예제 #1
0
파일: eval.c 프로젝트: AxFab/nasm
static expr *expr3(int critical)
{
    expr *e, *f;

    e = expr4(critical);
    if (!e)
        return NULL;

    while (i == TOKEN_SHL || i == TOKEN_SHR) {
        int j = i;
        i = scan(scpriv, tokval);
        f = expr4(critical);
        if (!f)
            return NULL;
        if (!(is_simple(e) || is_just_unknown(e)) ||
            !(is_simple(f) || is_just_unknown(f))) {
            nasm_error(ERR_NONFATAL, "shift operator may only be applied to"
                  " scalar values");
        } else if (is_just_unknown(e) || is_just_unknown(f)) {
            e = unknown_expr();
        } else
            switch (j) {
            case TOKEN_SHL:
                e = scalarvect(reloc_value(e) << reloc_value(f));
                break;
            case TOKEN_SHR:
                e = scalarvect(((uint64_t)reloc_value(e)) >>
                               reloc_value(f));
                break;
            }
    }
    return e;
}
예제 #2
0
파일: eval.c 프로젝트: AxFab/nasm
static expr *expr2(int critical)
{
    expr *e, *f;

    e = expr3(critical);
    if (!e)
        return NULL;

    while (i == '&') {
        i = scan(scpriv, tokval);
        f = expr3(critical);
        if (!f)
            return NULL;
        if (!(is_simple(e) || is_just_unknown(e)) ||
            !(is_simple(f) || is_just_unknown(f))) {
            nasm_error(ERR_NONFATAL, "`&' operator may only be applied to"
                  " scalar values");
        }
        if (is_just_unknown(e) || is_just_unknown(f))
            e = unknown_expr();
        else
            e = scalarvect(reloc_value(e) & reloc_value(f));
    }
    return e;
}
예제 #3
0
파일: eval.c 프로젝트: AxFab/nasm
static expr *rexp1(int critical)
{
    expr *e, *f;

    e = rexp2(critical);
    if (!e)
        return NULL;

    while (i == TOKEN_DBL_XOR) {
        i = scan(scpriv, tokval);
        f = rexp2(critical);
        if (!f)
            return NULL;
        if (!(is_simple(e) || is_just_unknown(e)) ||
            !(is_simple(f) || is_just_unknown(f))) {
            nasm_error(ERR_NONFATAL, "`^' operator may only be applied to"
                  " scalar values");
        }

        if (is_just_unknown(e) || is_just_unknown(f))
            e = unknown_expr();
        else
            e = scalarvect((int64_t)(!reloc_value(e) ^ !reloc_value(f)));
    }
    return e;
}
예제 #4
0
파일: eval.c 프로젝트: AxFab/nasm
static expr *rexp3(int critical)
{
    expr *e, *f;
    int64_t v;

    e = expr0(critical);
    if (!e)
        return NULL;

    while (i == TOKEN_EQ || i == TOKEN_LT || i == TOKEN_GT ||
           i == TOKEN_NE || i == TOKEN_LE || i == TOKEN_GE) {
        int j = i;
        i = scan(scpriv, tokval);
        f = expr0(critical);
        if (!f)
            return NULL;

        e = add_vectors(e, scalar_mult(f, -1L, false));

        switch (j) {
        case TOKEN_EQ:
        case TOKEN_NE:
            if (is_unknown(e))
                v = -1;         /* means unknown */
            else if (!is_really_simple(e) || reloc_value(e) != 0)
                v = (j == TOKEN_NE);    /* unequal, so return true if NE */
            else
                v = (j == TOKEN_EQ);    /* equal, so return true if EQ */
            break;
        default:
            if (is_unknown(e))
                v = -1;         /* means unknown */
            else if (!is_really_simple(e)) {
                nasm_error(ERR_NONFATAL,
                      "`%s': operands differ by a non-scalar",
                      (j == TOKEN_LE ? "<=" : j == TOKEN_LT ? "<" : j ==
                       TOKEN_GE ? ">=" : ">"));
                v = 0;          /* must set it to _something_ */
            } else {
                int64_t vv = reloc_value(e);
                if (vv == 0)
                    v = (j == TOKEN_LE || j == TOKEN_GE);
                else if (vv > 0)
                    v = (j == TOKEN_GE || j == TOKEN_GT);
                else            /* vv < 0 */
                    v = (j == TOKEN_LE || j == TOKEN_LT);
            }
            break;
        }

