2 * x86 instruction analysis
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
18 * Copyright (C) IBM Corporation, 2002, 2004, 2009
22 #include <linux/string.h>
29 /* Verify next sizeof(t) bytes can be on the same instruction */
30 #define validate_next(t, insn, n) \
31 ((insn)->next_byte + sizeof(t) + n < (insn)->end_kaddr)
33 #define __get_next(t, insn) \
34 ({ t r = *(t*)insn->next_byte; insn->next_byte += sizeof(t); r; })
36 #define __peek_nbyte_next(t, insn, n) \
37 ({ t r = *(t*)((insn)->next_byte + n); r; })
39 #define get_next(t, insn) \
40 ({ if (unlikely(!validate_next(t, insn, 0))) goto err_out; __get_next(t, insn); })
42 #define peek_nbyte_next(t, insn, n) \
43 ({ if (unlikely(!validate_next(t, insn, n))) goto err_out; __peek_nbyte_next(t, insn, n); })
45 #define peek_next(t, insn) peek_nbyte_next(t, insn, 0)
48 * insn_init() - initialize struct insn
49 * @insn: &struct insn to be initialized
50 * @kaddr: address (in kernel memory) of instruction (or copy thereof)
51 * @x86_64: !0 for 64-bit kernel or 64-bit app
53 void insn_init(struct insn
*insn
, const void *kaddr
, int buf_len
, int x86_64
)
55 memset(insn
, 0, sizeof(*insn
));
57 insn
->end_kaddr
= kaddr
+ buf_len
;
58 insn
->next_byte
= kaddr
;
59 insn
->x86_64
= x86_64
? 1 : 0;
68 * insn_get_prefixes - scan x86 instruction prefix bytes
69 * @insn: &struct insn containing instruction
71 * Populates the @insn->prefixes bitmap, and updates @insn->next_byte
72 * to point to the (first) opcode. No effect if @insn->prefixes.got
75 void insn_get_prefixes(struct insn
*insn
)
77 struct insn_field
*prefixes
= &insn
->prefixes
;
87 b
= peek_next(insn_byte_t
, insn
);
88 attr
= inat_get_opcode_attribute(b
);
89 while (inat_is_legacy_prefix(attr
)) {
90 /* Skip if same prefix */
91 for (i
= 0; i
< nb
; i
++)
92 if (prefixes
->bytes
[i
] == b
)
95 /* Invalid instruction */
97 prefixes
->bytes
[nb
++] = b
;
98 if (inat_is_address_size_prefix(attr
)) {
99 /* address size switches 2/4 or 4/8 */
101 insn
->addr_bytes
^= 12;
103 insn
->addr_bytes
^= 6;
104 } else if (inat_is_operand_size_prefix(attr
)) {
105 /* oprand size switches 2/4 */
106 insn
->opnd_bytes
^= 6;
112 b
= peek_next(insn_byte_t
, insn
);
113 attr
= inat_get_opcode_attribute(b
);
115 /* Set the last prefix */
116 if (lb
&& lb
!= insn
->prefixes
.bytes
[3]) {
117 if (unlikely(insn
->prefixes
.bytes
[3])) {
118 /* Swap the last prefix */
119 b
= insn
->prefixes
.bytes
[3];
120 for (i
= 0; i
< nb
; i
++)
121 if (prefixes
->bytes
[i
] == lb
)
122 prefixes
->bytes
[i
] = b
;
124 insn
->prefixes
.bytes
[3] = lb
;
127 /* Decode REX prefix */
129 b
= peek_next(insn_byte_t
, insn
);
130 attr
= inat_get_opcode_attribute(b
);
131 if (inat_is_rex_prefix(attr
)) {
132 insn
->rex_prefix
.value
= b
;
133 insn
->rex_prefix
.nbytes
= 1;
136 /* REX.W overrides opnd_size */
137 insn
->opnd_bytes
= 8;
140 insn
->rex_prefix
.got
= 1;
142 /* Decode VEX prefix */
143 b
= peek_next(insn_byte_t
, insn
);
144 attr
= inat_get_opcode_attribute(b
);
145 if (inat_is_vex_prefix(attr
)) {
146 insn_byte_t b2
