def classify(self): if patterns.decimal_pattern.match(self.value): self.emit_type = 'numeric' self.int_value = int(self.value) t = hex(self.int_value) self.nbits = 4 * len(t[2:]) if vclassify(): msgb("CLASSIFY", "%s as decimal values" % (self.value)) return if patterns.hex_pattern.match(self.value): self.emit_type = 'numeric' self.int_value = int(self.value, 16) self.nbits = 4 * (len(self.value) - 2 ) # drop the 0x, convert nibbles to bits if vclassify(): msgb("CLASSIFY", "%s as hex" % (self.value)) return if patterns.letter_and_underscore_pattern.match(self.value): self.emit_type = 'letters' t = self.value t = genutil.no_underscores(t) self.nbits = len(t) if vclassify(): msgb("CLASSIFY", "%s as letters" % (self.value)) return b = patterns.binary_pattern.match(self.value) # leading "0b" if b: self.emit_type = 'numeric' t = '0b' + b.group('bits') # pattern match strips out 0b self.int_value = genutil.make_numeric(t) bits_str = genutil.make_binary(t) self.nbits = len(bits_str) if vclassify(): msgb( "CLASSIFY", "%s as explicit-binary -> int = %d nbits=%d [%s,%s]" % (self.value, self.int_value, self.nbits, t, bits_str)) return if patterns.bits_and_letters_underscore_pattern.match(self.value): self.emit_type = 'letters' v = genutil.no_underscores(self.value) self.nbits = len(v) if vclassify(): msgb("CLASSIFY", "%s as mixed-letters" % (self.value)) return if patterns.simple_number_pattern.match(self.value): self.emit_type = 'numeric' self.int_value = genutil.make_numeric(self.value) t = hex(self.int_value) self.nbits = 4 * len(t[2:]) if vclassify(): msgb("CLASSIFY", "%s as simple-number" % (self.value)) return genutil.die("unknown pattern")
def _parse_map_line(s): # shlex allows for quoted substrings containing spaces as # individual args. t = shlex.split(s.strip()) if len(t) != 10: _die("Bad map description line: [{}]".format(s)) mi = map_info_t() mi.map_name = t[0] mi.space = t[1] mi.legacy_escape = t[2] mi.legacy_opcode = t[3] mi.map_id = t[4] mi.modrm = t[5] mi.imm8 = t[6] mi.imm32 = t[7] mi.opcpos = t[8] mi.search_pattern = t[9] if mi.space not in ['legacy', 'vex', 'evex', 'xop']: _die("Bad map description encoding space [{}]".format(s)) if mi.space == 'legacy': if genutil.is_hex(mi.legacy_escape): pass elif mi.legacy_escape != 'N/A': _die("Bad map description legacy escape [{}]".format(s)) if genutil.is_hex(mi.legacy_opcode): pass elif mi.legacy_opcode != 'N/A': _die("Bad map description legacy opcode [{}]".format(s)) if mi.map_id == 'N/A': _die("Bad map description map-id [{}]".format(s)) else: if mi.legacy_escape != 'N/A': _die("Bad map description legacy escape [{}]".format(s)) if mi.legacy_opcode != 'N/A': _die("Bad map description legacy opcode [{}]".format(s)) if genutil.numeric(mi.map_id): mi.map_id = genutil.make_numeric(mi.map_id) else: _die("Bad map description map id [{}]".format(s)) if mi.modrm not in ['var', 'yes', 'no']: _die("Bad map description modrm specifier [{}]".format(s)) if mi.imm8 not in ['var', 'yes', 'no']: _die("Bad map description imm8 specifier [{}]".format(s)) if mi.imm32 not in ['var', 'yes', 'no']: _die("Bad map description imm32 specifier [{}]".format(s)) if genutil.numeric(mi.opcpos): mi.opcpos = genutil.make_numeric(mi.opcpos) else: _die("Bad map description opcode position specifier [{}]".format(s)) # we want the longer patterns first when we sort the map_info_t. mi.priority = 100 - len(mi.search_pattern) return mi
