1163 lines
35 KiB
C
1163 lines
35 KiB
C
/* write.c - emit .o file
|
||
Copyright (C) 1986, 1987, 1990, 1991 Free Software Foundation, Inc.
|
||
|
||
This file is part of GAS, the GNU Assembler.
|
||
|
||
GAS 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 1, or (at your option)
|
||
any later version.
|
||
|
||
GAS 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 GAS; see the file COPYING. If not, write to
|
||
the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */
|
||
|
||
/* static const char rcsid[] = "$Id$"; */
|
||
|
||
/*
|
||
|
||
This thing should be set up to do byteordering correctly. But...
|
||
|
||
In order to cross-assemble the target machine must have an a.out header
|
||
similar to the one in a.out.h on THIS machine. Byteorder doesn't matter,
|
||
we take special care of it, but the numbers must be the same SIZE (# of
|
||
bytes) and in the same PLACE. If this is not true, you will have some
|
||
trouble.
|
||
*/
|
||
|
||
#include "as.h"
|
||
|
||
#include "subsegs.h"
|
||
#include "obstack.h"
|
||
#include "output-file.h"
|
||
|
||
/* Hook for machine dependent relocation information output routine.
|
||
If not defined, the variable is allocated in BSS (Fortran common model).
|
||
If some other module defines it, we will see their value. */
|
||
|
||
void (*md_emit_relocations)();
|
||
|
||
/*
|
||
* In: length of relocation (or of address) in chars: 1, 2 or 4.
|
||
* Out: GNU LD relocation length code: 0, 1, or 2.
|
||
*/
|
||
|
||
unsigned char
|
||
nbytes_r_length [] = {
|
||
42, 0, 1, 42, 2
|
||
};
|
||
|
||
|
||
static struct frag *text_frag_root;
|
||
static struct frag *data_frag_root;
|
||
|
||
static struct frag *text_last_frag; /* Last frag in segment. */
|
||
static struct frag *data_last_frag; /* Last frag in segment. */
|
||
|
||
static object_headers headers;
|
||
|
||
long string_byte_count;
|
||
|
||
static char *the_object_file;
|
||
|
||
char *next_object_file_charP; /* Tracks object file bytes. */
|
||
|
||
int magic_number_for_object_file = DEFAULT_MAGIC_NUMBER_FOR_OBJECT_FILE;
|
||
|
||
/* static long length; JF unused */ /* String length, including trailing '\0'. */
|
||
|
||
|
||
#ifdef __STDC__
|
||
|
||
static int is_dnrange(struct frag *f1, struct frag *f2);
|
||
static long fixup_segment(fixS *fixP, segT this_segment_type);
|
||
static relax_addressT relax_align(relax_addressT address, long alignment);
|
||
static void relax_segment(struct frag *segment_frag_root, segT segment_type);
|
||
|
||
#else
|
||
|
||
static int is_dnrange();
|
||
static long fixup_segment();
|
||
static relax_addressT relax_align();
|
||
static void relax_segment();
|
||
|
||
#endif /* __STDC__ */
|
||
|
||
/*
|
||
* fix_new()
|
||
*
|
||
* Create a fixS in obstack 'notes'.
|
||
*/
|
||
fixS *fix_new(frag, where, size, add_symbol, sub_symbol, offset, pcrel, r_type)
|
||
fragS *frag; /* Which frag? */
|
||
int where; /* Where in that frag? */
|
||
short int size; /* 1, 2 or 4 usually. */
|
||
symbolS *add_symbol; /* X_add_symbol. */
|
||
symbolS *sub_symbol; /* X_subtract_symbol. */
|
||
long offset; /* X_add_number. */
|
||
int pcrel; /* TRUE if PC-relative relocation. */
|
||
enum reloc_type r_type; /* Relocation type */
|
||
{
|
||
register fixS * fixP;
|
||
|
||
fixP = (fixS *)obstack_alloc(¬es,sizeof(fixS));
|
||
|
||
fixP->fx_frag = frag;
|
||
fixP->fx_where = where;
|
||
fixP->fx_size = size;
|
||
fixP->fx_addsy = add_symbol;
|
||
fixP->fx_subsy = sub_symbol;
|
||
fixP->fx_offset = offset;
|
||
fixP->fx_pcrel = pcrel;
|
||
fixP->fx_r_type = r_type;
|
||
fixP->fx_next = NULL;
|
||
|
||
/* JF these 'cuz of the NS32K stuff */
|
||
fixP->fx_im_disp = 0;
|
||
fixP->fx_pcrel_adjust = 0;
|
||
fixP->fx_bsr = 0;
|
||
fixP->fx_bit_fixP = 0;
|
||
|
||
if (*seg_fix_tailP)
|
||
(*seg_fix_tailP)->fx_next = fixP;
|
||
else
|
||
*seg_fix_rootP = fixP;
|
||
*seg_fix_tailP = fixP;
|
||
fixP->fx_callj = 0;
|
||
return fixP;
|
||
}
|
||
|
||
void write_object_file() {
|
||
register struct frchain * frchainP; /* Track along all frchains. */
|
||
register fragS * fragP; /* Track along all frags. */
|
||
register struct frchain * next_frchainP;
|
||
register fragS * * prev_fragPP;
|
||
/* register char * name; */
|
||
/* symbolS *symbolP; */
|
||
/* register symbolS ** symbolPP; */
|
||
/* register fixS * fixP; JF unused */
|
||
unsigned int data_siz;
|
||
|
||
#ifdef DONTDEF
|
||
void gdb_emit();
|
||
void gdb_end();
|
||
#endif
|
||
long object_file_size;
|
||
|
||
#ifdef VMS
|
||
/*
|
||
* Under VMS we try to be compatible with VAX-11 "C". Thus, we
|
||
* call a routine to check for the definition of the procedure
|
||
* "_main", and if so -- fix it up so that it can be program
|
||
* entry point.
