]> git.saurik.com Git - apple/xnu.git/blobdiff - osfmk/kern/task.c
xnu-1504.7.4.tar.gz
[apple/xnu.git] / osfmk / kern / task.c
index 6911821503806905d95c37b99610733c6ee761e5..aedd993a153d758dc05f6531d01a8e199f1719d7 100644 (file)
@@ -1,31 +1,29 @@
 /*
- * Copyright (c) 2000-2005 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2009 Apple Inc. All rights reserved.
  *
- * @APPLE_LICENSE_OSREFERENCE_HEADER_START@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
- * This file contains Original Code and/or Modifications of Original Code 
- * as defined in and that are subject to the Apple Public Source License 
- * Version 2.0 (the 'License'). You may not use this file except in 
- * compliance with the License.  The rights granted to you under the 
- * License may not be used to create, or enable the creation or 
- * redistribution of, unlawful or unlicensed copies of an Apple operating 
- * system, or to circumvent, violate, or enable the circumvention or 
- * violation of, any terms of an Apple operating system software license 
- * agreement.
- *
- * Please obtain a copy of the License at 
- * http://www.opensource.apple.com/apsl/ and read it before using this 
- * file.
- *
- * The Original Code and all software distributed under the License are 
- * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER 
- * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, 
- * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, 
- * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. 
- * Please see the License for the specific language governing rights and 
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ * 
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ * 
+ * The Original Code and all software distributed under the License are
+ * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
  * limitations under the License.
- *
- * @APPLE_LICENSE_OSREFERENCE_HEADER_END@
+ * 
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /*
  * @OSF_FREE_COPYRIGHT@
  * improvements that they make and grant CSL redistribution rights.
  *
  */
+/*
+ * NOTICE: This file was modified by McAfee Research in 2004 to introduce
+ * support for mandatory and extensible security protections.  This notice
+ * is included in support of clause 2.2 (b) of the Apple Public License,
+ * Version 2.0.
+ * Copyright (c) 2005 SPARTA, Inc.
+ */
 
 #include <mach_kdb.h>
-#include <mach_host.h>
-#include <mach_prof.h>
 #include <fast_tas.h>
 #include <platforms.h>
 
 #include <kern/host.h>
 #include <kern/clock.h>
 #include <kern/timer.h>
-#include <kern/profile.h>
 #include <kern/assert.h>
 #include <kern/sync_lock.h>
+#include <kern/affinity.h>
 
 #include <vm/pmap.h>
 #include <vm/vm_map.h>
 #include <vm/vm_kern.h>                /* for kernel_map, ipc_kernel_map */
 #include <vm/vm_pageout.h>
-#include <vm/vm_protos.h>      /* for vm_map_remove_commpage64 */
+#include <vm/vm_protos.h>
 
 #if    MACH_KDB
 #include <ddb/db_sym.h>
 #include <ppc/hw_perfmon.h>
 #endif
 
+
 /*
  * Exported interfaces
  */
 #include <mach/mach_host_server.h>
 #include <mach/host_security_server.h>
 #include <mach/mach_port_server.h>
+#include <mach/security_server.h>
+
+#include <vm/vm_shared_region.h>
+
+#if CONFIG_MACF_MACH
+#include <security/mac_mach_internal.h>
+#endif
+
+#if CONFIG_COUNTERS
+#include <pmc/pmc.h>
+#endif /* CONFIG_COUNTERS */
 
-#include <vm/task_working_set.h>
-#include <vm/vm_shared_memory_server.h>
+task_t                 kernel_task;
+zone_t                 task_zone;
+lck_attr_t      task_lck_attr;
+lck_grp_t       task_lck_grp;
+lck_grp_attr_t  task_lck_grp_attr;
 
-task_t kernel_task;
-zone_t task_zone;
+int task_max = CONFIG_TASK_MAX; /* Max number of tasks */
 
 /* Forwards */
 
@@ -166,6 +183,9 @@ kern_return_t       task_set_ledger(
                        ledger_t        wired,
                        ledger_t        paged);
 
+int check_for_tasksuspend(
+                       task_t task);
+
 void
 task_backing_store_privileged(
                        task_t task)
@@ -176,119 +196,110 @@ task_backing_store_privileged(
        return;
 }
 
-void
-task_working_set_disable(task_t task)
-{
-       struct tws_hash *ws;
-
-       task_lock(task);
-       ws = task->dynamic_working_set;
-       task->dynamic_working_set = NULL;
-       task_unlock(task);
-       if (ws) {
-               tws_hash_ws_flush(ws);
-               tws_hash_destroy(ws);
-       }
-}
 
