daily update
[binutils.git] / bfd / syms.c
blob0ff94164ed5e28dfb78a7049980d30e9c55ca096
1 /* Generic symbol-table support for the BFD library.
2 Copyright 1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
3 2000, 2001, 2002, 2003, 2004
4 Free Software Foundation, Inc.
5 Written by Cygnus Support.
7 This file is part of BFD, the Binary File Descriptor library.
9 This program is free software; you can redistribute it and/or modify
10 it under the terms of the GNU General Public License as published by
11 the Free Software Foundation; either version 2 of the License, or
12 (at your option) any later version.
14 This program is distributed in the hope that it will be useful,
15 but WITHOUT ANY WARRANTY; without even the implied warranty of
16 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
17 GNU General Public License for more details.
19 You should have received a copy of the GNU General Public License
20 along with this program; if not, write to the Free Software
21 Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. */
24 SECTION
25 Symbols
27 BFD tries to maintain as much symbol information as it can when
28 it moves information from file to file. BFD passes information
29 to applications though the <<asymbol>> structure. When the
30 application requests the symbol table, BFD reads the table in
31 the native form and translates parts of it into the internal
32 format. To maintain more than the information passed to
33 applications, some targets keep some information ``behind the
34 scenes'' in a structure only the particular back end knows
35 about. For example, the coff back end keeps the original
36 symbol table structure as well as the canonical structure when
37 a BFD is read in. On output, the coff back end can reconstruct
38 the output symbol table so that no information is lost, even
39 information unique to coff which BFD doesn't know or
40 understand. If a coff symbol table were read, but were written
41 through an a.out back end, all the coff specific information
42 would be lost. The symbol table of a BFD
43 is not necessarily read in until a canonicalize request is
44 made. Then the BFD back end fills in a table provided by the
45 application with pointers to the canonical information. To
46 output symbols, the application provides BFD with a table of
47 pointers to pointers to <<asymbol>>s. This allows applications
48 like the linker to output a symbol as it was read, since the ``behind
49 the scenes'' information will be still available.
50 @menu
51 @* Reading Symbols::
52 @* Writing Symbols::
53 @* Mini Symbols::
54 @* typedef asymbol::
55 @* symbol handling functions::
56 @end menu
58 INODE
59 Reading Symbols, Writing Symbols, Symbols, Symbols
60 SUBSECTION
61 Reading symbols
63 There are two stages to reading a symbol table from a BFD:
64 allocating storage, and the actual reading process. This is an
65 excerpt from an application which reads the symbol table:
67 | long storage_needed;
68 | asymbol **symbol_table;
69 | long number_of_symbols;
70 | long i;
72 | storage_needed = bfd_get_symtab_upper_bound (abfd);
74 | if (storage_needed < 0)
75 | FAIL
77 | if (storage_needed == 0)
78 | return;
80 | symbol_table = xmalloc (storage_needed);
81 | ...
82 | number_of_symbols =
83 | bfd_canonicalize_symtab (abfd, symbol_table);
85 | if (number_of_symbols < 0)
86 | FAIL
88 | for (i = 0; i < number_of_symbols; i++)
89 | process_symbol (symbol_table[i]);
91 All storage for the symbols themselves is in an objalloc
92 connected to the BFD; it is freed when the BFD is closed.
94 INODE
95 Writing Symbols, Mini Symbols, Reading Symbols, Symbols
96 SUBSECTION
97 Writing symbols
99 Writing of a symbol table is automatic when a BFD open for
100 writing is closed. The application attaches a vector of
101 pointers to pointers to symbols to the BFD being written, and
102 fills in the symbol count. The close and cleanup code reads
103 through the table provided and performs all the necessary
104 operations. The BFD output code must always be provided with an
105 ``owned'' symbol: one which has come from another BFD, or one
106 which has been created using <<bfd_make_empty_symbol>>. Here is an
107 example showing the creation of a symbol table with only one element:
109 | #include "bfd.h"
110 | int main (void)
112 | bfd *abfd;
113 | asymbol *ptrs[2];
114 | asymbol *new;
116 | abfd = bfd_openw ("foo","a.out-sunos-big");
117 | bfd_set_format (abfd, bfd_object);
118 | new = bfd_make_empty_symbol (abfd);
119 | new->name = "dummy_symbol";
120 | new->section = bfd_make_section_old_way (abfd, ".text");
121 | new->flags = BSF_GLOBAL;
122 | new->value = 0x12345;
124 | ptrs[0] = new;
125 | ptrs[1] = 0;
127 | bfd_set_symtab (abfd, ptrs, 1);
128 | bfd_close (abfd);
129 | return 0;
132 | ./makesym
133 | nm foo
134 | 00012345 A dummy_symbol
136 Many formats cannot represent arbitrary symbol information; for
137 instance, the <<a.out>> object format does not allow an
138 arbitrary number of sections. A symbol pointing to a section
139 which is not one of <<.text>>, <<.data>> or <<.bss>> cannot
140 be described.
142 INODE
143 Mini Symbols, typedef asymbol, Writing Symbols, Symbols
144 SUBSECTION
145 Mini Symbols
147 Mini symbols provide read-only access to the symbol table.
148 They use less memory space, but require more time to access.
149 They can be useful for tools like nm or objdump, which may
150 have to handle symbol tables of extremely large executables.
152 The <<bfd_read_minisymbols>> function will read the symbols
153 into memory in an internal form. It will return a <<void *>>
154 pointer to a block of memory, a symbol count, and the size of
155 each symbol. The pointer is allocated using <<malloc>>, and
156 should be freed by the caller when it is no longer needed.
