]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/i386/pcb.c
xnu-6153.11.26.tar.gz
[apple/xnu.git] / osfmk / i386 / pcb.c
index caf0c68dbad7aa92530c5a86d6de32f70e6d6418..9ece881bd134066069eb2a935eed3b78757fe424 100644 (file)
@@ -1,8 +1,8 @@
 /*
- * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2016 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@
  */
-/* 
+/*
  * Mach Operating System
  * Copyright (c) 1991,1990 Carnegie Mellon University
  * All Rights Reserved.
- * 
+ *
  * Permission to use, copy, modify and distribute this software and its
  * documentation is hereby granted, provided that both the copyright
  * notice and this permission notice appear in all copies of the
  * software, derivative works or modified versions, and any portions
  * thereof, and that both notices appear in supporting documentation.
- * 
+ *
  * CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS"
  * CONDITION.  CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR
  * ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE.
- * 
+ *
  * Carnegie Mellon requests users of this software to return to
- * 
+ *
  *  Software Distribution Coordinator  or  Software.Distribution@CS.CMU.EDU
  *  School of Computer Science
  *  Carnegie Mellon University
  *  Pittsburgh PA 15213-3890
- * 
+ *
  * any improvements or extensions that they make and grant Carnegie Mellon
  * the rights to redistribute these changes.
  */
 
-#include <mach_rt.h>
 #include <mach_debug.h>
 #include <mach_ldebug.h>
 
@@ -77,6 +76,7 @@
 #include <kern/assert.h>
 #include <kern/spl.h>
 #include <kern/machine.h>
+#include <kern/kpc.h>
 #include <ipc/ipc_port.h>
 #include <vm/vm_kern.h>
 #include <vm/vm_map.h>
 #include <i386/misc_protos.h>
 #include <i386/mp_desc.h>
 #include <i386/thread.h>
-#if defined(__i386__)
-#include <i386/fpu.h>
-#endif
 #include <i386/machine_routines.h>
 #include <i386/lapic.h> /* LAPIC_PMC_SWI_VECTOR */
+#include <i386/seg.h>
 
-#if CONFIG_COUNTERS
-#include <pmc/pmc.h>
-#endif /* CONFIG_COUNTERS */
+#if HYPERVISOR
+#include <kern/hv_support.h>
+#endif
 
 /*
  * Maps state flavor to number of words in the state:
  */
 unsigned int _MachineStateCount[] = {
-       /* FLAVOR_LIST */
-        0,
-       x86_THREAD_STATE32_COUNT,
-       x86_FLOAT_STATE32_COUNT,
-       x86_EXCEPTION_STATE32_COUNT,
-       x86_THREAD_STATE64_COUNT,
-       x86_FLOAT_STATE64_COUNT,
-       x86_EXCEPTION_STATE64_COUNT,
-       x86_THREAD_STATE_COUNT,
-       x86_FLOAT_STATE_COUNT,
-       x86_EXCEPTION_STATE_COUNT,
-       0,
-       x86_SAVED_STATE32_COUNT,
-       x86_SAVED_STATE64_COUNT,
-       x86_DEBUG_STATE32_COUNT,
-       x86_DEBUG_STATE64_COUNT,
-       x86_DEBUG_STATE_COUNT
+       [x86_THREAD_STATE32]            = x86_THREAD_STATE32_COUNT,
+       [x86_THREAD_STATE64]            = x86_THREAD_STATE64_COUNT,
+       [x86_THREAD_FULL_STATE64]       = x86_THREAD_FULL_STATE64_COUNT,
+       [x86_THREAD_STATE]              = x86_THREAD_STATE_COUNT,
+       [x86_FLOAT_STATE32]             = x86_FLOAT_STATE32_COUNT,
+       [x86_FLOAT_STATE64]             = x86_FLOAT_STATE64_COUNT,
+       [x86_FLOAT_STATE]               = x86_FLOAT_STATE_COUNT,
+       [x86_EXCEPTION_STATE32]         = x86_EXCEPTION_STATE32_COUNT,
+       [x86_EXCEPTION_STATE64]         = x86_EXCEPTION_STATE64_COUNT,
+       [x86_EXCEPTION_STATE]           = x86_EXCEPTION_STATE_COUNT,
+       [x86_DEBUG_STATE32]             = x86_DEBUG_STATE32_COUNT,
+       [x86_DEBUG_STATE64]             = x86_DEBUG_STATE64_COUNT,
+       [x86_DEBUG_STATE]               = x86_DEBUG_STATE_COUNT,
+       [x86_AVX_STATE32]               = x86_AVX_STATE32_COUNT,
+       [x86_AVX_STATE64]               = x86_AVX_STATE64_COUNT,
+       [x86_AVX_STATE]                 = x86_AVX_STATE_COUNT,
+       [x86_AVX512_STATE32]            = x86_AVX512_STATE32_COUNT,
+       [x86_AVX512_STATE64]            = x86_AVX512_STATE64_COUNT,
+       [x86_AVX512_STATE]              = x86_AVX512_STATE_COUNT,
+       [x86_PAGEIN_STATE]              = x86_PAGEIN_STATE_COUNT
 };
 
-zone_t         iss_zone;               /* zone for saved_state area */
-zone_t         ids_zone;               /* zone for debug_state area */
+zone_t          iss_zone;               /* zone for saved_state area */
+zone_t          ids_zone;               /* zone for debug_state area */
 
 /* Forward */
 
-extern void            Thread_continue(void);
-extern void            Load_context(
-                               thread_t                        thread);
+extern void             Thread_continue(void);
+extern void             Load_context(
+       thread_t                        thread) __attribute__((noreturn));
 
 static void
 get_exception_state32(thread_t thread, x86_exception_state32_t *es);
@@ -143,65 +144,36 @@ static void
 get_thread_state32(thread_t thread, x86_thread_state32_t *ts);
 
 static void
-get_thread_state64(thread_t thread, x86_thread_state64_t *ts);
+get_thread_state64(thread_t thread, void *ts, boolean_t full);
 
 static int
 set_thread_state32(thread_t thread, x86_thread_state32_t *ts);
 
 static int
-set_thread_state64(thread_t thread, x86_thread_state64_t *ts);
-
-#if CONFIG_COUNTERS
-static inline void
-machine_pmc_cswitch(thread_t /* old */, thread_t /* new */);
-
-static inline boolean_t
-machine_thread_pmc_eligible(thread_t);
+set_thread_state64(thread_t thread, void *ts, boolean_t full);
 
+#if HYPERVISOR
 static inline void
-pmc_swi(thread_t /* old */, thread_t /*new */);
-
-static inline boolean_t
-machine_thread_pmc_eligible(thread_t t) {
-       /*
-        * NOTE: Task-level reservations are propagated to child threads via
-        * thread_create_internal.  Any mutation of task reservations forces a
-        * recalculate of t_chud (for the pmc flag) for all threads in that task.
-        * Consequently, we can simply check the current thread's flag against
-        * THREAD_PMC_FLAG.  If the result is non-zero, we SWI for a PMC switch.
-        */
-       return (t != NULL) ? ((t->t_chud & THREAD_PMC_FLAG) ? TRUE : FALSE) : FALSE;
-}
-
-static inline void
-pmc_swi(thread_t old, thread_t new) {
-       current_cpu_datap()->csw_old_thread = old;
-       current_cpu_datap()->csw_new_thread = new;
-       pal_pmc_swi();
-}
-
-static inline void
-machine_pmc_cswitch(thread_t old, thread_t new) {
-       if (machine_thread_pmc_eligible(old) || machine_thread_pmc_eligible(new)) {
-               pmc_swi(old, new);
+ml_hv_cswitch(thread_t old, thread_t new)
+{
+       if (old->hv_thread_target) {
+               hv_callbacks.preempt(old->hv_thread_target);
        }
-}
 
-void ml_get_csw_threads(thread_t *old, thread_t *new) {
-       *old = current_cpu_datap()->csw_old_thread;
-       *new = current_cpu_datap()->csw_new_thread;
+       if (new->hv_thread_target) {
+               hv_callbacks.dispatch(new->hv_thread_target);
+       }
 }
-
-#endif /* CONFIG_COUNTERS */
+#endif
 
 /*
- * Don't let an illegal value for dr7 get set. Specifically,
- * check for undefined settings.  Setting these bit patterns
+ * Don't let an illegal value for the lower 32-bits of dr7 get set.
+ * Specifically, check for undefined settings.  Setting these bit patterns
  * result in undefined behaviour and can lead to an unexpected
  * TRCTRAP.
  */
 static boolean_t
-dr7_is_valid(uint32_t *dr7)
+dr7d_is_valid(uint32_t *dr7d)
 {
        int i;
        uint32_t mask1, mask2;
@@ -210,114 +182,67 @@ dr7_is_valid(uint32_t *dr7)
         * If the DE bit is set in CR4, R/W0-3 can be pattern
         * "10B" to indicate i/o reads and write
         */
-       if (!(get_cr4() & CR4_DE))
-               for (i = 0, mask1 = 0x3<<16, mask2 = 0x2<<16; i < 4; 
-                               i++, mask1 <<= 4, mask2 <<= 4)
-                       if ((*dr7 & mask1) == mask2)
-                               return (FALSE);
-
-       /*
-        * len0-3 pattern "10B" is ok for len on Merom and newer processors
-        * (it signifies an 8-byte wide region). We use the 64bit capability
-        * of the processor in lieu of the more laborious model/family checks
-        * as all 64-bit capable processors so far support this.
-        * Reject an attempt to use this on 64-bit incapable processors.
-        */
-       if (current_cpu_datap()->cpu_is64bit == FALSE)
-               for (i = 0, mask1 = 0x3<<18, mask2 = 0x2<<18; i < 4; 
-                               i++, mask1 <<= 4, mask2 <<= 4)
-                       if ((*dr7 & mask1) == mask2)
-                               return (FALSE);
+       if (!(get_cr4() & CR4_DE)) {
+               for (i = 0, mask1 = 0x3 << 16, mask2 = 0x2 << 16; i < 4;
+                   i++, mask1 <<= 4, mask2 <<= 4) {
+                       if ((*dr7d & mask1) == mask2) {
+                               return FALSE;
+                       }
+               }
+       }
 
