#include "dwarf2.h"
#include "dwarf.h"
#include "gdb/gdb-index.h"
+#include "filenames.h"
#include <assert.h>
#undef MAX
int do_debug_addr;
int do_debug_cu_index;
int do_wide;
+int do_debug_links;
+int do_follow_links;
int dwarf_cutoff_level = -1;
unsigned long dwarf_start_die;
static unsigned int shndx_pool_size = 0;
static unsigned int shndx_pool_used = 0;
+/* Pointer to a separate file containing extra debug information. */
+static void * separate_debug_file = NULL;
+static const char * separate_debug_filename = NULL;
+
/* For version 2 package files, each set contains an array of section offsets
and an array of section sizes, giving the offset and size of the
contribution from a CU or TU within one of the debug sections.
* low_bits = tmp;
}
+static const char *
+fetch_alt_indirect_string (dwarf_vma offset)
+{
+ struct dwarf_section * section;
+ const char * ret;
+
+ if (! do_follow_links)
+ return "";
+
+ if (separate_debug_file == NULL)
+ return _("<following link not possible>");
+
+ if (! load_debug_section (separate_debug_str, separate_debug_file))
+ return _("<could not load separate string section>");
+
+ section = &debug_displays [separate_debug_str].section;
+ if (section->start == NULL)
+ return _("<no .debug_str section>");
+
+ if (offset >= section->size)
+ {
+ warn (_("DW_FORM_GNU_strp_alt offset too big: %s\n"), dwarf_vmatoa ("x", offset));
+ return _("<offset is too big>");
+ }
+
+ ret = (const char *) (section->start + offset);
+ /* Unfortunately we cannot rely upon the .debug_str section ending with a
+ NUL byte. Since our caller is expecting to receive a well formed C
+ string we test for the lack of a terminating byte here. */
+ if (strnlen ((const char *) ret, section->size - offset)
+ == section->size - offset)
+ return _("<no NUL byte at end of .debug_str section>");
+
+ return ret;
+}
+
static unsigned char *
-read_and_display_attr_value (unsigned long attribute,
- unsigned long form,
- dwarf_signed_vma implicit_const,
- unsigned char * data,
- unsigned char * end,
- dwarf_vma cu_offset,
- dwarf_vma pointer_size,
- dwarf_vma offset_size,
- int dwarf_version,
- debug_info * debug_info_p,
- int do_loc,
- struct dwarf_section * section,
- struct cu_tu_set * this_set, char delimiter)
+read_and_display_attr_value (unsigned long attribute,
+ unsigned long form,
+ dwarf_signed_vma implicit_const,
+ unsigned char * data,
+ unsigned char * end,
+ dwarf_vma cu_offset,
+ dwarf_vma pointer_size,
+ dwarf_vma offset_size,
+ int dwarf_version,
+ debug_info * debug_info_p,
+ int do_loc,
+ struct dwarf_section * section,
+ struct cu_tu_set * this_set,
+ char delimiter)
{
dwarf_vma uvalue = 0;
unsigned char *block_start = NULL;
case DW_FORM_GNU_ref_alt:
if (!do_loc)
printf ("%c<alt 0x%s>", delimiter, dwarf_vmatoa ("x",uvalue));
+ /* FIXME: Follow the reference... */
break;
case DW_FORM_ref1:
case DW_FORM_GNU_strp_alt:
if (!do_loc)
- printf (_("%c(alt indirect string, offset: 0x%s)"), delimiter,
- dwarf_vmatoa ("x", uvalue));
+ {
+ printf (_("%c(alt indirect string, offset: 0x%s) %s"), delimiter,
+ dwarf_vmatoa ("x", uvalue),
+ fetch_alt_indirect_string (uvalue));
+ }
break;
case DW_FORM_indirect:
}
static unsigned char *
-read_and_display_attr (unsigned long attribute,
- unsigned long form,
- dwarf_signed_vma implicit_const,
- unsigned char * data,
- unsigned char * end,
- dwarf_vma cu_offset,
- dwarf_vma pointer_size,
- dwarf_vma offset_size,
- int dwarf_version,
- debug_info * debug_info_p,
- int do_loc,
- struct dwarf_section * section,
- struct cu_tu_set * this_set)
+read_and_display_attr (unsigned long attribute,
+ unsigned long form,
+ dwarf_signed_vma implicit_const,
+ unsigned char * data,
+ unsigned char * end,
+ dwarf_vma cu_offset,
+ dwarf_vma pointer_size,
+ dwarf_vma offset_size,
+ int dwarf_version,
+ debug_info * debug_info_p,
+ int do_loc,
+ struct dwarf_section * section,
+ struct cu_tu_set * this_set)
{
if (!do_loc)
printf (" %-18s:", get_AT_name (attribute));
return data;
}
+/* Like load_debug_section, but if the ordinary call fails, and we are
+ following debug links, and we have been able to load a separate debug
+ info file, then attempt to load the requested section from the separate
+ file. */
+
+static bfd_boolean
+load_debug_section_with_follow (enum dwarf_section_display_enum sec_enum,
+ void * data)
+{
+ if (load_debug_section (sec_enum, data))
+ {
+ if (data == separate_debug_file)
+ debug_displays[sec_enum].section.filename = separate_debug_filename;
+
+ /* FIXME: We should check to see if there is a separate debug info file
+ that also contains this section, and if so, issue a warning. */
+ return TRUE;
+ }
+
+ if (do_follow_links && separate_debug_file != NULL)
+ if (load_debug_section (sec_enum, separate_debug_file))
+ {
+ debug_displays[sec_enum].section.filename = separate_debug_filename;
+ return TRUE;
+ }
+
+ return FALSE;
+}
+
+static void
+introduce (struct dwarf_section * section, bfd_boolean raw)
+{
+ if (raw)
+ {
+ if (do_follow_links && section->filename)
+ printf (_("Raw dump of debug contents of section %s (loaded from %s):\n\n"),
+ section->name, section->filename);
+ else
+ printf (_("Raw dump of debug contents of section %s:\n\n"), section->name);
+ }
+ else
+ {
+ if (do_follow_links && section->filename)
+ printf (_("Contents of the %s section (loaded from %s):\n\n"),
+ section->name, section->filename);
+ else
+ printf (_("Contents of the %s section:\n\n"), section->name);
+ }
+}
+
/* Process the contents of a .debug_info section. If do_loc is non-zero
then we are scanning for location lists and we do not want to display
anything to the user. If do_types is non-zero, we are processing
if (!do_loc)
{
- if (dwarf_start_die == 0)
- printf (_("Contents of the %s section:\n\n"), section->name);
-
- load_debug_section (str, file);
- load_debug_section (line_str, file);
- load_debug_section (str_dwo, file);
- load_debug_section (str_index, file);
- load_debug_section (str_index_dwo, file);
- load_debug_section (debug_addr, file);
+ load_debug_section_with_follow (str, file);
+ load_debug_section_with_follow (line_str, file);
+ load_debug_section_with_follow (str_dwo, file);
+ load_debug_section_with_follow (str_index, file);
+ load_debug_section_with_follow (str_index_dwo, file);
+ load_debug_section_with_follow (debug_addr, file);
}
- load_debug_section (abbrev_sec, file);
+ load_debug_section_with_follow (abbrev_sec, file);
if (debug_displays [abbrev_sec].section.start == NULL)
{
warn (_("Unable to locate %s section!\n"),
- debug_displays [abbrev_sec].section.name);
+ debug_displays [abbrev_sec].section.uncompressed_name);
return 0;
}
+ if (!do_loc && dwarf_start_die == 0)
+ introduce (section, FALSE);
+
for (section_begin = start, unit = 0; start < end; unit++)
{
DWARF2_Internal_CompUnit compunit;
/* If this is a DWARF package file, load the CU and TU indexes. */
(void) load_cu_tu_indexes (file);
- if (load_debug_section (info, file)
+ if (load_debug_section_with_follow (info, file)
&& process_debug_info (&debug_displays [info].section, file, abbrev, 1, 0))
return num_debug_info_entries;
- if (load_debug_section (info_dwo, file)
+ if (load_debug_section_with_follow (info_dwo, file)
&& process_debug_info (&debug_displays [info_dwo].section, file,
abbrev_dwo, 1, 0))
return num_debug_info_entries;
}
static unsigned char *
-display_formatted_table (unsigned char *data,
- unsigned char *start, unsigned char *end,
- const DWARF2_Internal_LineInfo *linfo,
- struct dwarf_section *section, const char *what)
+display_formatted_table (unsigned char * data,
+ unsigned char * start,
+ unsigned char * end,
+ const DWARF2_Internal_LineInfo * linfo,
+ struct dwarf_section * section,
+ const char * what)
{
unsigned char *format_start, format_count, *format, formati;
dwarf_vma data_count, datai;
}
static int
-display_debug_lines_raw (struct dwarf_section *section,
- unsigned char *data,
- unsigned char *end, void *file)
+display_debug_lines_raw (struct dwarf_section * section,
+ unsigned char * data,
+ unsigned char * end,
+ void * file)
{
unsigned char *start = section->start;
int verbose_view = 0;
- printf (_("Raw dump of debug contents of section %s:\n\n"),
- section->name);
+ introduce (section, TRUE);
while (data < end)
{
if (linfo.li_version >= 5)
{
- load_debug_section (line_str, file);
+ load_debug_section_with_follow (line_str, file);
data = display_formatted_table (data, start, end, &linfo, section,
_("Directory"));
/* Output a decoded representation of the .debug_line section. */
static int
-display_debug_lines_decoded (struct dwarf_section *section,
- unsigned char *data,
- unsigned char *end, void *fileptr)
+display_debug_lines_decoded (struct dwarf_section * section,
+ unsigned char * data,
+ unsigned char * end,
+ void * fileptr)
{
static DWARF2_Internal_LineInfo saved_linfo;
- printf (_("Decoded dump of debug contents of section %s:\n\n"),
- section->name);
+ introduce (section, FALSE);
while (data < end)
{
dwarf_vma formati, entryi;
unsigned int bytes_read;
- load_debug_section (line_str, fileptr);
+ load_debug_section_with_follow (line_str, fileptr);
/* Skip directories format. */
SAFE_BYTE_GET_AND_INC (format_count, data, 1, end);
we test for that later on. */
load_debug_info (file);
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
while (start < end)
{
unsigned int bytes_read;
enum dwarf_macinfo_record_type op;
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
while (curr < end)
{
unsigned char *extended_op_buf[256];
unsigned int bytes_read;
- load_debug_section (str, file);
- load_debug_section (line, file);
+ load_debug_section_with_follow (str, file);
+ load_debug_section_with_follow (line, file);
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
while (curr < end)
{
unsigned char *start = section->start;
unsigned char *end = start + section->size;
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
do
{
if (!locs_sorted)
array = (unsigned int *) xcmalloc (num_loc_list, sizeof (unsigned int));
- printf (_("Contents of the %s section:\n\n"), section->name);
+
+ introduce (section, FALSE);
+
if (reloc_at (section, 0))
printf (_(" Warning: This section has relocations - addresses seen here may not be accurate.\n\n"));
+
printf (_(" Offset Begin End Expression\n"));
seen_first_offset = 0;
return 0;
}
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
while (bytes)
{
unsigned char *start = section->start;
unsigned char *end = start + section->size;
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
/* It does not matter if this load fails,
we test for that later on. */
return 0;
}
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
/* PR 17531: file: cf38d01b.
We use xcalloc because a corrupt file may not have initialised all of the
warn (_("Range lists in %s section start at 0x%lx\n"),
section->name, (unsigned long) range_entries[0].ranges_offset);
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
+
printf (_(" Offset Begin End\n"));
for (i = 0; i < num_range_list; i++)
const char *bad_reg = _("bad register: ");
unsigned int saved_eh_addr_size = eh_addr_size;
- printf (_("Contents of the %s section:\n"), section->name);
+ introduce (section, FALSE);
while (start < end)
{
const unsigned char *const section_end = section->start + section->size;
unsigned char *unit_end;
- printf (_("Contents of the %s section:\n"), section->name);
+ introduce (section, FALSE);
- load_debug_section (str, file);
+ load_debug_section_with_follow (str, file);
for (; hdrptr < section_end; hdrptr = unit_end)
{
}
static int
+display_debug_links (struct dwarf_section * section,
+ void * file ATTRIBUTE_UNUSED)
+{
+ const unsigned char * filename;
+ unsigned int filelen;
+
+ introduce (section, FALSE);
+
+ /* The .gnu_debuglink section is formatted as:
+ (c-string) Filename.
+ (padding) If needed to reach a 4 byte boundary.
+ (uint32_t) CRC32 value.
+
+ The .gun_debugaltlink section is formatted as:
+ (c-string) Filename.
+ (binary) Build-ID. */
+
+ filename = section->start;
+ filelen = strnlen ((const char *) filename, section->size);
+ if (filelen == section->size)
+ {
+ warn (_("The debuglink filename is corrupt/missing\n"));
+ return 0;
+ }
+
+ printf (_(" Separate debug info file: %s\n"), filename);
+
+ if (const_strneq (section->name, ".gnu_debuglink"))
+ {
+ unsigned int crc32;
+ unsigned int crc_offset;
+
+ crc_offset = filelen + 1;
+ crc_offset = (crc_offset + 3) & ~3;
+ if (crc_offset + 4 > section->size)
+ {
+ warn (_("CRC offset missing/truncated\n"));
+ return 0;
+ }
+
+ crc32 = byte_get (filename + crc_offset, 4);
+
+ printf (_(" CRC value: %#x\n"), crc32);
+
+ if (crc_offset + 4 < section->size)
+ {
+ warn (_("There are %#lx extraneous bytes at the end of the section\n"),
+ (long)(section->size - (crc_offset + 4)));
+ return 0;
+ }
+ }
+ else /* const_strneq (section->name, ".gnu_debugaltlink") */
+ {
+ const unsigned char * build_id = section->start + filelen + 1;
+ bfd_size_type build_id_len = section->size - (filelen + 1);
+ bfd_size_type printed;
+
+ /* FIXME: Should we support smaller build-id notes ? */
+ if (build_id_len < 0x14)
+ {
+ warn (_("Build-ID is too short (%#lx bytes)\n"), (long) build_id_len);
+ return 0;
+ }
+
+ printed = printf (_(" Build-ID (%#lx bytes):"), (long) build_id_len);
+
+ if (do_wide || build_id_len < ((80 - printed) / 3))
+ for (printed = 0; printed < build_id_len; ++printed)
+ printf (" %02x", build_id[printed]);
+ else
+ {
+ for (printed = 0; printed < build_id_len; ++printed)
+ {
+ if (printed % (80 / 3) == 0)
+ putchar ('\n');
+ printf (" %02x", build_id[printed]);
+ }
+ }
+ putchar ('\n');
+ }
+
+ putchar ('\n');
+ return 1;
+}
+
+static int
display_gdb_index (struct dwarf_section *section,
void *file ATTRIBUTE_UNUSED)
{
/* The documentation for the format of this file is in gdb/dwarf2read.c. */
- printf (_("Contents of the %s section:\n"), section->name);
+ introduce (section, FALSE);
if (section->size < 6 * sizeof (uint32_t))
{
if (do_display)
{
- printf (_("Contents of the %s section:\n\n"), section->name);
+ introduce (section, FALSE);
+
printf (_(" Version: %d\n"), version);
if (version >= 2)
printf (_(" Number of columns: %d\n"), ncols);
{
cu_tu_indexes_read = TRUE;
- if (load_debug_section (dwp_cu_index, file))
+ if (load_debug_section_with_follow (dwp_cu_index, file))
if (! process_cu_tu_index (&debug_displays [dwp_cu_index].section, 0))
cu_tu_indexes_read = FALSE;
- if (load_debug_section (dwp_tu_index, file))
+ if (load_debug_section_with_follow (dwp_tu_index, file))
if (! process_cu_tu_index (&debug_displays [dwp_tu_index].section, 0))
cu_tu_indexes_read = FALSE;
}
/* Like malloc, but takes two parameters like calloc.
Verifies that the first parameter is not too large.
Note: does *not* initialise the allocated memory to zero. */
+
void *
cmalloc (size_t nmemb, size_t size)
{
/* Like xmalloc, but takes two parameters like calloc.
Verifies that the first parameter is not too large.
Note: does *not* initialise the allocated memory to zero. */
+
void *
xcmalloc (size_t nmemb, size_t size)
{
/* Like xrealloc, but takes three parameters.
Verifies that the second parameter is not too large.
Note: does *not* initialise any new memory to zero. */
+
void *
xcrealloc (void *ptr, size_t nmemb, size_t size)
{
/* Check for overflow. */
if (nmemb >= ~(size_t) 0 / size)
{
- fprintf (stderr,
- _("Attempt to re-allocate an array with an excessive number of elements: 0x%lx\n"),
- (long) nmemb);
+ error (_("Attempt to re-allocate an array with an excessive number of elements: 0x%lx\n"),
+ (long) nmemb);
xexit (1);
}
}
/* Like xcalloc, but verifies that the first parameter is not too large. */
+
void *
xcalloc2 (size_t nmemb, size_t size)
{
/* Check for overflow. */
if (nmemb >= ~(size_t) 0 / size)
{
- fprintf (stderr,
- _("Attempt to allocate a zero'ed array with an excessive number of elements: 0x%lx\n"),
- (long) nmemb);
+ error (_("Attempt to allocate a zero'ed array with an excessive number of elements: 0x%lx\n"),
+ (long) nmemb);
xexit (1);
}
return xcalloc (nmemb, size);
}
+static unsigned long
+calc_gnu_debuglink_crc32 (unsigned long crc,
+ const unsigned char * buf,
+ bfd_size_type len)
+{
+ static const unsigned long crc32_table[256] =
+ {
+ 0x00000000, 0x77073096, 0xee0e612c, 0x990951ba, 0x076dc419,
+ 0x706af48f, 0xe963a535, 0x9e6495a3, 0x0edb8832, 0x79dcb8a4,
+ 0xe0d5e91e, 0x97d2d988, 0x09b64c2b, 0x7eb17cbd, 0xe7b82d07,
+ 0x90bf1d91, 0x1db71064, 0x6ab020f2, 0xf3b97148, 0x84be41de,
+ 0x1adad47d, 0x6ddde4eb, 0xf4d4b551, 0x83d385c7, 0x136c9856,
+ 0x646ba8c0, 0xfd62f97a, 0x8a65c9ec, 0x14015c4f, 0x63066cd9,
+ 0xfa0f3d63, 0x8d080df5, 0x3b6e20c8, 0x4c69105e, 0xd56041e4,
+ 0xa2677172, 0x3c03e4d1, 0x4b04d447, 0xd20d85fd, 0xa50ab56b,
+ 0x35b5a8fa, 0x42b2986c, 0xdbbbc9d6, 0xacbcf940, 0x32d86ce3,
+ 0x45df5c75, 0xdcd60dcf, 0xabd13d59, 0x26d930ac, 0x51de003a,
+ 0xc8d75180, 0xbfd06116, 0x21b4f4b5, 0x56b3c423, 0xcfba9599,
+ 0xb8bda50f, 0x2802b89e, 0x5f058808, 0xc60cd9b2, 0xb10be924,
+ 0x2f6f7c87, 0x58684c11, 0xc1611dab, 0xb6662d3d, 0x76dc4190,
+ 0x01db7106, 0x98d220bc, 0xefd5102a, 0x71b18589, 0x06b6b51f,
+ 0x9fbfe4a5, 0xe8b8d433, 0x7807c9a2, 0x0f00f934, 0x9609a88e,
+ 0xe10e9818, 0x7f6a0dbb, 0x086d3d2d, 0x91646c97, 0xe6635c01,
+ 0x6b6b51f4, 0x1c6c6162, 0x856530d8, 0xf262004e, 0x6c0695ed,
+ 0x1b01a57b, 0x8208f4c1, 0xf50fc457, 0x65b0d9c6, 0x12b7e950,
+ 0x8bbeb8ea, 0xfcb9887c, 0x62dd1ddf, 0x15da2d49, 0x8cd37cf3,
+ 0xfbd44c65, 0x4db26158, 0x3ab551ce, 0xa3bc0074, 0xd4bb30e2,
+ 0x4adfa541, 0x3dd895d7, 0xa4d1c46d, 0xd3d6f4fb, 0x4369e96a,
+ 0x346ed9fc, 0xad678846, 0xda60b8d0, 0x44042d73, 0x33031de5,
+ 0xaa0a4c5f, 0xdd0d7cc9, 0x5005713c, 0x270241aa, 0xbe0b1010,
+ 0xc90c2086, 0x5768b525, 0x206f85b3, 0xb966d409, 0xce61e49f,
+ 0x5edef90e, 0x29d9c998, 0xb0d09822, 0xc7d7a8b4, 0x59b33d17,
+ 0x2eb40d81, 0xb7bd5c3b, 0xc0ba6cad, 0xedb88320, 0x9abfb3b6,
+ 0x03b6e20c, 0x74b1d29a, 0xead54739, 0x9dd277af, 0x04db2615,
+ 0x73dc1683, 0xe3630b12, 0x94643b84, 0x0d6d6a3e, 0x7a6a5aa8,
+ 0xe40ecf0b, 0x9309ff9d, 0x0a00ae27, 0x7d079eb1, 0xf00f9344,
+ 0x8708a3d2, 0x1e01f268, 0x6906c2fe, 0xf762575d, 0x806567cb,
+ 0x196c3671, 0x6e6b06e7, 0xfed41b76, 0x89d32be0, 0x10da7a5a,
+ 0x67dd4acc, 0xf9b9df6f, 0x8ebeeff9, 0x17b7be43, 0x60b08ed5,
+ 0xd6d6a3e8, 0xa1d1937e, 0x38d8c2c4, 0x4fdff252, 0xd1bb67f1,
+ 0xa6bc5767, 0x3fb506dd, 0x48b2364b, 0xd80d2bda, 0xaf0a1b4c,
+ 0x36034af6, 0x41047a60, 0xdf60efc3, 0xa867df55, 0x316e8eef,
+ 0x4669be79, 0xcb61b38c, 0xbc66831a, 0x256fd2a0, 0x5268e236,
+ 0xcc0c7795, 0xbb0b4703, 0x220216b9, 0x5505262f, 0xc5ba3bbe,
+ 0xb2bd0b28, 0x2bb45a92, 0x5cb36a04, 0xc2d7ffa7, 0xb5d0cf31,
+ 0x2cd99e8b, 0x5bdeae1d, 0x9b64c2b0, 0xec63f226, 0x756aa39c,
+ 0x026d930a, 0x9c0906a9, 0xeb0e363f, 0x72076785, 0x05005713,
+ 0x95bf4a82, 0xe2b87a14, 0x7bb12bae, 0x0cb61b38, 0x92d28e9b,
+ 0xe5d5be0d, 0x7cdcefb7, 0x0bdbdf21, 0x86d3d2d4, 0xf1d4e242,
+ 0x68ddb3f8, 0x1fda836e, 0x81be16cd, 0xf6b9265b, 0x6fb077e1,
+ 0x18b74777, 0x88085ae6, 0xff0f6a70, 0x66063bca, 0x11010b5c,
+ 0x8f659eff, 0xf862ae69, 0x616bffd3, 0x166ccf45, 0xa00ae278,
+ 0xd70dd2ee, 0x4e048354, 0x3903b3c2, 0xa7672661, 0xd06016f7,
+ 0x4969474d, 0x3e6e77db, 0xaed16a4a, 0xd9d65adc, 0x40df0b66,
+ 0x37d83bf0, 0xa9bcae53, 0xdebb9ec5, 0x47b2cf7f, 0x30b5ffe9,
+ 0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 0xbad03605,
+ 0xcdd70693, 0x54de5729, 0x23d967bf, 0xb3667a2e, 0xc4614ab8,
+ 0x5d681b02, 0x2a6f2b94, 0xb40bbe37, 0xc30c8ea1, 0x5a05df1b,
+ 0x2d02ef8d
+ };
+ const unsigned char *end;
+
+ crc = ~crc & 0xffffffff;
+ for (end = buf + len; buf < end; ++ buf)
+ crc = crc32_table[(crc ^ *buf) & 0xff] ^ (crc >> 8);
+ return ~crc & 0xffffffff;
+}
+
+typedef bfd_boolean (* check_func_type) (const char *, void *);
+typedef const char * (* parse_func_type) (struct dwarf_section *, void *);
+
+static bfd_boolean
+check_gnu_debuglink (const char * pathname, void * crc_pointer)
+{
+ static unsigned char buffer [8 * 1024];
+ FILE * f;
+ bfd_size_type count;
+ unsigned long crc = 0;
+ void * sep_data;
+
+ sep_data = open_debug_file (pathname);
+ if (sep_data == NULL)
+ return FALSE;
+
+ /* Yes - we are opening the file twice... */
+ f = fopen (pathname, "rb");
+ if (f == NULL)
+ {
+ /* Paranoia: This should never happen. */
+ close_debug_file (sep_data);
+ warn (_("Unable to reopen separate debug info file: %s\n"), pathname);
+ return FALSE;
+ }
+
+ while ((count = fread (buffer, 1, sizeof (buffer), f)) > 0)
+ crc = calc_gnu_debuglink_crc32 (crc, buffer, count);
+
+ fclose (f);
+
+ if (crc != * (unsigned long *) crc_pointer)
+ {
+ close_debug_file (sep_data);
+ warn (_("Separate debug info file %s found, but CRC does not match - ignoring\n"),
+ pathname);
+ return FALSE;
+ }
+
+ return TRUE;
+}
+
+static const char *
+parse_gnu_debuglink (struct dwarf_section * section, void * data)
+{
+ const char * name;
+ unsigned int crc_offset;
+ unsigned long * crc32 = (unsigned long *) data;
+
+ /* The name is first.
+ The CRC value is stored after the filename, aligned up to 4 bytes. */
+ name = (const char *) section->start;
+
+ crc_offset = strnlen (name, section->size) + 1;
+ crc_offset = (crc_offset + 3) & ~3;
+ if (crc_offset + 4 > section->size)
+ return NULL;
+
+ * crc32 = byte_get (section->start + crc_offset, 4);
+ return name;
+}
+
+static bfd_boolean
+check_gnu_debugaltlink (const char * filename, void * data ATTRIBUTE_UNUSED)
+{
+ void * sep_data = open_debug_file (filename);
+
+ if (sep_data == NULL)
+ return FALSE;
+
+ /* FIXME: We should now extract the build-id in the separate file
+ and check it... */
+
+ return TRUE;
+}
+
+typedef struct build_id_data
+{
+ bfd_size_type len;
+ const unsigned char * data;
+} Build_id_data;
+
+static const char *
+parse_gnu_debugaltlink (struct dwarf_section * section, void * data)
+{
+ const char * name;
+ bfd_size_type namelen;
+ bfd_size_type id_len;
+ Build_id_data * build_id_data;
+
+ /* The name is first.
+ The build-id follows immediately, with no padding, up to the section's end. */
+
+ name = (const char *) section->start;
+ namelen = strnlen (name, section->size) + 1;
+ if (namelen >= section->size)
+ return NULL;
+
+ id_len = section->size - namelen;
+ if (id_len < 0x14)
+ return NULL;
+
+ build_id_data = calloc (1, sizeof * build_id_data);
+ if (build_id_data == NULL)
+ return NULL;
+
+ build_id_data->len = id_len;
+ build_id_data->data = section->start + namelen;
+
+ * (Build_id_data **) data = build_id_data;
+
+ return name;
+}
+
+static void *
+load_separate_debug_info (const char * main_filename,
+ struct dwarf_section * link,
+ parse_func_type parse_func,
+ check_func_type check_func,
+ void * func_data)
+{
+ const char * separate_filename;
+ char * debugfile;
+ char * canon_dir;
+ size_t canon_dirlen;
+ size_t dirlen;
+
+ if ((separate_filename = parse_func (link, func_data)) == NULL)
+ {
+ warn (_("Corrupt debuglink section: %s\n"),
+ link->name ? link->name : link->uncompressed_name);
+ return FALSE;
+ }
+
+ /* Attempt to locate the separate file.
+ This should duplicate the logic in bfd/opncls.c:find_separate_debug_file(). */
+
+ canon_dir = lrealpath (main_filename);
+
+ for (canon_dirlen = strlen (canon_dir); canon_dirlen > 0; canon_dirlen--)
+ if (IS_DIR_SEPARATOR (canon_dir[canon_dirlen - 1]))
+ break;
+ canon_dir[canon_dirlen] = '\0';
+
+#ifndef DEBUGDIR
+#define DEBUGDIR "/lib/debug"
+#endif
+#ifndef EXTRA_DEBUG_ROOT1
+#define EXTRA_DEBUG_ROOT1 "/usr/lib/debug"
+#endif
+#ifndef EXTRA_DEBUG_ROOT2
+#define EXTRA_DEBUG_ROOT2 "/usr/lib/debug/usr"
+#endif
+
+ debugfile = (char *) malloc (strlen (DEBUGDIR) + 1
+ + canon_dirlen
+ + strlen (".debug/")
+#ifdef EXTRA_DEBUG_ROOT1
+ + strlen (EXTRA_DEBUG_ROOT1)
+#endif
+#ifdef EXTRA_DEBUG_ROOT2
+ + strlen (EXTRA_DEBUG_ROOT2)
+#endif
+ + strlen (separate_filename)
+ + 1);
+ if (debugfile == NULL)
+ {
+ warn (_("Out of memory"));
+ return NULL;
+ }
+
+ /* First try in the current directory. */
+ sprintf (debugfile, "%s", separate_filename);
+ if (check_func (debugfile, func_data))
+ goto found;
+
+ /* Then try in a subdirectory called .debug. */
+ sprintf (debugfile, ".debug/%s", separate_filename);
+ if (check_func (debugfile, func_data))
+ goto found;
+
+ /* Then try in the same directory as the original file. */
+ sprintf (debugfile, "%s%s", canon_dir, separate_filename);
+ if (check_func (debugfile, func_data))
+ goto found;
+
+ /* And the .debug subdirectory of that directory. */
+ sprintf (debugfile, "%s.debug/%s", canon_dir, separate_filename);
+ if (check_func (debugfile, func_data))
+ goto found;
+
+#ifdef EXTRA_DEBUG_ROOT1
+ /* Try the first extra debug file root. */
+ sprintf (debugfile, "%s/%s", EXTRA_DEBUG_ROOT1, separate_filename);
+ if (check_func (debugfile, func_data))
+ goto found;
+#endif
+
+#ifdef EXTRA_DEBUG_ROOT2
+ /* Try the second extra debug file root. */
+ sprintf (debugfile, "%s/%s", EXTRA_DEBUG_ROOT2, separate_filename);
+ if (check_func (debugfile, func_data))
+ goto found;
+#endif
+
+ /* Then try in the global debugfile directory. */
+ strcpy (debugfile, DEBUGDIR);
+ dirlen = strlen (DEBUGDIR) - 1;
+ if (dirlen > 0 && DEBUGDIR[dirlen] != '/')
+ strcat (debugfile, "/");
+ strcat (debugfile, (const char *) separate_filename);
+
+ if (check_func (debugfile, func_data))
+ goto found;
+
+ /* Failed to find the file. */
+ warn (_("could not find separate debug file '%s'\n"), separate_filename);
+ warn (_("tried: %s\n"), debugfile);
+
+#ifdef EXTRA_DEBUG_ROOT2
+ sprintf (debugfile, "%s/%s", EXTRA_DEBUG_ROOT2, separate_filename);
+ warn (_("tried: %s\n"), debugfile);
+#endif
+
+#ifdef EXTRA_DEBUG_ROOT1
+ sprintf (debugfile, "%s/%s", EXTRA_DEBUG_ROOT1, separate_filename);
+ warn (_("tried: %s\n"), debugfile);
+#endif
+
+ sprintf (debugfile, "%s.debug/%s", canon_dir, separate_filename);
+ warn (_("tried: %s\n"), debugfile);
+
+ sprintf (debugfile, "%s%s", canon_dir, separate_filename);
+ warn (_("tried: %s\n"), debugfile);
+
+ sprintf (debugfile, ".debug/%s", separate_filename);
+ warn (_("tried: %s\n"), debugfile);
+
+ sprintf (debugfile, "%s", separate_filename);
+ warn (_("tried: %s\n"), debugfile);
+
+ free (canon_dir);
+ free (debugfile);
+ return NULL;
+
+ found:
+ free (canon_dir);
+
+ /* Now open the file.... */
+ if ((separate_debug_file = open_debug_file (debugfile)) == NULL)
+ {
+ warn (_("failed to open separate debug file: %s\n"), debugfile);
+ free (debugfile);
+ return FALSE;
+ }
+
+ /* FIXME: We do not check to see if there are any other separate debug info
+ files that would also match. */
+
+ printf (_("%s: Found separate debug info file: %s\n\n"), main_filename, debugfile);
+ separate_debug_filename = debugfile;
+
+ /* Do not free debugfile - it might be referenced inside
+ the structure returned by open_debug_file(). */
+ return separate_debug_file;
+}
+
+/* Load a separate debug info file, if it exists.