|
||
*/
|
||
VMS_Check_For_Main();
|
||
#endif /* VMS */
|
||
/*
|
||
* After every sub-segment, we fake an ".align ...". This conforms to BSD4.2
|
||
* brane-damage. We then fake ".fill 0" because that is the kind of frag
|
||
* that requires least thought. ".align" frags like to have a following
|
||
* frag since that makes calculating their intended length trivial.
|
||
*/
|
||
#define SUB_SEGMENT_ALIGN (2)
|
||
for (frchainP = frchain_root; frchainP; frchainP = frchainP->frch_next) {
|
||
#ifdef VMS
|
||
/*
|
||
* Under VAX/VMS, the linker (and PSECT specifications)
|
||
* take care of correctly aligning the segments.
|
||
* Doing the alignment here (on initialized data) can
|
||
* mess up the calculation of global data PSECT sizes.
|
||
*/
|
||
#undef SUB_SEGMENT_ALIGN
|
||
#define SUB_SEGMENT_ALIGN ((frchainP->frch_seg != SEG_DATA) ? 2 : 0)
|
||
#endif /* VMS */
|
||
subseg_new (frchainP->frch_seg, frchainP->frch_subseg);
|
||
frag_align (SUB_SEGMENT_ALIGN, 0);
|
||
/* frag_align will have left a new frag. */
|
||
/* Use this last frag for an empty ".fill". */
|
||
/*
|
||
* For this segment ...
|
||
* Create a last frag. Do not leave a "being filled in frag".
|
||
*/
|
||
frag_wane (frag_now);
|
||
frag_now->fr_fix = 0;
|
||
know( frag_now->fr_next == NULL );
|
||
/* know( frags . obstack_c_base == frags . obstack_c_next_free ); */
|
||
/* Above shows we haven't left a half-completed object on obstack. */
|
||
} /* walk the frag chain */
|
||
|
||
/*
|
||
* From now on, we don't care about sub-segments.
|
||
* Build one frag chain for each segment. Linked thru fr_next.
|
||
* We know that there is at least 1 text frchain & at least 1 data frchain.
|
||
*/
|
||
prev_fragPP = &text_frag_root;
|
||
for (frchainP = frchain_root; frchainP; frchainP = next_frchainP) {
|
||
know( frchainP->frch_root );
|
||
* prev_fragPP = frchainP->frch_root;
|
||
prev_fragPP = & frchainP->frch_last->fr_next;
|
||
|
||
if (((next_frchainP = frchainP->frch_next) == NULL)
|
||
|| next_frchainP == data0_frchainP) {
|
||
prev_fragPP = & data_frag_root;
|
||
if (next_frchainP) {
|
||
text_last_frag = frchainP->frch_last;
|
||
} else {
|
||
data_last_frag = frchainP->frch_last;
|
||
}
|
||
}
|
||
} /* walk the frag chain */
|
||
|
||
/*
|
||
* We have two segments. If user gave -R flag, then we must put the
|
||
* data frags into the text segment. Do this before relaxing so
|
||
* we know to take advantage of -R and make shorter addresses.
|
||
*/
|
||
if (flagseen[ 'R' ]) {
|
||
fixS *tmp;
|
||
|
||
text_last_frag->fr_next = data_frag_root;
|
||
text_last_frag = data_last_frag;
|
||
data_last_frag = NULL;
|
||
data_frag_root = NULL;
|
||
if (text_fix_root) {
|
||
for (tmp = text_fix_root; tmp->fx_next; tmp = tmp->fx_next) ;;
|
||
tmp->fx_next=data_fix_root;
|
||
} else
|
||
text_fix_root=data_fix_root;
|
||
data_fix_root=NULL;
|
||
}
|
||
|
||
relax_segment(text_frag_root, SEG_TEXT);
|
||
relax_segment(data_frag_root, SEG_DATA);
|
||
/*
|
||
* Now the addresses of frags are correct within the segment.
|
||
*/
|
||
|
||
know(text_last_frag->fr_type == rs_fill && text_last_frag->fr_offset == 0);
|
||
H_SET_TEXT_SIZE(&headers,text_last_frag->fr_address);
|
||
text_last_frag->fr_address=H_GET_TEXT_SIZE(&headers);
|
||
|
||
/*
|
||
* Join the 2 segments into 1 huge segment.
|
||
* To do this, re-compute every rn_address in the SEG_DATA frags.
|
||
* Then join the data frags after the text frags.
|
||
*
|
||
* Determine a_data [length of data segment].
|
||
*/
|
||
if (data_frag_root) {
|
||
register relax_addressT slide;
|
||
|
||
know( text_last_frag->fr_type == rs_fill && text_last_frag->fr_offset == 0 );
|
||
H_SET_DATA_SIZE(&headers, data_last_frag->fr_address);
|
||
data_last_frag->fr_address = H_GET_DATA_SIZE(&headers);
|
||
slide = H_GET_TEXT_SIZE(&headers); /* & in file of the data segment. */
|
||
|
||
for (fragP = data_frag_root;
|
||
fragP;
|
||
fragP = fragP->fr_next)
|
||
{
|
||
fragP->fr_address += slide;
|
||
}
|
||
know( text_last_frag );
|
||
text_last_frag->fr_next = data_frag_root;
|
||
} else {
|
||
H_SET_DATA_SIZE(&headers,0);
|
||
data_siz = 0;
|
||
}
|
||
|
||
bss_address_frag.fr_address = H_GET_TEXT_SIZE(&headers) +
|
||
H_GET_DATA_SIZE(&headers);
|
||
|
||
H_SET_BSS_SIZE(&headers,local_bss_counter);
|
||
|
||
/*
|
||
*
|
||
* Crawl the symbol chain.
|
||
*
|
||
* For each symbol whose value depends on a frag, take the address of
|
||
* that frag and subsume it into the value of the symbol.
|
||
* After this, there is just one way to lookup a symbol value.
|
||
* Values are left in their final state for object file emission.
|
||
* We adjust the values of 'L' local symbols, even if we do
|
||
* not intend to emit them to the object file, because their values
|
||
* are needed for fix-ups.
|
||
*
|
||
* Unless we saw a -L flag, remove all symbols that begin with 'L'
|
||
* from the symbol chain. (They are still pointed to by the fixes.)
