64b384e1e8
bfd/ * archures.c (bfd_mach_i386_nacl): Fix definition so it doesn't collide with bfd_mach_l1om. * bfd-in2.h: Regenerate. * elf32-i386.c (elf32_i386_nacl_elf_object_p): New function. (elf_backend_object_p): Use that in elf32-i386-nacl definition. * elf64-x86-64.c (elf64_x86_64_nacl_elf_object_p): New function. (elf_backend_object_p): Use that in elf64-x86-64-nacl definition. (elf32_x86_64_nacl_elf_object_p): New function. (elf_backend_object_p): Use that in elf32-x86-64-nacl definition. binutils/ * objdump.c (dump_dwarf): Grok bfd_mach_x86_64_nacl and bfd_mach_x64_32_nacl as equivalent to bfd_mach_x86_64. ld/testsuite/ * ld-x86-64/x86-64.exp (mixed1, mixed2): Loosen error string match so it accepts "i386:nacl" in place of "i386". * ld-x86-64/ilp32-2.d: Likewise. * ld-x86-64/ilp32-3.d: Likewise. * ld-x86-64/lp64-2.d: Likewise. * ld-x86-64/lp64-3.d: Likewise.
3698 lines
95 KiB
C
3698 lines
95 KiB
C
/* objdump.c -- dump information about an object file.
|
||
Copyright 1990-2013 Free Software Foundation, Inc.
|
||
|
||
This file is part of GNU Binutils.
|
||
|
||
This program is free software; you can redistribute it and/or modify
|
||
it under the terms of the GNU General Public License as published by
|
||
the Free Software Foundation; either version 3, or (at your option)
|
||
any later version.
|
||
|
||
This program is distributed in the hope that it will be useful,
|
||
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||
GNU General Public License for more details.
|
||
|
||
You should have received a copy of the GNU General Public License
|
||
along with this program; if not, write to the Free Software
|
||
Foundation, 51 Franklin Street - Fifth Floor, Boston,
|
||
MA 02110-1301, USA. */
|
||
|
||
|
||
/* Objdump overview.
|
||
|
||
Objdump displays information about one or more object files, either on
|
||
their own, or inside libraries. It is commonly used as a disassembler,
|
||
but it can also display information about file headers, symbol tables,
|
||
relocations, debugging directives and more.
|
||
|
||
The flow of execution is as follows:
|
||
|
||
1. Command line arguments are checked for control switches and the
|
||
information to be displayed is selected.
|
||
|
||
2. Any remaining arguments are assumed to be object files, and they are
|
||
processed in order by display_bfd(). If the file is an archive each
|
||
of its elements is processed in turn.
|
||
|
||
3. The file's target architecture and binary file format are determined
|
||
by bfd_check_format(). If they are recognised, then dump_bfd() is
|
||
called.
|
||
|
||
4. dump_bfd() in turn calls separate functions to display the requested
|
||
item(s) of information(s). For example disassemble_data() is called if
|
||
a disassembly has been requested.
|
||
|
||
When disassembling the code loops through blocks of instructions bounded
|
||
by symbols, calling disassemble_bytes() on each block. The actual
|
||
disassembling is done by the libopcodes library, via a function pointer
|
||
supplied by the disassembler() function. */
|
||
|
||
#include "sysdep.h"
|
||
#include "bfd.h"
|
||
#include "elf-bfd.h"
|
||
#include "progress.h"
|
||
#include "bucomm.h"
|
||
#include "elfcomm.h"
|
||
#include "dwarf.h"
|
||
#include "getopt.h"
|
||
#include "safe-ctype.h"
|
||
#include "dis-asm.h"
|
||
#include "libiberty.h"
|
||
#include "demangle.h"
|
||
#include "filenames.h"
|
||
#include "debug.h"
|
||
#include "budbg.h"
|
||
#include "objdump.h"
|
||
|
||
#ifdef HAVE_MMAP
|
||
#include <sys/mman.h>
|
||
#endif
|
||
|
||
/* Internal headers for the ELF .stab-dump code - sorry. */
|
||
#define BYTES_IN_WORD 32
|
||
#include "aout/aout64.h"
|
||
|
||
/* Exit status. */
|
||
static int exit_status = 0;
|
||
|
||
static char *default_target = NULL; /* Default at runtime. */
|
||
|
||
/* The following variables are set based on arguments passed on the
|
||
command line. */
|
||
static int show_version = 0; /* Show the version number. */
|
||
static int dump_section_contents; /* -s */
|
||
static int dump_section_headers; /* -h */
|
||
static bfd_boolean dump_file_header; /* -f */
|
||
static int dump_symtab; /* -t */
|
||
static int dump_dynamic_symtab; /* -T */
|
||
static int dump_reloc_info; /* -r */
|
||
static int dump_dynamic_reloc_info; /* -R */
|
||
static int dump_ar_hdrs; /* -a */
|
||
static int dump_private_headers; /* -p */
|
||
static char *dump_private_options; /* -P */
|
||
static int prefix_addresses; /* --prefix-addresses */
|
||
static int with_line_numbers; /* -l */
|
||
static bfd_boolean with_source_code; /* -S */
|
||
static int show_raw_insn; /* --show-raw-insn */
|
||
static int dump_dwarf_section_info; /* --dwarf */
|
||
static int dump_stab_section_info; /* --stabs */
|
||
static int do_demangle; /* -C, --demangle */
|
||
static bfd_boolean disassemble; /* -d */
|
||
static bfd_boolean disassemble_all; /* -D */
|
||
static int disassemble_zeroes; /* --disassemble-zeroes */
|
||
static bfd_boolean formats_info; /* -i */
|
||
static int wide_output; /* -w */
|
||
static int insn_width; /* --insn-width */
|
||
static bfd_vma start_address = (bfd_vma) -1; /* --start-address */
|
||
static bfd_vma stop_address = (bfd_vma) -1; /* --stop-address */
|
||
static int dump_debugging; /* --debugging */
|
||
static int dump_debugging_tags; /* --debugging-tags */
|
||
static int suppress_bfd_header;
|
||
static int dump_special_syms = 0; /* --special-syms */
|
||
static bfd_vma adjust_section_vma = 0; /* --adjust-vma */
|
||
static int file_start_context = 0; /* --file-start-context */
|
||
static bfd_boolean display_file_offsets;/* -F */
|
||
static const char *prefix; /* --prefix */
|
||
static int prefix_strip; /* --prefix-strip */
|
||
static size_t prefix_length;
|
||
|
||
/* A structure to record the sections mentioned in -j switches. */
|
||
struct only
|
||
{
|
||
const char * name; /* The name of the section. */
|
||
bfd_boolean seen; /* A flag to indicate that the section has been found in one or more input files. */
|
||
struct only * next; /* Pointer to the next structure in the list. */
|
||
};
|
||
/* Pointer to an array of 'only' structures.
|
||
This pointer is NULL if the -j switch has not been used. */
|
||
static struct only * only_list = NULL;
|
||
|
||
/* Variables for handling include file path table. */
|
||
static const char **include_paths;
|
||
static int include_path_count;
|
||
|
||
/* Extra info to pass to the section disassembler and address printing
|
||
function. */
|
||
struct objdump_disasm_info
|
||
{
|
||
bfd * abfd;
|
||
asection * sec;
|
||
bfd_boolean require_sec;
|
||
arelent ** dynrelbuf;
|
||
long dynrelcount;
|
||
disassembler_ftype disassemble_fn;
|
||
arelent * reloc;
|
||
};
|
||
|
||
/* Architecture to disassemble for, or default if NULL. */
|
||
static char *machine = NULL;
|
||
|
||
/* Target specific options to the disassembler. */
|
||
static char *disassembler_options = NULL;
|
||
|
||
/* Endianness to disassemble for, or default if BFD_ENDIAN_UNKNOWN. */
|
||
static enum bfd_endian endian = BFD_ENDIAN_UNKNOWN;
|
||
|
||
/* The symbol table. */
|
||
static asymbol **syms;
|
||
|
||
/* Number of symbols in `syms'. */
|
||
static long symcount = 0;
|
||
|
||
/* The sorted symbol table. */
|
||
static asymbol **sorted_syms;
|
||
|
||
/* Number of symbols in `sorted_syms'. */
|
||
static long sorted_symcount = 0;
|
||
|
||
/* The dynamic symbol table. */
|
||
static asymbol **dynsyms;
|
||
|
||
/* The synthetic symbol table. */
|
||
static asymbol *synthsyms;
|
||
static long synthcount = 0;
|
||
|
||
/* Number of symbols in `dynsyms'. */
|
||
static long dynsymcount = 0;
|
||
|
||
static bfd_byte *stabs;
|
||
static bfd_size_type stab_size;
|
||
|
||
static char *strtab;
|
||
static bfd_size_type stabstr_size;
|
||
|
||
static bfd_boolean is_relocatable = FALSE;
|
||
|
||
/* Handlers for -P/--private. */
|
||
static const struct objdump_private_desc * const objdump_private_vectors[] =
|
||
{
|
||
OBJDUMP_PRIVATE_VECTORS
|
||
NULL
|
||
};
|
||
|
||
static void usage (FILE *, int) ATTRIBUTE_NORETURN;
|
||
static void
|
||
usage (FILE *stream, int status)
|
||
{
|
||
fprintf (stream, _("Usage: %s <option(s)> <file(s)>\n"), program_name);
|
||
fprintf (stream, _(" Display information from object <file(s)>.\n"));
|
||
fprintf (stream, _(" At least one of the following switches must be given:\n"));
|
||
fprintf (stream, _("\
|
||
-a, --archive-headers Display archive header information\n\
|
||
-f, --file-headers Display the contents of the overall file header\n\
|
||
-p, --private-headers Display object format specific file header contents\n\
|
||
-P, --private=OPT,OPT... Display object format specific contents\n\
|
||
-h, --[section-]headers Display the contents of the section headers\n\
|
||
-x, --all-headers Display the contents of all headers\n\
|
||
-d, --disassemble Display assembler contents of executable sections\n\
|
||
-D, --disassemble-all Display assembler contents of all sections\n\
|
||
-S, --source Intermix source code with disassembly\n\
|
||
-s, --full-contents Display the full contents of all sections requested\n\
|
||
-g, --debugging Display debug information in object file\n\
|
||
-e, --debugging-tags Display debug information using ctags style\n\
|
||
-G, --stabs Display (in raw form) any STABS info in the file\n\
|
||
-W[lLiaprmfFsoRt] or\n\
|
||
--dwarf[=rawline,=decodedline,=info,=abbrev,=pubnames,=aranges,=macro,=frames,\n\
|
||
=frames-interp,=str,=loc,=Ranges,=pubtypes,\n\
|
||
=gdb_index,=trace_info,=trace_abbrev,=trace_aranges,\n\
|
||
=addr,=cu_index]\n\
|
||
Display DWARF info in the file\n\
|
||
-t, --syms Display the contents of the symbol table(s)\n\
|
||
-T, --dynamic-syms Display the contents of the dynamic symbol table\n\
|
||
-r, --reloc Display the relocation entries in the file\n\
|
||
-R, --dynamic-reloc Display the dynamic relocation entries in the file\n\
|
||
@<file> Read options from <file>\n\
|
||
-v, --version Display this program's version number\n\
|
||
-i, --info List object formats and architectures supported\n\
|
||
-H, --help Display this information\n\
|
||
"));
|
||
if (status != 2)
|
||
{
|
||
const struct objdump_private_desc * const *desc;
|
||
|
||
fprintf (stream, _("\n The following switches are optional:\n"));
|
||
fprintf (stream, _("\
|
||
-b, --target=BFDNAME Specify the target object format as BFDNAME\n\
|
||
-m, --architecture=MACHINE Specify the target architecture as MACHINE\n\
|
||
-j, --section=NAME Only display information for section NAME\n\
|
||
-M, --disassembler-options=OPT Pass text OPT on to the disassembler\n\
|
||
-EB --endian=big Assume big endian format when disassembling\n\
|
||
-EL --endian=little Assume little endian format when disassembling\n\
|
||
--file-start-context Include context from start of file (with -S)\n\
|
||
-I, --include=DIR Add DIR to search list for source files\n\
|
||
-l, --line-numbers Include line numbers and filenames in output\n\
|
||
-F, --file-offsets Include file offsets when displaying information\n\
|
||
-C, --demangle[=STYLE] Decode mangled/processed symbol names\n\
|
||
The STYLE, if specified, can be `auto', `gnu',\n\
|
||
`lucid', `arm', `hp', `edg', `gnu-v3', `java'\n\
|
||
or `gnat'\n\
|
||
-w, --wide Format output for more than 80 columns\n\
|
||
-z, --disassemble-zeroes Do not skip blocks of zeroes when disassembling\n\
|
||
--start-address=ADDR Only process data whose address is >= ADDR\n\
|
||
--stop-address=ADDR Only process data whose address is <= ADDR\n\
|
||
--prefix-addresses Print complete address alongside disassembly\n\
|
||
--[no-]show-raw-insn Display hex alongside symbolic disassembly\n\
|
||
--insn-width=WIDTH Display WIDTH bytes on a single line for -d\n\
|
||
--adjust-vma=OFFSET Add OFFSET to all displayed section addresses\n\
|
||
--special-syms Include special symbols in symbol dumps\n\
|
||
--prefix=PREFIX Add PREFIX to absolute paths for -S\n\
|
||
--prefix-strip=LEVEL Strip initial directory names for -S\n"));
|
||
fprintf (stream, _("\
|
||
--dwarf-depth=N Do not display DIEs at depth N or greater\n\
|
||
--dwarf-start=N Display DIEs starting with N, at the same depth\n\
|
||
or deeper\n\
|
||
--dwarf-check Make additional dwarf internal consistency checks.\
|
||
\n\n"));
|
||
list_supported_targets (program_name, stream);
|
||
list_supported_architectures (program_name, stream);
|
||
|
||
disassembler_usage (stream);
|
||
|
||
if (objdump_private_vectors[0] != NULL)
|
||
{
|
||
fprintf (stream,
|
||
_("\nOptions supported for -P/--private switch:\n"));
|
||
for (desc = objdump_private_vectors; *desc != NULL; desc++)
|
||
(*desc)->help (stream);
|
||
}
|
||
}
|
||
if (REPORT_BUGS_TO[0] && status == 0)
|
||
fprintf (stream, _("Report bugs to %s.\n"), REPORT_BUGS_TO);
|
||
exit (status);
|
||
}
|
||
|
||
/* 150 isn't special; it's just an arbitrary non-ASCII char value. */
|
||
enum option_values
|
||
{
|
||
OPTION_ENDIAN=150,
|
||
OPTION_START_ADDRESS,
|
||
OPTION_STOP_ADDRESS,
|
||
OPTION_DWARF,
|
||
OPTION_PREFIX,
|
||
OPTION_PREFIX_STRIP,
|
||
OPTION_INSN_WIDTH,
|
||
OPTION_ADJUST_VMA,
|
||
OPTION_DWARF_DEPTH,
|
||
OPTION_DWARF_CHECK,
|
||
OPTION_DWARF_START
|
||
};
|
||
|
||
static struct option long_options[]=
|
||
{
|
||
{"adjust-vma", required_argument, NULL, OPTION_ADJUST_VMA},
|
||
{"all-headers", no_argument, NULL, 'x'},
|
||
{"private-headers", no_argument, NULL, 'p'},
|
||
{"private", required_argument, NULL, 'P'},
|
||
{"architecture", required_argument, NULL, 'm'},
|
||
{"archive-headers", no_argument, NULL, 'a'},
|
||
{"debugging", no_argument, NULL, 'g'},
|
||
{"debugging-tags", no_argument, NULL, 'e'},
|
||
{"demangle", optional_argument, NULL, 'C'},
|
||
{"disassemble", no_argument, NULL, 'd'},
|
||
{"disassemble-all", no_argument, NULL, 'D'},
|
||
{"disassembler-options", required_argument, NULL, 'M'},
|
||
{"disassemble-zeroes", no_argument, NULL, 'z'},
|
||
{"dynamic-reloc", no_argument, NULL, 'R'},
|
||
{"dynamic-syms", no_argument, NULL, 'T'},
|
||
{"endian", required_argument, NULL, OPTION_ENDIAN},
|
||
{"file-headers", no_argument, NULL, 'f'},
|
||
{"file-offsets", no_argument, NULL, 'F'},
|
||
{"file-start-context", no_argument, &file_start_context, 1},
|
||
{"full-contents", no_argument, NULL, 's'},
|
||
{"headers", no_argument, NULL, 'h'},
|
||
{"help", no_argument, NULL, 'H'},
|
||
{"info", no_argument, NULL, 'i'},
|
||
{"line-numbers", no_argument, NULL, 'l'},
|
||
{"no-show-raw-insn", no_argument, &show_raw_insn, -1},
|
||
{"prefix-addresses", no_argument, &prefix_addresses, 1},
|
||
{"reloc", no_argument, NULL, 'r'},
|
||
{"section", required_argument, NULL, 'j'},
|
||
{"section-headers", no_argument, NULL, 'h'},
|
||
{"show-raw-insn", no_argument, &show_raw_insn, 1},
|
||
{"source", no_argument, NULL, 'S'},
|
||
{"special-syms", no_argument, &dump_special_syms, 1},
|
||
{"include", required_argument, NULL, 'I'},
|
||
{"dwarf", optional_argument, NULL, OPTION_DWARF},
|
||
{"stabs", no_argument, NULL, 'G'},
|
||
{"start-address", required_argument, NULL, OPTION_START_ADDRESS},
|
||
{"stop-address", required_argument, NULL, OPTION_STOP_ADDRESS},
|
||
{"syms", no_argument, NULL, 't'},
|
||
{"target", required_argument, NULL, 'b'},
|
||
{"version", no_argument, NULL, 'V'},
|
||
{"wide", no_argument, NULL, 'w'},
|
||
{"prefix", required_argument, NULL, OPTION_PREFIX},
|
||
{"prefix-strip", required_argument, NULL, OPTION_PREFIX_STRIP},
|
||
{"insn-width", required_argument, NULL, OPTION_INSN_WIDTH},
|
||
{"dwarf-depth", required_argument, 0, OPTION_DWARF_DEPTH},
|
||
{"dwarf-start", required_argument, 0, OPTION_DWARF_START},
|
||
{"dwarf-check", no_argument, 0, OPTION_DWARF_CHECK},
|
||
{0, no_argument, 0, 0}
|
||
};
|
||
|
||
static void
|
||
nonfatal (const char *msg)
|
||
{
|
||
bfd_nonfatal (msg);
|
||
exit_status = 1;
|
||
}
|
||
|
||
/* Returns TRUE if the specified section should be dumped. */
|
||
|
||
static bfd_boolean
|
||
process_section_p (asection * section)
|
||
{
|
||
struct only * only;
|
||
|
||
if (only_list == NULL)
|
||
return TRUE;
|
||
|
||
for (only = only_list; only; only = only->next)
|
||
if (strcmp (only->name, section->name) == 0)
|
||
{
|
||
only->seen = TRUE;
|
||
return TRUE;
|
||
}
|
||
|
||
return FALSE;
|
||
}
|
||
|
||
/* Add an entry to the 'only' list. */
|
||
|
||
static void
|
||
add_only (char * name)
|
||
{
|
||
struct only * only;
|
||
|
||
/* First check to make sure that we do not
|
||
already have an entry for this name. */
|
||
for (only = only_list; only; only = only->next)
|
||
if (strcmp (only->name, name) == 0)
|
||
return;
|
||
|
||
only = xmalloc (sizeof * only);
|
||
only->name = name;
|
||
only->seen = FALSE;
|
||
only->next = only_list;
|
||
only_list = only;
|
||
}
|
||
|
||
/* Release the memory used by the 'only' list.
