97d40d3d4a
us up to version 2.17.50.20070703, at the last GPLv2 commit. Amongst others, this added upstream support for some FreeBSD-specific things that we previously had to manually hack in, such as the OSABI label support, and so on. There are also quite a number of new files, some for cpu's (e.g. SPU) that we may or may not be interested in, but those can be cleaned up later on, if needed.
1004 lines
25 KiB
C
1004 lines
25 KiB
C
/* BFD backend for Extended Tektronix Hex Format objects.
|
|
Copyright 1992, 1993, 1994, 1995, 1996, 1998, 1999, 2000, 2001, 2002,
|
|
2003, 2004, 2007 Free Software Foundation, Inc.
|
|
Written by Steve Chamberlain of Cygnus Support <sac@cygnus.com>.
|
|
|
|
This file is part of BFD, the Binary File Descriptor library.
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
it under the terms of the GNU General Public License as published by
|
|
the Free Software Foundation; either version 2 of the License, or
|
|
(at your option) any later version.
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program; if not, write to the Free Software
|
|
Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA 02110-1301, USA. */
|
|
|
|
/* SUBSECTION
|
|
Tektronix Hex Format handling
|
|
|
|
DESCRIPTION
|
|
|
|
Tek Hex records can hold symbols and data, but not
|
|
relocations. Their main application is communication with
|
|
devices like PROM programmers and ICE equipment.
|
|
|
|
It seems that the sections are described as being really big,
|
|
the example I have says that the text section is 0..ffffffff.
|
|
BFD would barf with this, many apps would try to alloc 4GB to
|
|
read in the file.
|
|
|
|
Tex Hex may contain many sections, but the data which comes in
|
|
has no tag saying which section it belongs to, so we create
|
|
one section for each block of data, called "blknnnn" which we
|
|
stick all the data into.
|
|
|
|
TekHex may come out of order and there is no header, so an
|
|
initial scan is required to discover the minimum and maximum
|
|
addresses used to create the vma and size of the sections we
|
|
create.
|
|
We read in the data into pages of CHUNK_MASK+1 size and read
|
|
them out from that whenever we need to.
|
|
|
|
Any number of sections may be created for output, we save them
|
|
up and output them when it's time to close the bfd.
|
|
|
|
A TekHex record looks like:
|
|
EXAMPLE
|
|
%<block length><type><checksum><stuff><cr>
|
|
|
|
DESCRIPTION
|
|
Where
|
|
o length
|
|
is the number of bytes in the record not including the % sign.
|
|
o type
|
|
is one of:
|
|
3) symbol record
|
|
6) data record
|
|
8) termination record
|
|
|
|
The data can come out of order, and may be discontigous. This is a
|
|
serial protocol, so big files are unlikely, so we keep a list of 8k chunks. */
|
|
|
|
#include "sysdep.h"
|
|
#include "bfd.h"
|
|
#include "libbfd.h"
|
|
#include "libiberty.h"
|
|
|
|
typedef struct
|
|
{
|
|
bfd_vma low;
|
|
bfd_vma high;
|
|
} addr_range_type;
|
|
|
|
typedef struct tekhex_symbol_struct
|
|
{
|
|
asymbol symbol;
|
|
struct tekhex_symbol_struct *prev;
|
|
} tekhex_symbol_type;
|
|
|
|
static const char digs[] = "0123456789ABCDEF";
|
|
|
|
static char sum_block[256];
|
|
|
|
#define NOT_HEX 20
|
|
#define NIBBLE(x) hex_value(x)
|
|
#define HEX(buffer) ((NIBBLE ((buffer)[0]) << 4) + NIBBLE ((buffer)[1]))
|
|
#define ISHEX(x) hex_p(x)
|
|
#define TOHEX(d, x) \
|
|
(d)[1] = digs[(x) & 0xf]; \
|
|
(d)[0] = digs[((x)>>4)&0xf];
|
|
|
|
/* Here's an example
|
|
