1 /*
   2  * Copyright (c) 1997, 2018, Oracle and/or its affiliates. All rights reserved.
   3  * Copyright (c) 2015, Red Hat Inc. All rights reserved.
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   6  * This code is free software; you can redistribute it and/or modify it
   7  * under the terms of the GNU General Public License version 2 only, as
   8  * published by the Free Software Foundation.
   9  *
  10  * This code is distributed in the hope that it will be useful, but WITHOUT
  11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
  13  * version 2 for more details (a copy is included in the LICENSE file that
  14  * accompanied this code).
  15  *
  16  * You should have received a copy of the GNU General Public License version
  17  * 2 along with this work; if not, write to the Free Software Foundation,
  18  * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
  19  *
  20  * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
  21  * or visit www.oracle.com if you need additional information or have any
  22  * questions.
  23  *
  24  */
  25 
  26 #include "precompiled.hpp"
  27 #include "asm/macroAssembler.hpp"
  28 #include "asm/macroAssembler.inline.hpp"
  29 #include "memory/resourceArea.hpp"
  30 #include "runtime/java.hpp"
  31 #include "runtime/stubCodeGenerator.hpp"
  32 #include "utilities/macros.hpp"
  33 #include "vm_version_aarch64.hpp"
  34 
  35 #include OS_HEADER_INLINE(os)
  36 
  37 #ifndef BUILTIN_SIM
  38 #include <sys/auxv.h>
  39 #include <asm/hwcap.h>
  40 #else
  41 #define getauxval(hwcap) 0
  42 #endif
  43 
  44 #ifndef HWCAP_AES
  45 #define HWCAP_AES   (1<<3)
  46 #endif
  47 
  48 #ifndef HWCAP_PMULL
  49 #define HWCAP_PMULL (1<<4)
  50 #endif
  51 
  52 #ifndef HWCAP_SHA1
  53 #define HWCAP_SHA1  (1<<5)
  54 #endif
  55 
  56 #ifndef HWCAP_SHA2
  57 #define HWCAP_SHA2  (1<<6)
  58 #endif
  59 
  60 #ifndef HWCAP_CRC32
  61 #define HWCAP_CRC32 (1<<7)
  62 #endif
  63 
  64 #ifndef HWCAP_ATOMICS
  65 #define HWCAP_ATOMICS (1<<8)
  66 #endif
  67 
  68 int VM_Version::_cpu;
  69 int VM_Version::_model;
  70 int VM_Version::_model2;
  71 int VM_Version::_variant;
  72 int VM_Version::_revision;
  73 int VM_Version::_stepping;
  74 VM_Version::PsrInfo VM_Version::_psr_info   = { 0, };
  75 
  76 static BufferBlob* stub_blob;
  77 static const int stub_size = 550;
  78 
  79 extern "C" {
  80   typedef void (*getPsrInfo_stub_t)(void*);
  81 }
  82 static getPsrInfo_stub_t getPsrInfo_stub = NULL;
  83 
  84 
  85 class VM_Version_StubGenerator: public StubCodeGenerator {
  86  public:
  87 
  88   VM_Version_StubGenerator(CodeBuffer *c) : StubCodeGenerator(c) {}
  89 
  90   address generate_getPsrInfo() {
  91     StubCodeMark mark(this, "VM_Version", "getPsrInfo_stub");
  92 #   define __ _masm->
  93     address start = __ pc();
  94 
  95 #ifdef BUILTIN_SIM
  96     __ c_stub_prolog(1, 0, MacroAssembler::ret_type_void);
  97 #endif
  98 
  99     // void getPsrInfo(VM_Version::PsrInfo* psr_info);
