int
fn_disassemble_xed(xed_syntax_enum_t syntax,
char* buf,
int buflen,
xed_decoded_inst_t* xedd,
xed_uint64_t runtime_instruction_address,
void* caller_data)
{
/* the caller_data is for symbolic disam */
char buffer[XED_TMP_BUF_LEN];
int blen = buflen;
int ok = xed_format_context(syntax, xedd,
buffer, XED_TMP_BUF_LEN,
runtime_instruction_address, caller_data);
if(ok)
blen = xed_strncpy(buf, buffer, buflen);
else
{
blen = buflen;
blen = xed_strncpy(buf, "Error disassembling ", blen);
blen = xed_strncat(buf, xed_syntax_enum_t2str(syntax), blen);
blen = xed_strncat(buf, " syntax.", blen);
}
return blen;
}
void xed_dot_graph_add_instruction(
xed_dot_graph_supp_t* gg,
xed_decoded_inst_t* xedd,
xed_uint64_t runtime_instr_addr,
void* caller_data)
{
/*
make a new node
for each operand:
if read:
make edge from src node for that reg to the new node
for each operand:
if write:
install this node as the writer
what about partial writes?
what about register nesting?
*/
char disasm_str[XED_DOT_TMP_BUF_LEN];
char* p = 0;
size_t alen = 0;
int ok;
xed_dot_node_t* n = 0;
xed_uint32_t remaining_buffer_bytes = XED_DOT_TMP_BUF_LEN;
// put addr on separate line in node label
#if defined(XED_WINDOWS)
ok = sprintf_s(disasm_str,
XED_DOT_TMP_BUF_LEN,
XED_FMT_LX "\\n",
runtime_instr_addr);
#else
ok = sprintf(disasm_str,
XED_FMT_LX "\\n",
runtime_instr_addr);
#endif
assert(ok > 0);
alen = strlen(disasm_str);
p = disasm_str + alen;
remaining_buffer_bytes -= XED_CAST(xed_uint32_t, alen);
ok = xed_format_context(gg->syntax,
xedd,
p,
remaining_buffer_bytes,
runtime_instr_addr,
caller_data);
if (!ok) {
(void)xed_strncpy(disasm_str,"???", XED_DOT_TMP_BUF_LEN);
}
n = xed_dot_node(gg->g, disasm_str);
add_read_operands(gg,xedd,n);
add_write_operands(gg,xedd,n);
}
static PyObject *dump_intel_format(instruction_t *self)
{
char buf[64];
/* In Pin versions before "pin-2.14-*", `xed_format_context()' takes 6
* parameters instead of 7. Unfortunately, there's no consistent way of
* dumping instructions between various XED versions and there's no easy
* way of detecting the actual XED version at compile time.
*/
xed_format_context(XED_SYNTAX_INTEL, self->decoded_inst, buf, sizeof(buf) - 1,
self->runtime_address, NULL, NULL);
return PyString_FromFormat("%p %s", (void *)self->runtime_address, buf);
}
int
main(int argc, char** argv) {
xed_error_enum_t xed_error;
xed_bool_t long_mode = 0;
xed_bool_t prot16 = 0;
unsigned int first_argv;
unsigned int bytes = 0;
unsigned char itext[XED_MAX_INSTRUCTION_BYTES];
int i;
unsigned int u;
xed_decoded_inst_t xedd;
#define BUFLEN 1000
char buffer[BUFLEN];
xed_bool_t ok;
unsigned int isyntax;
xed_syntax_enum_t syntax;
xed_machine_mode_enum_t mmode;
xed_address_width_enum_t stack_addr_width;
xed_encoder_request_t* enc_req;
xed_uint8_t array[XED_MAX_INSTRUCTION_BYTES];
unsigned int enc_olen, ilen = XED_MAX_INSTRUCTION_BYTES;
xed_error_enum_t encode_okay;
xed_bool_t change_to_long_mode = 0;
xed_tables_init();
xed_set_verbosity( 99 );
if (argc > 2 && strcmp(argv[1], "-64") == 0)
long_mode = 1;
else if (argc > 2 && strcmp(argv[1], "-16") == 0)
prot16 = 1;
if (long_mode) {
first_argv = 2;
mmode=XED_MACHINE_MODE_LONG_64;
stack_addr_width =XED_ADDRESS_WIDTH_64b;
}
else if (prot16) {
first_argv = 2;
mmode=XED_MACHINE_MODE_LEGACY_16;
stack_addr_width =XED_ADDRESS_WIDTH_16b;
}
else {
first_argv=1;
mmode=XED_MACHINE_MODE_LEGACY_32;
stack_addr_width =XED_ADDRESS_WIDTH_32b;
}
