]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/i386/mp_desc.c
xnu-4570.41.2.tar.gz
[apple/xnu.git] / osfmk / i386 / mp_desc.c
index 03cc903d75d6fa2ea12f4047ffe0c4304be02a3d..788e71663931347b217f1dd0b3550d4f6c7cb776 100644 (file)
@@ -1,23 +1,29 @@
 /*
- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2012 Apple Inc. All rights reserved.
  *
- * @APPLE_LICENSE_HEADER_START@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
- * The contents of this file constitute Original Code as defined in and
- * are subject to the Apple Public Source License Version 1.1 (the
- * "License").  You may not use this file except in compliance with the
- * License.  Please obtain a copy of the License at
- * http://www.apple.com/publicsource and read it before using this file.
+ * 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.
  * 
- * This Original Code and all software distributed under the License are
- * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * 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 OR NON-INFRINGEMENT.  Please see the
- * License for the specific language governing rights and limitations
- * under the License.
+ * 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_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * @OSF_COPYRIGHT@
 /*
  */
 
-
 #include <kern/cpu_number.h>
 #include <kern/kalloc.h>
 #include <kern/cpu_data.h>
+#include <mach/mach_types.h>
 #include <mach/machine.h>
+#include <mach/vm_map.h>
+#include <mach/machine/vm_param.h>
 #include <vm/vm_kern.h>
+#include <vm/vm_map.h>
 
+#include <i386/bit_routines.h>
 #include <i386/mp_desc.h>
-#include <i386/lock.h>
 #include <i386/misc_protos.h>
 #include <i386/mp.h>
 #include <i386/pmap.h>
+#include <i386/postcode.h>
+#include <i386/pmap_internal.h>
+#if CONFIG_MCA
+#include <i386/machine_check.h>
+#endif
 
 #include <kern/misc_protos.h>
 
-#include <mach_kdb.h>
-
-/*
- * The i386 needs an interrupt stack to keep the PCB stack from being
- * overrun by interrupts.  All interrupt stacks MUST lie at lower addresses
- * than any thread`s kernel stack.
- */
+#if MONOTONIC
+#include <kern/monotonic.h>
+#endif /* MONOTONIC */
+#include <san/kasan.h>
+
+#define K_INTR_GATE (ACC_P|ACC_PL_K|ACC_INTR_GATE)
+#define U_INTR_GATE (ACC_P|ACC_PL_U|ACC_INTR_GATE)
+
+// Declare macros that will declare the externs
+#define TRAP(n, name)          extern void *name ;
+#define TRAP_ERR(n, name)      extern void *name ;
+#define TRAP_SPC(n, name)      extern void *name ;
+#define TRAP_IST1(n, name)     extern void *name ;
+#define TRAP_IST2(n, name)     extern void *name ;
+#define INTERRUPT(n)           extern void *_intr_ ## n ;
+#define USER_TRAP(n, name)     extern void *name ;
+#define USER_TRAP_SPC(n, name) extern void *name ;
+
+// Include the table to declare the externs
+#include "../x86_64/idt_table.h"
+
+// Undef the macros, then redefine them so we can declare the table
+#undef TRAP
+#undef TRAP_ERR
+#undef TRAP_SPC
+#undef TRAP_IST1
+#undef TRAP_IST2
+#undef INTERRUPT
+#undef USER_TRAP
+#undef USER_TRAP_SPC
+
+#define TRAP(n, name)                  \
+       [n] = {                         \
+               (uintptr_t)&name,       \
+               KERNEL64_CS,            \
+               0,                      \
+               K_INTR_GATE,            \
+               0                       \
+       },
+
+#define TRAP_ERR TRAP
+#define TRAP_SPC TRAP
+
+#define TRAP_IST1(n, name) \
+       [n] = {                         \
+               (uintptr_t)&name,       \
+               KERNEL64_CS,            \
+               1,                      \
+               K_INTR_GATE,            \
+               0                       \
+       },
+
+#define TRAP_IST2(n, name) \
+       [n] = {                         \
+               (uintptr_t)&name,       \
+               KERNEL64_CS,            \
+               2,                      \
+               K_INTR_GATE,            \
+               0                       \
+       },
+
+#define INTERRUPT(n) \
+       [n] = {                         \
+               (uintptr_t)&_intr_ ## n,\
+               KERNEL64_CS,            \
+               0,                      \
+               K_INTR_GATE,            \
+               0                       \
+       },
+
+#define USER_TRAP(n, name) \
+       [n] = {                         \
+               (uintptr_t)&name,       \
+               KERNEL64_CS,            \
+               0,                      \
+               U_INTR_GATE,            \
+               0                       \
+       },
+
+#define USER_TRAP_SPC USER_TRAP
+
+// Declare the table using the macros we just set up
+struct fake_descriptor64 master_idt64[IDTSZ]
+       __attribute__ ((section("__HIB,__desc")))
+       __attribute__ ((aligned(PAGE_SIZE))) = {
+#include "../x86_64/idt_table.h"
+};
 
 /*
  * First cpu`s interrupt stack.
  */
-extern char            intstack[];     /* bottom */
-extern char            eintstack[];    /* top */
+extern uint32_t                low_intstack[];         /* bottom */
+extern uint32_t                low_eintstack[];        /* top */
 
