nasm/disasm.c

758 lines
21 KiB
C

/* disasm.c where all the _work_ gets done in the Netwide Disassembler
*
* The Netwide Assembler is copyright (C) 1996 Simon Tatham and
* Julian Hall. All rights reserved. The software is
* redistributable under the licence given in the file "Licence"
* distributed in the NASM archive.
*
* initial version 27/iii/95 by Simon Tatham
*/
#include <stdio.h>
#include <string.h>
#include "nasm.h"
#include "disasm.h"
#include "sync.h"
#include "insns.h"
#include "names.c"
extern struct itemplate **itable[];
/*
* Flags that go into the `segment' field of `insn' structures
* during disassembly.
*/
#define SEG_RELATIVE 1
#define SEG_32BIT 2
#define SEG_RMREG 4
#define SEG_DISP8 8
#define SEG_DISP16 16
#define SEG_DISP32 32
#define SEG_NODISP 64
#define SEG_SIGNED 128
static int whichreg(long regflags, int regval)
{
static int reg32[] = {
R_EAX, R_ECX, R_EDX, R_EBX, R_ESP, R_EBP, R_ESI, R_EDI };
static int reg16[] = {
R_AX, R_CX, R_DX, R_BX, R_SP, R_BP, R_SI, R_DI };
static int reg8[] = {
R_AL, R_CL, R_DL, R_BL, R_AH, R_CH, R_DH, R_BH };
static int sreg[] = {
R_ES, R_CS, R_SS, R_DS, R_FS, R_GS, 0, 0 };
static int creg[] = {
R_CR0, 0, R_CR2, R_CR3, R_CR4, 0, 0, 0 };
static int dreg[] = {
R_DR0, R_DR1, R_DR2, R_DR3, 0, 0, R_DR6, R_DR7 };
static int treg[] = {
0, 0, 0, R_TR3, R_TR4, R_TR5, R_TR6, R_TR7 };
static int fpureg[] = {
R_ST0, R_ST1, R_ST2, R_ST3, R_ST4, R_ST5, R_ST6, R_ST7 };
static int mmxreg[] = {
R_MM0, R_MM1, R_MM2, R_MM3, R_MM4, R_MM5, R_MM6, R_MM7 };
static int xmmreg[] = {
R_XMM0, R_XMM1, R_XMM2, R_XMM3, R_XMM4, R_XMM5, R_XMM6, R_XMM7 };
if (!(REG_AL & ~regflags))
return R_AL;
if (!(REG_AX & ~regflags))
return R_AX;
if (!(REG_EAX & ~regflags))
return R_EAX;
if (!(REG_DX & ~regflags))
return R_DX;
if (!(REG_CL & ~regflags))
return R_CL;
if (!(REG_CX & ~regflags))
return R_CX;
if (!(REG_ECX & ~regflags))
return R_ECX;
if (!(REG_CR4 & ~regflags))
return R_CR4;
if (!(FPU0 & ~regflags))
return R_ST0;
if (!(REG_CS & ~regflags))
return R_CS;
if (!(REG_DESS & ~regflags))
return (regval == 0 || regval == 2 || regval == 3 ? sreg[regval] : 0);
if (!(REG_FSGS & ~regflags))
return (regval == 4 || regval == 5 ? sreg[regval] : 0);
if (!((REGMEM|BITS8) & ~regflags))
return reg8[regval];
if (!((REGMEM|BITS16) & ~regflags))
return reg16[regval];
if (!((REGMEM|BITS32) & ~regflags))
return reg32[regval];
if (!(REG_SREG & ~regflags))
return sreg[regval];
if (!(REG_CREG & ~regflags))
return creg[regval];
if (!(REG_DREG & ~regflags))
return dreg[regval];
if (!(REG_TREG & ~regflags))
return treg[regval];
if (!(FPUREG & ~regflags))
return fpureg[regval];
if (!(MMXREG & ~regflags))
return mmxreg[regval];
if (!(XMMREG & ~regflags))
return xmmreg[regval];
return 0;
}
static char *whichcond(int condval)
{
static int conds[] = {
C_O, C_NO, C_C, C_NC, C_Z, C_NZ, C_NA, C_A,
C_S, C_NS, C_PE, C_PO, C_L, C_NL, C_NG, C_G
};
return conditions[conds[condval]];
}
/*
* Process an effective address (ModRM) specification.
