1#! /usr/bin/env perl 2# Copyright 2005-2024 The OpenSSL Project Authors. All Rights Reserved. 3# 4# Licensed under the Apache License 2.0 (the "License"). You may not use 5# this file except in compliance with the License. You can obtain a copy 6# in the file LICENSE in the source distribution or at 7# https://www.openssl.org/source/license.html 8 9 10# Ascetic x86_64 AT&T to MASM/NASM assembler translator by <appro>. 11# 12# Why AT&T to MASM and not vice versa? Several reasons. Because AT&T 13# format is way easier to parse. Because it's simpler to "gear" from 14# Unix ABI to Windows one [see cross-reference "card" at the end of 15# file]. Because Linux targets were available first... 16# 17# In addition the script also "distills" code suitable for GNU 18# assembler, so that it can be compiled with more rigid assemblers, 19# such as Solaris /usr/ccs/bin/as. 20# 21# This translator is not designed to convert *arbitrary* assembler 22# code from AT&T format to MASM one. It's designed to convert just 23# enough to provide for dual-ABI OpenSSL modules development... 24# There *are* limitations and you might have to modify your assembler 25# code or this script to achieve the desired result... 26# 27# Currently recognized limitations: 28# 29# - can't use multiple ops per line; 30# 31# Dual-ABI styling rules. 32# 33# 1. Adhere to Unix register and stack layout [see cross-reference 34# ABI "card" at the end for explanation]. 35# 2. Forget about "red zone," stick to more traditional blended 36# stack frame allocation. If volatile storage is actually required 37# that is. If not, just leave the stack as is. 38# 3. Functions tagged with ".type name,@function" get crafted with 39# unified Win64 prologue and epilogue automatically. If you want 40# to take care of ABI differences yourself, tag functions as 41# ".type name,@abi-omnipotent" instead. 42# 4. To optimize the Win64 prologue you can specify number of input 43# arguments as ".type name,@function,N." Keep in mind that if N is 44# larger than 6, then you *have to* write "abi-omnipotent" code, 45# because >6 cases can't be addressed with unified prologue. 46# 5. Name local labels as .L*, do *not* use dynamic labels such as 1: 47# (sorry about latter). 48# 6. Don't use [or hand-code with .byte] "rep ret." "ret" mnemonic is 49# required to identify the spots, where to inject Win64 epilogue! 50# But on the pros, it's then prefixed with rep automatically:-) 51# 7. Stick to explicit ip-relative addressing. If you have to use 52# GOTPCREL addressing, stick to mov symbol@GOTPCREL(%rip),%r??. 53# Both are recognized and translated to proper Win64 addressing 54# modes. 55# 56# 8. In order to provide for structured exception handling unified 57# Win64 prologue copies %rsp value to %rax. For further details 58# see SEH paragraph at the end. 59# 9. .init segment is allowed to contain calls to functions only. 60# a. If function accepts more than 4 arguments *and* >4th argument 61# is declared as non 64-bit value, do clear its upper part. 62 63 64use strict; 65 66my $flavour = shift; 67my $output = shift; 68if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } 69 70open STDOUT,">$output" || die "can't open $output: $!" 71 if (defined($output)); 72 73my $gas=1; $gas=0 if ($output =~ /\.asm$/); 74my $elf=1; $elf=0 if (!$gas); 75my $win64=0; 76my $prefix=""; 77my $decor=".L"; 78 79my $masmref=8 + 50727*2**-32; # 8.00.50727 shipped with VS2005 80my $masm=0; 81my $PTR=" PTR"; 82 83my $nasmref=2.03; 84my $nasm=0; 85 86# GNU as indicator, as opposed to $gas, which indicates acceptable 87# syntax 88my $gnuas=0; 89 90if ($flavour eq "mingw64") { $gas=1; $elf=0; $win64=1; 91 $prefix=`echo __USER_LABEL_PREFIX__ | $ENV{CC} -E -P -`; 92 $prefix =~ s|\R$||; # Better chomp 93 } 94elsif ($flavour eq "macosx") { $gas=1; $elf=0; $prefix="_"; $decor="L\$"; } 95elsif ($flavour eq "masm") { $gas=0; $elf=0; $masm=$masmref; $win64=1; $decor="\$L\$"; } 96elsif ($flavour eq "nasm") { $gas=0; $elf=0; $nasm=$nasmref; $win64=1; $decor="\$L\$"; $PTR=""; } 97elsif (!$gas) 98{ if ($ENV{ASM} =~ m/nasm/ && `nasm -v` =~ m/version ([0-9]+)\.([0-9]+)/i) 99 { $nasm = $1 + $2*0.01; $PTR=""; } 100 elsif (`ml64 2>&1` =~ m/Version ([0-9]+)\.([0-9]+)(\.([0-9]+))?/) 101 { $masm = $1 + $2*2**-16 + $4*2**-32; } 102 die "no assembler found on %PATH%" if (!($nasm || $masm)); 103 $win64=1; 104 $elf=0; 105 $decor="\$L\$"; 106} 107# Find out if we're using GNU as 108elsif (`$ENV{CC} -Wa,-v -c -o /dev/null -x assembler /dev/null 2>&1` 109 =~ /GNU assembler version ([2-9]\.[0-9]+)/) 110{ 111 $gnuas=1; 112} 113elsif (`$ENV{CC} --version 2>/dev/null` 114 =~ /(clang .*|Intel.*oneAPI .*)/) 115{ 116 $gnuas=1; 117} 118elsif (`$ENV{CC} -V 2>/dev/null` 119 =~ /nvc .*/) 120{ 121 $gnuas=1; 122} 123 124my $cet_property; 125if ($flavour =~ /elf/) { 126 # Always generate .note.gnu.property section for ELF outputs to 127 # mark Intel CET support since all input files must be marked 128 # with Intel CET support in order for linker to mark output with 129 # Intel CET support. 130 my $p2align=3; $p2align=2 if ($flavour eq "elf32"); 131 my $section='.note.gnu.property, #alloc'; 132 $section='".note.gnu.property", "a"' if $gnuas; 133 $cet_property = <<_____; 134 .section $section 135 .p2align $p2align 136 .long 1f - 0f 137 .long 4f - 1f 138 .long 5 1390: 140 # "GNU" encoded with .byte, since .asciz isn't supported 141 # on Solaris. 142 .byte 0x47 143 .byte 0x4e 144 .byte 0x55 145 .byte 0 1461: 147 .p2align $p2align 148 .long 0xc0000002 149 .long 3f - 2f 1502: 151 .long 3 1523: 153 .p2align $p2align 1544: 155_____ 156} 157 158my $current_segment; 159# 160# I could not find equivalent of .previous directive for MASM (Microsoft 161# assembler ML). Using of .previous got introduced to .pl files with 162# placing of various constants into .rodata sections (segments). 163# Each .rodata section is terminated by .previous directive which 164# restores the preceding section to .rodata: 165# 166# .text 167# ; this is is the text section/segment 168# .rodata 169# ; constant definitions go here 170# .previous 171# ; the .text section which precedes .rodata got restored here 172# 173# The equivalent form for masm reads as follows: 174# 175# .text$ SEGMENT ALIGN(256) 'CODE' 176# ; this is is the text section/segment 177# .text$ ENDS 178# .rdata SEGMENT READONLY ALIGN(64) 179# ; constant definitions go here 180# .rdata$ ENDS 181# .text$ SEGMENT ALIGN(256) 'CODE' 182# ; text section follows 183# .text$ ENDS 184# 185# The .previous directive typically terminates .roadata segments/sections which 186# hold definitions of constants. In order to place constants into .rdata 187# segments when using masm we need to introduce a segment_stack array so we can 188# emit proper ENDS directive whenever we see .previous. 