nasm/parser.c

959 lines
32 KiB
C
Raw Normal View History

2002-05-01 04:51:32 +08:00
/* parser.c source line parser for the Netwide Assembler
*
* The Netwide Assembler is copyright (C) 1996 Simon Tatham and
* Julian Hall. All rights reserved. The software is
* redistributable under the license given in the file "LICENSE"
2002-05-01 04:51:32 +08:00
* distributed in the NASM archive.
*
* initial version 27/iii/95 by Simon Tatham
*/
#include "compiler.h"
2002-05-01 04:51:32 +08:00
#include <stdio.h>
#include <stdlib.h>
#include <stddef.h>
#include <string.h>
#include <ctype.h>
#include <inttypes.h>
2002-05-01 04:51:32 +08:00
#include "nasm.h"
2002-09-13 00:34:06 +08:00
#include "insns.h"
2002-05-01 04:51:32 +08:00
#include "nasmlib.h"
#include "stdscan.h"
2002-05-01 04:51:32 +08:00
#include "parser.h"
#include "float.h"
#include "tables.h"
2002-05-01 04:51:32 +08:00
2005-01-16 06:15:51 +08:00
extern int in_abs_seg; /* ABSOLUTE segment flag */
extern int32_t abs_seg; /* ABSOLUTE segment */
extern int32_t abs_offset; /* ABSOLUTE segment offset */
2005-01-16 06:15:51 +08:00
static int is_comma_next(void);
2002-05-01 04:51:32 +08:00
static int i;
static struct tokenval tokval;
static efunc error;
2005-01-16 06:15:51 +08:00
static struct ofmt *outfmt; /* Structure of addresses of output routines */
static struct location *location; /* Pointer to current line's segment,offset */
2002-05-01 04:53:55 +08:00
void parser_global_info(struct ofmt *output, struct location * locp)
2002-05-01 04:53:55 +08:00
{
outfmt = output;
location = locp;
}
2002-05-01 04:51:53 +08:00
static int prefix_slot(enum prefixes prefix)
{
switch (prefix) {
case P_WAIT:
return PPS_WAIT;
case R_CS:
case R_DS:
case R_SS:
case R_ES:
case R_FS:
case R_GS:
return PPS_SEG;
case P_LOCK:
case P_REP:
case P_REPE:
case P_REPZ:
case P_REPNE:
case P_REPNZ:
return PPS_LREP;
case P_O16:
case P_O32:
case P_O64:
case P_OSP:
return PPS_OSIZE;
case P_A16:
case P_A32:
case P_A64:
case P_ASP:
return PPS_ASIZE;
default:
error(ERR_PANIC, "Invalid value %d passed to prefix_slot()", prefix);
return -1;
}
}
static void process_size_override(insn * result, int operand)
{
if (tasm_compatible_mode) {
switch ((int)tokval.t_integer) {
/* For TASM compatibility a size override inside the
* brackets changes the size of the operand, not the
* address type of the operand as it does in standard
* NASM syntax. Hence:
*
* mov eax,[DWORD val]
*
* is valid syntax in TASM compatibility mode. Note that
* you lose the ability to override the default address
* type for the instruction, but we never use anything
* but 32-bit flat model addressing in our code.