*/
static unsigned char *do_ea (unsigned char *data, int modrm, int asize,
int segsize, operand *op)
{
int mod, rm, scale, index, base;
mod = (modrm >> 6) & 03;
rm = modrm & 07;
if (mod == 3) { /* pure register version */
op->basereg = rm;
op->segment |= SEG_RMREG;
return data;
}
op->addr_size = 0;
if (asize == 16) {
/*
* <mod> specifies the displacement size (none, byte or
* word), and <rm> specifies the register combination.
* Exception: mod=0,rm=6 does not specify [BP] as one might
* expect, but instead specifies [disp16].
*/
op->indexreg = op->basereg = -1;
op->scale = 1; /* always, in 16 bits */
switch (rm) {
case 0: op->basereg = R_BX; op->indexreg = R_SI; break;
case 1: op->basereg = R_BX; op->indexreg = R_DI; break;
case 2: op->basereg = R_BP; op->indexreg = R_SI; break;
case 3: op->basereg = R_BP; op->indexreg = R_DI; break;
case 4: op->basereg = R_SI; break;
case 5: op->basereg = R_DI; break;
case 6: op->basereg = R_BP; break;
case 7: op->basereg = R_BX; break;
}
if (rm == 6 && mod == 0) { /* special case */
op->basereg = -1;
if (segsize != 16)
op->addr_size = 16;
mod = 2; /* fake disp16 */
}
switch (mod) {
case 0:
op->segment |= SEG_NODISP;
break;
case 1:
op->segment |= SEG_DISP8;
op->offset = (signed char) *data++;
break;
case 2:
op->segment |= SEG_DISP16;
op->offset = *data++;
op->offset |= (*data++) << 8;
break;
}
return data;
} else {
/*
* Once again, <mod> specifies displacement size (this time
* none, byte or *dword*), while <rm> specifies the base
* register. Again, [EBP] is missing, replaced by a pure
* disp32 (this time that's mod=0,rm=*5*). However, rm=4
* indicates not a single base register, but instead the
* presence of a SIB byte...
*/
op->indexreg = -1;
switch (rm) {
case 0: op->basereg = R_EAX; break;
case 1: op->basereg = R_ECX; break;
case 2: op->basereg = R_EDX; break;
case 3: op->basereg = R_EBX; break;
case 5: op->basereg = R_EBP; break;
case 6: op->basereg = R_ESI; break;
case 7: op->basereg = R_EDI; break;
}
if (rm == 5 && mod == 0) {
op->basereg = -1;
if (segsize != 32)
op->addr_size = 32;
mod = 2; /* fake disp32 */
}
if (rm == 4) { /* process SIB */
scale = (*data >> 6) & 03;
index = (*data >> 3) & 07;
base = *data & 07;
data++;
op->scale = 1 << scale;
switch (index) {
case 0: op->indexreg = R_EAX; break;
case 1: op->indexreg = R_ECX; break;
case 2: op->indexreg = R_EDX; break;
case 3: op->indexreg = R_EBX; break;
case 4: op->indexreg = -1; break;
case 5: op->indexreg = R_EBP; break;
case 6: op->indexreg = R_ESI; break;
case 7: op->indexreg = R_EDI; break;
}
switch (base) {
case 0: op->basereg = R_EAX; break;
case 1: op->basereg = R_ECX; break;
case 2: op->basereg = R_EDX; break;
case 3: op->basereg = R_EBX; break;
case 4: op->basereg = R_ESP; break;
case 6: op->basereg = R_ESI; break;
case 7: op->basereg = R_EDI; break;
case 5:
if (mod == 0) {
mod = 2;
op->basereg = -1;
} else
op->basereg = R_EBP;
break;
}
}
switch (mod) {
case 0:
op->segment |= SEG_NODISP;
break;
case 1:
op->segment |= SEG_DISP8;
op->offset = (signed char) *data++;
break;
case 2:
op->segment |= SEG_DISP32;
op->offset = *data++;
op->offset |= (*data++) << 8;
op->offset |= ((long) *data++) << 16;
op->offset |= ((long) *data++) << 24;
break;
}
return data;
}
}
/*
* Determine whether the instruction template in t corresponds to the data
* stream in data. Return the number of bytes matched if so.