189# 190# The code is tailored to work current set of .pl/asm files. There are some 191# inconsistencies. For example .text section is the first section in all those 192# files except ecp_nistz256. So we need to take that into account. 193# 194# ; stack is empty 195# .text 196# ; push '.text ' section twice, the stack looks as 197# ; follows: 198# ; ('.text', '.text') 199# .rodata 200# ; pop() so we can generate proper 'ENDS' for masm. 201# ; stack looks like: 202# ; ('.text') 203# ; push '.rodata', so we can create corresponding ENDS for masm. 204# ; stack looks like: 205# ; ('.rodata', '.text') 206# .previous 207# ; pop() '.rodata' from stack, so we create '.rodata ENDS' 208# ; in masm flavour. For nasm flavour we just pop() because 209# ; nasm does not use .rodata ENDS to close the current section 210# ; the stack content is like this: 211# ; ('.text', '.text') 212# ; pop() again to find a previous section we need to restore. 213# ; Depending on flavour we either generate .section .text 214# ; or .text SEGMENT. The stack looks like: 215# ; ('.text') 216# 217my @segment_stack = (); 218my $current_function; 219my %globals; 220 221{ package opcode; # pick up opcodes 222 sub re { 223 my ($class, $line) = @_; 224 my $self = {}; 225 my $ret; 226 227 if ($$line =~ /^([a-z][a-z0-9]*)/i) { 228 bless $self,$class; 229 $self->{op} = $1; 230 $ret = $self; 231 $$line = substr($$line,@+[0]); $$line =~ s/^\s+//; 232 233 undef $self->{sz}; 234 if ($self->{op} =~ /^(movz)x?([bw]).*/) { # movz is pain... 235 $self->{op} = $1; 236 $self->{sz} = $2; 237 } elsif ($self->{op} =~ /call|jmp/) { 238 $self->{sz} = ""; 239 } elsif ($self->{op} =~ /^p/ && $' !~ /^(ush|op|insrw)/) { # SSEn 240 $self->{sz} = ""; 241 } elsif ($self->{op} =~ /^[vk]/) { # VEX or k* such as kmov 242 $self->{sz} = ""; 243 } elsif ($self->{op} =~ /mov[dq]/ && $$line =~ /%xmm/) { 244 $self->{sz} = ""; 245 } elsif ($self->{op} =~ /([a-z]{3,})([qlwb])$/) { 246 $self->{op} = $1; 247 $self->{sz} = $2; 248 } 249 } 250 $ret; 251 } 252 sub size { 253 my ($self, $sz) = @_; 254 $self->{sz} = $sz if (defined($sz) && !defined($self->{sz})); 255 $self->{sz}; 256 } 257 sub out { 258 my $self = shift; 259 if ($gas) { 260 if ($self->{op} eq "movz") { # movz is pain... 261 sprintf "%s%s%s",$self->{op},$self->{sz},shift; 262 } elsif ($self->{op} =~ /^set/) { 263 "$self->{op}"; 264 } elsif ($self->{op} eq "ret") { 265 my $epilogue = ""; 266 if ($win64 && $current_function->{abi} eq "svr4") { 267 $epilogue = "movq 8(%rsp),%rdi\n\t" . 268 "movq 16(%rsp),%rsi\n\t"; 269 } 270 $epilogue . ".byte 0xf3,0xc3"; 271 } elsif ($self->{op} eq "call" && !$elf && $current_segment eq ".init") { 272 ".p2align\t3\n\t.quad"; 273 } else { 274 "$self->{op}$self->{sz}"; 275 } 276 } else { 277 $self->{op} =~ s/^movz/movzx/; 278 if ($self->{op} eq "ret") { 279 $self->{op} = ""; 280 if ($win64 && $current_function->{abi} eq "svr4") { 281 $self->{op} = "mov rdi,QWORD$PTR\[8+rsp\]\t;WIN64 epilogue\n\t". 282 "mov rsi,QWORD$PTR\[16+rsp\]\n\t"; 283 } 284 $self->{op} .= "DB\t0F3h,0C3h\t\t;repret"; 285 } elsif ($self->{op} =~ /^(pop|push)f/) { 286 $self->{op} .= $self->{sz}; 287 } elsif ($self->{op} eq "call" && $current_segment eq ".CRT\$XCU") { 288 $self->{op} = "\tDQ"; 289 } 290 $self->{op}; 291 } 292 } 293 sub mnemonic { 294 my ($self, $op) = @_; 295 $self->{op}=$op if (defined($op)); 296 $self->{op}; 297 } 298} 299{ package const; # pick up constants, which start with $ 300 sub re { 301 my ($class, $line) = @_; 302 my $self = {}; 303 my $ret; 304 305 if ($$line =~ /^\$([^,]+)/) { 306 bless $self, $class; 307 $self->{value} = $1; 308 $ret = $self; 309 $$line = substr($$line,@+[0]); $$line =~ s/^\s+//; 310 } 311 $ret; 312 } 313 sub out { 314 my $self = shift; 315 316 $self->{value} =~ s/\b(0b[0-1]+)/oct($1)/eig; 317 if ($gas) { 318 # Solaris /usr/ccs/bin/as can't handle multiplications 319 # in $self->{value} 320 my $value = $self->{value}; 321 no warnings; # oct might complain about overflow, ignore here... 322 $value =~ s/(?<![\w\$\.])(0x?[0-9a-f]+)/oct($1)/egi; 323 if ($value =~ s/([0-9]+\s*[\*\/\%]\s*[0-9]+)/eval($1)/eg) { 324 $self->{value} = $value; 325 } 326 sprintf "\$%s",$self->{value}; 327 } else { 328 my $value = $self->{value}; 329 $value =~ s/0x([0-9a-f]+)/0$1h/ig if ($masm); 330 sprintf "%s",$value; 331 } 332 } 333} 334{ package ea; # pick up effective addresses: expr(%reg,%reg,scale) 335 336 my %szmap = ( b=>"BYTE$PTR", w=>"WORD$PTR", 337 l=>"DWORD$PTR", d=>"DWORD$PTR", 338 q=>"QWORD$PTR", o=>"OWORD$PTR", 339 x=>"XMMWORD$PTR", y=>"YMMWORD$PTR", 340 z=>"ZMMWORD$PTR" ) if (!$gas); 341 342 sub re { 343 my ($class, $line, $opcode) = @_; 344 my $self = {}; 345 my $ret; 346 347 # optional * ----vvv--- appears in indirect jmp/call 348 if ($$line =~ /^(\*?)([^\(,]*)\(([%\w,]+)\)((?:{[^}]+})*)/) { 349 bless $self, $class; 350 $self->{asterisk} = $1; 351 $self->{label} = $2; 352 ($self->{base},$self->{index},$self->{scale})=split(/,/,$3); 353 $self->{scale} = 1 if (!defined($self->{scale})); 354 $self->{opmask} = $4; 355 $ret = $self; 356 $$line = substr($$line,@+[0]); $$line =~ s/^\s+//; 357 358 if ($win64 && $self->{label} =~ s/\@GOTPCREL//) { 359 die if ($opcode->mnemonic() ne "mov"); 360 $opcode->mnemonic("lea"); 361 } 362 $self->{base} =~ s/^%//; 363 $self->{index} =~ s/^%// if (defined($self->{index})); 364 $self->{opcode} = $opcode; 365 } 366 $ret; 367 } 368 sub size {} 369 sub out { 370 my ($self, $sz) = @_; 371 372 $self->{label} =~ s/([_a-z][_a-z0-9]*)/$globals{$1} or $1/gei; 373 $self->{label} =~ s/\.L/$decor/g; 374 375 # Silently convert all EAs to 64-bit. This is required for 376 # elder GNU assembler and results in more compact code, 377 # *but* most importantly AES module depends on this feature! 378 $self->{index} =~ s/^[er](.?[0-9xpi])[d]?$/r\1/; 379 $self->{base} =~ s/^[er](.?[0-9xpi])[d]?$/r\1/; 380 381 # Solaris /usr/ccs/bin/as can't handle multiplications 382 # in $self->{label}... 383 use integer; 384 $self->{label} =~ s/(?<![\w\$\.])(0x?[0-9a-f]+)/oct($1)/egi; 385 $self->{label} =~ s/\b([0-9]+\s*[\*\/\%]\s*[0-9]+)\b/eval($1)/eg; 386 387 # Some assemblers insist on signed presentation of 32-bit 388 # offsets, but sign extension is a tricky business in perl... 389 if ((1<<31)<<1) { 390 $self->{label} =~ s/\b([0-9]+)\b/$1<<32>>32/eg; 391 } else { 392 $self->{label} =~ s/\b([0-9]+)\b/$1>>0/eg; 393 } 394 395 # if base register is %rbp or %r13, see if it's possible to 396 # flip base and index registers [for better performance] 397 if (!$self->{label} && $self->{index} && $self->{scale}==1 && 398 $self->{base} =~ /(rbp|r13)/) { 399 $self->{base} = $self->{index}; $self->{index} = $1; 400 } 401 402 if ($gas) { 403 $self->{label} =~ s/^___imp_/__imp__/ if ($flavour eq "mingw64"); 404 405 if (defined($self->{index})) { 406 sprintf "%s%s(%s,%%%s,%d)%s", 407 $self->{asterisk},$self->{label}, 408 $self->{base}?"