 void
 task_set_64bit(
                task_t task,
                boolean_t is64bit)
 {
-       if(is64bit) {
-               /* LP64todo - no task working set for 64-bit */
+#if defined(__i386__) || defined(__x86_64__)
+       thread_t thread;
+#endif /* __i386__ */
+       int     vm_flags = 0;
+
+       if (is64bit) {
+               if (task_has_64BitAddr(task))
+                       return;
+
                task_set_64BitAddr(task);
-               task_working_set_disable(task);
-               task->map->max_offset = MACH_VM_MAX_ADDRESS;
        } else {
+               if ( !task_has_64BitAddr(task))
+                       return;
+
                /*
                 * Deallocate all memory previously allocated
                 * above the 32-bit address space, since it won't
                 * be accessible anymore.
                 */
-               /* LP64todo - make this clean */
+               /* remove regular VM map entries & pmap mappings */
+               (void) vm_map_remove(task->map,
+                                    (vm_map_offset_t) VM_MAX_ADDRESS,
+                                    MACH_VM_MAX_ADDRESS,
+                                    0);
 #ifdef __ppc__
-               vm_map_remove_commpage64(task->map);
+               /*
+                * PPC51: ppc64 is limited to 51-bit addresses.
+                * Memory mapped above that limit is handled specially
+                * at the pmap level, so let pmap clean the commpage mapping
+                * explicitly...
+                */
                pmap_unmap_sharedpage(task->map->pmap); /* Unmap commpage */
-#endif
+               /* ... and avoid regular pmap cleanup */
+               vm_flags |= VM_MAP_REMOVE_NO_PMAP_CLEANUP;
+#endif /* __ppc__ */
+               /* remove the higher VM mappings */
                (void) vm_map_remove(task->map,
-                                    (vm_map_offset_t) VM_MAX_ADDRESS,
                                     MACH_VM_MAX_ADDRESS,
-                                    VM_MAP_NO_FLAGS);
+                                    0xFFFFFFFFFFFFF000ULL,
+                                    vm_flags);
                task_clear_64BitAddr(task);
-               task->map->max_offset = (vm_map_offset_t)VM_MAX_ADDRESS;
        }
+       /* FIXME: On x86, the thread save state flavor can diverge from the
+        * task's 64-bit feature flag due to the 32-bit/64-bit register save
+        * state dichotomy. Since we can be pre-empted in this interval,
+        * certain routines may observe the thread as being in an inconsistent
+        * state with respect to its task's 64-bitness.
+        */
+#if defined(__i386__) || defined(__x86_64__)
+       task_lock(task);
+       queue_iterate(&task->threads, thread, thread_t, task_threads) {
+               thread_mtx_lock(thread);
+               machine_thread_switch_addrmode(thread);
+               thread_mtx_unlock(thread);
+       }
+       task_unlock(task);
+#endif /* __i386__ */
+}
+
+
+void
+task_set_dyld_info(task_t task, mach_vm_address_t addr, mach_vm_size_t size)
+{
+       task_lock(task);
+       task->all_image_info_addr = addr;
+       task->all_image_info_size = size;
+       task_unlock(task);
 }
 
 void
 task_init(void)
 {
+
+       lck_grp_attr_setdefault(&task_lck_grp_attr);
+       lck_grp_init(&task_lck_grp, "task", &task_lck_grp_attr);
+       lck_attr_setdefault(&task_lck_attr);
+       lck_mtx_init(&tasks_threads_lock, &task_lck_grp, &task_lck_attr);
+
        task_zone = zinit(
                        sizeof(struct task),
-                       TASK_MAX * sizeof(struct task),
+                       task_max * sizeof(struct task),
                        TASK_CHUNK * sizeof(struct task),
                        "tasks");
 
        /*
         * Create the kernel task as the first task.
         */
-       if (task_create_internal(TASK_NULL, FALSE, &kernel_task) != KERN_SUCCESS)
+#ifdef __LP64__
+       if (task_create_internal(TASK_NULL, FALSE, TRUE, &kernel_task) != KERN_SUCCESS)
+#else
+       if (task_create_internal(TASK_NULL, FALSE, FALSE, &kernel_task) != KERN_SUCCESS)
+#endif
                panic("task_init\n");
 
        vm_map_deallocate(kernel_task->map);
        kernel_task->map = kernel_map;
 }
 
-#if    MACH_HOST
-
-#if 0
-static void
-task_freeze(
-       task_t task)
-{
-       task_lock(task);
-       /*
-        *      If may_assign is false, task is already being assigned,
-        *      wait for that to finish.
-        */
-       while (task->may_assign == FALSE) {
-               wait_result_t res;
-
-               task->assign_active = TRUE;
-               res = thread_sleep_mutex((event_t) &task->assign_active,
-                                        &task->lock, THREAD_UNINT);
-               assert(res == THREAD_AWAKENED);
-       }
-       task->may_assign = FALSE;
-       task_unlock(task);
-       return;
-}
-#else
-#define thread_freeze(thread)  assert(task->processor_set == &default_pset)
-#endif
-
-#if 0
-static void
-task_unfreeze(
-       task_t task)
-{
-       task_lock(task);
-       assert(task->may_assign == FALSE);
-       task->may_assign = TRUE;
-       if (task->assign_active == TRUE) {
-               task->assign_active = FALSE;
-               thread_wakeup((event_t)&task->assign_active);
-       }
-       task_unlock(task);
-       return;
-}
-#else
-#define thread_unfreeze(thread)        assert(task->processor_set == &default_pset)
-#endif
-
-#endif /* MACH_HOST */
-
 /*
  * Create a task running in the kernel address space.  It may
  * have its own map of size mem_size and may have ipc privileges.
@@ -305,65 +316,55 @@ kernel_task_create(
 
 kern_return_t
 task_create(
-       task_t                                  parent_task,
+       task_t                          parent_task,
        __unused ledger_port_array_t    ledger_ports,
-       __unused  mach_msg_type_number_t        num_ledger_ports,
-       boolean_t                               inherit_memory,
-       task_t                                  *child_task)    /* OUT */
+       __unused mach_msg_type_number_t num_ledger_ports,
+       __unused boolean_t              inherit_memory,
+       __unused task_t                 *child_task)    /* OUT */
 {
        if (parent_task == TASK_NULL)
                return(KERN_INVALID_ARGUMENT);
 
-       return task_create_internal(
-                       parent_task, inherit_memory, child_task);
+       /*
+        * No longer supported: too many calls assume that a task has a valid
+        * process attached.
+        */
+       return(KERN_FAILURE);
 }
 
 kern_return_t
 host_security_create_task_token(
        host_security_t                 host_security,
-       task_t                                  parent_task,
-       security_token_t                        sec_token,
-       audit_token_t                           audit_token,
-       host_priv_t                             host_priv,
+       task_t                          parent_task,
+       __unused security_token_t       sec_token,
+       __unused audit_token_t          audit_token,
+       __unused host_priv_t            host_priv,
        __unused ledger_port_array_t    ledger_ports,
        __unused mach_msg_type_number_t num_ledger_ports,
-       boolean_t                               inherit_memory,
-       task_t                                  *child_task)    /* OUT */
+       __unused boolean_t              inherit_memory,
+       __unused task_t                 *child_task)    /* OUT */
 {
-        kern_return_t          result;
-        
        if (parent_task == TASK_NULL)
                return(KERN_INVALID_ARGUMENT);
 
        if (host_security == HOST_NULL)
                return(KERN_INVALID_SECURITY);
 