*/
case S_BYTE:
result->oprs[operand].type |= BITS8;
break;
case S_WORD:
result->oprs[operand].type |= BITS16;
break;
case S_DWORD:
case S_LONG:
result->oprs[operand].type |= BITS32;
break;
case S_QWORD:
result->oprs[operand].type |= BITS64;
break;
case S_TWORD:
result->oprs[operand].type |= BITS80;
break;
case S_OWORD:
result->oprs[operand].type |= BITS128;
break;
default:
error(ERR_NONFATAL,
"invalid operand size specification");
break;
}
} else {
/* Standard NASM compatible syntax */
switch ((int)tokval.t_integer) {
case S_NOSPLIT:
result->oprs[operand].eaflags |= EAF_TIMESTWO;
break;
case S_REL:
result->oprs[operand].eaflags |= EAF_REL;
break;
case S_ABS:
result->oprs[operand].eaflags |= EAF_ABS;
break;
case S_BYTE:
result->oprs[operand].disp_size = 8;
result->oprs[operand].eaflags |= EAF_BYTEOFFS;
break;
case P_A16:
case P_A32:
case P_A64:
if (result->prefixes[PPS_ASIZE] &&
result->prefixes[PPS_ASIZE] != tokval.t_integer)
error(ERR_NONFATAL,
"conflicting address size specifications");
else
result->prefixes[PPS_ASIZE] = tokval.t_integer;
break;
case S_WORD:
result->oprs[operand].disp_size = 16;
result->oprs[operand].eaflags |= EAF_WORDOFFS;
break;
case S_DWORD:
case S_LONG:
result->oprs[operand].disp_size = 32;
result->oprs[operand].eaflags |= EAF_WORDOFFS;
break;
case S_QWORD:
result->oprs[operand].disp_size = 64;
result->oprs[operand].eaflags |= EAF_WORDOFFS;
break;
default:
error(ERR_NONFATAL, "invalid size specification in"
" effective address");
break;
}
}
}
insn *parse_line(int pass, char *buffer, insn * result,
2005-01-16 06:15:51 +08:00
efunc errfunc, evalfunc evaluate, ldfunc ldef)
2002-05-01 04:53:55 +08:00
{
2002-05-01 04:51:32 +08:00
int operand;
int critical;
2002-05-01 04:52:49 +08:00
struct eval_hints hints;
int j;
bool first;
bool insn_is_label = false;
2002-05-01 04:51:32 +08:00
restart_parse:
first = true;
result->forw_ref = false;
2002-05-01 04:51:32 +08:00
error = errfunc;
2002-05-01 04:52:49 +08:00
stdscan_reset();
stdscan_bufptr = buffer;
i = stdscan(NULL, &tokval);
2002-05-01 04:51:32 +08:00
2005-01-16 06:15:51 +08:00
result->label = NULL; /* Assume no label */
result->eops = NULL; /* must do this, whatever happens */
result->operands = 0; /* must initialize this */
2002-05-01 04:51:32 +08:00
2005-01-16 06:15:51 +08:00
if (i == 0) { /* blank line - ignore */
result->opcode = -1; /* and no instruction either */
return result;
2002-05-01 04:51:32 +08:00
}
if (i != TOKEN_ID && i != TOKEN_INSN && i != TOKEN_PREFIX &&
(i != TOKEN_REG || (REG_SREG & ~nasm_reg_flags[tokval.t_integer]))) {
2005-01-16 06:15:51 +08:00
error(ERR_NONFATAL, "label or instruction expected"
" at start of line");
result->opcode = -1;
return result;
2002-05-01 04:51:32 +08:00
}
if (i == TOKEN_ID || (insn_is_label && i == TOKEN_INSN)) {
/* there's a label here */
first = false;
2005-01-16 06:15:51 +08:00
result->label = tokval.t_charptr;
i = stdscan(NULL, &tokval);
if (i == ':') { /* skip over the optional colon */
i = stdscan(NULL, &tokval);
} else if (i == 0) {
error(ERR_WARNING | ERR_WARN_OL | ERR_PASS1,
"label alone on a line without a colon might be in error");
}
if (i != TOKEN_INSN || tokval.t_integer != I_EQU) {
/*
* FIXME: location->segment could be NO_SEG, in which case
* it is possible we should be passing 'abs_seg'. Look into this.
* Work out whether that is *really* what we should be doing.
* Generally fix things. I think this is right as it is, but
* am still not certain.
*/
ldef(result->label, in_abs_seg ? abs_seg : location->segment,
location->offset, NULL, true, false, outfmt, errfunc);
2005-01-16 06:15:51 +08:00
}
2002-05-01 04:53:55 +08:00
}
2002-05-01 04:51:32 +08:00
2005-01-16 06:15:51 +08:00
if (i == 0) {
result->opcode = -1; /* this line contains just a label */
return result;
2002-05-01 04:51:32 +08:00
}
for (j = 0; j < MAXPREFIX; j++)
result->prefixes[j] = P_none;
2002-05-01 04:52:08 +08:00
result->times = 1L;
2002-05-01 04:51:32 +08:00
while (i == TOKEN_PREFIX ||
(i == TOKEN_REG && !(REG_SREG & ~nasm_reg_flags[tokval.t_integer])))
2002-05-01 04:53:55 +08:00
{
first = false;
2005-01-16 06:15:51 +08:00
/*
* Handle special case: the TIMES prefix.