|
||
*
|
||
* Count the remaining symbols.
|
||
* Assign a symbol number to each symbol.
|
||
* Count the number of string-table chars we will emit.
|
||
* Put this info into the headers as appropriate.
|
||
*
|
||
*/
|
||
know(zero_address_frag.fr_address == 0);
|
||
string_byte_count = sizeof(string_byte_count);
|
||
|
||
obj_crawl_symbol_chain(&headers);
|
||
|
||
if (string_byte_count == sizeof(string_byte_count)) {
|
||
string_byte_count = 0;
|
||
} /* if no strings, then no count. */
|
||
|
||
H_SET_STRING_SIZE(&headers, string_byte_count);
|
||
|
||
/*
|
||
* Addresses of frags now reflect addresses we use in the object file.
|
||
* Symbol values are correct.
|
||
* Scan the frags, converting any ".org"s and ".align"s to ".fill"s.
|
||
* Also converting any machine-dependent frags using md_convert_frag();
|
||
*/
|
||
subseg_change(SEG_TEXT, 0);
|
||
|
||
for (fragP = text_frag_root; fragP; fragP = fragP->fr_next) {
|
||
switch (fragP->fr_type) {
|
||
case rs_align:
|
||
case rs_org:
|
||
fragP->fr_type = rs_fill;
|
||
know( fragP->fr_var == 1 );
|
||
know( fragP->fr_next );
|
||
fragP->fr_offset
|
||
= fragP->fr_next->fr_address
|
||
- fragP->fr_address
|
||
- fragP->fr_fix;
|
||
break;
|
||
|
||
case rs_fill:
|
||
break;
|
||
|
||
case rs_machine_dependent:
|
||
md_convert_frag (fragP);
|
||
/*
|
||
* After md_convert_frag, we make the frag into a ".space 0".
|
||
* Md_convert_frag() should set up any fixSs and constants
|
||
* required.
|
||
*/
|
||
frag_wane (fragP);
|
||
break;
|
||
|
||
#ifndef WORKING_DOT_WORD
|
||
case rs_broken_word: {
|
||
struct broken_word *lie;
|
||
extern md_short_jump_size;
|
||
extern md_long_jump_size;
|
||
|
||
if (fragP->fr_subtype) {
|
||
fragP->fr_fix+=md_short_jump_size;
|
||
for (lie=(struct broken_word *)(fragP->fr_symbol);lie && lie->dispfrag==fragP;lie=lie->next_broken_word)
|
||
if (lie->added==1)
|
||
fragP->fr_fix+=md_long_jump_size;
|
||
}
|
||
frag_wane(fragP);
|
||
}
|
||
break;
|
||
#endif
|
||
|
||
default:
|
||
BAD_CASE( fragP->fr_type );
|
||
break;
|
||
} /* switch (fr_type) */
|
||
} /* for each frag. */
|
||
|
||
#ifndef WORKING_DOT_WORD
|
||
{
|
||
struct broken_word *lie;
|
||
struct broken_word **prevP;
|
||
|
||
prevP= &broken_words;
|
||
for (lie=broken_words; lie; lie=lie->next_broken_word)
|
||
if (!lie->added) {
|
||
#ifdef TC_NS32K
|
||
fix_new_ns32k(lie->frag,
|
||
lie->word_goes_here - lie->frag->fr_literal,
|
||
2,
|
||
lie->add,
|
||
lie->sub,
|
||
lie->addnum,
|
||
0, 0, 2, 0, 0);
|
||
#else /* TC_NS32K */
|
||
fix_new( lie->frag, lie->word_goes_here - lie->frag->fr_literal,
|
||
2, lie->add,
|
||
lie->sub, lie->addnum,
|
||
0, NO_RELOC);
|
||
#endif /* TC_NS32K */
|
||
/* md_number_to_chars(lie->word_goes_here,
|
||
S_GET_VALUE(lie->add)
|
||
+ lie->addnum
|
||
- S_GET_VALUE(lie->sub),
|
||
2); */
|
||
*prevP=lie->next_broken_word;
|
||
} else
|
||
prevP= &(lie->next_broken_word);
|
||
|
||
for (lie=broken_words;lie;) {
|
||
struct broken_word *untruth;
|
||
char *table_ptr;
|
||
long table_addr;
|
||
long from_addr,
|
||
to_addr;
|
||
int n,
|
||
m;
|
||
|
||
extern md_short_jump_size;
|
||
extern md_long_jump_size;
|
||
|
||
fragP=lie->dispfrag;
|
||
|
||
/* Find out how many broken_words go here */
|
||
n=0;
|
||
for (untruth=lie;untruth && untruth->dispfrag==fragP;untruth=untruth->next_broken_word)
|
||
if (untruth->added==1)
|
||
n++;
|
||
|
||
table_ptr=lie->dispfrag->fr_opcode;
|
||
table_addr=lie->dispfrag->fr_address+(table_ptr - lie->dispfrag->fr_literal);
|
||
/* Create the jump around the long jumps */
|
||
/* This is a short jump from table_ptr+0 to table_ptr+n*long_jump_size */
|
||
from_addr=table_addr;
|
||
to_addr = table_addr + md_short_jump_size + n * md_long_jump_size;
|
||
md_create_short_jump(table_ptr, from_addr, to_addr, lie->dispfrag, lie->add);
|
||
table_ptr+=md_short_jump_size;
|
||
table_addr+=md_short_jump_size;
|
||
|
||
for (m=0;lie && lie->dispfrag==fragP;m++,lie=lie->next_broken_word) {
|
||
if (lie->added==2)
|
||
continue;
|
||
/* Patch the jump table */
|
||
/* This is the offset from ??? to table_ptr+0 */
|
||
to_addr = table_addr
|
||
- S_GET_VALUE(lie->sub);
|
||
md_number_to_chars(lie->word_goes_here,to_addr,2);
|
||
for (untruth=lie->next_broken_word;untruth && untruth->dispfrag==fragP;untruth=untruth->next_broken_word) {
|
||
if (untruth->use_jump==lie)
|
||
md_number_to_chars(untruth->word_goes_here,to_addr,2);
|
||
}
|
||
|
||
/* Install the long jump */
|
||
/* this is a long jump from table_ptr+0 to the final target */
|
||
from_addr=table_addr;
|
||
to_addr=S_GET_VALUE(lie->add) + lie->addnum;
|
||
md_create_long_jump(table_ptr,from_addr,to_addr,lie->dispfrag,lie->add);
|
||
table_ptr+=md_long_jump_size;
|
||
table_addr+=md_long_jump_size;
|
||
}
|
||
}
|
||
}
|
||
#endif /* not WORKING_DOT_WORD */
|
||
|
||
#ifndef VMS
|
||
{ /* not vms */
|
||
/*
|
||
* Scan every FixS performing fixups. We had to wait until now to do
|
||
* this because md_convert_frag() may have made some fixSs.