-       result = task_create_internal(
-                       parent_task, inherit_memory, child_task);
-
-        if (result != KERN_SUCCESS)
-                return(result);
-
-       result = host_security_set_task_token(host_security,
-                                             *child_task,
-                                             sec_token,
-                                             audit_token,
-                                             host_priv);
-
-       if (result != KERN_SUCCESS)
-               return(result);
-
-       return(result);
+       /*
+        * No longer supported.
+        */
+       return(KERN_FAILURE);
 }
 
 kern_return_t
 task_create_internal(
        task_t          parent_task,
        boolean_t       inherit_memory,
+       boolean_t       is_64bit,
        task_t          *child_task)            /* OUT */
 {
-       task_t          new_task;
-       processor_set_t pset;
+       task_t                  new_task;
+       vm_shared_region_t      shared_region;
 
        new_task = (task_t) zalloc(task_zone);
 
@@ -373,14 +374,19 @@ task_create_internal(
        /* one ref for just being alive; one for our caller */
        new_task->ref_count = 2;
 
+       /* if inherit_memory is true, parent_task MUST not be NULL */
        if (inherit_memory)
                new_task->map = vm_map_fork(parent_task->map);
        else
-               new_task->map = vm_map_create(pmap_create(0),
+               new_task->map = vm_map_create(pmap_create(0, is_64bit),
                                        (vm_map_offset_t)(VM_MIN_ADDRESS),
                                        (vm_map_offset_t)(VM_MAX_ADDRESS), TRUE);
 
-       mutex_init(&new_task->lock, 0);
+       /* Inherit memlock limit from parent */
+       if (parent_task)
+               vm_map_set_user_wire_limit(new_task->map, (vm_size_t)parent_task->map->user_wire_limit);
+
+       lck_mtx_init(&new_task->lock, &task_lck_grp, &task_lck_attr);
        queue_init(&new_task->threads);
        new_task->suspend_count = 0;
        new_task->thread_count = 0;
@@ -388,7 +394,8 @@ task_create_internal(
        new_task->user_stop_count = 0;
        new_task->role = TASK_UNSPECIFIED;
        new_task->active = TRUE;
-       new_task->user_data = 0;
+       new_task->halting = FALSE;
+       new_task->user_data = NULL;
        new_task->faults = 0;
        new_task->cow_faults = 0;
        new_task->pageins = 0;
@@ -397,18 +404,20 @@ task_create_internal(
        new_task->syscalls_mach = 0;
        new_task->priv_flags = 0;
        new_task->syscalls_unix=0;
-       new_task->csw=0;
+       new_task->c_switch = new_task->p_switch = new_task->ps_switch = 0;
        new_task->taskFeatures[0] = 0;                          /* Init task features */
        new_task->taskFeatures[1] = 0;                          /* Init task features */
-       new_task->dynamic_working_set = 0;
-
-       task_working_set_create(new_task, TWS_SMALL_HASH_LINE_COUNT, 
-                               0, TWS_HASH_STYLE_DEFAULT);
 
 #ifdef MACH_BSD
-       new_task->bsd_info = 0;
+       new_task->bsd_info = NULL;
 #endif /* MACH_BSD */
 
+#if defined(__i386__) || defined(__x86_64__)
+       new_task->i386_ldt = 0;
+       new_task->task_debug = NULL;
+
+#endif
+
 #ifdef __ppc__
        if(BootProcInfo.pf.Available & pf64Bit) new_task->taskFeatures[0] |= tf64BitData;       /* If 64-bit machine, show we have 64-bit registers at least */
 #endif
@@ -418,35 +427,33 @@ task_create_internal(
        new_task->semaphores_owned = 0;
        new_task->lock_sets_owned = 0;
 
-#if    MACH_HOST
-       new_task->may_assign = TRUE;
-       new_task->assign_active = FALSE;
-#endif /* MACH_HOST */
+#if CONFIG_MACF_MACH
+       new_task->label = labelh_new(1);
+       mac_task_label_init (&new_task->maclabel);
+#endif
 
        ipc_task_init(new_task, parent_task);
 
        new_task->total_user_time = 0;
        new_task->total_system_time = 0;
 
-       task_prof_init(new_task);
+       new_task->vtimers = 0;
 
-       if (parent_task != TASK_NULL) {
-#if    MACH_HOST
-               /*
-                * Freeze the parent, so that parent_task->processor_set
-                * cannot change.
-                */
-               task_freeze(parent_task);
-#endif /* MACH_HOST */
-               pset = parent_task->processor_set;
-               if (!pset->active)
-                       pset = &default_pset;
+       new_task->shared_region = NULL;
 
+       new_task->affinity_space = NULL;
+
+#if CONFIG_COUNTERS
+       new_task->t_chud = 0U;
+#endif
+
+       if (parent_task != TASK_NULL) {
                new_task->sec_token = parent_task->sec_token;
                new_task->audit_token = parent_task->audit_token;
 
-               shared_region_mapping_ref(parent_task->system_shared_region);
-               new_task->system_shared_region = parent_task->system_shared_region;
+               /* inherit the parent's shared region */
+               shared_region = vm_shared_region_get(parent_task);
+               vm_shared_region_set(new_task, shared_region);
 
                new_task->wired_ledger_port = ledger_copy(
                        convert_port_to_ledger(parent_task->wired_ledger_port));
@@ -454,14 +461,29 @@ task_create_internal(
                        convert_port_to_ledger(parent_task->paged_ledger_port));
                if(task_has_64BitAddr(parent_task))
                        task_set_64BitAddr(new_task);
+               new_task->all_image_info_addr = parent_task->all_image_info_addr;
+               new_task->all_image_info_size = parent_task->all_image_info_size;
+
+#if defined(__i386__) || defined(__x86_64__)
+               if (inherit_memory && parent_task->i386_ldt)
+                       new_task->i386_ldt = user_ldt_copy(parent_task->i386_ldt);
+#endif
+               if (inherit_memory && parent_task->affinity_space)
+                       task_affinity_create(parent_task, new_task);
+
+               new_task->pset_hint = parent_task->pset_hint = task_choose_pset(parent_task);
        }
        else {
-               pset = &default_pset;
-
                new_task->sec_token = KERNEL_SECURITY_TOKEN;
                new_task->audit_token = KERNEL_AUDIT_TOKEN;
                new_task->wired_ledger_port = ledger_copy(root_wired_ledger);
                new_task->paged_ledger_port = ledger_copy(root_paged_ledger);
+#ifdef __LP64__
+               if(is_64bit)
+                       task_set_64BitAddr(new_task);
+#endif
+
+               new_task->pset_hint = PROCESSOR_SET_NULL;
        }
 