 /*
  * Per-cpu data area pointers.
- * The master cpu (cpu 0) has its data area statically allocated;
- * others are allocated dynamically and this array is updated at runtime.
  */
-cpu_data_t     cpu_data_master;
-cpu_data_t     *cpu_data_ptr[MAX_CPUS] = { [0] &cpu_data_master };
+cpu_data_t cpshadows[MAX_CPUS] __attribute__((aligned(64))) __attribute__((section("__HIB, __desc")));
+cpu_data_t scdatas[MAX_CPUS] __attribute__((aligned(64))) = {
+       [0].cpu_this = &scdatas[0],
+       [0].cpu_nanotime = &pal_rtc_nanotime_info,
+       [0].cpu_int_stack_top = (vm_offset_t) low_eintstack,
+       [0].cd_shadow = &cpshadows[0]
+};
+cpu_data_t *cpu_data_master = &scdatas[0];
 
-decl_simple_lock_data(,cpu_lock);      /* protects real_ncpus */
+cpu_data_t     *cpu_data_ptr[MAX_CPUS] = { [0] = &scdatas[0] };
+
+decl_simple_lock_data(,ncpus_lock);    /* protects real_ncpus */
 unsigned int   real_ncpus = 1;
 unsigned int   max_ncpus = MAX_CPUS;
 
+extern void hi64_sysenter(void);
+extern void hi64_syscall(void);
+
+typedef struct {
+       struct real_descriptor pcldts[LDTSZ];
+} cldt_t;
+
+cpu_desc_table64_t scdtables[MAX_CPUS] __attribute__((aligned(64))) __attribute__((section("__HIB, __desc")));
+cpu_fault_stack_t scfstks[MAX_CPUS] __attribute__((aligned(64))) __attribute__((section("__HIB, __desc")));
+
+cldt_t *dyn_ldts;
+
 /*
  * Multiprocessor i386/i486 systems use a separate copy of the
  * GDT, IDT, LDT, and kernel TSS per processor.  The first three
@@ -105,107 +217,330 @@ unsigned int    max_ncpus = MAX_CPUS;
  * Allocate and initialize the per-processor descriptor tables.
  */
 
-struct fake_descriptor ldt_desc_pattern = {
-       (unsigned int) 0,
-       LDTSZ * sizeof(struct fake_descriptor) - 1,
+/*
+ * This is the expanded, 64-bit variant of the kernel LDT descriptor.
+ * When switching to 64-bit mode this replaces KERNEL_LDT entry
+ * and the following empty slot. This enables the LDT to be referenced
+ * in the uber-space remapping window on the kernel.
+ */
+struct fake_descriptor64 kernel_ldt_desc64 = {
        0,
-       ACC_P|ACC_PL_K|ACC_LDT
-};
-struct fake_descriptor tss_desc_pattern = {
-       (unsigned int) 0,
-       sizeof(struct i386_tss),
+       LDTSZ_MIN*sizeof(struct fake_descriptor)-1,
        0,
-       ACC_P|ACC_PL_K|ACC_TSS
+       ACC_P|ACC_PL_K|ACC_LDT,
+       0
 };
 
-struct fake_descriptor cpudata_desc_pattern = {
-       (unsigned int) 0,
-       sizeof(cpu_data_t)-1,
-       SZ_32,
-       ACC_P|ACC_PL_K|ACC_DATA_W
+/*
+ * This is the expanded, 64-bit variant of the kernel TSS descriptor.
+ * It is follows pattern of the KERNEL_LDT.
+ */
+struct fake_descriptor64 kernel_tss_desc64 = {
+       0,
+       sizeof(struct x86_64_tss)-1,
+       0,
+       ACC_P|ACC_PL_K|ACC_TSS,
+       0
 };
 
