]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/i386/i386_init.c
xnu-4903.221.2.tar.gz
[apple/xnu.git] / osfmk / i386 / i386_init.c
index 887624b23c4b73d2177113190d910d111fea419b..8eb6b7edf6bfbac5244410f60c09888bf1a52b08 100644 (file)
@@ -1,31 +1,29 @@
 /*
 /*
- * Copyright (c) 2003-2005 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2003-2016 Apple Inc. All rights reserved.
  *
  *
- * @APPLE_LICENSE_OSREFERENCE_HEADER_START@
+ * @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 
- * compliance with the License.  The rights granted to you under the 
- * License may not be used to create, or enable the creation or 
- * redistribution of, 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, 
- * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, 
- * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. 
- * Please see the License for the specific language governing rights and 
+ * 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
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * 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,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
+ * 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.
  * limitations under the License.
- *
- * @APPLE_LICENSE_OSREFERENCE_HEADER_END@
+ * 
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * @OSF_COPYRIGHT@
  */
 /*
  * @OSF_COPYRIGHT@
@@ -56,8 +54,6 @@
  * the rights to redistribute these changes.
  */
 
  * the rights to redistribute these changes.
  */
 
-#include <platforms.h>
-#include <mach_kdb.h>
 
 #include <mach/i386/vm_param.h>
 
 
 #include <mach/i386/vm_param.h>
 
 #include <kern/xpr.h>
 #include <kern/cpu_data.h>
 #include <kern/processor.h>
 #include <kern/xpr.h>
 #include <kern/cpu_data.h>
 #include <kern/processor.h>
+#include <sys/kdebug.h>
 #include <console/serial_protos.h>
 #include <vm/vm_page.h>
 #include <vm/pmap.h>
 #include <vm/vm_kern.h>
 #include <console/serial_protos.h>
 #include <vm/vm_page.h>
 #include <vm/pmap.h>
 #include <vm/vm_kern.h>
+#include <machine/pal_routines.h>
 #include <i386/fpu.h>
 #include <i386/pmap.h>
 #include <i386/fpu.h>
 #include <i386/pmap.h>
-#include <i386/ipl.h>
 #include <i386/misc_protos.h>
 #include <i386/misc_protos.h>
+#include <i386/cpu_threads.h>
 #include <i386/cpuid.h>
 #include <i386/cpuid.h>
+#include <i386/lapic.h>
 #include <i386/mp.h>
 #include <i386/mp_desc.h>
 #include <i386/mp.h>
 #include <i386/mp_desc.h>
+#if CONFIG_MTRR
+#include <i386/mtrr.h>
+#endif
 #include <i386/machine_routines.h>
 #include <i386/machine_routines.h>
+#if CONFIG_MCA
+#include <i386/machine_check.h>
+#endif
+#include <i386/ucode.h>
 #include <i386/postcode.h>
 #include <i386/Diagnostics.h>
 #include <i386/pmCPU.h>
 #include <i386/tsc.h>
 #include <i386/postcode.h>
 #include <i386/Diagnostics.h>
 #include <i386/pmCPU.h>
 #include <i386/tsc.h>
-#include <i386/hpet.h>
-#if    MACH_KDB
-#include <ddb/db_aout.h>
-#endif /* MACH_KDB */
-#include <ddb/tr.h>
+#include <i386/locks.h> /* LcksOpts */
+#if DEBUG
+#include <machine/pal_routines.h>
+#endif
+
+#if MONOTONIC
+#include <kern/monotonic.h>
+#endif /* MONOTONIC */
+
+#include <san/kasan.h>
+
+#if DEBUG
+#define DBG(x...)       kprintf(x)
+#else
+#define DBG(x...)
+#endif
 
 
-static boot_args *kernelBootArgs;
+int                    debug_task;
 
 
-extern int disableConsoleOutput;
-extern const char version[];
-extern const char version_variant[];
-extern int nx_enabled;
+static boot_args       *kernelBootArgs;
 
 
-extern int noVMX;      /* if set, rosetta should not emulate altivec */
+extern int             disableConsoleOutput;
+extern const char      version[];
+extern const char      version_variant[];
+extern int             nx_enabled;
 
