1 /* Generic BFD support for file formats.
2 Copyright (C) 1990-2024 Free Software Foundation, Inc.
3 Written by Cygnus Support.
5 This file is part of BFD, the Binary File Descriptor library.
7 This program is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 3 of the License, or
10 (at your option) any later version.
12 This program is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with this program; if not, write to the Free Software
19 Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston,
20 MA 02110-1301, USA. */
27 A format is a BFD concept of high level file contents type. The
28 formats supported by BFD are:
32 The BFD may contain data, symbols, relocations and debug info.
36 The BFD contains other BFDs and an optional index.
40 The BFD contains the result of an executable core dump.
50 /* IMPORT from targets.c. */
51 extern const size_t _bfd_target_vector_entries
;
58 bool bfd_check_format (bfd *abfd, bfd_format format);
61 Verify if the file attached to the BFD @var{abfd} is compatible
62 with the format @var{format} (i.e., one of <<bfd_object>>,
63 <<bfd_archive>> or <<bfd_core>>).
65 If the BFD has been set to a specific target before the
66 call, only the named target and format combination is
67 checked. If the target has not been set, or has been set to
68 <<default>>, then all the known target backends is
69 interrogated to determine a match. If the default target
70 matches, it is used. If not, exactly one target must recognize
71 the file, or an error results.
73 The function returns <<TRUE>> on success, otherwise <<FALSE>>
74 with one of the following error codes:
76 o <<bfd_error_invalid_operation>> -
77 if <<format>> is not one of <<bfd_object>>, <<bfd_archive>> or
80 o <<bfd_error_system_call>> -
81 if an error occured during a read - even some file mismatches
82 can cause bfd_error_system_calls.
84 o <<file_not_recognised>> -
85 none of the backends recognised the file format.
87 o <<bfd_error_file_ambiguously_recognized>> -
88 more than one backend recognised the file format.
90 When calling bfd_check_format (or bfd_check_format_matches),
91 any underlying file descriptor will be kept open for the
92 duration of the call. This is done to avoid races when
93 another thread calls bfd_cache_close_all. In this scenario,
94 the thread calling bfd_check_format must call bfd_cache_close
99 bfd_check_format (bfd
*abfd
, bfd_format format
)
101 return bfd_check_format_matches (abfd
, format
, NULL
);
109 const struct bfd_iovec
*iovec
;
111 const struct bfd_arch_info
*arch_info
;
112 const struct bfd_build_id
*build_id
;
114 struct bfd_section
*sections
;
115 struct bfd_section
*section_last
;
116 unsigned int section_count
;
117 unsigned int section_id
;
118 unsigned int symcount
;
120 bfd_vma start_address
;
121 struct bfd_hash_table section_htab
;
124 /* When testing an object for compatibility with a particular target
125 back-end, the back-end object_p function needs to set up certain
126 fields in the bfd on successfully recognizing the object. This
127 typically happens in a piecemeal fashion, with failures possible at
128 many points. On failure, the bfd is supposed to be restored to its
129 initial state, which is virtually impossible. However, restoring a
130 subset of the bfd state works in practice. This function stores
134 bfd_preserve_save (bfd
*abfd
, struct bfd_preserve
*preserve
,
137 preserve
->tdata
= abfd
->tdata
.any
;
138 preserve
->arch_info
= abfd
->arch_info
;
139 preserve
->flags
= abfd
->flags
;
140 preserve
->iovec
= abfd
->iovec
;
141 preserve
->iostream
= abfd
->iostream
;
142 preserve
->sections
= abfd
->sections
;
143 preserve
->section_last
= abfd
->section_last
;
144 preserve
->section_count
= abfd
->section_count
;
145 preserve
->section_id
= _bfd_section_id
;
146 preserve
->symcount
= abfd
->symcount
;
147 preserve
->read_only
= abfd
->read_only
;
148 preserve
->start_address
= abfd
->start_address
;
149 preserve
->section_htab
= abfd
->section_htab
;
150 preserve
->marker
= bfd_alloc (abfd
, 1);
151 preserve
->build_id
= abfd
->build_id
;
152 preserve
->cleanup
= cleanup
;
153 if (preserve
->marker
== NULL
)
156 return bfd_hash_table_init (&abfd
->section_htab
, bfd_section_hash_newfunc
,
157 sizeof (struct section_hash_entry
));
160 /* A back-end object_p function may flip a bfd from file backed to
161 in-memory, eg. pe_ILF_object_p. In that case to restore the
162 original IO state we need to reopen the file. Conversely, if we
163 are restoring a previously matched pe ILF format and have been
164 checking further target matches using file IO then we need to close
165 the file and detach the bfd from the cache lru list. */
168 io_reinit (bfd
*abfd
, struct bfd_preserve
*preserve
)
170 if (abfd
->iovec
!= preserve
->iovec
)
172 /* Handle file backed to in-memory transition. bfd_cache_close
173 won't do anything unless abfd->iovec is the cache_iovec.