%3A6C6480004E56FFFC4E717063B0AEFFFC6D0652AEFFFC60F24E5E4E75
|
|
%1B3709T_SEGMENT1108FFFFFFFF
|
|
%2B3AB9T_SEGMENT7Dgcc_compiled$1087hello$c10
|
|
%373829T_SEGMENT80int$t1$r1$$214741080char$t2$r2$0$12710
|
|
%373769T_SEGMENT80long$int$t3$r1$$1080unsigned$int$t4$10
|
|
%373CA9T_SEGMENT80long$unsigned$in1080short$int$t6$r1$10
|
|
%373049T_SEGMENT80long$long$int$t71080short$unsigned$i10
|
|
%373A29T_SEGMENT80long$long$unsign1080signed$char$t10$10
|
|
%373D69T_SEGMENT80unsigned$char$t11080float$t12$r1$4$010
|
|
%373D19T_SEGMENT80double$t13$r1$8$1080long$double$t14$10
|
|
%2734D9T_SEGMENT8Bvoid$t15$151035_main10
|
|
%2F3CA9T_SEGMENT81$1081$1681$1E81$21487main$F110
|
|
%2832F9T_SEGMENT83i$18FFFFFFFC81$1481$214
|
|
%07 8 10 10
|
|
|
|
explanation:
|
|
%3A6C6480004E56FFFC4E717063B0AEFFFC6D0652AEFFFC60F24E5E4E75
|
|
^ ^^ ^ ^-data
|
|
| || +------ 4 char integer 0x8000
|
|
| |+-------- checksum
|
|
| +--------- type 6 (data record)
|
|
+----------- length 3a chars
|
|
<---------------------- 3a (58 chars) ------------------->
|
|
|
|
%1B3709T_SEGMENT1108FFFFFFFF
|
|
^ ^^ ^- 8 character integer 0xffffffff
|
|
| |+- 1 character integer 0
|
|
| +-- type 1 symbol (section definition)
|
|
+------------ 9 char symbol T_SEGMENT
|
|
|
|
%2B3AB9T_SEGMENT7Dgcc_compiled$1087hello$c10
|
|
%373829T_SEGMENT80int$t1$r1$$214741080char$t2$r2$0$12710
|
|
%373769T_SEGMENT80long$int$t3$r1$$1080unsigned$int$t4$10
|
|
%373CA9T_SEGMENT80long$unsigned$in1080short$int$t6$r1$10
|
|
%373049T_SEGMENT80long$long$int$t71080short$unsigned$i10
|
|
%373A29T_SEGMENT80long$long$unsign1080signed$char$t10$10
|
|
%373D69T_SEGMENT80unsigned$char$t11080float$t12$r1$4$010
|
|
%373D19T_SEGMENT80double$t13$r1$8$1080long$double$t14$10
|
|
%2734D9T_SEGMENT8Bvoid$t15$151035_main10
|
|
%2F3CA9T_SEGMENT81$1081$1681$1E81$21487main$F110
|
|
%2832F9T_SEGMENT83i$18FFFFFFFC81$1481$214
|
|
%0781010
|
|
|
|
Turns into
|
|
sac@thepub$ ./objdump -dx -m m68k f
|
|
|
|
f: file format tekhex
|
|
-----x--- 9/55728 -134219416 Sep 29 15:13 1995 f
|
|
architecture: UNKNOWN!, flags 0x00000010:
|
|
HAS_SYMS
|
|
start address 0x00000000
|
|
SECTION 0 [D00000000] : size 00020000 vma 00000000 align 2**0
|
|
ALLOC, LOAD
|
|
SECTION 1 [D00008000] : size 00002001 vma 00008000 align 2**0
|
|
|
|
SECTION 2 [T_SEGMENT] : size ffffffff vma 00000000 align 2**0
|
|
|
|
SYMBOL TABLE:
|
|
00000000 g T_SEGMENT gcc_compiled$
|
|
00000000 g T_SEGMENT hello$c
|
|
00000000 g T_SEGMENT int$t1$r1$$21474
|
|
00000000 g T_SEGMENT char$t2$r2$0$127
|
|
00000000 g T_SEGMENT long$int$t3$r1$$
|
|
00000000 g T_SEGMENT unsigned$int$t4$
|
|
00000000 g T_SEGMENT long$unsigned$in
|
|
00000000 g T_SEGMENT short$int$t6$r1$
|
|
00000000 g T_SEGMENT long$long$int$t7
|
|
00000000 g T_SEGMENT short$unsigned$i
|
|
00000000 g T_SEGMENT long$long$unsign
|
|
00000000 g T_SEGMENT signed$char$t10$
|
|
00000000 g T_SEGMENT unsigned$char$t1
|
|
00000000 g T_SEGMENT float$t12$r1$4$0
|
|
00000000 g T_SEGMENT double$t13$r1$8$
|
|
00000000 g T_SEGMENT long$double$t14$
|
|
00000000 g T_SEGMENT void$t15$15
|
|
00000000 g T_SEGMENT _main
|
|
00000000 g T_SEGMENT $
|
|
00000000 g T_SEGMENT $
|
|
00000000 g T_SEGMENT $
|
|
00000010 g T_SEGMENT $
|
|
00000000 g T_SEGMENT main$F1
|
|
fcffffff g T_SEGMENT i$1
|
|
00000000 g T_SEGMENT $
|
|
00000010 g T_SEGMENT $
|
|
|
|
RELOCATION RECORDS FOR [D00000000]: (none)
|
|
|
|
RELOCATION RECORDS FOR [D00008000]: (none)
|
|
|
|
RELOCATION RECORDS FOR [T_SEGMENT]: (none)
|
|
|
|
Disassembly of section D00000000:
|
|
...
