2 # Copyright 2005-2018 The OpenSSL Project Authors. All Rights Reserved.
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
10 # $output is the last argument if it looks like a file (it has an extension)
11 # $flavour is the first argument if it doesn't look like a file
12 $output = $#ARGV >= 0 && $ARGV[$#ARGV] =~ m|\.\w+$| ? pop : undef;
13 $flavour = $#ARGV >= 0 && $ARGV[0] !~ m|\.| ? shift : undef;
15 $win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
17 $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
18 ( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
19 ( $xlate="${dir}perlasm/x86_64-xlate.pl" and -f $xlate) or
20 die "can't locate x86_64-xlate.pl";
22 open OUT,"| \"$^X\" \"$xlate\" $flavour \"$output\""
23 or die "can't call $xlate: $!";
26 ($arg1,$arg2,$arg3,$arg4)=$win64?("%rcx","%rdx","%r8", "%r9") : # Win64 order
27 ("%rdi","%rsi","%rdx","%rcx"); # Unix order
30 .extern OPENSSL_cpuid_setup
31 .hidden OPENSSL_cpuid_setup
33 call OPENSSL_cpuid_setup
35 .hidden OPENSSL_ia32cap_P
36 .comm OPENSSL_ia32cap_P,16,4
40 .globl OPENSSL_atomic_add
41 .type OPENSSL_atomic_add,\@abi-omnipotent
46 .Lspin: leaq ($arg2,%rax),%r8
51 .byte 0x48,0x98 # cltq/cdqe
54 .size OPENSSL_atomic_add,.-OPENSSL_atomic_add
57 .type OPENSSL_rdtsc,\@abi-omnipotent
66 .size OPENSSL_rdtsc,.-OPENSSL_rdtsc
68 .globl OPENSSL_ia32_cpuid
69 .type OPENSSL_ia32_cpuid,\@function,1
73 mov %rbx,%r8 # save %rbx
74 .cfi_register %rbx,%r8
77 mov %rax,8(%rdi) # clear extended feature flags
79 mov %eax,%r11d # max value for standard query level
82 cmp \$0x756e6547,%ebx # "Genu"
85 cmp \$0x49656e69,%edx # "ineI"
88 cmp \$0x6c65746e,%ecx # "ntel"
90 or %eax,%r9d # 0 indicates Intel CPU
93 cmp \$0x68747541,%ebx # "Auth"
96 cmp \$0x69746E65,%edx # "enti"
99 cmp \$0x444D4163,%ecx # "cAMD"
101 or %eax,%r10d # 0 indicates AMD CPU
105 mov \$0x80000000,%eax
107 cmp \$0x80000001,%eax
110 mov \$0x80000001,%eax
113 and \$0x00000801,%r9d # isolate AMD XOP bit, 1<<11
115 cmp \$0x80000008,%r10d
118 mov \$0x80000008,%eax
120 movzb %cl,%r10 # number of cores - 1
121 inc %r10 # number of cores
125 bt \$28,%edx # test hyper-threading bit
127 shr \$16,%ebx # number of logical processors
130 and \$0xefffffff,%edx # ~(1<<28)
139 mov \$0,%ecx # query L1D
143 and \$0xfff,%r10d # number of cores -1 per L1D
148 movd %eax,%xmm0 # put aside processor id
149 and \$0xbfefffff,%edx # force reserved bits to 0
152 or \$0x40000000,%edx # set reserved bit#30 on Intel CPUs
154 cmp \$15,%ah # examine Family ID
156 or \$0x00100000,%edx # set reserved bit#20 to engage RC4_CHAR
160 and \$0x0fff0ff0,%eax
161 cmp \$0x00050670,%eax # Knights Landing
163 cmp \$0x00080650,%eax # Knights Mill (according to sde)
166 and \$0xfbffffff,%ecx # clear XSAVE flag to mimic Silvermont
169 bt \$28,%edx # test hyper-threading bit
171 and \$0xefffffff,%edx # ~(1<<28)
175 or \$0x10000000,%edx # 1<<28
177 cmp \$1,%bl # see if cache is shared
179 and \$0xefffffff,%edx # ~(1<<28)
181 and \$0x00000800,%r9d # isolate AMD XOP flag
182 and \$0xfffff7ff,%ecx
183 or %ecx,%r9d # merge AMD XOP flag
185 mov %edx,%r10d # %r9d:%r10d is copy of %ecx:%edx
188 jb .Lno_extended_info
192 bt \$26,%r9d # check XSAVE bit, cleared on Knights
194 and \$0xfff7ffff,%ebx # clear ADCX/ADOX flag
