/*

 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.

 *

 * This code is free software; you can redistribute it and/or modify it

 * under the terms of the GNU General Public License version 2 only, as

 * published by the Free Software Foundation.

 *

 * This code is distributed in the hope that it will be useful, but WITHOUT

 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or

 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License

 * version 2 for more details (a copy is included in the LICENSE file that

 * accompanied this code).

 *

 * You should have received a copy of the GNU General Public License version

 * 2 along with this work; if not, write to the Free Software Foundation,

 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.

 *

 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA

 * or visit www.oracle.com if you need additional information or have any

 * questions.

 *

 */

#include "precompiled.hpp"

#include "asm/macroAssembler.hpp"

#include "asm/macroAssembler.inline.hpp"

#include "memory/resourceArea.hpp"

#include "runtime/java.hpp"

#include "runtime/stubCodeGenerator.hpp"

#include "vm_version_x86.hpp"

#ifdef TARGET_OS_FAMILY_linux

# include "os_linux.inline.hpp"

#endif

#ifdef TARGET_OS_FAMILY_solaris

# include "os_solaris.inline.hpp"

#endif

#ifdef TARGET_OS_FAMILY_windows

# include "os_windows.inline.hpp"

#endif

#ifdef TARGET_OS_FAMILY_bsd

# include "os_bsd.inline.hpp"

#endif

int VM_Version::_cpu;

int VM_Version::_model;

int VM_Version::_stepping;

int VM_Version::_cpuFeatures;

const char*           VM_Version::_features_str = "";

VM_Version::CpuidInfo VM_Version::_cpuid_info   = { 0, };

// Address of instruction which causes SEGV

address VM_Version::_cpuinfo_segv_addr = 0;

// Address of instruction after the one which causes SEGV

address VM_Version::_cpuinfo_cont_addr = 0;

static BufferBlob* stub_blob;

static const int stub_size = 600;

extern "C" {

  typedef void (*get_cpu_info_stub_t)(void*);

}

static get_cpu_info_stub_t get_cpu_info_stub = NULL;

class VM_Version_StubGenerator: public StubCodeGenerator {

 public:

  VM_Version_StubGenerator(CodeBuffer *c) : StubCodeGenerator(c) {}

  address generate_get_cpu_info() {

    // Flags to test CPU type.

    const uint32_t HS_EFL_AC           = 0x40000;

    const uint32_t HS_EFL_ID           = 0x200000;

    // Values for when we don't have a CPUID instruction.

    const int      CPU_FAMILY_SHIFT = 8;

    const uint32_t CPU_FAMILY_386   = (3 << CPU_FAMILY_SHIFT);

    const uint32_t CPU_FAMILY_486   = (4 << CPU_FAMILY_SHIFT);

    Label detect_486, cpu486, detect_586, std_cpuid1, std_cpuid4;

    Label sef_cpuid, ext_cpuid, ext_cpuid1, ext_cpuid5, ext_cpuid7, done;

    StubCodeMark mark(this, "VM_Version", "get_cpu_info_stub");

#   define __ _masm->

    address start = __ pc();

    //

    // void get_cpu_info(VM_Version::CpuidInfo* cpuid_info);

    //

    // LP64: rcx and rdx are first and second argument registers on windows

    __ push(rbp);

#ifdef _LP64

    __ mov(rbp, c_rarg0); // cpuid_info address

#else

    __ movptr(rbp, Address(rsp, 8)); // cpuid_info address

#endif

    __ push(rbx);

    __ push(rsi);

    __ pushf();          // preserve rbx, and flags

    __ pop(rax);

    __ push(rax);

    __ mov(rcx, rax);

    //

    // if we are unable to change the AC flag, we have a 386

    //

    __ xorl(rax, HS_EFL_AC);

    __ push(rax);

    __ popf();

    __ pushf();

    __ pop(rax);

    __ cmpptr(rax, rcx);

    __ jccb(Assembler::notEqual, detect_486);

    __ movl(rax, CPU_FAMILY_386);

    __ movl(Address(rbp, in_bytes(VM_Version::std_cpuid1_offset())), rax);

    __ jmp(done);

    //

    // If we are unable to change the ID flag, we have a 486 which does

    // not support the "cpuid" instruction.