|
|
00008000 ($+)7ff0 linkw fp,#-4
|
|
00008004 ($+)7ff4 nop
|
|
00008006 ($+)7ff6 movel #99,d0
|
|
00008008 ($+)7ff8 cmpl fp@(-4),d0
|
|
0000800c ($+)7ffc blts 00008014 ($+)8004
|
|
0000800e ($+)7ffe addql #1,fp@(-4)
|
|
00008012 ($+)8002 bras 00008006 ($+)7ff6
|
|
00008014 ($+)8004 unlk fp
|
|
00008016 ($+)8006 rts
|
|
... */
|
|
|
|
static void
|
|
tekhex_init (void)
|
|
{
|
|
unsigned int i;
|
|
static bfd_boolean inited = FALSE;
|
|
int val;
|
|
|
|
if (! inited)
|
|
{
|
|
inited = TRUE;
|
|
hex_init ();
|
|
val = 0;
|
|
for (i = 0; i < 10; i++)
|
|
sum_block[i + '0'] = val++;
|
|
|
|
for (i = 'A'; i <= 'Z'; i++)
|
|
sum_block[i] = val++;
|
|
|
|
sum_block['$'] = val++;
|
|
sum_block['%'] = val++;
|
|
sum_block['.'] = val++;
|
|
sum_block['_'] = val++;
|
|
for (i = 'a'; i <= 'z'; i++)
|
|
sum_block[i] = val++;
|
|
}
|
|
}
|
|
|
|
/* The maximum number of bytes on a line is FF. */
|
|
#define MAXCHUNK 0xff
|
|
/* The number of bytes we fit onto a line on output. */
|
|
#define CHUNK 21
|
|
|
|
/* We cannot output our tekhexords as we see them, we have to glue them
|
|
together, this is done in this structure : */
|
|
|
|
struct tekhex_data_list_struct
|
|
{
|
|
unsigned char *data;
|
|
bfd_vma where;
|
|
bfd_size_type size;
|
|
struct tekhex_data_list_struct *next;
|
|
|
|
};
|
|
typedef struct tekhex_data_list_struct tekhex_data_list_type;
|
|
|
|
#define CHUNK_MASK 0x1fff
|
|
|
|
struct data_struct
|
|
{
|
|
char chunk_data[CHUNK_MASK + 1];
|
|
char chunk_init[CHUNK_MASK + 1];
|
|
bfd_vma vma;
|
|
struct data_struct *next;
|
|
};
|
|
|
|
typedef struct tekhex_data_struct
|
|
{
|
|
tekhex_data_list_type *head;
|
|
unsigned int type;
|
|
struct tekhex_symbol_struct *symbols;
|
|
struct data_struct *data;
|
|
} tdata_type;
|
|
|
|
#define enda(x) (x->vma + x->size)
|
|
|
|
static bfd_boolean
|
|
getvalue (char **srcp, bfd_vma *valuep)
|
|
{
|
|
char *src = *srcp;
|
|
bfd_vma value = 0;
|
|
unsigned int len;
|
|
|
|
if (!ISHEX (*src))
|
|
return FALSE;
|
|
|
|
len = hex_value (*src++);
|
|
if (len == 0)
|
|
len = 16;
|
|
while (len--)
|
|
{
|
|
if (!ISHEX (*src))
|
|
return FALSE;
|
|
value = value << 4 | hex_value (*src++);
|
|
}
|
|
|
|
*srcp = src;
|
|
*valuep = value;
|
|
return TRUE;
|
|
}
|
|
|
|
static bfd_boolean
|
|
getsym (char *dstp, char **srcp, unsigned int *lenp)
|
|
{
|
|
char *src = *srcp;
|
|
unsigned int i;
|
|
unsigned int len;
|
|
|
|
if (!ISHEX (*src))
|
|
return FALSE;
|
|
|
|
len = hex_value (*src++);
|
|
if (len == 0)
|
|
len = 16;
|
|
for (i = 0; i < len; i++)
|
|
dstp[i] = src[i];
|
|
dstp[i] = 0;
|
|
*srcp = src + i;
|
|
*lenp = len;
|
|
return TRUE;
|
|
}
|
|
|
|
static struct data_struct *
|
|
find_chunk (bfd *abfd, bfd_vma vma)
|
|
{
|
|
struct data_struct *d = abfd->tdata.tekhex_data->data;
|
|
|
|
vma &= ~CHUNK_MASK;
|
|
while (d && (d->vma) != vma)
|
|
d = d->next;
|
|
|
|
if (!d)
|
|
{
|
|
/* No chunk for this address, so make one up. */
|
|
d = bfd_zalloc (abfd, (bfd_size_type) sizeof (struct data_struct));
|
|
|
|
if (!d)
|
|
return NULL;
|
|
|
|
d->next = abfd->tdata.tekhex_data->data;
|
|
d->vma = vma;
|
|
abfd->tdata.tekhex_data->data = d;
|
|
}
|
|
return d;
|
|
}
|
|
|
|
static void
|
|
insert_byte (bfd *abfd, int value, bfd_vma addr)
|
|
{
|
|
/* Find the chunk that this byte needs and put it in. */
|
|
struct data_struct *d = find_chunk (abfd, addr);
|
|
|
|
d->chunk_data[addr & CHUNK_MASK] = value;
|
|
d->chunk_init[addr & CHUNK_MASK] = 1;
|
|
}
