/*
* The SEG operator: calculate the segment part of a relocatable
* value. Return NULL, as usual, if an error occurs. Report the
* error too.
*/
static expr *segment_part(expr * e)
{
int32_t seg;
if (is_unknown(e))
return unknown_expr();
if (!is_reloc(e)) {
nasm_error(ERR_NONFATAL, "cannot apply SEG to a non-relocatable value");
return NULL;
}
seg = reloc_seg(e);
if (seg == NO_SEG) {
nasm_error(ERR_NONFATAL, "cannot apply SEG to a non-relocatable value");
return NULL;
} else if (seg & SEG_ABS) {
return scalarvect(seg & ~SEG_ABS);
} else if (seg & 1) {
nasm_error(ERR_NONFATAL, "SEG applied to something which"
" is already a segment base");
return NULL;
} else {
int32_t base = ofmt->segbase(seg + 1);
begintemp();
addtotemp((base == NO_SEG ? EXPR_UNKNOWN : EXPR_SEGBASE + base),
1L);
return finishtemp();
}
}
insn *parse_line (int pass, char *buffer, insn *result,
efunc errfunc, evalfunc evaluate, evalinfofunc einfo) {
int operand;
int critical;
struct eval_hints hints;
result->forw_ref = FALSE;
error = errfunc;
einfo ("", 0L, 0L);
stdscan_reset();
stdscan_bufptr = buffer;
i = stdscan(NULL, &tokval);
result->eops = NULL; /* must do this, whatever happens */
result->operands = 0; /* must initialise this */
if (i==0) { /* blank line - ignore */
result->label = NULL; /* so, no label on it */
result->opcode = -1; /* and no instruction either */
return result;
}
if (i != TOKEN_ID && i != TOKEN_INSN && i != TOKEN_PREFIX &&
(i!=TOKEN_REG || (REG_SREG & ~reg_flags[tokval.t_integer]))) {
error (ERR_NONFATAL, "label or instruction expected"
" at start of line");
result->label = NULL;
result->opcode = -1;
return result;
}
if (i == TOKEN_ID) { /* there's a label here */
result->label = tokval.t_charptr;
einfo (result->label, 0L, 0L);
i = stdscan(NULL, &tokval);
if (i == ':') { /* skip over the optional colon */
i = stdscan(NULL, &tokval);
} else if (i == 0 && pass == 1) {
error (ERR_WARNING|ERR_WARN_OL,
"label alone on a line without a colon might be in error");
}
} else /* no label; so, moving swiftly on */
result->label = NULL;
if (i==0) {
result->opcode = -1; /* this line contains just a label */
return result;
}
result->nprefix = 0;
result->times = 1L;
while (i == TOKEN_PREFIX ||
(i==TOKEN_REG && !(REG_SREG & ~reg_flags[tokval.t_integer]))) {
/*
* 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, pass, 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)
error(ERR_NONFATAL, "TIMES value %d is negative",
value->value);
}
} else {
if (result->nprefix == MAXPREFIX)
error (ERR_NONFATAL,
"instruction has more than %d prefixes", MAXPREFIX);
else
result->prefixes[result->nprefix++] = tokval.t_integer;
i = stdscan(NULL, &tokval);
}
}
if (i != TOKEN_INSN) {
if (result->nprefix > 0 && i == 0) {
/*
* 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;
}
}
result->opcode = tokval.t_integer;
result->condition = tokval.t_inttwo;
/*
* RESB, RESW and RESD cannot be satisfied with incorrectly
* 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. Nasty, but there's nothing we can
* do about it.
*
* For the moment, EQU has the same difficulty, so we'll
* include that.
*/
if (result->opcode == I_RESB ||
result->opcode == I_RESW ||
result->opcode == I_RESD ||
result->opcode == I_RESQ ||
result->opcode == I_REST ||
result->opcode == I_EQU)
critical = pass;
else
critical = (pass==2 ? 2 : 0);
if (result->opcode == I_DB ||
result->opcode == I_DW ||
result->opcode == I_DD ||
result->opcode == I_DQ ||
result->opcode == I_DT ||
result->opcode == I_INCBIN) {
extop *eop, **tail = &result->eops, **fixptr;
int oper_num = 0;
/*
* Begin to read the DB/DW/DD/DQ/DT operands.
*/
while (1) {
i = stdscan(NULL, &tokval);
if (i == 0)
break;
fixptr = tail;
eop = *tail = nasm_malloc(sizeof(extop));
tail = &eop->next;
eop->next = NULL;
eop->type = EOT_NOTHING;
oper_num++;
if (i == TOKEN_NUM && tokval.t_charptr && is_comma_next()) {
eop->type = EOT_DB_STRING;
eop->stringval = tokval.t_charptr;
eop->stringlen = tokval.t_inttwo;
i = stdscan(NULL, &tokval); /* eat the comma */
continue;
}
if (i == TOKEN_FLOAT || i == '-') {
long sign = +1L;
if (i == '-') {
char *save = stdscan_bufptr;
i = stdscan(NULL, &tokval);
sign = -1L;
if (i != TOKEN_FLOAT) {
stdscan_bufptr = save;
i = tokval.t_type = '-';
}
}
if (i == TOKEN_FLOAT) {
eop->type = EOT_DB_STRING;
if (result->opcode == I_DD)
eop->stringlen = 4;
else if (result->opcode == I_DQ)
eop->stringlen = 8;
else if (result->opcode == I_DT)
eop->stringlen = 10;
else {
error(ERR_NONFATAL, "floating-point constant"
" encountered in `D%c' instruction",
result->opcode == I_DW ? 'W' : 'B');
eop->type = EOT_NOTHING;
}
eop = nasm_realloc(eop, sizeof(extop)+eop->stringlen);
tail = &eop->next;
*fixptr = eop;
eop->stringval = (char *)eop + sizeof(extop);
if (!float_const (tokval.t_charptr, sign,
(unsigned char *)eop->stringval,
eop->stringlen, error))
eop->type = EOT_NOTHING;
i = stdscan(NULL, &tokval); /* eat the comma */
continue;
}
}
/* anything else */ {
expr *value;
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");
else
return result;
/*
* If we reach here, one of the above errors happened.
