1 /* disasm.c where all the _work_ gets done in the Netwide Disassembler
3 * The Netwide Assembler is copyright (C) 1996 Simon Tatham and
4 * Julian Hall. All rights reserved. The software is
5 * redistributable under the licence given in the file "Licence"
6 * distributed in the NASM archive.
8 * initial version 27/iii/95 by Simon Tatham
21 extern struct itemplate
**itable
[];
24 * Flags that go into the `segment' field of `insn' structures
27 #define SEG_RELATIVE 1
34 #define SEG_SIGNED 128
36 static int whichreg(long regflags
, int regval
) {
37 static int reg32
[] = {
38 R_EAX
, R_ECX
, R_EDX
, R_EBX
, R_ESP
, R_EBP
, R_ESI
, R_EDI
};
39 static int reg16
[] = {
40 R_AX
, R_CX
, R_DX
, R_BX
, R_SP
, R_BP
, R_SI
, R_DI
};
42 R_AL
, R_CL
, R_DL
, R_BL
, R_AH
, R_CH
, R_DH
, R_BH
};
44 R_ES
, R_CS
, R_SS
, R_DS
, R_FS
, R_GS
, 0, 0 };
46 R_CR0
, 0, R_CR2
, R_CR3
, R_CR4
, 0, 0, 0 };
48 R_DR0
, R_DR1
, R_DR2
, R_DR3
, 0, 0, R_DR6
, R_DR7
};
50 0, 0, 0, R_TR3
, R_TR4
, R_TR5
, R_TR6
, R_TR7
};
51 static int fpureg
[] = {
52 R_ST0
, R_ST1
, R_ST2
, R_ST3
, R_ST4
, R_ST5
, R_ST6
, R_ST7
};
53 static int mmxreg
[] = {
54 R_MM0
, R_MM1
, R_MM2
, R_MM3
, R_MM4
, R_MM5
, R_MM6
, R_MM7
};
56 if (!(REG_AL
& ~regflags
))
58 if (!(REG_AX
& ~regflags
))
60 if (!(REG_EAX
& ~regflags
))
62 if (!(REG_DX
& ~regflags
))
64 if (!(REG_CL
& ~regflags
))
66 if (!(REG_CX
& ~regflags
))
68 if (!(REG_ECX
& ~regflags
))
70 if (!(REG_CR4
& ~regflags
))
72 if (!(FPU0
& ~regflags
))
74 if (!((REGMEM
|BITS8
) & ~regflags
))
76 if (!((REGMEM
|BITS16
) & ~regflags
))
78 if (!((REGMEM
|BITS32
) & ~regflags
))
80 if (!(REG_SREG
& ~regflags
))
82 if (!(REG_CREG
& ~regflags
))
84 if (!(REG_DREG
& ~regflags
))
86 if (!(REG_TREG
& ~regflags
))
88 if (!(FPUREG
& ~regflags
))
89 return fpureg
[regval
];
90 if (!(MMXREG
& ~regflags
))
91 return mmxreg
[regval
];
95 static char *whichcond(int condval
) {
96 static int conds
[] = {
97 C_O
, C_NO
, C_B
, C_AE
, C_E
, C_NE
, C_BE
, C_A
,
98 C_S
, C_NS
, C_PE
, C_PO
, C_L
, C_GE
, C_LE
, C_G
100 return conditions
[conds
[condval
]];
104 * Process an effective address (ModRM) specification.
106 static unsigned char *do_ea (unsigned char *data
, int modrm
, int asize
,
107 int segsize
, operand
*op
) {
108 int mod
, rm
, scale
, index
, base
;
110 mod
= (modrm
>> 6) & 03;
113 if (mod
== 3) { /* pure register version */
115 op
->segment
|= SEG_RMREG
;
123 * <mod> specifies the displacement size (none, byte or
124 * word), and <rm> specifies the register combination.
125 * Exception: mod=0,rm=6 does not specify [BP] as one might
126 * expect, but instead specifies [disp16].