*/
if (i == TOKEN_PREFIX && tokval.t_integer == P_TIMES) {
expr *value;
i = stdscan(NULL, &tokval);
value =
evaluate(stdscan, NULL, &tokval, NULL, pass0, error, NULL);
i = tokval.t_type;
if (!value) { /* but, error in evaluator */
result->opcode = -1; /* unrecoverable parse error: */
return result; /* ignore this instruction */
}
if (!is_simple(value)) {
error(ERR_NONFATAL,
"non-constant argument supplied to TIMES");
result->times = 1L;
} else {
result->times = value->value;
if (value->value < 0 && pass0 == 2) {
2005-01-16 06:15:51 +08:00
error(ERR_NONFATAL, "TIMES value %d is negative",
value->value);
result->times = 0;
}
}
} else {
int slot = prefix_slot(tokval.t_integer);
if (result->prefixes[slot]) {
if (result->prefixes[slot] == tokval.t_integer)
error(ERR_WARNING,
"instruction has redundant prefixes");
else
error(ERR_NONFATAL,
"instruction has conflicting prefixes");
}
result->prefixes[slot] = tokval.t_integer;
2005-01-16 06:15:51 +08:00
i = stdscan(NULL, &tokval);
}
2002-05-01 04:51:32 +08:00
}
if (i != TOKEN_INSN) {
int j;
enum prefixes pfx;
for (j = 0; j < MAXPREFIX; j++)
if ((pfx = result->prefixes[j]) != P_none)
break;
if (i == 0 && pfx != P_none) {
2005-01-16 06:15:51 +08:00
/*
* Instruction prefixes are present, but no actual
* instruction. This is allowed: at this point we
* invent a notional instruction of RESB 0.
*/
result->opcode = I_RESB;
result->operands = 1;
result->oprs[0].type = IMMEDIATE;
result->oprs[0].offset = 0L;
result->oprs[0].segment = result->oprs[0].wrt = NO_SEG;
return result;
} else {
error(ERR_NONFATAL, "parser: instruction expected");
result->opcode = -1;
return result;
}
2002-05-01 04:51:32 +08:00
}
result->opcode = tokval.t_integer;
result->condition = tokval.t_inttwo;
/*
* INCBIN cannot be satisfied with incorrectly
2002-05-01 04:51:32 +08:00
* evaluated operands, since the correct values _must_ be known
* on the first pass. Hence, even in pass one, we set the
* `critical' flag on calling evaluate(), so that it will bomb
* out on undefined symbols.
2002-05-01 04:51:32 +08:00
*/
if (result->opcode == I_INCBIN) {
critical = (pass0 < 2 ? 1 : 2);
2005-01-16 06:15:51 +08:00
} else
critical = (pass == 2 ? 2 : 0);
2002-05-01 04:51:32 +08:00
if (result->opcode == I_DB || result->opcode == I_DW ||
result->opcode == I_DD || result->opcode == I_DQ ||
result->opcode == I_DT || result->opcode == I_DO ||
result->opcode == I_DY || result->opcode == I_INCBIN) {
2005-01-16 06:15:51 +08:00
extop *eop, **tail = &result->eops, **fixptr;
int oper_num = 0;
int32_t sign;
2005-01-16 06:15:51 +08:00
result->eops_float = false;
2005-01-16 06:15:51 +08:00
/*
* Begin to read the DB/DW/DD/DQ/DT/DO/INCBIN operands.