|
||
*/
|
||
H_SET_RELOCATION_SIZE(&headers,
|
||
md_reloc_size * fixup_segment(text_fix_root, SEG_TEXT),
|
||
md_reloc_size * fixup_segment(data_fix_root, SEG_DATA));
|
||
|
||
/* FIXME move this stuff into the pre-write-hook */
|
||
H_SET_MAGIC_NUMBER(&headers, magic_number_for_object_file);
|
||
H_SET_ENTRY_POINT(&headers,0);
|
||
|
||
#ifdef EXEC_MACHINE_TYPE
|
||
H_SET_MACHINE_TYPE(&headers,EXEC_MACHINE_TYPE);
|
||
#endif
|
||
#ifdef EXEC_VERSION
|
||
H_SET_VERSION(&headers,EXEC_VERSION);
|
||
#endif
|
||
|
||
obj_pre_write_hook(&headers); /* extra coff stuff */
|
||
|
||
if ((had_warnings() && flagseen['Z'])
|
||
|| had_errors() > 0) {
|
||
if (flagseen['Z']) {
|
||
as_warn("%d error%s, %d warning%s, generating bad object file.\n",
|
||
had_errors(), had_errors() == 1 ? "" : "s",
|
||
had_warnings(), had_warnings() == 1 ? "" : "s");
|
||
} else {
|
||
as_fatal("%d error%s, %d warning%s, no object file generated.\n",
|
||
had_errors(), had_errors() == 1 ? "" : "s",
|
||
had_warnings(), had_warnings() == 1 ? "" : "s");
|
||
} /* on want output */
|
||
} /* on error condition */
|
||
|
||
object_file_size = H_GET_FILE_SIZE(&headers);
|
||
next_object_file_charP = the_object_file = xmalloc(object_file_size);
|
||
|
||
output_file_create(out_file_name);
|
||
|
||
obj_header_append(&next_object_file_charP, &headers);
|
||
|
||
/*
|
||
* Emit code.
|
||
*/
|
||
for (fragP = text_frag_root; fragP; fragP = fragP->fr_next) {
|
||
register long count;
|
||
register char * fill_literal;
|
||
register long fill_size;
|
||
|
||
know( fragP->fr_type == rs_fill );
|
||
append (& next_object_file_charP, fragP->fr_literal, (unsigned long)fragP->fr_fix);
|
||
fill_literal= fragP->fr_literal + fragP->fr_fix;
|
||
fill_size = fragP->fr_var;
|
||
know( fragP->fr_offset >= 0 );
|
||
for (count = fragP->fr_offset; count; count --)
|
||
append (& next_object_file_charP, fill_literal, (unsigned long)fill_size);
|
||
} /* for each code frag. */
|
||
|
||
/*
|
||
* Emit relocations.
|
||
*/
|
||
obj_emit_relocations(&next_object_file_charP, text_fix_root, (relax_addressT)0);
|
||
|
||
#ifdef TC_I960
|
||
/* Make addresses in data relocation directives relative to beginning of
|
||
* first data fragment, not end of last text fragment: alignment of the
|
||
* start of the data segment may place a gap between the segments.
|
||
*/
|
||
obj_emit_relocations(&next_object_file_charP, data_fix_root, data0_frchainP->frch_root->fr_address);
|
||
#else /* TC_I960 */
|
||
obj_emit_relocations(&next_object_file_charP, data_fix_root, text_last_frag->fr_address);
|
||
#endif /* TC_I960 */
|
||
|
||
/*
|
||
* Emit line number entries.
|
||
*/
|
||
OBJ_EMIT_LINENO(&next_object_file_charP, lineno_rootP, the_object_file);
|
||
|
||
/*
|
||
* Emit symbols.
|
||
*/
|
||
obj_emit_symbols(&next_object_file_charP, symbol_rootP);
|
||
|
||
/*
|
||
* Emit strings.
|
||
*/
|
||
|
||
if (string_byte_count > 0) {
|
||
obj_emit_strings(&next_object_file_charP);
|
||
} /* only if we have a string table */
|
||
|
||
know(next_object_file_charP == the_object_file + object_file_size);
|
||
/* Write the data to the file */
|
||
output_file_append(the_object_file,object_file_size,out_file_name);
|
||
|
||
#ifdef DONTDEF
|
||
if (flagseen['G']) /* GDB symbol file to be appended? */
|
||
{
|
||
gdb_emit (out_file_name);
|
||
gdb_end ();
|
||
}
|
||
#endif /* DONTDEF */
|
||
|
||
output_file_close(out_file_name);
|
||
} /* non vms output */
|
||
#else /* VMS */
|
||
/*
|
||
* Now do the VMS-dependent part of writing the object file
|
||
*/
|
||
VMS_write_object_file(text_siz, data_siz, text_frag_root, data_frag_root);
|
||
#endif /* VMS */
|
||
} /* write_object_file() */
|
||
|
||
/*
|
||
* relax_segment()
|
||
*
|
||
* Now we have a segment, not a crowd of sub-segments, we can make fr_address
|
||
* values.
|
||
*
|
||
* Relax the frags.
|
||
*
|
||
* After this, all frags in this segment have addresses that are correct
|
||
* within the segment. Since segments live in different file addresses,
|
||
* these frag addresses may not be the same as final object-file addresses.