* Throw the instruction away.
*/
result->opcode = -1;
return result;
}
return result;
}
/* right. Now we begin to parse the operands. There may be up to three
* of these, separated by commas, and terminated by a zero token. */
for (operand = 0; operand < 3; operand++) {
expr *value; /* used most of the time */
int mref; /* is this going to be a memory ref? */
int bracket; /* is it a [] mref, or a & mref? */
result->oprs[operand].addr_size = 0;/* have to zero this whatever */
result->oprs[operand].eaflags = 0; /* and this */
i = stdscan(NULL, &tokval);
if (i == 0) break; /* end of operands: get out of here */
result->oprs[operand].type = 0; /* so far, no override */
while (i == TOKEN_SPECIAL) {/* size specifiers */
switch ((int)tokval.t_integer) {
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_TO:
result->oprs[operand].type |= TO;
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;
}
i = stdscan(NULL, &tokval);
}
if (i == '[' || i == '&') { /* memory reference */
mref = TRUE;
bracket = (i == '[');
i = stdscan(NULL, &tokval);
if (i == TOKEN_SPECIAL) { /* check for address size override */
switch ((int)tokval.t_integer) {
case S_NOSPLIT:
result->oprs[operand].eaflags |= EAF_TIMESTWO;
break;
case S_BYTE:
result->oprs[operand].eaflags |= EAF_BYTEOFFS;
break;
case S_WORD:
result->oprs[operand].addr_size = 16;
result->oprs[operand].eaflags |= EAF_WORDOFFS;
break;
case S_DWORD:
case S_LONG:
result->oprs[operand].addr_size = 32;
result->oprs[operand].eaflags |= EAF_WORDOFFS;
break;
default:
error (ERR_NONFATAL, "invalid size specification in"
" effective address");
}
i = stdscan(NULL, &tokval);
}
} else { /* immediate operand, or register */
mref = FALSE;
bracket = FALSE; /* placate optimisers */
}
value = evaluate (stdscan, NULL, &tokval,
&result->forw_ref, critical, error, &hints);
i = tokval.t_type;
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 & ~reg_flags[value->type])
error (ERR_NONFATAL, "invalid segment override");
else if (result->nprefix == MAXPREFIX)
error (ERR_NONFATAL,
"instruction has more than %d prefixes",
MAXPREFIX);
else
result->prefixes[result->nprefix++] = value->type;
i = stdscan(NULL, &tokval); /* then skip the colon */
if (i == TOKEN_SPECIAL) { /* another check for size override */
switch ((int)tokval.t_integer) {
case S_WORD:
result->oprs[operand].addr_size = 16;
break;
case S_DWORD:
case S_LONG:
result->oprs[operand].addr_size = 32;
break;
default:
error (ERR_NONFATAL, "invalid size specification in"
" effective address");
}
i = stdscan(NULL, &tokval);
}
value = evaluate (stdscan, NULL, &tokval,
&result->forw_ref, critical, error, &hints);
i = tokval.t_type;
/* 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 */
long o; /* offset */
b = i = -1, o = s = 0;
result->oprs[operand].hintbase = hints.base;
result->oprs[operand].hinttype = hints.type;
if (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 (e->value != 1) { /* it has to be indexreg */
if (i != -1) { /* but it can't be */
error(ERR_NONFATAL, "invalid effective address");
result->opcode = -1;
return result;
} else
i = e->type, s = e->value;
} else { /* it can be basereg */
if (b != -1) /* or can it? */
i = e->type, s = 1;
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, "invalid effective address");
result->opcode = -1;
return result;
} else {
if (e->type == EXPR_UNKNOWN) {
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, "invalid effective address");
result->opcode = -1;
return result;
}
while (e->type && e->value == 0)
e++;
if (e->type && e->value != 1) {
error (ERR_NONFATAL, "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, "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, "invalid effective address");
result->opcode = -1;
return result;
}
result->oprs[operand].type |= MEMORY;
if (b==-1 && (i==-1 || s==0))
result->oprs[operand].type |= MEM_OFFS;
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].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 */
} 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) && reloc_value(value)==1)
result->oprs[operand].type |= UNITY;
} else { /* it's a register */
if (value->type>=EXPR_SIMPLE || value->value!=1) {
error (ERR_NONFATAL, "invalid operand type");
result->opcode = -1;
return result;
}
/* clear overrides, except TO which applies to FPU regs */
result->oprs[operand].type &= TO;
result->oprs[operand].type |= REGISTER;
result->oprs[operand].type |= reg_flags[value->type];
result->oprs[operand].basereg = value->type;
}
}
}
result->operands = operand; /* set operand count */
while (operand<3) /* clear remaining operands */
result->oprs[operand++].type = 0;
/*
* Transform RESW, RESD, RESQ, REST into RESB.
