X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/1c79356b52d46aa6b508fb032f5ae709b1f2897b..c18c124eaa464aaaa5549e99e5a70fc9cbb50944:/osfmk/kern/task.c?ds=sidebyside diff --git a/osfmk/kern/task.c b/osfmk/kern/task.c index 0eaf47eaa..999dfefcf 100644 --- a/osfmk/kern/task.c +++ b/osfmk/kern/task.c @@ -1,23 +1,29 @@ /* - * Copyright (c) 2000 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2010 Apple Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. + * This file contains Original Code and/or Modifications of Original Code + * as defined in and that are subject to the Apple Public Source License + * Version 2.0 (the 'License'). You may not use this file except in + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * unlawful or unlicensed copies of an Apple operating system, or to + * circumvent, violate, or enable the circumvention or violation of, any + * terms of an Apple operating system software license agreement. * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * Please obtain a copy of the License at + * http://www.opensource.apple.com/apsl/ and read it before using this file. + * + * The Original Code and all software distributed under the License are + * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. + * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. + * Please see the License for the specific language governing rights and + * limitations under the License. * - * @APPLE_LICENSE_HEADER_END@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* * @OSF_FREE_COPYRIGHT@ @@ -72,49 +78,59 @@ * 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 -#include -#include -#include -#include -#include - +#include #include +#include #include #include #include #include #include -#include -#include + +#include +#include #include #include +#include + +#include #include #include #include #include +#include #include #include #include #include /* for thread_wakeup */ -#include -#include /*** ??? fix so this can be removed ***/ #include -#include #include -#include /* for kernel_map, ipc_kernel_map */ -#include +#include +#include #include #include -#if MACH_KDB -#include -#endif /* MACH_KDB */ +#include +#include +#if CONFIG_TELEMETRY +#include +#endif -#if TASK_SWAPPER -#include -#endif /* TASK_SWAPPER */ +#include +#include +#include /* for kernel_map, ipc_kernel_map */ +#include +#include +#include +#include /* * Exported interfaces */ @@ -122,214 +138,588 @@ #include #include #include +#include + +#include + +#if CONFIG_COUNTERS +#include +#endif /* CONFIG_COUNTERS */ + +#include +#include + +#if CONFIG_ATM +#include +#endif + +#include + +#if KPERF +extern int kpc_force_all_ctrs(task_t, int); +#endif + +uint32_t qos_override_mode; + +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; + +/* Flag set by core audio when audio is playing. Used to stifle EXC_RESOURCE generation when active. */ +int audio_active = 0; + +zinfo_usage_store_t tasks_tkm_private; +zinfo_usage_store_t tasks_tkm_shared; + +/* A container to accumulate statistics for expired tasks */ +expired_task_statistics_t dead_task_statistics; +lck_spin_t dead_task_statistics_lock; + +ledger_template_t task_ledger_template = NULL; + +struct _task_ledger_indices task_ledgers __attribute__((used)) = + {-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, -1, + { 0 /* initialized at runtime */}, +#ifdef CONFIG_BANK + -1, -1, +#endif + }; + +void init_task_ledgers(void); +void task_footprint_exceeded(int warning, __unused const void *param0, __unused const void *param1); +void task_wakeups_rate_exceeded(int warning, __unused const void *param0, __unused const void *param1); +void __attribute__((noinline)) THIS_PROCESS_IS_CAUSING_TOO_MANY_WAKEUPS__SENDING_EXC_RESOURCE(void); +void __attribute__((noinline)) THIS_PROCESS_CROSSED_HIGH_WATERMARK__SENDING_EXC_RESOURCE(int max_footprint_mb); +int coredump(void *core_proc, int reserve_mb, int ignore_ulimit); + +kern_return_t task_suspend_internal(task_t); +kern_return_t task_resume_internal(task_t); + +void proc_init_cpumon_params(void); + +// Warn tasks when they hit 80% of their memory limit. +#define PHYS_FOOTPRINT_WARNING_LEVEL 80 + +#define TASK_WAKEUPS_MONITOR_DEFAULT_LIMIT 150 /* wakeups per second */ +#define TASK_WAKEUPS_MONITOR_DEFAULT_INTERVAL 300 /* in seconds. */ -task_t kernel_task; -zone_t task_zone; +/* + * Level (in terms of percentage of the limit) at which the wakeups monitor triggers telemetry. + * + * (ie when the task's wakeups rate exceeds 70% of the limit, start taking user + * stacktraces, aka micro-stackshots) + */ +#define TASK_WAKEUPS_MONITOR_DEFAULT_USTACKSHOTS_TRIGGER 70 + +int task_wakeups_monitor_interval; /* In seconds. Time period over which wakeups rate is observed */ +int task_wakeups_monitor_rate; /* In hz. Maximum allowable wakeups per task before EXC_RESOURCE is sent */ + +int task_wakeups_monitor_ustackshots_trigger_pct; /* Percentage. Level at which we start gathering telemetry. */ + +int disable_exc_resource; /* Global override to supress EXC_RESOURCE for resource monitor violations. */ + +int max_task_footprint = 0; /* Per-task limit on physical memory consumption */ +#if MACH_ASSERT +int pmap_ledgers_panic = 1; +#endif /* MACH_ASSERT */ + +int task_max = CONFIG_TASK_MAX; /* Max number of tasks */ + +int hwm_user_cores = 0; /* high watermark violations generate user core files */ + +#ifdef MACH_BSD +extern void proc_getexecutableuuid(void *, unsigned char *, unsigned long); +extern int proc_pid(struct proc *p); +extern int proc_selfpid(void); +extern char *proc_name_address(struct proc *p); +#if CONFIG_JETSAM +extern void memorystatus_on_ledger_footprint_exceeded(int warning, const int max_footprint_mb); +#endif +#endif +#if MACH_ASSERT +extern int pmap_ledgers_panic; +#endif /* MACH_ASSERT */ /* Forwards */ void task_hold_locked( task_t task); void task_wait_locked( - task_t task); + task_t task, + boolean_t until_not_runnable); void task_release_locked( task_t task); -void task_collect_scan(void); void task_free( task_t task ); void task_synchronizer_destroy_all( task_t task); -void task_subsystem_destroy_all( - task_t task); -kern_return_t task_set_ledger( - task_t task, - ledger_t wired, - ledger_t paged); +int check_for_tasksuspend( + task_t task); void -task_init(void) +task_backing_store_privileged( + task_t task) { - task_zone = zinit( - sizeof(struct task), - TASK_MAX * sizeof(struct task), - TASK_CHUNK * sizeof(struct task), - "tasks"); + task_lock(task); + task->priv_flags |= VM_BACKING_STORE_PRIV; + task_unlock(task); + return; +} - eml_init(); - /* - * Create the kernel task as the first task. - * Task_create_local must assign to kernel_task as a side effect, - * for other initialization. (:-() +void +task_set_64bit( + task_t task, + boolean_t is64bit) +{ +#if defined(__i386__) || defined(__x86_64__) || defined(__arm64__) + thread_t thread; +#endif /* defined(__i386__) || defined(__x86_64__) || defined(__arm64__) */ + + task_lock(task); + + if (is64bit) { + if (task_has_64BitAddr(task)) + goto