158 The function <<bfd_minisymbol_to_symbol>> will take a pointer
159 to a minisymbol, and a pointer to a structure returned by
160 <<bfd_make_empty_symbol>>, and return a <<asymbol>> structure.
161 The return value may or may not be the same as the value from
162 <<bfd_make_empty_symbol>> which was passed in.
167 DOCDD
168 INODE
169 typedef asymbol, symbol handling functions, Mini Symbols, Symbols
173 SUBSECTION
174 typedef asymbol
176 An <<asymbol>> has the form:
181 CODE_FRAGMENT
184 .typedef struct bfd_symbol
186 . {* A pointer to the BFD which owns the symbol. This information
187 . is necessary so that a back end can work out what additional
188 . information (invisible to the application writer) is carried
189 . with the symbol.
191 . This field is *almost* redundant, since you can use section->owner
192 . instead, except that some symbols point to the global sections
193 . bfd_{abs,com,und}_section. This could be fixed by making
194 . these globals be per-bfd (or per-target-flavor). FIXME. *}
195 . struct bfd *the_bfd; {* Use bfd_asymbol_bfd(sym) to access this field. *}
197 . {* The text of the symbol. The name is left alone, and not copied; the
198 . application may not alter it. *}
199 . const char *name;
201 . {* The value of the symbol. This really should be a union of a
202 . numeric value with a pointer, since some flags indicate that
203 . a pointer to another symbol is stored here. *}
204 . symvalue value;
206 . {* Attributes of a symbol. *}
207 .#define BSF_NO_FLAGS 0x00
209 . {* The symbol has local scope; <<static>> in <<C>>. The value
210 . is the offset into the section of the data. *}
211 .#define BSF_LOCAL 0x01
213 . {* The symbol has global scope; initialized data in <<C>>. The
214 . value is the offset into the section of the data. *}
215 .#define BSF_GLOBAL 0x02
217 . {* The symbol has global scope and is exported. The value is
218 . the offset into the section of the data. *}
219 .#define BSF_EXPORT BSF_GLOBAL {* No real difference. *}
221 . {* A normal C symbol would be one of:
222 . <<BSF_LOCAL>>, <<BSF_FORT_COMM>>, <<BSF_UNDEFINED>> or
223 . <<BSF_GLOBAL>>. *}
225 . {* The symbol is a debugging record. The value has an arbitrary
226 . meaning, unless BSF_DEBUGGING_RELOC is also set. *}
227 .#define BSF_DEBUGGING 0x08
229 . {* The symbol denotes a function entry point. Used in ELF,
230 . perhaps others someday. *}
231 .#define BSF_FUNCTION 0x10
233 . {* Used by the linker. *}
234 .#define BSF_KEEP 0x20
235 .#define BSF_KEEP_G 0x40
237 . {* A weak global symbol, overridable without warnings by
238 . a regular global symbol of the same name. *}
239 .#define BSF_WEAK 0x80
241 . {* This symbol was created to point to a section, e.g. ELF's
242 . STT_SECTION symbols. *}
243 .#define BSF_SECTION_SYM 0x100
245 . {* The symbol used to be a common symbol, but now it is
246 . allocated. *}
247 .#define BSF_OLD_COMMON 0x200
249 . {* The default value for common data. *}
250 .#define BFD_FORT_COMM_DEFAULT_VALUE 0
252 . {* In some files the type of a symbol sometimes alters its
253 . location in an output file - ie in coff a <<ISFCN>> symbol
254 . which is also <<C_EXT>> symbol appears where it was
255 . declared and not at the end of a section. This bit is set
256 . by the target BFD part to convey this information. *}
257 .#define BSF_NOT_AT_END 0x400
259 . {* Signal that the symbol is the label of constructor section. *}
260 .#define BSF_CONSTRUCTOR 0x800
262 . {* Signal that the symbol is a warning symbol. The name is a
263 . warning. The name of the next symbol is the one to warn about;
264 . if a reference is made to a symbol with the same name as the next
265 . symbol, a warning is issued by the linker. *}
266 .#define BSF_WARNING 0x1000
268 . {* Signal that the symbol is indirect. This symbol is an indirect
269 . pointer to the symbol with the same name as the next symbol. *}
270 .#define BSF_INDIRECT 0x2000
272 . {* BSF_FILE marks symbols that contain a file name. This is used
273 . for ELF STT_FILE symbols. *}
274 .#define BSF_FILE 0x4000
276 . {* Symbol is from dynamic linking information. *}
277 .#define BSF_DYNAMIC 0x8000
279 . {* The symbol denotes a data object. Used in ELF, and perhaps
280 . others someday. *}
281 .#define BSF_OBJECT 0x10000
283 . {* This symbol is a debugging symbol. The value is the offset
284 . into the section of the data. BSF_DEBUGGING should be set
285 . as well. *}
286 .#define BSF_DEBUGGING_RELOC 0x20000
288 . {* This symbol is thread local. Used in ELF. *}
289 .#define BSF_THREAD_LOCAL 0x40000
291 . flagword flags;
293 . {* A pointer to the section to which this symbol is
294 . relative. This will always be non NULL, there are special
295 . sections for undefined and absolute symbols. *}
296 . struct bfd_section *section;
298 . {* Back end special data. *}
299 . union
301 . void *p;
302 . bfd_vma i;
304 . udata;
306 .asymbol;
310 #include "bfd.h"
311 #include "sysdep.h"
312 #include "libbfd.h"
313 #include "safe-ctype.h"
314 #include "bfdlink.h"
315 #include "aout/stab_gnu.h"
318 DOCDD
319 INODE
320 symbol handling functions, , typedef asymbol, Symbols
321 SUBSECTION
322 Symbol handling functions
326 FUNCTION
327 bfd_get_symtab_upper_bound
329 DESCRIPTION
330 Return the number of bytes required to store a vector of pointers
331 to <<asymbols>> for all the symbols in the BFD @var{abfd},
332 including a terminal NULL pointer. If there are no symbols in
333 the BFD, then return 0. If an error occurs, return -1.