+ Returns the data pointer that is the result of calling open_debug_file
+ on the separate debug info file. */
+
+void *
+load_separate_debug_file (void * file, const char * filename)
+{
+ if (! do_follow_links)
+ return NULL;
+
+ /* FIXME: We do not check for the presence of both link sections in the same file. */
+ /* FIXME: We do not check the separate debug info file to see if it too contains debuglinks. */
+ /* FIXME: We do not check for the presence of multiple, same-name debuglink sections. */
+
+ /* We try the alt version first as that is blessed by the DWARF5 standard. */
+ if (load_debug_section (gnu_debugaltlink, file))
+ {
+ Build_id_data * build_id_data;
+
+ return load_separate_debug_info (filename,
+ & debug_displays[gnu_debugaltlink].section,
+ parse_gnu_debugaltlink,
+ check_gnu_debugaltlink,
+ & build_id_data);
+ }
+
+ if (load_debug_section (gnu_debuglink, file))
+ {
+ unsigned long crc32;
+
+ return load_separate_debug_info (filename,
+ & debug_displays[gnu_debuglink].section,
+ parse_gnu_debuglink,
+ check_gnu_debuglink,
+ & crc32);
+ }
+
+ do_follow_links = 0;
+ return NULL;
+}
+
void
free_debug_memory (void)
{
debug_information = NULL;
alloc_num_debug_info_entries = num_debug_info_entries = 0;
}
+
+ if (separate_debug_file != NULL)
+ {
+ close_debug_file (separate_debug_file);
+ separate_debug_file = NULL;
+
+ free ((void *) separate_debug_filename);
+ separate_debug_filename = NULL;
+ }
}
void
{ "aranges", & do_debug_aranges, 1 },
{ "cu_index", & do_debug_cu_index, 1 },
{ "decodedline", & do_debug_lines, FLAG_DEBUG_LINES_DECODED },
+ { "follow-links", & do_follow_links, 1 },
{ "frames", & do_debug_frames, 1 },
{ "frames-interp", & do_debug_frames_interp, 1 },
/* The special .gdb_index section. */
{ "gdb_index", & do_gdb_index, 1 },
{ "info", & do_debug_info, 1 },
{ "line", & do_debug_lines, FLAG_DEBUG_LINES_RAW }, /* For backwards compatibility. */
+ { "links", & do_debug_links, 1 },
{ "loc", & do_debug_loc, 1 },
{ "macro", & do_debug_macinfo, 1 },
{ "pubnames", & do_debug_pubnames, 1 },
while (letters[lindex])
switch (letters[lindex++])
{
- case 'i':
- do_debug_info = 1;
- break;
-
- case 'a':
- do_debug_abbrevs = 1;
- break;
-
- case 'l':
- do_debug_lines |= FLAG_DEBUG_LINES_RAW;
- break;
-
- case 'L':
- do_debug_lines |= FLAG_DEBUG_LINES_DECODED;
- break;
-
- case 'p':
- do_debug_pubnames = 1;
- break;
-
- case 't':
- do_debug_pubtypes = 1;
- break;
-
- case 'r':
- do_debug_aranges = 1;
- break;
-
- case 'R':
- do_debug_ranges = 1;
- break;
-
- case 'F':
- do_debug_frames_interp = 1;
- /* Fall through. */
- case 'f':
- do_debug_frames = 1;
- break;
-
- case 'm':
- do_debug_macinfo = 1;
- break;
-
- case 's':
- do_debug_str = 1;
- break;
-
- case 'o':
- do_debug_loc = 1;
- break;
+ case 'A': do_debug_addr = 1; break;
+ case 'a': do_debug_abbrevs = 1; break;
+ case 'c': do_debug_cu_index = 1; break;
+ case 'F': do_debug_frames_interp = 1; /* Fall through. */
+ case 'f': do_debug_frames = 1; break;
+ case 'g': do_gdb_index = 1; break;
+ case 'i': do_debug_info = 1; break;
+ case 'K': do_follow_links = 1; break;
+ case 'k': do_debug_links = 1; break;
+ case 'l': do_debug_lines |= FLAG_DEBUG_LINES_RAW; break;
+ case 'L': do_debug_lines |= FLAG_DEBUG_LINES_DECODED; break;
+ case 'm': do_debug_macinfo = 1; break;
+ case 'o': do_debug_loc = 1; break;
+ case 'p': do_debug_pubnames = 1; break;
+ case 'R': do_debug_ranges = 1; break;
+ case 'r': do_debug_aranges = 1; break;
+ case 's': do_debug_str = 1; break;
+ case 'T': do_trace_aranges = 1; break;
+ case 't': do_debug_pubtypes = 1; break;
+ case 'U': do_trace_info = 1; break;
+ case 'u': do_trace_abbrevs = 1; break;
default:
warn (_("Unrecognized debug option '%s'\n"), letters);
do_trace_aranges = 1;
do_debug_addr = 1;
do_debug_cu_index = 1;
+ do_follow_links = 1;
+ do_debug_links = 1;
}
+#define NO_ABBREVS NULL, NULL, NULL, 0, 0, 0, NULL, 0, NULL
+#define ABBREV(N) NULL, NULL, NULL, 0, 0, N, NULL, 0, NULL
+
+/* N.B. The order here must match the order in section_display_enum. */
+
struct dwarf_section_display debug_displays[] =
{
- { { ".debug_abbrev", ".zdebug_abbrev", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_abbrev, &do_debug_abbrevs, FALSE },
- { { ".debug_aranges", ".zdebug_aranges", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_aranges, &do_debug_aranges, TRUE },
- { { ".debug_frame", ".zdebug_frame", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_frames, &do_debug_frames, TRUE },
- { { ".debug_info", ".zdebug_info", NULL, NULL, 0, 0, abbrev, NULL, 0, NULL },
- display_debug_info, &do_debug_info, TRUE },
- { { ".debug_line", ".zdebug_line", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_lines, &do_debug_lines, TRUE },
- { { ".debug_pubnames", ".zdebug_pubnames", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_pubnames, &do_debug_pubnames, FALSE },
- { { ".debug_gnu_pubnames", ".zdebug_gnu_pubnames", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_gnu_pubnames, &do_debug_pubnames, FALSE },
- { { ".eh_frame", "", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_frames, &do_debug_frames, TRUE },
- { { ".debug_macinfo", ".zdebug_macinfo", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_macinfo, &do_debug_macinfo, FALSE },
- { { ".debug_macro", ".zdebug_macro", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_macro, &do_debug_macinfo, TRUE },
- { { ".debug_str", ".zdebug_str", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_str, &do_debug_str, FALSE },
- { { ".debug_line_str", ".zdebug_line_str", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_str, &do_debug_str, FALSE },
- { { ".debug_loc", ".zdebug_loc", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_loc, &do_debug_loc, TRUE },
- { { ".debug_loclists", ".zdebug_loclists", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_loc, &do_debug_loc, TRUE },
- { { ".debug_pubtypes", ".zdebug_pubtypes", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_pubnames, &do_debug_pubtypes, FALSE },
- { { ".debug_gnu_pubtypes", ".zdebug_gnu_pubtypes", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_gnu_pubnames, &do_debug_pubtypes, FALSE },
- { { ".debug_ranges", ".zdebug_ranges", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_ranges, &do_debug_ranges, TRUE },
- { { ".debug_rnglists", ".zdebug_rnglists", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_ranges, &do_debug_ranges, TRUE },
- { { ".debug_static_func", ".zdebug_static_func", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_not_supported, NULL, FALSE },
- { { ".debug_static_vars", ".zdebug_static_vars", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_not_supported, NULL, FALSE },
- { { ".debug_types", ".zdebug_types", NULL, NULL, 0, 0, abbrev, NULL, 0, NULL },
- display_debug_types, &do_debug_info, TRUE },
- { { ".debug_weaknames", ".zdebug_weaknames", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_not_supported, NULL, FALSE },
- { { ".gdb_index", "", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_gdb_index, &do_gdb_index, FALSE },
- { { ".debug_names", "", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_names, &do_gdb_index, FALSE },
- { { ".trace_info", "", NULL, NULL, 0, 0, trace_abbrev, NULL, 0, NULL },
- display_trace_info, &do_trace_info, TRUE },
- { { ".trace_abbrev", "", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_abbrev, &do_trace_abbrevs, FALSE },
- { { ".trace_aranges", "", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_aranges, &do_trace_aranges, FALSE },
- { { ".debug_info.dwo", ".zdebug_info.dwo", NULL, NULL, 0, 0, abbrev_dwo, NULL, 0, NULL },
- display_debug_info, &do_debug_info, TRUE },
- { { ".debug_abbrev.dwo", ".zdebug_abbrev.dwo", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_abbrev, &do_debug_abbrevs, FALSE },
- { { ".debug_types.dwo", ".zdebug_types.dwo", NULL, NULL, 0, 0, abbrev_dwo, NULL, 0, NULL },
- display_debug_types, &do_debug_info, TRUE },
- { { ".debug_line.dwo", ".zdebug_line.dwo", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_lines, &do_debug_lines, TRUE },
- { { ".debug_loc.dwo", ".zdebug_loc.dwo", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_loc, &do_debug_loc, TRUE },
- { { ".debug_macro.dwo", ".zdebug_macro.dwo", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_macro, &do_debug_macinfo, TRUE },
- { { ".debug_macinfo.dwo", ".zdebug_macinfo.dwo", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_macinfo, &do_debug_macinfo, FALSE },
- { { ".debug_str.dwo", ".zdebug_str.dwo", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_str, &do_debug_str, TRUE },
- { { ".debug_str_offsets", ".zdebug_str_offsets", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_str_offsets, NULL, FALSE },
- { { ".debug_str_offsets.dwo", ".zdebug_str_offsets.dwo", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_str_offsets, NULL, FALSE },
- { { ".debug_addr", ".zdebug_addr", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_debug_addr, &do_debug_addr, TRUE },
- { { ".debug_cu_index", "", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_cu_index, &do_debug_cu_index, FALSE },
- { { ".debug_tu_index", "", NULL, NULL, 0, 0, 0, NULL, 0, NULL },
- display_cu_index, &do_debug_cu_index, FALSE },
+ { { ".debug_abbrev", ".zdebug_abbrev", NO_ABBREVS }, display_debug_abbrev, &do_debug_abbrevs, FALSE },
+ { { ".debug_aranges", ".zdebug_aranges", NO_ABBREVS }, display_debug_aranges, &do_debug_aranges, TRUE },
+ { { ".debug_frame", ".zdebug_frame", NO_ABBREVS }, display_debug_frames, &do_debug_frames, TRUE },
+ { { ".debug_info", ".zdebug_info", ABBREV (abbrev)}, display_debug_info, &do_debug_info, TRUE },
+ { { ".debug_line", ".zdebug_line", NO_ABBREVS }, display_debug_lines, &do_debug_lines, TRUE },
+ { { ".debug_pubnames", ".zdebug_pubnames", NO_ABBREVS }, display_debug_pubnames, &do_debug_pubnames, FALSE },
+ { { ".debug_gnu_pubnames", ".zdebug_gnu_pubnames", NO_ABBREVS }, display_debug_gnu_pubnames, &do_debug_pubnames, FALSE },
+ { { ".eh_frame", "", NO_ABBREVS }, display_debug_frames, &do_debug_frames, TRUE },
+ { { ".debug_macinfo", ".zdebug_macinfo", NO_ABBREVS }, display_debug_macinfo, &do_debug_macinfo, FALSE },
+ { { ".debug_macro", ".zdebug_macro", NO_ABBREVS }, display_debug_macro, &do_debug_macinfo, TRUE },
+ { { ".debug_str", ".zdebug_str", NO_ABBREVS }, display_debug_str, &do_debug_str, FALSE },
+ { { ".debug_line_str", ".zdebug_line_str", NO_ABBREVS }, display_debug_str, &do_debug_str, FALSE },
+ { { ".debug_loc", ".zdebug_loc", NO_ABBREVS }, display_debug_loc, &do_debug_loc, TRUE },
+ { { ".debug_loclists", ".zdebug_loclists", NO_ABBREVS }, display_debug_loc, &do_debug_loc, TRUE },
+ { { ".debug_pubtypes", ".zdebug_pubtypes", NO_ABBREVS }, display_debug_pubnames, &do_debug_pubtypes, FALSE },
+ { { ".debug_gnu_pubtypes", ".zdebug_gnu_pubtypes", NO_ABBREVS }, display_debug_gnu_pubnames, &do_debug_pubtypes, FALSE },
+ { { ".debug_ranges", ".zdebug_ranges", NO_ABBREVS }, display_debug_ranges, &do_debug_ranges, TRUE },
+ { { ".debug_rnglists", ".zdebug_rnglists", NO_ABBREVS }, display_debug_ranges, &do_debug_ranges, TRUE },
+ { { ".debug_static_func", ".zdebug_static_func", NO_ABBREVS }, display_debug_not_supported, NULL, FALSE },
+ { { ".debug_static_vars", ".zdebug_static_vars", NO_ABBREVS }, display_debug_not_supported, NULL, FALSE },
+ { { ".debug_types", ".zdebug_types", ABBREV (abbrev) }, display_debug_types, &do_debug_info, TRUE },
+ { { ".debug_weaknames", ".zdebug_weaknames", NO_ABBREVS }, display_debug_not_supported, NULL, FALSE },
+ { { ".gdb_index", "", NO_ABBREVS }, display_gdb_index, &do_gdb_index, FALSE },
+ { { ".debug_names", "", NO_ABBREVS }, display_debug_names, &do_gdb_index, FALSE },
+ { { ".trace_info", "", ABBREV (trace_abbrev) }, display_trace_info, &do_trace_info, TRUE },
+ { { ".trace_abbrev", "", NO_ABBREVS }, display_debug_abbrev, &do_trace_abbrevs, FALSE },
+ { { ".trace_aranges", "", NO_ABBREVS }, display_debug_aranges, &do_trace_aranges, FALSE },
+ { { ".debug_info.dwo", ".zdebug_info.dwo", ABBREV (abbrev_dwo) }, display_debug_info, &do_debug_info, TRUE },
+ { { ".debug_abbrev.dwo", ".zdebug_abbrev.dwo", NO_ABBREVS }, display_debug_abbrev, &do_debug_abbrevs, FALSE },
+ { { ".debug_types.dwo", ".zdebug_types.dwo", ABBREV (abbrev_dwo) }, display_debug_types, &do_debug_info, TRUE },
+ { { ".debug_line.dwo", ".zdebug_line.dwo", NO_ABBREVS }, display_debug_lines, &do_debug_lines, TRUE },
+ { { ".debug_loc.dwo", ".zdebug_loc.dwo", NO_ABBREVS }, display_debug_loc, &do_debug_loc, TRUE },
+ { { ".debug_macro.dwo", ".zdebug_macro.dwo", NO_ABBREVS }, display_debug_macro, &do_debug_macinfo, TRUE },
+ { { ".debug_macinfo.dwo", ".zdebug_macinfo.dwo", NO_ABBREVS }, display_debug_macinfo, &do_debug_macinfo, FALSE },
+ { { ".debug_str.dwo", ".zdebug_str.dwo", NO_ABBREVS }, display_debug_str, &do_debug_str, TRUE },
+ { { ".debug_str_offsets", ".zdebug_str_offsets", NO_ABBREVS }, display_debug_str_offsets, NULL, FALSE },
+ { { ".debug_str_offsets.dwo", ".zdebug_str_offsets.dwo", NO_ABBREVS }, display_debug_str_offsets, NULL, FALSE },
+ { { ".debug_addr", ".zdebug_addr", NO_ABBREVS }, display_debug_addr, &do_debug_addr, TRUE },
+ { { ".debug_cu_index", "", NO_ABBREVS }, display_cu_index, &do_debug_cu_index, FALSE },
+ { { ".debug_tu_index", "", NO_ABBREVS }, display_cu_index, &do_debug_cu_index, FALSE },
+ { { ".gnu_debuglink", "", NO_ABBREVS }, display_debug_links, &do_debug_links, FALSE },
+ { { ".gnu_debugaltlink", "", NO_ABBREVS }, display_debug_links, &do_debug_links, FALSE },
+ /* Separate debug info files can containt their own .debug_str section,
+ and this might be in *addition* to a .debug_str section already present
+ in the main file. Hence we need to have two entries for .debug_str. */
+ { { ".debug_str", ".zdebug_str", NO_ABBREVS }, display_debug_str, &do_debug_str, FALSE },
};
/* A static assertion. */
typedef struct elf_section_list
{
- Elf_Internal_Shdr * hdr;
- struct elf_section_list * next;
+ Elf_Internal_Shdr * hdr;
+ struct elf_section_list * next;
} elf_section_list;
+/* Flag bits indicating particular types of dump. */
+#define HEX_DUMP (1 << 0) /* The -x command line switch. */
+#define DISASS_DUMP (1 << 1) /* The -i command line switch. */
+#define DEBUG_DUMP (1 << 2) /* The -w command line switch. */
+#define STRING_DUMP (1 << 3) /* The -p command line switch. */
+#define RELOC_DUMP (1 << 4) /* The -R command line switch. */
+
+typedef unsigned char dump_type;
+
+/* A linked list of the section names for which dumps were requested. */
+struct dump_list_entry
+{
+ char * name;
+ dump_type type;
+ struct dump_list_entry * next;
+};
+
+typedef struct filedata
+{
+ const char * file_name;
+ FILE * handle;
+ bfd_size_type file_size;
+ Elf_Internal_Ehdr file_header;
+ Elf_Internal_Shdr * section_headers;
+ Elf_Internal_Phdr * program_headers;
+ char * string_table;
+ unsigned long string_table_length;
+ /* A dynamic array of flags indicating for which sections a dump of
+ some kind has been requested. It is reset on a per-object file
+ basis and then initialised from the cmdline_dump_sects array,
+ the results of interpreting the -w switch, and the
+ dump_sects_byname list. */
+ dump_type * dump_sects;
+ unsigned int num_dump_sects;
+} Filedata;
+
char * program_name = "readelf";
+
static unsigned long archive_file_offset;
static unsigned long archive_file_size;
-static bfd_size_type current_file_size;
static unsigned long dynamic_addr;
static bfd_size_type dynamic_size;
static size_t dynamic_nent;
static char * dynamic_strings;
static unsigned long dynamic_strings_length;
-static char * string_table;
-static unsigned long string_table_length;
static unsigned long num_dynamic_syms;
static Elf_Internal_Sym * dynamic_symbols;
static Elf_Internal_Syminfo * dynamic_syminfo;
static bfd_vma dynamic_info[DT_ENCODING];
static bfd_vma dynamic_info_DT_GNU_HASH;
static bfd_vma version_info[16];
-static Elf_Internal_Ehdr elf_header;
-static Elf_Internal_Shdr * section_headers;
-static Elf_Internal_Phdr * program_headers;
static Elf_Internal_Dyn * dynamic_section;
static elf_section_list * symtab_shndx_list;
static bfd_boolean show_name = FALSE;
struct group_list
{
- struct group_list * next;
- unsigned int section_index;
+ struct group_list * next;
+ unsigned int section_index;
};
struct group
{
- struct group_list * root;
- unsigned int group_index;
+ struct group_list * root;
+ unsigned int group_index;
};
-static size_t group_count;
-static struct group * section_groups;
-static struct group ** section_headers_groups;
-
-
-/* Flag bits indicating particular types of dump. */
-#define HEX_DUMP (1 << 0) /* The -x command line switch. */
-#define DISASS_DUMP (1 << 1) /* The -i command line switch. */
-#define DEBUG_DUMP (1 << 2) /* The -w command line switch. */
-#define STRING_DUMP (1 << 3) /* The -p command line switch. */
-#define RELOC_DUMP (1 << 4) /* The -R command line switch. */
-
-typedef unsigned char dump_type;
-
-/* A linked list of the section names for which dumps were requested. */
-struct dump_list_entry
-{
- char * name;
- dump_type type;
- struct dump_list_entry * next;
-};
-static struct dump_list_entry * dump_sects_byname;
+static size_t group_count;
+static struct group * section_groups;
+static struct group ** section_headers_groups;
/* A dynamic array of flags indicating for which sections a dump
has been requested via command line switches. */
-static dump_type * cmdline_dump_sects = NULL;
-static unsigned int num_cmdline_dump_sects = 0;
-
-/* A dynamic array of flags indicating for which sections a dump of
- some kind has been requested. It is reset on a per-object file
- basis and then initialised from the cmdline_dump_sects array,
- the results of interpreting the -w switch, and the
- dump_sects_byname list. */
-static dump_type * dump_sects = NULL;
-static unsigned int num_dump_sects = 0;
+static Filedata cmdline;
+static struct dump_list_entry * dump_sects_byname;
/* How to print a vma value. */
typedef enum print_mode
};
static const char * get_symbol_version_string
- (FILE *, bfd_boolean, const char *, unsigned long, unsigned,
+ (Filedata *, bfd_boolean, const char *, unsigned long, unsigned,
Elf_Internal_Sym *, enum versioned_symbol_info *, unsigned short *);
#define UNKNOWN -1
#define SECTION_NAME(X) \
((X) == NULL ? _("<none>") \
- : string_table == NULL ? _("<no-name>") \
- : ((X)->sh_name >= string_table_length ? _("<corrupt>") \
- : string_table + (X)->sh_name))
+ : filedata->string_table == NULL ? _("<no-strings>") \
+ : ((X)->sh_name >= filedata->string_table_length ? _("<corrupt>") \
+ : filedata->string_table + (X)->sh_name))
#define DT_VERSIONTAGIDX(tag) (DT_VERNEEDNUM - (tag)) /* Reverse order! */
#define REMOVE_ARCH_BITS(ADDR) \
do \
{ \
- if (elf_header.e_machine == EM_ARM) \
+ if (filedata->file_header.e_machine == EM_ARM) \
(ADDR) &= ~1; \
} \
while (0)
return ret;
}
-/* Retrieve NMEMB structures, each SIZE bytes long from FILE starting at OFFSET +
- the offset of the current archive member, if we are examining an archive.
- Put the retrieved data into VAR, if it is not NULL. Otherwise allocate a buffer
- using malloc and fill that. In either case return the pointer to the start of
- the retrieved data or NULL if something went wrong. If something does go wrong
- and REASON is not NULL then emit an error message using REASON as part of the
- context. */
+/* Retrieve NMEMB structures, each SIZE bytes long from FILEDATA starting at
+ OFFSET + the offset of the current archive member, if we are examining an
+ archive. Put the retrieved data into VAR, if it is not NULL. Otherwise
+ allocate a buffer using malloc and fill that. In either case return the
+ pointer to the start of the retrieved data or NULL if something went wrong.
+ If something does go wrong and REASON is not NULL then emit an error
+ message using REASON as part of the context. */
static void *
-get_data (void * var, FILE * file, unsigned long offset, bfd_size_type size,
- bfd_size_type nmemb, const char * reason)
+get_data (void * var,
+ Filedata * filedata,
+ unsigned long offset,
+ bfd_size_type size,
+ bfd_size_type nmemb,
+ const char * reason)
{
void * mvar;
bfd_size_type amt = size * nmemb;
/* Be kind to memory chekers (eg valgrind, address sanitizer) by not
attempting to allocate memory when the read is bound to fail. */
- if (amt > current_file_size
- || offset + archive_file_offset + amt > current_file_size)
+ if (amt > filedata->file_size
+ || offset + archive_file_offset + amt > filedata->file_size)
{
if (reason)
error (_("Reading %s bytes extends past end of file for %s\n"),
return NULL;
}
- if (fseek (file, archive_file_offset + offset, SEEK_SET))
+ if (fseek (filedata->handle, archive_file_offset + offset, SEEK_SET))
{
if (reason)
error (_("Unable to seek to 0x%lx for %s\n"),
((char *) mvar)[amt] = '\0';
}
- if (fread (mvar, (size_t) size, (size_t) nmemb, file) != nmemb)
+ if (fread (mvar, (size_t) size, (size_t) nmemb, filedata->handle) != nmemb)
{
if (reason)
error (_("Unable to read in %s bytes of %s\n"),
to print multibyte characters, it just interprets them as hex values. */
static const char *
-printable_section_name (const Elf_Internal_Shdr * sec)
+printable_section_name (Filedata * filedata, const Elf_Internal_Shdr * sec)
{
#define MAX_PRINT_SEC_NAME_LEN 128
static char sec_name_buf [MAX_PRINT_SEC_NAME_LEN + 1];
}
static const char *
-printable_section_name_from_index (unsigned long ndx)
+printable_section_name_from_index (Filedata * filedata, unsigned long ndx)
{
- if (ndx >= elf_header.e_shnum)
+ if (ndx >= filedata->file_header.e_shnum)
return _("<corrupt>");
- return printable_section_name (section_headers + ndx);
+ return printable_section_name (filedata, filedata->section_headers + ndx);
}
/* Return a pointer to section NAME, or NULL if no such section exists. */
static Elf_Internal_Shdr *
-find_section (const char * name)
+find_section (Filedata * filedata, const char * name)
{
unsigned int i;
- for (i = 0; i < elf_header.e_shnum; i++)
- if (streq (SECTION_NAME (section_headers + i), name))
- return section_headers + i;
+ assert (filedata->section_headers != NULL);
+
+ for (i = 0; i < filedata->file_header.e_shnum; i++)
+ if (streq (SECTION_NAME (filedata->section_headers + i), name))
+ return filedata->section_headers + i;
return NULL;
}
section exists. */
static Elf_Internal_Shdr *
-find_section_by_address (bfd_vma addr)
+find_section_by_address (Filedata * filedata, bfd_vma addr)
{
unsigned int i;
- for (i = 0; i < elf_header.e_shnum; i++)
+ for (i = 0; i < filedata->file_header.e_shnum; i++)
{
- Elf_Internal_Shdr *sec = section_headers + i;
+ Elf_Internal_Shdr *sec = filedata->section_headers + i;
+
if (addr >= sec->sh_addr && addr < sec->sh_addr + sec->sh_size)
return sec;
}
}
static Elf_Internal_Shdr *
-find_section_by_type (unsigned int type)
+find_section_by_type (Filedata * filedata, unsigned int type)
{
unsigned int i;
- for (i = 0; i < elf_header.e_shnum; i++)
+ for (i = 0; i < filedata->file_header.e_shnum; i++)
{
- Elf_Internal_Shdr *sec = section_headers + i;
+ Elf_Internal_Shdr *sec = filedata->section_headers + i;
+
if (sec->sh_type == type)
return sec;
}
restricted to the list of sections given in SET. */
static Elf_Internal_Shdr *
-find_section_in_set (const char * name, unsigned int * set)
+find_section_in_set (Filedata * filedata, const char * name, unsigned int * set)
{
unsigned int i;
while ((i = *set++) > 0)
{
/* See PR 21156 for a reproducer. */
- if (i >= elf_header.e_shnum)
+ if (i >= filedata->file_header.e_shnum)
continue; /* FIXME: Should we issue an error message ? */
- if (streq (SECTION_NAME (section_headers + i), name))
- return section_headers + i;
+ if (streq (SECTION_NAME (filedata->section_headers + i), name))
+ return filedata->section_headers + i;
}
}
- return find_section (name);
+ return find_section (filedata, name);
}
/* Read an unsigned LEB128 encoded value from DATA.
many places. */
static inline bfd_boolean
-is_ia64_vms (void)
+is_ia64_vms (Filedata * filedata)
{
- return elf_header.e_machine == EM_IA_64
- && elf_header.e_ident[EI_OSABI] == ELFOSABI_OPENVMS;
+ return filedata->file_header.e_machine == EM_IA_64
+ && filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_OPENVMS;
}
/* Guess the relocation size commonly used by the specific machines. */
}
}
-/* Load RELA type relocations from FILE at REL_OFFSET extending for REL_SIZE bytes.
+/* Load RELA type relocations from FILEDATA at REL_OFFSET extending for REL_SIZE bytes.
Returns TRUE upon success, FALSE otherwise. If successful then a
pointer to a malloc'ed buffer containing the relocs is placed in *RELASP,
and the number of relocs loaded is placed in *NRELASP. It is the caller's
responsibility to free the allocated buffer. */
static bfd_boolean
-slurp_rela_relocs (FILE * file,
- unsigned long rel_offset,
- unsigned long rel_size,
- Elf_Internal_Rela ** relasp,
- unsigned long * nrelasp)
+slurp_rela_relocs (Filedata * filedata,
+ unsigned long rel_offset,
+ unsigned long rel_size,
+ Elf_Internal_Rela ** relasp,
+ unsigned long * nrelasp)
{
Elf_Internal_Rela * relas;
size_t nrelas;
{
Elf32_External_Rela * erelas;
- erelas = (Elf32_External_Rela *) get_data (NULL, file, rel_offset, 1,
+ erelas = (Elf32_External_Rela *) get_data (NULL, filedata, rel_offset, 1,
rel_size, _("32-bit relocation data"));
if (!erelas)
return FALSE;
{
Elf64_External_Rela * erelas;
- erelas = (Elf64_External_Rela *) get_data (NULL, file, rel_offset, 1,
+ erelas = (Elf64_External_Rela *) get_data (NULL, filedata, rel_offset, 1,
rel_size, _("64-bit relocation data"));
if (!erelas)
return FALSE;
warning. We know that if we do not have a 64 bit data
type that we will never execute this code anyway. */
#ifdef BFD64
- if (elf_header.e_machine == EM_MIPS
- && elf_header.e_ident[EI_DATA] != ELFDATA2MSB)
+ if (filedata->file_header.e_machine == EM_MIPS
+ && filedata->file_header.e_ident[EI_DATA] != ELFDATA2MSB)
{
/* In little-endian objects, r_info isn't really a
64-bit little-endian value: it has a 32-bit
return TRUE;
}
-/* Load REL type relocations from FILE at REL_OFFSET extending for REL_SIZE bytes.
+/* Load REL type relocations from FILEDATA at REL_OFFSET extending for REL_SIZE bytes.