*/
switch (result->opcode) {
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;
}
return result;
}
insn *parse_line(char *buffer, insn *result)
{
int pass = 1; /* This used to be an argument. it's hardcoded now because we always want it as 1 */
bool insn_is_label = false;
struct eval_hints hints;
int operand;
int critical;
bool first;
bool recover;
int j;
restart_parse:
first = true;
result->forw_ref = false;
stdscan_reset();
stdscan_set(buffer);
i = stdscan(NULL, &tokval);
result->label = NULL; /* Assume no label */
result->eops = NULL; /* must do this, whatever happens */
result->operands = 0; /* must initialize this */
/* Ignore blank lines */
if (i == TOKEN_EOS) {
result->opcode = I_none;
return result;
}
if (i != TOKEN_ID &&
i != TOKEN_INSN &&
i != TOKEN_PREFIX &&
(i != TOKEN_REG || !IS_SREG(tokval.t_integer))) {
nasm_error(ERR_NONFATAL,
"label or instruction expected at start of line");
result->opcode = I_none;
return result;
}
if (i == TOKEN_ID || (insn_is_label && i == TOKEN_INSN)) {
/* there's a label here */
first = false;
result->label = tokval.t_charptr;
i = stdscan(NULL, &tokval);
if (i == ':') { /* skip over the optional colon */
i = stdscan(NULL, &tokval);
} else if (i == 0) {
nasm_error(ERR_WARNING | ERR_WARN_OL | ERR_PASS1,
"label alone on a line without a colon might be in error");
}
}
/* Just a label here */
if (i == TOKEN_EOS) {
result->opcode = I_none;
return result;
}
for (j = 0; j < MAXPREFIX; j++)
result->prefixes[j] = P_none;
result->times = 1L;
while (i == TOKEN_PREFIX ||
(i == TOKEN_REG && IS_SREG(tokval.t_integer))) {
first = false;
/*
* 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, nasm_error, NULL);
i = tokval.t_type;
if (!value) { /* but, error in evaluator */
result->opcode = I_none; /* unrecoverable parse error: */
return result; /* ignore this instruction */
}
if (!is_simple(value)) {
nasm_error(ERR_NONFATAL,
"non-constant argument supplied to TIMES");
result->times = 1L;
} else {
result->times = value->value;
if (value->value < 0 && pass0 == 2) {
nasm_error(ERR_NONFATAL, "TIMES value %"PRId64" 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)
nasm_error(ERR_WARNING | ERR_PASS1,
"instruction has redundant prefixes");
else
nasm_error(ERR_NONFATAL,
"instruction has conflicting prefixes");
}
result->prefixes[slot] = tokval.t_integer;
i = stdscan(NULL, &tokval);
}
}
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) {
/*
* 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 {
nasm_error(ERR_NONFATAL, "parser: instruction expected");
result->opcode = I_none;
return result;
}
}
result->opcode = tokval.t_integer;
result->condition = tokval.t_inttwo;
/*
* INCBIN cannot be satisfied with incorrectly
* 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.
*/
if (result->opcode == I_INCBIN) {
critical = (pass0 < 2 ? 1 : 2);
} else
critical = (pass == 2 ? 2 : 0);
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) {
extop *eop, **tail = &result->eops, **fixptr;
int oper_num = 0;
int32_t sign;
result->eops_float = false;
/*
* Begin to read the DB/DW/DD/DQ/DT/DO/INCBIN operands.
*/
while (1) {
i = stdscan(NULL, &tokval);
if (i == TOKEN_EOS)
break;
else if (first && i == ':') {
insn_is_label = true;
goto restart_parse;
}
first = false;
fixptr = tail;
eop = *tail = nasm_malloc(sizeof(extop));
tail = &eop->next;
eop->next = NULL;
eop->type = EOT_NOTHING;
oper_num++;
sign = +1;
/*
* is_comma_next() here is to distinguish this from
* a string used as part of an expression...
*/
if (i == TOKEN_STR && is_comma_next()) {
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) {
nasm_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) {
nasm_error(ERR_NONFATAL, "invalid string for transform");
eop->type = EOT_NOTHING;
}
}
if (parens && i && i != ')') {
i = stdscan(NULL, &tokval);
if (i != ')') {
nasm_error(ERR_NONFATAL, "unterminated %s function",
funcname);
}
}
if (i && i != ',')
i = stdscan(NULL, &tokval);
} else if (i == '-' || i == '+') {
char *save = stdscan_get();
int token = i;
sign = (i == '-') ? -1 : 1;
i = stdscan(NULL, &tokval);
if (i != TOKEN_FLOAT) {
stdscan_set(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;
eop->stringlen = idata_bytes(result->opcode);
if (eop->stringlen > 16) {
nasm_error(ERR_NONFATAL, "floating-point constant"
" encountered in DY instruction");
eop->stringlen = 0;
} else if (eop->stringlen < 1) {
nasm_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;
}
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, nasm_error))
eop->type = EOT_NOTHING;
i = stdscan(NULL, &tokval); /* eat the comma */
} else {
/* anything else, assume it is an expression */
expr *value;
is_expression:
value = evaluate(stdscan, NULL, &tokval, NULL,
critical, nasm_error, NULL);
i = tokval.t_type;
if (!value) { /* error in evaluator */
result->opcode = I_none; /* 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 {
nasm_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 == TOKEN_EOS) /* also could be EOL */
break;
if (i != ',') {
nasm_error(ERR_NONFATAL, "comma expected after operand %d",
oper_num);
result->opcode = I_none;/* 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)
nasm_error(ERR_NONFATAL, "`incbin' expects a file name");
else if (result->eops->next &&
result->eops->next->type != EOT_DB_NUMBER)
nasm_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)
nasm_error(ERR_NONFATAL, "`incbin': third parameter is"
" non-numeric");
else if (result->eops->next && result->eops->next->next &&
result->eops->next->next->next)
nasm_error(ERR_NONFATAL,
"`incbin': more than three parameters");
else
return result;
/*
* If we reach here, one of the above errors happened.
* Throw the instruction away.