*/
static int matches (struct itemplate *t, unsigned char *data, int asize,
int osize, int segsize, int rep, insn *ins)
{
unsigned char * r = (unsigned char *)(t->code);
unsigned char * origdata = data;
int a_used = FALSE, o_used = FALSE;
int drep = 0;
if ( rep == 0xF2 )
drep = P_REPNE;
else if ( rep == 0xF3 )
drep = P_REP;
while (*r)
{
int c = *r++;
if (c >= 01 && c <= 03) {
while (c--)
if (*r++ != *data++)
return FALSE;
}
if (c == 04) {
switch (*data++) {
case 0x07: ins->oprs[0].basereg = 0; break;
case 0x17: ins->oprs[0].basereg = 2; break;
case 0x1F: ins->oprs[0].basereg = 3; break;
default: return FALSE;
}
}
if (c == 05) {
switch (*data++) {
case 0xA1: ins->oprs[0].basereg = 4; break;
case 0xA9: ins->oprs[0].basereg = 5; break;
default: return FALSE;
}
}
if (c == 06) {
switch (*data++) {
case 0x06: ins->oprs[0].basereg = 0; break;
case 0x0E: ins->oprs[0].basereg = 1; break;
case 0x16: ins->oprs[0].basereg = 2; break;
case 0x1E: ins->oprs[0].basereg = 3; break;
default: return FALSE;
}
}
if (c == 07) {
switch (*data++) {
case 0xA0: ins->oprs[0].basereg = 4; break;
case 0xA8: ins->oprs[0].basereg = 5; break;
default: return FALSE;
}
}
if (c >= 010 && c <= 012) {
int t = *r++, d = *data++;
if (d < t || d > t+7)
return FALSE;
else {
ins->oprs[c-010].basereg = d-t;
ins->oprs[c-010].segment |= SEG_RMREG;
}
}
if (c == 017)
if (*data++)
return FALSE;
if (c >= 014 && c <= 016) {
ins->oprs[c-014].offset = (signed char) *data++;
ins->oprs[c-014].segment |= SEG_SIGNED;
}
if (c >= 020 && c <= 022)
ins->oprs[c-020].offset = *data++;
if (c >= 024 && c <= 026)
ins->oprs[c-024].offset = *data++;
if (c >= 030 && c <= 032) {
ins->oprs[c-030].offset = *data++;
ins->oprs[c-030].offset |= (*data++ << 8);
}
if (c >= 034 && c <= 036) {
ins->oprs[c-034].offset = *data++;
ins->oprs[c-034].offset |= (*data++ << 8);
if (osize == 32) {
ins->oprs[c-034].offset |= (((long) *data++) << 16);
ins->oprs[c-034].offset |= (((long) *data++) << 24);
}
if (segsize != asize)
ins->oprs[c-034].addr_size = asize;
}
if (c >= 040 && c <= 042) {
ins->oprs[c-040].offset = *data++;
ins->oprs[c-040].offset |= (*data++ << 8);
ins->oprs[c-040].offset |= (((long) *data++) << 16);
ins->oprs[c-040].offset |= (((long) *data++) << 24);
}
if (c >= 044 && c <= 046) {
ins->oprs[c-044].offset = *data++;
ins->oprs[c-044].offset |= (*data++ << 8);
if (asize == 32) {
ins->oprs[c-044].offset |= (((long) *data++) << 16);
ins->oprs[c-044].offset |= (((long) *data++) << 24);
}
if (segsize != asize)
ins->oprs[c-044].addr_size = asize;
}
if (c >= 050 && c <= 052) {
ins->oprs[c-050].offset = (signed char) *data++;
ins->oprs[c-050].segment |= SEG_RELATIVE;
}
if (c >= 060 && c <= 062) {
ins->oprs[c-060].offset = *data++;
ins->oprs[c-060].offset |= (*data++ << 8);
ins->oprs[c-060].segment |= SEG_RELATIVE;
ins->oprs[c-060].segment &= ~SEG_32BIT;
}
if (c >= 064 && c <= 066) {
ins->oprs[c-064].offset = *data++;
ins->oprs[c-064].offset |= (*data++ << 8);