%$self->{base}":"", 409 $self->{index},$self->{scale}, 410 $self->{opmask}; 411 } else { 412 sprintf "%s%s(%%%s)%s", $self->{asterisk},$self->{label}, 413 $self->{base},$self->{opmask}; 414 } 415 } else { 416 $self->{label} =~ s/\./\$/g; 417 $self->{label} =~ s/(?<![\w\$\.])0x([0-9a-f]+)/0$1h/ig; 418 $self->{label} = "($self->{label})" if ($self->{label} =~ /[\*\+\-\/]/); 419 420 my $mnemonic = $self->{opcode}->mnemonic(); 421 ($self->{asterisk}) && ($sz="q") || 422 ($mnemonic =~ /^v?mov([qd])$/) && ($sz=$1) || 423 ($mnemonic =~ /^v?pinsr([qdwb])$/) && ($sz=$1) || 424 ($mnemonic =~ /^vpbroadcast([qdwb])$/) && ($sz=$1) || 425 ($mnemonic =~ /^v(?!perm)[a-z]+[fi]128$/) && ($sz="x"); 426 427 $self->{opmask} =~ s/%(k[0-7])/$1/; 428 429 if (defined($self->{index})) { 430 sprintf "%s[%s%s*%d%s]%s",$szmap{$sz}, 431 $self->{label}?"$self->{label}+":"", 432 $self->{index},$self->{scale}, 433 $self->{base}?"+$self->{base}":"", 434 $self->{opmask}; 435 } elsif ($self->{base} eq "rip") { 436 sprintf "%s[%s]",$szmap{$sz},$self->{label}; 437 } else { 438 sprintf "%s[%s%s]%s", $szmap{$sz}, 439 $self->{label}?"$self->{label}+":"", 440 $self->{base},$self->{opmask}; 441 } 442 } 443 } 444} 445{ package register; # pick up registers, which start with %. 446 sub re { 447 my ($class, $line, $opcode) = @_; 448 my $self = {}; 449 my $ret; 450 451 # optional * ----vvv--- appears in indirect jmp/call 452 if ($$line =~ /^(\*?)%(\w+)((?:{[^}]+})*)/) { 453 bless $self,$class; 454 $self->{asterisk} = $1; 455 $self->{value} = $2; 456 $self->{opmask} = $3; 457 $opcode->size($self->size()); 458 $ret = $self; 459 $$line = substr($$line,@+[0]); $$line =~ s/^\s+//; 460 } 461 $ret; 462 } 463 sub size { 464 my $self = shift; 465 my $ret; 466 467 if ($self->{value} =~ /^r[\d]+b$/i) { $ret="b"; } 468 elsif ($self->{value} =~ /^r[\d]+w$/i) { $ret="w"; } 469 elsif ($self->{value} =~ /^r[\d]+d$/i) { $ret="l"; } 470 elsif ($self->{value} =~ /^r[\w]+$/i) { $ret="q"; } 471 elsif ($self->{value} =~ /^[a-d][hl]$/i){ $ret="b"; } 472 elsif ($self->{value} =~ /^[\w]{2}l$/i) { $ret="b"; } 473 elsif ($self->{value} =~ /^[\w]{2}$/i) { $ret="w"; } 474 elsif ($self->{value} =~ /^e[a-z]{2}$/i){ $ret="l"; } 475 476 $ret; 477 } 478 sub out { 479 my $self = shift; 480 if ($gas) { sprintf "%s%%%s%s", $self->{asterisk}, 481 $self->{value}, 482 $self->{opmask}; } 483 else { $self->{opmask} =~ s/%(k[0-7])/$1/; 484 $self->{value}.$self->{opmask}; } 485 } 486} 487{ package label; # pick up labels, which end with : 488 sub re { 489 my ($class, $line) = @_; 490 my $self = {}; 491 my $ret; 492 493 if ($$line =~ /(^[\.\w]+)\:/) { 494 bless $self,$class; 495 $self->{value} = $1; 496 $ret = $self; 497 $$line = substr($$line,@+[0]); $$line =~ s/^\s+//; 498 499 $self->{value} =~ s/^\.L/$decor/; 500 } 501 $ret; 502 } 503 sub out { 504 my $self = shift; 505 506 if ($gas) { 507 my $func = ($globals{$self->{value}} or $self->{value}) . ":"; 508 if ($win64 && $current_function->{name} eq $self->{value} 509 && $current_function->{abi} eq "svr4") { 510 $func .= "\n"; 511 $func .= " movq %rdi,8(%rsp)\n"; 512 $func .= " movq %rsi,16(%rsp)\n"; 513 $func .= " movq %rsp,%rax\n"; 514 $func .= "${decor}SEH_begin_$current_function->{name}:\n"; 515 my $narg = $current_function->{narg}; 516 $narg=6 if (!defined($narg)); 517 $func .= " movq %rcx,%rdi\n" if ($narg>0); 518 $func .= " movq %rdx,%rsi\n" if ($narg>1); 519 $func .= " movq %r8,%rdx\n" if ($narg>2); 520 $func .= " movq %r9,%rcx\n" if ($narg>3); 521 $func .= " movq 40(%rsp),%r8\n" if ($narg>4); 522 $func .= " movq 48(%rsp),%r9\n" if ($narg>5); 523 } 524 $func; 525 } elsif ($self->{value} ne "$current_function->{name}") { 526 # Make all labels in masm global. 527 $self->{value} .= ":" if ($masm); 528 $self->{value} . ":"; 529 } elsif ($win64 && $current_function->{abi} eq "svr4") { 530 my $func = "$current_function->{name}" . 531 ($nasm ? ":" : "\tPROC $current_function->{scope}") . 532 "\n"; 533 $func .= " mov QWORD$PTR\[8+rsp\],rdi\t;WIN64 prologue\n"; 534 $func .= " mov QWORD$PTR\[16+rsp\],rsi\n"; 535 $func .= " mov rax,rsp\n"; 536 $func .= "${decor}SEH_begin_$current_function->{name}:"; 537 $func .= ":" if ($masm); 538 $func .= "\n"; 539 my $narg = $current_function->{narg}; 540 $narg=6 if (!defined($narg)); 541 $func .= " mov rdi,rcx\n" if ($narg>0); 542 $func .= " mov rsi,rdx\n" if ($narg>1); 543 $func .= " mov rdx,r8\n" if ($narg>2); 544 $func .= " mov rcx,r9\n" if ($narg>3); 545 $func .= " mov r8,QWORD$PTR\[40+rsp\]\n" if ($narg>4); 546 $func .= " mov r9,QWORD$PTR\[48+rsp\]\n" if ($narg>5); 547 $func .= "\n"; 548 } else { 549 "$current_function->{name}". 550 ($nasm ? ":" : "\tPROC $current_function->{scope}"); 551 } 552 } 553} 554{ package expr; # pick up expressions 555 sub re { 556 my ($class, $line, $opcode) = @_; 557 my $self = {}; 558 my $ret; 559 560 if ($$line =~ /(^[^,]+)/) { 561 bless $self,$class; 562 $self->{value} = $1; 563 $ret = $self; 564 $$line = substr($$line,@+[0]); $$line =~ s/^\s+//; 565 566 $self->{value} =~ s/\@PLT// if (!$elf); 567 $self->{value} =~ s/([_a-z][_a-z0-9]*)/$globals{$1} or $1/gei; 568 $self->{value} =~ s/\.L/$decor/g; 569 $self->{opcode} = $opcode; 570 } 571 $ret; 572 } 573 sub out { 574 my $self = shift; 575 if ($nasm && $self->{opcode}->mnemonic()=~m/^j(?![re]cxz)/) { 576 "NEAR ".$self->{value}; 577 } else { 578 $self->{value}; 579 } 580 } 581} 582{ package cfi_directive; 583 # CFI directives annotate instructions that are significant for 584 # stack unwinding procedure compliant with DWARF specification, 585 # see http://dwarfstd.org/. Besides naturally expected for this 586 # script platform-specific filtering function, this module adds 587 # three auxiliary synthetic directives not recognized by [GNU] 588 # assembler: 589 # 590 # - .cfi_push to annotate push instructions in prologue, which 591 # translates to .cfi_adjust_cfa_offset (if needed) and 592 # .cfi_offset; 593 # - .cfi_pop to annotate pop instructions in epilogue, which 594 # translates to .cfi_adjust_cfa_offset (if needed) and 595 # .cfi_restore; 596 # - [and most notably] .cfi_cfa_expression which encodes 597 # DW_CFA_def_cfa_expression and passes it to .cfi_escape as 598 # byte vector; 599 # 600 # CFA expressions were introduced in DWARF specification version 601 # 3 and describe how to deduce CFA, Canonical Frame Address. This 602 # becomes handy if your stack frame is variable and you can't 603 # spare register for [previous] frame pointer. Suggested directive 604 # syntax is made-up mix of DWARF operator suffixes [subset of] 605 # and references to registers with optional bias. Following example 606 # describes offloaded *original* stack pointer at specific offset 607 # from *current* stack pointer: 608 # 609 # .cfi_cfa_expression %rsp+40,deref,+8 610 # 611 # Final +8 has everything to do with the fact that CFA is defined 612 # as reference to top of caller's stack, and on x86_64 call to 613 # subroutine pushes 8-byte return address. In other words original 614 # stack pointer upon entry to a subroutine is 8 bytes off from CFA. 615 616 # Below constants are taken from "DWARF Expressions" section of the 617 # DWARF specification, section is numbered 7.7 in versions 3 and 4. 