        /*
         * if we are doing an instruction execution break (indicated
         * by r/w[x] being "00B"), then the len[x] must also be set
         * to "00B"
         */
-       for (i = 0; i < 4; i++)
-               if (((((*dr7 >> (16 + i*4))) & 0x3) == 0) &&
-                               ((((*dr7 >> (18 + i*4))) & 0x3) != 0))
-                       return (FALSE);
+       for (i = 0; i < 4; i++) {
+               if (((((*dr7d >> (16 + i * 4))) & 0x3) == 0) &&
+                   ((((*dr7d >> (18 + i * 4))) & 0x3) != 0)) {
+                       return FALSE;
+               }
+       }
 
        /*
         * Intel docs have these bits fixed.
         */
-       *dr7 |= 0x1 << 10; /* set bit 10 to 1 */
-       *dr7 &= ~(0x1 << 11); /* set bit 11 to 0 */
-       *dr7 &= ~(0x1 << 12); /* set bit 12 to 0 */
-       *dr7 &= ~(0x1 << 14); /* set bit 14 to 0 */
-       *dr7 &= ~(0x1 << 15); /* set bit 15 to 0 */
+       *dr7d |= 0x1 << 10; /* set bit 10 to 1 */
+       *dr7d &= ~(0x1 << 11); /* set bit 11 to 0 */
+       *dr7d &= ~(0x1 << 12); /* set bit 12 to 0 */
+       *dr7d &= ~(0x1 << 14); /* set bit 14 to 0 */
+       *dr7d &= ~(0x1 << 15); /* set bit 15 to 0 */
 
        /*
         * We don't allow anything to set the global breakpoints.
         */
 
-       if (*dr7 & 0x2)
-               return (FALSE);
-
-       if (*dr7 & (0x2<<2))
-               return (FALSE);
+       if (*dr7d & 0x2) {
+               return FALSE;
+       }
 
-       if (*dr7 & (0x2<<4))
-               return (FALSE);
+       if (*dr7d & (0x2 << 2)) {
+               return FALSE;
+       }
 
-       if (*dr7 & (0x2<<6))
-               return (FALSE);
+       if (*dr7d & (0x2 << 4)) {
+               return FALSE;
+       }
 
-       return (TRUE);
-}
+       if (*dr7d & (0x2 << 6)) {
+               return FALSE;
+       }
 
-static inline void
-set_live_debug_state32(cpu_data_t *cdp, x86_debug_state32_t *ds)
-{
-       __asm__ volatile ("movl %0,%%db0" : :"r" (ds->dr0));
-       __asm__ volatile ("movl %0,%%db1" : :"r" (ds->dr1));
-       __asm__ volatile ("movl %0,%%db2" : :"r" (ds->dr2));
-       __asm__ volatile ("movl %0,%%db3" : :"r" (ds->dr3));
-       if (cpu_mode_is64bit())
-               cdp->cpu_dr7 = ds->dr7;
+       return TRUE;
 }
 
 extern void set_64bit_debug_regs(x86_debug_state64_t *ds);
 
-static inline void
-set_live_debug_state64(cpu_data_t *cdp, x86_debug_state64_t *ds)
-{
-       /*
-        * We need to enter 64-bit mode in order to set the full
-        * width of these registers
-        */
-       set_64bit_debug_regs(ds);
-       cdp->cpu_dr7 = ds->dr7;
-}
-
 boolean_t
-debug_state_is_valid32(x86_debug_state32_t *ds) 
+debug_state_is_valid32(x86_debug_state32_t *ds)
 {
-       if (!dr7_is_valid(&ds->dr7))
+       if (!dr7d_is_valid(&ds->dr7)) {
                return FALSE;
-
-#if defined(__i386__)
-       /*
-        * Only allow local breakpoints and make sure they are not
-        * in the trampoline code.
-        */
-       if (ds->dr7 & 0x1)
-               if (ds->dr0 >= (unsigned long)HIGH_MEM_BASE)
-                       return FALSE;
-
-       if (ds->dr7 & (0x1<<2))
-               if (ds->dr1 >= (unsigned long)HIGH_MEM_BASE)
-                       return FALSE;
-
-       if (ds->dr7 & (0x1<<4))
-               if (ds->dr2 >= (unsigned long)HIGH_MEM_BASE)
-                       return FALSE;
-
-       if (ds->dr7 & (0x1<<6))
-               if (ds->dr3 >= (unsigned long)HIGH_MEM_BASE)
-                       return FALSE;
-#endif
+       }
 
        return TRUE;
 }
@@ -325,28 +250,40 @@ debug_state_is_valid32(x86_debug_state32_t *ds)
 boolean_t
 debug_state_is_valid64(x86_debug_state64_t *ds)
 {
-       if (!dr7_is_valid((uint32_t *)&ds->dr7))
+       if (!dr7d_is_valid((uint32_t *)&ds->dr7)) {
                return FALSE;
+       }
 
        /*
         * Don't allow the user to set debug addresses above their max
         * value
         */
-       if (ds->dr7 & 0x1)
-               if (ds->dr0 >= VM_MAX_PAGE_ADDRESS)
+       if (ds->dr7 & 0x1) {
+               if (ds->dr0 >= VM_MAX_PAGE_ADDRESS) {
                        return FALSE;
+               }
+       }
 
-       if (ds->dr7 & (0x1<<2))
-               if (ds->dr1 >= VM_MAX_PAGE_ADDRESS)
+       if (ds->dr7 & (0x1 << 2)) {
+               if (ds->dr1 >= VM_MAX_PAGE_ADDRESS) {
                        return FALSE;
+               }
+       }
 
-       if (ds->dr7 & (0x1<<4))
-               if (ds->dr2 >= VM_MAX_PAGE_ADDRESS)
+       if (ds->dr7 & (0x1 << 4)) {
+               if (ds->dr2 >= VM_MAX_PAGE_ADDRESS) {
                        return FALSE;
+               }
+       }
 
-       if (ds->dr7 & (0x1<<6))
-               if (ds->dr3 >= VM_MAX_PAGE_ADDRESS)
+       if (ds->dr7 & (0x1 << 6)) {
+               if (ds->dr3 >= VM_MAX_PAGE_ADDRESS) {
                        return FALSE;
+               }
+       }
+
+       /* For x86-64, we must ensure the upper 32-bits of DR7 are clear */
+       ds->dr7 &= 0xffffffffULL;
 
        return TRUE;
 }
@@ -355,68 +292,73 @@ debug_state_is_valid64(x86_debug_state64_t *ds)
 static kern_return_t
 set_debug_state32(thread_t thread, x86_debug_state32_t *ds)
 {
-       x86_debug_state32_t *ids;
+       x86_debug_state32_t *new_ids;
        pcb_t pcb;
 
        pcb = THREAD_TO_PCB(thread);
-       ids = pcb->ids;
 
        if (debug_state_is_valid32(ds) != TRUE) {
                return KERN_INVALID_ARGUMENT;
        }
 
-       if (ids == NULL) {
-               ids = zalloc(ids_zone);
-               bzero(ids, sizeof *ids);
+       if (pcb->ids == NULL) {
+               new_ids = zalloc(ids_zone);
+               bzero(new_ids, sizeof *new_ids);
 
-               simple_lock(&pcb->lock);
+               simple_lock(&pcb->lock, LCK_GRP_NULL);
                /* make sure it wasn't already alloc()'d elsewhere */
                if (pcb->ids == NULL) {
-                       pcb->ids = ids;
+                       pcb->ids = new_ids;
                        simple_unlock(&pcb->lock);
                } else {
                        simple_unlock(&pcb->lock);
-                       zfree(ids_zone, ids);
+                       zfree(ids_zone, new_ids);
                }
        }
 
 
-       copy_debug_state32(ds, ids, FALSE);
+       copy_debug_state32(ds, pcb->ids, FALSE);
 
-       return (KERN_SUCCESS);
+       return KERN_SUCCESS;
 }
 
 static kern_return_t
 set_debug_state64(thread_t thread, x86_debug_state64_t *ds)
 {
-       x86_debug_state64_t *ids;
+       x86_debug_state64_t *new_ids;
        pcb_t pcb;
 
        pcb = THREAD_TO_PCB(thread);
-       ids = pcb->ids;
 
        if (debug_state_is_valid64(ds) != TRUE) {
                return KERN_INVALID_ARGUMENT;
        }
 
-       if (ids == NULL) {
-               ids = zalloc(ids_zone);
-               bzero(ids, sizeof *ids);
+       if (pcb->ids == NULL) {
+               new_ids = zalloc(ids_zone);
+               bzero(new_ids, sizeof *new_ids);
+
+#if HYPERVISOR
+               if (thread->hv_thread_target) {
+                       hv_callbacks.volatile_state(thread->hv_thread_target,
+                           HV_DEBUG_STATE);
+               }
+#endif
 
-               simple_lock(&pcb->lock);
+               simple_lock(&pcb->lock, LCK_GRP_NULL);
                /* make sure it wasn't already alloc()'d elsewhere */
                if (pcb->ids == NULL) {
-                       pcb->ids = ids;
+                       pcb->ids = new_ids;
                        simple_unlock(&pcb->lock);
                } else {
                        simple_unlock(&pcb->lock);
-                       zfree(ids_zone, ids);
+                       zfree(ids_zone, new_ids);
                }
        }
 
-       copy_debug_state64(ds, ids, FALSE);
+       copy_debug_state64(ds, pcb->ids, FALSE);
 
-       return (KERN_SUCCESS);
+       return KERN_SUCCESS;
 }
 
 static void
@@ -428,8 +370,9 @@ get_debug_state32(thread_t thread, x86_debug_state32_t *ds)
 
        if (saved_state) {
                copy_debug_state32(saved_state, ds, TRUE);
-       } else
+       } else {
                bzero(ds, sizeof *ds);
+       }
 }
 
 static void
@@ -441,8 +384,9 @@ get_debug_state64(thread_t thread, x86_debug_state64_t *ds)
 
        if (saved_state) {
                copy_debug_state64(saved_state, ds, TRUE);
-       } else
+       } else {
                bzero(ds, sizeof *ds);
+       }
 }
 