 100 
 101     address entry = __ pc();
 102 
 103     __ enter();
 104 
 105     __ get_dczid_el0(rscratch1);
 106     __ strw(rscratch1, Address(c_rarg0, in_bytes(VM_Version::dczid_el0_offset())));
 107 
 108     __ get_ctr_el0(rscratch1);
 109     __ strw(rscratch1, Address(c_rarg0, in_bytes(VM_Version::ctr_el0_offset())));
 110 
 111     __ leave();
 112     __ ret(lr);
 113 
 114 #   undef __
 115 
 116     return start;
 117   }
 118 };
 119 
 120 
 121 void VM_Version::get_processor_features() {
 122   _supports_cx8 = true;
 123   _supports_atomic_getset4 = true;
 124   _supports_atomic_getadd4 = true;
 125   _supports_atomic_getset8 = true;
 126   _supports_atomic_getadd8 = true;
 127 
 128   getPsrInfo_stub(&_psr_info);
 129 
 130   int dcache_line = VM_Version::dcache_line_size();
 131 
 132   if (FLAG_IS_DEFAULT(AllocatePrefetchDistance))
 133     FLAG_SET_DEFAULT(AllocatePrefetchDistance, 3*dcache_line);
 134   if (FLAG_IS_DEFAULT(AllocatePrefetchStepSize))
 135     FLAG_SET_DEFAULT(AllocatePrefetchStepSize, dcache_line);
 136   if (FLAG_IS_DEFAULT(PrefetchScanIntervalInBytes))
 137     FLAG_SET_DEFAULT(PrefetchScanIntervalInBytes, 3*dcache_line);
 138   if (FLAG_IS_DEFAULT(PrefetchCopyIntervalInBytes))
 139     FLAG_SET_DEFAULT(PrefetchCopyIntervalInBytes, 3*dcache_line);
 140   if (FLAG_IS_DEFAULT(SoftwarePrefetchHintDistance))
 141     FLAG_SET_DEFAULT(SoftwarePrefetchHintDistance, 3*dcache_line);
 142 
 143   if (PrefetchCopyIntervalInBytes != -1 &&
 144        ((PrefetchCopyIntervalInBytes & 7) || (PrefetchCopyIntervalInBytes >= 32768))) {
 145     warning("PrefetchCopyIntervalInBytes must be -1, or a multiple of 8 and < 32768");
 146     PrefetchCopyIntervalInBytes &= ~7;
 147     if (PrefetchCopyIntervalInBytes >= 32768)
 148       PrefetchCopyIntervalInBytes = 32760;
 149   }
 150 
 151   if (AllocatePrefetchDistance !=-1 && (AllocatePrefetchDistance & 7)) {
 152     warning("AllocatePrefetchDistance must be multiple of 8");
 153     AllocatePrefetchDistance &= ~7;
 154   }
 155 
 156   if (AllocatePrefetchStepSize & 7) {
 157     warning("AllocatePrefetchStepSize must be multiple of 8");
 158     AllocatePrefetchStepSize &= ~7;
 159   }
 160 
 161   if (SoftwarePrefetchHintDistance != -1 &&
 162        (SoftwarePrefetchHintDistance & 7)) {
 163     warning("SoftwarePrefetchHintDistance must be -1, or a multiple of 8");
 164     SoftwarePrefetchHintDistance &= ~7;
 165   }
 166 
 167   unsigned long auxv = getauxval(AT_HWCAP);
 168 
 169   char buf[512];
 170 
 171   _features = auxv;
 172 
 173   int cpu_lines = 0;
 174   if (FILE *f = fopen("/proc/cpuinfo", "r")) {
 175     char buf[128], *p;
 176     while (fgets(buf, sizeof (buf), f) != NULL) {
 177       if (p = strchr(buf, ':')) {
 178         long v = strtol(p+1, NULL, 0);
 179         if (strncmp(buf, "CPU implementer", sizeof "CPU implementer" - 1) == 0) {
 180           _cpu = v;
 181           cpu_lines++;
 182         } else if (strncmp(buf, "CPU variant", sizeof "CPU variant" - 1) == 0) {