+/*
+ * Convert a descriptor from fake to real format.
+ *
+ * Fake descriptor format:
+ *     bytes 0..3              base 31..0
+ *     bytes 4..5              limit 15..0
+ *     byte  6                 access byte 2 | limit 19..16
+ *     byte  7                 access byte 1
+ *
+ * Real descriptor format:
+ *     bytes 0..1              limit 15..0
+ *     bytes 2..3              base 15..0
+ *     byte  4                 base 23..16
+ *     byte  5                 access byte 1
+ *     byte  6                 access byte 2 | limit 19..16
+ *     byte  7                 base 31..24
+ *
+ * Fake gate format:
+ *     bytes 0..3              offset
+ *     bytes 4..5              selector
+ *     byte  6                 word count << 4 (to match fake descriptor)
+ *     byte  7                 access byte 1
+ *
+ * Real gate format:
+ *     bytes 0..1              offset 15..0
+ *     bytes 2..3              selector
+ *     byte  4                 word count
+ *     byte  5                 access byte 1
+ *     bytes 6..7              offset 31..16
+ */
+void
+fix_desc(void *d, int num_desc) {
+       //early_kprintf("fix_desc(%x, %x)\n", d, num_desc);
+       uint8_t *desc = (uint8_t*) d;
+
+       do {
+               if ((desc[7] & 0x14) == 0x04) { /* gate */
+                       uint32_t offset;
+                       uint16_t selector;
+                       uint8_t wordcount;
+                       uint8_t acc;
+                       
+                       offset = *((uint32_t*)(desc));
+                       selector = *((uint32_t*)(desc+4));
+                       wordcount = desc[6] >> 4;
+                       acc = desc[7];
+
+                       *((uint16_t*)desc) = offset & 0xFFFF;
+                       *((uint16_t*)(desc+2)) = selector;
+                       desc[4] = wordcount;
+                       desc[5] = acc;
+                       *((uint16_t*)(desc+6)) = offset >> 16;
+
+               } else { /* descriptor */
+                       uint32_t base;
+                       uint16_t limit;
+                       uint8_t acc1, acc2;
+
+                       base = *((uint32_t*)(desc));
+                       limit = *((uint16_t*)(desc+4));
+                       acc2 = desc[6];
+                       acc1 = desc[7];
+
+                       *((uint16_t*)(desc)) = limit;
+                       *((uint16_t*)(desc+2)) = base & 0xFFFF;
+                       desc[4] = (base >> 16) & 0xFF;
+                       desc[5] = acc1;
+                       desc[6] = acc2;
+                       desc[7] = base >> 24;
+               }
+               desc += 8;
+       } while (--num_desc);
+}
+
 void
-mp_desc_init(
-       cpu_data_t      *cdp,
-       boolean_t       is_boot_cpu)
+fix_desc64(void *descp, int count)
 {
-       struct mp_desc_table    *mpt = cdp->cpu_desc_tablep;
-       cpu_desc_index_t        *cdt = &cdp->cpu_desc_index;
+       struct fake_descriptor64        *fakep;
+       union {
+               struct real_gate64              gate;
+               struct real_descriptor64        desc;
+       }                               real;
+       int                             i;
+
+       fakep = (struct fake_descriptor64 *) descp;
+       
+       for (i = 0; i < count; i++, fakep++) {
+               /*
+                * Construct the real decriptor locally.
+                */
+
+               bzero((void *) &real, sizeof(real));
+
+               switch (fakep->access & ACC_TYPE) {
+               case 0:
+                       break;
+               case ACC_CALL_GATE:
+               case ACC_INTR_GATE:
+               case ACC_TRAP_GATE:
+                       real.gate.offset_low16 = (uint16_t)(fakep->offset64 & 0xFFFF);
+                       real.gate.selector16 = fakep->lim_or_seg & 0xFFFF;
+                       real.gate.IST = fakep->size_or_IST & 0x7;
+                       real.gate.access8 = fakep->access;
+                       real.gate.offset_high16 = (uint16_t)((fakep->offset64>>16) & 0xFFFF);
+                       real.gate.offset_top32 = (uint32_t)(fakep->offset64>>32);
+                       break;
+               default:        /* Otherwise */
+                       real.desc.limit_low16 = fakep->lim_or_seg & 0xFFFF;
+                       real.desc.base_low16 = (uint16_t)(fakep->offset64 & 0xFFFF);
+                       real.desc.base_med8 = (uint8_t)((fakep->offset64 >> 16) & 0xFF);
+                       real.desc.access8 = fakep->access;
+                       real.desc.limit_high4 = (fakep->lim_or_seg >> 16) & 0xFF;
+                       real.desc.granularity4 = fakep->size_or_IST;
+                       real.desc.base_high8 = (uint8_t)((fakep->offset64 >> 24) & 0xFF);
+                       real.desc.base_top32 = (uint32_t)(fakep->offset64>>32);
+               }
 