2005-01-16 06:15:51 +08:00
*/
while (1) {
i = stdscan(NULL, &tokval);
if (i == 0)
break;
else if (first && i == ':') {
insn_is_label = true;
goto restart_parse;
}
first = false;
2005-01-16 06:15:51 +08:00
fixptr = tail;
eop = *tail = nasm_malloc(sizeof(extop));
tail = &eop->next;
eop->next = NULL;
eop->type = EOT_NOTHING;
oper_num++;
sign = +1;
2005-01-16 06:15:51 +08:00
/* is_comma_next() here is to distinguish this from
a string used as part of an expression... */
if (i == TOKEN_STR && is_comma_next()) {
2005-01-16 06:15:51 +08:00
eop->type = EOT_DB_STRING;
eop->stringval = tokval.t_charptr;
eop->stringlen = tokval.t_inttwo;
i = stdscan(NULL, &tokval); /* eat the comma */
} else if (i == TOKEN_STRFUNC) {
bool parens = false;
const char *funcname = tokval.t_charptr;
enum strfunc func = tokval.t_integer;
i = stdscan(NULL, &tokval);
if (i == '(') {
parens = true;
i = stdscan(NULL, &tokval);
}
if (i != TOKEN_STR) {
error(ERR_NONFATAL,
"%s must be followed by a string constant",
funcname);
eop->type = EOT_NOTHING;
} else {
eop->type = EOT_DB_STRING_FREE;
eop->stringlen =
string_transform(tokval.t_charptr, tokval.t_inttwo,
&eop->stringval, func);
if (eop->stringlen == (size_t)-1) {
error(ERR_NONFATAL, "invalid string for transform");
eop->type = EOT_NOTHING;
}
}
if (parens && i && i != ')') {
i = stdscan(NULL, &tokval);
if (i != ')') {
error(ERR_NONFATAL, "unterminated %s function",
funcname);
}
}
if (i && i != ',')
i = stdscan(NULL, &tokval);
} else if (i == '-' || i == '+') {
char *save = stdscan_bufptr;
int token = i;
sign = (i == '-') ? -1 : 1;
i = stdscan(NULL, &tokval);
if (i != TOKEN_FLOAT) {
stdscan_bufptr = save;
i = tokval.t_type = token;
goto is_expression;
} else {
goto is_float;
}
} else if (i == TOKEN_FLOAT) {
is_float:
eop->type = EOT_DB_STRING;
result->eops_float = true;
switch (result->opcode) {
case I_DB:
eop->stringlen = 1;
break;
case I_DW:
eop->stringlen = 2;
break;
case I_DD:
eop->stringlen = 4;
break;
case I_DQ:
eop->stringlen = 8;
break;
case I_DT:
eop->stringlen = 10;
break;
case I_DO:
eop->stringlen = 16;
break;
case I_DY:
error(ERR_NONFATAL, "floating-point constant"
" encountered in DY instruction");
eop->stringlen = 0;
break;
default:
error(ERR_NONFATAL, "floating-point constant"
" encountered in unknown instruction");
/*
* fix suggested by Pedro Gimeno... original line
* was:
* eop->type = EOT_NOTHING;
*/
eop->stringlen = 0;
break;
}
eop = nasm_realloc(eop, sizeof(extop) + eop->stringlen);
tail = &eop->next;
*fixptr = eop;
eop->stringval = (char *)eop + sizeof(extop);
if (!eop->stringlen ||
!float_const(tokval.t_charptr, sign,
(uint8_t *)eop->stringval,
eop->stringlen, error))
eop->type = EOT_NOTHING;
i = stdscan(NULL, &tokval); /* eat the comma */
} else {
/* anything else, assume it is an expression */
2005-01-16 06:15:51 +08:00
expr *value;
is_expression:
2005-01-16 06:15:51 +08:00
value = evaluate(stdscan, NULL, &tokval, NULL,
critical, error, NULL);
i = tokval.t_type;
if (!value) { /* error in evaluator */
result->opcode = -1; /* unrecoverable parse error: */
return result; /* ignore this instruction */
}
if (is_unknown(value)) {
eop->type = EOT_DB_NUMBER;
eop->offset = 0; /* doesn't matter what we put */
eop->segment = eop->wrt = NO_SEG; /* likewise */
} else if (is_reloc(value)) {
eop->type = EOT_DB_NUMBER;
eop->offset = reloc_value(value);
eop->segment = reloc_seg(value);
eop->wrt = reloc_wrt(value);
} else {
error(ERR_NONFATAL,
"operand %d: expression is not simple"
" or relocatable", oper_num);
}
}
/*
* We're about to call stdscan(), which will eat the
* comma that we're currently sitting on between
* arguments. However, we'd better check first that it
* _is_ a comma.