if (osize == 32) {
ins->oprs[c-064].offset |= (((long) *data++) << 16);
ins->oprs[c-064].offset |= (((long) *data++) << 24);
ins->oprs[c-064].segment |= SEG_32BIT;
} else
ins->oprs[c-064].segment &= ~SEG_32BIT;
ins->oprs[c-064].segment |= SEG_RELATIVE;
if (segsize != osize) {
ins->oprs[c-064].type =
(ins->oprs[c-064].type & NON_SIZE)
| ((osize == 16) ? BITS16 : BITS32);
}
}
if (c >= 070 && c <= 072) {
ins->oprs[c-070].offset = *data++;
ins->oprs[c-070].offset |= (*data++ << 8);
ins->oprs[c-070].offset |= (((long) *data++) << 16);
ins->oprs[c-070].offset |= (((long) *data++) << 24);
ins->oprs[c-070].segment |= SEG_32BIT | SEG_RELATIVE;
}
if (c >= 0100 && c < 0130) {
int modrm = *data++;
ins->oprs[c & 07].basereg = (modrm >> 3) & 07;
ins->oprs[c & 07].segment |= SEG_RMREG;
data = do_ea (data, modrm, asize, segsize,
&ins->oprs[(c >> 3) & 07]);
}
if (c >= 0130 && c <= 0132) {
ins->oprs[c-0130].offset = *data++;
ins->oprs[c-0130].offset |= (*data++ << 8);
}
if (c >= 0140 && c <= 0142) {
ins->oprs[c-0140].offset = *data++;
ins->oprs[c-0140].offset |= (*data++ << 8);
ins->oprs[c-0140].offset |= (((long) *data++) << 16);
ins->oprs[c-0140].offset |= (((long) *data++) << 24);
}
if (c >= 0200 && c <= 0277) {
int modrm = *data++;
if (((modrm >> 3) & 07) != (c & 07))
return FALSE; /* spare field doesn't match up */
data = do_ea (data, modrm, asize, segsize,
&ins->oprs[(c >> 3) & 07]);
}
if (c >= 0300 && c <= 0302) {
if (asize)
ins->oprs[c-0300].segment |= SEG_32BIT;
else
ins->oprs[c-0300].segment &= ~SEG_32BIT;
a_used = TRUE;
}
if (c == 0310) {
if (asize == 32)
return FALSE;
else
a_used = TRUE;
}
if (c == 0311) {
if (asize == 16)
return FALSE;
else
a_used = TRUE;
}
if (c == 0312) {
if (asize != segsize)
return FALSE;
else
a_used = TRUE;
}
if (c == 0320) {
if (osize == 32)
return FALSE;
else
o_used = TRUE;
}
if (c == 0321) {
if (osize == 16)
return FALSE;
else
o_used = TRUE;
}
if (c == 0322) {
if (osize != segsize)
return FALSE;
else
o_used = TRUE;
}
if (c == 0330) {
int t = *r++, d = *data++;
if (d < t || d > t+15)
return FALSE;
else
ins->condition = d - t;
}
if (c == 0331) {
if ( rep )
return FALSE;
}
if (c == 0332) {
if (drep == P_REP)
drep = P_REPE;
}
if (c == 0333) {
if ( rep != 0xF3 )
return FALSE;
drep = 0;
}
}
/*
* Check for unused rep or a/o prefixes.
*/
ins->nprefix = 0;
if (drep)
ins->prefixes[ins->nprefix++] = drep;
if (!a_used && asize != segsize)
ins->prefixes[ins->nprefix++] = (asize == 16 ? P_A16 : P_A32);
if (!o_used && osize != segsize)
ins->prefixes[ins->nprefix++] = (osize == 16 ? P_O16 : P_O32);
return data - origdata;
}
long disasm (unsigned char *data, char *output, int segsize, long offset,
int autosync, unsigned long prefer)
{
struct itemplate **p, **best_p;
int length, best_length = 0;
char *segover;
int rep, lock, asize, osize, i, slen, colon;
unsigned char *origdata;
int works;
insn tmp_ins, ins;
unsigned long goodness, best;
/*
* Scan for prefixes.