-       if (is_boot_cpu) {
-           /*
-            * Master CPU uses the tables built at boot time.
-            * Just set the TSS and GDT pointers.
-            */
-           cdt->cdi_ktss = &ktss;
-#if    MACH_KDB
-           cdt->cdi_dbtss = &dbtss;
-#endif /* MACH_KDB */
-           cdt->cdi_gdt = gdt;
-           cdt->cdi_idt = idt;
-           cdt->cdi_ldt = ldt;
-
-       } else {
-
-           cdt->cdi_ktss = &mpt->ktss;
-           cdt->cdi_gdt = mpt->gdt;
-           cdt->cdi_idt = mpt->idt;
-           cdt->cdi_ldt = mpt->ldt;
-
-           /*
-            * Copy the tables
-            */
-           bcopy((char *)idt,
-                 (char *)mpt->idt,
-                 sizeof(idt));
-           bcopy((char *)gdt,
-                 (char *)mpt->gdt,
-                 sizeof(gdt));
-           bcopy((char *)ldt,
-                 (char *)mpt->ldt,
-                 sizeof(ldt));
-           bzero((char *)&mpt->ktss,
-                 sizeof(struct i386_tss));
-
-#if    MACH_KDB
-           cdt->cdi_dbtss = &dbtss;
-           bcopy((char *)&dbtss,
-                 (char *)&mpt->dbtss,
-                 sizeof(struct i386_tss));
-#endif /* MACH_KDB */
-
-           /*
-            * Fix up the entries in the GDT to point to
-            * this LDT and this TSS.
-            */
-           mpt->gdt[sel_idx(KERNEL_LDT)] = ldt_desc_pattern;
-           mpt->gdt[sel_idx(KERNEL_LDT)].offset = (vm_offset_t) mpt->ldt;
-           fix_desc(&mpt->gdt[sel_idx(KERNEL_LDT)], 1);
-
-           mpt->gdt[sel_idx(KERNEL_TSS)] = tss_desc_pattern;
-           mpt->gdt[sel_idx(KERNEL_TSS)].offset = (vm_offset_t) &mpt->ktss;
-           fix_desc(&mpt->gdt[sel_idx(KERNEL_TSS)], 1);
-
-           mpt->gdt[sel_idx(CPU_DATA_GS)] = cpudata_desc_pattern;
-           mpt->gdt[sel_idx(CPU_DATA_GS)].offset = (vm_offset_t) cdp;
-           fix_desc(&mpt->gdt[sel_idx(CPU_DATA_GS)], 1);
-
-#if    MACH_KDB
-           mpt->gdt[sel_idx(DEBUG_TSS)] = tss_desc_pattern;
-           mpt->gdt[sel_idx(DEBUG_TSS)].offset = (vm_offset_t) &mpt->dbtss;
-           fix_desc(&mpt->gdt[sel_idx(DEBUG_TSS)], 1);
-
-           mpt->dbtss.esp0 = (int)(db_task_stack_store +
-                   (INTSTACK_SIZE * (cpu + 1)) - sizeof (natural_t));
-           mpt->dbtss.esp = mpt->dbtss.esp0;
-           mpt->dbtss.eip = (int)&db_task_start;
-#endif /* MACH_KDB */
-
-           mpt->ktss.ss0 = KERNEL_DS;
-           mpt->ktss.io_bit_map_offset = 0x0FFF;       /* no IO bitmap */
+               /*
+                * Now copy back over the fake structure.
+                */
+               bcopy((void *) &real, (void *) fakep, sizeof(real));
+       }
+}
+
+extern unsigned mldtsz;
+void
+cpu_desc_init(cpu_data_t *cdp)
+{
+       cpu_desc_index_t        *cdi = &cdp->cpu_desc_index;
+
+       if (cdp == cpu_data_master) {
+               /*
+                * Populate the double-mapped 'u' and base 'b' fields in the
+                * KTSS with I/G/LDT and sysenter stack data.
+                */
+               cdi->cdi_ktssu = (void *)DBLMAP(&master_ktss64);
+               cdi->cdi_ktssb = (void *)&master_ktss64;
+               cdi->cdi_sstku = (vm_offset_t) DBLMAP(&master_sstk.top);
+               cdi->cdi_sstkb = (vm_offset_t) &master_sstk.top;
+
+               cdi->cdi_gdtu.ptr = (void *)DBLMAP((uintptr_t) &master_gdt);
+               cdi->cdi_gdtb.ptr = (void *)&master_gdt;
+               cdi->cdi_idtu.ptr  = (void *)DBLMAP((uintptr_t) &master_idt64);
+               cdi->cdi_idtb.ptr  = (void *)((uintptr_t) &master_idt64);
+               cdi->cdi_ldtu  = (struct fake_descriptor *) (void *) DBLMAP((uintptr_t)&master_ldt[0]);
+               cdi->cdi_ldtb  = (struct fake_descriptor *) (void *) &master_ldt[0];
+
+               /* Replace the expanded LDTs and TSS slots in the GDT */
+               kernel_ldt_desc64.offset64 = (uintptr_t) cdi->cdi_ldtu;
+               *(struct fake_descriptor64 *) &master_gdt[sel_idx(KERNEL_LDT)] =
+                       kernel_ldt_desc64;
+               *(struct fake_descriptor64 *) &master_gdt[sel_idx(USER_LDT)] =
+                       kernel_ldt_desc64;
+               kernel_tss_desc64.offset64 = (uintptr_t) DBLMAP(&master_ktss64);
+               *(struct fake_descriptor64 *) &master_gdt[sel_idx(KERNEL_TSS)] =
+                       kernel_tss_desc64;
+
+               /* Fix up the expanded descriptors for 64-bit. */
+               fix_desc64((void *) &master_idt64, IDTSZ);
+               fix_desc64((void *) &master_gdt[sel_idx(KERNEL_LDT)], 1);
+               fix_desc64((void *) &master_gdt[sel_idx(USER_LDT)], 1);
+               fix_desc64((void *) &master_gdt[sel_idx(KERNEL_TSS)], 1);
+
+               /*
+                * Set the NMI/fault stacks as IST2/IST1 in the 64-bit TSS
+                */
+               master_ktss64.ist2 = (uintptr_t) low_eintstack;
+               master_ktss64.ist1 = (uintptr_t) low_eintstack - sizeof(x86_64_intr_stack_frame_t);
+       } else if (cdi->cdi_ktssu == NULL) {    /* Skipping re-init on wake */
+               cpu_desc_table64_t      *cdt = (cpu_desc_table64_t *) cdp->cpu_desc_tablep;
+
+               cdi->cdi_idtu.ptr  = (void *)DBLMAP((uintptr_t) &master_idt64);
+