128 op
->indexreg
= op
->basereg
= -1;
129 op
->scale
= 1; /* always, in 16 bits */
131 case 0: op
->basereg
= R_BX
; op
->indexreg
= R_SI
; break;
132 case 1: op
->basereg
= R_BX
; op
->indexreg
= R_DI
; break;
133 case 2: op
->basereg
= R_BP
; op
->indexreg
= R_SI
; break;
134 case 3: op
->basereg
= R_BP
; op
->indexreg
= R_DI
; break;
135 case 4: op
->basereg
= R_SI
; break;
136 case 5: op
->basereg
= R_DI
; break;
137 case 6: op
->basereg
= R_BP
; break;
138 case 7: op
->basereg
= R_BX
; break;
140 if (rm
== 6 && mod
== 0) { /* special case */
144 mod
= 2; /* fake disp16 */
148 op
->segment
|= SEG_NODISP
;
151 op
->segment
|= SEG_DISP8
;
152 op
->offset
= (signed char) *data
++;
155 op
->segment
|= SEG_DISP16
;
156 op
->offset
= *data
++;
157 op
->offset
|= (*data
++) << 8;
163 * Once again, <mod> specifies displacement size (this time
164 * none, byte or *dword*), while <rm> specifies the base
165 * register. Again, [EBP] is missing, replaced by a pure
166 * disp32 (this time that's mod=0,rm=*5*). However, rm=4
167 * indicates not a single base register, but instead the
168 * presence of a SIB byte...
172 case 0: op
->basereg
= R_EAX
; break;
173 case 1: op
->basereg
= R_ECX
; break;
174 case 2: op
->basereg
= R_EDX
; break;
175 case 3: op
->basereg
= R_EBX
; break;
176 case 5: op
->basereg
= R_EBP
; break;
177 case 6: op
->basereg
= R_ESI
; break;
178 case 7: op
->basereg
= R_EDI
; break;
180 if (rm
== 5 && mod
== 0) {
184 mod
= 2; /* fake disp32 */
186 if (rm
== 4) { /* process SIB */
187 scale
= (*data
>> 6) & 03;
188 index
= (*data
>> 3) & 07;
192 op
->scale
= 1 << scale
;
194 case 0: op
->indexreg
= R_EAX
; break;
195 case 1: op
->indexreg
= R_ECX
; break;
196 case 2: op
->indexreg
= R_EDX
; break;
197 case 3: op
->indexreg
= R_EBX
; break;
198 case 4: op
->indexreg
= -1; break;
199 case 5: op
->indexreg
= R_EBP
; break;
200 case 6: op
->indexreg
= R_ESI
; break;
201 case 7: op
->indexreg
= R_EDI
; break;
205 case 0: op
->basereg
= R_EAX
; break;
206 case 1: op
->basereg
= R_ECX
; break;
207 case 2: op
->basereg
= R_EDX
; break;
208 case 3: op
->basereg
= R_EBX
; break;
209 case 4: op
->basereg
= R_ESP
; break;
210 case 6: op
->basereg
= R_ESI
; break;
211 case 7: op
->basereg
= R_EDI
; break;
223 op
->segment
|= SEG_NODISP
;
226 op
->segment
|= SEG_DISP8
;
227 op
->offset
= (signed char) *data
++;
230 op
->segment
|= SEG_DISP32
;
231 op
->offset
= *data
++;
232 op
->offset
|= (*data
++) << 8;
233 op
->offset
|= ((long) *data
++) << 16;
234 op
->offset
|= ((long) *data
++) << 24;
242 * Determine whether the code string in r corresponds to the data
243 * stream in data. Return the number of bytes matched if so.