|
||
PR 11225: Issue a warning message for unseen sections.
|
||
Only do this if none of the sections were seen. This is mainly to support
|
||
tools like the GAS testsuite where an object file is dumped with a list of
|
||
generic section names known to be present in a range of different file
|
||
formats. */
|
||
|
||
static void
|
||
free_only_list (void)
|
||
{
|
||
bfd_boolean at_least_one_seen = FALSE;
|
||
struct only * only;
|
||
struct only * next;
|
||
|
||
if (only_list == NULL)
|
||
return;
|
||
|
||
for (only = only_list; only; only = only->next)
|
||
if (only->seen)
|
||
{
|
||
at_least_one_seen = TRUE;
|
||
break;
|
||
}
|
||
|
||
for (only = only_list; only; only = next)
|
||
{
|
||
if (! at_least_one_seen)
|
||
{
|
||
non_fatal (_("section '%s' mentioned in a -j option, "
|
||
"but not found in any input file"),
|
||
only->name);
|
||
exit_status = 1;
|
||
}
|
||
next = only->next;
|
||
free (only);
|
||
}
|
||
}
|
||
|
||
|
||
static void
|
||
dump_section_header (bfd *abfd, asection *section,
|
||
void *ignored ATTRIBUTE_UNUSED)
|
||
{
|
||
char *comma = "";
|
||
unsigned int opb = bfd_octets_per_byte (abfd);
|
||
|
||
/* Ignore linker created section. See elfNN_ia64_object_p in
|
||
bfd/elfxx-ia64.c. */
|
||
if (section->flags & SEC_LINKER_CREATED)
|
||
return;
|
||
|
||
/* PR 10413: Skip sections that we are ignoring. */
|
||
if (! process_section_p (section))
|
||
return;
|
||
|
||
printf ("%3d %-13s %08lx ", section->index,
|
||
bfd_get_section_name (abfd, section),
|
||
(unsigned long) bfd_section_size (abfd, section) / opb);
|
||
bfd_printf_vma (abfd, bfd_get_section_vma (abfd, section));
|
||
printf (" ");
|
||
bfd_printf_vma (abfd, section->lma);
|
||
printf (" %08lx 2**%u", (unsigned long) section->filepos,
|
||
bfd_get_section_alignment (abfd, section));
|
||
if (! wide_output)
|
||
printf ("\n ");
|
||
printf (" ");
|
||
|
||
#define PF(x, y) \
|
||
if (section->flags & x) { printf ("%s%s", comma, y); comma = ", "; }
|
||
|
||
PF (SEC_HAS_CONTENTS, "CONTENTS");
|
||
PF (SEC_ALLOC, "ALLOC");
|
||
PF (SEC_CONSTRUCTOR, "CONSTRUCTOR");
|
||
PF (SEC_LOAD, "LOAD");
|
||
PF (SEC_RELOC, "RELOC");
|
||
PF (SEC_READONLY, "READONLY");
|
||
PF (SEC_CODE, "CODE");
|
||
PF (SEC_DATA, "DATA");
|
||
PF (SEC_ROM, "ROM");
|
||
PF (SEC_DEBUGGING, "DEBUGGING");
|
||
PF (SEC_NEVER_LOAD, "NEVER_LOAD");
|
||
PF (SEC_EXCLUDE, "EXCLUDE");
|
||
PF (SEC_SORT_ENTRIES, "SORT_ENTRIES");
|
||
if (bfd_get_arch (abfd) == bfd_arch_tic54x)
|
||
{
|
||
PF (SEC_TIC54X_BLOCK, "BLOCK");
|
||
PF (SEC_TIC54X_CLINK, "CLINK");
|
||
}
|
||
PF (SEC_SMALL_DATA, "SMALL_DATA");
|
||
if (bfd_get_flavour (abfd) == bfd_target_coff_flavour)
|
||
PF (SEC_COFF_SHARED, "SHARED");
|
||
PF (SEC_THREAD_LOCAL, "THREAD_LOCAL");
|
||
PF (SEC_GROUP, "GROUP");
|
||
|
||
if ((section->flags & SEC_LINK_ONCE) != 0)
|
||
{
|
||
const char *ls;
|
||
struct coff_comdat_info *comdat;
|
||
|
||
switch (section->flags & SEC_LINK_DUPLICATES)
|
||
{
|
||
default:
|
||
abort ();
|
||
case SEC_LINK_DUPLICATES_DISCARD:
|
||
ls = "LINK_ONCE_DISCARD";
|
||
break;
|
||
case SEC_LINK_DUPLICATES_ONE_ONLY:
|
||
ls = "LINK_ONCE_ONE_ONLY";
|
||
break;
|
||
case SEC_LINK_DUPLICATES_SAME_SIZE:
|
||
ls = "LINK_ONCE_SAME_SIZE";
|
||
break;
|
||
case SEC_LINK_DUPLICATES_SAME_CONTENTS:
|
||
ls = "LINK_ONCE_SAME_CONTENTS";
|
||
break;
|
||
}
|
||
printf ("%s%s", comma, ls);
|
||
|
||
comdat = bfd_coff_get_comdat_section (abfd, section);
|
||
if (comdat != NULL)
|
||
printf (" (COMDAT %s %ld)", comdat->name, comdat->symbol);
|
||
|
||
comma = ", ";
|
||
}
|
||
|
||
printf ("\n");
|
||
#undef PF
|
||
}
|
||
|
||
static void
|
||
dump_headers (bfd *abfd)
|
||
{
|
||
printf (_("Sections:\n"));
|
||
|
||
#ifndef BFD64
|
||
printf (_("Idx Name Size VMA LMA File off Algn"));
|
||
#else
|
||
/* With BFD64, non-ELF returns -1 and wants always 64 bit addresses. */
|
||
if (bfd_get_arch_size (abfd) == 32)
|
||
printf (_("Idx Name Size VMA LMA File off Algn"));
|
||
else
|
||
printf (_("Idx Name Size VMA LMA File off Algn"));
|
||
#endif
|
||
|
||
if (wide_output)
|
||
printf (_(" Flags"));
|
||
printf ("\n");
|
||
|
||
bfd_map_over_sections (abfd, dump_section_header, NULL);
|
||
}
|
||
|
||
static asymbol **
|
||
slurp_symtab (bfd *abfd)
|
||
{
|
||
asymbol **sy = NULL;
|
||
long storage;
|
||
|
||
if (!(bfd_get_file_flags (abfd) & HAS_SYMS))
|
||
{
|
||
symcount = 0;
|
||
return NULL;
|
||
}
|
||
|
||
storage = bfd_get_symtab_upper_bound (abfd);
|
||
if (storage < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
if (storage)
|
||
sy = (asymbol **) xmalloc (storage);
|
||
|
||
symcount = bfd_canonicalize_symtab (abfd, sy);
|
||
if (symcount < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
return sy;
|
||
}
|
||
|
||
/* Read in the dynamic symbols. */
|
||
|
||
static asymbol **
|
||
slurp_dynamic_symtab (bfd *abfd)
|
||
{
|
||
asymbol **sy = NULL;
|
||
long storage;
|
||
|
||
storage = bfd_get_dynamic_symtab_upper_bound (abfd);
|
||
if (storage < 0)
|
||
{
|
||
if (!(bfd_get_file_flags (abfd) & DYNAMIC))
|
||
{
|
||
non_fatal (_("%s: not a dynamic object"), bfd_get_filename (abfd));
|
||
exit_status = 1;
|
||
dynsymcount = 0;
|
||
return NULL;
|
||
}
|
||
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
}
|
||
if (storage)
|
||
sy = (asymbol **) xmalloc (storage);
|
||
|
||
dynsymcount = bfd_canonicalize_dynamic_symtab (abfd, sy);
|
||
if (dynsymcount < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
return sy;
|
||
}
|
||
|
||
/* Filter out (in place) symbols that are useless for disassembly.
|
||
COUNT is the number of elements in SYMBOLS.
|
||
Return the number of useful symbols. */
|
||
|
||
static long
|
||
remove_useless_symbols (asymbol **symbols, long count)
|
||
{
|
||
asymbol **in_ptr = symbols, **out_ptr = symbols;
|
||
|
||
while (--count >= 0)
|
||
{
|
||
asymbol *sym = *in_ptr++;
|
||
|
||
if (sym->name == NULL || sym->name[0] == '\0')
|
||
continue;
|
||
if (sym->flags & (BSF_DEBUGGING | BSF_SECTION_SYM))
|
||
continue;
|
||
if (bfd_is_und_section (sym->section)
|
||
|| bfd_is_com_section (sym->section))
|
||
continue;
|
||
|
||
*out_ptr++ = sym;
|
||
}
|
||
return out_ptr - symbols;
|
||
}
|
||
|
||
/* Sort symbols into value order. */
|
||
|
||
static int
|
||
compare_symbols (const void *ap, const void *bp)
|
||
{
|
||
const asymbol *a = * (const asymbol **) ap;
|
||
const asymbol *b = * (const asymbol **) bp;
|
||
const char *an;
|
||
const char *bn;
|
||
size_t anl;
|
||
size_t bnl;
|
||
bfd_boolean af;
|
||
bfd_boolean bf;
|
||
flagword aflags;
|
||
flagword bflags;
|
||
|
||
if (bfd_asymbol_value (a) > bfd_asymbol_value (b))
|
||
return 1;
|
||
else if (bfd_asymbol_value (a) < bfd_asymbol_value (b))
|
||
return -1;
|
||
|
||
if (a->section > b->section)
|
||
return 1;
|
||
else if (a->section < b->section)
|
||
return -1;
|
||
|
||
an = bfd_asymbol_name (a);
|
||
bn = bfd_asymbol_name (b);
|
||
anl = strlen (an);
|
||
bnl = strlen (bn);
|
||
|
||
/* The symbols gnu_compiled and gcc2_compiled convey no real
|
||
information, so put them after other symbols with the same value. */
|
||
af = (strstr (an, "gnu_compiled") != NULL
|
||
|| strstr (an, "gcc2_compiled") != NULL);
|
||
bf = (strstr (bn, "gnu_compiled") != NULL
|
||
|| strstr (bn, "gcc2_compiled") != NULL);
|
||
|
||
if (af && ! bf)
|
||
return 1;
|
||
if (! af && bf)
|
||
return -1;
|
||
|
||
/* We use a heuristic for the file name, to try to sort it after
|
||
more useful symbols. It may not work on non Unix systems, but it
|
||
doesn't really matter; the only difference is precisely which
|
||
symbol names get printed. */
|
||
|
||
#define file_symbol(s, sn, snl) \
|
||
(((s)->flags & BSF_FILE) != 0 \
|
||
|| ((sn)[(snl) - 2] == '.' \
|
||
&& ((sn)[(snl) - 1] == 'o' \
|
||
|| (sn)[(snl) - 1] == 'a')))
|
||
|
||
af = file_symbol (a, an, anl);
|
||
bf = file_symbol (b, bn, bnl);
|
||
|
||
if (af && ! bf)
|
||
return 1;
|
||
if (! af && bf)
|
||
return -1;
|
||
|
||
/* Try to sort global symbols before local symbols before function
|
||
symbols before debugging symbols. */
|
||
|
||
aflags = a->flags;
|
||
bflags = b->flags;
|
||
|
||
if ((aflags & BSF_DEBUGGING) != (bflags & BSF_DEBUGGING))
|
||
{
|
||
if ((aflags & BSF_DEBUGGING) != 0)
|
||
return 1;
|
||
else
|
||
return -1;
|
||
}
|
||
if ((aflags & BSF_FUNCTION) != (bflags & BSF_FUNCTION))
|
||
{
|
||
if ((aflags & BSF_FUNCTION) != 0)
|
||
return -1;
|
||
else
|
||
return 1;
|
||
}
|
||
if ((aflags & BSF_LOCAL) != (bflags & BSF_LOCAL))
|
||
{
|
||
if ((aflags & BSF_LOCAL) != 0)
|
||
return 1;
|
||
else
|
||
return -1;
|
||
}
|
||
if ((aflags & BSF_GLOBAL) != (bflags & BSF_GLOBAL))
|
||
{
|
||
if ((aflags & BSF_GLOBAL) != 0)
|
||
return -1;
|
||
else
|
||
return 1;
|
||
}
|
||
|
||
/* Symbols that start with '.' might be section names, so sort them
|
||
after symbols that don't start with '.'. */
|
||
if (an[0] == '.' && bn[0] != '.')
|
||
return 1;
|
||
if (an[0] != '.' && bn[0] == '.')
|
||
return -1;
|
||
|
||
/* Finally, if we can't distinguish them in any other way, try to
|
||
get consistent results by sorting the symbols by name. */
|
||
return strcmp (an, bn);
|
||
}
|
||
|
||
/* Sort relocs into address order. */
|
||
|
||
static int
|
||
compare_relocs (const void *ap, const void *bp)
|
||
{
|
||
const arelent *a = * (const arelent **) ap;
|
||
const arelent *b = * (const arelent **) bp;
|
||
|
||
if (a->address > b->address)
|
||
return 1;
|
||
else if (a->address < b->address)
|
||
return -1;
|
||
|
||
/* So that associated relocations tied to the same address show up
|
||
in the correct order, we don't do any further sorting. */
|
||
if (a > b)
|
||
return 1;
|
||
else if (a < b)
|
||
return -1;
|
||
else
|
||
return 0;
|
||
}
|
||
|
||
/* Print an address (VMA) to the output stream in INFO.