 
-void cpu_stack_set(void);
+/*
+ * Set initial values so that ml_phys_* routines can use the booter's ID mapping
+ * to touch physical space before the kernel's physical aperture exists.
+ */
+uint64_t               physmap_base = 0;
+uint64_t               physmap_max = 4*GB;
+
+pd_entry_t             *KPTphys;
+pd_entry_t             *IdlePTD;
+pdpt_entry_t           *IdlePDPT;
+pml4_entry_t           *IdlePML4;
 
 
+char *physfree;
+void idt64_remap(void);
+
+/*
+ * Note: ALLOCPAGES() can only be used safely within Idle_PTs_init()
+ * due to the mutation of physfree.
+ */
+static void *
+ALLOCPAGES(int npages)
+{
+       uintptr_t tmp = (uintptr_t)physfree;
+       bzero(physfree, npages * PAGE_SIZE);
+       physfree += npages * PAGE_SIZE;
+       tmp += VM_MIN_KERNEL_ADDRESS & ~LOW_4GB_MASK;
+       return (void *)tmp;
+}
+
+static void
+fillkpt(pt_entry_t *base, int prot, uintptr_t src, int index, int count)
+{
+       int i;
+       for (i=0; i<count; i++) {
+               base[index] = src | prot | INTEL_PTE_VALID;
+               src += PAGE_SIZE;
+               index++;
+       }
+}
+
+extern pmap_paddr_t first_avail;
+
+int break_kprintf = 0;
+
+uint64_t
+x86_64_pre_sleep(void)
+{
+       IdlePML4[0] = IdlePML4[KERNEL_PML4_INDEX];
+       uint64_t oldcr3 = get_cr3_raw();
+       set_cr3_raw((uint32_t) (uintptr_t)ID_MAP_VTOP(IdlePML4));
+       return oldcr3;
+}
+
+void
+x86_64_post_sleep(uint64_t new_cr3)
+{
+       IdlePML4[0] = 0;
+       set_cr3_raw((uint32_t) new_cr3);
+}
+
+
+
+
+// Set up the physical mapping - NPHYSMAP GB of memory mapped at a high address
+// NPHYSMAP is determined by the maximum supported RAM size plus 4GB to account
+// the PCI hole (which is less 4GB but not more).
+
+/* Compile-time guard: NPHYSMAP is capped to 256GiB, accounting for
+ * randomisation
+ */
+extern int maxphymapsupported[NPHYSMAP <= (PTE_PER_PAGE/2) ? 1 : -1];
+
+static void
+physmap_init(void)
+{
+       pt_entry_t *physmapL3 = ALLOCPAGES(1);
+       struct {
+               pt_entry_t entries[PTE_PER_PAGE];
+       } * physmapL2 = ALLOCPAGES(NPHYSMAP);
+
+       uint64_t i;
+       uint8_t phys_random_L3 = early_random() & 0xFF;
+
+       /* We assume NX support. Mark all levels of the PHYSMAP NX
+        * to avoid granting executability via a single bit flip.
+        */
+#if DEVELOPMENT || DEBUG
+       uint32_t reg[4];
+       do_cpuid(0x80000000, reg);
+       if (reg[eax] >= 0x80000001) {
+               do_cpuid(0x80000001, reg);
+               assert(reg[edx] & CPUID_EXTFEATURE_XD);
+       }
+#endif /* DEVELOPMENT || DEBUG */
+
+       for(i = 0; i < NPHYSMAP; i++) {
+               physmapL3[i + phys_random_L3] =
+                               ((uintptr_t)ID_MAP_VTOP(&physmapL2[i]))
+                               | INTEL_PTE_VALID
+                               | INTEL_PTE_NX
+                               | INTEL_PTE_WRITE;
+
+               uint64_t j;
+               for(j = 0; j < PTE_PER_PAGE; j++) {
+                       physmapL2[i].entries[j] =
+                           ((i * PTE_PER_PAGE + j) << PDSHIFT)
+                                                       | INTEL_PTE_PS
+                                                       | INTEL_PTE_VALID
+                                                       | INTEL_PTE_NX
+                                                       | INTEL_PTE_WRITE;
+               }
+       }
+
+       IdlePML4[KERNEL_PHYSMAP_PML4_INDEX] =
+                                       ((uintptr_t)ID_MAP_VTOP(physmapL3))
+                                       | INTEL_PTE_VALID
+                                       | INTEL_PTE_NX
+                                       | INTEL_PTE_WRITE;
+
+       physmap_base = KVADDR(KERNEL_PHYSMAP_PML4_INDEX, phys_random_L3, 0, 0);
+       physmap_max = physmap_base + NPHYSMAP * GB;
