-#!/usr/bin/env perl
+#! /usr/bin/env perl
+# Copyright 2004-2016 The OpenSSL Project Authors. All Rights Reserved.
+#
+# Licensed under the OpenSSL license (the "License"). You may not use
+# this file except in compliance with the License. You can obtain a copy
+# in the file LICENSE in the source distribution or at
+# https://www.openssl.org/source/license.html
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC, "${dir}perlasm", "perlasm");
require "x86asm.pl";
+$output = pop;
+open OUT,">$output";
+*STDOUT=*OUT;
+
&asm_init($ARGV[0],"x86cpuid");
for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
&pushf ();
&pop ("eax");
&xor ("ecx","eax");
- &bt ("ecx",21);
- &jnc (&label("generic"));
&xor ("eax","eax");
+ &mov ("esi",&wparam(0));
+ &mov (&DWP(8,"esi"),"eax"); # clear extended feature flags
+ &bt ("ecx",21);
+ &jnc (&label("nocpuid"));
&cpuid ();
&mov ("edi","eax"); # max value for standard query level
&inc ("esi"); # number of cores
&mov ("eax",1);
+ &xor ("ecx","ecx");
&cpuid ();
&bt ("edx",28);
&jnc (&label("generic"));
&set_label("intel");
&cmp ("edi",4);
- &mov ("edi",-1);
+ &mov ("esi",-1);
&jb (&label("nocacheinfo"));
&mov ("eax",4);
&mov ("ecx",0); # query L1D
&cpuid ();
- &mov ("edi","eax");
- &shr ("edi",14);
- &and ("edi",0xfff); # number of cores -1 per L1D
+ &mov ("esi","eax");
+ &shr ("esi",14);
+ &and ("esi",0xfff); # number of cores -1 per L1D
&set_label("nocacheinfo");
&mov ("eax",1);
+ &xor ("ecx","ecx");
&cpuid ();
+ &and ("edx",0xbfefffff); # force reserved bits #20, #30 to 0
&cmp ("ebp",0);
- &jne (&label("notP4"));
+ &jne (&label("notintel"));
+ &or ("edx",1<<30); # set reserved bit#30 on Intel CPUs
&and (&HB("eax"),15); # familiy ID
&cmp (&HB("eax"),15); # P4?
- &jne (&label("notP4"));
- &or ("edx",1<<20); # use reserved bit to engage RC4_CHAR
-&set_label("notP4");
+ &jne (&label("notintel"));
+ &or ("edx",1<<20); # set reserved bit#20 to engage RC4_CHAR
+&set_label("notintel");
&bt ("edx",28); # test hyper-threading bit
&jnc (&label("generic"));
&and ("edx",0xefffffff);
- &cmp ("edi",0);
+ &cmp ("esi",0);
&je (&label("generic"));
&or ("edx",0x10000000);
&set_label("generic");
&and ("ebp",1<<11); # isolate AMD XOP flag
- &and ("ecx",~(1<<11));
- &mov ("esi","edx");
+ &and ("ecx",0xfffff7ff); # force 11th bit to 0
+ &mov ("esi","edx"); # %ebp:%esi is copy of %ecx:%edx
&or ("ebp","ecx"); # merge AMD XOP flag
- &bt ("ecx",26); # check XSAVE bit
- &jnc (&label("done"));
- &bt ("ecx",27); # check OSXSAVE bit
- &jnc (&label("clear_xmm"));
+ &cmp ("edi",7);
+ &mov ("edi",&wparam(0));
+ &jb (&label("no_extended_info"));
+ &mov ("eax",7);
+ &xor ("ecx","ecx");
+ &cpuid ();
+ &mov (&DWP(8,"edi"),"ebx"); # save extended feature flag
+&set_label("no_extended_info");
+
+ &bt ("ebp",27); # check OSXSAVE bit
+ &jnc (&label("clear_avx"));
&xor ("ecx","ecx");
&data_byte(0x0f,0x01,0xd0); # xgetbv
&and ("eax",6);
&cmp ("eax",2);
&je (&label("clear_avx"));
&set_label("clear_xmm");
- &and ("ebp",~(1<<25|1<<1)); # clear AESNI and PCLMULQDQ bits
- &and ("esi",~(1<<24)); # clear FXSR
+ &and ("ebp",0xfdfffffd); # clear AESNI and PCLMULQDQ bits
+ &and ("esi",0xfeffffff); # clear FXSR
&set_label("clear_avx");
- &and ("ebp",~(1<<28|1<<12|1<<11));# clear AVX, FMA and AMD XOP bits
+ &and ("ebp",0xefffe7ff); # clear AVX, FMA and AMD XOP bits
+ &and (&DWP(8,"edi"),0xffffffdf); # clear AVX2
&set_label("done");
&mov ("eax","esi");
&mov ("edx","ebp");
+&set_label("nocpuid");
&function_end("OPENSSL_ia32_cpuid");
&external_label("OPENSSL_ia32cap_P");
&jnz (&label("nohalt")); # not enough privileges
&pushf ();
- &pop ("eax")
+ &pop ("eax");
&bt ("eax",9);
&jnc (&label("nohalt")); # interrupts are disabled
&function_begin_B("OPENSSL_far_spin");
&pushf ();
- &pop ("eax")
+ &pop ("eax");
&bt ("eax",9);
&jnc (&label("nospin")); # interrupts are disabled
&set_label("spin");
&lea ("ebx",&DWP(0,"eax","ecx"));
&nop ();
- &data_word(0x1ab10ff0); # lock; cmpxchg %ebx,(%edx) # %eax is envolved and is always reloaded
+ &data_word(0x1ab10ff0); # lock; cmpxchg %ebx,(%edx) # %eax is involved and is always reloaded
&jne (&label("spin"));
&mov ("eax","ebx"); # OpenSSL expects the new value
&pop ("ebx");