 /*
@@ -465,16 +409,24 @@ consider_machine_adjust(void)
  */
 void
 machine_load_context(
-       thread_t                new)
+       thread_t                new)
 {
-#if CONFIG_COUNTERS
-       machine_pmc_cswitch(NULL, new);
-#endif
        new->machine.specFlags |= OnProc;
        act_machine_switch_pcb(NULL, new);
        Load_context(new);
 }
 
+static inline void
+pmap_switch_context(thread_t ot, thread_t nt, int cnum)
+{
+       pmap_assert(ml_get_interrupts_enabled() == FALSE);
+       vm_map_t nmap = nt->map, omap = ot->map;
+       if ((omap != nmap) || (nmap->pmap->pagezero_accessible)) {
+               PMAP_DEACTIVATE_MAP(omap, ot, cnum);
+               PMAP_ACTIVATE_MAP(nmap, nt, cnum);
+       }
+}
+
 /*
  * Switch to a new thread.
  * Save the old thread`s kernel state or continuation,
@@ -482,29 +434,29 @@ machine_load_context(
  */
 thread_t
 machine_switch_context(
-       thread_t                        old,
-       thread_continue_t       continuation,
-       thread_t                        new)
+       thread_t                        old,
+       thread_continue_t       continuation,
+       thread_t                        new)
 {
-#if MACH_RT
-        assert(current_cpu_datap()->cpu_active_stack == old->kernel_stack);
-#endif
-#if CONFIG_COUNTERS
-       machine_pmc_cswitch(old, new);
-#endif
+       assert(current_cpu_datap()->cpu_active_stack == old->kernel_stack);
+
+#if KPC
+       kpc_off_cpu(old);
+#endif /* KPC */
+
        /*
         *      Save FP registers if in use.
         */
-       fpu_save_context(old);
+       fpu_switch_context(old, new);
 
        old->machine.specFlags &= ~OnProc;
        new->machine.specFlags |= OnProc;
 
        /*
-        * Monitor the stack depth and report new max,
+        * Monitor the stack depth and report new max,
         * not worrying about races.
         */
-       vm_offset_t     depth = current_stack_depth();
+       vm_offset_t     depth = current_stack_depth();
        if (depth > kernel_stack_depth_max) {
                kernel_stack_depth_max = depth;
                KERNEL_DEBUG_CONSTANT(
@@ -516,28 +468,38 @@ machine_switch_context(
         *      Switch address maps if need be, even if not switching tasks.
         *      (A server activation may be "borrowing" a client map.)
         */
-       PMAP_SWITCH_CONTEXT(old, new, cpu_number());
+       pmap_switch_context(old, new, cpu_number());
 
        /*
         *      Load the rest of the user state for the new thread
         */
        act_machine_switch_pcb(old, new);
 
-       return(Switch_context(old, continuation, new));
+#if HYPERVISOR
+       ml_hv_cswitch(old, new);
+#endif
+
+       return Switch_context(old, continuation, new);
 }
 
-thread_t        
+boolean_t
+machine_thread_on_core(thread_t thread)
+{
+       return thread->machine.specFlags & OnProc;
+}
+
+thread_t
 machine_processor_shutdown(
-       thread_t        thread,
-       void            (*doshutdown)(processor_t),
-       processor_t     processor)
+       thread_t        thread,
+       void            (*doshutdown)(processor_t),
+       processor_t     processor)
 {
 #if CONFIG_VMX
        vmx_suspend();
 #endif
-       fpu_save_context(thread);
-       PMAP_SWITCH_CONTEXT(thread, processor->idle_thread, cpu_number());
-       return(Shutdown_context(thread, doshutdown, processor));
+       fpu_switch_context(thread, NULL);
+       pmap_switch_context(thread, processor->idle_thread, cpu_number());
+       return Shutdown_context(thread, doshutdown, processor);
 }
 
 
@@ -549,16 +511,17 @@ kern_return_t
 machine_thread_state_initialize(
        thread_t thread)
 {
-    /*
-     * If there's an fpu save area, free it.
-     * The initialized state will then be lazily faulted-in, if required.
-     * And if we're target, re-arm the no-fpu trap.
-     */
+       /*
+        * If there's an fpu save area, free it.
+        * The initialized state will then be lazily faulted-in, if required.
+        * And if we're target, re-arm the no-fpu trap.
+        */
        if (thread->machine.ifps) {
                (void) fpu_set_fxstate(thread, NULL, x86_FLOAT_STATE64);
 
-               if (thread == current_thread())
+               if (thread == current_thread()) {
                        clear_fpu();
+               }
        }
 
        if (thread->machine.ids) {
@@ -566,7 +529,7 @@ machine_thread_state_initialize(
                thread->machine.ids = NULL;
        }
 
-       return  KERN_SUCCESS;
+       return KERN_SUCCESS;
 }
 
 uint32_t
@@ -600,38 +563,38 @@ get_eflags_exportmask(void)
  *                        for either 32bit or 64bit tasks
  */
 
+
 static void
 get_exception_state64(thread_t thread, x86_exception_state64_t *es)
 {
-        x86_saved_state64_t *saved_state;
+       x86_saved_state64_t *saved_state;
 
-        saved_state = USER_REGS64(thread);
+       saved_state = USER_REGS64(thread);
 
        es->trapno = saved_state->isf.trapno;
        es->cpu = saved_state->isf.cpu;
        es->err = (typeof(es->err))saved_state->isf.err;
        es->faultvaddr = saved_state->cr2;
-}              
+}
 
 static void
 get_exception_state32(thread_t thread, x86_exception_state32_t *es)
 {
-        x86_saved_state32_t *saved_state;
+       x86_saved_state32_t *saved_state;
 
-        saved_state = USER_REGS32(thread);
+       saved_state = USER_REGS32(thread);
 
        es->trapno = saved_state->trapno;
        es->cpu = saved_state->cpu;
        es->err = saved_state->err;
        es->faultvaddr = saved_state->cr2;
-}              
+}
 
 
 static int
 set_thread_state32(thread_t thread, x86_thread_state32_t *ts)
 {
-        x86_saved_state32_t    *saved_state;
+       x86_saved_state32_t     *saved_state;
 
        pal_register_cache_state(thread, DIRTY);
 
@@ -641,11 +604,6 @@ set_thread_state32(thread_t thread, x86_thread_state32_t *ts)
         * Scrub segment selector values:
         */
        ts->cs = USER_CS;
-#ifdef __i386__
-       if (ts->ss == 0) ts->ss = USER_DS;
-       if (ts->ds == 0) ts->ds = USER_DS;
-       if (ts->es == 0) ts->es = USER_DS;
-#else /* __x86_64__ */
        /*
         * On a 64 bit kernel, we always override the data segments,
         * as the actual selector numbers have changed. This also
@@ -655,16 +613,19 @@ set_thread_state32(thread_t thread, x86_thread_state32_t *ts)
        ts->ss = USER_DS;
        ts->ds = USER_DS;
        ts->es = USER_DS;
-#endif
+
+       /* Set GS to CTHREAD only if's been established */
+       ts->gs = thread->machine.cthread_self ? USER_CTHREAD : NULL_SEG;
 
        /* Check segment selectors are safe */
        if (!valid_user_segment_selectors(ts->cs,
-                                         ts->ss,
-                                         ts->ds,
-                                         ts->es,
-                                         ts->fs,
-                                         ts->gs))
-               return(KERN_INVALID_ARGUMENT);
+           ts->ss,
+           ts->ds,
+           ts->es,
+           ts->fs,
+           ts->gs)) {
+               return KERN_INVALID_ARGUMENT;
+       }
 
        saved_state->eax = ts->eax;
        saved_state->ebx = ts->ebx;
@@ -688,24 +649,33 @@ set_thread_state32(thread_t thread, x86_thread_state32_t *ts)
         * ensure that the user returns via iret
         * - which is signaled thusly:
         */
-       if ((saved_state->efl & EFL_TF) && saved_state->cs == SYSENTER_CS)
+       if ((saved_state->efl & EFL_TF) && saved_state->cs == SYSENTER_CS) {
                saved_state->cs = SYSENTER_TF_CS;
+       }
 
-       return(KERN_SUCCESS);
+       return KERN_SUCCESS;
 }
 
 static int
-set_thread_state64(thread_t thread, x86_thread_state64_t *ts)
+set_thread_state64(thread_t thread, void *state, int full)
 {
-        x86_saved_state64_t    *saved_state;
+       x86_thread_state64_t *ts;
+       x86_saved_state64_t     *saved_state;
+
+       if (full == TRUE) {
+               ts = &((x86_thread_full_state64_t *)state)->ss64;
+       } else {
+               ts = (x86_thread_state64_t *)state;
+       }
 
        pal_register_cache_state(thread, DIRTY);
 
        saved_state = USER_REGS64(thread);
 
        if (!IS_USERADDR64_CANONICAL(ts->rsp) ||
-           !IS_USERADDR64_CANONICAL(ts->rip))
-               return(KERN_INVALID_ARGUMENT);
+           !IS_USERADDR64_CANONICAL(ts->rip)) {
+               return KERN_INVALID_ARGUMENT;
+       }
 
        saved_state->r8 = ts->r8;
        saved_state->r9 = ts->r9;
@@ -725,11 +695,22 @@ set_thread_state64(thread_t thread, x86_thread_state64_t *ts)
        saved_state->isf.rsp = ts->rsp;
        saved_state->isf.rflags = (ts->rflags & ~EFL_USER_CLEAR) | EFL_USER_SET;
        saved_state->isf.rip = ts->rip;
-       saved_state->isf.cs = USER64_CS;
+
+       if (full == FALSE) {
+               saved_state->isf.cs = USER64_CS;
+       } else {
+               saved_state->isf.cs = ((x86_thread_full_state64_t *)ts)->ss64.cs;
+               saved_state->isf.ss = ((x86_thread_full_state64_t *)ts)->ss;
+               saved_state->ds = (uint32_t)((x86_thread_full_state64_t *)ts)->ds;
+               saved_state->es = (uint32_t)((x86_thread_full_state64_t *)ts)->es;
+               machine_thread_set_tsd_base(thread,
+                   ((x86_thread_full_state64_t *)ts)->gsbase);
+       }
+
        saved_state->fs = (uint32_t)ts->fs;
        saved_state->gs = (uint32_t)ts->gs;
 