174 Don't be tempted to call iovec->bclose here. We don't want
175 to call memory_bclose, which would free the bim. The bim
176 must be kept if bfd_check_format_matches is going to decide
177 later that the PE format needing it is in fact the correct
179 bfd_cache_close (abfd
);
180 abfd
->iovec
= preserve
->iovec
;
181 abfd
->iostream
= preserve
->iostream
;
183 /* Handle in-memory to file backed transition. */
184 if ((abfd
->flags
& BFD_CLOSED_BY_CACHE
) != 0
185 && (abfd
->flags
& BFD_IN_MEMORY
) != 0
186 && (preserve
->flags
& BFD_CLOSED_BY_CACHE
) == 0
187 && (preserve
->flags
& BFD_IN_MEMORY
) == 0)
188 bfd_open_file (abfd
);
190 abfd
->flags
= preserve
->flags
;
193 /* Clear out a subset of BFD state. */
196 bfd_reinit (bfd
*abfd
, unsigned int section_id
,
197 struct bfd_preserve
*preserve
, bfd_cleanup cleanup
)
199 _bfd_section_id
= section_id
;
202 abfd
->tdata
.any
= NULL
;
203 abfd
->arch_info
= &bfd_default_arch_struct
;
204 io_reinit (abfd
, preserve
);
207 abfd
->start_address
= 0;
208 abfd
->build_id
= NULL
;
209 bfd_section_list_clear (abfd
);
212 /* Restores bfd state saved by bfd_preserve_save. */
215 bfd_preserve_restore (bfd
*abfd
, struct bfd_preserve
*preserve
)
217 bfd_hash_table_free (&abfd
->section_htab
);
219 abfd
->tdata
.any
= preserve
->tdata
;
220 abfd
->arch_info
= preserve
->arch_info
;
221 io_reinit (abfd
, preserve
);
222 abfd
->section_htab
= preserve
->section_htab
;
223 abfd
->sections
= preserve
->sections
;
224 abfd
->section_last
= preserve
->section_last
;
225 abfd
->section_count
= preserve
->section_count
;
226 _bfd_section_id
= preserve
->section_id
;
227 abfd
->symcount
= preserve
->symcount
;
228 abfd
->read_only
= preserve
->read_only
;
229 abfd
->start_address
= preserve
->start_address
;
230 abfd
->build_id
= preserve
->build_id
;
232 /* bfd_release frees all memory more recently bfd_alloc'd than
233 its arg, as well as its arg. */
234 bfd_release (abfd
, preserve
->marker
);
235 preserve
->marker
= NULL
;
236 return preserve
->cleanup
;
239 /* Called when the bfd state saved by bfd_preserve_save is no longer
243 bfd_preserve_finish (bfd
*abfd ATTRIBUTE_UNUSED
, struct bfd_preserve
*preserve
)
245 if (preserve
->cleanup
)
247 /* Run the cleanup, assuming that all it will need is the
248 tdata at the time the cleanup was returned. */
249 void *tdata
= abfd
->tdata
.any
;
250 abfd
->tdata
.any
= preserve
->tdata
;
251 preserve
->cleanup (abfd
);
252 abfd
->tdata
.any
= tdata
;
254 /* It would be nice to be able to free more memory here, eg. old
255 tdata, but that's not possible since these blocks are sitting
256 inside bfd_alloc'd memory. The section hash is on a separate