    //

    __ bind(detect_486);

    __ mov(rax, rcx);

    __ xorl(rax, HS_EFL_ID);

    __ push(rax);

    __ popf();

    __ pushf();

    __ pop(rax);

    __ cmpptr(rcx, rax);

    __ jccb(Assembler::notEqual, detect_586);

    __ bind(cpu486);

    __ movl(rax, CPU_FAMILY_486);

    __ movl(Address(rbp, in_bytes(VM_Version::std_cpuid1_offset())), rax);

    __ jmp(done);

    //

    // At this point, we have a chip which supports the "cpuid" instruction

    //

    __ bind(detect_586);

    __ xorl(rax, rax);

    __ cpuid();

    __ orl(rax, rax);

    __ jcc(Assembler::equal, cpu486);   // if cpuid doesn't support an input

                                        // value of at least 1, we give up and

                                        // assume a 486

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::std_cpuid0_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    __ cmpl(rax, 0xa);                  // Is cpuid(0xB) supported?

    __ jccb(Assembler::belowEqual, std_cpuid4);

    //

    // cpuid(0xB) Processor Topology

    //

    __ movl(rax, 0xb);

    __ xorl(rcx, rcx);   // Threads level

    __ cpuid();

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::tpl_cpuidB0_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    __ movl(rax, 0xb);

    __ movl(rcx, 1);     // Cores level

    __ cpuid();

    __ push(rax);

    __ andl(rax, 0x1f);  // Determine if valid topology level

    __ orl(rax, rbx);    // eax[4:0] | ebx[0:15] == 0 indicates invalid level

    __ andl(rax, 0xffff);

    __ pop(rax);

    __ jccb(Assembler::equal, std_cpuid4);

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::tpl_cpuidB1_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    __ movl(rax, 0xb);

    __ movl(rcx, 2);     // Packages level

    __ cpuid();

    __ push(rax);

    __ andl(rax, 0x1f);  // Determine if valid topology level

    __ orl(rax, rbx);    // eax[4:0] | ebx[0:15] == 0 indicates invalid level

    __ andl(rax, 0xffff);

    __ pop(rax);

    __ jccb(Assembler::equal, std_cpuid4);

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::tpl_cpuidB2_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    //

    // cpuid(0x4) Deterministic cache params

    //

    __ bind(std_cpuid4);

    __ movl(rax, 4);

    __ cmpl(rax, Address(rbp, in_bytes(VM_Version::std_cpuid0_offset()))); // Is cpuid(0x4) supported?

    __ jccb(Assembler::greater, std_cpuid1);

    __ xorl(rcx, rcx);   // L1 cache

    __ cpuid();

    __ push(rax);

    __ andl(rax, 0x1f);  // Determine if valid cache parameters used

    __ orl(rax, rax);    // eax[4:0] == 0 indicates invalid cache

    __ pop(rax);

    __ jccb(Assembler::equal, std_cpuid1);

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::dcp_cpuid4_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    //

    // Standard cpuid(0x1)

    //

    __ bind(std_cpuid1);

    __ movl(rax, 1);

    __ cpuid();

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::std_cpuid1_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    //

    // Check if OS has enabled XGETBV instruction to access XCR0

    // (OSXSAVE feature flag) and CPU supports AVX

    //

    __ andl(rcx, 0x18000000); // cpuid1 bits osxsave | avx

    __ cmpl(rcx, 0x18000000);

    __ jccb(Assembler::notEqual, sef_cpuid); // jump if AVX is not supported

    //

    // XCR0, XFEATURE_ENABLED_MASK register

    //

    __ xorl(rcx, rcx);   // zero for XCR0 register

    __ xgetbv();

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::xem_xcr0_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rdx);

    __ andl(rax, 0x6); // xcr0 bits sse | ymm

    __ cmpl(rax, 0x6);

    __ jccb(Assembler::notEqual, sef_cpuid); // jump if AVX is not supported

    //

    // Some OSs have a bug when upper 128bits of YMM

    // registers are not restored after a signal processing.

    // Generate SEGV here (reference through NULL)

    // and check upper YMM bits after it.