335 .#define bfd_get_symtab_upper_bound(abfd) \
336 . BFD_SEND (abfd, _bfd_get_symtab_upper_bound, (abfd))
341 FUNCTION
342 bfd_is_local_label
344 SYNOPSIS
345 bfd_boolean bfd_is_local_label (bfd *abfd, asymbol *sym);
347 DESCRIPTION
348 Return TRUE if the given symbol @var{sym} in the BFD @var{abfd} is
349 a compiler generated local label, else return FALSE.
352 bfd_boolean
353 bfd_is_local_label (bfd *abfd, asymbol *sym)
355 /* The BSF_SECTION_SYM check is needed for IA-64, where every label that
356 starts with '.' is local. This would accidentally catch section names
357 if we didn't reject them here. */
358 if ((sym->flags & (BSF_GLOBAL | BSF_WEAK | BSF_FILE | BSF_SECTION_SYM)) != 0)
359 return FALSE;
360 if (sym->name == NULL)
361 return FALSE;
362 return bfd_is_local_label_name (abfd, sym->name);
366 FUNCTION
367 bfd_is_local_label_name
369 SYNOPSIS
370 bfd_boolean bfd_is_local_label_name (bfd *abfd, const char *name);
372 DESCRIPTION
373 Return TRUE if a symbol with the name @var{name} in the BFD
374 @var{abfd} is a compiler generated local label, else return
375 FALSE. This just checks whether the name has the form of a
376 local label.
378 .#define bfd_is_local_label_name(abfd, name) \
379 . BFD_SEND (abfd, _bfd_is_local_label_name, (abfd, name))
384 FUNCTION
385 bfd_is_target_special_symbol
387 SYNOPSIS
388 bfd_boolean bfd_is_target_special_symbol (bfd *abfd, asymbol *sym);
390 DESCRIPTION
391 Return TRUE iff a symbol @var{sym} in the BFD @var{abfd} is something
392 special to the particular target represented by the BFD. Such symbols
393 should normally not be mentioned to the user.
395 .#define bfd_is_target_special_symbol(abfd, sym) \
396 . BFD_SEND (abfd, _bfd_is_target_special_symbol, (abfd, sym))
401 FUNCTION
402 bfd_canonicalize_symtab
404 DESCRIPTION
405 Read the symbols from the BFD @var{abfd}, and fills in
406 the vector @var{location} with pointers to the symbols and
407 a trailing NULL.
408 Return the actual number of symbol pointers, not
409 including the NULL.
411 .#define bfd_canonicalize_symtab(abfd, location) \
412 . BFD_SEND (abfd, _bfd_canonicalize_symtab, (abfd, location))
417 FUNCTION
418 bfd_set_symtab
420 SYNOPSIS
421 bfd_boolean bfd_set_symtab
422 (bfd *abfd, asymbol **location, unsigned int count);
424 DESCRIPTION
425 Arrange that when the output BFD @var{abfd} is closed,
426 the table @var{location} of @var{count} pointers to symbols
427 will be written.
430 bfd_boolean
431 bfd_set_symtab (bfd *abfd, asymbol **location, unsigned int symcount)
433 if (abfd->format != bfd_object || bfd_read_p (abfd))
435 bfd_set_error (bfd_error_invalid_operation);
436 return FALSE;
439 bfd_get_outsymbols (abfd) = location;
440 bfd_get_symcount (abfd) = symcount;
441 return TRUE;
445 FUNCTION
446 bfd_print_symbol_vandf
448 SYNOPSIS
449 void bfd_print_symbol_vandf (bfd *abfd, void *file, asymbol *symbol);
451 DESCRIPTION
452 Print the value and flags of the @var{symbol} supplied to the
453 stream @var{file}.
455 void
456 bfd_print_symbol_vandf (bfd *abfd, void *arg, asymbol *symbol)
458 FILE *file = arg;
460 flagword type = symbol->flags;
462 if (symbol->section != NULL)
463 bfd_fprintf_vma (abfd, file, symbol->value + symbol->section->vma);
464 else
465 bfd_fprintf_vma (abfd, file, symbol->value);
467 /* This presumes that a symbol can not be both BSF_DEBUGGING and
468 BSF_DYNAMIC, nor more than one of BSF_FUNCTION, BSF_FILE, and
469 BSF_OBJECT. */
470 fprintf (file, " %c%c%c%c%c%c%c",
471 ((type & BSF_LOCAL)
472 ? (type & BSF_GLOBAL) ? '!' : 'l'
473 : (type & BSF_GLOBAL) ? 'g' : ' '),
474 (type & BSF_WEAK) ? 'w' : ' ',
475 (type & BSF_CONSTRUCTOR) ? 'C' : ' ',
476 (type & BSF_WARNING) ? 'W' : ' ',
477 (type & BSF_INDIRECT) ? 'I' : ' ',
478 (type & BSF_DEBUGGING) ? 'd' : (type & BSF_DYNAMIC) ? 'D' : ' ',
479 ((type & BSF_FUNCTION)
480 ? 'F'
481 : ((type & BSF_FILE)
482 ? 'f'
483 : ((type & BSF_OBJECT) ? 'O' : ' '))));
487 FUNCTION
488 bfd_make_empty_symbol
490 DESCRIPTION
491 Create a new <<asymbol>> structure for the BFD @var{abfd}
492 and return a pointer to it.