        if (v == -1)
            e = unknown_expr();
        else
            e = scalarvect(v);
    }
    return e;
}
예제 #5
0
파일: parser.c 프로젝트: fido2478/epspectra
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;
}
예제 #6
0
파일: outelf.c 프로젝트: sunank200/ATOM-OS
static void elf_deflabel(char *name, long segment, long offset,
                         int is_global, char *special)
{
    int pos = strslen;
    struct Symbol *sym;
    int special_used = FALSE;

#if defined(DEBUG) && DEBUG>2
    fprintf(stderr,
            " elf_deflabel: %s, seg=%ld, off=%ld, is_global=%d, %s\n",
            name, segment, offset, is_global, special);
#endif
    if (name[0] == '.' && name[1] == '.' && name[2] != '@') {
        /*
         * This is a NASM special symbol. We never allow it into
         * the ELF symbol table, even if it's a valid one. If it
         * _isn't_ a valid one, we should barf immediately.
         */
        if (strcmp(name, "..gotpc") && strcmp(name, "..gotoff") &&
            strcmp(name, "..got") && strcmp(name, "..plt") &&
            strcmp(name, "..sym"))
            error(ERR_NONFATAL, "unrecognised special symbol `%s'", name);
        return;
    }

    if (is_global == 3) {
        struct Symbol **s;
        /*
         * Fix up a forward-reference symbol size from the first
         * pass.
         */
        for (s = &fwds; *s; s = &(*s)->nextfwd)
            if (!strcmp((*s)->name, name)) {
                struct tokenval tokval;
                expr *e;
                char *p = special;

                while (*p && !isspace(*p))
                    p++;
                while (*p && isspace(*p))
                    p++;
                stdscan_reset();
                stdscan_bufptr = p;
                tokval.t_type = TOKEN_INVALID;
                e = evaluate(stdscan, NULL, &tokval, NULL, 1, error, NULL);
                if (e) {
                    if (!is_simple(e))
                        error(ERR_NONFATAL, "cannot use relocatable"
                              " expression as symbol size");
                    else
                        (*s)->size = reloc_value(e);
                }

                /*
                 * Remove it from the list of unresolved sizes.
                 */
                nasm_free((*s)->name);
                *s = (*s)->nextfwd;
                return;
            }
        return;                 /* it wasn't an important one */
    }

    saa_wbytes(strs, name, (long)(1 + strlen(name)));
    strslen += 1 + strlen(name);

    sym = saa_wstruct(syms);

    sym->strpos = pos;
    sym->type = is_global ? SYM_GLOBAL : 0;
    sym->size = 0;
    if (segment == NO_SEG)
        sym->section = SHN_ABS;
    else {
        int i;
        sym->section = SHN_UNDEF;
        if (nsects == 0 && segment == def_seg) {
            int tempint;
            if (segment != elf_section_names(".text", 2, &tempint))
                error(ERR_PANIC,
                      "strange segment conditions in ELF driver");
            sym->section = nsects;
        } else {
            for (i = 0; i < nsects; i++)
                if (segment == sects[i]->index) {
                    sym->section = i + 1;
                    break;
                }
        }
    }

    if (is_global == 2) {
        sym->size = offset;
        sym->value = 0;
        sym->section = SHN_COMMON;
        /*
         * We have a common variable. Check the special text to see
         * if it's a valid number and power of two; if so, store it
         * as the alignment for the common variable.
         */
        if (special) {
            int err;
            sym->value = readnum(special, &err);
            if (err)
                error(ERR_NONFATAL, "alignment constraint `%s' is not a"
                      " valid number", special);
            else if ((sym->value | (sym->value - 1)) != 2 * sym->value - 1)
                error(ERR_NONFATAL, "alignment constraint `%s' is not a"
                      " power of two", special);
        }
        special_used = TRUE;
    } else
        sym->value = (sym->section == SHN_UNDEF ? 0 : offset);