*/
if (i == 0) /* also could be EOL */
break;
if (i != ',') {
error(ERR_NONFATAL, "comma expected after operand %d",
oper_num);
result->opcode = -1; /* unrecoverable parse error: */
return result; /* ignore this instruction */
}
}
if (result->opcode == I_INCBIN) {
/*
* Correct syntax for INCBIN is that there should be
* one string operand, followed by one or two numeric
* operands.
*/
if (!result->eops || result->eops->type != EOT_DB_STRING)
error(ERR_NONFATAL, "`incbin' expects a file name");
else if (result->eops->next &&
result->eops->next->type != EOT_DB_NUMBER)
error(ERR_NONFATAL, "`incbin': second parameter is",
" non-numeric");
else if (result->eops->next && result->eops->next->next &&
result->eops->next->next->type != EOT_DB_NUMBER)
error(ERR_NONFATAL, "`incbin': third parameter is",
" non-numeric");
else if (result->eops->next && result->eops->next->next &&
result->eops->next->next->next)
error(ERR_NONFATAL,
"`incbin': more than three parameters");
2002-05-01 04:53:55 +08:00
else
2005-01-16 06:15:51 +08:00
return result;
/*
* If we reach here, one of the above errors happened.
* Throw the instruction away.
*/
result->opcode = -1;
return result;
} else /* DB ... */ if (oper_num == 0)
error(ERR_WARNING | ERR_PASS1,
"no operand for data declaration");
else
result->operands = oper_num;
return result;
2002-05-01 04:51:32 +08:00
}
/* right. Now we begin to parse the operands. There may be up to four
2002-05-01 04:51:32 +08:00
* of these, separated by commas, and terminated by a zero token. */
for (operand = 0; operand < MAX_OPERANDS; operand++) {
expr *value; /* used most of the time */
2005-01-16 06:15:51 +08:00
int mref; /* is this going to be a memory ref? */
int bracket; /* is it a [] mref, or a & mref? */
int setsize = 0;
result->oprs[operand].disp_size = 0; /* have to zero this whatever */
2005-01-16 06:15:51 +08:00
result->oprs[operand].eaflags = 0; /* and this */
result->oprs[operand].opflags = 0;
i = stdscan(NULL, &tokval);
if (i == 0)
break; /* end of operands: get out of here */
else if (first && i == ':') {
insn_is_label = true;
goto restart_parse;
}
first = false;
2005-01-16 06:15:51 +08:00
result->oprs[operand].type = 0; /* so far, no override */
while (i == TOKEN_SPECIAL) { /* size specifiers */
switch ((int)tokval.t_integer) {
case S_BYTE:
if (!setsize) /* we want to use only the first */
result->oprs[operand].type |= BITS8;
setsize = 1;
break;
case S_WORD:
if (!setsize)
result->oprs[operand].type |= BITS16;
setsize = 1;
break;
case S_DWORD:
case S_LONG:
if (!setsize)
result->oprs[operand].type |= BITS32;
setsize = 1;
break;
case S_QWORD:
if (!setsize)
result->oprs[operand].type |= BITS64;
setsize = 1;
break;
case S_TWORD:
if (!setsize)
result->oprs[operand].type |= BITS80;
setsize = 1;
break;
case S_OWORD:
if (!setsize)
result->oprs[operand].type |= BITS128;
setsize = 1;
break;
case S_YWORD:
if (!setsize)
result->oprs[operand].type |= BITS256;
setsize = 1;
break;
2005-01-16 06:15:51 +08:00
case S_TO:
result->oprs[operand].type |= TO;
break;
case S_STRICT:
result->oprs[operand].type |= STRICT;
break;
case S_FAR:
result->oprs[operand].type |= FAR;
break;
case S_NEAR:
result->oprs[operand].type |= NEAR;
break;
case S_SHORT:
result->oprs[operand].type |= SHORT;
break;
default:
error(ERR_NONFATAL, "invalid operand size specification");
}
i = stdscan(NULL, &tokval);
}
if (i == '[' || i == '&') { /* memory reference */
mref = true;
2005-01-16 06:15:51 +08:00
bracket = (i == '[');
i = stdscan(NULL, &tokval); /* then skip the colon */
while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
process_size_override(result, operand);
i = stdscan(NULL, &tokval);
2005-01-16 06:15:51 +08:00
}
} else { /* immediate operand, or register */
mref = false;
bracket = false; /* placate optimisers */
2005-01-16 06:15:51 +08:00
}
if ((result->oprs[operand].type & FAR) && !mref &&
