mips/masm/gen.c

#include <stdlib.h>
#include <merror.h>
#include <melf.h>
#include <mips32.h>
#include <mips32r6.h>

#include "tab.h"
#include "gen.h"
#include "parse.h"

///
/// section table
///

static void section_get_default_perm(struct section *sec, const char *name)
{
	#define __LEN 7
	static const struct perms {
		char *name;
		bool read;
		bool write;
		bool execute;
		int alignment;
	} defaults[__LEN] = {
		{".text", true, false, true, 4},
		{".code", true, false, true, 4},
		{".data", true, true, false, 1},
		{".stack", true, true, false, 1},
		{".rodata", true, false, false, 1},
		{".bss", true, true, false, 1},
		{".robss", true, false, false, 1},
	};

	for (int i = 0; i < __LEN; i++) {
		const struct perms *p = &defaults[i];
		if (strcasecmp(name, p->name) != 0)
			continue;
		sec->read = p->read;
		sec->write = p->write;
		sec->execute = p->execute;
		sec->align = p->alignment;
		break;
	}

}

static int section_get(struct generator *gen, struct section **res,
		       const struct string *const name)
{
	/// find the section if it exists
	for (size_t i = 0; i < gen->sections_len; i++) {
		struct section *sec = &gen->sections[i];
		if (sec->name.len != name->len)
			continue;
		if (strcmp(sec->name.str, name->str) != 0)
			continue;
		*res = sec;
		return M_SUCCESS;
	}

	/// allocate a new one if it doesnt
	size_t size = gen->sections_size ? gen->sections_size * 2 : 8;
	void *new = realloc(gen->sections, size * sizeof(struct section));
	if (new == NULL) {
		PERROR("cannot realloc");
		return M_ERROR;
	}

	gen->sections_size = size;
	gen->sections = new;

	struct section *sec = &gen->sections[gen->sections_len++];

	// alloc reftab
	if (reftab_init(&sec->reftab))
		return M_ERROR;

	// copy name
	if (string_clone(&sec->name, name))
		return M_ERROR;

	// set defaults
	sec->len = 0;
	sec->size = 0;
	sec->align = 1;
	sec->data = NULL;
	sec->read = true;
	sec->write = true;
	sec->execute = false;
	section_get_default_perm(sec, name->str);

	*res = sec;
	return M_SUCCESS;
}

static int section_extend(struct section *section, size_t space)
{
	size_t newlen = section->len + space;
	if (newlen < section->size)
		return M_SUCCESS;

	size_t size = section->size ? section->size * 2 + newlen : newlen * 2;
	void *new = realloc(section->data, size);
	if (new == NULL) {
		PERROR("cannot realloc");
		return M_ERROR;
	}
	section->size = size;
	section->data = new;

	return M_SUCCESS;
}

static int section_push(struct section *section, void *data, size_t len)
{
	size_t newlen = section->len + len;
	size_t zeros = newlen % section->align;
	if (zeros)
		zeros = section->align - zeros;

	if (section_extend(section, len + zeros))
		return M_ERROR;

	memset(section->data + section->len, 0, zeros);
	memcpy(section->data + section->len + zeros, data, len);
	section->len += len + zeros;

	return M_SUCCESS;
}

static int section_zero(struct section *section, size_t len)
{
	size_t zeros = section->len % section->align;
	if (zeros)
		zeros = section->align - zeros;

	if (section_extend(section, len + zeros))
		return M_ERROR;

	memset(section->data + section->len, 0, len + zeros);
	section->len += len + zeros;

	return M_SUCCESS;
}

void section_free(struct section *section)
{
	reftab_free(&section->reftab);
	string_free(&section->name);
	free(section->data);
}

///
/// generation functions
///

static void print_curr_line(struct generator *gen,
			    const struct expr *const expr)
{
	int line = expr->line_no,
	    len = expr->byte_end - expr->byte_start,
	    nl = true,
	    c = EOF;
	FILE *file = gen->parser.lexer.file;

	fseek(file, expr->byte_start, SEEK_SET);

	while (len--) {
		c = getc(file);
		if (c == EOF || c == '\0')
			break;
		if (nl) {
			fprintf(stderr, "\t%d | ", line);
			line++;
			nl = false;
		}
		if (c == '\n')
			nl = true;
		putc(c, stderr);
	}