def _read_constant_tables(lines, tables): """Read lines from lines until a new header or a blank line is reached""" nlines = 0 y = None for line in lines: nlines += 1 line = line.strip() line = re.sub(r'#.*', '', line) m = constant_table_t.match_blank.match(line) if m: continue m = constant_table_t.match_header.match(line) if m: # found next header name = m.group('name') y = None for t in tables: if t.name == name: y = t if not y: y = constant_table_t() tables.append(y) y.name = name y.operand = m.group('operand') continue m = constant_table_t.match_pair.match(line) if m: value = m.group('value') symbol = m.group('symbol') numeric_value = genutil.make_numeric(value) #print "INPUT: [%s] [%s]" % (value,symbol) if not y: genutil.die( "Malformed constant table line {}: [{}]\n\n".format( nlines, line)) y.value_string_pairs.append((numeric_value, symbol)) continue m = constant_table_t.match_pair_error.match(line) if m: value = m.group('value') numeric_value = genutil.make_numeric(value) if not y: genutil.die( "Malformed constant table line {}: [{}]\n\n".format( nlines, line)) y.value_string_pairs.append((numeric_value, None)) continue else: genutil.die("Could not parse line {}: [{}]\n\n".format( nlines, line))
def parse_opcode(self, op_str): # has side effects of settting self.missing_bits and self.incomplete val = None if genutil.numeric(op_str): val = genutil.make_numeric(op_str) # special check for partial binary numbers as opcodes if genutil.is_binary(op_str): bin_str = re.sub('^0b', '', op_str) bin_str = re.sub('_', '', bin_str) #if it is bin string, it might be incomplete, and we #assume that given bits are msb. Hence we have to #shift left if len(bin_str) > 8: ildutil.ild_err("Unexpectedly long binary opcode: %s" % bin_str) if len(bin_str) < 8: self.missing_bits = 8-len(bin_str) val = val << (self.missing_bits) self.incomplete_opcode = True _msg('incomplete opcode for iclass %s, pttrn %s' % (self.iclass, self.ptrn)) return val
def set_ext_opcode(self): #inline captures MOD[mm] REG[rrr] RM[nnn] #or REG[111] etc. -- can be constant for w in self.ptrn_wrds: pb = reg_binding_pattern.match(w) if pb: bits = pb.group('bits') self.ext_opcode = genutil.make_numeric(bits) return
def read(self, lines): """Read lines from lines until a new header or a blank line is reached""" started = False while 1: if not lines: break self.nlines += 1 line = lines[0] line = line.strip() line = re.sub(r'#.*','',line) m= constant_table_t.match_blank.match(line) if m: del lines[0] continue m= constant_table_t.match_header.match(line) if m: if started: return else: started = True del lines[0] self.name = m.group('name') self.operand = m.group('operand') continue m = constant_table_t.match_pair.match(line) if m: value = m.group('value') symbol = m.group('symbol') numeric_value = genutil.make_numeric(value) #print "INPUT: [%s] [%s]" % (value,symbol) self.value_string_pairs.append((numeric_value,symbol)) del lines[0] continue m = constant_table_t.match_pair_error.match(line) if m: value = m.group('value') numeric_value = genutil.make_numeric(value) self.value_string_pairs.append((numeric_value,None)) del lines[0] continue else: genutil.die("Could not parse line %d: [%s]\n\n" % (self.nlines,line))
def _build_constraint(self, nonterminal): ''' build a dict that represents the values that the operands can have. e.g. for the constraint EASZ=3 MODE!=3 the cdict is {EASZ:[3],MODE:[0,1,2]} ''' for rule in nonterminal.rules: constraints = {} for cond in rule.conditions.and_conditions: key = cond.field_name if cond.equals: if cond.rvalue.null(): constraints[key] = [self.reg2int[_xed_reg_invalid]] elif cond.rvalue.any_valid(): #will be gathered later continue else: if self._cond_is_UIMM0_1(cond): key = self._replace_UIMM0_1_cond() val = cond.rvalue.value if self._is_reg_type(key): constraints[key] = [self.reg2int[val]] else: constraints[key] = [genutil.make_numeric(val)] else: #we have != condition, we need to calculate all the #possible values based on the width of the field and remove #the the unwanted value. #need to deep copy since we modify the list, and we want #to preserve the original all_vals = list(copy.deepcopy(self.state_space[key])) val = genutil.make_numeric(cond.rvalue.value) all_vals.remove(val) constraints[key] = all_vals rule.cdict = constraints