|
|
|
|
/* The first pass is to find the names of all the sections, and see
|
|
how big the data is. */
|
|
|
|
static bfd_boolean
|
|
first_phase (bfd *abfd, int type, char *src)
|
|
{
|
|
asection *section = bfd_abs_section_ptr;
|
|
unsigned int len;
|
|
bfd_vma val;
|
|
char sym[17]; /* A symbol can only be 16chars long. */
|
|
|
|
switch (type)
|
|
{
|
|
case '6':
|
|
/* Data record - read it and store it. */
|
|
{
|
|
bfd_vma addr;
|
|
|
|
if (!getvalue (&src, &addr))
|
|
return FALSE;
|
|
|
|
while (*src)
|
|
{
|
|
insert_byte (abfd, HEX (src), addr);
|
|
src += 2;
|
|
addr++;
|
|
}
|
|
}
|
|
|
|
return TRUE;
|
|
case '3':
|
|
/* Symbol record, read the segment. */
|
|
if (!getsym (sym, &src, &len))
|
|
return FALSE;
|
|
section = bfd_get_section_by_name (abfd, sym);
|
|
if (section == NULL)
|
|
{
|
|
char *n = bfd_alloc (abfd, (bfd_size_type) len + 1);
|
|
|
|
if (!n)
|
|
return FALSE;
|
|
memcpy (n, sym, len + 1);
|
|
section = bfd_make_section (abfd, n);
|
|
}
|
|
while (*src)
|
|
{
|
|
switch (*src)
|
|
{
|
|
case '1': /* Section range. */
|
|
src++;
|
|
if (!getvalue (&src, §ion->vma))
|
|
return FALSE;
|
|
if (!getvalue (&src, &val))
|
|
return FALSE;
|
|
section->size = val - section->vma;
|
|
section->flags = SEC_HAS_CONTENTS | SEC_LOAD | SEC_ALLOC;
|
|
break;
|
|
case '0':
|
|
case '2':
|
|
case '3':
|
|
case '4':
|
|
case '6':
|
|
case '7':
|
|
case '8':
|
|
/* Symbols, add to section. */
|
|
{
|
|
bfd_size_type amt = sizeof (tekhex_symbol_type);
|
|
tekhex_symbol_type *new = bfd_alloc (abfd, amt);
|
|
char stype = (*src);
|
|
|
|
if (!new)
|
|
return FALSE;
|
|
new->symbol.the_bfd = abfd;
|
|
src++;
|
|
abfd->symcount++;
|
|
abfd->flags |= HAS_SYMS;
|
|
new->prev = abfd->tdata.tekhex_data->symbols;
|
|
abfd->tdata.tekhex_data->symbols = new;
|
|
if (!getsym (sym, &src, &len))
|
|
return FALSE;
|
|
new->symbol.name = bfd_alloc (abfd, (bfd_size_type) len + 1);
|
|
if (!new->symbol.name)
|
|
return FALSE;
|
|
memcpy ((char *) (new->symbol.name), sym, len + 1);
|
|
new->symbol.section = section;
|
|
if (stype <= '4')
|
|
new->symbol.flags = (BSF_GLOBAL | BSF_EXPORT);
|
|
else
|
|
new->symbol.flags = BSF_LOCAL;
|
|
if (!getvalue (&src, &val))
|
|
return FALSE;
|
|
new->symbol.value = val - section->vma;
|
|
break;
|
|
}
|
|
default:
|
|
return FALSE;
|
|
}
|
|
}
|
|
}
|
|
|
|
return TRUE;
|
|
}
|
|
|
|
/* Pass over a tekhex, calling one of the above functions on each
|
|
record. */
|
|
|
|
static bfd_boolean
|
|
pass_over (bfd *abfd, bfd_boolean (*func) (bfd *, int, char *))
|
|
{
|
|
unsigned int chars_on_line;
|
|
bfd_boolean eof = FALSE;
|
|
|
|
/* To the front of the file. */
|
|
if (bfd_seek (abfd, (file_ptr) 0, SEEK_SET) != 0)
|
|
return FALSE;
|
|
while (! eof)
|
|
{
|
|
char src[MAXCHUNK];
|
|
char type;
|
|
|
|
/* Find first '%'. */
|
|
eof = (bfd_boolean) (bfd_bread (src, (bfd_size_type) 1, abfd) != 1);
|
|
while (*src != '%' && !eof)
|
|
eof = (bfd_boolean) (bfd_bread (src, (bfd_size_type) 1, abfd) != 1);
|
|
|
|
if (eof)
|
|
break;
|
|
|
|
/* Fetch the type and the length and the checksum. */
|
|
if (bfd_bread (src, (bfd_size_type) 5, abfd) != 5)
|
|
return FALSE;
|
|
|
|
type = src[2];
|
|
|
|
if (!ISHEX (src[0]) || !ISHEX (src[1]))
|
|
break;
|
|
|
|
/* Already read five chars. */
|
|
chars_on_line = HEX (src) - 5;
|
|
|
|
if (chars_on_line >= MAXCHUNK)
|
|
return FALSE;
|
|
|
|
if (bfd_bread (src, (bfd_size_type) chars_on_line, abfd) != chars_on_line)
|
|
return FALSE;