258 bfd_hash_table_free (&preserve
->section_htab
);
259 preserve
->marker
= NULL
;
263 print_warnmsg (struct per_xvec_message
**list
)
265 for (struct per_xvec_message
*warn
= *list
; warn
; warn
= warn
->next
)
266 _bfd_error_handler ("%s", warn
->message
);
270 clear_warnmsg (struct per_xvec_message
**list
)
272 struct per_xvec_message
*warn
= *list
;
275 struct per_xvec_message
*next
= warn
->next
;
282 /* Free all the storage in LIST. Note that the first element of LIST
283 is special and is assumed to be stack-allocated. TARG is used for
284 re-issuing warning messages. If TARG is PER_XVEC_NO_TARGET, then
285 it acts like a sort of wildcard -- messages are only reissued if
286 they are all associated with a single BFD target, regardless of
287 which one it is. If TARG is anything else, then only messages
288 associated with TARG are emitted. */
291 print_and_clear_messages (struct per_xvec_messages
*list
,
292 const bfd_target
*targ
)
294 struct per_xvec_messages
*iter
= list
;
296 if (targ
== PER_XVEC_NO_TARGET
&& list
->next
== NULL
)
297 print_warnmsg (&list
->messages
);
301 struct per_xvec_messages
*next
= iter
->next
;
303 if (iter
->targ
== targ
)
304 print_warnmsg (&iter
->messages
);
305 clear_warnmsg (&iter
->messages
);
312 /* This a copy of lto_section defined in GCC (lto-streamer.h). */
316 int16_t major_version
;
317 int16_t minor_version
;
318 unsigned char slim_object
;
320 /* Flags is a private field that is not defined publicly. */
324 /* Set lto_type in ABFD. */
327 bfd_set_lto_type (bfd
*abfd ATTRIBUTE_UNUSED
)
329 #if BFD_SUPPORTS_PLUGINS
330 if (abfd
->format
== bfd_object
331 && abfd
->lto_type
== lto_non_object
332 && (abfd
->flags
& (DYNAMIC
| EXEC_P
)) == 0)
335 enum bfd_lto_object_type type
= lto_non_ir_object
;
336 struct lto_section lsection
;
337 /* GCC uses .gnu.lto_.lto.<some_hash> as a LTO bytecode information
339 for (sec
= abfd
->sections
; sec
!= NULL
; sec
= sec
->next
)
340 if (startswith (sec
->name
, ".gnu.lto_.lto.")
341 && bfd_get_section_contents (abfd
, sec
, &lsection
, 0,
342 sizeof (struct lto_section
)))
344 if (lsection
.slim_object
)
345 type
= lto_slim_ir_object
;
347 type
= lto_fat_ir_object
;
351 abfd
->lto_type
= type
;
358 bfd_check_format_matches
361 bool bfd_check_format_matches
362 (bfd *abfd, bfd_format format, char ***matching);
365 Like <<bfd_check_format>>, except when it returns FALSE with
366 <<bfd_errno>> set to <<bfd_error_file_ambiguously_recognized>>. In that
367 case, if @var{matching} is not NULL, it will be filled in with
368 a NULL-terminated list of the names of the formats that matched,
369 allocated with <<malloc>>.