Returns TRUE upon success, FALSE otherwise. If successful then a
pointer to a malloc'ed buffer containing the relocs is placed in *RELSP,
and the number of relocs loaded is placed in *NRELSP. It is the caller's
responsibility to free the allocated buffer. */
static bfd_boolean
-slurp_rel_relocs (FILE * file,
- unsigned long rel_offset,
- unsigned long rel_size,
- Elf_Internal_Rela ** relsp,
- unsigned long * nrelsp)
+slurp_rel_relocs (Filedata * filedata,
+ unsigned long rel_offset,
+ unsigned long rel_size,
+ Elf_Internal_Rela ** relsp,
+ unsigned long * nrelsp)
{
Elf_Internal_Rela * rels;
size_t nrels;
{
Elf32_External_Rel * erels;
- erels = (Elf32_External_Rel *) get_data (NULL, file, rel_offset, 1,
+ erels = (Elf32_External_Rel *) get_data (NULL, filedata, rel_offset, 1,
rel_size, _("32-bit relocation data"));
if (!erels)
return FALSE;
{
Elf64_External_Rel * erels;
- erels = (Elf64_External_Rel *) get_data (NULL, file, rel_offset, 1,
+ erels = (Elf64_External_Rel *) get_data (NULL, filedata, rel_offset, 1,
rel_size, _("64-bit relocation data"));
if (!erels)
return FALSE;
warning. We know that if we do not have a 64 bit data
type that we will never execute this code anyway. */
#ifdef BFD64
- if (elf_header.e_machine == EM_MIPS
- && elf_header.e_ident[EI_DATA] != ELFDATA2MSB)
+ if (filedata->file_header.e_machine == EM_MIPS
+ && filedata->file_header.e_ident[EI_DATA] != ELFDATA2MSB)
{
/* In little-endian objects, r_info isn't really a
64-bit little-endian value: it has a 32-bit
/* Returns the reloc type extracted from the reloc info field. */
static unsigned int
-get_reloc_type (bfd_vma reloc_info)
+get_reloc_type (Filedata * filedata, bfd_vma reloc_info)
{
if (is_32bit_elf)
return ELF32_R_TYPE (reloc_info);
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_MIPS:
/* Note: We assume that reloc_info has already been adjusted for us. */
}
static inline bfd_boolean
-uses_msp430x_relocs (void)
+uses_msp430x_relocs (Filedata * filedata)
{
return
- elf_header.e_machine == EM_MSP430 /* Paranoia. */
+ filedata->file_header.e_machine == EM_MSP430 /* Paranoia. */
/* GCC uses osabi == ELFOSBI_STANDALONE. */
- && (((elf_header.e_flags & EF_MSP430_MACH) == E_MSP430_MACH_MSP430X)
+ && (((filedata->file_header.e_flags & EF_MSP430_MACH) == E_MSP430_MACH_MSP430X)
/* TI compiler uses ELFOSABI_NONE. */
- || (elf_header.e_ident[EI_OSABI] == ELFOSABI_NONE));
+ || (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_NONE));
}
/* Display the contents of the relocation data found at the specified
offset. */
static bfd_boolean
-dump_relocations (FILE * file,
- unsigned long rel_offset,
- unsigned long rel_size,
- Elf_Internal_Sym * symtab,
- unsigned long nsyms,
- char * strtab,
- unsigned long strtablen,
- int is_rela,
- bfd_boolean is_dynsym)
+dump_relocations (Filedata * filedata,
+ unsigned long rel_offset,
+ unsigned long rel_size,
+ Elf_Internal_Sym * symtab,
+ unsigned long nsyms,
+ char * strtab,
+ unsigned long strtablen,
+ int is_rela,
+ bfd_boolean is_dynsym)
{
unsigned long i;
Elf_Internal_Rela * rels;
bfd_boolean res = TRUE;
if (is_rela == UNKNOWN)
- is_rela = guess_is_rela (elf_header.e_machine);
+ is_rela = guess_is_rela (filedata->file_header.e_machine);
if (is_rela)
{
- if (!slurp_rela_relocs (file, rel_offset, rel_size, &rels, &rel_size))
+ if (!slurp_rela_relocs (filedata, rel_offset, rel_size, &rels, &rel_size))
return FALSE;
}
else
{
- if (!slurp_rel_relocs (file, rel_offset, rel_size, &rels, &rel_size))
+ if (!slurp_rel_relocs (filedata, rel_offset, rel_size, &rels, &rel_size))
return FALSE;
}
offset = rels[i].r_offset;
inf = rels[i].r_info;
- type = get_reloc_type (inf);
+ type = get_reloc_type (filedata, inf);
symtab_index = get_reloc_symindex (inf);
if (is_32bit_elf)
#endif
}
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
default:
rtype = NULL;
break;
case EM_MSP430:
- if (uses_msp430x_relocs ())
+ if (uses_msp430x_relocs (filedata))
{
rtype = elf_msp430x_reloc_type (type);
break;
else
printf (do_wide ? "%-22s" : "%-17.17s", rtype);
- if (elf_header.e_machine == EM_ALPHA
+ if (filedata->file_header.e_machine == EM_ALPHA
&& rtype != NULL
&& streq (rtype, "R_ALPHA_LITUSE")
&& is_rela)
psym = symtab + symtab_index;
version_string
- = get_symbol_version_string (file, is_dynsym,
+ = get_symbol_version_string (filedata, is_dynsym,
strtab, strtablen,
symtab_index,
psym,
if (ELF_ST_TYPE (psym->st_info) == STT_SECTION)
{
- if (psym->st_shndx < elf_header.e_shnum)
- sec_name = SECTION_NAME (section_headers + psym->st_shndx);
+ if (psym->st_shndx < filedata->file_header.e_shnum)
+ sec_name = SECTION_NAME (filedata->section_headers + psym->st_shndx);
else if (psym->st_shndx == SHN_ABS)
sec_name = "ABS";
else if (psym->st_shndx == SHN_COMMON)
sec_name = "COMMON";
- else if ((elf_header.e_machine == EM_MIPS
+ else if ((filedata->file_header.e_machine == EM_MIPS
&& psym->st_shndx == SHN_MIPS_SCOMMON)
- || (elf_header.e_machine == EM_TI_C6000
+ || (filedata->file_header.e_machine == EM_TI_C6000
&& psym->st_shndx == SHN_TIC6X_SCOMMON))
sec_name = "SCOMMON";
- else if (elf_header.e_machine == EM_MIPS
+ else if (filedata->file_header.e_machine == EM_MIPS
&& psym->st_shndx == SHN_MIPS_SUNDEFINED)
sec_name = "SUNDEF";
- else if ((elf_header.e_machine == EM_X86_64
- || elf_header.e_machine == EM_L1OM
- || elf_header.e_machine == EM_K1OM)
+ else if ((filedata->file_header.e_machine == EM_X86_64
+ || filedata->file_header.e_machine == EM_L1OM
+ || filedata->file_header.e_machine == EM_K1OM)
&& psym->st_shndx == SHN_X86_64_LCOMMON)
sec_name = "LARGE_COMMON";
- else if (elf_header.e_machine == EM_IA_64
- && elf_header.e_ident[EI_OSABI] == ELFOSABI_HPUX
+ else if (filedata->file_header.e_machine == EM_IA_64
+ && filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_HPUX
&& psym->st_shndx == SHN_IA_64_ANSI_COMMON)
sec_name = "ANSI_COM";
- else if (is_ia64_vms ()
+ else if (is_ia64_vms (filedata)
&& psym->st_shndx == SHN_IA_64_VMS_SYMVEC)
sec_name = "VMS_SYMVEC";
else
printf ("%" BFD_VMA_FMT "x", off);
}
- if (elf_header.e_machine == EM_SPARCV9
+ if (filedata->file_header.e_machine == EM_SPARCV9
&& rtype != NULL
&& streq (rtype, "R_SPARC_OLO10"))
printf (" + %lx", (unsigned long) ELF64_R_TYPE_DATA (inf));
putchar ('\n');
#ifdef BFD64
- if (! is_32bit_elf && elf_header.e_machine == EM_MIPS)
+ if (! is_32bit_elf && filedata->file_header.e_machine == EM_MIPS)
{
bfd_vma type2 = ELF64_MIPS_R_TYPE2 (inf);
bfd_vma type3 = ELF64_MIPS_R_TYPE3 (inf);
}
static const char *
-get_dynamic_type (unsigned long type)
+get_dynamic_type (Filedata * filedata, unsigned long type)
{
static char buff[64];
{
const char * result;
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_MIPS:
case EM_MIPS_RS3_LE:
result = get_nios2_dynamic_type (type);
break;
default:
- if (elf_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
+ if (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
result = get_solaris_dynamic_type (type);
else
result = NULL;
snprintf (buff, sizeof (buff), _("Processor Specific: %lx"), type);
}
else if (((type >= DT_LOOS) && (type <= DT_HIOS))
- || (elf_header.e_machine == EM_PARISC
+ || (filedata->file_header.e_machine == EM_PARISC
&& (type >= OLD_DT_LOOS) && (type <= OLD_DT_HIOS)))
{
const char * result;
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_PARISC:
result = get_parisc_dynamic_type (type);
result = get_ia64_dynamic_type (type);
break;
default:
- if (elf_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
+ if (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
result = get_solaris_dynamic_type (type);
else
result = NULL;
}
static char *
-get_machine_flags (unsigned e_flags, unsigned e_machine)
+get_machine_flags (Filedata * filedata, unsigned e_flags, unsigned e_machine)
{
static char buf[1024];
strcat (buf, ", constant gp");
if ((e_flags & EF_IA_64_ABSOLUTE))
strcat (buf, ", absolute");
- if (elf_header.e_ident[EI_OSABI] == ELFOSABI_OPENVMS)
+ if (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_OPENVMS)
{
if ((e_flags & EF_IA_64_VMS_LINKAGES))
strcat (buf, ", vms_linkages");
}
static const char *
-get_osabi_name (unsigned int osabi)
+get_osabi_name (Filedata * filedata, unsigned int osabi)
{
static char buff[32];
case ELFOSABI_OPENVOS: return "Stratus Technologies OpenVOS";
default:
if (osabi >= 64)
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_ARM:
switch (osabi)
}
static const char *
-get_segment_type (unsigned long p_type)
+get_segment_type (Filedata * filedata, unsigned long p_type)
{
static char buff[32];
{
const char * result;
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_AARCH64:
result = get_aarch64_segment_type (p_type);
{
const char * result;
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_PARISC:
result = get_parisc_segment_type (p_type);
result = get_ia64_segment_type (p_type);
break;
default:
- if (elf_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
+ if (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
result = get_solaris_segment_type (p_type);
else
result = NULL;
}
static const char *
-get_ia64_section_type_name (unsigned int sh_type)
+get_ia64_section_type_name (Filedata * filedata, unsigned int sh_type)
{
/* If the top 8 bits are 0x78 the next 8 are the os/abi ID. */
if ((sh_type & 0xFF000000) == SHT_IA_64_LOPSREG)
- return get_osabi_name ((sh_type & 0x00FF0000) >> 16);
+ return get_osabi_name (filedata, (sh_type & 0x00FF0000) >> 16);
switch (sh_type)
{
}
static const char *
-get_section_type_name (unsigned int sh_type)
+get_section_type_name (Filedata * filedata, unsigned int sh_type)
{
static char buff[32];
const char * result;
default:
if ((sh_type >= SHT_LOPROC) && (sh_type <= SHT_HIPROC))
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_ARC:
case EM_ARC_COMPACT:
result = get_parisc_section_type_name (sh_type);
break;
case EM_IA_64:
- result = get_ia64_section_type_name (sh_type);
+ result = get_ia64_section_type_name (filedata, sh_type);
break;
case EM_X86_64:
case EM_L1OM:
}
else if ((sh_type >= SHT_LOOS) && (sh_type <= SHT_HIOS))
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_IA_64:
- result = get_ia64_section_type_name (sh_type);
+ result = get_ia64_section_type_name (filedata, sh_type);
break;
default:
- if (elf_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
+ if (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
result = get_solaris_section_type (sh_type);
else
{
}
else if ((sh_type >= SHT_LOUSER) && (sh_type <= SHT_HIUSER))
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_V800:
case EM_V850:
-R --relocated-dump=<number|name>\n\
Dump the contents of section <number|name> as relocated bytes\n\
-z --decompress Decompress section before dumping it\n\
- -w[lLiaprmfFsoRt] or\n\
+ -w[lLiaprmfFsoRtUuTgAckK] or\n\
--debug-dump[=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 the contents of DWARF2 debug sections\n"));
+ =addr,=cu_index,=links,=follow-links]\n\
+ Display the contents of DWARF debug sections\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\
the first time. */
static void
-request_dump_bynumber (unsigned int section, dump_type type)
+request_dump_bynumber (Filedata * filedata, unsigned int section, dump_type type)
{
- if (section >= num_dump_sects)
+ if (section >= filedata->num_dump_sects)
{
dump_type * new_dump_sects;
new_dump_sects = (dump_type *) calloc (section + 1,
- sizeof (* dump_sects));
+ sizeof (* new_dump_sects));
if (new_dump_sects == NULL)
error (_("Out of memory allocating dump request table.\n"));
else
{
- if (dump_sects)
+ if (filedata->dump_sects)
{
/* Copy current flag settings. */
- memcpy (new_dump_sects, dump_sects, num_dump_sects * sizeof (* dump_sects));
+ memcpy (new_dump_sects, filedata->dump_sects,
+ filedata->num_dump_sects * sizeof (* new_dump_sects));
- free (dump_sects);
+ free (filedata->dump_sects);
}
- dump_sects = new_dump_sects;
- num_dump_sects = section + 1;
+ filedata->dump_sects = new_dump_sects;
+ filedata->num_dump_sects = section + 1;
}
}
- if (dump_sects)
- dump_sects[section] |= type;
-
- return;
+ if (filedata->dump_sects)
+ filedata->dump_sects[section] |= type;
}
/* Request a dump by section name. */
}
static inline void
-request_dump (dump_type type)
+request_dump (Filedata * filedata, dump_type type)
{
int section;
char * cp;
section = strtoul (optarg, & cp, 0);
if (! *cp && section >= 0)
- request_dump_bynumber (section, type);
+ request_dump_bynumber (filedata, section, type);
else
request_dump_byname (optarg, type);
}
-
static void
-parse_args (int argc, char ** argv)
+parse_args (Filedata * filedata, int argc, char ** argv)
{
int c;
do_archive_index = TRUE;
break;
case 'x':
- request_dump (HEX_DUMP);
+ request_dump (filedata, HEX_DUMP);
break;
case 'p':
- request_dump (STRING_DUMP);
+ request_dump (filedata, STRING_DUMP);
break;
case 'R':
- request_dump (RELOC_DUMP);
+ request_dump (filedata, RELOC_DUMP);
break;
case 'z':
decompress_dumps = TRUE;
break;
#ifdef SUPPORT_DISASSEMBLY
case 'i':
- request_dump (DISASS_DUMP);
+ request_dump (filedata, DISASS_DUMP);
break;
#endif
case 'v':
}
}
-/* Decode the data held in 'elf_header'. */
+/* Decode the data held in 'filedata->file_header'. */
static bfd_boolean
-process_file_header (void)
+process_file_header (Filedata * filedata)
{
- if ( elf_header.e_ident[EI_MAG0] != ELFMAG0
- || elf_header.e_ident[EI_MAG1] != ELFMAG1
- || elf_header.e_ident[EI_MAG2] != ELFMAG2
- || elf_header.e_ident[EI_MAG3] != ELFMAG3)
+ Elf_Internal_Ehdr * header = & filedata->file_header;
+
+ if ( header->e_ident[EI_MAG0] != ELFMAG0
+ || header->e_ident[EI_MAG1] != ELFMAG1
+ || header->e_ident[EI_MAG2] != ELFMAG2
+ || header->e_ident[EI_MAG3] != ELFMAG3)
{
error
(_("Not an ELF file - it has the wrong magic bytes at the start\n"));
return FALSE;
}
- init_dwarf_regnames (elf_header.e_machine);
+ init_dwarf_regnames (header->e_machine);
if (do_header)
{
printf (_("ELF Header:\n"));
printf (_(" Magic: "));
for (i = 0; i < EI_NIDENT; i++)
- printf ("%2.2x ", elf_header.e_ident[i]);
+ printf ("%2.2x ", header->e_ident[i]);
printf ("\n");
printf (_(" Class: %s\n"),
- get_elf_class (elf_header.e_ident[EI_CLASS]));
+ get_elf_class (header->e_ident[EI_CLASS]));
printf (_(" Data: %s\n"),
- get_data_encoding (elf_header.e_ident[EI_DATA]));
+ get_data_encoding (header->e_ident[EI_DATA]));
printf (_(" Version: %d %s\n"),
- elf_header.e_ident[EI_VERSION],
- (elf_header.e_ident[EI_VERSION] == EV_CURRENT
+ header->e_ident[EI_VERSION],
+ (header->e_ident[EI_VERSION] == EV_CURRENT
? "(current)"
- : (elf_header.e_ident[EI_VERSION] != EV_NONE
+ : (header->e_ident[EI_VERSION] != EV_NONE
? _("<unknown: %lx>")
: "")));
printf (_(" OS/ABI: %s\n"),
- get_osabi_name (elf_header.e_ident[EI_OSABI]));
+ get_osabi_name (filedata, header->e_ident[EI_OSABI]));
printf (_(" ABI Version: %d\n"),
- elf_header.e_ident[EI_ABIVERSION]);
+ header->e_ident[EI_ABIVERSION]);
printf (_(" Type: %s\n"),
- get_file_type (elf_header.e_type));
+ get_file_type (header->e_type));
printf (_(" Machine: %s\n"),
- get_machine_name (elf_header.e_machine));
+ get_machine_name (header->e_machine));
printf (_(" Version: 0x%lx\n"),
- (unsigned long) elf_header.e_version);
+ (unsigned long) header->e_version);
printf (_(" Entry point address: "));
- print_vma ((bfd_vma) elf_header.e_entry, PREFIX_HEX);
+ print_vma ((bfd_vma) header->e_entry, PREFIX_HEX);
printf (_("\n Start of program headers: "));
- print_vma ((bfd_vma) elf_header.e_phoff, DEC);
+ print_vma ((bfd_vma) header->e_phoff, DEC);
printf (_(" (bytes into file)\n Start of section headers: "));
- print_vma ((bfd_vma) elf_header.e_shoff, DEC);
+ print_vma ((bfd_vma) header->e_shoff, DEC);
printf (_(" (bytes into file)\n"));
printf (_(" Flags: 0x%lx%s\n"),
- (unsigned long) elf_header.e_flags,
- get_machine_flags (elf_header.e_flags, elf_header.e_machine));
+ (unsigned long) header->e_flags,
+ get_machine_flags (filedata, header->e_flags, header->e_machine));
printf (_(" Size of this header: %ld (bytes)\n"),
- (long) elf_header.e_ehsize);
+ (long) header->e_ehsize);
printf (_(" Size of program headers: %ld (bytes)\n"),
- (long) elf_header.e_phentsize);
+ (long) header->e_phentsize);
printf (_(" Number of program headers: %ld"),
- (long) elf_header.e_phnum);
- if (section_headers != NULL
- && elf_header.e_phnum == PN_XNUM
- && section_headers[0].sh_info != 0)
- printf (" (%ld)", (long) section_headers[0].sh_info);
+ (long) header->e_phnum);
+ if (filedata->section_headers != NULL
+ && header->e_phnum == PN_XNUM
+ && filedata->section_headers[0].sh_info != 0)
+ printf (" (%ld)", (long) filedata->section_headers[0].sh_info);
putc ('\n', stdout);
printf (_(" Size of section headers: %ld (bytes)\n"),
- (long) elf_header.e_shentsize);
+ (long) header->e_shentsize);
printf (_(" Number of section headers: %ld"),
- (long) elf_header.e_shnum);
- if (section_headers != NULL && elf_header.e_shnum == SHN_UNDEF)
- printf (" (%ld)", (long) section_headers[0].sh_size);
+ (long) header->e_shnum);
+ if (filedata->section_headers != NULL && header->e_shnum == SHN_UNDEF)
+ printf (" (%ld)", (long) filedata->section_headers[0].sh_size);
putc ('\n', stdout);
printf (_(" Section header string table index: %ld"),
- (long) elf_header.e_shstrndx);
- if (section_headers != NULL
- && elf_header.e_shstrndx == (SHN_XINDEX & 0xffff))
- printf (" (%u)", section_headers[0].sh_link);
- else if (elf_header.e_shstrndx != SHN_UNDEF
- && elf_header.e_shstrndx >= elf_header.e_shnum)
+ (long) header->e_shstrndx);
+ if (filedata->section_headers != NULL
+ && header->e_shstrndx == (SHN_XINDEX & 0xffff))
+ printf (" (%u)", filedata->section_headers[0].sh_link);
+ else if (header->e_shstrndx != SHN_UNDEF
+ && header->e_shstrndx >= header->e_shnum)
printf (_(" <corrupt: out of range>"));
putc ('\n', stdout);
}
- if (section_headers != NULL)
+ if (filedata->section_headers != NULL)
{
- if (elf_header.e_phnum == PN_XNUM
- && section_headers[0].sh_info != 0)
- elf_header.e_phnum = section_headers[0].sh_info;
- if (elf_header.e_shnum == SHN_UNDEF)
- elf_header.e_shnum = section_headers[0].sh_size;
- if (elf_header.e_shstrndx == (SHN_XINDEX & 0xffff))
- elf_header.e_shstrndx = section_headers[0].sh_link;
- else if (elf_header.e_shstrndx >= elf_header.e_shnum)
- elf_header.e_shstrndx = SHN_UNDEF;
- free (section_headers);
- section_headers = NULL;
+ if (header->e_phnum == PN_XNUM
+ && filedata->section_headers[0].sh_info != 0)
+ header->e_phnum = filedata->section_headers[0].sh_info;
+ if (header->e_shnum == SHN_UNDEF)
+ header->e_shnum = filedata->section_headers[0].sh_size;
+ if (header->e_shstrndx == (SHN_XINDEX & 0xffff))
+ header->e_shstrndx = filedata->section_headers[0].sh_link;
+ else if (header->e_shstrndx >= header->e_shnum)
+ header->e_shstrndx = SHN_UNDEF;
+ free (filedata->section_headers);
+ filedata->section_headers = NULL;
}
return TRUE;
}
+/* Read in the program headers from FILEDATA and store them in PHEADERS.
+ Returns TRUE upon success, FALSE otherwise. Loads 32-bit headers. */
+
static bfd_boolean
-get_32bit_program_headers (FILE * file, Elf_Internal_Phdr * pheaders)
+get_32bit_program_headers (Filedata * filedata, Elf_Internal_Phdr * pheaders)
{
Elf32_External_Phdr * phdrs;
Elf32_External_Phdr * external;
Elf_Internal_Phdr * internal;
unsigned int i;
- unsigned int size = elf_header.e_phentsize;
- unsigned int num = elf_header.e_phnum;
+ unsigned int size = filedata->file_header.e_phentsize;
+ unsigned int num = filedata->file_header.e_phnum;
/* PR binutils/17531: Cope with unexpected section header sizes. */
if (size == 0 || num == 0)
if (size > sizeof * phdrs)
warn (_("The e_phentsize field in the ELF header is larger than the size of an ELF program header\n"));
- phdrs = (Elf32_External_Phdr *) get_data (NULL, file, elf_header.e_phoff,
+ phdrs = (Elf32_External_Phdr *) get_data (NULL, filedata, filedata->file_header.e_phoff,
size, num, _("program headers"));
if (phdrs == NULL)
return FALSE;
for (i = 0, internal = pheaders, external = phdrs;
- i < elf_header.e_phnum;
+ i < filedata->file_header.e_phnum;
i++, internal++, external++)
{
internal->p_type = BYTE_GET (external->p_type);
return TRUE;
}
+/* Read in the program headers from FILEDATA and store them in PHEADERS.
+ Returns TRUE upon success, FALSE otherwise. Loads 64-bit headers. */
+
static bfd_boolean
-get_64bit_program_headers (FILE * file, Elf_Internal_Phdr * pheaders)
+get_64bit_program_headers (Filedata * filedata, Elf_Internal_Phdr * pheaders)
{
Elf64_External_Phdr * phdrs;
Elf64_External_Phdr * external;
Elf_Internal_Phdr * internal;
unsigned int i;
- unsigned int size = elf_header.e_phentsize;
- unsigned int num = elf_header.e_phnum;
+ unsigned int size = filedata->file_header.e_phentsize;
+ unsigned int num = filedata->file_header.e_phnum;
/* PR binutils/17531: Cope with unexpected section header sizes. */
if (size == 0 || num == 0)
if (size > sizeof * phdrs)
warn (_("The e_phentsize field in the ELF header is larger than the size of an ELF program header\n"));
- phdrs = (Elf64_External_Phdr *) get_data (NULL, file, elf_header.e_phoff,
+ phdrs = (Elf64_External_Phdr *) get_data (NULL, filedata, filedata->file_header.e_phoff,
size, num, _("program headers"));
if (!phdrs)
return FALSE;
for (i = 0, internal = pheaders, external = phdrs;
- i < elf_header.e_phnum;
+ i < filedata->file_header.e_phnum;
i++, internal++, external++)
{
internal->p_type = BYTE_GET (external->p_type);
/* Returns TRUE if the program headers were read into `program_headers'. */
static bfd_boolean
-get_program_headers (FILE * file)
+get_program_headers (Filedata * filedata)
{
Elf_Internal_Phdr * phdrs;
/* Check cache of prior read. */
- if (program_headers != NULL)
+ if (filedata->program_headers != NULL)
return TRUE;
/* Be kind to memory checkers by looking for
e_phnum values which we know must be invalid. */
- if (elf_header.e_phnum
+ if (filedata->file_header.e_phnum
* (is_32bit_elf ? sizeof (Elf32_External_Phdr) : sizeof (Elf64_External_Phdr))
- >= current_file_size)
+ >= filedata->file_size)
{
error (_("Too many program headers - %#x - the file is not that big\n"),
- elf_header.e_phnum);
+ filedata->file_header.e_phnum);
return FALSE;
}
- phdrs = (Elf_Internal_Phdr *) cmalloc (elf_header.e_phnum,
+ phdrs = (Elf_Internal_Phdr *) cmalloc (filedata->file_header.e_phnum,
sizeof (Elf_Internal_Phdr));
if (phdrs == NULL)
{
error (_("Out of memory reading %u program headers\n"),
- elf_header.e_phnum);
+ filedata->file_header.e_phnum);
return FALSE;
}
if (is_32bit_elf
- ? get_32bit_program_headers (file, phdrs)
- : get_64bit_program_headers (file, phdrs))
+ ? get_32bit_program_headers (filedata, phdrs)
+ : get_64bit_program_headers (filedata, phdrs))
{
- program_headers = phdrs;
+ filedata->program_headers = phdrs;
return TRUE;
}
/* Returns TRUE if the program headers were loaded. */
static bfd_boolean
-process_program_headers (FILE * file)
+process_program_headers (Filedata * filedata)
{
Elf_Internal_Phdr * segment;
unsigned int i;
Elf_Internal_Phdr * previous_load = NULL;
- if (elf_header.e_phnum == 0)
+ if (filedata->file_header.e_phnum == 0)
{
/* PR binutils/12467. */
- if (elf_header.e_phoff != 0)
+ if (filedata->file_header.e_phoff != 0)
{
warn (_("possibly corrupt ELF header - it has a non-zero program"
" header offset, but no program headers\n"));
if (do_segments && !do_header)
{
- printf (_("\nElf file type is %s\n"), get_file_type (elf_header.e_type));
- printf (_("Entry point 0x%s\n"), bfd_vmatoa ("x", elf_header.e_entry));
+ printf (_("\nElf file type is %s\n"), get_file_type (filedata->file_header.e_type));
+ printf (_("Entry point 0x%s\n"), bfd_vmatoa ("x", filedata->file_header.e_entry));
printf (ngettext ("There is %d program header, starting at offset %s\n",
"There are %d program headers, starting at offset %s\n",
- elf_header.e_phnum),
- elf_header.e_phnum,
- bfd_vmatoa ("u", elf_header.e_phoff));
+ filedata->file_header.e_phnum),
+ filedata->file_header.e_phnum,
+ bfd_vmatoa ("u", filedata->file_header.e_phoff));
}
- if (! get_program_headers (file))
+ if (! get_program_headers (filedata))
return TRUE;
if (do_segments)
{
- if (elf_header.e_phnum > 1)
+ if (filedata->file_header.e_phnum > 1)
printf (_("\nProgram Headers:\n"));
else
printf (_("\nProgram Headers:\n"));
dynamic_addr = 0;
dynamic_size = 0;
- for (i = 0, segment = program_headers;
- i < elf_header.e_phnum;
+ for (i = 0, segment = filedata->program_headers;
+ i < filedata->file_header.e_phnum;
i++, segment++)
{
if (do_segments)
{
- printf (" %-14.14s ", get_segment_type (segment->p_type));
+ printf (" %-14.14s ", get_segment_type (filedata, segment->p_type));
if (is_32bit_elf)
{
/* PR 20815 - Verify that the program header is loaded into memory. */
if (i > 0 && previous_load != NULL)
error (_("the PHDR segment must occur before any LOAD segment\n"));
- if (elf_header.e_machine != EM_PARISC)
+ if (filedata->file_header.e_machine != EM_PARISC)
{
unsigned int j;
- for (j = 1; j < elf_header.e_phnum; j++)
- if (program_headers[j].p_vaddr <= segment->p_vaddr
- && (program_headers[j].p_vaddr + program_headers[j].p_memsz)
+ for (j = 1; j < filedata->file_header.e_phnum; j++)
+ if (filedata->program_headers[j].p_vaddr <= segment->p_vaddr
+ && (filedata->program_headers[j].p_vaddr
+ + filedata->program_headers[j].p_memsz)
>= (segment->p_vaddr + segment->p_filesz))
break;
- if (j == elf_header.e_phnum)
+ if (j == filedata->file_header.e_phnum)
error (_("the PHDR segment is not covered by a LOAD segment\n"));
}
break;
/* Try to locate the .dynamic section. If there is
a section header table, we can easily locate it. */
- if (section_headers != NULL)
+ if (filedata->section_headers != NULL)
{
Elf_Internal_Shdr * sec;
- sec = find_section (".dynamic");
+ sec = find_section (filedata, ".dynamic");
if (sec == NULL || sec->sh_size == 0)
{
/* A corresponding .dynamic section is expected, but on
IA-64/OpenVMS it is OK for it to be missing. */
- if (!is_ia64_vms ())
+ if (!is_ia64_vms (filedata))
error (_("no .dynamic section in the dynamic segment\n"));
break;
}
segment. Check this after matching against the section headers
so we don't warn on debuginfo file (which have NOBITS .dynamic
sections). */
- if (dynamic_addr + dynamic_size >= current_file_size)
+ if (dynamic_addr + dynamic_size >= filedata->file_size)
{
error (_("the dynamic segment offset + size exceeds the size of the file\n"));
dynamic_addr = dynamic_size = 0;
break;
case PT_INTERP:
- if (fseek (file, archive_file_offset + (long) segment->p_offset,
+ if (fseek (filedata->handle, archive_file_offset + (long) segment->p_offset,
SEEK_SET))
error (_("Unable to find program interpreter name\n"));
else
error (_("Internal error: failed to create format string to display program interpreter\n"));
program_interpreter[0] = 0;
- if (fscanf (file, fmt, program_interpreter) <= 0)
+ if (fscanf (filedata->handle, fmt, program_interpreter) <= 0)
error (_("Unable to read program interpreter name\n"));
if (do_segments)
}
}
- if (do_segments && section_headers != NULL && string_table != NULL)
+ if (do_segments
+ && filedata->section_headers != NULL
+ && filedata->string_table != NULL)
{
printf (_("\n Section to Segment mapping:\n"));
printf (_(" Segment Sections...\n"));
- for (i = 0; i < elf_header.e_phnum; i++)
+ for (i = 0; i < filedata->file_header.e_phnum; i++)
{
unsigned int j;
Elf_Internal_Shdr * section;
- segment = program_headers + i;
- section = section_headers + 1;
+ segment = filedata->program_headers + i;
+ section = filedata->section_headers + 1;
printf (" %2.2d ", i);
- for (j = 1; j < elf_header.e_shnum; j++, section++)
+ for (j = 1; j < filedata->file_header.e_shnum; j++, section++)
{
if (!ELF_TBSS_SPECIAL (section, segment)
&& ELF_SECTION_IN_SEGMENT_STRICT (section, segment))
- printf ("%s ", printable_section_name (section));
+ printf ("%s ", printable_section_name (filedata, section));
}
putc ('\n',stdout);
/* Find the file offset corresponding to VMA by using the program headers. */
static long
-offset_from_vma (FILE * file, bfd_vma vma, bfd_size_type size)
+offset_from_vma (Filedata * filedata, bfd_vma vma, bfd_size_type size)
{
Elf_Internal_Phdr * seg;
- if (! get_program_headers (file))
+ if (! get_program_headers (filedata))
{
warn (_("Cannot interpret virtual addresses without program headers.\n"));
return (long) vma;
}
- for (seg = program_headers;
- seg < program_headers + elf_header.e_phnum;
+ for (seg = filedata->program_headers;
+ seg < filedata->program_headers + filedata->file_header.e_phnum;
++seg)
{
if (seg->p_type != PT_LOAD)
}
-/* Allocate memory and load the sections headers into the global pointer
- SECTION_HEADERS. If PROBE is true, this is just a probe and we do not
- generate any error messages if the load fails. */
+/* Allocate memory and load the sections headers into FILEDATA->filedata->section_headers.