out; + task_set_64BitAddr(task); + } else { + if ( !task_has_64BitAddr(task)) + goto out; + task_clear_64BitAddr(task); + } + /* 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 (task_create_local( - TASK_NULL, FALSE, FALSE, &kernel_task) != KERN_SUCCESS) - panic("task_init\n"); - vm_map_deallocate(kernel_task->map); - kernel_task->map = kernel_map; -#if MACH_ASSERT - if (watchacts & WA_TASK) - printf("task_init: kernel_task = %x map=%x\n", - kernel_task, kernel_map); -#endif /* MACH_ASSERT */ +#if defined(__i386__) || defined(__x86_64__) || defined(__arm64__) + queue_iterate(&task->threads, thread, thread_t, task_threads) { + thread_mtx_lock(thread); + machine_thread_switch_addrmode(thread); + thread_mtx_unlock(thread); + } +#endif /* defined(__i386__) || defined(__x86_64__) || defined(__arm64__) */ + +out: + task_unlock(task); } -#if MACH_HOST + void -task_freeze( - task_t task) +task_set_dyld_info(task_t task, mach_vm_address_t addr, mach_vm_size_t size) { task_lock(task); - /* - * If may_assign is false, task is already being assigned, - * wait for that to finish. - */ - while (task->may_assign == FALSE) { - task->assign_active = TRUE; - thread_sleep_mutex((event_t) &task->assign_active, - &task->lock, THREAD_INTERRUPTIBLE); - task_lock(task); - } - task->may_assign = FALSE; + task->all_image_info_addr = addr; + task->all_image_info_size = size; task_unlock(task); +} - return; +void +task_atm_reset(__unused task_t task) { + +#if CONFIG_ATM + if (task->atm_context != NULL) { + atm_task_descriptor_destroy(task->atm_context); + task->atm_context = NULL; + } +#endif + +} + +#if TASK_REFERENCE_LEAK_DEBUG +#include + +decl_simple_lock_data(static,task_ref_lock); +static btlog_t *task_ref_btlog; +#define TASK_REF_OP_INCR 0x1 +#define TASK_REF_OP_DECR 0x2 + +#define TASK_REF_BTDEPTH 7 + +static void +task_ref_lock_lock(void *context) +{ + simple_lock((simple_lock_t)context); +} +static void +task_ref_lock_unlock(void *context) +{ + simple_unlock((simple_lock_t)context); } void -task_unfreeze( - task_t task) +task_reference_internal(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); + void * bt[TASK_REF_BTDEPTH]; + int numsaved = 0; - return; + numsaved = OSBacktrace(bt, TASK_REF_BTDEPTH); + + (void)hw_atomic_add(&(task)->ref_count, 1); + btlog_add_entry(task_ref_btlog, task, TASK_REF_OP_INCR, + bt, numsaved); } -#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. - */ -kern_return_t -kernel_task_create( - task_t parent_task, - vm_offset_t map_base, - vm_size_t map_size, - task_t *child_task) +uint32_t +task_deallocate_internal(task_t task) { - kern_return_t result; - task_t new_task; - vm_map_t old_map; + void * bt[TASK_REF_BTDEPTH]; + int numsaved = 0; - /* - * Create the task. - */ - result = task_create_local(parent_task, FALSE, TRUE, &new_task); - if (result != KERN_SUCCESS) - return (result); + numsaved = OSBacktrace(bt, TASK_REF_BTDEPTH); + + btlog_add_entry(task_ref_btlog, task, TASK_REF_OP_DECR, + bt, numsaved); + return hw_atomic_sub(&(task)->ref_count, 1); +} + +#endif /* TASK_REFERENCE_LEAK_DEBUG */ + +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_CHUNK * sizeof(struct task), + "tasks"); + + zone_change(task_zone, Z_NOENCRYPT, TRUE); /* - * Task_create_local creates the task with a user-space map. - * We attempt to replace the map and free it afterwards; else - * task_deallocate will free it (can NOT set map to null before - * task_deallocate, this impersonates a norma placeholder task). - * _Mark the memory as pageable_ -- this is what we - * want for images (like servers) loaded into the kernel. + * Configure per-task memory limit. + * The boot-arg is interpreted as Megabytes, + * and takes precedence over the device tree. + * Setting the boot-arg to 0 disables task limits. */ - if (map_size == 0) { - vm_map_deallocate(new_task->map); - new_task->map = kernel_map; - *child_task = new_task; - } else { - old_map = new_task->map; - if ((result = kmem_suballoc(kernel_map, &map_base, - map_size, TRUE, FALSE, - &new_task->map)) != KERN_SUCCESS) { + if (!PE_parse_boot_argn("max_task_pmem", &max_task_footprint, + sizeof (max_task_footprint))) { + /* + * No limit was found in boot-args, so go look in the device tree. + */ + if (!PE_get_default("kern.max_task_pmem", &max_task_footprint, + sizeof(max_task_footprint))) { /* - * New task created with ref count of 2 -- decrement by - * one to force task deletion. + * No limit was found in device tree. */ - printf("kmem_suballoc(%x,%x,%x,1,0,&new) Fails\n", - kernel_map, map_base, map_size); - --new_task->ref_count; - task_deallocate(new_task); - return (result); + max_task_footprint = 0; } - vm_map_deallocate(old_map); - *child_task = new_task; } - return (KERN_SUCCESS); + + if (max_task_footprint != 0) { +#if CONFIG_JETSAM + if (max_task_footprint < 50) { + printf("Warning: max_task_pmem %d below minimum.\n", + max_task_footprint); + max_task_footprint = 50; + } + printf("Limiting task physical memory footprint to %d MB\n", + max_task_footprint); + max_task_footprint *= 1024 * 1024; // Convert MB to bytes +#else + printf("Warning: max_task_footprint specified, but jetsam not configured; ignoring.\n"); +#endif + } + +#if MACH_ASSERT + PE_parse_boot_argn("pmap_ledgers_panic", &pmap_ledgers_panic, + sizeof (pmap_ledgers_panic)); +#endif /* MACH_ASSERT */ + + if (!PE_parse_boot_argn("hwm_user_cores", &hwm_user_cores, + sizeof (hwm_user_cores))) { + hwm_user_cores = 0; + } + + if (PE_parse_boot_argn("qos_override_mode", &qos_override_mode, sizeof(qos_override_mode))) { + printf("QOS override mode: 0x%08x\n", qos_override_mode); + } else { + qos_override_mode = QOS_OVERRIDE_MODE_FINE_GRAINED_OVERRIDE_BUT_SINGLE_MUTEX_OVERRIDE; + } + + proc_init_cpumon_params(); + + if (!PE_parse_boot_argn("task_wakeups_monitor_rate", &task_wakeups_monitor_rate, sizeof (task_wakeups_monitor_rate))) { + task_wakeups_monitor_rate = TASK_WAKEUPS_MONITOR_DEFAULT_LIMIT; + } + + if (!PE_parse_boot_argn("task_wakeups_monitor_interval", &task_wakeups_monitor_interval, sizeof (task_wakeups_monitor_interval))) { + task_wakeups_monitor_interval = TASK_WAKEUPS_MONITOR_DEFAULT_INTERVAL; + } + + if (!PE_parse_boot_argn("task_wakeups_monitor_ustackshots_trigger_pct", &task_wakeups_monitor_ustackshots_trigger_pct, + sizeof (task_wakeups_monitor_ustackshots_trigger_pct))) { + task_wakeups_monitor_ustackshots_trigger_pct = TASK_WAKEUPS_MONITOR_DEFAULT_USTACKSHOTS_TRIGGER; + } + + if (!PE_parse_boot_argn("disable_exc_resource", &disable_exc_resource, + sizeof (disable_exc_resource))) { + disable_exc_resource = 0; + } + +/* + * If we have coalitions, coalition_init() will call init_task_ledgers() as it + * sets up the ledgers for the default coalition. If we don't have coalitions, + * then we have to call it now. + */ +#if CONFIG_COALITIONS + assert(task_ledger_template); +#else /* CONFIG_COALITIONS */ + init_task_ledgers(); +#endif /* CONFIG_COALITIONS */ + +#if TASK_REFERENCE_LEAK_DEBUG + simple_lock_init(&task_ref_lock, 