494 This routine is necessary because each back end has private
495 information surrounding the <<asymbol>>. Building your own
496 <<asymbol>> and pointing to it will not create the private
497 information, and will cause problems later on.
499 .#define bfd_make_empty_symbol(abfd) \
500 . BFD_SEND (abfd, _bfd_make_empty_symbol, (abfd))
505 FUNCTION
506 _bfd_generic_make_empty_symbol
508 SYNOPSIS
509 asymbol *_bfd_generic_make_empty_symbol (bfd *);
511 DESCRIPTION
512 Create a new <<asymbol>> structure for the BFD @var{abfd}
513 and return a pointer to it. Used by core file routines,
514 binary back-end and anywhere else where no private info
515 is needed.
518 asymbol *
519 _bfd_generic_make_empty_symbol (bfd *abfd)
521 bfd_size_type amt = sizeof (asymbol);
522 asymbol *new = bfd_zalloc (abfd, amt);
523 if (new)
524 new->the_bfd = abfd;
525 return new;
529 FUNCTION
530 bfd_make_debug_symbol
532 DESCRIPTION
533 Create a new <<asymbol>> structure for the BFD @var{abfd},
534 to be used as a debugging symbol. Further details of its use have
535 yet to be worked out.
537 .#define bfd_make_debug_symbol(abfd,ptr,size) \
538 . BFD_SEND (abfd, _bfd_make_debug_symbol, (abfd, ptr, size))
542 struct section_to_type
544 const char *section;
545 char type;
548 /* Map section names to POSIX/BSD single-character symbol types.
549 This table is probably incomplete. It is sorted for convenience of
550 adding entries. Since it is so short, a linear search is used. */
551 static const struct section_to_type stt[] =
553 {".bss", 'b'},
554 {"code", 't'}, /* MRI .text */
555 {".data", 'd'},
556 {"*DEBUG*", 'N'},
557 {".debug", 'N'}, /* MSVC's .debug (non-standard debug syms) */
558 {".drectve", 'i'}, /* MSVC's .drective section */
559 {".edata", 'e'}, /* MSVC's .edata (export) section */
560 {".fini", 't'}, /* ELF fini section */
561 {".idata", 'i'}, /* MSVC's .idata (import) section */
562 {".init", 't'}, /* ELF init section */
563 {".pdata", 'p'}, /* MSVC's .pdata (stack unwind) section */
564 {".rdata", 'r'}, /* Read only data. */
565 {".rodata", 'r'}, /* Read only data. */
566 {".sbss", 's'}, /* Small BSS (uninitialized data). */
567 {".scommon", 'c'}, /* Small common. */
568 {".sdata", 'g'}, /* Small initialized data. */
569 {".text", 't'},
570 {"vars", 'd'}, /* MRI .data */
571 {"zerovars", 'b'}, /* MRI .bss */
572 {0, 0}
575 /* Return the single-character symbol type corresponding to
576 section S, or '?' for an unknown COFF section.
578 Check for any leading string which matches, so .text5 returns
579 't' as well as .text */
581 static char
582 coff_section_type (const char *s)
584 const struct section_to_type *t;
586 for (t = &stt[0]; t->section; t++)
587 if (!strncmp (s, t->section, strlen (t->section)))
588 return t->type;
590 return '?';
593 /* Return the single-character symbol type corresponding to section
594 SECTION, or '?' for an unknown section. This uses section flags to
595 identify sections.
597 FIXME These types are unhandled: c, i, e, p. If we handled these also,
598 we could perhaps obsolete coff_section_type. */
600 static char
601 decode_section_type (const struct bfd_section *section)
603 if (section->flags & SEC_CODE)
604 return 't';
605 if (section->flags & SEC_DATA)
607 if (section->flags & SEC_READONLY)
608 return 'r';
609 else if (section->flags & SEC_SMALL_DATA)
610 return 'g';
611 else
612 return 'd';
614 if ((section->flags & SEC_HAS_CONTENTS) == 0)
616 if (section->flags & SEC_SMALL_DATA)
617 return 's';
618 else
619 return 'b';
621 if (section->flags & SEC_DEBUGGING)
622 return 'N';
623 if ((section->flags & SEC_HAS_CONTENTS) && (section->flags & SEC_READONLY))
624 return 'n';
626 return '?';
630 FUNCTION
631 bfd_decode_symclass
633 DESCRIPTION
634 Return a character corresponding to the symbol
635 class of @var{symbol}, or '?' for an unknown class.
637 SYNOPSIS
638 int bfd_decode_symclass (asymbol *symbol);
641 bfd_decode_symclass (asymbol *symbol)
643 char c;
645 if (bfd_is_com_section (symbol->section))
646 return 'C';
647 if (bfd_is_und_section (symbol->section))
649 if (symbol->flags & BSF_WEAK)
651 /* If weak, determine if it's specifically an object
652 or non-object weak. */
653 if (symbol->flags & BSF_OBJECT)
654 return 'v';
655 else
656 return 'w';
658 else
659 return 'U';
661 if (bfd_is_ind_section (symbol->section))
662 return 'I';
663 if (symbol->flags & BSF_WEAK)
665 /* If weak, determine if it's specifically an object
666 or non-object weak. */
667 if (symbol->flags & BSF_OBJECT)
668 return 'V';
669 else
670 return 'W';
672 if (!(symbol->flags & (BSF_GLOBAL | BSF_LOCAL)))
673 return '?';
675 if (bfd_is_abs_section (symbol->section))
676 c = 'a';
677 else if (symbol->section)
679 c = coff_section_type (symbol->section->name);
680 if (c == '?')