+ If PROBE is true, this is just a probe and we do not generate any error
+ messages if the load fails. */
static bfd_boolean
-get_32bit_section_headers (FILE * file, bfd_boolean probe)
+get_32bit_section_headers (Filedata * filedata, bfd_boolean probe)
{
Elf32_External_Shdr * shdrs;
Elf_Internal_Shdr * internal;
- unsigned int i;
- unsigned int size = elf_header.e_shentsize;
- unsigned int num = probe ? 1 : elf_header.e_shnum;
+ unsigned int i;
+ unsigned int size = filedata->file_header.e_shentsize;
+ unsigned int num = probe ? 1 : filedata->file_header.e_shnum;
/* PR binutils/17531: Cope with unexpected section header sizes. */
if (size == 0 || num == 0)
if (!probe && size > sizeof * shdrs)
warn (_("The e_shentsize field in the ELF header is larger than the size of an ELF section header\n"));
- shdrs = (Elf32_External_Shdr *) get_data (NULL, file, elf_header.e_shoff,
+ shdrs = (Elf32_External_Shdr *) get_data (NULL, filedata, filedata->file_header.e_shoff,
size, num,
probe ? NULL : _("section headers"));
if (shdrs == NULL)
return FALSE;
- if (section_headers != NULL)
- free (section_headers);
- section_headers = (Elf_Internal_Shdr *) cmalloc (num,
- sizeof (Elf_Internal_Shdr));
- if (section_headers == NULL)
+ free (filedata->section_headers);
+ filedata->section_headers = (Elf_Internal_Shdr *)
+ cmalloc (num, sizeof (Elf_Internal_Shdr));
+ if (filedata->section_headers == NULL)
{
if (!probe)
error (_("Out of memory reading %u section headers\n"), num);
return FALSE;
}
- for (i = 0, internal = section_headers;
+ for (i = 0, internal = filedata->section_headers;
i < num;
i++, internal++)
{
return TRUE;
}
+/* Like get_32bit_section_headers, except that it fetches 64-bit headers. */
+
static bfd_boolean
-get_64bit_section_headers (FILE * file, bfd_boolean probe)
+get_64bit_section_headers (Filedata * filedata, bfd_boolean probe)
{
- Elf64_External_Shdr * shdrs;
- Elf_Internal_Shdr * internal;
- unsigned int i;
- unsigned int size = elf_header.e_shentsize;
- unsigned int num = probe ? 1 : elf_header.e_shnum;
+ Elf64_External_Shdr * shdrs;
+ Elf_Internal_Shdr * internal;
+ unsigned int i;
+ unsigned int size = filedata->file_header.e_shentsize;
+ unsigned int num = probe ? 1 : filedata->file_header.e_shnum;
/* PR binutils/17531: Cope with unexpected section header sizes. */
if (size == 0 || num == 0)
return FALSE;
+
if (size < sizeof * shdrs)
{
if (! probe)
error (_("The e_shentsize field in the ELF header is less than the size of an ELF section header\n"));
return FALSE;
}
+
if (! probe && size > sizeof * shdrs)
warn (_("The e_shentsize field in the ELF header is larger than the size of an ELF section header\n"));
- shdrs = (Elf64_External_Shdr *) get_data (NULL, file, elf_header.e_shoff,
+ shdrs = (Elf64_External_Shdr *) get_data (NULL, filedata,
+ filedata->file_header.e_shoff,
size, num,
probe ? NULL : _("section headers"));
if (shdrs == NULL)
return FALSE;
- if (section_headers != NULL)
- free (section_headers);
- section_headers = (Elf_Internal_Shdr *) cmalloc (num,
- sizeof (Elf_Internal_Shdr));
- if (section_headers == NULL)
+ free (filedata->section_headers);
+ filedata->section_headers = (Elf_Internal_Shdr *)
+ cmalloc (num, sizeof (Elf_Internal_Shdr));
+ if (filedata->section_headers == NULL)
{
if (! probe)
error (_("Out of memory reading %u section headers\n"), num);
return FALSE;
}
- for (i = 0, internal = section_headers;
+ for (i = 0, internal = filedata->section_headers;
i < num;
i++, internal++)
{
}
static Elf_Internal_Sym *
-get_32bit_elf_symbols (FILE * file,
- Elf_Internal_Shdr * section,
- unsigned long * num_syms_return)
+get_32bit_elf_symbols (Filedata * filedata,
+ Elf_Internal_Shdr * section,
+ unsigned long * num_syms_return)
{
unsigned long number = 0;
Elf32_External_Sym * esyms = NULL;
if (section->sh_entsize == 0 || section->sh_entsize > section->sh_size)
{
error (_("Section %s has an invalid sh_entsize of 0x%lx\n"),
- printable_section_name (section), (unsigned long) section->sh_entsize);
+ printable_section_name (filedata, section),
+ (unsigned long) section->sh_entsize);
goto exit_point;
}
- if (section->sh_size > current_file_size)
+ if (section->sh_size > filedata->file_size)
{
error (_("Section %s has an invalid sh_size of 0x%lx\n"),
- printable_section_name (section), (unsigned long) section->sh_size);
+ printable_section_name (filedata, section),
+ (unsigned long) section->sh_size);
goto exit_point;
}
{
error (_("Size (0x%lx) of section %s is not a multiple of its sh_entsize (0x%lx)\n"),
(unsigned long) section->sh_size,
- printable_section_name (section),
+ printable_section_name (filedata, section),
(unsigned long) section->sh_entsize);
goto exit_point;
}
- esyms = (Elf32_External_Sym *) get_data (NULL, file, section->sh_offset, 1,
+ esyms = (Elf32_External_Sym *) get_data (NULL, filedata, section->sh_offset, 1,
section->sh_size, _("symbols"));
if (esyms == NULL)
goto exit_point;
shndx = NULL;
for (entry = symtab_shndx_list; entry != NULL; entry = entry->next)
- if (entry->hdr->sh_link == (unsigned long) (section - section_headers))
+ if (entry->hdr->sh_link == (unsigned long) (section - filedata->section_headers))
{
- shndx = (Elf_External_Sym_Shndx *) get_data (NULL, file,
+ shndx = (Elf_External_Sym_Shndx *) get_data (NULL, filedata,
entry->hdr->sh_offset,
1, entry->hdr->sh_size,
_("symbol table section indicies"));
else if (entry->hdr->sh_size / sizeof (Elf_External_Sym_Shndx) < number)
{
error (_("Index section %s has an sh_size of 0x%lx - expected 0x%lx\n"),
- printable_section_name (entry->hdr),
+ printable_section_name (filedata, entry->hdr),
(unsigned long) entry->hdr->sh_size,
(unsigned long) section->sh_size);
goto exit_point;
}
static Elf_Internal_Sym *
-get_64bit_elf_symbols (FILE * file,
- Elf_Internal_Shdr * section,
- unsigned long * num_syms_return)
+get_64bit_elf_symbols (Filedata * filedata,
+ Elf_Internal_Shdr * section,
+ unsigned long * num_syms_return)
{
unsigned long number = 0;
Elf64_External_Sym * esyms = NULL;
if (section->sh_entsize == 0 || section->sh_entsize > section->sh_size)
{
error (_("Section %s has an invalid sh_entsize of 0x%lx\n"),
- printable_section_name (section),
+ printable_section_name (filedata, section),
(unsigned long) section->sh_entsize);
goto exit_point;
}
- if (section->sh_size > current_file_size)
+ if (section->sh_size > filedata->file_size)
{
error (_("Section %s has an invalid sh_size of 0x%lx\n"),
- printable_section_name (section),
+ printable_section_name (filedata, section),
(unsigned long) section->sh_size);
goto exit_point;
}
{
error (_("Size (0x%lx) of section %s is not a multiple of its sh_entsize (0x%lx)\n"),
(unsigned long) section->sh_size,
- printable_section_name (section),
+ printable_section_name (filedata, section),
(unsigned long) section->sh_entsize);
goto exit_point;
}
- esyms = (Elf64_External_Sym *) get_data (NULL, file, section->sh_offset, 1,
+ esyms = (Elf64_External_Sym *) get_data (NULL, filedata, section->sh_offset, 1,
section->sh_size, _("symbols"));
if (!esyms)
goto exit_point;
shndx = NULL;
for (entry = symtab_shndx_list; entry != NULL; entry = entry->next)
- if (entry->hdr->sh_link == (unsigned long) (section - section_headers))
+ if (entry->hdr->sh_link == (unsigned long) (section - filedata->section_headers))
{
- shndx = (Elf_External_Sym_Shndx *) get_data (NULL, file,
+ shndx = (Elf_External_Sym_Shndx *) get_data (NULL, filedata,
entry->hdr->sh_offset,
1, entry->hdr->sh_size,
_("symbol table section indicies"));
else if (entry->hdr->sh_size / sizeof (Elf_External_Sym_Shndx) < number)
{
error (_("Index section %s has an sh_size of 0x%lx - expected 0x%lx\n"),
- printable_section_name (entry->hdr),
+ printable_section_name (filedata, entry->hdr),
(unsigned long) entry->hdr->sh_size,
(unsigned long) section->sh_size);
goto exit_point;
}
static const char *
-get_elf_section_flags (bfd_vma sh_flags)
+get_elf_section_flags (Filedata * filedata, bfd_vma sh_flags)
{
static char buff[1024];
char * p = buff;
default:
sindex = -1;
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_IA_64:
if (flag == SHF_IA_64_SHORT)
else if (flag == SHF_IA_64_NORECOV)
sindex = 11;
#ifdef BFD64
- else if (elf_header.e_ident[EI_OSABI] == ELFOSABI_OPENVMS)
+ else if (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_OPENVMS)
switch (flag)
{
case SHF_IA_64_VMS_GLOBAL: sindex = 12; break;
case SHF_GNU_MBIND: *p = 'D'; break;
default:
- if ((elf_header.e_machine == EM_X86_64
- || elf_header.e_machine == EM_L1OM
- || elf_header.e_machine == EM_K1OM)
+ if ((filedata->file_header.e_machine == EM_X86_64
+ || filedata->file_header.e_machine == EM_L1OM
+ || filedata->file_header.e_machine == EM_K1OM)
&& flag == SHF_X86_64_LARGE)
*p = 'l';
- else if (elf_header.e_machine == EM_ARM
+ else if (filedata->file_header.e_machine == EM_ARM
&& flag == SHF_ARM_PURECODE)
*p = 'y';
- else if (elf_header.e_machine == EM_PPC
+ else if (filedata->file_header.e_machine == EM_PPC
&& flag == SHF_PPC_VLE)
*p = 'v';
else if (flag & SHF_MASKOS)
}
static bfd_boolean
-process_section_headers (FILE * file)
+process_section_headers (Filedata * filedata)
{
Elf_Internal_Shdr * section;
unsigned int i;
- section_headers = NULL;
+ filedata->section_headers = NULL;
- if (elf_header.e_shnum == 0)
+ if (filedata->file_header.e_shnum == 0)
{
/* PR binutils/12467. */
- if (elf_header.e_shoff != 0)
+ if (filedata->file_header.e_shoff != 0)
{
warn (_("possibly corrupt ELF file header - it has a non-zero"
" section header offset, but no section headers\n"));
"starting at offset 0x%lx:\n",
"There are %d section headers, "
"starting at offset 0x%lx:\n",
- elf_header.e_shnum),
- elf_header.e_shnum, (unsigned long) elf_header.e_shoff);
+ filedata->file_header.e_shnum),
+ filedata->file_header.e_shnum,
+ (unsigned long) filedata->file_header.e_shoff);
if (is_32bit_elf)
{
- if (! get_32bit_section_headers (file, FALSE))
+ if (! get_32bit_section_headers (filedata, FALSE))
return FALSE;
}
else
{
- if (! get_64bit_section_headers (file, FALSE))
+ if (! get_64bit_section_headers (filedata, FALSE))
return FALSE;
}
/* Read in the string table, so that we have names to display. */
- if (elf_header.e_shstrndx != SHN_UNDEF
- && elf_header.e_shstrndx < elf_header.e_shnum)
+ if (filedata->file_header.e_shstrndx != SHN_UNDEF
+ && filedata->file_header.e_shstrndx < filedata->file_header.e_shnum)
{
- section = section_headers + elf_header.e_shstrndx;
+ section = filedata->section_headers + filedata->file_header.e_shstrndx;
if (section->sh_size != 0)
{
- string_table = (char *) get_data (NULL, file, section->sh_offset,
- 1, section->sh_size,
- _("string table"));
+ filedata->string_table = (char *) get_data (NULL, filedata, section->sh_offset,
+ 1, section->sh_size,
+ _("string table"));
- string_table_length = string_table != NULL ? section->sh_size : 0;
+ filedata->string_table_length = filedata->string_table != NULL ? section->sh_size : 0;
}
}
symtab_shndx_list = NULL;
eh_addr_size = is_32bit_elf ? 4 : 8;
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_MIPS:
case EM_MIPS_RS3_LE:
earlier compilers provided no way of distinguishing ILP32 objects
from LP64 objects, so if there's any doubt, we should assume that
the official LP64 form is being used. */
- if ((elf_header.e_flags & EF_MIPS_ABI) == E_MIPS_ABI_EABI64
- && find_section (".gcc_compiled_long32") == NULL)
+ if ((filedata->file_header.e_flags & EF_MIPS_ABI) == E_MIPS_ABI_EABI64
+ && find_section (filedata, ".gcc_compiled_long32") == NULL)
eh_addr_size = 8;
break;
case EM_H8_300:
case EM_H8_300H:
- switch (elf_header.e_flags & EF_H8_MACH)
+ switch (filedata->file_header.e_flags & EF_H8_MACH)
{
case E_H8_MACH_H8300:
case E_H8_MACH_H8300HN:
case EM_M32C_OLD:
case EM_M32C:
- switch (elf_header.e_flags & EF_M32C_CPU_MASK)
+ switch (filedata->file_header.e_flags & EF_M32C_CPU_MASK)
{
case EF_M32C_CPU_M16C:
eh_addr_size = 2;
CHECK_ENTSIZE_VALUES (section, i, sizeof (Elf32_External_##type), \
sizeof (Elf64_External_##type))
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum;
i++, section++)
{
char * name = SECTION_NAME (section);
}
CHECK_ENTSIZE (section, i, Sym);
- dynamic_symbols = GET_ELF_SYMBOLS (file, section, & num_dynamic_syms);
+ dynamic_symbols = GET_ELF_SYMBOLS (filedata, section, & num_dynamic_syms);
}
else if (section->sh_type == SHT_STRTAB
&& streq (name, ".dynstr"))
continue;
}
- dynamic_strings = (char *) get_data (NULL, file, section->sh_offset,
+ dynamic_strings = (char *) get_data (NULL, filedata, section->sh_offset,
1, section->sh_size,
_("dynamic strings"));
dynamic_strings_length = dynamic_strings == NULL ? 0 : section->sh_size;
else if (section->sh_type == SHT_SYMTAB_SHNDX)
{
elf_section_list * entry = xmalloc (sizeof * entry);
+
entry->hdr = section;
entry->next = symtab_shndx_list;
symtab_shndx_list = entry;
|| (do_debug_cu_index && const_strneq (name, "cu_index"))
|| (do_debug_cu_index && const_strneq (name, "tu_index"))
)
- request_dump_bynumber (i, DEBUG_DUMP);
+ request_dump_bynumber (filedata, i, DEBUG_DUMP);
}
/* Linkonce section to be combined with .debug_info at link time. */
else if ((do_debugging || do_debug_info)
&& const_strneq (name, ".gnu.linkonce.wi."))
- request_dump_bynumber (i, DEBUG_DUMP);
+ request_dump_bynumber (filedata, i, DEBUG_DUMP);
else if (do_debug_frames && streq (name, ".eh_frame"))
- request_dump_bynumber (i, DEBUG_DUMP);
+ request_dump_bynumber (filedata, i, DEBUG_DUMP);
else if (do_gdb_index && (streq (name, ".gdb_index")
|| streq (name, ".debug_names")))
- request_dump_bynumber (i, DEBUG_DUMP);
+ request_dump_bynumber (filedata, i, DEBUG_DUMP);
/* Trace sections for Itanium VMS. */
else if ((do_debugging || do_trace_info || do_trace_abbrevs
|| do_trace_aranges)
|| (do_trace_abbrevs && streq (name, "abbrev"))
|| (do_trace_aranges && streq (name, "aranges"))
)
- request_dump_bynumber (i, DEBUG_DUMP);
+ request_dump_bynumber (filedata, i, DEBUG_DUMP);
}
+ else if ((do_debugging || do_debug_links)
+ && (const_strneq (name, ".gnu_debuglink")
+ || const_strneq (name, ".gnu_debugaltlink")))
+ request_dump_bynumber (filedata, i, DEBUG_DUMP);
}
if (! do_sections)
return TRUE;
- if (elf_header.e_shnum > 1)
+ if (filedata->file_header.e_shnum > 1)
printf (_("\nSection Headers:\n"));
else
printf (_("\nSection Header:\n"));
if (do_section_details)
printf (_(" Flags\n"));
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum;
i++, section++)
{
/* Run some sanity checks on the section header. */
case SHT_REL:
case SHT_RELA:
if (section->sh_link < 1
- || section->sh_link >= elf_header.e_shnum
- || (section_headers[section->sh_link].sh_type != SHT_SYMTAB
- && section_headers[section->sh_link].sh_type != SHT_DYNSYM))
+ || section->sh_link >= filedata->file_header.e_shnum
+ || (filedata->section_headers[section->sh_link].sh_type != SHT_SYMTAB
+ && filedata->section_headers[section->sh_link].sh_type != SHT_DYNSYM))
warn (_("[%2u]: Link field (%u) should index a symtab section.\n"),
i, section->sh_link);
break;
case SHT_GNU_verdef:
case SHT_GNU_LIBLIST:
if (section->sh_link < 1
- || section->sh_link >= elf_header.e_shnum
- || section_headers[section->sh_link].sh_type != SHT_STRTAB)
+ || section->sh_link >= filedata->file_header.e_shnum
+ || filedata->section_headers[section->sh_link].sh_type != SHT_STRTAB)
warn (_("[%2u]: Link field (%u) should index a string section.\n"),
i, section->sh_link);
break;
case SHT_REL:
case SHT_RELA:
if (section->sh_info < 1
- || section->sh_info >= elf_header.e_shnum
- || (section_headers[section->sh_info].sh_type != SHT_PROGBITS
- && section_headers[section->sh_info].sh_type != SHT_NOBITS
- && section_headers[section->sh_info].sh_type != SHT_NOTE
- && section_headers[section->sh_info].sh_type != SHT_INIT_ARRAY
+ || section->sh_info >= filedata->file_header.e_shnum
+ || (filedata->section_headers[section->sh_info].sh_type != SHT_PROGBITS
+ && filedata->section_headers[section->sh_info].sh_type != SHT_NOBITS
+ && filedata->section_headers[section->sh_info].sh_type != SHT_NOTE
+ && filedata->section_headers[section->sh_info].sh_type != SHT_INIT_ARRAY
/* FIXME: Are other section types valid ? */
- && section_headers[section->sh_info].sh_type < SHT_LOOS))
+ && filedata->section_headers[section->sh_info].sh_type < SHT_LOOS))
{
if (section->sh_info == 0
&& (streq (SECTION_NAME (section), ".rel.dyn")
;
else if (section->sh_flags & SHF_INFO_LINK)
{
- if (section->sh_info < 1 || section->sh_info >= elf_header.e_shnum)
+ if (section->sh_info < 1 || section->sh_info >= filedata->file_header.e_shnum)
warn (_("[%2u]: Expected link to another section in info field"), i);
}
else if (section->sh_type < SHT_LOOS
}
/* Check the sh_size field. */
- if (section->sh_size > current_file_size
+ if (section->sh_size > filedata->file_size
&& section->sh_type != SHT_NOBITS
&& section->sh_type != SHT_NULL
&& section->sh_type < SHT_LOOS)
printf (" [%2u] ", i);
if (do_section_details)
- printf ("%s\n ", printable_section_name (section));
+ printf ("%s\n ", printable_section_name (filedata, section));
else
print_symbol (-17, SECTION_NAME (section));
printf (do_wide ? " %-15s " : " %-15.15s ",
- get_section_type_name (section->sh_type));
+ get_section_type_name (filedata, section->sh_type));
if (is_32bit_elf)
{
if (do_section_details)
fputs (" ", stdout);
else
- printf (" %3s ", get_elf_section_flags (section->sh_flags));
+ printf (" %3s ", get_elf_section_flags (filedata, section->sh_flags));
- if (section->sh_link >= elf_header.e_shnum)
+ if (section->sh_link >= filedata->file_header.e_shnum)
{
link_too_big = "";
/* The sh_link value is out of range. Normally this indicates
an error but it can have special values in Solaris binaries. */
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_386:
case EM_IAMCU:
if (do_section_details)
fputs (" ", stdout);
else
- printf (" %3s ", get_elf_section_flags (section->sh_flags));
+ printf (" %3s ", get_elf_section_flags (filedata, section->sh_flags));
printf ("%2u %3u ", section->sh_link, section->sh_info);
else if (do_section_details)
{
printf (" %-15.15s ",
- get_section_type_name (section->sh_type));
+ get_section_type_name (filedata, section->sh_type));
print_vma (section->sh_addr, LONG_HEX);
if ((long) section->sh_offset == section->sh_offset)
printf (" %16.16lx", (unsigned long) section->sh_offset);
printf (" ");
print_vma (section->sh_entsize, LONG_HEX);
- printf (" %3s ", get_elf_section_flags (section->sh_flags));
+ printf (" %3s ", get_elf_section_flags (filedata, section->sh_flags));
printf (" %2u %3u %lu\n",
section->sh_link,
if (do_section_details)
{
- printf (" %s\n", get_elf_section_flags (section->sh_flags));
+ printf (" %s\n", get_elf_section_flags (filedata, section->sh_flags));
if ((section->sh_flags & SHF_COMPRESSED) != 0)
{
/* Minimum section size is 12 bytes for 32-bit compression
unsigned char buf[24];
assert (sizeof (buf) >= sizeof (Elf64_External_Chdr));
- if (get_data (&buf, (FILE *) file, section->sh_offset, 1,
+ if (get_data (&buf, filedata, section->sh_offset, 1,
sizeof (buf), _("compression header")))
{
Elf_Internal_Chdr chdr;
W (write), A (alloc), X (execute), M (merge), S (strings), I (info),\n\
L (link order), O (extra OS processing required), G (group), T (TLS),\n\
C (compressed), x (unknown), o (OS specific), E (exclude),\n "));
- if (elf_header.e_machine == EM_X86_64
- || elf_header.e_machine == EM_L1OM
- || elf_header.e_machine == EM_K1OM)
+ if (filedata->file_header.e_machine == EM_X86_64
+ || filedata->file_header.e_machine == EM_L1OM
+ || filedata->file_header.e_machine == EM_K1OM)
printf (_("l (large), "));
- else if (elf_header.e_machine == EM_ARM)
+ else if (filedata->file_header.e_machine == EM_ARM)
printf (_("y (purecode), "));
- else if (elf_header.e_machine == EM_PPC)
+ else if (filedata->file_header.e_machine == EM_PPC)
printf (_("v (VLE), "));
printf ("p (processor specific)\n");
}
}
static bfd_boolean
-process_section_groups (FILE * file)
+process_section_groups (Filedata * filedata)
{
Elf_Internal_Shdr * section;
unsigned int i;
if (!do_unwind && !do_section_groups)
return TRUE;
- if (elf_header.e_shnum == 0)
+ if (filedata->file_header.e_shnum == 0)
{
if (do_section_groups)
printf (_("\nThere are no sections to group in this file.\n"));
return TRUE;
}
- if (section_headers == NULL)
+ if (filedata->section_headers == NULL)
{
error (_("Section headers are not available!\n"));
/* PR 13622: This can happen with a corrupt ELF header. */
return FALSE;
}
- section_headers_groups = (struct group **) calloc (elf_header.e_shnum,
+ section_headers_groups = (struct group **) calloc (filedata->file_header.e_shnum,
sizeof (struct group *));
if (section_headers_groups == NULL)
{
error (_("Out of memory reading %u section group headers\n"),
- elf_header.e_shnum);
+ filedata->file_header.e_shnum);
return FALSE;
}
/* Scan the sections for the group section. */
group_count = 0;
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum;
i++, section++)
if (section->sh_type == SHT_GROUP)
group_count++;
num_syms = 0;
strtab = NULL;
strtab_size = 0;
- for (i = 0, section = section_headers, group = section_groups;
- i < elf_header.e_shnum;
+ for (i = 0, section = filedata->section_headers, group = section_groups;
+ i < filedata->file_header.e_shnum;
i++, section++)
{
if (section->sh_type == SHT_GROUP)
{
- const char * name = printable_section_name (section);
+ const char * name = printable_section_name (filedata, section);
const char * group_name;
unsigned char * start;
unsigned char * indices;
Elf_Internal_Sym * sym;
/* Get the symbol table. */
- if (section->sh_link >= elf_header.e_shnum
- || ((sec = section_headers + section->sh_link)->sh_type
+ if (section->sh_link >= filedata->file_header.e_shnum
+ || ((sec = filedata->section_headers + section->sh_link)->sh_type
!= SHT_SYMTAB))
{
error (_("Bad sh_link in group section `%s'\n"), name);
symtab_sec = sec;
if (symtab)
free (symtab);
- symtab = GET_ELF_SYMBOLS (file, symtab_sec, & num_syms);
+ symtab = GET_ELF_SYMBOLS (filedata, symtab_sec, & num_syms);
}
if (symtab == NULL)
if (ELF_ST_TYPE (sym->st_info) == STT_SECTION)
{
if (sym->st_shndx == 0
- || sym->st_shndx >= elf_header.e_shnum)
+ || sym->st_shndx >= filedata->file_header.e_shnum)
{
error (_("Bad sh_info in group section `%s'\n"), name);
continue;
}
- group_name = SECTION_NAME (section_headers + sym->st_shndx);
+ group_name = SECTION_NAME (filedata->section_headers + sym->st_shndx);
strtab_sec = NULL;
if (strtab)
free (strtab);
else
{
/* Get the string table. */
- if (symtab_sec->sh_link >= elf_header.e_shnum)
+ if (symtab_sec->sh_link >= filedata->file_header.e_shnum)
{
strtab_sec = NULL;
if (strtab)
strtab_size = 0;
}
else if (strtab_sec
- != (sec = section_headers + symtab_sec->sh_link))
+ != (sec = filedata->section_headers + symtab_sec->sh_link))
{
strtab_sec = sec;
if (strtab)
free (strtab);
- strtab = (char *) get_data (NULL, file, strtab_sec->sh_offset,
+ strtab = (char *) get_data (NULL, filedata, strtab_sec->sh_offset,
1, strtab_sec->sh_size,
_("string table"));
strtab_size = strtab != NULL ? strtab_sec->sh_size : 0;
if (section->sh_entsize > section->sh_size)
{
error (_("Section %s has sh_entsize (0x%lx) which is larger than its size (0x%lx)\n"),
- printable_section_name (section),
+ printable_section_name (filedata, section),
(unsigned long) section->sh_entsize,
(unsigned long) section->sh_size);
break;
}
- start = (unsigned char *) get_data (NULL, file, section->sh_offset,
+ start = (unsigned char *) get_data (NULL, filedata, section->sh_offset,
1, section->sh_size,
_("section data"));
if (start == NULL)
entry = byte_get (indices, 4);
indices += 4;
- if (entry >= elf_header.e_shnum)
+ if (entry >= filedata->file_header.e_shnum)
{
static unsigned num_group_errors = 0;
if (num_group_errors ++ < 10)
{
error (_("section [%5u] in group section [%5u] > maximum section [%5u]\n"),
- entry, i, elf_header.e_shnum - 1);
+ entry, i, filedata->file_header.e_shnum - 1);
if (num_group_errors == 10)
warn (_("Further error messages about overlarge group section indicies suppressed\n"));
}
if (do_section_groups)
{
- sec = section_headers + entry;
- printf (" [%5u] %s\n", entry, printable_section_name (sec));
+ sec = filedata->section_headers + entry;
+ printf (" [%5u] %s\n", entry, printable_section_name (filedata, sec));
}
g = (struct group_list *) xmalloc (sizeof (struct group_list));
library). */
static bfd_boolean
-dump_ia64_vms_dynamic_fixups (FILE * file,
- struct ia64_vms_dynfixup * fixup,
- const char * strtab,
- unsigned int strtab_sz)
+dump_ia64_vms_dynamic_fixups (Filedata * filedata,
+ struct ia64_vms_dynfixup * fixup,
+ const char * strtab,
+ unsigned int strtab_sz)
{
Elf64_External_VMS_IMAGE_FIXUP * imfs;
long i;
const char * lib_name;
- imfs = get_data (NULL, file, dynamic_addr + fixup->fixup_rela_off,
+ imfs = get_data (NULL, filedata, dynamic_addr + fixup->fixup_rela_off,
1, fixup->fixup_rela_cnt * sizeof (*imfs),
_("dynamic section image fixups"));
if (!imfs)
/* Display IA-64 OpenVMS dynamic relocations (used to relocate an image). */
static bfd_boolean
-dump_ia64_vms_dynamic_relocs (FILE *file, struct ia64_vms_dynimgrela *imgrela)
+dump_ia64_vms_dynamic_relocs (Filedata * filedata, struct ia64_vms_dynimgrela *imgrela)
{
Elf64_External_VMS_IMAGE_RELA *imrs;
long i;
- imrs = get_data (NULL, file, dynamic_addr + imgrela->img_rela_off,
+ imrs = get_data (NULL, filedata, dynamic_addr + imgrela->img_rela_off,
1, imgrela->img_rela_cnt * sizeof (*imrs),
_("dynamic section image relocations"));
if (!imrs)
/* Display IA-64 OpenVMS dynamic relocations and fixups. */
static bfd_boolean
-process_ia64_vms_dynamic_relocs (FILE *file)
+process_ia64_vms_dynamic_relocs (Filedata * filedata)
{
struct ia64_vms_dynfixup fixup;
struct ia64_vms_dynimgrela imgrela;
case DT_STRSZ:
strtab_sz = entry->d_un.d_val;
if (strtab == NULL)
- strtab = get_data (NULL, file, dynamic_addr + strtab_off,
+ strtab = get_data (NULL, filedata, dynamic_addr + strtab_off,
1, strtab_sz, _("dynamic string section"));
break;
break;
case DT_IA_64_VMS_FIXUP_RELA_OFF:
fixup.fixup_rela_off = entry->d_un.d_val;
- if (! dump_ia64_vms_dynamic_fixups (file, &fixup, strtab, strtab_sz))
+ if (! dump_ia64_vms_dynamic_fixups (filedata, &fixup, strtab, strtab_sz))
res = FALSE;
break;
case DT_IA_64_VMS_IMG_RELA_CNT:
break;
case DT_IA_64_VMS_IMG_RELA_OFF:
imgrela.img_rela_off = entry->d_un.d_val;
- if (! dump_ia64_vms_dynamic_relocs (file, &imgrela))
+ if (! dump_ia64_vms_dynamic_relocs (filedata, &imgrela))
res = FALSE;
break;
/* Process the reloc section. */
static bfd_boolean
-process_relocs (FILE * file)
+process_relocs (Filedata * filedata)
{
unsigned long rel_size;
unsigned long rel_offset;
(_("\n'%s' relocation section at offset 0x%lx contains %ld bytes:\n"),
name, rel_offset, rel_size);
- dump_relocations (file,
- offset_from_vma (file, rel_offset, rel_size),
+ dump_relocations (filedata,
+ offset_from_vma (filedata, rel_offset, rel_size),
rel_size,
dynamic_symbols, num_dynamic_syms,
dynamic_strings, dynamic_strings_length,
}
}
- if (is_ia64_vms ())
- if (process_ia64_vms_dynamic_relocs (file))
+ if (is_ia64_vms (filedata))
+ if (process_ia64_vms_dynamic_relocs (filedata))
has_dynamic_reloc = TRUE;
if (! has_dynamic_reloc)
unsigned long i;
bfd_boolean found = FALSE;
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum;
i++, section++)
{
if ( section->sh_type != SHT_RELA
printf (_("\nRelocation section "));
- if (string_table == NULL)
+ if (filedata->string_table == NULL)
printf ("%d", section->sh_name);
else
- printf ("'%s'", printable_section_name (section));
+ printf ("'%s'", printable_section_name (filedata, section));
num_rela = rel_size / section->sh_entsize;
printf (ngettext (" at offset 0x%lx contains %lu entry:\n",
is_rela = section->sh_type == SHT_RELA;
if (section->sh_link != 0
- && section->sh_link < elf_header.e_shnum)
+ && section->sh_link < filedata->file_header.e_shnum)
{
Elf_Internal_Shdr * symsec;
Elf_Internal_Sym * symtab;
unsigned long strtablen = 0;
char * strtab = NULL;
- symsec = section_headers + section->sh_link;
+ symsec = filedata->section_headers + section->sh_link;
if (symsec->sh_type != SHT_SYMTAB
&& symsec->sh_type != SHT_DYNSYM)
continue;
- symtab = GET_ELF_SYMBOLS (file, symsec, & nsyms);
+ symtab = GET_ELF_SYMBOLS (filedata, symsec, & nsyms);
if (symtab == NULL)
continue;
if (symsec->sh_link != 0
- && symsec->sh_link < elf_header.e_shnum)
+ && symsec->sh_link < filedata->file_header.e_shnum)
{
- strsec = section_headers + symsec->sh_link;
+ strsec = filedata->section_headers + symsec->sh_link;
- strtab = (char *) get_data (NULL, file, strsec->sh_offset,
+ strtab = (char *) get_data (NULL, filedata, strsec->sh_offset,
1, strsec->sh_size,
_("string table"));
strtablen = strtab == NULL ? 0 : strsec->sh_size;
}
- dump_relocations (file, rel_offset, rel_size,
+ dump_relocations (filedata, rel_offset, rel_size,
symtab, nsyms, strtab, strtablen,
is_rela,
symsec->sh_type == SHT_DYNSYM);
free (symtab);
}
else
- dump_relocations (file, rel_offset, rel_size,
+ dump_relocations (filedata, rel_offset, rel_size,
NULL, 0, NULL, 0, is_rela,
FALSE /* is_dynamic */);
#define ABSADDR(a) \
((a).section \
- ? section_headers [(a).section].sh_addr + (a).offset \
+ ? filedata->section_headers [(a).section].sh_addr + (a).offset \
: (a).offset)
/* Find the nearest symbol at or below ADDR. Returns the symbol
name, if found, and the offset from the symbol to ADDR. */
static void
-find_symbol_for_address (Elf_Internal_Sym * symtab,
- unsigned long nsyms,
- const char * strtab,
- unsigned long strtab_size,
- struct absaddr addr,
- const char ** symname,
- bfd_vma * offset)
+find_symbol_for_address (Filedata * filedata,
+ Elf_Internal_Sym * symtab,
+ unsigned long nsyms,
+ const char * strtab,
+ unsigned long strtab_size,
+ struct absaddr addr,
+ const char ** symname,
+ bfd_vma * offset)
{
bfd_vma dist = 0x100000;