|
||
If SKIP_ZEROES is TRUE, omit leading zeroes. */
|
||
|
||
static void
|
||
objdump_print_value (bfd_vma vma, struct disassemble_info *inf,
|
||
bfd_boolean skip_zeroes)
|
||
{
|
||
char buf[30];
|
||
char *p;
|
||
struct objdump_disasm_info *aux;
|
||
|
||
aux = (struct objdump_disasm_info *) inf->application_data;
|
||
bfd_sprintf_vma (aux->abfd, buf, vma);
|
||
if (! skip_zeroes)
|
||
p = buf;
|
||
else
|
||
{
|
||
for (p = buf; *p == '0'; ++p)
|
||
;
|
||
if (*p == '\0')
|
||
--p;
|
||
}
|
||
(*inf->fprintf_func) (inf->stream, "%s", p);
|
||
}
|
||
|
||
/* Print the name of a symbol. */
|
||
|
||
static void
|
||
objdump_print_symname (bfd *abfd, struct disassemble_info *inf,
|
||
asymbol *sym)
|
||
{
|
||
char *alloc;
|
||
const char *name;
|
||
|
||
alloc = NULL;
|
||
name = bfd_asymbol_name (sym);
|
||
if (do_demangle && name[0] != '\0')
|
||
{
|
||
/* Demangle the name. */
|
||
alloc = bfd_demangle (abfd, name, DMGL_ANSI | DMGL_PARAMS);
|
||
if (alloc != NULL)
|
||
name = alloc;
|
||
}
|
||
|
||
if (inf != NULL)
|
||
(*inf->fprintf_func) (inf->stream, "%s", name);
|
||
else
|
||
printf ("%s", name);
|
||
|
||
if (alloc != NULL)
|
||
free (alloc);
|
||
}
|
||
|
||
/* Locate a symbol given a bfd and a section (from INFO->application_data),
|
||
and a VMA. If INFO->application_data->require_sec is TRUE, then always
|
||
require the symbol to be in the section. Returns NULL if there is no
|
||
suitable symbol. If PLACE is not NULL, then *PLACE is set to the index
|
||
of the symbol in sorted_syms. */
|
||
|
||
static asymbol *
|
||
find_symbol_for_address (bfd_vma vma,
|
||
struct disassemble_info *inf,
|
||
long *place)
|
||
{
|
||
/* @@ Would it speed things up to cache the last two symbols returned,
|
||
and maybe their address ranges? For many processors, only one memory
|
||
operand can be present at a time, so the 2-entry cache wouldn't be
|
||
constantly churned by code doing heavy memory accesses. */
|
||
|
||
/* Indices in `sorted_syms'. */
|
||
long min = 0;
|
||
long max_count = sorted_symcount;
|
||
long thisplace;
|
||
struct objdump_disasm_info *aux;
|
||
bfd *abfd;
|
||
asection *sec;
|
||
unsigned int opb;
|
||
bfd_boolean want_section;
|
||
|
||
if (sorted_symcount < 1)
|
||
return NULL;
|
||
|
||
aux = (struct objdump_disasm_info *) inf->application_data;
|
||
abfd = aux->abfd;
|
||
sec = aux->sec;
|
||
opb = inf->octets_per_byte;
|
||
|
||
/* Perform a binary search looking for the closest symbol to the
|
||
required value. We are searching the range (min, max_count]. */
|
||
while (min + 1 < max_count)
|
||
{
|
||
asymbol *sym;
|
||
|
||
thisplace = (max_count + min) / 2;
|
||
sym = sorted_syms[thisplace];
|
||
|
||
if (bfd_asymbol_value (sym) > vma)
|
||
max_count = thisplace;
|
||
else if (bfd_asymbol_value (sym) < vma)
|
||
min = thisplace;
|
||
else
|
||
{
|
||
min = thisplace;
|
||
break;
|
||
}
|
||
}
|
||
|
||
/* The symbol we want is now in min, the low end of the range we
|
||
were searching. If there are several symbols with the same
|
||
value, we want the first one. */
|
||
thisplace = min;
|
||
while (thisplace > 0
|
||
&& (bfd_asymbol_value (sorted_syms[thisplace])
|
||
== bfd_asymbol_value (sorted_syms[thisplace - 1])))
|
||
--thisplace;
|
||
|
||
/* Prefer a symbol in the current section if we have multple symbols
|
||
with the same value, as can occur with overlays or zero size
|
||
sections. */
|
||
min = thisplace;
|
||
while (min < max_count
|
||
&& (bfd_asymbol_value (sorted_syms[min])
|
||
== bfd_asymbol_value (sorted_syms[thisplace])))
|
||
{
|
||
if (sorted_syms[min]->section == sec
|
||
&& inf->symbol_is_valid (sorted_syms[min], inf))
|
||
{
|
||
thisplace = min;
|
||
|
||
if (place != NULL)
|
||
*place = thisplace;
|
||
|
||
return sorted_syms[thisplace];
|
||
}
|
||
++min;
|
||
}
|
||
|
||
/* If the file is relocatable, and the symbol could be from this
|
||
section, prefer a symbol from this section over symbols from
|
||
others, even if the other symbol's value might be closer.
|
||
|
||
Note that this may be wrong for some symbol references if the
|
||
sections have overlapping memory ranges, but in that case there's
|
||
no way to tell what's desired without looking at the relocation
|
||
table.
|
||
|
||
Also give the target a chance to reject symbols. */
|
||
want_section = (aux->require_sec
|
||
|| ((abfd->flags & HAS_RELOC) != 0
|
||
&& vma >= bfd_get_section_vma (abfd, sec)
|
||
&& vma < (bfd_get_section_vma (abfd, sec)
|
||
+ bfd_section_size (abfd, sec) / opb)));
|
||
if ((sorted_syms[thisplace]->section != sec && want_section)
|
||
|| ! inf->symbol_is_valid (sorted_syms[thisplace], inf))
|
||
{
|
||
long i;
|
||
long newplace = sorted_symcount;
|
||
|
||
for (i = min - 1; i >= 0; i--)
|
||
{
|
||
if ((sorted_syms[i]->section == sec || !want_section)
|
||
&& inf->symbol_is_valid (sorted_syms[i], inf))
|
||
{
|
||
if (newplace == sorted_symcount)
|
||
newplace = i;
|
||
|
||
if (bfd_asymbol_value (sorted_syms[i])
|
||
!= bfd_asymbol_value (sorted_syms[newplace]))
|
||
break;
|
||
|
||
/* Remember this symbol and keep searching until we reach
|
||
an earlier address. */
|
||
newplace = i;
|
||
}
|
||
}
|
||
|
||
if (newplace != sorted_symcount)
|
||
thisplace = newplace;
|
||
else
|
||
{
|
||
/* We didn't find a good symbol with a smaller value.
|
||
Look for one with a larger value. */
|
||
for (i = thisplace + 1; i < sorted_symcount; i++)
|
||
{
|
||
if ((sorted_syms[i]->section == sec || !want_section)
|
||
&& inf->symbol_is_valid (sorted_syms[i], inf))
|
||
{
|
||
thisplace = i;
|
||
break;
|
||
}
|
||
}
|
||
}
|
||
|
||
if ((sorted_syms[thisplace]->section != sec && want_section)
|
||
|| ! inf->symbol_is_valid (sorted_syms[thisplace], inf))
|
||
/* There is no suitable symbol. */
|
||
return NULL;
|
||
}
|
||
|
||
if (place != NULL)
|
||
*place = thisplace;
|
||
|
||
return sorted_syms[thisplace];
|
||
}
|
||
|
||
/* Print an address and the offset to the nearest symbol. */
|
||
|
||
static void
|
||
objdump_print_addr_with_sym (bfd *abfd, asection *sec, asymbol *sym,
|
||
bfd_vma vma, struct disassemble_info *inf,
|
||
bfd_boolean skip_zeroes)
|
||
{
|
||
objdump_print_value (vma, inf, skip_zeroes);
|
||
|
||
if (sym == NULL)
|
||
{
|
||
bfd_vma secaddr;
|
||
|
||
(*inf->fprintf_func) (inf->stream, " <%s",
|
||
bfd_get_section_name (abfd, sec));
|
||
secaddr = bfd_get_section_vma (abfd, sec);
|
||
if (vma < secaddr)
|
||
{
|
||
(*inf->fprintf_func) (inf->stream, "-0x");
|
||
objdump_print_value (secaddr - vma, inf, TRUE);
|
||
}
|
||
else if (vma > secaddr)
|
||
{
|
||
(*inf->fprintf_func) (inf->stream, "+0x");
|
||
objdump_print_value (vma - secaddr, inf, TRUE);
|
||
}
|
||
(*inf->fprintf_func) (inf->stream, ">");
|
||
}
|
||
else
|
||
{
|
||
(*inf->fprintf_func) (inf->stream, " <");
|
||
objdump_print_symname (abfd, inf, sym);
|
||
if (bfd_asymbol_value (sym) > vma)
|
||
{
|
||
(*inf->fprintf_func) (inf->stream, "-0x");
|
||
objdump_print_value (bfd_asymbol_value (sym) - vma, inf, TRUE);
|
||
}
|
||
else if (vma > bfd_asymbol_value (sym))
|
||
{
|
||
(*inf->fprintf_func) (inf->stream, "+0x");
|
||
objdump_print_value (vma - bfd_asymbol_value (sym), inf, TRUE);
|
||
}
|
||
(*inf->fprintf_func) (inf->stream, ">");
|
||
}
|
||
|
||
if (display_file_offsets)
|
||
inf->fprintf_func (inf->stream, _(" (File Offset: 0x%lx)"),
|
||
(long int)(sec->filepos + (vma - sec->vma)));
|
||
}
|
||
|
||
/* Print an address (VMA), symbolically if possible.
|
||
If SKIP_ZEROES is TRUE, don't output leading zeroes. */
|
||
|
||
static void
|
||
objdump_print_addr (bfd_vma vma,
|
||
struct disassemble_info *inf,
|
||
bfd_boolean skip_zeroes)
|
||
{
|
||
struct objdump_disasm_info *aux;
|
||
asymbol *sym = NULL;
|
||
bfd_boolean skip_find = FALSE;
|
||
|
||
aux = (struct objdump_disasm_info *) inf->application_data;
|
||
|
||
if (sorted_symcount < 1)
|
||
{
|
||
(*inf->fprintf_func) (inf->stream, "0x");
|
||
objdump_print_value (vma, inf, skip_zeroes);
|
||
|
||
if (display_file_offsets)
|
||
inf->fprintf_func (inf->stream, _(" (File Offset: 0x%lx)"),
|
||
(long int)(aux->sec->filepos + (vma - aux->sec->vma)));
|
||
return;
|
||
}
|
||
|
||
if (aux->reloc != NULL
|
||
&& aux->reloc->sym_ptr_ptr != NULL
|
||
&& * aux->reloc->sym_ptr_ptr != NULL)
|
||
{
|
||
sym = * aux->reloc->sym_ptr_ptr;
|
||
|
||
/* Adjust the vma to the reloc. */
|
||
vma += bfd_asymbol_value (sym);
|
||
|
||
if (bfd_is_und_section (bfd_get_section (sym)))
|
||
skip_find = TRUE;
|
||
}
|
||
|
||
if (!skip_find)
|
||
sym = find_symbol_for_address (vma, inf, NULL);
|
||
|
||
objdump_print_addr_with_sym (aux->abfd, aux->sec, sym, vma, inf,
|
||
skip_zeroes);
|
||
}
|
||
|
||
/* Print VMA to INFO. This function is passed to the disassembler
|
||
routine. */
|
||
|
||
static void
|
||
objdump_print_address (bfd_vma vma, struct disassemble_info *inf)
|
||
{
|
||
objdump_print_addr (vma, inf, ! prefix_addresses);
|
||
}
|
||
|
||
/* Determine if the given address has a symbol associated with it. */
|
||
|
||
static int
|
||
objdump_symbol_at_address (bfd_vma vma, struct disassemble_info * inf)
|
||
{
|
||
asymbol * sym;
|
||
|
||
sym = find_symbol_for_address (vma, inf, NULL);
|
||
|
||
return (sym != NULL && (bfd_asymbol_value (sym) == vma));
|
||
}
|
||
|
||
/* Hold the last function name and the last line number we displayed
|
||
in a disassembly. */
|
||
|
||
static char *prev_functionname;
|
||
static unsigned int prev_line;
|
||
static unsigned int prev_discriminator;
|
||
|
||
/* We keep a list of all files that we have seen when doing a
|
||
disassembly with source, so that we know how much of the file to
|
||
display. This can be important for inlined functions. */
|
||
|
||
struct print_file_list
|
||
{
|
||
struct print_file_list *next;
|
||
const char *filename;
|
||
const char *modname;
|
||
const char *map;
|
||
size_t mapsize;
|
||
const char **linemap;
|
||
unsigned maxline;
|
||
unsigned last_line;
|
||
int first;
|
||
};
|
||
|
||
static struct print_file_list *print_files;
|
||
|
||
/* The number of preceding context lines to show when we start
|
||
displaying a file for the first time. */
|
||
|
||
#define SHOW_PRECEDING_CONTEXT_LINES (5)
|
||
|
||
/* Read a complete file into memory. */
|
||
|
||
static const char *
|
||
slurp_file (const char *fn, size_t *size)
|
||
{
|
||
#ifdef HAVE_MMAP
|
||
int ps = getpagesize ();
|
||
size_t msize;
|
||
#endif
|
||
const char *map;
|
||
struct stat st;
|
||
int fd = open (fn, O_RDONLY | O_BINARY);
|
||
|
||
if (fd < 0)
|
||
return NULL;
|
||
if (fstat (fd, &st) < 0)
|
||
{
|
||
close (fd);
|
||
return NULL;
|
||
}
|
||
*size = st.st_size;
|
||
#ifdef HAVE_MMAP
|
||
msize = (*size + ps - 1) & ~(ps - 1);
|
||
map = mmap (NULL, msize, PROT_READ, MAP_SHARED, fd, 0);
|
||
if (map != (char *) -1L)
|
||
{
|
||
close (fd);
|
||
return map;
|
||
}
|
||
#endif
|
||
map = (const char *) malloc (*size);
|
||
if (!map || (size_t) read (fd, (char *) map, *size) != *size)
|
||
{
|
||
free ((void *) map);
|
||
map = NULL;
|
||
}
|
||
close (fd);
|
||
return map;
|
||
}
|
||
|
||
#define line_map_decrease 5
|
||
|
||
/* Precompute array of lines for a mapped file. */
|
||
|
||
static const char **
|
||
index_file (const char *map, size_t size, unsigned int *maxline)
|
||
{
|
||
const char *p, *lstart, *end;
|
||
int chars_per_line = 45; /* First iteration will use 40. */
|
||
unsigned int lineno;
|
||
const char **linemap = NULL;
|
||
unsigned long line_map_size = 0;
|
||
|
||
lineno = 0;
|
||
lstart = map;
|
||
end = map + size;
|
||
|
||
for (p = map; p < end; p++)
|
||
{
|
||
if (*p == '\n')
|
||
{
|
||
if (p + 1 < end && p[1] == '\r')
|
||
p++;
|
||
}
|
||
else if (*p == '\r')
|
||
{
|
||
if (p + 1 < end && p[1] == '\n')
|
||
p++;
|
||
}
|
||
else
|
||
continue;
|
||
|
||
/* End of line found. */
|
||
|
||
if (linemap == NULL || line_map_size < lineno + 1)
|
||
{
|
||
unsigned long newsize;
|
||
|
||
chars_per_line -= line_map_decrease;
|
||
if (chars_per_line <= 1)
|
||
chars_per_line = 1;
|
||
line_map_size = size / chars_per_line + 1;
|
||
if (line_map_size < lineno + 1)
|
||
line_map_size = lineno + 1;
|
||
newsize = line_map_size * sizeof (char *);
|
||
linemap = (const char **) xrealloc (linemap, newsize);
|
||
}
|
||
|
||
linemap[lineno++] = lstart;
|
||
lstart = p + 1;
|
||
}
|
||
|
||
*maxline = lineno;
|
||
return linemap;
|
||
}
|
||
|
||
/* Tries to open MODNAME, and if successful adds a node to print_files
|
||
linked list and returns that node. Returns NULL on failure. */
|
||
|
||
static struct print_file_list *
|
||
try_print_file_open (const char *origname, const char *modname)
|
||
{
|
||
struct print_file_list *p;
|
||
|
||
p = (struct print_file_list *) xmalloc (sizeof (struct print_file_list));
|
||
|
||
p->map = slurp_file (modname, &p->mapsize);
|
||
if (p->map == NULL)
|
||
{
|
||
free (p);
|
||
return NULL;
|
||
}
|
||
|
||
p->linemap = index_file (p->map, p->mapsize, &p->maxline);
|
||
p->last_line = 0;
|
||
p->filename = origname;
|
||
p->modname = modname;
|
||
p->next = print_files;
|
||
p->first = 1;
|
||
print_files = p;
|
||
return p;
|
||
}
|
||
|
||
/* If the source file, as described in the symtab, is not found
|
||
try to locate it in one of the paths specified with -I
|
||
If found, add location to print_files linked list. */
|
||
|
||
static struct print_file_list *
|
||
update_source_path (const char *filename)
|
||
{
|
||
struct print_file_list *p;