+               cdi->cdi_ktssu = (void *)DBLMAP(&cdt->ktss);
+               cdi->cdi_ktssb = (void *)(&cdt->ktss);
+               cdi->cdi_sstku = (vm_offset_t)DBLMAP(&cdt->sstk.top);
+               cdi->cdi_sstkb = (vm_offset_t)(&cdt->sstk.top);
+               cdi->cdi_ldtu  = (void *)LDTALIAS(cdp->cpu_ldtp);
+               cdi->cdi_ldtb  = (void *)(cdp->cpu_ldtp);
+
+               /*
+                * Copy the tables
+                */
+               bcopy((char *)master_gdt, (char *)cdt->gdt, sizeof(master_gdt));
+               bcopy((char *)master_ldt, (char *)cdp->cpu_ldtp, mldtsz);
+               bcopy((char *)&master_ktss64, (char *)&cdt->ktss, sizeof(struct x86_64_tss));
+               cdi->cdi_gdtu.ptr  = (void *)DBLMAP(cdt->gdt);
+               cdi->cdi_gdtb.ptr  = (void *)(cdt->gdt);
+               /*
+                * Fix up the entries in the GDT to point to
+                * this LDT and this TSS.
+                * Note reuse of global 'kernel_ldt_desc64, which is not
+                * concurrency-safe. Higher level synchronization is expected
+                */
+               kernel_ldt_desc64.offset64 = (uintptr_t) cdi->cdi_ldtu;
+               *(struct fake_descriptor64 *) &cdt->gdt[sel_idx(KERNEL_LDT)] =
+                       kernel_ldt_desc64;
+               fix_desc64(&cdt->gdt[sel_idx(KERNEL_LDT)], 1);
+
+               kernel_ldt_desc64.offset64 = (uintptr_t) cdi->cdi_ldtu;
+               *(struct fake_descriptor64 *) &cdt->gdt[sel_idx(USER_LDT)] =
+                       kernel_ldt_desc64;
+               fix_desc64(&cdt->gdt[sel_idx(USER_LDT)], 1);
+
+               kernel_tss_desc64.offset64 = (uintptr_t) cdi->cdi_ktssu;
+               *(struct fake_descriptor64 *) &cdt->gdt[sel_idx(KERNEL_TSS)] =
+                       kernel_tss_desc64;
+               fix_desc64(&cdt->gdt[sel_idx(KERNEL_TSS)], 1);
+
+               /* Set (zeroed) fault stack as IST1, NMI intr stack IST2 */
+               uint8_t *cfstk = &scfstks[cdp->cpu_number].fstk[0];
+               cdt->fstkp = cfstk;
+               bzero((void *) cfstk, FSTK_SZ);
+               cdt->ktss.ist2 = DBLMAP((uint64_t)cdt->fstkp + FSTK_SZ);
+               cdt->ktss.ist1 = cdt->ktss.ist2 - sizeof(x86_64_intr_stack_frame_t);
        }
+
+       /* Require that the top of the sysenter stack is 16-byte aligned */
+       if ((cdi->cdi_sstku % 16) != 0)
+               panic("cpu_desc_init() sysenter stack not 16-byte aligned");
 }
+void
+cpu_desc_load(cpu_data_t *cdp)
+{
+       cpu_desc_index_t        *cdi = &cdp->cpu_desc_index;
+
+       postcode(CPU_DESC_LOAD_ENTRY);
+
+       /* Stuff the kernel per-cpu data area address into the MSRs */
+       postcode(CPU_DESC_LOAD_GS_BASE);
+       wrmsr64(MSR_IA32_GS_BASE, (uintptr_t) cdp);
+       postcode(CPU_DESC_LOAD_KERNEL_GS_BASE);
+       wrmsr64(MSR_IA32_KERNEL_GS_BASE, (uintptr_t) cdp);
+
+       /*
+        * Ensure the TSS segment's busy bit is clear. This is required
+        * for the case of reloading descriptors at wake to avoid
+        * their complete re-initialization.
+        */
+       gdt_desc_p(KERNEL_TSS)->access &= ~ACC_TSS_BUSY;
+
+       /* Load the GDT, LDT, IDT and TSS */
+       cdi->cdi_gdtb.size = sizeof(struct real_descriptor)*GDTSZ - 1;
+       cdi->cdi_gdtu.size = cdi->cdi_gdtb.size;
+       cdi->cdi_idtb.size = 0x1000 + cdp->cpu_number;
+       cdi->cdi_idtu.size = cdi->cdi_idtb.size;
+
+       postcode(CPU_DESC_LOAD_GDT);
+       lgdt((uintptr_t *) &cdi->cdi_gdtu);
+       postcode(CPU_DESC_LOAD_IDT);
+       lidt((uintptr_t *) &cdi->cdi_idtu);
+       postcode(CPU_DESC_LOAD_LDT);
+       lldt(KERNEL_LDT);
+       postcode(CPU_DESC_LOAD_TSS);
+       set_tr(KERNEL_TSS);
+
+#if GPROF // Hack to enable mcount to work on K64
+       __asm__ volatile("mov %0, %%gs" : : "rm" ((unsigned short)(KERNEL_DS)));
+#endif
+       postcode(CPU_DESC_LOAD_EXIT);
+}
+
+/*
+ * Set MSRs for sysenter/sysexit and syscall/sysret for 64-bit.
+ */
+void
+cpu_syscall_init(cpu_data_t *cdp)
+{
+#if MONOTONIC
+       mt_cpu_up(cdp);
+#else /* MONOTONIC */
+#pragma unused(cdp)
+#endif /* !MONOTONIC */
+       wrmsr64(MSR_IA32_SYSENTER_CS, SYSENTER_CS); 
+       wrmsr64(MSR_IA32_SYSENTER_EIP, DBLMAP((uintptr_t) hi64_sysenter));
+       wrmsr64(MSR_IA32_SYSENTER_ESP, current_cpu_datap()->cpu_desc_index.cdi_sstku);
+       /* Enable syscall/sysret */
+       wrmsr64(MSR_IA32_EFER, rdmsr64(MSR_IA32_EFER) | MSR_IA32_EFER_SCE);
+
+       /*
+        * MSRs for 64-bit syscall/sysret
+        * Note USER_CS because sysret uses this + 16 when returning to
+        * 64-bit code.
+        */
+       wrmsr64(MSR_IA32_LSTAR, DBLMAP((uintptr_t) hi64_syscall));
+       wrmsr64(MSR_IA32_STAR, (((uint64_t)USER_CS) << 48) | (((uint64_t)KERNEL64_CS) << 32));
+       /*
+        * Emulate eflags cleared by sysenter but note that
+        * we also clear the trace trap to avoid the complications
+        * of single-stepping into a syscall. The nested task bit
+        * is also cleared to avoid a spurious "task switch"
+        * should we choose to return via an IRET.
+        */
+       wrmsr64(MSR_IA32_FMASK, EFL_DF|EFL_IF|EFL_TF|EFL_NT);
+
+}
+extern vm_offset_t dyn_dblmap(vm_offset_t, vm_offset_t);
+uint64_t ldt_alias_offset;
 