*/
asize = osize = segsize;
segover = NULL;
rep = lock = 0;
origdata = data;
for (;;) {
if (*data == 0xF3 || *data == 0xF2)
rep = *data++;
else if (*data == 0xF0)
lock = *data++;
else if (*data == 0x2E || *data == 0x36 || *data == 0x3E ||
*data == 0x26 || *data == 0x64 || *data == 0x65) {
switch (*data++) {
case 0x2E: segover = "cs"; break;
case 0x36: segover = "ss"; break;
case 0x3E: segover = "ds"; break;
case 0x26: segover = "es"; break;
case 0x64: segover = "fs"; break;
case 0x65: segover = "gs"; break;
}
} else if (*data == 0x66)
osize = 48 - segsize, data++;
else if (*data == 0x67)
asize = 48 - segsize, data++;
else
break;
}
tmp_ins.oprs[0].segment = tmp_ins.oprs[1].segment =
tmp_ins.oprs[2].segment =
tmp_ins.oprs[0].addr_size = tmp_ins.oprs[1].addr_size =
tmp_ins.oprs[2].addr_size = (segsize == 16 ? 0 : SEG_32BIT);
tmp_ins.condition = -1;
best = ~0UL; /* Worst possible */
best_p = NULL;
for (p = itable[*data]; *p; p++) {
if ( (length = matches(*p, data, asize, osize,
segsize, rep, &tmp_ins)) ) {
works = TRUE;
/*
* Final check to make sure the types of r/m match up.
*/
for (i = 0; i < (*p)->operands; i++) {
if (
/* 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) &&
!(REGNORM & ~(*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))) {
works = FALSE;
break;
}
}
if (works) {
goodness = ((*p)->flags & IF_PFMASK) ^ prefer;
if ( goodness < best ) {
/* This is the best one found so far */
best = goodness;
best_p = p;
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;
if (lock)
slen += sprintf(output+slen, "lock ");
for (i = 0; i < ins.nprefix; i++)
switch (ins.prefixes[i]) {
case P_REP: slen += sprintf(output+slen, "rep "); break;
case P_REPE: slen += sprintf(output+slen, "repe "); break;
case P_REPNE: slen += sprintf(output+slen, "repne "); break;
case P_A16: slen += sprintf(output+slen, "a16 "); break;
case P_A32: slen += sprintf(output+slen, "a32 "); break;
case P_O16: slen += sprintf(output+slen, "o16 "); break;
case P_O32: slen += sprintf(output+slen, "o32 "); break;
}
for (i = 0; i < elements(ico); i++)
if ((*p)->opcode == ico[i]) {
slen += sprintf(output+slen, "%s%s", icn[i],
whichcond(ins.condition));
break;
}
if (i >= elements(ico))
slen += sprintf(output+slen, "%s", insn_names[(*p)->opcode]);
colon = FALSE;
length += data - origdata; /* fix up for prefixes */
for (i=0; i<(*p)->operands; i++) {
output[slen++] = (colon ? ':' : i==0 ? ' ' : ',');
if (ins.oprs[i].segment & SEG_RELATIVE) {
ins.oprs[i].offset += offset + length;
/*
* sort out wraparound
*/
if (!(ins.oprs[i].segment & SEG_32BIT))
ins.oprs[i].offset &= 0xFFFF;
/*
* add sync marker, if autosync is on
*/
if (autosync)
add_sync (ins.oprs[i].offset, 0L);
}
if ((*p)->opd[i] & COLON)
colon = TRUE;
else
colon = FALSE;
if (((*p)->opd[i] & (REGISTER | FPUREG)) ||
(ins.oprs[i].segment & SEG_RMREG))
{
ins.oprs[i].basereg = whichreg ((*p)->opd[i],
ins.oprs[i].basereg);
if ( (*p)->opd[i] & TO )
slen += sprintf(output+slen, "to ");