+       DBG("Physical address map base: 0x%qx\n", physmap_base);
+       DBG("Physical map idlepml4[%d]: 0x%llx\n",
+               KERNEL_PHYSMAP_PML4_INDEX, IdlePML4[KERNEL_PHYSMAP_PML4_INDEX]);
+}
+
+void doublemap_init(void);
+
+static void
+Idle_PTs_init(void)
+{
+       /* Allocate the "idle" kernel page tables: */
+       KPTphys  = ALLOCPAGES(NKPT);            /* level 1 */
+       IdlePTD  = ALLOCPAGES(NPGPTD);          /* level 2 */
+       IdlePDPT = ALLOCPAGES(1);               /* level 3 */
+       IdlePML4 = ALLOCPAGES(1);               /* level 4 */
+
+       // Fill the lowest level with everything up to physfree
+       fillkpt(KPTphys,
+               INTEL_PTE_WRITE, 0, 0, (int)(((uintptr_t)physfree) >> PAGE_SHIFT));
+
+       /* IdlePTD */
+       fillkpt(IdlePTD,
+               INTEL_PTE_WRITE, (uintptr_t)ID_MAP_VTOP(KPTphys), 0, NKPT);
+
+       // IdlePDPT entries
+       fillkpt(IdlePDPT,
+               INTEL_PTE_WRITE, (uintptr_t)ID_MAP_VTOP(IdlePTD), 0, NPGPTD);
+
+       // IdlePML4 single entry for kernel space.
+       fillkpt(IdlePML4 + KERNEL_PML4_INDEX,
+               INTEL_PTE_WRITE, (uintptr_t)ID_MAP_VTOP(IdlePDPT), 0, 1);
+       
+       postcode(VSTART_PHYSMAP_INIT);
+
+       physmap_init();
+       doublemap_init();
+       idt64_remap();
+
+       postcode(VSTART_SET_CR3);
+
+       // Switch to the page tables..
+       set_cr3_raw((uintptr_t)ID_MAP_VTOP(IdlePML4));
+
+}
+
+extern void vstart_trap_handler;
+
+#define BOOT_TRAP_VECTOR(t)                            \
+       [t] = {                                         \
+               (uintptr_t) &vstart_trap_handler,       \
+               KERNEL64_CS,                            \
+               0,                                      \
+               ACC_P|ACC_PL_K|ACC_INTR_GATE,           \
+               0                                       \
+       },
+
+/* Recursive macro to iterate 0..31 */
+#define L0(x,n)         x(n)
+#define L1(x,n)         L0(x,n-1)     L0(x,n)
+#define L2(x,n)  L1(x,n-2)     L1(x,n)
+#define L3(x,n)  L2(x,n-4)     L2(x,n)
+#define L4(x,n)  L3(x,n-8)     L3(x,n)
+#define L5(x,n)  L4(x,n-16)    L4(x,n)
+#define FOR_0_TO_31(x) L5(x,31)
+
+/*
+ * Bootstrap IDT. Active only during early startup.
+ * Only the trap vectors are defined since interrupts are masked.
+ * All traps point to a common handler.
+ */
+struct fake_descriptor64 master_boot_idt64[IDTSZ]
+       __attribute__((section("__HIB,__desc")))
+       __attribute__((aligned(PAGE_SIZE))) = {
+       FOR_0_TO_31(BOOT_TRAP_VECTOR)
+};
+
+static void
+vstart_idt_init(void)
+{
+       x86_64_desc_register_t  vstart_idt = {
+                                       sizeof(master_boot_idt64),
+                                       master_boot_idt64 };
+       
+       fix_desc64(master_boot_idt64, 32);
+       lidt((void *)&vstart_idt);
+}
+
+/*
+ * vstart() is called in the natural mode (64bit for K64, 32 for K32)
+ * on a set of bootstrap pagetables which use large, 2MB pages to map 
+ * all of physical memory in both. See idle_pt.c for details.
+ *
+ * In K64 this identity mapping is mirrored the top and bottom 512GB 
+ * slots of PML4.
+ *
+ * The bootstrap processor called with argument boot_args_start pointing to
+ * the boot-args block. The kernel's (4K page) page tables are allocated and
+ * initialized before switching to these.
+ *
+ * Non-bootstrap processors are called with argument boot_args_start NULL.
+ * These processors switch immediately to the existing kernel page tables.
+ */
+__attribute__((noreturn))
+void
+vstart(vm_offset_t boot_args_start)
+{
+       boolean_t       is_boot_cpu = !(boot_args_start == 0);
+       int             cpu = 0;
+       uint32_t        lphysfree;
+
+       postcode(VSTART_ENTRY);
+
+       if (is_boot_cpu) {