|
|
|
|
/* Put a null at the end. */
|
|
src[chars_on_line] = 0;
|
|
|
|
if (!func (abfd, type, src))
|
|
return FALSE;
|
|
}
|
|
|
|
return TRUE;
|
|
}
|
|
|
|
static long
|
|
tekhex_canonicalize_symtab (bfd *abfd, asymbol **table)
|
|
{
|
|
tekhex_symbol_type *p = abfd->tdata.tekhex_data->symbols;
|
|
unsigned int c = bfd_get_symcount (abfd);
|
|
|
|
table[c] = 0;
|
|
while (p)
|
|
{
|
|
table[--c] = &(p->symbol);
|
|
p = p->prev;
|
|
}
|
|
|
|
return bfd_get_symcount (abfd);
|
|
}
|
|
|
|
static long
|
|
tekhex_get_symtab_upper_bound (bfd *abfd)
|
|
{
|
|
return (abfd->symcount + 1) * (sizeof (struct tekhex_asymbol_struct *));
|
|
|
|
}
|
|
|
|
static bfd_boolean
|
|
tekhex_mkobject (bfd *abfd)
|
|
{
|
|
tdata_type *tdata;
|
|
|
|
tdata = bfd_alloc (abfd, (bfd_size_type) sizeof (tdata_type));
|
|
if (!tdata)
|
|
return FALSE;
|
|
abfd->tdata.tekhex_data = tdata;
|
|
tdata->type = 1;
|
|
tdata->head = NULL;
|
|
tdata->symbols = NULL;
|
|
tdata->data = NULL;
|
|
return TRUE;
|
|
}
|
|
|
|
/* Return TRUE if the file looks like it's in TekHex format. Just look
|
|
for a percent sign and some hex digits. */
|
|
|
|
static const bfd_target *
|
|
tekhex_object_p (bfd *abfd)
|
|
{
|
|
char b[4];
|
|
|
|
tekhex_init ();
|
|
|
|
if (bfd_seek (abfd, (file_ptr) 0, SEEK_SET) != 0
|
|
|| bfd_bread (b, (bfd_size_type) 4, abfd) != 4)
|
|
return NULL;
|
|
|
|
if (b[0] != '%' || !ISHEX (b[1]) || !ISHEX (b[2]) || !ISHEX (b[3]))
|
|
return NULL;
|
|
|
|
tekhex_mkobject (abfd);
|
|
|
|
if (!pass_over (abfd, first_phase))
|
|
return NULL;
|
|
|
|
return abfd->xvec;
|
|
}
|
|
|
|
static void
|
|
move_section_contents (bfd *abfd,
|
|
asection *section,
|
|
const void * locationp,
|
|
file_ptr offset,
|
|
bfd_size_type count,
|
|
bfd_boolean get)
|
|
{
|
|
bfd_vma addr;
|
|
char *location = (char *) locationp;
|
|
bfd_vma prev_number = 1; /* Nothing can have this as a high bit. */
|
|
struct data_struct *d = NULL;
|
|
|
|
BFD_ASSERT (offset == 0);
|
|
for (addr = section->vma; count != 0; count--, addr++)
|
|
{
|
|
/* Get high bits of address. */
|
|
bfd_vma chunk_number = addr & ~(bfd_vma) CHUNK_MASK;
|
|
bfd_vma low_bits = addr & CHUNK_MASK;
|
|
|
|
if (chunk_number != prev_number)
|
|
/* Different chunk, so move pointer. */
|
|
d = find_chunk (abfd, chunk_number);
|
|
|
|
if (get)
|
|
{
|
|
if (d->chunk_init[low_bits])
|
|
*location = d->chunk_data[low_bits];
|
|
else
|
|
*location = 0;
|
|
}
|
|
else
|
|
{
|
|
d->chunk_data[low_bits] = *location;
|
|
d->chunk_init[low_bits] = (*location != 0);
|
|
}
|
|
|
|
location++;
|
|
}
|
|
}
|
|
|
|
static bfd_boolean
|
|
tekhex_get_section_contents (bfd *abfd,
|
|
asection *section,
|
|
void * locationp,
|
|
file_ptr offset,
|
|
bfd_size_type count)
|
|
{
|
|
if (section->flags & (SEC_LOAD | SEC_ALLOC))
|
|
{
|
|
move_section_contents (abfd, section, locationp, offset, count, TRUE);
|
|
return TRUE;
|
|
}
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
static bfd_boolean
|
|
tekhex_set_arch_mach (bfd *abfd,
|
|
enum bfd_architecture arch,
|
|
unsigned long machine)
|
|
{
|
|
return bfd_default_set_arch_mach (abfd, arch, machine);
|
|
}
|
|
|
|
/* We have to save up all the Tekhexords for a splurge before output. */
|
|
|
|
static bfd_boolean
|
|
tekhex_set_section_contents (bfd *abfd,
|
|
sec_ptr section,
|
|
const void * locationp,
|
|
file_ptr offset,
|
|
bfd_size_type bytes_to_do)
|
|
{
|
|
if (! abfd->output_has_begun)
|
|
{
|
|