    if (sym->type == SYM_GLOBAL) {
        /*
         * There's a problem here that needs fixing.
         * If sym->section == SHN_ABS, then the first line of the
         * else section causes a core dump, because its a reference
         * beyond the end of the section array.
         * This behaviour is exhibited by this code:
         *     GLOBAL crash_nasm
         *     crash_nasm equ 0
         *
         * I'm not sure how to procede, because I haven't got the
         * first clue about how ELF works, so I don't know what to
         * do with it. Furthermore, I'm not sure what the rest of this
         * section of code does. Help?
         *
         * For now, I'll see if doing absolutely nothing with it will
         * work...
         */
        if (sym->section == SHN_UNDEF || sym->section == SHN_COMMON) {
            bsym = raa_write(bsym, segment, nglobs);
        } else if (sym->section != SHN_ABS) {
            /*
             * This is a global symbol; so we must add it to the linked
             * list of global symbols in its section. We'll push it on
             * the beginning of the list, because it doesn't matter
             * much which end we put it on and it's easier like this.
             *
             * In addition, we check the special text for symbol
             * type and size information.
             */
            sym->next = sects[sym->section - 1]->gsyms;
            sects[sym->section - 1]->gsyms = sym;

            if (special) {
                int n = strcspn(special, " ");

                if (!nasm_strnicmp(special, "function", n))
                    sym->type |= SYM_FUNCTION;
                else if (!nasm_strnicmp(special, "data", n) ||
                         !nasm_strnicmp(special, "object", n))
                    sym->type |= SYM_DATA;
                else
                    error(ERR_NONFATAL, "unrecognised symbol type `%.*s'",
                          n, special);
                if (special[n]) {
                    struct tokenval tokval;
                    expr *e;
                    int fwd = FALSE;
                    char *saveme = stdscan_bufptr;      /* bugfix? fbk 8/10/00 */

                    while (special[n] && isspace(special[n]))
                        n++;
                    /*
                     * We have a size expression; attempt to
                     * evaluate it.
                     */
                    stdscan_reset();
                    stdscan_bufptr = special + n;
                    tokval.t_type = TOKEN_INVALID;
                    e = evaluate(stdscan, NULL, &tokval, &fwd, 0, error,
                                 NULL);
                    if (fwd) {
                        sym->nextfwd = fwds;
                        fwds = sym;
                        sym->name = nasm_strdup(name);
                    } else if (e) {
                        if (!is_simple(e))
                            error(ERR_NONFATAL, "cannot use relocatable"
                                  " expression as symbol size");
                        else
                            sym->size = reloc_value(e);
                    }
                    stdscan_bufptr = saveme;    /* bugfix? fbk 8/10/00 */
                }
                special_used = TRUE;
            }
        }
        sym->globnum = nglobs;
        nglobs++;
    } else
        nlocals++;

    if (special && !special_used)
        error(ERR_NONFATAL, "no special symbol features supported here");
}
예제 #7
0
파일: parser.c 프로젝트: gurry/ProcPlay
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;
}
예제 #8
0
static void aout_deflabel (char *name, long segment, long offset,
			   int is_global, char *special) 
{
    int pos = strslen+4;
    struct Symbol *sym;
    int special_used = FALSE;

    if (name[0] == '.' && name[1] == '.' && name[2] != '@') {
	/*
	 * This is a NASM special symbol. We never allow it into
	 * the a.out symbol table, even if it's a valid one. If it
	 * _isn't_ a valid one, we should barf immediately.
	 */
	if (strcmp(name, "..gotpc") && strcmp(name, "..gotoff") &&
	    strcmp(name, "..got") && strcmp(name, "..plt") &&
	    strcmp(name, "..sym"))
	    error (ERR_NONFATAL, "unrecognised special symbol `%s'", name);
	return;
    }

    if (is_global == 3) {
	struct Symbol **s;
	/*
	 * Fix up a forward-reference symbol size from the first
	 * pass.
	 */
	for (s = &fwds; *s; s = &(*s)->nextfwd)
	    if (!strcmp((*s)->name, name)) {
		struct tokenval tokval;
		expr *e;
		char *p = special;

		while (*p && !isspace(*p)) p++;
		while (*p && isspace(*p)) p++;
		stdscan_reset();
		stdscan_bufptr = p;
		tokval.t_type = TOKEN_INVALID;
		e = evaluate(stdscan, NULL, &tokval, NULL, 1, error, NULL);
		if (e) {
		    if (!is_simple(e))
			error (ERR_NONFATAL, "cannot use relocatable"
			       " expression as symbol size");
		    else
			(*s)->size = reloc_value(e);
		}