 183           _variant = v;
 184         } else if (strncmp(buf, "CPU part", sizeof "CPU part" - 1) == 0) {
 185           if (_model != v)  _model2 = _model;
 186           _model = v;
 187         } else if (strncmp(buf, "CPU revision", sizeof "CPU revision" - 1) == 0) {
 188           _revision = v;
 189         }
 190       }
 191     }
 192     fclose(f);
 193   }
 194 
 195   // Enable vendor specific features
 196 
 197   // ThunderX
 198   if (_cpu == CPU_CAVIUM && (_model == 0xA1)) {
 199     if (_variant == 0) _features |= CPU_DMB_ATOMICS;
 200     if (FLAG_IS_DEFAULT(AvoidUnalignedAccesses)) {
 201       FLAG_SET_DEFAULT(AvoidUnalignedAccesses, true);
 202     }
 203     if (FLAG_IS_DEFAULT(UseSIMDForMemoryOps)) {
 204       FLAG_SET_DEFAULT(UseSIMDForMemoryOps, (_variant > 0));
 205     }
 206     if (FLAG_IS_DEFAULT(UseSIMDForArrayEquals)) {
 207       FLAG_SET_DEFAULT(UseSIMDForArrayEquals, false);
 208     }
 209   }
 210 
 211   // ThunderX2
 212   if ((_cpu == CPU_CAVIUM && (_model == 0xAF)) ||
 213       (_cpu == CPU_BROADCOM && (_model == 0x516))) {
 214     if (FLAG_IS_DEFAULT(AvoidUnalignedAccesses)) {
 215       FLAG_SET_DEFAULT(AvoidUnalignedAccesses, true);
 216     }
 217     if (FLAG_IS_DEFAULT(UseSIMDForMemoryOps)) {
 218       FLAG_SET_DEFAULT(UseSIMDForMemoryOps, true);
 219     }
 220   }
 221 
 222   // Cortex A53
 223   if (_cpu == CPU_ARM && (_model == 0xd03 || _model2 == 0xd03)) {
 224     _features |= CPU_A53MAC;
 225     if (FLAG_IS_DEFAULT(UseSIMDForArrayEquals)) {
 226       FLAG_SET_DEFAULT(UseSIMDForArrayEquals, false);
 227     }
 228   }
 229 
 230   // Cortex A73
 231   if (_cpu == CPU_ARM && (_model == 0xd09 || _model2 == 0xd09)) {
 232     if (FLAG_IS_DEFAULT(SoftwarePrefetchHintDistance)) {
 233       FLAG_SET_DEFAULT(SoftwarePrefetchHintDistance, -1);
 234     }
 235     // A73 is faster with short-and-easy-for-speculative-execution-loop
 236     if (FLAG_IS_DEFAULT(UseSimpleArrayEquals)) {
 237       FLAG_SET_DEFAULT(UseSimpleArrayEquals, true);
 238     }
 239   }
 240 
 241   if (_cpu == CPU_ARM && (_model == 0xd07 || _model2 == 0xd07)) _features |= CPU_STXR_PREFETCH;
 242   // If an olde style /proc/cpuinfo (cpu_lines == 1) then if _model is an A57 (0xd07)
 243   // we assume the worst and assume we could be on a big little system and have
 244   // undisclosed A53 cores which we could be swapped to at any stage
 245   if (_cpu == CPU_ARM && cpu_lines == 1 && _model == 0xd07) _features |= CPU_A53MAC;
 246 
 247   sprintf(buf, "0x%02x:0x%x:0x%03x:%d", _cpu, _variant, _model, _revision);
 248   if (_model2) sprintf(buf+strlen(buf), "(0x%03x)", _model2);
 249   if (auxv & HWCAP_ASIMD) strcat(buf, ", simd");
 250   if (auxv & HWCAP_CRC32) strcat(buf, ", crc");
 251   if (auxv & HWCAP_AES)   strcat(buf, ", aes");
 252   if (auxv & HWCAP_SHA1)  strcat(buf, ", sha1");
 253   if (auxv & HWCAP_SHA2)  strcat(buf, ", sha256");
 254   if (auxv & HWCAP_ATOMICS) strcat(buf, ", lse");
 255 
