diff --git a/crypto/perlasm/x86_64-xlate.pl b/crypto/perlasm/x86_64-xlate.pl index 91017c69f..a6b718b5e 100755 --- a/crypto/perlasm/x86_64-xlate.pl +++ b/crypto/perlasm/x86_64-xlate.pl @@ -625,7 +625,7 @@ my %globals; /\.comm/ && do { my @str=split(/,\s*/,$line); my $v=undef; if ($nasm) { - $v.="common $prefix@str[0] @str[1]:near"; + $v.="common $prefix@str[0] @str[1]"; } else { $v="$current_segment\tENDS\n" if ($current_segment); $current_segment = ".data"; @@ -757,6 +757,19 @@ my $pclmulqdq = sub { } }; +my $rdrand = sub { + if (shift =~ /%[er](\w+)/) { + my @opcode=(); + my $dst=$1; + if ($dst !~ /[0-9]+/) { $dst = $regrm{"%e$dst"}; } + rex(\@opcode,0,$1,8); + push @opcode,0x0f,0xc7,0xf0|($dst&7); + @opcode; + } else { + (); + } +}; + if ($nasm) { print <<___; default rel diff --git a/crypto/perlasm/x86asm.pl b/crypto/perlasm/x86asm.pl index efa7b61ce..b7ddb8137 100644 --- a/crypto/perlasm/x86asm.pl +++ b/crypto/perlasm/x86asm.pl @@ -123,6 +123,14 @@ sub ::pclmulqdq { &::generic("pclmulqdq",@_); } } +sub ::rdrand +{ my ($dst)=@_; + if ($dst =~ /(e[a-dsd][ixp])/) + { &::data_byte(0x0f,0xc7,0xf0|$regrm{$dst}); } + else + { &::generic("rdrand",@_); } +} + # label management $lbdecor="L"; # local label decoration, set by package $label="000"; diff --git a/crypto/x86_64cpuid.pl b/crypto/x86_64cpuid.pl index 7b76522bd..26b8e5cc0 100644 --- a/crypto/x86_64cpuid.pl +++ b/crypto/x86_64cpuid.pl @@ -7,7 +7,11 @@ if ($flavour =~ /\./) { $output = $flavour; undef $flavour; } $win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/); $0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1; -open STDOUT,"| $^X ${dir}perlasm/x86_64-xlate.pl $flavour $output"; +( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or +( $xlate="${dir}perlasm/x86_64-xlate.pl" and -f $xlate) or +die "can't locate x86_64-xlate.pl"; + +open STDOUT,"| $^X $xlate $flavour $output"; ($arg1,$arg2,$arg3,$arg4)=$win64?("%rcx","%rdx","%r8", "%r9") : # Win64 order ("%rdi","%rsi","%rdx","%rcx"); # Unix order @@ -349,4 +353,21 @@ OPENSSL_instrument_bus2: ___ } +print<<___; +.globl OPENSSL_ia32_rdrand +.type OPENSSL_ia32_rdrand,\@abi-omnipotent +.align 16 +OPENSSL_ia32_rdrand: + mov \$8,%ecx +.Loop_rdrand: + rdrand %rax + jc .Lbreak_rdrand + loop .Loop_rdrand +.Lbreak_rdrand: + cmp \$0,%rax + cmove %rcx,%rax + ret +.size OPENSSL_ia32_rdrand,.-OPENSSL_ia32_rdarnd +___ + close STDOUT; # flush diff --git a/crypto/x86cpuid.pl b/crypto/x86cpuid.pl index 70c8a2d67..6595ff35f 100644 --- a/crypto/x86cpuid.pl +++ b/crypto/x86cpuid.pl @@ -442,6 +442,18 @@ my $max = "ebp"; &function_end("OPENSSL_instrument_bus2"); } +&function_begin_B("OPENSSL_ia32_rdrand"); + &mov ("ecx",8); +&set_label("loop"); + &rdrand ("eax"); + &jc (&label("break")); + &loop (&label("loop")); +&set_label("break"); + &cmp ("eax",0); + &cmove ("eax","ecx"); + &ret (); +&function_end_B("OPENSSL_ia32_rdrand"); + &initseg("OPENSSL_cpuid_setup"); &asm_finish(); diff --git a/doc/crypto/OPENSSL_ia32cap.pod b/doc/crypto/OPENSSL_ia32cap.pod index 3f6458c6b..17189ed84 100644 --- a/doc/crypto/OPENSSL_ia32cap.pod +++ b/doc/crypto/OPENSSL_ia32cap.pod @@ -51,6 +51,8 @@ moment of this writing following bits are significant: =item bit #60 denoting AVX extension; +=item bit #62 denoting availability of RDRAND instruction; + For example, clearing bit #26 at run-time disables high-performance SSE2 code present in the crypto library, while clearing bit #24 disables SSE2 code operating on 128-bit XMM register bank. You might