+               /*
+                * Set-up temporary trap handlers during page-table set-up.
+                */
+               vstart_idt_init();
+               postcode(VSTART_IDT_INIT);
+
+               /*
+                * Get startup parameters.
+                */
+               kernelBootArgs = (boot_args *)boot_args_start;
+               lphysfree = kernelBootArgs->kaddr + kernelBootArgs->ksize;
+               physfree = (void *)(uintptr_t)((lphysfree + PAGE_SIZE - 1) &~ (PAGE_SIZE - 1));
+
+#if DEVELOPMENT || DEBUG
+               pal_serial_init();
+#endif
+               DBG("revision      0x%x\n", kernelBootArgs->Revision);
+               DBG("version       0x%x\n", kernelBootArgs->Version);
+               DBG("command line  %s\n", kernelBootArgs->CommandLine);
+               DBG("memory map    0x%x\n", kernelBootArgs->MemoryMap);
+               DBG("memory map sz 0x%x\n", kernelBootArgs->MemoryMapSize);
+               DBG("kaddr         0x%x\n", kernelBootArgs->kaddr);
+               DBG("ksize         0x%x\n", kernelBootArgs->ksize);
+               DBG("physfree      %p\n", physfree);
+               DBG("bootargs: %p, &ksize: %p &kaddr: %p\n",
+                       kernelBootArgs, 
+                       &kernelBootArgs->ksize,
+                       &kernelBootArgs->kaddr);
+               DBG("SMBIOS mem sz 0x%llx\n", kernelBootArgs->PhysicalMemorySize);
+
+               /*
+                * Setup boot args given the physical start address.
+                * Note: PE_init_platform needs to be called before Idle_PTs_init
+                * because access to the DeviceTree is required to read the
+                * random seed before generating a random physical map slide.
+                */
+               kernelBootArgs = (boot_args *)
+                   ml_static_ptovirt(boot_args_start);
+               DBG("i386_init(0x%lx) kernelBootArgs=%p\n",
+                   (unsigned long)boot_args_start, kernelBootArgs);
+
+#if KASAN
+               kasan_reserve_memory(kernelBootArgs);
+#endif
+
+               PE_init_platform(FALSE, kernelBootArgs);
+               postcode(PE_INIT_PLATFORM_D);
+
+               Idle_PTs_init();
+               postcode(VSTART_IDLE_PTS_INIT);
+
+#if KASAN
+               /* Init kasan and map whatever was stolen from physfree */
+               kasan_init();
+               kasan_notify_stolen((uintptr_t)ml_static_ptovirt((vm_offset_t)physfree));
+#endif
+
+#if MONOTONIC
+               mt_early_init();
+#endif /* MONOTONIC */
+
+               first_avail = (vm_offset_t)ID_MAP_VTOP(physfree);
+
+               cpu_data_alloc(TRUE);
+
+               cpu_desc_init(cpu_datap(0));
+               postcode(VSTART_CPU_DESC_INIT);
+               cpu_desc_load(cpu_datap(0));
+
+               postcode(VSTART_CPU_MODE_INIT);
+               cpu_syscall_init(cpu_datap(0)); /* cpu_syscall_init() will be
+                                                * invoked on the APs
+                                                * via i386_init_slave()
+                                                */
+       } else {
+               /* Switch to kernel's page tables (from the Boot PTs) */
+               set_cr3_raw((uintptr_t)ID_MAP_VTOP(IdlePML4));
+               /* Find our logical cpu number */
+               cpu = lapic_to_cpu[(LAPIC_READ(ID)>>LAPIC_ID_SHIFT) & LAPIC_ID_MASK];
+               DBG("CPU: %d, GSBASE initial value: 0x%llx\n", cpu, rdmsr64(MSR_IA32_GS_BASE));
+               cpu_desc_load(cpu_datap(cpu));
+       }
+
+       postcode(VSTART_EXIT);
+       x86_init_wrapper(is_boot_cpu ? (uintptr_t) i386_init
+                                    : (uintptr_t) i386_init_slave,
+                        cpu_datap(cpu)->cpu_int_stack_top);
+}
+
+void
+pstate_trace(void)
+{
+}
 
 /*
  *     Cpu initialization.  Running virtual, but without MACH VM
 
 /*
  *     Cpu initialization.  Running virtual, but without MACH VM
- *     set up.  First C routine called.
+ *     set up.
  */
 void
  */
 void
-i386_init(vm_offset_t boot_args_start)
+i386_init(void)
 {
        unsigned int    maxmem;
        uint64_t        maxmemtouse;
 {
        unsigned int    maxmem;
        uint64_t        maxmemtouse;
-       unsigned int    cpus;
-       boolean_t       legacy_mode;
+       unsigned int    cpus = 0;
+       boolean_t       fidn;
+       boolean_t       IA32e = TRUE;
 
        postcode(I386_INIT_ENTRY);
 
 
        postcode(I386_INIT_ENTRY);
 
-       i386_macho_zerofill();
+       pal_i386_init();
+       tsc_init();
+       rtclock_early_init();   /* mach_absolute_time() now functionsl */
 
 
-       /*
-        * Setup boot args given the physical start address.
-        */
-       kernelBootArgs = (boot_args *)
-               ml_static_ptovirt(boot_args_start);
-        kernelBootArgs->MemoryMap = (uint32_t)
-               ml_static_ptovirt((vm_offset_t)kernelBootArgs->MemoryMap);
-        kernelBootArgs->deviceTreeP = (uint32_t)
-               ml_static_ptovirt((vm_offset_t)kernelBootArgs->deviceTreeP);
+       kernel_debug_string_early("i386_init");
+       pstate_trace();
+
+#if CONFIG_MCA
+       /* Initialize machine-check handling */
+       mca_cpu_init();
+#endif
 
        master_cpu = 0;
 
        master_cpu = 0;
-       (void) cpu_data_alloc(TRUE);
        cpu_init();
        cpu_init();
-       postcode(CPU_INIT_D);
 
 
-       /* init processor performance control */
-       pmsInit();
-       
-       PE_init_platform(FALSE, kernelBootArgs);
-       postcode(PE_INIT_PLATFORM_D);
+       postcode(CPU_INIT_D);
 
        printf_init();                  /* Init this in case we need debugger */
        panic_init();                   /* Init this in case we need debugger */
 
        /* setup debugging output if one has been chosen */
 
        printf_init();                  /* Init this in case we need debugger */
        panic_init();                   /* Init this in case we need debugger */
 
        /* setup debugging output if one has been chosen */
+       kernel_debug_string_early("PE_init_kprintf");
        PE_init_kprintf(FALSE);
 
        PE_init_kprintf(FALSE);
 
-       if (!PE_parse_boot_arg("diag", &dgWork.dgFlags))
+       kernel_debug_string_early("kernel_early_bootstrap");
+       kernel_early_bootstrap();
+
+       if (!PE_parse_boot_argn("diag", &dgWork.dgFlags, sizeof (dgWork.dgFlags)))
                dgWork.dgFlags = 0;
 
        serialmode = 0;
                dgWork.dgFlags = 0;
 
        serialmode = 0;
-       if(PE_parse_boot_arg("serial", &serialmode)) {
+       if (PE_parse_boot_argn("serial", &serialmode, sizeof(serialmode))) {
                /* We want a serial keyboard and/or console */
                kprintf("Serial mode specified: %08X\n", serialmode);
                /* We want a serial keyboard and/or console */
                kprintf("Serial mode specified: %08X\n", serialmode);
+               int force_sync = serialmode & SERIALMODE_SYNCDRAIN;
+               if (force_sync || PE_parse_boot_argn("drain_uart_sync", &force_sync, sizeof(force_sync))) {
+                       if (force_sync) {
+                               serialmode |= SERIALMODE_SYNCDRAIN;
+                               kprintf(
+                                   "WARNING: Forcing uart driver to output synchronously."
+                                   "printf()s/IOLogs will impact kernel performance.\n"
+                                   "You are advised to avoid using 'drain_uart_sync' boot-arg.\n");
+                       }
+               }
        }
        }
-       if(serialmode & 1) {
+       if (serialmode & SERIALMODE_OUTPUT) {
                (void)switch_to_serial_console();
                (void)switch_to_serial_console();
-               disableConsoleOutput = FALSE;   /* Allow printfs to happen */
+               disableConsoleOutput = FALSE; /* Allow printfs to happen */
        }
 
        /* setup console output */
        }
 
        /* setup console output */
+       kernel_debug_string_early("PE_init_printf");
        PE_init_printf(FALSE);
 
        kprintf("version_variant = %s\n", version_variant);
        kprintf("version         = %s\n", version);
        PE_init_printf(FALSE);
 
        kprintf("version_variant = %s\n", version_variant);
        kprintf("version         = %s\n", version);
-
-       /*   
-        * VM initialization, after this we're using page tables...
-        * The maximum number of cpus must be set beforehand.
-        */
-       if (!PE_parse_boot_arg("maxmem", &maxmem))
-               maxmemtouse=0;
+       
+       if (!PE_parse_boot_argn("maxmem", &maxmem, sizeof (maxmem)))
+               maxmemtouse = 0;
        else
        else
-               maxmemtouse = ((uint64_t)maxmem) * (uint64_t)(1024 * 1024);
+               maxmemtouse = ((uint64_t)maxmem) * MB;
 
 
-       if (PE_parse_boot_arg("cpus", &cpus)) {
+       if (PE_parse_boot_argn("cpus", &cpus, sizeof (cpus))) {
                if ((0 < cpus) && (cpus < max_ncpus))
                         max_ncpus = cpus;
        }
                if ((0 < cpus) && (cpus < max_ncpus))
                         max_ncpus = cpus;
        }
@@ -189,43 +537,238 @@ i386_init(vm_offset_t boot_args_start)
        /*
         * debug support for > 4G systems
         */
        /*
         * debug support for > 4G systems
         */
-       if (!PE_parse_boot_arg("himemory_mode", &vm_himemory_mode))
-               vm_himemory_mode = 0;
+       PE_parse_boot_argn("himemory_mode", &vm_himemory_mode, sizeof (vm_himemory_mode));
+       if (vm_himemory_mode != 0)
+               kprintf("himemory_mode: %d\n", vm_himemory_mode);
+
+       if (!PE_parse_boot_argn("immediate_NMI", &fidn, sizeof (fidn)))
+               force_immediate_debugger_NMI = FALSE;
+       else
+               force_immediate_debugger_NMI = fidn;
+
+#if DEBUG
+       nanoseconds_to_absolutetime(URGENCY_NOTIFICATION_ASSERT_NS, &urgency_notification_assert_abstime_threshold);
+#endif
+       PE_parse_boot_argn("urgency_notification_abstime",
+           &urgency_notification_assert_abstime_threshold,
+           sizeof(urgency_notification_assert_abstime_threshold));
 
 
-       /*
-        * At this point we check whether we are a 64-bit processor
-        * and that we're not restricted to legacy mode, 32-bit operation.
-        */
-       boolean_t IA32e = FALSE;
-       if (cpuid_extfeatures() & CPUID_EXTFEATURE_EM64T) {
-               kprintf("EM64T supported");
-               if (PE_parse_boot_arg("-legacy", &legacy_mode)) {
-                       kprintf(" but legacy mode forced\n");
-               } else {
-                       IA32e = TRUE;
-                       kprintf(" and will be enabled\n");
-               }
-       }
        if (!(cpuid_extfeatures() & CPUID_EXTFEATURE_XD))
                nx_enabled = 0;
 
        if (!(cpuid_extfeatures() & CPUID_EXTFEATURE_XD))
                nx_enabled = 0;
 
+       /*   
+        * VM initialization, after this we're using page tables...
+        * Thn maximum number of cpus must be set beforehand.
+        */
+       kernel_debug_string_early("i386_vm_init");
        i386_vm_init(maxmemtouse, IA32e, kernelBootArgs);
 
        i386_vm_init(maxmemtouse, IA32e, kernelBootArgs);
 
-       if ( ! PE_parse_boot_arg("novmx", &noVMX))
-               noVMX = 0;      /* OK to support Altivec in rosetta? */
-
-       tsc_init();
-       hpet_init();
-       power_management_init();
-
-       PE_init_platform(TRUE, kernelBootArgs);
-
        /* create the console for verbose or pretty mode */
        /* create the console for verbose or pretty mode */
+       /* Note: doing this prior to tsc_init() allows for graceful panic! */
+       PE_init_platform(TRUE, kernelBootArgs);
        PE_create_console();
 
        PE_create_console();
 
+       kernel_debug_string_early("power_management_init");
+       power_management_init();
        processor_bootstrap();
        thread_bootstrap();
 
        processor_bootstrap();
        thread_bootstrap();
 
+       pstate_trace();
+       kernel_debug_string_early("machine_startup");
        machine_startup();
        machine_startup();
+       pstate_trace();
+}
+
+static void
+do_init_slave(boolean_t fast_restart)
+{
+       void    *init_param     = FULL_SLAVE_INIT;
+
+       postcode(I386_INIT_SLAVE);
+
+       if (!fast_restart) {
+               /* Ensure that caching and write-through are enabled */
+               set_cr0(get_cr0() & ~(CR0_NW|CR0_CD));
+  
+               DBG("i386_init_slave() CPU%d: phys (%d) active.\n",
+                   get_cpu_number(), get_cpu_phys_number());
+  
+               assert(!ml_get_interrupts_enabled());
+  
+               cpu_syscall_init(current_cpu_datap());
+               pmap_cpu_init();
+  
+#if CONFIG_MCA
+               mca_cpu_init();
+#endif
+  
+               LAPIC_INIT();
+               lapic_configure();
+               LAPIC_DUMP();
+               LAPIC_CPU_MAP_DUMP();
+  
+               init_fpu();
+  
+#if CONFIG_MTRR
+               mtrr_update_cpu();
+#endif
+               /* update CPU microcode */
+               ucode_update_wake();
+       } else
+           init_param = FAST_SLAVE_INIT;
+
+#if CONFIG_VMX
+       /* resume VT operation */
+       vmx_resume(FALSE);
+#endif
+
+#if CONFIG_MTRR
+       if (!fast_restart)
+           pat_init();
+#endif
+
+       cpu_thread_init();      /* not strictly necessary */
+
+       cpu_init();     /* Sets cpu_running which starter cpu waits for */
+       slave_main(init_param);
+  
+       panic("do_init_slave() returned from slave_main()");
+}
+
+/*
+ * i386_init_slave() is called from pstart.
+ * We're in the cpu's interrupt stack with interrupts disabled.
+ * At this point we are in legacy mode. We need to switch on IA32e
+ * if the mode is set to 64-bits.
+ */
+void
+i386_init_slave(void)
+{
+       do_init_slave(FALSE);
+}
 
