1#!/bin/awk -f 2# gen-insn-attr-x86.awk: Instruction attribute table generator 3# Written by Masami Hiramatsu <mhiramat@redhat.com> 4# 5# Usage: awk -f gen-insn-attr-x86.awk x86-opcode-map.txt > inat-tables.c 6 7# Awk implementation sanity check 8function check_awk_implement() { 9 if (sprintf("%x", 0) != "0") 10 return "Your awk has a printf-format problem." 11 return "" 12} 13 14# Clear working vars 15function clear_vars() { 16 delete table 17 delete lptable2 18 delete lptable1 19 delete lptable3 20 eid = -1 # escape id 21 gid = -1 # group id 22 aid = -1 # AVX id 23 tname = "" 24} 25 26BEGIN { 27 # Implementation error checking 28 awkchecked = check_awk_implement() 29 if (awkchecked != "") { 30 print "Error: " awkchecked > "/dev/stderr" 31 print "Please try to use gawk." > "/dev/stderr" 32 exit 1 33 } 34 35 # Setup generating tables 36 print "/* x86 opcode map generated from x86-opcode-map.txt */" 37 print "/* Do not change this code. */\n" 38 ggid = 1 39 geid = 1 40 gaid = 0 41 delete etable 42 delete gtable 43 delete atable 44 45 opnd_expr = "^[A-Za-z/]" 46 ext_expr = "^\\(" 47 sep_expr = "^\\|$" 48 group_expr = "^Grp[0-9A-Za-z]+" 49 50 imm_expr = "^[IJAOL][a-z]" 51 imm_flag["Ib"] = "INAT_MAKE_IMM(INAT_IMM_BYTE)" 52 imm_flag["Jb"] = "INAT_MAKE_IMM(INAT_IMM_BYTE)" 53 imm_flag["Iw"] = "INAT_MAKE_IMM(INAT_IMM_WORD)" 54 imm_flag["Id"] = "INAT_MAKE_IMM(INAT_IMM_DWORD)" 55 imm_flag["Iq"] = "INAT_MAKE_IMM(INAT_IMM_QWORD)" 56 imm_flag["Ap"] = "INAT_MAKE_IMM(INAT_IMM_PTR)" 57 imm_flag["Iz"] = "INAT_MAKE_IMM(INAT_IMM_VWORD32)" 58 imm_flag["Jz"] = "INAT_MAKE_IMM(INAT_IMM_VWORD32)" 59 imm_flag["Iv"] = "INAT_MAKE_IMM(INAT_IMM_VWORD)" 60 imm_flag["Ob"] = "INAT_MOFFSET" 61 imm_flag["Ov"] = "INAT_MOFFSET" 62 imm_flag["Lx"] = "INAT_MAKE_IMM(INAT_IMM_BYTE)" 63 64 modrm_expr = "^([CDEGMNPQRSUVW/][a-z]+|NTA|T[012])" 65 force64_expr = "\\([df]64\\)" 66 rex_expr = "^REX(\\.[XRWB]+)*" 67 fpu_expr = "^ESC" # TODO 68 69 lprefix1_expr = "\\((66|!F3)\\)" 70 lprefix2_expr = "\\(F3\\)" 71 lprefix3_expr = "\\((F2|!F3|66&F2)\\)" 72 lprefix_expr = "\\((66|F2|F3)\\)" 73 max_lprefix = 4 74 75 # All opcodes starting with lower-case 'v' or with (v1) superscript 76 # accepts VEX prefix 77 vexok_opcode_expr = "^v.*" 78 vexok_expr = "\\(v1\\)" 79 # All opcodes with (v) superscript supports *only* VEX prefix 80 vexonly_expr = "\\(v\\)" 81 82 prefix_expr = "\\(Prefix\\)" 83 prefix_num["Operand-Size"] = "INAT_PFX_OPNDSZ" 84 prefix_num["REPNE"] = "INAT_PFX_REPNE" 85 prefix_num["REP/REPE"] = "INAT_PFX_REPE" 86 prefix_num["XACQUIRE"] = "INAT_PFX_REPNE" 87 prefix_num["XRELEASE"] = "INAT_PFX_REPE" 88 prefix_num["LOCK"] = "INAT_PFX_LOCK" 89 prefix_num["SEG=CS"] = "INAT_PFX_CS" 90 prefix_num["SEG=DS"] = "INAT_PFX_DS" 91 prefix_num["SEG=ES"] = "INAT_PFX_ES" 92 prefix_num["SEG=FS"] = "INAT_PFX_FS" 93 prefix_num["SEG=GS"] = "INAT_PFX_GS" 94 prefix_num["SEG=SS"] = "INAT_PFX_SS" 95 prefix_num["Address-Size"] = "INAT_PFX_ADDRSZ" 96 prefix_num["VEX+1byte"] = "INAT_PFX_VEX2" 97 prefix_num["VEX+2byte"] = "INAT_PFX_VEX3" 98 99 clear_vars() 100} 101 102function semantic_error(msg) { 103 print "Semantic error at " NR ": " msg > "/dev/stderr" 104 exit 1 105} 106 107function debug(msg) { 108 print "DEBUG: " msg 109} 110 111function array_size(arr, i,c) { 112 c = 0 113 for (i in arr) 114 c++ 115 return c 116} 117 118/^Table:/ { 119 print "/* " $0 " */" 120 if (tname != "") 121 semantic_error("Hit Table: before EndTable:."); 122} 123 124/^Referrer:/ { 125 if (NF != 1) { 126 # escape opcode table 127 ref = "" 128 for (i = 2; i <= NF; i++) 129 ref = ref $i 130 eid = escape[ref] 131 tname = sprintf("inat_escape_table_%d", eid) 132 } 133} 134 135/^AVXcode:/ { 136 if (NF != 1) { 137 # AVX/escape opcode table 138 aid = $2 139 if (gaid <= aid) 140 gaid = aid + 1 141 if (tname == "") # AVX only opcode table 142 tname = sprintf("inat_avx_table_%d", $2) 143 } 144 if (aid == -1 && eid == -1) # primary opcode table 145 tname = "inat_primary_table" 146} 147 148/^GrpTable:/ { 149 print "/* " $0 " */" 150 if (!($2 in group)) 151 semantic_error("No group: " $2 ) 152 gid = group[$2] 153 tname = "inat_group_table_" gid 154} 155 156function print_table(tbl,name,fmt,n) 157{ 158 print "const insn_attr_t " name " = {" 159 for (i = 0; i < n; i++) { 160 id = sprintf(fmt, i) 161 if (tbl[id]) 162 print " [" id "] = " tbl[id] "," 163 } 164 print "};" 165} 166 167/^EndTable/ { 168 if (gid != -1) { 169 # print group tables 170 if (array_size(table) != 0) { 171 print_table(table, tname "[INAT_GROUP_TABLE_SIZE]", 172 "0x%x", 8) 173 gtable[gid,0] = tname 174 } 175 if (array_size(lptable1) != 0) { 176 print_table(lptable1, tname "_1[INAT_GROUP_TABLE_SIZE]", 177 "0x%x", 8) 178 gtable[gid,1] = tname "_1" 179 } 180 if (array_size(lptable2) != 0) { 181 print_table(lptable2, tname "_2[INAT_GROUP_TABLE_SIZE]", 182 "0x%x", 8) 183 gtable[gid,2] = tname "_2" 184 } 185 if (array_size(lptable3) != 0) { 186 print_table(lptable3, tname "_3[INAT_GROUP_TABLE_SIZE]", 187 "0x%x", 8) 188 gtable[gid,3] = tname "_3" 189 } 190 } else { 191 # print primary/escaped tables 192 if (array_size(table) != 0) { 193 print_table(table, tname "[INAT_OPCODE_TABLE_SIZE]", 194 "0x%02x", 256) 195 etable[eid,0] = tname 196 if (aid >= 0) 197 atable[aid,0] = tname 198 } 199 if (array_size(lptable1) != 0) { 200 print_table(lptable1,tname "_1[INAT_OPCODE_TABLE_SIZE]", 201 "0x%02x", 256) 202 etable[eid,1] = tname "_1" 203 if (aid >= 0) 204 atable[aid,1] = tname "_1" 205 } 206 if (array_size(lptable2) != 0) { 207 print_table(lptable2,tname "_2[INAT_OPCODE_TABLE_SIZE]", 208 "0x%02x", 256) 209 etable[eid,2] = tname "_2" 210 if (aid >= 0) 211 atable[aid,2] = tname "_2" 212 } 213 if (array_size(lptable3) != 0) { 214 print_table(lptable3,tname "_3[INAT_OPCODE_TABLE_SIZE]", 215 "0x%02x", 256) 216 etable[eid,3] = tname "_3" 217 if (aid >= 0) 218 atable[aid,3] = tname "_3" 219 } 220 } 221 print "" 222 clear_vars() 223} 224 225function add_flags(old,new) { 226 if (old && new) 227 return old " | " new 228 else if (old) 229 return old 230 else 231 return new 232} 233 234# convert operands to flags. 235function convert_operands(count,opnd, i,j,imm,mod) 236{ 237 imm = null 238 mod = null 239 for (j = 1; j <= count; j++) { 240 i = opnd[j] 241 if (match(i, imm_expr) == 1) { 242 if (!imm_flag[i]) 243 semantic_error("Unknown imm opnd: " i) 244 if (imm) { 245 if (i != "Ib") 246 semantic_error("Second IMM error") 247 imm = add_flags(imm, "INAT_SCNDIMM") 248 } else 249 imm = imm_flag[i] 250 } else if (match(i, modrm_expr)) 251 mod = "INAT_MODRM" 252 } 253 return add_flags(imm, mod) 254} 255 256/^[0-9a-f]+:/ { 257 if (NR == 1) 258 next 259 # get index 260 idx = "0x" substr($1, 1, index($1,":") - 1) 261 if (idx in table) 262 semantic_error("Redefine " idx " in " tname) 263 264 # check if escaped opcode 265 if ("escape" == $2) { 266 if ($3 != "#") 