 256   _features_string = os::strdup(buf);
 257 
 258   if (FLAG_IS_DEFAULT(UseCRC32)) {
 259     UseCRC32 = (auxv & HWCAP_CRC32) != 0;
 260   }
 261   if (UseCRC32 && (auxv & HWCAP_CRC32) == 0) {
 262     warning("UseCRC32 specified, but not supported on this CPU");
 263   }
 264 
 265   if (FLAG_IS_DEFAULT(UseAdler32Intrinsics)) {
 266     FLAG_SET_DEFAULT(UseAdler32Intrinsics, true);
 267   }
 268 
 269   if (UseVectorizedMismatchIntrinsic) {
 270     warning("UseVectorizedMismatchIntrinsic specified, but not available on this CPU.");
 271     FLAG_SET_DEFAULT(UseVectorizedMismatchIntrinsic, false);
 272   }
 273 
 274   if (auxv & HWCAP_ATOMICS) {
 275     if (FLAG_IS_DEFAULT(UseLSE))
 276       FLAG_SET_DEFAULT(UseLSE, true);
 277   } else {
 278     if (UseLSE) {
 279       warning("UseLSE specified, but not supported on this CPU");
 280     }
 281   }
 282 
 283   if (auxv & HWCAP_AES) {
 284     UseAES = UseAES || FLAG_IS_DEFAULT(UseAES);
 285     UseAESIntrinsics =
 286         UseAESIntrinsics || (UseAES && FLAG_IS_DEFAULT(UseAESIntrinsics));
 287     if (UseAESIntrinsics && !UseAES) {
 288       warning("UseAESIntrinsics enabled, but UseAES not, enabling");
 289       UseAES = true;
 290     }
 291   } else {
 292     if (UseAES) {
 293       warning("UseAES specified, but not supported on this CPU");
 294     }
 295     if (UseAESIntrinsics) {
 296       warning("UseAESIntrinsics specified, but not supported on this CPU");
 297     }
 298   }
 299 
 300   if (UseAESCTRIntrinsics) {
 301     warning("AES/CTR intrinsics are not available on this CPU");
 302     FLAG_SET_DEFAULT(UseAESCTRIntrinsics, false);
 303   }
 304 
 305   if (FLAG_IS_DEFAULT(UseCRC32Intrinsics)) {
 306     UseCRC32Intrinsics = true;
 307   }
 308 
 309   if (auxv & HWCAP_CRC32) {
 310     if (FLAG_IS_DEFAULT(UseCRC32CIntrinsics)) {
 311       FLAG_SET_DEFAULT(UseCRC32CIntrinsics, true);
 312     }
 313   } else if (UseCRC32CIntrinsics) {
 314     warning("CRC32C is not available on the CPU");
 315     FLAG_SET_DEFAULT(UseCRC32CIntrinsics, false);
 316   }
 317 
 318   if (FLAG_IS_DEFAULT(UseFMA)) {
 319     FLAG_SET_DEFAULT(UseFMA, true);
 320   }
 321 
 322   if (auxv & (HWCAP_SHA1 | HWCAP_SHA2)) {
 323     if (FLAG_IS_DEFAULT(UseSHA)) {
 324       FLAG_SET_DEFAULT(UseSHA, true);
 325     }
 326   } else if (UseSHA) {
 327     warning("SHA instructions are not available on this CPU");
 328     FLAG_SET_DEFAULT(UseSHA, false);
 329   }
 330 
 331   if (UseSHA && (auxv & HWCAP_SHA1)) {
 332     if (FLAG_IS_DEFAULT(UseSHA1Intrinsics)) {
 333       FLAG_SET_DEFAULT(UseSHA1Intrinsics, true);
 334     }
 335   } else if (UseSHA1Intrinsics) {
 336     warning("Intrinsics for SHA-1 crypto hash functions not available on this CPU.");
 337     FLAG_SET_DEFAULT(UseSHA1Intrinsics, false);
 338   }
 339 
 340   if (UseSHA && (auxv & HWCAP_SHA2)) {
 341     if (FLAG_IS_DEFAULT(UseSHA256Intrinsics)) {
 342       FLAG_SET_DEFAULT(UseSHA256Intrinsics, true);
 343     }