xed_decoded_inst_zero(&xedd);
xed_decoded_inst_set_mode(&xedd, mmode, stack_addr_width);
for( i=first_argv ;i < argc; i++) {
xed_uint8_t x = (xed_uint8_t)(xed_atoi_hex(argv[i]));
assert(bytes < XED_MAX_INSTRUCTION_BYTES);
itext[bytes++] = x;
}
if (bytes == 0) {
fprintf(stderr, "Must supply some hex bytes\n");
exit(1);
}
printf("PARSING BYTES: ");
for( u=0;u<bytes; u++)
printf("%02x ", XED_STATIC_CAST(unsigned int,itext[u]));
printf("\n");
xed_error = xed_decode(&xedd,
XED_REINTERPRET_CAST(const xed_uint8_t*,itext),
bytes);
switch(xed_error)
{
case XED_ERROR_NONE:
break;
case XED_ERROR_BUFFER_TOO_SHORT:
fprintf(stderr,"Not enough bytes provided\n");
exit(1);
case XED_ERROR_GENERAL_ERROR:
fprintf(stderr,"Could not decode given input.\n");
exit(1);
default:
fprintf(stderr,"Unhandled error code %s\n",
xed_error_enum_t2str(xed_error));
exit(1);
}
//memset(buffer,0,BUFLEN);
xed_decoded_inst_dump(&xedd,buffer, BUFLEN);
printf("%s\n",buffer);
for(isyntax= XED_SYNTAX_XED; isyntax < XED_SYNTAX_LAST; isyntax++) {
syntax = XED_STATIC_CAST(xed_syntax_enum_t, isyntax);
ok = xed_format_context(syntax, &xedd, buffer, BUFLEN, 0, 0, 0);
if (ok)
printf("%s syntax: %s\n", xed_syntax_enum_t2str(syntax), buffer);
else
printf("Error disassembling %s syntax\n",
xed_syntax_enum_t2str(syntax));
}
printf("\n\nPreparing to encode ...\n");
enc_req = &xedd; // they are basically the same now
// convert decode structure to proper encode structure
xed_encoder_request_init_from_decode(&xedd);
change_to_long_mode = 0;
if (change_to_long_mode) {
// change to 64b mode
xed_state_t state;
xed_operand_values_t* ov;
xed_state_init2(&state,XED_MACHINE_MODE_LONG_64,XED_ADDRESS_WIDTH_64b);
ov = xed_decoded_inst_operands(&xedd);
xed_operand_values_set_mode(ov, &state);
xed_encoder_request_set_effective_address_size(enc_req, 64);
// need to fix base regs...
//xed_operand_values_set_operand_reg(ov, XED_OPERAND_BASE0, XED_REG_RSI);
xed_encoder_request_set_effective_operand_width(enc_req, 32);
}
printf("Encoding...\n");
encode_okay = xed_encode(enc_req, array, ilen, &enc_olen);
if (encode_okay != XED_ERROR_NONE) {
printf("Error code = %s\n", xed_error_enum_t2str(encode_okay));
}
else {
unsigned int j;
printf("Encodable: ");
for(j=0;j<enc_olen;j++) {
printf("%02x", array[j]);
}
printf("\n");
}
return 0;
}
void Disasm::disasm(std::ostream& out, uint8_t* codeStartAddr,
uint8_t* codeEndAddr) {
#ifdef HAVE_LIBXED
auto const endClr = m_opts.m_color.empty() ? "" : ANSI_COLOR_END;
char codeStr[MAX_INSTR_ASM_LEN];
xed_uint8_t *frontier;
xed_decoded_inst_t xedd;
uint64_t codeBase = uint64_t(codeStartAddr);
uint64_t ip;
// Decode and print each instruction
for (frontier = codeStartAddr, ip = (uint64_t)codeStartAddr;
frontier < codeEndAddr; ) {
xed_decoded_inst_zero_set_mode(&xedd, &m_xedState);
xed_decoded_inst_set_input_chip(&xedd, XED_CHIP_INVALID);
xed_error_enum_t xed_error = xed_decode(&xedd, frontier, 15);
if (xed_error != XED_ERROR_NONE) error("disasm error: xed_decode failed");
// Get disassembled instruction in codeStr
auto const syntax = m_opts.m_forceAttSyntax ? XED_SYNTAX_ATT
: s_xed_syntax;
if (!xed_format_context(syntax, &xedd, codeStr,
MAX_INSTR_ASM_LEN, ip, nullptr)) {
error("disasm error: xed_format_context failed");
}
uint32_t instrLen = xed_decoded_inst_get_length(&xedd);
// If it's a jump, we're printing relative offsets, and the dest
// is within the range we're printing, add the dest as a relative
// offset.
