8db.c

这是一个同样来自贝尔实验室的和UNIX有着渊源的操作系统, 其简洁的设计和实现易于我们学习和理解

				ip->disp = s&0xffff;
			}
			if (igets(map, ip, (ushort*)&ip->seg) < 0)
				return 0;
			ip->jumptype = PTR;
			break;
		case AUXMM:	/* Multi-byte op code; prefix determines table selection */
			if (igetc(map, ip, &c) < 0)
				return 0;
			obase = (Optable*)op->proto;
			switch (ip->opre) {
			case 0x66:	op = optab660F; break;
			case 0xF2:	op = optabF20F; break;
			case 0xF3:	op = optabF30F; break;
			default:	op = nil; break;
			}
			if(op != nil && op[c].proto != nil)
				obase = op;
			norex = 1;	/* no more rex prefixes */
			/* otherwise the optab entry captures it */
			goto newop;
		case AUX:	/* Multi-byte op code - Auxiliary table */
			obase = (Optable*)op->proto;
			if (igetc(map, ip, &c) < 0)
				return 0;
			goto newop;
		case OPRE:	/* Instr Prefix or media op */
			ip->opre = c;
			/* fall through */
		case PRE:	/* Instr Prefix */
			ip->prefix = (char*)op->proto;
			if (igetc(map, ip, &c) < 0)
				return 0;
			if (ip->opre && c == 0x0F)
				ip->prefix = 0;
			goto newop;
		case SEG:	/* Segment Prefix */
			ip->segment = (char*)op->proto;
			if (igetc(map, ip, &c) < 0)
				return 0;
			goto newop;
		case OPOVER:	/* Operand size override */
			ip->opre = c;
			ip->osize = 'W';
			if (igetc(map, ip, &c) < 0)
				return 0;
			if (c == 0x0F)
				ip->osize = 'L';
			else if (ip->amd64 && (c&0xF0) == 0x40)
				ip->osize = 'Q';
			goto newop;
		case ADDOVER:	/* Address size override */
			ip->asize = 0;
			if (igetc(map, ip, &c) < 0)
				return 0;
			goto newop;
		case JUMP:	/* mark instruction as JUMP or RET */
		case RET:
			ip->jumptype = op->operand[i];
			break;
		default:
			werrstr("bad operand type %d", op->operand[i]);
			return 0;
		}
	}
	return op;
}

#pragma	varargck	argpos	bprint		2

static void
bprint(Instr *ip, char *fmt, ...)
{
	va_list arg;

	va_start(arg, fmt);
	ip->curr = vseprint(ip->curr, ip->end, fmt, arg);
	va_end(arg);
}

/*
 *  if we want to call 16 bit regs AX,BX,CX,...
 *  and 32 bit regs EAX,EBX,ECX,... then
 *  change the defs of ANAME and ONAME to:
 *  #define	ANAME(ip)	((ip->asize == 'E' ? "E" : "")
 *  #define	ONAME(ip)	((ip)->osize == 'L' ? "E" : "")
 */
#define	ANAME(ip)	""
#define	ONAME(ip)	""

static char *reg[] =  {
[AX]	"AX",
[CX]	"CX",
[DX]	"DX",
[BX]	"BX",
[SP]	"SP",
[BP]	"BP",
[SI]	"SI",
[DI]	"DI",

	/* amd64 */
[R8]	"R8",
[R9]	"R9",
[R10]	"R10",
[R11]	"R11",
[R12]	"R12",
[R13]	"R13",
[R14]	"R14",
[R15]	"R15",
};

static char *breg[] = { "AL", "CL", "DL", "BL", "AH", "CH", "DH", "BH" };
static char *breg64[] = { "AL", "CL", "DL", "BL", "SPB", "BPB", "SIB", "DIB",
	"R8B", "R9B", "R10B", "R11B", "R12B", "R13B", "R14B", "R15B" };
static char *sreg[] = { "ES", "CS", "SS", "DS", "FS", "GS" };

static void
plocal(Instr *ip)
{
	int ret;
	long offset;
	Symbol s;
	char *reg;

	offset = ip->disp;
	if (!findsym(ip->addr, CTEXT, &s) || !findlocal(&s, FRAMENAME, &s)) {
		bprint(ip, "%lux(SP)", offset);
		return;
	}

	if (s.value > ip->disp) {
		ret = getauto(&s, s.value-ip->disp-mach->szaddr, CAUTO, &s);
		reg = "(SP)";
	} else {
		offset -= s.value;
		ret = getauto(&s, offset, CPARAM, &s);
		reg = "(FP)";
	}
	if (ret)
		bprint(ip, "%s+", s.name);
	else
		offset = ip->disp;
	bprint(ip, "%lux%s", offset, reg);
}

static int
isjmp(Instr *ip)
{
	switch(ip->jumptype){
	case Iwds:
	case Jbs:
	case JUMP:
		return 1;
	default:
		return 0;
	}
}

/*
 * This is too smart for its own good, but it really is nice
 * to have accurate translations when debugging, and it
 * helps us identify which code is different in binaries that
 * are changed on sources.
 */
static int
issymref(Instr *ip, Symbol *s, long w, long val)
{
	Symbol next, tmp;
	long isstring, size;

	if (isjmp(ip))
		return 1;
	if (s->class==CTEXT && w==0)
		return 1;
	if (s->class==CDATA) {
		/* use first bss symbol (or "end") rather than edata */
		if (s->name[0]=='e' && strcmp(s->name, "edata") == 0){
			if((s ->index >= 0 && globalsym(&tmp, s->index+1) && tmp.value==s->value)
			|| (s->index > 0 && globalsym(&tmp, s->index-1) && tmp.value==s->value))
				*s = tmp;
		}
		if (w == 0)
			return 1;
		for (next=*s; next.value==s->value; next=tmp)
			if (!globalsym(&tmp, next.index+1))
				break;
		size = next.value - s->value;
		if (w >= size)
			return 0;
		if (w > size-w)
			w = size-w;
		/* huge distances are usually wrong except in .string */
		isstring = (s->name[0]=='.' && strcmp(s->name, ".string") == 0);
		if (w > 8192 && !isstring)
			return 0;
		/* medium distances are tricky - look for constants */
		/* near powers of two */
		if ((val&(val-1)) == 0 || (val&(val+1)) == 0)
			return 0;
		return 1;
	}
	return 0;
}

static void
immediate(Instr *ip, vlong val)
{
	Symbol s;
	long w;

	if (findsym(val, CANY, &s)) {		/* TO DO */
		w = val - s.value;
		if (w < 0)
			w = -w;
		if (issymref(ip, &s, w, val)) {
			if (w)
				bprint(ip, "%s+%lux(SB)", s.name, w);
			else
				bprint(ip, "%s(SB)", s.name);
			return;
		}
/*
		if (s.class==CDATA && globalsym(&s, s.index+1)) {
			w = s.value - val;
			if (w < 0)
				w = -w;
			if (w < 4096) {
				bprint(ip, "%s-%lux(SB)", s.name, w);
				return;
			}
		}
*/
	}
	if((ip->rex & REXW) == 0)
		bprint(ip, "%lux", (long)val);
	else
		bprint(ip, "%llux", val);
}

static void
pea(Instr *ip)
{
	if (ip->mod == 3) {
		if (ip->osize == 'B')
			bprint(ip, (ip->rex & REXB? breg64: breg)[ip->base]);
		else if(ip->rex & REXB)
			bprint(ip, "%s%s", ANAME(ip), reg[ip->base+8]);
		else
			bprint(ip, "%s%s", ANAME(ip), reg[ip->base]);
		return;
	}
	if (ip->segment)
		bprint(ip, ip->segment);
	if (ip->asize == 'E' && ip->base == SP)
		plocal(ip);
	else {
		if (ip->base < 0)
			immediate(ip, ip->disp);
		else {
			bprint(ip, "%lux", ip->disp);
			if(ip->rip)
				bprint(ip, "(RIP)");
			bprint(ip,"(%s%s)", ANAME(ip), reg[ip->rex&REXB? ip->base+8: ip->base]);
		}
	}
	if (ip->index >= 0)
		bprint(ip,"(%s%s*%d)", ANAME(ip), reg[ip->rex&REXX? ip->index+8: ip->index], 1<<ip->ss);
}

static void
prinstr(Instr *ip, char *fmt)
{
	vlong v;

	if (ip->prefix)
		bprint(ip, "%s ", ip->prefix);
	for (; *fmt && ip->curr < ip->end; fmt++) {
		if (*fmt != '%'){
			*ip->curr++ = *fmt;
			continue;
		}
		switch(*++fmt){
		case '%':
			*ip->curr++ = '%';
			break;
		case 'A':
			bprint(ip, "%s", ANAME(ip));
			break;
		case 'C':
			bprint(ip, "CR%d", ip->reg);
			break;
		case 'D':
			if (ip->reg < 4 || ip->reg == 6 || ip->reg == 7)
				bprint(ip, "DR%d",ip->reg);
			else
				bprint(ip, "???");
			break;
		case 'I':
			bprint(ip, "$");
			immediate(ip, ip->imm2);
			break;
		case 'O':
			bprint(ip,"%s", ONAME(ip));
			break;
		case 'i':
			bprint(ip, "$");
			v = ip->imm;
			if(ip->rex & REXW)
				v = ip->imm64;
			immediate(ip, v);
			break;
		case 'R':
			bprint(ip, "%s%s", ONAME(ip), reg[ip->rex&REXR? ip->reg+8: ip->reg]);
			break;
		case 'S':
			if(ip->osize == 'Q' || ip->osize == 'L' && ip->rex & REXW)
				bprint(ip, "Q");
			else
				bprint(ip, "%c", ip->osize);
			break;
		case 's':
			if(ip->opre == 0 || ip->opre == 0x66)
				bprint(ip, "P");
			else
				bprint(ip, "S");
			if(ip->opre == 0xf2 || ip->opre == 0x66)
				bprint(ip, "D");
			else
				bprint(ip, "S");
			break;
		case 'T':
			if (ip->reg == 6 || ip->reg == 7)
				bprint(ip, "TR%d",ip->reg);
			else
				bprint(ip, "???");
			break;
		case 'W':
			if (ip->osize == 'Q' || ip->osize == 'L' && ip->rex & REXW)
				bprint(ip, "CDQE");
			else if (ip->osize == 'L')
				bprint(ip,"CWDE");
			else
				bprint(ip, "CBW");
			break;
		case 'd':
			bprint(ip,"%ux:%lux",ip->seg,ip->disp);
			break;
		case 'm':
			if (ip->mod == 3 && ip->osize != 'B') {
				if(fmt[1] != '*'){
					if(ip->opre != 0) {
						bprint(ip, "X%d", ip->rex&REXB? ip->base+8: ip->base);
						break;
					}
				} else
					fmt++;
				bprint(ip, "M%d", ip->base);
				break;
			}
			pea(ip);
			break;
		case 'e':
			pea(ip);
			break;
		case 'f':
			bprint(ip, "F%d", ip->base);
			break;
		case 'g':
			if (ip->reg < 6)
				bprint(ip,"%s",sreg[ip->reg]);
			else
				bprint(ip,"???");
			break;
		case 'p':
			/*
			 * signed immediate in the ulong ip->imm.
			 */
			v = (long)ip->imm;
			immediate(ip, v+ip->addr+ip->n);
			break;
		case 'r':
			if (ip->osize == 'B')
				bprint(ip,"%s", (ip->rex? breg64: breg)[ip->rex&REXR? ip->reg+8: ip->reg]);
			else
				bprint(ip, reg[ip->rex&REXR? ip->reg+8: ip->reg]);
			break;
		case 'w':
			if (ip->osize == 'Q' || ip->rex & REXW)
				bprint(ip, "CQO");
			else if (ip->osize == 'L')
				bprint(ip,"CDQ");
			else
				bprint(ip, "CWD");
			break;
		case 'M':
			if(ip->opre != 0)
				bprint(ip, "X%d", ip->rex&REXR? ip->reg+8: ip->reg);
			else
				bprint(ip, "M%d", ip->reg);
			break;
		case 'x':
			if (ip->mod == 3 && ip->osize != 'B') {
				bprint(ip, "X%d", ip->rex&REXB? ip->base+8: ip->base);
				break;
			}
			pea(ip);
			break;
		case 'X':
			bprint(ip, "X%d", ip->rex&REXR? ip->reg+8: ip->reg);
			break;
		default:
			bprint(ip, "%%%c", *fmt);
			break;
		}
	}
	*ip->curr = 0;		/* there's always room for 1 byte */
}

static int
i386inst(Map *map, uvlong pc, char modifier, char *buf, int n)
{
	Instr instr;
	Optable *op;

	USED(modifier);
	op = mkinstr(map, &instr, pc);
	if (op == 0) {
		errstr(buf, n);
		return -1;
	}
	instr.curr = buf;
	instr.end = buf+n-1;
	prinstr(&instr, op->proto);
	return instr.n;
}

static int
i386das(Map *map, uvlong pc, char *buf, int n)
{
	Instr instr;
	int i;

	if (mkinstr(map, &instr, pc) == 0) {
		errstr(buf, n);
		return -1;
	}
	for(i = 0; i < instr.n && n > 2; i++) {
		_hexify(buf, instr.mem[i], 1);
		buf += 2;
		n -= 2;
	}
	*buf = 0;
	return instr.n;
}

static int
i386instlen(Map *map, uvlong pc)
{
	Instr i;

	if (mkinstr(map, &i, pc))
		return i.n;
	return -1;
}

static int
i386foll(Map *map, uvlong pc, Rgetter rget, uvlong *foll)
{
	Instr i;
	Optable *op;
	ushort s;
	uvlong l, addr;
	vlong v;
	int n;

	op = mkinstr(map, &i, pc);
	if (!op)
		return -1;

	n = 0;

	switch(i.jumptype) {
	case RET:		/* RETURN or LEAVE */
	case Iw:		/* RETURN */
		if (strcmp(op->proto, "LEAVE") == 0) {
			if (geta(map, (*rget)(map, "BP"), &l) < 0)
				return -1;
		} else if (geta(map, (*rget)(map, mach->sp), &l) < 0)
			return -1;
		foll[0] = l;
		return 1;
	case Iwds:		/* pc relative JUMP or CALL*/
	case Jbs:		/* pc relative JUMP or CALL */
		v = (long)i.imm;
		foll[0] = pc+v+i.n;
		n = 1;
		break;
	case PTR:		/* seg:displacement JUMP or CALL */
		foll[0] = (i.seg<<4)+i.disp;
		return 1;
	case JUMP:		/* JUMP or CALL EA */

		if(i.mod == 3) {
			foll[0] = (*rget)(map, reg[i.rex&REXB? i.base+8: i.base]);
			return 1;
		}
			/* calculate the effective address */
		addr = i.disp;
		if (i.base >= 0) {
			if (geta(map, (*rget)(map, reg[i.rex&REXB? i.base+8: i.base]), &l) < 0)
				return -1;
			addr += l;
		}
		if (i.index >= 0) {
			if (geta(map, (*rget)(map, reg[i.rex&REXX? i.index+8: i.index]), &l) < 0)
				return -1;
			addr += l*(1<<i.ss);
		}
			/* now retrieve a seg:disp value at that address */
		if (get2(map, addr, &s) < 0)			/* seg */
			return -1;
		foll[0] = s<<4;
		addr += 2;
		if (i.asize == 'L') {
			if (geta(map, addr, &l) < 0)		/* disp32 */
				return -1;
			foll[0] += l;
		} else {					/* disp16 */
			if (get2(map, addr, &s) < 0)
				return -1;
			foll[0] += s;
		}
		return 1;
	default:
		break;
	}		
	if (strncmp(op->proto,"JMP", 3) == 0 || strncmp(op->proto,"CALL", 4) == 0)
		return 1;
	foll[n++] = pc+i.n;
	return n;
}