result->opcode != I_JMP && result->opcode != I_CALL) {
error(ERR_NONFATAL, "invalid use of FAR operand specifier");
}
value = evaluate(stdscan, NULL, &tokval,
&result->oprs[operand].opflags,
critical, error, &hints);
i = tokval.t_type;
if (result->oprs[operand].opflags & OPFLAG_FORWARD) {
result->forw_ref = true;
2005-01-16 06:15:51 +08:00
}
if (!value) { /* error in evaluator */
result->opcode = -1; /* unrecoverable parse error: */
return result; /* ignore this instruction */
}
if (i == ':' && mref) { /* it was seg:offset */
/*
* Process the segment override.
*/
if (value[1].type != 0 || value->value != 1 ||
REG_SREG & ~nasm_reg_flags[value->type])
2005-01-16 06:15:51 +08:00
error(ERR_NONFATAL, "invalid segment override");
else if (result->prefixes[PPS_SEG])
2005-01-16 06:15:51 +08:00
error(ERR_NONFATAL,
"instruction has conflicting segment overrides");
else {
result->prefixes[PPS_SEG] = value->type;
if (!(REG_FSGS & ~nasm_reg_flags[value->type]))
result->oprs[operand].eaflags |= EAF_FSGS;
}
2005-01-16 06:15:51 +08:00
i = stdscan(NULL, &tokval); /* then skip the colon */
while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
process_size_override(result, operand);
2005-01-16 06:15:51 +08:00
i = stdscan(NULL, &tokval);
}
value = evaluate(stdscan, NULL, &tokval,
&result->oprs[operand].opflags,
critical, error, &hints);
i = tokval.t_type;
if (result->oprs[operand].opflags & OPFLAG_FORWARD) {
result->forw_ref = true;
2005-01-16 06:15:51 +08:00
}
/* and get the offset */
if (!value) { /* but, error in evaluator */
result->opcode = -1; /* unrecoverable parse error: */
return result; /* ignore this instruction */
}
}
if (mref && bracket) { /* find ] at the end */
if (i != ']') {
error(ERR_NONFATAL, "parser: expecting ]");
do { /* error recovery again */
i = stdscan(NULL, &tokval);
} while (i != 0 && i != ',');
} else /* we got the required ] */
i = stdscan(NULL, &tokval);
} else { /* immediate operand */
if (i != 0 && i != ',' && i != ':') {
error(ERR_NONFATAL, "comma or end of line expected");
do { /* error recovery */
i = stdscan(NULL, &tokval);
} while (i != 0 && i != ',');
} else if (i == ':') {
result->oprs[operand].type |= COLON;
}
}
/* now convert the exprs returned from evaluate() into operand
* descriptions... */
if (mref) { /* it's a memory reference */
expr *e = value;
int b, i, s; /* basereg, indexreg, scale */
2007-04-16 12:56:06 +08:00
int64_t o; /* offset */
2005-01-16 06:15:51 +08:00
b = i = -1, o = s = 0;
result->oprs[operand].hintbase = hints.base;
result->oprs[operand].hinttype = hints.type;
if (e->type && e->type <= EXPR_REG_END) { /* this bit's a register */
if (e->value == 1) /* in fact it can be basereg */
b = e->type;
else /* no, it has to be indexreg */
i = e->type, s = e->value;
e++;
}
if (e->type && e->type <= EXPR_REG_END) { /* it's a 2nd register */
if (b != -1) /* If the first was the base, ... */
i = e->type, s = e->value; /* second has to be indexreg */
else if (e->value != 1) { /* If both want to be index */
error(ERR_NONFATAL,
"beroset-p-592-invalid effective address");
result->opcode = -1;
return result;
} else
b = e->type;
e++;
}
if (e->type != 0) { /* is there an offset? */
if (e->type <= EXPR_REG_END) { /* in fact, is there an error? */
error(ERR_NONFATAL,
"beroset-p-603-invalid effective address");
result->opcode = -1;
return result;
} else {
if (e->type == EXPR_UNKNOWN) {
result->oprs[operand].opflags |= OPFLAG_UNKNOWN;
2005-01-16 06:15:51 +08:00
o = 0; /* doesn't matter what */
result->oprs[operand].wrt = NO_SEG; /* nor this */
result->oprs[operand].segment = NO_SEG; /* or this */
while (e->type)
e++; /* go to the end of the line */
} else {
if (e->type == EXPR_SIMPLE) {
o = e->value;
e++;
}
if (e->type == EXPR_WRT) {
result->oprs[operand].wrt = e->value;
e++;
} else
result->oprs[operand].wrt = NO_SEG;
/*
* Look for a segment base type.