*/
result->opcode = I_none;
return result;
} else /* DB ... */ if (oper_num == 0)
nasm_error(ERR_WARNING | ERR_PASS1,
"no operand for data declaration");
else
result->operands = oper_num;
return result;
}
/*
* Now we begin to parse the operands. There may be up to four
* 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 */
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 */
result->oprs[operand].eaflags = 0; /* and this */
result->oprs[operand].opflags = 0;
i = stdscan(NULL, &tokval);
if (i == TOKEN_EOS)
break; /* end of operands: get out of here */
else if (first && i == ':') {
insn_is_label = true;
goto restart_parse;
}
first = false;
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;
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:
nasm_error(ERR_NONFATAL, "invalid operand size specification");
}
i = stdscan(NULL, &tokval);
}
if (i == '[' || i == '&') { /* memory reference */
mref = true;
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);
}
} else { /* immediate operand, or register */
mref = false;
bracket = false; /* placate optimisers */
}
if ((result->oprs[operand].type & FAR) && !mref &&
result->opcode != I_JMP && result->opcode != I_CALL) {
nasm_error(ERR_NONFATAL, "invalid use of FAR operand specifier");
}
value = evaluate(stdscan, NULL, &tokval,
&result->oprs[operand].opflags,
critical, nasm_error, &hints);
i = tokval.t_type;
if (result->oprs[operand].opflags & OPFLAG_FORWARD) {
result->forw_ref = true;
}
if (!value) { /* nasm_error in evaluator */
result->opcode = I_none; /* 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 ||
!IS_SREG(value->type))
nasm_error(ERR_NONFATAL, "invalid segment override");
else if (result->prefixes[PPS_SEG])
nasm_error(ERR_NONFATAL,
"instruction has conflicting segment overrides");
else {
result->prefixes[PPS_SEG] = value->type;
if (IS_FSGS(value->type))
result->oprs[operand].eaflags |= EAF_FSGS;
}
i = stdscan(NULL, &tokval); /* then skip the colon */
while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
process_size_override(result, operand);
i = stdscan(NULL, &tokval);
}
value = evaluate(stdscan, NULL, &tokval,
&result->oprs[operand].opflags,
critical, nasm_error, &hints);
i = tokval.t_type;
if (result->oprs[operand].opflags & OPFLAG_FORWARD) {
result->forw_ref = true;
}
/* and get the offset */
if (!value) { /* but, error in evaluator */
result->opcode = I_none; /* unrecoverable parse error: */
return result; /* ignore this instruction */
}
}
recover = false;
if (mref && bracket) { /* find ] at the end */
if (i != ']') {
nasm_error(ERR_NONFATAL, "parser: expecting ]");
recover = true;
} else { /* we got the required ] */
i = stdscan(NULL, &tokval);
if (i != 0 && i != ',') {
nasm_error(ERR_NONFATAL, "comma or end of line expected");
recover = true;
}
}
} else { /* immediate operand */
if (i != 0 && i != ',' && i != ':') {
nasm_error(ERR_NONFATAL, "comma, colon or end of line expected");
recover = true;
} else if (i == ':') {
result->oprs[operand].type |= COLON;
}
}
if (recover) {
do { /* error recovery */
i = stdscan(NULL, &tokval);
} while (i != 0 && i != ',');
}
/*
* 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 */
int64_t o; /* offset */
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 */
nasm_error(ERR_NONFATAL,
"beroset-p-592-invalid effective address");
result->opcode = I_none;
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? */
nasm_error(ERR_NONFATAL,
"beroset-p-603-invalid effective address");
result->opcode = I_none;
return result;
} else {
if (e->type == EXPR_UNKNOWN) {
result->oprs[operand].opflags |= OPFLAG_UNKNOWN;
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) {
nasm_error(ERR_NONFATAL,
"beroset-p-630-invalid effective address");
result->opcode = I_none;
return result;
}
while (e->type && e->value == 0)
e++;
if (e->type && e->value != 1) {
nasm_error(ERR_NONFATAL,
"beroset-p-637-invalid effective address");
result->opcode = I_none;
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) {
nasm_error(ERR_NONFATAL,
"beroset-p-650-invalid effective address");
result->opcode = I_none;
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! */
nasm_error(ERR_NONFATAL,
"beroset-p-663-invalid effective address");
result->opcode = I_none;
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;
}
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;
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 | UDWORD64 | SDWORD64;
}
} 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;
if ((uint64_t)v64 <= UINT64_C(0xffffffff))
result->oprs[operand].type |= UDWORD64;
if (v64 >= -INT64_C(0x80000000) &&
v64 <= INT64_C(0x7fffffff))
result->oprs[operand].type |= SDWORD64;
}
}
} else { /* it's a register */
unsigned int rs;
if (value->type >= EXPR_SIMPLE || value->value != 1) {
nasm_error(ERR_NONFATAL, "invalid operand type");
result->opcode = I_none;
return result;
}
/*
* check that its only 1 register, not an expression...
*/
for (i = 1; value[i].type; i++)
if (value[i].value) {
nasm_error(ERR_NONFATAL, "invalid operand type");
result->opcode = I_none;
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;
} else
rs = 0;
result->oprs[operand].type &= TO;
result->oprs[operand].type |= REGISTER;
result->oprs[operand].type |= nasm_reg_flags[value->type];
result->oprs[operand].basereg = value->type;
if (rs && (result->oprs[operand].type & SIZE_MASK) != rs)
nasm_error(ERR_WARNING | ERR_PASS1,
"register size specification ignored");
}
}
}
result->operands = operand; /* set operand count */
/* clear remaining operands */
while (operand < MAX_OPERANDS)
result->oprs[operand++].type = 0;
/*
* Transform RESW, RESD, RESQ, REST, RESO, RESY into RESB.
