xnu-6153.11.26.tar.gz

[apple/xnu.git] / osfmk / i386 / trap.c

diff --git a/osfmk/i386/trap.c b/osfmk/i386/trap.c
index 5cd9b390d8dd8e7dfe3644a7571e8f1ff9ec9397..bfc24c4aa93e45faf1235cf34669a3eabfacc287 100644 (file)
--- a/osfmk/i386/trap.c
+++ b/osfmk/i386/trap.c
@@ -1,8 +1,8 @@
 /*
 /*

- * Copyright (c) 2000-2008 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2018 Apple Inc. All rights reserved.

  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@

- * 
+ *

  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
  * 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


@@ -11,10 +11,10 @@
  * 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.
  * 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.
  * 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,
  * 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,


@@ -22,46 +22,44 @@
  * 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.
  * 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@
  */
 /*
  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*

-* @OSF_COPYRIGHT@
-*/
-/* 
-* Mach Operating System
-* Copyright (c) 1991,1990,1989,1988 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.
-*/
+ * @OSF_COPYRIGHT@
+ */

 /*
 /*

-*/
+ * Mach Operating System
+ * Copyright (c) 1991,1990,1989,1988 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.
+ */
+/*
+ */

 
 /*
 
 /*

-* Hardware trap/fault handler.
+ * Hardware trap/fault handler.

  */
 
  */
 

-#include <mach_kdb.h>
-#include <mach_kgdb.h>

 #include <mach_kdp.h>
 #include <mach_ldebug.h>
 
 #include <mach_kdp.h>
 #include <mach_ldebug.h>
 


@@ -91,20 +89,12 @@
 #include <kern/spl.h>
 #include <kern/misc_protos.h>
 #include <kern/debug.h>
 #include <kern/spl.h>
 #include <kern/misc_protos.h>
 #include <kern/debug.h>

-
+#if CONFIG_TELEMETRY
+#include <kern/telemetry.h>
+#endif

 #include <sys/kdebug.h>
 #include <sys/kdebug.h>

-
-#if    MACH_KGDB
-#include <kgdb/kgdb_defs.h>
-#endif /* MACH_KGDB */
-
-#if    MACH_KDB
-#include <debug.h>
-#include <ddb/db_watch.h>
-#include <ddb/db_run.h>
-#include <ddb/db_break.h>
-#include <ddb/db_trap.h>
-#endif /* MACH_KDB */
+#include <kperf/kperf.h>
+#include <prng/random.h>

 
 #include <string.h>
 
 
 #include <string.h>
 


@@ -117,24 +107,18 @@
 #include <mach/i386/syscall_sw.h>
 
 #include <libkern/OSDebug.h>
 #include <mach/i386/syscall_sw.h>
 
 #include <libkern/OSDebug.h>

+#include <i386/cpu_threads.h>
+#include <machine/pal_routines.h>

 
 

-extern void throttle_lowpri_io(boolean_t);
-
+extern void throttle_lowpri_io(int);
+extern void kprint_state(x86_saved_state64_t *saved_state);

 
 /*
  * Forward declarations
  */
 static void user_page_fault_continue(kern_return_t kret);
 
 /*
  * Forward declarations
  */
 static void user_page_fault_continue(kern_return_t kret);

-#ifdef __i386__
-static void panic_trap(x86_saved_state32_t *saved_state);
-static void set_recovery_ip(x86_saved_state32_t *saved_state, vm_offset_t ip);
-#else
-static void panic_trap(x86_saved_state64_t *saved_state);
+static void panic_trap(x86_saved_state64_t *saved_state, uint32_t pl, kern_return_t fault_result) __dead2;

 static void set_recovery_ip(x86_saved_state64_t *saved_state, vm_offset_t ip);
 static void set_recovery_ip(x86_saved_state64_t *saved_state, vm_offset_t ip);

-#endif
-
-volatile perfCallback perfTrapHook = NULL; /* Pointer to CHUD trap hook routine */
-volatile perfCallback perfASTHook  = NULL; /* Pointer to CHUD AST hook routine */

 
 #if CONFIG_DTRACE
 /* See <rdar://problem/4613924> */
 
 #if CONFIG_DTRACE
 /* See <rdar://problem/4613924> */


@@ -143,184 +127,110 @@ perfCallback tempDTraceTrapHook = NULL; /* Pointer to DTrace fbt trap hook routi
 extern boolean_t dtrace_tally_fault(user_addr_t);
 #endif
 
 extern boolean_t dtrace_tally_fault(user_addr_t);
 #endif
 

+extern boolean_t pmap_smep_enabled;
+extern boolean_t pmap_smap_enabled;
+
+__attribute__((noreturn))

 void
 thread_syscall_return(
 void
 thread_syscall_return(

-        kern_return_t ret)
+       kern_return_t ret)

 {
 {

-        thread_t       thr_act = current_thread();
-       boolean_t       is_mach;
-       int             code;
+       thread_t        thr_act = current_thread();
+       boolean_t       is_mach;
+       int             code;
+
+       pal_register_cache_state(thr_act, DIRTY);

 
 

+       if (thread_is_64bit_addr(thr_act)) {
+               x86_saved_state64_t     *regs;

 
 

-        if (thread_is_64bit(thr_act)) {
-               x86_saved_state64_t     *regs;
-               

                regs = USER_REGS64(thr_act);
 
                code = (int) (regs->rax & SYSCALL_NUMBER_MASK);
                is_mach = (regs->rax & SYSCALL_CLASS_MASK)
                regs = USER_REGS64(thr_act);
 
                code = (int) (regs->rax & SYSCALL_NUMBER_MASK);
                is_mach = (regs->rax & SYSCALL_CLASS_MASK)

-                           == (SYSCALL_CLASS_MACH << SYSCALL_CLASS_SHIFT);
+                   == (SYSCALL_CLASS_MACH << SYSCALL_CLASS_SHIFT);

                if (kdebug_enable && is_mach) {
                if (kdebug_enable && is_mach) {

-                       /* Mach trap */
-                       KERNEL_DEBUG_CONSTANT(
-                             MACHDBG_CODE(DBG_MACH_EXCP_SC,code)|DBG_FUNC_END,
-                             ret, 0, 0, 0, 0);
+                       /* Mach trap */
+                       KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
+                           MACHDBG_CODE(DBG_MACH_EXCP_SC, code) | DBG_FUNC_END,
+                           ret, 0, 0, 0, 0);

                }
                regs->rax = ret;
 #if DEBUG
                }
                regs->rax = ret;
 #if DEBUG

-               if (is_mach)
+               if (is_mach) {

                        DEBUG_KPRINT_SYSCALL_MACH(
                                "thread_syscall_return: 64-bit mach ret=%u\n",
                                ret);
                        DEBUG_KPRINT_SYSCALL_MACH(
                                "thread_syscall_return: 64-bit mach ret=%u\n",
                                ret);

-               else
+               } else {

                        DEBUG_KPRINT_SYSCALL_UNIX(
                                "thread_syscall_return: 64-bit unix ret=%u\n",
                                ret);
                        DEBUG_KPRINT_SYSCALL_UNIX(
                                "thread_syscall_return: 64-bit unix ret=%u\n",
                                ret);

+               }

 #endif
        } else {
 #endif
        } else {

-               x86_saved_state32_t     *regs;
-               
+               x86_saved_state32_t     *regs;
+

                regs = USER_REGS32(thr_act);
 
                code = ((int) regs->eax);
                is_mach = (code < 0);
                if (kdebug_enable && is_mach) {
                regs = USER_REGS32(thr_act);
 
                code = ((int) regs->eax);
                is_mach = (code < 0);
                if (kdebug_enable && is_mach) {

-                       /* Mach trap */
-                       KERNEL_DEBUG_CONSTANT(
-                             MACHDBG_CODE(DBG_MACH_EXCP_SC,-code)|DBG_FUNC_END,
-                             ret, 0, 0, 0, 0);
+                       /* Mach trap */
+                       KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
+                           MACHDBG_CODE(DBG_MACH_EXCP_SC, -code) | DBG_FUNC_END,
+                           ret, 0, 0, 0, 0);

                }
                regs->eax = ret;
 #if DEBUG
                }
                regs->eax = ret;
 #if DEBUG

-               if (is_mach)
+               if (is_mach) {

                        DEBUG_KPRINT_SYSCALL_MACH(
                                "thread_syscall_return: 32-bit mach ret=%u\n",
                                ret);
                        DEBUG_KPRINT_SYSCALL_MACH(
                                "thread_syscall_return: 32-bit mach ret=%u\n",
                                ret);

-               else
+               } else {

                        DEBUG_KPRINT_SYSCALL_UNIX(
                                "thread_syscall_return: 32-bit unix ret=%u\n",
                                ret);
                        DEBUG_KPRINT_SYSCALL_UNIX(
                                "thread_syscall_return: 32-bit unix ret=%u\n",
                                ret);

+               }

 #endif
        }
 #endif
        }

-       throttle_lowpri_io(TRUE);
-
-       thread_exception_return();
-        /*NOTREACHED*/
-}
-
-
-#if    MACH_KDB
-boolean_t      debug_all_traps_with_kdb = FALSE;
-extern struct db_watchpoint *db_watchpoint_list;
-extern boolean_t db_watchpoints_inserted;
-extern boolean_t db_breakpoints_inserted;

 
 

-void
-thread_kdb_return(void)
-{
-       thread_t                thr_act = current_thread();
-       x86_saved_state_t       *iss = USER_STATE(thr_act);
-
-
-        if (is_saved_state64(iss)) {
-               x86_saved_state64_t     *regs;
-               
-               regs = saved_state64(iss);
-
-               if (kdb_trap(regs->isf.trapno, (int)regs->isf.err, (void *)regs)) {
-                       thread_exception_return();
-                       /*NOTREACHED*/
-               }
+#if DEBUG || DEVELOPMENT
+       kern_allocation_name_t
+       prior __assert_only = thread_get_kernel_state(thr_act)->allocation_name;
+       assertf(prior == NULL, "thread_set_allocation_name(\"%s\") not cleared", kern_allocation_get_name(prior));
+#endif /* DEBUG || DEVELOPMENT */

 
 

-       } else {
-               x86_saved_state32_t     *regs;
-               
-               regs = saved_state32(iss);
+       throttle_lowpri_io(1);

 
 

-               if (kdb_trap(regs->trapno, regs->err, (void *)regs)) {
-                       thread_exception_return();
-                       /*NOTREACHED*/
-               }
-       }
+       thread_exception_return();
+       /*NOTREACHED*/

 }
 
 }
 

-#endif /* MACH_KDB */

 
 

-void
+static inline void

 user_page_fault_continue(
 user_page_fault_continue(

-                        kern_return_t  kr)
+       kern_return_t  kr)

 {
 {

-       thread_t        thread = current_thread();
-       ast_t           *myast;
-       boolean_t       intr;
-       user_addr_t     vaddr;
-
-
-#if    MACH_KDB
-       x86_saved_state_t *regs = USER_STATE(thread);
-       int             err;
-       int             trapno;
-
-       assert((is_saved_state32(regs) && !thread_is_64bit(thread)) ||
-              (is_saved_state64(regs) &&  thread_is_64bit(thread)));
-#endif
+       thread_t        thread = current_thread();
+       user_addr_t     vaddr;

 
 

-        if (thread_is_64bit(thread)) {
-               x86_saved_state64_t     *uregs;
+       if (thread_is_64bit_addr(thread)) {
+               x86_saved_state64_t     *uregs;

 
                uregs = USER_REGS64(thread);
 
 
                uregs = USER_REGS64(thread);
 

-#if    MACH_KDB
-               trapno = uregs->isf.trapno;
-               err = (int)uregs->isf.err;
-#endif

                vaddr = (user_addr_t)uregs->cr2;
        } else {
                vaddr = (user_addr_t)uregs->cr2;
        } else {

-               x86_saved_state32_t     *uregs;
+               x86_saved_state32_t     *uregs;

 
                uregs = USER_REGS32(thread);
 
 
                uregs = USER_REGS32(thread);
 