618 my %DW_OP_simple = ( # no-arg operators, mapped directly 619 deref => 0x06, dup => 0x12, 620 drop => 0x13, over => 0x14, 621 pick => 0x15, swap => 0x16, 622 rot => 0x17, xderef => 0x18, 623 624 abs => 0x19, and => 0x1a, 625 div => 0x1b, minus => 0x1c, 626 mod => 0x1d, mul => 0x1e, 627 neg => 0x1f, not => 0x20, 628 or => 0x21, plus => 0x22, 629 shl => 0x24, shr => 0x25, 630 shra => 0x26, xor => 0x27, 631 ); 632 633 my %DW_OP_complex = ( # used in specific subroutines 634 constu => 0x10, # uleb128 635 consts => 0x11, # sleb128 636 plus_uconst => 0x23, # uleb128 637 lit0 => 0x30, # add 0-31 to opcode 638 reg0 => 0x50, # add 0-31 to opcode 639 breg0 => 0x70, # add 0-31 to opcole, sleb128 640 regx => 0x90, # uleb28 641 fbreg => 0x91, # sleb128 642 bregx => 0x92, # uleb128, sleb128 643 piece => 0x93, # uleb128 644 ); 645 646 # Following constants are defined in x86_64 ABI supplement, for 647 # example available at https://www.uclibc.org/docs/psABI-x86_64.pdf, 648 # see section 3.7 "Stack Unwind Algorithm". 649 my %DW_reg_idx = ( 650 "%rax"=>0, "%rdx"=>1, "%rcx"=>2, "%rbx"=>3, 651 "%rsi"=>4, "%rdi"=>5, "%rbp"=>6, "%rsp"=>7, 652 "%r8" =>8, "%r9" =>9, "%r10"=>10, "%r11"=>11, 653 "%r12"=>12, "%r13"=>13, "%r14"=>14, "%r15"=>15 654 ); 655 656 my ($cfa_reg, $cfa_rsp); 657 my @cfa_stack; 658 659 # [us]leb128 format is variable-length integer representation base 660 # 2^128, with most significant bit of each byte being 0 denoting 661 # *last* most significant digit. See "Variable Length Data" in the 662 # DWARF specification, numbered 7.6 at least in versions 3 and 4. 663 sub sleb128 { 664 use integer; # get right shift extend sign 665 666 my $val = shift; 667 my $sign = ($val < 0) ? -1 : 0; 668 my @ret = (); 669 670 while(1) { 671 push @ret, $val&0x7f; 672 673 # see if remaining bits are same and equal to most 674 # significant bit of the current digit, if so, it's 675 # last digit... 676 last if (($val>>6) == $sign); 677 678 @ret[-1] |= 0x80; 679 $val >>= 7; 680 } 681 682 return @ret; 683 } 684 sub uleb128 { 685 my $val = shift; 686 my @ret = (); 687 688 while(1) { 689 push @ret, $val&0x7f; 690 691 # see if it's last significant digit... 692 last if (($val >>= 7) == 0); 693 694 @ret[-1] |= 0x80; 695 } 696 697 return @ret; 698 } 699 sub const { 700 my $val = shift; 701 702 if ($val >= 0 && $val < 32) { 703 return ($DW_OP_complex{lit0}+$val); 704 } 705 return ($DW_OP_complex{consts}, sleb128($val)); 706 } 707 sub reg { 708 my $val = shift; 709 710 return if ($val !~ m/^(%r\w+)(?:([\+\-])((?:0x)?[0-9a-f]+))?/); 711 712 my $reg = $DW_reg_idx{$1}; 713 my $off = eval ("0 $2 $3"); 714 715 return (($DW_OP_complex{breg0} + $reg), sleb128($off)); 716 # Yes, we use DW_OP_bregX+0 to push register value and not 717 # DW_OP_regX, because latter would require even DW_OP_piece, 718 # which would be a waste under the circumstances. If you have 719 # to use DWP_OP_reg, use "regx:N"... 720 } 721 sub cfa_expression { 722 my $line = shift; 723 my @ret; 724 725 foreach my $token (split(/,\s*/,$line)) { 726 if ($token =~ /^%r/) { 727 push @ret,reg($token); 728 } elsif ($token =~ /((?:0x)?[0-9a-f]+)\((%r\w+)\)/) { 729 push @ret,reg("$2+$1"); 730 } elsif ($token =~ /(\w+):(\-?(?:0x)?[0-9a-f]+)(U?)/i) { 731 my $i = 1*eval($2); 732 push @ret,$DW_OP_complex{$1}, ($3 ? uleb128($i) : sleb128($i)); 733 } elsif (my $i = 1*eval($token) or $token eq "0") { 734 if ($token =~ /^\+/) { 735 push @ret,$DW_OP_complex{plus_uconst},uleb128($i); 736 } else { 737 push @ret,const($i); 738 } 739 } else { 740 push @ret,$DW_OP_simple{$token}; 741 } 742 } 743 744 # Finally we return DW_CFA_def_cfa_expression, 15, followed by 745 # length of the expression and of course the expression itself. 746 return (15,scalar(@ret),@ret); 747 } 748 sub re { 749 my ($class, $line) = @_; 750 my $self = {}; 751 my $ret; 752 753 if ($$line =~ s/^\s*\.cfi_(\w+)\s*//) { 754 bless $self,$class; 755 $ret = $self; 756 undef $self->{value}; 757 my $dir = $1; 758 759 SWITCH: for ($dir) { 760 # What is $cfa_rsp? Effectively it's difference between %rsp 761 # value and current CFA, Canonical Frame Address, which is 762 # why it starts with -8. Recall that CFA is top of caller's 763 # stack... 764 /startproc/ && do { ($cfa_reg, $cfa_rsp) = ("%rsp", -8); last; }; 765 /endproc/ && do { ($cfa_reg, $cfa_rsp) = ("%rsp", 0); 766 # .cfi_remember_state directives that are not 767 # matched with .cfi_restore_state are 768 # unnecessary. 769 die "unpaired .cfi_remember_state" if (@cfa_stack); 770 last; 771 }; 772 /def_cfa_register/ 773 && do { $cfa_reg = $$line; last; }; 774 /def_cfa_offset/ 775 && do { $cfa_rsp = -1*eval($$line) if ($cfa_reg eq "%rsp"); 776 last; 777 }; 778 /adjust_cfa_offset/ 779 && do { $cfa_rsp -= 1*eval($$line) if ($cfa_reg eq "%rsp"); 780 last; 781 }; 782 /def_cfa/ && do { if ($$line =~ /(%r\w+)\s*,\s*(.+)/) { 783 $cfa_reg = $1; 784 $cfa_rsp = -1*eval($2) if ($cfa_reg eq "%rsp"); 785 } 786 last; 787 }; 788 /push/ && do { $dir = undef; 789 $cfa_rsp -= 8; 790 if ($cfa_reg eq "%rsp") { 791 $self->{value} = ".cfi_adjust_cfa_offset\t8\n"; 792 } 793 $self->{value} .= ".cfi_offset\t$$line,$cfa_rsp"; 794 last; 795 }; 796 /pop/ && do { $dir = undef; 797 $cfa_rsp += 8; 798 if ($cfa_reg eq "%rsp") { 799 $self->{value} = ".cfi_adjust_cfa_offset\t-8\n"; 800 } 801 $self->{value} .= ".cfi_restore\t$$line"; 802 last; 803 }; 804 /cfa_expression/ 805 && do { $dir = undef; 806 $self->{value} = ".cfi_escape\t" . 807 join(",", map(sprintf("0x%02x", $_), 808 cfa_expression($$line))); 809 last; 810 }; 811 /remember_state/ 812 && do { push @cfa_stack, [$cfa_reg, $cfa_rsp]; 813 last; 814 }; 815 /restore_state/ 816 && do { ($cfa_reg, $cfa_rsp) = @{pop @cfa_stack}; 817 last; 818 }; 819 } 820 821 $self->{value} = ".cfi_$dir\t$$line" if ($dir); 822 823 $$line = ""; 824 } 825 826 return $ret; 827 } 828 sub out { 829 my $self = shift; 830 return ($elf ? $self->{value} : undef); 831 } 832} 833{ package directive; # pick up directives, which start with . 834 sub re { 835 my ($class, $line) = @_; 836 my $self = {}; 837 my $ret; 838 my $dir; 839 840 # chain-call to cfi_directive 841 $ret = cfi_directive->re($line) and return $ret; 842 843 if ($$line =~ /^\s*(\.