Elf_Internal_Sym * sym;
};
static bfd_boolean
-dump_ia64_unwind (struct ia64_unw_aux_info * aux)
+dump_ia64_unwind (Filedata * filedata, struct ia64_unw_aux_info * aux)
{
struct ia64_unw_table_entry * tp;
unsigned long j, nfuns;
const unsigned char * end;
const char * procname;
- find_symbol_for_address (aux->funtab, aux->nfuns, aux->strtab,
+ find_symbol_for_address (filedata, aux->funtab, aux->nfuns, aux->strtab,
aux->strtab_size, tp->start, &procname, &offset);
fputs ("\n<", stdout);
}
static bfd_boolean
-slurp_ia64_unwind_table (FILE * file,
- struct ia64_unw_aux_info * aux,
- Elf_Internal_Shdr * sec)
+slurp_ia64_unwind_table (Filedata * filedata,
+ struct ia64_unw_aux_info * aux,
+ Elf_Internal_Shdr * sec)
{
unsigned long size, nrelas, i;
Elf_Internal_Phdr * seg;
/* First, find the starting address of the segment that includes
this section: */
- if (elf_header.e_phnum)
+ if (filedata->file_header.e_phnum)
{
- if (! get_program_headers (file))
+ if (! get_program_headers (filedata))
return FALSE;
- for (seg = program_headers;
- seg < program_headers + elf_header.e_phnum;
+ for (seg = filedata->program_headers;
+ seg < filedata->program_headers + filedata->file_header.e_phnum;
++seg)
{
if (seg->p_type != PT_LOAD)
/* Second, build the unwind table from the contents of the unwind section: */
size = sec->sh_size;
- table = (unsigned char *) get_data (NULL, file, sec->sh_offset, 1, size,
+ table = (unsigned char *) get_data (NULL, filedata, sec->sh_offset, 1, size,
_("unwind table"));
if (!table)
return FALSE;
free (table);
/* Third, apply any relocations to the unwind table: */
- for (relsec = section_headers;
- relsec < section_headers + elf_header.e_shnum;
+ for (relsec = filedata->section_headers;
+ relsec < filedata->section_headers + filedata->file_header.e_shnum;
++relsec)
{
if (relsec->sh_type != SHT_RELA
- || relsec->sh_info >= elf_header.e_shnum
- || section_headers + relsec->sh_info != sec)
+ || relsec->sh_info >= filedata->file_header.e_shnum
+ || filedata->section_headers + relsec->sh_info != sec)
continue;
- if (!slurp_rela_relocs (file, relsec->sh_offset, relsec->sh_size,
+ if (!slurp_rela_relocs (filedata, relsec->sh_offset, relsec->sh_size,
& rela, & nrelas))
{
free (aux->table);
for (rp = rela; rp < rela + nrelas; ++rp)
{
- relname = elf_ia64_reloc_type (get_reloc_type (rp->r_info));
+ relname = elf_ia64_reloc_type (get_reloc_type (filedata, rp->r_info));
sym = aux->symtab + get_reloc_symindex (rp->r_info);
/* PR 17531: file: 9fa67536. */
if (relname == NULL)
{
- warn (_("Skipping unknown relocation type: %u\n"), get_reloc_type (rp->r_info));
+ warn (_("Skipping unknown relocation type: %u\n"),
+ get_reloc_type (filedata, rp->r_info));
continue;
}
}
static bfd_boolean
-ia64_process_unwind (FILE * file)
+ia64_process_unwind (Filedata * filedata)
{
Elf_Internal_Shdr * sec;
Elf_Internal_Shdr * unwsec = NULL;
memset (& aux, 0, sizeof (aux));
- for (i = 0, sec = section_headers; i < elf_header.e_shnum; ++i, ++sec)
+ for (i = 0, sec = filedata->section_headers; i < filedata->file_header.e_shnum; ++i, ++sec)
{
if (sec->sh_type == SHT_SYMTAB
- && sec->sh_link < elf_header.e_shnum)
+ && sec->sh_link < filedata->file_header.e_shnum)
{
- aux.symtab = GET_ELF_SYMBOLS (file, sec, & aux.nsyms);
+ aux.symtab = GET_ELF_SYMBOLS (filedata, sec, & aux.nsyms);
- strsec = section_headers + sec->sh_link;
+ strsec = filedata->section_headers + sec->sh_link;
if (aux.strtab != NULL)
{
error (_("Multiple auxillary string tables encountered\n"));
free (aux.strtab);
res = FALSE;
}
- aux.strtab = (char *) get_data (NULL, file, strsec->sh_offset,
+ aux.strtab = (char *) get_data (NULL, filedata, strsec->sh_offset,
1, strsec->sh_size,
_("string table"));
aux.strtab_size = aux.strtab != NULL ? strsec->sh_size : 0;
char * suffix;
size_t len, len2;
- for (i = unwstart, sec = section_headers + unwstart, unwsec = NULL;
- i < elf_header.e_shnum; ++i, ++sec)
+ for (i = unwstart, sec = filedata->section_headers + unwstart, unwsec = NULL;
+ i < filedata->file_header.e_shnum; ++i, ++sec)
if (sec->sh_type == SHT_IA_64_UNWIND)
{
unwsec = sec;
if (section_headers_groups == NULL
|| section_headers_groups [i] == NULL)
- i = elf_header.e_shnum;
+ i = filedata->file_header.e_shnum;
else
{
g = section_headers_groups [i]->root;
for (; g != NULL; g = g->next)
{
- sec = section_headers + g->section_index;
+ sec = filedata->section_headers + g->section_index;
if (streq (SECTION_NAME (sec), ELF_STRING_ia64_unwind_info))
break;
}
if (g == NULL)
- i = elf_header.e_shnum;
+ i = filedata->file_header.e_shnum;
}
}
else if (strneq (SECTION_NAME (unwsec), ELF_STRING_ia64_unwind_once, len))
/* .gnu.linkonce.ia64unw.FOO -> .gnu.linkonce.ia64unwi.FOO. */
len2 = sizeof (ELF_STRING_ia64_unwind_info_once) - 1;
suffix = SECTION_NAME (unwsec) + len;
- for (i = 0, sec = section_headers; i < elf_header.e_shnum;
+ for (i = 0, sec = filedata->section_headers; i < filedata->file_header.e_shnum;
++i, ++sec)
if (strneq (SECTION_NAME (sec), ELF_STRING_ia64_unwind_info_once, len2)
&& streq (SECTION_NAME (sec) + len2, suffix))
suffix = "";
if (strneq (SECTION_NAME (unwsec), ELF_STRING_ia64_unwind, len))
suffix = SECTION_NAME (unwsec) + len;
- for (i = 0, sec = section_headers; i < elf_header.e_shnum;
+ for (i = 0, sec = filedata->section_headers; i < filedata->file_header.e_shnum;
++i, ++sec)
if (strneq (SECTION_NAME (sec), ELF_STRING_ia64_unwind_info, len2)
&& streq (SECTION_NAME (sec) + len2, suffix))
break;
}
- if (i == elf_header.e_shnum)
+ if (i == filedata->file_header.e_shnum)
{
printf (_("\nCould not find unwind info section for "));
- if (string_table == NULL)
+ if (filedata->string_table == NULL)
printf ("%d", unwsec->sh_name);
else
- printf ("'%s'", printable_section_name (unwsec));
+ printf ("'%s'", printable_section_name (filedata, unwsec));
}
else
{
aux.info_addr = sec->sh_addr;
- aux.info = (unsigned char *) get_data (NULL, file, sec->sh_offset, 1,
+ aux.info = (unsigned char *) get_data (NULL, filedata, sec->sh_offset, 1,
sec->sh_size,
_("unwind info"));
aux.info_size = aux.info == NULL ? 0 : sec->sh_size;
printf (_("\nUnwind section "));
- if (string_table == NULL)
+ if (filedata->string_table == NULL)
printf ("%d", unwsec->sh_name);
else
- printf ("'%s'", printable_section_name (unwsec));
+ printf ("'%s'", printable_section_name (filedata, unwsec));
printf (_(" at offset 0x%lx contains %lu entries:\n"),
(unsigned long) unwsec->sh_offset,
(unsigned long) (unwsec->sh_size / (3 * eh_addr_size)));
- if (slurp_ia64_unwind_table (file, & aux, unwsec)
+ if (slurp_ia64_unwind_table (filedata, & aux, unwsec)
&& aux.table_len > 0)
- dump_ia64_unwind (& aux);
+ dump_ia64_unwind (filedata, & aux);
if (aux.table)
free ((char *) aux.table);
};
static bfd_boolean
-dump_hppa_unwind (struct hppa_unw_aux_info * aux)
+dump_hppa_unwind (Filedata * filedata, struct hppa_unw_aux_info * aux)
{
struct hppa_unw_table_entry * tp;
unsigned long j, nfuns;
bfd_vma offset;
const char * procname;
- find_symbol_for_address (aux->funtab, aux->nfuns, aux->strtab,
+ find_symbol_for_address (filedata, aux->funtab, aux->nfuns, aux->strtab,
aux->strtab_size, tp->start, &procname,
&offset);
}
static bfd_boolean
-slurp_hppa_unwind_table (FILE * file,
- struct hppa_unw_aux_info * aux,
- Elf_Internal_Shdr * sec)
+slurp_hppa_unwind_table (Filedata * filedata,
+ struct hppa_unw_aux_info * aux,
+ Elf_Internal_Shdr * sec)
{
unsigned long size, unw_ent_size, nentries, nrelas, i;
Elf_Internal_Phdr * seg;
/* First, find the starting address of the segment that includes
this section. */
- if (elf_header.e_phnum)
+ if (filedata->file_header.e_phnum)
{
- if (! get_program_headers (file))
+ if (! get_program_headers (filedata))
return FALSE;
- for (seg = program_headers;
- seg < program_headers + elf_header.e_phnum;
+ for (seg = filedata->program_headers;
+ seg < filedata->program_headers + filedata->file_header.e_phnum;
++seg)
{
if (seg->p_type != PT_LOAD)
/* Second, build the unwind table from the contents of the unwind
section. */
size = sec->sh_size;
- table = (unsigned char *) get_data (NULL, file, sec->sh_offset, 1, size,
+ table = (unsigned char *) get_data (NULL, filedata, sec->sh_offset, 1, size,
_("unwind table"));
if (!table)
return FALSE;
free (table);
/* Third, apply any relocations to the unwind table. */
- for (relsec = section_headers;
- relsec < section_headers + elf_header.e_shnum;
+ for (relsec = filedata->section_headers;
+ relsec < filedata->section_headers + filedata->file_header.e_shnum;
++relsec)
{
if (relsec->sh_type != SHT_RELA
- || relsec->sh_info >= elf_header.e_shnum
- || section_headers + relsec->sh_info != sec)
+ || relsec->sh_info >= filedata->file_header.e_shnum
+ || filedata->section_headers + relsec->sh_info != sec)
continue;
- if (!slurp_rela_relocs (file, relsec->sh_offset, relsec->sh_size,
+ if (!slurp_rela_relocs (filedata, relsec->sh_offset, relsec->sh_size,
& rela, & nrelas))
return FALSE;
for (rp = rela; rp < rela + nrelas; ++rp)
{
- relname = elf_hppa_reloc_type (get_reloc_type (rp->r_info));
+ relname = elf_hppa_reloc_type (get_reloc_type (filedata, rp->r_info));
sym = aux->symtab + get_reloc_symindex (rp->r_info);
/* R_PARISC_SEGREL32 or R_PARISC_SEGREL64. */
}
static bfd_boolean
-hppa_process_unwind (FILE * file)
+hppa_process_unwind (Filedata * filedata)
{
struct hppa_unw_aux_info aux;
Elf_Internal_Shdr * unwsec = NULL;
unsigned long i;
bfd_boolean res = TRUE;
- if (string_table == NULL)
+ if (filedata->string_table == NULL)
return FALSE;
memset (& aux, 0, sizeof (aux));
- for (i = 0, sec = section_headers; i < elf_header.e_shnum; ++i, ++sec)
+ for (i = 0, sec = filedata->section_headers; i < filedata->file_header.e_shnum; ++i, ++sec)
{
if (sec->sh_type == SHT_SYMTAB
- && sec->sh_link < elf_header.e_shnum)
+ && sec->sh_link < filedata->file_header.e_shnum)
{
- aux.symtab = GET_ELF_SYMBOLS (file, sec, & aux.nsyms);
+ aux.symtab = GET_ELF_SYMBOLS (filedata, sec, & aux.nsyms);
- strsec = section_headers + sec->sh_link;
+ strsec = filedata->section_headers + sec->sh_link;
if (aux.strtab != NULL)
{
error (_("Multiple auxillary string tables encountered\n"));
free (aux.strtab);
res = FALSE;
}
- aux.strtab = (char *) get_data (NULL, file, strsec->sh_offset,
+ aux.strtab = (char *) get_data (NULL, filedata, strsec->sh_offset,
1, strsec->sh_size,
_("string table"));
aux.strtab_size = aux.strtab != NULL ? strsec->sh_size : 0;
if (!unwsec)
printf (_("\nThere are no unwind sections in this file.\n"));
- for (i = 0, sec = section_headers; i < elf_header.e_shnum; ++i, ++sec)
+ for (i = 0, sec = filedata->section_headers; i < filedata->file_header.e_shnum; ++i, ++sec)
{
if (streq (SECTION_NAME (sec), ".PARISC.unwind"))
{
unsigned long num_unwind = sec->sh_size / (2 * eh_addr_size + 8);
+
printf (ngettext ("\nUnwind section '%s' at offset 0x%lx "
"contains %lu entry:\n",
"\nUnwind section '%s' at offset 0x%lx "
"contains %lu entries:\n",
num_unwind),
- printable_section_name (sec),
+ printable_section_name (filedata, sec),
(unsigned long) sec->sh_offset,
num_unwind);
- if (! slurp_hppa_unwind_table (file, &aux, sec))
+ if (! slurp_hppa_unwind_table (filedata, &aux, sec))
res = FALSE;
if (aux.table_len > 0)
{
- if (! dump_hppa_unwind (&aux))
+ if (! dump_hppa_unwind (filedata, &aux))
res = FALSE;
}
struct arm_unw_aux_info
{
- FILE * file; /* The file containing the unwind sections. */
+ Filedata * filedata; /* The file containing the unwind sections. */
Elf_Internal_Sym * symtab; /* The file's symbol table. */
unsigned long nsyms; /* Number of symbols. */
Elf_Internal_Sym * funtab; /* Sorted table of STT_FUNC symbols. */
};
static const char *
-arm_print_vma_and_name (struct arm_unw_aux_info *aux,
- bfd_vma fn, struct absaddr addr)
+arm_print_vma_and_name (Filedata * filedata,
+ struct arm_unw_aux_info * aux,
+ bfd_vma fn,
+ struct absaddr addr)
{
const char *procname;
bfd_vma sym_offset;
if (addr.section == SHN_UNDEF)
addr.offset = fn;
- find_symbol_for_address (aux->funtab, aux->nfuns, aux->strtab,
+ find_symbol_for_address (filedata, aux->funtab, aux->nfuns, aux->strtab,
aux->strtab_size, addr, &procname,
&sym_offset);
5) Return TRUE upon success, FALSE otherwise. */
static bfd_boolean
-get_unwind_section_word (struct arm_unw_aux_info * aux,
+get_unwind_section_word (Filedata * filedata,
+ struct arm_unw_aux_info * aux,
struct arm_section * arm_sec,
Elf_Internal_Shdr * sec,
bfd_vma word_offset,
arm_free_section (arm_sec);
arm_sec->sec = sec;
- arm_sec->data = get_data (NULL, aux->file, sec->sh_offset, 1,
+ arm_sec->data = get_data (NULL, aux->filedata, sec->sh_offset, 1,
sec->sh_size, _("unwind data"));
arm_sec->rela = NULL;
arm_sec->nrelas = 0;
- for (relsec = section_headers;
- relsec < section_headers + elf_header.e_shnum;
+ for (relsec = filedata->section_headers;
+ relsec < filedata->section_headers + filedata->file_header.e_shnum;
++relsec)
{
- if (relsec->sh_info >= elf_header.e_shnum
- || section_headers + relsec->sh_info != sec
+ if (relsec->sh_info >= filedata->file_header.e_shnum
+ || filedata->section_headers + relsec->sh_info != sec
/* PR 15745: Check the section type as well. */
|| (relsec->sh_type != SHT_REL
&& relsec->sh_type != SHT_RELA))
arm_sec->rel_type = relsec->sh_type;
if (relsec->sh_type == SHT_REL)
{
- if (!slurp_rel_relocs (aux->file, relsec->sh_offset,
+ if (!slurp_rel_relocs (aux->filedata, relsec->sh_offset,
relsec->sh_size,
& arm_sec->rela, & arm_sec->nrelas))
return FALSE;
}
else /* relsec->sh_type == SHT_RELA */
{
- if (!slurp_rela_relocs (aux->file, relsec->sh_offset,
+ if (!slurp_rela_relocs (aux->filedata, relsec->sh_offset,
relsec->sh_size,
& arm_sec->rela, & arm_sec->nrelas))
return FALSE;
prelval = offset - (arm_sec->sec->sh_addr + rp->r_offset);
/* Check that we are processing the expected reloc type. */
- if (elf_header.e_machine == EM_ARM)
+ if (filedata->file_header.e_machine == EM_ARM)
{
relname = elf_arm_reloc_type (ELF32_R_TYPE (rp->r_info));
if (relname == NULL)
continue;
}
}
- else if (elf_header.e_machine == EM_TI_C6000)
+ else if (filedata->file_header.e_machine == EM_TI_C6000)
{
relname = elf_tic6x_reloc_type (ELF32_R_TYPE (rp->r_info));
if (relname == NULL)
if (remaining == 0 && more_words) \
{ \
data_offset += 4; \
- if (! get_unwind_section_word (aux, data_arm_sec, data_sec, \
+ if (! get_unwind_section_word (filedata, aux, data_arm_sec, data_sec, \
data_offset, & word, & addr, NULL)) \
return FALSE; \
remaining = 4; \
printf ("0x%02x ", OP)
static bfd_boolean
-decode_arm_unwind_bytecode (struct arm_unw_aux_info * aux,
+decode_arm_unwind_bytecode (Filedata * filedata,
+ struct arm_unw_aux_info * aux,
unsigned int word,
unsigned int remaining,
unsigned int more_words,
}
static bfd_boolean
-decode_tic6x_unwind_bytecode (struct arm_unw_aux_info * aux,
+decode_tic6x_unwind_bytecode (Filedata * filedata,
+ struct arm_unw_aux_info * aux,
unsigned int word,
unsigned int remaining,
unsigned int more_words,
}
static bfd_vma
-arm_expand_prel31 (bfd_vma word, bfd_vma where)
+arm_expand_prel31 (Filedata * filedata, bfd_vma word, bfd_vma where)
{
bfd_vma offset;
if (offset & 0x40000000)
offset |= ~ (bfd_vma) 0x7fffffff;
- if (elf_header.e_machine == EM_TI_C6000)
+ if (filedata->file_header.e_machine == EM_TI_C6000)
offset <<= 1;
return offset + where;
}
static bfd_boolean
-decode_arm_unwind (struct arm_unw_aux_info * aux,
+decode_arm_unwind (Filedata * filedata,
+ struct arm_unw_aux_info * aux,
unsigned int word,
unsigned int remaining,
bfd_vma data_offset,
here will always be 0. So we also pass in the sym_name
parameter so that we can find the symbol associated with
the personality routine. */
- if (! get_unwind_section_word (aux, data_arm_sec, data_sec, data_offset,
+ if (! get_unwind_section_word (filedata, aux, data_arm_sec, data_sec, data_offset,
& word, & addr, & sym_name))
return FALSE;
bfd_vma fn;
const char *procname;
- fn = arm_expand_prel31 (word, data_sec->sh_addr + data_offset);
+ fn = arm_expand_prel31 (filedata, word, data_sec->sh_addr + data_offset);
printf (_(" Personality routine: "));
if (fn == 0
&& addr.section == SHN_UNDEF && addr.offset == 0
}
}
else
- procname = arm_print_vma_and_name (aux, fn, addr);
+ procname = arm_print_vma_and_name (filedata, aux, fn, addr);
fputc ('\n', stdout);
/* The GCC personality routines use the standard compact
-- ----- ----- ----
1 0 index Data for personalityRoutine[index] */
- if (elf_header.e_machine == EM_ARM
+ if (filedata->file_header.e_machine == EM_ARM
&& (word & 0x70000000))
{
warn (_("Corrupt ARM compact model table entry: %x \n"), word);
}
}
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_ARM:
if (per_index < 3)
{
- if (! decode_arm_unwind_bytecode (aux, word, remaining, more_words,
+ if (! decode_arm_unwind_bytecode (filedata, aux, word, remaining, more_words,
data_offset, data_sec, data_arm_sec))
res = FALSE;
}
case EM_TI_C6000:
if (per_index < 3)
{
- if (! decode_tic6x_unwind_bytecode (aux, word, remaining, more_words,
+ if (! decode_tic6x_unwind_bytecode (filedata, aux, word, remaining, more_words,
data_offset, data_sec, data_arm_sec))
res = FALSE;
}
default:
error (_("Unsupported architecture type %d encountered when decoding unwind table\n"),
- elf_header.e_machine);
+ filedata->file_header.e_machine);
res = FALSE;
}
}
static bfd_boolean
-dump_arm_unwind (struct arm_unw_aux_info *aux, Elf_Internal_Shdr *exidx_sec)
+dump_arm_unwind (Filedata * filedata,
+ struct arm_unw_aux_info * aux,
+ Elf_Internal_Shdr * exidx_sec)
{
struct arm_section exidx_arm_sec, extab_arm_sec;
unsigned int i, exidx_len;
fputc ('\n', stdout);
- if (! get_unwind_section_word (aux, & exidx_arm_sec, exidx_sec,
+ if (! get_unwind_section_word (filedata, aux, & exidx_arm_sec, exidx_sec,
8 * i, & exidx_fn, & fn_addr, NULL)
- || ! get_unwind_section_word (aux, & exidx_arm_sec, exidx_sec,
+ || ! get_unwind_section_word (filedata, aux, & exidx_arm_sec, exidx_sec,
8 * i + 4, & exidx_entry, & entry_addr, NULL))
{
free (aux->funtab);
res = FALSE;
}
- fn = arm_expand_prel31 (exidx_fn, exidx_sec->sh_addr + 8 * i);
+ fn = arm_expand_prel31 (filedata, exidx_fn, exidx_sec->sh_addr + 8 * i);
- arm_print_vma_and_name (aux, fn, fn_addr);
+ arm_print_vma_and_name (filedata, aux, fn, fn_addr);
fputs (": ", stdout);
if (exidx_entry == 1)
{
print_vma (exidx_entry, PREFIX_HEX);
fputc ('\n', stdout);
- decode_arm_unwind (aux, exidx_entry, 4, 0, NULL, NULL);
+ decode_arm_unwind (filedata, aux, exidx_entry, 4, 0, NULL, NULL);
}
else
{
Elf_Internal_Shdr *table_sec;
fputs ("@", stdout);
- table = arm_expand_prel31 (exidx_entry, exidx_sec->sh_addr + 8 * i + 4);
+ table = arm_expand_prel31 (filedata, exidx_entry, exidx_sec->sh_addr + 8 * i + 4);
print_vma (table, PREFIX_HEX);
printf ("\n");
/* Locate the matching .ARM.extab. */
if (entry_addr.section != SHN_UNDEF
- && entry_addr.section < elf_header.e_shnum)
+ && entry_addr.section < filedata->file_header.e_shnum)
{
- table_sec = section_headers + entry_addr.section;
+ table_sec = filedata->section_headers + entry_addr.section;
table_offset = entry_addr.offset;
/* PR 18879 */
if (table_offset > table_sec->sh_size
{
warn (_("Unwind entry contains corrupt offset (0x%lx) into section %s\n"),
(unsigned long) table_offset,
- printable_section_name (table_sec));
+ printable_section_name (filedata, table_sec));
res = FALSE;
continue;
}
}
else
{
- table_sec = find_section_by_address (table);
+ table_sec = find_section_by_address (filedata, table);
if (table_sec != NULL)
table_offset = table - table_sec->sh_addr;
}
continue;
}
- if (! decode_arm_unwind (aux, 0, 0, table_offset, table_sec,
+ if (! decode_arm_unwind (filedata, aux, 0, 0, table_offset, table_sec,
&extab_arm_sec))
res = FALSE;
}
/* Used for both ARM and C6X unwinding tables. */
static bfd_boolean
-arm_process_unwind (FILE *file)
+arm_process_unwind (Filedata * filedata)
{
struct arm_unw_aux_info aux;
Elf_Internal_Shdr *unwsec = NULL;
unsigned int sec_type;
bfd_boolean res = TRUE;
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_ARM:
sec_type = SHT_ARM_EXIDX;
default:
error (_("Unsupported architecture type %d encountered when processing unwind table\n"),
- elf_header.e_machine);
+ filedata->file_header.e_machine);
return FALSE;
}
- if (string_table == NULL)
+ if (filedata->string_table == NULL)
return FALSE;
memset (& aux, 0, sizeof (aux));
- aux.file = file;
+ aux.filedata = filedata;
- for (i = 0, sec = section_headers; i < elf_header.e_shnum; ++i, ++sec)
+ for (i = 0, sec = filedata->section_headers; i < filedata->file_header.e_shnum; ++i, ++sec)
{
- if (sec->sh_type == SHT_SYMTAB && sec->sh_link < elf_header.e_shnum)
+ if (sec->sh_type == SHT_SYMTAB && sec->sh_link < filedata->file_header.e_shnum)
{
- aux.symtab = GET_ELF_SYMBOLS (file, sec, & aux.nsyms);
+ aux.symtab = GET_ELF_SYMBOLS (filedata, sec, & aux.nsyms);
- strsec = section_headers + sec->sh_link;
+ strsec = filedata->section_headers + sec->sh_link;
/* PR binutils/17531 file: 011-12666-0.004. */
if (aux.strtab != NULL)
free (aux.strtab);
res = FALSE;
}
- aux.strtab = get_data (NULL, file, strsec->sh_offset,
+ aux.strtab = get_data (NULL, filedata, strsec->sh_offset,
1, strsec->sh_size, _("string table"));
aux.strtab_size = aux.strtab != NULL ? strsec->sh_size : 0;
}
if (unwsec == NULL)
printf (_("\nThere are no unwind sections in this file.\n"));
else
- for (i = 0, sec = section_headers; i < elf_header.e_shnum; ++i, ++sec)
+ for (i = 0, sec = filedata->section_headers; i < filedata->file_header.e_shnum; ++i, ++sec)
{
if (sec->sh_type == sec_type)
{
"\nUnwind section '%s' at offset 0x%lx "
"contains %lu entries:\n",
num_unwind),
- printable_section_name (sec),
+ printable_section_name (filedata, sec),
(unsigned long) sec->sh_offset,
num_unwind);
- if (! dump_arm_unwind (&aux, sec))
+ if (! dump_arm_unwind (filedata, &aux, sec))
res = FALSE;
}
}
}
static bfd_boolean
-process_unwind (FILE * file)
+process_unwind (Filedata * filedata)
{
struct unwind_handler
{
unsigned int machtype;
- bfd_boolean (* handler)(FILE *);
+ bfd_boolean (* handler)(Filedata *);
} handlers[] =
{
{ EM_ARM, arm_process_unwind },
return TRUE;
for (i = 0; handlers[i].handler != NULL; i++)
- if (elf_header.e_machine == handlers[i].machtype)
- return handlers[i].handler (file);
+ if (filedata->file_header.e_machine == handlers[i].machtype)
+ return handlers[i].handler (filedata);
printf (_("\nThe decoding of unwind sections for machine type %s is not currently supported.\n"),
- get_machine_name (elf_header.e_machine));
+ get_machine_name (filedata->file_header.e_machine));
return TRUE;
}
}
static bfd_boolean
-get_32bit_dynamic_section (FILE * file)
+get_32bit_dynamic_section (Filedata * filedata)
{
Elf32_External_Dyn * edyn;
Elf32_External_Dyn * ext;
Elf_Internal_Dyn * entry;
- edyn = (Elf32_External_Dyn *) get_data (NULL, file, dynamic_addr, 1,
+ edyn = (Elf32_External_Dyn *) get_data (NULL, filedata, dynamic_addr, 1,
dynamic_size, _("dynamic section"));
if (!edyn)
return FALSE;
}
static bfd_boolean
-get_64bit_dynamic_section (FILE * file)
+get_64bit_dynamic_section (Filedata * filedata)
{
Elf64_External_Dyn * edyn;
Elf64_External_Dyn * ext;
Elf_Internal_Dyn * entry;
/* Read in the data. */
- edyn = (Elf64_External_Dyn *) get_data (NULL, file, dynamic_addr, 1,
+ edyn = (Elf64_External_Dyn *) get_data (NULL, filedata, dynamic_addr, 1,
dynamic_size, _("dynamic section"));
if (!edyn)
return FALSE;
/* Parse and display the contents of the dynamic section. */
static bfd_boolean
-process_dynamic_section (FILE * file)
+process_dynamic_section (Filedata * filedata)
{
Elf_Internal_Dyn * entry;
if (is_32bit_elf)
{
- if (! get_32bit_dynamic_section (file))
+ if (! get_32bit_dynamic_section (filedata))
return FALSE;
}
else
{
- if (! get_64bit_dynamic_section (file))
+ if (! get_64bit_dynamic_section (filedata))
return FALSE;
}
we default to reading in the entire file (!) and
processing that. This is overkill, I know, but it
should work. */
- section.sh_offset = offset_from_vma (file, entry->d_un.d_val, 0);
- if ((bfd_size_type) section.sh_offset > current_file_size)
+ section.sh_offset = offset_from_vma (filedata, entry->d_un.d_val, 0);
+ if ((bfd_size_type) section.sh_offset > filedata->file_size)
{
/* See PR 21379 for a reproducer. */
error (_("Invalid DT_SYMTAB entry: %lx"), (long) section.sh_offset);
if (archive_file_offset != 0)
section.sh_size = archive_file_size - section.sh_offset;
else
- {
- if (fseek (file, 0, SEEK_END))
- error (_("Unable to seek to end of file!\n"));
-
- section.sh_size = ftell (file) - section.sh_offset;
- }
+ section.sh_size = filedata->file_size - section.sh_offset;
if (is_32bit_elf)
section.sh_entsize = sizeof (Elf32_External_Sym);
else
section.sh_entsize = sizeof (Elf64_External_Sym);
- section.sh_name = string_table_length;
+ section.sh_name = filedata->string_table_length;
- dynamic_symbols = GET_ELF_SYMBOLS (file, §ion, & num_dynamic_syms);
+ dynamic_symbols = GET_ELF_SYMBOLS (filedata, §ion, & num_dynamic_syms);
if (num_dynamic_syms < 1)
{
error (_("Unable to determine the number of symbols to load\n"));
processing that. This is overkill, I know, but it
should work. */
- offset = offset_from_vma (file, entry->d_un.d_val, 0);
+ offset = offset_from_vma (filedata, entry->d_un.d_val, 0);
if (archive_file_offset != 0)
str_tab_len = archive_file_size - offset;
else
- {
- if (fseek (file, 0, SEEK_END))
- error (_("Unable to seek to end of file\n"));
- str_tab_len = ftell (file) - offset;
- }
+ str_tab_len = filedata->file_size;
if (str_tab_len < 1)
{
continue;
}
- dynamic_strings = (char *) get_data (NULL, file, offset, 1,
+ dynamic_strings = (char *) get_data (NULL, filedata, offset, 1,
str_tab_len,
_("dynamic string table"));
dynamic_strings_length = dynamic_strings == NULL ? 0 : str_tab_len;
else if (entry->d_tag == DT_SYMINSZ)
syminsz = entry->d_un.d_val;
else if (entry->d_tag == DT_SYMINFO)
- dynamic_syminfo_offset = offset_from_vma (file, entry->d_un.d_val,
+ dynamic_syminfo_offset = offset_from_vma (filedata, entry->d_un.d_val,
syminsz);
}
/* There is a syminfo section. Read the data. */
extsyminfo = (Elf_External_Syminfo *)
- get_data (NULL, file, dynamic_syminfo_offset, 1, syminsz,
+ get_data (NULL, filedata, dynamic_syminfo_offset, 1, syminsz,
_("symbol information"));
if (!extsyminfo)
return FALSE;
putchar (' ');
print_vma (entry->d_tag, FULL_HEX);
- dtype = get_dynamic_type (entry->d_tag);
+ dtype = get_dynamic_type (filedata, entry->d_tag);
printf (" (%s)%*s", dtype,
((is_32bit_elf ? 27 : 19) - (int) strlen (dtype)), " ");
}
case DT_PLTREL:
dynamic_info[entry->d_tag] = entry->d_un.d_val;
if (do_dynamic)
- puts (get_dynamic_type (entry->d_un.d_val));
+ puts (get_dynamic_type (filedata, entry->d_un.d_val));
break;
case DT_NULL :
if (do_dynamic)
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_MIPS:
case EM_MIPS_RS3_LE:
/* Display the contents of the version sections. */
static bfd_boolean
-process_version_sections (FILE * file)
+process_version_sections (Filedata * filedata)
{
Elf_Internal_Shdr * section;
unsigned i;
if (! do_version)
return TRUE;
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum;
i++, section++)
{
switch (section->sh_type)
"\nVersion definition section '%s' "
"contains %u entries:\n",
section->sh_info),
- printable_section_name (section),
+ printable_section_name (filedata, section),
section->sh_info);
printf (_(" Addr: 0x"));
printf_vma (section->sh_addr);
printf (_(" Offset: %#08lx Link: %u (%s)\n"),
(unsigned long) section->sh_offset, section->sh_link,
- printable_section_name_from_index (section->sh_link));
+ printable_section_name_from_index (filedata, section->sh_link));
edefs = (Elf_External_Verdef *)
- get_data (NULL, file, section->sh_offset, 1,section->sh_size,
+ get_data (NULL, filedata, section->sh_offset, 1,section->sh_size,
_("version definition section"));
if (!edefs)
break;
"\nVersion needs section '%s' "
"contains %u entries:\n",
section->sh_info),
- printable_section_name (section), section->sh_info);
+ printable_section_name (filedata, section), section->sh_info);
printf (_(" Addr: 0x"));
printf_vma (section->sh_addr);
printf (_(" Offset: %#08lx Link: %u (%s)\n"),
(unsigned long) section->sh_offset, section->sh_link,
- printable_section_name_from_index (section->sh_link));
+ printable_section_name_from_index (filedata, section->sh_link));
- eneed = (Elf_External_Verneed *) get_data (NULL, file,
+ eneed = (Elf_External_Verneed *) get_data (NULL, filedata,
section->sh_offset, 1,
section->sh_size,
_("Version Needs section"));
unsigned long num_syms;
long off;
- if (section->sh_link >= elf_header.e_shnum)
+ if (section->sh_link >= filedata->file_header.e_shnum)
break;
- link_section = section_headers + section->sh_link;
+ link_section = filedata->section_headers + section->sh_link;
total = section->sh_size / sizeof (Elf_External_Versym);
- if (link_section->sh_link >= elf_header.e_shnum)
+ if (link_section->sh_link >= filedata->file_header.e_shnum)
break;
found = TRUE;
- symbols = GET_ELF_SYMBOLS (file, link_section, & num_syms);
+ symbols = GET_ELF_SYMBOLS (filedata, link_section, & num_syms);
if (symbols == NULL)
break;
- string_sec = section_headers + link_section->sh_link;
+ string_sec = filedata->section_headers + link_section->sh_link;
- strtab = (char *) get_data (NULL, file, string_sec->sh_offset, 1,
+ strtab = (char *) get_data (NULL, filedata, string_sec->sh_offset, 1,
string_sec->sh_size,
_("version string table"));
if (!strtab)
"\nVersion symbols section '%s' "
"contains %lu entries:\n",
total),
- printable_section_name (section), (unsigned long) total);
+ printable_section_name (filedata, section), (unsigned long) total);
printf (_(" Addr: "));
printf_vma (section->sh_addr);
printf (_(" Offset: %#08lx Link: %u (%s)\n"),
(unsigned long) section->sh_offset, section->sh_link,
- printable_section_name (link_section));
+ printable_section_name (filedata, link_section));