 
+/*
+ * i386_init_slave_fast() is called from pmCPUHalt.
+ * We're running on the idle thread and need to fix up
+ * some accounting and get it so that the scheduler sees this
+ * CPU again.
+ */
+void
+i386_init_slave_fast(void)
+{
+       do_init_slave(TRUE);
+}
+
+#include <libkern/kernel_mach_header.h>
+
+/* TODO: Evaluate global PTEs for the double-mapped translations */
+
+uint64_t dblmap_base, dblmap_max;
+kernel_segment_command_t *hdescseg;
+
+pt_entry_t *dblmapL3;
+unsigned int dblallocs;
+uint64_t dblmap_dist;
+extern uint64_t idt64_hndl_table0[];
+
+
+void doublemap_init(void) {
+       dblmapL3 = ALLOCPAGES(1); // for 512 1GiB entries
+       dblallocs++;
+
+       struct {
+               pt_entry_t entries[PTE_PER_PAGE];
+       } * dblmapL2 = ALLOCPAGES(1); // for 512 2MiB entries
+       dblallocs++;
+
+       dblmapL3[0] = ((uintptr_t)ID_MAP_VTOP(&dblmapL2[0]))
+           | INTEL_PTE_VALID
+           | INTEL_PTE_WRITE;
+
+       hdescseg = getsegbynamefromheader(&_mh_execute_header, "__HIB");
+
+       vm_offset_t hdescb = hdescseg->vmaddr;
+       unsigned long hdescsz = hdescseg->vmsize;
+       unsigned long hdescszr = round_page_64(hdescsz);
+       vm_offset_t hdescc = hdescb, hdesce = hdescb + hdescszr;
+
+       kernel_section_t *thdescsect = getsectbynamefromheader(&_mh_execute_header, "__HIB", "__text");
+       vm_offset_t thdescb = thdescsect->addr;
+       unsigned long thdescsz = thdescsect->size;
+       unsigned long thdescszr = round_page_64(thdescsz);
+       vm_offset_t thdesce = thdescb + thdescszr;
+
+       assert((hdescb & 0xFFF) == 0);
+       /* Mirror HIB translations into the double-mapped pagetable subtree*/
+       for(int i = 0; hdescc < hdesce; i++) {
+               struct {
+                       pt_entry_t entries[PTE_PER_PAGE];
+               } * dblmapL1 = ALLOCPAGES(1);
+               dblallocs++;
+               dblmapL2[0].entries[i] = ((uintptr_t)ID_MAP_VTOP(&dblmapL1[0])) | INTEL_PTE_VALID | INTEL_PTE_WRITE | INTEL_PTE_REF;
+               int hdescn = (int) ((hdesce - hdescc) / PAGE_SIZE);
+               for (int j = 0; j < MIN(PTE_PER_PAGE, hdescn); j++) {
+                       uint64_t template = INTEL_PTE_VALID;
+                       if ((hdescc >= thdescb) && (hdescc < thdesce)) {
+                               /* executable */
+                       } else {
+                               template |= INTEL_PTE_WRITE | INTEL_PTE_NX ; /* Writeable, NX */
+                       }
+                       dblmapL1[0].entries[j] = ((uintptr_t)ID_MAP_VTOP(hdescc)) | template;
+                       hdescc += PAGE_SIZE;
+               }
+       }
+
+       IdlePML4[KERNEL_DBLMAP_PML4_INDEX] = ((uintptr_t)ID_MAP_VTOP(dblmapL3)) | INTEL_PTE_VALID | INTEL_PTE_WRITE | INTEL_PTE_REF;
+
+       dblmap_base = KVADDR(KERNEL_DBLMAP_PML4_INDEX, dblmapL3, 0, 0);
+       dblmap_max = dblmap_base + hdescszr;
+       /* Calculate the double-map distance, which accounts for the current
+        * KASLR slide
+        */
+
+       dblmap_dist = dblmap_base - hdescb;
+       idt64_hndl_table0[1] = DBLMAP(idt64_hndl_table0[1]);
+       idt64_hndl_table0[6] = (uint64_t)(uintptr_t)&kernel_stack_mask;
+
+       extern cpu_data_t cpshadows[], scdatas[];
+       uintptr_t cd1 = (uintptr_t) &cpshadows[0];
+       uintptr_t cd2 = (uintptr_t) &scdatas[0];
+/* Record the displacement from the kernel's per-CPU data pointer, eventually
+ * programmed into GSBASE, to the "shadows" in the doublemapped
+ * region. These are not aliases, but separate physical allocations
+ * containing data required in the doublemapped trampolines.
+*/
+       idt64_hndl_table0[2] = dblmap_dist + cd1 - cd2;
+
+       DBG("Double map base: 0x%qx\n", dblmap_base);
+       DBG("double map idlepml4[%d]: 0x%llx\n", KERNEL_DBLMAP_PML4_INDEX, IdlePML4[KERNEL_DBLMAP_PML4_INDEX]);
+       assert(LDTSZ > LDTSZ_MIN);
+}
+
+vm_offset_t dyn_dblmap(vm_offset_t, vm_offset_t);
+
+#include <i386/pmap_internal.h>
+
+/* Use of this routine is expected to be synchronized by callers
+ * Creates non-executable aliases.
+ */
+vm_offset_t dyn_dblmap(vm_offset_t cva, vm_offset_t sz) {
+       vm_offset_t ava = dblmap_max;
+
+       assert((sz & PAGE_MASK) == 0);
+       assert(cva != 0);
+
+       pmap_alias(ava, cva, cva + sz, VM_PROT_READ | VM_PROT_WRITE, PMAP_EXPAND_OPTIONS_ALIASMAP);
+       dblmap_max += sz;
+       return (ava - cva);
+}
+/* Adjust offsets interior to the bootstrap interrupt descriptor table to redirect
+ * control to the double-mapped interrupt vectors. The IDTR proper will be
+ * programmed via cpu_desc_load()
+ */
+void idt64_remap(void) {
+       for (int i = 0; i < IDTSZ; i++) {
+               master_idt64[i].offset64 = DBLMAP(master_idt64[i].offset64);
+       }
 }
 }