*/
switch (result->opcode) {
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;
}
return result;
}
insn *parse_line(int pass, char *buffer, insn *result, ldfunc ldef)
{
bool insn_is_label = false;
struct eval_hints hints;
int opnum;
int critical;
bool first;
bool recover;
restart_parse:
first = true;
result->forw_ref = false;
stdscan_reset();
stdscan_set(buffer);
i = stdscan(NULL, &tokval);
result->label = NULL; /* Assume no label */
result->eops = NULL; /* must do this, whatever happens */
result->operands = 0; /* must initialize this */
result->evex_rm = 0; /* Ensure EVEX rounding mode is reset */
result->evex_brerop = -1; /* Reset EVEX broadcasting/ER op position */
/* Ignore blank lines */
if (i == TOKEN_EOS)
goto fail;
if (i != TOKEN_ID &&
i != TOKEN_INSN &&
i != TOKEN_PREFIX &&
(i != TOKEN_REG || !IS_SREG(tokval.t_integer))) {
nasm_error(ERR_NONFATAL,
"label or instruction expected at start of line");
goto fail;
}
if (i == TOKEN_ID || (insn_is_label && i == TOKEN_INSN)) {
/* there's a label here */
first = false;
result->label = tokval.t_charptr;
i = stdscan(NULL, &tokval);
if (i == ':') { /* skip over the optional colon */
i = stdscan(NULL, &tokval);
} else if (i == 0) {
nasm_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);
}
}
/* Just a label here */
if (i == TOKEN_EOS)
goto fail;
nasm_build_assert(P_none != 0);
memset(result->prefixes, P_none, sizeof(result->prefixes));
result->times = 1L;
while (i == TOKEN_PREFIX ||
(i == TOKEN_REG && IS_SREG(tokval.t_integer))) {
first = false;
/*
* 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, nasm_error, NULL);
i = tokval.t_type;
if (!value) /* Error in evaluator */
goto fail;
if (!is_simple(value)) {
nasm_error(ERR_NONFATAL,
"non-constant argument supplied to TIMES");
result->times = 1L;
} else {
result->times = value->value;
if (value->value < 0 && pass0 == 2) {
nasm_error(ERR_NONFATAL, "TIMES value %"PRId64" 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)
nasm_error(ERR_WARNING | ERR_PASS1,
"instruction has redundant prefixes");
else
nasm_error(ERR_NONFATAL,
"instruction has conflicting prefixes");
}
result->prefixes[slot] = tokval.t_integer;
i = stdscan(NULL, &tokval);
}
}
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) {
/*
* 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 {
nasm_error(ERR_NONFATAL, "parser: instruction expected");
goto fail;
}
}
result->opcode = tokval.t_integer;
result->condition = tokval.t_inttwo;
/*
* INCBIN cannot be satisfied with incorrectly
* 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.
*/
if (result->opcode == I_INCBIN) {
critical = (pass0 < 2 ? 1 : 2);
} else
critical = (pass == 2 ? 2 : 0);
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_DZ ||
result->opcode == I_INCBIN) {
extop *eop, **tail = &result->eops, **fixptr;
int oper_num = 0;
int32_t sign;
result->eops_float = false;
/*
* Begin to read the DB/DW/DD/DQ/DT/DO/DY/DZ/INCBIN operands.
*/
while (1) {
i = stdscan(NULL, &tokval);
if (i == TOKEN_EOS)
break;
else if (first && i == ':') {
insn_is_label = true;
goto restart_parse;
}
first = false;
fixptr = tail;
eop = *tail = nasm_malloc(sizeof(extop));
tail = &eop->next;
eop->next = NULL;
eop->type = EOT_NOTHING;
oper_num++;
sign = +1;
/*
* is_comma_next() here is to distinguish this from
* a string used as part of an expression...
*/
if (i == TOKEN_STR && is_comma_next()) {
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) {
nasm_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) {
nasm_error(ERR_NONFATAL, "invalid string for transform");
eop->type = EOT_NOTHING;
}
}
if (parens && i && i != ')') {
i = stdscan(NULL, &tokval);
if (i != ')') {
nasm_error(ERR_NONFATAL, "unterminated %s function",
funcname);
}
}
if (i && i != ',')
i = stdscan(NULL, &tokval);
} else if (i == '-' || i == '+') {
char *save = stdscan_get();
int token = i;
sign = (i == '-') ? -1 : 1;
i = stdscan(NULL, &tokval);
if (i != TOKEN_FLOAT) {
stdscan_set(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;
eop->stringlen = idata_bytes(result->opcode);
if (eop->stringlen > 16) {
nasm_error(ERR_NONFATAL, "floating-point constant"
" encountered in DY or DZ instruction");
eop->stringlen = 0;
} else if (eop->stringlen < 1) {
nasm_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;
}
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, nasm_error))
eop->type = EOT_NOTHING;
i = stdscan(NULL, &tokval); /* eat the comma */
} else {
/* anything else, assume it is an expression */
expr *value;
is_expression:
value = evaluate(stdscan, NULL, &tokval, NULL,
critical, nasm_error, NULL);
i = tokval.t_type;
if (!value) /* Error in evaluator */
goto fail;
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 {
nasm_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 == TOKEN_EOS) /* also could be EOL */
break;
if (i != ',') {
nasm_error(ERR_NONFATAL, "comma expected after operand %d",
oper_num);
goto fail;
}
}
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)
nasm_error(ERR_NONFATAL, "`incbin' expects a file name");
else if (result->eops->next &&
result->eops->next->type != EOT_DB_NUMBER)
nasm_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)
nasm_error(ERR_NONFATAL, "`incbin': third parameter is"
" non-numeric");
else if (result->eops->next && result->eops->next->next &&
result->eops->next->next->next)
nasm_error(ERR_NONFATAL,
"`incbin': more than three parameters");
else
return result;
/*
* If we reach here, one of the above errors happened.
* Throw the instruction away.
*/
goto fail;
} else /* DB ... */ if (oper_num == 0)
nasm_error(ERR_WARNING | ERR_PASS1,
"no operand for data declaration");
else
result->operands = oper_num;
return result;
}
/*
* Now we begin to parse the operands. There may be up to four
* of these, separated by commas, and terminated by a zero token.