		/*
		 * Remove it from the list of unresolved sizes.
		 */
		nasm_free ((*s)->name);
		*s = (*s)->nextfwd;
		return;
	    }
	return;			       /* it wasn't an important one */
    }

    saa_wbytes (strs, name, (long)(1+strlen(name)));
    strslen += 1+strlen(name);

    sym = saa_wstruct (syms);

    sym->strpos = pos;
    sym->type = is_global ? SYM_GLOBAL : 0;
    sym->segment = segment;
    if (segment == NO_SEG)
	sym->type |= SECT_ABS;
    else if (segment == stext.index) {
	sym->type |= SECT_TEXT;
	if (is_global) {
	    sym->next = stext.gsyms;
	    stext.gsyms = sym;
	} else if (!stext.asym)
	    stext.asym = sym;
    } else if (segment == sdata.index) {
	sym->type |= SECT_DATA;
	if (is_global) {
	    sym->next = sdata.gsyms;
	    sdata.gsyms = sym;
	} else if (!sdata.asym)
	    sdata.asym = sym;
    } else if (segment == sbss.index) {
	sym->type |= SECT_BSS;
	if (is_global) {
	    sym->next = sbss.gsyms;
	    sbss.gsyms = sym;
	} else if (!sbss.asym)
	    sbss.asym = sym;
    } else
	sym->type = SYM_GLOBAL;
    if (is_global == 2)
	sym->value = offset;
    else
	sym->value = (sym->type == SYM_GLOBAL ? 0 : offset);

    if (is_global && sym->type != SYM_GLOBAL) {
	/*
	 * Global symbol exported _from_ this module. We must check
	 * the special text for type information.
	 */

	if (special) {
	    int n = strcspn(special, " ");

	    if (!nasm_strnicmp(special, "function", n))
		sym->type |= SYM_FUNCTION;
	    else if (!nasm_strnicmp(special, "data", n) ||
		     !nasm_strnicmp(special, "object", n))
		sym->type |= SYM_DATA;
	    else
		error(ERR_NONFATAL, "unrecognised symbol type `%.*s'",
		      n, special);
	    if (special[n]) {
		struct tokenval tokval;
		expr *e;
		int fwd = FALSE;
		char *saveme=stdscan_bufptr;  /* bugfix? fbk 8/10/00 */

		if (!bsd) {
		    error(ERR_NONFATAL, "Linux a.out does not support"
			  " symbol size information");
		} else {
		    while (special[n] && isspace(special[n]))
			n++;
		    /*
		     * We have a size expression; attempt to
		     * evaluate it.
		     */
		    sym->type |= SYM_WITH_SIZE;
		    stdscan_reset();
		    stdscan_bufptr = special+n;
		    tokval.t_type = TOKEN_INVALID;
		    e = evaluate(stdscan, NULL, &tokval, &fwd, 0, error, NULL);
		    if (fwd) {
			sym->nextfwd = fwds;
			fwds = sym;
			sym->name = nasm_strdup(name);
		    } else if (e) {
			if (!is_simple(e))
			    error (ERR_NONFATAL, "cannot use relocatable"
				   " expression as symbol size");
			else
			    sym->size = reloc_value(e);
		    }
		}
		stdscan_bufptr=saveme;     /* bugfix? fbk 8/10/00 */
	    }
	    special_used = TRUE;
	}
    }

