📄 disasm.c
字号:
char *segover;
int i, slen, colon, n;
uint8_t *origdata;
int works;
insn tmp_ins, ins;
uint32_t goodness, best;
int best_pref;
struct prefix_info prefix;
bool end_prefix;
memset(&ins, 0, sizeof ins);
/*
* Scan for prefixes.
*/
memset(&prefix, 0, sizeof prefix);
prefix.asize = segsize;
prefix.osize = (segsize == 64) ? 32 : segsize;
segover = NULL;
origdata = data;
ix = itable;
end_prefix = false;
while (!end_prefix) {
switch (*data) {
case 0xF2:
case 0xF3:
prefix.rep = *data++;
break;
case 0xF0:
prefix.lock = *data++;
break;
case 0x2E:
segover = "cs", prefix.seg = *data++;
break;
case 0x36:
segover = "ss", prefix.seg = *data++;
break;
case 0x3E:
segover = "ds", prefix.seg = *data++;
break;
case 0x26:
segover = "es", prefix.seg = *data++;
break;
case 0x64:
segover = "fs", prefix.seg = *data++;
break;
case 0x65:
segover = "gs", prefix.seg = *data++;
break;
case 0x66:
prefix.osize = (segsize == 16) ? 32 : 16;
prefix.osp = *data++;
break;
case 0x67:
prefix.asize = (segsize == 32) ? 16 : 32;
prefix.asp = *data++;
break;
case 0xC4:
case 0xC5:
if (segsize == 64 || (data[1] & 0xc0) == 0xc0) {
prefix.vex[0] = *data++;
prefix.vex[1] = *data++;
prefix.rex = REX_V;
if (prefix.vex[0] == 0xc4) {
prefix.vex[2] = *data++;
prefix.rex |= (~prefix.vex[1] >> 5) & 7; /* REX_RXB */
prefix.rex |= (prefix.vex[2] >> (7-3)) & REX_W;
prefix.vex_m = prefix.vex[1] & 0x1f;
prefix.vex_v = (~prefix.vex[2] >> 3) & 15;
prefix.vex_lp = prefix.vex[2] & 7;
} else {
prefix.rex |= (~prefix.vex[1] >> (7-2)) & REX_R;
prefix.vex_m = 1;
prefix.vex_v = (~prefix.vex[1] >> 3) & 15;
prefix.vex_lp = prefix.vex[1] & 7;
}
ix = itable_VEX[prefix.vex_m][prefix.vex_lp];
}
end_prefix = true;
break;
case REX_P + 0x0:
case REX_P + 0x1:
case REX_P + 0x2:
case REX_P + 0x3:
case REX_P + 0x4:
case REX_P + 0x5:
case REX_P + 0x6:
case REX_P + 0x7:
case REX_P + 0x8:
case REX_P + 0x9:
case REX_P + 0xA:
case REX_P + 0xB:
case REX_P + 0xC:
case REX_P + 0xD:
case REX_P + 0xE:
case REX_P + 0xF:
if (segsize == 64) {
prefix.rex = *data++;
if (prefix.rex & REX_W)
prefix.osize = 64;
}
end_prefix = true;
break;
default:
end_prefix = true;
break;
}
}
best = -1; /* Worst possible */
best_p = NULL;
best_pref = INT_MAX;
if (!ix)
return 0; /* No instruction table at all... */
dp = data;
ix += *dp++;
while (ix->n == -1) {
ix = (const struct disasm_index *)ix->p + *dp++;
}
p = (const struct itemplate * const *)ix->p;
for (n = ix->n; n; n--, p++) {
if ((length = matches(*p, data, &prefix, segsize, &tmp_ins))) {
works = true;
/*
* Final check to make sure the types of r/m match up.
* XXX: Need to make sure this is actually correct.
*/
for (i = 0; i < (*p)->operands; i++) {
if (!((*p)->opd[i] & SAME_AS) &&
(
/* If it's a mem-only EA but we have a
register, die. */
((tmp_ins.oprs[i].segment & SEG_RMREG) &&
!(MEMORY & ~(*p)->opd[i])) ||
/* If it's a reg-only EA but we have a memory
ref, die. */
(!(tmp_ins.oprs[i].segment & SEG_RMREG) &&
!(REG_EA & ~(*p)->opd[i]) &&
!((*p)->opd[i] & REG_SMASK)) ||
/* Register type mismatch (eg FS vs REG_DESS):
die. */
((((*p)->opd[i] & (REGISTER | FPUREG)) ||
(tmp_ins.oprs[i].segment & SEG_RMREG)) &&
!whichreg((*p)->opd[i],
tmp_ins.oprs[i].basereg, tmp_ins.rex))
)) {
works = false;
break;
}
}
/*
* Note: we always prefer instructions which incorporate
* prefixes in the instructions themselves. This is to allow
* e.g. PAUSE to be preferred to REP NOP, and deal with
* MMX/SSE instructions where prefixes are used to select
* between MMX and SSE register sets or outright opcode
* selection.