|
||
const char *fname;
|
||
int i;
|
||
|
||
p = try_print_file_open (filename, filename);
|
||
if (p != NULL)
|
||
return p;
|
||
|
||
if (include_path_count == 0)
|
||
return NULL;
|
||
|
||
/* Get the name of the file. */
|
||
fname = lbasename (filename);
|
||
|
||
/* If file exists under a new path, we need to add it to the list
|
||
so that show_line knows about it. */
|
||
for (i = 0; i < include_path_count; i++)
|
||
{
|
||
char *modname = concat (include_paths[i], "/", fname, (const char *) 0);
|
||
|
||
p = try_print_file_open (filename, modname);
|
||
if (p)
|
||
return p;
|
||
|
||
free (modname);
|
||
}
|
||
|
||
return NULL;
|
||
}
|
||
|
||
/* Print a source file line. */
|
||
|
||
static void
|
||
print_line (struct print_file_list *p, unsigned int linenum)
|
||
{
|
||
const char *l;
|
||
size_t len;
|
||
|
||
--linenum;
|
||
if (linenum >= p->maxline)
|
||
return;
|
||
l = p->linemap [linenum];
|
||
/* Test fwrite return value to quiet glibc warning. */
|
||
len = strcspn (l, "\n\r");
|
||
if (len == 0 || fwrite (l, len, 1, stdout) == 1)
|
||
putchar ('\n');
|
||
}
|
||
|
||
/* Print a range of source code lines. */
|
||
|
||
static void
|
||
dump_lines (struct print_file_list *p, unsigned int start, unsigned int end)
|
||
{
|
||
if (p->map == NULL)
|
||
return;
|
||
while (start <= end)
|
||
{
|
||
print_line (p, start);
|
||
start++;
|
||
}
|
||
}
|
||
|
||
/* Show the line number, or the source line, in a disassembly
|
||
listing. */
|
||
|
||
static void
|
||
show_line (bfd *abfd, asection *section, bfd_vma addr_offset)
|
||
{
|
||
const char *filename;
|
||
const char *functionname;
|
||
unsigned int linenumber;
|
||
unsigned int discriminator;
|
||
bfd_boolean reloc;
|
||
|
||
if (! with_line_numbers && ! with_source_code)
|
||
return;
|
||
|
||
if (! bfd_find_nearest_line_discriminator (abfd, section, syms, addr_offset,
|
||
&filename, &functionname,
|
||
&linenumber, &discriminator))
|
||
return;
|
||
|
||
if (filename != NULL && *filename == '\0')
|
||
filename = NULL;
|
||
if (functionname != NULL && *functionname == '\0')
|
||
functionname = NULL;
|
||
|
||
if (filename
|
||
&& IS_ABSOLUTE_PATH (filename)
|
||
&& prefix)
|
||
{
|
||
char *path_up;
|
||
const char *fname = filename;
|
||
char *path = (char *) alloca (prefix_length + PATH_MAX + 1);
|
||
|
||
if (prefix_length)
|
||
memcpy (path, prefix, prefix_length);
|
||
path_up = path + prefix_length;
|
||
|
||
/* Build relocated filename, stripping off leading directories
|
||
from the initial filename if requested. */
|
||
if (prefix_strip > 0)
|
||
{
|
||
int level = 0;
|
||
const char *s;
|
||
|
||
/* Skip selected directory levels. */
|
||
for (s = fname + 1; *s != '\0' && level < prefix_strip; s++)
|
||
if (IS_DIR_SEPARATOR(*s))
|
||
{
|
||
fname = s;
|
||
level++;
|
||
}
|
||
}
|
||
|
||
/* Update complete filename. */
|
||
strncpy (path_up, fname, PATH_MAX);
|
||
path_up[PATH_MAX] = '\0';
|
||
|
||
filename = path;
|
||
reloc = TRUE;
|
||
}
|
||
else
|
||
reloc = FALSE;
|
||
|
||
if (with_line_numbers)
|
||
{
|
||
if (functionname != NULL
|
||
&& (prev_functionname == NULL
|
||
|| strcmp (functionname, prev_functionname) != 0))
|
||
printf ("%s():\n", functionname);
|
||
if (linenumber > 0 && (linenumber != prev_line ||
|
||
(discriminator != prev_discriminator)))
|
||
{
|
||
if (discriminator > 0)
|
||
printf ("%s:%u (discriminator %u)\n", filename == NULL ? "???" : filename,
|
||
linenumber, discriminator);
|
||
else
|
||
printf ("%s:%u\n", filename == NULL ? "???" : filename, linenumber);
|
||
}
|
||
}
|
||
|
||
if (with_source_code
|
||
&& filename != NULL
|
||
&& linenumber > 0)
|
||
{
|
||
struct print_file_list **pp, *p;
|
||
unsigned l;
|
||
|
||
for (pp = &print_files; *pp != NULL; pp = &(*pp)->next)
|
||
if (filename_cmp ((*pp)->filename, filename) == 0)
|
||
break;
|
||
p = *pp;
|
||
|
||
if (p == NULL)
|
||
{
|
||
if (reloc)
|
||
filename = xstrdup (filename);
|
||
p = update_source_path (filename);
|
||
}
|
||
|
||
if (p != NULL && linenumber != p->last_line)
|
||
{
|
||
if (file_start_context && p->first)
|
||
l = 1;
|
||
else
|
||
{
|
||
l = linenumber - SHOW_PRECEDING_CONTEXT_LINES;
|
||
if (l >= linenumber)
|
||
l = 1;
|
||
if (p->last_line >= l && p->last_line <= linenumber)
|
||
l = p->last_line + 1;
|
||
}
|
||
dump_lines (p, l, linenumber);
|
||
p->last_line = linenumber;
|
||
p->first = 0;
|
||
}
|
||
}
|
||
|
||
if (functionname != NULL
|
||
&& (prev_functionname == NULL
|
||
|| strcmp (functionname, prev_functionname) != 0))
|
||
{
|
||
if (prev_functionname != NULL)
|
||
free (prev_functionname);
|
||
prev_functionname = (char *) xmalloc (strlen (functionname) + 1);
|
||
strcpy (prev_functionname, functionname);
|
||
}
|
||
|
||
if (linenumber > 0 && linenumber != prev_line)
|
||
prev_line = linenumber;
|
||
|
||
if (discriminator != prev_discriminator)
|
||
prev_discriminator = discriminator;
|
||
}
|
||
|
||
/* Pseudo FILE object for strings. */
|
||
typedef struct
|
||
{
|
||
char *buffer;
|
||
size_t pos;
|
||
size_t alloc;
|
||
} SFILE;
|
||
|
||
/* sprintf to a "stream". */
|
||
|
||
static int ATTRIBUTE_PRINTF_2
|
||
objdump_sprintf (SFILE *f, const char *format, ...)
|
||
{
|
||
size_t n;
|
||
va_list args;
|
||
|
||
while (1)
|
||
{
|
||
size_t space = f->alloc - f->pos;
|
||
|
||
va_start (args, format);
|
||
n = vsnprintf (f->buffer + f->pos, space, format, args);
|
||
va_end (args);
|
||
|
||
if (space > n)
|
||
break;
|
||
|
||
f->alloc = (f->alloc + n) * 2;
|
||
f->buffer = (char *) xrealloc (f->buffer, f->alloc);
|
||
}
|
||
f->pos += n;
|
||
|
||
return n;
|
||
}
|
||
|
||
/* The number of zeroes we want to see before we start skipping them.
|
||
The number is arbitrarily chosen. */
|
||
|
||
#define DEFAULT_SKIP_ZEROES 8
|
||
|
||
/* The number of zeroes to skip at the end of a section. If the
|
||
number of zeroes at the end is between SKIP_ZEROES_AT_END and
|
||
SKIP_ZEROES, they will be disassembled. If there are fewer than
|
||
SKIP_ZEROES_AT_END, they will be skipped. This is a heuristic
|
||
attempt to avoid disassembling zeroes inserted by section
|
||
alignment. */
|
||
|
||
#define DEFAULT_SKIP_ZEROES_AT_END 3
|
||
|
||
/* Disassemble some data in memory between given values. */
|
||
|
||
static void
|
||
disassemble_bytes (struct disassemble_info * inf,
|
||
disassembler_ftype disassemble_fn,
|
||
bfd_boolean insns,
|
||
bfd_byte * data,
|
||
bfd_vma start_offset,
|
||
bfd_vma stop_offset,
|
||
bfd_vma rel_offset,
|
||
arelent *** relppp,
|
||
arelent ** relppend)
|
||
{
|
||
struct objdump_disasm_info *aux;
|
||
asection *section;
|
||
int octets_per_line;
|
||
int skip_addr_chars;
|
||
bfd_vma addr_offset;
|
||
unsigned int opb = inf->octets_per_byte;
|
||
unsigned int skip_zeroes = inf->skip_zeroes;
|
||
unsigned int skip_zeroes_at_end = inf->skip_zeroes_at_end;
|
||
int octets = opb;
|
||
SFILE sfile;
|
||
|
||
aux = (struct objdump_disasm_info *) inf->application_data;
|
||
section = aux->sec;
|
||
|
||
sfile.alloc = 120;
|
||
sfile.buffer = (char *) xmalloc (sfile.alloc);
|
||
sfile.pos = 0;
|
||
|
||
if (insn_width)
|
||
octets_per_line = insn_width;
|
||
else if (insns)
|
||
octets_per_line = 4;
|
||
else
|
||
octets_per_line = 16;
|
||
|
||
/* Figure out how many characters to skip at the start of an
|
||
address, to make the disassembly look nicer. We discard leading
|
||
zeroes in chunks of 4, ensuring that there is always a leading
|
||
zero remaining. */
|
||
skip_addr_chars = 0;
|
||
if (! prefix_addresses)
|
||
{
|
||
char buf[30];
|
||
|
||
bfd_sprintf_vma (aux->abfd, buf, section->vma + section->size / opb);
|
||
|
||
while (buf[skip_addr_chars] == '0')
|
||
++skip_addr_chars;
|
||
|
||
/* Don't discard zeros on overflow. */
|
||
if (buf[skip_addr_chars] == '\0' && section->vma != 0)
|
||
skip_addr_chars = 0;
|
||
|
||
if (skip_addr_chars != 0)
|
||
skip_addr_chars = (skip_addr_chars - 1) & -4;
|
||
}
|
||
|
||
inf->insn_info_valid = 0;
|
||
|
||
addr_offset = start_offset;
|
||
while (addr_offset < stop_offset)
|
||
{
|
||
bfd_vma z;
|
||
bfd_boolean need_nl = FALSE;
|
||
int previous_octets;
|
||
|
||
/* Remember the length of the previous instruction. */
|
||
previous_octets = octets;
|
||
octets = 0;
|
||
|
||
/* Make sure we don't use relocs from previous instructions. */
|
||
aux->reloc = NULL;
|
||
|
||
/* If we see more than SKIP_ZEROES octets of zeroes, we just
|
||
print `...'. */
|
||
for (z = addr_offset * opb; z < stop_offset * opb; z++)
|
||
if (data[z] != 0)
|
||
break;
|
||
if (! disassemble_zeroes
|
||
&& (inf->insn_info_valid == 0
|
||
|| inf->branch_delay_insns == 0)
|
||
&& (z - addr_offset * opb >= skip_zeroes
|
||
|| (z == stop_offset * opb &&
|
||
z - addr_offset * opb < skip_zeroes_at_end)))
|
||
{
|
||
/* If there are more nonzero octets to follow, we only skip
|
||
zeroes in multiples of 4, to try to avoid running over
|
||
the start of an instruction which happens to start with
|
||
zero. */
|
||
if (z != stop_offset * opb)
|
||
z = addr_offset * opb + ((z - addr_offset * opb) &~ 3);
|
||
|
||
octets = z - addr_offset * opb;
|
||
|
||
/* If we are going to display more data, and we are displaying
|
||
file offsets, then tell the user how many zeroes we skip
|
||
and the file offset from where we resume dumping. */
|
||
if (display_file_offsets && ((addr_offset + (octets / opb)) < stop_offset))
|
||
printf ("\t... (skipping %d zeroes, resuming at file offset: 0x%lx)\n",
|
||
octets / opb,
|
||
(unsigned long) (section->filepos
|
||
+ (addr_offset + (octets / opb))));
|
||
else
|
||
printf ("\t...\n");
|
||
}
|
||
else
|
||
{
|
||
char buf[50];
|
||
int bpc = 0;
|
||
int pb = 0;
|
||
|
||
if (with_line_numbers || with_source_code)
|
||
show_line (aux->abfd, section, addr_offset);
|
||
|
||
if (! prefix_addresses)
|
||
{
|
||
char *s;
|
||
|
||
bfd_sprintf_vma (aux->abfd, buf, section->vma + addr_offset);
|
||
for (s = buf + skip_addr_chars; *s == '0'; s++)
|
||
*s = ' ';
|
||
if (*s == '\0')
|
||
*--s = '0';
|
||
printf ("%s:\t", buf + skip_addr_chars);
|
||
}
|
||
else
|
||
{
|
||
aux->require_sec = TRUE;
|
||
objdump_print_address (section->vma + addr_offset, inf);
|
||
aux->require_sec = FALSE;
|
||
putchar (' ');
|
||
}
|
||
|
||
if (insns)
|
||
{
|
||
sfile.pos = 0;
|
||
inf->fprintf_func = (fprintf_ftype) objdump_sprintf;
|
||
inf->stream = &sfile;
|
||
inf->bytes_per_line = 0;
|
||
inf->bytes_per_chunk = 0;
|
||
inf->flags = disassemble_all ? DISASSEMBLE_DATA : 0;
|
||
if (machine)
|
||
inf->flags |= USER_SPECIFIED_MACHINE_TYPE;
|
||
|
||
if (inf->disassembler_needs_relocs
|
||
&& (bfd_get_file_flags (aux->abfd) & EXEC_P) == 0
|
||
&& (bfd_get_file_flags (aux->abfd) & DYNAMIC) == 0
|
||
&& *relppp < relppend)
|
||
{
|
||
bfd_signed_vma distance_to_rel;
|
||
|
||
distance_to_rel = (**relppp)->address
|
||
- (rel_offset + addr_offset);
|
||
|
||
/* Check to see if the current reloc is associated with
|
||
the instruction that we are about to disassemble. */
|
||
if (distance_to_rel == 0
|
||
/* FIXME: This is wrong. We are trying to catch
|
||
relocs that are addressed part way through the
|
||
current instruction, as might happen with a packed
|
||
VLIW instruction. Unfortunately we do not know the
|
||
length of the current instruction since we have not
|
||
disassembled it yet. Instead we take a guess based
|
||
upon the length of the previous instruction. The
|
||
proper solution is to have a new target-specific
|
||
disassembler function which just returns the length
|
||
of an instruction at a given address without trying
|
||
to display its disassembly. */
|
||
|| (distance_to_rel > 0
|
||
&& distance_to_rel < (bfd_signed_vma) (previous_octets/ opb)))
|
||
{
|
||
inf->flags |= INSN_HAS_RELOC;
|
||
aux->reloc = **relppp;
|
||
}
|
||
}
|
||
|
||
octets = (*disassemble_fn) (section->vma + addr_offset, inf);
|
||
inf->fprintf_func = (fprintf_ftype) fprintf;
|
||
inf->stream = stdout;
|
||
if (insn_width == 0 && inf->bytes_per_line != 0)
|
||
octets_per_line = inf->bytes_per_line;
|
||
if (octets < (int) opb)
|
||
{
|
||
if (sfile.pos)
|
||
printf ("%s\n", sfile.buffer);
|
||
if (octets >= 0)
|
||
{
|
||
non_fatal (_("disassemble_fn returned length %d"),
|
||
octets);
|
||
exit_status = 1;
|
||
}
|
||
break;
|
||
}
|
||
}
|
||
else
|
||
{
|
||
bfd_vma j;
|
||
|
||
octets = octets_per_line;
|
||
if (addr_offset + octets / opb > stop_offset)
|
||
octets = (stop_offset - addr_offset) * opb;
|
||
|
||
for (j = addr_offset * opb; j < addr_offset * opb + octets; ++j)
|
||
{
|
||
if (ISPRINT (data[j]))
|
||
buf[j - addr_offset * opb] = data[j];
|
||
else
|
||
buf[j - addr_offset * opb] = '.';
|
||
}
|
||
buf[j - addr_offset * opb] = '\0';
|
||
}
|
||
|
||
if (prefix_addresses
|
||
? show_raw_insn > 0
|
||
: show_raw_insn >= 0)
|
||
{
|
||
bfd_vma j;
|
||
|
||
/* If ! prefix_addresses and ! wide_output, we print
|
||
octets_per_line octets per line. */
|
||
pb = octets;
|
||
if (pb > octets_per_line && ! prefix_addresses && ! wide_output)
|
||
pb = octets_per_line;
|
||
|
||
if (inf->bytes_per_chunk)
|
||
bpc = inf->bytes_per_chunk;
|
||
else
|
||
bpc = 1;
|
||
|
||
for (j = addr_offset * opb; j < addr_offset * opb + pb; j += bpc)
|
||
{
|
||
int k;
|
||
|
||
if (bpc > 1 && inf->display_endian == BFD_ENDIAN_LITTLE)
|
||
{
|
||
for (k = bpc - 1; k >= 0; k--)
|
||
printf ("%02x", (unsigned) data[j + k]);
|
||
putchar (' ');
|
||
}
|
||
else
|
||
{