681 c = decode_section_type (symbol->section);
683 else
684 return '?';
685 if (symbol->flags & BSF_GLOBAL)
686 c = TOUPPER (c);
687 return c;
689 /* We don't have to handle these cases just yet, but we will soon:
690 N_SETV: 'v';
691 N_SETA: 'l';
692 N_SETT: 'x';
693 N_SETD: 'z';
694 N_SETB: 's';
695 N_INDR: 'i';
700 FUNCTION
701 bfd_is_undefined_symclass
703 DESCRIPTION
704 Returns non-zero if the class symbol returned by
705 bfd_decode_symclass represents an undefined symbol.
706 Returns zero otherwise.
708 SYNOPSIS
709 bfd_boolean bfd_is_undefined_symclass (int symclass);
712 bfd_boolean
713 bfd_is_undefined_symclass (int symclass)
715 return symclass == 'U' || symclass == 'w' || symclass == 'v';
719 FUNCTION
720 bfd_symbol_info
722 DESCRIPTION
723 Fill in the basic info about symbol that nm needs.
724 Additional info may be added by the back-ends after
725 calling this function.
727 SYNOPSIS
728 void bfd_symbol_info (asymbol *symbol, symbol_info *ret);
731 void
732 bfd_symbol_info (asymbol *symbol, symbol_info *ret)
734 ret->type = bfd_decode_symclass (symbol);
736 if (bfd_is_undefined_symclass (ret->type))
737 ret->value = 0;
738 else
739 ret->value = symbol->value + symbol->section->vma;
741 ret->name = symbol->name;
745 FUNCTION
746 bfd_copy_private_symbol_data
748 SYNOPSIS
749 bfd_boolean bfd_copy_private_symbol_data
750 (bfd *ibfd, asymbol *isym, bfd *obfd, asymbol *osym);
752 DESCRIPTION
753 Copy private symbol information from @var{isym} in the BFD
754 @var{ibfd} to the symbol @var{osym} in the BFD @var{obfd}.
755 Return <<TRUE>> on success, <<FALSE>> on error. Possible error
756 returns are:
758 o <<bfd_error_no_memory>> -
759 Not enough memory exists to create private data for @var{osec}.
761 .#define bfd_copy_private_symbol_data(ibfd, isymbol, obfd, osymbol) \
762 . BFD_SEND (obfd, _bfd_copy_private_symbol_data, \
763 . (ibfd, isymbol, obfd, osymbol))
767 /* The generic version of the function which returns mini symbols.
768 This is used when the backend does not provide a more efficient
769 version. It just uses BFD asymbol structures as mini symbols. */
771 long
772 _bfd_generic_read_minisymbols (bfd *abfd,
773 bfd_boolean dynamic,
774 void **minisymsp,
775 unsigned int *sizep)
777 long storage;
778 asymbol **syms = NULL;
779 long symcount;
781 if (dynamic)
782 storage = bfd_get_dynamic_symtab_upper_bound (abfd);
783 else
784 storage = bfd_get_symtab_upper_bound (abfd);
785 if (storage < 0)
786 goto error_return;
787 if (storage == 0)
788 return 0;
790 syms = bfd_malloc (storage);
791 if (syms == NULL)
792 goto error_return;
794 if (dynamic)
795 symcount = bfd_canonicalize_dynamic_symtab (abfd, syms);
796 else
797 symcount = bfd_canonicalize_symtab (abfd, syms);
798 if (symcount < 0)
799 goto error_return;
801 *minisymsp = syms;
802 *sizep = sizeof (asymbol *);
803 return symcount;
805 error_return:
806 bfd_set_error (bfd_error_no_symbols);
807 if (syms != NULL)
808 free (syms);
809 return -1;
812 /* The generic version of the function which converts a minisymbol to
813 an asymbol. We don't worry about the sym argument we are passed;
814 we just return the asymbol the minisymbol points to. */
816 asymbol *
817 _bfd_generic_minisymbol_to_symbol (bfd *abfd ATTRIBUTE_UNUSED,
818 bfd_boolean dynamic ATTRIBUTE_UNUSED,
819 const void *minisym,
820 asymbol *sym ATTRIBUTE_UNUSED)
822 return *(asymbol **) minisym;
825 /* Look through stabs debugging information in .stab and .stabstr
826 sections to find the source file and line closest to a desired
827 location. This is used by COFF and ELF targets. It sets *pfound
828 to TRUE if it finds some information. The *pinfo field is used to
829 pass cached information in and out of this routine; this first time
830 the routine is called for a BFD, *pinfo should be NULL. The value
831 placed in *pinfo should be saved with the BFD, and passed back each
832 time this function is called. */
834 /* We use a cache by default. */
836 #define ENABLE_CACHING
838 /* We keep an array of indexentry structures to record where in the
839 stabs section we should look to find line number information for a
840 particular address. */
842 struct indexentry
844 bfd_vma val;
845 bfd_byte *stab;
846 bfd_byte *str;
847 char *directory_name;
848 char *file_name;
849 char *function_name;
852 /* Compare two indexentry structures. This is called via qsort. */
854 static int
855 cmpindexentry (const void *a, const void *b)
857 const struct indexentry *contestantA = a;
858 const struct indexentry *contestantB = b;