370 Then the user may choose a format and try again.
372 When done with the list that @var{matching} points to, the caller
377 bfd_check_format_matches (bfd
*abfd
, bfd_format format
, char ***matching
)
379 extern const bfd_target binary_vec
;
380 #if BFD_SUPPORTS_PLUGINS
381 extern const bfd_target plugin_vec
;
383 const bfd_target
* const *target
;
384 const bfd_target
**matching_vector
= NULL
;
385 const bfd_target
*save_targ
, *right_targ
, *ar_right_targ
, *match_targ
;
386 int match_count
, best_count
, best_match
;
388 unsigned int initial_section_id
= _bfd_section_id
;
389 struct bfd_preserve preserve
, preserve_match
;
390 bfd_cleanup cleanup
= NULL
;
391 struct per_xvec_messages messages
= { abfd
, PER_XVEC_NO_TARGET
, NULL
, NULL
};
392 struct per_xvec_messages
*orig_messages
;
393 bool old_in_format_matches
;
395 if (matching
!= NULL
)
398 if (!bfd_read_p (abfd
)
399 || (unsigned int) abfd
->format
>= (unsigned int) bfd_type_end
)
401 bfd_set_error (bfd_error_invalid_operation
);
405 if (abfd
->format
!= bfd_unknown
)
407 bfd_set_lto_type (abfd
);
408 return abfd
->format
== format
;
411 if (matching
!= NULL
|| *bfd_associated_vector
!= NULL
)
415 amt
= sizeof (*matching_vector
) * 2 * _bfd_target_vector_entries
;
416 matching_vector
= (const bfd_target
**) bfd_malloc (amt
);
417 if (!matching_vector
)
421 /* Avoid clashes with bfd_cache_close_all running in another
423 if (!bfd_cache_set_uncloseable (abfd
, true, &old_in_format_matches
))
426 /* Presume the answer is yes. */
427 abfd
->format
= format
;
428 save_targ
= abfd
->xvec
;
430 /* Don't report errors on recursive calls checking the first element
432 orig_messages
= _bfd_set_error_handler_caching (&messages
);
434 preserve_match
.marker
= NULL
;
435 if (!bfd_preserve_save (abfd
, &preserve
, NULL
))
438 /* If the target type was explicitly specified, just check that target. */
439 if (!abfd
->target_defaulted
)
441 if (bfd_seek (abfd
, 0, SEEK_SET
) != 0) /* rewind! */
444 cleanup
= BFD_SEND_FMT (abfd
, _bfd_check_format
, (abfd
));
449 /* For a long time the code has dropped through to check all
450 targets if the specified target was wrong. I don't know why,
451 and I'm reluctant to change it. However, in the case of an
452 archive, it can cause problems. If the specified target does
453 not permit archives (e.g., the binary target), then we should
454 not allow some other target to recognize it as an archive, but
455 should instead allow the specified target to recognize it as an
456 object. When I first made this change, it broke the PE target,
457 because the specified pei-i386 target did not recognize the
458 actual pe-i386 archive. Since there may be other problems of
459 this sort, I changed this test to check only for the binary
461 if (format
== bfd_archive
&& save_targ
== &binary_vec
)
465 /* Since the target type was defaulted, check them all in the hope
466 that one will be uniquely recognized. */
468 ar_right_targ
= NULL
;
473 ar_match_index
= _bfd_target_vector_entries
;
475 for (target
= bfd_target_vector
; *target
!= NULL
; target
++)
479 /* The binary target matches anything, so don't return it when
480 searching. Don't match the plugin target if we have another
481 alternative since we want to properly set the input format
482 before allowing a plugin to claim the file. Also, don't
483 check the default target twice. */
484 if (*target
== &binary_vec
485 #if BFD_SUPPORTS_PLUGINS
486 || (match_count
!= 0 && *target
== &plugin_vec
)
488 || (!abfd
->target_defaulted
&& *target
== save_targ
))
491 /* If we already tried a match, the bfd is modified and may
492 have sections attached, which will confuse the next
493 _bfd_check_format call. */
494 bfd_reinit (abfd
, initial_section_id
, &preserve
, cleanup
);
495 /* Free bfd_alloc memory too. If we have matched and preserved
496 a target then the high water mark is that much higher. */
497 if (preserve_match
.marker
)
498 high_water
= &preserve_match
.marker
;