196 movd %xmm0,%eax # restore processor id
197 and \$0x0fff0ff0,%eax
198 cmp \$0x00050650,%eax # Skylake-X
200 and \$0xfffeffff,%ebx # ~(1<<16)
201 # suppress AVX512F flag on Skylake-X
203 mov %ebx,8(%rdi) # save extended feature flags
207 bt \$27,%r9d # check OSXSAVE bit
210 .byte 0x0f,0x01,0xd0 # xgetbv
211 and \$0xe6,%eax # isolate XMM, YMM and ZMM state support
214 andl \$0x3fdeffff,8(%rdi) # ~(1<<31|1<<30|1<<21|1<<16)
215 # clear AVX512F+BW+VL+FIMA, all of
216 # them are EVEX-encoded, which requires
217 # ZMM state support even if one uses
218 # only XMM and YMM :-(
219 and \$6,%eax # isolate XMM and YMM state support
223 mov \$0xefffe7ff,%eax # ~(1<<28|1<<12|1<<11)
224 and %eax,%r9d # clear AVX, FMA and AMD XOP bits
225 mov \$0x3fdeffdf,%eax # ~(1<<31|1<<30|1<<21|1<<16|1<<5)
226 and %eax,8(%rdi) # clear AVX2 and AVX512* bits
230 mov %r8,%rbx # restore %rbx
235 .size OPENSSL_ia32_cpuid,.-OPENSSL_ia32_cpuid
237 .globl OPENSSL_cleanse
238 .type OPENSSL_cleanse,\@abi-omnipotent
272 .size OPENSSL_cleanse,.-OPENSSL_cleanse
275 .type CRYPTO_memcmp,\@abi-omnipotent
308 .size CRYPTO_memcmp,.-CRYPTO_memcmp
311 print<<___ if (!$win64);
312 .globl OPENSSL_wipe_cpu
313 .type OPENSSL_wipe_cpu,\@abi-omnipotent
344 .size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
346 print<<___ if ($win64);
347 .globl OPENSSL_wipe_cpu
348 .type OPENSSL_wipe_cpu,\@abi-omnipotent
365 .size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
373 my $redzone=win64?8:-8;
376 .globl OPENSSL_instrument_bus
377 .type OPENSSL_instrument_bus,\@abi-omnipotent
379 OPENSSL_instrument_bus:
381 mov $arg1,$out # tribute to Win64
385 rdtsc # collect 1st tick
386 mov %eax,$lasttick # lasttick = tick
387 mov \$0,$lastdiff # lastdiff = 0
408 .size OPENSSL_instrument_bus,.-OPENSSL_instrument_bus
410 .globl OPENSSL_instrument_bus2
411 .type OPENSSL_instrument_bus2,\@abi-omnipotent
413 OPENSSL_instrument_bus2:
415 mov $arg1,$out # tribute to Win64
418 mov $cnt,$redzone(%rsp)
420 rdtsc # collect 1st tick
421 mov %eax,$lasttick # lasttick = tick
422 mov \$0,$lastdiff # lastdiff = 0
428 rdtsc # collect 1st diff
430 sub $lasttick,%eax # diff
431 mov %edx,$lasttick # lasttick = tick
432 mov %eax,$lastdiff # lastdiff = diff
436 add %eax,($out) # accumulate diff
443 sub $lasttick,%eax # diff
444 mov %edx,$lasttick # lasttick = tick
446 mov %eax,$lastdiff # lastdiff = diff
449 sub %rdx,$cnt # conditional --$cnt
450 lea ($out,%rdx,4),$out # conditional ++$out
454 mov $redzone(%rsp),%rax
458 .size OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
465 .globl OPENSSL_ia32_${rdop}_bytes
466 .type OPENSSL_ia32_${rdop}_bytes,\@abi-omnipotent
468 OPENSSL_ia32_${rdop}_bytes:
470 xor %rax, %rax # return value
472 je .Ldone_${rdop}_bytes
477 jc .Lbreak_${rdop}_bytes
479 jnz .Loop_${rdop}_bytes
480 jmp .Ldone_${rdop}_bytes
483 .Lbreak_${rdop}_bytes:
485 jb .Ltail_${rdop}_bytes
490 jz .Ldone_${rdop}_bytes
492 jmp .Loop_${rdop}_bytes
495 .Ltail_${rdop}_bytes:
501 jnz .Ltail_${rdop}_bytes
503 .Ldone_${rdop}_bytes:
504 xor %r10,%r10 # Clear sensitive data from register
507 .size OPENSSL_ia32_${rdop}_bytes,.-OPENSSL_ia32_${rdop}_bytes
510 gen_random("rdrand");
511 gen_random("rdseed");
513 close STDOUT or die "error closing STDOUT"; # flush