-       return(KERN_SUCCESS);
+       return KERN_SUCCESS;
 }
 
 
@@ -737,7 +718,7 @@ set_thread_state64(thread_t thread, x86_thread_state64_t *ts)
 static void
 get_thread_state32(thread_t thread, x86_thread_state32_t *ts)
 {
-        x86_saved_state32_t    *saved_state;
+       x86_saved_state32_t     *saved_state;
 
        pal_register_cache_state(thread, VALID);
 
@@ -763,9 +744,16 @@ get_thread_state32(thread_t thread, x86_thread_state32_t *ts)
 
 
 static void
-get_thread_state64(thread_t thread, x86_thread_state64_t *ts)
+get_thread_state64(thread_t thread, void *state, boolean_t full)
 {
-        x86_saved_state64_t    *saved_state;
+       x86_thread_state64_t    *ts;
+       x86_saved_state64_t     *saved_state;
+
+       if (full == TRUE) {
+               ts = &((x86_thread_full_state64_t *)state)->ss64;
+       } else {
+               ts = (x86_thread_state64_t *)state;
+       }
 
        pal_register_cache_state(thread, VALID);
 
@@ -790,10 +778,59 @@ get_thread_state64(thread_t thread, x86_thread_state64_t *ts)
        ts->rflags = saved_state->isf.rflags;
        ts->rip = saved_state->isf.rip;
        ts->cs = saved_state->isf.cs;
+
+       if (full == TRUE) {
+               ((x86_thread_full_state64_t *)state)->ds = saved_state->ds;
+               ((x86_thread_full_state64_t *)state)->es = saved_state->es;
+               ((x86_thread_full_state64_t *)state)->ss = saved_state->isf.ss;
+               ((x86_thread_full_state64_t *)state)->gsbase =
+                   thread->machine.cthread_self;
+       }
+
        ts->fs = saved_state->fs;
        ts->gs = saved_state->gs;
 }
 
+kern_return_t
+machine_thread_state_convert_to_user(
+       __unused thread_t thread,
+       __unused thread_flavor_t flavor,
+       __unused thread_state_t tstate,
+       __unused mach_msg_type_number_t *count)
+{
+       // No conversion to userspace representation on this platform
+       return KERN_SUCCESS;
+}
+
+kern_return_t
+machine_thread_state_convert_from_user(
+       __unused thread_t thread,
+       __unused thread_flavor_t flavor,
+       __unused thread_state_t tstate,
+       __unused mach_msg_type_number_t count)
+{
+       // No conversion from userspace representation on this platform
+       return KERN_SUCCESS;
+}
+
+kern_return_t
+machine_thread_siguctx_pointer_convert_to_user(
+       __unused thread_t thread,
+       __unused user_addr_t *uctxp)
+{
+       // No conversion to userspace representation on this platform
+       return KERN_SUCCESS;
+}
+
+kern_return_t
+machine_thread_function_pointers_convert_from_user(
+       __unused thread_t thread,
+       __unused user_addr_t *fptrs,
+       __unused uint32_t count)
+{
+       // No conversion from userspace representation on this platform
+       return KERN_SUCCESS;
+}
 
 /*
  *     act_machine_set_state:
@@ -811,25 +848,32 @@ machine_thread_set_state(
        switch (flavor) {
        case x86_SAVED_STATE32:
        {
-               x86_saved_state32_t     *state;
-               x86_saved_state32_t     *saved_state;
+               x86_saved_state32_t     *state;
+               x86_saved_state32_t     *saved_state;
 
-               if (count < x86_SAVED_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
-        
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (count < x86_SAVED_STATE32_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_saved_state32_t *) tstate;
 
+               /*
+                * Refuse to allow 64-bit processes to set
+                * 32-bit state.
+                */
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
                /* Check segment selectors are safe */
                if (!valid_user_segment_selectors(state->cs,
-                                       state->ss,
-                                       state->ds,
-                                       state->es,
-                                       state->fs,
-                                       state->gs))
+                   state->ss,
+                   state->ds,
+                   state->es,
+                   state->fs,
+                   state->gs)) {
                        return KERN_INVALID_ARGUMENT;
+               }
 
                pal_register_cache_state(thr_act, DIRTY);
 
@@ -855,8 +899,9 @@ machine_thread_set_state(
                 * ensure that the user returns via iret
                 * - which is signaled thusly:
                 */
-               if ((saved_state->efl & EFL_TF) && state->cs == SYSENTER_CS)
+               if ((saved_state->efl & EFL_TF) && state->cs == SYSENTER_CS) {
                        state->cs = SYSENTER_TF_CS;
+               }
 
                /*
                 * User setting segment registers.
@@ -876,14 +921,16 @@ machine_thread_set_state(
 
        case x86_SAVED_STATE64:
        {
-               x86_saved_state64_t     *state;
-               x86_saved_state64_t     *saved_state;
+               x86_saved_state64_t     *state;
+               x86_saved_state64_t     *saved_state;
 
-               if (count < x86_SAVED_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (count < x86_SAVED_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (!thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_saved_state64_t *) tstate;
 
@@ -894,13 +941,15 @@ machine_thread_set_state(
                 * restore the segment registers--hence they are no
                 * longer relevant for validation.
                 */
-               if (!valid_user_code_selector(state->isf.cs))
-                       return KERN_INVALID_ARGUMENT;
-               
+               if (!valid_user_code_selector(state->isf.cs)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
                /* Check pc and stack are canonical addresses */
                if (!IS_USERADDR64_CANONICAL(state->isf.rsp) ||
-                   !IS_USERADDR64_CANONICAL(state->isf.rip))
+                   !IS_USERADDR64_CANONICAL(state->isf.rip)) {
                        return KERN_INVALID_ARGUMENT;
+               }
 
                pal_register_cache_state(thr_act, DIRTY);
 
@@ -944,119 +993,163 @@ machine_thread_set_state(
        }
 
        case x86_FLOAT_STATE32:
+       case x86_AVX_STATE32:
+       case x86_AVX512_STATE32:
        {
-               if (count != x86_FLOAT_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (count != _MachineStateCount[flavor]) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                return fpu_set_fxstate(thr_act, tstate, flavor);
        }
 
        case x86_FLOAT_STATE64:
+       case x86_AVX_STATE64:
+       case x86_AVX512_STATE64:
        {
-               if (count != x86_FLOAT_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (count != _MachineStateCount[flavor]) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if ( !thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                return fpu_set_fxstate(thr_act, tstate, flavor);
        }
 
        case x86_FLOAT_STATE:
-       {   
+       {
                x86_float_state_t       *state;
 
-               if (count != x86_FLOAT_STATE_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (count != x86_FLOAT_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_float_state_t *)tstate;
                if (state->fsh.flavor == x86_FLOAT_STATE64 && state->fsh.count == x86_FLOAT_STATE64_COUNT &&
-                   thread_is_64bit(thr_act)) {
+                   thread_is_64bit_addr(thr_act)) {
                        return fpu_set_fxstate(thr_act, (thread_state_t)&state->ufs.fs64, x86_FLOAT_STATE64);
                }
                if (state->fsh.flavor == x86_FLOAT_STATE32 && state->fsh.count == x86_FLOAT_STATE32_COUNT &&
-                   !thread_is_64bit(thr_act)) {
-                       return fpu_set_fxstate(thr_act, (thread_state_t)&state->ufs.fs32, x86_FLOAT_STATE32); 
+                   !thread_is_64bit_addr(thr_act)) {
+                       return fpu_set_fxstate(thr_act, (thread_state_t)&state->ufs.fs32, x86_FLOAT_STATE32);
                }
-               return(KERN_INVALID_ARGUMENT);
+               return KERN_INVALID_ARGUMENT;
        }
 
-       case x86_AVX_STATE32:
+       case x86_AVX_STATE:
+       case x86_AVX512_STATE:
        {
-               if (count != x86_AVX_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               x86_avx_state_t       *state;
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (count != _MachineStateCount[flavor]) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               return fpu_set_fxstate(thr_act, tstate, flavor);
+               state = (x86_avx_state_t *)tstate;
+               /* Flavors are defined to have sequential values: 32-bit, 64-bit, non-specific */
+               /* 64-bit flavor? */
+               if (state->ash.flavor == (flavor - 1) &&
+                   state->ash.count == _MachineStateCount[flavor - 1] &&
+                   thread_is_64bit_addr(thr_act)) {
+                       return fpu_set_fxstate(thr_act,
+                                  (thread_state_t)&state->ufs.as64,
+                                  flavor - 1);
+               }
+               /* 32-bit flavor? */
+               if (state->ash.flavor == (flavor - 2) &&
+                   state->ash.count == _MachineStateCount[flavor - 2] &&
+                   !thread_is_64bit_addr(thr_act)) {
+                       return fpu_set_fxstate(thr_act,
+                                  (thread_state_t)&state->ufs.as32,
+                                  flavor - 2);
+               }
+               return KERN_INVALID_ARGUMENT;
        }
 
-       case x86_AVX_STATE64:
+       case x86_THREAD_STATE32:
        {
-               if (count != x86_AVX_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (count != x86_THREAD_STATE32_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (!thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               return fpu_set_fxstate(thr_act, tstate, flavor);
+               return set_thread_state32(thr_act, (x86_thread_state32_t *)tstate);
        }
 
-       case x86_THREAD_STATE32: 
+       case x86_THREAD_STATE64:
        {
-               if (count != x86_THREAD_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (count != x86_THREAD_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               return set_thread_state32(thr_act, (x86_thread_state32_t *)tstate);
+               return set_thread_state64(thr_act, tstate, FALSE);
        }
 
-       case x86_THREAD_STATE64: 
+       case x86_THREAD_FULL_STATE64:
        {
-               if (count != x86_THREAD_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (count != x86_THREAD_FULL_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (!thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               return set_thread_state64(thr_act, (x86_thread_state64_t *)tstate);
+               /* If this process does not have a custom LDT, return failure */
+               if (thr_act->task->i386_ldt == 0) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
+               return set_thread_state64(thr_act, tstate, TRUE);
        }
+
        case x86_THREAD_STATE:
        {
                x86_thread_state_t      *state;
 