}

static int gen_directive_whb(struct generator *gen, const void *data,
			     uint32_t count, uint32_t len)
{
	// TODO: endianess
	for (uint32_t i = 0; i < count; i++) {
		void *ptr = (char *) data + (len * i);
		if (section_push(gen->current, ptr, len))
			return M_ERROR;
	}

	return M_SUCCESS;
}

static int gen_directive(struct generator *gen,
			 const struct expr *const e)
{
	const struct expr_directive *const expr = &e->directive;
	int res = M_SUCCESS;

	switch (expr->type) {
	case EXPR_DIRECTIVE_ALIGN:
		if (expr->align < 1) {
			ERROR("alignment cannot be zero");
			print_curr_line(gen, e);
			return M_ERROR;
		}
		gen->current->align = expr->align;
		break;
	case EXPR_DIRECTIVE_SPACE:
		res = section_zero(gen->current, expr->space);
		break;
	case EXPR_DIRECTIVE_WORD:
		res = gen_directive_whb(gen, expr->words, expr->len,
			   sizeof(uint32_t));
		break;
	case EXPR_DIRECTIVE_HALF:
		res = gen_directive_whb(gen, expr->halfs, expr->len,
			   sizeof(uint16_t));
		break;
	case EXPR_DIRECTIVE_BYTE:
		res = gen_directive_whb(gen, expr->bytes, expr->len,
			   sizeof(uint8_t));
		break;
	case EXPR_DIRECTIVE_SECTION:
		res = section_get(gen, &gen->current, &expr->section);
		break;
	case EXPR_DIRECTIVE_EXTERN: {
		struct symbol *sym;
		res = symtab_find_or_stub(&gen->symtab, &sym, &expr->label);
		if (res == M_SUCCESS)
			sym->type = SYM_EXTERN;
		break;
	}
	case EXPR_DIRECTIVE_GLOBL: {
		struct symbol *sym;
		res = symtab_find_or_stub(&gen->symtab, &sym, &expr->label);
		if (res == M_SUCCESS)
			sym->type = SYM_GLOBAL;
		break;
	}
	case EXPR_DIRECTIVE_ASCII:
		res = section_push(gen->current, expr->string.str,
		     expr->string.len - 1);
		break;
	case EXPR_DIRECTIVE_ASCIIZ:
		res = section_push(gen->current, expr->string.str,
		     expr->string.len);
		break;
	}

	return res;
}

static int gen_constant(struct generator *gen, struct expr_const *const expr)
{
	(void) gen;
	(void) expr;

	ERROR("constants not yet implemented");
	return M_ERROR;
}

static enum grammer_type get_gmr_type(const char *name, size_t *len)
{
	#define CHK(part, str) { \
		if (strncasecmp(str, name, strlen(str)) == 0) { \
			*len = strlen(str); \
			return GMR_ ##part; \
		}} \

	CHK(RD, "rd")
	CHK(RS, "rs")
	CHK(RT, "rt")
	CHK(IMMD, "immd")
	CHK(OFFSET_BASE, "offset(base)")
	CHK(OFFSET, "offset")
	CHK(TARGET, "target")
	CHK(HI, "hi")
	CHK(LO, "lo")

	#undef CHK

	ERROR("!!! BUG: this should never hit !!!");
	exit(1);
}

static int parse_register(enum mips32_register *reg, struct string *name)
{
	int len = name->len;
	int c0 = len > 0 ? name->str[0] : '\0',
	    c1 = len > 1 ? name->str[1] : '\0',
	    c2 = len > 2 ? name->str[2] : '\0',
	    c3 = len > 3 ? name->str[3] : '\0';

	// $zero
	if (c0 == 'z') {
		if (c1 == 'e' && c2 == 'r' && c3 == 'o') {
			*reg = MIPS32_REG_ZERO;
			return M_SUCCESS;
		}
	}

	// $a0-a3 $at
	else if (c0 == 'a') {
		if (c1 == 't') {
			*reg = MIPS32_REG_AT;
			return M_SUCCESS;
		}
		if (c1 >= '0' && c1 <= '3') {
			*reg = MIPS32_REG_A0;
			*reg += c1 - '0';
			return M_SUCCESS;
		}
	}

