parse_~1.c

操作系统源代码

	case '&':
	case '|':
	case '^':
	case S_LEFTSHIFT:
	case S_RIGHTSHIFT:
		(*pn)++;
		a1= e;
		if ((a2= ack_get_C_expression(pn)) == nil) {
			del_expr(a1);
			return nil;
		}
		e= new_expr();
		e->operator= t->symbol;
		e->left= a1;
		e->right= a2;
	}
	return e;
}

static expression_t *ack_get_operand(int *pn, int deref)
/* get something like: (memory), offset(base)(index*scale), or simpler. */
{
	expression_t *e, *offset, *base, *index;
	token_t *t;
	int c;

	/* Is it (memory)? */
	if (get_token(*pn)->symbol == '('
		&& ((t= get_token(*pn + 1))->type != T_WORD
			|| !isregister(t->name))
	) {
		/* A memory dereference. */
		(*pn)++;
		if ((offset= ack_get_C_expression(pn)) == nil) return nil;
		if (get_token(*pn)->symbol != ')') {
			parse_err(1, t, "operand syntax error\n");
			del_expr(offset);
			return nil;
		}
		(*pn)++;
		e= new_expr();
		e->operator= '(';
		e->middle= offset;
		return e;
	}

	/* #constant? */
	if (dialect == NCC && deref
			&& ((c= get_token(*pn)->symbol) == '#' || c == '*')) {
		/* NCC: mov ax,#constant  ->  ACK: mov ax,constant */
		(*pn)++;
		return ack_get_C_expression(pn);
	}

	/* @address? */
	if (dialect == NCC && get_token(*pn)->symbol == '@') {
		/* NCC: jmp @address  ->  ACK: jmp (address) */
		(*pn)++;
		if ((offset= ack_get_operand(pn, deref)) == nil) return nil;
		e= new_expr();
		e->operator= '(';
		e->middle= offset;
		return e;
	}

	/* Offset? */
	if (get_token(*pn)->symbol != '(') {
		/* There is an offset. */
		if ((offset= ack_get_C_expression(pn)) == nil) return nil;
	} else {
		/* No offset. */
		offset= nil;
	}

	/* (base)? */
	if (get_token(*pn)->symbol == '('
		&& (t= get_token(*pn + 1))->type == T_WORD
		&& isregister(t->name)
		&& get_token(*pn + 2)->symbol == ')'
	) {
		/* A base register expression. */
		base= new_expr();
		base->operator= 'B';
		base->name= copystr(t->name);
		(*pn)+= 3;
	} else {
		/* No base register expression. */
		base= nil;
	}

	/* (index*scale)? */
	if (get_token(*pn)->symbol == '(') {
		/* An index most likely. */
		token_t *m= nil;

		if (!(		/* This must be true: */
			(t= get_token(*pn + 1))->type == T_WORD
			&& isregister(t->name)
			&& (get_token(*pn + 2)->symbol == ')' || (
				get_token(*pn + 2)->symbol == '*'
				&& (m= get_token(*pn + 3))->type == T_WORD
				&& strchr("1248", m->name[0]) != nil
				&& m->name[1] == 0
				&& get_token(*pn + 4)->symbol == ')'
			))
		)) {
			/* Alas it isn't */
			parse_err(1, t, "operand syntax error\n");
			del_expr(offset);
			del_expr(base);
			return nil;
		}
		/* Found an index. */
		index= new_expr();
		index->operator= m == nil ? '1' : m->name[0];
		index->name= copystr(t->name);
		(*pn)+= (m == nil ? 3 : 5);
	} else {
		/* No index. */
		index= nil;
	}

	if (dialect == NCC && deref && base == nil && index == nil
		&& !(offset != nil && offset->operator == 'W'
					&& isregister(offset->name))
	) {
		/* NCC: mov ax,thing  ->  ACK mov ax,(thing) */
		e= new_expr();
		e->operator= '(';
		e->middle= offset;
		return e;
	}

	if (base == nil && index == nil) {
		/* Return a lone offset as is. */
		e= offset;
	} else {
		e= new_expr();
		e->operator= 'O';
		e->left= offset;
		e->middle= base;
		e->right= index;
	}
	return e;
}

static expression_t *ack_get_oplist(int *pn, int deref)
/* Get a comma (or colon for jmpf and callf) separated list of instruction
 * operands.
 */
{
	expression_t *e, *o1, *o2;
	token_t *t;