std::string jmpComment;
auto const cat = xed_decoded_inst_get_category(&xedd);
if (cat == XED_CATEGORY_COND_BR || cat == XED_CATEGORY_UNCOND_BR) {
if (m_opts.m_relativeOffset) {
auto disp = uint64_t(frontier + instrLen +
xed_decoded_inst_get_branch_displacement(&xedd) -
codeBase);
if (disp < uint64_t(codeEndAddr - codeStartAddr)) {
jmpComment = folly::format(" # {:#x}", disp).str();
}
}
}
for (int i = 0; i < m_opts.m_indentLevel; ++i) {
out << ' ';
}
out << m_opts.m_color;
if (m_opts.m_addresses) {
const char* fmt = m_opts.m_relativeOffset ? "{:3x}: " : "{:#10x}: ";
out << folly::format(fmt, ip - (m_opts.m_relativeOffset ? codeBase : 0));
}
if (m_opts.m_printEncoding) {
// print encoding, like in objdump
unsigned posi = 0;
for (; posi < instrLen; ++posi) {
out << folly::format("{:02x} ", (uint8_t)frontier[posi]);
}
for (; posi < 16; ++posi) {
out << " ";
}
}
out << codeStr << jmpComment << endClr << '\n';
frontier += instrLen;
ip += instrLen;
}
#else
out << "This binary was compiled without disassembly support\n";
#endif // HAVE_LIBXED
}
static string
disassemble(UINT64 start, UINT64 stop) {
UINT64 pc = start;
xed_state_t dstate;
xed_syntax_enum_t syntax = XED_SYNTAX_INTEL;
xed_error_enum_t xed_error;
xed_decoded_inst_t xedd;
ostringstream os;
if (sizeof(ADDRINT) == 4)
xed_state_init(&dstate,
XED_MACHINE_MODE_LEGACY_32,
XED_ADDRESS_WIDTH_32b,
XED_ADDRESS_WIDTH_32b);
else
xed_state_init(&dstate,
XED_MACHINE_MODE_LONG_64,
XED_ADDRESS_WIDTH_64b,
XED_ADDRESS_WIDTH_64b);
/*while( pc < stop )*/ {
xed_decoded_inst_zero_set_mode(&xedd, &dstate);
UINT32 len = 15;
if (stop - pc < 15)
len = stop-pc;
xed_error = xed_decode(&xedd, reinterpret_cast<const UINT8*>(pc), len);
bool okay = (xed_error == XED_ERROR_NONE);
iostream::fmtflags fmt = os.flags();
os << std::setfill('0')
<< "XDIS "
<< std::hex
<< std::setw(sizeof(ADDRINT)*2)
<< pc
<< std::dec
<< ": "
<< std::setfill(' ')
<< std::setw(4);
if (okay) {
char buffer[200];
unsigned int dec_len, sp;
os << xed_extension_enum_t2str(xed_decoded_inst_get_extension(&xedd));
dec_len = xed_decoded_inst_get_length(&xedd);
print_hex_line(buffer, reinterpret_cast<UINT8*>(pc), dec_len);
os << " " << buffer;
for ( sp=dec_len; sp < 12; sp++) // pad out the instruction bytes
os << " ";
os << " ";
memset(buffer,0,200);
int dis_okay = xed_format_context(syntax, &xedd, buffer, 200, pc, 0, 0);
if (dis_okay)
os << buffer << endl;
else
os << "Error disasassembling pc 0x" << std::hex << pc << std::dec << endl;
pc += dec_len;
}
else { // print the byte and keep going.