*/
for (opnum = 0; opnum < MAX_OPERANDS; opnum++) {
operand *op = &result->oprs[opnum];
expr *value; /* used most of the time */
bool mref; /* is this going to be a memory ref? */
bool bracket; /* is it a [] mref, or a & mref? */
bool mib; /* compound (mib) mref? */
int setsize = 0;
decoflags_t brace_flags = 0; /* flags for decorators in braces */
op->disp_size = 0; /* have to zero this whatever */
op->eaflags = 0; /* and this */
op->opflags = 0;
op->decoflags = 0;
i = stdscan(NULL, &tokval);
if (i == TOKEN_EOS)
break; /* end of operands: get out of here */
else if (first && i == ':') {
insn_is_label = true;
goto restart_parse;
}
first = false;
op->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 */
op->type |= BITS8;
setsize = 1;
break;
case S_WORD:
if (!setsize)
op->type |= BITS16;
setsize = 1;
break;
case S_DWORD:
case S_LONG:
if (!setsize)
op->type |= BITS32;
setsize = 1;
break;
case S_QWORD:
if (!setsize)
op->type |= BITS64;
setsize = 1;
break;
case S_TWORD:
if (!setsize)
op->type |= BITS80;
setsize = 1;
break;
case S_OWORD:
if (!setsize)
op->type |= BITS128;
setsize = 1;
break;
case S_YWORD:
if (!setsize)
op->type |= BITS256;
setsize = 1;
break;
case S_ZWORD:
if (!setsize)
op->type |= BITS512;
setsize = 1;
break;
case S_TO:
op->type |= TO;
break;
case S_STRICT:
op->type |= STRICT;
break;
case S_FAR:
op->type |= FAR;
break;
case S_NEAR:
op->type |= NEAR;
break;
case S_SHORT:
op->type |= SHORT;
break;
default:
nasm_error(ERR_NONFATAL, "invalid operand size specification");
}
i = stdscan(NULL, &tokval);
}
if (i == '[' || i == '&') { /* memory reference */
mref = true;
bracket = (i == '[');
i = stdscan(NULL, &tokval); /* then skip the colon */
while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
process_size_override(result, op);
i = stdscan(NULL, &tokval);
}
/* when a comma follows an opening bracket - [ , eax*4] */
if (i == ',') {
/* treat as if there is a zero displacement virtually */
tokval.t_type = TOKEN_NUM;
tokval.t_integer = 0;
stdscan_set(stdscan_get() - 1); /* rewind the comma */
}
} else { /* immediate operand, or register */
mref = false;
bracket = false; /* placate optimisers */
}
if ((op->type & FAR) && !mref &&
result->opcode != I_JMP && result->opcode != I_CALL) {
nasm_error(ERR_NONFATAL, "invalid use of FAR operand specifier");
}
value = evaluate(stdscan, NULL, &tokval,
&op->opflags,
critical, nasm_error, &hints);
i = tokval.t_type;
if (op->opflags & OPFLAG_FORWARD) {
result->forw_ref = true;
}
if (!value) /* Error in evaluator */
goto fail;
if (i == ':' && mref) { /* it was seg:offset */
/*
* Process the segment override.
*/
if (value[1].type != 0 ||
value->value != 1 ||
!IS_SREG(value->type))
nasm_error(ERR_NONFATAL, "invalid segment override");
else if (result->prefixes[PPS_SEG])
nasm_error(ERR_NONFATAL,
"instruction has conflicting segment overrides");
else {
result->prefixes[PPS_SEG] = value->type;
if (IS_FSGS(value->type))
op->eaflags |= EAF_FSGS;
}
i = stdscan(NULL, &tokval); /* then skip the colon */
while (i == TOKEN_SPECIAL || i == TOKEN_PREFIX) {
process_size_override(result, op);
i = stdscan(NULL, &tokval);
}
value = evaluate(stdscan, NULL, &tokval,
&op->opflags,
critical, nasm_error, &hints);
i = tokval.t_type;
if (op->opflags & OPFLAG_FORWARD) {
result->forw_ref = true;
}
/* and get the offset */
if (!value) /* Error in evaluator */
goto fail;
}
mib = false;
if (mref && bracket && i == ',') {
/* [seg:base+offset,index*scale] syntax (mib) */
operand o1, o2; /* Partial operands */
if (parse_mref(&o1, value))
goto fail;
i = stdscan(NULL, &tokval); /* Eat comma */
value = evaluate(stdscan, NULL, &tokval, &op->opflags,
critical, nasm_error, &hints);
i = tokval.t_type;
if (parse_mref(&o2, value))
goto fail;
if (o2.basereg != -1 && o2.indexreg == -1) {
o2.indexreg = o2.basereg;
o2.scale = 1;
o2.basereg = -1;
}
if (o1.indexreg != -1 || o2.basereg != -1 || o2.offset != 0 ||
o2.segment != NO_SEG || o2.wrt != NO_SEG) {
nasm_error(ERR_NONFATAL, "invalid mib expression");
goto fail;
}
op->basereg = o1.basereg;
op->indexreg = o2.indexreg;
op->scale = o2.scale;
op->offset = o1.offset;
op->segment = o1.segment;
op->wrt = o1.wrt;
if (op->basereg != -1) {
op->hintbase = op->basereg;
op->hinttype = EAH_MAKEBASE;
} else if (op->indexreg != -1) {
op->hintbase = op->indexreg;
op->hinttype = EAH_NOTBASE;
} else {
op->hintbase = -1;
op->hinttype = EAH_NOHINT;
}
mib = true;
}
recover = false;
if (mref && bracket) { /* find ] at the end */
if (i != ']') {
nasm_error(ERR_NONFATAL, "parser: expecting ]");
recover = true;
} else { /* we got the required ] */
i = stdscan(NULL, &tokval);
if ((i == TOKEN_DECORATOR) || (i == TOKEN_OPMASK)) {
/*
* according to AVX512 spec, broacast or opmask decorator
* is expected for memory reference operands
*/
if (tokval.t_flag & TFLAG_BRDCAST) {
brace_flags |= GEN_BRDCAST(0) |
VAL_BRNUM(tokval.t_integer - BRC_1TO8);
i = stdscan(NULL, &tokval);
} else if (i == TOKEN_OPMASK) {
brace_flags |= VAL_OPMASK(nasm_regvals[tokval.t_integer]);
i = stdscan(NULL, &tokval);
} else {
nasm_error(ERR_NONFATAL, "broadcast or opmask "
"decorator expected inside braces");
recover = true;
}
}
if (i != 0 && i != ',') {
nasm_error(ERR_NONFATAL, "comma or end of line expected");
recover = true;
}
}
} else { /* immediate operand */
if (i != 0 && i != ',' && i != ':' &&
i != TOKEN_DECORATOR && i != TOKEN_OPMASK) {
nasm_error(ERR_NONFATAL, "comma, colon, decorator or end of "
"line expected after operand");
recover = true;
} else if (i == ':') {
op->type |= COLON;
} else if (i == TOKEN_DECORATOR || i == TOKEN_OPMASK) {
/* parse opmask (and zeroing) after an operand */
recover = parse_braces(&brace_flags);
}
}
if (recover) {
do { /* error recovery */
i = stdscan(NULL, &tokval);
} while (i != 0 && i != ',');
}
/*
* now convert the exprs returned from evaluate()
* into operand descriptions...