	if ((e= ack_get_operand(pn, deref)) == nil) return nil;

	if ((t= get_token(*pn))->symbol == ',' || t->symbol == ':') {
		o1= e;
		(*pn)++;
		if ((o2= ack_get_oplist(pn, deref)) == nil) {
			del_expr(o1);
			return nil;
		}
		e= new_expr();
		e->operator= ',';
		e->left= o1;
		e->right= o2;
	}
	return e;
}

static asm86_t *ack_get_statement(void)
/* Get a pseudo op or machine instruction with arguments. */
{
	token_t *t= get_token(0);
	asm86_t *a;
	mnemonic_t *m;
	int n;
	int prefix_seen;
	int oaz_prefix;
	int deref;

	assert(t->type == T_WORD);

	if (strcmp(t->name, ".sect") == 0) {
		/* .sect .text etc.  Accept only four segment names. */
		skip_token(1);
		t= get_token(0);
		if (t->type != T_WORD || (
			strcmp(t->name, ".text") != 0
			&& strcmp(t->name, ".rom") != 0
			&& strcmp(t->name, ".data") != 0
			&& strcmp(t->name, ".bss") != 0
			&& strcmp(t->name, ".end") != 0
		)) {
			parse_err(1, t, "weird section name to .sect\n");
			return nil;
		}
	}
	a= new_asm86();

	/* Process instruction prefixes. */
	oaz_prefix= 0;
	for (prefix_seen= 0;; prefix_seen= 1) {
		if (strcmp(t->name, "o16") == 0) {
			if (use16()) {
				parse_err(1, t, "o16 in an 8086 section\n");
			}
			oaz_prefix|= OPZ;
		} else
		if (strcmp(t->name, "o32") == 0) {
			if (use32()) {
				parse_err(1, t, "o32 in an 80386 section\n");
			}
			oaz_prefix|= OPZ;
		} else
		if (strcmp(t->name, "a16") == 0) {
			if (use16()) {
				parse_err(1, t, "a16 in an 8086 section\n");
			}
			oaz_prefix|= ADZ;
		} else
		if (strcmp(t->name, "a32") == 0) {
			if (use32()) {
				parse_err(1, t, "a32 in an 80386 section\n");
			}
			oaz_prefix|= ADZ;
		} else
		if (strcmp(t->name, "rep") == 0
			|| strcmp(t->name, "repe") == 0
			|| strcmp(t->name, "repne") == 0
			|| strcmp(t->name, "repz") == 0
			|| strcmp(t->name, "repnz") == 0
		) {
			if (a->rep != ONCE) {
				parse_err(1, t,
					"can't have more than one rep\n");
			}
			switch (t->name[3]) {
			case 0:		a->rep= REP;	break;
			case 'e':
			case 'z':	a->rep= REPE;	break;
			case 'n':	a->rep= REPNE;	break;
			}
		} else
		if (strchr("cdefgs", t->name[0]) != nil
					&& strcmp(t->name+1, "seg") == 0) {
			if (a->seg != DEFSEG) {
				parse_err(1, t,
				"can't have more than one segment prefix\n");
			}
			switch (t->name[0]) {
			case 'c':	a->seg= CSEG;	break;
			case 'd':	a->seg= DSEG;	break;
			case 'e':	a->seg= ESEG;	break;
			case 'f':	a->seg= FSEG;	break;
			case 'g':	a->seg= GSEG;	break;
			case 's':	a->seg= SSEG;	break;
			}
		} else
		if (!prefix_seen) {
			/* No prefix here, get out! */
			break;
		} else {
			/* No more prefixes, next must be an instruction. */
			if (t->type != T_WORD
				|| (m= search_mnem(t->name)) == nil
				|| m->optype == PSEUDO
			) {
				parse_err(1, t,
		"machine instruction expected after instruction prefix\n");
				del_asm86(a);
				return nil;
			}
			if (oaz_prefix != 0 && m->optype != JUMP
						&& m->optype != WORD) {
				parse_err(1, t,
			"'%s' can't have an operand size prefix\n", m->name);
			}
			break;
		}

		/* Skip the prefix and extra newlines. */
		do {
			skip_token(1);
		} while ((t= get_token(0))->symbol == ';');
	}

	/* All the readahead being done upsets the line counter. */
	a->line= t->line;