0); + task_ref_btlog = btlog_create(100000, + TASK_REF_BTDEPTH, + task_ref_lock_lock, + task_ref_lock_unlock, + &task_ref_lock); + assert(task_ref_btlog); +#endif + + /* + * Create the kernel task as the first task. + */ +#ifdef __LP64__ + if (task_create_internal(TASK_NULL, COALITION_NULL, FALSE, TRUE, &kernel_task) != KERN_SUCCESS) +#else + if (task_create_internal(TASK_NULL, COALITION_NULL, FALSE, FALSE, &kernel_task) != KERN_SUCCESS) +#endif + panic("task_init\n"); + + vm_map_deallocate(kernel_task->map); + kernel_task->map = kernel_map; + lck_spin_init(&dead_task_statistics_lock, &task_lck_grp, &task_lck_attr); + +} + +/* + * Create a task running in the kernel address space. It may + * have its own map of size mem_size and may have ipc privileges. + */ +kern_return_t +kernel_task_create( + __unused task_t parent_task, + __unused vm_offset_t map_base, + __unused vm_size_t map_size, + __unused task_t *child_task) +{ + return (KERN_INVALID_ARGUMENT); } kern_return_t task_create( - task_t parent_task, - ledger_port_array_t ledger_ports, - mach_msg_type_number_t num_ledger_ports, - boolean_t inherit_memory, - task_t *child_task) /* OUT */ + task_t parent_task, + __unused ledger_port_array_t ledger_ports, + __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_local( - parent_task, inherit_memory, FALSE, 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, - host_priv_t host_priv, - ledger_port_array_t ledger_ports, - mach_msg_type_number_t num_ledger_ports, - boolean_t inherit_memory, - task_t *child_task) /* OUT */ -{ - kern_return_t result; - + host_security_t host_security, + 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, + __unused boolean_t inherit_memory, + __unused task_t *child_task) /* OUT */ +{ if (parent_task == TASK_NULL) return(KERN_INVALID_ARGUMENT); if (host_security == HOST_NULL) return(KERN_INVALID_SECURITY); - result = task_create_local( - parent_task, inherit_memory, FALSE, child_task); + /* + * No longer supported. + */ + return(KERN_FAILURE); +} + +/* + * Task ledgers + * ------------ + * + * phys_footprint + * Physical footprint: This is the sum of: + * + internal + * + internal_compressed + * + iokit_mapped + * - alternate_accounting + * + * internal + * The task's anonymous memory, which on iOS is always resident. + * + * internal_compressed + * Amount of this task's internal memory which is held by the compressor. + * Such memory is no longer actually resident for the task [i.e., resident in its pmap], + * and could be either decompressed back into memory, or paged out to storage, depending + * on our implementation. + * + * iokit_mapped + * IOKit mappings: The total size of all IOKit mappings in this task, regardless of + clean/dirty or internal/external state]. + * + * alternate_accounting + * The number of internal dirty pages which are part of IOKit mappings. By definition, these pages + * are counted in both internal *and* iokit_mapped, so we must subtract them from the total to avoid + * double counting. + */ +void +init_task_ledgers(void) +{ + ledger_template_t t; + + assert(task_ledger_template == NULL); + assert(kernel_task == TASK_NULL); + + if ((t = ledger_template_create("Per-task ledger")) == NULL) + panic("couldn't create task ledger template"); + + task_ledgers.cpu_time = ledger_entry_add(t, "cpu_time", "sched", "ns"); + task_ledgers.tkm_private = ledger_entry_add(t, "tkm_private", + "physmem", "bytes"); + task_ledgers.tkm_shared = ledger_entry_add(t, "tkm_shared", "physmem", + "bytes"); + task_ledgers.phys_mem = ledger_entry_add(t, "phys_mem", "physmem", + "bytes"); + task_ledgers.wired_mem = ledger_entry_add(t, "wired_mem", "physmem", + "bytes"); + task_ledgers.internal = ledger_entry_add(t, "internal", "physmem", + "bytes"); + task_ledgers.iokit_mapped = ledger_entry_add(t, "iokit_mapped", "mappings", + "bytes"); + task_ledgers.alternate_accounting = ledger_entry_add(t, "alternate_accounting", "physmem", + "bytes"); + task_ledgers.phys_footprint = ledger_entry_add(t, "phys_footprint", "physmem", + "bytes"); + task_ledgers.internal_compressed = ledger_entry_add(t, "internal_compressed", "physmem", + "bytes"); + task_ledgers.purgeable_volatile = ledger_entry_add(t, "purgeable_volatile", "physmem", "bytes"); + task_ledgers.purgeable_nonvolatile = ledger_entry_add(t, "purgeable_nonvolatile", "physmem", "bytes"); + task_ledgers.purgeable_volatile_compressed = ledger_entry_add(t, "purgeable_volatile_compress", "physmem", "bytes"); + task_ledgers.purgeable_nonvolatile_compressed = ledger_entry_add(t, "purgeable_nonvolatile_compress", "physmem", "bytes"); + task_ledgers.platform_idle_wakeups = ledger_entry_add(t, "platform_idle_wakeups", "power", + "count"); + task_ledgers.interrupt_wakeups = ledger_entry_add(t, "interrupt_wakeups", "power", + "count"); + + sfi_class_id_t class_id, ledger_alias; + for (class_id = SFI_CLASS_UNSPECIFIED; class_id < MAX_SFI_CLASS_ID; class_id++) { + task_ledgers.sfi_wait_times[class_id] = -1; + } + + /* don't account for UNSPECIFIED */ + for (class_id = SFI_CLASS_UNSPECIFIED + 1; class_id < MAX_SFI_CLASS_ID; class_id++) { + ledger_alias = sfi_get_ledger_alias_for_class(class_id); + if (ledger_alias != SFI_CLASS_UNSPECIFIED) { + /* Check to see if alias has been registered yet */ + if (task_ledgers.sfi_wait_times[ledger_alias] != -1) { + task_ledgers.sfi_wait_times[class_id] = task_ledgers.sfi_wait_times[ledger_alias]; + } else { + /* Otherwise, initialize it first */ + task_ledgers.sfi_wait_times[class_id] = task_ledgers.sfi_wait_times[ledger_alias] = sfi_ledger_entry_add(t, ledger_alias); + } + } else { + task_ledgers.sfi_wait_times[class_id] = sfi_ledger_entry_add(t, class_id); + } + + if (task_ledgers.sfi_wait_times[class_id] < 0) { + panic("couldn't create entries for task ledger template for SFI class 0x%x", class_id); + } + } + +#ifdef CONFIG_BANK + task_ledgers.cpu_time_billed_to_me = ledger_entry_add(t, "cpu_time_billed_to_me", "sched", "ns"); + task_ledgers.cpu_time_billed_to_others = ledger_entry_add(t, "cpu_time_billed_to_others", "sched", "ns"); +#endif - if (result != KERN_SUCCESS) - return(result); + assert(task_ledgers.sfi_wait_times[MAX_SFI_CLASS_ID -1] != -1); + + if ((task_ledgers.cpu_time < 0) || + (task_ledgers.tkm_private < 0) || + (task_ledgers.tkm_shared < 0) || + (task_ledgers.phys_mem < 0) || + (task_ledgers.wired_mem < 0) || + (task_ledgers.internal < 0) || + (task_ledgers.iokit_mapped < 0) || + (task_ledgers.alternate_accounting < 0) || + (task_ledgers.phys_footprint < 0) || + (task_ledgers.internal_compressed < 0) || + (task_ledgers.purgeable_volatile < 0) || + (task_ledgers.purgeable_nonvolatile < 0) || + (task_ledgers.purgeable_volatile_compressed < 0) || + (task_ledgers.purgeable_nonvolatile_compressed < 0) || + (task_ledgers.platform_idle_wakeups < 0) || + (task_ledgers.interrupt_wakeups < 0) +#ifdef CONFIG_BANK + || (task_ledgers.cpu_time_billed_to_me < 0) || (task_ledgers.cpu_time_billed_to_others < 0) +#endif + ) { + panic("couldn't create entries