\w+)/) { 844 bless $self,$class; 845 $dir = $1; 846 $ret = $self; 847 undef $self->{value}; 848 $$line = substr($$line,@+[0]); $$line =~ s/^\s+//; 849 850 SWITCH: for ($dir) { 851 /\.global|\.globl|\.extern/ 852 && do { $globals{$$line} = $prefix . $$line; 853 $$line = $globals{$$line} if ($prefix); 854 last; 855 }; 856 /\.type/ && do { my ($sym,$type,$narg) = split(',',$$line); 857 if ($type eq "\@function") { 858 undef $current_function; 859 $current_function->{name} = $sym; 860 $current_function->{abi} = "svr4"; 861 $current_function->{narg} = $narg; 862 $current_function->{scope} = defined($globals{$sym})?"PUBLIC":"PRIVATE"; 863 } elsif ($type eq "\@abi-omnipotent") { 864 undef $current_function; 865 $current_function->{name} = $sym; 866 $current_function->{scope} = defined($globals{$sym})?"PUBLIC":"PRIVATE"; 867 } 868 $$line =~ s/\@abi\-omnipotent/\@function/; 869 $$line =~ s/\@function.*/\@function/; 870 last; 871 }; 872 /\.asciz/ && do { if ($$line =~ /^"(.*)"$/) { 873 $dir = ".byte"; 874 $$line = join(",",unpack("C*",$1),0); 875 } 876 last; 877 }; 878 /\.rva|\.long|\.quad|\.byte/ 879 && do { $$line =~ s/([_a-z][_a-z0-9]*)/$globals{$1} or $1/gei; 880 $$line =~ s/\.L/$decor/g; 881 last; 882 }; 883 } 884 885 if ($gas) { 886 $self->{value} = $dir . "\t" . $$line; 887 888 if ($dir =~ /\.extern/) { 889 $self->{value} = ""; # swallow extern 890 } elsif (!$elf && $dir =~ /\.type/) { 891 $self->{value} = ""; 892 $self->{value} = ".def\t" . ($globals{$1} or $1) . ";\t" . 893 (defined($globals{$1})?".scl 2;":".scl 3;") . 894 "\t.type 32;\t.endef" 895 if ($win64 && $$line =~ /([^,]+),\@function/); 896 } elsif (!$elf && $dir =~ /\.size/) { 897 $self->{value} = ""; 898 if (defined($current_function)) { 899 $self->{value} .= "${decor}SEH_end_$current_function->{name}:" 900 if ($win64 && $current_function->{abi} eq "svr4"); 901 undef $current_function; 902 } 903 } elsif (!$elf && $dir =~ /\.align/) { 904 $self->{value} = ".p2align\t" . (log($$line)/log(2)); 905 } elsif ($dir eq ".section") { 906 # 907 # get rid off align option, it's not supported/tolerated 908 # by gcc. openssl project introduced the option as an aid 909 # to deal with nasm/masm assembly. 910 # 911 $self->{value} =~ s/(.+)\s+align\s*=.*$/$1/; 912 $current_segment = pop(@segment_stack); 913 if (not $current_segment) { 914 # if no previous section is defined, then assume .text 915 # so code does not land in .data section by accident. 916 # this deals with inconsistency of perl-assembly files. 917 push(@segment_stack, ".text"); 918 } 919 # 920 # $$line may still contains align= option. We do care 921 # about section type here. 922 # 923 $current_segment = $$line; 924 $current_segment =~ s/([^\s]+).*$/$1/; 925 push(@segment_stack, $current_segment); 926 if (!$elf && $current_segment eq ".rodata") { 927 if ($flavour eq "macosx") { $self->{value} = ".section\t__DATA,__const"; } 928 elsif ($flavour eq "mingw64") { $self->{value} = ".section\t.rodata"; } 929 } 930 if (!$elf && $current_segment eq ".init") { 931 if ($flavour eq "macosx") { $self->{value} = ".mod_init_func"; } 932 elsif ($flavour eq "mingw64") { $self->{value} = ".section\t.ctors"; } 933 } 934 } elsif ($dir =~ /\.(text|data)/) { 935 $current_segment = pop(@segment_stack); 936 if (not $current_segment) { 937 # if no previous section is defined, then assume .text 938 # so code does not land in .data section by accident. 939 # this deals with inconsistency of perl-assembly files. 940 push(@segment_stack, ".text"); 941 } 942 $current_segment=".$1"; 943 push(@segment_stack, $current_segment); 944 } elsif ($dir =~ /\.hidden/) { 945 if ($flavour eq "macosx") { $self->{value} = ".private_extern\t$prefix$$line"; } 946 elsif ($flavour eq "mingw64") { $self->{value} = ""; } 947 } elsif ($dir =~ /\.comm/) { 948 $self->{value} = "$dir\t$prefix$$line"; 949 $self->{value} =~ s|,([0-9]+),([0-9]+)$|",$1,".log($2)/log(2)|e if ($flavour eq "macosx"); 950 } elsif ($dir =~ /\.previous/) { 951 pop(@segment_stack); #pop ourselves 952 # just peek at the top of the stack here 953 $current_segment = @segment_stack[0]; 954 if (not $current_segment) { 955 # if no previous segment was defined assume .text so 956 # the code does not accidentally land in .data section. 957 $current_segment = ".text"; 958 push(@segment_stack, $current_segment); 959 } 960 $self->{value} = $current_segment if ($flavour eq "mingw64"); 961 } 962 $$line = ""; 963 return $self; 964 } 965 966 # non-gas case or nasm/masm 967 SWITCH: for ($dir) { 968 /\.text/ && do { my $v=undef; 969 if ($nasm) { 970 $current_segment = pop(@segment_stack); 971 if (not $current_segment) { 972 push(@segment_stack, ".text"); 973 } 974 $v="section .text code align=64\n"; 975 $current_segment = ".text"; 976 push(@segment_stack, $current_segment); 977 } else { 978 $current_segment = pop(@segment_stack); 979 if (not $current_segment) { 980 push(@segment_stack, ".text\$"); 981 } 982 $v="$current_segment\tENDS\n" if ($current_segment); 983 $current_segment = ".text\$"; 984 push(@segment_stack, $current_segment); 985 $v.="$current_segment\tSEGMENT "; 986 $v.=$masm>=$masmref ? "ALIGN(256)" : "PAGE"; 987 $v.=" 'CODE'"; 988 } 989 $self->{value} = $v; 990 last; 991 }; 992 /\.data/ && do { my $v=undef; 993 if ($nasm) { 994 $v="section .data data align=8\n"; 995 } else { 996 $current_segment = pop(@segment_stack); 997 $v="$current_segment\tENDS\n" if ($current_segment); 998 $current_segment = "_DATA"; 999 push(@segment_stack, $current_segment); 1000 $v.="$current_segment\tSEGMENT"; 1001 } 1002 $self->{value} = $v; 1003 last; 1004 }; 1005 /\.section/ && do { my $v=undef; 1006 my $align=undef; 1007 # 1008 # $$line may currently contain something like this 1009 # .rodata align = 64 1010 # align part is optional 1011 # 1012 $align = $$line; 1013 $align =~ s/(.*)(align\s*=\s*\d+$)/$2/; 1014 $$line =~ s/(.*)(\s+align\s*=\s*\d+$)/$1/; 1015 $$line =~ s/,.*//; 1016 $$line = ".CRT\$XCU" if ($$line eq ".init"); 1017 $$line = ".rdata" if ($$line eq ".rodata"); 1018 if ($nasm) { 1019 $current_segment = pop(@segment_stack); 1020 if (not $current_segment) { 1021 # 1022 # This is a hack which deals with ecp_nistz256-x86_64.pl, 1023 # The precomputed curve is stored in the first section 1024 # in .asm file. Pushing extra .text section here 1025 # allows our poor man's solution to stick to assumption 1026 # .text section is always the first. 1027 # 1028 push(@segment_stack, ".text"); 1029 } 1030 $v="section $$line"; 1031 if ($$line=~/\.([prx])data/) { 1032 if ($align =~ /align\s*=\s*(\d+)/) { 1033 $v.= " rdata align=$1" ; 1034 } else { 1035 $v.=" rdata align="; 1036 $v.=$1 eq "p"? 4 : 8; 1037 } 1038 } elsif ($$line=~/\.CRT\$/i) { 1039 $v.=" rdata align=8"; 1040 } 1041 } else { 1042 $current_segment = pop(@segment_stack); 1043 if (not $current_segment) { 1044 # 1045 # same hack for masm to keep ecp_nistz256-x86_64.pl 1046 # happy. 1047 # 1048 push(@segment_stack, ".text\$"); 1049 } 1050 $v="$current_segment\tENDS\n" if ($current_segment); 1051 $v.="$$line\tSEGMENT"; 1052 if ($$line=~/\.([prx])data/) { 1053 $v.=" READONLY"; 1054 if ($align =~ /align\s*=\s*(\d+)$/) { 1055 $v.=" ALIGN($1)" if ($masm>=$masmref); 1056 } else { 1057 $v.=" ALIGN(".($1 eq "p" ? 