-#if    MACH_KDB
-               trapno = uregs->trapno;
-               err = uregs->err;
-#endif

                vaddr = uregs->cr2;
        }
 
                vaddr = uregs->cr2;
        }
 

-       if ((kr == KERN_SUCCESS) || (kr == KERN_ABORTED)) {
-#if    MACH_KDB
-               if (!db_breakpoints_inserted) {
-                       db_set_breakpoints();
-               }
-               if (db_watchpoint_list &&
-                   db_watchpoints_inserted &&
-                   (err & T_PF_WRITE) &&
-                   db_find_watchpoint(thread->map,
-                                      (vm_offset_t)vaddr,
-                                      saved_state32(regs)))
-                       kdb_trap(T_WATCHPOINT, 0, saved_state32(regs));
-#endif /* MACH_KDB */
-               intr = ml_set_interrupts_enabled(FALSE);
-               myast = ast_pending();
-               while (*myast & AST_ALL) {
-                       ast_taken(AST_ALL, intr);
-                       ml_set_interrupts_enabled(FALSE);
-                       myast = ast_pending();
-               }
-               ml_set_interrupts_enabled(intr);
-
-               thread_exception_return();
-               /*NOTREACHED*/
-       }
-
-#if    MACH_KDB
-       if (debug_all_traps_with_kdb &&
-           kdb_trap(trapno, err, saved_state32(regs))) {
-               thread_exception_return();
-               /*NOTREACHED*/
-       }
-#endif /* MACH_KDB */

 
 

+       /* PAL debug hook */
+       pal_dbg_page_fault( thread, vaddr, kr );

 
        i386_exception(EXC_BAD_ACCESS, kr, vaddr);
        /*NOTREACHED*/
 
        i386_exception(EXC_BAD_ACCESS, kr, vaddr);
        /*NOTREACHED*/


@@ -330,34 +240,33 @@ user_page_fault_continue(
  * Fault recovery in copyin/copyout routines.
  */
 struct recovery {
  * Fault recovery in copyin/copyout routines.
  */
 struct recovery {

-       uintptr_t       fault_addr;
-       uintptr_t       recover_addr;
+       uintptr_t       fault_addr;
+       uintptr_t       recover_addr;

 };
 
 };
 

-extern struct recovery recover_table[];
-extern struct recovery recover_table_end[];
+extern struct recovery  recover_table[];
+extern struct recovery  recover_table_end[];
+
+const char *    trap_type[] = {TRAP_NAMES};
+unsigned        TRAP_TYPES = sizeof(trap_type) / sizeof(trap_type[0]);

 
 

-const char *   trap_type[] = {TRAP_NAMES};
-unsigned       TRAP_TYPES = sizeof(trap_type)/sizeof(trap_type[0]);
+extern void     PE_incoming_interrupt(int interrupt);

 
 #if defined(__x86_64__) && DEBUG
 
 #if defined(__x86_64__) && DEBUG

-static void
-print_state(x86_saved_state64_t        *saved_state)
+void
+kprint_state(x86_saved_state64_t        *saved_state)

 {
        kprintf("current_cpu_datap() 0x%lx\n", (uintptr_t)current_cpu_datap());
        kprintf("Current GS base MSR 0x%llx\n", rdmsr64(MSR_IA32_GS_BASE));
        kprintf("Kernel  GS base MSR 0x%llx\n", rdmsr64(MSR_IA32_KERNEL_GS_BASE));
        kprintf("state at 0x%lx:\n", (uintptr_t) saved_state);
 
 {
        kprintf("current_cpu_datap() 0x%lx\n", (uintptr_t)current_cpu_datap());
        kprintf("Current GS base MSR 0x%llx\n", rdmsr64(MSR_IA32_GS_BASE));
        kprintf("Kernel  GS base MSR 0x%llx\n", rdmsr64(MSR_IA32_KERNEL_GS_BASE));
        kprintf("state at 0x%lx:\n", (uintptr_t) saved_state);
 

-       kprintf("      rdi    0x%llx\n", saved_state->rdi);        
-       kprintf("      rsi    0x%llx\n", saved_state->rsi);    
+       kprintf("      rdi    0x%llx\n", saved_state->rdi);
+       kprintf("      rsi    0x%llx\n", saved_state->rsi);

        kprintf("      rdx    0x%llx\n", saved_state->rdx);
        kprintf("      r10    0x%llx\n", saved_state->r10);
        kprintf("      r8     0x%llx\n", saved_state->r8);
        kprintf("      rdx    0x%llx\n", saved_state->rdx);
        kprintf("      r10    0x%llx\n", saved_state->r10);
        kprintf("      r8     0x%llx\n", saved_state->r8);

-       kprintf("      r9     0x%llx\n", saved_state->r9);     
-       kprintf("      v_arg6 0x%llx\n", saved_state->v_arg6);
-       kprintf("      v_arg7 0x%llx\n", saved_state->v_arg7);
-       kprintf("      v_arg8 0x%llx\n", saved_state->v_arg8);
+       kprintf("      r9     0x%llx\n", saved_state->r9);

 
        kprintf("      cr2    0x%llx\n", saved_state->cr2);
        kprintf("real  cr2    0x%lx\n", get_cr2());
 
        kprintf("      cr2    0x%llx\n", saved_state->cr2);
        kprintf("real  cr2    0x%lx\n", get_cr2());


@@ -384,20 +293,60 @@ print_state(x86_saved_state64_t   *saved_state)
        kprintf("  isf.rsp    0x%llx\n", saved_state->isf.rsp);
        kprintf("  isf.ss     0x%llx\n", saved_state->isf.ss);
 }
        kprintf("  isf.rsp    0x%llx\n", saved_state->isf.rsp);
        kprintf("  isf.ss     0x%llx\n", saved_state->isf.ss);
 }

+#endif
+
+

 /*
 /*

- * K64 debug - fatal handler for debug code in the trap vectors.
+ * Non-zero indicates latency assert is enabled and capped at valued
+ * absolute time units.

  */
  */

-extern void
-panic_idt64(x86_saved_state_t *rsp);
+
+uint64_t interrupt_latency_cap = 0;
+boolean_t ilat_assert = FALSE;
+

 void
 void

-panic_idt64(x86_saved_state_t *rsp)
+interrupt_latency_tracker_setup(void)

 {
 {

-       print_state(saved_state64(rsp));
-       panic("panic_idt64");
+       uint32_t ilat_cap_us;
+       if (PE_parse_boot_argn("interrupt_latency_cap_us", &ilat_cap_us, sizeof(ilat_cap_us))) {
+               interrupt_latency_cap = ilat_cap_us * NSEC_PER_USEC;
+               nanoseconds_to_absolutetime(interrupt_latency_cap, &interrupt_latency_cap);
+       } else {
+               interrupt_latency_cap = LockTimeOut;
+       }
+       PE_parse_boot_argn("-interrupt_latency_assert_enable", &ilat_assert, sizeof(ilat_assert));
+}
+
+void
+interrupt_reset_latency_stats(void)
+{
+       uint32_t i;
+       for (i = 0; i < real_ncpus; i++) {
+               cpu_data_ptr[i]->cpu_max_observed_int_latency =
+                   cpu_data_ptr[i]->cpu_max_observed_int_latency_vector = 0;
+       }
+}
+
+void
+interrupt_populate_latency_stats(char *buf, unsigned bufsize)
+{
+       uint32_t i, tcpu = ~0;
+       uint64_t cur_max = 0;
+
+       for (i = 0; i < real_ncpus; i++) {
+               if (cur_max < cpu_data_ptr[i]->cpu_max_observed_int_latency) {
+                       cur_max = cpu_data_ptr[i]->cpu_max_observed_int_latency;
+                       tcpu = i;
+               }
+       }
+
+       if (tcpu < real_ncpus) {
+               snprintf(buf, bufsize, "0x%x 0x%x 0x%llx", tcpu, cpu_data_ptr[tcpu]->cpu_max_observed_int_latency_vector, cpu_data_ptr[tcpu]->cpu_max_observed_int_latency);
+       }

 }
 }

-#endif

 
 

-extern void    PE_incoming_interrupt(int interrupt);
+uint32_t interrupt_timer_coalescing_enabled = 1;
+uint64_t interrupt_coalesced_timers;

 
 /*
  * Handle interrupts:
 
 /*
  * Handle interrupts:


@@ -407,71 +356,149 @@ extern void      PE_incoming_interrupt(int interrupt);
 void
 interrupt(x86_saved_state_t *state)
 {
 void
 interrupt(x86_saved_state_t *state)
 {

-       uint64_t        rip;
-       uint64_t        rsp;
-       int             interrupt_num;
-       boolean_t       user_mode = FALSE;
-
+       uint64_t        rip;
+       uint64_t        rsp;
+       int             interrupt_num;
+       boolean_t       user_mode = FALSE;
+       int             ipl;
+       int             cnum = cpu_number();
+       cpu_data_t      *cdp = cpu_data_ptr[cnum];
+       int             itype = DBG_INTR_TYPE_UNKNOWN;
+       int             handled;
+
+       x86_saved_state64_t     *state64 = saved_state64(state);
+       rip = state64->isf.rip;
+       rsp = state64->isf.rsp;
+       interrupt_num = state64->isf.trapno;
+       if (state64->isf.cs & 0x03) {
+               user_mode = TRUE;
+       }

 
 

-        if (is_saved_state64(state) == TRUE) {
-               x86_saved_state64_t     *state64;
+       if (cpu_data_ptr[cnum]->lcpu.package->num_idle == topoParms.nLThreadsPerPackage) {
+               cpu_data_ptr[cnum]->cpu_hwIntpexits[interrupt_num]++;
+       }

 
 

-               state64 = saved_state64(state);
-               rip = state64->isf.rip;
-               rsp = state64->isf.rsp;
-               interrupt_num = state64->isf.trapno;
-#ifdef __x86_64__
-               if(state64->isf.cs & 0x03)
-#endif
-                       user_mode = TRUE;
+       if (interrupt_num == (LAPIC_DEFAULT_INTERRUPT_BASE + LAPIC_INTERPROCESSOR_INTERRUPT)) {
+               itype = DBG_INTR_TYPE_IPI;
+       } else if (interrupt_num == (LAPIC_DEFAULT_INTERRUPT_BASE + LAPIC_TIMER_INTERRUPT)) {
+               itype = DBG_INTR_TYPE_TIMER;

        } else {
        } else {

-               x86_saved_state32_t     *state32;
-
-               state32 = saved_state32(state);
-               if (state32->cs & 0x03)
-                       user_mode = TRUE;
-               rip = state32->eip;
-               rsp = state32->uesp;
-               interrupt_num = state32->trapno;
+               itype = DBG_INTR_TYPE_OTHER;
+       }
+
+       KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
+           MACHDBG_CODE(DBG_MACH_EXCP_INTR, 0) | DBG_FUNC_START,
+           interrupt_num,
+           (user_mode ? rip : VM_KERNEL_UNSLIDE(rip)),
+           user_mode, itype, 0);
+
+       SCHED_STATS_INTERRUPT(current_processor());
+
+#if CONFIG_TELEMETRY
+       if (telemetry_needs_record) {
+               telemetry_mark_curthread(user_mode, FALSE);

        }
        }

+#endif

 
 

-       KERNEL_DEBUG_CONSTANT(
-               MACHDBG_CODE(DBG_MACH_EXCP_INTR, 0) | DBG_FUNC_START,
-               interrupt_num, (long) rip, user_mode, 0, 0);
+       ipl = get_preemption_level();

 
        /*
         * Handle local APIC interrupts
         * else call platform expert for devices.
 
        /*
         * Handle local APIC interrupts
         * else call platform expert for devices.