 cpu_data_t *
 cpu_data_alloc(boolean_t is_boot_cpu)
@@ -215,43 +550,43 @@ cpu_data_alloc(boolean_t is_boot_cpu)
 
        if (is_boot_cpu) {
                assert(real_ncpus == 1);
-               simple_lock_init(&cpu_lock, 0);
-               cdp = &cpu_data_master;
+               cdp = cpu_datap(0);
                if (cdp->cpu_processor == NULL) {
+                       simple_lock_init(&ncpus_lock, 0);
                        cdp->cpu_processor = cpu_processor_alloc(TRUE);
+#if NCOPY_WINDOWS > 0
                        cdp->cpu_pmap = pmap_cpu_alloc(TRUE);
-                       cdp->cpu_this = cdp;
-                       cdp->cpu_int_stack_top = (vm_offset_t) eintstack;
-                       mp_desc_init(cdp, TRUE);
+#endif
                }
                return cdp;
        }
 
-       /* Check count before making allocations */
-       if (real_ncpus >= max_ncpus)
-               return NULL;
+       boolean_t do_ldt_alloc = FALSE;
+       simple_lock(&ncpus_lock);
+       int cnum = real_ncpus;
+       real_ncpus++;
+       if (dyn_ldts == NULL) {
+               do_ldt_alloc = TRUE;
+       }
+       simple_unlock(&ncpus_lock);
 
        /*
         * Allocate per-cpu data:
         */
-       ret = kmem_alloc(kernel_map, 
-                        (vm_offset_t *) &cdp, sizeof(cpu_data_t));
-       if (ret != KERN_SUCCESS) {
-               printf("cpu_data_alloc() failed, ret=%d\n", ret);
-               goto abort;
-       }
+
+       cdp = &scdatas[cnum];
        bzero((void*) cdp, sizeof(cpu_data_t));
        cdp->cpu_this = cdp;
-
+       cdp->cpu_number = cnum;
+       cdp->cd_shadow = &cpshadows[cnum];
        /*
         * Allocate interrupt stack:
         */
        ret = kmem_alloc(kernel_map, 
                         (vm_offset_t *) &cdp->cpu_int_stack_top,
-                        INTSTACK_SIZE);
+                        INTSTACK_SIZE, VM_KERN_MEMORY_CPU);
        if (ret != KERN_SUCCESS) {
-               printf("cpu_data_alloc() int stack failed, ret=%d\n", ret);
-               goto abort;
+               panic("cpu_data_alloc() int stack failed, ret=%d\n", ret);
        }
        bzero((void*) cdp->cpu_int_stack_top, INTSTACK_SIZE);
        cdp->cpu_int_stack_top += INTSTACK_SIZE;
@@ -259,57 +594,285 @@ cpu_data_alloc(boolean_t is_boot_cpu)
        /*
         * Allocate descriptor table:
         */
-       ret = kmem_alloc(kernel_map, 
-                        (vm_offset_t *) &cdp->cpu_desc_tablep,
-                        sizeof(struct mp_desc_table));
-       if (ret != KERN_SUCCESS) {
-               printf("cpu_data_alloc() desc_table failed, ret=%d\n", ret);
-               goto abort;
-       }
 