*/
if (works) {
int i, nprefix;
goodness = ((*p)->flags & IF_PFMASK) ^ prefer;
nprefix = 0;
for (i = 0; i < MAXPREFIX; i++)
if (tmp_ins.prefixes[i])
nprefix++;
if (nprefix < best_pref ||
(nprefix == best_pref && goodness < best)) {
/* This is the best one found so far */
best = goodness;
best_p = p;
best_pref = nprefix;
best_length = length;
ins = tmp_ins;
}
}
}
}
if (!best_p)
return 0; /* no instruction was matched */
/* Pick the best match */
p = best_p;
length = best_length;
slen = 0;
/* TODO: snprintf returns the value that the string would have if
* the buffer were long enough, and not the actual length of
* the returned string, so each instance of using the return
* value of snprintf should actually be checked to assure that
* the return value is "sane." Maybe a macro wrapper could
* be used for that purpose.
*/
for (i = 0; i < MAXPREFIX; i++)
switch (ins.prefixes[i]) {
case P_LOCK:
slen += snprintf(output + slen, outbufsize - slen, "lock ");
break;
case P_REP:
slen += snprintf(output + slen, outbufsize - slen, "rep ");
break;
case P_REPE:
slen += snprintf(output + slen, outbufsize - slen, "repe ");
break;
case P_REPNE:
slen += snprintf(output + slen, outbufsize - slen, "repne ");
break;
case P_A16:
slen += snprintf(output + slen, outbufsize - slen, "a16 ");
break;
case P_A32:
slen += snprintf(output + slen, outbufsize - slen, "a32 ");
break;
case P_A64:
slen += snprintf(output + slen, outbufsize - slen, "a64 ");
break;
case P_O16:
slen += snprintf(output + slen, outbufsize - slen, "o16 ");
break;
case P_O32:
slen += snprintf(output + slen, outbufsize - slen, "o32 ");
break;
case P_O64:
slen += snprintf(output + slen, outbufsize - slen, "o64 ");
break;
default:
break;
}
i = (*p)->opcode;
if (i >= FIRST_COND_OPCODE)
slen += snprintf(output + slen, outbufsize - slen, "%s%s",
nasm_insn_names[i], condition_name[ins.condition]);
else
slen += snprintf(output + slen, outbufsize - slen, "%s",
nasm_insn_names[i]);
colon = false;
length += data - origdata; /* fix up for prefixes */
for (i = 0; i < (*p)->operands; i++) {
opflags_t t = (*p)->opd[i];
const operand *o = &ins.oprs[i];
int64_t offs;
if (t & SAME_AS) {
o = &ins.oprs[t & ~SAME_AS];
t = (*p)->opd[t & ~SAME_AS];
}
output[slen++] = (colon ? ':' : i == 0 ? ' ' : ',');
offs = o->offset;
if (o->segment & SEG_RELATIVE) {
offs += offset + length;
/*
* sort out wraparound
*/
if (!(o->segment & (SEG_32BIT|SEG_64BIT)))
offs &= 0xffff;
else if (segsize != 64)
offs &= 0xffffffff;
/*
* add sync marker, if autosync is on
*/
if (autosync)
add_sync(offs, 0L);
}
if (t & COLON)
colon = true;
else
colon = false;
if ((t & (REGISTER | FPUREG)) ||
(o->segment & SEG_RMREG)) {
enum reg_enum reg;
reg = whichreg(t, o->basereg, ins.rex);
if (t & TO)
slen += snprintf(output + slen, outbufsize - slen, "to ");
slen += snprintf(output + slen, outbufsize - slen, "%s",
nasm_reg_names[reg-EXPR_REG_START]);
} else if (!(UNITY & ~t)) {
output[slen++] = '1';
} else if (t & IMMEDIATE) {
if (t & BITS8) {
slen +=
snprintf(output + slen, outbufsize - slen, "byte ");
if (o->segment & SEG_SIGNED) {
if (offs < 0) {
offs *= -1;
output[slen++] = '-';
} else
output[slen++] = '+';
}
} else if (t & BITS16) {
slen +=
snprintf(output + slen, outbufsize - slen, "word ");
} else if (t & BITS32) {
slen +=
snprintf(output + slen, outbufsize - slen, "dword ");
} else if (t & BITS64) {
slen +=
snprintf(output + slen, outbufsize - slen, "qword ");
} else if (t & NEAR) {
slen +=
snprintf(output + slen, outbufsize - slen, "near ");
} else if (t & SHORT) {
slen +=