-        */ 
-       if (!lapic_interrupt(interrupt_num, state))
-               PE_incoming_interrupt(interrupt_num);
+        */
+       handled = lapic_interrupt(interrupt_num, state);
+
+       if (!handled) {
+               if (interrupt_num == (LAPIC_DEFAULT_INTERRUPT_BASE + LAPIC_CMCI_INTERRUPT)) {
+                       /*
+                        * CMCI can be signalled on any logical processor, and the kexts
+                        * that implement handling CMCI use IOKit to register handlers for
+                        * the CMCI vector, so if we see a CMCI, do not encode a CPU
+                        * number in bits 8:31 (since the vector is the same regardless of
+                        * the handling CPU).
+                        */
+                       PE_incoming_interrupt(interrupt_num);
+               } else if (cnum <= lapic_max_interrupt_cpunum) {
+                       PE_incoming_interrupt((cnum << 8) | interrupt_num);
+               }
+       }
+
+       if (__improbable(get_preemption_level() != ipl)) {
+               panic("Preemption level altered by interrupt vector 0x%x: initial 0x%x, final: 0x%x\n", interrupt_num, ipl, get_preemption_level());
+       }
+
+
+       if (__improbable(cdp->cpu_nested_istack)) {
+               cdp->cpu_nested_istack_events++;
+       } else {
+               uint64_t ctime = mach_absolute_time();
+               uint64_t int_latency = ctime - cdp->cpu_int_event_time;
+               uint64_t esdeadline, ehdeadline;
+               /* Attempt to process deferred timers in the context of
+                * this interrupt, unless interrupt time has already exceeded
+                * TCOAL_ILAT_THRESHOLD.
+                */
+#define TCOAL_ILAT_THRESHOLD (30000ULL)
+
+               if ((int_latency < TCOAL_ILAT_THRESHOLD) &&
+                   interrupt_timer_coalescing_enabled) {
+                       esdeadline = cdp->rtclock_timer.queue.earliest_soft_deadline;
+                       ehdeadline = cdp->rtclock_timer.deadline;
+                       if ((ctime >= esdeadline) && (ctime < ehdeadline)) {
+                               interrupt_coalesced_timers++;
+                               TCOAL_DEBUG(0x88880000 | DBG_FUNC_START, ctime, esdeadline, ehdeadline, interrupt_coalesced_timers, 0);
+                               rtclock_intr(state);
+                               TCOAL_DEBUG(0x88880000 | DBG_FUNC_END, ctime, esdeadline, interrupt_coalesced_timers, 0, 0);
+                       } else {
+                               TCOAL_DEBUG(0x77770000, ctime, cdp->rtclock_timer.queue.earliest_soft_deadline, cdp->rtclock_timer.deadline, interrupt_coalesced_timers, 0);
+                       }
+               }

 
 

-       KERNEL_DEBUG_CONSTANT(
-               MACHDBG_CODE(DBG_MACH_EXCP_INTR, 0) | DBG_FUNC_END,
-               0, 0, 0, 0, 0);
+               if (__improbable(ilat_assert && (int_latency > interrupt_latency_cap) && !machine_timeout_suspended())) {
+                       panic("Interrupt vector 0x%x exceeded interrupt latency threshold, 0x%llx absolute time delta, prior signals: 0x%x, current signals: 0x%x", interrupt_num, int_latency, cdp->cpu_prior_signals, cdp->cpu_signals);
+               }
+
+               if (__improbable(int_latency > cdp->cpu_max_observed_int_latency)) {
+                       cdp->cpu_max_observed_int_latency = int_latency;
+                       cdp->cpu_max_observed_int_latency_vector = interrupt_num;
+               }
+       }

 
        /*
         * Having serviced the interrupt first, look at the interrupted stack depth.
         */
        if (!user_mode) {
 
        /*
         * Having serviced the interrupt first, look at the interrupted stack depth.
         */
        if (!user_mode) {

-               uint64_t depth = current_cpu_datap()->cpu_kernel_stack
-                                + sizeof(struct x86_kernel_state)
-                                + sizeof(struct i386_exception_link *)
-                                - rsp;
-               if (depth > kernel_stack_depth_max) {
+               uint64_t depth = cdp->cpu_kernel_stack
+                   + sizeof(struct thread_kernel_state)
+                   + sizeof(struct i386_exception_link *)
+                   - rsp;
+               if (__improbable(depth > kernel_stack_depth_max)) {

                        kernel_stack_depth_max = (vm_offset_t)depth;
                        KERNEL_DEBUG_CONSTANT(
                                MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_DEPTH),
                        kernel_stack_depth_max = (vm_offset_t)depth;
                        KERNEL_DEBUG_CONSTANT(
                                MACHDBG_CODE(DBG_MACH_SCHED, MACH_STACK_DEPTH),

-                               (long) depth, (long) rip, 0, 0, 0);
+                               (long) depth, (long) VM_KERNEL_UNSLIDE(rip), 0, 0, 0);

                }
        }
                }
        }

+
+       if (cnum == master_cpu) {
+               ml_entropy_collect();
+       }
+
+#if KPERF
+       kperf_interrupt();
+#endif /* KPERF */
+
+       KDBG_RELEASE(MACHDBG_CODE(DBG_MACH_EXCP_INTR, 0) | DBG_FUNC_END,
+           interrupt_num);
+
+       assert(ml_get_interrupts_enabled() == FALSE);

 }
 
 static inline void
 reset_dr7(void)
 {
        long dr7 = 0x400; /* magic dr7 reset value; 32 bit on i386, 64 bit on x86_64 */
 }
 
 static inline void
 reset_dr7(void)
 {
        long dr7 = 0x400; /* magic dr7 reset value; 32 bit on i386, 64 bit on x86_64 */

-       __asm__ volatile("mov %0,%%dr7" : : "r" (dr7));
+       __asm__ volatile ("mov %0,%%dr7" : : "r" (dr7));

 }
 #if MACH_KDP
 unsigned kdp_has_active_watchpoints = 0;
 }
 #if MACH_KDP
 unsigned kdp_has_active_watchpoints = 0;


@@ -487,79 +514,51 @@ unsigned kdp_has_active_watchpoints = 0;
 
 void
 kernel_trap(
 
 void
 kernel_trap(

-       x86_saved_state_t       *state)
+       x86_saved_state_t       *state,
+       uintptr_t *lo_spp)

 {
 {

-#ifdef __i386__
-       x86_saved_state32_t     *saved_state;
-#else
-       x86_saved_state64_t     *saved_state;
-#endif
-       int                     code;
-       user_addr_t             vaddr;
-       int                     type;
-       vm_map_t                map = 0;        /* protected by T_PAGE_FAULT */
-       kern_return_t           result = KERN_FAILURE;
-       thread_t                thread;
-       ast_t                   *myast;
+       x86_saved_state64_t     *saved_state;
+       int                     code;
+       user_addr_t             vaddr;
+       int                     type;
+       vm_map_t                map = 0;        /* protected by T_PAGE_FAULT */
+       kern_return_t           result = KERN_FAILURE;
+       kern_return_t           fault_result = KERN_SUCCESS;
+       thread_t                thread;

        boolean_t               intr;
        boolean_t               intr;

-       vm_prot_t               prot;
-        struct recovery                *rp;
-       vm_offset_t             kern_ip;
+       vm_prot_t               prot;
+       struct recovery         *rp;
+       vm_offset_t             kern_ip;

 #if NCOPY_WINDOWS > 0
 #if NCOPY_WINDOWS > 0

-       int                     fault_in_copy_window = -1;
+       int                     fault_in_copy_window = -1;

 #endif
 #endif

-       int                     is_user = 0;
-#if MACH_KDB
-       pt_entry_t              *pte;
-#endif /* MACH_KDB */
-       
+       int                     is_user;
+       int                     trap_pl = get_preemption_level();
+

        thread = current_thread();
 
        thread = current_thread();
 

-#ifdef __i386__
-       if (is_saved_state64(state))
-               panic("kernel_trap(%p) with 64-bit state", state);
-       saved_state = saved_state32(state);
-       vaddr = (user_addr_t)saved_state->cr2;
-       type  = saved_state->trapno;
-       code  = saved_state->err & 0xffff;
-       intr  = (saved_state->efl & EFL_IF) != 0;       /* state of ints at trap */
-       kern_ip = (vm_offset_t)saved_state->eip;
-#else
-       if (is_saved_state32(state))
+       if (__improbable(is_saved_state32(state))) {

                panic("kernel_trap(%p) with 32-bit state", state);
                panic("kernel_trap(%p) with 32-bit state", state);

+       }

        saved_state = saved_state64(state);
        saved_state = saved_state64(state);

+
+       /* Record cpu where state was captured */
+       saved_state->isf.cpu = cpu_number();
+

        vaddr = (user_addr_t)saved_state->cr2;
        type  = saved_state->isf.trapno;
        code  = (int)(saved_state->isf.err & 0xffff);
        vaddr = (user_addr_t)saved_state->cr2;
        type  = saved_state->isf.trapno;
        code  = (int)(saved_state->isf.err & 0xffff);

-       intr  = (saved_state->isf.rflags & EFL_IF) != 0;        /* state of ints at trap */
+       intr  = (saved_state->isf.rflags & EFL_IF) != 0;        /* state of ints at trap */

        kern_ip = (vm_offset_t)saved_state->isf.rip;
        kern_ip = (vm_offset_t)saved_state->isf.rip;

-#endif
-
-       myast = ast_pending();

 
 

-       perfCallback fn = perfASTHook;
-       if (fn) {
-               if (*myast & AST_CHUD_ALL)
-                       fn(type, NULL, 0, 0);
-       } else
-               *myast &= ~AST_CHUD_ALL;
+       is_user = (vaddr < VM_MAX_USER_PAGE_ADDRESS);

 
 

+#if CONFIG_DTRACE

        /*
        /*

-        * Is there a hook?
+        * Is there a DTrace hook?

         */
         */

-       fn = perfTrapHook;
-       if (fn) {
-               if (fn(type, NULL, 0, 0) == KERN_SUCCESS) {
-                       /*
-                        * If it succeeds, we are done...
-                        */
-                       return;
-               }
-       }
-
-#if CONFIG_DTRACE
-       if (tempDTraceTrapHook) {
-               if (tempDTraceTrapHook(type, state, 0, 0) == KERN_SUCCESS) {
+       if (__improbable(tempDTraceTrapHook != NULL)) {
+               if (tempDTraceTrapHook(type, state, lo_spp, 0) == KERN_SUCCESS) {

                        /*
                         * If it succeeds, we are done...
                         */
                        /*
                         * If it succeeds, we are done...
                         */


@@ -573,25 +572,33 @@ kernel_trap(
         * on preemption below.  but we do want to re-enable interrupts
         * as soon we possibly can to hold latency down
         */
         * on preemption below.  but we do want to re-enable interrupts
         * as soon we possibly can to hold latency down
         */

-       if (T_PREEMPT == type) {
-               ast_taken(AST_PREEMPTION, FALSE);
+       if (__improbable(T_PREEMPT == type)) {
+               ast_taken_kernel();

 
 

-               KERNEL_DEBUG_CONSTANT((MACHDBG_CODE(DBG_MACH_EXCP_KTRAP_x86, type)) | DBG_FUNC_NONE,
-                                     0, 0, 0, kern_ip, 0);
+               KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
+                   (MACHDBG_CODE(DBG_MACH_EXCP_KTRAP_x86, type)) | DBG_FUNC_NONE,
+                   0, 0, 0, VM_KERNEL_UNSLIDE(kern_ip), 0);

                return;
        }
                return;
        }

-       
+
+       user_addr_t     kd_vaddr = is_user ? vaddr : VM_KERNEL_UNSLIDE(vaddr);
+       KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
+           (MACHDBG_CODE(DBG_MACH_EXCP_KTRAP_x86, type)) | DBG_FUNC_NONE,
+           (unsigned)(kd_vaddr >> 32), (unsigned)kd_vaddr, is_user,
+           VM_KERNEL_UNSLIDE(kern_ip), 0);
+
+

        if (T_PAGE_FAULT == type) {
                /*
                 * assume we're faulting in the kernel map
                 */
                map = kernel_map;
 
        if (T_PAGE_FAULT == type) {
                /*
                 * assume we're faulting in the kernel map
                 */
                map = kernel_map;
 

-               if (thread != THREAD_NULL && thread->map != kernel_map) {
+               if (__probable(thread != THREAD_NULL && thread->map != kernel_map)) {

 #if NCOPY_WINDOWS > 0
 #if NCOPY_WINDOWS > 0

-                       vm_offset_t     copy_window_base;
-                       vm_offset_t     kvaddr;
-                       int             window_index;
+                       vm_offset_t     copy_window_base;
+                       vm_offset_t     kvaddr;
+                       int             window_index;

 
                        kvaddr = (vm_offset_t)vaddr;
                        /*
 
                        kvaddr = (vm_offset_t)vaddr;
                        /*


@@ -601,183 +608,140 @@ kernel_trap(
                         * we only need to look at the window
                         * associated with this processor
                         */
                         * we only need to look at the window
                         * associated with this processor
                         */

-                       copy_window_base = current_cpu_datap()->cpu_copywindow_base;
-
-                       if (kvaddr >= copy_window_base && kvaddr < (copy_window_base + (NBPDE * NCOPY_WINDOWS)) ) {
+                       copy_window_base = current_cpu_datap()->cpu_copywindow_base;

 
 

-                               window_index = (kvaddr - copy_window_base) / NBPDE;
+                       if (kvaddr >= copy_window_base && kvaddr < (copy_window_base + (NBPDE * NCOPY_WINDOWS))) {
+                               window_index = (int)((kvaddr - copy_window_base) / NBPDE);

 
                                if (thread->machine.copy_window[window_index].user_base != (user_addr_t)-1) {
 
                                if (thread->machine.copy_window[window_index].user_base != (user_addr_t)-1) {

-
-                                       kvaddr -= (copy_window_base + (NBPDE * window_index));
-                                       vaddr = thread->machine.copy_window[window_index].user_base + kvaddr;
+                                       kvaddr -= (copy_window_base + (NBPDE * window_index));
+                                       vaddr = thread->machine.copy_window[window_index].user_base + kvaddr;