        if (kernel_task == TASK_NULL) {
@@ -472,14 +494,11 @@ task_create_internal(
                new_task->priority = BASEPRI_DEFAULT;
                new_task->max_priority = MAXPRI_USER;
        }
-
-       pset_lock(pset);
-       pset_add_task(pset, new_task);
-       pset_unlock(pset);
-#if    MACH_HOST
-       if (parent_task != TASK_NULL)
-               task_unfreeze(parent_task);
-#endif /* MACH_HOST */
+       
+       lck_mtx_lock(&tasks_threads_lock);
+       queue_enter(&tasks, new_task, task_t, tasks);
+       tasks_count++;
+       lck_mtx_unlock(&tasks_threads_lock);
 
        if (vm_backing_store_low && parent_task != NULL)
                new_task->priv_flags |= (parent_task->priv_flags&VM_BACKING_STORE_PRIV);
@@ -499,29 +518,41 @@ void
 task_deallocate(
        task_t          task)
 {
-       processor_set_t         pset;
-
        if (task == TASK_NULL)
            return;
 
        if (task_deallocate_internal(task) > 0)
                return;
 
-       pset = task->processor_set;
-       pset_deallocate(pset);
-
-       if(task->dynamic_working_set)
-               tws_hash_destroy(task->dynamic_working_set);
-
        ipc_task_terminate(task);
 
+       if (task->affinity_space)
+               task_affinity_deallocate(task);
+
        vm_map_deallocate(task->map);
        is_release(task->itk_space);
 
-       task_prof_deallocate(task);
+       lck_mtx_destroy(&task->lock, &task_lck_grp);
+
+#if CONFIG_MACF_MACH
+       labelh_release(task->label);
+#endif
        zfree(task_zone, task);
 }
 
+/*
+ *     task_name_deallocate:
+ *
+ *     Drop a reference on a task name.
+ */
+void
+task_name_deallocate(
+       task_name_t             task_name)
+{
+       return(task_deallocate((task_t)task_name));
+}
+
+
 /*
  *     task_terminate:
  *
@@ -546,7 +577,6 @@ kern_return_t
 task_terminate_internal(
        task_t                  task)
 {
-       processor_set_t         pset;
        thread_t                        thread, self;
        task_t                          self_task;
        boolean_t                       interrupt_save;
@@ -634,11 +664,15 @@ task_terminate_internal(
         */
        ipc_space_destroy(task->itk_space);
 
-/* LP64todo - make this clean */
 #ifdef __ppc__
-       vm_map_remove_commpage64(task->map);
+       /*
+        * PPC51: ppc64 is limited to 51-bit addresses.
+        */
        pmap_unmap_sharedpage(task->map->pmap);         /* Unmap commpage */
-#endif
+#endif /* __ppc__ */
+
+       if (vm_map_has_4GB_pagezero(task->map))
+               vm_map_clear_4GB_pagezero(task->map);
 
        /*
         * If the current thread is a member of the task
@@ -648,21 +682,18 @@ task_terminate_internal(
         * expense of removing the address space regions
         * at reap time, we do it explictly here.
         */
-       vm_map_remove(task->map, task->map->min_offset,
-                                                               task->map->max_offset, VM_MAP_NO_FLAGS);
-
-       shared_region_mapping_dealloc(task->system_shared_region);
+       vm_map_remove(task->map,
+                     task->map->min_offset,
+                     task->map->max_offset,
+                     VM_MAP_NO_FLAGS);
 
-       /*
-        * Flush working set here to avoid I/O in reaper thread
-        */
-       if (task->dynamic_working_set)
-               tws_hash_ws_flush(task->dynamic_working_set);
+       /* release our shared region */
+       vm_shared_region_set(task, NULL);
 
-       pset = task->processor_set;
-       pset_lock(pset);
-       pset_remove_task(pset,task);
-       pset_unlock(pset);
+       lck_mtx_lock(&tasks_threads_lock);
+       queue_remove(&tasks, task, task_t, tasks);
+       tasks_count--;
+       lck_mtx_unlock(&tasks_threads_lock);
 
        /*
         * We no longer need to guard against being aborted, so restore
@@ -683,16 +714,15 @@ task_terminate_internal(
 }
 
 /*
- * task_halt:
+ * task_start_halt:
  *
  *     Shut the current task down (except for the current thread) in
  *     preparation for dramatic changes to the task (probably exec).
- *     We hold the task, terminate all other threads in the task and
- *     wait for them to terminate, clean up the portspace, and when
- *     all done, let the current thread go.
+ *     We hold the task and mark all other threads in the task for
+ *     termination.
  */
 kern_return_t
-task_halt(
+task_start_halt(
        task_t          task)
 {
        thread_t        thread, self;
@@ -706,7 +736,7 @@ task_halt(
 
        task_lock(task);
 
-       if (!task->active || !self->active) {
+       if (task->halting || !task->active || !self->active) {
                /*
                 *      Task or current thread is already being terminated.
                 *      Hurry up and return out of the current kernel context
@@ -718,7 +748,10 @@ task_halt(
                return (KERN_FAILURE);
        }
 