245 static int matches (unsigned char *r
, unsigned char *data
, int asize
,
246 int osize
, int segsize
, insn
*ins
) {
247 unsigned char *origdata
= data
;
248 int a_used
= FALSE
, o_used
= FALSE
;
252 if (c
>= 01 && c
<= 03) {
259 case 0x07: ins
->oprs
[0].basereg
= 0; break;
260 case 0x17: ins
->oprs
[0].basereg
= 2; break;
261 case 0x1F: ins
->oprs
[0].basereg
= 3; break;
262 default: return FALSE
;
267 case 0xA1: ins
->oprs
[0].basereg
= 4; break;
268 case 0xA9: ins
->oprs
[0].basereg
= 5; break;
269 default: return FALSE
;
274 case 0x06: ins
->oprs
[0].basereg
= 0; break;
275 case 0x0E: ins
->oprs
[0].basereg
= 1; break;
276 case 0x16: ins
->oprs
[0].basereg
= 2; break;
277 case 0x1E: ins
->oprs
[0].basereg
= 3; break;
278 default: return FALSE
;
283 case 0xA0: ins
->oprs
[0].basereg
= 4; break;
284 case 0xA8: ins
->oprs
[0].basereg
= 5; break;
285 default: return FALSE
;
288 if (c
>= 010 && c
<= 012) {
289 int t
= *r
++, d
= *data
++;
290 if (d
< t
|| d
> t
+7)
293 ins
->oprs
[c
-010].basereg
= d
-t
;
294 ins
->oprs
[c
-010].segment
|= SEG_RMREG
;
300 if (c
>= 014 && c
<= 016) {
301 ins
->oprs
[c
-014].offset
= (signed char) *data
++;
302 ins
->oprs
[c
-014].segment
|= SEG_SIGNED
;
304 if (c
>= 020 && c
<= 022)
305 ins
->oprs
[c
-020].offset
= *data
++;
306 if (c
>= 024 && c
<= 026)
307 ins
->oprs
[c
-024].offset
= *data
++;
308 if (c
>= 030 && c
<= 032) {
309 ins
->oprs
[c
-030].offset
= *data
++;
310 ins
->oprs
[c
-030].offset
|= (*data
++ << 8);
312 if (c
>= 034 && c
<= 036) {
313 ins
->oprs
[c
-034].offset
= *data
++;
314 ins
->oprs
[c
-034].offset
|= (*data
++ << 8);
316 ins
->oprs
[c
-034].offset
|= (((long) *data
++) << 16);
317 ins
->oprs
[c
-034].offset
|= (((long) *data
++) << 24);
319 if (segsize
!= asize
)
320 ins
->oprs
[c
-034].addr_size
= asize
;
322 if (c
>= 040 && c
<= 042) {
323 ins
->oprs
[c
-040].offset
= *data
++;
324 ins
->oprs
[c
-040].offset
|= (*data
++ << 8);
325 ins
->oprs
[c
-040].offset
|= (((long) *data
++) << 16);
326 ins
->oprs
[c
-040].offset
|= (((long) *data
++) << 24);
328 if (c
>= 050 && c
<= 052) {
329 ins
->oprs
[c
-050].offset
= (signed char) *data
++;
330 ins
->oprs
[c
-050].segment
|= SEG_RELATIVE
;
332 if (c
>= 060 && c
<= 062) {
333 ins
->oprs
[c
-060].offset
= *data
++;
334 ins
->oprs
[c
-060].offset
|= (*data
++ << 8);
335 ins
->oprs
[c
-060].segment
|= SEG_RELATIVE
;
336 ins
->oprs
[c
-060].segment
&= ~SEG_32BIT
;
338 if (c
>= 064 && c
<= 066) {
339 ins
->oprs
[c
-064].offset
= *data
++;
340 ins
->oprs
[c
-064].offset
|= (*data
++ << 8);
342 ins
->oprs
[c
-064].offset
|= (((long) *data
++) << 16);
343 ins
->oprs
[c
-064].offset
|= (((long) *data
++) << 24);
344 ins
->oprs
[c
