1331 
1332 	return dup_cnt;
1333 }
1334 
1335 static const char *btf_dump_resolve_name(struct btf_dump *d, __u32 id,
1336 					 struct hashmap *name_map)
1337 {
1338 	struct btf_dump_type_aux_state *s = &d->type_states[id];
1339 	const struct btf_type *t = btf__type_by_id(d->btf, id);
1340 	const char *orig_name = btf_name_of(d, t->name_off);
1341 	const char **cached_name = &d->cached_names[id];
1342 	size_t dup_cnt;
1343 
1344 	if (t->name_off == 0)
1345 		return "";
1346 
1347 	if (s->name_resolved)
1348 		return *cached_name ? *cached_name : orig_name;
1349 
1350 	dup_cnt = btf_dump_name_dups(d, name_map, orig_name);
1351 	if (dup_cnt > 1) {
1352 		const size_t max_len = 256;
1353 		char new_name[max_len];
1354 
1355 		snprintf(new_name, max_len, "%s___%zu", orig_name, dup_cnt);
1356 		*cached_name = strdup(new_name);
1357 	}
1358 
1359 	s->name_resolved = 1;
1360 	return *cached_name ? *cached_name : orig_name;
1361 }
1362 
1363 static const char *btf_dump_type_name(struct btf_dump *d, __u32 id)
1364 {
1365 	return btf_dump_resolve_name(d, id, d->type_names);
1366 }
1367 
1368 static const char *btf_dump_ident_name(struct btf_dump *d, __u32 id)
1369 {
1370 	return btf_dump_resolve_name(d, id, d->ident_names);
1371 }
1372