+       task->halting = TRUE;
+
        if (task->thread_count > 1) {
+
                /*
                 * Mark all the threads to keep them from starting any more
                 * user-level execution.  The thread_terminate_internal code
@@ -737,15 +770,47 @@ task_halt(
 
                task_release_locked(task);
        }
+       task_unlock(task);
+       return KERN_SUCCESS;
+}
+
+
+/*
+ * task_complete_halt:
+ *
+ *     Complete task halt by waiting for threads to terminate, then clean
+ *     up task resources (VM, port namespace, etc...) and then let the
+ *     current thread go in the (practically empty) task context.
+ */
+void
+task_complete_halt(task_t task)
+{
+       task_lock(task);
+       assert(task->halting);
+       assert(task == current_task());
 
        /*
         *      Give the machine dependent code a chance
-        *      to perform cleanup before ripping apart
-        *      the task.
+        *      to perform cleanup of task-level resources
+        *      associated with the current thread before
+        *      ripping apart the task.
+        *
+        *      This must be done with the task locked.
         */
        machine_thread_terminate_self();
 
-       task_unlock(task);
+       /*
+        *      Wait for the other threads to get shut down.
+        *      When the last other thread is reaped, we'll be
+        *      worken up.
+        */
+       if (task->thread_count > 1) {
+               assert_wait((event_t)&task->halting, THREAD_UNINT);
+               task_unlock(task);
+               thread_block(THREAD_CONTINUE_NULL);
+       } else {
+               task_unlock(task);
+       }
 
        /*
         *      Destroy all synchronizers owned by the task.
@@ -765,7 +830,7 @@ task_halt(
        vm_map_remove(task->map, task->map->min_offset,
                      task->map->max_offset, VM_MAP_NO_FLAGS);
 
-       return (KERN_SUCCESS);
+       task->halting = FALSE;
 }
 
 /*
@@ -922,7 +987,7 @@ task_threads(
        mach_msg_type_number_t  *count)
 {
        mach_msg_type_number_t  actual;
-       thread_t                                *threads;
+       thread_t                                *thread_list;
        thread_t                                thread;
        vm_size_t                               size, size_needed;
        void                                    *addr;
@@ -931,7 +996,7 @@ task_threads(
        if (task == TASK_NULL)
                return (KERN_INVALID_ARGUMENT);
 
-       size = 0; addr = 0;
+       size = 0; addr = NULL;
 
        for (;;) {
                task_lock(task);
@@ -966,14 +1031,14 @@ task_threads(
        }
 
        /* OK, have memory and the task is locked & active */
-       threads = (thread_t *)addr;
+       thread_list = (thread_t *)addr;
 
        i = j = 0;
 
        for (thread = (thread_t)queue_first(&task->threads); i < actual;
                                ++i, thread = (thread_t)queue_next(&thread->task_threads)) {
                thread_reference_internal(thread);
-               threads[j++] = thread;
+               thread_list[j++] = thread;
        }
 
        assert(queue_end(&task->threads, (queue_entry_t)thread));
@@ -987,7 +1052,7 @@ task_threads(
        if (actual == 0) {
                /* no threads, so return null pointer and deallocate memory */
 
-               *threads_out = 0;
+               *threads_out = NULL;
                *count = 0;
 
                if (size != 0)
@@ -1002,23 +1067,23 @@ task_threads(
                        newaddr = kalloc(size_needed);
                        if (newaddr == 0) {
                                for (i = 0; i < actual; ++i)
-                                       thread_deallocate(threads[i]);
+                                       thread_deallocate(thread_list[i]);
                                kfree(addr, size);
                                return (KERN_RESOURCE_SHORTAGE);
                        }
 
                        bcopy(addr, newaddr, size_needed);
                        kfree(addr, size);
-                       threads = (thread_t *)newaddr;
+                       thread_list = (thread_t *)newaddr;
                }
 
-               *threads_out = threads;
+               *threads_out = thread_list;
                *count = actual;
 
                /* do the conversion that Mig should handle */
 
                for (i = 0; i < actual; ++i)
-                       ((ipc_port_t *) threads)[i] = convert_thread_to_port(threads[i]);
+                       ((ipc_port_t *) thread_list)[i] = convert_thread_to_port(thread_list[i]);
        }
 
        return (KERN_SUCCESS);
@@ -1206,40 +1271,61 @@ task_info(
        task_info_t                             task_info_out,
        mach_msg_type_number_t  *task_info_count)
 {
+       kern_return_t error = KERN_SUCCESS;
+
        if (task == TASK_NULL)
                return (KERN_INVALID_ARGUMENT);
 
+       task_lock(task);
+
+       if ((task != current_task()) && (!task->active)) {
+               task_unlock(task);
+               return (KERN_INVALID_ARGUMENT);
+       }
+
        switch (flavor) {
 
        case TASK_BASIC_INFO_32:
+       case TASK_BASIC2_INFO_32:
        {
                task_basic_info_32_t    basic_info;
-               vm_map_t                        map;
+               vm_map_t                                map;
+               clock_sec_t                             secs;
+               clock_usec_t                    usecs;
 
-               if (*task_info_count < TASK_BASIC_INFO_32_COUNT)
-                   return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < TASK_BASIC_INFO_32_COUNT) {
+                   error = KERN_INVALID_ARGUMENT;
+                   break;
+               }
 
                basic_info = (task_basic_info_32_t)task_info_out;
 
                map = (task == kernel_task)? kernel_map: task->map;
-               basic_info->virtual_size  = CAST_DOWN(vm_offset_t,map->size);
-               basic_info->resident_size = pmap_resident_count(map->pmap)
-                                                  * PAGE_SIZE;
+               basic_info->virtual_size = (typeof(basic_info->virtual_size))map->size;
+               if (flavor == TASK_BASIC2_INFO_32) {
+                       /*
+                        * The "BASIC2" flavor gets the maximum resident
+                        * size instead of the current resident size...
+                        */
+                       basic_info->resident_size = pmap_resident_max(map->pmap);
+               } else {
+                       basic_info->resident_size = pmap_resident_count(map->pmap);
+               }
+               basic_info->resident_size *= PAGE_SIZE;
 
-               task_lock(task);
                basic_info->policy = ((task != kernel_task)?
                                                                                  POLICY_TIMESHARE: POLICY_RR);
                basic_info->suspend_count = task->user_stop_count;
 