snprintf(output + slen, outbufsize - slen, "short ");
}
slen +=
snprintf(output + slen, outbufsize - slen, "0x%"PRIx64"",
offs);
} else if (!(MEM_OFFS & ~t)) {
slen +=
snprintf(output + slen, outbufsize - slen,
"[%s%s%s0x%"PRIx64"]",
(segover ? segover : ""),
(segover ? ":" : ""),
(o->disp_size == 64 ? "qword " :
o->disp_size == 32 ? "dword " :
o->disp_size == 16 ? "word " : ""), offs);
segover = NULL;
} else if (!(REGMEM & ~t)) {
int started = false;
if (t & BITS8)
slen +=
snprintf(output + slen, outbufsize - slen, "byte ");
if (t & BITS16)
slen +=
snprintf(output + slen, outbufsize - slen, "word ");
if (t & BITS32)
slen +=
snprintf(output + slen, outbufsize - slen, "dword ");
if (t & BITS64)
slen +=
snprintf(output + slen, outbufsize - slen, "qword ");
if (t & BITS80)
slen +=
snprintf(output + slen, outbufsize - slen, "tword ");
if (t & BITS128)
slen +=
snprintf(output + slen, outbufsize - slen, "oword ");
if (t & BITS256)
slen +=
snprintf(output + slen, outbufsize - slen, "yword ");
if (t & FAR)
slen += snprintf(output + slen, outbufsize - slen, "far ");
if (t & NEAR)
slen +=
snprintf(output + slen, outbufsize - slen, "near ");
output[slen++] = '[';
if (o->disp_size)
slen += snprintf(output + slen, outbufsize - slen, "%s",
(o->disp_size == 64 ? "qword " :
o->disp_size == 32 ? "dword " :
o->disp_size == 16 ? "word " :
""));
if (o->eaflags & EAF_REL)
slen += snprintf(output + slen, outbufsize - slen, "rel ");
if (segover) {
slen +=
snprintf(output + slen, outbufsize - slen, "%s:",
segover);
segover = NULL;
}
if (o->basereg != -1) {
slen += snprintf(output + slen, outbufsize - slen, "%s",
nasm_reg_names[(o->basereg-EXPR_REG_START)]);
started = true;
}
if (o->indexreg != -1) {
if (started)
output[slen++] = '+';
slen += snprintf(output + slen, outbufsize - slen, "%s",
nasm_reg_names[(o->indexreg-EXPR_REG_START)]);
if (o->scale > 1)
slen +=
snprintf(output + slen, outbufsize - slen, "*%d",
o->scale);
started = true;
}
if (o->segment & SEG_DISP8) {
const char *prefix;
uint8_t offset = offs;
if ((int8_t)offset < 0) {
prefix = "-";
offset = -offset;
} else {
prefix = "+";
}
slen +=
snprintf(output + slen, outbufsize - slen, "%s0x%"PRIx8"",
prefix, offset);
} else if (o->segment & SEG_DISP16) {
const char *prefix;
uint16_t offset = offs;
if ((int16_t)offset < 0 && started) {
offset = -offset;
prefix = "-";
} else {
prefix = started ? "+" : "";
}
slen +=
snprintf(output + slen, outbufsize - slen,
"%s0x%"PRIx16"", prefix, offset);
} else if (o->segment & SEG_DISP32) {
if (prefix.asize == 64) {
const char *prefix;
uint64_t offset = (int64_t)(int32_t)offs;
if ((int32_t)offs < 0 && started) {
offset = -offset;
prefix = "-";
} else {
prefix = started ? "+" : "";
}
slen +=
snprintf(output + slen, outbufsize - slen,
"%s0x%"PRIx64"", prefix, offset);
} else {
const char *prefix;
uint32_t offset = offs;
if ((int32_t) offset < 0 && started) {
offset = -offset;
prefix = "-";
} else {
prefix = started ? "+" : "";
}
slen +=
snprintf(output + slen, outbufsize - slen,
"%s0x%"PRIx32"", prefix, offset);
}
}
output[slen++] = ']';
} else {
slen +=
snprintf(output + slen, outbufsize - slen, "<operand%d>",
i);
}
}
output[slen] = '\0';
if (segover) { /* unused segment override */
char *p = output;
int count = slen + 1;
while (count--)
p[count + 3] = p[count];
strncpy(output, segover, 2);
output[2] = ' ';
}
return length;
}
int32_t eatbyte(uint8_t *data, char *output, int outbufsize)
{
snprintf(output, outbufsize, "db 0x%02X", *data);
return 1;
}
⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -