]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/i386/cpuid.c
xnu-7195.101.1.tar.gz
[apple/xnu.git] / osfmk / i386 / cpuid.c
index 29d7ef51eb5515f93b559d7d26e2caa8400ad8cd..90844be06efc9b144cabc0cb7c84cea8ea70f53a 100644 (file)
@@ -1,8 +1,8 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2020 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
- * 
+ *
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
  * unlawful or unlicensed copies of an Apple operating system, or to
  * circumvent, violate, or enable the circumvention or violation of, any
  * terms of an Apple operating system software license agreement.
- * 
+ *
  * Please obtain a copy of the License at
  * http://www.opensource.apple.com/apsl/ and read it before using this file.
- * 
+ *
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  * Please see the License for the specific language governing rights and
  * limitations under the License.
- * 
+ *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * @OSF_COPYRIGHT@
  */
-#include <platforms.h>
-#include <mach_kdb.h>
+#include <vm/vm_page.h>
 #include <pexpert/pexpert.h>
 
-#include "cpuid.h"
-#if MACH_KDB
-#include <i386/db_machdep.h>
-#include <ddb/db_aout.h>
-#include <ddb/db_access.h>
-#include <ddb/db_sym.h>
-#include <ddb/db_variables.h>
-#include <ddb/db_command.h>
-#include <ddb/db_output.h>
-#include <ddb/db_expr.h>
+#include <i386/cpu_threads.h>
+#include <i386/cpuid.h>
+#include <i386/machine_routines.h>
+
+int force_tecs_at_idle;
+int tecs_mode_supported;
+
+static  boolean_t       cpuid_dbg
+#if DEBUG
+        = TRUE;
+#else
+        = FALSE;
 #endif
+#define DBG(x...)                       \
+       do {                            \
+               if (cpuid_dbg)          \
+                       kprintf(x);     \
+       } while (0)                     \
 
-#define min(a,b) ((a) < (b) ? (a) : (b))
-#define quad(hi,lo)    (((uint64_t)(hi)) << 32 | (lo))
+#define min(a, b) ((a) < (b) ? (a) : (b))
+#define quad(hi, lo)     (((uint64_t)(hi)) << 32 | (lo))
 
-#define bit(n)         (1UL << (n))
-#define bitmask(h,l)   ((bit(h)|(bit(h)-1)) & ~(bit(l)-1))
-#define bitfield(x,h,l)        (((x) & bitmask(h,l)) >> l)
+/*
+ * Leaf 2 cache descriptor encodings.
+ */
+typedef enum {
+       _NULL_,         /* NULL (empty) descriptor */
+       CACHE,          /* Cache */
+       TLB,            /* TLB */
+       STLB,           /* Shared second-level unified TLB */
+       PREFETCH        /* Prefetch size */
+} cpuid_leaf2_desc_type_t;
+
+typedef enum {
+       NA,             /* Not Applicable */
+       FULLY,          /* Fully-associative */
+       TRACE,          /* Trace Cache (P4 only) */
+       INST,           /* Instruction TLB */
+       DATA,           /* Data TLB */
+       DATA0,          /* Data TLB, 1st level */
+       DATA1,          /* Data TLB, 2nd level */
+       L1,             /* L1 (unified) cache */
+       L1_INST,        /* L1 Instruction cache */
+       L1_DATA,        /* L1 Data cache */
+       L2,             /* L2 (unified) cache */
+       L3,             /* L3 (unified) cache */
+       L2_2LINESECTOR, /* L2 (unified) cache with 2 lines per sector */
+       L3_2LINESECTOR, /* L3(unified) cache with 2 lines per sector */
+       SMALL,          /* Small page TLB */
+       LARGE,          /* Large page TLB */
+       BOTH            /* Small and Large page TLB */
+} cpuid_leaf2_qualifier_t;
+
+typedef struct cpuid_cache_descriptor {
+       uint8_t         value;          /* descriptor code */
+       uint8_t         type;           /* cpuid_leaf2_desc_type_t */
+       uint8_t         level;          /* level of cache/TLB hierachy */
+       uint8_t         ways;           /* wayness of cache */
+       uint16_t        size;           /* cachesize or TLB pagesize */
+       uint16_t        entries;        /* number of TLB entries or linesize */
+} cpuid_cache_descriptor_t;
 
 /*
- * CPU identification routines.
- *
- * Note that this code assumes a processor that supports the
- * 'cpuid' instruction.
+ * These multipliers are used to encode 1*K .. 64*M in a 16 bit size field
+ */
+#define K       (1)
+#define M       (1024)
+
+/*
+ * Intel cache descriptor table:
  */
+static cpuid_cache_descriptor_t intel_cpuid_leaf2_descriptor_table[] = {
+//     -------------------------------------------------------
+//     value   type    level           ways    size    entries
+//     -------------------------------------------------------
+       { 0x00, _NULL_, NA, NA, NA, NA  },
+       { 0x01, TLB, INST, 4, SMALL, 32  },
+       { 0x02, TLB, INST, FULLY, LARGE, 2   },
+       { 0x03, TLB, DATA, 4, SMALL, 64  },
+       { 0x04, TLB, DATA, 4, LARGE, 8   },
+       { 0x05, TLB, DATA1, 4, LARGE, 32  },
+       { 0x06, CACHE, L1_INST, 4, 8 * K, 32  },
+       { 0x08, CACHE, L1_INST, 4, 16 * K, 32  },
+       { 0x09, CACHE, L1_INST, 4, 32 * K, 64  },
+       { 0x0A, CACHE, L1_DATA, 2, 8 * K, 32  },
+       { 0x0B, TLB, INST, 4, LARGE, 4   },
+       { 0x0C, CACHE, L1_DATA, 4, 16 * K, 32  },
+       { 0x0D, CACHE, L1_DATA, 4, 16 * K, 64  },
+       { 0x0E, CACHE, L1_DATA, 6, 24 * K, 64  },
+       { 0x21, CACHE, L2, 8, 256 * K, 64  },
+       { 0x22, CACHE, L3_2LINESECTOR, 4, 512 * K, 64  },
+       { 0x23, CACHE, L3_2LINESECTOR, 8, 1 * M, 64  },
+       { 0x25, CACHE, L3_2LINESECTOR, 8, 2 * M, 64  },
+       { 0x29, CACHE, L3_2LINESECTOR, 8, 4 * M, 64  },
+       { 0x2C, CACHE, L1_DATA, 8, 32 * K, 64  },
+       { 0x30, CACHE, L1_INST, 8, 32 * K, 64  },
+       { 0x40, CACHE, L2, NA, 0, NA  },
+       { 0x41, CACHE, L2, 4, 128 * K, 32  },
+       { 0x42, CACHE, L2, 4, 256 * K, 32  },
+       { 0x43, CACHE, L2, 4, 512 * K, 32  },
+       { 0x44, CACHE, L2, 4, 1 * M, 32  },
+       { 0x45, CACHE, L2, 4, 2 * M, 32  },
+       { 0x46, CACHE, L3, 4, 4 * M, 64  },
+       { 0x47, CACHE, L3, 8, 8 * M, 64  },
+       { 0x48, CACHE, L2, 12, 3 * M, 64  },
+       { 0x49, CACHE, L2, 16, 4 * M, 64  },
+       { 0x4A, CACHE, L3, 12, 6 * M, 64  },
+       { 0x4B, CACHE, L3, 16, 8 * M, 64  },
+       { 0x4C, CACHE, L3, 12, 12 * M, 64  },
+       { 0x4D, CACHE, L3, 16, 16 * M, 64  },
+       { 0x4E, CACHE, L2, 24, 6 * M, 64  },
+       { 0x4F, TLB, INST, NA, SMALL, 32  },
+       { 0x50, TLB, INST, NA, BOTH, 64  },
+       { 0x51, TLB, INST, NA, BOTH, 128 },
+       { 0x52, TLB, INST, NA, BOTH, 256 },
+       { 0x55, TLB, INST, FULLY, BOTH, 7   },
+       { 0x56, TLB, DATA0, 4, LARGE, 16  },
+       { 0x57, TLB, DATA0, 4, SMALL, 16  },
+       { 0x59, TLB, DATA0, FULLY, SMALL, 16  },
+       { 0x5A, TLB, DATA0, 4, LARGE, 32  },
+       { 0x5B, TLB, DATA, NA, BOTH, 64  },
+       { 0x5C, TLB, DATA, NA, BOTH, 128 },
+       { 0x5D, TLB, DATA, NA, BOTH, 256 },
+       { 0x60, CACHE, L1, 16 * K, 8, 64  },
+       { 0x61, CACHE, L1, 4, 8 * K, 64  },
+       { 0x62, CACHE, L1, 4, 16 * K, 64  },
+       { 0x63, CACHE, L1, 4, 32 * K, 64  },
+       { 0x70, CACHE, TRACE, 8, 12 * K, NA  },
+       { 0x71, CACHE, TRACE, 8, 16 * K, NA  },
+       { 0x72, CACHE, TRACE, 8, 32 * K, NA  },
+       { 0x76, TLB, INST, NA, BOTH, 8   },
+       { 0x78, CACHE, L2, 4, 1 * M, 64  },
+       { 0x79, CACHE, L2_2LINESECTOR, 8, 128 * K, 64  },
+       { 0x7A, CACHE, L2_2LINESECTOR, 8, 256 * K, 64  },
+       { 0x7B, CACHE, L2_2LINESECTOR, 8, 512 * K, 64  },
+       { 0x7C, CACHE, L2_2LINESECTOR, 8, 1 * M, 64  },
+       { 0x7D, CACHE, L2, 8, 2 * M, 64  },
+       { 0x7F, CACHE, L2, 2, 512 * K, 64  },
+       { 0x80, CACHE, L2, 8, 512 * K, 64  },
+       { 0x82, CACHE, L2, 8, 256 * K, 32  },
+       { 0x83, CACHE, L2, 8, 512 * K, 32  },
+       { 0x84, CACHE, L2, 8, 1 * M, 32  },
+       { 0x85, CACHE, L2, 8, 2 * M, 32  },
+       { 0x86, CACHE, L2, 4, 512 * K, 64  },
+       { 0x87, CACHE, L2, 8, 1 * M, 64  },
+       { 0xB0, TLB, INST, 4, SMALL, 128 },
+       { 0xB1, TLB, INST, 4, LARGE, 8   },
+       { 0xB2, TLB, INST, 4, SMALL, 64  },
+       { 0xB3, TLB, DATA, 4, SMALL, 128 },
+       { 0xB4, TLB, DATA1, 4, SMALL, 256 },
+       { 0xB5, TLB, DATA1, 8, SMALL, 64  },
+       { 0xB6, TLB, DATA1, 8, SMALL, 128 },
+       { 0xBA, TLB, DATA1, 4, BOTH, 64  },
+       { 0xC1, STLB, DATA1, 8, SMALL, 1024},
+       { 0xCA, STLB, DATA1, 4, SMALL, 512 },
+       { 0xD0, CACHE, L3, 4, 512 * K, 64  },
+       { 0xD1, CACHE, L3, 4, 1 * M, 64  },
+       { 0xD2, CACHE, L3, 4, 2 * M, 64  },
+       { 0xD3, CACHE, L3, 4, 4 * M, 64  },
+       { 0xD4, CACHE, L3, 4, 8 * M, 64  },
+       { 0xD6, CACHE, L3, 8, 1 * M, 64  },
+       { 0xD7, CACHE, L3, 8, 2 * M, 64  },
+       { 0xD8, CACHE, L3, 8, 4 * M, 64  },
+       { 0xD9, CACHE, L3, 8, 8 * M, 64  },
+       { 0xDA, CACHE, L3, 8, 12 * M, 64  },
+       { 0xDC, CACHE, L3, 12, 1536 * K, 64  },
+       { 0xDD, CACHE, L3, 12, 3 * M, 64  },
+       { 0xDE, CACHE, L3, 12, 6 * M, 64  },
+       { 0xDF, CACHE, L3, 12, 12 * M, 64  },
+       { 0xE0, CACHE, L3, 12, 18 * M, 64  },
+       { 0xE2, CACHE, L3, 16, 2 * M, 64  },
+       { 0xE3, CACHE, L3, 16, 4 * M, 64  },
+       { 0xE4, CACHE, L3, 16, 8 * M, 64  },
+       { 0xE5, CACHE, L3, 16, 16 * M, 64  },
+       { 0xE6, CACHE, L3, 16, 24 * M, 64  },
+       { 0xF0, PREFETCH, NA, NA, 64, NA  },
+       { 0xF1, PREFETCH, NA, NA, 128, NA  },
+       { 0xFF, CACHE, NA, NA, 0, NA  }
+};
+#define INTEL_LEAF2_DESC_NUM (sizeof(intel_cpuid_leaf2_descriptor_table) / \
+                               sizeof(cpuid_cache_descriptor_t))
+
+boolean_t cpuid_tsx_disabled = false;   /* true if XNU disabled TSX */
+boolean_t cpuid_tsx_supported = false;
 
-static unsigned int    cpuid_maxcpuid;
+static void do_cwas(i386_cpu_info_t *cpuinfo, boolean_t on_slave);
+static void cpuid_do_precpuid_was(void);
 
-static i386_cpu_info_t cpuid_cpu_info;
+#if DEBUG || DEVELOPMENT
+static void cpuid_vmm_detect_pv_interface(i386_vmm_info_t *info_p, const char *signature,
+    bool (*)(i386_vmm_info_t*, const uint32_t, const uint32_t));
+static bool cpuid_vmm_detect_applepv_features(i386_vmm_info_t *info_p, const uint32_t base, const uint32_t max_leaf);
+#endif /* DEBUG || DEVELOPMENT */
 
-uint32_t               cpuid_feature;          /* XXX obsolescent for compat */
+static inline cpuid_cache_descriptor_t *
+cpuid_leaf2_find(uint8_t value)
+{
+       unsigned int    i;
+
+       for (i = 0; i < INTEL_LEAF2_DESC_NUM; i++) {
+               if (intel_cpuid_leaf2_descriptor_table[i].value == value) {
+                       return &intel_cpuid_leaf2_descriptor_table[i];
+               }
+       }
+       return NULL;
+}
 
 /*
- * We only identify Intel CPUs here.  Adding support
- * for others would be straightforward.
+ * CPU identification routines.
  */
-static void    set_cpu_generic(i386_cpu_info_t *);
-static void    set_cpu_intel(i386_cpu_info_t *);
-static void    set_cpu_amd(i386_cpu_info_t *);
-static void    set_cpu_nsc(i386_cpu_info_t *);
-static void    set_cpu_unknown(i386_cpu_info_t *);
-
-struct {
-       const char      *vendor;
-       void            (* func)(i386_cpu_info_t *);
-} cpu_vendors[] = {
-       {CPUID_VID_INTEL,       set_cpu_intel},
-       {CPUID_VID_AMD,         set_cpu_amd},
-       {CPUID_VID_NSC,         set_cpu_nsc},
-       {0,                     set_cpu_unknown}
+
+static i386_cpu_info_t  cpuid_cpu_info;
+static i386_cpu_info_t  *cpuid_cpu_infop = NULL;
+
+static void
+cpuid_fn(uint32_t selector, uint32_t *result)
+{
+       do_cpuid(selector, result);
+       DBG("cpuid_fn(0x%08x) eax:0x%08x ebx:0x%08x ecx:0x%08x edx:0x%08x\n",
+           selector, result[0], result[1], result[2], result[3]);
+}
+
+static const char *cache_type_str[LCACHE_MAX] = {
+       "Lnone", "L1I", "L1D", "L2U", "L3U"
 };
 
-void
-cpuid_get_info(i386_cpu_info_t *info_p)
+static void
+do_cwas(i386_cpu_info_t *cpuinfo, boolean_t on_slave)
 {
-       uint32_t        cpuid_result[4];
-       int             i;
+       extern int force_thread_policy_tecs;
+       cwa_classifier_e wa_reqd;
 
-       bzero((void *)info_p, sizeof(i386_cpu_info_t));
+       /*
+        * Workaround for reclaiming perf counter 3 due to TSX memory ordering erratum.
+        * This workaround does not support being forcibly set (since an MSR must be
+        * enumerated, lest we #GP when forced to access it.)
+        *
+        * Note that if disabling TSX is supported, disablement is prefered over forcing
+        * TSX transactions to abort.
+        */
+       if (cpuid_wa_required(CPU_INTEL_TSXDA) == CWA_ON) {
+               /* This must be executed on all logical processors */
+               wrmsr64(MSR_IA32_TSX_CTRL, MSR_IA32_TSXCTRL_TSX_CPU_CLEAR | MSR_IA32_TSXCTRL_RTM_DISABLE);
+       } else if (cpuid_wa_required(CPU_INTEL_TSXFA) == CWA_ON) {
+               /* This must be executed on all logical processors */
+               wrmsr64(MSR_IA32_TSX_FORCE_ABORT,
+                   rdmsr64(MSR_IA32_TSX_FORCE_ABORT) | MSR_IA32_TSXFA_RTM_FORCE_ABORT);
+       }
 
-       /* do cpuid 0 to get vendor */
-       do_cpuid(0, cpuid_result);
-       cpuid_maxcpuid = cpuid_result[eax];
-       bcopy((char *)&cpuid_result[ebx], &info_p->cpuid_vendor[0], 4); /* ug */
-       bcopy((char *)&cpuid_result[ecx], &info_p->cpuid_vendor[8], 4);
-       bcopy((char *)&cpuid_result[edx], &info_p->cpuid_vendor[4], 4);
-       info_p->cpuid_vendor[12] = 0;
+       if (((wa_reqd = cpuid_wa_required(CPU_INTEL_SRBDS)) & CWA_ON) != 0 &&
+           ((wa_reqd & CWA_FORCE_ON) == CWA_ON ||
+           (cpuinfo->cpuid_leaf7_extfeatures & CPUID_LEAF7_EXTFEATURE_SRBDS_CTRL) != 0)) {
+               /* This must be executed on all logical processors */
+               uint64_t mcuoptctrl = rdmsr64(MSR_IA32_MCU_OPT_CTRL);
+               mcuoptctrl |= MSR_IA32_MCUOPTCTRL_RNGDS_MITG_DIS;
+               wrmsr64(MSR_IA32_MCU_OPT_CTRL, mcuoptctrl);
+       }
 
-       /* look up vendor */
-       for (i = 0; ; i++) {
-               if ((cpu_vendors[i].vendor == 0) ||
-                   (!strcmp(cpu_vendors[i].vendor, info_p->cpuid_vendor))) {
-                       cpu_vendors[i].func(info_p);
-                       break;
+       if (on_slave) {
+               return;
+       }
+
+       switch (cpuid_wa_required(CPU_INTEL_SEGCHK)) {
+       case CWA_FORCE_ON:
+               force_thread_policy_tecs = 1;
+
+               /* If hyperthreaded, enable idle workaround */
+               if (cpuinfo->thread_count > cpuinfo->core_count) {
+                       force_tecs_at_idle = 1;
                }
+
+               OS_FALLTHROUGH;
+       case CWA_ON:
+               tecs_mode_supported = 1;
+               break;
+
+       case CWA_FORCE_OFF:
+       case CWA_OFF:
+               tecs_mode_supported = 0;
+               force_tecs_at_idle = 0;
+               force_thread_policy_tecs = 0;
+               break;
+
+       default:
+               break;
        }
 }
 
-/*
- * Cache descriptor table. Each row has the form:
- *        (descriptor_value,           cache,  size,           linesize,
- *                             description)
- * Note: the CACHE_DESC macro does not expand description text in the kernel.
- */
-static cpuid_cache_desc_t cpuid_cache_desc_tab[] = {
-CACHE_DESC(CPUID_CACHE_ITLB_4K,        Lnone,  0,              0, \
-       "Instruction TLB, 4K, pages 4-way set associative, 64 entries"),
-CACHE_DESC(CPUID_CACHE_ITLB_4M,        Lnone,  0,              0, \
-       "Instruction TLB, 4M, pages 4-way set associative, 2 entries"),
-CACHE_DESC(CPUID_CACHE_DTLB_4K,        Lnone,  0,              0, \
-       "Data TLB, 4K pages, 4-way set associative, 64 entries"),
-CACHE_DESC(CPUID_CACHE_DTLB_4M,        Lnone,  0,              0, \
-       "Data TLB, 4M pages, 4-way set associative, 8 entries"),
-CACHE_DESC(CPUID_CACHE_ITLB_64,        Lnone,  0,              0, \
-       "Instruction TLB, 4K and 2M or 4M pages, 64 entries"),
-CACHE_DESC(CPUID_CACHE_ITLB_128,       Lnone,  0,              0, \
-       "Instruction TLB, 4K and 2M or 4M pages, 128 entries"),
-CACHE_DESC(CPUID_CACHE_ITLB_256,       Lnone,  0,              0, \
-       "Instruction TLB, 4K and 2M or 4M pages, 256 entries"),
-CACHE_DESC(CPUID_CACHE_DTLB_64,                Lnone,  0,              0, \
-       "Data TLB, 4K and 4M pages, 64 entries"),
-CACHE_DESC(CPUID_CACHE_DTLB_128,       Lnone,  0,              0, \
-       "Data TLB, 4K and 4M pages, 128 entries"),
-CACHE_DESC(CPUID_CACHE_DTLB_256,       Lnone,  0,              0, \
-       "Data TLB, 4K and 4M pages, 256 entries"),
-CACHE_DESC(CPUID_CACHE_ITLB_128_4,      Lnone,  0,              0, \
-        "Instruction TLB, 4K pages, 4-way set associative, 128 entries"),
-CACHE_DESC(CPUID_CACHE_DTLB_128_4,      Lnone,  0,              0, \
-        "Data TLB, 4K pages, 4-way set associative, 128 entries"),
-CACHE_DESC(CPUID_CACHE_ICACHE_8K,      L1I,    8*1024,         32, \
-       "Instruction L1 cache, 8K, 4-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_DCACHE_8K,      L1D,    8*1024,         32, \
-       "Data L1 cache, 8K, 2-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_ICACHE_16K,     L1I,    16*1024,         32, \
-       "Instruction L1 cache, 16K, 4-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_DCACHE_16K,     L1D,    16*1024,        32, \
-       "Data L1 cache, 16K, 4-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_DCACHE_8K_64,   L1D,    8*1024,         64, \
-       "Data L1 cache, 8K, 4-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_DCACHE_16K_64,  L1D,    16*1024,        64, \
-       "Data L1 cache, 16K, 4-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_DCACHE_32K_64,  L1D,    32*1024,        64, \
-       "Data L1 cache, 32K, 4-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_DCACHE_32K,      L1D,    32*1024,        64, \
-        "Data L1 cache, 32K, 8-way set assocative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_ICACHE_32K,      L1I,    32*1024,        64, \
-        "Instruction L1 cache, 32K, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_DCACHE_16K_8,    L1D,    16*1024,        64, \
-        "Data L1 cache, 16K, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_TRACE_12K,      L1I,    12*1024,        64, \
-       "Trace cache, 12K-uop, 8-way set associative"),
-CACHE_DESC(CPUID_CACHE_TRACE_16K,      L1I,    16*1024,        64, \
-       "Trace cache, 16K-uop, 8-way set associative"),
-CACHE_DESC(CPUID_CACHE_TRACE_32K,      L1I,    32*1024,        64, \
-       "Trace cache, 32K-uop, 8-way set associative"),
-CACHE_DESC(CPUID_CACHE_UCACHE_128K,    L2U,    128*1024,       32, \
-       "Unified L2 cache, 128K, 4-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_256K,    L2U,    128*1024,       32, \
-       "Unified L2 cache, 256K, 4-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_512K,    L2U,    512*1024,       32, \
-       "Unified L2 cache, 512K, 4-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_1M,      L2U,    1*1024*1024,    32, \
-       "Unified L2 cache, 1M, 4-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_2M,      L2U,    2*1024*1024,    32, \
-       "Unified L2 cache, 2M, 4-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_4M,      L2U,    4*1024*1024,    64, \
-       "Unified L2 cache, 4M, 16-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_128K_64, L2U,    128*1024,       64, \
-       "Unified L2 cache, 128K, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_256K_64, L2U,    256*1024,       64, \
-       "Unified L2 cache, 256K, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_512K_64, L2U,    512*1024,       64, \
-       "Unified L2 cache, 512K, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_1M_64,   L2U,    1*1024*1024,    64, \
-       "Unified L2 cache, 1M, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_256K_32, L2U,    256*1024,       32, \
-       "Unified L2 cache, 256K, 8-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_512K_32, L2U,    512*1024,       32, \
-       "Unified L2 cache, 512K, 8-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_1M_32,   L2U,    1*1024*1024,    32, \
-       "Unified L2 cache, 1M, 8-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_2M_32,   L2U,    2*1024*1024,    32, \
-       "Unified L2 cache, 2M, 8-way set associative, 32byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_1M_64_4,  L2U,    1*1024*1024,    64, \
-        "Unified L2 cache, 1M, 4-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_2M_64,    L2U,    2*1024*1024,    64, \
-        "Unified L2 cache, 2M, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_512K_64_2,L2U,    512*1024,       64, \
-        "Unified L2 cache, 512K, 2-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_512K_64_4,L2U,    512*1024,       64, \
-        "Unified L2 cache, 512K, 4-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_1M_64_8,  L2U,    1*1024*1024,    64, \
-        "Unified L2 cache, 1M, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_128K_S4,  L2U,    128*1024,       64, \
-        "Unified L2 sectored cache, 128K, 4-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_128K_S2,  L2U,    128*1024,       64, \
-        "Unified L2 sectored cache, 128K, 2-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_UCACHE_256K_S4,  L2U,    256*1024,       64, \
-        "Unified L2 sectored cache, 256K, 4-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_L3CACHE_512K,    L3U,    512*1024,       64, \
-        "Unified L3 cache, 512K, 4-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_L3CACHE_1M,      L3U,    1*1024*1024,    64, \
-        "Unified L3 cache, 1M, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_L3CACHE_2M,      L3U,    2*1024*1024,    64, \
-        "Unified L3 cache, 2M, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_L3CACHE_4M,      L3U,    4*1024*1024,    64, \
-        "Unified L3 cache, 4M, 8-way set associative, 64byte line size"),
-CACHE_DESC(CPUID_CACHE_PREFETCH_64,     Lnone,  0,              0,  \
-        "64-Byte Prefetching"),
-CACHE_DESC(CPUID_CACHE_PREFETCH_128,    Lnone,  0,              0,  \
-        "128-Byte Prefetching"),
-CACHE_DESC(CPUID_CACHE_NOCACHE, Lnone, 0, 0, \
-        "No L2 cache or, if valid L2 cache, no L3 cache"),
-CACHE_DESC(CPUID_CACHE_NULL, Lnone, 0, 0, \
-       (char *)0),
-};
+void
+cpuid_do_was(void)
+{
+       do_cwas(cpuid_info(), TRUE);
+}
 
-static const char * get_intel_model_string( i386_cpu_info_t * info_p, cpu_type_t* type, cpu_subtype_t* subtype)
-{
-       *type = CPU_TYPE_X86;
-       *subtype = CPU_SUBTYPE_X86_ARCH1;
-
-    /* check for brand id string */
-    switch(info_p->cpuid_brand) {
-        case CPUID_BRAND_UNSUPPORTED:
-            /* brand ID not supported; use alternate method. */
-            switch(info_p->cpuid_family) {
-                case CPUID_FAMILY_486:
-                    return "Intel 486";
-                case CPUID_FAMILY_586:
-                    return "Intel Pentium";
-                case CPUID_FAMILY_686:
-                    switch(info_p->cpuid_model) {
-                        case CPUID_MODEL_P6:
-                            return "Intel Pentium Pro";
-                        case CPUID_MODEL_PII:
-                            return "Intel Pentium II";
-                        case CPUID_MODEL_P65:
-                        case CPUID_MODEL_P66:
-                            return "Intel Celeron";
-                        case CPUID_MODEL_P67:
-                        case CPUID_MODEL_P68:
-                        case CPUID_MODEL_P6A:
-                        case CPUID_MODEL_P6B:
-                            return "Intel Pentium III";
-                        case CPUID_MODEL_PM9:
-                        case CPUID_MODEL_PMD:
-                            return "Intel Pentium M";
-                        default:
-                            return "Unknown Intel P6 Family";
-                    }
-                case CPUID_FAMILY_EXTENDED:
-                    switch (info_p->cpuid_extfamily) {
-                        case CPUID_EXTFAMILY_PENTIUM4:
-                           *subtype = CPU_SUBTYPE_PENTIUM_4;
-                            return "Intel Pentium 4";
-                                               default:
-                                   return "Unknown Intel Extended Family";
-                    }
-                default:
-                    return "Unknown Intel Family";
-            }
-            break;
-        case CPUID_BRAND_CELERON_1:
-        case CPUID_BRAND_CELERON_A:
-        case CPUID_BRAND_CELERON_14:
-            return "Intel Celeron";
-        case CPUID_BRAND_PENTIUM_III_2:
-        case CPUID_BRAND_PENTIUM_III_4:
-            return "Pentium III";
-        case CPUID_BRAND_PIII_XEON:
-                       if (info_p->cpuid_signature == 0x6B1) {
-                               return "Intel Celeron";
-                       } else {
-                               return "Intel Pentium III Xeon";
-                       }
-        case CPUID_BRAND_PENTIUM_III_M:
-            return "Mobile Intel Pentium III-M";
-        case CPUID_BRAND_M_CELERON_7:
-        case CPUID_BRAND_M_CELERON_F:
-        case CPUID_BRAND_M_CELERON_13:
-        case CPUID_BRAND_M_CELERON_17:
-            return "Mobile Intel Celeron";
-        case CPUID_BRAND_PENTIUM4_8:
-        case CPUID_BRAND_PENTIUM4_9:
-           *subtype = CPU_SUBTYPE_PENTIUM_4;
-            return "Intel Pentium 4";
-        case CPUID_BRAND_XEON:
-            return "Intel Xeon";
-        case CPUID_BRAND_XEON_MP:
-            return "Intel Xeon MP";
-        case CPUID_BRAND_PENTIUM4_M:
-                       if (info_p->cpuid_signature == 0xF13) {
-                               return "Intel Xeon";
-                       } else {
-                               *subtype = CPU_SUBTYPE_PENTIUM_4;
-                               return "Mobile Intel Pentium 4";
-                       }
-        case CPUID_BRAND_CELERON_M:
-            return "Intel Celeron M";
-        case CPUID_BRAND_PENTIUM_M:
-            return "Intel Pentium M";
-        case CPUID_BRAND_MOBILE_15:
-        case CPUID_BRAND_MOBILE_17:
-            return "Mobile Intel";
-    }        
-    return "Unknown Intel";
-}
-
-static void set_intel_cache_info( i386_cpu_info_t * info_p )
-{
-       uint32_t        cpuid_result[4];
-        uint32_t        l1d_cache_linesize = 0;
-       unsigned int    i;
-       unsigned int    j;
-
-       /* get processor cache descriptor info */
-       do_cpuid(2, cpuid_result);
+/* this function is Intel-specific */
+static void
+cpuid_set_cache_info( i386_cpu_info_t * info_p )
+{
+       uint32_t        cpuid_result[4];
+       uint32_t        reg[4];
+       uint32_t        index;
+       uint32_t        linesizes[LCACHE_MAX];
+       unsigned int    i;
+       unsigned int    j;
+       boolean_t       cpuid_deterministic_supported = FALSE;
+
+       DBG("cpuid_set_cache_info(%p)\n", info_p);
+
+       bzero( linesizes, sizeof(linesizes));
+
+       /* Get processor cache descriptor info using leaf 2.  We don't use
+        * this internally, but must publish it for KEXTs.
+        */
+       cpuid_fn(2, cpuid_result);
        for (j = 0; j < 4; j++) {
-               if ((cpuid_result[j] >> 31) == 1)       /* bit31 is validity */
+               if ((cpuid_result[j] >> 31) == 1) {     /* bit31 is validity */
                        continue;
-               ((uint32_t *) info_p->cache_info)[j] = cpuid_result[j];
+               }
+               ((uint32_t *)(void *)info_p->cache_info)[j] = cpuid_result[j];
        }
        /* first byte gives number of cpuid calls to get all descriptors */
        for (i = 1; i < info_p->cache_info[0]; i++) {
-               if (i*16 > sizeof(info_p->cache_info))
+               if (i * 16 > sizeof(info_p->cache_info)) {
                        break;
-               do_cpuid(2, cpuid_result);
+               }
+               cpuid_fn(2, cpuid_result);
                for (j = 0; j < 4; j++) {
-                       if ((cpuid_result[j] >> 31) == 1) 
+                       if ((cpuid_result[j] >> 31) == 1) {
                                continue;
-                       ((uint32_t *) info_p->cache_info)[4*i+j] =
-                               cpuid_result[j];
+                       }
+                       ((uint32_t *)(void *)info_p->cache_info)[4 * i + j] =
+                           cpuid_result[j];
                }
        }
 
-       /* decode the descriptors looking for L1/L2/L3 size info */
-       for (i = 1; i < sizeof(info_p->cache_info); i++) {
-               cpuid_cache_desc_t      *descp;
-               uint8_t                 desc = info_p->cache_info[i];
+       /*
+        * Get cache info using leaf 4, the "deterministic cache parameters."
+        * Most processors Mac OS X supports implement this flavor of CPUID.
+        * Loop over each cache on the processor.
+        */
+       cpuid_fn(0, cpuid_result);
+       if (cpuid_result[eax] >= 4) {
+               cpuid_deterministic_supported = TRUE;
+       }
 
-               if (desc == CPUID_CACHE_NULL)
-                       continue;
-               for (descp = cpuid_cache_desc_tab;
-                       descp->value != CPUID_CACHE_NULL; descp++) {
-                       if (descp->value != desc)
-                               continue;
-                       info_p->cache_size[descp->type] = descp->size;
-                       if (descp->type == L2U)
-                               info_p->cache_linesize = descp->linesize;
-                        if (descp->type == L1D)
-                                l1d_cache_linesize = descp->linesize;
+       for (index = 0; cpuid_deterministic_supported; index++) {
+               cache_type_t    type = Lnone;
+               uint32_t        cache_type;
+               uint32_t        cache_level;
+               uint32_t        cache_sharing;
+               uint32_t        cache_linesize;
+               uint32_t        cache_sets;
+               uint32_t        cache_associativity;
+               uint32_t        cache_size;
+               uint32_t        cache_partitions;
+               uint32_t        colors;
+
+               reg[eax] = 4;           /* cpuid request 4 */
+               reg[ecx] = index;       /* index starting at 0 */
+               cpuid(reg);
+               DBG("cpuid(4) index=%d eax=0x%x\n", index, reg[eax]);
+               cache_type = bitfield32(reg[eax], 4, 0);
+               if (cache_type == 0) {
+                       break;          /* no more caches */
+               }
+               cache_level             = bitfield32(reg[eax], 7, 5);
+               cache_sharing           = bitfield32(reg[eax], 25, 14) + 1;
+               info_p->cpuid_cores_per_package
+                       = bitfield32(reg[eax], 31, 26) + 1;
+               cache_linesize          = bitfield32(reg[ebx], 11, 0) + 1;
+               cache_partitions        = bitfield32(reg[ebx], 21, 12) + 1;
+               cache_associativity     = bitfield32(reg[ebx], 31, 22) + 1;
+               cache_sets              = bitfield32(reg[ecx], 31, 0) + 1;
+
+               /* Map type/levels returned by CPUID into cache_type_t */
+               switch (cache_level) {
+               case 1:
+                       type = cache_type == 1 ? L1D :
+                           cache_type == 2 ? L1I :
+                           Lnone;
+                       break;
+               case 2:
+                       type = cache_type == 3 ? L2U :
+                           Lnone;
                        break;
+               case 3:
+                       type = cache_type == 3 ? L3U :
+                           Lnone;
+                       break;
+               default:
+                       type = Lnone;
+               }
+
+               /* The total size of a cache is:
+                *      ( linesize * sets * associativity * partitions )
+                */
+               if (type != Lnone) {
+                       cache_size = cache_linesize * cache_sets *
+                           cache_associativity * cache_partitions;
+                       info_p->cache_size[type] = cache_size;
+                       info_p->cache_sharing[type] = cache_sharing;
+                       info_p->cache_partitions[type] = cache_partitions;
+                       linesizes[type] = cache_linesize;
+
+                       DBG(" cache_size[%s]      : %d\n",
+                           cache_type_str[type], cache_size);
+                       DBG(" cache_sharing[%s]   : %d\n",
+                           cache_type_str[type], cache_sharing);
+                       DBG(" cache_partitions[%s]: %d\n",
+                           cache_type_str[type], cache_partitions);
+
+                       /*
+                        * Overwrite associativity determined via
+                        * CPUID.0x80000006 -- this leaf is more
+                        * accurate
+                        */
+                       if (type == L2U) {
+                               info_p->cpuid_cache_L2_associativity = cache_associativity;
+                       }
+                       /*
+                        * Adjust #sets to account for the N CBos
+                        * This is because addresses are hashed across CBos
+                        */
+                       if (type == L3U && info_p->core_count) {
+                               cache_sets = cache_sets / info_p->core_count;
+                       }
+
+                       /* Compute the number of page colors for this cache,
+                        * which is:
+                        *      ( linesize * sets ) / page_size
+                        *
+                        * To help visualize this, consider two views of a
+                        * physical address.  To the cache, it is composed
+                        * of a line offset, a set selector, and a tag.
+                        * To VM, it is composed of a page offset, a page
+                        * color, and other bits in the pageframe number:
+                        *
+                        *           +-----------------+---------+--------+
+                        *  cache:   |       tag       |   set   | offset |
+                        *           +-----------------+---------+--------+
+                        *
+                        *           +-----------------+-------+----------+
+                        *  VM:      |    don't care   | color | pg offset|
+                        *           +-----------------+-------+----------+
+                        *
+                        * The color is those bits in (set+offset) not covered
+                        * by the page offset.
+                        */
+                       colors = (cache_linesize * cache_sets) >> 12;
+
+                       if (colors > vm_cache_geometry_colors) {
+                               vm_cache_geometry_colors = colors;
+                       }
                }
        }
-       /* For P-IIIs, L2 could be 256k or 512k but we can't tell */ 
-       if (info_p->cache_size[L2U] == 0 &&
-           info_p->cpuid_family == 0x6 && info_p->cpuid_model == 0xb) {
-               info_p->cache_size[L2U] = 256*1024;
-               info_p->cache_linesize = 32;
+       DBG(" vm_cache_geometry_colors: %d\n", vm_cache_geometry_colors);
+
+       /*
+        * If deterministic cache parameters are not available, use
+        * something else
+        */
+       if (info_p->cpuid_cores_per_package == 0) {
+               info_p->cpuid_cores_per_package = 1;
+
+               /* cpuid define in 1024 quantities */
+               info_p->cache_size[L2U] = info_p->cpuid_cache_size * 1024;
+               info_p->cache_sharing[L2U] = 1;
+               info_p->cache_partitions[L2U] = 1;
+
+               linesizes[L2U] = info_p->cpuid_cache_linesize;
+
+               DBG(" cache_size[L2U]      : %d\n",
+                   info_p->cache_size[L2U]);
+               DBG(" cache_sharing[L2U]   : 1\n");
+               DBG(" cache_partitions[L2U]: 1\n");
+               DBG(" linesizes[L2U]       : %d\n",
+                   info_p->cpuid_cache_linesize);
        }
-        /* If we have no L2 cache, use the L1 data cache line size */
-        if (info_p->cache_size[L2U] == 0)
-            info_p->cache_linesize = l1d_cache_linesize;
 
        /*
-        * Get cache sharing info if available.
+        * What linesize to publish?  We use the L2 linesize if any,
+        * else the L1D.
         */
-       do_cpuid(0, cpuid_result);
-       if (cpuid_result[eax] >= 4) {
-               uint32_t        reg[4];
-               uint32_t        index;
-               for (index = 0;; index++) {
-                       /*
-                        * Scan making calls for cpuid with %eax = 4
-                        * to get info about successive cache levels
-                        * until a null type is returned.
-                        */
-                       cache_type_t    type = Lnone;
-                       uint32_t        cache_type;
-                       uint32_t        cache_level;
-                       uint32_t        cache_sharing;
-
-                       reg[eax] = 4;           /* cpuid request 4 */
-                       reg[ecx] = index;       /* index starting at 0 */
-                       cpuid(reg);
-//kprintf("cpuid(4) index=%d eax=%p\n", index, reg[eax]);
-                       cache_type = bitfield(reg[eax], 4, 0);
-                       if (cache_type == 0)
-                               break;          /* done with cache info */
-                       cache_level   = bitfield(reg[eax],  7,  5);
-                       cache_sharing = bitfield(reg[eax], 25, 14);
-                       info_p->cpuid_cores_per_package = 
-                                       bitfield(reg[eax], 31, 26) + 1;
-                       switch (cache_level) {
-                       case 1:
-                               type = cache_type == 1 ? L1D :
-                                      cache_type == 2 ? L1I :
-                                                        Lnone;
+       if (linesizes[L2U]) {
+               info_p->cache_linesize = linesizes[L2U];
+       } else if (linesizes[L1D]) {
+               info_p->cache_linesize = linesizes[L1D];
+       } else {
+               panic("no linesize");
+       }
+       DBG(" cache_linesize    : %d\n", info_p->cache_linesize);
+
+       /*
+        * Extract and publish TLB information from Leaf 2 descriptors.
+        */
+       DBG(" %ld leaf2 descriptors:\n", sizeof(info_p->cache_info));
+       for (i = 1; i < sizeof(info_p->cache_info); i++) {
+               cpuid_cache_descriptor_t        *descp;
+               int                             id;
+               int                             level;
+               int                             page;
+
+               DBG(" 0x%02x", info_p->cache_info[i]);
+               descp = cpuid_leaf2_find(info_p->cache_info[i]);
+               if (descp == NULL) {
+                       continue;
+               }
+
+               switch (descp->type) {
+               case TLB:
+                       page = (descp->size == SMALL) ? TLB_SMALL : TLB_LARGE;
+                       /* determine I or D: */
+                       switch (descp->level) {
+                       case INST:
+                               id = TLB_INST;
                                break;
-                       case 2:
-                               type = cache_type == 3 ? L2U :
-                                                        Lnone;
+                       case DATA:
+                       case DATA0:
+                       case DATA1:
+                               id = TLB_DATA;
+                               break;
+                       default:
+                               continue;
+                       }
+                       /* determine level: */
+                       switch (descp->level) {
+                       case DATA1:
+                               level = 1;
                                break;
-                       case 3:
-                               type = cache_type == 3 ? L3U :
-                                                        Lnone;
+                       default:
+                               level = 0;
                        }
-                       if (type != Lnone)
-                               info_p->cache_sharing[type] = cache_sharing + 1;
-               } 
-       }
-}
-
-static void set_cpu_intel( i386_cpu_info_t * info_p )
-{
-    set_cpu_generic(info_p);
-    set_intel_cache_info(info_p);
-    info_p->cpuid_model_string = get_intel_model_string(info_p, &info_p->cpuid_cpu_type, &info_p->cpuid_cpu_subtype);
-}
-
-static const char * get_amd_model_string( i386_cpu_info_t * info_p, cpu_type_t* type, cpu_subtype_t* subtype )
-{
-       *type = CPU_TYPE_X86;
-       *subtype = CPU_SUBTYPE_X86_ARCH1;
-
-    /* check for brand id string */
-    switch (info_p->cpuid_family)
-    {
-        case CPUID_FAMILY_486:
-            switch (info_p->cpuid_model) {
-                case CPUID_MODEL_AM486_DX:
-                case CPUID_MODEL_AM486_DX2:
-                case CPUID_MODEL_AM486_DX2WB:
-                case CPUID_MODEL_AM486_DX4:
-                case CPUID_MODEL_AM486_DX4WB:
-                    return "Am486";
-                case CPUID_MODEL_AM486_5X86:
-                case CPUID_MODEL_AM486_5X86WB:
-                    return "Am5x86";
-            }
-            break;
-        case CPUID_FAMILY_586:
-            switch (info_p->cpuid_model) {
-                case CPUID_MODEL_K5M0:
-                case CPUID_MODEL_K5M1:
-                case CPUID_MODEL_K5M2:
-                case CPUID_MODEL_K5M3:
-                    return "AMD-K5";
-                case CPUID_MODEL_K6M6:
-                case CPUID_MODEL_K6M7:
-                    return "AMD-K6";
-                case CPUID_MODEL_K6_2:
-                    return "AMD-K6-2";
-                case CPUID_MODEL_K6_III:
-                    return "AMD-K6-III";
-            }
-            break;
-        case CPUID_FAMILY_686:
-            switch (info_p->cpuid_model) {
-                case CPUID_MODEL_ATHLON_M1:
-                case CPUID_MODEL_ATHLON_M2:
-                case CPUID_MODEL_ATHLON_M4:
-                case CPUID_MODEL_ATHLON_M6:
-                case CPUID_MODEL_ATHLON_M8:
-                case CPUID_MODEL_ATHLON_M10:
-                    return "AMD Athlon";
-                case CPUID_MODEL_DURON_M3:
-                case CPUID_MODEL_DURON_M7:
-                    return "AMD Duron";
-                default:
-                    return "Unknown AMD Athlon";
-            }
-        case CPUID_FAMILY_EXTENDED:
-            switch (info_p->cpuid_model) {
-                case CPUID_MODEL_ATHLON64:
-                    return "AMD Athlon 64";
-                case CPUID_MODEL_OPTERON:
-                    return "AMD Opteron";
-                default:
-                    return "Unknown AMD-64";
-            }
-    }
-    return "Unknown AMD";
-}
-
-static void set_amd_cache_info( i386_cpu_info_t * info_p )
-{
-    uint32_t   cpuid_result[4];
-
-    /* It would make sense to fill in info_p->cache_info with complete information
-     * on the TLBs and data cache associativity, lines, etc, either by mapping
-     * to the Intel tags (if possible), or replacing cache_info with a generic
-     * mechanism.  But right now, nothing makes use of that information (that I know
-     * of).
-     */
-
-    /* L1 Cache and TLB Information */
-    do_cpuid(0x80000005, cpuid_result);
-    
-    /* EAX: TLB Information for 2-Mbyte and 4-MByte Pages */
-    /* (ignore) */
-    
-    /* EBX: TLB Information for 4-Kbyte Pages */
-    /* (ignore) */
-    
-    /* ECX: L1 Data Cache Information */
-    info_p->cache_size[L1D] = ((cpuid_result[ecx] >> 24) & 0xFF) * 1024;
-    info_p->cache_linesize = (cpuid_result[ecx] & 0xFF);
-    
-    /* EDX: L1 Instruction Cache Information */
-    info_p->cache_size[L1I] = ((cpuid_result[edx] >> 24) & 0xFF) * 1024;
-
-    /* L2 Cache Information */
-    do_cpuid(0x80000006, cpuid_result);
-    
-    /* EAX: L2 TLB Information for 2-Mbyte and 4-Mbyte Pages */
-    /* (ignore) */
-    
-    /* EBX: L2 TLB Information for 4-Kbyte Pages */
-    /* (ignore) */
-    
-    /* ECX: L2 Cache Information */
-    info_p->cache_size[L2U] = ((cpuid_result[ecx] >> 16) & 0xFFFF) * 1024;
-    if (info_p->cache_size[L2U] > 0)
-        info_p->cache_linesize = cpuid_result[ecx] & 0xFF;
-}
-
-static void set_cpu_amd( i386_cpu_info_t * info_p )
-{
-    set_cpu_generic(info_p);
-    set_amd_cache_info(info_p);
-    info_p->cpuid_model_string = get_amd_model_string(info_p, &info_p->cpuid_cpu_type, &info_p->cpuid_cpu_subtype);
-}
-
-static void set_cpu_nsc( i386_cpu_info_t * info_p )
-{
-    set_cpu_generic(info_p);
-    set_amd_cache_info(info_p);
-
-    /* check for brand id string */
-    if (info_p->cpuid_family == CPUID_FAMILY_586 && info_p->cpuid_model == CPUID_MODEL_GX1) {
-        info_p->cpuid_model_string = "AMD Geode GX1";
-    } else if (info_p->cpuid_family == CPUID_FAMILY_586 && info_p->cpuid_model == CPUID_MODEL_GX2) {
-        info_p->cpuid_model_string = "AMD Geode GX";
-    } else {
-        info_p->cpuid_model_string = "Unknown National Semiconductor";
-    }
-    info_p->cpuid_cpu_type = CPU_TYPE_X86;
-    info_p->cpuid_cpu_subtype = CPU_SUBTYPE_X86_ARCH1;
+                       info_p->cpuid_tlb[id][page][level] = descp->entries;
+                       break;
+               case STLB:
+                       info_p->cpuid_stlb = descp->entries;
+               }
+       }
+       DBG("\n");
 }
 
 static void
-set_cpu_generic(i386_cpu_info_t *info_p)
+cpuid_set_generic_info(i386_cpu_info_t *info_p)
 {
-       uint32_t        cpuid_result[4];
-       uint32_t        max_extid;
-        char            str[128], *p;
+       uint32_t        reg[4];
+       char            str[128], *p;
+
+       DBG("cpuid_set_generic_info(%p)\n", info_p);
+
+       /* do cpuid 0 to get vendor */
+       cpuid_fn(0, reg);
+       info_p->cpuid_max_basic = reg[eax];
+       bcopy((char *)&reg[ebx], &info_p->cpuid_vendor[0], 4); /* ug */
+       bcopy((char *)&reg[ecx], &info_p->cpuid_vendor[8], 4);
+       bcopy((char *)&reg[edx], &info_p->cpuid_vendor[4], 4);
+       info_p->cpuid_vendor[12] = 0;
 
        /* get extended cpuid results */
-       do_cpuid(0x80000000, cpuid_result);
-       max_extid = cpuid_result[eax];
+       cpuid_fn(0x80000000, reg);
+       info_p->cpuid_max_ext = reg[eax];
 