|
||
*/
|
||
#ifndef VMS
|
||
static
|
||
#endif /* not VMS */
|
||
void relax_segment(segment_frag_root, segment_type)
|
||
struct frag * segment_frag_root;
|
||
segT segment_type; /* SEG_DATA or SEG_TEXT */
|
||
{
|
||
register struct frag * fragP;
|
||
register relax_addressT address;
|
||
/* register relax_addressT old_address; JF unused */
|
||
/* register relax_addressT new_address; JF unused */
|
||
|
||
know( segment_type == SEG_DATA || segment_type == SEG_TEXT );
|
||
|
||
/* In case md_estimate_size_before_relax() wants to make fixSs. */
|
||
subseg_change(segment_type, 0);
|
||
|
||
/*
|
||
* For each frag in segment: count and store (a 1st guess of) fr_address.
|
||
*/
|
||
address = 0;
|
||
for ( fragP = segment_frag_root; fragP; fragP = fragP->fr_next )
|
||
{
|
||
fragP->fr_address = address;
|
||
address += fragP->fr_fix;
|
||
switch (fragP->fr_type)
|
||
{
|
||
case rs_fill:
|
||
address += fragP->fr_offset * fragP->fr_var;
|
||
break;
|
||
|
||
case rs_align:
|
||
address += relax_align(address, fragP->fr_offset);
|
||
break;
|
||
|
||
case rs_org:
|
||
/*
|
||
* Assume .org is nugatory. It will grow with 1st relax.
|
||
*/
|
||
break;
|
||
|
||
case rs_machine_dependent:
|
||
address += md_estimate_size_before_relax(fragP, segment_type);
|
||
break;
|
||
|
||
#ifndef WORKING_DOT_WORD
|
||
/* Broken words don't concern us yet */
|
||
case rs_broken_word:
|
||
break;
|
||
#endif
|
||
|
||
default:
|
||
BAD_CASE( fragP->fr_type );
|
||
break;
|
||
} /* switch(fr_type) */
|
||
} /* for each frag in the segment */
|
||
|
||
/*
|
||
* Do relax().
|
||
*/
|
||
{
|
||
register long stretch; /* May be any size, 0 or negative. */
|
||
/* Cumulative number of addresses we have */
|
||
/* relaxed this pass. */
|
||
/* We may have relaxed more than one address. */
|
||
register long stretched; /* Have we stretched on this pass? */
|
||
/* This is 'cuz stretch may be zero, when,
|
||
in fact some piece of code grew, and
|
||
another shrank. If a branch instruction
|
||
doesn't fit anymore, we could be scrod */
|
||
|
||
do
|
||
{
|
||
stretch = stretched = 0;
|
||
for (fragP = segment_frag_root; fragP; fragP = fragP->fr_next)
|
||
{
|
||
register long growth = 0;
|
||
register unsigned long was_address;
|
||
/* register long var; */
|
||
register long offset;
|
||
register symbolS *symbolP;
|
||
register long target;
|
||
register long after;
|
||
register long aim;
|
||
|
||
was_address = fragP->fr_address;
|
||
address = fragP->fr_address += stretch;
|
||
symbolP = fragP->fr_symbol;
|
||
offset = fragP->fr_offset;
|
||
/* var = fragP->fr_var; */
|
||
switch (fragP->fr_type)
|
||
{
|
||
case rs_fill: /* .fill never relaxes. */
|
||
growth = 0;
|
||
break;
|
||
|
||
#ifndef WORKING_DOT_WORD
|
||
/* JF: This is RMS's idea. I do *NOT* want to be blamed
|
||
for it I do not want to write it. I do not want to have
|
||
anything to do with it. This is not the proper way to
|
||
implement this misfeature. */
|
||
case rs_broken_word:
|
||
{
|
||
struct broken_word *lie;
|
||
struct broken_word *untruth;
|
||
extern int md_short_jump_size;
|
||
extern int md_long_jump_size;
|
||
|
||
/* Yes this is ugly (storing the broken_word pointer
|
||
in the symbol slot). Still, this whole chunk of
|
||
code is ugly, and I don't feel like doing anything
|
||
about it. Think of it as stubbornness in action */
|
||
growth=0;
|
||
for (lie=(struct broken_word *)(fragP->fr_symbol);
|
||
lie && lie->dispfrag==fragP;
|
||
lie=lie->next_broken_word) {
|
||
|
||
if (lie->added)
|
||
continue;
|
||
offset= lie->add->sy_frag->fr_address+ S_GET_VALUE(lie->add) + lie->addnum -
|
||
(lie->sub->sy_frag->fr_address+ S_GET_VALUE(lie->sub));
|
||
if (offset<=-32768 || offset>=32767) {
|
||
if (flagseen['k'])
|
||
as_warn(".word %s-%s+%ld didn't fit",
|
||
S_GET_NAME(lie->add),
|
||
S_GET_NAME(lie->sub),
|
||
lie->addnum);
|
||
lie->added=1;
|
||
if (fragP->fr_subtype==0) {
|
||
fragP->fr_subtype++;
|
||
growth+=md_short_jump_size;
|
||
}
|
||
for (untruth=lie->next_broken_word;untruth && untruth->dispfrag==lie->dispfrag;untruth=untruth->next_broken_word)
|
||
if ((untruth->add->sy_frag == lie->add->sy_frag)
|
||
&& S_GET_VALUE(untruth->add) == S_GET_VALUE(lie->add)) {
|
||
untruth->added=2;
|
||
untruth->use_jump=lie;
|
||
}
|
||
growth+=md_long_jump_size;
|
||
}
|
||
}
|
||
}
|
||
break;
|
||
#endif
|
||
case rs_align:
|
||
growth = relax_align ((relax_addressT)(address + fragP->fr_fix), offset)
|
||
- relax_align ((relax_addressT)(was_address + fragP->fr_fix), offset);
|
||
break;
|
||
|
||
case rs_org:
|
||
target = offset;
|
||
if (symbolP)
|
||
{
|
||
know((S_GET_SEGMENT(symbolP) == SEG_ABSOLUTE) ||
|
||
(S_GET_SEGMENT(symbolP) == SEG_DATA) ||
|
||
