1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2006 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #ifndef _CTFTOOLS_H
27 #define	_CTFTOOLS_H
28 
29 /*
30  * Functions and data structures used in the manipulation of stabs and CTF data
31  */
32 
33 #include <stdio.h>
34 #include <stdlib.h>
35 #include <stdarg.h>
36 #include <libelf.h>
37 #include <gelf.h>
38 #include <pthread.h>
39 
40 #include <sys/ccompile.h>
41 
42 #ifdef __cplusplus
43 extern "C" {
44 #endif
45 
46 #include "list.h"
47 #include "hash.h"
48 
49 #ifndef DEBUG_LEVEL
50 #define	DEBUG_LEVEL 0
51 #endif
52 #ifndef DEBUG_PARSE
53 #define	DEBUG_PARSE 0
54 #endif
55 
56 #ifndef DEBUG_STREAM
57 #define	DEBUG_STREAM stderr
58 #endif
59 
60 #ifndef MAX
61 #define	MAX(a, b) 		((a) < (b) ? (b) : (a))
62 #endif
63 
64 #ifndef MIN
65 #define	MIN(a, b) 		((a) > (b) ? (b) : (a))
66 #endif
67 
68 #ifndef TRUE
69 #define	TRUE	1
70 #endif
71 #ifndef FALSE
72 #define	FALSE	0
73 #endif
74 
75 #define	CTF_ELF_SCN_NAME	".SUNW_ctf"
76 
77 #define	CTF_LABEL_LASTIDX	-1
78 
79 #define	CTF_DEFAULT_LABEL	"*** No Label Provided ***"
80 
81 /*
82  * Default hash sizes
83  */
84 #define	TDATA_LAYOUT_HASH_SIZE	8191	/* A tdesc hash based on layout */
85 #define	TDATA_ID_HASH_SIZE	997	/* A tdesc hash based on type id */
86 #define	IIDESC_HASH_SIZE	8191	/* Hash of iidesc's */
87 
88 /*
89  * The default function argument array size.  We'll realloc the array larger
90  * if we need to, but we want a default value that will allow us to avoid
91  * reallocation in the common case.
92  */
93 #define	FUNCARG_DEF	5
94 
95 extern const char *progname;
96 extern int debug_level;
97 extern int debug_parse;
98 extern char *curhdr;
99 
100 /*
101  * This is a partial copy of the stab.h that DevPro includes with their
102  * compiler.
103  */
104 typedef struct stab {
105 	uint32_t	n_strx;
106 	uint8_t		n_type;
107 	int8_t		n_other;
108 	int16_t		n_desc;
109 	uint32_t	n_value;
110 } stab_t;
111 
112 #define	N_GSYM	0x20	/* global symbol: name,,0,type,0 */
113 #define	N_FUN	0x24	/* procedure: name,,0,linenumber,0 */
114 #define	N_STSYM	0x26	/* static symbol: name,,0,type,0 or section relative */
115 #define	N_LCSYM	0x28	/* .lcomm symbol: name,,0,type,0 or section relative */
116 #define	N_ROSYM	0x2c	/* ro_data: name,,0,type,0 or section relative */
117 #define	N_OPT	0x3c	/* compiler options */
118 #define	N_RSYM	0x40	/* register sym: name,,0,type,register */
119 #define	N_SO	0x64	/* source file name: name,,0,0,0 */
120 #define	N_LSYM	0x80	/* local sym: name,,0,type,offset */
121 #define	N_SOL	0x84	/* #included file name: name,,0,0,0 */
122 #define	N_PSYM	0xa0	/* parameter: name,,0,type,offset */
123 #define	N_LBRAC	0xc0	/* left bracket: 0,,0,nesting level,function relative */
124 #define	N_RBRAC	0xe0	/* right bracket: 0,,0,nesting level,func relative */
125 #define	N_BINCL 0x82	/* header file: name,,0,0,0 */
126 #define	N_EINCL 0xa2	/* end of include file */
127 
128 /*
129  * Nodes in the type tree
130  *
131  * Each node consists of a single tdesc_t, with one of several auxiliary
132  * structures linked in via the `data' union.