-               if (count != x86_THREAD_STATE_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (count != x86_THREAD_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_thread_state_t *)tstate;
 
                if (state->tsh.flavor == x86_THREAD_STATE64 &&
                    state->tsh.count == x86_THREAD_STATE64_COUNT &&
-                   thread_is_64bit(thr_act)) {
-                       return set_thread_state64(thr_act, &state->uts.ts64);
+                   thread_is_64bit_addr(thr_act)) {
+                       return set_thread_state64(thr_act, &state->uts.ts64, FALSE);
+               } else if (state->tsh.flavor == x86_THREAD_FULL_STATE64 &&
+                   state->tsh.count == x86_THREAD_FULL_STATE64_COUNT &&
+                   thread_is_64bit_addr(thr_act) && thr_act->task->i386_ldt != 0) {
+                       return set_thread_state64(thr_act, &state->uts.ts64, TRUE);
                } else if (state->tsh.flavor == x86_THREAD_STATE32 &&
-                          state->tsh.count == x86_THREAD_STATE32_COUNT &&
-                          !thread_is_64bit(thr_act)) {
+                   state->tsh.count == x86_THREAD_STATE32_COUNT &&
+                   !thread_is_64bit_addr(thr_act)) {
                        return set_thread_state32(thr_act, &state->uts.ts32);
-               } else
-                       return(KERN_INVALID_ARGUMENT);
-
-               break;
+               } else {
+                       return KERN_INVALID_ARGUMENT;
+               }
        }
        case x86_DEBUG_STATE32:
        {
                x86_debug_state32_t *state;
                kern_return_t ret;
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_debug_state32_t *)tstate;
 
@@ -1069,8 +1162,9 @@ machine_thread_set_state(
                x86_debug_state64_t *state;
                kern_return_t ret;
 
-               if (!thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_debug_state64_t *)tstate;
 
@@ -1083,30 +1177,53 @@ machine_thread_set_state(
                x86_debug_state_t *state;
                kern_return_t ret = KERN_INVALID_ARGUMENT;
 
-               if (count != x86_DEBUG_STATE_COUNT)
-                       return (KERN_INVALID_ARGUMENT);
+               if (count != x86_DEBUG_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_debug_state_t *)tstate;
                if (state->dsh.flavor == x86_DEBUG_STATE64 &&
-                               state->dsh.count == x86_DEBUG_STATE64_COUNT &&
-                               thread_is_64bit(thr_act)) {
+                   state->dsh.count == x86_DEBUG_STATE64_COUNT &&
+                   thread_is_64bit_addr(thr_act)) {
                        ret = set_debug_state64(thr_act, &state->uds.ds64);
-               }
-               else
-                       if (state->dsh.flavor == x86_DEBUG_STATE32 &&
-                           state->dsh.count == x86_DEBUG_STATE32_COUNT &&
-                           !thread_is_64bit(thr_act)) {
-                               ret = set_debug_state32(thr_act, &state->uds.ds32);
+               } else if (state->dsh.flavor == x86_DEBUG_STATE32 &&
+                   state->dsh.count == x86_DEBUG_STATE32_COUNT &&
+                   !thread_is_64bit_addr(thr_act)) {
+                       ret = set_debug_state32(thr_act, &state->uds.ds32);
                }
                return ret;
        }
        default:
-               return(KERN_INVALID_ARGUMENT);
+               return KERN_INVALID_ARGUMENT;
        }
 
-       return(KERN_SUCCESS);
+       return KERN_SUCCESS;
 }
 
+mach_vm_address_t
+machine_thread_pc(thread_t thr_act)
+{
+       if (thread_is_64bit_addr(thr_act)) {
+               return (mach_vm_address_t)USER_REGS64(thr_act)->isf.rip;
+       } else {
+               return (mach_vm_address_t)USER_REGS32(thr_act)->eip;
+       }
+}
+
+void
+machine_thread_reset_pc(thread_t thr_act, mach_vm_address_t pc)
+{
+       pal_register_cache_state(thr_act, DIRTY);
+
+       if (thread_is_64bit_addr(thr_act)) {
+               if (!IS_USERADDR64_CANONICAL(pc)) {
+                       pc = 0;
+               }
+               USER_REGS64(thr_act)->isf.rip = (uint64_t)pc;
+       } else {
+               USER_REGS32(thr_act)->eip = (uint32_t)pc;
+       }
+}
 
 
 /*
@@ -1122,46 +1239,99 @@ machine_thread_get_state(
        thread_state_t tstate,
        mach_msg_type_number_t *count)
 {
+       switch (flavor) {
+       case THREAD_STATE_FLAVOR_LIST:
+       {
+               if (*count < 3) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-       switch (flavor)  {
-
-           case THREAD_STATE_FLAVOR_LIST:
-           {
-               if (*count < 3)
-                       return (KERN_INVALID_ARGUMENT);
-
-               tstate[0] = i386_THREAD_STATE;
+               tstate[0] = i386_THREAD_STATE;
                tstate[1] = i386_FLOAT_STATE;
                tstate[2] = i386_EXCEPTION_STATE;
 
                *count = 3;
                break;
-           }
+       }
 
-           case THREAD_STATE_FLAVOR_LIST_NEW:
-           {
-               if (*count < 4)
-                       return (KERN_INVALID_ARGUMENT);
+       case THREAD_STATE_FLAVOR_LIST_NEW:
+       {
+               if (*count < 4) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               tstate[0] = x86_THREAD_STATE;
+               tstate[0] = x86_THREAD_STATE;
                tstate[1] = x86_FLOAT_STATE;
                tstate[2] = x86_EXCEPTION_STATE;
                tstate[3] = x86_DEBUG_STATE;
 
                *count = 4;
                break;
-           }
+       }
 
-           case x86_SAVED_STATE32:
-           {
-               x86_saved_state32_t     *state;
-               x86_saved_state32_t     *saved_state;
+       case THREAD_STATE_FLAVOR_LIST_10_9:
+       {
+               if (*count < 5) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               tstate[0] = x86_THREAD_STATE;
+               tstate[1] = x86_FLOAT_STATE;
+               tstate[2] = x86_EXCEPTION_STATE;
+               tstate[3] = x86_DEBUG_STATE;
+               tstate[4] = x86_AVX_STATE;
+
+               *count = 5;
+               break;
+       }
+
+       case THREAD_STATE_FLAVOR_LIST_10_13:
+       {
+               if (*count < 6) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               tstate[0] = x86_THREAD_STATE;
+               tstate[1] = x86_FLOAT_STATE;
+               tstate[2] = x86_EXCEPTION_STATE;
+               tstate[3] = x86_DEBUG_STATE;
+               tstate[4] = x86_AVX_STATE;
+               tstate[5] = x86_AVX512_STATE;
+
+               *count = 6;
+               break;
+       }
+
+       case THREAD_STATE_FLAVOR_LIST_10_15:
+       {
+               if (*count < 7) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               tstate[0] = x86_THREAD_STATE;
+               tstate[1] = x86_FLOAT_STATE;
+               tstate[2] = x86_EXCEPTION_STATE;
+               tstate[3] = x86_DEBUG_STATE;
+               tstate[4] = x86_AVX_STATE;
+               tstate[5] = x86_AVX512_STATE;
+               tstate[6] = x86_PAGEIN_STATE;
+
+               *count = 7;
+               break;
+       }
 
-               if (*count < x86_SAVED_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+       case x86_SAVED_STATE32:
+       {
+               x86_saved_state32_t     *state;
+               x86_saved_state32_t     *saved_state;
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count < x86_SAVED_STATE32_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_saved_state32_t *) tstate;
                saved_state = USER_REGS32(thr_act);
@@ -1177,18 +1347,20 @@ machine_thread_get_state(
 
                *count = x86_SAVED_STATE32_COUNT;
                break;
-           }
+       }
 
-           case x86_SAVED_STATE64:
-           {
-               x86_saved_state64_t     *state;
-               x86_saved_state64_t     *saved_state;
+       case x86_SAVED_STATE64:
+       {
+               x86_saved_state64_t     *state;
+               x86_saved_state64_t     *saved_state;
 
-               if (*count < x86_SAVED_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count < x86_SAVED_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (!thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_saved_state64_t *)tstate;
                saved_state = USER_REGS64(thr_act);
@@ -1197,158 +1369,227 @@ machine_thread_get_state(
                 * First, copy everything:
                 */
                *state = *saved_state;
+               state->ds = saved_state->ds & 0xffff;
+               state->es = saved_state->es & 0xffff;
                state->fs = saved_state->fs & 0xffff;
                state->gs = saved_state->gs & 0xffff;
 
                *count = x86_SAVED_STATE64_COUNT;
                break;
-           }
+       }
 
-           case x86_FLOAT_STATE32:
-           {
-               if (*count < x86_FLOAT_STATE32_COUNT) 
-                       return(KERN_INVALID_ARGUMENT);
+       case x86_FLOAT_STATE32:
+       {
+               if (*count < x86_FLOAT_STATE32_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                *count = x86_FLOAT_STATE32_COUNT;
 
                return fpu_get_fxstate(thr_act, tstate, flavor);
-           }
+       }
 
-           case x86_FLOAT_STATE64:
-           {
-               if (*count < x86_FLOAT_STATE64_COUNT) 
-                       return(KERN_INVALID_ARGUMENT);
+       case x86_FLOAT_STATE64:
+       {
+               if (*count < x86_FLOAT_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if ( !thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                *count = x86_FLOAT_STATE64_COUNT;
 
                return fpu_get_fxstate(thr_act, tstate, flavor);
-           }
+       }
 
-           case x86_FLOAT_STATE:
-           {
-               x86_float_state_t       *state;
-               kern_return_t           kret;
+       case x86_FLOAT_STATE:
+       {
+               x86_float_state_t       *state;
+               kern_return_t           kret;
 
-               if (*count < x86_FLOAT_STATE_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count < x86_FLOAT_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_float_state_t *)tstate;
 
                /*
-                * no need to bzero... currently 
+                * no need to bzero... currently
                 * x86_FLOAT_STATE64_COUNT == x86_FLOAT_STATE32_COUNT
                 */
-               if (thread_is_64bit(thr_act)) {
-                       state->fsh.flavor = x86_FLOAT_STATE64;
-                       state->fsh.count  = x86_FLOAT_STATE64_COUNT;
+               if (thread_is_64bit_addr(thr_act)) {
+                       state->fsh.flavor = x86_FLOAT_STATE64;
+                       state->fsh.count  = x86_FLOAT_STATE64_COUNT;
 
