X-Git-Url: https://git.openssl.org/gitweb/?p=openssl.git;a=blobdiff_plain;f=crypto%2Fx86cpuid.pl;h=ef1216a8b2ed4cc8ee5fc35667114ce36017b1aa;hp=0513398739f0dc04a2f8e2c2fa514bee31fe62ee;hb=a2ea9f3ecc4887add86c21774af80654b41707ca;hpb=5fabb88a7816f19090384e45bb8f2a22c7f290fb diff --git a/crypto/x86cpuid.pl b/crypto/x86cpuid.pl index 0513398739..ef1216a8b2 100644 --- a/crypto/x86cpuid.pl +++ b/crypto/x86cpuid.pl @@ -19,9 +19,11 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } &pushf (); &pop ("eax"); &xor ("ecx","eax"); - &bt ("ecx",21); - &jnc (&label("done")); &xor ("eax","eax"); + &bt ("ecx",21); + &jnc (&label("nocpuid")); + &mov ("esi",&wparam(0)); + &mov (&DWP(8,"esi"),"eax"); # clear 3rd word &cpuid (); &mov ("edi","eax"); # max value for standard query level @@ -51,7 +53,14 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } # AMD specific &mov ("eax",0x80000000); &cpuid (); - &cmp ("eax",0x80000008); + &cmp ("eax",0x80000001); + &jb (&label("intel")); + &mov ("esi","eax"); + &mov ("eax",0x80000001); + &cpuid (); + &or ("ebp","ecx"); + &and ("ebp",1<<11|1); # isolate XOP bit + &cmp ("esi",0x80000008); &jb (&label("intel")); &mov ("eax",0x80000008); @@ -60,17 +69,28 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } &inc ("esi"); # number of cores &mov ("eax",1); + &xor ("ecx","ecx"); &cpuid (); &bt ("edx",28); - &jnc (&label("done")); + &jnc (&label("generic")); &shr ("ebx",16); &and ("ebx",0xff); &cmp ("ebx","esi"); - &ja (&label("done")); + &ja (&label("generic")); &and ("edx",0xefffffff); # clear hyper-threading bit - &jmp (&label("done")); + &jmp (&label("generic")); &set_label("intel"); + &cmp ("edi",7); + &jb (&label("cacheinfo")); + + &mov ("esi",&wparam(0)); + &mov ("eax",7); + &xor ("ecx","ecx"); + &cpuid (); + &mov (&DWP(8,"esi"),"ebx"); + +&set_label("cacheinfo"); &cmp ("edi",4); &mov ("edi",-1); &jb (&label("nocacheinfo")); @@ -84,28 +104,55 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } &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("done")); + &jnc (&label("generic")); &and ("edx",0xefffffff); &cmp ("edi",0); - &je (&label("done")); + &je (&label("generic")); &or ("edx",0x10000000); &shr ("ebx",16); &cmp (&LB("ebx"),1); - &ja (&label("done")); + &ja (&label("generic")); &and ("edx",0xefffffff); # clear hyper-threading bit if not + +&set_label("generic"); + &and ("ebp",1<<11); # isolate AMD XOP flag + &and ("ecx",0xfffff7ff); # force 11th bit to 0 + &mov ("esi","edx"); + &or ("ebp","ecx"); # merge AMD XOP flag + + &bt ("ecx",27); # check OSXSAVE bit + &jnc (&label("clear_avx")); + &xor ("ecx","ecx"); + &data_byte(0x0f,0x01,0xd0); # xgetbv + &and ("eax",6); + &cmp ("eax",6); + &je (&label("done")); + &cmp ("eax",2); + &je (&label("clear_avx")); +&set_label("clear_xmm"); + &and ("ebp",0xfdfffffd); # clear AESNI and PCLMULQDQ bits + &and ("esi",0xfeffffff); # clear FXSR +&set_label("clear_avx"); + &and ("ebp",0xefffe7ff); # clear AVX, FMA and AMD XOP bits + &mov ("edi",&wparam(0)); + &and (&DWP(8,"edi"),0xffffffdf); # clear AVX2 &set_label("done"); - &mov ("eax","edx"); - &mov ("edx","ecx"); + &mov ("eax","esi"); + &mov ("edx","ebp"); +&set_label("nocpuid"); &function_end("OPENSSL_ia32_cpuid"); &external_label("OPENSSL_ia32cap_P"); @@ -134,7 +181,7 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } &jnz (&label("nohalt")); # not enough privileges &pushf (); - &pop ("eax") + &pop ("eax"); &bt ("eax",9); &jnc (&label("nohalt")); # interrupts are disabled @@ -165,7 +212,7 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } &function_begin_B("OPENSSL_far_spin"); &pushf (); - &pop ("eax") + &pop ("eax"); &bt ("eax",9); &jnc (&label("nospin")); # interrupts are disabled @@ -199,8 +246,9 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } &bt (&DWP(0,"ecx"),1); &jnc (&label("no_x87")); if ($sse2) { - &bt (&DWP(0,"ecx"),26); - &jnc (&label("no_sse2")); + &and ("ecx",1<<26|1<<24); # check SSE2 and FXSR bits + &cmp ("ecx",1<<26|1<<24); + &jne (&label("no_sse2")); &pxor ("xmm0","xmm0"); &pxor ("xmm1","xmm1"); &pxor ("xmm2","xmm2"); @@ -248,7 +296,7 @@ for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); } # 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"); @@ -331,7 +379,7 @@ my $max = "ebp"; &mov ($lasttick,"eax"); # lasttick = tick &mov ($lastdiff,0); # lastdiff = 0 &clflush(&DWP(0,$out)); - &lock (); + &data_byte(0xf0); # lock &add (&DWP(0,$out),$lastdiff); &jmp (&label("loop")); @@ -342,7 +390,7 @@ my $max = "ebp"; &mov ($lasttick,"edx"); # lasttick = tick &mov ($lastdiff,"eax"); # lastdiff = diff &clflush(&DWP(0,$out)); - &lock (); + &data_byte(0xf0); # lock &add (&DWP(0,$out),"eax"); # accumulate diff &lea ($out,&DWP(4,$out)); # ++$out &sub ($cnt,1); # --$cnt @@ -371,7 +419,7 @@ my $max = "ebp"; &mov ($lastdiff,0); # lastdiff = 0 &clflush(&DWP(0,$out)); - &lock (); + &data_byte(0xf0); # lock &add (&DWP(0,$out),$lastdiff); &rdtsc (); # collect 1st diff @@ -383,7 +431,7 @@ my $max = "ebp"; &set_label("loop2",16); &clflush(&DWP(0,$out)); - &lock (); + &data_byte(0xf0); # lock &add (&DWP(0,$out),"eax"); # accumulate diff &sub ($max,1); @@ -409,6 +457,21 @@ 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"); +&hidden("OPENSSL_cpuid_setup"); +&hidden("OPENSSL_ia32cap_P"); + &asm_finish();