= peek_nbyte_next(insn_byte_t
, insn
, 1);
149 * In 32-bits mode, if the [7:6] bits (mod bits of
150 * ModRM) on the second byte are not 11b, it is
153 if (X86_MODRM_MOD(b2
) != 3)
156 insn
->vex_prefix
.bytes
[0] = b
;
157 insn
->vex_prefix
.bytes
[1] = b2
;
158 if (inat_is_vex3_prefix(attr
)) {
159 b2
= peek_nbyte_next(insn_byte_t
, insn
, 2);
160 insn
->vex_prefix
.bytes
[2] = b2
;
161 insn
->vex_prefix
.nbytes
= 3;
162 insn
->next_byte
+= 3;
163 if (insn
->x86_64
&& X86_VEX_W(b2
))
164 /* VEX.W overrides opnd_size */
165 insn
->opnd_bytes
= 8;
167 insn
->vex_prefix
.nbytes
= 2;
168 insn
->next_byte
+= 2;
172 insn
->vex_prefix
.got
= 1;
181 * insn_get_opcode - collect opcode(s)
182 * @insn: &struct insn containing instruction
184 * Populates @insn->opcode, updates @insn->next_byte to point past the
185 * opcode byte(s), and set @insn->attr (except for groups).
186 * If necessary, first collects any preceding (prefix) bytes.
187 * Sets @insn->opcode.value = opcode1. No effect if @insn->opcode.got
190 void insn_get_opcode(struct insn
*insn
)
192 struct insn_field
*opcode
= &insn
->opcode
;
197 if (!insn
->prefixes
.got
)
198 insn_get_prefixes(insn
);
200 /* Get first opcode */
201 op
= get_next(insn_byte_t
, insn
);
202 opcode
->bytes
[0] = op
;
205 /* Check if there is VEX prefix or not */
206 if (insn_is_avx(insn
)) {
208 m
= insn_vex_m_bits(insn
);
209 p
= insn_vex_p_bits(insn
);
210 insn
->attr
= inat_get_avx_attribute(op
, m
, p
);
211 if (!inat_accept_vex(insn
->attr
) && !inat_is_group(insn
->attr
))
212 insn
->attr
= 0; /* This instruction is bad */
213 goto end
; /* VEX has only 1 byte for opcode */
216 insn
->attr
= inat_get_opcode_attribute(op
);
217 while (inat_is_escape(insn
->attr
)) {
218 /* Get escaped opcode */
219 op
= get_next(insn_byte_t
, insn
);
220 opcode
->bytes
[opcode
->nbytes
++] = op
;
221 pfx_id
= insn_last_prefix_id(insn
);
222 insn
->attr
= inat_get_escape_attribute(op
, pfx_id
, insn
->attr
);
224 if (inat_must_vex(insn
->attr
))
225 insn
->attr
= 0; /* This instruction is bad */
234 * insn_get_modrm - collect ModRM byte, if any
235 * @insn: &struct insn containing instruction
237 * Populates @insn->modrm and updates @insn->next_byte to point past the
238 * ModRM byte, if any. If necessary, first collects the preceding bytes
239 * (prefixes and opcode(s)). No effect if @insn->modrm.got is already 1.
241 void insn_get_modrm(struct insn
*insn
)
243 struct insn_field
*modrm
= &insn
->modrm
;
244 insn_byte_t pfx_id
, mod
;
247 if (!insn
->opcode
.got
)
248 insn_get_opcode(insn
);
250 if (inat_has_modrm(insn
->attr
)) {
251 mod
= get_next(insn_byte_t
, insn
);
254 if (inat_is_group(insn
->attr
)) {
255 pfx_id
= insn_last_prefix_id(insn
);
256 insn
->attr
= inat_get_group_attribute(mod
, pfx_id
,
258 if (insn_is_avx(insn
) && !inat_accept_vex(insn
->attr
))
259 insn
->attr
= 0; /* This is bad */
263 if (insn
->x86_64
&& inat_is_force64(insn
->attr
))
264 insn
->opnd_bytes
= 8;
273 * insn_rip_relative() - Does instruction use RIP-relative addressing mode?