- off = offset_from_vma (file,
+ off = offset_from_vma (filedata,
version_info[DT_VERSIONTAGIDX (DT_VERSYM)],
total * sizeof (short));
- edata = (unsigned char *) get_data (NULL, file, off, total,
+ edata = (unsigned char *) get_data (NULL, filedata, off, total,
sizeof (short),
_("version symbol data"));
if (!edata)
unsigned long offset;
offset = offset_from_vma
- (file, version_info[DT_VERSIONTAGIDX (DT_VERNEED)],
+ (filedata, version_info[DT_VERSIONTAGIDX (DT_VERNEED)],
sizeof (Elf_External_Verneed));
do
Elf_External_Vernaux evna;
unsigned long a_off;
- if (get_data (&evn, file, offset, sizeof (evn), 1,
+ if (get_data (&evn, filedata, offset, sizeof (evn), 1,
_("version need")) == NULL)
break;
do
{
- if (get_data (&evna, file, a_off, sizeof (evna),
+ if (get_data (&evna, filedata, a_off, sizeof (evna),
1, _("version need aux (2)")) == NULL)
{
ivna.vna_next = 0;
unsigned long offset;
offset = offset_from_vma
- (file, version_info[DT_VERSIONTAGIDX (DT_VERDEF)],
+ (filedata, version_info[DT_VERSIONTAGIDX (DT_VERDEF)],
sizeof evd);
do
{
- if (get_data (&evd, file, offset, sizeof (evd), 1,
+ if (get_data (&evd, filedata, offset, sizeof (evd), 1,
_("version def")) == NULL)
{
ivd.vd_next = 0;
ivd.vd_aux = BYTE_GET (evd.vd_aux);
- if (get_data (&evda, file,
+ if (get_data (&evda, filedata,
offset - ivd.vd_next + ivd.vd_aux,
sizeof (evda), 1,
_("version def aux")) == NULL)
}
static const char *
-get_symbol_binding (unsigned int binding)
+get_symbol_binding (Filedata * filedata, unsigned int binding)
{
static char buff[32];
else if (binding >= STB_LOOS && binding <= STB_HIOS)
{
if (binding == STB_GNU_UNIQUE
- && (elf_header.e_ident[EI_OSABI] == ELFOSABI_GNU
+ && (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_GNU
/* GNU is still using the default value 0. */
- || elf_header.e_ident[EI_OSABI] == ELFOSABI_NONE))
+ || filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_NONE))
return "UNIQUE";
snprintf (buff, sizeof (buff), _("<OS specific>: %d"), binding);
}
}
static const char *
-get_symbol_type (unsigned int type)
+get_symbol_type (Filedata * filedata, unsigned int type)
{
static char buff[32];
default:
if (type >= STT_LOPROC && type <= STT_HIPROC)
{
- if (elf_header.e_machine == EM_ARM && type == STT_ARM_TFUNC)
+ if (filedata->file_header.e_machine == EM_ARM && type == STT_ARM_TFUNC)
return "THUMB_FUNC";
- if (elf_header.e_machine == EM_SPARCV9 && type == STT_REGISTER)
+ if (filedata->file_header.e_machine == EM_SPARCV9 && type == STT_REGISTER)
return "REGISTER";
- if (elf_header.e_machine == EM_PARISC && type == STT_PARISC_MILLI)
+ if (filedata->file_header.e_machine == EM_PARISC && type == STT_PARISC_MILLI)
return "PARISC_MILLI";
snprintf (buff, sizeof (buff), _("<processor specific>: %d"), type);
}
else if (type >= STT_LOOS && type <= STT_HIOS)
{
- if (elf_header.e_machine == EM_PARISC)
+ if (filedata->file_header.e_machine == EM_PARISC)
{
if (type == STT_HP_OPAQUE)
return "HP_OPAQUE";
}
if (type == STT_GNU_IFUNC
- && (elf_header.e_ident[EI_OSABI] == ELFOSABI_GNU
- || elf_header.e_ident[EI_OSABI] == ELFOSABI_FREEBSD
+ && (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_GNU
+ || filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_FREEBSD
/* GNU is still using the default value 0. */
- || elf_header.e_ident[EI_OSABI] == ELFOSABI_NONE))
+ || filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_NONE))
return "IFUNC";
snprintf (buff, sizeof (buff), _("<OS specific>: %d"), type);
}
static const char *
-get_ia64_symbol_other (unsigned int other)
+get_ia64_symbol_other (Filedata * filedata, unsigned int other)
{
- if (is_ia64_vms ())
+ if (is_ia64_vms (filedata))
{
static char res[32];
res[0] = 0;
/* Function types is for images and .STB files only. */
- switch (elf_header.e_type)
+ switch (filedata->file_header.e_type)
{
case ET_DYN:
case ET_EXEC:
}
static const char *
-get_symbol_other (unsigned int other)
+get_symbol_other (Filedata * filedata, unsigned int other)
{
const char * result = NULL;
static char buff [32];
if (other == 0)
return "";
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_MIPS:
result = get_mips_symbol_other (other);
break;
case EM_IA_64:
- result = get_ia64_symbol_other (other);
+ result = get_ia64_symbol_other (filedata, other);
break;
case EM_PPC64:
result = get_ppc64_symbol_other (other);
}
static const char *
-get_symbol_index_type (unsigned int type)
+get_symbol_index_type (Filedata * filedata, unsigned int type)
{
static char buff[32];
case SHN_COMMON: return "COM";
default:
if (type == SHN_IA_64_ANSI_COMMON
- && elf_header.e_machine == EM_IA_64
- && elf_header.e_ident[EI_OSABI] == ELFOSABI_HPUX)
+ && filedata->file_header.e_machine == EM_IA_64
+ && filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_HPUX)
return "ANSI_COM";
- else if ((elf_header.e_machine == EM_X86_64
- || elf_header.e_machine == EM_L1OM
- || elf_header.e_machine == EM_K1OM)
+ else if ((filedata->file_header.e_machine == EM_X86_64
+ || filedata->file_header.e_machine == EM_L1OM
+ || filedata->file_header.e_machine == EM_K1OM)
&& type == SHN_X86_64_LCOMMON)
return "LARGE_COM";
else if ((type == SHN_MIPS_SCOMMON
- && elf_header.e_machine == EM_MIPS)
+ && filedata->file_header.e_machine == EM_MIPS)
|| (type == SHN_TIC6X_SCOMMON
- && elf_header.e_machine == EM_TI_C6000))
+ && filedata->file_header.e_machine == EM_TI_C6000))
return "SCOM";
else if (type == SHN_MIPS_SUNDEFINED
- && elf_header.e_machine == EM_MIPS)
+ && filedata->file_header.e_machine == EM_MIPS)
return "SUND";
else if (type >= SHN_LOPROC && type <= SHN_HIPROC)
sprintf (buff, "PRC[0x%04x]", type & 0xffff);
sprintf (buff, "OS [0x%04x]", type & 0xffff);
else if (type >= SHN_LORESERVE)
sprintf (buff, "RSV[0x%04x]", type & 0xffff);
- else if (type >= elf_header.e_shnum)
+ else if (type >= filedata->file_header.e_shnum)
sprintf (buff, _("bad section index[%3d]"), type);
else
sprintf (buff, "%3d", type);
}
static bfd_vma *
-get_dynamic_data (FILE * file, bfd_size_type number, unsigned int ent_size)
+get_dynamic_data (Filedata * filedata, bfd_size_type number, unsigned int ent_size)
{
unsigned char * e_data;
bfd_vma * i_data;
/* Be kind to memory chekers (eg valgrind, address sanitizer) by not
attempting to allocate memory when the read is bound to fail. */
- if (ent_size * number > current_file_size)
+ if (ent_size * number > filedata->file_size)
{
error (_("Invalid number of dynamic entries: %s\n"),
bfd_vmatoa ("u", number));
return NULL;
}
- if (fread (e_data, ent_size, (size_t) number, file) != number)
+ if (fread (e_data, ent_size, (size_t) number, filedata->handle) != number)
{
error (_("Unable to read in %s bytes of dynamic data\n"),
bfd_vmatoa ("u", number * ent_size));
}
static void
-print_dynamic_symbol (bfd_vma si, unsigned long hn)
+print_dynamic_symbol (Filedata * filedata, bfd_vma si, unsigned long hn)
{
Elf_Internal_Sym * psym;
int n;
putchar (' ');
print_vma (psym->st_size, DEC_5);
- printf (" %-7s", get_symbol_type (ELF_ST_TYPE (psym->st_info)));
- printf (" %-6s", get_symbol_binding (ELF_ST_BIND (psym->st_info)));
+ printf (" %-7s", get_symbol_type (filedata, ELF_ST_TYPE (psym->st_info)));
+ printf (" %-6s", get_symbol_binding (filedata, ELF_ST_BIND (psym->st_info)));
- if (elf_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
+ if (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
printf (" %-7s", get_solaris_symbol_visibility (psym->st_other));
else
{
table being generated, but for the moment this case is very
rare. */
if (psym->st_other ^ vis)
- printf (" [%s] ", get_symbol_other (psym->st_other ^ vis));
+ printf (" [%s] ", get_symbol_other (filedata, psym->st_other ^ vis));
}
- printf (" %3.3s ", get_symbol_index_type (psym->st_shndx));
+ printf (" %3.3s ", get_symbol_index_type (filedata, psym->st_shndx));
if (VALID_DYNAMIC_NAME (psym->st_name))
print_symbol (25, GET_DYNAMIC_NAME (psym->st_name));
else
}
static const char *
-get_symbol_version_string (FILE * file,
+get_symbol_version_string (Filedata * filedata,
bfd_boolean is_dynsym,
const char * strtab,
unsigned long int strtab_size,
|| version_info[DT_VERSIONTAGIDX (DT_VERSYM)] == 0)
return NULL;
- offset = offset_from_vma (file, version_info[DT_VERSIONTAGIDX (DT_VERSYM)],
+ offset = offset_from_vma (filedata, version_info[DT_VERSIONTAGIDX (DT_VERSYM)],
sizeof data + si * sizeof (vers_data));
- if (get_data (&data, file, offset + si * sizeof (vers_data),
+ if (get_data (&data, filedata, offset + si * sizeof (vers_data),
sizeof (data), 1, _("version data")) == NULL)
return NULL;
Elf_External_Verdaux evda;
unsigned long off;
- off = offset_from_vma (file,
+ off = offset_from_vma (filedata,
version_info[DT_VERSIONTAGIDX (DT_VERDEF)],
sizeof (Elf_External_Verdef));
{
Elf_External_Verdef evd;
- if (get_data (&evd, file, off, sizeof (evd), 1,
+ if (get_data (&evd, filedata, off, sizeof (evd), 1,
_("version def")) == NULL)
{
ivd.vd_ndx = 0;
off -= ivd.vd_next;
off += ivd.vd_aux;
- if (get_data (&evda, file, off, sizeof (evda), 1,
+ if (get_data (&evda, filedata, off, sizeof (evda), 1,
_("version def aux")) != NULL)
{
ivda.vda_name = BYTE_GET (evda.vda_name);
Elf_Internal_Verneed ivn;
Elf_Internal_Vernaux ivna;
- offset = offset_from_vma (file,
+ offset = offset_from_vma (filedata,
version_info[DT_VERSIONTAGIDX (DT_VERNEED)],
sizeof evn);
do
{
unsigned long vna_off;
- if (get_data (&evn, file, offset, sizeof (evn), 1,
+ if (get_data (&evn, filedata, offset, sizeof (evn), 1,
_("version need")) == NULL)
{
ivna.vna_next = 0;
{
Elf_External_Vernaux evna;
- if (get_data (&evna, file, vna_off, sizeof (evna), 1,
+ if (get_data (&evna, filedata, vna_off, sizeof (evna), 1,
_("version need aux (3)")) == NULL)
{
ivna.vna_next = 0;
/* Dump the symbol table. */
static bfd_boolean
-process_symbol_table (FILE * file)
+process_symbol_table (Filedata * filedata)
{
Elf_Internal_Shdr * section;
bfd_size_type nbuckets = 0;
unsigned char nc[8];
unsigned int hash_ent_size = 4;
- if ((elf_header.e_machine == EM_ALPHA
- || elf_header.e_machine == EM_S390
- || elf_header.e_machine == EM_S390_OLD)
- && elf_header.e_ident[EI_CLASS] == ELFCLASS64)
+ if ((filedata->file_header.e_machine == EM_ALPHA
+ || filedata->file_header.e_machine == EM_S390
+ || filedata->file_header.e_machine == EM_S390_OLD)
+ && filedata->file_header.e_ident[EI_CLASS] == ELFCLASS64)
hash_ent_size = 8;
- if (fseek (file,
+ if (fseek (filedata->handle,
(archive_file_offset
- + offset_from_vma (file, dynamic_info[DT_HASH],
+ + offset_from_vma (filedata, dynamic_info[DT_HASH],
sizeof nb + sizeof nc)),
SEEK_SET))
{
goto no_hash;
}
- if (fread (nb, hash_ent_size, 1, file) != 1)
+ if (fread (nb, hash_ent_size, 1, filedata->handle) != 1)
{
error (_("Failed to read in number of buckets\n"));
goto no_hash;
}
- if (fread (nc, hash_ent_size, 1, file) != 1)
+ if (fread (nc, hash_ent_size, 1, filedata->handle) != 1)
{
error (_("Failed to read in number of chains\n"));
goto no_hash;
nbuckets = byte_get (nb, hash_ent_size);
nchains = byte_get (nc, hash_ent_size);
- buckets = get_dynamic_data (file, nbuckets, hash_ent_size);
- chains = get_dynamic_data (file, nchains, hash_ent_size);
+ buckets = get_dynamic_data (filedata, nbuckets, hash_ent_size);
+ chains = get_dynamic_data (filedata, nchains, hash_ent_size);
no_hash:
if (buckets == NULL || chains == NULL)
bfd_vma i, maxchain = 0xffffffff, bitmaskwords;
bfd_vma buckets_vma;
- if (fseek (file,
+ if (fseek (filedata->handle,
(archive_file_offset
- + offset_from_vma (file, dynamic_info_DT_GNU_HASH,
+ + offset_from_vma (filedata, dynamic_info_DT_GNU_HASH,
sizeof nb)),
SEEK_SET))
{
goto no_gnu_hash;
}
- if (fread (nb, 16, 1, file) != 1)
+ if (fread (nb, 16, 1, filedata->handle) != 1)
{
error (_("Failed to read in number of buckets\n"));
goto no_gnu_hash;
else
buckets_vma += bitmaskwords * 8;
- if (fseek (file,
+ if (fseek (filedata->handle,
(archive_file_offset
- + offset_from_vma (file, buckets_vma, 4)),
+ + offset_from_vma (filedata, buckets_vma, 4)),
SEEK_SET))
{
error (_("Unable to seek to start of dynamic information\n"));
goto no_gnu_hash;
}
- gnubuckets = get_dynamic_data (file, ngnubuckets, 4);
+ gnubuckets = get_dynamic_data (filedata, ngnubuckets, 4);
if (gnubuckets == NULL)
goto no_gnu_hash;
maxchain -= gnusymidx;
- if (fseek (file,
+ if (fseek (filedata->handle,
(archive_file_offset
- + offset_from_vma (file, buckets_vma
+ + offset_from_vma (filedata, buckets_vma
+ 4 * (ngnubuckets + maxchain), 4)),
SEEK_SET))
{
do
{
- if (fread (nb, 4, 1, file) != 1)
+ if (fread (nb, 4, 1, filedata->handle) != 1)
{
error (_("Failed to determine last chain length\n"));
goto no_gnu_hash;
}
while ((byte_get (nb, 4) & 1) == 0);
- if (fseek (file,
+ if (fseek (filedata->handle,
(archive_file_offset
- + offset_from_vma (file, buckets_vma + 4 * ngnubuckets, 4)),
+ + offset_from_vma (filedata, buckets_vma + 4 * ngnubuckets, 4)),
SEEK_SET))
{
error (_("Unable to seek to start of dynamic information\n"));
goto no_gnu_hash;
}
- gnuchains = get_dynamic_data (file, maxchain, 4);
+ gnuchains = get_dynamic_data (filedata, maxchain, 4);
ngnuchains = maxchain;
no_gnu_hash:
{
for (si = buckets[hn]; si > 0; si = chains[si])
{
- print_dynamic_symbol (si, hn);
+ print_dynamic_symbol (filedata, si, hn);
if (si >= nchains || visited[si])
{
error (_("histogram chain is corrupt\n"));
do
{
- print_dynamic_symbol (si, hn);
+ print_dynamic_symbol (filedata, si, hn);
si++;
}
while (off < ngnuchains && (gnuchains[off++] & 1) == 0);
}
}
else if ((do_dyn_syms || (do_syms && !do_using_dynamic))
- && section_headers != NULL)
+ && filedata->section_headers != NULL)
{
unsigned int i;
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum;
i++, section++)
{
unsigned int si;
if (section->sh_entsize == 0)
{
printf (_("\nSymbol table '%s' has a sh_entsize of zero!\n"),
- printable_section_name (section));
+ printable_section_name (filedata, section));
continue;
}
printf (ngettext ("\nSymbol table '%s' contains %lu entry:\n",
"\nSymbol table '%s' contains %lu entries:\n",
num_syms),
- printable_section_name (section),
+ printable_section_name (filedata, section),
num_syms);
if (is_32bit_elf)
else
printf (_(" Num: Value Size Type Bind Vis Ndx Name\n"));
- symtab = GET_ELF_SYMBOLS (file, section, & num_syms);
+ symtab = GET_ELF_SYMBOLS (filedata, section, & num_syms);
if (symtab == NULL)
continue;
- if (section->sh_link == elf_header.e_shstrndx)
+ if (section->sh_link == filedata->file_header.e_shstrndx)
{
- strtab = string_table;
- strtab_size = string_table_length;
+ strtab = filedata->string_table;
+ strtab_size = filedata->string_table_length;
}
- else if (section->sh_link < elf_header.e_shnum)
+ else if (section->sh_link < filedata->file_header.e_shnum)
{
Elf_Internal_Shdr * string_sec;
- string_sec = section_headers + section->sh_link;
+ string_sec = filedata->section_headers + section->sh_link;
- strtab = (char *) get_data (NULL, file, string_sec->sh_offset,
+ strtab = (char *) get_data (NULL, filedata, string_sec->sh_offset,
1, string_sec->sh_size,
_("string table"));
strtab_size = strtab != NULL ? string_sec->sh_size : 0;
print_vma (psym->st_value, LONG_HEX);
putchar (' ');
print_vma (psym->st_size, DEC_5);
- printf (" %-7s", get_symbol_type (ELF_ST_TYPE (psym->st_info)));
- printf (" %-6s", get_symbol_binding (ELF_ST_BIND (psym->st_info)));
- if (elf_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
+ printf (" %-7s", get_symbol_type (filedata, ELF_ST_TYPE (psym->st_info)));
+ printf (" %-6s", get_symbol_binding (filedata, ELF_ST_BIND (psym->st_info)));
+ if (filedata->file_header.e_ident[EI_OSABI] == ELFOSABI_SOLARIS)
printf (" %-7s", get_solaris_symbol_visibility (psym->st_other));
else
{
Note - displaying this information disrupts the layout of the
table being generated, but for the moment this case is very rare. */
if (psym->st_other ^ vis)
- printf (" [%s] ", get_symbol_other (psym->st_other ^ vis));
+ printf (" [%s] ", get_symbol_other (filedata, psym->st_other ^ vis));
}
- printf (" %4s ", get_symbol_index_type (psym->st_shndx));
+ printf (" %4s ", get_symbol_index_type (filedata, psym->st_shndx));
print_symbol (25, psym->st_name < strtab_size
? strtab + psym->st_name : _("<corrupt>"));
version_string
- = get_symbol_version_string (file,
+ = get_symbol_version_string (filedata,
section->sh_type == SHT_DYNSYM,
strtab, strtab_size, si,
psym, &sym_info, &vna_other);
if (ELF_ST_BIND (psym->st_info) == STB_LOCAL
&& si >= section->sh_info
/* Irix 5 and 6 MIPS binaries are known to ignore this requirement. */
- && elf_header.e_machine != EM_MIPS
+ && filedata->file_header.e_machine != EM_MIPS
/* Solaris binaries have been found to violate this requirement as
well. Not sure if this is a bug or an ABI requirement. */
- && elf_header.e_ident[EI_OSABI] != ELFOSABI_SOLARIS)
+ && filedata->file_header.e_ident[EI_OSABI] != ELFOSABI_SOLARIS)
warn (_("local symbol %u found at index >= %s's sh_info value of %u\n"),
- si, printable_section_name (section), section->sh_info);
+ si, printable_section_name (filedata, section), section->sh_info);
}
free (symtab);
- if (strtab != string_table)
+ if (strtab != filedata->string_table)
free (strtab);
}
}
}
static bfd_boolean
-process_syminfo (FILE * file ATTRIBUTE_UNUSED)
+process_syminfo (Filedata * filedata ATTRIBUTE_UNUSED)
{
unsigned int i;
discarded. */
static bfd_boolean
-target_specific_reloc_handling (Elf_Internal_Rela * reloc,
- unsigned char * start,
- unsigned char * end,
- Elf_Internal_Sym * symtab,
- unsigned long num_syms)
+target_specific_reloc_handling (Filedata * filedata,
+ Elf_Internal_Rela * reloc,
+ unsigned char * start,
+ unsigned char * end,
+ Elf_Internal_Sym * symtab,
+ unsigned long num_syms)
{
unsigned int reloc_type = 0;
unsigned long sym_index = 0;
if (reloc)
{
- reloc_type = get_reloc_type (reloc->r_info);
+ reloc_type = get_reloc_type (filedata, reloc->r_info);
sym_index = get_reloc_symindex (reloc->r_info);
}
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_MSP430:
case EM_MSP430_OLD:
switch (reloc_type)
{
case 10: /* R_MSP430_SYM_DIFF */
- if (uses_msp430x_relocs ())
+ if (uses_msp430x_relocs (filedata))
break;
/* Fall through. */
case 21: /* R_MSP430X_SYM_DIFF */
case 5: /* R_MSP430_16_BYTE */
case 9: /* R_MSP430_8 */
- if (uses_msp430x_relocs ())
+ if (uses_msp430x_relocs (filedata))
break;
goto handle_sym_diff;
case 2: /* R_MSP430_ABS16 */
case 15: /* R_MSP430X_ABS16 */
- if (! uses_msp430x_relocs ())
+ if (! uses_msp430x_relocs (filedata))
break;
goto handle_sym_diff;
reloc naming functions. */
static bfd_boolean
-is_32bit_abs_reloc (unsigned int reloc_type)
+is_32bit_abs_reloc (Filedata * filedata, unsigned int reloc_type)
{
/* Please keep this table alpha-sorted for ease of visual lookup. */
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_386:
case EM_IAMCU:
static unsigned int prev_warn = 0;
/* Avoid repeating the same warning multiple times. */
- if (prev_warn != elf_header.e_machine)
+ if (prev_warn != filedata->file_header.e_machine)
error (_("Missing knowledge of 32-bit reloc types used in DWARF sections of machine number %d\n"),
- elf_header.e_machine);
- prev_warn = elf_header.e_machine;
+ filedata->file_header.e_machine);
+ prev_warn = filedata->file_header.e_machine;
return FALSE;
}
}
a 32-bit pc-relative RELA relocation used in DWARF debug sections. */
static bfd_boolean
-is_32bit_pcrel_reloc (unsigned int reloc_type)
+is_32bit_pcrel_reloc (Filedata * filedata, unsigned int reloc_type)
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
/* Please keep this table alpha-sorted for ease of visual lookup. */
{
case EM_386:
a 64-bit absolute RELA relocation used in DWARF debug sections. */
static bfd_boolean
-is_64bit_abs_reloc (unsigned int reloc_type)
+is_64bit_abs_reloc (Filedata * filedata, unsigned int reloc_type)
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_AARCH64:
return reloc_type == 257; /* R_AARCH64_ABS64. */
a 64-bit pc-relative RELA relocation used in DWARF debug sections. */
static bfd_boolean
-is_64bit_pcrel_reloc (unsigned int reloc_type)
+is_64bit_pcrel_reloc (Filedata * filedata, unsigned int reloc_type)
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_AARCH64:
return reloc_type == 260; /* R_AARCH64_PREL64. */
a 24-bit absolute RELA relocation used in DWARF debug sections. */
static bfd_boolean
-is_24bit_abs_reloc (unsigned int reloc_type)
+is_24bit_abs_reloc (Filedata * filedata, unsigned int reloc_type)
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_CYGNUS_MN10200:
case EM_MN10200:
a 16-bit absolute RELA relocation used in DWARF debug sections. */
static bfd_boolean
-is_16bit_abs_reloc (unsigned int reloc_type)
+is_16bit_abs_reloc (Filedata * filedata, unsigned int reloc_type)
{
/* Please keep this table alpha-sorted for ease of visual lookup. */
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_ARC:
case EM_ARC_COMPACT:
case EM_MN10300:
return reloc_type == 2; /* R_MN10300_16. */
case EM_MSP430:
- if (uses_msp430x_relocs ())
+ if (uses_msp430x_relocs (filedata))
return reloc_type == 2; /* R_MSP430_ABS16. */
/* Fall through. */
case EM_MSP430_OLD:
relocation entries (possibly formerly used for SHT_GROUP sections). */
static bfd_boolean
-is_none_reloc (unsigned int reloc_type)
+is_none_reloc (Filedata * filedata, unsigned int reloc_type)
{
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_386: /* R_386_NONE. */
case EM_68K: /* R_68K_NONE. */
more relocations ? */
static bfd_boolean
-apply_relocations (void * file,
+apply_relocations (Filedata * filedata,
const Elf_Internal_Shdr * section,
unsigned char * start,
bfd_size_type size,
* num_relocs_return = 0;
}
- if (elf_header.e_type != ET_REL)
+ if (filedata->file_header.e_type != ET_REL)
/* No relocs to apply. */
return TRUE;
/* Find the reloc section associated with the section. */
- for (relsec = section_headers;
- relsec < section_headers + elf_header.e_shnum;
+ for (relsec = filedata->section_headers;
+ relsec < filedata->section_headers + filedata->file_header.e_shnum;
++relsec)
{
bfd_boolean is_rela;
Elf_Internal_Sym * sym;
if ((relsec->sh_type != SHT_RELA && relsec->sh_type != SHT_REL)
- || relsec->sh_info >= elf_header.e_shnum
- || section_headers + relsec->sh_info != section
+ || relsec->sh_info >= filedata->file_header.e_shnum
+ || filedata->section_headers + relsec->sh_info != section
|| relsec->sh_size == 0
- || relsec->sh_link >= elf_header.e_shnum)
+ || relsec->sh_link >= filedata->file_header.e_shnum)
continue;
is_rela = relsec->sh_type == SHT_RELA;
if (is_rela)
{
- if (!slurp_rela_relocs ((FILE *) file, relsec->sh_offset,
+ if (!slurp_rela_relocs (filedata, relsec->sh_offset,
relsec->sh_size, & relocs, & num_relocs))
return FALSE;
}
else
{
- if (!slurp_rel_relocs ((FILE *) file, relsec->sh_offset,
+ if (!slurp_rel_relocs (filedata, relsec->sh_offset,
relsec->sh_size, & relocs, & num_relocs))
return FALSE;
}
/* SH uses RELA but uses in place value instead of the addend field. */
- if (elf_header.e_machine == EM_SH)
+ if (filedata->file_header.e_machine == EM_SH)
is_rela = FALSE;
- symsec = section_headers + relsec->sh_link;
+ symsec = filedata->section_headers + relsec->sh_link;
if (symsec->sh_type != SHT_SYMTAB
&& symsec->sh_type != SHT_DYNSYM)
return FALSE;
- symtab = GET_ELF_SYMBOLS ((FILE *) file, symsec, & num_syms);
+ symtab = GET_ELF_SYMBOLS (filedata, symsec, & num_syms);
for (rp = relocs; rp < relocs + num_relocs; ++rp)
{
unsigned char * rloc;
unsigned long sym_index;
- reloc_type = get_reloc_type (rp->r_info);
+ reloc_type = get_reloc_type (filedata, rp->r_info);
- if (target_specific_reloc_handling (rp, start, end, symtab, num_syms))
+ if (target_specific_reloc_handling (filedata, rp, start, end, symtab, num_syms))
continue;
- else if (is_none_reloc (reloc_type))
+ else if (is_none_reloc (filedata, reloc_type))
continue;
- else if (is_32bit_abs_reloc (reloc_type)
- || is_32bit_pcrel_reloc (reloc_type))
+ else if (is_32bit_abs_reloc (filedata, reloc_type)
+ || is_32bit_pcrel_reloc (filedata, reloc_type))
reloc_size = 4;
- else if (is_64bit_abs_reloc (reloc_type)
- || is_64bit_pcrel_reloc (reloc_type))
+ else if (is_64bit_abs_reloc (filedata, reloc_type)
+ || is_64bit_pcrel_reloc (filedata, reloc_type))
reloc_size = 8;
- else if (is_24bit_abs_reloc (reloc_type))
+ else if (is_24bit_abs_reloc (filedata, reloc_type))
reloc_size = 3;
- else if (is_16bit_abs_reloc (reloc_type))
+ else if (is_16bit_abs_reloc (filedata, reloc_type))
reloc_size = 2;
else
{
static unsigned int prev_reloc = 0;
+
if (reloc_type != prev_reloc)
warn (_("unable to apply unsupported reloc type %d to section %s\n"),
- reloc_type, printable_section_name (section));
+ reloc_type, printable_section_name (filedata, section));
prev_reloc = reloc_type;
res = FALSE;
continue;
{
warn (_("skipping invalid relocation offset 0x%lx in section %s\n"),
(unsigned long) rp->r_offset,
- printable_section_name (section));
+ printable_section_name (filedata, section));
res = FALSE;
continue;
}
if (sym_index >= num_syms)
{
warn (_("skipping invalid relocation symbol index 0x%lx in section %s\n"),
- sym_index, printable_section_name (section));
+ sym_index, printable_section_name (filedata, section));
res = FALSE;
continue;
}
&& ELF_ST_TYPE (sym->st_info) > STT_SECTION)
{
warn (_("skipping unexpected symbol type %s in section %s relocation %ld\n"),
- get_symbol_type (ELF_ST_TYPE (sym->st_info)),
- printable_section_name (relsec),
+ get_symbol_type (filedata, ELF_ST_TYPE (sym->st_info)),
+ printable_section_name (filedata, relsec),
(long int)(rp - relocs));
res = FALSE;
continue;
/* R_XTENSA_32, R_PJ_DATA_DIR32 and R_D30V_32_NORMAL are
partial_inplace. */
if (!is_rela
- || (elf_header.e_machine == EM_XTENSA
+ || (filedata->file_header.e_machine == EM_XTENSA
&& reloc_type == 1)
- || ((elf_header.e_machine == EM_PJ
- || elf_header.e_machine == EM_PJ_OLD)
+ || ((filedata->file_header.e_machine == EM_PJ
+ || filedata->file_header.e_machine == EM_PJ_OLD)
&& reloc_type == 1)
- || ((elf_header.e_machine == EM_D30V
- || elf_header.e_machine == EM_CYGNUS_D30V)
+ || ((filedata->file_header.e_machine == EM_D30V
+ || filedata->file_header.e_machine == EM_CYGNUS_D30V)
&& reloc_type == 12))
addend += byte_get (rloc, reloc_size);
- if (is_32bit_pcrel_reloc (reloc_type)
- || is_64bit_pcrel_reloc (reloc_type))
+ if (is_32bit_pcrel_reloc (filedata, reloc_type)
+ || is_64bit_pcrel_reloc (filedata, reloc_type))
{
/* On HPPA, all pc-relative relocations are biased by 8. */
- if (elf_header.e_machine == EM_PARISC)
+ if (filedata->file_header.e_machine == EM_PARISC)
addend -= 8;
byte_put (rloc, (addend + sym->st_value) - rp->r_offset,
reloc_size);
free (symtab);
/* Let the target specific reloc processing code know that
we have finished with these relocs. */
- target_specific_reloc_handling (NULL, NULL, NULL, NULL, 0);
+ target_specific_reloc_handling (filedata, NULL, NULL, NULL, NULL, 0);
if (relocs_return)
{
#ifdef SUPPORT_DISASSEMBLY
static bfd_boolean
-disassemble_section (Elf_Internal_Shdr * section, FILE * file)
+disassemble_section (Elf_Internal_Shdr * section, Filedata * filedata)
{
- printf (_("\nAssembly dump of section %s\n"), printable_section_name (section));
+ printf (_("\nAssembly dump of section %s\n"), printable_section_name (filedata, section));
/* FIXME: XXX -- to be done --- XXX */
to a malloc'ed buffer or NULL if something went wrong. */
static char *
-get_section_contents (Elf_Internal_Shdr * section, FILE * file)
+get_section_contents (Elf_Internal_Shdr * section, Filedata * filedata)
{
- bfd_size_type num_bytes;
-
- num_bytes = section->sh_size;
+ bfd_size_type num_bytes = section->sh_size;
if (num_bytes == 0 || section->sh_type == SHT_NOBITS)
{
printf (_("Section '%s' has no data to dump.\n"),
- printable_section_name (section));
+ printable_section_name (filedata, section));
return NULL;
}
- return (char *) get_data (NULL, file, section->sh_offset, 1, num_bytes,
+ return (char *) get_data (NULL, filedata, section->sh_offset, 1, num_bytes,
_("section contents"));
}
/* Uncompresses a section that was compressed using zlib, in place. */
static bfd_boolean
-uncompress_section_contents (unsigned char **buffer,
- dwarf_size_type uncompressed_size,
- dwarf_size_type *size)
+uncompress_section_contents (unsigned char ** buffer,
+ dwarf_size_type uncompressed_size,
+ dwarf_size_type * size)
{
dwarf_size_type compressed_size = *size;
unsigned char * compressed_buffer = *buffer;
}
static bfd_boolean
-dump_section_as_strings (Elf_Internal_Shdr * section, FILE * file)
+dump_section_as_strings (Elf_Internal_Shdr * section, Filedata * filedata)
{
Elf_Internal_Shdr * relsec;
bfd_size_type num_bytes;
unsigned char * start;
bfd_boolean some_strings_shown;
- real_start = start = (unsigned char *) get_section_contents (section, file);
+ real_start = start = (unsigned char *) get_section_contents (section, filedata);
if (start == NULL)
/* PR 21820: Do not fail if the section was empty. */
return (section->sh_size == 0 || section->sh_type == SHT_NOBITS) ? TRUE : FALSE;
num_bytes = section->sh_size;
- printf (_("\nString dump of section '%s':\n"), printable_section_name (section));
+ printf (_("\nString dump of section '%s':\n"), printable_section_name (filedata, section));
if (decompress_dumps)
{
if (chdr.ch_type != ELFCOMPRESS_ZLIB)
{
warn (_("section '%s' has unsupported compress type: %d\n"),
- printable_section_name (section), chdr.ch_type);
+ printable_section_name (filedata, section), chdr.ch_type);
return FALSE;
}
else if (chdr.ch_addralign != section->sh_addralign)
{
warn (_("compressed section '%s' is corrupted\n"),
- printable_section_name (section));
+ printable_section_name (filedata, section));
return FALSE;
}
uncompressed_size = chdr.ch_size;
else
{
error (_("Unable to decompress section %s\n"),
- printable_section_name (section));
+ printable_section_name (filedata, section));
return FALSE;
}
}
case and issue a warning message in order to avoid confusion.