-       simple_lock(&cpu_lock);
-       if (real_ncpus >= max_ncpus) {
-               simple_unlock(&cpu_lock);
-               goto abort;
+       cdp->cpu_desc_tablep = (struct cpu_desc_table *) &scdtables[cnum];
+       /*
+        * Allocate LDT
+        */
+       if (do_ldt_alloc) {
+               boolean_t do_ldt_free = FALSE;
+               vm_offset_t sldtoffset = 0;
+               /*
+                * Allocate LDT
+                */
+               vm_offset_t ldtalloc = 0, ldtallocsz = round_page_64(MAX_CPUS * sizeof(struct real_descriptor) * LDTSZ);
+               ret = kmem_alloc(kernel_map, (vm_offset_t *) &ldtalloc, ldtallocsz, VM_KERN_MEMORY_CPU);
+               if (ret != KERN_SUCCESS) {
+                       panic("cpu_data_alloc() ldt failed, kmem_alloc=%d\n", ret);
+               }
+
+               simple_lock(&ncpus_lock);
+               if (dyn_ldts == NULL) {
+                       dyn_ldts = (cldt_t *)ldtalloc;
+               } else {
+                       do_ldt_free = TRUE;
+               }
+               simple_unlock(&ncpus_lock);
+
+               if (do_ldt_free) {
+                       kmem_free(kernel_map, ldtalloc, ldtallocsz);
+               } else {
+                       /* CPU registration and startup are expected to execute
+                        * serially, as invoked by the platform driver.
+                        * Create trampoline alias of LDT region.
+                        */
+                       sldtoffset = dyn_dblmap(ldtalloc, ldtallocsz);
+                       ldt_alias_offset = sldtoffset;
+               }
        }
-       cpu_data_ptr[real_ncpus] = cdp;
-       cdp->cpu_number = real_ncpus;
-       real_ncpus++;
-       simple_unlock(&cpu_lock);
-       
-       kprintf("cpu_data_alloc(%d) 0x%x desc_table: 0x%x "
-               "int_stack: 0x%x-0x%x\n",
-               cdp->cpu_number, cdp, cdp->cpu_desc_tablep,
-               cdp->cpu_int_stack_top - INTSTACK_SIZE, cdp->cpu_int_stack_top);
+       cdp->cpu_ldtp = &dyn_ldts[cnum].pcldts[0];
+
+#if CONFIG_MCA
+       /* Machine-check shadow register allocation. */
+       mca_cpu_alloc(cdp);
+#endif
+
+       /*
+        * Before this cpu has been assigned a real thread context,
+        * we give it a fake, unique, non-zero thread id which the locking
+        * primitives use as their lock value.
+        * Note that this does not apply to the boot processor, cpu 0, which
+        * transitions to a thread context well before other processors are
+        * started.
+        */
+       cdp->cpu_active_thread = (thread_t) (uintptr_t) cdp->cpu_number;
+
+       cdp->cpu_nanotime = &pal_rtc_nanotime_info;
+
+       kprintf("cpu_data_alloc(%d) %p desc_table: %p "
+               "ldt: %p "
+               "int_stack: 0x%lx-0x%lx\n",
+               cdp->cpu_number, cdp, cdp->cpu_desc_tablep, cdp->cpu_ldtp,
+               (long)(cdp->cpu_int_stack_top - INTSTACK_SIZE), (long)(cdp->cpu_int_stack_top));
+       cpu_data_ptr[cnum] = cdp;
 
        return cdp;
 
-abort:
-       if (cdp) {
-               if (cdp->cpu_desc_tablep)
-                       kfree((void *) cdp->cpu_desc_tablep,
-                               sizeof(*cdp->cpu_desc_tablep));
-               if (cdp->cpu_int_stack_top)
-                       kfree((void *) (cdp->cpu_int_stack_top - INTSTACK_SIZE),
-                               INTSTACK_SIZE);
-               kfree((void *) cdp, sizeof(*cdp));
-       }
-       return NULL;
+}
+
+boolean_t
+valid_user_data_selector(uint16_t selector)
+{
+    sel_t      sel = selector_to_sel(selector);
+    
+    if (selector == 0)
+       return (TRUE);
+
+    if (sel.ti == SEL_LDT)
+       return (TRUE);
+    else if (sel.index < GDTSZ) {
+       if ((gdt_desc_p(selector)->access & ACC_PL_U) == ACC_PL_U)
+           return (TRUE);
+    }
+    return (FALSE);
+}
+
+boolean_t
+valid_user_code_selector(uint16_t selector)
+{
+    sel_t      sel = selector_to_sel(selector);
+    
+    if (selector == 0)
+       return (FALSE);
+
+    if (sel.ti == SEL_LDT) {
+       if (sel.rpl == USER_PRIV)
+           return (TRUE);
+    }
+    else if (sel.index < GDTSZ && sel.rpl == USER_PRIV) {
+       if ((gdt_desc_p(selector)->access & ACC_PL_U) == ACC_PL_U)
+           return (TRUE);
+       /* Explicitly validate the system code selectors
+        * even if not instantaneously privileged,
+        * since they are dynamically re-privileged
+        * at context switch
+        */
+       if ((selector == USER_CS) || (selector == USER64_CS))
+               return (TRUE);
+    }
+
+    return (FALSE);
+}
+
+boolean_t
+valid_user_stack_selector(uint16_t selector)
+{
+    sel_t      sel = selector_to_sel(selector);
+    
+    if (selector == 0)
+       return (FALSE);
+
+    if (sel.ti == SEL_LDT) {
+       if (sel.rpl == USER_PRIV)
+           return (TRUE);
+    }
+    else if (sel.index < GDTSZ && sel.rpl == USER_PRIV) {
+       if ((gdt_desc_p(selector)->access & ACC_PL_U) == ACC_PL_U)
+           return (TRUE);
+    }
+               
+    return (FALSE);
 }
 
 boolean_t
 valid_user_segment_selectors(uint16_t cs,
-                            uint16_t ss,
-                            uint16_t ds,
-                            uint16_t es,
-                            uint16_t fs,
-                            uint16_t gs)
+               uint16_t ss,
+               uint16_t ds,
+               uint16_t es,
+               uint16_t fs,
+               uint16_t gs)
 {      
        return valid_user_code_selector(cs)  &&
-              valid_user_stack_selector(ss) &&
-              valid_user_data_selector(ds)  &&
-              valid_user_data_selector(es)  &&
-              valid_user_data_selector(fs)  &&
-              valid_user_data_selector(gs);
+               valid_user_stack_selector(ss) &&
+               valid_user_data_selector(ds)  &&
+               valid_user_data_selector(es)  &&
+               valid_user_data_selector(fs)  &&
+               valid_user_data_selector(gs);
 }
 