UINT8 memval = *reinterpret_cast<UINT8*>(pc);
os << "???? " // no extension
<< std::hex
<< std::setw(2)
<< std::setfill('0')
<< static_cast<UINT32>(memval)
<< std::endl;
pc += 1;
}
os.flags(fmt);
}
return os.str();
}
int
main(int argc, char** argv)
{
xed_error_enum_t xed_error;
xed_bool_t long_mode = 0;
unsigned int bytes = 0;
unsigned char itext[XED_MAX_INSTRUCTION_BYTES];
int i;
unsigned int u;
xed_decoded_inst_t xedd;
#define BUFLEN 1000
char buffer[BUFLEN];
xed_bool_t ok;
unsigned int isyntax;
xed_syntax_enum_t syntax;
xed_machine_mode_enum_t mmode;
xed_address_width_enum_t stack_addr_width;
xed_format_options_t format_options;
// one time initialization
xed_tables_init();
xed_set_verbosity( 99 );
memset(&format_options,0, sizeof(format_options));
format_options.hex_address_before_symbolic_name=0;
format_options.xml_a=0;
format_options.omit_unit_scale=0;
format_options.no_sign_extend_signed_immediates=0;
for(i=1;i<argc;i++) {
if (strcmp(argv[i], "-xml") == 0)
format_options.xml_a=1;
else if (strcmp(argv[i], "-no-unit-scale") == 0)
format_options.omit_unit_scale=1;
else if (strcmp(argv[i], "-no-sign-extend") == 0)
format_options.no_sign_extend_signed_immediates=1;
else if (strcmp(argv[i], "-64") == 0)
long_mode = 1;
else
break;
}
xed_format_set_options( format_options );
/// begin processing of instructions...
if (long_mode) {
mmode=XED_MACHINE_MODE_LONG_64;
stack_addr_width =XED_ADDRESS_WIDTH_64b;
}
else {
mmode=XED_MACHINE_MODE_LEGACY_32;
stack_addr_width =XED_ADDRESS_WIDTH_32b;
}
xed_decoded_inst_zero(&xedd);
xed_decoded_inst_set_mode(&xedd, mmode, stack_addr_width);
for( ;i < argc; i++) {
xed_uint8_t x = (xed_uint8_t)(xed_atoi_hex(argv[i]));
assert(bytes < XED_MAX_INSTRUCTION_BYTES);
itext[bytes++] = x;
}
if (bytes == 0) {
fprintf(stderr, "Must supply some hex bytes\n");
exit(1);
}
printf("PARSING BYTES: ");
for( u=0;u<bytes; u++)
printf("%02x ", XED_STATIC_CAST(unsigned int,itext[u]));
printf("\n");
xed_error = xed_decode(&xedd,
XED_REINTERPRET_CAST(const xed_uint8_t*,itext),
bytes);
switch(xed_error)
{
case XED_ERROR_NONE:
break;
case XED_ERROR_BUFFER_TOO_SHORT:
fprintf(stderr,"Not enough bytes provided\n");
exit(1);
case XED_ERROR_GENERAL_ERROR:
fprintf(stderr,"Could not decode given input.\n");
exit(1);
default:
fprintf(stderr,"Unhandled error code %s\n",
xed_error_enum_t2str(xed_error));
exit(1);
}
xed_decoded_inst_dump(&xedd,buffer, BUFLEN);
printf("%s\n",buffer);
for(isyntax= XED_SYNTAX_XED; isyntax < XED_SYNTAX_LAST; isyntax++) {
syntax = XED_STATIC_CAST(xed_syntax_enum_t, isyntax);
ok = xed_format_context(syntax, &xedd, buffer, BUFLEN, 0, 0, 0);
if (ok)
printf("%s syntax: %s\n", xed_syntax_enum_t2str(syntax), buffer);
else
printf("Error disassembling %s syntax\n", xed_syntax_enum_t2str(syntax));
}
return 0;
}
void OfflineX86Code::disasm(FILE* file,
TCA fileStartAddr,
TCA codeStartAddr,
uint64_t codeLen,
const PerfEventsMap<TCA>& perfEvents,
BCMappingInfo bcMappingInfo,
bool printAddr /* =true */,
bool printBinary /* =false */) {
char codeStr[MAX_INSTR_ASM_LEN];
xed_uint8_t* code = (xed_uint8_t*) alloca(codeLen);