FIXME: Maybe we ought to have an option that dumps a section with
relocs applied ? */
- for (relsec = section_headers;
- relsec < section_headers + elf_header.e_shnum;
+ for (relsec = filedata->section_headers;
+ relsec < filedata->section_headers + filedata->file_header.e_shnum;
++relsec)
{
if ((relsec->sh_type != SHT_RELA && relsec->sh_type != SHT_REL)
- || relsec->sh_info >= elf_header.e_shnum
- || section_headers + relsec->sh_info != section
+ || relsec->sh_info >= filedata->file_header.e_shnum
+ || filedata->section_headers + relsec->sh_info != section
|| relsec->sh_size == 0
- || relsec->sh_link >= elf_header.e_shnum)
+ || relsec->sh_link >= filedata->file_header.e_shnum)
continue;
printf (_(" Note: This section has relocations against it, but these have NOT been applied to this dump.\n"));
}
static bfd_boolean
-dump_section_as_bytes (Elf_Internal_Shdr * section,
- FILE * file,
- bfd_boolean relocate)
+dump_section_as_bytes (Elf_Internal_Shdr * section,
+ Filedata * filedata,
+ bfd_boolean relocate)
{
Elf_Internal_Shdr * relsec;
bfd_size_type bytes;
unsigned char * real_start;
unsigned char * start;
- real_start = start = (unsigned char *) get_section_contents (section, file);
+ real_start = start = (unsigned char *) get_section_contents (section, filedata);
if (start == NULL)
/* PR 21820: Do not fail if the section was empty. */
return (section->sh_size == 0 || section->sh_type == SHT_NOBITS) ? TRUE : FALSE;
section_size = section->sh_size;
- printf (_("\nHex dump of section '%s':\n"), printable_section_name (section));
+ printf (_("\nHex dump of section '%s':\n"), printable_section_name (filedata, section));
if (decompress_dumps)
{
if (chdr.ch_type != ELFCOMPRESS_ZLIB)
{
warn (_("section '%s' has unsupported compress type: %d\n"),
- printable_section_name (section), chdr.ch_type);
+ printable_section_name (filedata, section), chdr.ch_type);
return FALSE;
}
else if (chdr.ch_addralign != section->sh_addralign)
{
warn (_("compressed section '%s' is corrupted\n"),
- printable_section_name (section));
+ printable_section_name (filedata, section));
return FALSE;
}
uncompressed_size = chdr.ch_size;
else
{
error (_("Unable to decompress section %s\n"),
- printable_section_name (section));
+ printable_section_name (filedata, section));
/* FIXME: Print the section anyway ? */
return FALSE;
}
if (relocate)
{
- if (! apply_relocations (file, section, start, section_size, NULL, NULL))
+ if (! apply_relocations (filedata, section, start, section_size, NULL, NULL))
return FALSE;
}
else
case and issue a warning message in order to avoid confusion.
FIXME: Maybe we ought to have an option that dumps a section with
relocs applied ? */
- for (relsec = section_headers;
- relsec < section_headers + elf_header.e_shnum;
+ for (relsec = filedata->section_headers;
+ relsec < filedata->section_headers + filedata->file_header.e_shnum;
++relsec)
{
if ((relsec->sh_type != SHT_RELA && relsec->sh_type != SHT_REL)
- || relsec->sh_info >= elf_header.e_shnum
- || section_headers + relsec->sh_info != section
+ || relsec->sh_info >= filedata->file_header.e_shnum
+ || filedata->section_headers + relsec->sh_info != section
|| relsec->sh_size == 0
- || relsec->sh_link >= elf_header.e_shnum)
+ || relsec->sh_link >= filedata->file_header.e_shnum)
continue;
printf (_(" NOTE: This section has relocations against it, but these have NOT been applied to this dump.\n"));
}
static bfd_boolean
-load_specific_debug_section (enum dwarf_section_display_enum debug,
- const Elf_Internal_Shdr * sec, void * file)
+load_specific_debug_section (enum dwarf_section_display_enum debug,
+ const Elf_Internal_Shdr * sec,
+ void * data)
{
struct dwarf_section * section = &debug_displays [debug].section;
char buf [64];
-
- /* If it is already loaded, do nothing. */
+ Filedata * filedata = (Filedata *) data;
+
if (section->start != NULL)
- return TRUE;
+ {
+ /* If it is already loaded, do nothing. */
+ if (streq (section->filename, filedata->file_name))
+ return TRUE;
+ free (section->start);
+ }
snprintf (buf, sizeof (buf), _("%s section data"), section->name);
section->address = sec->sh_addr;
section->user_data = NULL;
- section->start = (unsigned char *) get_data (NULL, (FILE *) file,
+ section->filename = filedata->file_name;
+ section->start = (unsigned char *) get_data (NULL, filedata,
sec->sh_offset, 1,
sec->sh_size, buf);
if (section->start == NULL)
else
{
error (_("Unable to decompress section %s\n"),
- printable_section_name (sec));
+ printable_section_name (filedata, sec));
return FALSE;
}
}
if (debug_displays [debug].relocate)
{
- if (! apply_relocations ((FILE *) file, sec, section->start, section->size,
+ if (! apply_relocations (filedata, sec, section->start, section->size,
& section->reloc_info, & section->num_relocs))
return FALSE;
}
static unsigned int * section_subset = NULL;
bfd_boolean
-load_debug_section (enum dwarf_section_display_enum debug, void * file)
+load_debug_section (enum dwarf_section_display_enum debug, void * data)
{
struct dwarf_section * section = &debug_displays [debug].section;
Elf_Internal_Shdr * sec;
+ Filedata * filedata = (Filedata *) data;
+
+ if (filedata->string_table == NULL)
+ {
+ Elf_Internal_Shdr * strs;
+
+ /* Read in the string table, so that we have section names to scan. */
+ strs = filedata->section_headers + filedata->file_header.e_shstrndx;
+
+ if (strs->sh_size != 0)
+ {
+ filedata->string_table = (char *) get_data (NULL, filedata, strs->sh_offset,
+ 1, strs->sh_size,
+ _("string table"));
+
+ filedata->string_table_length = filedata->string_table != NULL ? strs->sh_size : 0;
+ }
+ }
/* Locate the debug section. */
- sec = find_section_in_set (section->uncompressed_name, section_subset);
+ sec = find_section_in_set (filedata, section->uncompressed_name, section_subset);
if (sec != NULL)
section->name = section->uncompressed_name;
else
{
- sec = find_section_in_set (section->compressed_name, section_subset);
+ sec = find_section_in_set (filedata, section->compressed_name, section_subset);
if (sec != NULL)
section->name = section->compressed_name;
}
if (section_subset != NULL)
free_debug_section (debug);
- return load_specific_debug_section (debug, sec, (FILE *) file);
+ return load_specific_debug_section (debug, sec, data);
}
void
}
static bfd_boolean
-display_debug_section (int shndx, Elf_Internal_Shdr * section, FILE * file)
+display_debug_section (int shndx, Elf_Internal_Shdr * section, Filedata * filedata)
{
char * name = SECTION_NAME (section);
- const char * print_name = printable_section_name (section);
+ const char * print_name = printable_section_name (filedata, section);
bfd_size_type length;
bfd_boolean result = TRUE;
int i;
|| streq (debug_displays[i].section.compressed_name, name))
{
struct dwarf_section * sec = &debug_displays [i].section;
- int secondary = (section != find_section (name));
+ int secondary = (section != find_section (filedata, name));
if (secondary)
free_debug_section ((enum dwarf_section_display_enum) i);
sec->name = sec->uncompressed_name;
else
sec->name = sec->compressed_name;
+
if (load_specific_debug_section ((enum dwarf_section_display_enum) i,
- section, file))
+ section, filedata))
{
/* If this debug section is part of a CU/TU set in a .dwp file,
restrict load_debug_section to the sections in that set. */
- section_subset = find_cu_tu_set (file, shndx);
+ section_subset = find_cu_tu_set (filedata, shndx);
- result &= debug_displays[i].display (sec, file);
+ result &= debug_displays[i].display (sec, filedata);
section_subset = NULL;
based on section name. */
static void
-initialise_dumps_byname (void)
+initialise_dumps_byname (Filedata * filedata)
{
struct dump_list_entry * cur;
unsigned int i;
bfd_boolean any = FALSE;
- for (i = 0; i < elf_header.e_shnum; i++)
- if (streq (SECTION_NAME (section_headers + i), cur->name))
+ for (i = 0; i < filedata->file_header.e_shnum; i++)
+ if (streq (SECTION_NAME (filedata->section_headers + i), cur->name))
{
- request_dump_bynumber (i, cur->type);
+ request_dump_bynumber (filedata, i, cur->type);
any = TRUE;
}
}
static bfd_boolean
-process_section_contents (FILE * file)
+process_section_contents (Filedata * filedata)
{
Elf_Internal_Shdr * section;
unsigned int i;
if (! do_dump)
return TRUE;
- initialise_dumps_byname ();
+ initialise_dumps_byname (filedata);
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum && i < num_dump_sects;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum && i < filedata->num_dump_sects;
i++, section++)
{
+ dump_type dump = filedata->dump_sects[i];
+
#ifdef SUPPORT_DISASSEMBLY
- if (dump_sects[i] & DISASS_DUMP)
- disassemble_section (section, file);
+ if (dump & DISASS_DUMP)
+ {
+ if (! disassemble_section (section, filedata))
+ res = FALSE;
+ }
#endif
- if (dump_sects[i] & HEX_DUMP)
+ if (dump & HEX_DUMP)
{
- if (! dump_section_as_bytes (section, file, FALSE))
+ if (! dump_section_as_bytes (section, filedata, FALSE))
res = FALSE;
}
- if (dump_sects[i] & RELOC_DUMP)
+ if (dump & RELOC_DUMP)
{
- if (! dump_section_as_bytes (section, file, TRUE))
+ if (! dump_section_as_bytes (section, filedata, TRUE))
res = FALSE;
}
- if (dump_sects[i] & STRING_DUMP)
+ if (dump & STRING_DUMP)
{
- if (! dump_section_as_strings (section, file))
+ if (! dump_section_as_strings (section, filedata))
res = FALSE;
}
- if (dump_sects[i] & DEBUG_DUMP)
+ if (dump & DEBUG_DUMP)
{
- if (! display_debug_section (i, section, file))
+ if (! display_debug_section (i, section, filedata))
res = FALSE;
}
}
/* Check to see if the user requested a
dump of a section that does not exist. */
- while (i < num_dump_sects)
+ while (i < filedata->num_dump_sects)
{
- if (dump_sects[i])
+ if (filedata->dump_sects[i])
{
warn (_("Section %d was not dumped because it does not exist!\n"), i);
res = FALSE;
}
static bfd_boolean
-process_attributes (FILE * file,
+process_attributes (Filedata * filedata,
const char * public_name,
unsigned int proc_type,
unsigned char * (* display_pub_attribute) (unsigned char *, const unsigned char * const),
bfd_boolean res = TRUE;
/* Find the section header so that we get the size. */
- for (i = 0, sect = section_headers;
- i < elf_header.e_shnum;
+ for (i = 0, sect = filedata->section_headers;
+ i < filedata->file_header.e_shnum;
i++, sect++)
{
unsigned char * contents;
if (sect->sh_type != proc_type && sect->sh_type != SHT_GNU_ATTRIBUTES)
continue;
- contents = (unsigned char *) get_data (NULL, file, sect->sh_offset, 1,
+ contents = (unsigned char *) get_data (NULL, filedata, sect->sh_offset, 1,
sect->sh_size, _("attributes"));
if (contents == NULL)
{
}
static bfd_boolean
-process_mips_specific (FILE * file)
+process_mips_specific (Filedata * filedata)
{
Elf_Internal_Dyn * entry;
Elf_Internal_Shdr *sect = NULL;
bfd_vma symtabno = 0;
bfd_boolean res = TRUE;
- if (! process_attributes (file, NULL, SHT_GNU_ATTRIBUTES, NULL,
+ if (! process_attributes (filedata, NULL, SHT_GNU_ATTRIBUTES, NULL,
display_mips_gnu_attribute))
res = FALSE;
- sect = find_section (".MIPS.abiflags");
+ sect = find_section (filedata, ".MIPS.abiflags");
if (sect != NULL)
{
}
else
{
- abiflags_ext = get_data (NULL, file, sect->sh_offset, 1,
+ abiflags_ext = get_data (NULL, filedata, sect->sh_offset, 1,
sect->sh_size, _("MIPS ABI Flags section"));
if (abiflags_ext)
{
if (dynamic_section == NULL)
{
/* No dynamic information available. See if there is static GOT. */
- sect = find_section (".got");
+ sect = find_section (filedata, ".got");
if (sect != NULL)
{
unsigned char *data_end;
addr_size = (is_32bit_elf ? 4 : 8);
end = pltgot + sect->sh_size;
- data = (unsigned char *) get_data (NULL, file, sect->sh_offset,
+ data = (unsigned char *) get_data (NULL, filedata, sect->sh_offset,
end - pltgot, 1,
_("Global Offset Table data"));
/* PR 12855: Null data is handled gracefully throughout. */
{
case DT_MIPS_LIBLIST:
liblist_offset
- = offset_from_vma (file, entry->d_un.d_val,
+ = offset_from_vma (filedata, entry->d_un.d_val,
liblistno * sizeof (Elf32_External_Lib));
break;
case DT_MIPS_LIBLISTNO:
liblistno = entry->d_un.d_val;
break;
case DT_MIPS_OPTIONS:
- options_offset = offset_from_vma (file, entry->d_un.d_val, 0);
+ options_offset = offset_from_vma (filedata, entry->d_un.d_val, 0);
break;
case DT_MIPS_CONFLICT:
conflicts_offset
- = offset_from_vma (file, entry->d_un.d_val,
+ = offset_from_vma (filedata, entry->d_un.d_val,
conflictsno * sizeof (Elf32_External_Conflict));
break;
case DT_MIPS_CONFLICTNO:
Elf32_External_Lib * elib;
size_t cnt;
- elib = (Elf32_External_Lib *) get_data (NULL, file, liblist_offset,
+ elib = (Elf32_External_Lib *) get_data (NULL, filedata, liblist_offset,
liblistno,
sizeof (Elf32_External_Lib),
_("liblist section data"));
Elf_Internal_Options * option;
size_t offset;
int cnt;
- sect = section_headers;
+ sect = filedata->section_headers;
/* Find the section header so that we get the size. */
- sect = find_section_by_type (SHT_MIPS_OPTIONS);
+ sect = find_section_by_type (filedata, SHT_MIPS_OPTIONS);
/* PR 17533 file: 012-277276-0.004. */
if (sect == NULL)
{
return FALSE;
}
- eopt = (Elf_External_Options *) get_data (NULL, file, options_offset, 1,
+ eopt = (Elf_External_Options *) get_data (NULL, filedata, options_offset, 1,
sect->sh_size, _("options"));
if (eopt)
{
printf (ngettext ("\nSection '%s' contains %d entry:\n",
"\nSection '%s' contains %d entries:\n",
cnt),
- printable_section_name (sect), cnt);
+ printable_section_name (filedata, sect), cnt);
option = iopt;
offset = 0;
break;
case ODK_REGINFO:
printf (" REGINFO ");
- if (elf_header.e_machine == EM_MIPS)
+ if (filedata->file_header.e_machine == EM_MIPS)
{
/* 32bit form. */
Elf32_External_RegInfo * ereg;
/* PR 21345 - print a slightly more helpful error message
if we are sure that the cmalloc will fail. */
- if (conflictsno * sizeof (* iconf) > current_file_size)
+ if (conflictsno * sizeof (* iconf) > filedata->file_size)
{
error (_("Overlarge number of conflicts detected: %lx\n"),
(long) conflictsno);
Elf32_External_Conflict * econf32;
econf32 = (Elf32_External_Conflict *)
- get_data (NULL, file, conflicts_offset, conflictsno,
+ get_data (NULL, filedata, conflicts_offset, conflictsno,
sizeof (* econf32), _("conflict"));
if (!econf32)
return FALSE;
Elf64_External_Conflict * econf64;
econf64 = (Elf64_External_Conflict *)
- get_data (NULL, file, conflicts_offset, conflictsno,
+ get_data (NULL, filedata, conflicts_offset, conflictsno,
sizeof (* econf64), _("conflict"));
if (!econf64)
return FALSE;
return FALSE;
}
- offset = offset_from_vma (file, pltgot, global_end - pltgot);
- data = (unsigned char *) get_data (NULL, file, offset,
+ offset = offset_from_vma (filedata, pltgot, global_end - pltgot);
+ data = (unsigned char *) get_data (NULL, filedata, offset,
global_end - pltgot, 1,
_("Global Offset Table data"));
/* PR 12855: Null data is handled gracefully throughout. */
print_vma (psym->st_value, LONG_HEX);
printf (" %-7s %3s ",
- get_symbol_type (ELF_ST_TYPE (psym->st_info)),
- get_symbol_index_type (psym->st_shndx));
+ get_symbol_type (filedata, ELF_ST_TYPE (psym->st_info)),
+ get_symbol_index_type (filedata, psym->st_shndx));
if (VALID_DYNAMIC_NAME (psym->st_name))
print_symbol (sym_width, GET_DYNAMIC_NAME (psym->st_name));
int addr_size, sym_width;
Elf_Internal_Rela * rels;
- rel_offset = offset_from_vma (file, jmprel, pltrelsz);
+ rel_offset = offset_from_vma (filedata, jmprel, pltrelsz);
if (pltrel == DT_RELA)
{
- if (!slurp_rela_relocs (file, rel_offset, pltrelsz, &rels, &count))
+ if (!slurp_rela_relocs (filedata, rel_offset, pltrelsz, &rels, &count))
return FALSE;
}
else
{
- if (!slurp_rel_relocs (file, rel_offset, pltrelsz, &rels, &count))
+ if (!slurp_rel_relocs (filedata, rel_offset, pltrelsz, &rels, &count))
return FALSE;
}
addr_size = (is_32bit_elf ? 4 : 8);
end = mips_pltgot + (2 + count) * addr_size;
- offset = offset_from_vma (file, mips_pltgot, end - mips_pltgot);
- data = (unsigned char *) get_data (NULL, file, offset, end - mips_pltgot,
+ offset = offset_from_vma (filedata, mips_pltgot, end - mips_pltgot);
+ data = (unsigned char *) get_data (NULL, filedata, offset, end - mips_pltgot,
1, _("Procedure Linkage Table data"));
if (data == NULL)
return FALSE;
print_vma (psym->st_value, LONG_HEX);
printf (" %-7s %3s ",
- get_symbol_type (ELF_ST_TYPE (psym->st_info)),
- get_symbol_index_type (psym->st_shndx));
+ get_symbol_type (filedata, ELF_ST_TYPE (psym->st_info)),
+ get_symbol_index_type (filedata, psym->st_shndx));
if (VALID_DYNAMIC_NAME (psym->st_name))
print_symbol (sym_width, GET_DYNAMIC_NAME (psym->st_name));
else
}
static bfd_boolean
-process_nds32_specific (FILE * file)
+process_nds32_specific (Filedata * filedata)
{
Elf_Internal_Shdr *sect = NULL;
- sect = find_section (".nds32_e_flags");
+ sect = find_section (filedata, ".nds32_e_flags");
if (sect != NULL)
{
unsigned int *flag;
printf ("\nNDS32 elf flags section:\n");
- flag = get_data (NULL, file, sect->sh_offset, 1,
+ flag = get_data (NULL, filedata, sect->sh_offset, 1,
sect->sh_size, _("NDS32 elf flags section"));
if (! flag)
}
static bfd_boolean
-process_gnu_liblist (FILE * file)
+process_gnu_liblist (Filedata * filedata)
{
Elf_Internal_Shdr * section;
Elf_Internal_Shdr * string_sec;
if (! do_arch)
return TRUE;
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum;
i++, section++)
{
switch (section->sh_type)
{
case SHT_GNU_LIBLIST:
- if (section->sh_link >= elf_header.e_shnum)
+ if (section->sh_link >= filedata->file_header.e_shnum)
break;
elib = (Elf32_External_Lib *)
- get_data (NULL, file, section->sh_offset, 1, section->sh_size,
+ get_data (NULL, filedata, section->sh_offset, 1, section->sh_size,
_("liblist section data"));
if (elib == NULL)
break;
}
- string_sec = section_headers + section->sh_link;
- strtab = (char *) get_data (NULL, file, string_sec->sh_offset, 1,
+ string_sec = filedata->section_headers + section->sh_link;
+ strtab = (char *) get_data (NULL, filedata, string_sec->sh_offset, 1,
string_sec->sh_size,
_("liblist string table"));
if (strtab == NULL
printf (ngettext ("\nLibrary list section '%s' contains %lu entries:\n",
"\nLibrary list section '%s' contains %lu entries:\n",
num_liblist),
- printable_section_name (section),
+ printable_section_name (filedata, section),
num_liblist);
puts (_(" Library Time Stamp Checksum Version Flags"));
}
static const char *
-get_note_type (unsigned e_type)
+get_note_type (Filedata * filedata, unsigned e_type)
{
static char buff[64];
- if (elf_header.e_type == ET_CORE)
+ if (filedata->file_header.e_type == ET_CORE)
switch (e_type)
{
case NT_AUXV:
}
static void
-print_gnu_property_note (Elf_Internal_Note * pnote)
+print_gnu_property_note (Filedata * filedata, Elf_Internal_Note * pnote)
{
unsigned char * ptr = (unsigned char *) pnote->descdata;
unsigned char * ptr_end = ptr + pnote->descsz;
if (type >= GNU_PROPERTY_LOPROC && type <= GNU_PROPERTY_HIPROC)
{
- if (elf_header.e_machine == EM_X86_64
- || elf_header.e_machine == EM_IAMCU
- || elf_header.e_machine == EM_386)
+ if (filedata->file_header.e_machine == EM_X86_64
+ || filedata->file_header.e_machine == EM_IAMCU
+ || filedata->file_header.e_machine == EM_386)
{
switch (type)
{
}
static bfd_boolean
-print_gnu_note (Elf_Internal_Note *pnote)
+print_gnu_note (Filedata * filedata, Elf_Internal_Note *pnote)
{
/* NB/ Keep this switch statement in sync with get_gnu_elf_note_type (). */
switch (pnote->type)
break;
case NT_GNU_PROPERTY_TYPE_0:
- print_gnu_property_note (pnote);
+ print_gnu_property_note (filedata, pnote);
break;
default:
}
static const char *
-get_freebsd_elfcore_note_type (unsigned e_type)
+get_freebsd_elfcore_note_type (Filedata * filedata, unsigned e_type)
{
switch (e_type)
{
case NT_FREEBSD_PTLWPINFO:
return _("NT_PTLWPINFO (ptrace_lwpinfo structure)");
}
- return get_note_type (e_type);
+ return get_note_type (filedata, e_type);
}
static const char *
-get_netbsd_elfcore_note_type (unsigned e_type)
+get_netbsd_elfcore_note_type (Filedata * filedata, unsigned e_type)
{
static char buff[64];
if (e_type == NT_NETBSDCORE_PROCINFO)
- {
- /* NetBSD core "procinfo" structure. */
- return _("NetBSD procinfo structure");
- }
+ return _("NetBSD procinfo structure");
/* As of Jan 2002 there are no other machine-independent notes
defined for NetBSD core files. If the note type is less
return buff;
}
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
/* On the Alpha, SPARC (32-bit and 64-bit), PT_GETREGS == mach+0
and PT_GETFPREGS == mach+2. */
that is attached to address OFFSET. */
static bfd_boolean
-print_symbol_for_build_attribute (FILE * file,
+print_symbol_for_build_attribute (Filedata * filedata,
unsigned long offset,
bfd_boolean is_open_attr)
{
- static FILE * saved_file = NULL;
+ static Filedata * saved_filedata = NULL;
static char * strtab;
static unsigned long strtablen;
static Elf_Internal_Sym * symtab;
Elf_Internal_Sym * saved_sym = NULL;
Elf_Internal_Sym * sym;
- if (section_headers != NULL
- && (saved_file == NULL || file != saved_file))
+ if (filedata->section_headers != NULL
+ && (saved_filedata == NULL || filedata != saved_filedata))
{
Elf_Internal_Shdr * symsec;
/* Load the symbol and string sections. */
- for (symsec = section_headers;
- symsec < section_headers + elf_header.e_shnum;
+ for (symsec = filedata->section_headers;
+ symsec < filedata->section_headers + filedata->file_header.e_shnum;
symsec ++)
{
if (symsec->sh_type == SHT_SYMTAB)
{
- symtab = GET_ELF_SYMBOLS (file, symsec, & nsyms);
+ symtab = GET_ELF_SYMBOLS (filedata, symsec, & nsyms);
- if (symsec->sh_link < elf_header.e_shnum)
+ if (symsec->sh_link < filedata->file_header.e_shnum)
{
- Elf_Internal_Shdr * strtab_sec = section_headers + symsec->sh_link;
+ Elf_Internal_Shdr * strtab_sec = filedata->section_headers + symsec->sh_link;
- strtab = (char *) get_data (NULL, file, strtab_sec->sh_offset,
+ strtab = (char *) get_data (NULL, filedata, strtab_sec->sh_offset,
1, strtab_sec->sh_size,
_("string table"));
strtablen = strtab != NULL ? strtab_sec->sh_size : 0;
}
}
}
- saved_file = file;
+ saved_filedata = filedata;
}
if (symtab == NULL || strtab == NULL)
}
static bfd_boolean
-print_gnu_build_attribute_description (Elf_Internal_Note * pnote,
- FILE * file)
+print_gnu_build_attribute_description (Elf_Internal_Note * pnote,
+ Filedata * filedata)
{
static unsigned long global_offset = 0;
unsigned long offset;
else
{
printf (_(" Applies to func at %#lx"), global_offset);
- return print_symbol_for_build_attribute (file, global_offset, is_open_attr);
+ return print_symbol_for_build_attribute (filedata, global_offset, is_open_attr);
}
}
printf (_(" Applies to func at %#lx"), offset);
}
- return print_symbol_for_build_attribute (file, offset, is_open_attr);
+ return print_symbol_for_build_attribute (filedata, offset, is_open_attr);
}
static bfd_boolean
if (bytes > sizeof (val))
{
- fprintf (stderr, "namesz %lx name %p namedata %p\n",
- pnote->namesz, name, pnote->namedata);
error (_("corrupt numeric name field: too many bytes in the value: %x\n"),
bytes);
bytes = sizeof (val);
static bfd_boolean
process_note (Elf_Internal_Note * pnote,
- FILE * file)
+ Filedata * filedata)
{
const char * name = pnote->namesz ? pnote->namedata : "(NONE)";
const char * nt;
if (pnote->namesz == 0)
/* If there is no note name, then use the default set of
note type strings. */
- nt = get_note_type (pnote->type);
+ nt = get_note_type (filedata, pnote->type);
else if (const_strneq (pnote->namedata, "GNU"))
/* GNU-specific object file notes. */
else if (const_strneq (pnote->namedata, "FreeBSD"))
/* FreeBSD-specific core file notes. */
- nt = get_freebsd_elfcore_note_type (pnote->type);
+ nt = get_freebsd_elfcore_note_type (filedata, pnote->type);
else if (const_strneq (pnote->namedata, "NetBSD-CORE"))
/* NetBSD-specific core file notes. */
- nt = get_netbsd_elfcore_note_type (pnote->type);
+ nt = get_netbsd_elfcore_note_type (filedata, pnote->type);
else if (const_strneq (pnote->namedata, "NetBSD"))
/* NetBSD-specific core file notes. */
else
/* Don't recognize this note name; just use the default set of
note type strings. */
- nt = get_note_type (pnote->type);
+ nt = get_note_type (filedata, pnote->type);
printf (" ");
if (const_strneq (pnote->namedata, "IPF/VMS"))
return print_ia64_vms_note (pnote);
else if (const_strneq (pnote->namedata, "GNU"))
- return print_gnu_note (pnote);
+ return print_gnu_note (filedata, pnote);
else if (const_strneq (pnote->namedata, "stapsdt"))