/* The first time around, allocate enough sections to hold all the chunks. */
|
|
asection *s = abfd->sections;
|
|
bfd_vma vma;
|
|
|
|
for (s = abfd->sections; s; s = s->next)
|
|
{
|
|
if (s->flags & SEC_LOAD)
|
|
{
|
|
for (vma = s->vma & ~(bfd_vma) CHUNK_MASK;
|
|
vma < s->vma + s->size;
|
|
vma += CHUNK_MASK)
|
|
find_chunk (abfd, vma);
|
|
}
|
|
}
|
|
}
|
|
|
|
if (section->flags & (SEC_LOAD | SEC_ALLOC))
|
|
{
|
|
move_section_contents (abfd, section, locationp, offset, bytes_to_do,
|
|
FALSE);
|
|
return TRUE;
|
|
}
|
|
|
|
return FALSE;
|
|
}
|
|
|
|
static void
|
|
writevalue (char **dst, bfd_vma value)
|
|
{
|
|
char *p = *dst;
|
|
int len;
|
|
int shift;
|
|
|
|
for (len = 8, shift = 28; shift; shift -= 4, len--)
|
|
{
|
|
if ((value >> shift) & 0xf)
|
|
{
|
|
*p++ = len + '0';
|
|
while (len)
|
|
{
|
|
*p++ = digs[(value >> shift) & 0xf];
|
|
shift -= 4;
|
|
len--;
|
|
}
|
|
*dst = p;
|
|
return;
|
|
|
|
}
|
|
}
|
|
*p++ = '1';
|
|
*p++ = '0';
|
|
*dst = p;
|
|
}
|
|
|
|
static void
|
|
writesym (char **dst, const char *sym)
|
|
{
|
|
char *p = *dst;
|
|
int len = (sym ? strlen (sym) : 0);
|
|
|
|
if (len >= 16)
|
|
{
|
|
*p++ = '0';
|
|
len = 16;
|
|
}
|
|
else
|
|
{
|
|
if (len == 0)
|
|
{
|
|
*p++ = '1';
|
|
sym = "$";
|
|
len = 1;
|
|
}
|
|
else
|
|
*p++ = digs[len];
|
|
}
|
|
|
|
while (len--)
|
|
*p++ = *sym++;
|
|
|
|
*dst = p;
|
|
}
|
|
|
|
static void
|
|
out (bfd *abfd, int type, char *start, char *end)
|
|
{
|
|
int sum = 0;
|
|
char *s;
|
|
char front[6];
|
|
bfd_size_type wrlen;
|
|
|
|
front[0] = '%';
|
|
TOHEX (front + 1, end - start + 5);
|
|
front[3] = type;
|
|
|
|
for (s = start; s < end; s++)
|
|
sum += sum_block[(unsigned char) *s];
|
|
|
|
sum += sum_block[(unsigned char) front[1]]; /* Length. */
|
|
sum += sum_block[(unsigned char) front[2]];
|
|
sum += sum_block[(unsigned char) front[3]]; /* Type. */
|
|
TOHEX (front + 4, sum);
|
|
if (bfd_bwrite (front, (bfd_size_type) 6, abfd) != 6)
|
|
abort ();
|
|
end[0] = '\n';
|
|
wrlen = end - start + 1;
|
|
if (bfd_bwrite (start, wrlen, abfd) != wrlen)
|
|
abort ();
|
|
}
|
|
|
|
static bfd_boolean
|
|
tekhex_write_object_contents (bfd *abfd)
|
|
{
|
|
char buffer[100];
|
|
asymbol **p;
|
|
asection *s;
|
|
struct data_struct *d;
|
|
|
|
tekhex_init ();
|
|
|
|
/* And the raw data. */
|
|
for (d = abfd->tdata.tekhex_data->data;
|
|
d != NULL;
|
|
d = d->next)
|
|
{
|
|
int low;
|
|
|
|
const int span = 32;
|
|
int addr;
|
|
|
|
/* Write it in blocks of 32 bytes. */
|
|
for (addr = 0; addr < CHUNK_MASK + 1; addr += span)
|
|
{
|
|
int need = 0;
|
|
|
|
/* Check to see if necessary. */
|
|
for (low = 0; !need && low < span; low++)
|
|
if (d->chunk_init[addr + low])
|
|
need = 1;
|
|
|
|
if (need)
|
|
{
|
|
char *dst = buffer;
|
|
|
|
writevalue (&dst, addr + d->vma);
|
|
for (low = 0; low < span; low++)
|
|
{
|
|
TOHEX (dst, d->chunk_data[addr + low]);
|
|
dst += 2;
|
|
}
|
|
out (abfd, '6', buffer, dst);
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Write all the section headers for the sections. */
|
|
for (s = abfd->sections; s != NULL; s = s->next)
|
|
{
|
|
char *dst = buffer;
|
|
|
|
writesym (&dst, s->name);
|
|
*dst++ = '1';
|
|
writevalue (&dst, s->vma);
|
|
writevalue (&dst, s->vma + s->size);
|
|
out (abfd, '3', buffer, dst);
|
|
}
|
|
|
|
/* And the symbols. */
|
|
if (abfd->outsymbols)
|
|
{
|
|
for (p = abfd->outsymbols; *p; p++)
|
|
{
|
|
int section_code = bfd_decode_symclass (*p);
|
|
|
|
if (section_code != '?')