def expand_partial_opcode(self): expanded = [] if self.incomplete_opcode: if 'RM[rrr]' in self.ptrn or 'REG[rrr]' in self.ptrn: _msg('Expanding opcode for %s' % self.iclass) #FIXME: MJC: was assuming hex for self.opcode (2012-06-19) opcode = genutil.make_numeric(self.opcode) #iterating through all legal variations of the incomplete #opcode. #Since the variant with all missing bits == 0 was already #created (it is the current self), we need to iterate through #1 to 2 ^ (self.missing_bits) for i in range(1, 2**self.missing_bits): new_ptrn = copy.deepcopy(self) new_ptrn.opcode = hex(opcode | i) expanded.append(new_ptrn) return expanded
def _find_opcodes(self): '''augment the records with information found by parsing the pattern''' map_pattern = re.compile(r'MAP=(?P<map>[0-6])') vex_prefix = re.compile(r'VEX_PREFIX=(?P<prefix>[0-9])') rep_prefix = re.compile(r'REP=(?P<prefix>[0-3])') osz_prefix = re.compile(r' OSZ=(?P<prefix>[01])') no_prefix = re.compile(r'REP=0 OSZ=0') rexw_prefix = re.compile(r'REXW=(?P<rexw>[01])') reg_required = re.compile(r'REG[\[](?P<reg>[b01]+)]') mod_required = re.compile(r'MOD[\[](?P<mod>[b01]+)]') mod_mem_required = re.compile(r'MOD!=3') mod_reg_required = re.compile(r'MOD=3') rm_val_required = re.compile(r'RM=(?P<rm>[0-9]+)') rm_required = re.compile( r'RM[\[](?P<rm>[b01]+)]') # RM[...] or SRM[...] mode_pattern = re.compile(r' MODE=(?P<mode>[012]+)') not64_pattern = re.compile(r' MODE!=2') for v in self.recs: if not hasattr(v, 'isa_set'): v.isa_set = v.extension v.undocumented = False if hasattr(v, 'comment'): if 'UNDOC' in v.comment: v.undocumented = True pattern = v.pattern.split() p0 = pattern[0] v.map = 0 v.space = 'legacy' if p0 == 'VEXVALID=1': v.space = 'vex' opcode = pattern[1] elif p0 == 'VEXVALID=2': v.space = 'evex' opcode = pattern[1] elif p0 == 'VEXVALID=4': #KNC v.space = 'evex.u0' opcode = pattern[1] elif p0 == 'VEXVALID=3': v.space = 'xop' opcode = pattern[1] else: # legacy maps and AMD 3dNow (if enabled) v.map, opcode = self._find_legacy_map_opcode(pattern) v.opcode = opcode v.partial_opcode = False v.amd_3dnow_opcode = None # conditional test avoids prefetches and FEMMS. if v.extension == '3DNOW' and v.category == '3DNOW': v.amd_3dnow_opcode = pattern[-1] mp = map_pattern.search(v.pattern) if mp: v.map = int(mp.group('map')) v.no_prefixes_allowed = False if no_prefix.search(v.pattern): v.no_prefixes_allowed = True v.osz_required = False osz = osz_prefix.search(v.pattern) if osz: if osz.group('prefix') == '1': v.osz_required = True v.f2_required = False v.f3_required = False rep = rep_prefix.search(v.pattern) if rep: if rep.group('prefix') == '2': v.f2_required = True elif rep.group('prefix') == '3': v.f3_required = True if v.space in ['evex', 'vex', 'xop']: vexp = vex_prefix.search(v.pattern) if vexp: if vexp.group('prefix') == '0': v.no_prefixes_allowed = True elif vexp.group('prefix') == '1': v.osz_required = True elif vexp.group('prefix') == '2': v.f2_required = True elif vexp.group('prefix') == '3': v.f3_required = True v.rexw_prefix = "unspecified" rexw = rexw_prefix.search(v.pattern) if rexw: v.rexw_prefix = rexw.group('rexw') # 0 or 1 v.reg_required = 