(S_GET_SEGMENT(symbolP) == SEG_TEXT));
|
||
know(symbolP->sy_frag);
|
||
know(!(S_GET_SEGMENT(symbolP) == SEG_ABSOLUTE) ||
|
||
symbolP->sy_frag==&zero_address_frag );
|
||
target +=
|
||
S_GET_VALUE(symbolP)
|
||
+ symbolP->sy_frag->fr_address;
|
||
}
|
||
know( fragP->fr_next );
|
||
after = fragP->fr_next->fr_address;
|
||
growth = ((target - after ) > 0) ? (target - after) : 0;
|
||
/* Growth may be -ve, but variable part */
|
||
/* of frag cannot have < 0 chars. */
|
||
/* That is, we can't .org backwards. */
|
||
|
||
growth -= stretch; /* This is an absolute growth factor */
|
||
break;
|
||
|
||
case rs_machine_dependent:
|
||
{
|
||
register const relax_typeS * this_type;
|
||
register const relax_typeS * start_type;
|
||
register relax_substateT next_state;
|
||
register relax_substateT this_state;
|
||
|
||
start_type = this_type
|
||
= md_relax_table + (this_state = fragP->fr_subtype);
|
||
target = offset;
|
||
if (symbolP)
|
||
{
|
||
know((S_GET_SEGMENT(symbolP) == SEG_ABSOLUTE) ||
|
||
(S_GET_SEGMENT(symbolP) == SEG_DATA) ||
|
||
(S_GET_SEGMENT(symbolP) == SEG_TEXT));
|
||
know(symbolP->sy_frag);
|
||
know(!(S_GET_SEGMENT(symbolP) == SEG_ABSOLUTE) ||
|
||
symbolP->sy_frag==&zero_address_frag );
|
||
target +=
|
||
S_GET_VALUE(symbolP)
|
||
+ symbolP->sy_frag->fr_address;
|
||
|
||
/* If frag has yet to be reached on this pass,
|
||
assume it will move by STRETCH just as we did.
|
||
If this is not so, it will be because some frag
|
||
between grows, and that will force another pass. */
|
||
|
||
/* JF was just address */
|
||
/* JF also added is_dnrange hack */
|
||
/* There's gotta be a better/faster/etc way
|
||
to do this. . . */
|
||
/* gnu@cygnus.com: I changed this from > to >=
|
||
because I ran into a zero-length frag (fr_fix=0)
|
||
which was created when the obstack needed a new
|
||
chunk JUST AFTER the opcode of a branch. Since
|
||
fr_fix is zero, fr_address of this frag is the same
|
||
as fr_address of the next frag. This
|
||
zero-length frag was variable and jumped to .+2
|
||
(in the next frag), but since the > comparison
|
||
below failed (the two were =, not >), "stretch"
|
||
was not added to the target. Stretch was 178, so
|
||
the offset appeared to be .-176 instead, which did
|
||
not fit into a byte branch, so the assembler
|
||
relaxed the branch to a word. This didn't compare
|
||
with what happened when the same source file was
|
||
assembled on other machines, which is how I found it.
|
||
You might want to think about what other places have
|
||
trouble with zero length frags... */
|
||
|
||
if (symbolP->sy_frag->fr_address >= was_address && is_dnrange(fragP,symbolP->sy_frag))
|
||
target += stretch;
|
||
|
||
}
|
||
aim = target - address - fragP->fr_fix;
|
||
/* The displacement is affected by the instruction size
|
||
* for the 32k architecture. I think we ought to be able
|
||
* to add fragP->fr_pcrel_adjust in all cases (it should be
|
||
* zero if not used), but just in case it breaks something
|
||
* else we'll put this inside #ifdef NS32K ... #endif
|
||
*/
|
||
#ifdef TC_NS32K
|
||
aim += fragP->fr_pcrel_adjust;
|
||
#endif /* TC_NS32K */
|
||
|
||
if (aim < 0)
|
||
{
|
||
/* Look backwards. */
|
||
for (next_state = this_type->rlx_more; next_state; )
|
||
{
|
||
if (aim >= this_type->rlx_backward)
|
||
next_state = 0;
|
||
else
|
||
{ /* Grow to next state. */
|
||
this_type = md_relax_table + (this_state = next_state);
|
||
next_state = this_type->rlx_more;
|
||
}
|
||
}
|
||
}
|
||
else
|
||
{
|
||
#ifdef DONTDEF
|
||
/* JF these next few lines of code are for the mc68020 which can't handle short
|
||
offsets of zero in branch instructions. What a kludge! */
|
||
if (aim==0 && this_state==(1<<2+0)) { /* FOO hard encoded from m.c */
|
||
aim=this_type->rlx_forward+1; /* Force relaxation into word mode */
|
||
}
|
||
#endif
|
||
/* JF end of 68020 code */
|
||
/* Look forwards. */
|
||
for (next_state = this_type->rlx_more; next_state; )
|
||
{
|
||
if (aim <= this_type->rlx_forward)
|
||
next_state = 0;
|
||
else
|
||
{ /* Grow to next state. */
|
||
this_type = md_relax_table + (this_state = next_state);
|
||
next_state = this_type->rlx_more;
|
||
}
|
||
}
|
||
}
|
||
if ((growth = this_type->rlx_length - start_type->rlx_length) != 0)
|
||
fragP->fr_subtype = this_state;
|
||
}
|
||
break;
|
||
|
||
default:
|
||
BAD_CASE( fragP->fr_type );
|
||
break;
|
||
}
|
||
if (growth) {
|
||
stretch += growth;
|
||
stretched++;
|
||
}
|
||
} /* For each frag in the segment. */
|
||
} while (stretched); /* Until nothing further to relax. */
|
||
}
|
||
|
||
/*
|
||
* We now have valid fr_address'es for each frag.
|
||
*/
|
||
|
||
/*
|
||
* All fr_address's are correct, relative to their own segment.