-               absolutetime_to_microtime(
-                                               task->total_user_time,
-                                                               &basic_info->user_time.seconds,
-                                                               &basic_info->user_time.microseconds);
-               absolutetime_to_microtime(
-                                               task->total_system_time,
-                                                               &basic_info->system_time.seconds,
-                                                               &basic_info->system_time.microseconds);
-               task_unlock(task);
+               absolutetime_to_microtime(task->total_user_time, &secs, &usecs);
+               basic_info->user_time.seconds = 
+                       (typeof(basic_info->user_time.seconds))secs;
+               basic_info->user_time.microseconds = usecs;
+
+               absolutetime_to_microtime(task->total_system_time, &secs, &usecs);
+               basic_info->system_time.seconds = 
+                       (typeof(basic_info->system_time.seconds))secs;
+               basic_info->system_time.microseconds = usecs;
 
                *task_info_count = TASK_BASIC_INFO_32_COUNT;
                break;
@@ -1248,32 +1334,36 @@ task_info(
        case TASK_BASIC_INFO_64:
        {
                task_basic_info_64_t    basic_info;
-               vm_map_t                        map;
+               vm_map_t                                map;
+               clock_sec_t                             secs;
+               clock_usec_t                    usecs;
 
-               if (*task_info_count < TASK_BASIC_INFO_64_COUNT)
-                   return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < TASK_BASIC_INFO_64_COUNT) {
+                   error = KERN_INVALID_ARGUMENT;
+                   break;
+               }
 
                basic_info = (task_basic_info_64_t)task_info_out;
 
                map = (task == kernel_task)? kernel_map: task->map;
                basic_info->virtual_size  = map->size;
-               basic_info->resident_size = (mach_vm_size_t)(pmap_resident_count(map->pmap)
-                                                  * PAGE_SIZE);
+               basic_info->resident_size =
+                       (mach_vm_size_t)(pmap_resident_count(map->pmap))
+                       * PAGE_SIZE_64;
 
-               task_lock(task);
                basic_info->policy = ((task != kernel_task)?
                                                                                  POLICY_TIMESHARE: POLICY_RR);
                basic_info->suspend_count = task->user_stop_count;
 
-               absolutetime_to_microtime(
-                                               task->total_user_time,
-                                                               &basic_info->user_time.seconds,
-                                                               &basic_info->user_time.microseconds);
-               absolutetime_to_microtime(
-                                               task->total_system_time,
-                                                               &basic_info->system_time.seconds,
-                                                               &basic_info->system_time.microseconds);
-               task_unlock(task);
+               absolutetime_to_microtime(task->total_user_time, &secs, &usecs);
+               basic_info->user_time.seconds = 
+                       (typeof(basic_info->user_time.seconds))secs;
+               basic_info->user_time.microseconds = usecs;
+
+               absolutetime_to_microtime(task->total_system_time, &secs, &usecs);
+               basic_info->system_time.seconds =
+                       (typeof(basic_info->system_time.seconds))secs;
+               basic_info->system_time.microseconds = usecs;
 
                *task_info_count = TASK_BASIC_INFO_64_COUNT;
                break;
@@ -1284,8 +1374,10 @@ task_info(
                register task_thread_times_info_t       times_info;
                register thread_t                                       thread;
 
-               if (*task_info_count < TASK_THREAD_TIMES_INFO_COUNT)
-                   return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < TASK_THREAD_TIMES_INFO_COUNT) {
+                   error = KERN_INVALID_ARGUMENT;
+                   break;
+               }
 
                times_info = (task_thread_times_info_t) task_info_out;
                times_info->user_time.seconds = 0;
@@ -1293,7 +1385,6 @@ task_info(
                times_info->system_time.seconds = 0;
                times_info->system_time.microseconds = 0;
 
-               task_lock(task);
 
                queue_iterate(&task->threads, thread, thread_t, task_threads) {
                    time_value_t        user_time, system_time;
@@ -1304,7 +1395,6 @@ task_info(
                    time_value_add(&times_info->system_time, &system_time);
                }
 
-               task_unlock(task);
 
                *task_info_count = TASK_THREAD_TIMES_INFO_COUNT;
                break;
@@ -1315,13 +1405,14 @@ task_info(
                task_absolutetime_info_t        info;
                register thread_t                       thread;
 
-               if (*task_info_count < TASK_ABSOLUTETIME_INFO_COUNT)
-                       return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < TASK_ABSOLUTETIME_INFO_COUNT) {
+                       error = KERN_INVALID_ARGUMENT;
+                       break;
+               }
 
                info = (task_absolutetime_info_t)task_info_out;
                info->threads_user = info->threads_system = 0;
 
-               task_lock(task);
 
                info->total_user = task->total_user_time;
                info->total_system = task->total_system_time;
@@ -1338,20 +1429,36 @@ task_info(
                        info->total_system += tval;
                }
 
-               task_unlock(task);
 
                *task_info_count = TASK_ABSOLUTETIME_INFO_COUNT;
                break;
        }
 
+       case TASK_DYLD_INFO:
+       {
+               task_dyld_info_t info;
+
+               if (*task_info_count < TASK_DYLD_INFO_COUNT) {
+                       error = KERN_INVALID_ARGUMENT;
+                       break;
+               }
+               info = (task_dyld_info_t)task_info_out;
+               info->all_image_info_addr = task->all_image_info_addr;
+               info->all_image_info_size = task->all_image_info_size;
+               *task_info_count = TASK_DYLD_INFO_COUNT;
+               break;
+       }
+
        /* OBSOLETE */
        case TASK_SCHED_FIFO_INFO:
        {
 
-               if (*task_info_count < POLICY_FIFO_BASE_COUNT)
-                       return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < POLICY_FIFO_BASE_COUNT) {
+                       error = KERN_INVALID_ARGUMENT;
+                       break;
+               }
 
-               return (KERN_INVALID_POLICY);
+               error = KERN_INVALID_POLICY;
        }
 
        /* OBSOLETE */
@@ -1359,19 +1466,19 @@ task_info(
        {
                register policy_rr_base_t       rr_base;
 