|
||
for (k = 0; k < bpc; k++)
|
||
printf ("%02x", (unsigned) data[j + k]);
|
||
putchar (' ');
|
||
}
|
||
}
|
||
|
||
for (; pb < octets_per_line; pb += bpc)
|
||
{
|
||
int k;
|
||
|
||
for (k = 0; k < bpc; k++)
|
||
printf (" ");
|
||
putchar (' ');
|
||
}
|
||
|
||
/* Separate raw data from instruction by extra space. */
|
||
if (insns)
|
||
putchar ('\t');
|
||
else
|
||
printf (" ");
|
||
}
|
||
|
||
if (! insns)
|
||
printf ("%s", buf);
|
||
else if (sfile.pos)
|
||
printf ("%s", sfile.buffer);
|
||
|
||
if (prefix_addresses
|
||
? show_raw_insn > 0
|
||
: show_raw_insn >= 0)
|
||
{
|
||
while (pb < octets)
|
||
{
|
||
bfd_vma j;
|
||
char *s;
|
||
|
||
putchar ('\n');
|
||
j = addr_offset * opb + pb;
|
||
|
||
bfd_sprintf_vma (aux->abfd, buf, section->vma + j / opb);
|
||
for (s = buf + skip_addr_chars; *s == '0'; s++)
|
||
*s = ' ';
|
||
if (*s == '\0')
|
||
*--s = '0';
|
||
printf ("%s:\t", buf + skip_addr_chars);
|
||
|
||
pb += octets_per_line;
|
||
if (pb > octets)
|
||
pb = octets;
|
||
for (; j < addr_offset * opb + pb; j += bpc)
|
||
{
|
||
int k;
|
||
|
||
if (bpc > 1 && inf->display_endian == BFD_ENDIAN_LITTLE)
|
||
{
|
||
for (k = bpc - 1; k >= 0; k--)
|
||
printf ("%02x", (unsigned) data[j + k]);
|
||
putchar (' ');
|
||
}
|
||
else
|
||
{
|
||
for (k = 0; k < bpc; k++)
|
||
printf ("%02x", (unsigned) data[j + k]);
|
||
putchar (' ');
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
if (!wide_output)
|
||
putchar ('\n');
|
||
else
|
||
need_nl = TRUE;
|
||
}
|
||
|
||
while ((*relppp) < relppend
|
||
&& (**relppp)->address < rel_offset + addr_offset + octets / opb)
|
||
{
|
||
if (dump_reloc_info || dump_dynamic_reloc_info)
|
||
{
|
||
arelent *q;
|
||
|
||
q = **relppp;
|
||
|
||
if (wide_output)
|
||
putchar ('\t');
|
||
else
|
||
printf ("\t\t\t");
|
||
|
||
objdump_print_value (section->vma - rel_offset + q->address,
|
||
inf, TRUE);
|
||
|
||
if (q->howto == NULL)
|
||
printf (": *unknown*\t");
|
||
else if (q->howto->name)
|
||
printf (": %s\t", q->howto->name);
|
||
else
|
||
printf (": %d\t", q->howto->type);
|
||
|
||
if (q->sym_ptr_ptr == NULL || *q->sym_ptr_ptr == NULL)
|
||
printf ("*unknown*");
|
||
else
|
||
{
|
||
const char *sym_name;
|
||
|
||
sym_name = bfd_asymbol_name (*q->sym_ptr_ptr);
|
||
if (sym_name != NULL && *sym_name != '\0')
|
||
objdump_print_symname (aux->abfd, inf, *q->sym_ptr_ptr);
|
||
else
|
||
{
|
||
asection *sym_sec;
|
||
|
||
sym_sec = bfd_get_section (*q->sym_ptr_ptr);
|
||
sym_name = bfd_get_section_name (aux->abfd, sym_sec);
|
||
if (sym_name == NULL || *sym_name == '\0')
|
||
sym_name = "*unknown*";
|
||
printf ("%s", sym_name);
|
||
}
|
||
}
|
||
|
||
if (q->addend)
|
||
{
|
||
bfd_signed_vma addend = q->addend;
|
||
if (addend < 0)
|
||
{
|
||
printf ("-0x");
|
||
addend = -addend;
|
||
}
|
||
else
|
||
printf ("+0x");
|
||
objdump_print_value (addend, inf, TRUE);
|
||
}
|
||
|
||
printf ("\n");
|
||
need_nl = FALSE;
|
||
}
|
||
++(*relppp);
|
||
}
|
||
|
||
if (need_nl)
|
||
printf ("\n");
|
||
|
||
addr_offset += octets / opb;
|
||
}
|
||
|
||
free (sfile.buffer);
|
||
}
|
||
|
||
static void
|
||
disassemble_section (bfd *abfd, asection *section, void *inf)
|
||
{
|
||
const struct elf_backend_data * bed;
|
||
bfd_vma sign_adjust = 0;
|
||
struct disassemble_info * pinfo = (struct disassemble_info *) inf;
|
||
struct objdump_disasm_info * paux;
|
||
unsigned int opb = pinfo->octets_per_byte;
|
||
bfd_byte * data = NULL;
|
||
bfd_size_type datasize = 0;
|
||
arelent ** rel_pp = NULL;
|
||
arelent ** rel_ppstart = NULL;
|
||
arelent ** rel_ppend;
|
||
unsigned long stop_offset;
|
||
asymbol * sym = NULL;
|
||
long place = 0;
|
||
long rel_count;
|
||
bfd_vma rel_offset;
|
||
unsigned long addr_offset;
|
||
|
||
/* Sections that do not contain machine
|
||
code are not normally disassembled. */
|
||
if (! disassemble_all
|
||
&& only_list == NULL
|
||
&& ((section->flags & (SEC_CODE | SEC_HAS_CONTENTS))
|
||
!= (SEC_CODE | SEC_HAS_CONTENTS)))
|
||
return;
|
||
|
||
if (! process_section_p (section))
|
||
return;
|
||
|
||
datasize = bfd_get_section_size (section);
|
||
if (datasize == 0)
|
||
return;
|
||
|
||
if (start_address == (bfd_vma) -1
|
||
|| start_address < section->vma)
|
||
addr_offset = 0;
|
||
else
|
||
addr_offset = start_address - section->vma;
|
||
|
||
if (stop_address == (bfd_vma) -1)
|
||
stop_offset = datasize / opb;
|
||
else
|
||
{
|
||
if (stop_address < section->vma)
|
||
stop_offset = 0;
|
||
else
|
||
stop_offset = stop_address - section->vma;
|
||
if (stop_offset > datasize / opb)
|
||
stop_offset = datasize / opb;
|
||
}
|
||
|
||
if (addr_offset >= stop_offset)
|
||
return;
|
||
|
||
/* Decide which set of relocs to use. Load them if necessary. */
|
||
paux = (struct objdump_disasm_info *) pinfo->application_data;
|
||
if (paux->dynrelbuf)
|
||
{
|
||
rel_pp = paux->dynrelbuf;
|
||
rel_count = paux->dynrelcount;
|
||
/* Dynamic reloc addresses are absolute, non-dynamic are section
|
||
relative. REL_OFFSET specifies the reloc address corresponding
|
||
to the start of this section. */
|
||
rel_offset = section->vma;
|
||
}
|
||
else
|
||
{
|
||
rel_count = 0;
|
||
rel_pp = NULL;
|
||
rel_offset = 0;
|
||
|
||
if ((section->flags & SEC_RELOC) != 0
|
||
&& (dump_reloc_info || pinfo->disassembler_needs_relocs))
|
||
{
|
||
long relsize;
|
||
|
||
relsize = bfd_get_reloc_upper_bound (abfd, section);
|
||
if (relsize < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
|
||
if (relsize > 0)
|
||
{
|
||
rel_ppstart = rel_pp = (arelent **) xmalloc (relsize);
|
||
rel_count = bfd_canonicalize_reloc (abfd, section, rel_pp, syms);
|
||
if (rel_count < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
|
||
/* Sort the relocs by address. */
|
||
qsort (rel_pp, rel_count, sizeof (arelent *), compare_relocs);
|
||
}
|
||
}
|
||
}
|
||
rel_ppend = rel_pp + rel_count;
|
||
|
||
data = (bfd_byte *) xmalloc (datasize);
|
||
|
||
bfd_get_section_contents (abfd, section, data, 0, datasize);
|
||
|
||
paux->sec = section;
|
||
pinfo->buffer = data;
|
||
pinfo->buffer_vma = section->vma;
|
||
pinfo->buffer_length = datasize;
|
||
pinfo->section = section;
|
||
|
||
/* Skip over the relocs belonging to addresses below the
|
||
start address. */
|
||
while (rel_pp < rel_ppend
|
||
&& (*rel_pp)->address < rel_offset + addr_offset)
|
||
++rel_pp;
|
||
|
||
printf (_("\nDisassembly of section %s:\n"), section->name);
|
||
|
||
/* Find the nearest symbol forwards from our current position. */
|
||
paux->require_sec = TRUE;
|
||
sym = (asymbol *) find_symbol_for_address (section->vma + addr_offset,
|
||
(struct disassemble_info *) inf,
|
||
&place);
|
||
paux->require_sec = FALSE;
|
||
|
||
/* PR 9774: If the target used signed addresses then we must make
|
||
sure that we sign extend the value that we calculate for 'addr'
|
||
in the loop below. */
|
||
if (bfd_get_flavour (abfd) == bfd_target_elf_flavour
|
||
&& (bed = get_elf_backend_data (abfd)) != NULL
|
||
&& bed->sign_extend_vma)
|
||
sign_adjust = (bfd_vma) 1 << (bed->s->arch_size - 1);
|
||
|
||
/* Disassemble a block of instructions up to the address associated with
|
||
the symbol we have just found. Then print the symbol and find the
|
||
next symbol on. Repeat until we have disassembled the entire section
|
||
or we have reached the end of the address range we are interested in. */
|
||
while (addr_offset < stop_offset)
|
||
{
|
||
bfd_vma addr;
|
||
asymbol *nextsym;
|
||
unsigned long nextstop_offset;
|
||
bfd_boolean insns;
|
||
|
||
addr = section->vma + addr_offset;
|
||
addr = ((addr & ((sign_adjust << 1) - 1)) ^ sign_adjust) - sign_adjust;
|
||
|
||
if (sym != NULL && bfd_asymbol_value (sym) <= addr)
|
||
{
|
||
int x;
|
||
|
||
for (x = place;
|
||
(x < sorted_symcount
|
||
&& (bfd_asymbol_value (sorted_syms[x]) <= addr));
|
||
++x)
|
||
continue;
|
||
|
||
pinfo->symbols = sorted_syms + place;
|
||
pinfo->num_symbols = x - place;
|
||
pinfo->symtab_pos = place;
|
||
}
|
||
else
|
||
{
|
||
pinfo->symbols = NULL;
|
||
pinfo->num_symbols = 0;
|
||
pinfo->symtab_pos = -1;
|
||
}
|
||
|
||
if (! prefix_addresses)
|
||
{
|
||
pinfo->fprintf_func (pinfo->stream, "\n");
|
||
objdump_print_addr_with_sym (abfd, section, sym, addr,
|
||
pinfo, FALSE);
|
||
pinfo->fprintf_func (pinfo->stream, ":\n");
|
||
}
|
||
|
||
if (sym != NULL && bfd_asymbol_value (sym) > addr)
|
||
nextsym = sym;
|
||
else if (sym == NULL)
|
||
nextsym = NULL;
|
||
else
|
||
{
|
||
#define is_valid_next_sym(SYM) \
|
||
((SYM)->section == section \
|
||
&& (bfd_asymbol_value (SYM) > bfd_asymbol_value (sym)) \
|
||
&& pinfo->symbol_is_valid (SYM, pinfo))
|
||
|
||
/* Search forward for the next appropriate symbol in
|
||
SECTION. Note that all the symbols are sorted
|
||
together into one big array, and that some sections
|
||
may have overlapping addresses. */
|
||
while (place < sorted_symcount
|
||
&& ! is_valid_next_sym (sorted_syms [place]))
|
||
++place;
|
||
|
||
if (place >= sorted_symcount)
|
||
nextsym = NULL;
|
||
else
|
||
nextsym = sorted_syms[place];
|
||
}
|
||
|
||
if (sym != NULL && bfd_asymbol_value (sym) > addr)
|
||
nextstop_offset = bfd_asymbol_value (sym) - section->vma;
|
||
else if (nextsym == NULL)
|
||
nextstop_offset = stop_offset;
|
||
else
|
||
nextstop_offset = bfd_asymbol_value (nextsym) - section->vma;
|
||
|
||
if (nextstop_offset > stop_offset
|
||
|| nextstop_offset <= addr_offset)
|
||
nextstop_offset = stop_offset;
|
||
|
||
/* If a symbol is explicitly marked as being an object
|
||
rather than a function, just dump the bytes without
|
||
disassembling them. */
|
||
if (disassemble_all
|
||
|| sym == NULL
|
||
|| sym->section != section
|
||
|| bfd_asymbol_value (sym) > addr
|
||
|| ((sym->flags & BSF_OBJECT) == 0
|
||
&& (strstr (bfd_asymbol_name (sym), "gnu_compiled")
|
||
== NULL)
|
||
&& (strstr (bfd_asymbol_name (sym), "gcc2_compiled")
|
||
== NULL))
|
||
|| (sym->flags & BSF_FUNCTION) != 0)
|
||
insns = TRUE;
|
||
else
|
||
insns = FALSE;
|
||
|
||
disassemble_bytes (pinfo, paux->disassemble_fn, insns, data,
|
||
addr_offset, nextstop_offset,
|
||
rel_offset, &rel_pp, rel_ppend);
|
||
|
||
addr_offset = nextstop_offset;
|
||
sym = nextsym;
|
||
}
|
||
|
||
free (data);
|
||
|
||
if (rel_ppstart != NULL)
|
||
free (rel_ppstart);
|
||
}
|
||
|
||
/* Disassemble the contents of an object file. */
|
||
|
||
static void
|
||
disassemble_data (bfd *abfd)
|
||
{
|
||
struct disassemble_info disasm_info;
|
||
struct objdump_disasm_info aux;
|
||
long i;
|
||
|
||
print_files = NULL;
|
||
prev_functionname = NULL;
|
||
prev_line = -1;
|
||
prev_discriminator = 0;
|
||
|
||
/* We make a copy of syms to sort. We don't want to sort syms
|
||
because that will screw up the relocs. */
|
||
sorted_symcount = symcount ? symcount : dynsymcount;
|
||
sorted_syms = (asymbol **) xmalloc ((sorted_symcount + synthcount)
|
||
* sizeof (asymbol *));
|
||
memcpy (sorted_syms, symcount ? syms : dynsyms,
|
||
sorted_symcount * sizeof (asymbol *));
|
||
|
||
sorted_symcount = remove_useless_symbols (sorted_syms, sorted_symcount);
|
||
|
||
for (i = 0; i < synthcount; ++i)
|
||
{
|
||
sorted_syms[sorted_symcount] = synthsyms + i;
|
||
++sorted_symcount;
|
||
}
|
||
|
||
/* Sort the symbols into section and symbol order. */
|
||
qsort (sorted_syms, sorted_symcount, sizeof (asymbol *), compare_symbols);
|
||
|
||
init_disassemble_info (&disasm_info, stdout, (fprintf_ftype) fprintf);
|
||
|
||
disasm_info.application_data = (void *) &aux;
|
||
aux.abfd = abfd;
|
||
aux.require_sec = FALSE;
|
||
aux.dynrelbuf = NULL;
|
||
aux.dynrelcount = 0;
|
||
aux.reloc = NULL;
|
||
|
||
disasm_info.print_address_func = objdump_print_address;
|
||
disasm_info.symbol_at_address_func = objdump_symbol_at_address;
|
||
|
||
if (machine != NULL)
|
||
{
|
||
const bfd_arch_info_type *inf = bfd_scan_arch (machine);
|
||
|
||
if (inf == NULL)
|
||
fatal (_("can't use supplied machine %s"), machine);
|
||
|
||
abfd->arch_info = inf;
|
||
}
|
||
|
||
if (endian != BFD_ENDIAN_UNKNOWN)
|
||
{
|
||
struct bfd_target *xvec;
|
||
|
||
xvec = (struct bfd_target *) xmalloc (sizeof (struct bfd_target));
|
||
memcpy (xvec, abfd->xvec, sizeof (struct bfd_target));
|
||
xvec->byteorder = endian;
|
||
abfd->xvec = xvec;
|
||
}
|
||
|
||
/* Use libopcodes to locate a suitable disassembler. */
|
||
aux.disassemble_fn = disassembler (abfd);
|
||
if (!aux.disassemble_fn)
|
||
{
|
||
non_fatal (_("can't disassemble for architecture %s\n"),
|
||
bfd_printable_arch_mach (bfd_get_arch (abfd), 0));
|
||
exit_status = 1;
|
||
return;
|
||
}
|
||
|
||
disasm_info.flavour = bfd_get_flavour (abfd);
|
||
disasm_info.arch = bfd_get_arch (abfd);
|
||
disasm_info.mach = bfd_get_mach (abfd);
|
||
disasm_info.disassembler_options = disassembler_options;
|
||
disasm_info.octets_per_byte = bfd_octets_per_byte (abfd);
|
||
disasm_info.skip_zeroes = DEFAULT_SKIP_ZEROES;
|
||
disasm_info.skip_zeroes_at_end = DEFAULT_SKIP_ZEROES_AT_END;
|
||
disasm_info.disassembler_needs_relocs = FALSE;
|
||
|
||
if (bfd_big_endian (abfd))
|
||
disasm_info.display_endian = disasm_info.endian = BFD_ENDIAN_BIG;
|
||
else if (bfd_little_endian (abfd))
|
||
disasm_info.display_endian = disasm_info.endian = BFD_ENDIAN_LITTLE;
|
||
else
|
||
/* ??? Aborting here seems too drastic. We could default to big or little
|
||
instead. */
|
||
disasm_info.endian = BFD_ENDIAN_UNKNOWN;
|
||
|
||