	// $v0-v1
	else if (c0 == 'v') {
		if (c1 >= '0' && c1 <= '1') {
			*reg = MIPS32_REG_V0;
			*reg += c1 - '0';
			return M_SUCCESS;
		}
	}

	// $t0-t9
	else if (c0 == 't') {
		if (c1 >= '0' && c1 <= '7') {
			*reg = MIPS32_REG_T0;
			*reg += c1 - '0';
			return M_SUCCESS;
		}
		// reg T8-T9 are not in order with T0-T7
		if (c1 >= '8' && c1 <= '9') {
			*reg = MIPS32_REG_T8;
			*reg += c1 - '8';
			return M_SUCCESS;
		}
	}

	// $s0-s7 $sp
	else if (c0 == 's') {
		if (c1 >= '0' && c1 <= '7') {
			*reg = MIPS32_REG_S0;
			*reg += c1 - '0';
			return M_SUCCESS;
		}
		if (c1 == 'p') {
			*reg = MIPS32_REG_SP;
			return M_SUCCESS;
		}
	}

	// $k0-k1
	else if (c0 == 'k') {
		if (c1 >= '0' && c1 <= '1') {
			*reg = MIPS32_REG_K0;
			*reg += c1 - '0';
			return M_SUCCESS;
		}
	}

	// $gp
	else if (c0 == 'g') {
		if (c1 == 'p') {
			*reg = MIPS32_REG_GP;
			return M_SUCCESS;
		}
	}

	// $fp
	else if (c0 == 'f') {
		if (c1 == 'p') {
			*reg = MIPS32_REG_FP;
			return M_SUCCESS;
		}
	}

	// $rp
	else if (c0 == 'r') {
		if (c1 == 'a') {
			*reg = MIPS32_REG_RA;
			return M_SUCCESS;
		}
	}

	// $0-31 (non aliased register names)
	else if (c0 >= '0' && c0 <= '9') {
		int i = c0 - '0';
		if (c1 >= '0' && c1 <= '9') {
			i *= 10;
			i += c1 - '0';
		}
		if (i <= 31) {
			*reg = i;
			return M_SUCCESS;
		}
	}

	ERROR("unknown register $%.*s", name->len, name->str);
	return M_ERROR;
}

static int gen_ins_read_state(struct generator *gen,
			      struct expr *const expr,
			      struct gen_ins_state *state,
			      struct mips32_grammer *grammer)
{
	char *ptr = grammer->grammer;
	uint32_t argi = 0;

	// read values into state
	while (*ptr != '\0') {

		if (argi >= expr->instruction.args_len) {
			ERROR("not enough arguments passed");
			print_curr_line(gen, expr);
			return M_ERROR;
		}
		struct expr_ins_arg *arg = &expr->instruction.args[argi++];

		size_t skip;
		switch (get_gmr_type(ptr, &skip)) {
		case GMR_RD:
			// rd
			if (arg->type != EXPR_INS_ARG_REGISTER) {
				ERROR("expected a register");
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			if (parse_register(&state->rd, &arg->reg)) {
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			break;
		case GMR_RS:
			// rs
			if (arg->type != EXPR_INS_ARG_REGISTER) {
				ERROR("expected a register");
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			if (parse_register(&state->rs, &arg->reg)) {
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			break;
		case GMR_RT:
			// rt
			if (arg->type != EXPR_INS_ARG_REGISTER) {
				ERROR("expected a register");
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			if (parse_register(&state->rt, &arg->reg)) {
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			break;
		case GMR_IMMD:
			// immd
			if (arg->type != EXPR_INS_ARG_IMMEDIATE) {
				ERROR("expected an immediate");
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			state->immd = arg->immd;
			break;
		case GMR_OFFSET:
			// offset
			state->offset = 0;
			if (arg->type == EXPR_INS_ARG_IMMEDIATE)
				state->offset = arg->immd;
			else if (arg->type == EXPR_INS_ARG_LABEL)
				state->label = &arg->label;
			else {
				ERROR("invalid instruction");
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			break;
		case GMR_OFFSET_BASE:
			// offset(base)
			if (arg->type != EXPR_INS_ARG_OFFSET) {
				ERROR("expected an offset($base)");
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			state->offset = arg->offset.immd;
			if (parse_register(&state->base, &arg->offset.reg)) {
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			break;
		case GMR_TARGET:
			// target
			state->target = 0;
			if (arg->type == EXPR_INS_ARG_IMMEDIATE)
				state->target = arg->immd;
			else if (arg->type == EXPR_INS_ARG_LABEL)
				state->label = &arg->label;
			else {
				ERROR("invalid instruction");
				print_curr_line(gen, expr);
				return M_ERROR;
			}
			break;
		default:
			break;
		}