-064].segment
|= SEG_32BIT
;
346 ins
->oprs
[c
-064].segment
&= ~SEG_32BIT
;
347 ins
->oprs
[c
-064].segment
|= SEG_RELATIVE
;
348 if (segsize
!= asize
)
349 ins
->oprs
[c
-064].addr_size
= asize
;
351 if (c
>= 070 && c
<= 072) {
352 ins
->oprs
[c
-070].offset
= *data
++;
353 ins
->oprs
[c
-070].offset
|= (*data
++ << 8);
354 ins
->oprs
[c
-070].offset
|= (((long) *data
++) << 16);
355 ins
->oprs
[c
-070].offset
|= (((long) *data
++) << 24);
356 ins
->oprs
[c
-070].segment
|= SEG_32BIT
| SEG_RELATIVE
;
358 if (c
>= 0100 && c
<= 0177) {
360 ins
->oprs
[c
& 07].basereg
= (modrm
>> 3) & 07;
361 ins
->oprs
[c
& 07].segment
|= SEG_RMREG
;
362 data
= do_ea (data
, modrm
, asize
, segsize
,
363 &ins
->oprs
[(c
>> 3) & 07]);
365 if (c
>= 0200 && c
<= 0277) {
367 if (((modrm
>> 3) & 07) != (c
& 07))
368 return FALSE
; /* spare field doesn't match up */
369 data
= do_ea (data
, modrm
, asize
, segsize
,
370 &ins
->oprs
[(c
>> 3) & 07]);
372 if (c
>= 0300 && c
<= 0302) {
374 ins
->oprs
[c
-0300].segment
|= SEG_32BIT
;
376 ins
->oprs
[c
-0300].segment
&= ~SEG_32BIT
;
392 if (asize
!= segsize
)
410 if (osize
!= segsize
)
416 int t
= *r
++, d
= *data
++;
417 if (d
< t
|| d
> t
+15)
420 ins
->condition
= d
- t
;
425 * Check for unused a/o prefixes.
428 if (!a_used
&& asize
!= segsize
)
429 ins
->prefixes
[ins
->nprefix
++] = (asize
== 16 ? P_A16
: P_A32
);
430 if (!o_used
&& osize
!= segsize
)
431 ins
->prefixes
[ins
->nprefix
++] = (osize
== 16 ? P_O16
: P_O32
);
433 return data
- origdata
;
436 long disasm (unsigned char *data
, char *output
, int segsize
, long offset
,
438 struct itemplate
**p
;
441 int rep
, lock
, asize
, osize
, i
, slen
, colon
;
442 unsigned char *origdata
;
449 asize
= osize
= segsize
;
454 if (*data
== 0xF3 || *data
== 0xF2)
456 else if (*data
== 0xF0)
458 else if (*data
== 0x2E || *data
== 0x36 || *data
== 0x3E ||
459 *data
== 0x26 || *data
== 0x64 || *data
== 0x65) {
461 case 0x2E: segover
= "cs"; break;
462 case 0x36: segover
= "ss"; break;
463 case 0x3E: segover
= "ds"; break;
464 case 0x26: segover
= "es"; break;
465 case 0x64: segover
= "fs"; break;
466 case 0x65: segover
= "gs"; break;
468 } else if (*data
== 0x66)
469 osize
= 48 - segsize
, data
++;
470 else if (*data
== 0x67)
471 asize
= 48 - segsize
, data
++;
476 ins
.oprs
[0].segment
= ins
.oprs
[1].segment
= ins
.oprs
[2].segment
=
477 ins
.oprs
[0].addr_size
= ins
.oprs
[1].addr_size
= ins
.oprs
[2].addr_size
=
478 (segsize
== 16 ? 0 : SEG_32BIT
);
481 for (p
= itable
[*data
]; *p
; p
++)
482 if ( (length
= matches((unsigned char *)((*p
)->code
), data
,
483 asize
, osize
, segsize
, &ins
)) ) {
486 * Final check to make sure the types of r/m match up.