4 : 8).")" if ($masm>=$masmref); 1058 } 1059 } elsif ($$line=~/\.CRT\$/i) { 1060 $v.=" READONLY "; 1061 $v.=$masm>=$masmref ? "ALIGN(8)" : "DWORD"; 1062 } 1063 } 1064 $current_segment = $$line; 1065 push(@segment_stack, $$line); 1066 $self->{value} = $v; 1067 last; 1068 }; 1069 /\.extern/ && do { $self->{value} = "EXTERN\t".$$line; 1070 $self->{value} .= ":NEAR" if ($masm); 1071 last; 1072 }; 1073 /\.globl|.global/ 1074 && do { $self->{value} = $masm?"PUBLIC":"global"; 1075 $self->{value} .= "\t".$$line; 1076 last; 1077 }; 1078 /\.size/ && do { if (defined($current_function)) { 1079 undef $self->{value}; 1080 if ($current_function->{abi} eq "svr4") { 1081 $self->{value}="${decor}SEH_end_$current_function->{name}:"; 1082 $self->{value}.=":\n" if($masm); 1083 } 1084 $self->{value}.="$current_function->{name}\tENDP" if($masm && $current_function->{name}); 1085 undef $current_function; 1086 } 1087 last; 1088 }; 1089 /\.align/ && do { my $max = ($masm && $masm>=$masmref) ? 256 : 4096; 1090 $self->{value} = "ALIGN\t".($$line>$max?$max:$$line); 1091 last; 1092 }; 1093 /\.(value|long|rva|quad)/ 1094 && do { my $sz = substr($1,0,1); 1095 my @arr = split(/,\s*/,$$line); 1096 my $last = pop(@arr); 1097 my $conv = sub { my $var=shift; 1098 $var=~s/^(0b[0-1]+)/oct($1)/eig; 1099 $var=~s/^0x([0-9a-f]+)/0$1h/ig if ($masm); 1100 if ($sz eq "D" && ($current_segment=~/.[px]data/ || $dir eq ".rva")) 1101 { $var=~s/^([_a-z\$\@][_a-z0-9\$\@]*)/$nasm?"$1 wrt ..imagebase":"imagerel $1"/egi; } 1102 $var; 1103 }; 1104 1105 $sz =~ tr/bvlrq/BWDDQ/; 1106 $self->{value} = "\tD$sz\t"; 1107 for (@arr) { $self->{value} .= &$conv($_).","; } 1108 $self->{value} .= &$conv($last); 1109 last; 1110 }; 1111 /\.byte/ && do { my @str=split(/,\s*/,$$line); 1112 map(s/(0b[0-1]+)/oct($1)/eig,@str); 1113 map(s/0x([0-9a-f]+)/0$1h/ig,@str) if ($masm); 1114 while ($#str>15) { 1115 $self->{value}.="DB\t" 1116 .join(",",@str[0..15])."\n"; 1117 foreach (0..15) { shift @str; } 1118 } 1119 $self->{value}.="DB\t" 1120 .join(",",@str) if (@str); 1121 last; 1122 }; 1123 /\.comm/ && do { my @str=split(/,\s*/,$$line); 1124 my $v=undef; 1125 if ($nasm) { 1126 $v.="common $prefix@str[0] @str[1]"; 1127 } else { 1128 $current_segment = pop(@segment_stack);; 1129 $v="$current_segment\tENDS\n" if ($current_segment); 1130 $current_segment = "_DATA"; 1131 push(@segment_stack, $current_segment); 1132 $v.="$current_segment\tSEGMENT\n"; 1133 $v.="COMM @str[0]:DWORD:".@str[1]/4; 1134 } 1135 $self->{value} = $v; 1136 last; 1137 }; 1138 /^.previous/ && do { 1139 my $v=undef; 1140 if ($nasm) { 1141 pop(@segment_stack); # pop ourselves, we don't need to emit END directive 1142 # pop section so we can emit proper .section name. 1143 $current_segment = pop(@segment_stack); 1144 $v="section $current_segment"; 1145 # Hack again: 1146 # push section/segment to stack. The .previous is currently paired 1147 # with .rodata only. We have to keep extra '.text' on stack for 1148 # situation where there is for example .pdata section 'terminated' 1149 # by new '.text' section. 1150 # 1151 push(@segment_stack, $current_segment); 1152 } else { 1153 $current_segment = pop(@segment_stack); 1154 $v="$current_segment\tENDS\n" if ($current_segment); 1155 $current_segment = pop(@segment_stack); 1156 if ($current_segment =~ /\.text\$/) { 1157 $v.="$current_segment\tSEGMENT "; 1158 $v.=$masm>=$masmref ? "ALIGN(256)" : "PAGE"; 1159 $v.=" 'CODE'"; 1160 push(@segment_stack, $current_segment); 1161 } 1162 } 1163 $self->{value} = $v; 1164 last; 1165 }; 1166 } 1167 $$line = ""; 1168 } 1169 1170 $ret; 1171 } 1172 sub out { 1173 my $self = shift; 1174 $self->{value}; 1175 } 1176} 1177 1178# Upon initial x86_64 introduction SSE>2 extensions were not introduced 1179# yet. In order not to be bothered by tracing exact assembler versions, 1180# but at the same time to provide a bare security minimum of AES-NI, we 1181# hard-code some instructions. Extensions past AES-NI on the other hand 1182# are traced by examining assembler version in individual perlasm 1183# modules... 1184 1185my %regrm = ( "%eax"=>0, "%ecx"=>1, "%edx"=>2, "%ebx"=>3, 1186 "%esp"=>4, "%ebp"=>5, "%esi"=>6, "%edi"=>7 ); 1187 1188sub rex { 1189 my $opcode=shift; 1190 my ($dst,$src,$rex)=@_; 1191 1192 $rex|=0x04 if($dst>=8); 1193 $rex|=0x01 if($src>=8); 1194 push @$opcode,($rex|0x40) if ($rex); 1195} 1196 1197my $movq = sub { # elderly gas can't handle inter-register movq 1198 my $arg = shift; 1199 my @opcode=(0x66); 1200 if ($arg =~ /%xmm([0-9]+),\s*%r(\w+)/) { 1201 my ($src,$dst)=($1,$2); 1202 if ($dst !~ /[0-9]+/) { $dst = $regrm{"%e$dst"}; } 1203 rex(\@opcode,$src,$dst,0x8); 1204 push @opcode,0x0f,0x7e; 1205 push @opcode,0xc0|(($src&7)<<3)|($dst&7); # ModR/M 1206 @opcode; 1207 } elsif ($arg =~ /%r(\w+),\s*%xmm([0-9]+)/) { 1208 my ($src,$dst)=($2,$1); 1209 if ($dst !~ /[0-9]+/) { $dst = $regrm{"%e$dst"}; } 1210 rex(\@opcode,$src,$dst,0x8); 1211 push @opcode,0x0f,0x6e; 1212 push @opcode,0xc0|(($src&7)<<3)|($dst&7); # ModR/M 1213 @opcode; 1214 } else { 1215 (); 1216 } 1217}; 1218 1219my $pextrd = sub { 1220 if (shift =~ /\$([0-9]+),\s*%xmm([0-9]+),\s*(%\w+)/) { 1221 my @opcode=(0x66); 1222 my $imm=$1; 1223 my $src=$2; 1224 my $dst=$3; 1225 if ($dst =~ /%r([0-9]+)d/) { $dst = $1; } 1226 elsif ($dst =~ /%e/) { $dst = $regrm{$dst}; } 1227 rex(\@opcode,$src,$dst); 1228 push @opcode,0x0f,0x3a,0x16; 1229 push @opcode,0xc0|(($src&7)<<3)|($dst&7); # ModR/M 1230 push @opcode,$imm; 1231 @opcode; 1232 } else { 1233 (); 1234 } 1235}; 1236 1237my $pinsrd = sub { 1238 if (shift =~ /\$([0-9]+),\s*(%\w+),\s*%xmm([0-9]+)/) { 1239 my @opcode=(0x66); 1240 my $imm=$1; 1241 my $src=$2; 1242 my $dst=$3; 1243 if ($src =~ /%r([0-9]+)/) { $src = $1; } 1244 elsif ($src =~ /%e/) { $src = $regrm{$src}; } 1245 rex(\@opcode,$dst,$src); 1246 push @opcode,0x0f,0x3a,0x22; 1247 push @opcode,0xc0|(($dst&7)<<3)|($src&7); # ModR/M 1248 push @opcode,$imm; 1249 @opcode; 1250 } else { 1251 (); 1252 } 1253}; 1254 1255my $pshufb = sub { 1256 if (shift =~ /%xmm([0-9]+),\s*%xmm([0-9]+)/) { 1257 my @opcode=(0x66); 1258 rex(\@opcode,$2,$1); 1259 push @opcode,0x0f,0x38,0x00; 1260 push @opcode,0xc0|($1&7)|(($2&7)<<3); # ModR/M 1261 @opcode; 1262 } else { 1263 (); 1264 } 1265}; 1266 1267my $palignr = sub { 1268 if (shift =~ /\$([0-9]+),\s*%xmm([0-9]+),\s*%xmm([0-9]+)/) { 1269 my @opcode=(0x66); 1270 rex(\@opcode,$3,$2); 1271 push @opcode,0x0f,0x3a,0x0f; 1272 push @opcode,0xc0|($2&7)|(($3&7)<<3); # ModR/M 1273 push @opcode,$1; 1274 @opcode; 1275 } else { 1276 (); 1277 } 1278}; 1279 1280my $pclmulqdq = sub { 1281 if (shift =~ /\$([x0-9a-f]+),\s*%xmm([0-9]+),\s*%xmm([0-9]+)/) { 1282 my @opcode=(0x66); 1283 rex(\@opcode,$3,$2); 1284 push @opcode,0x0f,0x3a,0x44; 1285 push @opcode,0xc0|($2&7)|(($3&7)<<3); # ModR/M 1286 my $c=$1; 1287 push @opcode,$c=~/^0/?oct($c):$c; 1288 @opcode; 1289 } else { 1290 (); 1291 } 1292}; 1293 1294my $rdrand = sub { 1295 if (shift =~ /%[er](\w+)/) { 1296 my @opcode=(); 1297 my $dst=$1; 1298 if ($dst !