*/
op->decoflags |= brace_flags;
if (mref) { /* it's a memory reference */
/* A mib reference was fully parsed already */
if (!mib) {
if (parse_mref(op, value))
goto fail;
op->hintbase = hints.base;
op->hinttype = hints.type;
}
mref_set_optype(op);
} else { /* it's not a memory reference */
if (is_just_unknown(value)) { /* it's immediate but unknown */
op->type |= IMMEDIATE;
op->opflags |= OPFLAG_UNKNOWN;
op->offset = 0; /* don't care */
op->segment = NO_SEG; /* don't care again */
op->wrt = NO_SEG; /* still don't care */
if(optimizing >= 0 && !(op->type & STRICT)) {
/* Be optimistic */
op->type |=
UNITY | SBYTEWORD | SBYTEDWORD | UDWORD | SDWORD;
}
} else if (is_reloc(value)) { /* it's immediate */
op->type |= IMMEDIATE;
op->offset = reloc_value(value);
op->segment = reloc_seg(value);
op->wrt = reloc_wrt(value);
if (is_simple(value)) {
uint64_t n = reloc_value(value);
if (n == 1)
op->type |= UNITY;
if (optimizing >= 0 &&
!(op->type & STRICT)) {
if ((uint32_t) (n + 128) <= 255)
op->type |= SBYTEDWORD;
if ((uint16_t) (n + 128) <= 255)
op->type |= SBYTEWORD;
if (n <= 0xFFFFFFFF)
op->type |= UDWORD;
if (n + 0x80000000 <= 0xFFFFFFFF)
op->type |= SDWORD;
}
}
} else if(value->type == EXPR_RDSAE) {
/*
* it's not an operand but a rounding or SAE decorator.
* put the decorator information in the (opflag_t) type field
* of previous operand.
*/
opnum--; op--;
switch (value->value) {
case BRC_RN:
case BRC_RU:
case BRC_RD:
case BRC_RZ:
case BRC_SAE:
op->decoflags |= (value->value == BRC_SAE ? SAE : ER);
result->evex_rm = value->value;
break;
default:
nasm_error(ERR_NONFATAL, "invalid decorator");
break;
}
} else { /* it's a register */
opflags_t rs;
if (value->type >= EXPR_SIMPLE || value->value != 1) {
nasm_error(ERR_NONFATAL, "invalid operand type");
goto fail;
}
/*
* check that its only 1 register, not an expression...
*/
for (i = 1; value[i].type; i++)
if (value[i].value) {
nasm_error(ERR_NONFATAL, "invalid operand type");
goto fail;
}
/* clear overrides, except TO which applies to FPU regs */
if (op->type & ~TO) {
/*
* we want to produce a warning iff the specified size
* is different from the register size
*/
rs = op->type & SIZE_MASK;
} else
rs = 0;
op->type &= TO;
op->type |= REGISTER;
op->type |= nasm_reg_flags[value->type];
op->decoflags |= brace_flags;
op->basereg = value->type;
if (rs && (op->type & SIZE_MASK) != rs)
nasm_error(ERR_WARNING | ERR_PASS1,
"register size specification ignored");
}
}
/* remember the position of operand having broadcasting/ER mode */
if (op->decoflags & (BRDCAST_MASK | ER | SAE))
result->evex_brerop = opnum;
}
result->operands = opnum; /* set operand count */
/* clear remaining operands */
while (opnum < MAX_OPERANDS)
result->oprs[opnum++].type = 0;
/*
* Transform RESW, RESD, RESQ, REST, RESO, RESY, RESZ into RESB.
*/
switch (result->opcode) {
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;
case I_RESZ:
result->opcode = I_RESB;
result->oprs[0].offset *= 64;
break;
default:
break;
}
return result;
fail:
result->opcode = I_none;
return result;
}
int main(int argc, char *argv[]) {
int i = 1, mode, hlen;
size_t n;
FILE *fp;
int tflag = 0, dflag = 0, bflag = 0, zflag = 0;
int tbase = 0, dbase = 0, bbase = 0, zbase = 0;
char *outfile = "a.o65";
int extract = 0;
if (argc <= 1) {
usage(stderr);
exit(1);
}
if (strstr(argv[1], "--help")) {
usage(stdout);
exit(0);
}
if (strstr(argv[1], "--version")) {
version(programname, progversion, author, copyright);
exit(0);
}
while(i<argc) {
if(argv[i][0]=='-') {
/* process options */
switch(argv[i][1]) {
case 'o':
if(argv[i][2]) outfile=argv[i]+2;
else outfile=argv[++i];
break;
case 'X':
extract=3;
break;
case 'b':
switch(argv[i][2]) {
case 't':
tflag= 1;
if(argv[i][3]) tbase = atoi(argv[i]+3);
else tbase = atoi(argv[++i]);
break;
case 'd':
dflag= 1;
if(argv[i][3]) dbase = atoi(argv[i]+3);
else dbase = atoi(argv[++i]);
break;
case 'b':
bflag= 1;
if(argv[i][3]) bbase = atoi(argv[i]+3);
else bbase = atoi(argv[++i]);
break;
case 'z':
zflag= 1;
if(argv[i][3]) zbase = atoi(argv[i]+3);
else zbase = atoi(argv[++i]);
break;
default:
printf("Unknown segment type '%c' - ignored!\n", argv[i][2]);
break;
}
break;
case 'x': /* extract segment */
switch(argv[i][2]) {
case 't':
extract = 1;
break;
case 'd':
extract = 2;
break;
case 'z':
case 'b':