'unspecified' reg = reg_required.search(v.pattern) if reg: v.reg_required = genutil.make_numeric(reg.group('reg')) v.rm_required = 'unspecified' rm = rm_required.search(v.pattern) # bit capture if rm: v.rm_required = genutil.make_numeric(rm.group('rm')) rm = rm_val_required.search(v.pattern) # equality if rm: v.rm_required = genutil.make_numeric(rm.group('rm')) v.mod_required = 'unspecified' mod = mod_required.search(v.pattern) if mod: v.mod_required = genutil.make_numeric(mod.group('mod')) mod = mod_mem_required.search(v.pattern) if mod: v.mod_required = '00/01/10' mod = mod_reg_required.search(v.pattern) if mod: v.mod_required = 3 v.has_modrm = False if 'MODRM' in v.pattern: # accounts for normal MODRM and various VSIB types v.has_modrm = True elif (v.reg_required != 'unspecified' or v.mod_required != 'unspecified'): v.has_modrm = True elif v.rm_required != 'unspecified': # avoid the partial opcode bytes which do not have MODRM if 'SRM[' not in v.pattern: v.has_modrm = True # 16/32/64b mode restrictions v.mode_restriction = 'unspecified' if not64_pattern.search(v.pattern): v.mode_restriction = 'not64' else: mode = mode_pattern.search(v.pattern) if mode: v.mode_restriction = int(mode.group('mode')) easz = 'aszall' if 'EASZ=1' in v.pattern: easz = 'a16' elif 'EASZ=2' in v.pattern: easz = 'a32' elif 'EASZ=3' in v.pattern: easz = 'a64' elif 'EASZ!=1' in v.pattern: easz = 'asznot16' v.easz = easz v.default_64b = False if 'DF64()' in v.pattern or 'CR_WIDTH()' in v.pattern: v.default_64b = True _set_eosz(v) v.scalar = False if hasattr(v, 'attributes'): v.attributes = v.attributes.upper() if 'SCALAR' in v.attributes: v.scalar = True else: v.attributes = '' if opcode.startswith('0x'): v.opcode_base10 = int(opcode, 16) elif opcode.startswith('0b'): # partial opcode.. 5 bits, shifted v.opcode_base10 = genutil.make_numeric(opcode) << 3 v.partial_opcode = True v.upper_nibble = int(v.opcode_base10 / 16) v.lower_nibble = v.opcode_base10 & 0xF
def parse_one_operand(w, default_vis='DEFAULT', xtypes=None, default_xtypes=None, internal=False): """Format examples: name=xxxxxy:{r,w,crw,rw,rcw}[:{EXPL,IMPL,SUPP,ECOND}][:{some oc2 code}][:{some xtype code}] name=NTLUR():{r,w,crw,rw,rcw}[:{EXPL,IMPL,SUPP,ECOND}][:{some oc2 code}][:{some xtype code}] oc2 can be before EXPL/IMPL/SUPP. oc2 is the width code. MEM{0,1}, PTR, RELBR, AGEN, IMM{0,1,2,3} xtype describes the number of data type and width of each element. If the xtype is omitted, xed will attempt to infer it from the oc2 code. ECOND is for encoder-only conditions. Completely ignored by the decoder. Default is read-only @param w: string @param w: an operand specification string @rtype operand_info_t @return a parsed operand """ if vopnd(): genutil.msge("PARSE-OPND: " + w) # get the r/w/rw info, if any vis = default_vis oc2 = None otype = None rw = 'r' cvt = [] invert = False lookupfn_name=None xtype = None multireg = 0 if colon_pattern.search(w): chunks = w.split(':') if vopnd(): genutil.msge("CHUNKS [%s]" % (",".join(chunks))) for i,c in enumerate(chunks): if vopnd(): genutil.msge("\tCHUNK %d %s" %( i,c)) if i == 0: a = c elif i == 1: rw = c if vopnd(): genutil.msge("\t\tSET rw to %s" % (rw)) elif (i == 2 or i == 3) and (c in ['IMPL', 'SUPP', 'EXPL', 'ECOND']): vis = c if vopnd(): genutil.msge("\t\tSET VIS to %s" % (vis)) else: # FIXME: somewhat sloppy error checking on input multi_reg_p=multireg_pattern.match(c) cp=convert_pattern.match(c) if multi_reg_p: multireg = int(multi_reg_p.group('nreg')) elif cp: cvt.append(cp.group('rhs')) elif oc2 == None and oc2_pattern.match(c): oc2 = c if vopnd(): genutil.msge("\t\tSET OC2 to %s" % (oc2)) elif oc2 and c in xtypes: xtype = c if vopnd(): genutil.msge("\t\tSET xtype to %s" % (xtype)) elif decimal_number_pattern.match(c): genutil.die("Bad number in %s" % (w)) else: genutil.die( "Bad oc2 pattern in %s when looking at %d chunk: %s " % (w,i,c) ) else: a = w if vis == 'ECOND': return None # From now on, use a, not w. if slash_pattern.search(a): genutil.die("Bad slash in operand") if xtype == None: # need to use the default xtype based on the oc2 width code if oc2: try: xtype = default_xtypes[oc2.upper()] except: s = '' for i,v in default_xtypes.items(): s += "\t%10s -> %10s\n" % (i,v) genutil.die("Parsing operand [%s]. Could not find default type for %s. xtypes=%s\nTypes=%s" % (w, oc2, str(xtypes), s)) else: # there was no oc2 type and no xtype. probably a nonterminal # lookup function xtype = 'INVALID' # look for X=y and X!=y and bare operands like MEM0. eqp = equals_pattern.search(a) neqp = not_equals_pattern.search(a) if eqp: (name,rhs) = eqp.group('lhs','rhs') if vopnd(): genutil.msge("PARSE-OPND:\t" + name + " + " + rhs) if double_parens_pattern.search(rhs): # NTLUF if vopnd(): genutil.msge("PARSE-OPND:\t nonterminal lookup function " + name + " <- " + rhs) # remove the parens nt_lookup_fn = double_parens_pattern.sub('',rhs) optype ='nt_lookup_fn' rhs = None lookupfn_name=nt_lookup_fn elif reg_pattern.match(rhs): optype = 'reg' elif error_pattern.match(rhs): optype = 'error' elif enum_pattern.match(rhs): # for storing XED_* enum values as RHS's of operand bindings optype = 'imm_const' elif (not genutil.numeric(rhs)) and az_cap_pattern.search(rhs): genutil.die("THIS SHOULD NOT HAPPEN: %s" % (rhs)) elif letters_underscore_pattern.match(rhs): rhs = list(rhs.replace('_','')) optype = 'imm' else: rhs = hex(genutil.make_numeric(rhs)) optype = 'imm_const' elif neqp: (name,rhs) = neqp.group('lhs','rhs') if vopnd(): genutil.msge("PARSE-OPND: (NOT EQUALS)\t" + name + " + " + rhs) invert = True if reg_pattern.match(rhs): optype = 'reg' elif az_cap_pattern.search(rhs): genutil.die("THIS SHOULD NOT HAPPEN") elif letters_underscore_pattern.match(rhs): genutil.die("Cannot have a != pattern with don't-care letters") else: rhs = hex(genutil.make_numeric(rhs)) optype = 'imm_const' elif mem_pattern.search(a): # memop name = a optype ='imm_const' rhs = '1' elif imm_token_pattern.search(a): # immediate placeholder name = a optype ='imm_const' rhs = '1' elif agen_pattern.search(a): # agen name = a optype ='imm_const' rhs = '1' elif relative_branch_pattern.search(a): name = a optype ='imm_const' rhs = '1' elif pointer_pattern.search(a): name = a optype ='imm_const' rhs = '1' elif xed_reset_pattern.search(a): # special marker that tells the traverser to restart this # nonterminal from the current position name = a optype ='xed_reset' rhs = '' vis = 'SUPP' elif double_parens_pattern.search(a): if vopnd(): genutil.msge("PARSE-OPND:\t unbound nonterminal lookup function " + a) # 2007-07-23 this code is not used genutil.die("UNBOUND NTLUF!: %s" % (a)) else: # macros -- these get rewritten later if vopnd(): genutil.msge("PARSE-OPND:\t flag-ish: " + a) name = a optype = 'flag' rhs = '' xop = operand_info_t(name, optype, rhs, rw=rw, invert=invert, vis=vis, oc2=oc2, cvt=cvt, xtype=xtype, lookupfn_name=lookupfn_name, internal=internal, multireg=multireg) return xop