~ /[0-9]+/) { $dst = $regrm{"%e$dst"}; } 1299 rex(\@opcode,0,$dst,8); 1300 push @opcode,0x0f,0xc7,0xf0|($dst&7); 1301 @opcode; 1302 } else { 1303 (); 1304 } 1305}; 1306 1307my $rdseed = sub { 1308 if (shift =~ /%[er](\w+)/) { 1309 my @opcode=(); 1310 my $dst=$1; 1311 if ($dst !~ /[0-9]+/) { $dst = $regrm{"%e$dst"}; } 1312 rex(\@opcode,0,$dst,8); 1313 push @opcode,0x0f,0xc7,0xf8|($dst&7); 1314 @opcode; 1315 } else { 1316 (); 1317 } 1318}; 1319 1320# Not all AVX-capable assemblers recognize AMD XOP extension. Since we 1321# are using only two instructions hand-code them in order to be excused 1322# from chasing assembler versions... 1323 1324sub rxb { 1325 my $opcode=shift; 1326 my ($dst,$src1,$src2,$rxb)=@_; 1327 1328 $rxb|=0x7<<5; 1329 $rxb&=~(0x04<<5) if($dst>=8); 1330 $rxb&=~(0x01<<5) if($src1>=8); 1331 $rxb&=~(0x02<<5) if($src2>=8); 1332 push @$opcode,$rxb; 1333} 1334 1335my $vprotd = sub { 1336 if (shift =~ /\$([x0-9a-f]+),\s*%xmm([0-9]+),\s*%xmm([0-9]+)/) { 1337 my @opcode=(0x8f); 1338 rxb(\@opcode,$3,$2,-1,0x08); 1339 push @opcode,0x78,0xc2; 1340 push @opcode,0xc0|($2&7)|(($3&7)<<3); # ModR/M 1341 my $c=$1; 1342 push @opcode,$c=~/^0/?oct($c):$c; 1343 @opcode; 1344 } else { 1345 (); 1346 } 1347}; 1348 1349my $vprotq = sub { 1350 if (shift =~ /\$([x0-9a-f]+),\s*%xmm([0-9]+),\s*%xmm([0-9]+)/) { 1351 my @opcode=(0x8f); 1352 rxb(\@opcode,$3,$2,-1,0x08); 1353 push @opcode,0x78,0xc3; 1354 push @opcode,0xc0|($2&7)|(($3&7)<<3); # ModR/M 1355 my $c=$1; 1356 push @opcode,$c=~/^0/?oct($c):$c; 1357 @opcode; 1358 } else { 1359 (); 1360 } 1361}; 1362 1363# Intel Control-flow Enforcement Technology extension. All functions and 1364# indirect branch targets will have to start with this instruction... 1365 1366my $endbranch = sub { 1367 (0xf3,0x0f,0x1e,0xfa); 1368}; 1369 1370######################################################################## 1371 1372if ($nasm) { 1373 print <<___; 1374default rel 1375%define XMMWORD 1376%define YMMWORD 1377%define ZMMWORD 1378___ 1379} elsif ($masm) { 1380 print <<___; 1381OPTION DOTNAME 1382___ 1383} 1384while(defined(my $line=<>)) { 1385 1386 $line =~ s|\R$||; # Better chomp 1387 1388 $line =~ s|[#!].*$||; # get rid of asm-style comments... 1389 $line =~ s|/\*.*\*/||; # ... and C-style comments... 1390 $line =~ s|^\s+||; # ... and skip whitespaces in beginning 1391 $line =~ s|\s+$||; # ... and at the end 1392 1393 if (my $label=label->re(\$line)) { print $label->out(); } 1394 1395 if (my $directive=directive->re(\$line)) { 1396 printf "%s",$directive->out(); 1397 } elsif (my $opcode=opcode->re(\$line)) { 1398 my $asm = eval("\$".$opcode->mnemonic()); 1399 1400 if ((ref($asm) eq 'CODE') && scalar(my @bytes=&$asm($line))) { 1401 print $gas?".byte\t":"DB\t",join(',',@bytes),"\n"; 1402 next; 1403 } 1404 1405 my @args; 1406 ARGUMENT: while (1) { 1407 my $arg; 1408 1409 ($arg=register->re(\$line, $opcode))|| 1410 ($arg=const->re(\$line)) || 1411 ($arg=ea->re(\$line, $opcode)) || 1412 ($arg=expr->re(\$line, $opcode)) || 1413 last ARGUMENT; 1414 1415 push @args,$arg; 1416 1417 last ARGUMENT if ($line !~ /^,/); 1418 1419 $line =~ s/^,\s*//; 1420 } # ARGUMENT: 1421 1422 if ($#args>=0) { 1423 my $insn; 1424 my $sz=$opcode->size(); 1425 1426 if ($gas) { 1427 $insn = $opcode->out($#args>=1?$args[$#args]->size():$sz); 1428 @args = map($_->out($sz),@args); 1429 printf "\t%s\t%s",$insn,join(",",@args); 1430 } else { 1431 $insn = $opcode->out(); 1432 foreach (@args) { 1433 my $arg = $_->out(); 1434 # $insn.=$sz compensates for movq, pinsrw, ... 1435 if ($arg =~ /^xmm[0-9]+$/) { $insn.=$sz; $sz="x" if(!$sz); last; } 1436 if ($arg =~ /^ymm[0-9]+$/) { $insn.=$sz; $sz="y" if(!$sz); last; } 1437 if ($arg =~ /^zmm[0-9]+$/) { $insn.=$sz; $sz="z" if(!$sz); last; } 1438 if ($arg =~ /^mm[0-9]+$/) { $insn.=$sz; $sz="q" if(!$sz); last; } 1439 } 1440 @args = reverse(@args); 1441 undef $sz if ($nasm && $opcode->mnemonic() eq "lea"); 1442 printf "\t%s\t%s",$insn,join(",",map($_->out($sz),@args)); 1443 } 1444 } else { 1445 printf "\t%s",$opcode->out(); 1446 } 1447 } 1448 1449 print $line,"\n"; 1450} 1451 1452print "$cet_property" if ($cet_property); 1453print "\n$current_segment\tENDS\n" if ($current_segment && $masm); 1454print "END\n" if ($masm); 1455 1456close STDOUT or die "error closing STDOUT: $!;" 1457 1458 ################################################# 1459# Cross-reference x86_64 ABI "card" 1460# 1461# Unix Win64 1462# %rax * * 1463# %rbx - - 1464# %rcx #4 #1 1465# %rdx #3 #2 1466# %rsi #2 - 1467# %rdi #1 - 1468# %rbp - - 1469# %rsp - - 1470# %r8 #5 #3 1471# %r9 #6 #4 1472# %r10 * * 1473# %r11 * * 1474# %r12 - - 1475# %r13 - - 1476# %r14 - - 1477# %r15 - - 1478# 1479# (*) volatile register 1480# (-) preserved by callee 1481# (#) Nth argument, volatile 1482# 1483# In Unix terms top of stack is argument transfer area for arguments 1484# which could not be accommodated in registers. Or in other words 7th 1485# [integer] argument resides at 8(%rsp) upon function entry point. 1486# 128 bytes above %rsp constitute a "red zone" which is not touched 1487# by signal handlers and can be used as temporal storage without 1488# allocating a frame. 1489# 1490# In Win64 terms N*8 bytes on top of stack is argument transfer area, 1491# which belongs to/can be overwritten by callee. N is the number of 1492# arguments passed to callee, *but* not less than 4! This means that 1493# upon function entry point 5th argument resides at 40(%rsp), as well 1494# as that 32 bytes from 8(%rsp) can always be used as temporal 1495# storage [without allocating a frame]. One can actually argue that 1496# one can assume a "red zone" above stack pointer under Win64 as well. 1497# Point is that at apparently no occasion Windows kernel would alter 1498# the area above user stack pointer in true asynchronous manner... 1499# 1500# All the above means that if assembler programmer adheres to Unix 1501# register and stack layout, but disregards the "red zone" existence, 1502# it's possible to use following prologue and epilogue to "gear" from 1503# Unix to Win64 ABI in leaf functions with not more than 6 arguments. 1504# 1505# omnipotent_function: 1506# ifdef WIN64 1507# movq %rdi,8(%rsp) 1508# movq %rsi,16(%rsp) 1509# movq %rcx,%rdi ; if 1st argument is actually present 1510# movq %rdx,%rsi ; if 2nd argument is actually ... 1511# movq %r8,%rdx ; if 3rd argument is ... 1512# movq %r9,%rcx ; if 4th argument ... 1513# movq 40(%rsp),%r8 ; if 5th ... 1514# movq 48(%rsp),%r9 ; if 6th ... 1515# endif 1516# ... 1517# ifdef WIN64 1518# movq 8(%rsp),%rdi 1519# movq 16(%rsp),%rsi 1520# endif 1521# ret 1522# 1523 ################################################# 1524# Win64 SEH, Structured Exception Handling. 