                        kret = fpu_get_fxstate(thr_act, (thread_state_t)&state->ufs.fs64, x86_FLOAT_STATE64);
                } else {
-                       state->fsh.flavor = x86_FLOAT_STATE32;
+                       state->fsh.flavor = x86_FLOAT_STATE32;
                        state->fsh.count  = x86_FLOAT_STATE32_COUNT;
 
                        kret = fpu_get_fxstate(thr_act, (thread_state_t)&state->ufs.fs32, x86_FLOAT_STATE32);
                }
                *count = x86_FLOAT_STATE_COUNT;
 
-               return(kret);
-           }
+               return kret;
+       }
 
        case x86_AVX_STATE32:
+       case x86_AVX512_STATE32:
        {
-               if (*count != x86_AVX_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count != _MachineStateCount[flavor]) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               *count = x86_AVX_STATE32_COUNT;
+               *count = _MachineStateCount[flavor];
 
                return fpu_get_fxstate(thr_act, tstate, flavor);
        }
 
        case x86_AVX_STATE64:
+       case x86_AVX512_STATE64:
        {
-               if (*count != x86_AVX_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count != _MachineStateCount[flavor]) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if ( !thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               *count = x86_AVX_STATE64_COUNT;
+               *count = _MachineStateCount[flavor];
 
                return fpu_get_fxstate(thr_act, tstate, flavor);
        }
 
-           case x86_THREAD_STATE32: 
-           {
-               if (*count < x86_THREAD_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+       case x86_AVX_STATE:
+       case x86_AVX512_STATE:
+       {
+               x86_avx_state_t         *state;
+               thread_state_t          fstate;
+
+               if (*count < _MachineStateCount[flavor]) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               *count = _MachineStateCount[flavor];
+               state = (x86_avx_state_t *)tstate;
+
+               bzero((char *)state, *count * sizeof(int));
+
+               if (thread_is_64bit_addr(thr_act)) {
+                       flavor -= 1;         /* 64-bit flavor */
+                       fstate = (thread_state_t) &state->ufs.as64;
+               } else {
+                       flavor -= 2;         /* 32-bit flavor */
+                       fstate = (thread_state_t) &state->ufs.as32;
+               }
+               state->ash.flavor = flavor;
+               state->ash.count  = _MachineStateCount[flavor];
+
+               return fpu_get_fxstate(thr_act, fstate, flavor);
+       }
+
+       case x86_THREAD_STATE32:
+       {
+               if (*count < x86_THREAD_STATE32_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                *count = x86_THREAD_STATE32_COUNT;
 
                get_thread_state32(thr_act, (x86_thread_state32_t *)tstate);
                break;
-           }
+       }
 
-           case x86_THREAD_STATE64:
-           {
-               if (*count < x86_THREAD_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+       case x86_THREAD_STATE64:
+       {
+               if (*count < x86_THREAD_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-                if ( !thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                *count = x86_THREAD_STATE64_COUNT;
 
-               get_thread_state64(thr_act, (x86_thread_state64_t *)tstate);
+               get_thread_state64(thr_act, tstate, FALSE);
+               break;
+       }
+
+       case x86_THREAD_FULL_STATE64:
+       {
+               if (*count < x86_THREAD_FULL_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               /* If this process does not have a custom LDT, return failure */
+               if (thr_act->task->i386_ldt == 0) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               *count = x86_THREAD_FULL_STATE64_COUNT;
+
+               get_thread_state64(thr_act, tstate, TRUE);
                break;
-           }
+       }
 
-           case x86_THREAD_STATE:
-           {
-               x86_thread_state_t      *state;
+       case x86_THREAD_STATE:
+       {
+               x86_thread_state_t      *state;
 
-               if (*count < x86_THREAD_STATE_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count < x86_THREAD_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_thread_state_t *)tstate;
 
                bzero((char *)state, sizeof(x86_thread_state_t));
 
-               if (thread_is_64bit(thr_act)) {
+               if (thread_is_64bit_addr(thr_act)) {
                        state->tsh.flavor = x86_THREAD_STATE64;
                        state->tsh.count  = x86_THREAD_STATE64_COUNT;
 
-                       get_thread_state64(thr_act, &state->uts.ts64);
+                       get_thread_state64(thr_act, &state->uts.ts64, FALSE);
                } else {
                        state->tsh.flavor = x86_THREAD_STATE32;
                        state->tsh.count  = x86_THREAD_STATE32_COUNT;
 
-                       get_thread_state32(thr_act, &state->uts.ts32);
+                       get_thread_state32(thr_act, &state->uts.ts32);
                }
                *count = x86_THREAD_STATE_COUNT;
 
                break;
-           }
+       }
 
 
-           case x86_EXCEPTION_STATE32:
-           {
-               if (*count < x86_EXCEPTION_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+       case x86_EXCEPTION_STATE32:
+       {
+               if (*count < x86_EXCEPTION_STATE32_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                *count = x86_EXCEPTION_STATE32_COUNT;
 
@@ -1359,15 +1600,17 @@ machine_thread_get_state(
                 */
                ((x86_exception_state32_t *)tstate)->cpu = 0;
                break;
-           }
+       }
 
-           case x86_EXCEPTION_STATE64:
-           {
-               if (*count < x86_EXCEPTION_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+       case x86_EXCEPTION_STATE64:
+       {
+               if (*count < x86_EXCEPTION_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if ( !thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                *count = x86_EXCEPTION_STATE64_COUNT;
 
@@ -1378,29 +1621,30 @@ machine_thread_get_state(
                 */
                ((x86_exception_state64_t *)tstate)->cpu = 0;
                break;
-           }
+       }
 
-           case x86_EXCEPTION_STATE:
-           {
-               x86_exception_state_t   *state;
+       case x86_EXCEPTION_STATE:
+       {
+               x86_exception_state_t   *state;
 
-               if (*count < x86_EXCEPTION_STATE_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count < x86_EXCEPTION_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_exception_state_t *)tstate;
 
                bzero((char *)state, sizeof(x86_exception_state_t));
 
-               if (thread_is_64bit(thr_act)) {
+               if (thread_is_64bit_addr(thr_act)) {
                        state->esh.flavor = x86_EXCEPTION_STATE64;
                        state->esh.count  = x86_EXCEPTION_STATE64_COUNT;
 
-                       get_exception_state64(thr_act, &state->ues.es64);
+                       get_exception_state64(thr_act, &state->ues.es64);
                } else {
                        state->esh.flavor = x86_EXCEPTION_STATE32;
                        state->esh.count  = x86_EXCEPTION_STATE32_COUNT;
 
-                       get_exception_state32(thr_act, &state->ues.es32);
+                       get_exception_state32(thr_act, &state->ues.es32);
                }
                *count = x86_EXCEPTION_STATE_COUNT;
 
@@ -1408,11 +1652,13 @@ machine_thread_get_state(
        }
        case x86_DEBUG_STATE32:
        {
-               if (*count < x86_DEBUG_STATE32_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count < x86_DEBUG_STATE32_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
-               if (thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                get_debug_state32(thr_act, (x86_debug_state32_t *)tstate);
 
@@ -1422,11 +1668,13 @@ machine_thread_get_state(
        }
        case x86_DEBUG_STATE64:
        {
-               if (*count < x86_DEBUG_STATE64_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
-               
-               if (!thread_is_64bit(thr_act))
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count < x86_DEBUG_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               if (!thread_is_64bit_addr(thr_act)) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                get_debug_state64(thr_act, (x86_debug_state64_t *)tstate);
 
@@ -1438,14 +1686,15 @@ machine_thread_get_state(
        {
                x86_debug_state_t   *state;
 
-               if (*count < x86_DEBUG_STATE_COUNT)
-                       return(KERN_INVALID_ARGUMENT);
+               if (*count < x86_DEBUG_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_debug_state_t *)tstate;
 
                bzero(state, sizeof *state);
 
-               if (thread_is_64bit(thr_act)) {
+               if (thread_is_64bit_addr(thr_act)) {
                        state->dsh.flavor = x86_DEBUG_STATE64;
                        state->dsh.count  = x86_DEBUG_STATE64_COUNT;
 
@@ -1459,36 +1708,52 @@ machine_thread_get_state(
                *count = x86_DEBUG_STATE_COUNT;
                break;
        }
+
+       case x86_PAGEIN_STATE:
+       {
+               if (*count < x86_PAGEIN_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
+
+               x86_pagein_state_t *state = (void *)tstate;
+
+               state->__pagein_error = thr_act->t_pagein_error;
+
+               *count = x86_PAGEIN_STATE_COUNT;
+               break;
+       }
        default:
-               return(KERN_INVALID_ARGUMENT);
+               return KERN_INVALID_ARGUMENT;
        }
 
-       return(KERN_SUCCESS);
+       return KERN_SUCCESS;
 }
 
 kern_return_t
 machine_thread_get_kern_state(
-               thread_t                thread,
-               thread_flavor_t         flavor,
-               thread_state_t          tstate,
-               mach_msg_type_number_t  *count)
+       thread_t                thread,
+       thread_flavor_t         flavor,
+       thread_state_t          tstate,
+       mach_msg_type_number_t  *count)
 {
-       x86_saved_state_t       *int_state = current_cpu_datap()->cpu_int_state;
+       x86_saved_state_t       *int_state = current_cpu_datap()->cpu_int_state;
 
        /*
         * This works only for an interrupted kernel thread
         */
-       if (thread != current_thread() || int_state == NULL)
+       if (thread != current_thread() || int_state == NULL) {
                return KERN_FAILURE;
+       }
 
        switch (flavor) {
-           case x86_THREAD_STATE32: {
+       case x86_THREAD_STATE32: {
                x86_thread_state32_t *state;
                x86_saved_state32_t *saved_state;
 
                if (!is_saved_state32(int_state) ||
-                   *count < x86_THREAD_STATE32_COUNT)
-                       return (KERN_INVALID_ARGUMENT);
+                   *count < x86_THREAD_STATE32_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_thread_state32_t *) tstate;
 