		// skip entry
		ptr += skip;

		// skip comma
		if (*ptr == ',') {
			ptr++;
			continue;
		} else if (*ptr == '\0') {
			break;
		} else {
			ERROR("!! BUG3: invalid splitting char %c !!!", *ptr);
			exit(1);
		}

	}

	return M_SUCCESS;
}

static int gen_ins_write_state(
	struct generator *gen,
	union mips32_instruction ins, // the instruction to modify
	struct gen_ins_state *state,  // the current read state
	char *grammer)		      // the gramemr to parse
{
	char *ptr = grammer;
	enum reference_type reftype = REF_NONE;

	// read values into state
	while (*ptr != '\0') {

		// parse next dsl entry
		size_t skip;
		enum grammer_type gmr = get_gmr_type(ptr, &skip);

		// check for dsl hardcoded register argument
		bool hardcoded = false;
		enum mips32_register hard_reg;
		if (*(ptr + skip) == '=') {
			// parse argument
			char *rptr = ptr + skip + 2;
			hardcoded = true;
			struct string regname;
			string_bss(&regname, rptr);
			if (parse_register(&hard_reg, &regname)) {
				ERROR("!!! BUG2: this should never hit !!!");
				exit(1);
			}
		}

		// skip till next comma
		for (;*ptr != '\0' && *ptr != ','; ptr++);
		if (*ptr == ',')
			ptr++;

		switch (gmr) {
		case GMR_RD:
			ins.rd = hardcoded ? hard_reg : state->rd;
			break;
		case GMR_RS:
			ins.rs = hardcoded ? hard_reg : state->rs;
			break;
		case GMR_RT:
			ins.rt = hardcoded ? hard_reg : state->rt;
			break;
		case GMR_IMMD:
			ins.immd = state->immd;
			break;
		case GMR_OFFSET:
			ins.offset = state->offset;
			reftype = REF_MIPS_PC16;
			break;
		case GMR_OFFSET_BASE:
			ins.offset = state->offset;
			ins.rs = state->base;
			reftype = REF_MIPS_16;
			break;
		case GMR_TARGET:
			ins.target = state->target;
			reftype = REF_MIPS_26;
			break;
		case GMR_HI:
			ins.immd = state->target >> 16;
			reftype = REF_MIPS_HI16;
			break;
		case GMR_LO:
			ins.immd = state->target & 0x0000FFFF;
			reftype = REF_MIPS_LO16;
			break;
		}
	}

	// get offset for reference (if needed)
	uint32_t offset = gen->current->len;
	size_t zeros = offset % gen->current->align;
	if (zeros)
		zeros = gen->current->align - zeros;
	offset += zeros;

	// write instructon to section
	uint32_t raw = B32(ins.raw);
	if (section_push(gen->current, &raw, sizeof(uint32_t))) {
		return M_ERROR;
	}

	// create reference (if needed)
	if (reftype != REF_NONE && state->label != NULL) {
		struct symbol *sym;

		if (symtab_find_or_stub(&gen->symtab, &sym, state->label))
			return M_ERROR;

		struct reference ref = {
			.type = reftype,
			.symbol = sym,
			.offset = offset
		};

		if (reftab_push(&gen->current->reftab, &ref)) {
			return M_ERROR;
		}
	}

	return M_SUCCESS;
}

static int gen_ins(struct generator *gen, struct expr *const expr)
{
	struct mips32_grammer *grammer = NULL;
	for (uint32_t i = 0; i < gen->grammers_len; i++) {
		struct mips32_grammer *temp = &gen->grammers[i];
		if (strcasecmp(temp->name, expr->instruction.name.str) != 0)
			continue;
		grammer = temp;
		break;
	}

	if (grammer == NULL) {
		ERROR("unknown instruction");
		print_curr_line(gen, expr);
		return M_ERROR;
	}

	struct gen_ins_state state;
	state.label = NULL;