133  */
134 
135 /* The type of tdesc_t node */
136 typedef enum stabtype {
137 	STABTYPE_FIRST, /* do not use */
138 	INTRINSIC,
139 	POINTER,
140 	REFERENCE,
141 	ARRAY,
142 	FUNCTION,
143 	STRUCT,
144 	UNION,
145 	CLASS,
146 	ENUM,
147 	FORWARD,
148 	TYPEDEF,
149 	TYPEDEF_UNRES,
150 	VOLATILE,
151 	CONST,
152 	RESTRICT,
153 	STABTYPE_LAST /* do not use */
154 } stabtype_t;
155 
156 typedef struct tdesc tdesc_t;
157 
158 /* Auxiliary structure for array tdesc_t */
159 typedef struct ardef {
160 	tdesc_t	*ad_contents;
161 	tdesc_t *ad_idxtype;
162 	uint_t	ad_nelems;
163 } ardef_t;
164 
165 /* Auxiliary structure for structure/union tdesc_t */
166 typedef struct mlist {
167 	int	ml_offset;	/* Offset from start of structure (in bits) */
168 	uint_t	ml_size;	/* Member size (in bits) */
169 	char	*ml_name;	/* Member name */
170 	struct	tdesc *ml_type;	/* Member type */
171 	struct	mlist *ml_next;	/* Next member */
172 } mlist_t;
173 
174 /* Auxiliary structure for enum tdesc_t */
175 typedef struct elist {
176 	char	*el_name;
177 	int	el_number;
178 	struct elist *el_next;
179 } elist_t;
180 
181 /* Auxiliary structure for intrinsics (integers and reals) */
182 typedef enum {
183 	INTR_INT,
184 	INTR_REAL
185 } intrtype_t;
186 
187 typedef struct intr {
188 	intrtype_t	intr_type;
189 	int		intr_signed;
190 	union {
191 			char _iformat;
192 			int _fformat;
193 	} _u;
194 	int		intr_offset;
195 	int		intr_nbits;
196 } intr_t;
197 
198 #define	intr_iformat _u._iformat
199 #define	intr_fformat _u._fformat
200 
201 typedef struct fnarg {
202 	char *fna_name;
203 	struct tdesc *fna_type;
204 } fnarg_t;
205 
206 #define	FN_F_GLOBAL	0x1
207 #define	FN_F_VARARGS	0x2
208 
209 typedef struct fndef {
210 	struct tdesc *fn_ret;
211 	uint_t fn_nargs;
212 	tdesc_t **fn_args;
213 	uint_t fn_vargs;
214 } fndef_t;
215 
216 typedef int32_t tid_t;
217 
218 /*
219  * The tdesc_t (Type DESCription) is the basic node type used in the stabs data
220  * structure.  Each data node gets a tdesc structure.  Each node is linked into
221  * a directed graph (think of it as a tree with multiple roots and multiple
222  * leaves), with the root nodes at the top, and intrinsics at the bottom.  The
223  * root nodes, which are pointed to by iidesc nodes, correspond to the types,
224  * globals, and statics defined by the stabs.
225  */
226 struct tdesc {
227 	char	*t_name;
228 	tdesc_t *t_next;	/* Name hash next pointer */
229 
230 	tid_t t_id;
231 	tdesc_t *t_hash;	/* ID hash next pointer */
232 
233 	stabtype_t t_type;
234 	int	t_size;	/* Size in bytes of object represented by this node */
235 
236 	union {
237 		intr_t	*intr;		/* int, real */
238 		tdesc_t *tdesc;		/* ptr, typedef, vol, const, restr */
239 		ardef_t *ardef;		/* array */
240 		mlist_t *members;	/* struct, union */
241 		elist_t *emem;		/* enum */
242 		fndef_t *fndef;		/* function - first is return type */
243 	} t_data;
244 
245 	int t_flags;
246 	int t_vgen;	/* Visitation generation (see traverse.c) */
247 	int t_emark;	/* Equality mark (see equiv_cb() in merge.c) */
248 };
249 
250 #define	t_intr		t_data.intr
251 #define	t_tdesc		t_data.tdesc
252 #define	t_ardef		t_data.ardef
253 #define	t_members	t_data.members
254 #define	t_emem		t_data.emem
255 #define	t_fndef		t_data.fndef
256 
257 #define	TDESC_F_ISROOT		0x1	/* Has an iidesc_t (see below) */
258 #define	TDESC_F_GLOBAL		0x2
259 #define	TDESC_F_RESOLVED	0x4
260 
261 /*
262  * iidesc_t (Interesting Item DESCription) nodes point to tdesc_t nodes that
263  * correspond to "interesting" stabs.  A stab is interesting if it defines a
264  * global or static variable, a global or static function, or a data type.