xed_uint8_t* frontier;
TCA ip;
TCA r10val = 0;
size_t currBC = 0;
if (codeLen == 0) return;
auto const offset = codeStartAddr - fileStartAddr;
if (fseek(file, offset, SEEK_SET)) {
error("disasm error: seeking file");
}
size_t readLen = fread(code, codeLen, 1, file);
if (readLen != 1) {
error("Failed to read {} bytes at offset {} from code file due to {}",
codeLen, offset, feof(file) ? "EOF" : "read error");
}
xed_decoded_inst_t xedd;
// Decode and print each instruction
for (frontier = code, ip = codeStartAddr; frontier < code + codeLen; ) {
xed_decoded_inst_zero_set_mode(&xedd, &xed_state);
xed_decoded_inst_set_input_chip(&xedd, XED_CHIP_INVALID);
xed_error_enum_t xed_error = xed_decode(&xedd, frontier, 15);
if (xed_error != XED_ERROR_NONE) break;
// Get disassembled instruction in codeStr
if (!xed_format_context(xed_syntax, &xedd, codeStr,
MAX_INSTR_ASM_LEN, (uint64_t)ip, nullptr
#if XED_ENCODE_ORDER_MAX_ENTRIES != 28 // Newer version of XED library
, 0
#endif
)) {
error("disasm error: xed_format_context failed");
}
// Annotate the x86 with its bytecode.
currBC = printBCMapping(bcMappingInfo, currBC, (TCA)ip);
if (printAddr) printf("%14p: ", ip);
uint32_t instrLen = xed_decoded_inst_get_length(&xedd);
if (printBinary) {
uint32_t i;
for (i=0; i < instrLen; i++) {
printf("%02X", frontier[i]);
}
for (; i < 16; i++) {
printf(" ");
}
}
// For calls, we try to figure out the destination symbol name.
// We look both at relative branches and the pattern:
// move r10, IMMEDIATE
// call r10
xed_iclass_enum_t iclass = xed_decoded_inst_get_iclass(&xedd);
string callDest = "";
if (iclass == XED_ICLASS_CALL_NEAR || iclass == XED_ICLASS_CALL_FAR) {
const xed_inst_t *xi = xed_decoded_inst_inst(&xedd);
always_assert(xed_inst_noperands(xi) >= 1);
const xed_operand_t *opnd = xed_inst_operand(xi, 0);
xed_operand_enum_t opndName = xed_operand_name(opnd);
if (opndName == XED_OPERAND_RELBR) {
if (xed_decoded_inst_get_branch_displacement_width(&xedd)) {
xed_int32_t disp = xed_decoded_inst_get_branch_displacement(&xedd);
TCA addr = ip + instrLen + disp;
callDest = getSymbolName(addr);
}
} else if (opndName == XED_OPERAND_REG0) {
if (xed_decoded_inst_get_reg(&xedd, opndName) == XED_REG_R10) {
callDest = getSymbolName(r10val);
}
}
} else if (iclass == XED_ICLASS_MOV) {
// Look for moves into r10 and keep r10val updated
const xed_inst_t* xi = xed_decoded_inst_inst(&xedd);
always_assert(xed_inst_noperands(xi) >= 2);
const xed_operand_t *destOpnd = xed_inst_operand(xi, 0);
xed_operand_enum_t destOpndName = xed_operand_name(destOpnd);
if (destOpndName == XED_OPERAND_REG0 &&
xed_decoded_inst_get_reg(&xedd, destOpndName) == XED_REG_R10) {
const xed_operand_t *srcOpnd = xed_inst_operand(xi, 1);
xed_operand_enum_t srcOpndName = xed_operand_name(srcOpnd);
if (srcOpndName == XED_OPERAND_IMM0) {
TCA addr = (TCA)xed_decoded_inst_get_unsigned_immediate(&xedd);
r10val = addr;
}
}
}
if (!perfEvents.empty()) {
printEventStats((TCA)ip, instrLen, perfEvents);
} else {
printf("%48s", "");
}
printf("%s%s\n", codeStr, callDest.c_str());
frontier += instrLen;
ip += instrLen;
}
}