 
                                        map = thread->map;
                                        fault_in_copy_window = window_index;
                                }
 
                                        map = thread->map;
                                        fault_in_copy_window = window_index;
                                }

-                               is_user = -1;

                        }
 #else
                        }
 #else

-                       if (vaddr < VM_MAX_USER_PAGE_ADDRESS) {
+                       if (__probable(vaddr < VM_MAX_USER_PAGE_ADDRESS)) {

                                /* fault occurred in userspace */
                                map = thread->map;
                                /* fault occurred in userspace */
                                map = thread->map;

-                               is_user = -1;
+
+                               /* Intercept a potential Supervisor Mode Execute
+                                * Protection fault. These criteria identify
+                                * both NX faults and SMEP faults, but both
+                                * are fatal. We avoid checking PTEs (racy).
+                                * (The VM could just redrive a SMEP fault, hence
+                                * the intercept).
+                                */
+                               if (__improbable((code == (T_PF_PROT | T_PF_EXECUTE)) &&
+                                   (pmap_smep_enabled) && (saved_state->isf.rip == vaddr))) {
+                                       goto debugger_entry;
+                               }
+
+                               /*
+                                * Additionally check for SMAP faults...
+                                * which are characterized by page-present and
+                                * the AC bit unset (i.e. not from copyin/out path).
+                                */
+                               if (__improbable(code & T_PF_PROT &&
+                                   pmap_smap_enabled &&
+                                   (saved_state->isf.rflags & EFL_AC) == 0)) {
+                                       goto debugger_entry;
+                               }
+

                                /*
                                 * If we're not sharing cr3 with the user
                                 * and we faulted in copyio,
                                 * then switch cr3 here and dismiss the fault.
                                 */
                                if (no_shared_cr3 &&
                                /*
                                 * If we're not sharing cr3 with the user
                                 * and we faulted in copyio,
                                 * then switch cr3 here and dismiss the fault.
                                 */
                                if (no_shared_cr3 &&

-                                   (thread->machine.specFlags&CopyIOActive) &&
-                                   map->pmap->pm_cr3 != get_cr3()) {
-                                       set_cr3(map->pmap->pm_cr3);
+                                   (thread->machine.specFlags & CopyIOActive) &&
+                                   map->pmap->pm_cr3 != get_cr3_base()) {
+                                       pmap_assert(current_cpu_datap()->cpu_pmap_pcid_enabled == FALSE);
+                                       set_cr3_raw(map->pmap->pm_cr3);

                                        return;
                                }
                                        return;
                                }

+                               if (__improbable(vaddr < PAGE_SIZE) &&
+                                   ((thread->machine.specFlags & CopyIOActive) == 0)) {
+                                       goto debugger_entry;
+                               }

                        }
 #endif
                }
        }
                        }
 #endif
                }
        }

-       KERNEL_DEBUG_CONSTANT(
-               (MACHDBG_CODE(DBG_MACH_EXCP_KTRAP_x86, type)) | DBG_FUNC_NONE,
-               (unsigned)(vaddr >> 32), (unsigned)vaddr, is_user, kern_ip, 0);
-

 
        (void) ml_set_interrupts_enabled(intr);
 
        switch (type) {
 
        (void) ml_set_interrupts_enabled(intr);
 
        switch (type) {

-
-           case T_NO_FPU:
+       case T_NO_FPU:

                fpnoextflt();
                return;
 
                fpnoextflt();
                return;
 

-           case T_FPU_FAULT:
+       case T_FPU_FAULT:

                fpextovrflt();
                return;
 
                fpextovrflt();
                return;
 

-           case T_FLOATING_POINT_ERROR:
+       case T_FLOATING_POINT_ERROR:

                fpexterrflt();
                return;
 
                fpexterrflt();
                return;
 

-           case T_SSE_FLOAT_ERROR:
-               fpSSEexterrflt();
+       case T_SSE_FLOAT_ERROR:
+               fpSSEexterrflt();

                return;
                return;

-           case T_DEBUG:
-#ifdef __i386__
-                   if ((saved_state->efl & EFL_TF) == 0 && NO_WATCHPOINTS)
-#else
-                   if ((saved_state->isf.rflags & EFL_TF) == 0 && NO_WATCHPOINTS)
-#endif
-                   {
-                           /* We've somehow encountered a debug
-                            * register match that does not belong
-                            * to the kernel debugger.
-                            * This isn't supposed to happen.
-                            */
-                           reset_dr7();
-                           return;
-                   }
-                   goto debugger_entry;
-#ifdef __x86_64__
-           case T_INT3:
-             goto debugger_entry;
-#endif
-           case T_PAGE_FAULT:
-               /*
-                * If the current map is a submap of the kernel map,
-                * and the address is within that map, fault on that
-                * map.  If the same check is done in vm_fault
-                * (vm_map_lookup), we may deadlock on the kernel map
-                * lock.
-                */

 
 

-               prot = VM_PROT_READ;
-
-               if (code & T_PF_WRITE)
-                       prot |= VM_PROT_WRITE;
-#if     PAE
-               if (code & T_PF_EXECUTE)
-                       prot |= VM_PROT_EXECUTE;
-#endif
+       case T_INVALID_OPCODE:
+               fpUDflt(kern_ip);
+               goto debugger_entry;

 
 

-#if    MACH_KDB
-               /*
-                * Check for watchpoint on kernel static data.
-                * vm_fault would fail in this case 
-                */
-               if (map == kernel_map && db_watchpoint_list && db_watchpoints_inserted &&
-                   (code & T_PF_WRITE) && vaddr < vm_map_max(map) &&
-                   ((*(pte = pmap_pte(kernel_pmap, (vm_map_offset_t)vaddr))) & INTEL_PTE_WRITE) == 0) {
-                       pmap_store_pte(
-                               pte,
-                               *pte | INTEL_PTE_VALID | INTEL_PTE_WRITE);
-                       /* XXX need invltlb here? */
-
-                       result = KERN_SUCCESS;
-                       goto look_for_watchpoints;
+       case T_DEBUG:
+               if ((saved_state->isf.rflags & EFL_TF) == 0 && NO_WATCHPOINTS) {
+                       /* We've somehow encountered a debug
+                        * register match that does not belong
+                        * to the kernel debugger.
+                        * This isn't supposed to happen.
+                        */
+                       reset_dr7();
+                       return;

                }
                }

-#endif /* MACH_KDB */
+               goto debugger_entry;
+       case T_INT3:
+               goto debugger_entry;
+       case T_PAGE_FAULT:

 
 #if CONFIG_DTRACE
 
 #if CONFIG_DTRACE

-               if (thread->options & TH_OPT_DTRACE) {  /* Executing under dtrace_probe? */
+               if (thread != THREAD_NULL && thread->t_dtrace_inprobe) { /* Executing under dtrace_probe? */

                        if (dtrace_tally_fault(vaddr)) { /* Should a fault under dtrace be ignored? */
                                /*
                                 * DTrace has "anticipated" the possibility of this fault, and has
                                 * established the suitable recovery state. Drop down now into the
                        if (dtrace_tally_fault(vaddr)) { /* Should a fault under dtrace be ignored? */
                                /*
                                 * DTrace has "anticipated" the possibility of this fault, and has
                                 * established the suitable recovery state. Drop down now into the

-                                * recovery handling code in "case T_GENERAL_PROTECTION:". 
+                                * recovery handling code in "case T_GENERAL_PROTECTION:".

                                 */
                                goto FALL_THROUGH;
                        }
                }
 #endif /* CONFIG_DTRACE */
 
                                 */
                                goto FALL_THROUGH;
                        }
                }
 #endif /* CONFIG_DTRACE */
 

-               result = vm_fault(map,
-                                 vm_map_trunc_page(vaddr),
-                                 prot,
-                                 FALSE, 
-                                 THREAD_UNINT, NULL, 0);
+               prot = VM_PROT_READ;

 
 

-#if    MACH_KDB
-               if (result == KERN_SUCCESS) {
-                       /*
-                        * Look for watchpoints
-                        */
-look_for_watchpoints:
-                       if (map == kernel_map && db_watchpoint_list && db_watchpoints_inserted && (code & T_PF_WRITE) &&
-                           db_find_watchpoint(map, vaddr, saved_state))
-                               kdb_trap(T_WATCHPOINT, 0, saved_state);
+               if (code & T_PF_WRITE) {
+                       prot |= VM_PROT_WRITE;

                }
                }

-#endif /* MACH_KDB */
+               if (code & T_PF_EXECUTE) {
+                       prot |= VM_PROT_EXECUTE;
+               }
+
+               fault_result = result = vm_fault(map,
+                   vaddr,
+                   prot,
+                   FALSE, VM_KERN_MEMORY_NONE,
+                   THREAD_UNINT, NULL, 0);

 
                if (result == KERN_SUCCESS) {
 #if NCOPY_WINDOWS > 0
                        if (fault_in_copy_window != -1) {
 
                if (result == KERN_SUCCESS) {
 #if NCOPY_WINDOWS > 0
                        if (fault_in_copy_window != -1) {

-                               pt_entry_t      *updp;
-                               pt_entry_t      *kpdp;
-
-                               /*
-                                * in case there was no page table assigned
-                                * for the user base address and the pmap
-                                * got 'expanded' due to this fault, we'll
-                                * copy in the descriptor 
-                                *
-                                * we're either setting the page table descriptor
-                                * to the same value or it was 0... no need
-                                * for a TLB flush in either case
-                                */
-
-                       ml_set_interrupts_enabled(FALSE);
-                       updp = pmap_pde(map->pmap, thread->machine.copy_window[fault_in_copy_window].user_base);
-                               assert(updp);
-                               if (0 == updp) panic("trap: updp 0"); /* XXX DEBUG */
-                               kpdp = current_cpu_datap()->cpu_copywindow_pdp;
-                               kpdp += fault_in_copy_window;
-
-#if JOE_DEBUG
-                               if (*kpdp && (*kpdp & PG_FRAME) != (*updp & PG_FRAME))
-                                       panic("kernel_fault: user pdp doesn't match - updp = 0x%qx, kpdp = 0x%qx\n", *updp, *kpdp);
-#endif
-                               pmap_store_pte(kpdp, *updp);
-
+                               ml_set_interrupts_enabled(FALSE);
+                               copy_window_fault(thread, map,
+                                   fault_in_copy_window);

                                (void) ml_set_interrupts_enabled(intr);
                        }
 #endif /* NCOPY_WINDOWS > 0 */
                                (void) ml_set_interrupts_enabled(intr);
                        }
 #endif /* NCOPY_WINDOWS > 0 */


@@ -790,17 +754,14 @@ look_for_watchpoints:
 FALL_THROUGH:
 #endif /* CONFIG_DTRACE */
 
 FALL_THROUGH:
 #endif /* CONFIG_DTRACE */
 

-           case T_GENERAL_PROTECTION:
-#if defined(__x86_64__) && DEBUG
-               print_state(saved_state);
-#endif
+       case T_GENERAL_PROTECTION:

                /*
                 * If there is a failure recovery address
                 * for this fault, go there.
                 */
                /*
                 * If there is a failure recovery address
                 * for this fault, go there.
                 */

-               for (rp = recover_table; rp < recover_table_end; rp++) {
-                       if (kern_ip == rp->fault_addr) {
-                               set_recovery_ip(saved_state, rp->recover_addr);
+               for (rp = recover_table; rp < recover_table_end; rp++) {
+                       if (kern_ip == rp->fault_addr) {
+                               set_recovery_ip(saved_state, rp->recover_addr);

                                return;
                        }
                }
                                return;
                        }
                }


@@ -808,25 +769,24 @@ FALL_THROUGH:
                /*
                 * Check thread recovery address also.
                 */
                /*
                 * Check thread recovery address also.
                 */

-               if (thread->recover) {
+               if (thread != THREAD_NULL && thread->recover) {

                        set_recovery_ip(saved_state, thread->recover);
                        thread->recover = 0;
                        return;
                }
                        set_recovery_ip(saved_state, thread->recover);
                        thread->recover = 0;
                        return;
                }

-               /*
-                * Unanticipated page-fault errors in kernel
-                * should not happen.
-                *
-                * fall through...
-                */
-
-           default:
+       /*
+        * Unanticipated page-fault errors in kernel
+        * should not happen.
+        *
+        * fall through...
+        */
+       default:

                /*
                 * Exception 15 is reserved but some chips may generate it
                 * spuriously. Seen at startup on AMD Athlon-64.
                 */
                /*
                 * Exception 15 is reserved but some chips may generate it
                 * spuriously. Seen at startup on AMD Athlon-64.
                 */