 344   } else if (UseSHA256Intrinsics) {
 345     warning("Intrinsics for SHA-224 and SHA-256 crypto hash functions not available on this CPU.");
 346     FLAG_SET_DEFAULT(UseSHA256Intrinsics, false);
 347   }
 348 
 349   if (UseSHA512Intrinsics) {
 350     warning("Intrinsics for SHA-384 and SHA-512 crypto hash functions not available on this CPU.");
 351     FLAG_SET_DEFAULT(UseSHA512Intrinsics, false);
 352   }
 353 
 354   if (!(UseSHA1Intrinsics || UseSHA256Intrinsics || UseSHA512Intrinsics)) {
 355     FLAG_SET_DEFAULT(UseSHA, false);
 356   }
 357 
 358   if (auxv & HWCAP_PMULL) {
 359     if (FLAG_IS_DEFAULT(UseGHASHIntrinsics)) {
 360       FLAG_SET_DEFAULT(UseGHASHIntrinsics, true);
 361     }
 362   } else if (UseGHASHIntrinsics) {
 363     warning("GHASH intrinsics are not available on this CPU");
 364     FLAG_SET_DEFAULT(UseGHASHIntrinsics, false);
 365   }
 366 
 367   if (is_zva_enabled()) {
 368     if (FLAG_IS_DEFAULT(UseBlockZeroing)) {
 369       FLAG_SET_DEFAULT(UseBlockZeroing, true);
 370     }
 371     if (FLAG_IS_DEFAULT(BlockZeroingLowLimit)) {
 372       FLAG_SET_DEFAULT(BlockZeroingLowLimit, 4 * VM_Version::zva_length());
 373     }
 374   } else if (UseBlockZeroing) {
 375     warning("DC ZVA is not available on this CPU");
 376     FLAG_SET_DEFAULT(UseBlockZeroing, false);
 377   }
 378 
 379   // This machine allows unaligned memory accesses
 380   if (FLAG_IS_DEFAULT(UseUnalignedAccesses)) {
 381     FLAG_SET_DEFAULT(UseUnalignedAccesses, true);
 382   }
 383 
 384   if (FLAG_IS_DEFAULT(UseBarriersForVolatile)) {
 385     UseBarriersForVolatile = (_features & CPU_DMB_ATOMICS) != 0;
 386   }
 387 
 388   if (FLAG_IS_DEFAULT(UsePopCountInstruction)) {
 389     UsePopCountInstruction = true;
 390   }
 391 
 392 #ifdef COMPILER2
 393   if (FLAG_IS_DEFAULT(UseMultiplyToLenIntrinsic)) {
 394     UseMultiplyToLenIntrinsic = true;
 395   }
 396 
 397   if (FLAG_IS_DEFAULT(UseSquareToLenIntrinsic)) {
 398     UseSquareToLenIntrinsic = true;
 399   }
 400 
 401   if (FLAG_IS_DEFAULT(UseMulAddIntrinsic)) {
 402     UseMulAddIntrinsic = true;
 403   }
 404 
 405   if (FLAG_IS_DEFAULT(UseMontgomeryMultiplyIntrinsic)) {
 406     UseMontgomeryMultiplyIntrinsic = true;
 407   }
 408   if (FLAG_IS_DEFAULT(UseMontgomerySquareIntrinsic)) {
 409     UseMontgomerySquareIntrinsic = true;
 410   }
 411 
 412   if (FLAG_IS_DEFAULT(OptoScheduling)) {
 413     OptoScheduling = true;
 414   }
 415 #endif
 416 }
 417 
 418 void VM_Version::initialize() {
 419   ResourceMark rm;
 420 
 421   stub_blob = BufferBlob::create("getPsrInfo_stub", stub_size);
 422   if (stub_blob == NULL) {
 423     vm_exit_during_initialization("Unable to allocate getPsrInfo_stub");
 424   }
 425 
 426   CodeBuffer c(stub_blob);
 427   VM_Version_StubGenerator g(&c);
 428   getPsrInfo_stub = CAST_TO_FN_PTR(getPsrInfo_stub_t,
 429                                    g.generate_getPsrInfo());
 430 
 431   get_processor_features();
 432 
 433   UNSUPPORTED_OPTION(CriticalJNINatives);
 434 }