/* Allow the target to customize the info structure. */
|
||
disassemble_init_for_target (& disasm_info);
|
||
|
||
/* Pre-load the dynamic relocs if we are going
|
||
to be dumping them along with the disassembly. */
|
||
if (dump_dynamic_reloc_info)
|
||
{
|
||
long relsize = bfd_get_dynamic_reloc_upper_bound (abfd);
|
||
|
||
if (relsize < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
|
||
if (relsize > 0)
|
||
{
|
||
aux.dynrelbuf = (arelent **) xmalloc (relsize);
|
||
aux.dynrelcount = bfd_canonicalize_dynamic_reloc (abfd,
|
||
aux.dynrelbuf,
|
||
dynsyms);
|
||
if (aux.dynrelcount < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
|
||
/* Sort the relocs by address. */
|
||
qsort (aux.dynrelbuf, aux.dynrelcount, sizeof (arelent *),
|
||
compare_relocs);
|
||
}
|
||
}
|
||
disasm_info.symtab = sorted_syms;
|
||
disasm_info.symtab_size = sorted_symcount;
|
||
|
||
bfd_map_over_sections (abfd, disassemble_section, & disasm_info);
|
||
|
||
if (aux.dynrelbuf != NULL)
|
||
free (aux.dynrelbuf);
|
||
free (sorted_syms);
|
||
}
|
||
|
||
static int
|
||
load_specific_debug_section (enum dwarf_section_display_enum debug,
|
||
asection *sec, void *file)
|
||
{
|
||
struct dwarf_section *section = &debug_displays [debug].section;
|
||
bfd *abfd = (bfd *) file;
|
||
bfd_boolean ret;
|
||
|
||
/* If it is already loaded, do nothing. */
|
||
if (section->start != NULL)
|
||
return 1;
|
||
|
||
section->address = 0;
|
||
section->size = bfd_get_section_size (sec);
|
||
section->start = NULL;
|
||
ret = bfd_get_full_section_contents (abfd, sec, §ion->start);
|
||
|
||
if (! ret)
|
||
{
|
||
free_debug_section (debug);
|
||
printf (_("\nCan't get contents for section '%s'.\n"),
|
||
section->name);
|
||
return 0;
|
||
}
|
||
|
||
if (is_relocatable && debug_displays [debug].relocate)
|
||
{
|
||
bfd_cache_section_contents (sec, section->start);
|
||
|
||
ret = bfd_simple_get_relocated_section_contents (abfd,
|
||
sec,
|
||
section->start,
|
||
syms) != NULL;
|
||
|
||
if (! ret)
|
||
{
|
||
free_debug_section (debug);
|
||
printf (_("\nCan't get contents for section '%s'.\n"),
|
||
section->name);
|
||
return 0;
|
||
}
|
||
}
|
||
|
||
return 1;
|
||
}
|
||
|
||
int
|
||
load_debug_section (enum dwarf_section_display_enum debug, void *file)
|
||
{
|
||
struct dwarf_section *section = &debug_displays [debug].section;
|
||
bfd *abfd = (bfd *) file;
|
||
asection *sec;
|
||
|
||
/* If it is already loaded, do nothing. */
|
||
if (section->start != NULL)
|
||
return 1;
|
||
|
||
/* Locate the debug section. */
|
||
sec = bfd_get_section_by_name (abfd, section->uncompressed_name);
|
||
if (sec != NULL)
|
||
section->name = section->uncompressed_name;
|
||
else
|
||
{
|
||
sec = bfd_get_section_by_name (abfd, section->compressed_name);
|
||
if (sec != NULL)
|
||
section->name = section->compressed_name;
|
||
}
|
||
if (sec == NULL)
|
||
return 0;
|
||
|
||
return load_specific_debug_section (debug, sec, file);
|
||
}
|
||
|
||
void
|
||
free_debug_section (enum dwarf_section_display_enum debug)
|
||
{
|
||
struct dwarf_section *section = &debug_displays [debug].section;
|
||
|
||
if (section->start == NULL)
|
||
return;
|
||
|
||
free ((char *) section->start);
|
||
section->start = NULL;
|
||
section->address = 0;
|
||
section->size = 0;
|
||
}
|
||
|
||
static void
|
||
dump_dwarf_section (bfd *abfd, asection *section,
|
||
void *arg ATTRIBUTE_UNUSED)
|
||
{
|
||
const char *name = bfd_get_section_name (abfd, section);
|
||
const char *match;
|
||
int i;
|
||
|
||
if (CONST_STRNEQ (name, ".gnu.linkonce.wi."))
|
||
match = ".debug_info";
|
||
else
|
||
match = name;
|
||
|
||
for (i = 0; i < max; i++)
|
||
if ((strcmp (debug_displays [i].section.uncompressed_name, match) == 0
|
||
|| strcmp (debug_displays [i].section.compressed_name, match) == 0)
|
||
&& debug_displays [i].enabled != NULL
|
||
&& *debug_displays [i].enabled)
|
||
{
|
||
struct dwarf_section *sec = &debug_displays [i].section;
|
||
|
||
if (strcmp (sec->uncompressed_name, match) == 0)
|
||
sec->name = sec->uncompressed_name;
|
||
else
|
||
sec->name = sec->compressed_name;
|
||
if (load_specific_debug_section ((enum dwarf_section_display_enum) i,
|
||
section, abfd))
|
||
{
|
||
debug_displays [i].display (sec, abfd);
|
||
|
||
if (i != info && i != abbrev)
|
||
free_debug_section ((enum dwarf_section_display_enum) i);
|
||
}
|
||
break;
|
||
}
|
||
}
|
||
|
||
/* Dump the dwarf debugging information. */
|
||
|
||
static void
|
||
dump_dwarf (bfd *abfd)
|
||
{
|
||
is_relocatable = (abfd->flags & (EXEC_P | DYNAMIC)) == 0;
|
||
|
||
eh_addr_size = bfd_arch_bits_per_address (abfd) / 8;
|
||
|
||
if (bfd_big_endian (abfd))
|
||
byte_get = byte_get_big_endian;
|
||
else if (bfd_little_endian (abfd))
|
||
byte_get = byte_get_little_endian;
|
||
else
|
||
abort ();
|
||
|
||
switch (bfd_get_arch (abfd))
|
||
{
|
||
case bfd_arch_i386:
|
||
switch (bfd_get_mach (abfd))
|
||
{
|
||
case bfd_mach_x86_64:
|
||
case bfd_mach_x86_64_intel_syntax:
|
||
case bfd_mach_x86_64_nacl:
|
||
case bfd_mach_x64_32:
|
||
case bfd_mach_x64_32_intel_syntax:
|
||
case bfd_mach_x64_32_nacl:
|
||
init_dwarf_regnames_x86_64 ();
|
||
break;
|
||
|
||
default:
|
||
init_dwarf_regnames_i386 ();
|
||
break;
|
||
}
|
||
break;
|
||
|
||
default:
|
||
break;
|
||
}
|
||
|
||
bfd_map_over_sections (abfd, dump_dwarf_section, NULL);
|
||
|
||
free_debug_memory ();
|
||
}
|
||
|
||
/* Read ABFD's stabs section STABSECT_NAME, and return a pointer to
|
||
it. Return NULL on failure. */
|
||
|
||
static char *
|
||
read_section_stabs (bfd *abfd, const char *sect_name, bfd_size_type *size_ptr)
|
||
{
|
||
asection *stabsect;
|
||
bfd_size_type size;
|
||
char *contents;
|
||
|
||
stabsect = bfd_get_section_by_name (abfd, sect_name);
|
||
if (stabsect == NULL)
|
||
{
|
||
printf (_("No %s section present\n\n"), sect_name);
|
||
return FALSE;
|
||
}
|
||
|
||
size = bfd_section_size (abfd, stabsect);
|
||
contents = (char *) xmalloc (size);
|
||
|
||
if (! bfd_get_section_contents (abfd, stabsect, contents, 0, size))
|
||
{
|
||
non_fatal (_("reading %s section of %s failed: %s"),
|
||
sect_name, bfd_get_filename (abfd),
|
||
bfd_errmsg (bfd_get_error ()));
|
||
exit_status = 1;
|
||
free (contents);
|
||
return NULL;
|
||
}
|
||
|
||
*size_ptr = size;
|
||
|
||
return contents;
|
||
}
|
||
|
||
/* Stabs entries use a 12 byte format:
|
||
4 byte string table index
|
||
1 byte stab type
|
||
1 byte stab other field
|
||
2 byte stab desc field
|
||
4 byte stab value
|
||
FIXME: This will have to change for a 64 bit object format. */
|
||
|
||
#define STRDXOFF (0)
|
||
#define TYPEOFF (4)
|
||
#define OTHEROFF (5)
|
||
#define DESCOFF (6)
|
||
#define VALOFF (8)
|
||
#define STABSIZE (12)
|
||
|
||
/* Print ABFD's stabs section STABSECT_NAME (in `stabs'),
|
||
using string table section STRSECT_NAME (in `strtab'). */
|
||
|
||
static void
|
||
print_section_stabs (bfd *abfd,
|
||
const char *stabsect_name,
|
||
unsigned *string_offset_ptr)
|
||
{
|
||
int i;
|
||
unsigned file_string_table_offset = 0;
|
||
unsigned next_file_string_table_offset = *string_offset_ptr;
|
||
bfd_byte *stabp, *stabs_end;
|
||
|
||
stabp = stabs;
|
||
stabs_end = stabp + stab_size;
|
||
|
||
printf (_("Contents of %s section:\n\n"), stabsect_name);
|
||
printf ("Symnum n_type n_othr n_desc n_value n_strx String\n");
|
||
|
||
/* Loop through all symbols and print them.
|
||
|
||
We start the index at -1 because there is a dummy symbol on
|
||
the front of stabs-in-{coff,elf} sections that supplies sizes. */
|
||
for (i = -1; stabp < stabs_end; stabp += STABSIZE, i++)
|
||
{
|
||
const char *name;
|
||
unsigned long strx;
|
||
unsigned char type, other;
|
||
unsigned short desc;
|
||
bfd_vma value;
|
||
|
||
strx = bfd_h_get_32 (abfd, stabp + STRDXOFF);
|
||
type = bfd_h_get_8 (abfd, stabp + TYPEOFF);
|
||
other = bfd_h_get_8 (abfd, stabp + OTHEROFF);
|
||
desc = bfd_h_get_16 (abfd, stabp + DESCOFF);
|
||
value = bfd_h_get_32 (abfd, stabp + VALOFF);
|
||
|
||
printf ("\n%-6d ", i);
|
||
/* Either print the stab name, or, if unnamed, print its number
|
||
again (makes consistent formatting for tools like awk). */
|
||
name = bfd_get_stab_name (type);
|
||
if (name != NULL)
|
||
printf ("%-6s", name);
|
||
else if (type == N_UNDF)
|
||
printf ("HdrSym");
|
||
else
|
||
printf ("%-6d", type);
|
||
printf (" %-6d %-6d ", other, desc);
|
||
bfd_printf_vma (abfd, value);
|
||
printf (" %-6lu", strx);
|
||
|
||
/* Symbols with type == 0 (N_UNDF) specify the length of the
|
||
string table associated with this file. We use that info
|
||
to know how to relocate the *next* file's string table indices. */
|
||
if (type == N_UNDF)
|
||
{
|
||
file_string_table_offset = next_file_string_table_offset;
|
||
next_file_string_table_offset += value;
|
||
}
|
||
else
|
||
{
|
||
/* Using the (possibly updated) string table offset, print the
|
||
string (if any) associated with this symbol. */
|
||
if ((strx + file_string_table_offset) < stabstr_size)
|
||
printf (" %s", &strtab[strx + file_string_table_offset]);
|
||
else
|
||
printf (" *");
|
||
}
|
||
}
|
||
printf ("\n\n");
|
||
*string_offset_ptr = next_file_string_table_offset;
|
||
}
|
||
|
||
typedef struct
|
||
{
|
||
const char * section_name;
|
||
const char * string_section_name;
|
||
unsigned string_offset;
|
||
}
|
||
stab_section_names;
|
||
|
||
static void
|
||
find_stabs_section (bfd *abfd, asection *section, void *names)
|
||
{
|
||
int len;
|
||
stab_section_names * sought = (stab_section_names *) names;
|
||
|
||
/* Check for section names for which stabsect_name is a prefix, to
|
||
handle .stab.N, etc. */
|
||
len = strlen (sought->section_name);
|
||
|
||
/* If the prefix matches, and the files section name ends with a
|
||
nul or a digit, then we match. I.e., we want either an exact
|
||
match or a section followed by a number. */
|
||
if (strncmp (sought->section_name, section->name, len) == 0
|
||
&& (section->name[len] == 0
|
||
|| (section->name[len] == '.' && ISDIGIT (section->name[len + 1]))))
|
||
{
|
||
if (strtab == NULL)
|
||
strtab = read_section_stabs (abfd, sought->string_section_name,
|
||
&stabstr_size);
|
||
|
||
if (strtab)
|
||
{
|
||
stabs = (bfd_byte *) read_section_stabs (abfd, section->name,
|
||
&stab_size);
|
||
if (stabs)
|
||
print_section_stabs (abfd, section->name, &sought->string_offset);
|
||
}
|
||
}
|
||
}
|
||
|
||
static void
|
||
dump_stabs_section (bfd *abfd, char *stabsect_name, char *strsect_name)
|
||
{
|
||
stab_section_names s;
|
||
|
||
s.section_name = stabsect_name;
|
||
s.string_section_name = strsect_name;
|
||
s.string_offset = 0;
|
||
|
||
bfd_map_over_sections (abfd, find_stabs_section, & s);
|
||
|
||
free (strtab);
|
||
strtab = NULL;
|
||
}
|
||
|
||
/* Dump the any sections containing stabs debugging information. */
|
||
|
||
static void
|
||
dump_stabs (bfd *abfd)
|
||
{
|
||
dump_stabs_section (abfd, ".stab", ".stabstr");
|
||
dump_stabs_section (abfd, ".stab.excl", ".stab.exclstr");
|
||
dump_stabs_section (abfd, ".stab.index", ".stab.indexstr");
|
||
|
||
/* For Darwin. */
|
||
dump_stabs_section (abfd, "LC_SYMTAB.stabs", "LC_SYMTAB.stabstr");
|
||
|
||
dump_stabs_section (abfd, "$GDB_SYMBOLS$", "$GDB_STRINGS$");
|
||
}
|
||
|
||
static void
|
||
dump_bfd_header (bfd *abfd)
|
||
{
|
||
char *comma = "";
|
||
|
||
printf (_("architecture: %s, "),
|
||
bfd_printable_arch_mach (bfd_get_arch (abfd),
|
||
bfd_get_mach (abfd)));
|
||
printf (_("flags 0x%08x:\n"), abfd->flags & ~BFD_FLAGS_FOR_BFD_USE_MASK);
|
||
|
||
#define PF(x, y) if (abfd->flags & x) {printf("%s%s", comma, y); comma=", ";}
|
||
PF (HAS_RELOC, "HAS_RELOC");
|
||
PF (EXEC_P, "EXEC_P");
|
||
PF (HAS_LINENO, "HAS_LINENO");
|
||
PF (HAS_DEBUG, "HAS_DEBUG");
|
||
PF (HAS_SYMS, "HAS_SYMS");
|
||
PF (HAS_LOCALS, "HAS_LOCALS");
|
||
PF (DYNAMIC, "DYNAMIC");
|
||
PF (WP_TEXT, "WP_TEXT");
|
||
PF (D_PAGED, "D_PAGED");
|
||
PF (BFD_IS_RELAXABLE, "BFD_IS_RELAXABLE");
|
||
PF (HAS_LOAD_PAGE, "HAS_LOAD_PAGE");
|
||
printf (_("\nstart address 0x"));
|
||
bfd_printf_vma (abfd, abfd->start_address);
|
||
printf ("\n");
|
||
}
|
||
|
||
|
||
static void
|
||
dump_bfd_private_header (bfd *abfd)
|
||
{
|
||
bfd_print_private_bfd_data (abfd, stdout);
|
||
}
|
||
|
||
static void
|
||
dump_target_specific (bfd *abfd)
|
||
{
|
||
const struct objdump_private_desc * const *desc;
|
||
struct objdump_private_option *opt;
|
||
char *e, *b;
|
||
|
||
/* Find the desc. */
|
||
for (desc = objdump_private_vectors; *desc != NULL; desc++)
|
||
if ((*desc)->filter (abfd))
|
||
break;
|
||
|
||
if (*desc == NULL)
|
||
{
|
||
non_fatal (_("option -P/--private not supported by this file"));
|
||
return;
|
||
}
|
||
|
||
/* Clear all options. */
|
||
for (opt = (*desc)->options; opt->name; opt++)
|
||
opt->selected = FALSE;
|
||
|
||
/* Decode options. */
|
||
b = dump_private_options;
|
||
do
|
||
{
|
||
e = strchr (b, ',');
|
||
|
||
if (e)
|
||
*e = 0;
|
||
|
||
for (opt = (*desc)->options; opt->name; opt++)
|
||
if (strcmp (opt->name, b) == 0)
|
||
{
|
||
opt->selected = TRUE;
|
||
break;
|
||
}
|
||
if (opt->name == NULL)
|
||
non_fatal (_("target specific dump '%s' not supported"), b);
|
||
|
||
if (e)
|
||
{
|
||
*e = ',';
|
||
b = e + 1;
|
||
}
|
||
}
|
||
while (e != NULL);
|
||
|
||
/* Dump. */
|
||
(*desc)->dump (abfd);
|
||
}
|
||
|
||
/* Display a section in hexadecimal format with associated characters.