for task ledger template"); + } - result = host_security_set_task_token(host_security, - *child_task, - sec_token, - host_priv); + ledger_track_maximum(t, task_ledgers.phys_footprint, 60); +#if MACH_ASSERT + if (pmap_ledgers_panic) { + ledger_panic_on_negative(t, task_ledgers.phys_footprint); + ledger_panic_on_negative(t, task_ledgers.internal); + ledger_panic_on_negative(t, task_ledgers.internal_compressed); + ledger_panic_on_negative(t, task_ledgers.iokit_mapped); + ledger_panic_on_negative(t, task_ledgers.alternate_accounting); + ledger_panic_on_negative(t, task_ledgers.purgeable_volatile); + ledger_panic_on_negative(t, task_ledgers.purgeable_nonvolatile); + ledger_panic_on_negative(t, task_ledgers.purgeable_volatile_compressed); + ledger_panic_on_negative(t, task_ledgers.purgeable_nonvolatile_compressed); + } +#endif /* MACH_ASSERT */ - if (result != KERN_SUCCESS) - return(result); +#if CONFIG_JETSAM + ledger_set_callback(t, task_ledgers.phys_footprint, task_footprint_exceeded, NULL, NULL); +#endif - return(result); + ledger_set_callback(t, task_ledgers.interrupt_wakeups, + task_wakeups_rate_exceeded, NULL, NULL); + + task_ledger_template = t; } kern_return_t -task_create_local( +task_create_internal( task_t parent_task, + coalition_t parent_coalition __unused, boolean_t inherit_memory, - boolean_t kernel_loaded, + 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; + ledger_t ledger = NULL; new_task = (task_t) zalloc(task_zone); @@ -339,264 +729,441 @@ task_create_local( /* one ref for just being alive; one for our caller */ new_task->ref_count = 2; + /* allocate with active entries */ + assert(task_ledger_template != NULL); + if ((ledger = ledger_instantiate(task_ledger_template, + LEDGER_CREATE_ACTIVE_ENTRIES)) == NULL) { + zfree(task_zone, new_task); + return(KERN_RESOURCE_SHORTAGE); + } + + new_task->ledger = ledger; + +#if defined(CONFIG_SCHED_MULTIQ) + new_task->sched_group = sched_group_create(); +#endif + + /* if inherit_memory is true, parent_task MUST not be NULL */ if (inherit_memory) - new_task->map = vm_map_fork(parent_task->map); + new_task->map = vm_map_fork(ledger, parent_task->map); else - new_task->map = vm_map_create(pmap_create(0), - round_page(VM_MIN_ADDRESS), - trunc_page(VM_MAX_ADDRESS), TRUE); + new_task->map = vm_map_create(pmap_create(ledger, 0, is_64bit), + (vm_map_offset_t)(VM_MIN_ADDRESS), + (vm_map_offset_t)(VM_MAX_ADDRESS), TRUE); + + /* 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); - mutex_init(&new_task->lock, ETAP_THREAD_TASK_NEW); - queue_init(&new_task->subsystem_list); - queue_init(&new_task->thr_acts); + lck_mtx_init(&new_task->lock, &task_lck_grp, &task_lck_attr); + queue_init(&new_task->threads); new_task->suspend_count = 0; - new_task->thr_act_count = 0; - new_task->res_act_count = 0; - new_task->active_act_count = 0; + new_task->thread_count = 0; + new_task->active_thread_count = 0; new_task->user_stop_count = 0; - new_task->importance = 0; + new_task->legacy_stop_count = 0; new_task->active = TRUE; - new_task->kernel_loaded = kernel_loaded; - 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; new_task->messages_sent = 0; new_task->messages_received = 0; 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->t_flags = 0; + new_task->importance = 0; + +#if CONFIG_ATM + new_task->atm_context = NULL; +#endif +#if CONFIG_BANK + new_task->bank_context = NULL; +#endif + + zinfo_task_init(new_task); #ifdef MACH_BSD - new_task->bsd_info = 0; + new_task->bsd_info = NULL; #endif /* MACH_BSD */ -#if TASK_SWAPPER - new_task->swap_state = TASK_SW_IN; - new_task->swap_flags = 0; - new_task->swap_ast_waiting = 0; - new_task->swap_stamp = sched_tick; - new_task->swap_rss = 0; - new_task->swap_nswap = 0; -#endif /* TASK_SWAPPER */ +#if CONFIG_JETSAM + if (max_task_footprint != 0) { + ledger_set_limit(ledger, task_ledgers.phys_footprint, max_task_footprint, PHYS_FOOTPRINT_WARNING_LEVEL); + } +#endif + + if (task_wakeups_monitor_rate != 0) { + uint32_t flags = WAKEMON_ENABLE | WAKEMON_SET_DEFAULTS; + int32_t rate; // Ignored because of WAKEMON_SET_DEFAULTS + task_wakeups_monitor_ctl(new_task, &flags, &rate); + } + +#if defined(__i386__) || defined(__x86_64__) + new_task->i386_ldt = 0; +#endif + + new_task->task_debug = NULL; queue_init(&new_task->semaphore_list); - queue_init(&new_task->lock_set_list); 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 */ - eml_task_reference(new_task, parent_task); ipc_task_init(new_task, parent_task); - new_task->total_user_time.seconds = 0; - new_task->total_user_time.microseconds = 0; - new_task->total_system_time.seconds = 0; - new_task->total_system_time.microseconds = 0; + 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->policy = parent_task->policy; + new_task->affinity_space = NULL; - new_task->priority = parent_task->priority; - new_task->max_priority = parent_task->max_priority; +#if CONFIG_COUNTERS + new_task->t_chud = 0U; +#endif - new_task->sec_token = parent_task->sec_token; + new_task->pidsuspended = FALSE; + new_task->frozen = FALSE; + new_task->changing_freeze_state = FALSE; + new_task->rusage_cpu_flags = 0; + new_task->rusage_cpu_percentage = 0; + new_task->rusage_cpu_interval = 0; + new_task->rusage_cpu_deadline = 0; + new_task->rusage_cpu_callt = NULL; +#if MACH_ASSERT + new_task->suspends_outstanding = 0; +#endif - shared_region_mapping_ref(parent_task->system_shared_region); - new_task->system_shared_region = parent_task->system_shared_region; +#if HYPERVISOR + new_task->hv_task_target = NULL; +#endif /* HYPERVISOR */ - new_task->wired_ledger_port = ledger_copy( - convert_port_to_ledger(parent_task->wired_ledger_port)); - new_task->paged_ledger_port = ledger_copy( - convert_port_to_ledger(parent_task->paged_ledger_port)); - } - else { - pset = &default_pset; - if (kernel_task == TASK_NULL) { - new_task->policy = POLICY_RR; + new_task->low_mem_notified_warn = 0; + new_task->low_mem_notified_critical = 0; + new_task->purged_memory_warn = 0; + new_task->purged_memory_critical = 0; + new_task->mem_notify_reserved = 0; +#if IMPORTANCE_INHERITANCE + new_task->task_imp_base = NULL; +#endif /* IMPORTANCE_INHERITANCE */ - new_task->priority = MINPRI_KERNBAND; - new_task->max_priority = MAXPRI_KERNBAND; - } - else { - new_task->policy = POLICY_TIMESHARE; +#if defined(__x86_64__) + new_task->uexc_range_start = new_task->uexc_range_size = new_task->uexc_handler = 0; +#endif - new_task->priority = BASEPRI_DEFAULT; - new_task->max_priority = MAXPRI_HIGHBAND; - } + new_task->requested_policy = default_task_requested_policy; + new_task->effective_policy = default_task_effective_policy; + new_task->pended_policy = default_task_pended_policy; - new_task->sec_token = KERNEL_SECURITY_TOKEN; - new_task->wired_ledger_port = ledger_copy(root_wired_ledger); - new_task->paged_ledger_port = ledger_copy(root_paged_ledger); - } - - 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 */ - -#if FAST_TAS - if (inherit_memory) { - new_task->fast_tas_base = parent_task->fast_tas_base; - new_task->fast_tas_end = parent_task->fast_tas_end; - } else { - new_task->fast_tas_base = (vm_offset_t)0; - new_task->fast_tas_end = (vm_offset_t)0; - } -#endif /* FAST_TAS */ + if (parent_task != TASK_NULL) { + new_task->sec_token = parent_task->sec_token; + new_task->audit_token = parent_task->audit_token; - ipc_task_enable(new_task); + /* inherit the parent's shared region */ + shared_region = vm_shared_region_get(parent_task); + vm_shared_region_set(new_task, shared_region); -#if TASK_SWAPPER - task_swapout_eligible(new_task); -#endif /* TASK_SWAPPER */ + 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 MACH_ASSERT - if (watchacts & WA_TASK) - printf("*** task_create_local(par=%x inh=%x) == 0x%x\n", - parent_task, inherit_memory, new_task); -#endif /* MACH_ASSERT */ +#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); - *child_task = new_task; - return(KERN_SUCCESS); -} + new_task->pset_hint = parent_task->pset_hint = task_choose_pset(parent_task); -/* - * task_free: - * - * Called by task_deallocate when the task's reference count drops to zero. - * Task is locked. - */ -void -task_free( - task_t task) -{ - processor_set_t pset; +#if IMPORTANCE_INHERITANCE + ipc_importance_task_t new_task_imp = IIT_NULL; -#if MACH_ASSERT - assert(task != 0); - if (watchacts & (WA_EXIT|WA_TASK)) - printf("task_free(%x(%d)) map ref %d\n", task, task->ref_count, - task->map->ref_count); -#endif /* MACH_ASSERT */ + if (task_is_marked_importance_donor(parent_task)) { + new_task_imp = ipc_importance_for_task(new_task, FALSE); + assert(IIT_NULL != new_task_imp); + ipc_importance_task_mark_donor(new_task_imp, TRUE); + } + /* Embedded doesn't want this to inherit */ + if (task_is_marked_importance_receiver(parent_task)) { + if (IIT_NULL == new_task_imp) + new_task_imp = ipc_importance_for_task(new_task, FALSE); + assert(IIT_NULL != new_task_imp); + ipc_importance_task_mark_receiver(new_task_imp, TRUE); + } + if (task_is_marked_importance_denap_receiver(parent_task)) { + if (IIT_NULL == new_task_imp) + new_task_imp = ipc_importance_for_task(new_task, FALSE); + assert(IIT_NULL != new_task_imp); + ipc_importance_task_mark_denap_receiver(new_task_imp, TRUE); + } + + if (IIT_NULL != new_task_imp) { + assert(new_task->task_imp_base == new_task_imp); + ipc_importance_task_release(new_task_imp); + } +#endif /* IMPORTANCE_INHERITANCE */ -#if TASK_SWAPPER - /* task_terminate guarantees that this task is off the list */ - assert((task->swap_state & TASK_SW_ELIGIBLE) == 0); -#endif /* TASK_SWAPPER */ + new_task->priority = BASEPRI_DEFAULT; + new_task->max_priority = MAXPRI_USER; - eml_task_deallocate(task); + new_task->requested_policy.t_apptype = parent_task->requested_policy.t_apptype; - /* - * Temporarily restore the reference we dropped above, then - * freeze the task so that the task->processor_set field - * cannot change. In the !MACH_HOST case, the logic can be - * simplified, since the default_pset is the only pset. - */ - ++task->ref_count; - task_unlock(task); -#if MACH_HOST - task_freeze(task); -#endif /* MACH_HOST */ + new_task->requested_policy.int_darwinbg = parent_task->requested_policy.int_darwinbg; + new_task->requested_policy.ext_darwinbg = parent_task->requested_policy.ext_darwinbg; + new_task->requested_policy.int_iotier = parent_task->requested_policy.int_iotier; + new_task->requested_policy.ext_iotier = parent_task->requested_policy.ext_iotier; + new_task->requested_policy.int_iopassive = parent_task->requested_policy.int_iopassive; + new_task->requested_policy.ext_iopassive = parent_task->requested_policy.ext_iopassive; + new_task->requested_policy.bg_iotier = parent_task->requested_policy.bg_iotier; + new_task->requested_policy.terminated = parent_task->requested_policy.terminated; + new_task->requested_policy.t_qos_clamp = parent_task->requested_policy.t_qos_clamp; + + task_policy_create(new_task, parent_task->requested_policy.t_boosted); + } else { + new_task->sec_token = KERNEL_SECURITY_TOKEN; + new_task->audit_token = KERNEL_AUDIT_TOKEN; +#ifdef __LP64__ + if(is_64bit) + task_set_64BitAddr(new_task); +#endif + new_task->all_image_info_addr = (mach_vm_address_t)0; + new_task->all_image_info_size = (mach_vm_size_t)0; + + new_task->pset_hint = PROCESSOR_SET_NULL; + + if (kernel_task == TASK_NULL) { + new_task->priority = BASEPRI_KERNEL; + new_task->max_priority = MAXPRI_KERNEL; + } else { + new_task->priority = BASEPRI_DEFAULT; + new_task->max_priority = MAXPRI_USER; + } + } + + new_task->coalition = COALITION_NULL; + +#if CONFIG_COALITIONS + if (parent_coalition) { + coalition_adopt_task(parent_coalition, new_task); + } else if (parent_task && parent_task->coalition) { + coalition_adopt_task(parent_task->coalition, new_task); + } else { + coalition_default_adopt_task(new_task); + } + + if (new_task->coalition == COALITION_NULL) { + panic("created task is not a member of any coalition"); + } +#endif /* CONFIG_COALITIONS */ + + /* Allocate I/O Statistics */ + new_task->task_io_stats = (io_stat_info_t)kalloc(sizeof(struct io_stat_info)); + assert(new_task->task_io_stats != NULL); + bzero(new_task->task_io_stats, sizeof(struct io_stat_info)); + + bzero(&(new_task->cpu_time_qos_stats), sizeof(struct _cpu_time_qos_stats)); + + bzero(&new_task->extmod_statistics, sizeof(new_task->extmod_statistics)); + new_task->task_timer_wakeups_bin_1 = new_task->task_timer_wakeups_bin_2 = 0; + new_task->task_gpu_ns = 0; + 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); + + new_task->task_volatile_objects = 0; + new_task->task_nonvolatile_objects = 0; + new_task->task_purgeable_disowning = FALSE; + new_task->task_purgeable_disowned = FALSE; + + ipc_task_enable(new_task); + + *child_task = new_task; + return(KERN_SUCCESS); +} + +int task_dropped_imp_count = 0; + +/* + * task_deallocate: + * + * Drop a reference on a task. + */ +void +task_deallocate( + task_t task) +{ + ledger_amount_t credit, debit, interrupt_wakeups, platform_idle_wakeups; + uint32_t refs; + + if (task == TASK_NULL) + return; + + refs = task_deallocate_internal(task); + +#if IMPORTANCE_INHERITANCE + if (refs > 1) + return; - pset = task->processor_set; - pset_lock(pset); - task_lock(task); - if (--task->ref_count > 0) { + if (refs == 1) { /* - * A new reference appeared (probably from the pset). - * Back out. Must unfreeze inline since we'already - * dropped our reference. + * If last ref potentially comes from the task's importance, + * disconnect it. But more task refs may be added before + * that completes, so wait for the reference to go to zero + * naturually (it may happen on a recursive task_deallocate() + * from the ipc_importance_disconnect_task() call). */ -#if MACH_HOST - 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); - } -#endif /* MACH_HOST */ - task_unlock(task); - pset_unlock(pset); + if (IIT_NULL != task->task_imp_base) + ipc_importance_disconnect_task(task); return; } - pset_remove_task(pset,task); - task_unlock(task); - pset_unlock(pset); - pset_deallocate(pset); +#else + if (refs > 0) + return; +#endif /* IMPORTANCE_INHERITANCE */ + + lck_mtx_lock(&tasks_threads_lock); + queue_remove(&terminated_tasks, task, task_t, tasks); + terminated_tasks_count--; + lck_mtx_unlock(&tasks_threads_lock); + + /* + * remove the reference on atm descriptor + */ + task_atm_reset(task); + +#if CONFIG_BANK + /* + * remove the reference on bank context + */ + if (task->bank_context != NULL) { + bank_task_destroy(task->bank_context); + task->bank_context = NULL; + } +#endif + + if (task->task_io_stats) + kfree(task->task_io_stats, sizeof(struct io_stat_info)); + + /* + * Give the machine dependent code a chance + * to perform cleanup before ripping apart + * the task. + */ + machine_task_terminate(task); ipc_task_terminate(task); - shared_region_mapping_dealloc(task->system_shared_region); - if (task->kernel_loaded) - vm_map_remove(kernel_map, task->map->min_offset, - task->map->max_offset, VM_MAP_NO_FLAGS); + if (task->affinity_space) + task_affinity_deallocate(task); + +#if MACH_ASSERT + if (task->ledger != NULL && + task->map != NULL && + task->map->pmap != NULL && + task->map->pmap->ledger != NULL) { + assert(task->ledger == task->map->pmap->ledger); + } +#endif /* MACH_ASSERT */ + + vm_purgeable_disown(task); + assert(task->task_purgeable_disowned); + if (task->task_volatile_objects != 0 || + task->task_nonvolatile_objects != 0) { + panic("task_deallocate(%p): " + "volatile_objects=%d nonvolatile_objects=%d\n", + task, + task->task_volatile_objects, + task->task_nonvolatile_objects); + } + vm_map_deallocate(task->map); is_release(task->itk_space); - task_prof_deallocate(task); - zfree(task_zone, (vm_offset_t) task); -} -void -task_deallocate( - task_t task) -{ - if (task != TASK_NULL) { - int c; + ledger_get_entries(task->ledger, task_ledgers.interrupt_wakeups, + &interrupt_wakeups, &debit); + ledger_get_entries(task->ledger, task_ledgers.platform_idle_wakeups, + &platform_idle_wakeups, &debit); - task_lock(task); - c = --task->ref_count; - if (c == 0) - task_free(task); /* unlocks task */ - else - task_unlock(task); +#if defined(CONFIG_SCHED_MULTIQ) + sched_group_destroy(task->sched_group); +#endif + + /* Accumulate statistics for dead tasks */ + lck_spin_lock(&dead_task_statistics_lock); + dead_task_statistics.total_user_time += task->total_user_time; + dead_task_statistics.total_system_time += task->total_system_time; + + dead_task_statistics.task_interrupt_wakeups += interrupt_wakeups; + dead_task_statistics.task_platform_idle_wakeups += platform_idle_wakeups; + + dead_task_statistics.task_timer_wakeups_bin_1 += task->task_timer_wakeups_bin_1; + dead_task_statistics.task_timer_wakeups_bin_2 += task->task_timer_wakeups_bin_2; + + lck_spin_unlock(&dead_task_statistics_lock); + lck_mtx_destroy(&task->lock, &task_lck_grp); + + if (!ledger_get_entries(task->ledger, task_ledgers.tkm_private, &credit, + &debit)) { + OSAddAtomic64(credit, (int64_t *)&tasks_tkm_private.alloc); + OSAddAtomic64(debit, (int64_t *)&tasks_tkm_private.free); + } + if (!ledger_get_entries(task->ledger, task_ledgers.tkm_shared, &credit, + &debit)) { + OSAddAtomic64(credit, (int64_t *)&tasks_tkm_shared.alloc); + OSAddAtomic64(debit, (int64_t *)&tasks_tkm_shared.free); } + ledger_dereference(task->ledger); + zinfo_task_free(task); + +#if TASK_REFERENCE_LEAK_DEBUG + btlog_remove_entries_for_element(task_ref_btlog, task); +#endif + +#if CONFIG_COALITIONS + if (!task->coalition) { + panic("deallocating task was not a member of any coalition"); + } + coalition_release(task->coalition); +#endif /* CONFIG_COALITIONS */ + + task->coalition = COALITION_NULL; + + zfree(task_zone, task); } +/* + * task_name_deallocate: + * + * Drop a reference on a task name. + */ void -task_reference( - task_t task) +task_name_deallocate( + task_name_t task_name) { - if (task != TASK_NULL) { - task_lock(task); - task->ref_count++; - task_unlock(task); - } + return(task_deallocate((task_t)task_name)); } -boolean_t -task_reference_try( - task_t task) +/* + * task_suspension_token_deallocate: + * + * Drop a reference on a task suspension token. + */ +void +task_suspension_token_deallocate( + task_suspension_token_t token) { - if (task != TASK_NULL) { - if (task_lock_try(task)) { - task->ref_count++; - task_unlock(task); - return TRUE; - } - } - return FALSE; + return(task_deallocate((task_t)token)); } /* @@ -611,70 +1178,90 @@ task_terminate( task_t task) { if (task == TASK_NULL) - return(KERN_INVALID_ARGUMENT); + return (KERN_INVALID_ARGUMENT); + if (task->bsd_info) - return(KERN_FAILURE); + return (KERN_FAILURE); + return (task_terminate_internal(task)); } +#if MACH_ASSERT +extern int proc_pid(struct proc *); +extern void proc_name_kdp(task_t t, char *buf, int size); +#endif /* MACH_ASSERT */ + +#define VM_MAP_PARTIAL_REAP 0x54 /* 0x150 */ +static void +__unused task_partial_reap(task_t task, __unused int pid) +{ + unsigned int reclaimed_resident = 0; + unsigned int reclaimed_compressed = 0; + uint64_t task_page_count; + + task_page_count = (get_task_phys_footprint(task) / PAGE_SIZE_64); + + KERNEL_DEBUG_CONSTANT((MACHDBG_CODE(DBG_MACH_VM, VM_MAP_PARTIAL_REAP) | DBG_FUNC_START), + pid, task_page_count, 0, 0, 0); + + vm_map_partial_reap(task->map, &reclaimed_resident, &reclaimed_compressed); + + KERNEL_DEBUG_CONSTANT((MACHDBG_CODE(DBG_MACH_VM, VM_MAP_PARTIAL_REAP) | DBG_FUNC_END), + pid, reclaimed_resident, reclaimed_compressed, 0, 0); +} + kern_return_t task_terminate_internal( - task_t task) + task_t task) { - thread_act_t thr_act, cur_thr_act; - task_t cur_task; + thread_t thread, self; + task_t self_task; + boolean_t interrupt_save; + int pid = 0; assert(task != kernel_task); - cur_thr_act = current_act(); - cur_task = cur_thr_act->task; - -#if TASK_SWAPPER - /* - * If task is not resident (swapped out, or being swapped - * out), we want to bring it back in (this can block). - * NOTE: The only way that this can happen in the current - * system is if the task is swapped while it has a thread - * in exit(), and the thread does not hit a clean point - * to swap itself before getting here. - * Terminating other tasks is another way to this code, but - * it is not yet fully supported. - * The task_swapin is unconditional. It used to be done - * only if the task is not resident. Swapping in a - * resident task will prevent it from being swapped out - * while it terminates. - */ - task_swapin(task, TRUE); /* TRUE means make it unswappable */ -#endif /* TASK_SWAPPER */ + self = current_thread(); + self_task = self->task; /* * Get the task locked and make sure that we are not racing * with someone else trying to terminate us. */ - if (task == cur_task) { + if (task == self_task) task_lock(task); - } else if (task < cur_task) { + else + if (task < self_task) { task_lock(task); - task_lock(cur_task); - } else { - task_lock(cur_task); + task_lock(self_task); + } + else { + task_lock(self_task); task_lock(task); } - if (!task->active || !cur_thr_act->active) { + if (!task->active) { /* - * Task or current act is already