860 if (contestantA->val < contestantB->val)
861 return -1;
862 else if (contestantA->val > contestantB->val)
863 return 1;
864 else
865 return 0;
868 /* A pointer to this structure is stored in *pinfo. */
870 struct stab_find_info
872 /* The .stab section. */
873 asection *stabsec;
874 /* The .stabstr section. */
875 asection *strsec;
876 /* The contents of the .stab section. */
877 bfd_byte *stabs;
878 /* The contents of the .stabstr section. */
879 bfd_byte *strs;
881 /* A table that indexes stabs by memory address. */
882 struct indexentry *indextable;
883 /* The number of entries in indextable. */
884 int indextablesize;
886 #ifdef ENABLE_CACHING
887 /* Cached values to restart quickly. */
888 struct indexentry *cached_indexentry;
889 bfd_vma cached_offset;
890 bfd_byte *cached_stab;
891 char *cached_file_name;
892 #endif
894 /* Saved ptr to malloc'ed filename. */
895 char *filename;
898 bfd_boolean
899 _bfd_stab_section_find_nearest_line (bfd *abfd,
900 asymbol **symbols,
901 asection *section,
902 bfd_vma offset,
903 bfd_boolean *pfound,
904 const char **pfilename,
905 const char **pfnname,
906 unsigned int *pline,
907 void **pinfo)
909 struct stab_find_info *info;
910 bfd_size_type stabsize, strsize;
911 bfd_byte *stab, *str;
912 bfd_byte *last_stab = NULL;
913 bfd_size_type stroff;
914 struct indexentry *indexentry;
915 char *file_name;
916 char *directory_name;
917 int saw_fun;
918 bfd_boolean saw_line, saw_func;
920 *pfound = FALSE;
921 *pfilename = bfd_get_filename (abfd);
922 *pfnname = NULL;
923 *pline = 0;
925 /* Stabs entries use a 12 byte format:
926 4 byte string table index
927 1 byte stab type
928 1 byte stab other field
929 2 byte stab desc field
930 4 byte stab value
931 FIXME: This will have to change for a 64 bit object format.
933 The stabs symbols are divided into compilation units. For the
934 first entry in each unit, the type of 0, the value is the length
935 of the string table for this unit, and the desc field is the
936 number of stabs symbols for this unit. */
938 #define STRDXOFF (0)
939 #define TYPEOFF (4)
940 #define OTHEROFF (5)
941 #define DESCOFF (6)
942 #define VALOFF (8)
943 #define STABSIZE (12)
945 info = *pinfo;
946 if (info != NULL)
948 if (info->stabsec == NULL || info->strsec == NULL)
950 /* No stabs debugging information. */
951 return TRUE;
954 stabsize = (info->stabsec->rawsize
955 ? info->stabsec->rawsize
956 : info->stabsec->size);
957 strsize = (info->strsec->rawsize
958 ? info->strsec->rawsize
959 : info->strsec->size);
961 else
963 long reloc_size, reloc_count;
964 arelent **reloc_vector;
965 int i;
966 char *name;
967 char *function_name;
968 bfd_size_type amt = sizeof *info;
970 info = bfd_zalloc (abfd, amt);
971 if (info == NULL)
972 return FALSE;
974 /* FIXME: When using the linker --split-by-file or
975 --split-by-reloc options, it is possible for the .stab and
976 .stabstr sections to be split. We should handle that. */
978 info->stabsec = bfd_get_section_by_name (abfd, ".stab");
979 info->strsec = bfd_get_section_by_name (abfd, ".stabstr");
981 if (info->stabsec == NULL || info->strsec == NULL)
983 /* No stabs debugging information. Set *pinfo so that we
984 can return quickly in the info != NULL case above. */
985 *pinfo = info;
986 return TRUE;
989 stabsize = (info->stabsec->rawsize
990 ? info->stabsec->rawsize
991 : info->stabsec->size);
992 strsize = (info->strsec->rawsize
993 ? info->strsec->rawsize
994 : info->strsec->size);
996 info->stabs = bfd_alloc (abfd, stabsize);
997 info->strs = bfd_alloc (abfd, strsize);
998 if (info->stabs == NULL || info->strs == NULL)
999 return FALSE;
1001 if (! bfd_get_section_contents (abfd, info->stabsec, info->stabs,
1002 0, stabsize)
1003 || ! bfd_get_section_contents (abfd, info->strsec, info->strs,
1004 0, strsize))
1005 return FALSE;
1007 /* If this is a relocatable object file, we have to relocate
1008 the entries in .stab. This should always be simple 32 bit
1009 relocations against symbols defined in this object file, so
1010 this should be no big deal. */
1011 reloc_size = bfd_get_reloc_upper_bound (abfd, info->stabsec);
1012 if (reloc_size < 0)
1013 return FALSE;
1014 reloc_vector = bfd_malloc (reloc_size);
1015 if (reloc_vector == NULL && reloc_size != 0)
1016 return FALSE;
1017 reloc_count = bfd_canonicalize_reloc (abfd, info->stabsec, reloc_vector,
1018 symbols);
1019 if (reloc_count < 0)
1021 if (reloc_vector != NULL)
1022 free (reloc_vector);
1023 return FALSE;
1025 if (reloc_count > 0)
1027 arelent **pr;