-               if (*task_info_count < POLICY_RR_BASE_COUNT)
-                       return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < POLICY_RR_BASE_COUNT) {
+                       error = KERN_INVALID_ARGUMENT;
+                       break;
+               }
 
                rr_base = (policy_rr_base_t) task_info_out;
 
-               task_lock(task);
                if (task != kernel_task) {
-                       task_unlock(task);
-                       return (KERN_INVALID_POLICY);
+                       error = KERN_INVALID_POLICY;
+                       break;
                }
 
                rr_base->base_priority = task->priority;
-               task_unlock(task);
 
                rr_base->quantum = std_quantum_us / 1000;
 
@@ -1384,19 +1491,19 @@ task_info(
        {
                register policy_timeshare_base_t        ts_base;
 
-               if (*task_info_count < POLICY_TIMESHARE_BASE_COUNT)
-                       return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < POLICY_TIMESHARE_BASE_COUNT) {
+                       error = KERN_INVALID_ARGUMENT;
+                       break;
+               }
 
                ts_base = (policy_timeshare_base_t) task_info_out;
 
-               task_lock(task);
                if (task == kernel_task) {
-                       task_unlock(task);
-                       return (KERN_INVALID_POLICY);
+                       error = KERN_INVALID_POLICY;
+                       break;
                }
 
                ts_base->base_priority = task->priority;
-               task_unlock(task);
 
                *task_info_count = POLICY_TIMESHARE_BASE_COUNT;
                break;
@@ -1406,14 +1513,14 @@ task_info(
        {
                register security_token_t       *sec_token_p;
 
-               if (*task_info_count < TASK_SECURITY_TOKEN_COUNT)
-                   return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < TASK_SECURITY_TOKEN_COUNT) {
+                   error = KERN_INVALID_ARGUMENT;
+                   break;
+               }
 
                sec_token_p = (security_token_t *) task_info_out;
 
-               task_lock(task);
                *sec_token_p = task->sec_token;
-               task_unlock(task);
 
                *task_info_count = TASK_SECURITY_TOKEN_COUNT;
                break;
@@ -1423,32 +1530,35 @@ task_info(
        {
                register audit_token_t  *audit_token_p;
 
-               if (*task_info_count < TASK_AUDIT_TOKEN_COUNT)
-                   return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < TASK_AUDIT_TOKEN_COUNT) {
+                   error = KERN_INVALID_ARGUMENT;
+                   break;
+               }
 
                audit_token_p = (audit_token_t *) task_info_out;
 
-               task_lock(task);
                *audit_token_p = task->audit_token;
-               task_unlock(task);
 
                *task_info_count = TASK_AUDIT_TOKEN_COUNT;
                break;
        }
             
        case TASK_SCHED_INFO:
-               return (KERN_INVALID_ARGUMENT);
+               error = KERN_INVALID_ARGUMENT;
 
        case TASK_EVENTS_INFO:
        {
                register task_events_info_t     events_info;
+               register thread_t                       thread;
 
-               if (*task_info_count < TASK_EVENTS_INFO_COUNT)
-                   return (KERN_INVALID_ARGUMENT);
+               if (*task_info_count < TASK_EVENTS_INFO_COUNT) {
+                  error = KERN_INVALID_ARGUMENT;
+                  break;
+               }
 
                events_info = (task_events_info_t) task_info_out;
 
-               task_lock(task);
+
                events_info->faults = task->faults;
                events_info->pageins = task->pageins;
                events_info->cow_faults = task->cow_faults;
@@ -1456,18 +1566,133 @@ task_info(
                events_info->messages_received = task->messages_received;
                events_info->syscalls_mach = task->syscalls_mach;
                events_info->syscalls_unix = task->syscalls_unix;
-               events_info->csw = task->csw;
-               task_unlock(task);
+
+               events_info->csw = task->c_switch;
+
+               queue_iterate(&task->threads, thread, thread_t, task_threads) {
+                       events_info->csw += thread->c_switch;
+               }
+
 
                *task_info_count = TASK_EVENTS_INFO_COUNT;
                break;
        }
+       case TASK_AFFINITY_TAG_INFO:
+       {
+               if (*task_info_count < TASK_AFFINITY_TAG_INFO_COUNT) {
+                   error = KERN_INVALID_ARGUMENT;
+                   break;
+               }
+
+               error = task_affinity_info(task, task_info_out, task_info_count);
+       }
 
        default:
-               return (KERN_INVALID_ARGUMENT);
+               error = KERN_INVALID_ARGUMENT;
+       }
+
+       task_unlock(task);
+       return (error);
+}
+
+void
+task_vtimer_set(
+       task_t          task,
+       integer_t       which)
+{
+       thread_t        thread;
+
+       /* assert(task == current_task()); */ /* bogus assert 4803227 4807483 */
+
+       task_lock(task);
+
+       task->vtimers |= which;
+
+       switch (which) {
+
+       case TASK_VTIMER_USER:
+               queue_iterate(&task->threads, thread, thread_t, task_threads) {
+                       thread->vtimer_user_save = timer_grab(&thread->user_timer);
+               }
+               break;
+
+       case TASK_VTIMER_PROF:
+               queue_iterate(&task->threads, thread, thread_t, task_threads) {
+                       thread->vtimer_prof_save = timer_grab(&thread->user_timer);
+                       thread->vtimer_prof_save += timer_grab(&thread->system_timer);
+               }
+               break;
+
+       case TASK_VTIMER_RLIM:
+               queue_iterate(&task->threads, thread, thread_t, task_threads) {
+                       thread->vtimer_rlim_save = timer_grab(&thread->user_timer);
+                       thread->vtimer_rlim_save += timer_grab(&thread->system_timer);
+               }
+               break;
+       }
+
+       task_unlock(task);
+}
+
+void
+task_vtimer_clear(
+       task_t          task,
+       integer_t       which)
+{
+       assert(task == current_task());
+
+       task_lock(task);
+
+       task->vtimers &= ~which;
+
+       task_unlock(task);
+}
+
+void
+task_vtimer_update(
+__unused
+       task_t          task,
+       integer_t       which,
+       uint32_t        *microsecs)
+{
+       thread_t        thread = current_thread();
+       uint32_t        tdelt;
+       clock_sec_t     secs;
+       uint64_t        tsum;
+
+       assert(task == current_task());
+
+       assert(task->vtimers & which);
+
+       secs = tdelt = 0;
+
+       switch (which) {
+
+       case TASK_VTIMER_USER:
+               tdelt = (uint32_t)timer_delta(&thread->user_timer,
+                                                               &thread->vtimer_user_save);
+               absolutetime_to_microtime(tdelt, &secs, microsecs);
+               break;
+
+       case TASK_VTIMER_PROF:
+               tsum = timer_grab(&thread->user_timer);
+               tsum += timer_grab(&thread->system_timer);
+               tdelt = (uint32_t)(tsum - thread->vtimer_prof_save);
+               absolutetime_to_microtime(tdelt, &secs, microsecs);
+               /* if the time delta is smaller than a usec, ignore */
+               if (*microsecs != 0)
+                       thread->vtimer_prof_save = tsum;
+               break;
+
+       case TASK_VTIMER_RLIM:
+               tsum = timer_grab(&thread->user_timer);
+               tsum += timer_grab(&thread->system_timer);
+               tdelt = (uint32_t)(tsum - thread->vtimer_rlim_save);
+               thread->vtimer_rlim_save = tsum;
+               absolutetime_to_microtime(tdelt, &secs, microsecs);
+               break;
        }
 