	// read in the values from the parser
	if (gen_ins_read_state(gen, expr, &state, grammer))
		return M_ERROR;

	// write the values into the instructions
	// ...and then the sections
	if (grammer->pseudo_len > 0) {
		// write pseudo
		for (int i = 0; i < grammer->pseudo_len; i++) {
			union mips32_instruction ins = gen->instructions[
				grammer->pseudo_grammer[i].enum_index];
			if (gen_ins_write_state(gen, ins, &state,
				grammer->pseudo_grammer[i].update))
				return M_ERROR;
		}
	} else {
		// write real
		union mips32_instruction ins
			= gen->instructions[grammer->enum_index];
		if (gen_ins_write_state(gen, ins, &state, grammer->grammer))
			return M_ERROR;
	}

	return M_SUCCESS;
}

static int gen_label(struct generator *gen, struct string *const label)
{
	uint32_t offset = gen->current->len;
	ptrdiff_t secidx = gen->current - gen->sections;
	size_t zeros = offset % gen->current->align;
	if (zeros)
		zeros = gen->current->align - zeros;
	offset += zeros;

	struct symbol *sym;
	/* update existing symbol (if exists) */
	if (symtab_find(&gen->symtab, &sym, label->str) == M_SUCCESS) {
		if (sym->secidx != SYM_SEC_STUB) {
			// symbols that are not labeled stub are fully defined,
			// it is a error to redefine them
			ERROR("redefined symbol '%s'", label->str);
			return M_ERROR;
		}
		sym->secidx = secidx;
		sym->offset = offset;
	/* create a new symbol */
	} else {
		struct symbol new = {
			.secidx = secidx,
			.offset = offset,
			.type = SYM_LOCAL,
		};
		if (string_clone(&new.name, label))
			return M_ERROR;
		if (symtab_push(&gen->symtab, &new)) {
			string_free(&new.name);
			return M_ERROR;
		}
	}

	return M_SUCCESS;
}

/* run codegen */
static int generate(struct generator *gen)
{
	struct expr expr;
	int res = M_SUCCESS;

	// get the next expression
	if ((res = parser_next(&gen->parser, &expr)))
		return  res;

	// if its not a segment directive
	// (and we dont have a section)
	// create the default
	if ((
		expr.type != EXPR_DIRECTIVE ||
		expr.directive.type != EXPR_DIRECTIVE_SECTION) &&
		gen->current == NULL) {
		// create .data section
		struct string temp = {
			.str = ".data",
			.len = 5,
			.size = 5,
			.allocated = false
		};
		if (section_get(gen, &gen->current, &temp)) {
			expr_free(&expr);
			return M_ERROR;
		}
	}

	res = M_SUCCESS;
	switch (expr.type) {
	case EXPR_DIRECTIVE:
		res = gen_directive(gen, &expr);
		break;
	case EXPR_CONSTANT:
		res = gen_constant(gen, &expr.constant);
		break;
	case EXPR_INS:
		res = gen_ins(gen, &expr);
		break;
	case EXPR_LABEL:
		res = gen_label(gen, &expr.label);
		break;
	}

	expr_free(&expr);
	return res;
}

/* run codegen with the mips32r6 specification */
int generate_mips32r6(struct generator *gen)
{
	gen->instructions_len = __MIPS32R6_INS_LEN;
	gen->instructions = mips32r6_instructions;
	gen->grammers_len = __MIPS32R6_GRAMMER_LEN;
	gen->grammers = mips32r6_grammers;

	int res;
	while (res = generate(gen), 1) {
		if (res == M_ERROR)
			return M_ERROR;
		if (res == M_EOF)
			break;
	}

	return M_SUCCESS;
}

int generator_init(const char *file, struct generator *gen)
{
	if (parser_init(file, &gen->parser))
		return M_ERROR;
	if (symtab_init(&gen->symtab))
		return M_ERROR;
	gen->sections = NULL;
	gen->sections_len = 0;
	gen->sections_size = 0;
	return M_SUCCESS;
}

void generator_free(struct generator *gen)
{
	parser_free(&gen->parser);
	symtab_free(&gen->symtab);
	for (size_t i = 0; i < gen->sections_len; i++)
		section_free(&gen->sections[i]);
	free(gen->sections);
}