-               if (type == 15) {
-                       kprintf("kernel_trap() ignoring spurious trap 15\n"); 
+               if (type == 15) {
+                       kprintf("kernel_trap() ignoring spurious trap 15\n");

                        return;
                }
 debugger_entry:
                        return;
                }
 debugger_entry:


@@ -836,319 +796,99 @@ debugger_entry:
                 * access through the debugger.
                 */
                sync_iss_to_iks(state);
                 * access through the debugger.
                 */
                sync_iss_to_iks(state);

-#if MACH_KDB
-restart_debugger:
-#endif /* MACH_KDB */          

 #if  MACH_KDP
 #if  MACH_KDP

-                if (current_debugger != KDB_CUR_DB) {
-                       if (kdp_i386_trap(type, saved_state, result, (vm_offset_t)vaddr))
-                               return;
-               } else {
-#endif /* MACH_KDP */
-#if MACH_KDB
-                       if (kdb_trap(type, code, saved_state)) {
-                               if (switch_debugger) {
-                                       current_debugger = KDP_CUR_DB;
-                                       switch_debugger = 0;
-                                       goto restart_debugger;
-                               }
-                               return;
-                       }
-#endif /* MACH_KDB */
-#if MACH_KDP
+               if (kdp_i386_trap(type, saved_state, result, (vm_offset_t)vaddr)) {
+                       return;

                }
 #endif
        }
                }
 #endif
        }

-
-       panic_trap(saved_state);
+       pal_cli();
+       panic_trap(saved_state, trap_pl, fault_result);

        /*
         * NO RETURN
         */
 }
 
        /*
         * NO RETURN
         */
 }
 

-
-#ifdef __i386__
-static void
-set_recovery_ip(x86_saved_state32_t  *saved_state, vm_offset_t ip)
-{
-        saved_state->eip = ip;
-}
-#else

 static void
 set_recovery_ip(x86_saved_state64_t  *saved_state, vm_offset_t ip)
 {
 static void
 set_recovery_ip(x86_saved_state64_t  *saved_state, vm_offset_t ip)
 {

-        saved_state->isf.rip = ip;
+       saved_state->isf.rip = ip;

 }
 }

-#endif
-

 
 

-#ifdef __i386__

 static void
 static void

-panic_trap(x86_saved_state32_t *regs)
+panic_trap(x86_saved_state64_t *regs, uint32_t pl, kern_return_t fault_result)

 {
 {

-       const char *trapname = "Unknown";
-       uint32_t        cr0 = get_cr0();
-       uint32_t        cr2 = get_cr2();
-       uint32_t        cr3 = get_cr3();
-       uint32_t        cr4 = get_cr4();
+       const char      *trapname = "Unknown";
+       pal_cr_t        cr0, cr2, cr3, cr4;
+       boolean_t       potential_smep_fault = FALSE, potential_kernel_NX_fault = FALSE;
+       boolean_t       potential_smap_fault = FALSE;
+
+       pal_get_control_registers( &cr0, &cr2, &cr3, &cr4 );
+       assert(ml_get_interrupts_enabled() == FALSE);
+       current_cpu_datap()->cpu_fatal_trap_state = regs;

        /*
         * Issue an I/O port read if one has been requested - this is an
         * event logic analyzers can use as a trigger point.
         */
        panic_io_port_read();
 
        /*
         * Issue an I/O port read if one has been requested - this is an
         * event logic analyzers can use as a trigger point.
         */
        panic_io_port_read();
 

-       kprintf("panic trap number 0x%x, eip 0x%x\n", regs->trapno, regs->eip);
-       kprintf("cr0 0x%08x cr2 0x%08x cr3 0x%08x cr4 0x%08x\n",
-               cr0, cr2, cr3, cr4);
-
-       if (regs->trapno < TRAP_TYPES)
-               trapname = trap_type[regs->trapno];
-#undef panic
-       panic("Kernel trap at 0x%08x, type %d=%s, registers:\n"
-             "CR0: 0x%08x, CR2: 0x%08x, CR3: 0x%08x, CR4: 0x%08x\n"
-             "EAX: 0x%08x, EBX: 0x%08x, ECX: 0x%08x, EDX: 0x%08x\n"
-             "CR2: 0x%08x, EBP: 0x%08x, ESI: 0x%08x, EDI: 0x%08x\n"
-             "EFL: 0x%08x, EIP: 0x%08x, CS:  0x%08x, DS:  0x%08x\n"
-             "Error code: 0x%08x\n",
-             regs->eip, regs->trapno, trapname, cr0, cr2, cr3, cr4,
-             regs->eax,regs->ebx,regs->ecx,regs->edx,
-             regs->cr2,regs->ebp,regs->esi,regs->edi,
-             regs->efl,regs->eip,regs->cs, regs->ds, regs->err);
-       /*
-        * This next statement is not executed,
-        * but it's needed to stop the compiler using tail call optimization
-        * for the panic call - which confuses the subsequent backtrace.
-        */
-       cr0 = 0;
-}
-#else
-static void
-panic_trap(x86_saved_state64_t *regs)
-{
-       const char      *trapname = "Unknown";
-       uint64_t        cr0 = get_cr0();
-       uint64_t        cr2 = get_cr2();
-       uint64_t        cr3 = get_cr3();
-       uint64_t        cr4 = get_cr4();
-
-       /*
-        * Issue an I/O port read if one has been requested - this is an
-        * event logic analyzers can use as a trigger point.
-        */
-       panic_io_port_read();
-
-       kprintf("panic trap number 0x%x, rip 0x%016llx\n",
-               regs->isf.trapno, regs->isf.rip);
+       kprintf("CPU %d panic trap number 0x%x, rip 0x%016llx\n",
+           cpu_number(), regs->isf.trapno, regs->isf.rip);

        kprintf("cr0 0x%016llx cr2 0x%016llx cr3 0x%016llx cr4 0x%016llx\n",
        kprintf("cr0 0x%016llx cr2 0x%016llx cr3 0x%016llx cr4 0x%016llx\n",

-               cr0, cr2, cr3, cr4);
-
-       if (regs->isf.trapno < TRAP_TYPES)
-               trapname = trap_type[regs->isf.trapno];
-#undef panic
-       panic("Kernel trap at 0x%016llx, type %d=%s, registers:\n"
-             "CR0: 0x%016llx, CR2: 0x%016llx, CR3: 0x%016llx, CR4: 0x%016llx\n"
-             "RAX: 0x%016llx, RBX: 0x%016llx, RCX: 0x%016llx, RDX: 0x%016llx\n"
-             "RSP: 0x%016llx, RBP: 0x%016llx, RSI: 0x%016llx, RDI: 0x%016llx\n"
-             "R8:  0x%016llx, R9:  0x%016llx, R10: 0x%016llx, R11: 0x%016llx\n"
-             "R12: 0x%016llx, R13: 0x%016llx, R14: 0x%016llx, R15: 0x%016llx\n"
-             "RFL: 0x%016llx, RIP: 0x%016llx, CS:  0x%016llx, SS:  0x%016llx\n"
-             "Error code: 0x%016llx\n",
-             regs->isf.rip, regs->isf.trapno, trapname,
-             cr0, cr2, cr3, cr4,
-             regs->rax, regs->rbx, regs->rcx, regs->rdx,
-             regs->isf.rsp, regs->rbp, regs->rsi, regs->rdi,
-             regs->r8,  regs->r9,  regs->r10, regs->r11,
-             regs->r12, regs->r13, regs->r14, regs->r15,
-             regs->isf.rflags, regs->isf.rip, regs->isf.cs,  regs->isf.ss,
-             regs->isf.err);
-       /*
-        * This next statement is not executed,
-        * but it's needed to stop the compiler using tail call optimization
-        * for the panic call - which confuses the subsequent backtrace.
-        */
-       cr0 = 0;
-}
-#endif
-
-extern void     kprintf_break_lock(void);
-
-#ifdef __i386__
-static void
-panic_32(__unused int code, __unused int pc, __unused const char *msg, boolean_t do_mca_dump, boolean_t do_bt)
-{
-       struct i386_tss *my_ktss = current_ktss();
-
-       /* Set postcode (DEBUG only) */
-       postcode(pc);
-
-       /*
-        * Issue an I/O port read if one has been requested - this is an
-        * event logic analyzers can use as a trigger point.
-        */
-       panic_io_port_read();
-
-       /*
-        * Break kprintf lock in case of recursion,
-        * and record originally faulted instruction address.
-        */
-       kprintf_break_lock();
-
-       if (do_mca_dump) {
-#if CONFIG_MCA
-               /*
-                * Dump the contents of the machine check MSRs (if any).
-                */
-               mca_dump();
-#endif
-       }
-
-#if MACH_KDP
-       /*
-        * Print backtrace leading to first fault:
-        */
-       if (do_bt)
-               panic_i386_backtrace((void *) my_ktss->ebp, 10, NULL, FALSE, NULL);
-#endif
-
-       panic("%s at 0x%08x, thread:%p, code:0x%x, "
-             "registers:\n"
-             "CR0: 0x%08x, CR2: 0x%08x, CR3: 0x%08x, CR4: 0x%08x\n"
-             "EAX: 0x%08x, EBX: 0x%08x, ECX: 0x%08x, EDX: 0x%08x\n"
-             "ESP: 0x%08x, EBP: 0x%08x, ESI: 0x%08x, EDI: 0x%08x\n"
-             "EFL: 0x%08x, EIP: 0x%08x\n",
-                 msg,
-             my_ktss->eip, current_thread(), code,
-             (uint32_t)get_cr0(), (uint32_t)get_cr2(), (uint32_t)get_cr3(), (uint32_t)get_cr4(),
-             my_ktss->eax, my_ktss->ebx, my_ktss->ecx, my_ktss->edx,
-             my_ktss->esp, my_ktss->ebp, my_ktss->esi, my_ktss->edi,
-             my_ktss->eflags, my_ktss->eip);
-}
-
-/*
- * Called from locore on a special reserved stack after a double-fault
- * is taken in kernel space.
- * Kernel stack overflow is one route here.
- */
-void
-panic_double_fault32(int code)
-{
-       panic_32(code, PANIC_DOUBLE_FAULT, "Double fault", FALSE, TRUE);
-}
-
-/*
- * Called from locore on a special reserved stack after a machine-check
- */
-void 
-panic_machine_check32(int code)
-{
-       panic_32(code, PANIC_MACHINE_CHECK, "Machine-check", TRUE, FALSE);
-}
-#endif /* __i386__ */
-
-static void
-panic_64(x86_saved_state_t *sp, __unused int pc, __unused const char *msg, boolean_t do_mca_dump)
-{
-       /* Set postcode (DEBUG only) */
-       postcode(pc);
+           cr0, cr2, cr3, cr4);

 
 

-       /*
-        * Issue an I/O port read if one has been requested - this is an
-        * event logic analyzers can use as a trigger point.
-        */
-       panic_io_port_read();
-
-       /*
-        * Break kprintf lock in case of recursion,
-        * and record originally faulted instruction address.
-        */
-       kprintf_break_lock();
-
-       if (do_mca_dump) {
-#if CONFIG_MCA
-               /*
-                * Dump the contents of the machine check MSRs (if any).
-                */
-               mca_dump();
-#endif
+       if (regs->isf.trapno < TRAP_TYPES) {
+               trapname = trap_type[regs->isf.trapno];