488 for (i
= 0; i
< (*p
)->operands
; i
++)
489 if (((ins
.oprs
[i
].segment
& SEG_RMREG
) &&
490 !(MEMORY
& ~(*p
)->opd
[i
])) ||
491 (!(ins
.oprs
[i
].segment
& SEG_RMREG
) &&
492 !(REGNORM
& ~(*p
)->opd
[i
]) &&
493 !((*p
)->opd
[i
] & REG_SMASK
)))
498 if (!length
|| !works
)
499 return 0; /* no instruction was matched */
504 slen
+= sprintf(output
+slen
, "rep%s ",
505 (rep
== 0xF2 ? "ne" :
506 (*p
)->opcode
== I_CMPSB
||
507 (*p
)->opcode
== I_CMPSW
||
508 (*p
)->opcode
== I_CMPSD
||
509 (*p
)->opcode
== I_SCASB
||
510 (*p
)->opcode
== I_SCASW
||
511 (*p
)->opcode
== I_SCASD
? "e" : ""));
514 slen
+= sprintf(output
+slen
, "lock ");
515 for (i
= 0; i
< ins
.nprefix
; i
++)
516 switch (ins
.prefixes
[i
]) {
517 case P_A16
: slen
+= sprintf(output
+slen
, "a16 "); break;
518 case P_A32
: slen
+= sprintf(output
+slen
, "a32 "); break;
519 case P_O16
: slen
+= sprintf(output
+slen
, "o16 "); break;
520 case P_O32
: slen
+= sprintf(output
+slen
, "o32 "); break;
523 for (i
= 0; i
< elements(ico
); i
++)
524 if ((*p
)->opcode
== ico
[i
]) {
525 slen
+= sprintf(output
+slen
, "%s%s", icn
[i
],
526 whichcond(ins
.condition
));
529 if (i
>= elements(ico
))
530 slen
+= sprintf(output
+slen
, "%s", insn_names
[(*p
)->opcode
]);
532 length
+= data
- origdata
; /* fix up for prefixes */
533 for (i
=0; i
<(*p
)->operands
; i
++) {
534 output
[slen
++] = (colon
? ':' : i
==0 ? ' ' : ',');
536 if (ins
.oprs
[i
].segment
& SEG_RELATIVE
) {
537 ins
.oprs
[i
].offset
+= offset
+ length
;
539 * sort out wraparound
541 if (!(ins
.oprs
[i
].segment
& SEG_32BIT
))
542 ins
.oprs
[i
].offset
&= 0xFFFF;
544 * add sync marker, if autosync is on
547 add_sync (ins
.oprs
[i
].offset
, 0L);
550 if ((*p
)->opd
[i
] & COLON
)
555 if (((*p
)->opd
[i
] & (REGISTER
| FPUREG
)) ||
556 (ins
.oprs
[i
].segment
& SEG_RMREG
)) {
557 ins
.oprs
[i
].basereg
= whichreg ((*p
)->opd
[i
],
558 ins
.oprs
[i
].basereg
);
559 slen
+= sprintf(output
+slen
, "%s",
560 reg_names
[ins
.oprs
[i
].basereg
]);
561 } else if (!(UNITY
& ~(*p
)->opd
[i
])) {
562 output
[slen
++] = '1';
563 } else if ( (*p
)->opd
[i
] & IMMEDIATE
) {
564 if ( (*p
)->opd
[i
] & BITS8
) {
565 slen
+= sprintf(output
+slen
, "byte ");
566 if (ins
.oprs
[i
].segment
& SEG_SIGNED
) {
567 if (ins
.oprs
[i
].offset
< 0) {
568 ins
.oprs
[i
].offset
*= -1;
569 output
[slen
++] = '-';
571 output
[slen
++] = '+';
573 } else if ( (*p
)->opd
[i
] & BITS16
) {
574 slen
+= sprintf(output
+slen
, "word ");
575 } else if ( (*p
)->opd
[i
] & BITS32
) {
576 slen
+= sprintf(output
+slen
, "dword ");