1029 for (pr = reloc_vector; *pr != NULL; pr++)
1031 arelent *r;
1032 unsigned long val;
1033 asymbol *sym;
1035 r = *pr;
1036 /* Ignore R_*_NONE relocs. */
1037 if (r->howto->dst_mask == 0)
1038 continue;
1040 if (r->howto->rightshift != 0
1041 || r->howto->size != 2
1042 || r->howto->bitsize != 32
1043 || r->howto->pc_relative
1044 || r->howto->bitpos != 0
1045 || r->howto->dst_mask != 0xffffffff)
1047 (*_bfd_error_handler)
1048 (_("Unsupported .stab relocation"));
1049 bfd_set_error (bfd_error_invalid_operation);
1050 if (reloc_vector != NULL)
1051 free (reloc_vector);
1052 return FALSE;
1055 val = bfd_get_32 (abfd, info->stabs + r->address);
1056 val &= r->howto->src_mask;
1057 sym = *r->sym_ptr_ptr;
1058 val += sym->value + sym->section->vma + r->addend;
1059 bfd_put_32 (abfd, (bfd_vma) val, info->stabs + r->address);
1063 if (reloc_vector != NULL)
1064 free (reloc_vector);
1066 /* First time through this function, build a table matching
1067 function VM addresses to stabs, then sort based on starting
1068 VM address. Do this in two passes: once to count how many
1069 table entries we'll need, and a second to actually build the
1070 table. */
1072 info->indextablesize = 0;
1073 saw_fun = 1;
1074 for (stab = info->stabs; stab < info->stabs + stabsize; stab += STABSIZE)
1076 if (stab[TYPEOFF] == (bfd_byte) N_SO)
1078 /* N_SO with null name indicates EOF */
1079 if (bfd_get_32 (abfd, stab + STRDXOFF) == 0)
1080 continue;
1082 /* if we did not see a function def, leave space for one. */
1083 if (saw_fun == 0)
1084 ++info->indextablesize;
1086 saw_fun = 0;
1088 /* two N_SO's in a row is a filename and directory. Skip */
1089 if (stab + STABSIZE < info->stabs + stabsize
1090 && *(stab + STABSIZE + TYPEOFF) == (bfd_byte) N_SO)
1092 stab += STABSIZE;
1095 else if (stab[TYPEOFF] == (bfd_byte) N_FUN)
1097 saw_fun = 1;
1098 ++info->indextablesize;
1102 if (saw_fun == 0)
1103 ++info->indextablesize;
1105 if (info->indextablesize == 0)
1106 return TRUE;
1107 ++info->indextablesize;
1109 amt = info->indextablesize;
1110 amt *= sizeof (struct indexentry);
1111 info->indextable = bfd_alloc (abfd, amt);
1112 if (info->indextable == NULL)
1113 return FALSE;
1115 file_name = NULL;
1116 directory_name = NULL;
1117 saw_fun = 1;
1119 for (i = 0, stroff = 0, stab = info->stabs, str = info->strs;
1120 i < info->indextablesize && stab < info->stabs + stabsize;
1121 stab += STABSIZE)
1123 switch (stab[TYPEOFF])
1125 case 0:
1126 /* This is the first entry in a compilation unit. */
1127 if ((bfd_size_type) ((info->strs + strsize) - str) < stroff)
1128 break;
1129 str += stroff;
1130 stroff = bfd_get_32 (abfd, stab + VALOFF);
1131 break;
1133 case N_SO:
1134 /* The main file name. */
1136 /* The following code creates a new indextable entry with
1137 a NULL function name if there were no N_FUNs in a file.
1138 Note that a N_SO without a file name is an EOF and
1139 there could be 2 N_SO following it with the new filename
1140 and directory. */
1141 if (saw_fun == 0)
1143 info->indextable[i].val = bfd_get_32 (abfd, last_stab + VALOFF);
1144 info->indextable[i].stab = last_stab;
1145 info->indextable[i].str = str;
1146 info->indextable[i].directory_name = directory_name;
1147 info->indextable[i].file_name = file_name;
1148 info->indextable[i].function_name = NULL;
1149 ++i;
1151 saw_fun = 0;
1153 file_name = (char *) str + bfd_get_32 (abfd, stab + STRDXOFF);
1154 if (*file_name == '\0')
1156 directory_name = NULL;
1157 file_name = NULL;
1158 saw_fun = 1;
1160 else
1162 last_stab = stab;
1163 if (stab + STABSIZE >= info->stabs + stabsize
1164 || *(stab + STABSIZE + TYPEOFF) != (bfd_byte) N_SO)
1166 directory_name = NULL;
1168 else
1170 /* Two consecutive N_SOs are a directory and a
1171 file name. */
1172 stab += STABSIZE;
1173 directory_name = file_name;
1174 file_name = ((char *) str
1175 + bfd_get_32 (abfd, stab + STRDXOFF));
1178 break;
1180 case N_SOL:
1181 /* The name of an include file. */
1182 file_name = (char *) str + bfd_get_32 (abfd, stab + STRDXOFF);
1183 break;
1185 case N_FUN:
1186 /* A function name. */
1187 saw_fun = 1;
1188 name = (char *) str + bfd_get_32 (abfd, stab + STRDXOFF);
1190 if (*name == '\0')
1191 name = NULL;
1193 function_name = name;
1195 if (name == NULL)
1196 continue;
1198 info->indextable[i].val = bfd_get_32 (abfd, stab + VALOFF);
1199 info->indextable[i].stab = stab;
1200 info->indextable[i].str = str;
1201 info->indextable[i].directory_name = directory_name;
1202 info->indextable[i].file_name = file_name;
1203 info->indextable[i].function_name = function_name;