        /* check to see if we can get brand string */
-       if (max_extid >= 0x80000004) {
+       if (info_p->cpuid_max_ext >= 0x80000004) {
                /*
                 * The brand string 48 bytes (max), guaranteed to
                 * be NUL terminated.
                 */
-               do_cpuid(0x80000002, cpuid_result);
-               bcopy((char *)cpuid_result, &str[0], 16);
-               do_cpuid(0x80000003, cpuid_result);
-               bcopy((char *)cpuid_result, &str[16], 16);
-               do_cpuid(0x80000004, cpuid_result);
-               bcopy((char *)cpuid_result, &str[32], 16);
+               cpuid_fn(0x80000002, reg);
+               bcopy((char *)reg, &str[0], 16);
+               cpuid_fn(0x80000003, reg);
+               bcopy((char *)reg, &str[16], 16);
+               cpuid_fn(0x80000004, reg);
+               bcopy((char *)reg, &str[32], 16);
                for (p = str; *p != '\0'; p++) {
-                       if (*p != ' ') break;
+                       if (*p != ' ') {
+                               break;
+                       }
+               }
+               strlcpy(info_p->cpuid_brand_string,
+                   p, sizeof(info_p->cpuid_brand_string));
+
+               if (!strncmp(info_p->cpuid_brand_string, CPUID_STRING_UNKNOWN,
+                   min(sizeof(info_p->cpuid_brand_string),
+                   strlen(CPUID_STRING_UNKNOWN) + 1))) {
+                       /*
+                        * This string means we have a firmware-programmable brand string,
+                        * and the firmware couldn't figure out what sort of CPU we have.
+                        */
+                       info_p->cpuid_brand_string[0] = '\0';
+               }
+       }
+
+       /* Get cache and addressing info. */
+       if (info_p->cpuid_max_ext >= 0x80000006) {
+               uint32_t assoc;
+               cpuid_fn(0x80000006, reg);
+               info_p->cpuid_cache_linesize   = bitfield32(reg[ecx], 7, 0);
+               assoc = bitfield32(reg[ecx], 15, 12);
+               /*
+                * L2 associativity is encoded, though in an insufficiently
+                * descriptive fashion, e.g. 24-way is mapped to 16-way.
+                * Represent a fully associative cache as 0xFFFF.
+                * Overwritten by associativity as determined via CPUID.4
+                * if available.
+                */
+               if (assoc == 6) {
+                       assoc = 8;
+               } else if (assoc == 8) {
+                       assoc = 16;
+               } else if (assoc == 0xF) {
+                       assoc = 0xFFFF;
                }
-               strncpy(info_p->cpuid_brand_string,
-                       p, sizeof(info_p->cpuid_brand_string)-1);
-               info_p->cpuid_brand_string[sizeof(info_p->cpuid_brand_string)-1] = '\0';
-
-                if (!strcmp(info_p->cpuid_brand_string, CPUID_STRING_UNKNOWN)) {
-                    /*
-                     * This string means we have a BIOS-programmable brand string,
-                     * and the BIOS couldn't figure out what sort of CPU we have.
-                     */
-                    info_p->cpuid_brand_string[0] = '\0';
-                }
-       }
-    
-       /* get processor signature and decode */
-       do_cpuid(1, cpuid_result);
-       info_p->cpuid_signature = cpuid_result[eax];
-       info_p->cpuid_stepping  = bitfield(cpuid_result[eax],  3,  0);
-       info_p->cpuid_model     = bitfield(cpuid_result[eax],  7,  4);
-       info_p->cpuid_family    = bitfield(cpuid_result[eax], 11,  8);
-       info_p->cpuid_type      = bitfield(cpuid_result[eax], 13, 12);
-       info_p->cpuid_extmodel  = bitfield(cpuid_result[eax], 19, 16);
-       info_p->cpuid_extfamily = bitfield(cpuid_result[eax], 27, 20);
-       info_p->cpuid_brand     = bitfield(cpuid_result[ebx],  7,  0);
-       info_p->cpuid_logical_per_package =
-                                 bitfield(cpuid_result[ebx], 23, 16);
-       info_p->cpuid_features  = quad(cpuid_result[ecx], cpuid_result[edx]);
-
-       if (max_extid >= 0x80000001) {
-               do_cpuid(0x80000001, cpuid_result);
+               info_p->cpuid_cache_L2_associativity = assoc;
+               info_p->cpuid_cache_size       = bitfield32(reg[ecx], 31, 16);
+               cpuid_fn(0x80000008, reg);
+               info_p->cpuid_address_bits_physical =
+                   bitfield32(reg[eax], 7, 0);
+               info_p->cpuid_address_bits_virtual =
+                   bitfield32(reg[eax], 15, 8);
+       }
+
+       /*
+        * Get processor signature and decode
+        * and bracket this with the approved procedure for reading the
+        * the microcode version number a.k.a. signature a.k.a. BIOS ID
+        */
+       wrmsr64(MSR_IA32_BIOS_SIGN_ID, 0);
+       cpuid_fn(1, reg);
+       info_p->cpuid_microcode_version =
+           (uint32_t) (rdmsr64(MSR_IA32_BIOS_SIGN_ID) >> 32);
+       info_p->cpuid_signature = reg[eax];
+       info_p->cpuid_stepping  = bitfield32(reg[eax], 3, 0);
+       info_p->cpuid_model     = bitfield32(reg[eax], 7, 4);
+       info_p->cpuid_family    = bitfield32(reg[eax], 11, 8);
+       info_p->cpuid_type      = bitfield32(reg[eax], 13, 12);
+       info_p->cpuid_extmodel  = bitfield32(reg[eax], 19, 16);
+       info_p->cpuid_extfamily = bitfield32(reg[eax], 27, 20);
+       info_p->cpuid_brand     = bitfield32(reg[ebx], 7, 0);
+       info_p->cpuid_features  = quad(reg[ecx], reg[edx]);
+
+       /* Get "processor flag"; necessary for microcode update matching */
+       info_p->cpuid_processor_flag = (rdmsr64(MSR_IA32_PLATFORM_ID) >> 50) & 0x7;
+
+       /* Fold extensions into family/model */
+       if (info_p->cpuid_family == 0x0f) {
+               info_p->cpuid_family += info_p->cpuid_extfamily;
+       }
+       if (info_p->cpuid_family == 0x0f || info_p->cpuid_family == 0x06) {
+               info_p->cpuid_model += (info_p->cpuid_extmodel << 4);
+       }
+
+       if (info_p->cpuid_features & CPUID_FEATURE_HTT) {
+               info_p->cpuid_logical_per_package =
+                   bitfield32(reg[ebx], 23, 16);
+       } else {
+               info_p->cpuid_logical_per_package = 1;
+       }
+
+       if (info_p->cpuid_max_ext >= 0x80000001) {
+               cpuid_fn(0x80000001, reg);
                info_p->cpuid_extfeatures =
-                               quad(cpuid_result[ecx], cpuid_result[edx]);
+                   quad(reg[ecx], reg[edx]);
+       }
+
+       DBG(" max_basic           : %d\n", info_p->cpuid_max_basic);
+       DBG(" max_ext             : 0x%08x\n", info_p->cpuid_max_ext);
+       DBG(" vendor              : %s\n", info_p->cpuid_vendor);
+       DBG(" brand_string        : %s\n", info_p->cpuid_brand_string);
+       DBG(" signature           : 0x%08x\n", info_p->cpuid_signature);
+       DBG(" stepping            : %d\n", info_p->cpuid_stepping);
+       DBG(" model               : %d\n", info_p->cpuid_model);
+       DBG(" family              : %d\n", info_p->cpuid_family);
+       DBG(" type                : %d\n", info_p->cpuid_type);
+       DBG(" extmodel            : %d\n", info_p->cpuid_extmodel);
+       DBG(" extfamily           : %d\n", info_p->cpuid_extfamily);
+       DBG(" brand               : %d\n", info_p->cpuid_brand);
+       DBG(" features            : 0x%016llx\n", info_p->cpuid_features);
+       DBG(" extfeatures         : 0x%016llx\n", info_p->cpuid_extfeatures);
+       DBG(" logical_per_package : %d\n", info_p->cpuid_logical_per_package);
+       DBG(" microcode_version   : 0x%08x\n", info_p->cpuid_microcode_version);
+
+       /* Fold in the Invariant TSC feature bit, if present */
+       if (info_p->cpuid_max_ext >= 0x80000007) {
+               cpuid_fn(0x80000007, reg);
+               info_p->cpuid_extfeatures |=
+                   reg[edx] & (uint32_t)CPUID_EXTFEATURE_TSCI;
+               DBG(" extfeatures         : 0x%016llx\n",
+                   info_p->cpuid_extfeatures);
+       }
+
+       if (info_p->cpuid_max_basic >= 0x5) {
+               cpuid_mwait_leaf_t      *cmp = &info_p->cpuid_mwait_leaf;
+
+               /*
+                * Extract the Monitor/Mwait Leaf info:
+                */
+               cpuid_fn(5, reg);
+               cmp->linesize_min = reg[eax];
+               cmp->linesize_max = reg[ebx];
+               cmp->extensions   = reg[ecx];
+               cmp->sub_Cstates  = reg[edx];
+               info_p->cpuid_mwait_leafp = cmp;
+
+               DBG(" Monitor/Mwait Leaf:\n");
+               DBG("  linesize_min : %d\n", cmp->linesize_min);
+               DBG("  linesize_max : %d\n", cmp->linesize_max);
+               DBG("  extensions   : %d\n", cmp->extensions);
+               DBG("  sub_Cstates  : 0x%08x\n", cmp->sub_Cstates);
+       }
+
+       if (info_p->cpuid_max_basic >= 0x6) {
+               cpuid_thermal_leaf_t    *ctp = &info_p->cpuid_thermal_leaf;
+
+               /*
+                * The thermal and Power Leaf:
+                */
+               cpuid_fn(6, reg);
+               ctp->sensor               = bitfield32(reg[eax], 0, 0);
+               ctp->dynamic_acceleration = bitfield32(reg[eax], 1, 1);
+               ctp->invariant_APIC_timer = bitfield32(reg[eax], 2, 2);
+               ctp->core_power_limits    = bitfield32(reg[eax], 4, 4);
+               ctp->fine_grain_clock_mod = bitfield32(reg[eax], 5, 5);
+               ctp->package_thermal_intr = bitfield32(reg[eax], 6, 6);
+               ctp->thresholds           = bitfield32(reg[ebx], 3, 0);
+               ctp->ACNT_MCNT            = bitfield32(reg[ecx], 0, 0);
+               ctp->hardware_feedback    = bitfield32(reg[ecx], 1, 1);
+               ctp->energy_policy        = bitfield32(reg[ecx], 3, 3);
+               info_p->cpuid_thermal_leafp = ctp;
+
+               DBG(" Thermal/Power Leaf:\n");
+               DBG("  sensor               : %d\n", ctp->sensor);
+               DBG("  dynamic_acceleration : %d\n", ctp->dynamic_acceleration);
+               DBG("  invariant_APIC_timer : %d\n", ctp->invariant_APIC_timer);
+               DBG("  core_power_limits    : %d\n", ctp->core_power_limits);
+               DBG("  fine_grain_clock_mod : %d\n", ctp->fine_grain_clock_mod);
+               DBG("  package_thermal_intr : %d\n", ctp->package_thermal_intr);
+               DBG("  thresholds           : %d\n", ctp->thresholds);
+               DBG("  ACNT_MCNT            : %d\n", ctp->ACNT_MCNT);
+               DBG("  ACNT2                : %d\n", ctp->hardware_feedback);
+               DBG("  energy_policy        : %d\n", ctp->energy_policy);
+       }
+
+       if (info_p->cpuid_max_basic >= 0xa) {
+               cpuid_arch_perf_leaf_t  *capp = &info_p->cpuid_arch_perf_leaf;
+
+               /*
+                * Architectural Performance Monitoring Leaf:
+                */
+               cpuid_fn(0xa, reg);
+               capp->version       = bitfield32(reg[eax], 7, 0);
+               capp->number        = bitfield32(reg[eax], 15, 8);
+               capp->width         = bitfield32(reg[eax], 23, 16);
+               capp->events_number = bitfield32(reg[eax], 31, 24);
+               capp->events        = reg[ebx];
+               capp->fixed_number  = bitfield32(reg[edx], 4, 0);
+               capp->fixed_width   = bitfield32(reg[edx], 12, 5);
+               info_p->cpuid_arch_perf_leafp = capp;
+
+               DBG(" Architectural Performance Monitoring Leaf:\n");
+               DBG("  version       : %d\n", capp->version);
+               DBG("  number        : %d\n", capp->number);
+               DBG("  width         : %d\n", capp->width);
+               DBG("  events_number : %d\n", capp->events_number);
+               DBG("  events        : %d\n", capp->events);
+               DBG("  fixed_number  : %d\n", capp->fixed_number);
+               DBG("  fixed_width   : %d\n", capp->fixed_width);
+       }
+
+       if (info_p->cpuid_max_basic >= 0xd) {
+               cpuid_xsave_leaf_t      *xsp;
+               /*
+                * XSAVE Features:
+                */
+               xsp = &info_p->cpuid_xsave_leaf[0];
+               info_p->cpuid_xsave_leafp = xsp;
+               xsp->extended_state[eax] = 0xd;
+               xsp->extended_state[ecx] = 0;
+               cpuid(xsp->extended_state);
+               DBG(" XSAVE Main leaf:\n");
+               DBG("  EAX           : 0x%x\n", xsp->extended_state[eax]);
+               DBG("  EBX           : 0x%x\n", xsp->extended_state[ebx]);
+               DBG("  ECX           : 0x%x\n", xsp->extended_state[ecx]);
+               DBG("  EDX           : 0x%x\n", xsp->extended_state[edx]);
+
+               xsp = &info_p->cpuid_xsave_leaf[1];
+               xsp->extended_state[eax] = 0xd;
+               xsp->extended_state[ecx] = 1;
+               cpuid(xsp->extended_state);
+               DBG(" XSAVE Sub-leaf1:\n");
+               DBG("  EAX           : 0x%x\n", xsp->extended_state[eax]);
+               DBG("  EBX           : 0x%x\n", xsp->extended_state[ebx]);
+               DBG("  ECX           : 0x%x\n", xsp->extended_state[ecx]);
+               DBG("  EDX           : 0x%x\n", xsp->extended_state[edx]);
+       }
+
+       if (info_p->cpuid_model >= CPUID_MODEL_IVYBRIDGE) {
+               /*
+                * Leaf7 Features:
+                */
+               cpuid_fn(0x7, reg);
+               info_p->cpuid_leaf7_features = quad(reg[ecx], reg[ebx]);
+               info_p->cpuid_leaf7_extfeatures = reg[edx];
+
+               cpuid_tsx_supported = (reg[ebx] & (CPUID_LEAF7_FEATURE_HLE | CPUID_LEAF7_FEATURE_RTM)) != 0;
+
+               DBG(" Feature Leaf7:\n");
+               DBG("  EBX           : 0x%x\n", reg[ebx]);
+               DBG("  ECX           : 0x%x\n", reg[ecx]);
+               DBG("  EDX           : 0x%x\n", reg[edx]);
+       }
+
+       if (info_p->cpuid_max_basic >= 0x15) {
+               /*
+                * TCS/CCC frequency leaf:
+                */
+               cpuid_fn(0x15, reg);
+               info_p->cpuid_tsc_leaf.denominator = reg[eax];
+               info_p->cpuid_tsc_leaf.numerator   = reg[ebx];
+
+               DBG(" TSC/CCC Information Leaf:\n");
+               DBG("  numerator     : 0x%x\n", reg[ebx]);
+               DBG("  denominator   : 0x%x\n", reg[eax]);
        }
 
        return;
 }
 
-static void
-set_cpu_unknown(__unused i386_cpu_info_t *info_p)
+static uint32_t
+cpuid_set_cpufamily(i386_cpu_info_t *info_p)
 {
-    info_p->cpuid_model_string = "Unknown";
+       uint32_t cpufamily = CPUFAMILY_UNKNOWN;
+
+       switch (info_p->cpuid_family) {
+       case 6:
+               switch (info_p->cpuid_model) {
+               case 23:
+                       cpufamily = CPUFAMILY_INTEL_PENRYN;
+                       break;
+               case CPUID_MODEL_NEHALEM:
+               case CPUID_MODEL_FIELDS:
+               case CPUID_MODEL_DALES:
+               case CPUID_MODEL_NEHALEM_EX:
+                       cpufamily = CPUFAMILY_INTEL_NEHALEM;
+                       break;
+               case CPUID_MODEL_DALES_32NM:
+               case CPUID_MODEL_WESTMERE:
+               case CPUID_MODEL_WESTMERE_EX:
+                       cpufamily = CPUFAMILY_INTEL_WESTMERE;
+                       break;
+               case CPUID_MODEL_SANDYBRIDGE:
+               case CPUID_MODEL_JAKETOWN:
+                       cpufamily = CPUFAMILY_INTEL_SANDYBRIDGE;
+                       break;
+               case CPUID_MODEL_IVYBRIDGE:
+               case CPUID_MODEL_IVYBRIDGE_EP:
+                       cpufamily = CPUFAMILY_INTEL_IVYBRIDGE;
+                       break;
+               case CPUID_MODEL_HASWELL:
+               case CPUID_MODEL_HASWELL_EP:
+               case CPUID_MODEL_HASWELL_ULT:
+               case CPUID_MODEL_CRYSTALWELL:
+                       cpufamily = CPUFAMILY_INTEL_HASWELL;
+                       break;
+               case CPUID_MODEL_BROADWELL:
+               case CPUID_MODEL_BRYSTALWELL:
+                       cpufamily = CPUFAMILY_INTEL_BROADWELL;
+                       break;
+               case CPUID_MODEL_SKYLAKE:
+               case CPUID_MODEL_SKYLAKE_DT:
+               case CPUID_MODEL_SKYLAKE_W:
+                       cpufamily = CPUFAMILY_INTEL_SKYLAKE;
+                       break;
+               case CPUID_MODEL_KABYLAKE:
+               case CPUID_MODEL_KABYLAKE_DT:
+                       cpufamily = CPUFAMILY_INTEL_KABYLAKE;
+                       break;
+               case CPUID_MODEL_ICELAKE:
+               case CPUID_MODEL_ICELAKE_H:
+               case CPUID_MODEL_ICELAKE_DT:
+                       cpufamily = CPUFAMILY_INTEL_ICELAKE;
+                       break;
+               }
+               break;
+       }
+
+       info_p->cpuid_cpufamily = cpufamily;
+       DBG("cpuid_set_cpufamily(%p) returning 0x%x\n", info_p, cpufamily);
+       return cpufamily;
 }
+/*
+ * Must be invoked either when executing single threaded, or with
+ * independent synchronization.
+ */
+void
+cpuid_set_info(void)
+{
+       i386_cpu_info_t         *info_p = &cpuid_cpu_info;
+       boolean_t               enable_x86_64h = TRUE;
+
+       /* Perform pre-cpuid workarounds (since their effects impact values returned via cpuid) */
+       cpuid_do_precpuid_was();
+
+       cpuid_set_generic_info(info_p);
+
+       /* verify we are running on a supported CPU */
+       if ((strncmp(CPUID_VID_INTEL, info_p->cpuid_vendor,
+           min(strlen(CPUID_STRING_UNKNOWN) + 1,
+           sizeof(info_p->cpuid_vendor)))) ||
+           (cpuid_set_cpufamily(info_p) == CPUFAMILY_UNKNOWN)) {
+               panic("Unsupported CPU");
+       }
+
+       info_p->cpuid_cpu_type = CPU_TYPE_X86;
+
+       if (!PE_parse_boot_argn("-enable_x86_64h", &enable_x86_64h, sizeof(enable_x86_64h))) {
+               boolean_t               disable_x86_64h = FALSE;
+
+               if (PE_parse_boot_argn("-disable_x86_64h", &disable_x86_64h, sizeof(disable_x86_64h))) {
+                       enable_x86_64h = FALSE;
+               }
+       }
+
+       if (enable_x86_64h &&
+           ((info_p->cpuid_features & CPUID_X86_64_H_FEATURE_SUBSET) == CPUID_X86_64_H_FEATURE_SUBSET) &&
+           ((info_p->cpuid_extfeatures & CPUID_X86_64_H_EXTFEATURE_SUBSET) == CPUID_X86_64_H_EXTFEATURE_SUBSET) &&
+           ((info_p->cpuid_leaf7_features & CPUID_X86_64_H_LEAF7_FEATURE_SUBSET) == CPUID_X86_64_H_LEAF7_FEATURE_SUBSET)) {
+               info_p->cpuid_cpu_subtype = CPU_SUBTYPE_X86_64_H;
+       } else {
+               info_p->cpuid_cpu_subtype = CPU_SUBTYPE_X86_ARCH1;
+       }
+       /* cpuid_set_cache_info must be invoked after set_generic_info */
+
+       /*
+        * Find the number of enabled cores and threads
+        * (which determines whether SMT/Hyperthreading is active).
+        */
+
+       /*
+        * Not all VMMs emulate MSR_CORE_THREAD_COUNT (0x35).
+        */
+       if (0 != (info_p->cpuid_features & CPUID_FEATURE_VMM) &&
+           PE_parse_boot_argn("-nomsr35h", NULL, 0)) {
+               info_p->core_count = 1;
+               info_p->thread_count = 1;
+               cpuid_set_cache_info(info_p);
+       } else {
+               switch (info_p->cpuid_cpufamily) {
+               case CPUFAMILY_INTEL_PENRYN:
+                       cpuid_set_cache_info(info_p);
+                       info_p->core_count   = info_p->cpuid_cores_per_package;
+                       info_p->thread_count = info_p->cpuid_logical_per_package;
+                       break;
+               case CPUFAMILY_INTEL_WESTMERE: {
+                       /*
+                        * This should be the same as Nehalem but an A0 silicon bug returns
+                        * invalid data in the top 12 bits. Hence, we use only bits [19..16]
+                        * rather than [31..16] for core count - which actually can't exceed 8.
+                        */
+                       uint64_t msr = rdmsr64(MSR_CORE_THREAD_COUNT);
+                       if (0 == msr) {
+                               /* Provide a non-zero default for some VMMs */
+                               msr = (1 << 16) | 1;
+                       }
+                       info_p->core_count   = bitfield32((uint32_t)msr, 19, 16);
+                       info_p->thread_count = bitfield32((uint32_t)msr, 15, 0);
+                       cpuid_set_cache_info(info_p);
+                       break;
+               }
+               default: {
+                       uint64_t msr = rdmsr64(MSR_CORE_THREAD_COUNT);
+                       if (0 == msr) {
+                               /* Provide a non-zero default for some VMMs */
+                               msr = (1 << 16) | 1;
+                       }
+                       info_p->core_count   = bitfield32((uint32_t)msr, 31, 16);
+                       info_p->thread_count = bitfield32((uint32_t)msr, 15, 0);
+                       cpuid_set_cache_info(info_p);
+                       break;
+               }
+               }
+       }
+
+       DBG("cpuid_set_info():\n");
+       DBG("  core_count   : %d\n", info_p->core_count);
+       DBG("  thread_count : %d\n", info_p->thread_count);
+       DBG("       cpu_type: 0x%08x\n", info_p->cpuid_cpu_type);
+       DBG("    cpu_subtype: 0x%08x\n", info_p->cpuid_cpu_subtype);
 