265  */
266 typedef enum iitype {
267 	II_NOT = 0,
268 	II_GFUN,	/* Global function */
269 	II_SFUN,	/* Static function */
270 	II_GVAR,	/* Global variable */
271 	II_SVAR,	/* Static variable */
272 	II_PSYM,	/* Function argument */
273 	II_SOU,		/* Struct or union */
274 	II_TYPE		/* Type (typedef) */
275 } iitype_t;
276 
277 typedef struct iidesc {
278 	iitype_t	ii_type;
279 	char		*ii_name;
280 	tdesc_t 	*ii_dtype;
281 	char		*ii_owner;	/* File that defined this node */
282 	int		ii_flags;
283 
284 	/* Function arguments (if any) */
285 	int		ii_nargs;
286 	tdesc_t 	**ii_args;
287 	int		ii_vargs;	/* Function uses varargs */
288 } iidesc_t;
289 
290 #define	IIDESC_F_USED	0x1	/* Write this iidesc out */
291 
292 /*
293  * labelent_t nodes identify labels and corresponding type ranges associated
294  * with them.  The label in a given labelent_t is associated with types with
295  * ids <= le_idx.
296  */
297 typedef struct labelent {
298 	char *le_name;
299 	int le_idx;
300 } labelent_t;
301 
302 /*
303  * The tdata_t (Type DATA) structure contains or references all type data for
304  * a given file or, during merging, several files.
305  */
306 typedef struct tdata {
307 	int	td_curemark;	/* Equality mark (see merge.c) */
308 	int	td_curvgen;	/* Visitation generation (see traverse.c) */
309 	int	td_nextid;	/* The ID for the next tdesc_t created */
310 	hash_t	*td_iihash;	/* The iidesc_t nodes for this file */
311 
312 	hash_t	*td_layouthash;	/* The tdesc nodes, hashed by structure */
313 	hash_t	*td_idhash;	/* The tdesc nodes, hashed by type id */
314 	list_t	*td_fwdlist;	/* All forward declaration tdesc nodes */
315 
316 	char	*td_parlabel;	/* Top label uniq'd against in parent */
317 	char	*td_parname;	/* Basename of parent */
318 	list_t	*td_labels;	/* Labels and their type ranges */
319 
320 	pthread_mutex_t td_mergelock;
321 
322 	int	td_ref;
323 } tdata_t;
324 
325 /*
326  * By design, the iidesc hash is heterogeneous.  The CTF emitter, on the
327  * other hand, needs to be able to access the elements of the list by type,
328  * and in a specific sorted order.  An iiburst holds these elements in that
329  * order.  (A burster is a machine that separates carbon-copy forms)
330  */
331 typedef struct iiburst {
332 	int iib_nfuncs;
333 	int iib_curfunc;
334 	iidesc_t **iib_funcs;
335 
336 	int iib_nobjts;
337 	int iib_curobjt;
338 	iidesc_t **iib_objts;
339 
340 	list_t *iib_types;
341 	int iib_maxtypeid;
342 
343 	tdata_t *iib_td;
344 	struct tdtrav_data *iib_tdtd; /* tdtrav_data_t */
345 } iiburst_t;
346 
347 typedef struct ctf_buf ctf_buf_t;
348 
349 typedef struct symit_data symit_data_t;
350 
351 /* fixup_tdescs.c */
352 void cvt_fixstabs(tdata_t *);
353 void cvt_fixups(tdata_t *, size_t);
354 
355 /* ctf.c */
356 caddr_t ctf_gen(iiburst_t *, size_t *, int);
357 tdata_t *ctf_load(char *, caddr_t, size_t, symit_data_t *, char *);
358 
359 /* iidesc.c */
360 iidesc_t *iidesc_new(char *);
361 int iidesc_hash(int, void *);
362 void iter_iidescs_by_name(tdata_t *, const char *,
363     int (*)(void *, void *), void *);
364 iidesc_t *iidesc_dup(iidesc_t *);
365 iidesc_t *iidesc_dup_rename(iidesc_t *, char const *, char const *);
366 void iidesc_add(hash_t *, iidesc_t *);