+#if NCOPY_WINDOWS > 0
+
+static vm_offset_t user_window_base = 0;
+
+void
+cpu_userwindow_init(int cpu)
+{
+       cpu_data_t              *cdp = cpu_data_ptr[cpu];
+       vm_offset_t             user_window;
+       vm_offset_t             vaddr;
+       int                     num_cpus;
+
+       num_cpus = ml_get_max_cpus();
+
+       if (cpu >= num_cpus)
+               panic("cpu_userwindow_init: cpu > num_cpus");
+
+       if (user_window_base == 0) {
+
+               if (vm_allocate(kernel_map, &vaddr,
+                                       (NBPDE * NCOPY_WINDOWS * num_cpus) + NBPDE,
+                                       VM_FLAGS_ANYWHERE | VM_MAKE_TAG(VM_KERN_MEMORY_CPU)) != KERN_SUCCESS)
+                       panic("cpu_userwindow_init: "
+                                       "couldn't allocate user map window");
+
+               /*
+                * window must start on a page table boundary
+                * in the virtual address space
+                */
+               user_window_base = (vaddr + (NBPDE - 1)) & ~(NBPDE - 1);
+
+               /*
+                * get rid of any allocation leading up to our
+                * starting boundary
+                */
+               vm_deallocate(kernel_map, vaddr, user_window_base - vaddr);
+
+               /*
+                * get rid of tail that we don't need
+                */
+               user_window = user_window_base +
+                                       (NBPDE * NCOPY_WINDOWS * num_cpus);
+
+               vm_deallocate(kernel_map, user_window,
+                               (vaddr +
+                                ((NBPDE * NCOPY_WINDOWS * num_cpus) + NBPDE)) -
+                                user_window);
+       }
+
+       user_window = user_window_base + (cpu * NCOPY_WINDOWS * NBPDE);
+
+       cdp->cpu_copywindow_base = user_window;
+       /*
+        * Abuse this pdp entry, the pdp now actually points to 
+        * an array of copy windows addresses.
+        */
+       cdp->cpu_copywindow_pdp  = pmap_pde(kernel_pmap, user_window);
+
+}
+
+void
+cpu_physwindow_init(int cpu)
+{
+       cpu_data_t              *cdp = cpu_data_ptr[cpu];
+        vm_offset_t            phys_window = cdp->cpu_physwindow_base;
+
+       if (phys_window == 0) {
+               if (vm_allocate(kernel_map, &phys_window,
+                               PAGE_SIZE, VM_FLAGS_ANYWHERE | VM_MAKE_TAG(VM_KERN_MEMORY_CPU))
+                               != KERN_SUCCESS)
+                       panic("cpu_physwindow_init: "
+                               "couldn't allocate phys map window");
+
+               /*
+                * make sure the page that encompasses the
+                * pte pointer we're interested in actually
+                * exists in the page table
+                */
+               pmap_expand(kernel_pmap, phys_window, PMAP_EXPAND_OPTIONS_NONE);
+
+               cdp->cpu_physwindow_base = phys_window;
+               cdp->cpu_physwindow_ptep = vtopte(phys_window);
+       }
+}
+#endif /* NCOPY_WINDOWS > 0 */
+
+/*
+ * Allocate a new interrupt stack for the boot processor from the
+ * heap rather than continue to use the statically allocated space.
+ * Also switch to a dynamically allocated cpu data area.
+ */
+void
+cpu_data_realloc(void)
+{
+       int             ret;
+       vm_offset_t     istk;
+       cpu_data_t      *cdp;
+       boolean_t       istate;
+
+       ret = kmem_alloc(kernel_map, &istk, INTSTACK_SIZE, VM_KERN_MEMORY_CPU);
+       if (ret != KERN_SUCCESS) {
+               panic("cpu_data_realloc() stack alloc, ret=%d\n", ret);
+       }
+       bzero((void*) istk, INTSTACK_SIZE);
+       istk += INTSTACK_SIZE;
+
+       cdp = &scdatas[0];
+
+       /* Copy old contents into new area and make fix-ups */
+       assert(cpu_number() == 0);
+       bcopy((void *) cpu_data_ptr[0], (void*) cdp, sizeof(cpu_data_t));
+       cdp->cpu_this = cdp;
+       cdp->cpu_int_stack_top = istk;
+       timer_call_queue_init(&cdp->rtclock_timer.queue);
+       cdp->cpu_desc_tablep = (struct cpu_desc_table *) &scdtables[0];
+       cpu_desc_table64_t      *cdt = (cpu_desc_table64_t *) cdp->cpu_desc_tablep;
+
+       uint8_t *cfstk = &scfstks[cdp->cpu_number].fstk[0];
+       cdt->fstkp = cfstk;
+       cfstk += FSTK_SZ;
+
+       /*
+        * With interrupts disabled commmit the new areas.
+        */
+       istate = ml_set_interrupts_enabled(FALSE);
+       cpu_data_ptr[0] = cdp;
+       master_ktss64.ist2 = DBLMAP((uintptr_t) cfstk);
+       master_ktss64.ist1 = DBLMAP((uintptr_t) cfstk - sizeof(x86_64_intr_stack_frame_t));
+       wrmsr64(MSR_IA32_GS_BASE, (uintptr_t) cdp);
+       wrmsr64(MSR_IA32_KERNEL_GS_BASE, (uintptr_t) cdp);
+       (void) ml_set_interrupts_enabled(istate);
+
+       kprintf("Reallocated master cpu data: %p,"
+               " interrupt stack: %p, fault stack: %p\n",
+               (void *) cdp, (void *) istk, (void *) cfstk);
+}