*/
if (e->type && e->type < EXPR_SEGBASE) {
error(ERR_NONFATAL,
"beroset-p-630-invalid effective address");
result->opcode = -1;
return result;
}
while (e->type && e->value == 0)
e++;
if (e->type && e->value != 1) {
error(ERR_NONFATAL,
"beroset-p-637-invalid effective address");
result->opcode = -1;
return result;
}
if (e->type) {
result->oprs[operand].segment =
e->type - EXPR_SEGBASE;
e++;
} else
result->oprs[operand].segment = NO_SEG;
while (e->type && e->value == 0)
e++;
if (e->type) {
error(ERR_NONFATAL,
"beroset-p-650-invalid effective address");
result->opcode = -1;
return result;
}
}
}
} else {
o = 0;
result->oprs[operand].wrt = NO_SEG;
result->oprs[operand].segment = NO_SEG;
}
if (e->type != 0) { /* there'd better be nothing left! */
error(ERR_NONFATAL,
"beroset-p-663-invalid effective address");
result->opcode = -1;
return result;
}
/* It is memory, but it can match any r/m operand */
result->oprs[operand].type |= MEMORY_ANY;
if (b == -1 && (i == -1 || s == 0)) {
int is_rel = globalbits == 64 &&
!(result->oprs[operand].eaflags & EAF_ABS) &&
((globalrel &&
!(result->oprs[operand].eaflags & EAF_FSGS)) ||
(result->oprs[operand].eaflags & EAF_REL));
result->oprs[operand].type |= is_rel ? IP_REL : MEM_OFFS;
}
2005-01-16 06:15:51 +08:00
result->oprs[operand].basereg = b;
result->oprs[operand].indexreg = i;
result->oprs[operand].scale = s;
result->oprs[operand].offset = o;
} else { /* it's not a memory reference */
if (is_just_unknown(value)) { /* it's immediate but unknown */
result->oprs[operand].type |= IMMEDIATE;
result->oprs[operand].opflags |= OPFLAG_UNKNOWN;
2005-01-16 06:15:51 +08:00
result->oprs[operand].offset = 0; /* don't care */
result->oprs[operand].segment = NO_SEG; /* don't care again */
result->oprs[operand].wrt = NO_SEG; /* still don't care */
if(optimizing >= 0 && !(result->oprs[operand].type & STRICT))
{
/* Be optimistic */
result->oprs[operand].type |= SBYTE16 | SBYTE32 | SBYTE64;
}
2005-01-16 06:15:51 +08:00
} else if (is_reloc(value)) { /* it's immediate */
result->oprs[operand].type |= IMMEDIATE;
result->oprs[operand].offset = reloc_value(value);
result->oprs[operand].segment = reloc_seg(value);
result->oprs[operand].wrt = reloc_wrt(value);
if (is_simple(value)) {
if (reloc_value(value) == 1)
result->oprs[operand].type |= UNITY;
if (optimizing >= 0 &&
!(result->oprs[operand].type & STRICT)) {
int64_t v64 = reloc_value(value);
int32_t v32 = (int32_t)v64;
int16_t v16 = (int16_t)v32;
if (v64 >= -128 && v64 <= 127)
result->oprs[operand].type |= SBYTE64;
if (v32 >= -128 && v32 <= 127)
result->oprs[operand].type |= SBYTE32;
if (v16 >= -128 && v16 <= 127)
result->oprs[operand].type |= SBYTE16;
2005-01-16 06:15:51 +08:00
}
}
} else { /* it's a register */
unsigned int rs;
2005-01-16 06:15:51 +08:00
if (value->type >= EXPR_SIMPLE || value->value != 1) {
error(ERR_NONFATAL, "invalid operand type");
result->opcode = -1;
return result;
}
/*
* check that its only 1 register, not an expression...