printf("Cannot extract segment type '%c' - ignored!\n", argv[i][2]);
break;
default:
printf("Unknown segment type '%c' - ignored!\n", argv[i][2]);
break;
}
break;
default:
fprintf(stderr,"reloc65: %s unknown option, use '-?' for help\n",argv[i]);
break;
}
} else {
struct stat fs;
file.fname=argv[i];
stat(argv[i], &fs);
file.fsize=fs.st_size;
file.buf=malloc(file.fsize);
if(!file.buf) {
fprintf(stderr,"Oops, no more memory!\n");
exit(1);
}
printf("reloc65: read file %s -> %s\n",argv[i],outfile);
fp = fopen(argv[i],"rb");
if(fp) {
n = fread(file.buf, 1, file.fsize, fp);
fclose(fp);
if((n>=file.fsize) && (!memcmp(file.buf, cmp, 5))) {
mode=file.buf[7]*256+file.buf[6];
if(mode & 0x2000) {
fprintf(stderr,"reloc65: %s: 32 bit size not supported\n", argv[i]);
} else
if(mode & 0x4000) {
fprintf(stderr,"reloc65: %s: pagewise relocation not supported\n", argv[i]);
} else {
hlen = BUF+read_options(file.buf+BUF);
file.tbase = file.buf[ 9]*256+file.buf[ 8];
file.tlen = file.buf[11]*256+file.buf[10];
file.tdiff = tflag? tbase - file.tbase : 0;
file.dbase = file.buf[13]*256+file.buf[12];
file.dlen = file.buf[15]*256+file.buf[14];
if (extract == 3) {
if (dflag) {
fprintf(stderr,"reloc65: %s: Warning: data segment address ignored for -X option\n", argv[i]);
}
dbase = file.tbase + file.tdiff + file.tlen;
file.ddiff = dbase - file.dbase;
} else {
file.ddiff = dflag? dbase - file.dbase : 0;
}
file.bbase = file.buf[17]*256+file.buf[16];
file.blen = file.buf[19]*256+file.buf[18];
file.bdiff = bflag? bbase - file.bbase : 0;
file.zbase = file.buf[21]*256+file.buf[20];
file.zlen = file.buf[23]*256+file.buf[21];
file.zdiff = zflag? zbase - file.zbase : 0;
file.segt = file.buf + hlen;
file.segd = file.segt + file.tlen;
file.utab = file.segd + file.dlen;
file.rttab = file.utab + read_undef(file.utab);
file.rdtab = reloc_seg(file.segt, file.tlen, file.rttab,
&file, extract);
file.extab = reloc_seg(file.segd, file.dlen, file.rdtab,
&file, extract);
reloc_globals(file.extab, &file);
if(tflag) {
file.buf[ 9]= (tbase>>8)&255;
file.buf[ 8]= tbase & 255;
}
if(dflag) {
file.buf[13]= (dbase>>8)&255;
file.buf[12]= dbase & 255;
}
if(bflag) {
file.buf[17]= (bbase>>8)&255;
file.buf[16]= bbase & 255;
}
if(zflag) {
file.buf[21]= (zbase>>8)&255;
file.buf[20]= zbase & 255;
}
fp = fopen(outfile, "wb");
if(fp) {
switch(extract) {
case 0: /* whole file */
fwrite(file.buf, 1, file.fsize, fp);
break;
case 1: /* text segment */
fwrite(file.segt, 1, file.tlen, fp);
break;
case 2: /* data segment */
fwrite(file.segd, 1, file.dlen, fp);
break;
case 3: /* text+data */
fwrite(file.segt, 1, file.tlen, fp);
fwrite(file.segd, 1, file.dlen, fp);
break;
}
fclose(fp);
} else {
fprintf(stderr,"reloc65: write '%s': %s\n",
outfile, strerror(errno));
}
}
} else {
fprintf(stderr,"reloc65: %s: not an o65 file!\n", argv[i]);
if(file.buf[0]==1 && file.buf[1]==8 && file.buf[3]==8) {
printf("%s: C64 BASIC executable (start address $0801)?\n", argv[i]);
} else
if(file.buf[0]==1 && file.buf[1]==4 && file.buf[3]==4) {
printf("%s: CBM PET BASIC executable (start address $0401)?\n", argv[i]);
}
}
} else {
expr *evaluate(scanner sc, void *scprivate, struct tokenval *tv,
int *fwref, int critical, struct eval_hints *hints)
{
expr *e;
expr *f = NULL;
hint = hints;
if (hint)
hint->type = EAH_NOHINT;
if (critical & CRITICAL) {
critical &= ~CRITICAL;
bexpr = rexp0;
} else
bexpr = expr0;
scan = sc;
scpriv = scprivate;
tokval = tv;
opflags = fwref;
if (tokval->t_type == TOKEN_INVALID)
i = scan(scpriv, tokval);
else
i = tokval->t_type;
while (ntempexprs) /* initialize temporary storage */
nasm_free(tempexprs[--ntempexprs]);
e = bexpr(critical);
if (!e)
return NULL;
if (i == TOKEN_WRT) {
i = scan(scpriv, tokval); /* eat the WRT */
f = expr6(critical);
if (!f)
return NULL;
}
e = scalar_mult(e, 1L, false); /* strip far-absolute segment part */
if (f) {
expr *g;
if (is_just_unknown(f))
g = unknown_expr();
else {
int64_t value;
begintemp();
if (!is_reloc(f)) {
nasm_error(ERR_NONFATAL, "invalid right-hand operand to WRT");
return NULL;
}
value = reloc_seg(f);
if (value == NO_SEG)
value = reloc_value(f) | SEG_ABS;
else if (!(value & SEG_ABS) && !(value % 2) && critical) {
nasm_error(ERR_NONFATAL, "invalid right-hand operand to WRT");
return NULL;
}
addtotemp(EXPR_WRT, value);
g = finishtemp();
}
e = add_vectors(e, g);
}
return e;
}