1525# 1526# Unlike on Unix systems(*) lack of Win64 stack unwinding information 1527# has undesired side-effect at run-time: if an exception is raised in 1528# assembler subroutine such as those in question (basically we're 1529# referring to segmentation violations caused by malformed input 1530# parameters), the application is briskly terminated without invoking 1531# any exception handlers, most notably without generating memory dump 1532# or any user notification whatsoever. This poses a problem. It's 1533# possible to address it by registering custom language-specific 1534# handler that would restore processor context to the state at 1535# subroutine entry point and return "exception is not handled, keep 1536# unwinding" code. Writing such handler can be a challenge... But it's 1537# doable, though requires certain coding convention. Consider following 1538# snippet: 1539# 1540# .type function,@function 1541# function: 1542# movq %rsp,%rax # copy rsp to volatile register 1543# pushq %r15 # save non-volatile registers 1544# pushq %rbx 1545# pushq %rbp 1546# movq %rsp,%r11 1547# subq %rdi,%r11 # prepare [variable] stack frame 1548# andq $-64,%r11 1549# movq %rax,0(%r11) # check for exceptions 1550# movq %r11,%rsp # allocate [variable] stack frame 1551# movq %rax,0(%rsp) # save original rsp value 1552# magic_point: 1553# ... 1554# movq 0(%rsp),%rcx # pull original rsp value 1555# movq -24(%rcx),%rbp # restore non-volatile registers 1556# movq -16(%rcx),%rbx 1557# movq -8(%rcx),%r15 1558# movq %rcx,%rsp # restore original rsp 1559# magic_epilogue: 1560# ret 1561# .size function,.-function 1562# 1563# The key is that up to magic_point copy of original rsp value remains 1564# in chosen volatile register and no non-volatile register, except for 1565# rsp, is modified. While past magic_point rsp remains constant till 1566# the very end of the function. In this case custom language-specific 1567# exception handler would look like this: 1568# 1569# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame, 1570# CONTEXT *context,DISPATCHER_CONTEXT *disp) 1571# { ULONG64 *rsp = (ULONG64 *)context->Rax; 1572# ULONG64 rip = context->Rip; 1573# 1574# if (rip >= magic_point) 1575# { rsp = (ULONG64 *)context->Rsp; 1576# if (rip < magic_epilogue) 1577# { rsp = (ULONG64 *)rsp[0]; 1578# context->Rbp = rsp[-3]; 1579# context->Rbx = rsp[-2]; 1580# context->R15 = rsp[-1]; 1581# } 1582# } 1583# context->Rsp = (ULONG64)rsp; 1584# context->Rdi = rsp[1]; 1585# context->Rsi = rsp[2]; 1586# 1587# memcpy (disp->ContextRecord,context,sizeof(CONTEXT)); 1588# RtlVirtualUnwind(UNW_FLAG_NHANDLER,disp->ImageBase, 1589# dips->ControlPc,disp->FunctionEntry,disp->ContextRecord, 1590# &disp->HandlerData,&disp->EstablisherFrame,NULL); 1591# return ExceptionContinueSearch; 1592# } 1593# 1594# It's appropriate to implement this handler in assembler, directly in 1595# function's module. In order to do that one has to know members' 1596# offsets in CONTEXT and DISPATCHER_CONTEXT structures and some constant 1597# values. Here they are: 1598# 1599# CONTEXT.Rax 120 1600# CONTEXT.Rcx 128 1601# CONTEXT.Rdx 136 1602# CONTEXT.Rbx 144 1603# CONTEXT.Rsp 152 1604# CONTEXT.Rbp 160 1605# CONTEXT.Rsi 168 1606# CONTEXT.Rdi 176 1607# CONTEXT.R8 184 1608# CONTEXT.R9 192 1609# CONTEXT.R10 200 1610# CONTEXT.R11 208 1611# CONTEXT.R12 216 1612# CONTEXT.R13 224 1613# CONTEXT.R14 232 1614# CONTEXT.R15 240 1615# CONTEXT.Rip 248 1616# CONTEXT.Xmm6 512 1617# sizeof(CONTEXT) 1232 1618# DISPATCHER_CONTEXT.ControlPc 0 1619# DISPATCHER_CONTEXT.ImageBase 8 1620# DISPATCHER_CONTEXT.FunctionEntry 16 1621# DISPATCHER_CONTEXT.EstablisherFrame 24 1622# DISPATCHER_CONTEXT.TargetIp 32 1623# DISPATCHER_CONTEXT.ContextRecord 40 1624# DISPATCHER_CONTEXT.LanguageHandler 48 1625# DISPATCHER_CONTEXT.HandlerData 56 1626# UNW_FLAG_NHANDLER 0 1627# ExceptionContinueSearch 1 1628# 1629# In order to tie the handler to the function one has to compose 1630# couple of structures: one for .xdata segment and one for .pdata. 1631# 1632# UNWIND_INFO structure for .xdata segment would be 1633# 1634# function_unwind_info: 1635# .byte 9,0,0,0 1636# .rva handler 1637# 1638# This structure designates exception handler for a function with 1639# zero-length prologue, no stack frame or frame register. 1640# 1641# To facilitate composing of .pdata structures, auto-generated "gear" 1642# prologue copies rsp value to rax and denotes next instruction with 1643# .LSEH_begin_{function_name} label. This essentially defines the SEH 1644# styling rule mentioned in the beginning. Position of this label is 1645# chosen in such manner that possible exceptions raised in the "gear" 1646# prologue would be accounted to caller and unwound from latter's frame. 1647# End of function is marked with respective .LSEH_end_{function_name} 1648# label. To summarize, .pdata segment would contain 1649# 1650# .rva .LSEH_begin_function 1651# .rva .LSEH_end_function 1652# .rva function_unwind_info 1653# 1654# Reference to function_unwind_info from .xdata segment is the anchor. 1655# In case you wonder why references are 32-bit .rvas and not 64-bit 1656# .quads. References put into these two segments are required to be 1657# *relative* to the base address of the current binary module, a.k.a. 1658# image base. No Win64 module, be it .exe or .dll, can be larger than 1659# 2GB and thus such relative references can be and are accommodated in 1660# 32 bits. 1661# 1662# Having reviewed the example function code, one can argue that "movq 1663# %rsp,%rax" above is redundant. It is not! Keep in mind that on Unix 1664# rax would contain an undefined value. If this "offends" you, use 1665# another register and refrain from modifying rax till magic_point is 1666# reached, i.e. as if it was a non-volatile register. If more registers 1667# are required prior [variable] frame setup is completed, note that 1668# nobody says that you can have only one "magic point." You can 1669# "liberate" non-volatile registers by denoting last stack off-load 1670# instruction and reflecting it in finer grade unwind logic in handler. 1671# After all, isn't it why it's called *language-specific* handler... 1672# 1673# SE handlers are also involved in unwinding stack when executable is 1674# profiled or debugged. Profiling implies additional limitations that 1675# are too subtle to discuss here. For now it's sufficient to say that 1676# in order to simplify handlers one should either a) offload original 1677# %rsp to stack (like discussed above); or b) if you have a register to 1678# spare for frame pointer, choose volatile one. 1679# 1680# (*) Note that we're talking about run-time, not debug-time. Lack of 1681# unwind information makes debugging hard on both Windows and 1682# Unix. "Unlike" refers to the fact that on Unix signal handler 1683# will always be invoked, core dumped and appropriate exit code 1684# returned to parent (for user notification). 1685