    //

    VM_Version::set_avx_cpuFeatures(); // Enable temporary to pass asserts

    intx saved_useavx = UseAVX;

    intx saved_usesse = UseSSE;

    UseAVX = 1;

    UseSSE = 2;

    // load value into all 32 bytes of ymm7 register

    __ movl(rcx, VM_Version::ymm_test_value());

    __ movdl(xmm0, rcx);

    __ pshufd(xmm0, xmm0, 0x00);

    __ vinsertf128h(xmm0, xmm0, xmm0);

    __ vmovdqu(xmm7, xmm0);

#ifdef _LP64

    __ vmovdqu(xmm8,  xmm0);

    __ vmovdqu(xmm15, xmm0);

#endif

    __ xorl(rsi, rsi);

    VM_Version::set_cpuinfo_segv_addr( __ pc() );

    // Generate SEGV

    __ movl(rax, Address(rsi, 0));

    VM_Version::set_cpuinfo_cont_addr( __ pc() );

    // Returns here after signal. Save xmm0 to check it later.

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::ymm_save_offset())));

    __ vmovdqu(Address(rsi,  0), xmm0);

    __ vmovdqu(Address(rsi, 32), xmm7);

#ifdef _LP64

    __ vmovdqu(Address(rsi, 64), xmm8);

    __ vmovdqu(Address(rsi, 96), xmm15);

#endif

    VM_Version::clean_cpuFeatures();

    UseAVX = saved_useavx;

    UseSSE = saved_usesse;

    //

    // cpuid(0x7) Structured Extended Features

    //

    __ bind(sef_cpuid);

    __ movl(rax, 7);

    __ cmpl(rax, Address(rbp, in_bytes(VM_Version::std_cpuid0_offset()))); // Is cpuid(0x7) supported?

    __ jccb(Assembler::greater, ext_cpuid);

    __ xorl(rcx, rcx);

    __ cpuid();

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::sef_cpuid7_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    //

    // Extended cpuid(0x80000000)

    //

    __ bind(ext_cpuid);

    __ movl(rax, 0x80000000);

    __ cpuid();

    __ cmpl(rax, 0x80000000);     // Is cpuid(0x80000001) supported?

    __ jcc(Assembler::belowEqual, done);

    __ cmpl(rax, 0x80000004);     // Is cpuid(0x80000005) supported?

    __ jccb(Assembler::belowEqual, ext_cpuid1);

    __ cmpl(rax, 0x80000006);     // Is cpuid(0x80000007) supported?

    __ jccb(Assembler::belowEqual, ext_cpuid5);

    __ cmpl(rax, 0x80000007);     // Is cpuid(0x80000008) supported?

    __ jccb(Assembler::belowEqual, ext_cpuid7);

    //

    // Extended cpuid(0x80000008)

    //

    __ movl(rax, 0x80000008);

    __ cpuid();

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::ext_cpuid8_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    //

    // Extended cpuid(0x80000007)

    //

    __ bind(ext_cpuid7);

    __ movl(rax, 0x80000007);

    __ cpuid();

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::ext_cpuid7_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    //

    // Extended cpuid(0x80000005)

    //

    __ bind(ext_cpuid5);

    __ movl(rax, 0x80000005);

    __ cpuid();

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::ext_cpuid5_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    //

    // Extended cpuid(0x80000001)

    //

    __ bind(ext_cpuid1);

    __ movl(rax, 0x80000001);

    __ cpuid();

    __ lea(rsi, Address(rbp, in_bytes(VM_Version::ext_cpuid1_offset())));

    __ movl(Address(rsi, 0), rax);

    __ movl(Address(rsi, 4), rbx);

    __ movl(Address(rsi, 8), rcx);

    __ movl(Address(rsi,12), rdx);

    //

    // return

    //

    __ bind(done);

    __ popf();

    __ pop(rsi);

    __ pop(rbx);

    __ pop(rbp);

    __ ret(0);

#   undef __

    return start;

  };

};

void VM_Version::get_cpu_info_wrapper() {

  get_cpu_info_stub(&_cpuid_info);

}

#ifndef CALL_TEST_FUNC_WITH_WRAPPER_IF_NEEDED

  #define CALL_TEST_FUNC_WITH_WRAPPER_IF_NEEDED(f) f()

#endif

void VM_Version::get_processor_features() {

  _cpu = 4; // 486 by default

  _model = 0;

  _stepping = 0;

  _cpuFeatures = 0;

  _logical_processors_per_package = 1;

  if (!Use486InstrsOnly) {

    // Get raw processor info

    // Some platforms (like Win*) need a wrapper around here

    // in order to properly handle SEGV for YMM registers test.