        }
 
        }
 

-#ifdef __i386__
-       /*
-        * Dump the interrupt stack frame at last kernel entry.
-        */
-       if (is_saved_state64(sp)) {
-               x86_saved_state64_t     *ss64p = saved_state64(sp);
-               panic("%s thread:%p, trapno:0x%x, err:0x%qx, "
-                     "registers:\n"
-                     "CR0: 0x%08x, CR2: 0x%08x, CR3: 0x%08x, CR4: 0x%08x\n"
-                     "RAX: 0x%016qx, RBX: 0x%016qx, RCX: 0x%016qx, RDX: 0x%016qx\n"
-                     "RSP: 0x%016qx, RBP: 0x%016qx, RSI: 0x%016qx, RDI: 0x%016qx\n"
-                     "R8:  0x%016qx, R9:  0x%016qx, R10: 0x%016qx, R11: 0x%016qx\n"
-                     "R12: 0x%016qx, R13: 0x%016qx, R14: 0x%016qx, R15: 0x%016qx\n"
-                     "RFL: 0x%016qx, RIP: 0x%016qx, CR2: 0x%016qx\n",
-                         msg,
-                     current_thread(), ss64p->isf.trapno, ss64p->isf.err,
-                     (uint32_t)get_cr0(), (uint32_t)get_cr2(), (uint32_t)get_cr3(), (uint32_t)get_cr4(),
-                     ss64p->rax, ss64p->rbx, ss64p->rcx, ss64p->rdx,
-                     ss64p->isf.rsp, ss64p->rbp, ss64p->rsi, ss64p->rdi,
-                     ss64p->r8, ss64p->r9, ss64p->r10, ss64p->r11,
-                     ss64p->r12, ss64p->r13, ss64p->r14, ss64p->r15,
-                     ss64p->isf.rflags, ss64p->isf.rip, ss64p->cr2);
-       } else {
-               x86_saved_state32_t     *ss32p = saved_state32(sp);
-               panic("%s at 0x%08x, thread:%p, trapno:0x%x, err:0x%x,"
-                     "registers:\n"
-                     "CR0: 0x%08x, CR2: 0x%08x, CR3: 0x%08x, CR4: 0x%08x\n"
-                     "EAX: 0x%08x, EBX: 0x%08x, ECX: 0x%08x, EDX: 0x%08x\n"
-                     "ESP: 0x%08x, EBP: 0x%08x, ESI: 0x%08x, EDI: 0x%08x\n"
-                     "EFL: 0x%08x, EIP: 0x%08x\n",
-                     msg,
-                         ss32p->eip, current_thread(), ss32p->trapno, ss32p->err,
-                     (uint32_t)get_cr0(), (uint32_t)get_cr2(), (uint32_t)get_cr3(), (uint32_t)get_cr4(),
-                     ss32p->eax, ss32p->ebx, ss32p->ecx, ss32p->edx,
-                     ss32p->uesp, ss32p->ebp, ss32p->esi, ss32p->edi,
-                     ss32p->efl, ss32p->eip);
+       if ((regs->isf.trapno == T_PAGE_FAULT) && (regs->isf.err == (T_PF_PROT | T_PF_EXECUTE)) && (regs->isf.rip == regs->cr2)) {
+               if (pmap_smep_enabled && (regs->isf.rip < VM_MAX_USER_PAGE_ADDRESS)) {
+                       potential_smep_fault = TRUE;
+               } else if (regs->isf.rip >= VM_MIN_KERNEL_AND_KEXT_ADDRESS) {
+                       potential_kernel_NX_fault = TRUE;
+               }
+       } else if (pmap_smap_enabled &&
+           regs->isf.trapno == T_PAGE_FAULT &&
+           regs->isf.err & T_PF_PROT &&
+           regs->cr2 < VM_MAX_USER_PAGE_ADDRESS &&
+           regs->isf.rip >= VM_MIN_KERNEL_AND_KEXT_ADDRESS) {
+               potential_smap_fault = TRUE;

        }
        }

-#else
-       x86_saved_state64_t *regs = saved_state64(sp);
-       panic("%s thread:%p at 0x%016llx, registers:\n"
-             "CR0: 0x%016lx, CR2: 0x%016lx, CR3: 0x%016lx, CR4: 0x%016lx\n"
-             "RAX: 0x%016llx, RBX: 0x%016llx, RCX: 0x%016llx, RDX: 0x%016llx\n"
-             "RSP: 0x%016llx, RBP: 0x%016llx, RSI: 0x%016llx, RDI: 0x%016llx\n"
-             "R8:  0x%016llx, R9:  0x%016llx, R10: 0x%016llx, R11: 0x%016llx\n"
-             "R12: 0x%016llx, R13: 0x%016llx, R14: 0x%016llx, R15: 0x%016llx\n"
-             "RFL: 0x%016llx, RIP: 0x%016llx, CS:  0x%016llx, SS:  0x%016llx\n"
-             "Error code: 0x%016llx\n",
-             msg,
-                 current_thread(), regs->isf.rip,
-             get_cr0(), get_cr2(), get_cr3(), get_cr4(),
-             regs->rax, regs->rbx, regs->rcx, regs->rdx,
-             regs->isf.rsp, regs->rbp, regs->rsi, regs->rdi,
-             regs->r8,  regs->r9,  regs->r10, regs->r11,
-             regs->r12, regs->r13, regs->r14, regs->r15,
-             regs->isf.rflags, regs->isf.rip, regs->isf.cs,  regs->isf.ss,
-             regs->isf.err);
-#endif
-}
-
-void
-panic_double_fault64(x86_saved_state_t *sp)
-{
-       panic_64(sp, PANIC_DOUBLE_FAULT, "Double fault", FALSE);
-
-}
-void
-
-panic_machine_check64(x86_saved_state_t *sp)
-{
-       panic_64(sp, PANIC_MACHINE_CHECK, "Machine Check", TRUE);

 
 

+#undef panic
+       panic("Kernel trap at 0x%016llx, type %d=%s, registers:\n"
+           "CR0: 0x%016llx, CR2: 0x%016llx, CR3: 0x%016llx, CR4: 0x%016llx\n"
+           "RAX: 0x%016llx, RBX: 0x%016llx, RCX: 0x%016llx, RDX: 0x%016llx\n"
+           "RSP: 0x%016llx, RBP: 0x%016llx, RSI: 0x%016llx, RDI: 0x%016llx\n"
+           "R8:  0x%016llx, R9:  0x%016llx, R10: 0x%016llx, R11: 0x%016llx\n"
+           "R12: 0x%016llx, R13: 0x%016llx, R14: 0x%016llx, R15: 0x%016llx\n"
+           "RFL: 0x%016llx, RIP: 0x%016llx, CS:  0x%016llx, SS:  0x%016llx\n"
+           "Fault CR2: 0x%016llx, Error code: 0x%016llx, Fault CPU: 0x%x%s%s%s%s, PL: %d, VF: %d\n",
+           regs->isf.rip, regs->isf.trapno, trapname,
+           cr0, cr2, cr3, cr4,
+           regs->rax, regs->rbx, regs->rcx, regs->rdx,
+           regs->isf.rsp, regs->rbp, regs->rsi, regs->rdi,
+           regs->r8, regs->r9, regs->r10, regs->r11,
+           regs->r12, regs->r13, regs->r14, regs->r15,
+           regs->isf.rflags, regs->isf.rip, regs->isf.cs & 0xFFFF,
+           regs->isf.ss & 0xFFFF, regs->cr2, regs->isf.err, regs->isf.cpu,
+           virtualized ? " VMM" : "",
+           potential_kernel_NX_fault ? " Kernel NX fault" : "",
+           potential_smep_fault ? " SMEP/User NX fault" : "",
+           potential_smap_fault ? " SMAP fault" : "",
+           pl,
+           fault_result);

 }
 
 #if CONFIG_DTRACE
 extern kern_return_t dtrace_user_probe(x86_saved_state_t *);
 #endif
 
 }
 
 #if CONFIG_DTRACE
 extern kern_return_t dtrace_user_probe(x86_saved_state_t *);
 #endif
 

+#if DEBUG
+uint32_t fsigs[2];
+uint32_t fsigns, fsigcs;
+#endif
+

 /*
  *     Trap from user mode.
  */
 /*
  *     Trap from user mode.
  */


@@ -1156,167 +896,167 @@ void
 user_trap(
        x86_saved_state_t *saved_state)
 {
 user_trap(
        x86_saved_state_t *saved_state)
 {

-       int                     exc;
-       int                     err;
-       mach_exception_code_t   code;
+       int                     exc;
+       int                     err;
+       mach_exception_code_t   code;

        mach_exception_subcode_t subcode;
        mach_exception_subcode_t subcode;

-       int                     type;
-       user_addr_t             vaddr;
-       vm_prot_t               prot;
-       thread_t                thread = current_thread();
-       ast_t                   *myast;
-       kern_return_t           kret;
-       user_addr_t             rip;
+       int                     type;
+       user_addr_t             vaddr;
+       vm_prot_t               prot;
+       thread_t                thread = current_thread();
+       kern_return_t           kret;
+       user_addr_t             rip;
+       unsigned long           dr6 = 0; /* 32 bit for i386, 64 bit for x86_64 */

 
 

-       assert((is_saved_state32(saved_state) && !thread_is_64bit(thread)) ||
-              (is_saved_state64(saved_state) &&  thread_is_64bit(thread)));
+       assert((is_saved_state32(saved_state) && !thread_is_64bit_addr(thread)) ||
+           (is_saved_state64(saved_state) && thread_is_64bit_addr(thread)));

 
        if (is_saved_state64(saved_state)) {
 
        if (is_saved_state64(saved_state)) {

-               x86_saved_state64_t     *regs;
+               x86_saved_state64_t     *regs;

 
                regs = saved_state64(saved_state);
 
 
                regs = saved_state64(saved_state);
 

+               /* Record cpu where state was captured */
+               regs->isf.cpu = cpu_number();
+

                type = regs->isf.trapno;
                err  = (int)regs->isf.err & 0xffff;
                vaddr = (user_addr_t)regs->cr2;
                rip   = (user_addr_t)regs->isf.rip;
        } else {
                type = regs->isf.trapno;
                err  = (int)regs->isf.err & 0xffff;
                vaddr = (user_addr_t)regs->cr2;
                rip   = (user_addr_t)regs->isf.rip;
        } else {

-               x86_saved_state32_t     *regs;
+               x86_saved_state32_t     *regs;

 
                regs = saved_state32(saved_state);
 
 
                regs = saved_state32(saved_state);
 

+               /* Record cpu where state was captured */
+               regs->cpu = cpu_number();
+

                type  = regs->trapno;
                err   = regs->err & 0xffff;
                vaddr = (user_addr_t)regs->cr2;
                rip   = (user_addr_t)regs->eip;
        }
 
                type  = regs->trapno;
                err   = regs->err & 0xffff;
                vaddr = (user_addr_t)regs->cr2;
                rip   = (user_addr_t)regs->eip;
        }
 

-       KERNEL_DEBUG_CONSTANT(
-               (MACHDBG_CODE(DBG_MACH_EXCP_UTRAP_x86, type)) | DBG_FUNC_NONE,
-               (unsigned)(vaddr>>32), (unsigned)vaddr,
-               (unsigned)(rip>>32), (unsigned)rip, 0);
+       if ((type == T_DEBUG) && thread->machine.ids) {
+               unsigned long clear = 0;
+               /* Stash and clear this processor's DR6 value, in the event
+                * this was a debug register match
+                */
+               __asm__ volatile ("mov %%db6, %0" : "=r" (dr6));
+               __asm__ volatile ("mov %0, %%db6" : : "r" (clear));
+       }
+
+       pal_sti();
+
+       KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
+           (MACHDBG_CODE(DBG_MACH_EXCP_UTRAP_x86, type)) | DBG_FUNC_NONE,
+           (unsigned)(vaddr >> 32), (unsigned)vaddr,
+           (unsigned)(rip >> 32), (unsigned)rip, 0);

 
        code = 0;
        subcode = 0;
        exc = 0;
 
 
        code = 0;
        subcode = 0;
        exc = 0;
 

-#if DEBUG_TRACE
-       kprintf("user_trap(0x%08x) type=%d vaddr=0x%016llx\n",
-               saved_state, type, vaddr);
-#endif
-       perfCallback fn = perfASTHook;
-       if (fn) {
-               myast = ast_pending();
-               if (*myast & AST_CHUD_ALL) {
-                       fn(type, saved_state, 0, 0);
-               }
-       }
-
-       /* Is there a hook? */
-       fn = perfTrapHook;
-       if (fn) {
-               if (fn(type, saved_state, 0, 0) == KERN_SUCCESS)
-                       return; /* If it succeeds, we are done... */
-       }
-
+#if CONFIG_DTRACE

        /*
         * DTrace does not consume all user traps, only INT_3's for now.
         * Avoid needlessly calling tempDTraceTrapHook here, and let the
         * INT_3 case handle them.
         */
        /*
         * DTrace does not consume all user traps, only INT_3's for now.
         * Avoid needlessly calling tempDTraceTrapHook here, and let the
         * INT_3 case handle them.
         */

+#endif
+

        DEBUG_KPRINT_SYSCALL_MASK(1,
        DEBUG_KPRINT_SYSCALL_MASK(1,

-               "user_trap: type=0x%x(%s) err=0x%x cr2=%p rip=%p\n",
-               type, trap_type[type], err, (void *)(long) vaddr, (void *)(long) rip);
+           "user_trap: type=0x%x(%s) err=0x%x cr2=%p rip=%p\n",
+           type, trap_type[type], err, (void *)(long) vaddr, (void *)(long) rip);

 
        switch (type) {
 
        switch (type) {

-
-           case T_DIVIDE_ERROR:
+       case T_DIVIDE_ERROR:

                exc = EXC_ARITHMETIC;
                code = EXC_I386_DIV;
                break;
 
                exc = EXC_ARITHMETIC;
                code = EXC_I386_DIV;
                break;
 