|
||
Each line prefixed by the zero padded address. */
|
||
|
||
static void
|
||
dump_section (bfd *abfd, asection *section, void *dummy ATTRIBUTE_UNUSED)
|
||
{
|
||
bfd_byte *data = 0;
|
||
bfd_size_type datasize;
|
||
bfd_size_type addr_offset;
|
||
bfd_size_type start_offset;
|
||
bfd_size_type stop_offset;
|
||
unsigned int opb = bfd_octets_per_byte (abfd);
|
||
/* Bytes per line. */
|
||
const int onaline = 16;
|
||
char buf[64];
|
||
int count;
|
||
int width;
|
||
|
||
if ((section->flags & SEC_HAS_CONTENTS) == 0)
|
||
return;
|
||
|
||
if (! process_section_p (section))
|
||
return;
|
||
|
||
if ((datasize = bfd_section_size (abfd, section)) == 0)
|
||
return;
|
||
|
||
/* Compute the address range to display. */
|
||
if (start_address == (bfd_vma) -1
|
||
|| start_address < section->vma)
|
||
start_offset = 0;
|
||
else
|
||
start_offset = start_address - section->vma;
|
||
|
||
if (stop_address == (bfd_vma) -1)
|
||
stop_offset = datasize / opb;
|
||
else
|
||
{
|
||
if (stop_address < section->vma)
|
||
stop_offset = 0;
|
||
else
|
||
stop_offset = stop_address - section->vma;
|
||
|
||
if (stop_offset > datasize / opb)
|
||
stop_offset = datasize / opb;
|
||
}
|
||
|
||
if (start_offset >= stop_offset)
|
||
return;
|
||
|
||
printf (_("Contents of section %s:"), section->name);
|
||
if (display_file_offsets)
|
||
printf (_(" (Starting at file offset: 0x%lx)"),
|
||
(unsigned long) (section->filepos + start_offset));
|
||
printf ("\n");
|
||
|
||
if (!bfd_get_full_section_contents (abfd, section, &data))
|
||
{
|
||
non_fatal (_("Reading section failed"));
|
||
return;
|
||
}
|
||
|
||
width = 4;
|
||
|
||
bfd_sprintf_vma (abfd, buf, start_offset + section->vma);
|
||
if (strlen (buf) >= sizeof (buf))
|
||
abort ();
|
||
|
||
count = 0;
|
||
while (buf[count] == '0' && buf[count+1] != '\0')
|
||
count++;
|
||
count = strlen (buf) - count;
|
||
if (count > width)
|
||
width = count;
|
||
|
||
bfd_sprintf_vma (abfd, buf, stop_offset + section->vma - 1);
|
||
if (strlen (buf) >= sizeof (buf))
|
||
abort ();
|
||
|
||
count = 0;
|
||
while (buf[count] == '0' && buf[count+1] != '\0')
|
||
count++;
|
||
count = strlen (buf) - count;
|
||
if (count > width)
|
||
width = count;
|
||
|
||
for (addr_offset = start_offset;
|
||
addr_offset < stop_offset; addr_offset += onaline / opb)
|
||
{
|
||
bfd_size_type j;
|
||
|
||
bfd_sprintf_vma (abfd, buf, (addr_offset + section->vma));
|
||
count = strlen (buf);
|
||
if ((size_t) count >= sizeof (buf))
|
||
abort ();
|
||
|
||
putchar (' ');
|
||
while (count < width)
|
||
{
|
||
putchar ('0');
|
||
count++;
|
||
}
|
||
fputs (buf + count - width, stdout);
|
||
putchar (' ');
|
||
|
||
for (j = addr_offset * opb;
|
||
j < addr_offset * opb + onaline; j++)
|
||
{
|
||
if (j < stop_offset * opb)
|
||
printf ("%02x", (unsigned) (data[j]));
|
||
else
|
||
printf (" ");
|
||
if ((j & 3) == 3)
|
||
printf (" ");
|
||
}
|
||
|
||
printf (" ");
|
||
for (j = addr_offset * opb;
|
||
j < addr_offset * opb + onaline; j++)
|
||
{
|
||
if (j >= stop_offset * opb)
|
||
printf (" ");
|
||
else
|
||
printf ("%c", ISPRINT (data[j]) ? data[j] : '.');
|
||
}
|
||
putchar ('\n');
|
||
}
|
||
free (data);
|
||
}
|
||
|
||
/* Actually display the various requested regions. */
|
||
|
||
static void
|
||
dump_data (bfd *abfd)
|
||
{
|
||
bfd_map_over_sections (abfd, dump_section, NULL);
|
||
}
|
||
|
||
/* Should perhaps share code and display with nm? */
|
||
|
||
static void
|
||
dump_symbols (bfd *abfd ATTRIBUTE_UNUSED, bfd_boolean dynamic)
|
||
{
|
||
asymbol **current;
|
||
long max_count;
|
||
long count;
|
||
|
||
if (dynamic)
|
||
{
|
||
current = dynsyms;
|
||
max_count = dynsymcount;
|
||
printf ("DYNAMIC SYMBOL TABLE:\n");
|
||
}
|
||
else
|
||
{
|
||
current = syms;
|
||
max_count = symcount;
|
||
printf ("SYMBOL TABLE:\n");
|
||
}
|
||
|
||
if (max_count == 0)
|
||
printf (_("no symbols\n"));
|
||
|
||
for (count = 0; count < max_count; count++)
|
||
{
|
||
bfd *cur_bfd;
|
||
|
||
if (*current == NULL)
|
||
printf (_("no information for symbol number %ld\n"), count);
|
||
|
||
else if ((cur_bfd = bfd_asymbol_bfd (*current)) == NULL)
|
||
printf (_("could not determine the type of symbol number %ld\n"),
|
||
count);
|
||
|
||
else if (process_section_p ((* current)->section)
|
||
&& (dump_special_syms
|
||
|| !bfd_is_target_special_symbol (cur_bfd, *current)))
|
||
{
|
||
const char *name = (*current)->name;
|
||
|
||
if (do_demangle && name != NULL && *name != '\0')
|
||
{
|
||
char *alloc;
|
||
|
||
/* If we want to demangle the name, we demangle it
|
||
here, and temporarily clobber it while calling
|
||
bfd_print_symbol. FIXME: This is a gross hack. */
|
||
alloc = bfd_demangle (cur_bfd, name, DMGL_ANSI | DMGL_PARAMS);
|
||
if (alloc != NULL)
|
||
(*current)->name = alloc;
|
||
bfd_print_symbol (cur_bfd, stdout, *current,
|
||
bfd_print_symbol_all);
|
||
if (alloc != NULL)
|
||
{
|
||
(*current)->name = name;
|
||
free (alloc);
|
||
}
|
||
}
|
||
else
|
||
bfd_print_symbol (cur_bfd, stdout, *current,
|
||
bfd_print_symbol_all);
|
||
printf ("\n");
|
||
}
|
||
|
||
current++;
|
||
}
|
||
printf ("\n\n");
|
||
}
|
||
|
||
static void
|
||
dump_reloc_set (bfd *abfd, asection *sec, arelent **relpp, long relcount)
|
||
{
|
||
arelent **p;
|
||
char *last_filename, *last_functionname;
|
||
unsigned int last_line;
|
||
unsigned int last_discriminator;
|
||
|
||
/* Get column headers lined up reasonably. */
|
||
{
|
||
static int width;
|
||
|
||
if (width == 0)
|
||
{
|
||
char buf[30];
|
||
|
||
bfd_sprintf_vma (abfd, buf, (bfd_vma) -1);
|
||
width = strlen (buf) - 7;
|
||
}
|
||
printf ("OFFSET %*s TYPE %*s VALUE \n", width, "", 12, "");
|
||
}
|
||
|
||
last_filename = NULL;
|
||
last_functionname = NULL;
|
||
last_line = 0;
|
||
last_discriminator = 0;
|
||
|
||
for (p = relpp; relcount && *p != NULL; p++, relcount--)
|
||
{
|
||
arelent *q = *p;
|
||
const char *filename, *functionname;
|
||
unsigned int linenumber;
|
||
unsigned int discriminator;
|
||
const char *sym_name;
|
||
const char *section_name;
|
||
bfd_vma addend2 = 0;
|
||
|
||
if (start_address != (bfd_vma) -1
|
||
&& q->address < start_address)
|
||
continue;
|
||
if (stop_address != (bfd_vma) -1
|
||
&& q->address > stop_address)
|
||
continue;
|
||
|
||
if (with_line_numbers
|
||
&& sec != NULL
|
||
&& bfd_find_nearest_line_discriminator (abfd, sec, syms, q->address,
|
||
&filename, &functionname,
|
||
&linenumber, &discriminator))
|
||
{
|
||
if (functionname != NULL
|
||
&& (last_functionname == NULL
|
||
|| strcmp (functionname, last_functionname) != 0))
|
||
{
|
||
printf ("%s():\n", functionname);
|
||
if (last_functionname != NULL)
|
||
free (last_functionname);
|
||
last_functionname = xstrdup (functionname);
|
||
}
|
||
|
||
if (linenumber > 0
|
||
&& (linenumber != last_line
|
||
|| (filename != NULL
|
||
&& last_filename != NULL
|
||
&& filename_cmp (filename, last_filename) != 0)
|
||
|| (discriminator != last_discriminator)))
|
||
{
|
||
if (discriminator > 0)
|
||
printf ("%s:%u\n", filename == NULL ? "???" : filename, linenumber);
|
||
else
|
||
printf ("%s:%u (discriminator %u)\n", filename == NULL ? "???" : filename,
|
||
linenumber, discriminator);
|
||
last_line = linenumber;
|
||
last_discriminator = discriminator;
|
||
if (last_filename != NULL)
|
||
free (last_filename);
|
||
if (filename == NULL)
|
||
last_filename = NULL;
|
||
else
|
||
last_filename = xstrdup (filename);
|
||
}
|
||
}
|
||
|
||
if (q->sym_ptr_ptr && *q->sym_ptr_ptr)
|
||
{
|
||
sym_name = (*(q->sym_ptr_ptr))->name;
|
||
section_name = (*(q->sym_ptr_ptr))->section->name;
|
||
}
|
||
else
|
||
{
|
||
sym_name = NULL;
|
||
section_name = NULL;
|
||
}
|
||
|
||
bfd_printf_vma (abfd, q->address);
|
||
if (q->howto == NULL)
|
||
printf (" *unknown* ");
|
||
else if (q->howto->name)
|
||
{
|
||
const char *name = q->howto->name;
|
||
|
||
/* R_SPARC_OLO10 relocations contain two addends.
|
||
But because 'arelent' lacks enough storage to
|
||
store them both, the 64-bit ELF Sparc backend
|
||
records this as two relocations. One R_SPARC_LO10
|
||
and one R_SPARC_13, both pointing to the same
|
||
address. This is merely so that we have some
|
||
place to store both addend fields.