274 * @insn: &struct insn containing instruction
276 * If necessary, first collects the instruction up to and including the
277 * ModRM byte. No effect if @insn->x86_64 is 0.
279 int insn_rip_relative(struct insn
*insn
)
281 struct insn_field
*modrm
= &insn
->modrm
;
286 insn_get_modrm(insn
);
288 * For rip-relative instructions, the mod field (top 2 bits)
289 * is zero and the r/m field (bottom 3 bits) is 0x5.
291 return (modrm
->nbytes
&& (modrm
->value
& 0xc7) == 0x5);
295 * insn_get_sib() - Get the SIB byte of instruction
296 * @insn: &struct insn containing instruction
298 * If necessary, first collects the instruction up to and including the
301 void insn_get_sib(struct insn
*insn
)
307 if (!insn
->modrm
.got
)
308 insn_get_modrm(insn
);
309 if (insn
->modrm
.nbytes
) {
310 modrm
= (insn_byte_t
)insn
->modrm
.value
;
311 if (insn
->addr_bytes
!= 2 &&
312 X86_MODRM_MOD(modrm
) != 3 && X86_MODRM_RM(modrm
) == 4) {
313 insn
->sib
.value
= get_next(insn_byte_t
, insn
);
314 insn
->sib
.nbytes
= 1;
325 * insn_get_displacement() - Get the displacement of instruction
326 * @insn: &struct insn containing instruction
328 * If necessary, first collects the instruction up to and including the
330 * Displacement value is sign-expanded.
332 void insn_get_displacement(struct insn
*insn
)
334 insn_byte_t mod
, rm
, base
;
336 if (insn
->displacement
.got
)
340 if (insn
->modrm
.nbytes
) {
342 * Interpreting the modrm byte:
343 * mod = 00 - no displacement fields (exceptions below)
344 * mod = 01 - 1-byte displacement field
345 * mod = 10 - displacement field is 4 bytes, or 2 bytes if
346 * address size = 2 (0x67 prefix in 32-bit mode)
347 * mod = 11 - no memory operand
349 * If address size = 2...
350 * mod = 00, r/m = 110 - displacement field is 2 bytes
352 * If address size != 2...
353 * mod != 11, r/m = 100 - SIB byte exists
354 * mod = 00, SIB base = 101 - displacement field is 4 bytes
355 * mod = 00, r/m = 101 - rip-relative addressing, displacement
358 mod
= X86_MODRM_MOD(insn
->modrm
.value
);
359 rm
= X86_MODRM_RM(insn
->modrm
.value
);
360 base
= X86_SIB_BASE(insn
->sib
.value
);
364 insn
->displacement
.value
= get_next(char, insn
);
365 insn
->displacement
.nbytes
= 1;
366 } else if (insn
->addr_bytes
== 2) {
367 if ((mod
== 0 && rm
== 6) || mod
== 2) {
368 insn
->displacement
.value
=
369 get_next(short, insn
);
370 insn
->displacement
.nbytes
= 2;
373 if ((mod
== 0 && rm
== 5) || mod
== 2 ||
374 (mod
== 0 && base
== 5)) {
375 insn
->displacement
.value
= get_next(int, insn
);
376 insn
->displacement
.nbytes
= 4;
381 insn
->displacement
.got
= 1;
387 /* Decode moffset16/32/64. Return 0 if failed */
388 static int __get_moffset(struct insn
*insn
)
390 switch (insn
->addr_bytes
) {
392 insn
->moffset1
.value
= get_next(short, insn
);
393 insn
->moffset1
.nbytes
= 2;
396 insn
->moffset1
.value
= get_next(int, insn
);
397 insn
->moffset1
.nbytes
= 4;
400 insn
->moffset1
.value
= get_next(int, insn
);
401 insn
->moffset1
.nbytes
= 4;
402 insn
->moffset2
.value
= get_next(int, insn
);
403 insn
->moffset2
.nbytes
= 4;
405 default: /* opnd_bytes must be modified manually */
408 insn
->moffset1
.got
= insn
->moffset2
.got
= 1;
416 /* Decode imm v32(Iz). Return 0 if failed */
417 static int __get_immv32(struct insn