|
|
{
|
|
/* Do not include debug symbols. */
|
|
asymbol *sym = *p;
|
|
char *dst = buffer;
|
|
|
|
writesym (&dst, sym->section->name);
|
|
|
|
switch (section_code)
|
|
{
|
|
case 'A':
|
|
*dst++ = '2';
|
|
break;
|
|
case 'a':
|
|
*dst++ = '6';
|
|
break;
|
|
case 'D':
|
|
case 'B':
|
|
case 'O':
|
|
*dst++ = '4';
|
|
break;
|
|
case 'd':
|
|
case 'b':
|
|
case 'o':
|
|
*dst++ = '8';
|
|
break;
|
|
case 'T':
|
|
*dst++ = '3';
|
|
break;
|
|
case 't':
|
|
*dst++ = '7';
|
|
break;
|
|
case 'C':
|
|
case 'U':
|
|
bfd_set_error (bfd_error_wrong_format);
|
|
return FALSE;
|
|
}
|
|
|
|
writesym (&dst, sym->name);
|
|
writevalue (&dst, sym->value + sym->section->vma);
|
|
out (abfd, '3', buffer, dst);
|
|
}
|
|
}
|
|
}
|
|
|
|
/* And the terminator. */
|
|
if (bfd_bwrite ("%0781010\n", (bfd_size_type) 9, abfd) != 9)
|
|
abort ();
|
|
return TRUE;
|
|
}
|
|
|
|
static int
|
|
tekhex_sizeof_headers (bfd *abfd ATTRIBUTE_UNUSED,
|
|
struct bfd_link_info *info ATTRIBUTE_UNUSED)
|
|
{
|
|
return 0;
|
|
}
|
|
|
|
static asymbol *
|
|
tekhex_make_empty_symbol (bfd *abfd)
|
|
{
|
|
bfd_size_type amt = sizeof (struct tekhex_symbol_struct);
|
|
tekhex_symbol_type *new = bfd_zalloc (abfd, amt);
|
|
|
|
if (!new)
|
|
return NULL;
|
|
new->symbol.the_bfd = abfd;
|
|
new->prev = NULL;
|
|
return &(new->symbol);
|
|
}
|
|
|
|
static void
|
|
tekhex_get_symbol_info (bfd *abfd ATTRIBUTE_UNUSED,
|
|
asymbol *symbol,
|
|
symbol_info *ret)
|
|
{
|
|
bfd_symbol_info (symbol, ret);
|
|
}
|
|
|
|
static void
|
|
tekhex_print_symbol (bfd *abfd,
|
|
void * filep,
|
|
asymbol *symbol,
|
|
bfd_print_symbol_type how)
|
|
{
|
|
FILE *file = (FILE *) filep;
|
|
|
|
switch (how)
|
|
{
|
|
case bfd_print_symbol_name:
|
|
fprintf (file, "%s", symbol->name);
|
|
break;
|
|
case bfd_print_symbol_more:
|
|
break;
|
|
|
|
case bfd_print_symbol_all:
|
|
{
|
|
const char *section_name = symbol->section->name;
|
|
|
|
bfd_print_symbol_vandf (abfd, (void *) file, symbol);
|
|
|
|
fprintf (file, " %-5s %s",
|
|
section_name, symbol->name);
|
|
}
|
|
}
|
|
}
|
|
|
|
#define tekhex_close_and_cleanup _bfd_generic_close_and_cleanup
|
|
#define tekhex_bfd_free_cached_info _bfd_generic_bfd_free_cached_info
|
|
#define tekhex_new_section_hook _bfd_generic_new_section_hook
|
|
#define tekhex_bfd_is_target_special_symbol ((bfd_boolean (*) (bfd *, asymbol *)) bfd_false)
|
|
#define tekhex_bfd_is_local_label_name bfd_generic_is_local_label_name
|
|
#define tekhex_get_lineno _bfd_nosymbols_get_lineno
|
|
#define tekhex_find_nearest_line _bfd_nosymbols_find_nearest_line
|
|
#define tekhex_find_inliner_info _bfd_nosymbols_find_inliner_info
|
|
#define tekhex_bfd_make_debug_symbol _bfd_nosymbols_bfd_make_debug_symbol
|
|