@@ -1516,15 +1781,16 @@ machine_thread_get_kern_state(
                *count = x86_THREAD_STATE32_COUNT;
 
                return KERN_SUCCESS;
-           }
-  
-           case x86_THREAD_STATE64: {
-               x86_thread_state64_t    *state;
-               x86_saved_state64_t     *saved_state;
+       }
+
+       case x86_THREAD_STATE64: {
+               x86_thread_state64_t    *state;
+               x86_saved_state64_t     *saved_state;
 
                if (!is_saved_state64(int_state) ||
-                   *count < x86_THREAD_STATE64_COUNT)
-                       return (KERN_INVALID_ARGUMENT);
+                   *count < x86_THREAD_STATE64_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_thread_state64_t *) tstate;
 
@@ -1557,13 +1823,14 @@ machine_thread_get_kern_state(
                *count = x86_THREAD_STATE64_COUNT;
 
                return KERN_SUCCESS;
-           }
-  
-           case x86_THREAD_STATE: {
+       }
+
+       case x86_THREAD_STATE: {
                x86_thread_state_t *state = NULL;
 
-               if (*count < x86_THREAD_STATE_COUNT)
-                       return (KERN_INVALID_ARGUMENT);
+               if (*count < x86_THREAD_STATE_COUNT) {
+                       return KERN_INVALID_ARGUMENT;
+               }
 
                state = (x86_thread_state_t *) tstate;
 
@@ -1629,7 +1896,7 @@ machine_thread_get_kern_state(
 
                *count = x86_THREAD_STATE_COUNT;
                return KERN_SUCCESS;
-           }
+       }
        }
        return KERN_FAILURE;
 }
@@ -1650,13 +1917,12 @@ machine_thread_switch_addrmode(thread_t thread)
         */
        machine_thread_create(thread, thread->task);
 
+       /* Adjust FPU state */
+       fpu_switch_addrmode(thread, task_has_64Bit_addr(thread->task));
+
        /* If we're switching ourselves, reset the pcb addresses etc. */
        if (thread == current_thread()) {
                boolean_t istate = ml_set_interrupts_enabled(FALSE);
-#if defined(__i386__)
-               if (current_cpu_datap()->cpu_active_cr3 != kernel_pmap->pm_cr3)
-                       pmap_load_kernel_cr3();
-#endif /* defined(__i386) */
                act_machine_switch_pcb(NULL, thread);
                ml_set_interrupts_enabled(istate);
        }
@@ -1675,27 +1941,6 @@ machine_set_current_thread(thread_t thread)
        current_cpu_datap()->cpu_active_thread = thread;
 }
 
-/*
- * This is called when a task is terminated, and also on exec().
- * Clear machine-dependent state that is stored on the task.
- */
-void
-machine_thread_terminate_self(void)
-{
-       task_t self_task = current_task();
-       if (self_task) {
-           user_ldt_t user_ldt = self_task->i386_ldt;
-           if (user_ldt != 0) {
-               self_task->i386_ldt = 0;
-               user_ldt_free(user_ldt);
-           }
-
-           if (self_task->task_debug != NULL) {
-               zfree(ids_zone, self_task->task_debug);
-               self_task->task_debug = NULL;
-           }    
-       }
-}
 
 /*
  * Perform machine-dependent per-thread initializations
@@ -1703,139 +1948,38 @@ machine_thread_terminate_self(void)
 void
 machine_thread_init(void)
 {
-       if (cpu_mode_is64bit()) {
-               assert(sizeof(x86_sframe_compat32_t) % 16 == 0);
-               iss_zone = zinit(sizeof(x86_sframe64_t),
-                               thread_max * sizeof(x86_sframe64_t),
-                               THREAD_CHUNK * sizeof(x86_sframe64_t),
-                               "x86_64 saved state");
-
-               ids_zone = zinit(sizeof(x86_debug_state64_t),
-                                thread_max * sizeof(x86_debug_state64_t),
-                                THREAD_CHUNK * sizeof(x86_debug_state64_t),
-                                "x86_64 debug state");
-
-       } else {
-               iss_zone = zinit(sizeof(x86_sframe32_t),
-                               thread_max * sizeof(x86_sframe32_t),
-                               THREAD_CHUNK * sizeof(x86_sframe32_t),
-                               "x86 saved state");
-               ids_zone = zinit(sizeof(x86_debug_state32_t),
-                               thread_max * (sizeof(x86_debug_state32_t)),
-                               THREAD_CHUNK * (sizeof(x86_debug_state32_t)),
-                               "x86 debug state");
-       }
-       fpu_module_init();
-}
-
-
-#if defined(__i386__)
-/*
- * Some routines for debugging activation code
- */
-static void    dump_handlers(thread_t);
-void           dump_regs(thread_t);
-int            dump_act(thread_t thr_act);
-
-static void
-dump_handlers(thread_t thr_act)
-{
-       ReturnHandler *rhp = thr_act->handlers;
-       int     counter = 0;
-
-       printf("\t");
-       while (rhp) {
-               if (rhp == &thr_act->special_handler){
-                       if (rhp->next)
-                               printf("[NON-Zero next ptr(%p)]", rhp->next);
-                       printf("special_handler()->");
-                       break;
-               }
-               printf("hdlr_%d(%p)->", counter, rhp->handler);
-               rhp = rhp->next;
-               if (++counter > 32) {
-                       printf("Aborting: HUGE handler chain\n");
-                       break;
-               }
-       }
-       printf("HLDR_NULL\n");
-}
-
-void
-dump_regs(thread_t thr_act)
-{
-       if (thread_is_64bit(thr_act)) {
-               x86_saved_state64_t     *ssp;
-
-               ssp = USER_REGS64(thr_act);
-
-               panic("dump_regs: 64bit tasks not yet supported");
-
-       } else {
-               x86_saved_state32_t     *ssp;
-
-               ssp = USER_REGS32(thr_act);
+       iss_zone = zinit(sizeof(x86_saved_state_t),
+           thread_max * sizeof(x86_saved_state_t),
+           THREAD_CHUNK * sizeof(x86_saved_state_t),
+           "x86_64 saved state");
 
-               /*
-                * Print out user register state
-                */
-               printf("\tRegs:\tedi=%x esi=%x ebp=%x ebx=%x edx=%x\n",
-                       ssp->edi, ssp->esi, ssp->ebp, ssp->ebx, ssp->edx);
+       ids_zone = zinit(sizeof(x86_debug_state64_t),
+           thread_max * sizeof(x86_debug_state64_t),
+           THREAD_CHUNK * sizeof(x86_debug_state64_t),
+           "x86_64 debug state");
 
-               printf("\t\tecx=%x eax=%x eip=%x efl=%x uesp=%x\n",
-                       ssp->ecx, ssp->eax, ssp->eip, ssp->efl, ssp->uesp);
-
-               printf("\t\tcs=%x ss=%x\n", ssp->cs, ssp->ss);
-       }
+       fpu_module_init();
 }
 
-int
-dump_act(thread_t thr_act)
-{
-       if (!thr_act)
-               return(0);
-
-       printf("thread(%p)(%d): task=%p(%d)\n",
-                       thr_act, thr_act->ref_count,
-                       thr_act->task,
-                       thr_act->task   ? thr_act->task->ref_count : 0);
-
-       printf("\tsusp=%d user_stop=%d active=%x ast=%x\n",
-                       thr_act->suspend_count, thr_act->user_stop_count,
-                       thr_act->active, thr_act->ast);
-       printf("\tpcb=%p\n", &thr_act->machine);
-
-       if (thr_act->kernel_stack) {
-               vm_offset_t stack = thr_act->kernel_stack;
 
-               printf("\tk_stk %lx  eip %x ebx %x esp %x iss %p\n",
-                       (long)stack, STACK_IKS(stack)->k_eip, STACK_IKS(stack)->k_ebx,
-                       STACK_IKS(stack)->k_esp, thr_act->machine.iss);
-       }
-
-       dump_handlers(thr_act);
-       dump_regs(thr_act);
-       return((int)thr_act);
-}
-#endif
 
 user_addr_t
 get_useraddr(void)
 {
-        thread_t thr_act = current_thread();
-        if (thread_is_64bit(thr_act)) {
-               x86_saved_state64_t     *iss64;
-               
+       thread_t thr_act = current_thread();
+
+       if (thread_is_64bit_addr(thr_act)) {
+               x86_saved_state64_t     *iss64;
+
                iss64 = USER_REGS64(thr_act);
 
-               return(iss64->isf.rip);
+               return iss64->isf.rip;
        } else {
-               x86_saved_state32_t     *iss32;
+               x86_saved_state32_t     *iss32;
 
                iss32 = USER_REGS32(thr_act);
 
-               return(iss32->eip);
+               return iss32->eip;
        }
 }
 
@@ -1849,14 +1993,14 @@ machine_stack_detach(thread_t thread)
        vm_offset_t     stack;
 
        KERNEL_DEBUG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_DETACH),
-                    (uintptr_t)thread_tid(thread), thread->priority,
-                    thread->sched_pri, 0,
-                    0);
+           (uintptr_t)thread_tid(thread), thread->priority,
+           thread->sched_pri, 0,
+           0);
 
        stack = thread->kernel_stack;
        thread->kernel_stack = 0;
 
-       return (stack);
+       return stack;
 }
 
 /*
@@ -1865,28 +2009,30 @@ machine_stack_detach(thread_t thread)
 
 void
 machine_stack_attach(
-       thread_t                thread,
-       vm_offset_t             stack)
+       thread_t                thread,
+       vm_offset_t             stack)
 {
        struct x86_kernel_state *statep;
 