*insn
)
419 switch (insn
->opnd_bytes
) {
421 insn
->immediate
.value
= get_next(short, insn
);
422 insn
->immediate
.nbytes
= 2;
426 insn
->immediate
.value
= get_next(int, insn
);
427 insn
->immediate
.nbytes
= 4;
429 default: /* opnd_bytes must be modified manually */
439 /* Decode imm v64(Iv/Ov), Return 0 if failed */
440 static int __get_immv(struct insn
*insn
)
442 switch (insn
->opnd_bytes
) {
444 insn
->immediate1
.value
= get_next(short, insn
);
445 insn
->immediate1
.nbytes
= 2;
448 insn
->immediate1
.value
= get_next(int, insn
);
449 insn
->immediate1
.nbytes
= 4;
452 insn
->immediate1
.value
= get_next(int, insn
);
453 insn
->immediate1
.nbytes
= 4;
454 insn
->immediate2
.value
= get_next(int, insn
);
455 insn
->immediate2
.nbytes
= 4;
457 default: /* opnd_bytes must be modified manually */
460 insn
->immediate1
.got
= insn
->immediate2
.got
= 1;
467 /* Decode ptr16:16/32(Ap) */
468 static int __get_immptr(struct insn
*insn
)
470 switch (insn
->opnd_bytes
) {
472 insn
->immediate1
.value
= get_next(short, insn
);
473 insn
->immediate1
.nbytes
= 2;
476 insn
->immediate1
.value
= get_next(int, insn
);
477 insn
->immediate1
.nbytes
= 4;
480 /* ptr16:64 is not exist (no segment) */
482 default: /* opnd_bytes must be modified manually */
485 insn
->immediate2
.value
= get_next(unsigned short, insn
);
486 insn
->immediate2
.nbytes
= 2;
487 insn
->immediate1
.got
= insn
->immediate2
.got
= 1;
495 * insn_get_immediate() - Get the immediates of instruction
496 * @insn: &struct insn containing instruction
498 * If necessary, first collects the instruction up to and including the
499 * displacement bytes.
500 * Basically, most of immediates are sign-expanded. Unsigned-value can be
501 * get by bit masking with ((1 << (nbytes * 8)) - 1)
503 void insn_get_immediate(struct insn
*insn
)
505 if (insn
->immediate
.got
)
507 if (!insn
->displacement
.got
)
508 insn_get_displacement(insn
);
510 if (inat_has_moffset(insn
->attr
)) {
511 if (!__get_moffset(insn
))
516 if (!inat_has_immediate(insn
->attr
))
520 switch (inat_immediate_size(insn
->attr
)) {
522 insn
->immediate
.value
= get_next(char, insn
);
523 insn
->immediate
.nbytes
= 1;
526 insn
->immediate
.value
= get_next(short, insn
);
527 insn
->immediate
.nbytes
= 2;
530 insn
->immediate
.value
= get_next(int, insn
);
531 insn
->immediate
.nbytes
= 4;
534 insn
->immediate1
.value
= get_next(int, insn
);
535 insn
->immediate1
.nbytes
= 4;
536 insn
->immediate2
.value
= get_next(int, insn
);
537 insn
->immediate2
.nbytes
= 4;
540 if (!__get_immptr(insn
))
543 case INAT_IMM_VWORD32
:
544 if (!__get_immv32(insn
))
548 if (!__get_immv(insn
))
552 /* Here, insn must have an immediate, but failed */
555 if (inat_has_second_immediate(insn
->attr
)) {
556 insn
->immediate2
.value
= get_next(char, insn
);
557 insn
->immediate2
.nbytes
= 1;
560 insn
->immediate
.got
= 1;
567 * insn_get_length() - Get the length of instruction
568 * @insn: &struct insn containing instruction
570 * If necessary, first collects the instruction up to and including the
573 void insn_get_length(struct insn
*insn
)
577 if (!insn
->immediate
.got
)
578 insn_get_immediate(insn
);
579 insn
->length
= (unsigned char)((unsigned long)insn
->next_byte
580 - (unsigned long)insn
->kaddr
);