	/* Read a machine instruction or pseudo op. */
	if ((m= search_mnem(t->name)) == nil) {
		parse_err(1, t, "unknown instruction\n");
		del_asm86(a);
		return nil;
	}
	a->opcode= m->opcode;
	a->optype= m->optype;
	a->oaz= oaz_prefix;

	switch (a->opcode) {
	case IN:
	case OUT:
	case INT:
		deref= 0;
		break;
	default:
		deref= (a->optype >= BYTE);
	}
	n= 1;
	if (get_token(1)->symbol != ';'
			&& (a->args= ack_get_oplist(&n, deref)) == nil) {
		del_asm86(a);
		return nil;
	}
	if (get_token(n)->symbol != ';') {
		parse_err(1, t, "garbage at end of instruction\n");
		del_asm86(a);
		return nil;
	}
	switch (a->opcode) {
	case DOT_ALIGN:
		/* Restrict .align to have a single numeric argument, some
		 * assemblers think of the argument as a power of two, so
		 * we need to be able to change the value.
		 */
		if (a->args == nil || a->args->operator != 'W'
					|| !isanumber(a->args->name)) {
			parse_err(1, t,
			  ".align is restricted to one numeric argument\n");
			del_asm86(a);
			return nil;
		}
		break;
	case JMPF:
	case CALLF:
		/* NCC jmpf off,seg  ->  ACK jmpf seg:off */
		if (dialect == NCC && a->args != nil
						&& a->args->operator == ',') {
			expression_t *t;

			t= a->args->left;
			a->args->left= a->args->right;
			a->args->right= t;
			break;
		}
		/*FALL THROUGH*/
	case JMP:
	case CALL:
		/* NCC jmp @(reg)  ->  ACK jmp (reg) */
		if (dialect == NCC && a->args != nil && (
			(a->args->operator == '('
				&& a->args->middle != nil
				&& a->args->middle->operator == 'O')
			|| (a->args->operator == 'O'
				&& a->args->left == nil
				&& a->args->middle != nil
				&& a->args->right == nil)
		)) {
			expression_t *t;

			t= a->args;
			a->args= a->args->middle;
			t->middle= nil;
			del_expr(t);
			if (a->args->operator == 'B') a->args->operator= 'W';
		}
		break;
	default:;
	}
	skip_token(n+1);
	return a;
}

asm86_t *ack_get_instruction(void)
{
	asm86_t *a= nil;
	expression_t *e;
	token_t *t;

	while ((t= get_token(0))->symbol == ';')
		skip_token(1);

	if (t->type == T_EOF) return nil;

	if (t->symbol == '#') {
		/* Preprocessor line and file change. */

		if ((t= get_token(1))->type != T_WORD || !isanumber(t->name)
			|| get_token(2)->type != T_STRING
		) {
			parse_err(1, t, "file not preprocessed?\n");
			zap();
		} else {
			set_file(get_token(2)->name,
				strtol(get_token(1)->name, nil, 0) - 1);

			/* GNU CPP adds extra cruft, simply zap the line. */
			zap();
		}
		a= ack_get_instruction();
	} else
	if (t->type == T_WORD && get_token(1)->symbol == ':') {
		/* A label definition. */
		a= new_asm86();
		a->line= t->line;
		a->opcode= DOT_LABEL;
		a->optype= PSEUDO;
		a->args= e= new_expr();
		e->operator= ':';
		e->name= copystr(t->name);
		skip_token(2);
	} else
	if (t->type == T_WORD && get_token(1)->symbol == '=') {
		int n= 2;

		if ((e= ack_get_C_expression(&n)) == nil) {
			zap();
			a= ack_get_instruction();
		} else
		if (get_token(n)->symbol != ';') {
			parse_err(1, t, "garbage after assignment\n");
			zap();
			a= ack_get_instruction();
		} else {
			a= new_asm86();
			a->line= t->line;
			a->opcode= DOT_EQU;
			a->optype= PSEUDO;
			a->args= new_expr();
			a->args->operator= '=';
			a->args->name= copystr(t->name);
			a->args->middle= e;
			skip_token(n+1);
		}
	} else
	if (t->type == T_WORD) {
		if ((a= ack_get_statement()) == nil) {
			zap();
			a= ack_get_instruction();
		}
	} else {
		parse_err(1, t, "syntax error\n");
		zap();
		a= ack_get_instruction();
	}
	return a;
}

asm86_t *ncc_get_instruction(void)
{
	return ack_get_instruction();
}