def __init__(self, arg_action): # field bindings (FB) are OD=value self.type = None # 'FB', 'emit', 'nt', 'error', 'nothing', 'return' self.field_name = None self.value = None self.nt = None self.ntluf = None self.int_value = None self.emit_type = None # 'numeric', 'letters', 'reg' self.nbits = 0 if vaction(): msgb("ARGACTION", arg_action) if arg_action in ['nothing', "NOTHING"]: self.type = 'nothing' return b = patterns.return_pattern.search(arg_action) if b: self.type = 'return' self.value = b.group('retval') return # in the inputs, "error" gets expanded to "ERROR=1" via the statebits. if arg_action == 'error' or arg_action == "ERROR" or arg_action == 'ERROR=1': self.type = 'error' return b = patterns.bit_expand_pattern.search(arg_action) if b: expanded = b.group('bitname') * int(b.group('count')) action = patterns.bit_expand_pattern.sub(expanded,arg_action) else: action = arg_action #msgerr("CHECKING: %s" % action) a = patterns.equals_pattern.search(action) if a: # field binding #msgerr("FIELD BINDING: %s" % action) self.field_name = a.group('lhs') rhs = a.group('rhs') if patterns.decimal_pattern.match(rhs) or \ patterns.binary_pattern.match(rhs) or \ patterns.hex_pattern.match(rhs): self.int_value = genutil.make_numeric(rhs) self.value = str(self.int_value) #msgb("SET RHS", "%s -> %s" % (rhs,self.value)) else: self.value = rhs self.type = 'FB' return nt = patterns.nt_name_pattern.match(action) if nt: # NTLUF or NT. Only shows up on decode-oriented rules self.nt = nt.group('ntname') self.type = 'nt' return ntluf = patterns.ntluf_name_pattern.match(action) if ntluf: # NTLUF or NT. Only shows up on decode-oriented rules self.ntluf = ntluf.group('ntname') self.type = 'ntluf' return cp = patterns.lhs_capture_pattern_end.match(action) if cp: self.type = 'emit' self.value = cp.group('bits') self.field_name = cp.group('name').lower() #msgerr("EMIT ACTION %s" % action) self.classify() return # simple byte encoding self.type = 'emit' self.field_name = None #msgerr("EMIT ACTION %s" % action) self.value = action self.classify()
def _find_opcodes(self): '''augment the records with information found by parsing the pattern''' map_pattern = re.compile(r'MAP=(?P<map>[0-6])') vex_prefix = re.compile(r'VEX_PREFIX=(?P<prefix>[0-9])') rep_prefix = re.compile(r'REP=(?P<prefix>[0-3])') osz_prefix = re.compile(r'OSZ=(?P<prefix>[01])') no_prefix = re.compile(r'REP=0 OSZ=0') rexw_prefix = re.compile(r'REXW=(?P<rexw>[01])') reg_required = re.compile(r'REG[[](?P<reg>[b01]+)]') mod_required = re.compile(r'MOD[[](?P<mod>[b01]+)]') mod_mem_required = re.compile(r'MOD!=3') rm_required = re.compile(r'RM[[](?P<rm>[b01]+)]') mode_pattern = re.compile(r' MODE=(?P<mode>[012]+)') not64_pattern = re.compile(r' MODE!=2') for v in self.recs: if not hasattr(v, 'isa_set'): v.isa_set = v.extension v.undocumented = False if hasattr(v, 'comment'): if 'UNDOC' in v.comment: v.undocumented = True pattern = v.pattern.split() p0 = pattern[0] v.map = 0 v.space = 'legacy' if p0 in ['0x0F']: if pattern[1] == '0x38': v.map = 2 opcode = pattern[2] elif pattern[1] == '0x3A': v.map = 3 opcode = pattern[2] else: v.map = 1 opcode = pattern[1] elif p0 == 'VEXVALID=1': v.space = 'vex' opcode = pattern[1] elif p0 == 'VEXVALID=2': v.space = 'evex' opcode = pattern[1] elif p0 == 'VEXVALID=4': #KNC v.space = 'evex.u0' opcode = pattern[1] elif p0 == 'VEXVALID=3': v.space = 'xop' opcode = pattern[1] else: opcode = p0 v.opcode = opcode v.partial_opcode = False mp = map_pattern.search(v.pattern) if mp: v.map = int(mp.group('map')) v.no_prefixes_allowed = False if no_prefix.search(v.pattern): v.no_prefixes_allowed = True v.osz_required = False osz = osz_prefix.search(v.pattern) if osz: if osz.group('prefix') == '1': v.osz_required = True v.f2_required = False v.f3_required = False rep = rep_prefix.search(v.pattern) if rep: if rep.group('prefix') == '2': v.f2_required = True elif rep.group('prefix') == '3': v.f3_required = True if v.space