+       info_p->cpuid_model_string = ""; /* deprecated */
+
+       /* Init CPU LBRs */
+       i386_lbr_init(info_p, true);
+
+       do_cwas(info_p, FALSE);
+}
 
-static struct {
-       uint64_t        mask;
-       const char      *name;
+static struct table {
+       uint64_t        mask;
+       const char      *name;
 } feature_map[] = {
-       {CPUID_FEATURE_FPU,   "FPU",},
-       {CPUID_FEATURE_VME,   "VME",},
-       {CPUID_FEATURE_DE,    "DE",},
-       {CPUID_FEATURE_PSE,   "PSE",},
-       {CPUID_FEATURE_TSC,   "TSC",},
-       {CPUID_FEATURE_MSR,   "MSR",},
-       {CPUID_FEATURE_PAE,   "PAE",},
-       {CPUID_FEATURE_MCE,   "MCE",},
-       {CPUID_FEATURE_CX8,   "CX8",},
-       {CPUID_FEATURE_APIC,  "APIC",},
-       {CPUID_FEATURE_SEP,   "SEP",},
-       {CPUID_FEATURE_MTRR,  "MTRR",},
-       {CPUID_FEATURE_PGE,   "PGE",},
-       {CPUID_FEATURE_MCA,   "MCA",},
-       {CPUID_FEATURE_CMOV,  "CMOV",},
-       {CPUID_FEATURE_PAT,   "PAT",},
-       {CPUID_FEATURE_PSE36, "PSE36",},
-       {CPUID_FEATURE_PSN,   "PSN",},
-       {CPUID_FEATURE_CLFSH, "CLFSH",},
-       {CPUID_FEATURE_DS,    "DS",},
-       {CPUID_FEATURE_ACPI,  "ACPI",},
-       {CPUID_FEATURE_MMX,   "MMX",},
-       {CPUID_FEATURE_FXSR,  "FXSR",},
-       {CPUID_FEATURE_SSE,   "SSE",},
-       {CPUID_FEATURE_SSE2,  "SSE2",},
-       {CPUID_FEATURE_SS,    "SS",},
-       {CPUID_FEATURE_HTT,   "HTT",},
-       {CPUID_FEATURE_TM,    "TM",},
-       {CPUID_FEATURE_SSE3,    "SSE3"},
+       {CPUID_FEATURE_FPU, "FPU"},
+       {CPUID_FEATURE_VME, "VME"},
+       {CPUID_FEATURE_DE, "DE"},
+       {CPUID_FEATURE_PSE, "PSE"},
+       {CPUID_FEATURE_TSC, "TSC"},
+       {CPUID_FEATURE_MSR, "MSR"},
+       {CPUID_FEATURE_PAE, "PAE"},
+       {CPUID_FEATURE_MCE, "MCE"},
+       {CPUID_FEATURE_CX8, "CX8"},
+       {CPUID_FEATURE_APIC, "APIC"},
+       {CPUID_FEATURE_SEP, "SEP"},
+       {CPUID_FEATURE_MTRR, "MTRR"},
+       {CPUID_FEATURE_PGE, "PGE"},
+       {CPUID_FEATURE_MCA, "MCA"},
+       {CPUID_FEATURE_CMOV, "CMOV"},
+       {CPUID_FEATURE_PAT, "PAT"},
+       {CPUID_FEATURE_PSE36, "PSE36"},
+       {CPUID_FEATURE_PSN, "PSN"},
+       {CPUID_FEATURE_CLFSH, "CLFSH"},
+       {CPUID_FEATURE_DS, "DS"},
+       {CPUID_FEATURE_ACPI, "ACPI"},
+       {CPUID_FEATURE_MMX, "MMX"},
+       {CPUID_FEATURE_FXSR, "FXSR"},
+       {CPUID_FEATURE_SSE, "SSE"},
+       {CPUID_FEATURE_SSE2, "SSE2"},
+       {CPUID_FEATURE_SS, "SS"},
+       {CPUID_FEATURE_HTT, "HTT"},
+       {CPUID_FEATURE_TM, "TM"},
+       {CPUID_FEATURE_PBE, "PBE"},
+       {CPUID_FEATURE_SSE3, "SSE3"},
+       {CPUID_FEATURE_PCLMULQDQ, "PCLMULQDQ"},
+       {CPUID_FEATURE_DTES64, "DTES64"},
        {CPUID_FEATURE_MONITOR, "MON"},
-       {CPUID_FEATURE_DSCPL,   "DSCPL"},
-       {CPUID_FEATURE_VMX,     "VMX"},
-       {CPUID_FEATURE_SMX,     "SMX"},
-       {CPUID_FEATURE_EST,     "EST"},
-       {CPUID_FEATURE_TM2,     "TM2"},
-       {CPUID_FEATURE_MNI,     "MNI"},
-       {CPUID_FEATURE_CID,     "CID"},
-       {CPUID_FEATURE_CX16,    "CX16"},
-       {CPUID_FEATURE_xTPR,    "TPR"},
+       {CPUID_FEATURE_DSCPL, "DSCPL"},
+       {CPUID_FEATURE_VMX, "VMX"},
+       {CPUID_FEATURE_SMX, "SMX"},
+       {CPUID_FEATURE_EST, "EST"},
+       {CPUID_FEATURE_TM2, "TM2"},
+       {CPUID_FEATURE_SSSE3, "SSSE3"},
+       {CPUID_FEATURE_CID, "CID"},
+       {CPUID_FEATURE_FMA, "FMA"},
+       {CPUID_FEATURE_CX16, "CX16"},
+       {CPUID_FEATURE_xTPR, "TPR"},
+       {CPUID_FEATURE_PDCM, "PDCM"},
+       {CPUID_FEATURE_SSE4_1, "SSE4.1"},
+       {CPUID_FEATURE_SSE4_2, "SSE4.2"},
+       {CPUID_FEATURE_x2APIC, "x2APIC"},
+       {CPUID_FEATURE_MOVBE, "MOVBE"},
+       {CPUID_FEATURE_POPCNT, "POPCNT"},
+       {CPUID_FEATURE_AES, "AES"},
+       {CPUID_FEATURE_VMM, "VMM"},
+       {CPUID_FEATURE_PCID, "PCID"},
+       {CPUID_FEATURE_XSAVE, "XSAVE"},
+       {CPUID_FEATURE_OSXSAVE, "OSXSAVE"},
+       {CPUID_FEATURE_SEGLIM64, "SEGLIM64"},
+       {CPUID_FEATURE_TSCTMR, "TSCTMR"},
+       {CPUID_FEATURE_AVX1_0, "AVX1.0"},
+       {CPUID_FEATURE_RDRAND, "RDRAND"},
+       {CPUID_FEATURE_F16C, "F16C"},
        {0, 0}
 },
-extfeature_map[] = {
+    extfeature_map[] = {
        {CPUID_EXTFEATURE_SYSCALL, "SYSCALL"},
-       {CPUID_EXTFEATURE_XD,      "XD"},
-       {CPUID_EXTFEATURE_EM64T,   "EM64T"},
-       {CPUID_EXTFEATURE_LAHF,    "LAHF"},
+       {CPUID_EXTFEATURE_XD, "XD"},
+       {CPUID_EXTFEATURE_1GBPAGE, "1GBPAGE"},
+       {CPUID_EXTFEATURE_EM64T, "EM64T"},
+       {CPUID_EXTFEATURE_LAHF, "LAHF"},
+       {CPUID_EXTFEATURE_LZCNT, "LZCNT"},
+       {CPUID_EXTFEATURE_PREFETCHW, "PREFETCHW"},
+       {CPUID_EXTFEATURE_RDTSCP, "RDTSCP"},
+       {CPUID_EXTFEATURE_TSCI, "TSCI"},
+       {0, 0}
+},
+    leaf7_feature_map[] = {
+       {CPUID_LEAF7_FEATURE_RDWRFSGS, "RDWRFSGS"},
+       {CPUID_LEAF7_FEATURE_TSCOFF, "TSC_THREAD_OFFSET"},
+       {CPUID_LEAF7_FEATURE_SGX, "SGX"},
+       {CPUID_LEAF7_FEATURE_BMI1, "BMI1"},
+       {CPUID_LEAF7_FEATURE_HLE, "HLE"},
+       {CPUID_LEAF7_FEATURE_AVX2, "AVX2"},
+       {CPUID_LEAF7_FEATURE_FDPEO, "FDPEO"},
+       {CPUID_LEAF7_FEATURE_SMEP, "SMEP"},
+       {CPUID_LEAF7_FEATURE_BMI2, "BMI2"},
+       {CPUID_LEAF7_FEATURE_ERMS, "ERMS"},
+       {CPUID_LEAF7_FEATURE_INVPCID, "INVPCID"},
+       {CPUID_LEAF7_FEATURE_RTM, "RTM"},
+       {CPUID_LEAF7_FEATURE_PQM, "PQM"},
+       {CPUID_LEAF7_FEATURE_FPU_CSDS, "FPU_CSDS"},
+       {CPUID_LEAF7_FEATURE_MPX, "MPX"},
+       {CPUID_LEAF7_FEATURE_PQE, "PQE"},
+       {CPUID_LEAF7_FEATURE_AVX512F, "AVX512F"},
+       {CPUID_LEAF7_FEATURE_AVX512DQ, "AVX512DQ"},
+       {CPUID_LEAF7_FEATURE_RDSEED, "RDSEED"},
+       {CPUID_LEAF7_FEATURE_ADX, "ADX"},
+       {CPUID_LEAF7_FEATURE_SMAP, "SMAP"},
+       {CPUID_LEAF7_FEATURE_AVX512IFMA, "AVX512IFMA"},
+       {CPUID_LEAF7_FEATURE_CLFSOPT, "CLFSOPT"},
+       {CPUID_LEAF7_FEATURE_CLWB, "CLWB"},
+       {CPUID_LEAF7_FEATURE_IPT, "IPT"},
+       {CPUID_LEAF7_FEATURE_AVX512CD, "AVX512CD"},
+       {CPUID_LEAF7_FEATURE_SHA, "SHA"},
+       {CPUID_LEAF7_FEATURE_AVX512BW, "AVX512BW"},
+       {CPUID_LEAF7_FEATURE_AVX512VL, "AVX512VL"},
+       {CPUID_LEAF7_FEATURE_PREFETCHWT1, "PREFETCHWT1"},
+       {CPUID_LEAF7_FEATURE_AVX512VBMI, "AVX512VBMI"},
+       {CPUID_LEAF7_FEATURE_UMIP, "UMIP"},
+       {CPUID_LEAF7_FEATURE_PKU, "PKU"},
+       {CPUID_LEAF7_FEATURE_OSPKE, "OSPKE"},
+       {CPUID_LEAF7_FEATURE_WAITPKG, "WAITPKG"},
+       {CPUID_LEAF7_FEATURE_GFNI, "GFNI"},
+       {CPUID_LEAF7_FEATURE_VAES, "VAES"},
+       {CPUID_LEAF7_FEATURE_VPCLMULQDQ, "VPCLMULQDQ"},
+       {CPUID_LEAF7_FEATURE_AVX512VNNI, "AVX512VNNI"},
+       {CPUID_LEAF7_FEATURE_AVX512BITALG, "AVX512BITALG"},
+       {CPUID_LEAF7_FEATURE_AVX512VPCDQ, "AVX512VPOPCNTDQ"},
+       {CPUID_LEAF7_FEATURE_RDPID, "RDPID"},
+       {CPUID_LEAF7_FEATURE_CLDEMOTE, "CLDEMOTE"},
+       {CPUID_LEAF7_FEATURE_MOVDIRI, "MOVDIRI"},
+       {CPUID_LEAF7_FEATURE_MOVDIRI64B, "MOVDIRI64B"},
+       {CPUID_LEAF7_FEATURE_SGXLC, "SGXLC"},
+       {0, 0}
+},
+    leaf7_extfeature_map[] = {
+       { CPUID_LEAF7_EXTFEATURE_AVX5124VNNIW, "AVX5124VNNIW" },
+       { CPUID_LEAF7_EXTFEATURE_AVX5124FMAPS, "AVX5124FMAPS" },
+       { CPUID_LEAF7_EXTFEATURE_FSREPMOV, "FSREPMOV" },
+       { CPUID_LEAF7_EXTFEATURE_MDCLEAR, "MDCLEAR" },
+       { CPUID_LEAF7_EXTFEATURE_TSXFA, "TSXFA" },
+       { CPUID_LEAF7_EXTFEATURE_IBRS, "IBRS" },
+       { CPUID_LEAF7_EXTFEATURE_STIBP, "STIBP" },
+       { CPUID_LEAF7_EXTFEATURE_L1DF, "L1DF" },
+       { CPUID_LEAF7_EXTFEATURE_ACAPMSR, "ACAPMSR" },
+       { CPUID_LEAF7_EXTFEATURE_CCAPMSR, "CCAPMSR" },
+       { CPUID_LEAF7_EXTFEATURE_SSBD, "SSBD" },
        {0, 0}
 };
 