500 high_water
= &preserve
.marker
;
501 bfd_release (abfd
, *high_water
);
502 *high_water
= bfd_alloc (abfd
, 1);
504 /* Change BFD's target temporarily. */
505 abfd
->xvec
= *target
;
507 if (bfd_seek (abfd
, 0, SEEK_SET
) != 0)
510 cleanup
= BFD_SEND_FMT (abfd
, _bfd_check_format
, (abfd
));
513 int match_priority
= abfd
->xvec
->match_priority
;
514 #if BFD_SUPPORTS_PLUGINS
515 /* If this object can be handled by a plugin, give that the
516 lowest priority; objects both handled by a plugin and
517 with an underlying object format will be claimed
518 separately by the plugin. */
519 if (*target
== &plugin_vec
)
520 match_priority
= (*target
)->match_priority
;
523 if (abfd
->format
!= bfd_archive
524 || (bfd_has_map (abfd
)
525 && bfd_get_error () != bfd_error_wrong_object_format
))
527 /* If this is the default target, accept it, even if
528 other targets might match. People who want those
529 other targets have to set the GNUTARGET variable. */
530 if (abfd
->xvec
== bfd_default_vector
[0])
534 matching_vector
[match_count
] = abfd
->xvec
;
537 if (match_priority
< best_match
)
539 best_match
= match_priority
;
542 if (match_priority
<= best_match
)
544 /* This format checks out as ok! */
545 right_targ
= abfd
->xvec
;
551 /* An archive with no armap or objects of the wrong
552 type. We want this target to match if we get no
554 if (ar_right_targ
!= bfd_default_vector
[0])
555 ar_right_targ
= *target
;
557 matching_vector
[ar_match_index
] = *target
;
561 if (preserve_match
.marker
== NULL
)
563 match_targ
= abfd
->xvec
;
564 if (!bfd_preserve_save (abfd
, &preserve_match
, cleanup
))
574 if (match_count
== 0)
576 /* Try partial matches. */
577 right_targ
= ar_right_targ
;
579 if (right_targ
== bfd_default_vector
[0])
585 match_count
= ar_match_index
- _bfd_target_vector_entries
;
587 if (matching_vector
&& match_count
> 1)
588 memcpy (matching_vector
,
589 matching_vector
+ _bfd_target_vector_entries
,
590 sizeof (*matching_vector
) * match_count
);
594 /* We have more than one equally good match. If any of the best
595 matches is a target in config.bfd targ_defvec or targ_selvecs,
599 const bfd_target
* const *assoc
= bfd_associated_vector
;
601 while ((right_targ
= *assoc
++) != NULL
)
606 if (matching_vector
[i
] == right_targ
607 && right_targ
->match_priority
<= best_match
)
618 /* We still have more than one equally good match, and at least some
619 of the targets support match priority. Choose the first of the
621 if (matching_vector
&& match_count
> 1 && best_count
!= match_count
)
625 for (i
= 0; i
< match_count
; i
++)
627 right_targ
= matching_vector
[i
];
628 if (right_targ
->match_priority
<= best_match
)
634 /* There is way too much undoing of half-known state here. We
635 really shouldn't iterate on live bfd's. Note that saving the
636 whole bfd and restoring it would be even worse; the first thing
637 you notice is that the cached bfd file position gets out of sync. */
638 if (preserve_match
.marker
!= NULL
)
639 cleanup
= bfd_preserve_restore (abfd
, &preserve_match
);
641 if (match_count
== 1)
643 abfd
->xvec
= right_targ
;
644 /* If we come out of the loop knowing that the last target that
645 matched is the one we want, then ABFD should still be in a usable
646 state (except possibly for XVEC). This is not just an
647 optimisation. In the case of plugins a match against the
648 plugin target can result in the bfd being changed such that
649 it no longer matches the plugin target, nor will it match
651 if (match_targ
!= right_targ
)
653 bfd_reinit (abfd
, initial_section_id
, &preserve
, cleanup
);
654 bfd_release (abfd
, preserve
.marker
);
655 if (bfd_seek (abfd
, 0, SEEK_SET
) != 0)
657 cleanup
= BFD_SEND_FMT (abfd
, _bfd_check_format
, (abfd
));
658 BFD_ASSERT (cleanup
!= NULL
);
662 /* If the file was opened for update, then `output_has_begun'
663 some time ago when the file was created. Do not recompute
664 sections sizes or alignments in _bfd_set_section_contents.