    /*
     * define the references from external-symbol segment numbers
     * to these symbol records.
     */
    if (segment != NO_SEG && segment != stext.index &&
	segment != sdata.index && segment != sbss.index)
	bsym = raa_write (bsym, segment, nsyms);
    sym->symnum = nsyms;

    nsyms++;
    if (sym->type & SYM_WITH_SIZE)
	nsyms++;		       /* and another for the size */

    if (special && !special_used)
	error(ERR_NONFATAL, "no special symbol features supported here");
}
예제 #9
0
파일: parser.c 프로젝트: projedi/nasm
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;
}
예제 #10
0
파일: eval.c 프로젝트: AxFab/nasm
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;
}
예제 #11
0
파일: eval.c 프로젝트: AxFab/nasm
static expr *expr6(int critical)
{
    int32_t type;
    expr *e;
    int32_t label_seg;
    int64_t label_ofs;
    int64_t tmpval;
    bool rn_warn;
    char *scope;

    switch (i) {
    case '-':
        i = scan(scpriv, tokval);
        e = expr6(critical);
        if (!e)
            return NULL;
        return scalar_mult(e, -1L, false);

    case '+':
        i = scan(scpriv, tokval);
        return expr6(critical);

    case '~':
        i = scan(scpriv, tokval);
        e = expr6(critical);
        if (!e)
            return NULL;
        if (is_just_unknown(e))
            return unknown_expr();
        else if (!is_simple(e)) {
            nasm_error(ERR_NONFATAL, "`~' operator may only be applied to"
                  " scalar values");
            return NULL;
        }
        return scalarvect(~reloc_value(e));

    case '!':
        i = scan(scpriv, tokval);
        e = expr6(critical);
        if (!e)
            return NULL;
        if (is_just_unknown(e))
            return unknown_expr();
        else if (!is_simple(e)) {
            nasm_error(ERR_NONFATAL, "`!' operator may only be applied to"
                  " scalar values");
            return NULL;
        }
        return scalarvect(!reloc_value(e));

    case TOKEN_IFUNC:
    {
        enum ifunc func = tokval->t_integer;
        i = scan(scpriv, tokval);
        e = expr6(critical);
        if (!e)
            return NULL;
        if (is_just_unknown(e))
            return unknown_expr();
        else if (!is_simple(e)) {
            nasm_error(ERR_NONFATAL, "function may only be applied to"
                  " scalar values");
            return NULL;
        }
        return scalarvect(eval_ifunc(reloc_value(e), func));
    }

    case TOKEN_SEG:
        i = scan(scpriv, tokval);
        e = expr6(critical);
        if (!e)
            return NULL;
        e = segment_part(e);
        if (!e)
            return NULL;
        if (is_unknown(e) && critical) {
            nasm_error(ERR_NONFATAL, "unable to determine segment base");
            return NULL;
        }
        return e;

    case TOKEN_FLOATIZE:
        return eval_floatize(tokval->t_integer);

    case TOKEN_STRFUNC:
        return eval_strfunc(tokval->t_integer);

    case '(':
        i = scan(scpriv, tokval);
        e = bexpr(critical);
        if (!e)
            return NULL;
        if (i != ')') {
            nasm_error(ERR_NONFATAL, "expecting `)'");
            return NULL;
        }
        i = scan(scpriv, tokval);
        return e;

    case TOKEN_NUM:
    case TOKEN_STR:
    case TOKEN_REG:
    case TOKEN_ID:
    case TOKEN_INSN:            /* Opcodes that occur here are really labels */
    case TOKEN_HERE:
    case TOKEN_BASE:
    case TOKEN_DECORATOR:
        begintemp();
        switch (i) {
        case TOKEN_NUM:
            addtotemp(EXPR_SIMPLE, tokval->t_integer);
            break;
        case TOKEN_STR:
            tmpval = readstrnum(tokval->t_charptr, tokval->t_inttwo, &rn_warn);
            if (rn_warn)
                nasm_error(ERR_WARNING|ERR_PASS1, "character constant too long");
            addtotemp(EXPR_SIMPLE, tmpval);
            break;
        case TOKEN_REG:
            addtotemp(tokval->t_integer, 1L);
            if (hint && hint->type == EAH_NOHINT)
                hint->base = tokval->t_integer, hint->type = EAH_MAKEBASE;
            break;
        case TOKEN_ID:
        case TOKEN_INSN:
        case TOKEN_HERE:
        case TOKEN_BASE:
            /*
             * If !location.known, this indicates that no
             * symbol, Here or Base references are valid because we
             * are in preprocess-only mode.
             */
            if (!location.known) {
                nasm_error(ERR_NONFATAL,
                      "%s not supported in preprocess-only mode",
                      (i == TOKEN_HERE ? "`$'" :
                       i == TOKEN_BASE ? "`$$'" :
                       "symbol references"));
                addtotemp(EXPR_UNKNOWN, 1L);
                break;
            }