267 semantic_error("No escaped name") 268 ref = "" 269 for (i = 4; i <= NF; i++) 270 ref = ref $i 271 if (ref in escape) 272 semantic_error("Redefine escape (" ref ")") 273 escape[ref] = geid 274 geid++ 275 table[idx] = "INAT_MAKE_ESCAPE(" escape[ref] ")" 276 next 277 } 278 279 variant = null 280 # converts 281 i = 2 282 while (i <= NF) { 283 opcode = $(i++) 284 delete opnds 285 ext = null 286 flags = null 287 opnd = null 288 # parse one opcode 289 if (match($i, opnd_expr)) { 290 opnd = $i 291 count = split($(i++), opnds, ",") 292 flags = convert_operands(count, opnds) 293 } 294 if (match($i, ext_expr)) 295 ext = $(i++) 296 if (match($i, sep_expr)) 297 i++ 298 else if (i < NF) 299 semantic_error($i " is not a separator") 300 301 # check if group opcode 302 if (match(opcode, group_expr)) { 303 if (!(opcode in group)) { 304 group[opcode] = ggid 305 ggid++ 306 } 307 flags = add_flags(flags, "INAT_MAKE_GROUP(" group[opcode] ")") 308 } 309 # check force(or default) 64bit 310 if (match(ext, force64_expr)) 311 flags = add_flags(flags, "INAT_FORCE64") 312 313 # check REX prefix 314 if (match(opcode, rex_expr)) 315 flags = add_flags(flags, "INAT_MAKE_PREFIX(INAT_PFX_REX)") 316 317 # check coprocessor escape : TODO 318 if (match(opcode, fpu_expr)) 319 flags = add_flags(flags, "INAT_MODRM") 320 321 # check VEX codes 322 if (match(ext, vexonly_expr)) 323 flags = add_flags(flags, "INAT_VEXOK | INAT_VEXONLY") 324 else if (match(ext, vexok_expr) || match(opcode, vexok_opcode_expr)) 325 flags = add_flags(flags, "INAT_VEXOK") 326 327 # check prefixes 328 if (match(ext, prefix_expr)) { 329 if (!prefix_num[opcode]) 330 semantic_error("Unknown prefix: " opcode) 331 flags = add_flags(flags, "INAT_MAKE_PREFIX(" prefix_num[opcode] ")") 332 } 333 if (length(flags) == 0) 334 continue 335 # check if last prefix 336 if (match(ext, lprefix1_expr)) { 337 lptable1[idx] = add_flags(lptable1[idx],flags) 338 variant = "INAT_VARIANT" 339 } 340 if (match(ext, lprefix2_expr)) { 341 lptable2[idx] = add_flags(lptable2[idx],flags) 342 variant = "INAT_VARIANT" 343 } 344 if (match(ext, lprefix3_expr)) { 345 lptable3[idx] = add_flags(lptable3[idx],flags) 346 variant = "INAT_VARIANT" 347 } 348 if (!match(ext, lprefix_expr)){ 349 table[idx] = add_flags(table[idx],flags) 350 } 351 } 352 if (variant) 353 table[idx] = add_flags(table[idx],variant) 354} 355 356END { 357 if (awkchecked != "") 358 exit 1 359 # print escape opcode map's array 360 print "/* Escape opcode map array */" 361 print "const insn_attr_t * const inat_escape_tables[INAT_ESC_MAX + 1]" \ 362 "[INAT_LSTPFX_MAX + 1] = {" 363 for (i = 0; i < geid; i++) 364 for (j = 0; j < max_lprefix; j++) 365 if (etable[i,j]) 366 print " ["i"]["j"] = "etable[i,j]"," 367 print "};\n" 368 # print group opcode map's array 369 print "/* Group opcode map array */" 370 print "const insn_attr_t * const inat_group_tables[INAT_GRP_MAX + 1]"\ 371 "[INAT_LSTPFX_MAX + 1] = {" 372 for (i = 0; i < ggid; i++) 373 for (j = 0; j < max_lprefix; j++) 374 if (gtable[i,j]) 375 print " ["i"]["j"] = "gtable[i,j]"," 376 print "};\n" 377 # print AVX opcode map's array 378 print "/* AVX opcode map array */" 379 print "const insn_attr_t * const inat_avx_tables[X86_VEX_M_MAX + 1]"\ 380 "[INAT_LSTPFX_MAX + 1] = {" 381 for (i = 0; i < gaid; i++) 382 for (j = 0; j < max_lprefix; j++) 383 if (atable[i,j]) 384 print " ["i"]["j"] = "atable[i,j]"," 385 print "};" 386} 387 388