        KERNEL_DEBUG(MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_ATTACH),
-                    (uintptr_t)thread_tid(thread), thread->priority,
-                    thread->sched_pri, 0, 0);
+           (uintptr_t)thread_tid(thread), thread->priority,
+           thread->sched_pri, 0, 0);
 
        assert(stack);
        thread->kernel_stack = stack;
+       thread_initialize_kernel_state(thread);
 
        statep = STACK_IKS(stack);
-#if defined(__x86_64__)
-       statep->k_rip = (unsigned long) Thread_continue;
-       statep->k_rbx = (unsigned long) thread_continue;
-       statep->k_rsp = (unsigned long) (STACK_IKS(stack) - 1);
-#else
-       statep->k_eip = (unsigned long) Thread_continue;
-       statep->k_ebx = (unsigned long) thread_continue;
-       statep->k_esp = (unsigned long) (STACK_IKS(stack) - 1);
-#endif
+
+       /*
+        * Reset the state of the thread to resume from a continuation,
+        * including resetting the stack and frame pointer to avoid backtracers
+        * seeing this temporary state and attempting to walk the defunct stack.
+        */
+       statep->k_rbp = (uint64_t) 0;
+       statep->k_rip = (uint64_t) Thread_continue;
+       statep->k_rbx = (uint64_t) thread_continue;
+       statep->k_rsp = (uint64_t) STACK_IKS(stack);
 
        return;
 }
@@ -1897,16 +2043,14 @@ machine_stack_attach(
 
 void
 machine_stack_handoff(thread_t old,
-             thread_t new)
+    thread_t new)
 {
        vm_offset_t     stack;
 
        assert(new);
        assert(old);
 
-#if CONFIG_COUNTERS
-       machine_pmc_cswitch(old, new);
-#endif
+       kpc_off_cpu(old);
 
        stack = old->kernel_stack;
        if (stack == old->reserved_stack) {
@@ -1921,15 +2065,20 @@ machine_stack_handoff(thread_t old,
         */
        new->kernel_stack = stack;
 
-       fpu_save_context(old);
-       
+       fpu_switch_context(old, new);
+
        old->machine.specFlags &= ~OnProc;
        new->machine.specFlags |= OnProc;
 
-       PMAP_SWITCH_CONTEXT(old, new, cpu_number());
+       pmap_switch_context(old, new, cpu_number());
        act_machine_switch_pcb(old, new);
 
+#if HYPERVISOR
+       ml_hv_cswitch(old, new);
+#endif
+
        machine_set_current_thread(new);
+       thread_initialize_kernel_state(new);
 
        return;
 }
@@ -1958,140 +2107,133 @@ act_thread_csave(void)
        mach_msg_type_number_t val;
        thread_t thr_act = current_thread();
 
-       if (thread_is_64bit(thr_act)) {
+       if (thread_is_64bit_addr(thr_act)) {
                struct x86_act_context64 *ic64;
 
                ic64 = (struct x86_act_context64 *)kalloc(sizeof(struct x86_act_context64));
 
-               if (ic64 == (struct x86_act_context64 *)NULL)
-                       return((void *)0);
+               if (ic64 == (struct x86_act_context64 *)NULL) {
+                       return (void *)0;
+               }
 
-               val = x86_SAVED_STATE64_COUNT; 
+               val = x86_SAVED_STATE64_COUNT;
                kret = machine_thread_get_state(thr_act, x86_SAVED_STATE64,
-                               (thread_state_t) &ic64->ss, &val);
+                   (thread_state_t) &ic64->ss, &val);
                if (kret != KERN_SUCCESS) {
                        kfree(ic64, sizeof(struct x86_act_context64));
-                       return((void *)0);
+                       return (void *)0;
                }
-               val = x86_FLOAT_STATE64_COUNT; 
+               val = x86_FLOAT_STATE64_COUNT;
                kret = machine_thread_get_state(thr_act, x86_FLOAT_STATE64,
-                               (thread_state_t) &ic64->fs, &val);
+                   (thread_state_t) &ic64->fs, &val);
                if (kret != KERN_SUCCESS) {
                        kfree(ic64, sizeof(struct x86_act_context64));
-                       return((void *)0);
+                       return (void *)0;
                }
 
                val = x86_DEBUG_STATE64_COUNT;
                kret = machine_thread_get_state(thr_act,
-                                               x86_DEBUG_STATE64,
-                                               (thread_state_t)&ic64->ds,
-                                               &val);
+                   x86_DEBUG_STATE64,
+                   (thread_state_t)&ic64->ds,
+                   &val);
                if (kret != KERN_SUCCESS) {
-                       kfree(ic64, sizeof(struct x86_act_context64));
-                       return((void *)0);
+                       kfree(ic64, sizeof(struct x86_act_context64));
+                       return (void *)0;
                }
-               return(ic64);
-
+               return ic64;
        } else {
                struct x86_act_context32 *ic32;
 
                ic32 = (struct x86_act_context32 *)kalloc(sizeof(struct x86_act_context32));
 
-               if (ic32 == (struct x86_act_context32 *)NULL)
-                       return((void *)0);
+               if (ic32 == (struct x86_act_context32 *)NULL) {
+                       return (void *)0;
+               }
 
-               val = x86_SAVED_STATE32_COUNT; 
+               val = x86_SAVED_STATE32_COUNT;
                kret = machine_thread_get_state(thr_act, x86_SAVED_STATE32,
-                               (thread_state_t) &ic32->ss, &val);
+                   (thread_state_t) &ic32->ss, &val);
                if (kret != KERN_SUCCESS) {
                        kfree(ic32, sizeof(struct x86_act_context32));
-                       return((void *)0);
+                       return (void *)0;
                }
-               val = x86_FLOAT_STATE32_COUNT; 
+               val = x86_FLOAT_STATE32_COUNT;
                kret = machine_thread_get_state(thr_act, x86_FLOAT_STATE32,
-                               (thread_state_t) &ic32->fs, &val);
+                   (thread_state_t) &ic32->fs, &val);
                if (kret != KERN_SUCCESS) {
                        kfree(ic32, sizeof(struct x86_act_context32));
-                       return((void *)0);
+                       return (void *)0;
                }
 
                val = x86_DEBUG_STATE32_COUNT;
                kret = machine_thread_get_state(thr_act,
-                                               x86_DEBUG_STATE32,
-                                               (thread_state_t)&ic32->ds,
-                                               &val);
+                   x86_DEBUG_STATE32,
+                   (thread_state_t)&ic32->ds,
+                   &val);
                if (kret != KERN_SUCCESS) {
-                       kfree(ic32, sizeof(struct x86_act_context32));
-                       return((void *)0);
+                       kfree(ic32, sizeof(struct x86_act_context32));
+                       return (void *)0;
                }
-               return(ic32);
+               return ic32;
        }
 }
 
 
-void 
+void
 act_thread_catt(void *ctx)
 {
-        thread_t thr_act = current_thread();
+       thread_t thr_act = current_thread();
        kern_return_t kret;
 
-       if (ctx == (void *)NULL)
-                               return;
+       if (ctx == (void *)NULL) {
+               return;
+       }
 
-        if (thread_is_64bit(thr_act)) {
-               struct x86_act_context64 *ic64;
+       if (thread_is_64bit_addr(thr_act)) {
+               struct x86_act_context64 *ic64;
 
-               ic64 = (struct x86_act_context64 *)ctx;
+               ic64 = (struct x86_act_context64 *)ctx;
 
                kret = machine_thread_set_state(thr_act, x86_SAVED_STATE64,
-                                               (thread_state_t) &ic64->ss, x86_SAVED_STATE64_COUNT);
+                   (thread_state_t) &ic64->ss, x86_SAVED_STATE64_COUNT);
                if (kret == KERN_SUCCESS) {
-                               machine_thread_set_state(thr_act, x86_FLOAT_STATE64,
-                                                        (thread_state_t) &ic64->fs, x86_FLOAT_STATE64_COUNT);
+                       machine_thread_set_state(thr_act, x86_FLOAT_STATE64,
+                           (thread_state_t) &ic64->fs, x86_FLOAT_STATE64_COUNT);
                }
                kfree(ic64, sizeof(struct x86_act_context64));
        } else {
-               struct x86_act_context32 *ic32;
+               struct x86_act_context32 *ic32;
 
-               ic32 = (struct x86_act_context32 *)ctx;
+               ic32 = (struct x86_act_context32 *)ctx;
 
                kret = machine_thread_set_state(thr_act, x86_SAVED_STATE32,
-                                               (thread_state_t) &ic32->ss, x86_SAVED_STATE32_COUNT);
+                   (thread_state_t) &ic32->ss, x86_SAVED_STATE32_COUNT);
                if (kret == KERN_SUCCESS) {
                        (void) machine_thread_set_state(thr_act, x86_FLOAT_STATE32,
-                                                (thread_state_t) &ic32->fs, x86_FLOAT_STATE32_COUNT);
+                           (thread_state_t) &ic32->fs, x86_FLOAT_STATE32_COUNT);
                }
                kfree(ic32, sizeof(struct x86_act_context32));
        }
 }
 
 
-void act_thread_cfree(__unused void *ctx)
+void
+act_thread_cfree(__unused void *ctx)
 {
        /* XXX - Unused */
 }
-void x86_toggle_sysenter_arg_store(thread_t thread, boolean_t valid);
-void x86_toggle_sysenter_arg_store(thread_t thread, boolean_t valid) {
-       thread->machine.arg_store_valid = valid;
-}
-
-boolean_t x86_sysenter_arg_store_isvalid(thread_t thread);
-
-boolean_t x86_sysenter_arg_store_isvalid(thread_t thread) {
-       return (thread->machine.arg_store_valid);
-}
 
 /*
  * Duplicate one x86_debug_state32_t to another.  "all" parameter
  * chooses whether dr4 and dr5 are copied (they are never meant
- * to be installed when we do machine_task_set_state() or 
+ * to be installed when we do machine_task_set_state() or
  * machine_thread_set_state()).
  */
 void
 copy_debug_state32(
-               x86_debug_state32_t *src,
-               x86_debug_state32_t *target,
-               boolean_t all)
+       x86_debug_state32_t *src,
+       x86_debug_state32_t *target,
+       boolean_t all)
 {
        if (all) {
                target->dr4 = src->dr4;
@@ -2109,14 +2251,14 @@ copy_debug_state32(
 /*
  * Duplicate one x86_debug_state64_t to another.  "all" parameter
  * chooses whether dr4 and dr5 are copied (they are never meant
- * to be installed when we do machine_task_set_state() or 
+ * to be installed when we do machine_task_set_state() or
  * machine_thread_set_state()).
  */
 void
 copy_debug_state64(
-               x86_debug_state64_t *src,
-               x86_debug_state64_t *target,
-               boolean_t all)
+       x86_debug_state64_t *src,
+       x86_debug_state64_t *target,
+       boolean_t all)
 {
        if (all) {
                target->dr4 = src->dr4;