            type = EXPR_SIMPLE; /* might get overridden by UNKNOWN */
            if (i == TOKEN_BASE) {
                label_seg = in_abs_seg ? abs_seg : location.segment;
                label_ofs = 0;
            } else if (i == TOKEN_HERE) {
                label_seg = in_abs_seg ? abs_seg : location.segment;
                label_ofs = in_abs_seg ? abs_offset : location.offset;
            } else {
                if (!lookup_label(tokval->t_charptr, &label_seg, &label_ofs)) {
                    scope = local_scope(tokval->t_charptr);
                    if (critical == 2) {
                        nasm_error(ERR_NONFATAL, "symbol `%s%s' undefined",
                              scope,tokval->t_charptr);
                        return NULL;
                    } else if (critical == 1) {
                        nasm_error(ERR_NONFATAL,
                              "symbol `%s%s' not defined before use",
                              scope,tokval->t_charptr);
                        return NULL;
                    } else {
                        if (opflags)
                            *opflags |= OPFLAG_FORWARD;
                        type = EXPR_UNKNOWN;
                        label_seg = NO_SEG;
                        label_ofs = 1;
                    }
                }
                if (opflags && is_extern(tokval->t_charptr))
                    *opflags |= OPFLAG_EXTERN;
            }
            addtotemp(type, label_ofs);
            if (label_seg != NO_SEG)
                addtotemp(EXPR_SEGBASE + label_seg, 1L);
            break;
        case TOKEN_DECORATOR:
            addtotemp(EXPR_RDSAE, tokval->t_integer);
            break;
        }
        i = scan(scpriv, tokval);
        return finishtemp();

    default:
        nasm_error(ERR_NONFATAL, "expression syntax error");
        return NULL;
    }
}
예제 #12
0
파일: eval.c 프로젝트: AxFab/nasm
static expr *expr5(int critical)
{
    expr *e, *f;

    e = expr6(critical);
    if (!e)
        return NULL;
    while (i == '*' || i == '/' || i == '%' ||
           i == TOKEN_SDIV || i == TOKEN_SMOD) {
        int j = i;
        i = scan(scpriv, tokval);
        f = expr6(critical);
        if (!f)
            return NULL;
        if (j != '*' && (!(is_simple(e) || is_just_unknown(e)) ||
                         !(is_simple(f) || is_just_unknown(f)))) {
            nasm_error(ERR_NONFATAL, "division operator may only be applied to"
                  " scalar values");
            return NULL;
        }
        if (j != '*' && !is_unknown(f) && reloc_value(f) == 0) {
            nasm_error(ERR_NONFATAL, "division by zero");
            return NULL;
        }
        switch (j) {
        case '*':
            if (is_simple(e))
                e = scalar_mult(f, reloc_value(e), true);
            else if (is_simple(f))
                e = scalar_mult(e, reloc_value(f), true);
            else if (is_just_unknown(e) && is_just_unknown(f))
                e = unknown_expr();
            else {
                nasm_error(ERR_NONFATAL, "unable to multiply two "
                      "non-scalar objects");
                return NULL;
            }
            break;
        case '/':
            if (is_just_unknown(e) || is_just_unknown(f))
                e = unknown_expr();
            else
                e = scalarvect(((uint64_t)reloc_value(e)) /
                               ((uint64_t)reloc_value(f)));
            break;
        case '%':
            if (is_just_unknown(e) || is_just_unknown(f))
                e = unknown_expr();
            else
                e = scalarvect(((uint64_t)reloc_value(e)) %
                               ((uint64_t)reloc_value(f)));
            break;
        case TOKEN_SDIV:
            if (is_just_unknown(e) || is_just_unknown(f))
                e = unknown_expr();
            else
                e = scalarvect(((int64_t)reloc_value(e)) /
                               ((int64_t)reloc_value(f)));
            break;
        case TOKEN_SMOD:
            if (is_just_unknown(e) || is_just_unknown(f))
                e = unknown_expr();
            else
                e = scalarvect(((int64_t)reloc_value(e)) %
                               ((int64_t)reloc_value(f)));
            break;
        }
    }
    return e;
}