-       return (KERN_SUCCESS);
 }
 
 /*
@@ -1494,7 +1719,7 @@ task_assign_default(
        task_t          task,
        boolean_t       assign_threads)
 {
-    return (task_assign(task, &default_pset, assign_threads));
+    return (task_assign(task, &pset0, assign_threads));
 }
 
 /*
@@ -1510,9 +1735,9 @@ task_get_assignment(
        if (!task->active)
                return(KERN_FAILURE);
 
-       *pset = task->processor_set;
-       pset_reference(*pset);
-       return(KERN_SUCCESS);
+       *pset = &pset0;
+
+       return (KERN_SUCCESS);
 }
 
 
@@ -1612,6 +1837,70 @@ task_synchronizer_destroy_all(task_t task)
        }
 }
 
+/*
+ * Install default (machine-dependent) initial thread state 
+ * on the task.  Subsequent thread creation will have this initial
+ * state set on the thread by machine_thread_inherit_taskwide().
+ * Flavors and structures are exactly the same as those to thread_set_state()
+ */
+kern_return_t 
+task_set_state(
+       task_t task, 
+       int flavor, 
+       thread_state_t state, 
+       mach_msg_type_number_t state_count)
+{
+       kern_return_t ret;
+
+       if (task == TASK_NULL) {
+               return (KERN_INVALID_ARGUMENT);
+       }
+
+       task_lock(task);
+
+       if (!task->active) {
+               task_unlock(task);
+               return (KERN_FAILURE);
+       }
+
+       ret = machine_task_set_state(task, flavor, state, state_count);
+
+       task_unlock(task);
+       return ret;
+}
+
+/*
+ * Examine the default (machine-dependent) initial thread state 
+ * on the task, as set by task_set_state().  Flavors and structures
+ * are exactly the same as those passed to thread_get_state().
+ */
+kern_return_t 
+task_get_state(
+       task_t  task, 
+       int     flavor,
+       thread_state_t state,
+       mach_msg_type_number_t *state_count)
+{
+       kern_return_t ret;
+
+       if (task == TASK_NULL) {
+               return (KERN_INVALID_ARGUMENT);
+       }
+
+       task_lock(task);
+
+       if (!task->active) {
+               task_unlock(task);
+               return (KERN_FAILURE);
+       }
+
+       ret = machine_task_get_state(task, flavor, state, state_count);
+
+       task_unlock(task);
+       return ret;
+}
+
+
 /*
  * We need to export some functions to other components that
  * are currently implemented in macros within the osfmk
@@ -1625,6 +1914,16 @@ boolean_t is_kerneltask(task_t t)
        return (FALSE);
 }
 
+int
+check_for_tasksuspend(task_t task)
+{
+
+       if (task == TASK_NULL)
+               return (0);
+
+       return (task->suspend_count > 0);
+}
+
 #undef current_task
 task_t current_task(void);
 task_t current_task(void)
@@ -1641,3 +1940,64 @@ task_reference(
        if (task != TASK_NULL)
                task_reference_internal(task);
 }
+
+#if CONFIG_MACF_MACH
+/*
+ * Protect 2 task labels against modification by adding a reference on
+ * both label handles. The locks do not actually have to be held while
+ * using the labels as only labels with one reference can be modified
+ * in place.
+ */
+
+void
+tasklabel_lock2(
+       task_t a,
+       task_t b)
+{
+       labelh_reference(a->label);
+       labelh_reference(b->label);
+}
+
+void
+tasklabel_unlock2(
+       task_t a,
+       task_t b)
+{
+       labelh_release(a->label);
+       labelh_release(b->label);
+}
+
+void
+mac_task_label_update_internal(
+       struct label    *pl,
+       struct task     *task)
+{
+
+       tasklabel_lock(task);
+       task->label = labelh_modify(task->label);
+       mac_task_label_update(pl, &task->maclabel);
+       tasklabel_unlock(task);
+       ip_lock(task->itk_self);
+       mac_port_label_update_cred(pl, &task->itk_self->ip_label);
+       ip_unlock(task->itk_self);
+}
+
+void
+mac_task_label_modify(
+       struct task     *task,
+       void            *arg,
+       void (*f)       (struct label *l, void *arg))
+{
+
+       tasklabel_lock(task);
+       task->label = labelh_modify(task->label);
+       (*f)(&task->maclabel, arg);
+       tasklabel_unlock(task);
+}
+
+struct label *
+mac_task_get_label(struct task *task)
+{
+       return (&task->maclabel);
+}
+#endif