#define tekhex_read_minisymbols _bfd_generic_read_minisymbols
|
|
#define tekhex_minisymbol_to_symbol _bfd_generic_minisymbol_to_symbol
|
|
#define tekhex_bfd_get_relocated_section_contents bfd_generic_get_relocated_section_contents
|
|
#define tekhex_bfd_relax_section bfd_generic_relax_section
|
|
#define tekhex_bfd_gc_sections bfd_generic_gc_sections
|
|
#define tekhex_bfd_merge_sections bfd_generic_merge_sections
|
|
#define tekhex_bfd_is_group_section bfd_generic_is_group_section
|
|
#define tekhex_bfd_discard_group bfd_generic_discard_group
|
|
#define tekhex_section_already_linked _bfd_generic_section_already_linked
|
|
#define tekhex_bfd_link_hash_table_create _bfd_generic_link_hash_table_create
|
|
#define tekhex_bfd_link_hash_table_free _bfd_generic_link_hash_table_free
|
|
#define tekhex_bfd_link_add_symbols _bfd_generic_link_add_symbols
|
|
#define tekhex_bfd_link_just_syms _bfd_generic_link_just_syms
|
|
#define tekhex_bfd_final_link _bfd_generic_final_link
|
|
#define tekhex_bfd_link_split_section _bfd_generic_link_split_section
|
|
#define tekhex_get_section_contents_in_window _bfd_generic_get_section_contents_in_window
|
|
|
|
const bfd_target tekhex_vec =
|
|
{
|
|
"tekhex", /* Name. */
|
|
bfd_target_tekhex_flavour,
|
|
BFD_ENDIAN_UNKNOWN, /* Target byte order. */
|
|
BFD_ENDIAN_UNKNOWN, /* Target headers byte order. */
|
|
(EXEC_P | /* Object flags. */
|
|
HAS_SYMS | HAS_LINENO | HAS_DEBUG |
|
|
HAS_RELOC | HAS_LOCALS | WP_TEXT | D_PAGED),
|
|
(SEC_CODE | SEC_DATA | SEC_ROM | SEC_HAS_CONTENTS
|
|
| SEC_ALLOC | SEC_LOAD | SEC_RELOC), /* Section flags. */
|
|
0, /* Leading underscore. */
|
|
' ', /* AR_pad_char. */
|
|
16, /* AR_max_namelen. */
|
|
bfd_getb64, bfd_getb_signed_64, bfd_putb64,
|
|
bfd_getb32, bfd_getb_signed_32, bfd_putb32,
|
|
bfd_getb16, bfd_getb_signed_16, bfd_putb16, /* Data. */
|
|
bfd_getb64, bfd_getb_signed_64, bfd_putb64,
|
|
bfd_getb32, bfd_getb_signed_32, bfd_putb32,
|
|
bfd_getb16, bfd_getb_signed_16, bfd_putb16, /* Headers. */
|
|
|
|
{
|
|
_bfd_dummy_target,
|
|
tekhex_object_p, /* bfd_check_format. */
|
|
_bfd_dummy_target,
|
|
_bfd_dummy_target,
|
|
},
|
|
{
|
|
bfd_false,
|
|
tekhex_mkobject,
|
|
_bfd_generic_mkarchive,
|
|
bfd_false,
|
|
},
|
|
{ /* bfd_write_contents. */
|
|
bfd_false,
|
|
tekhex_write_object_contents,
|
|
_bfd_write_archive_contents,
|
|
bfd_false,
|
|
},
|
|
|
|
BFD_JUMP_TABLE_GENERIC (tekhex),
|
|
BFD_JUMP_TABLE_COPY (_bfd_generic),
|
|
BFD_JUMP_TABLE_CORE (_bfd_nocore),
|
|
BFD_JUMP_TABLE_ARCHIVE (_bfd_noarchive),
|
|
BFD_JUMP_TABLE_SYMBOLS (tekhex),
|
|
BFD_JUMP_TABLE_RELOCS (_bfd_norelocs),
|
|
BFD_JUMP_TABLE_WRITE (tekhex),
|
|
BFD_JUMP_TABLE_LINK (tekhex),
|
|
BFD_JUMP_TABLE_DYNAMIC (_bfd_nodynamic),
|
|
|
|
NULL,
|
|
|
|
NULL
|
|
};
|