|
||
|
||
Undo this transformation, otherwise the output
|
||
will be confusing. */
|
||
if (abfd->xvec->flavour == bfd_target_elf_flavour
|
||
&& elf_tdata(abfd)->elf_header->e_machine == EM_SPARCV9
|
||
&& relcount > 1
|
||
&& !strcmp (q->howto->name, "R_SPARC_LO10"))
|
||
{
|
||
arelent *q2 = *(p + 1);
|
||
if (q2 != NULL
|
||
&& q2->howto
|
||
&& q->address == q2->address
|
||
&& !strcmp (q2->howto->name, "R_SPARC_13"))
|
||
{
|
||
name = "R_SPARC_OLO10";
|
||
addend2 = q2->addend;
|
||
p++;
|
||
}
|
||
}
|
||
printf (" %-16s ", name);
|
||
}
|
||
else
|
||
printf (" %-16d ", q->howto->type);
|
||
|
||
if (sym_name)
|
||
{
|
||
objdump_print_symname (abfd, NULL, *q->sym_ptr_ptr);
|
||
}
|
||
else
|
||
{
|
||
if (section_name == NULL)
|
||
section_name = "*unknown*";
|
||
printf ("[%s]", section_name);
|
||
}
|
||
|
||
if (q->addend)
|
||
{
|
||
bfd_signed_vma addend = q->addend;
|
||
if (addend < 0)
|
||
{
|
||
printf ("-0x");
|
||
addend = -addend;
|
||
}
|
||
else
|
||
printf ("+0x");
|
||
bfd_printf_vma (abfd, addend);
|
||
}
|
||
if (addend2)
|
||
{
|
||
printf ("+0x");
|
||
bfd_printf_vma (abfd, addend2);
|
||
}
|
||
|
||
printf ("\n");
|
||
}
|
||
|
||
if (last_filename != NULL)
|
||
free (last_filename);
|
||
if (last_functionname != NULL)
|
||
free (last_functionname);
|
||
}
|
||
|
||
static void
|
||
dump_relocs_in_section (bfd *abfd,
|
||
asection *section,
|
||
void *dummy ATTRIBUTE_UNUSED)
|
||
{
|
||
arelent **relpp;
|
||
long relcount;
|
||
long relsize;
|
||
|
||
if ( bfd_is_abs_section (section)
|
||
|| bfd_is_und_section (section)
|
||
|| bfd_is_com_section (section)
|
||
|| (! process_section_p (section))
|
||
|| ((section->flags & SEC_RELOC) == 0))
|
||
return;
|
||
|
||
relsize = bfd_get_reloc_upper_bound (abfd, section);
|
||
if (relsize < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
|
||
printf ("RELOCATION RECORDS FOR [%s]:", section->name);
|
||
|
||
if (relsize == 0)
|
||
{
|
||
printf (" (none)\n\n");
|
||
return;
|
||
}
|
||
|
||
relpp = (arelent **) xmalloc (relsize);
|
||
relcount = bfd_canonicalize_reloc (abfd, section, relpp, syms);
|
||
|
||
if (relcount < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
else if (relcount == 0)
|
||
printf (" (none)\n\n");
|
||
else
|
||
{
|
||
printf ("\n");
|
||
dump_reloc_set (abfd, section, relpp, relcount);
|
||
printf ("\n\n");
|
||
}
|
||
free (relpp);
|
||
}
|
||
|
||
static void
|
||
dump_relocs (bfd *abfd)
|
||
{
|
||
bfd_map_over_sections (abfd, dump_relocs_in_section, NULL);
|
||
}
|
||
|
||
static void
|
||
dump_dynamic_relocs (bfd *abfd)
|
||
{
|
||
long relsize;
|
||
arelent **relpp;
|
||
long relcount;
|
||
|
||
relsize = bfd_get_dynamic_reloc_upper_bound (abfd);
|
||
if (relsize < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
|
||
printf ("DYNAMIC RELOCATION RECORDS");
|
||
|
||
if (relsize == 0)
|
||
printf (" (none)\n\n");
|
||
else
|
||
{
|
||
relpp = (arelent **) xmalloc (relsize);
|
||
relcount = bfd_canonicalize_dynamic_reloc (abfd, relpp, dynsyms);
|
||
|
||
if (relcount < 0)
|
||
bfd_fatal (bfd_get_filename (abfd));
|
||
else if (relcount == 0)
|
||
printf (" (none)\n\n");
|
||
else
|
||
{
|
||
printf ("\n");
|
||
dump_reloc_set (abfd, NULL, relpp, relcount);
|
||
printf ("\n\n");
|
||
}
|
||
free (relpp);
|
||
}
|
||
}
|
||
|
||
/* Creates a table of paths, to search for source files. */
|
||
|
||
static void
|
||
add_include_path (const char *path)
|
||
{
|
||
if (path[0] == 0)
|
||
return;
|
||
include_path_count++;
|
||
include_paths = (const char **)
|
||
xrealloc (include_paths, include_path_count * sizeof (*include_paths));
|
||
#ifdef HAVE_DOS_BASED_FILE_SYSTEM
|
||
if (path[1] == ':' && path[2] == 0)
|
||
path = concat (path, ".", (const char *) 0);
|
||
#endif
|
||
include_paths[include_path_count - 1] = path;
|
||
}
|
||
|
||
static void
|
||
adjust_addresses (bfd *abfd ATTRIBUTE_UNUSED,
|
||
asection *section,
|
||
void *arg)
|
||
{
|
||
if ((section->flags & SEC_DEBUGGING) == 0)
|
||
{
|
||
bfd_boolean *has_reloc_p = (bfd_boolean *) arg;
|
||
section->vma += adjust_section_vma;
|
||
if (*has_reloc_p)
|
||
section->lma += adjust_section_vma;
|
||
}
|
||
}
|
||
|
||
/* Dump selected contents of ABFD. */
|
||
|
||
static void
|
||
dump_bfd (bfd *abfd)
|
||
{
|
||
/* If we are adjusting section VMA's, change them all now. Changing
|
||
the BFD information is a hack. However, we must do it, or
|
||
bfd_find_nearest_line will not do the right thing. */
|
||
if (adjust_section_vma != 0)
|
||
{
|
||
bfd_boolean has_reloc = (abfd->flags & HAS_RELOC);
|
||
bfd_map_over_sections (abfd, adjust_addresses, &has_reloc);
|
||
}
|
||
|
||
if (! dump_debugging_tags && ! suppress_bfd_header)
|
||
printf (_("\n%s: file format %s\n"), bfd_get_filename (abfd),
|
||
abfd->xvec->name);
|
||
if (dump_ar_hdrs)
|
||
print_arelt_descr (stdout, abfd, TRUE);
|
||
if (dump_file_header)
|
||
dump_bfd_header (abfd);
|
||
if (dump_private_headers)
|
||
dump_bfd_private_header (abfd);
|
||
if (dump_private_options != NULL)
|
||
dump_target_specific (abfd);
|
||
if (! dump_debugging_tags && ! suppress_bfd_header)
|
||
putchar ('\n');
|
||
|
||
if (dump_symtab
|
||
|| dump_reloc_info
|
||
|| disassemble
|
||
|| dump_debugging
|
||
|| dump_dwarf_section_info)
|
||
syms = slurp_symtab (abfd);
|
||
|
||
if (dump_section_headers)
|
||
dump_headers (abfd);
|
||
|
||
if (dump_dynamic_symtab || dump_dynamic_reloc_info
|
||
|| (disassemble && bfd_get_dynamic_symtab_upper_bound (abfd) > 0))
|
||
dynsyms = slurp_dynamic_symtab (abfd);
|
||
if (disassemble)
|
||
{
|
||
synthcount = bfd_get_synthetic_symtab (abfd, symcount, syms,
|
||
dynsymcount, dynsyms, &synthsyms);
|
||
if (synthcount < 0)
|
||
synthcount = 0;
|
||
}
|
||
|
||
if (dump_symtab)
|
||
dump_symbols (abfd, FALSE);
|
||
if (dump_dynamic_symtab)
|
||
dump_symbols (abfd, TRUE);
|
||
if (dump_dwarf_section_info)
|
||
dump_dwarf (abfd);
|
||
if (dump_stab_section_info)
|
||
dump_stabs (abfd);
|
||
if (dump_reloc_info && ! disassemble)
|
||
dump_relocs (abfd);
|
||
if (dump_dynamic_reloc_info && ! disassemble)
|
||
dump_dynamic_relocs (abfd);
|
||
if (dump_section_contents)
|
||
dump_data (abfd);
|
||
if (disassemble)
|
||
disassemble_data (abfd);
|
||
|
||
if (dump_debugging)
|
||
{
|
||
void *dhandle;
|
||
|
||
dhandle = read_debugging_info (abfd, syms, symcount, TRUE);
|
||
if (dhandle != NULL)
|
||
{
|
||
if (!print_debugging_info (stdout, dhandle, abfd, syms,
|
||
bfd_demangle,
|
||
dump_debugging_tags ? TRUE : FALSE))
|
||
{
|
||
non_fatal (_("%s: printing debugging information failed"),
|
||
bfd_get_filename (abfd));
|
||
exit_status = 1;
|
||
}
|
||
}
|
||
/* PR 6483: If there was no STABS or IEEE debug
|
||
info in the file, try DWARF instead. */
|
||
else if (! dump_dwarf_section_info)
|
||
{
|
||
dwarf_select_sections_all ();
|
||
dump_dwarf (abfd);
|
||
}
|
||
}
|
||
|
||
if (syms)
|
||
{
|
||
free (syms);
|
||
syms = NULL;
|
||
}
|
||
|
||
if (dynsyms)
|
||
{
|
||
free (dynsyms);
|
||
dynsyms = NULL;
|
||
}
|
||
|
||
if (synthsyms)
|
||
{
|
||
free (synthsyms);
|
||
synthsyms = NULL;
|
||
}
|
||
|
||
symcount = 0;
|
||
dynsymcount = 0;
|
||
synthcount = 0;
|
||
}
|
||
|
||
static void
|
||
display_object_bfd (bfd *abfd)
|
||
{
|
||
char **matching;
|
||
|
||
if (bfd_check_format_matches (abfd, bfd_object, &matching))
|
||
{
|
||
dump_bfd (abfd);
|
||
return;
|
||
}
|
||
|
||
if (bfd_get_error () == bfd_error_file_ambiguously_recognized)
|
||
{
|
||
nonfatal (bfd_get_filename (abfd));
|
||
list_matching_formats (matching);
|
||
free (matching);
|
||
return;
|
||
}
|
||
|
||
if (bfd_get_error () != bfd_error_file_not_recognized)
|
||
{
|
||
nonfatal (bfd_get_filename (abfd));
|
||
return;
|
||
}
|
||
|
||
if (bfd_check_format_matches (abfd, bfd_core, &matching))
|
||
{
|
||
dump_bfd (abfd);
|
||
return;
|
||
}
|
||
|
||
nonfatal (bfd_get_filename (abfd));
|
||
|
||
if (bfd_get_error () == bfd_error_file_ambiguously_recognized)
|
||
{
|
||
list_matching_formats (matching);
|
||
free (matching);
|
||
}
|
||
}
|
||
|
||
static void
|
||
display_any_bfd (bfd *file, int level)
|
||
{
|
||
/* Decompress sections unless dumping the section contents. */
|
||
if (!dump_section_contents)
|
||
file->flags |= BFD_DECOMPRESS;
|
||
|
||
/* If the file is an archive, process all of its elements. */
|
||
if (bfd_check_format (file, bfd_archive))
|
||
{
|
||
bfd *arfile = NULL;
|
||
bfd *last_arfile = NULL;
|
||
|
||
if (level == 0)
|
||
printf (_("In archive %s:\n"), bfd_get_filename (file));
|
||
else
|
||
printf (_("In nested archive %s:\n"), bfd_get_filename (file));
|
||
|
||
for (;;)
|
||
{
|
||
bfd_set_error (bfd_error_no_error);
|
||
|
||
arfile = bfd_openr_next_archived_file (file, arfile);
|
||
if (arfile == NULL)
|
||
{
|
||
if (bfd_get_error () != bfd_error_no_more_archived_files)
|
||
nonfatal (bfd_get_filename (file));
|
||
break;
|
||
}
|
||
|
||
display_any_bfd (arfile, level + 1);
|
||
|
||
if (last_arfile != NULL)
|
||
bfd_close (last_arfile);
|
||
last_arfile = arfile;
|
||
}
|
||
|
||
if (last_arfile != NULL)
|
||
bfd_close (last_arfile);
|
||
}
|
||
else
|
||
display_object_bfd (file);
|
||
}
|
||
|
||
static void
|
||
display_file (char *filename, char *target)
|
||
{
|
||
bfd *file;
|
||
|
||
if (get_file_size (filename) < 1)
|
||
{
|
||
exit_status = 1;
|
||
return;
|
||
}
|
||
|
||
file = bfd_openr (filename, target);
|
||
if (file == NULL)
|
||
{
|
||
nonfatal (filename);
|
||
return;
|
||
}
|
||
|
||
display_any_bfd (file, 0);
|
||
|
||
bfd_close (file);
|
||
}
|
||
|
||
int
|
||
main (int argc, char **argv)
|
||
{
|
||
int c;
|
||
char *target = default_target;
|
||
bfd_boolean seenflag = FALSE;
|
||
|
||
#if defined (HAVE_SETLOCALE)
|
||
#if defined (HAVE_LC_MESSAGES)
|
||
setlocale (LC_MESSAGES, "");
|
||
#endif
|
||
setlocale (LC_CTYPE, "");
|
||
#endif
|
||
|
||
bindtextdomain (PACKAGE, LOCALEDIR);
|
||
textdomain (PACKAGE);
|
||
|
||
program_name = *argv;
|
||
xmalloc_set_program_name (program_name);
|
||
|
||
START_PROGRESS (program_name, 0);
|
||
|
||
expandargv (&argc, &argv);
|
||
|
||
bfd_init ();
|
||
set_default_bfd_target ();
|
||
|
||
while ((c = getopt_long (argc, argv,
|
||
"pP:ib:m:M:VvCdDlfFaHhrRtTxsSI:j:wE:zgeGW::",
|
||
long_options, (int *) 0))
|
||
!= EOF)
|
||
{
|
||
switch (c)
|
||
{
|
||
case 0:
|
||
break; /* We've been given a long option. */
|
||
case 'm':
|
||
machine = optarg;
|
||
break;
|
||
case 'M':
|
||
if (disassembler_options)
|
||
/* Ignore potential memory leak for now. */
|
||
disassembler_options = concat (disassembler_options, ",",
|
||
optarg, (const char *) NULL);
|
||
else
|
||
disassembler_options = optarg;
|
||
break;
|
||
case 'j':
|
||
add_only (optarg);
|
||
break;
|
||
case 'F':
|
||
display_file_offsets = TRUE;
|
||
break;
|
||
case 'l':
|
||
with_line_numbers = TRUE;
|
||
break;
|
||
case 'b':
|
||
target = optarg;
|
||
break;
|
||
case 'C':
|
||
do_demangle = TRUE;
|
||
if (optarg != NULL)
|
||
{
|
||
enum demangling_styles style;
|
||
|
||
style = cplus_demangle_name_to_style (optarg);
|
||
if (style == unknown_demangling)
|
||
fatal (_("unknown demangling style `%s'"),
|
||
optarg);
|
||
|
||
cplus_demangle_set_style (style);
|
||
}
|
||
break;
|
||
case 'w':
|
||
wide_output = TRUE;
|
||
break;
|
||
case OPTION_ADJUST_VMA:
|
||
adjust_section_vma = parse_vma (optarg, "--adjust-vma");
|
||
break;
|
||
case OPTION_START_ADDRESS:
|
||
start_address = parse_vma (optarg, "--start-address");
|
||
if ((stop_address != (bfd_vma) -1) && stop_address <= start_address)
|
||
fatal (_("error: the start address should be before the end address"));
|
||
break;
|
||
case OPTION_STOP_ADDRESS:
|
||
stop_address = parse_vma (optarg, "--stop-address");
|
||
if ((start_address != (bfd_vma) -1) && stop_address <= start_address)
|
||
fatal (_("error: the stop address should be after the start address"));
|
||
break;
|
||
case OPTION_PREFIX:
|
||
prefix = optarg;
|
||
prefix_length = strlen (prefix);
|
||
/* Remove an unnecessary trailing '/' */
|
||
while (IS_DIR_SEPARATOR (prefix[prefix_length - 1]))
|
||
prefix_length--;
|
||
break;
|
||
case OPTION_PREFIX_STRIP:
|
||
prefix_strip = atoi (optarg);
|
||
if (prefix_strip < 0)
|
||
fatal (_("error: prefix strip must be non-negative"));
|
||
break;
|
||
case OPTION_INSN_WIDTH:
|
||
insn_width = strtoul (optarg, NULL, 0);
|
||
if (insn_width <= 0)
|
||
fatal (_("error: instruction width must be positive"));
|
||
break;
|
||
case 'E':
|
||
if (strcmp (optarg, "B") == 0)
|
||
endian = BFD_ENDIAN_BIG;
|
||
else if (strcmp (optarg, "L") == 0)
|
||
endian = BFD_ENDIAN_LITTLE;
|
||
else
|
||
{
|
||
nonfatal (_("unrecognized -E option"));
|
||
usage (stderr, 1);
|
||
}
|
||
break;
|
||
case OPTION_ENDIAN:
|
||
if (strncmp (optarg, "big", strlen (optarg)) == 0)
|
||
endian = BFD_ENDIAN_BIG;
|
||
else if (strncmp (optarg, "little", strlen (optarg)) == 0)
|
||
endian = BFD_ENDIAN_LITTLE;
|
||
else
|
||
{
|
||
non_fatal (_("unrecognized --endian type `%s'"), optarg);
|
||
exit_status = 1;
|
||
usage (stderr, 1);
|
||
}
|
||
break;
|
||
|
||
case 'f':
|
||
dump_file_header = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'i':
|
||
formats_info = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'I':
|
||
add_include_path (optarg);
|
||
break;
|
||
case 'p':
|
||
dump_private_headers = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'P':
|
||
dump_private_options = optarg;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'x':
|
||
dump_private_headers = TRUE;
|
||
dump_symtab = TRUE;
|
||
dump_reloc_info = TRUE;
|
||
dump_file_header = TRUE;
|
||
dump_ar_hdrs = TRUE;
|
||
dump_section_headers = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 't':
|
||
dump_symtab = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'T':
|
||
dump_dynamic_symtab = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'd':
|
||
disassemble = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'z':
|
||
disassemble_zeroes = TRUE;
|
||
break;
|
||
case 'D':
|
||
disassemble = TRUE;
|
||
disassemble_all = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'S':
|
||
disassemble = TRUE;
|
||
with_source_code = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'g':
|
||
dump_debugging = 1;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'e':
|
||
dump_debugging = 1;
|
||
dump_debugging_tags = 1;
|
||
do_demangle = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'W':
|
||
dump_dwarf_section_info = TRUE;
|
||
seenflag = TRUE;
|
||
if (optarg)
|
||
dwarf_select_sections_by_letters (optarg);
|
||
else
|
||
dwarf_select_sections_all ();
|
||
break;
|
||
case OPTION_DWARF:
|
||
dump_dwarf_section_info = TRUE;
|
||
seenflag = TRUE;
|
||
if (optarg)
|
||
dwarf_select_sections_by_names (optarg);
|
||
else
|
||
dwarf_select_sections_all ();
|
||
break;
|
||
case OPTION_DWARF_DEPTH:
|
||
{
|
||
char *cp;
|
||
dwarf_cutoff_level = strtoul (optarg, & cp, 0);
|
||
}
|
||
break;
|
||
case OPTION_DWARF_START:
|
||
{
|
||
char *cp;
|
||
dwarf_start_die = strtoul (optarg, & cp, 0);
|
||
suppress_bfd_header = 1;
|
||
}
|
||
break;
|
||
case OPTION_DWARF_CHECK:
|
||
dwarf_check = TRUE;
|
||
break;
|
||
case 'G':
|
||
dump_stab_section_info = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 's':
|
||
dump_section_contents = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'r':
|
||
dump_reloc_info = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'R':
|
||
dump_dynamic_reloc_info = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'a':
|
||
dump_ar_hdrs = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'h':
|
||
dump_section_headers = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
case 'v':
|
||
case 'V':
|
||
show_version = TRUE;
|
||
seenflag = TRUE;
|
||
break;
|
||
|
||
case 'H':
|
||
usage (stdout, 0);
|
||
/* No need to set seenflag or to break - usage() does not return. */
|
||
default:
|
||
usage (stderr, 1);
|
||
}
|
||
}
|
||
|
||
if (show_version)
|
||
print_version ("objdump");
|
||
|
||
if (!seenflag)
|
||
usage (stderr, 2);
|
||
|
||
if (formats_info)
|
||
exit_status = display_info ();
|
||
else
|
||
{
|
||
if (optind == argc)
|
||
display_file ("a.out", target);
|
||
else
|
||
for (; optind < argc;)
|
||
display_file (argv[optind++], target);
|
||
}
|
||
|
||
free_only_list ();
|
||
|
||
END_PROGRESS (program_name);
|
||
|
||
return exit_status;
|
||
}
|