-char *
-cpuid_get_feature_names(uint64_t features, char *buf, unsigned buf_len)
+static char *
+cpuid_get_names(struct table *map, uint64_t bits, char *buf, unsigned buf_len)
 {
-       int     len = -1;
-       char    *p = buf;
-       int     i;
+       size_t  len = 0;
+       char    *p = buf;
+       int     i;
 
-       for (i = 0; feature_map[i].mask != 0; i++) {
-               if ((features & feature_map[i].mask) == 0)
+       for (i = 0; map[i].mask != 0; i++) {
+               if ((bits & map[i].mask) == 0) {
                        continue;
-               if (len > 0)
+               }
+               if (len && ((size_t) (p - buf) < (buf_len - 1))) {
                        *p++ = ' ';
-               len = min(strlen(feature_map[i].name), (buf_len-1) - (p-buf));
-               if (len == 0)
+               }
+               len = min(strlen(map[i].name), (size_t)((buf_len - 1) - (p - buf)));
+               if (len == 0) {
                        break;
-               bcopy(feature_map[i].name, p, len);
+               }
+               bcopy(map[i].name, p, len);
                p += len;
        }
        *p = '\0';
        return buf;
 }
 
+i386_cpu_info_t *
+cpuid_info(void)
+{
+       /* Set-up the cpuid_info stucture lazily */
+       if (cpuid_cpu_infop == NULL) {
+               PE_parse_boot_argn("-cpuid", &cpuid_dbg, sizeof(cpuid_dbg));
+               cpuid_set_info();
+               cpuid_cpu_infop = &cpuid_cpu_info;
+       }
+       return cpuid_cpu_infop;
+}
+
+char *
+cpuid_get_feature_names(uint64_t features, char *buf, unsigned buf_len)
+{
+       return cpuid_get_names(feature_map, features, buf, buf_len);
+}
+
 char *
 cpuid_get_extfeature_names(uint64_t extfeatures, char *buf, unsigned buf_len)
 {
-       int     len = -1;
-       char    *p = buf;
-       int     i;
+       return cpuid_get_names(extfeature_map, extfeatures, buf, buf_len);
+}
 
-       for (i = 0; extfeature_map[i].mask != 0; i++) {
-               if ((extfeatures & extfeature_map[i].mask) == 0)
-                       continue;
-               if (len > 0)
-                       *p++ = ' ';
-               len = min(strlen(extfeature_map[i].name), (buf_len-1)-(p-buf));
-               if (len == 0)
-                       break;
-               bcopy(extfeature_map[i].name, p, len);
-               p += len;
-       }
-       *p = '\0';
-       return buf;
+char *
+cpuid_get_leaf7_feature_names(uint64_t features, char *buf, unsigned buf_len)
+{
+       return cpuid_get_names(leaf7_feature_map, features, buf, buf_len);
+}
+
+char *
+cpuid_get_leaf7_extfeature_names(uint64_t features, char *buf, unsigned buf_len)
+{
+       return cpuid_get_names(leaf7_extfeature_map, features, buf, buf_len);
 }
 
 void
 cpuid_feature_display(
-       const char      *header)
+       const char      *header)
 {
-       char    buf[256];
+       char    buf[320];
 
-       kprintf("%s: %s\n", header,
-                 cpuid_get_feature_names(cpuid_features(),
-                                               buf, sizeof(buf)));
+       kprintf("%s: %s", header,
+           cpuid_get_feature_names(cpuid_features(), buf, sizeof(buf)));
+       if (cpuid_leaf7_features()) {
+               kprintf(" %s", cpuid_get_leaf7_feature_names(
+                           cpuid_leaf7_features(), buf, sizeof(buf)));
+       }
+       if (cpuid_leaf7_extfeatures()) {
+               kprintf(" %s", cpuid_get_leaf7_extfeature_names(
+                           cpuid_leaf7_extfeatures(), buf, sizeof(buf)));
+       }
+       kprintf("\n");
        if (cpuid_features() & CPUID_FEATURE_HTT) {
-#define s_if_plural(n) ((n > 1) ? "s" : "")
+#define s_if_plural(n)  ((n > 1) ? "s" : "")
                kprintf("  HTT: %d core%s per package;"
-                            " %d logical cpu%s per package\n",
-                       cpuid_cpu_info.cpuid_cores_per_package,
-                       s_if_plural(cpuid_cpu_info.cpuid_cores_per_package),
-                       cpuid_cpu_info.cpuid_logical_per_package,
-                       s_if_plural(cpuid_cpu_info.cpuid_logical_per_package));
+                   " %d logical cpu%s per package\n",
+                   cpuid_cpu_infop->cpuid_cores_per_package,
+                   s_if_plural(cpuid_cpu_infop->cpuid_cores_per_package),
+                   cpuid_cpu_infop->cpuid_logical_per_package,
+                   s_if_plural(cpuid_cpu_infop->cpuid_logical_per_package));
        }
 }
 
 void
 cpuid_extfeature_display(
-       const char      *header)
+       const char      *header)
 {
-       char    buf[256];
+       char    buf[256];
 
        kprintf("%s: %s\n", header,
-                 cpuid_get_extfeature_names(cpuid_extfeatures(),
-                                               buf, sizeof(buf)));
+           cpuid_get_extfeature_names(cpuid_extfeatures(),
+           buf, sizeof(buf)));
 }
 
 void
 cpuid_cpu_display(
-       const char      *header)
+       const char      *header)
 {
-    if (cpuid_cpu_info.cpuid_brand_string[0] != '\0') {
-       kprintf("%s: %s\n", header, cpuid_cpu_info.cpuid_brand_string);
-    }
+       if (cpuid_cpu_infop->cpuid_brand_string[0] != '\0') {
+               kprintf("%s: %s\n", header, cpuid_cpu_infop->cpuid_brand_string);
+       }
 }
 
 unsigned int
 cpuid_family(void)
 {
-       return cpuid_cpu_info.cpuid_family;
+       return cpuid_info()->cpuid_family;
+}
+
+uint32_t
+cpuid_cpufamily(void)
+{
+       return cpuid_info()->cpuid_cpufamily;
 }
 
 cpu_type_t
 cpuid_cputype(void)
 {
-       return cpuid_cpu_info.cpuid_cpu_type;
+       return cpuid_info()->cpuid_cpu_type;
 }
 
 cpu_subtype_t
 cpuid_cpusubtype(void)
 {
-       return cpuid_cpu_info.cpuid_cpu_subtype;
+       return cpuid_info()->cpuid_cpu_subtype;
 }
 
 uint64_t
 cpuid_features(void)
 {
        static int checked = 0;
-       char    fpu_arg[16] = { 0 };
+       char    fpu_arg[20] = { 0 };
+
+       (void) cpuid_info();
        if (!checked) {
-                   /* check for boot-time fpu limitations */
-                       if (PE_parse_boot_arg("_fpu", &fpu_arg[0])) {
-                               printf("limiting fpu features to: %s\n", fpu_arg);
-                               if (!strncmp("387", fpu_arg, sizeof "387") || !strncmp("mmx", fpu_arg, sizeof "mmx")) {
-                                       printf("no sse or sse2\n");
-                                       cpuid_cpu_info.cpuid_features &= ~(CPUID_FEATURE_SSE | CPUID_FEATURE_SSE2 | CPUID_FEATURE_FXSR);
-                               } else if (!strncmp("sse", fpu_arg, sizeof "sse")) {
-                                       printf("no sse2\n");
-                                       cpuid_cpu_info.cpuid_features &= ~(CPUID_FEATURE_SSE2);
-                               }
+               /* check for boot-time fpu limitations */
+               if (PE_parse_boot_argn("_fpu", &fpu_arg[0], sizeof(fpu_arg))) {
+                       printf("limiting fpu features to: %s\n", fpu_arg);
+                       if (!strncmp("387", fpu_arg, sizeof("387")) || !strncmp("mmx", fpu_arg, sizeof("mmx"))) {
+                               printf("no sse or sse2\n");
+                               cpuid_cpu_infop->cpuid_features &= ~(CPUID_FEATURE_SSE | CPUID_FEATURE_SSE2 | CPUID_FEATURE_FXSR);
+                       } else if (!strncmp("sse", fpu_arg, sizeof("sse"))) {
+                               printf("no sse2\n");
+                               cpuid_cpu_infop->cpuid_features &= ~(CPUID_FEATURE_SSE2);
                        }
-                       checked = 1;
+               }
+               checked = 1;
        }
-       return cpuid_cpu_info.cpuid_features;
+       return cpuid_cpu_infop->cpuid_features;
 }
 
 uint64_t
 cpuid_extfeatures(void)
 {
-       return cpuid_cpu_info.cpuid_extfeatures;
+       return cpuid_info()->cpuid_extfeatures;
 }
-i386_cpu_info_t        *
-cpuid_info(void)
+
+uint64_t
+cpuid_leaf7_features(void)
 {
-       return &cpuid_cpu_info;
+       return cpuid_info()->cpuid_leaf7_features;
 }
 
-void
-cpuid_set_info(void)
+uint64_t
+cpuid_leaf7_extfeatures(void)
 {
-       cpuid_get_info(&cpuid_cpu_info);
+       return cpuid_info()->cpuid_leaf7_extfeatures;
 }
 
-#if MACH_KDB
+const char *
+cpuid_vmm_family_string(void)
+{
+       switch (cpuid_vmm_info()->cpuid_vmm_family) {
+       case CPUID_VMM_FAMILY_NONE:
+               return "None";
+
+       case CPUID_VMM_FAMILY_VMWARE:
+               return "VMWare";
 
-/*
- *     Display the cpuid
- * *           
- *     cp
- */
-void 
-db_cpuid(__unused db_expr_t addr,
-        __unused int have_addr,
-        __unused db_expr_t count,
-        __unused char *modif)
+       case CPUID_VMM_FAMILY_PARALLELS:
+               return "Parallels";
+
+       case CPUID_VMM_FAMILY_HYVE:
+               return "xHyve";
+
+       case CPUID_VMM_FAMILY_HVF:
+               return "HVF";
+
+       case CPUID_VMM_FAMILY_KVM:
+               return "KVM";
+
+       case CPUID_VMM_FAMILY_UNKNOWN:
+       /*FALLTHROUGH*/
+       default:
+               return "Unknown VMM";
+       }
+}
+
+static i386_vmm_info_t  *_cpuid_vmm_infop = NULL;
+static i386_vmm_info_t  _cpuid_vmm_info;
+
+static void
+cpuid_init_vmm_info(i386_vmm_info_t *info_p)
 {
+       uint32_t        reg[4], maxbasic_regs[4];
+       uint32_t        max_vmm_leaf;
 
-       uint32_t        i, mid;
-       uint32_t        cpid[4];
+       bzero(info_p, sizeof(*info_p));
 
-       do_cpuid(0, cpid);      /* Get the first cpuid which is the number of
-                                * basic ids */
-       db_printf("%08X - %08X %08X %08X %08X\n",
-               0, cpid[eax], cpid[ebx], cpid[ecx], cpid[edx]);
+       if (!cpuid_vmm_present()) {
+               return;
+       }
+
+       DBG("cpuid_init_vmm_info(%p)\n", info_p);
 