-           case T_DEBUG:
-               {
-                       pcb_t   pcb;
-                       long clear = 0; /* 32 bit for i386, 64 bit for x86_64 */
+       case T_DEBUG:
+       {
+               pcb_t   pcb;
+               /*
+                * Update the PCB with this processor's DR6 value
+                * in the event this was a debug register match.
+                */
+               pcb = THREAD_TO_PCB(thread);
+               if (pcb->ids) {

                        /*
                        /*

-                        * get dr6 and set it in the thread's pcb before
-                        * returning to userland
+                        * We can get and set the status register
+                        * in 32-bit mode even on a 64-bit thread
+                        * because the high order bits are not
+                        * used on x86_64

                         */
                         */

-                       pcb = thread->machine.pcb;
-                       if (pcb->ids) {
-                               /*
-                                * We can get and set the status register
-                                * in 32-bit mode even on a 64-bit thread
-                                * because the high order bits are not
-                                * used on x86_64
-                                */
-                               unsigned long dr6_temp; /* 32 bit for i386, 64 bit for x86_64 */
-                               __asm__ volatile ("mov %%db6, %0" : "=r" (dr6_temp)); /* Register constraint by necessity */
-                               if (thread_is_64bit(thread)) {
-                                       x86_debug_state64_t *ids = pcb->ids;
-                                       ids->dr6 = dr6_temp;
-                               } else { /* 32 bit thread */
-                                       x86_debug_state32_t *ids = pcb->ids;
-                                       ids->dr6 = (uint32_t) dr6_temp;
-                               }
-                               __asm__ volatile ("mov %0, %%db6" : : "r" (clear));
+                       if (thread_is_64bit_addr(thread)) {
+                               x86_debug_state64_t *ids = pcb->ids;
+                               ids->dr6 = dr6;
+                       } else {         /* 32 bit thread */
+                               x86_debug_state32_t *ids = pcb->ids;
+                               ids->dr6 = (uint32_t) dr6;

                        }
                        }

-                       exc = EXC_BREAKPOINT;
-                       code = EXC_I386_SGL;
-                       break;

                }
                }

-           case T_INT3:
+               exc = EXC_BREAKPOINT;
+               code = EXC_I386_SGL;
+               break;
+       }
+       case T_INT3:

 #if CONFIG_DTRACE
 #if CONFIG_DTRACE

-               if (dtrace_user_probe(saved_state) == KERN_SUCCESS)
+               if (dtrace_user_probe(saved_state) == KERN_SUCCESS) {

                        return; /* If it succeeds, we are done... */
                        return; /* If it succeeds, we are done... */

+               }

 #endif
                exc = EXC_BREAKPOINT;
                code = EXC_I386_BPT;
                break;
 
 #endif
                exc = EXC_BREAKPOINT;
                code = EXC_I386_BPT;
                break;
 

-           case T_OVERFLOW:
+       case T_OVERFLOW:

                exc = EXC_ARITHMETIC;
                code = EXC_I386_INTO;
                break;
 
                exc = EXC_ARITHMETIC;
                code = EXC_I386_INTO;
                break;
 

-           case T_OUT_OF_BOUNDS:
+       case T_OUT_OF_BOUNDS:

                exc = EXC_SOFTWARE;
                code = EXC_I386_BOUND;
                break;
 
                exc = EXC_SOFTWARE;
                code = EXC_I386_BOUND;
                break;
 

-           case T_INVALID_OPCODE:
+       case T_INVALID_OPCODE:
+#if !defined(RC_HIDE_XNU_J137)
+               fpUDflt(rip);   /* May return from exception directly */
+#endif

                exc = EXC_BAD_INSTRUCTION;
                code = EXC_I386_INVOP;
                break;
 
                exc = EXC_BAD_INSTRUCTION;
                code = EXC_I386_INVOP;
                break;
 

-           case T_NO_FPU:
+       case T_NO_FPU:

                fpnoextflt();
                return;
 
                fpnoextflt();
                return;
 

-           case T_FPU_FAULT:
+       case T_FPU_FAULT:

                fpextovrflt(); /* Propagates exception directly, doesn't return */
                return;
 
                fpextovrflt(); /* Propagates exception directly, doesn't return */
                return;
 

-           case T_INVALID_TSS: /* invalid TSS == iret with NT flag set */
+       case T_INVALID_TSS:     /* invalid TSS == iret with NT flag set */

                exc = EXC_BAD_INSTRUCTION;
                code = EXC_I386_INVTSSFLT;
                subcode = err;
                break;
 
                exc = EXC_BAD_INSTRUCTION;
                code = EXC_I386_INVTSSFLT;
                subcode = err;
                break;
 

-           case T_SEGMENT_NOT_PRESENT:
+       case T_SEGMENT_NOT_PRESENT:

                exc = EXC_BAD_INSTRUCTION;
                code = EXC_I386_SEGNPFLT;
                subcode = err;
                break;
 
                exc = EXC_BAD_INSTRUCTION;
                code = EXC_I386_SEGNPFLT;
                subcode = err;
                break;
 

-           case T_STACK_FAULT:
+       case T_STACK_FAULT:

                exc = EXC_BAD_INSTRUCTION;
                code = EXC_I386_STKFLT;
                subcode = err;
                break;
 
                exc = EXC_BAD_INSTRUCTION;
                code = EXC_I386_STKFLT;
                subcode = err;
                break;
 

-           case T_GENERAL_PROTECTION:
+       case T_GENERAL_PROTECTION:

                /*
                 * There's a wide range of circumstances which generate this
                 * class of exception. From user-space, many involve bad
                /*
                 * There's a wide range of circumstances which generate this
                 * class of exception. From user-space, many involve bad


@@ -1330,43 +1070,80 @@ user_trap(
                 * to EXC_BAD_ACCESS (and thence SIGSEGV) also - rather than
                 * EXC_BAD_INSTRUCTION which is more accurate. We just can't
                 * win!
                 * to EXC_BAD_ACCESS (and thence SIGSEGV) also - rather than
                 * EXC_BAD_INSTRUCTION which is more accurate. We just can't
                 * win!

-                */ 
+                */

                exc = EXC_BAD_ACCESS;
                code = EXC_I386_GPFLT;
                subcode = err;
                break;
 
                exc = EXC_BAD_ACCESS;
                code = EXC_I386_GPFLT;
                subcode = err;
                break;
 

-           case T_PAGE_FAULT:
+       case T_PAGE_FAULT:
+       {

                prot = VM_PROT_READ;
 
                prot = VM_PROT_READ;
 

-               if (err & T_PF_WRITE)
-                       prot |= VM_PROT_WRITE;
-#if     PAE
-               if (err & T_PF_EXECUTE)
-                       prot |= VM_PROT_EXECUTE;
+               if (err & T_PF_WRITE) {
+                       prot |= VM_PROT_WRITE;
+               }
+               if (__improbable(err & T_PF_EXECUTE)) {
+                       prot |= VM_PROT_EXECUTE;
+               }
+#if DEVELOPMENT || DEBUG
+               uint32_t fsig = 0;
+               fsig = thread_fpsimd_hash(thread);
+#if DEBUG
+               fsigs[0] = fsig;
+#endif
+#endif
+               kret = vm_fault(thread->map,
+                   vaddr,
+                   prot, FALSE, VM_KERN_MEMORY_NONE,
+                   THREAD_ABORTSAFE, NULL, 0);
+#if DEVELOPMENT || DEBUG
+               if (fsig) {
+                       uint32_t fsig2 = thread_fpsimd_hash(thread);
+#if DEBUG
+                       fsigcs++;
+                       fsigs[1] = fsig2;
+#endif
+                       if (fsig != fsig2) {
+                               panic("FP/SIMD state hash mismatch across fault thread: %p 0x%x->0x%x", thread, fsig, fsig2);
+                       }
+               } else {
+#if DEBUG
+                       fsigns++;
+#endif
+               }

 #endif
 #endif

-               kret = vm_fault(thread->map, vm_map_trunc_page(vaddr),
-                                prot, FALSE,
-                                THREAD_ABORTSAFE, NULL, 0);
+               if (__probable((kret == KERN_SUCCESS) || (kret == KERN_ABORTED))) {
+                       thread_exception_return();
+                       /*NOTREACHED*/
+               }

 
 

-               user_page_fault_continue(kret);
-       
-               /* NOTREACHED */
-               break;
+               /*
+                * For a user trap, vm_fault() should never return KERN_FAILURE.
+                * If it does, we're leaking preemption disables somewhere in the kernel.
+                */
+               if (__improbable(kret == KERN_FAILURE)) {
+                       panic("vm_fault() KERN_FAILURE from user fault on thread %p", thread);
+               }

 
 

-           case T_SSE_FLOAT_ERROR:
+               user_page_fault_continue(kret);
+       }       /* NOTREACHED */
+       break;
+
+       case T_SSE_FLOAT_ERROR:

                fpSSEexterrflt(); /* Propagates exception directly, doesn't return */
                return;
 
 
                fpSSEexterrflt(); /* Propagates exception directly, doesn't return */
                return;
 
 

-           case T_FLOATING_POINT_ERROR:
+       case T_FLOATING_POINT_ERROR:

                fpexterrflt(); /* Propagates exception directly, doesn't return */
                return;
 
                fpexterrflt(); /* Propagates exception directly, doesn't return */
                return;
 

-           case T_DTRACE_RET:
+       case T_DTRACE_RET:

 #if CONFIG_DTRACE
 #if CONFIG_DTRACE

-               if (dtrace_user_probe(saved_state) == KERN_SUCCESS)
+               if (dtrace_user_probe(saved_state) == KERN_SUCCESS) {

                        return; /* If it succeeds, we are done... */
                        return; /* If it succeeds, we are done... */

+               }

 #endif
                /*
                 * If we get an INT 0x7f when we do not expect to,
 #endif
                /*
                 * If we get an INT 0x7f when we do not expect to,


@@ -1376,17 +1153,8 @@ user_trap(
                code = EXC_I386_INVOP;
                break;
 
                code = EXC_I386_INVOP;
                break;
 

-           default:
-#if     MACH_KGDB
-               Debugger("Unanticipated user trap");
-               return;
-#endif  /* MACH_KGDB */
-#if    MACH_KDB
-               if (kdb_trap(type, err, saved_state32(saved_state)))
-                   return;
-#endif /* MACH_KDB */
+       default:

                panic("Unexpected user trap, type %d", type);
                panic("Unexpected user trap, type %d", type);

-               return;

        }
        /* Note: Codepaths that directly return from user_trap() have pending
         * ASTs processed in locore
        }
        /* Note: Codepaths that directly return from user_trap() have pending
         * ASTs processed in locore


@@ -1395,31 +1163,6 @@ user_trap(
        /* NOTREACHED */
 }
 
        /* NOTREACHED */
 }
 

-
-/*
- * Handle AST traps for i386.
- * Check for delayed floating-point exception from
- * AT-bus machines.
- */
-
-extern void     log_thread_action (thread_t, char *);
-
-void
-i386_astintr(int preemption)
-{
-       ast_t           mask = AST_ALL;
-       spl_t           s;
-
-       if (preemption)
-               mask = AST_PREEMPTION;
-
-       s = splsched();
-
-       ast_taken(mask, s);
-
-       splx(s);
-}
-

 /*
  * Handle exceptions for i386.
  *
 /*
  * Handle exceptions for i386.
  *


@@ -1432,131 +1175,51 @@ i386_astintr(int preemption)
  */
 void
 i386_exception(
  */
 void
 i386_exception(

-       int     exc,
+       int     exc,

        mach_exception_code_t code,
        mach_exception_subcode_t subcode)
 {
        mach_exception_data_type_t   codes[EXCEPTION_CODE_MAX];
 
        DEBUG_KPRINT_SYSCALL_MACH("i386_exception: exc=%d code=0x%llx subcode=0x%llx\n",
        mach_exception_code_t code,
        mach_exception_subcode_t subcode)
 {
        mach_exception_data_type_t   codes[EXCEPTION_CODE_MAX];
 
        DEBUG_KPRINT_SYSCALL_MACH("i386_exception: exc=%d code=0x%llx subcode=0x%llx\n",

-                                                         exc, code, subcode);
-       codes[0] = code;                /* new exception interface */
+           exc, code, subcode);
+       codes[0] = code;                /* new exception interface */

        codes[1] = subcode;
        exception_triage(exc, codes, 2);
        /*NOTREACHED*/
 }
 
 
        codes[1] = subcode;
        exception_triage(exc, codes, 2);
        /*NOTREACHED*/
 }
 
 