return print_stapsdt_note (pnote);
else if (const_strneq (pnote->namedata, "CORE"))
|| strchr ("*$!+", pnote->namedata[0]) != NULL)
&& (pnote->type == NT_GNU_BUILD_ATTRIBUTE_OPEN
|| pnote->type == NT_GNU_BUILD_ATTRIBUTE_FUNC))
- return print_gnu_build_attribute_description (pnote, file);
+ return print_gnu_build_attribute_description (pnote, filedata);
if (pnote->descsz)
{
}
static bfd_boolean
-process_notes_at (FILE * file,
- Elf_Internal_Shdr * section,
- bfd_vma offset,
- bfd_vma length)
+process_notes_at (Filedata * filedata,
+ Elf_Internal_Shdr * section,
+ bfd_vma offset,
+ bfd_vma length)
{
Elf_External_Note * pnotes;
Elf_External_Note * external;
if (section)
{
- pnotes = (Elf_External_Note *) get_section_contents (section, file);
+ pnotes = (Elf_External_Note *) get_section_contents (section, filedata);
if (pnotes)
{
- if (! apply_relocations (file, section, (unsigned char *) pnotes, length, NULL, NULL))
+ if (! apply_relocations (filedata, section, (unsigned char *) pnotes, length, NULL, NULL))
return FALSE;
}
}
else
- pnotes = (Elf_External_Note *) get_data (NULL, file, offset, 1, length,
+ pnotes = (Elf_External_Note *) get_data (NULL, filedata, offset, 1, length,
_("notes"));
if (pnotes == NULL)
return FALSE;
external = pnotes;
if (section)
- printf (_("\nDisplaying notes found in: %s\n"), printable_section_name (section));
+ printf (_("\nDisplaying notes found in: %s\n"), printable_section_name (filedata, section));
else
printf (_("\nDisplaying notes found at file offset 0x%08lx with length 0x%08lx:\n"),
(unsigned long) offset, (unsigned long) length);
char * temp = NULL;
size_t data_remaining = end - (char *) external;
- if (!is_ia64_vms ())
+ if (!is_ia64_vms (filedata))
{
/* PR binutils/15191
Make sure that there is enough data to read. */
inote.namedata[inote.namesz] = 0;
}
- if (! process_note (& inote, file))
+ if (! process_note (& inote, filedata))
res = FALSE;
if (temp != NULL)
}
static bfd_boolean
-process_corefile_note_segments (FILE * file)
+process_corefile_note_segments (Filedata * filedata)
{
Elf_Internal_Phdr * segment;
unsigned int i;
bfd_boolean res = TRUE;
- if (! get_program_headers (file))
+ if (! get_program_headers (filedata))
return TRUE;
- for (i = 0, segment = program_headers;
- i < elf_header.e_phnum;
+ for (i = 0, segment = filedata->program_headers;
+ i < filedata->file_header.e_phnum;
i++, segment++)
{
if (segment->p_type == PT_NOTE)
- if (! process_notes_at (file, NULL,
+ if (! process_notes_at (filedata, NULL,
(bfd_vma) segment->p_offset,
(bfd_vma) segment->p_filesz))
res = FALSE;
}
static bfd_boolean
-process_v850_notes (FILE * file, bfd_vma offset, bfd_vma length)
+process_v850_notes (Filedata * filedata, bfd_vma offset, bfd_vma length)
{
Elf_External_Note * pnotes;
Elf_External_Note * external;
if (length <= 0)
return FALSE;
- pnotes = (Elf_External_Note *) get_data (NULL, file, offset, 1, length,
+ pnotes = (Elf_External_Note *) get_data (NULL, filedata, offset, 1, length,
_("v850 notes"));
if (pnotes == NULL)
return FALSE;
}
static bfd_boolean
-process_note_sections (FILE * file)
+process_note_sections (Filedata * filedata)
{
Elf_Internal_Shdr * section;
unsigned long i;
unsigned int n = 0;
bfd_boolean res = TRUE;
- for (i = 0, section = section_headers;
- i < elf_header.e_shnum && section != NULL;
+ for (i = 0, section = filedata->section_headers;
+ i < filedata->file_header.e_shnum && section != NULL;
i++, section++)
{
if (section->sh_type == SHT_NOTE)
{
- if (! process_notes_at (file, section,
+ if (! process_notes_at (filedata, section,
(bfd_vma) section->sh_offset,
(bfd_vma) section->sh_size))
res = FALSE;
n++;
}
- if (( elf_header.e_machine == EM_V800
- || elf_header.e_machine == EM_V850
- || elf_header.e_machine == EM_CYGNUS_V850)
+ if (( filedata->file_header.e_machine == EM_V800
+ || filedata->file_header.e_machine == EM_V850
+ || filedata->file_header.e_machine == EM_CYGNUS_V850)
&& section->sh_type == SHT_RENESAS_INFO)
{
- if (! process_v850_notes (file,
+ if (! process_v850_notes (filedata,
(bfd_vma) section->sh_offset,
(bfd_vma) section->sh_size))
res = FALSE;
if (n == 0)
/* Try processing NOTE segments instead. */
- return process_corefile_note_segments (file);
+ return process_corefile_note_segments (filedata);
return res;
}
static bfd_boolean
-process_notes (FILE * file)
+process_notes (Filedata * filedata)
{
/* If we have not been asked to display the notes then do nothing. */
if (! do_notes)
return TRUE;
- if (elf_header.e_type != ET_CORE)
- return process_note_sections (file);
+ if (filedata->file_header.e_type != ET_CORE)
+ return process_note_sections (filedata);
/* No program headers means no NOTE segment. */
- if (elf_header.e_phnum > 0)
- return process_corefile_note_segments (file);
+ if (filedata->file_header.e_phnum > 0)
+ return process_corefile_note_segments (filedata);
printf (_("No note segments present in the core file.\n"));
return TRUE;
}
static bfd_boolean
-process_arch_specific (FILE * file)
+process_arch_specific (Filedata * filedata)
{
if (! do_arch)
return TRUE;
- switch (elf_header.e_machine)
+ switch (filedata->file_header.e_machine)
{
case EM_ARC:
case EM_ARC_COMPACT:
case EM_ARC_COMPACT2:
- return process_attributes (file, "ARC", SHT_ARC_ATTRIBUTES,
+ return process_attributes (filedata, "ARC", SHT_ARC_ATTRIBUTES,
display_arc_attribute,
display_generic_attribute);
case EM_ARM:
- return process_attributes (file, "aeabi", SHT_ARM_ATTRIBUTES,
+ return process_attributes (filedata, "aeabi", SHT_ARM_ATTRIBUTES,
display_arm_attribute,
display_generic_attribute);
case EM_MIPS:
case EM_MIPS_RS3_LE:
- return process_mips_specific (file);
+ return process_mips_specific (filedata);
case EM_MSP430:
- return process_attributes (file, "mspabi", SHT_MSP430_ATTRIBUTES,
- display_msp430x_attribute,
- display_generic_attribute);
+ return process_attributes (filedata, "mspabi", SHT_MSP430_ATTRIBUTES,
+ display_msp430x_attribute,
+ display_generic_attribute);
case EM_NDS32:
- return process_nds32_specific (file);
+ return process_nds32_specific (filedata);
case EM_PPC:
case EM_PPC64:
- return process_attributes (file, NULL, SHT_GNU_ATTRIBUTES, NULL,
+ return process_attributes (filedata, NULL, SHT_GNU_ATTRIBUTES, NULL,
display_power_gnu_attribute);
case EM_S390:
case EM_S390_OLD:
- return process_attributes (file, NULL, SHT_GNU_ATTRIBUTES, NULL,
+ return process_attributes (filedata, NULL, SHT_GNU_ATTRIBUTES, NULL,
display_s390_gnu_attribute);
case EM_SPARC:
case EM_SPARC32PLUS:
case EM_SPARCV9:
- return process_attributes (file, NULL, SHT_GNU_ATTRIBUTES, NULL,
+ return process_attributes (filedata, NULL, SHT_GNU_ATTRIBUTES, NULL,
display_sparc_gnu_attribute);
case EM_TI_C6000:
- return process_attributes (file, "c6xabi", SHT_C6000_ATTRIBUTES,
+ return process_attributes (filedata, "c6xabi", SHT_C6000_ATTRIBUTES,
display_tic6x_attribute,
display_generic_attribute);
default:
- return process_attributes (file, "gnu", SHT_GNU_ATTRIBUTES,
+ return process_attributes (filedata, "gnu", SHT_GNU_ATTRIBUTES,
display_public_gnu_attributes,
display_generic_attribute);
}
}
static bfd_boolean
-get_file_header (FILE * file)
+get_file_header (Filedata * filedata)
{
/* Read in the identity array. */
- if (fread (elf_header.e_ident, EI_NIDENT, 1, file) != 1)
+ if (fread (filedata->file_header.e_ident, EI_NIDENT, 1, filedata->handle) != 1)
return FALSE;
/* Determine how to read the rest of the header. */
- switch (elf_header.e_ident[EI_DATA])
+ switch (filedata->file_header.e_ident[EI_DATA])
{
default:
case ELFDATANONE:
}
/* For now we only support 32 bit and 64 bit ELF files. */
- is_32bit_elf = (elf_header.e_ident[EI_CLASS] != ELFCLASS64);
+ is_32bit_elf = (filedata->file_header.e_ident[EI_CLASS] != ELFCLASS64);
/* Read in the rest of the header. */
if (is_32bit_elf)
{
Elf32_External_Ehdr ehdr32;
- if (fread (ehdr32.e_type, sizeof (ehdr32) - EI_NIDENT, 1, file) != 1)
+ if (fread (ehdr32.e_type, sizeof (ehdr32) - EI_NIDENT, 1, filedata->handle) != 1)
return FALSE;
- elf_header.e_type = BYTE_GET (ehdr32.e_type);
- elf_header.e_machine = BYTE_GET (ehdr32.e_machine);
- elf_header.e_version = BYTE_GET (ehdr32.e_version);
- elf_header.e_entry = BYTE_GET (ehdr32.e_entry);
- elf_header.e_phoff = BYTE_GET (ehdr32.e_phoff);
- elf_header.e_shoff = BYTE_GET (ehdr32.e_shoff);
- elf_header.e_flags = BYTE_GET (ehdr32.e_flags);
- elf_header.e_ehsize = BYTE_GET (ehdr32.e_ehsize);
- elf_header.e_phentsize = BYTE_GET (ehdr32.e_phentsize);
- elf_header.e_phnum = BYTE_GET (ehdr32.e_phnum);
- elf_header.e_shentsize = BYTE_GET (ehdr32.e_shentsize);
- elf_header.e_shnum = BYTE_GET (ehdr32.e_shnum);
- elf_header.e_shstrndx = BYTE_GET (ehdr32.e_shstrndx);
+ filedata->file_header.e_type = BYTE_GET (ehdr32.e_type);
+ filedata->file_header.e_machine = BYTE_GET (ehdr32.e_machine);
+ filedata->file_header.e_version = BYTE_GET (ehdr32.e_version);
+ filedata->file_header.e_entry = BYTE_GET (ehdr32.e_entry);
+ filedata->file_header.e_phoff = BYTE_GET (ehdr32.e_phoff);
+ filedata->file_header.e_shoff = BYTE_GET (ehdr32.e_shoff);
+ filedata->file_header.e_flags = BYTE_GET (ehdr32.e_flags);
+ filedata->file_header.e_ehsize = BYTE_GET (ehdr32.e_ehsize);
+ filedata->file_header.e_phentsize = BYTE_GET (ehdr32.e_phentsize);
+ filedata->file_header.e_phnum = BYTE_GET (ehdr32.e_phnum);
+ filedata->file_header.e_shentsize = BYTE_GET (ehdr32.e_shentsize);
+ filedata->file_header.e_shnum = BYTE_GET (ehdr32.e_shnum);
+ filedata->file_header.e_shstrndx = BYTE_GET (ehdr32.e_shstrndx);
}
else
{
return FALSE;
}
- if (fread (ehdr64.e_type, sizeof (ehdr64) - EI_NIDENT, 1, file) != 1)
+ if (fread (ehdr64.e_type, sizeof (ehdr64) - EI_NIDENT, 1, filedata->handle) != 1)
return FALSE;
- elf_header.e_type = BYTE_GET (ehdr64.e_type);
- elf_header.e_machine = BYTE_GET (ehdr64.e_machine);
- elf_header.e_version = BYTE_GET (ehdr64.e_version);
- elf_header.e_entry = BYTE_GET (ehdr64.e_entry);
- elf_header.e_phoff = BYTE_GET (ehdr64.e_phoff);
- elf_header.e_shoff = BYTE_GET (ehdr64.e_shoff);
- elf_header.e_flags = BYTE_GET (ehdr64.e_flags);
- elf_header.e_ehsize = BYTE_GET (ehdr64.e_ehsize);
- elf_header.e_phentsize = BYTE_GET (ehdr64.e_phentsize);
- elf_header.e_phnum = BYTE_GET (ehdr64.e_phnum);
- elf_header.e_shentsize = BYTE_GET (ehdr64.e_shentsize);
- elf_header.e_shnum = BYTE_GET (ehdr64.e_shnum);
- elf_header.e_shstrndx = BYTE_GET (ehdr64.e_shstrndx);
+ filedata->file_header.e_type = BYTE_GET (ehdr64.e_type);
+ filedata->file_header.e_machine = BYTE_GET (ehdr64.e_machine);
+ filedata->file_header.e_version = BYTE_GET (ehdr64.e_version);
+ filedata->file_header.e_entry = BYTE_GET (ehdr64.e_entry);
+ filedata->file_header.e_phoff = BYTE_GET (ehdr64.e_phoff);
+ filedata->file_header.e_shoff = BYTE_GET (ehdr64.e_shoff);
+ filedata->file_header.e_flags = BYTE_GET (ehdr64.e_flags);
+ filedata->file_header.e_ehsize = BYTE_GET (ehdr64.e_ehsize);
+ filedata->file_header.e_phentsize = BYTE_GET (ehdr64.e_phentsize);
+ filedata->file_header.e_phnum = BYTE_GET (ehdr64.e_phnum);
+ filedata->file_header.e_shentsize = BYTE_GET (ehdr64.e_shentsize);
+ filedata->file_header.e_shnum = BYTE_GET (ehdr64.e_shnum);
+ filedata->file_header.e_shstrndx = BYTE_GET (ehdr64.e_shstrndx);
}
- if (elf_header.e_shoff)
+ if (filedata->file_header.e_shoff)
{
/* There may be some extensions in the first section header. Don't
bomb if we can't read it. */
if (is_32bit_elf)
- get_32bit_section_headers (file, TRUE);
+ get_32bit_section_headers (filedata, TRUE);
else
- get_64bit_section_headers (file, TRUE);
+ get_64bit_section_headers (filedata, TRUE);
}
return TRUE;
}
+static void
+close_file (Filedata * filedata)
+{
+ if (filedata)
+ {
+ if (filedata->handle)
+ fclose (filedata->handle);
+ free (filedata);
+ }
+}
+
+void
+close_debug_file (void * data)
+{
+ close_file ((Filedata *) data);
+}
+
+static Filedata *
+open_file (const char * pathname)
+{
+ struct stat statbuf;
+ Filedata * filedata = NULL;
+
+ if (stat (pathname, & statbuf) < 0
+ || ! S_ISREG (statbuf.st_mode))
+ goto fail;
+
+ filedata = calloc (1, sizeof * filedata);
+ if (filedata == NULL)
+ goto fail;
+
+ filedata->handle = fopen (pathname, "rb");
+ if (filedata->handle == NULL)
+ goto fail;
+
+ filedata->file_size = (bfd_size_type) statbuf.st_size;
+ filedata->file_name = pathname;
+
+ if (! get_file_header (filedata))
+ goto fail;
+
+ if (filedata->file_header.e_shoff)
+ {
+ bfd_boolean res;
+
+ /* Read the section headers again, this time for real. */
+ if (is_32bit_elf)
+ res = get_32bit_section_headers (filedata, FALSE);
+ else
+ res = get_64bit_section_headers (filedata, FALSE);
+
+ if (!res)
+ goto fail;
+ }
+
+ return filedata;
+
+ fail:
+ if (filedata)
+ {
+ if (filedata->handle)
+ fclose (filedata->handle);
+ free (filedata);
+ }
+ return NULL;
+}
+
+void *
+open_debug_file (const char * pathname)
+{
+ return open_file (pathname);
+}
+
/* Process one ELF object file according to the command line options.
This file may actually be stored in an archive. The file is
positioned at the start of the ELF object. Returns TRUE if no
problems were encountered, FALSE otherwise. */
static bfd_boolean
-process_object (char * file_name, FILE * file)
+process_object (Filedata * filedata)
{
+ Filedata * separates;
unsigned int i;
bfd_boolean res = TRUE;
- if (! get_file_header (file))
+ if (! get_file_header (filedata))
{
- error (_("%s: Failed to read file header\n"), file_name);
+ error (_("%s: Failed to read file header\n"), filedata->file_name);
return FALSE;
}
/* Process the file. */
if (show_name)
- printf (_("\nFile: %s\n"), file_name);
+ printf (_("\nFile: %s\n"), filedata->file_name);
/* Initialise the dump_sects array from the cmdline_dump_sects array.
Note we do this even if cmdline_dump_sects is empty because we
must make sure that the dump_sets array is zeroed out before each
object file is processed. */
- if (num_dump_sects > num_cmdline_dump_sects)
- memset (dump_sects, 0, num_dump_sects * sizeof (* dump_sects));
+ if (filedata->num_dump_sects > cmdline.num_dump_sects)
+ memset (filedata->dump_sects, 0, filedata->num_dump_sects * sizeof (* filedata->dump_sects));
- if (num_cmdline_dump_sects > 0)
+ if (cmdline.num_dump_sects > 0)
{
- if (num_dump_sects == 0)
+ if (filedata->num_dump_sects == 0)
/* A sneaky way of allocating the dump_sects array. */
- request_dump_bynumber (num_cmdline_dump_sects, 0);
+ request_dump_bynumber (filedata, cmdline.num_dump_sects, 0);
- assert (num_dump_sects >= num_cmdline_dump_sects);
- memcpy (dump_sects, cmdline_dump_sects,
- num_cmdline_dump_sects * sizeof (* dump_sects));
+ assert (filedata->num_dump_sects >= cmdline.num_dump_sects);
+ memcpy (filedata->dump_sects, cmdline.dump_sects,
+ cmdline.num_dump_sects * sizeof (* filedata->dump_sects));
}
- if (! process_file_header ())
+ if (! process_file_header (filedata))
return FALSE;
- if (! process_section_headers (file))
+ if (! process_section_headers (filedata))
{
/* Without loaded section headers we cannot process lots of things. */
do_unwind = do_version = do_dump = do_arch = FALSE;
do_syms = do_dyn_syms = do_reloc = FALSE;
}
- if (! process_section_groups (file))
+ if (! process_section_groups (filedata))
/* Without loaded section groups we cannot process unwind. */
do_unwind = FALSE;
- if (process_program_headers (file))
- process_dynamic_section (file);
+ if (process_program_headers (filedata))
+ process_dynamic_section (filedata);
else
res = FALSE;
- if (! process_relocs (file))
+ if (! process_relocs (filedata))
res = FALSE;
- if (! process_unwind (file))
+ if (! process_unwind (filedata))
res = FALSE;
- if (! process_symbol_table (file))
+ if (! process_symbol_table (filedata))
res = FALSE;
- if (! process_syminfo (file))
+ if (! process_syminfo (filedata))
res = FALSE;
- if (! process_version_sections (file))
+ if (! process_version_sections (filedata))
res = FALSE;
- if (! process_section_contents (file))
+ separates = load_separate_debug_file (filedata, filedata->file_name);
+
+ if (! process_section_contents (filedata))
res = FALSE;
- if (! process_notes (file))
+ if (separates)
+ {
+ if (! process_section_headers (separates))
+ res = FALSE;
+ else if (! process_section_contents (separates))
+ res = FALSE;
+ }
+
+ if (! process_notes (filedata))
res = FALSE;
- if (! process_gnu_liblist (file))
+ if (! process_gnu_liblist (filedata))
res = FALSE;
- if (! process_arch_specific (file))
+ if (! process_arch_specific (filedata))
res = FALSE;
- if (program_headers)
- {
- free (program_headers);
- program_headers = NULL;
- }
+ free (filedata->program_headers);
+ filedata->program_headers = NULL;
- if (section_headers)
- {
- free (section_headers);
- section_headers = NULL;
- }
+ free (filedata->section_headers);
+ filedata->section_headers = NULL;
- if (string_table)
- {
- free (string_table);
- string_table = NULL;
- string_table_length = 0;
- }
+ free (filedata->string_table);
+ filedata->string_table = NULL;
+ filedata->string_table_length = 0;
if (dynamic_strings)
{
Returns TRUE upon success, FALSE otherwise. */
static bfd_boolean
-process_archive (char * file_name, FILE * file, bfd_boolean is_thin_archive)
+process_archive (Filedata * filedata, bfd_boolean is_thin_archive)
{
struct archive_info arch;
struct archive_info nested_arch;
nested_arch.sym_table = NULL;
nested_arch.longnames = NULL;
- if (setup_archive (&arch, file_name, file, is_thin_archive, do_archive_index) != 0)
+ if (setup_archive (&arch, filedata->file_name, filedata->handle,
+ is_thin_archive, do_archive_index) != 0)
{
ret = FALSE;
goto out;
if (do_archive_index)
{
if (arch.sym_table == NULL)
- error (_("%s: unable to dump the index as none was found\n"), file_name);
+ error (_("%s: unable to dump the index as none was found\n"), filedata->file_name);
else
{
unsigned long i, l;
unsigned long current_pos;
printf (_("Index of archive %s: (%lu entries, 0x%lx bytes in the symbol table)\n"),
- file_name, (unsigned long) arch.index_num, arch.sym_size);
- current_pos = ftell (file);
+ filedata->file_name, (unsigned long) arch.index_num, arch.sym_size);
+
+ current_pos = ftell (filedata->handle);
for (i = l = 0; i < arch.index_num; i++)
{
if (l >= arch.sym_size)
{
error (_("%s: end of the symbol table reached before the end of the index\n"),
- file_name);
+ filedata->file_name);
ret = FALSE;
break;
}
"but without corresponding entries in "
"the index table\n",
arch.sym_size - l),
- file_name, arch.sym_size - l);
+ filedata->file_name, arch.sym_size - l);
ret = FALSE;
}
- if (fseek (file, current_pos, SEEK_SET) != 0)
+ if (fseek (filedata->handle, current_pos, SEEK_SET) != 0)
{
- error (_("%s: failed to seek back to start of object files in the archive\n"), file_name);
+ error (_("%s: failed to seek back to start of object files in the archive\n"),
+ filedata->file_name);
ret = FALSE;
goto out;
}
char * qualified_name;
/* Read the next archive header. */
- if (fseek (file, arch.next_arhdr_offset, SEEK_SET) != 0)
+ if (fseek (filedata->handle, arch.next_arhdr_offset, SEEK_SET) != 0)
{
- error (_("%s: failed to seek to next archive header\n"), file_name);
+ error (_("%s: failed to seek to next archive header\n"), filedata->file_name);
return FALSE;
}
- got = fread (&arch.arhdr, 1, sizeof arch.arhdr, file);
+ got = fread (&arch.arhdr, 1, sizeof arch.arhdr, filedata->handle);
if (got != sizeof arch.arhdr)
{
if (got == 0)
break;
- error (_("%s: failed to read archive header\n"), file_name);
+ error (_("%s: failed to read archive header\n"), filedata->file_name);
ret = FALSE;
break;
}
name = get_archive_member_name (&arch, &nested_arch);
if (name == NULL)
{
- error (_("%s: bad archive file name\n"), file_name);
+ error (_("%s: bad archive file name\n"), filedata->file_name);
ret = FALSE;
break;
}
qualified_name = make_qualified_name (&arch, &nested_arch, name);
if (qualified_name == NULL)
{
- error (_("%s: bad archive file name\n"), file_name);
+ error (_("%s: bad archive file name\n"), filedata->file_name);
ret = FALSE;
break;
}
if (is_thin_archive && arch.nested_member_origin == 0)
{
/* This is a proxy for an external member of a thin archive. */
- FILE * member_file;
- char * member_file_name = adjust_relative_path (file_name, name, namelen);
+ Filedata * member_filedata;
+ char * member_file_name = adjust_relative_path
+ (filedata->file_name, name, namelen);
if (member_file_name == NULL)
{
break;
}
- member_file = fopen (member_file_name, "rb");
- if (member_file == NULL)
+ member_filedata = open_file (member_file_name);
+ if (member_filedata == NULL)
{
error (_("Input file '%s' is not readable.\n"), member_file_name);
free (member_file_name);
}
archive_file_offset = arch.nested_member_origin;
+ member_filedata->file_name = qualified_name;
- if (! process_object (qualified_name, member_file))
+ if (! process_object (member_filedata))
ret = FALSE;
- fclose (member_file);
+ close_file (member_filedata);
free (member_file_name);
}
else if (is_thin_archive)
{
+ Filedata thin_filedata = { 0 };
+
/* PR 15140: Allow for corrupt thin archives. */
if (nested_arch.file == NULL)
{
error (_("%s: contains corrupt thin archive: %s\n"),
- file_name, name);
+ filedata->file_name, name);
ret = FALSE;
break;
}
break;
}
- if (! process_object (qualified_name, nested_arch.file))
+ thin_filedata.handle = nested_arch.file;
+ thin_filedata.file_name = qualified_name;
+
+ if (! process_object (& thin_filedata))
ret = FALSE;
}
else
archive_file_offset = arch.next_arhdr_offset;
arch.next_arhdr_offset += archive_file_size;
- if (! process_object (qualified_name, file))
+ if (! process_object (filedata))
ret = FALSE;
}
- if (dump_sects != NULL)
+ if (filedata->dump_sects != NULL)
{
- free (dump_sects);
- dump_sects = NULL;
- num_dump_sects = 0;
+ free (filedata->dump_sects);
+ filedata->dump_sects = NULL;
+ filedata->num_dump_sects = 0;
}
free (qualified_name);
static bfd_boolean
process_file (char * file_name)
{
- FILE * file;
+ Filedata * filedata = NULL;
struct stat statbuf;
char armag[SARMAG];
bfd_boolean ret = TRUE;
return FALSE;
}
- file = fopen (file_name, "rb");
- if (file == NULL)
+ filedata = calloc (1, sizeof * filedata);
+ if (filedata == NULL)
+ {
+ error (_("Out of memory allocating file data structure\n"));
+ return FALSE;
+ }
+
+ filedata->file_name = file_name;
+ filedata->handle = fopen (file_name, "rb");
+ if (filedata->handle == NULL)
{
error (_("Input file '%s' is not readable.\n"), file_name);
+ free (filedata);
return FALSE;
}
- if (fread (armag, SARMAG, 1, file) != 1)
+ if (fread (armag, SARMAG, 1, filedata->handle) != 1)
{
error (_("%s: Failed to read file's magic number\n"), file_name);
- fclose (file);
+ fclose (filedata->handle);
+ free (filedata);
return FALSE;
}
- current_file_size = (bfd_size_type) statbuf.st_size;
+ filedata->file_size = (bfd_size_type) statbuf.st_size;
if (memcmp (armag, ARMAG, SARMAG) == 0)
{
- if (! process_archive (file_name, file, FALSE))
+ if (! process_archive (filedata, FALSE))
ret = FALSE;
}
else if (memcmp (armag, ARMAGT, SARMAG) == 0)
{
- if ( ! process_archive (file_name, file, TRUE))
+ if ( ! process_archive (filedata, TRUE))
ret = FALSE;
}
else
error (_("File %s is not an archive so its index cannot be displayed.\n"),
file_name);
- rewind (file);
+ rewind (filedata->handle);
archive_file_size = archive_file_offset = 0;
- if (! process_object (file_name, file))
+ if (! process_object (filedata))
ret = FALSE;
}
- fclose (file);
- current_file_size = 0;
+ fclose (filedata->handle);
+ free (filedata);
return ret;
}
}
/* Needed by the i386 disassembler. */
+
void
db_task_printsym (unsigned int addr)
{
expandargv (&argc, &argv);
- parse_args (argc, argv);
-
- if (num_dump_sects > 0)
- {
- /* Make a copy of the dump_sects array. */
- cmdline_dump_sects = (dump_type *)
- malloc (num_dump_sects * sizeof (* dump_sects));
- if (cmdline_dump_sects == NULL)
- error (_("Out of memory allocating dump request table.\n"));
- else
- {
- memcpy (cmdline_dump_sects, dump_sects,
- num_dump_sects * sizeof (* dump_sects));
- num_cmdline_dump_sects = num_dump_sects;
- }
- }
+ cmdline.file_name = "<cmdline>";
+ parse_args (& cmdline, argc, argv);
if (optind < (argc - 1))
show_name = TRUE;
if (! process_file (argv[optind++]))
err = TRUE;
- if (dump_sects != NULL)
- free (dump_sects);
- if (cmdline_dump_sects != NULL)
- free (cmdline_dump_sects);
+ if (cmdline.dump_sects != NULL)
+ free (cmdline.dump_sects);
return err ? EXIT_FAILURE : EXIT_SUCCESS;
}