|
||
* We have made all the fixS we will ever make.
|
||
*/
|
||
} /* relax_segment() */
|
||
|
||
/*
|
||
* Relax_align. Advance location counter to next address that has 'alignment'
|
||
* lowest order bits all 0s.
|
||
*/
|
||
|
||
/* How many addresses does the .align take? */
|
||
static relax_addressT relax_align(address, alignment)
|
||
register relax_addressT address; /* Address now. */
|
||
register long alignment; /* Alignment (binary). */
|
||
{
|
||
relax_addressT mask;
|
||
relax_addressT new_address;
|
||
|
||
mask = ~ ( (~0) << alignment );
|
||
new_address = (address + mask) & (~ mask);
|
||
return (new_address - address);
|
||
} /* relax_align() */
|
||
|
||
/* fixup_segment()
|
||
|
||
Go through all the fixS's in a segment and see which ones can be
|
||
handled now. (These consist of fixS where we have since discovered
|
||
the value of a symbol, or the address of the frag involved.)
|
||
For each one, call md_apply_fix to put the fix into the frag data.
|
||
|
||
Result is a count of how many relocation structs will be needed to
|
||
handle the remaining fixS's that we couldn't completely handle here.
|
||
These will be output later by emit_relocations(). */
|
||
|
||
static long fixup_segment(fixP, this_segment_type)
|
||
register fixS * fixP;
|
||
segT this_segment_type; /* N_TYPE bits for segment. */
|
||
{
|
||
register long seg_reloc_count;
|
||
register symbolS *add_symbolP;
|
||
register symbolS *sub_symbolP;
|
||
register long add_number;
|
||
register int size;
|
||
register char *place;
|
||
register long where;
|
||
register char pcrel;
|
||
register fragS *fragP;
|
||
register segT add_symbol_segment = SEG_ABSOLUTE;
|
||
fixS *topP = fixP;
|
||
|
||
|
||
seg_reloc_count = 0;
|
||
|
||
for ( ; fixP; fixP = fixP->fx_next) {
|
||
fragP = fixP->fx_frag;
|
||
know(fragP);
|
||
where = fixP->fx_where;
|
||
place = fragP->fr_literal + where;
|
||
size = fixP->fx_size;
|
||
add_symbolP = fixP->fx_addsy;
|
||
#ifdef TC_I960
|
||
if (fixP->fx_callj && TC_S_IS_CALLNAME(add_symbolP)) {
|
||
/* Relocation should be done via the
|
||
associated 'bal' entry point
|
||
symbol. */
|
||
|
||
if (!TC_S_IS_BALNAME(tc_get_bal_of_call(add_symbolP))) {
|
||
as_bad("No 'bal' entry point for leafproc %s",
|
||
S_GET_NAME(add_symbolP));
|
||
continue;
|
||
}
|
||
fixP->fx_addsy = add_symbolP = tc_get_bal_of_call(add_symbolP);
|
||
} /* callj relocation */
|
||
#endif
|
||
sub_symbolP = fixP->fx_subsy;
|
||
add_number = fixP->fx_offset;
|
||
pcrel = fixP->fx_pcrel;
|
||
|
||
if (add_symbolP) {
|
||
add_symbol_segment = S_GET_SEGMENT(add_symbolP);
|
||
} /* if there is an addend */
|
||
|
||
if (sub_symbolP) {
|
||
if (!add_symbolP) {
|
||
/* Its just -sym */
|
||
if (S_GET_SEGMENT(sub_symbolP) != SEG_ABSOLUTE) {
|
||
as_bad("Negative of non-absolute symbol %s", S_GET_NAME(sub_symbolP));
|
||
} /* not absolute */
|
||
|
||
add_number -= S_GET_VALUE(sub_symbolP);
|
||
|
||
/* if sub_symbol is in the same segment that add_symbol
|
||
and add_symbol is either in DATA, TEXT, BSS or ABSOLUTE */
|
||
} else if ((S_GET_SEGMENT(sub_symbolP) == add_symbol_segment)
|
||
&& ((add_symbol_segment == SEG_DATA)
|
||
|| (add_symbol_segment == SEG_TEXT)
|
||
|| (add_symbol_segment == SEG_BSS)
|
||
|| (add_symbol_segment == SEG_ABSOLUTE))) {
|
||
/* Difference of 2 symbols from same segment. */
|
||
/* Can't make difference of 2 undefineds: 'value' means */
|
||
/* something different for N_UNDF. */
|
||
#ifdef TC_I960
|
||
/* Makes no sense to use the difference of 2 arbitrary symbols
|
||
* as the target of a call instruction.
|
||
*/
|
||
if (fixP->fx_callj) {
|
||
as_bad("callj to difference of 2 symbols");
|
||
}
|
||
#endif /* TC_I960 */
|
||
add_number += S_GET_VALUE(add_symbolP) -
|
||
S_GET_VALUE(sub_symbolP);
|
||
|
||
add_symbolP = NULL;
|
||
fixP->fx_addsy = NULL;
|
||
} else {
|
||
/* Different segments in subtraction. */
|
||
know(!(S_IS_EXTERNAL(sub_symbolP) && (S_GET_SEGMENT(sub_symbolP) == SEG_ABSOLUTE)));
|
||
|
||
if ((S_GET_SEGMENT(sub_symbolP) == SEG_ABSOLUTE)) {
|
||
add_number -= S_GET_VALUE(sub_symbolP);
|
||
} else {
|
||
as_bad("Can't emit reloc {- %s-seg symbol \"%s\"} @ file address %d.",
|
||
segment_name(S_GET_SEGMENT(sub_symbolP)),
|
||
S_GET_NAME(sub_symbolP), fragP->fr_address + where);
|
||
} /* if absolute */
|
||
}
|
||
} /* if sub_symbolP */
|
||
|
||
if (add_symbolP) {
|
||
if (add_symbol_segment == this_segment_type && pcrel) {
|
||
/*
|
||
* This fixup was made when the symbol's segment was
|
||
* SEG_UNKNOWN, but it is now in the local segment.
|
||
* So we know how to do the address without relocation.