-
-void
-kernel_preempt_check(void)
-{
-       ast_t           *myast;
-       boolean_t       intr;
-
-       /*
-        * disable interrupts to both prevent pre-emption
-        * and to keep the ast state from changing via
-        * an interrupt handler making something runnable
-        */
-       intr = ml_set_interrupts_enabled(FALSE);
-
-       myast = ast_pending();
-
-       if ((*myast & AST_URGENT) && intr == TRUE && get_interrupt_level() == 0) {
-               /*
-                * can handle interrupts and preemptions 
-                * at this point
-                */
-               ml_set_interrupts_enabled(intr);
-
-               /*
-                * now cause the PRE-EMPTION trap
-                */
-               __asm__ volatile ("     int     $0xff");
-       } else {
-               /*
-                * if interrupts were already disabled or
-                * we're in an interrupt context, we can't
-                * preempt...  of course if AST_URGENT
-                * isn't set we also don't want to
-                */
-               ml_set_interrupts_enabled(intr);
-       }
-}
-
-#if    MACH_KDB
-
-extern void    db_i386_state(x86_saved_state32_t *regs);
-
-#include <ddb/db_output.h>
-
-void 
-db_i386_state(
-       x86_saved_state32_t *regs)
-{
-       db_printf("eip  %8x\n", regs->eip);
-       db_printf("trap %8x\n", regs->trapno);
-       db_printf("err  %8x\n", regs->err);
-       db_printf("efl  %8x\n", regs->efl);
-       db_printf("ebp  %8x\n", regs->ebp);
-       db_printf("esp  %8x\n", regs->cr2);
-       db_printf("uesp %8x\n", regs->uesp);
-       db_printf("cs   %8x\n", regs->cs & 0xff);
-       db_printf("ds   %8x\n", regs->ds & 0xff);
-       db_printf("es   %8x\n", regs->es & 0xff);
-       db_printf("fs   %8x\n", regs->fs & 0xff);
-       db_printf("gs   %8x\n", regs->gs & 0xff);
-       db_printf("ss   %8x\n", regs->ss & 0xff);
-       db_printf("eax  %8x\n", regs->eax);
-       db_printf("ebx  %8x\n", regs->ebx);
-       db_printf("ecx  %8x\n", regs->ecx);
-       db_printf("edx  %8x\n", regs->edx);
-       db_printf("esi  %8x\n", regs->esi);
-       db_printf("edi  %8x\n", regs->edi);
-}
-
-#endif /* MACH_KDB */
-
-/* Synchronize a thread's i386_kernel_state (if any) with the given
- * i386_saved_state_t obtained from the trap/IPI handler; called in
+/* Synchronize a thread's x86_kernel_state (if any) with the given
+ * x86_saved_state_t obtained from the trap/IPI handler; called in

  * kernel_trap() prior to entering the debugger, and when receiving
  * kernel_trap() prior to entering the debugger, and when receiving

- * an "MP_KDP" IPI.
+ * an "MP_KDP" IPI. Called with null saved_state if an incoming IPI
+ * was detected from the kernel while spinning with interrupts masked.

  */
  */

-  
+

 void
 sync_iss_to_iks(x86_saved_state_t *saved_state)
 {
 void
 sync_iss_to_iks(x86_saved_state_t *saved_state)
 {

-       struct x86_kernel_state *iks;
+       struct x86_kernel_state *iks = NULL;

        vm_offset_t kstack;
        boolean_t record_active_regs = FALSE;
 
        vm_offset_t kstack;
        boolean_t record_active_regs = FALSE;
 

-       if ((kstack = current_thread()->kernel_stack) != 0) {
-#ifdef __i386__
-               x86_saved_state32_t     *regs = saved_state32(saved_state);
-#else
-               x86_saved_state64_t     *regs = saved_state64(saved_state);
-#endif
+       /* The PAL may have a special way to sync registers */
+       if (saved_state && saved_state->flavor == THREAD_STATE_NONE) {
+               pal_get_kern_regs( saved_state );
+       }

 
 

-               iks = STACK_IKS(kstack);
+       if (current_thread() != NULL &&
+           (kstack = current_thread()->kernel_stack) != 0) {
+               x86_saved_state64_t     *regs = saved_state64(saved_state);

 
 

+               iks = STACK_IKS(kstack);

 
 

-                /* Did we take the trap/interrupt in kernel mode? */
-#ifdef __i386__
-               if (regs == USER_REGS32(current_thread()))
-                       record_active_regs = TRUE;
-               else {
-                       iks->k_ebx = regs->ebx;
-                       iks->k_esp = (int)regs;
-                       iks->k_ebp = regs->ebp;
-                       iks->k_edi = regs->edi;
-                       iks->k_esi = regs->esi;
-                       iks->k_eip = regs->eip;
-               }
-#else
-               if (regs == USER_REGS64(current_thread()))
-                       record_active_regs = TRUE;
-               else {
+               /* Did we take the trap/interrupt in kernel mode? */
+               if (saved_state == NULL || /* NULL => polling in kernel */
+                   regs == USER_REGS64(current_thread())) {
+                       record_active_regs = TRUE;
+               } else {

                        iks->k_rbx = regs->rbx;
                        iks->k_rsp = regs->isf.rsp;
                        iks->k_rbp = regs->rbp;
                        iks->k_rbx = regs->rbx;
                        iks->k_rsp = regs->isf.rsp;
                        iks->k_rbp = regs->rbp;


@@ -1566,34 +1229,22 @@ sync_iss_to_iks(x86_saved_state_t *saved_state)
                        iks->k_r15 = regs->r15;
                        iks->k_rip = regs->isf.rip;
                }
                        iks->k_r15 = regs->r15;
                        iks->k_rip = regs->isf.rip;
                }

-#endif

        }
 
        if (record_active_regs == TRUE) {
        }
 
        if (record_active_regs == TRUE) {

-#ifdef __i386__

                /* Show the trap handler path */
                /* Show the trap handler path */

-               __asm__ volatile("movl %%ebx, %0" : "=m" (iks->k_ebx));
-               __asm__ volatile("movl %%esp, %0" : "=m" (iks->k_esp));
-               __asm__ volatile("movl %%ebp, %0" : "=m" (iks->k_ebp));
-               __asm__ volatile("movl %%edi, %0" : "=m" (iks->k_edi));
-               __asm__ volatile("movl %%esi, %0" : "=m" (iks->k_esi));
+               __asm__ volatile ("movq %%rbx, %0" : "=m" (iks->k_rbx));
+               __asm__ volatile ("movq %%rsp, %0" : "=m" (iks->k_rsp));
+               __asm__ volatile ("movq %%rbp, %0" : "=m" (iks->k_rbp));
+               __asm__ volatile ("movq %%r12, %0" : "=m" (iks->k_r12));
+               __asm__ volatile ("movq %%r13, %0" : "=m" (iks->k_r13));
+               __asm__ volatile ("movq %%r14, %0" : "=m" (iks->k_r14));
+               __asm__ volatile ("movq %%r15, %0" : "=m" (iks->k_r15));

                /* "Current" instruction pointer */
                /* "Current" instruction pointer */

-               __asm__ volatile("movl $1f, %0\n1:" : "=m" (iks->k_eip));
-#else
-               /* Show the trap handler path */
-               __asm__ volatile("movq %%rbx, %0" : "=m" (iks->k_rbx));
-               __asm__ volatile("movq %%rsp, %0" : "=m" (iks->k_rsp));
-               __asm__ volatile("movq %%rbp, %0" : "=m" (iks->k_rbp));
-               __asm__ volatile("movq %%r12, %0" : "=m" (iks->k_r12));
-               __asm__ volatile("movq %%r13, %0" : "=m" (iks->k_r13));
-               __asm__ volatile("movq %%r14, %0" : "=m" (iks->k_r14));
-               __asm__ volatile("movq %%r15, %0" : "=m" (iks->k_r15));
-               /* "Current" instruction pointer */
-               __asm__ volatile("leaq 1f(%%rip), %%rax; mov %%rax, %0\n1:"
-                                : "=m" (iks->k_rip)
-                                :
-                                : "rax");
-#endif
+               __asm__ volatile ("leaq 1f(%%rip), %%rax; mov %%rax, %0\n1:"
+                                  : "=m" (iks->k_rip)
+                                  :
+                                  : "rax");

        }
 }
 
        }
 }
 


@@ -1604,32 +1255,52 @@ sync_iss_to_iks(x86_saved_state_t *saved_state)
  * or user space.
  */
 void
  * or user space.
  */
 void

-sync_iss_to_iks_unconditionally(__unused x86_saved_state_t *saved_state) {
+sync_iss_to_iks_unconditionally(__unused x86_saved_state_t *saved_state)
+{

        struct x86_kernel_state *iks;
        vm_offset_t kstack;
 
        if ((kstack = current_thread()->kernel_stack) != 0) {
                iks = STACK_IKS(kstack);
        struct x86_kernel_state *iks;
        vm_offset_t kstack;
 
        if ((kstack = current_thread()->kernel_stack) != 0) {
                iks = STACK_IKS(kstack);

-#ifdef __i386__
-               /* Display the trap handler path */
-               __asm__ volatile("movl %%ebx, %0" : "=m" (iks->k_ebx));
-               __asm__ volatile("movl %%esp, %0" : "=m" (iks->k_esp));
-               __asm__ volatile("movl %%ebp, %0" : "=m" (iks->k_ebp));
-               __asm__ volatile("movl %%edi, %0" : "=m" (iks->k_edi));
-               __asm__ volatile("movl %%esi, %0" : "=m" (iks->k_esi));
-               /* "Current" instruction pointer */
-               __asm__ volatile("movl $1f, %0\n1:" : "=m" (iks->k_eip));
-#else

                /* Display the trap handler path */
                /* Display the trap handler path */

-               __asm__ volatile("movq %%rbx, %0" : "=m" (iks->k_rbx));
-               __asm__ volatile("movq %%rsp, %0" : "=m" (iks->k_rsp));
-               __asm__ volatile("movq %%rbp, %0" : "=m" (iks->k_rbp));
-               __asm__ volatile("movq %%r12, %0" : "=m" (iks->k_r12));
-               __asm__ volatile("movq %%r13, %0" : "=m" (iks->k_r13));
-               __asm__ volatile("movq %%r14, %0" : "=m" (iks->k_r14));
-               __asm__ volatile("movq %%r15, %0" : "=m" (iks->k_r15));
+               __asm__ volatile ("movq %%rbx, %0" : "=m" (iks->k_rbx));
+               __asm__ volatile ("movq %%rsp, %0" : "=m" (iks->k_rsp));
+               __asm__ volatile ("movq %%rbp, %0" : "=m" (iks->k_rbp));
+               __asm__ volatile ("movq %%r12, %0" : "=m" (iks->k_r12));
+               __asm__ volatile ("movq %%r13, %0" : "=m" (iks->k_r13));
+               __asm__ volatile ("movq %%r14, %0" : "=m" (iks->k_r14));
+               __asm__ volatile ("movq %%r15, %0" : "=m" (iks->k_r15));

                /* "Current" instruction pointer */
                /* "Current" instruction pointer */

-               __asm__ volatile("leaq 1f(%%rip), %%rax; mov %%rax, %0\n1:" : "=m" (iks->k_rip)::"rax");
+               __asm__ volatile ("leaq 1f(%%rip), %%rax; mov %%rax, %0\n1:" : "=m" (iks->k_rip)::"rax");
+       }
+}
+
+#if DEBUG
+#define TERI 1

 #endif
 #endif

+
+#if TERI
+extern void     thread_exception_return_internal(void) __dead2;
+
+void
+thread_exception_return(void)
+{
+       thread_t thread = current_thread();
+       ml_set_interrupts_enabled(FALSE);
+       if (thread_is_64bit_addr(thread) != task_has_64Bit_addr(thread->task)) {
+               panic("Task/thread bitness mismatch %p %p, task: %d, thread: %d", thread, thread->task, thread_is_64bit_addr(thread), task_has_64Bit_addr(thread->task));

        }
        }

+
+       if (thread_is_64bit_addr(thread)) {
+               if ((gdt_desc_p(USER64_CS)->access & ACC_PL_U) == 0) {
+                       panic("64-GDT mismatch %p, descriptor: %p", thread, gdt_desc_p(USER64_CS));
+               }
+       } else {
+               if ((gdt_desc_p(USER_CS)->access & ACC_PL_U) == 0) {
+                       panic("32-GDT mismatch %p, descriptor: %p", thread, gdt_desc_p(USER_CS));
+               }
+       }
+       assert(get_preemption_level() == 0);
+       thread_exception_return_internal();

 }
 }

+#endif