# arguments is 1 or 2!
&function_begin_B("OPENSSL_indirect_call");
{
- my $i,$max=7; # $max has to be chosen as 4*n-1
+ my ($max,$i)=(7,); # $max has to be chosen as 4*n-1
# in order to preserve eventual
# stack alignment
&push ("ebp");
&ret ();
&function_end_B("OPENSSL_cleanse");
+&function_begin_B("CRYPTO_memcmp");
+ &push ("esi");
+ &push ("edi");
+ &mov ("esi",&wparam(0));
+ &mov ("edi",&wparam(1));
+ &mov ("ecx",&wparam(2));
+ &xor ("eax","eax");
+ &xor ("edx","edx");
+ &cmp ("ecx",0);
+ &je (&label("no_data"));
+&set_label("loop");
+ &mov ("dl",&BP(0,"esi"));
+ &lea ("esi",&DWP(1,"esi"));
+ &xor ("dl",&BP(0,"edi"));
+ &lea ("edi",&DWP(1,"edi"));
+ &or ("al","dl");
+ &dec ("ecx");
+ &jnz (&label("loop"));
+ &neg ("eax");
+ &shr ("eax",31);
+&set_label("no_data");
+ &pop ("edi");
+ &pop ("esi");
+ &ret ();
+&function_end_B("CRYPTO_memcmp");
{
my $lasttick = "esi";
my $lastdiff = "ebx";
&function_end("OPENSSL_instrument_bus2");
}
+sub gen_random {
+my $rdop = shift;
+&function_begin_B("OPENSSL_ia32_${rdop}");
+ &mov ("ecx",8);
+&set_label("loop");
+ &${rdop}("eax");
+ &jc (&label("break"));
+ &loop (&label("loop"));
+&set_label("break");
+ &cmp ("eax",0);
+ &cmove ("eax","ecx");
+ &ret ();
+&function_end_B("OPENSSL_ia32_${rdop}");
+
+&function_begin_B("OPENSSL_ia32_${rdop}_bytes");
+ &push ("edi");
+ &push ("ebx");
+ &xor ("eax","eax"); # return value
+ &mov ("edi",&wparam(0));
+ &mov ("ebx",&wparam(1));
+
+ &cmp ("ebx",0);
+ &je (&label("done"));
+
+ &mov ("ecx",8);
+&set_label("loop");
+ &${rdop}("edx");
+ &jc (&label("break"));
+ &loop (&label("loop"));
+ &jmp (&label("done"));
+
+&set_label("break",16);
+ &cmp ("ebx",4);
+ &jb (&label("tail"));
+ &mov (&DWP(0,"edi"),"edx");
+ &lea ("edi",&DWP(4,"edi"));
+ &add ("eax",4);
+ &sub ("ebx",4);
+ &jz (&label("done"));
+ &mov ("ecx",8);
+ &jmp (&label("loop"));
+
+&set_label("tail",16);
+ &mov (&BP(0,"edi"),"dl");
+ &lea ("edi",&DWP(1,"edi"));
+ &inc ("eax");
+ &shr ("edx",8);
+ &dec ("ebx");
+ &jnz (&label("tail"));
+
+&set_label("done");
+ &pop ("ebx");
+ &pop ("edi");
+ &ret ();
+&function_end_B("OPENSSL_ia32_${rdop}_bytes");
+}
+&gen_random("rdrand");
+&gen_random("rdseed");
+
&initseg("OPENSSL_cpuid_setup");
+&hidden("OPENSSL_cpuid_setup");
+&hidden("OPENSSL_ia32cap_P");
+
&asm_finish();
+
+close STDOUT;