|
||
*/
|
||
#ifdef TC_I960
|
||
/* reloc_callj() may replace a 'call' with a 'calls' or a 'bal',
|
||
* in which cases it modifies *fixP as appropriate. In the case
|
||
* of a 'calls', no further work is required, and *fixP has been
|
||
* set up to make the rest of the code below a no-op.
|
||
*/
|
||
reloc_callj(fixP);
|
||
#endif /* TC_I960 */
|
||
|
||
add_number += S_GET_VALUE(add_symbolP);
|
||
add_number -= md_pcrel_from (fixP);
|
||
pcrel = 0; /* Lie. Don't want further pcrel processing. */
|
||
fixP->fx_addsy = NULL; /* No relocations please. */
|
||
} else {
|
||
switch (add_symbol_segment) {
|
||
case SEG_ABSOLUTE:
|
||
#ifdef TC_I960
|
||
reloc_callj(fixP); /* See comment about reloc_callj() above*/
|
||
#endif /* TC_I960 */
|
||
add_number += S_GET_VALUE(add_symbolP);
|
||
fixP->fx_addsy = NULL;
|
||
add_symbolP = NULL;
|
||
break;
|
||
|
||
case SEG_BSS:
|
||
case SEG_DATA:
|
||
case SEG_TEXT:
|
||
seg_reloc_count ++;
|
||
add_number += S_GET_VALUE(add_symbolP);
|
||
break;
|
||
|
||
case SEG_UNKNOWN:
|
||
#ifdef TC_I960
|
||
if ((int)fixP->fx_bit_fixP == 13) {
|
||
/* This is a COBR instruction. They have only a
|
||
* 13-bit displacement and are only to be used
|
||
* for local branches: flag as error, don't generate
|
||
* relocation.
|
||
*/
|
||
as_bad("can't use COBR format with external label");
|
||
fixP->fx_addsy = NULL; /* No relocations please. */
|
||
continue;
|
||
} /* COBR */
|
||
#endif /* TC_I960 */
|
||
/* FIXME-SOON: I think this is trash, but I'm not sure. xoxorich. */
|
||
#ifdef comment
|
||
#ifdef OBJ_COFF
|
||
if (S_IS_COMMON(add_symbolP))
|
||
add_number += S_GET_VALUE(add_symbolP);
|
||
#endif /* OBJ_COFF */
|
||
#endif /* comment */
|
||
|
||
++seg_reloc_count;
|
||
break;
|
||
|
||
default:
|
||
BAD_CASE(add_symbol_segment);
|
||
break;
|
||
} /* switch on symbol seg */
|
||
} /* if not in local seg */
|
||
} /* if there was a + symbol */
|
||
|
||
if (pcrel) {
|
||
add_number -= md_pcrel_from(fixP);
|
||
if (add_symbolP == 0) {
|
||
fixP->fx_addsy = & abs_symbol;
|
||
++seg_reloc_count;
|
||
} /* if there's an add_symbol */
|
||
} /* if pcrel */
|
||
|
||
if (!fixP->fx_bit_fixP) {
|
||
if ((size==1 &&
|
||
(add_number& ~0xFF) && (add_number&~0xFF!=(-1&~0xFF))) ||
|
||
(size==2 &&
|
||
(add_number& ~0xFFFF) && (add_number&~0xFFFF!=(-1&~0xFFFF)))) {
|
||
as_bad("Value of %d too large for field of %d bytes at 0x%x",
|
||
add_number, size, fragP->fr_address + where);
|
||
} /* generic error checking */
|
||
} /* not a bit fix */
|
||
|
||
md_apply_fix(fixP, add_number);
|
||
} /* For each fixS in this segment. */
|
||
|
||
#ifdef OBJ_COFF
|
||
#ifdef TC_I960
|
||
/* two relocs per callj under coff. */
|
||
for (fixP = topP; fixP; fixP = fixP->fx_next) {
|
||
if (fixP->fx_callj && fixP->fx_addsy != 0) {
|
||
++seg_reloc_count;
|
||
} /* if callj and not already fixed. */
|
||
} /* for each fix */
|
||
#endif /* TC_I960 */
|
||
#endif /* OBJ_COFF */
|
||
return(seg_reloc_count);
|
||
} /* fixup_segment() */
|
||
|
||
|
||
static int is_dnrange(f1,f2)
|
||
struct frag *f1;
|
||
struct frag *f2;
|
||
{
|
||
while (f1) {
|
||
if (f1->fr_next==f2)
|
||
return 1;
|
||
f1=f1->fr_next;
|
||
}
|
||
return 0;
|
||
} /* is_dnrange() */
|
||
|
||
/* Append a string onto another string, bumping the pointer along. */
|
||
void
|
||
append (charPP, fromP, length)
|
||
char **charPP;
|
||
char *fromP;
|
||
unsigned long length;
|
||
{
|
||
if (length) { /* Don't trust bcopy() of 0 chars. */
|
||
bcopy(fromP, *charPP, (int) length);
|
||
*charPP += length;
|
||
}
|
||
}
|
||
|
||
int section_alignment[SEG_MAXIMUM_ORDINAL];
|
||
|
||
/*
|
||
* This routine records the largest alignment seen for each segment.
|
||
* If the beginning of the segment is aligned on the worst-case
|
||
* boundary, all of the other alignments within it will work. At
|
||
* least one object format really uses this info.
|
||
*/
|
||
void record_alignment(seg, align)
|
||
segT seg; /* Segment to which alignment pertains */
|
||
int align; /* Alignment, as a power of 2
|
||
* (e.g., 1 => 2-byte boundary, 2 => 4-byte boundary, etc.)
|
||
*/
|
||
{
|
||
|
||
if ( align > section_alignment[(int) seg] ){
|
||
section_alignment[(int) seg] = align;
|
||
} /* if highest yet */
|
||
|
||
return;
|
||
} /* record_alignment() */
|
||
|
||
/*
|
||
* Local Variables:
|
||
* comment-column: 0
|
||
* fill-column: 131
|
||
* End:
|
||
*/
|
||
|
||
/* end of write.c */
|