1204 ++i;
1205 break;
1209 if (saw_fun == 0)
1211 info->indextable[i].val = bfd_get_32 (abfd, last_stab + VALOFF);
1212 info->indextable[i].stab = last_stab;
1213 info->indextable[i].str = str;
1214 info->indextable[i].directory_name = directory_name;
1215 info->indextable[i].file_name = file_name;
1216 info->indextable[i].function_name = NULL;
1217 ++i;
1220 info->indextable[i].val = (bfd_vma) -1;
1221 info->indextable[i].stab = info->stabs + stabsize;
1222 info->indextable[i].str = str;
1223 info->indextable[i].directory_name = NULL;
1224 info->indextable[i].file_name = NULL;
1225 info->indextable[i].function_name = NULL;
1226 ++i;
1228 info->indextablesize = i;
1229 qsort (info->indextable, (size_t) i, sizeof (struct indexentry),
1230 cmpindexentry);
1232 *pinfo = info;
1235 /* We are passed a section relative offset. The offsets in the
1236 stabs information are absolute. */
1237 offset += bfd_get_section_vma (abfd, section);
1239 #ifdef ENABLE_CACHING
1240 if (info->cached_indexentry != NULL
1241 && offset >= info->cached_offset
1242 && offset < (info->cached_indexentry + 1)->val)
1244 stab = info->cached_stab;
1245 indexentry = info->cached_indexentry;
1246 file_name = info->cached_file_name;
1248 else
1249 #endif
1251 long low, high;
1252 long mid = -1;
1254 /* Cache non-existent or invalid. Do binary search on
1255 indextable. */
1256 indexentry = NULL;
1258 low = 0;
1259 high = info->indextablesize - 1;
1260 while (low != high)
1262 mid = (high + low) / 2;
1263 if (offset >= info->indextable[mid].val
1264 && offset < info->indextable[mid + 1].val)
1266 indexentry = &info->indextable[mid];
1267 break;
1270 if (info->indextable[mid].val > offset)
1271 high = mid;
1272 else
1273 low = mid + 1;
1276 if (indexentry == NULL)
1277 return TRUE;
1279 stab = indexentry->stab + STABSIZE;
1280 file_name = indexentry->file_name;
1283 directory_name = indexentry->directory_name;
1284 str = indexentry->str;
1286 saw_line = FALSE;
1287 saw_func = FALSE;
1288 for (; stab < (indexentry+1)->stab; stab += STABSIZE)
1290 bfd_boolean done;
1291 bfd_vma val;
1293 done = FALSE;
1295 switch (stab[TYPEOFF])
1297 case N_SOL:
1298 /* The name of an include file. */
1299 val = bfd_get_32 (abfd, stab + VALOFF);
1300 if (val <= offset)
1302 file_name = (char *) str + bfd_get_32 (abfd, stab + STRDXOFF);
1303 *pline = 0;
1305 break;
1307 case N_SLINE:
1308 case N_DSLINE:
1309 case N_BSLINE:
1310 /* A line number. If the function was specified, then the value
1311 is relative to the start of the function. Otherwise, the
1312 value is an absolute address. */
1313 val = ((indexentry->function_name ? indexentry->val : 0)
1314 + bfd_get_32 (abfd, stab + VALOFF));
1315 /* If this line starts before our desired offset, or if it's
1316 the first line we've been able to find, use it. The
1317 !saw_line check works around a bug in GCC 2.95.3, which emits
1318 the first N_SLINE late. */
1319 if (!saw_line || val <= offset)
1321 *pline = bfd_get_16 (abfd, stab + DESCOFF);
1323 #ifdef ENABLE_CACHING
1324 info->cached_stab = stab;
1325 info->cached_offset = val;
1326 info->cached_file_name = file_name;
1327 info->cached_indexentry = indexentry;
1328 #endif
1330 if (val > offset)
1331 done = TRUE;
1332 saw_line = TRUE;
1333 break;
1335 case N_FUN:
1336 case N_SO:
1337 if (saw_func || saw_line)
1338 done = TRUE;
1339 saw_func = TRUE;
1340 break;
1343 if (done)
1344 break;
1347 *pfound = TRUE;
1349 if (file_name == NULL || IS_ABSOLUTE_PATH (file_name)
1350 || directory_name == NULL)
1351 *pfilename = file_name;
1352 else
1354 size_t dirlen;
1356 dirlen = strlen (directory_name);
1357 if (info->filename == NULL
1358 || strncmp (info->filename, directory_name, dirlen) != 0
1359 || strcmp (info->filename + dirlen, file_name) != 0)
1361 size_t len;
1363 if (info->filename != NULL)
1364 free (info->filename);
1365 len = strlen (file_name) + 1;
1366 info->filename = bfd_malloc (dirlen + len);
1367 if (info->filename == NULL)
1368 return FALSE;
1369 memcpy (info->filename, directory_name, dirlen);
1370 memcpy (info->filename + dirlen, file_name, len);
1373 *pfilename = info->filename;
1376 if (indexentry->function_name != NULL)
1378 char *s;
1380 /* This will typically be something like main:F(0,1), so we want
1381 to clobber the colon. It's OK to change the name, since the
1382 string is in our own local storage anyhow. */
1383 s = strchr (indexentry->function_name, ':');
1384 if (s != NULL)
1385 *s = '\0';
1387 *pfnname = indexentry->function_name;
1390 return TRUE;