in ['evex', 'vex', 'xop']: vexp = vex_prefix.search(v.pattern) if vexp: if vexp.group('prefix') == '0': v.no_prefixes_allowed = True elif vexp.group('prefix') == '1': v.osz_required = True elif vexp.group('prefix') == '2': v.f2_required = True elif vexp.group('prefix') == '3': v.f3_required = True v.rexw_prefix = "unspecified" rexw = rexw_prefix.search(v.pattern) if rexw: v.rexw_prefix = rexw.group('rexw') # 0 or 1 v.reg_required = 'unspecified' reg = reg_required.search(v.pattern) if reg: v.reg_required = genutil.make_numeric(reg.group('reg')) v.rm_required = 'unspecified' rm = rm_required.search(v.pattern) if rm: v.rm_required = genutil.make_numeric(rm.group('rm')) v.mod_required = 'unspecified' mod = mod_required.search(v.pattern) if mod: v.mod_required = genutil.make_numeric(mod.group('mod')) mod = mod_mem_required.search(v.pattern) if mod: v.mod_required = '00/01/10' # 16/32/64b mode restrictions v.mode_restriction = 'unspecified' if not64_pattern.search(v.pattern): v.mode_restriction = 'not64' else: mode = mode_pattern.search(v.pattern) if mode: v.mode_restriction = int(mode.group('mode')) v.scalar = False if hasattr(v, 'attributes'): v.attributes = v.attributes.upper() if 'SCALAR' in v.attributes: v.scalar = True if opcode.startswith('0x'): nopcode = int(opcode, 16) elif opcode.startswith('0b'): # partial opcode.. 5 bits, shifted nopcode = genutil.make_numeric(opcode) << 3 v.partial_opcode = True v.upper_nibble = int(nopcode / 16) v.lower_nibble = nopcode & 0xF
def _parse_map_line(s): global _map_info_fields # shlex allows for quoted substrings containing spaces as # individual args. t = shlex.split(s.strip()) if len(t) != len(_map_info_fields): _die("Bad map description line: [{}]".format(s)) mi = map_info_t() for i, fld in enumerate(_map_info_fields): setattr(mi, fld, t[i]) # this gets used in function names so must only be legal characters mi.map_name = re.sub('-', '_', mi.map_name) if mi.space == 'legacy': if mi.legacy_escape != 'N/A': mi.legacy_escape_int = int(mi.legacy_escape, 16) if mi.legacy_opcode != 'N/A': mi.legacy_opcode_int = int(mi.legacy_opcode, 16) else: mi.legacy_opcode_int = None mi.map_id_fixup = False if mi.space not in ['legacy', 'vex', 'evex', 'xop', 'knc']: _die("Bad map description encoding space [{}]".format(s)) if mi.space == 'legacy': if genutil.is_hex(mi.legacy_escape): pass elif mi.legacy_escape != 'N/A': _die("Bad map description legacy escape [{}]".format(s)) if genutil.is_hex(mi.legacy_opcode): pass elif mi.legacy_opcode != 'N/A': _die("Bad map description legacy opcode [{}]".format(s)) if mi.map_id == 'N/A': _die("Bad map description map-id [{}]".format(s)) elif genutil.numeric(mi.map_id): mi.map_id = genutil.make_numeric(mi.map_id) else: mi.map_id_fixup = True else: if mi.legacy_escape != 'N/A': _die("Bad map description legacy escape [{}]".format(s)) if mi.legacy_opcode != 'N/A': _die("Bad map description legacy opcode [{}]".format(s)) if genutil.numeric(mi.map_id): mi.map_id = genutil.make_numeric(mi.map_id) else: _die("Bad map description map id [{}]".format(s)) if mi.disp not in ['var', 'no']: _die("Bad map description disp specifier [{}]".format(s)) if mi.modrm not in ['var', 'yes', 'no']: _die("Bad map description modrm specifier [{}]".format(s)) if mi.imm not in ['var', '0', '1', '2', '4']: _die("Bad map description imm specifier [{}]".format(s)) if genutil.numeric(mi.opcpos): mi.opcpos = genutil.make_numeric(mi.opcpos) else: _die("Bad map description opcode position specifier [{}]".format(s)) # we want the longer patterns first when we sort the map_info_t. mi.priority = 100 - len(mi.search_pattern) return mi