    CALL_TEST_FUNC_WITH_WRAPPER_IF_NEEDED(get_cpu_info_wrapper);

    assert_is_initialized();

    _cpu = extended_cpu_family();

    _model = extended_cpu_model();

    _stepping = cpu_stepping();

    if (cpu_family() > 4) { // it supports CPUID

      _cpuFeatures = feature_flags();

      // Logical processors are only available on P4s and above,

      // and only if hyperthreading is available.

      _logical_processors_per_package = logical_processor_count();

    }

  }

  _supports_cx8 = supports_cmpxchg8();

  // xchg and xadd instructions

  _supports_atomic_getset4 = true;

  _supports_atomic_getadd4 = true;

  LP64_ONLY(_supports_atomic_getset8 = true);

  LP64_ONLY(_supports_atomic_getadd8 = true);

#ifdef _LP64

  // OS should support SSE for x64 and hardware should support at least SSE2.

  if (!VM_Version::supports_sse2()) {

    vm_exit_during_initialization("Unknown x64 processor: SSE2 not supported");

  }

  // in 64 bit the use of SSE2 is the minimum

  if (UseSSE < 2) UseSSE = 2;

#endif

#ifdef AMD64

  // flush_icache_stub have to be generated first.

  // That is why Icache line size is hard coded in ICache class,

  // see icache_x86.hpp. It is also the reason why we can't use

  // clflush instruction in 32-bit VM since it could be running

  // on CPU which does not support it.

  //

  // The only thing we can do is to verify that flushed

  // ICache::line_size has correct value.

  guarantee(_cpuid_info.std_cpuid1_edx.bits.clflush != 0, "clflush is not supported");

  // clflush_size is size in quadwords (8 bytes).

  guarantee(_cpuid_info.std_cpuid1_ebx.bits.clflush_size == 8, "such clflush size is not supported");

#endif

  // If the OS doesn't support SSE, we can't use this feature even if the HW does

  if (!os::supports_sse())

    _cpuFeatures &= ~(CPU_SSE|CPU_SSE2|CPU_SSE3|CPU_SSSE3|CPU_SSE4A|CPU_SSE4_1|CPU_SSE4_2);

  if (UseSSE < 4) {

    _cpuFeatures &= ~CPU_SSE4_1;

    _cpuFeatures &= ~CPU_SSE4_2;

  }

  if (UseSSE < 3) {

    _cpuFeatures &= ~CPU_SSE3;

    _cpuFeatures &= ~CPU_SSSE3;

    _cpuFeatures &= ~CPU_SSE4A;

  }

  if (UseSSE < 2)

    _cpuFeatures &= ~CPU_SSE2;

  if (UseSSE < 1)

    _cpuFeatures &= ~CPU_SSE;

  if (UseAVX < 2)

    _cpuFeatures &= ~CPU_AVX2;

  if (UseAVX < 1)

    _cpuFeatures &= ~CPU_AVX;

  if (!UseAES && !FLAG_IS_DEFAULT(UseAES))

    _cpuFeatures &= ~CPU_AES;

  if (logical_processors_per_package() == 1) {

    // HT processor could be installed on a system which doesn't support HT.

    _cpuFeatures &= ~CPU_HT;

  }

  char buf[256];

  jio_snprintf(buf, sizeof(buf), "(%u cores per cpu, %u threads per core) family %d model %d stepping %d%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s",

               cores_per_cpu(), threads_per_core(),

               cpu_family(), _model, _stepping,

               (supports_cmov() ? ", cmov" : ""),

               (supports_cmpxchg8() ? ", cx8" : ""),

               (supports_fxsr() ? ", fxsr" : ""),

               (supports_mmx()  ? ", mmx"  : ""),

               (supports_sse()  ? ", sse"  : ""),

               (supports_sse2() ? ", sse2" : ""),

               (supports_sse3() ? ", sse3" : ""),

               (supports_ssse3()? ", ssse3": ""),

               (supports_sse4_1() ? ", sse4.1" : ""),

               (supports_sse4_2() ? ", sse4.2" : ""),