slen += sprintf(output+slen, "%s",
reg_names[ins.oprs[i].basereg-EXPR_REG_START]);
} else if (!(UNITY & ~(*p)->opd[i])) {
output[slen++] = '1';
} else if ( (*p)->opd[i] & IMMEDIATE ) {
if ( (*p)->opd[i] & BITS8 ) {
slen += sprintf(output+slen, "byte ");
if (ins.oprs[i].segment & SEG_SIGNED) {
if (ins.oprs[i].offset < 0) {
ins.oprs[i].offset *= -1;
output[slen++] = '-';
} else
output[slen++] = '+';
}
} else if ( (*p)->opd[i] & BITS16 ) {
slen += sprintf(output+slen, "word ");
} else if ( (*p)->opd[i] & BITS32 ) {
slen += sprintf(output+slen, "dword ");
} else if ( (*p)->opd[i] & NEAR ) {
slen += sprintf(output+slen, "near ");
} else if ( (*p)->opd[i] & SHORT ) {
slen += sprintf(output+slen, "short ");
}
slen += sprintf(output+slen, "0x%lx", ins.oprs[i].offset);
} else if ( !(MEM_OFFS & ~(*p)->opd[i]) ) {
slen += sprintf(output+slen, "[%s%s%s0x%lx]",
(segover ? segover : ""),
(segover ? ":" : ""),
(ins.oprs[i].addr_size == 32 ? "dword " :
ins.oprs[i].addr_size == 16 ? "word " : ""),
ins.oprs[i].offset);
segover = NULL;
} else if ( !(REGMEM & ~(*p)->opd[i]) ) {
int started = FALSE;
if ( (*p)->opd[i] & BITS8 )
slen += sprintf(output+slen, "byte ");
if ( (*p)->opd[i] & BITS16 )
slen += sprintf(output+slen, "word ");
if ( (*p)->opd[i] & BITS32 )
slen += sprintf(output+slen, "dword ");
if ( (*p)->opd[i] & BITS64 )
slen += sprintf(output+slen, "qword ");
if ( (*p)->opd[i] & BITS80 )
slen += sprintf(output+slen, "tword ");
if ( (*p)->opd[i] & FAR )
slen += sprintf(output+slen, "far ");
if ( (*p)->opd[i] & NEAR )
slen += sprintf(output+slen, "near ");
output[slen++] = '[';
if (ins.oprs[i].addr_size)
slen += sprintf(output+slen, "%s",
(ins.oprs[i].addr_size == 32 ? "dword " :
ins.oprs[i].addr_size == 16 ? "word " : ""));
if (segover) {
slen += sprintf(output+slen, "%s:", segover);
segover = NULL;
}
if (ins.oprs[i].basereg != -1) {
slen += sprintf(output+slen, "%s",
reg_names[(ins.oprs[i].basereg -
EXPR_REG_START)]);
started = TRUE;
}
if (ins.oprs[i].indexreg != -1) {
if (started)
output[slen++] = '+';
slen += sprintf(output+slen, "%s",
reg_names[(ins.oprs[i].indexreg -
EXPR_REG_START)]);
if (ins.oprs[i].scale > 1)
slen += sprintf(output+slen, "*%d", ins.oprs[i].scale);
started = TRUE;
}
if (ins.oprs[i].segment & SEG_DISP8) {
int sign = '+';
if (ins.oprs[i].offset & 0x80) {
ins.oprs[i].offset = - (signed char) ins.oprs[i].offset;
sign = '-';
}
slen += sprintf(output+slen, "%c0x%lx", sign,
ins.oprs[i].offset);
} else if (ins.oprs[i].segment & SEG_DISP16) {
if (started)
output[slen++] = '+';
slen += sprintf(output+slen, "0x%lx", ins.oprs[i].offset);
} else if (ins.oprs[i].segment & SEG_DISP32) {
if (started)
output[slen++] = '+';
slen += sprintf(output+slen, "0x%lx", ins.oprs[i].offset);
}
output[slen++] = ']';
} else {
slen += sprintf(output+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;
}
long eatbyte (unsigned char *data, char *output)
{
sprintf(output, "db 0x%02X", *data);
return 1;
}