577 } else if ( (*p
)->opd
[i
] & NEAR
) {
578 slen
+= sprintf(output
+slen
, "near ");
579 } else if ( (*p
)->opd
[i
] & SHORT
) {
580 slen
+= sprintf(output
+slen
, "short ");
582 slen
+= sprintf(output
+slen
, "0x%lx", ins
.oprs
[i
].offset
);
583 } else if ( !(MEM_OFFS
& ~(*p
)->opd
[i
]) ) {
584 slen
+= sprintf(output
+slen
, "[%s%s%s0x%lx]",
585 (segover
? segover
: ""),
586 (segover
? ":" : ""),
587 (ins
.oprs
[i
].addr_size
== 32 ? "dword " :
588 ins
.oprs
[i
].addr_size
== 16 ? "word " : ""),
591 } else if ( !(REGMEM
& ~(*p
)->opd
[i
]) ) {
593 if ( (*p
)->opd
[i
] & BITS8
)
594 slen
+= sprintf(output
+slen
, "byte ");
595 if ( (*p
)->opd
[i
] & BITS16
)
596 slen
+= sprintf(output
+slen
, "word ");
597 if ( (*p
)->opd
[i
] & BITS32
)
598 slen
+= sprintf(output
+slen
, "dword ");
599 if ( (*p
)->opd
[i
] & BITS64
)
600 slen
+= sprintf(output
+slen
, "qword ");
601 if ( (*p
)->opd
[i
] & BITS80
)
602 slen
+= sprintf(output
+slen
, "tword ");
603 if ( (*p
)->opd
[i
] & FAR
)
604 slen
+= sprintf(output
+slen
, "far ");
605 if ( (*p
)->opd
[i
] & NEAR
)
606 slen
+= sprintf(output
+slen
, "near ");
607 output
[slen
++] = '[';
608 if (ins
.oprs
[i
].addr_size
)
609 slen
+= sprintf(output
+slen
, "%s",
610 (ins
.oprs
[i
].addr_size
== 32 ? "dword " :
611 ins
.oprs
[i
].addr_size
== 16 ? "word " : ""));
613 slen
+= sprintf(output
+slen
, "%s:", segover
);
616 if (ins
.oprs
[i
].basereg
!= -1) {
617 slen
+= sprintf(output
+slen
, "%s",
618 reg_names
[ins
.oprs
[i
].basereg
]);
621 if (ins
.oprs
[i
].indexreg
!= -1) {
623 output
[slen
++] = '+';
624 slen
+= sprintf(output
+slen
, "%s",
625 reg_names
[ins
.oprs
[i
].indexreg
]);
626 if (ins
.oprs
[i
].scale
> 1)
627 slen
+= sprintf(output
+slen
, "*%d", ins
.oprs
[i
].scale
);
630 if (ins
.oprs
[i
].segment
& SEG_DISP8
) {
632 if (ins
.oprs
[i
].offset
& 0x80) {
633 ins
.oprs
[i
].offset
= - (signed char) ins
.oprs
[i
].offset
;
636 slen
+= sprintf(output
+slen
, "%c0x%lx", sign
,
638 } else if (ins
.oprs
[i
].segment
& SEG_DISP16
) {
640 output
[slen
++] = '+';
641 slen
+= sprintf(output
+slen
, "0x%lx", ins
.oprs
[i
].offset
);
642 } else if (ins
.oprs
[i
].segment
& SEG_DISP32
) {
644 output
[slen
++] = '+';
645 slen
+= sprintf(output
+slen
, "0x%lx", ins
.oprs
[i
].offset
);
647 output
[slen
++] = ']';
649 slen
+= sprintf(output
+slen
, "<operand%d>", i
);
653 if (segover
) { /* unused segment override */
657 p
[count
+3] = p
[count
];
658 strncpy (output
, segover
, 2);
664 long eatbyte (unsigned char *data
, char *output
) {
665 sprintf(output
, "db 0x%02X", *data
);