367 void iidesc_free(void *, void *);
368 int iidesc_count_type(void *, void *);
369 void iidesc_stats(hash_t *);
370 int iidesc_dump(iidesc_t *);
371 
372 /* input.c */
373 typedef enum source_types {
374 	SOURCE_NONE 	= 0,
375 	SOURCE_UNKNOWN	= 1,
376 	SOURCE_C	= 2,
377 	SOURCE_S	= 4
378 } source_types_t;
379 
380 source_types_t built_source_types(Elf *, const char *);
381 int count_files(char **, int);
382 int read_ctf(char **, int, char *, int (*)(tdata_t *, char *, void *),
383     void *, int);
384 int read_ctf_save_cb(tdata_t *, char *, void *);
385 symit_data_t *symit_new(Elf *, const char *);
386 void symit_reset(symit_data_t *);
387 char *symit_curfile(symit_data_t *);
388 GElf_Sym *symit_next(symit_data_t *, int);
389 char *symit_name(symit_data_t *);
390 void symit_free(symit_data_t *);
391 
392 /* merge.c */
393 void merge_into_master(tdata_t *, tdata_t *, tdata_t *, int);
394 
395 /* output.c */
396 #define	CTF_FUZZY_MATCH	0x1 /* match local symbols to global CTF */
397 #define	CTF_USE_DYNSYM	0x2 /* use .dynsym not .symtab */
398 #define	CTF_COMPRESS	0x4 /* compress CTF output */
399 #define	CTF_KEEP_STABS	0x8 /* keep .stabs sections */
400 #define	CTF_SWAP_BYTES	0x10 /* target byte order is different from host */
401 
402 void write_ctf(tdata_t *, const char *, const char *, int);
403 
404 /* parse.c */
405 void parse_init(tdata_t *);
406 void parse_finish(tdata_t *);
407 int parse_stab(stab_t *, char *, iidesc_t **);
408 tdesc_t *lookup(int);
409 tdesc_t *lookupname(const char *);
410 void check_hash(void);
411 void resolve_typed_bitfields(void);
412 
413 /* stabs.c */
414 int stabs_read(tdata_t *, Elf *, char *);
415 
416 /* dwarf.c */
417 int dw_read(tdata_t *, Elf *, char *);
418 const char *dw_tag2str(uint_t);
419 
420 /* tdata.c */
421 tdata_t *tdata_new(void);
422 void tdata_free(tdata_t *);
423 void tdata_build_hashes(tdata_t *td);
424 const char *tdesc_name(tdesc_t *);
425 int tdesc_idhash(int, void *);
426 int tdesc_idcmp(void *, void *);
427 int tdesc_namehash(int, void *);
428 int tdesc_namecmp(void *, void *);
429 int tdesc_layouthash(int, void *);
430 int tdesc_layoutcmp(void *, void *);
431 void tdesc_free(tdesc_t *);
432 void tdata_label_add(tdata_t *, const char *, int);
433 labelent_t *tdata_label_top(tdata_t *);
434 int tdata_label_find(tdata_t *, char *);
435 void tdata_label_free(tdata_t *);
436 void tdata_merge(tdata_t *, tdata_t *);
437 void tdata_label_newmax(tdata_t *, int);
438 
439 /* util.c */
440 int streq(const char *, const char *);
441 int findelfsecidx(Elf *, const char *, const char *);
442 size_t elf_ptrsz(Elf *);
443 char *mktmpname(const char *, const char *);
444 void terminate(const char *, ...) __printflike(1, 2) __dead;
445 void aborterr(const char *, ...) __printflike(1, 2) __dead;
446 void set_terminate_cleanup(void (*)(void));
447 void elfterminate(const char *, const char *, ...) __printflike(2, 3) __dead;
448 void warning(const char *, ...) __printflike(1, 2);
449 void vadebug(int, const char *, va_list) __printflike(2, 0);
450 void debug(int, const char *, ...) __printflike(2, 3);
451 
452 
453 void watch_dump(int);
454 void watch_set(void *, int);
455 
456 #ifdef __cplusplus
457 }
458 #endif
459 
460 #endif /* _CTFTOOLS_H */
461