665 We can not set this flag until after checking the format,
666 because it will interfere with creation of BFD sections. */
667 if (abfd
->direction
== both_direction
)
668 abfd
->output_has_begun
= true;
670 free (matching_vector
);
671 if (preserve_match
.marker
!= NULL
)
672 bfd_preserve_finish (abfd
, &preserve_match
);
673 bfd_preserve_finish (abfd
, &preserve
);
674 _bfd_restore_error_handler_caching (orig_messages
);
676 print_and_clear_messages (&messages
, abfd
->xvec
);
678 bfd_set_lto_type (abfd
);
680 /* File position has moved, BTW. */
681 return bfd_cache_set_uncloseable (abfd
, old_in_format_matches
, NULL
);
684 if (match_count
== 0)
687 bfd_set_error (bfd_error_file_not_recognized
);
691 abfd
->xvec
= save_targ
;
692 abfd
->format
= bfd_unknown
;
693 free (matching_vector
);
697 /* Restore original target type and format. */
698 abfd
->xvec
= save_targ
;
699 abfd
->format
= bfd_unknown
;
700 bfd_set_error (bfd_error_file_ambiguously_recognized
);
704 *matching
= (char **) matching_vector
;
705 matching_vector
[match_count
] = NULL
;
706 /* Return target names. This is a little nasty. Maybe we
707 should do another bfd_malloc? */
708 while (--match_count
>= 0)
710 const char *name
= matching_vector
[match_count
]->name
;
711 *(const char **) &matching_vector
[match_count
] = name
;
715 free (matching_vector
);
719 if (preserve_match
.marker
!= NULL
)
720 bfd_preserve_finish (abfd
, &preserve_match
);
721 bfd_preserve_restore (abfd
, &preserve
);
722 _bfd_restore_error_handler_caching (orig_messages
);
723 print_and_clear_messages (&messages
, PER_XVEC_NO_TARGET
);
724 bfd_cache_set_uncloseable (abfd
, old_in_format_matches
, NULL
);
733 bool bfd_set_format (bfd *abfd, bfd_format format);
736 This function sets the file format of the BFD @var{abfd} to the
737 format @var{format}. If the target set in the BFD does not
738 support the format requested, the format is invalid, or the BFD
739 is not open for writing, then an error occurs.
743 bfd_set_format (bfd
*abfd
, bfd_format format
)
745 if (bfd_read_p (abfd
)
746 || (unsigned int) abfd
->format
>= (unsigned int) bfd_type_end
)
748 bfd_set_error (bfd_error_invalid_operation
);
752 if (abfd
->format
!= bfd_unknown
)
753 return abfd
->format
== format
;
755 /* Presume the answer is yes. */
756 abfd
->format
= format
;
758 if (!BFD_SEND_FMT (abfd
, _bfd_set_format
, (abfd
)))
760 abfd
->format
= bfd_unknown
;
772 const char *bfd_format_string (bfd_format format);
775 Return a pointer to a const string
776 <<invalid>>, <<object>>, <<archive>>, <<core>>, or <<unknown>>,
777 depending upon the value of @var{format}.
781 bfd_format_string (bfd_format format
)
783 if (((int) format
< (int) bfd_unknown
)
784 || ((int) format
>= (int) bfd_type_end
))
790 return "object"; /* Linker/assembler/compiler output. */
792 return "archive"; /* Object archive file. */
794 return "core"; /* Core dump. */