-       mid = cpid[eax];        /* Set the number */
-       for (i = 1; i <= mid; i++) {    /* Dump 'em out */
-               do_cpuid(i, cpid);      /* Get the next */
-               db_printf("%08X - %08X %08X %08X %08X\n",
-                       i, cpid[eax], cpid[ebx], cpid[ecx], cpid[edx]);
+       /*
+        * Get the highest basic leaf value, then save the cpuid details for that leaf
+        * for comparison with the [ostensible] VMM leaf.
+        */
+       cpuid_fn(0, reg);
+       cpuid_fn(reg[eax], maxbasic_regs);
+
+       /* do cpuid 0x40000000 to get VMM vendor */
+       cpuid_fn(0x40000000, reg);
+
+       /*
+        * If leaf 0x40000000 is non-existent, cpuid will return the values as
+        * if the highest basic leaf was requested, so compare to those values
+        * we just retrieved to see if no vmm is present.
+        */
+       if (bcmp(reg, maxbasic_regs, sizeof(reg)) == 0) {
+               info_p->cpuid_vmm_family = CPUID_VMM_FAMILY_NONE;
+               DBG(" vmm_vendor          : NONE\n");
+               return;
        }
-       db_printf("\n");
 
-       do_cpuid(0x80000000, cpid);     /* Get the first extended cpuid which
-                                        * is the number of extended ids */
-       db_printf("%08X - %08X %08X %08X %08X\n",
-               0x80000000, cpid[eax], cpid[ebx], cpid[ecx], cpid[edx]);
+       max_vmm_leaf = reg[eax];
+       bcopy((char *)&reg[ebx], &info_p->cpuid_vmm_vendor[0], 4);
+       bcopy((char *)&reg[ecx], &info_p->cpuid_vmm_vendor[4], 4);
+       bcopy((char *)&reg[edx], &info_p->cpuid_vmm_vendor[8], 4);
+       info_p->cpuid_vmm_vendor[12] = '\0';
+
+       if (0 == strcmp(info_p->cpuid_vmm_vendor, CPUID_VMM_ID_VMWARE)) {
+               /* VMware identification string: kb.vmware.com/kb/1009458 */
+               info_p->cpuid_vmm_family = CPUID_VMM_FAMILY_VMWARE;
+       } else if (0 == bcmp(info_p->cpuid_vmm_vendor, CPUID_VMM_ID_PARALLELS, 12)) {
+               /* Parallels identification string */
+               info_p->cpuid_vmm_family = CPUID_VMM_FAMILY_PARALLELS;
+       } else if (0 == bcmp(info_p->cpuid_vmm_vendor, CPUID_VMM_ID_HYVE, 12)) {
+               /* bhyve/xhyve identification string */
+               info_p->cpuid_vmm_family = CPUID_VMM_FAMILY_HYVE;
+       } else if (0 == bcmp(info_p->cpuid_vmm_vendor, CPUID_VMM_ID_HVF, 12)) {
+               /* HVF identification string */
+               info_p->cpuid_vmm_family = CPUID_VMM_FAMILY_HVF;
+       } else if (0 == bcmp(info_p->cpuid_vmm_vendor, CPUID_VMM_ID_KVM, 12)) {
+               /* KVM identification string */
+               info_p->cpuid_vmm_family = CPUID_VMM_FAMILY_KVM;
+       } else {
+               info_p->cpuid_vmm_family = CPUID_VMM_FAMILY_UNKNOWN;
+       }
+
+       /* VMM generic leaves: https://lkml.org/lkml/2008/10/1/246 */
+       if (max_vmm_leaf >= 0x40000010) {
+               cpuid_fn(0x40000010, reg);
 
-       mid = cpid[eax];        /* Set the number */
-       for (i = 0x80000001; i <= mid; i++) {   /* Dump 'em out */
-               do_cpuid(i, cpid);      /* Get the next */
-               db_printf("%08X - %08X %08X %08X %08X\n",
-                       i, cpid[eax], cpid[ebx], cpid[ecx], cpid[edx]);
+               info_p->cpuid_vmm_tsc_frequency = reg[eax];
+               info_p->cpuid_vmm_bus_frequency = reg[ebx];
        }
-}
 
+#if DEBUG || DEVELOPMENT
+       cpuid_vmm_detect_pv_interface(info_p, APPLEPV_SIGNATURE, &cpuid_vmm_detect_applepv_features);
 #endif
+
+       DBG(" vmm_vendor          : %s\n", info_p->cpuid_vmm_vendor);
+       DBG(" vmm_family          : %u\n", info_p->cpuid_vmm_family);
+       DBG(" vmm_bus_frequency   : %u\n", info_p->cpuid_vmm_bus_frequency);
+       DBG(" vmm_tsc_frequency   : %u\n", info_p->cpuid_vmm_tsc_frequency);
+}
+
+boolean_t
+cpuid_vmm_present(void)
+{
+       return (cpuid_features() & CPUID_FEATURE_VMM) ? TRUE : FALSE;
+}
+
+i386_vmm_info_t *
+cpuid_vmm_info(void)
+{
+       if (_cpuid_vmm_infop == NULL) {
+               cpuid_init_vmm_info(&_cpuid_vmm_info);
+               _cpuid_vmm_infop = &_cpuid_vmm_info;
+       }
+       return _cpuid_vmm_infop;
+}
+
+uint32_t
+cpuid_vmm_family(void)
+{
+       return cpuid_vmm_info()->cpuid_vmm_family;
+}
+
+#if DEBUG || DEVELOPMENT
+uint64_t
+cpuid_vmm_get_applepv_features(void)
+{
+       return cpuid_vmm_info()->cpuid_vmm_applepv_features;
+}
+#endif /* DEBUG || DEVELOPMENT */
+
+cwa_classifier_e
+cpuid_wa_required(cpu_wa_e wa)
+{
+       i386_cpu_info_t *info_p = &cpuid_cpu_info;
+       static uint64_t bootarg_cpu_wa_enables = 0;
+       static uint64_t bootarg_cpu_wa_disables = 0;
+       static int bootargs_overrides_processed = 0;
+       uint32_t        reg[4];
+
+       if (!bootargs_overrides_processed) {
+               if (!PE_parse_boot_argn("cwae", &bootarg_cpu_wa_enables, sizeof(bootarg_cpu_wa_enables))) {
+                       bootarg_cpu_wa_enables = 0;
+               }
+
+               if (!PE_parse_boot_argn("cwad", &bootarg_cpu_wa_disables, sizeof(bootarg_cpu_wa_disables))) {
+                       bootarg_cpu_wa_disables = 0;
+               }
+               bootargs_overrides_processed = 1;
+       }
+
+       if (bootarg_cpu_wa_enables & (1 << wa)) {
+               return CWA_FORCE_ON;
+       }
+
+       if (bootarg_cpu_wa_disables & (1 << wa)) {
+               return CWA_FORCE_OFF;
+       }
+
+       switch (wa) {
+       case CPU_INTEL_SEGCHK:
+               /* First, check to see if this CPU requires the workaround */
+               if ((info_p->cpuid_leaf7_extfeatures & CPUID_LEAF7_EXTFEATURE_ACAPMSR) != 0) {
+                       /* We have ARCHCAP, so check it for either RDCL_NO or MDS_NO */
+                       uint64_t archcap_msr = rdmsr64(MSR_IA32_ARCH_CAPABILITIES);
+                       if ((archcap_msr & (MSR_IA32_ARCH_CAPABILITIES_RDCL_NO | MSR_IA32_ARCH_CAPABILITIES_MDS_NO)) != 0) {
+                               /* Workaround not needed */
+                               return CWA_OFF;
+                       }
+               }
+
+               if ((info_p->cpuid_leaf7_extfeatures & CPUID_LEAF7_EXTFEATURE_MDCLEAR) != 0) {
+                       return CWA_ON;
+               }
+
+               /*
+                * If the CPU supports the ARCHCAP MSR and neither the RDCL_NO bit nor the MDS_NO
+                * bit are set, OR the CPU does not support the ARCHCAP MSR and the CPU does
+                * not enumerate the presence of the enhanced VERW instruction, report
+                * that the workaround should not be enabled.
+                */
+               break;
+
+       case CPU_INTEL_TSXFA:
+               /*
+                * Note that if TSX was disabled in cpuid_do_precpuid_was(), the cached cpuid
+                * info will indicate that RTM is *not* supported and this workaround will not
+                * be enabled.
+                */
+               /*
+                * Otherwise, if the CPU supports both TSX(HLE) and FORCE_ABORT, return that
+                * the workaround should be enabled.
+                */
+               if ((info_p->cpuid_leaf7_extfeatures & CPUID_LEAF7_EXTFEATURE_TSXFA) != 0 &&
+                   (info_p->cpuid_leaf7_features & CPUID_LEAF7_FEATURE_RTM) != 0) {
+                       return CWA_ON;
+               }
+               break;
+
+       case CPU_INTEL_TSXDA:
+               /*
+                * Since this workaround might be requested before cpuid_set_info() is complete,
+                * we need to invoke cpuid directly when looking for the required bits.
+                */
+               cpuid_fn(0x7, reg);
+               if (reg[edx] & CPUID_LEAF7_EXTFEATURE_ACAPMSR) {
+                       uint64_t archcap_msr = rdmsr64(MSR_IA32_ARCH_CAPABILITIES);
+                       /*
+                        * If this CPU supports TSX (HLE being the proxy for TSX detection) AND it does
+                        * not include a hardware fix for TAA and it supports the TSX_CTRL MSR, disable TSX entirely.
+                        * (Note this can be overridden (above) if the cwad boot-arg's value has bit 2 set.)
+                        */
+                       if ((reg[ebx] & CPUID_LEAF7_FEATURE_HLE) != 0 &&
+                           (archcap_msr & (MSR_IA32_ARCH_CAPABILITIES_TAA_NO | MSR_IA32_ARCH_CAPABILITIES_TSX_CTRL))
+                           == MSR_IA32_ARCH_CAPABILITIES_TSX_CTRL) {
+                               return CWA_ON;
+                       }
+               }
+               break;
+
+       case CPU_INTEL_SRBDS:
+               /*
+                * SRBDS mitigations are enabled by default.  CWA_ON returned here indicates
+                * the caller should disable the mitigation.  Mitigations should be disabled
+                * at least for CPUs that advertise MDS_NO *and* (either TAA_NO is set OR TSX
+                * has been disabled).
+                */
+               if ((info_p->cpuid_leaf7_extfeatures & CPUID_LEAF7_EXTFEATURE_SRBDS_CTRL) != 0) {
+                       if ((info_p->cpuid_leaf7_extfeatures & CPUID_LEAF7_EXTFEATURE_ACAPMSR) != 0) {
+                               uint64_t archcap_msr = rdmsr64(MSR_IA32_ARCH_CAPABILITIES);
+                               if ((archcap_msr & MSR_IA32_ARCH_CAPABILITIES_MDS_NO) != 0 &&
+                                   ((archcap_msr & MSR_IA32_ARCH_CAPABILITIES_TAA_NO) != 0 ||
+                                   cpuid_tsx_disabled)) {
+                                       return CWA_ON;
+                               }
+                       }
+               }
+               break;
+
+       default:
+               break;
+       }
+
+       return CWA_OFF;
+}
+
+static void
+cpuid_do_precpuid_was(void)
+{
+       /*
+        * Note that care must be taken not to use any data from the cached cpuid data since it is
+        * likely uninitialized at this point.  That includes calling functions that make use of
+        * that data as well.
+        */
+
+       /* Note the TSX disablement, we do not support force-on since it depends on MSRs being present */
+       if (cpuid_wa_required(CPU_INTEL_TSXDA) == CWA_ON) {
+               /* This must be executed on all logical processors */
+               wrmsr64(MSR_IA32_TSX_CTRL, MSR_IA32_TSXCTRL_TSX_CPU_CLEAR | MSR_IA32_TSXCTRL_RTM_DISABLE);
+               cpuid_tsx_disabled = true;
+       }
+}
+
+
+#if DEBUG || DEVELOPMENT
+
+/*
+ * Hunt for Apple Paravirtualization support in the hypervisor class leaves [0x4000_0000-0x4001_0000].
+ * Hypervisor interfaces are expected to be found at 0x100 boundaries for compatibility.
+ */
+
+static bool
+cpuid_vmm_detect_applepv_features(i386_vmm_info_t *info_p, const uint32_t base, const uint32_t max_leaf)
+{
+       if ((max_leaf - base) < APPLEPV_LEAF_INDEX_MAX) {
+               return false;
+       }
+
+       /*
+        * Issue cpuid to make sure the interface supports "AH#1" features.
+        * This avoids a possible collision with "Hv#1" used by Hyper-V.
+        */
+       uint32_t reg[4];
+       char interface[5];
+       cpuid_fn(base + APPLEPV_INTERFACE_LEAF_INDEX, reg);
+       memcpy(&interface[0], &reg[eax], 4);
+       interface[4] = '\0';
+       if (0 == strcmp(interface, APPLEPV_INTERFACE)) {
+               cpuid_fn(base + APPLEPV_FEATURES_LEAF_INDEX, reg);
+               info_p->cpuid_vmm_applepv_features = quad(reg[ecx], reg[edx]);
+               return true;
+       }
+       return false;
+}
+
+static void
+cpuid_vmm_detect_pv_interface(i386_vmm_info_t *info_p, const char *signature,
+    bool (*searcher)(i386_vmm_info_t*, const uint32_t, const uint32_t))
+{
+       int hcalls;
+       if (PE_parse_boot_argn("hcalls", &hcalls, sizeof(hcalls)) &&
+           hcalls == 0) {
+               return;
+       }
+
+       assert(info_p);
+       /*
+        * Look for PV interface matching signature
+        */
+       for (uint32_t base = 0x40000100; base < 0x40010000; base += 0x100) {
+               uint32_t reg[4];
+               char vendor[13];
+
+               cpuid_fn(base, reg);
+               memcpy(&vendor[0], &reg[ebx], 4);
+               memcpy(&vendor[4], &reg[ecx], 4);
+               memcpy(&vendor[8], &reg[edx], 4);
+               vendor[12] = '\0';
+               if ((0 == strcmp(vendor, signature)) &&
+                   (reg[eax] - base) < 0x100 &&
+                   (*searcher)(info_p, base, reg[eax])) {
+                       break;
+               }
+       }
+}
+
+#endif /* DEBUG || DEVELOPMENT */