*/
for (i = 1; value[i].type; i++)
if (value[i].value) {
error(ERR_NONFATAL, "invalid operand type");
result->opcode = -1;
return result;
}
/* clear overrides, except TO which applies to FPU regs */
if (result->oprs[operand].type & ~TO) {
/*
* we want to produce a warning iff the specified size
* is different from the register size
*/
rs = result->oprs[operand].type & SIZE_MASK;
2005-01-16 06:15:51 +08:00
} else
rs = 0;
2005-01-16 06:15:51 +08:00
result->oprs[operand].type &= TO;
result->oprs[operand].type |= REGISTER;
result->oprs[operand].type |= nasm_reg_flags[value->type];
2005-01-16 06:15:51 +08:00
result->oprs[operand].basereg = value->type;
if (rs && (result->oprs[operand].type & SIZE_MASK) != rs)
2005-01-16 06:15:51 +08:00
error(ERR_WARNING | ERR_PASS1,
"register size specification ignored");
}
}
2002-05-01 04:51:32 +08:00
}
2005-01-16 06:15:51 +08:00
result->operands = operand; /* set operand count */
2002-05-01 04:51:32 +08:00
/* clear remaining operands */
while (operand < MAX_OPERANDS)
result->oprs[operand++].type = 0;
2002-05-01 04:51:32 +08:00
/*
* Transform RESW, RESD, RESQ, REST, RESO, RESY into RESB.
2002-05-01 04:51:32 +08:00
*/
switch (result->opcode) {
2005-01-16 06:15:51 +08:00
case I_RESW:
result->opcode = I_RESB;
result->oprs[0].offset *= 2;
break;
case I_RESD:
result->opcode = I_RESB;
result->oprs[0].offset *= 4;
break;
case I_RESQ:
result->opcode = I_RESB;
result->oprs[0].offset *= 8;
break;
case I_REST:
result->opcode = I_RESB;
result->oprs[0].offset *= 10;
break;
case I_RESO:
result->opcode = I_RESB;
result->oprs[0].offset *= 16;
break;
case I_RESY:
result->opcode = I_RESB;
result->oprs[0].offset *= 32;
break;
default:
break;
2002-05-01 04:51:32 +08:00
}
return result;
}
2005-01-16 06:15:51 +08:00
static int is_comma_next(void)
2002-05-01 04:53:55 +08:00
{
char *p;
2002-05-01 04:52:49 +08:00
int i;
struct tokenval tv;
2002-05-01 04:51:32 +08:00
2002-05-01 04:52:49 +08:00
p = stdscan_bufptr;
2005-01-16 06:15:51 +08:00
i = stdscan(NULL, &tv);
2002-05-01 04:52:49 +08:00
stdscan_bufptr = p;
return (i == ',' || i == ';' || !i);
2002-05-01 04:51:32 +08:00
}
2005-01-16 06:15:51 +08:00
void cleanup_insn(insn * i)
2002-05-01 04:53:55 +08:00
{
2002-05-01 04:51:32 +08:00
extop *e;
while ((e = i->eops)) {
i->eops = e->next;
if (e->type == EOT_DB_STRING_FREE)
nasm_free(e->stringval);
2005-01-16 06:15:51 +08:00
nasm_free(e);
2002-05-01 04:51:32 +08:00
}
}