being terminated. + * Task is already being terminated. * Just return an error. If we are dying, this will * just get us to our AST special handler and that * will get us to finalize the termination of ourselves. */ task_unlock(task); - if (cur_task != task) - task_unlock(cur_task); - return(KERN_FAILURE); + if (self_task != task) + task_unlock(self_task); + + return (KERN_FAILURE); } - if (cur_task != task) - task_unlock(cur_task); + + if (self_task != task) + task_unlock(self_task); + + /* + * Make sure the current thread does not get aborted out of + * the waits inside these operations. + */ + interrupt_save = thread_interrupt_level(THREAD_UNINT); /* * Indicate that we want all the threads to stop executing @@ -688,47 +1275,60 @@ task_terminate_internal( task->active = FALSE; ipc_task_disable(task); +#if CONFIG_TELEMETRY /* - * Terminate each activation in the task. - * - * Each terminated activation will run it's special handler - * when its current kernel context is unwound. That will - * clean up most of the thread resources. Then it will be - * handed over to the reaper, who will finally remove the - * thread from the task list and free the structures. - * - * We can't terminate the current activation yet, because - * it has to wait for the others in an interruptible state. - * We may also block interruptibly during the rest of the - * cleanup. Wait until the very last to terminate ourself. - * - * But if we have virtual machine state, we need to clean - * that up now, because it may be holding wirings the task's - * map that would get stuck in the vm_map_remove() below. - */ - queue_iterate(&task->thr_acts, thr_act, thread_act_t, thr_acts) { - if (thr_act != cur_thr_act) - thread_terminate_internal(thr_act); - else - act_virtual_machine_destroy(thr_act); + * Notify telemetry that this task is going away. + */ + telemetry_task_ctl_locked(task, TF_TELEMETRY, 0); +#endif + + /* + * Terminate each thread in the task. + */ + queue_iterate(&task->threads, thread, thread_t, task_threads) { + thread_terminate_internal(thread); } + +#ifdef MACH_BSD + if (task->bsd_info != NULL) { + pid = proc_pid(task->bsd_info); + } +#endif /* MACH_BSD */ + task_unlock(task); + proc_set_task_policy(task, THREAD_NULL, TASK_POLICY_ATTRIBUTE, + TASK_POLICY_TERMINATED, TASK_POLICY_ENABLE); + + /* Early object reap phase */ + +// PR-17045188: Revisit implementation +// task_partial_reap(task, pid); + + /* * Destroy all synchronizers owned by the task. */ task_synchronizer_destroy_all(task); - /* - * Deallocate all subsystems owned by the task. - */ - task_subsystem_destroy_all(task); - /* * Destroy the IPC space, leaving just a reference for it. */ - if (!task->kernel_loaded) - ipc_space_destroy(task->itk_space); + ipc_space_terminate(task->itk_space); + +#if 00 + /* if some ledgers go negative on tear-down again... */ + ledger_disable_panic_on_negative(task->map->pmap->ledger, + task_ledgers.phys_footprint); + ledger_disable_panic_on_negative(task->map->pmap->ledger, + task_ledgers.internal); + ledger_disable_panic_on_negative(task->map->pmap->ledger, + task_ledgers.internal_compressed); + ledger_disable_panic_on_negative(task->map->pmap->ledger, + task_ledgers.iokit_mapped); + ledger_disable_panic_on_negative(task->map->pmap->ledger, + task_ledgers.alternate_accounting); +#endif /* * If the current thread is a member of the task @@ -738,56 +1338,88 @@ task_terminate_internal( * expense of removing the address space regions * at reap time, we do it explictly here. */ - (void) vm_map_remove(task->map, - task->map->min_offset, - task->map->max_offset, VM_MAP_NO_FLAGS); + vm_map_remove(task->map, + task->map->min_offset, + task->map->max_offset, + VM_MAP_NO_FLAGS); + + /* release our shared region */ + vm_shared_region_set(task, NULL); + +#if MACH_ASSERT + /* + * Identify the pmap's process, in case the pmap ledgers drift + * and we have to report it. + */ + char procname[17]; + if (task->bsd_info) { + pid = proc_pid(task->bsd_info); + proc_name_kdp(task, procname, sizeof (procname)); + } else { + pid = 0; + strlcpy(procname, "", sizeof (procname)); + } + pmap_set_process(task->map->pmap, pid, procname); +#endif /* MACH_ASSERT */ + + lck_mtx_lock(&tasks_threads_lock); + queue_remove(&tasks, task, task_t, tasks); + queue_enter(&terminated_tasks, task, task_t, tasks); + tasks_count--; + terminated_tasks_count++; + lck_mtx_unlock(&tasks_threads_lock); + + /* + * We no longer need to guard against being aborted, so restore + * the previous interruptible state. + */ + thread_interrupt_level(interrupt_save); + +#if KPERF + /* force the task to release all ctrs */ + if (task->t_chud & TASK_KPC_FORCED_ALL_CTRS) + kpc_force_all_ctrs(task, 0); +#endif + +#if CONFIG_COALITIONS + /* + * Leave our coalition. (drop activation but not reference) + */ + coalition_remove_task(task); +#endif /* - * Finally, mark ourself for termination and then - * deallocate the task's reference to itself. + * Get rid of the task active reference on itself. */ - if (task == cur_task) - thread_terminate(cur_thr_act); task_deallocate(task); - return(KERN_SUCCESS); + return (KERN_SUCCESS); } /* - * task_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. + * 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 and mark all other threads in the task for + * termination. */ kern_return_t -task_halt( +task_start_halt( task_t task) { - thread_act_t thr_act, cur_thr_act; - task_t cur_task; + thread_t thread, self; assert(task != kernel_task); - cur_thr_act = current_act(); - cur_task = cur_thr_act->task; + self = current_thread(); - if (task != cur_task) { - return(KERN_INVALID_ARGUMENT); - } - -#if TASK_SWAPPER - /* - * If task is not resident (swapped out, or being swapped - * out), we want to bring it back in and make it unswappable. - * This can block, so do it early. - */ - task_swapin(task, TRUE); /* TRUE means make it unswappable */ -#endif /* TASK_SWAPPER */ + if (task != self->task) + return (KERN_INVALID_ARGUMENT); task_lock(task); - if (!task->active || !cur_thr_act->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 @@ -795,10 +1427,14 @@ task_halt( * ourselves. */ task_unlock(task); - return(KERN_FAILURE); + + return (KERN_FAILURE); } - if (task->thr_act_count > 1) { + 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 @@ -808,55 +1444,74 @@ task_halt( task_hold_locked(task); /* - * Terminate all the other activations in the task. - * - * Each terminated activation will run it's special handler - * when its current kernel context is unwound. That will - * clean up most of the thread resources. Then it will be - * handed over to the reaper, who will finally remove the - * thread from the task list and free the structures. - * - * If the current thread has any virtual machine state - * associated with it, clean that up now before we try - * to clean up the task VM and port spaces. + * Terminate all the other threads in the task. */ - queue_iterate(&task->thr_acts, thr_act, thread_act_t,thr_acts) { - if (thr_act != cur_thr_act) - thread_terminate_internal(thr_act); - else - act_virtual_machine_destroy(thr_act); + queue_iterate(&task->threads, thread, thread_t, task_threads) { + if (thread != self) + thread_terminate_internal(thread); } + 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()); /* - * Destroy all synchronizers owned by the task. + * Wait for the other threads to get shut down. + * When the last other thread is reaped, we'll be + * woken up. */ - task_synchronizer_destroy_all(task); + 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); + } /* - * Deallocate all subsystems owned by the task. + * Give the machine dependent code a chance + * to perform cleanup of task-level resources + * associated with the current thread before + * ripping apart the task. */ - task_subsystem_destroy_all(task); + machine_task_terminate(task); /* - * Destroy the IPC space, leaving just a reference for it. + * Destroy all synchronizers owned by the task. */ -#if 0 - if (!task->kernel_loaded) - ipc_space_clean(task->itk_space); -#endif + task_synchronizer_destroy_all(task); + + /* + * Destroy the contents of the IPC space, leaving just + * a reference for it. + */ + ipc_space_clean(task->itk_space); /* * Clean out the address space, as we are going to be * getting a new one. */ - (void) vm_map_remove(task->map, - task->map->min_offset, - task->map->max_offset, VM_MAP_NO_FLAGS); + vm_map_remove(task->map, task->map->min_offset, + task->map->max_offset, VM_MAP_NO_FLAGS); - return KERN_SUCCESS; + task->halting = FALSE; } /* @@ -870,21 +1525,22 @@ task_halt( */ void task_hold_locked( - register task_t task) + register task_t task) { - register thread_act_t thr_act; + register thread_t thread; assert(task->active); - task->suspend_count++; + if (task->suspend_count++ > 0) + return; /* - * Iterate through all the thread_act's and hold them. + * Iterate through all the threads and hold them. */ - queue_iterate(&task->thr_acts, thr_act, thread_act_t, thr_acts) { - act_lock_thread(thr_act); - thread_hold(thr_act); - act_unlock_thread(thr_act); + queue_iterate(&task->threads, thread, thread_t, task_threads) { + thread_mtx_lock(thread); + thread_hold(thread); + thread_mtx_unlock(thread); } } @@ -900,25 +1556,51 @@ task_hold_locked( * CONDITIONS: the caller holds a reference on the task */ kern_return_t -task_hold(task_t task) +task_hold( + register task_t task) +{ + if (task == TASK_NULL) + return (KERN_INVALID_ARGUMENT); + + task_lock(task); + + if (!task->active) { + task_unlock(task); + + return (KERN_FAILURE); + } + + task_hold_locked(task); + task_unlock(task); + + return (KERN_SUCCESS); +} + +kern_return_t +task_wait( + task_t task, + boolean_t until_not_runnable) { - kern_return_t kret; - if (task == TASK_NULL) return (KERN_INVALID_ARGUMENT); + task_lock(task); + if (!task->active) { task_unlock(task); + return (KERN_FAILURE); } - task_hold_locked(task); - task_unlock(task); - return(KERN_SUCCESS); + task_wait_locked(task, until_not_runnable); + task_unlock(task); + + return (KERN_SUCCESS); } /* - * Routine: task_wait_locked + * task_wait_locked: + * * Wait for all threads in task to stop. * * Conditions: @@ -926,27 +1608,24 @@ task_hold(task_t task) */ void task_wait_locked( - register task_t task) + register task_t task, + boolean_t until_not_runnable) { - register thread_act_t thr_act, cur_thr_act; + register thread_t thread, self; assert(task->active); assert(task->suspend_count > 0); - cur_thr_act = current_act(); + self = current_thread(); + /* - * Iterate through all the thread's and wait for them to + * Iterate through all the threads and wait for them to * stop. Do not wait for the current thread if it is within * the task. */ - queue_iterate(&task->thr_acts, thr_act, thread_act_t, thr_acts) { - if (thr_act != cur_thr_act) { - thread_shuttle_t thr_shuttle; - - thr_shuttle = act_lock_thread(thr_act); - thread_wait(thr_shuttle); - act_unlock_thread(thr_act); - } + queue_iterate(&task->threads, thread, thread_t, task_threads) { + if (thread != self) + thread_wait(thread, until_not_runnable); } } @@ -959,24 +1638,20 @@ task_wait_locked( */ void task_release_locked( - register task_t task) + register task_t task) { - register thread_act_t thr_act; + register thread_t thread; assert(task->active); + assert(task->suspend_count > 0); - task->suspend_count--; - assert(task->suspend_count >= 0); + if (--task->suspend_count > 0) + return; - /* - * Iterate through all the thread_act's and hold them. - * Do not hold the current thread_act if it is within the - * task. - */ - queue_iterate(&task->thr_acts, thr_act, thread_act_t, thr_acts) { - act_lock_thread(thr_act); - thread_release(thr_act); - act_unlock_thread(thr_act); + queue_iterate(&task->threads, thread, thread_t, task_threads) { + thread_mtx_lock(thread); + thread_release(thread); + thread_mtx_unlock(thread); } } @@ -989,56 +1664,59 @@ task_release_locked( * CONDITIONS: The caller holds a reference to the task */ kern_return_t -task_release(task_t task) +task_release( + task_t task) { - kern_return_t kret; - if (task == TASK_NULL) return (KERN_INVALID_ARGUMENT); + task_lock(task); + if (!task->active) { task_unlock(task); + return (KERN_FAILURE); } - task_release_locked(task); - task_unlock(task); - return(KERN_SUCCESS); -} + task_release_locked(task); + task_unlock(task); + + return (KERN_SUCCESS); +} kern_return_t task_threads( - task_t task, - thread_act_array_t *thr_act_list, + task_t task, + thread_act_array_t *threads_out, mach_msg_type_number_t *count) { - unsigned int actual; /* this many thr_acts */ - thread_act_t thr_act; - thread_act_t *thr_acts; - thread_t thread; - int i, j; - - vm_size_t size, size_needed; - vm_offset_t addr; + mach_msg_type_number_t actual; + thread_t *thread_list; + thread_t thread; + vm_size_t size, size_needed; + void *addr; + unsigned int i, j; if (task == TASK_NULL) - return KERN_INVALID_ARGUMENT; + return (KERN_INVALID_ARGUMENT); - size = 0; addr = 0; + size = 0; addr = NULL; for (;;) { task_lock(task); if (!task->active) { task_unlock(task); + if (size != 0) kfree(addr, size); - return KERN_FAILURE; + + return (KERN_FAILURE); } - actual = task->thr_act_count; + actual = task->thread_count; /* do we have the memory we need? */ - size_needed = actual * sizeof(mach_port_t); + size_needed = actual * sizeof (mach_port_t); if (size_needed <= size) break; @@ -1053,95 +1731,99 @@ task_threads( addr = kalloc(size); if (addr == 0) - return KERN_RESOURCE_SHORTAGE; + return (KERN_RESOURCE_SHORTAGE); } /* OK, have memory and the task is locked & active */ - thr_acts = (thread_act_t *) addr; - - for (i = j = 0, thr_act = (thread_act_t) queue_first(&task->thr_acts); - i < actual; - i++, thr_act = (thread_act_t) queue_next(&thr_act->thr_acts)) { - act_lock(thr_act); - if (thr_act->ref_count > 0) { - act_locked_act_reference(thr_act); - thr_acts[j++] = thr_act; - } - act_unlock(thr_act); + 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); + thread_list[j++] = thread; } - assert(queue_end(&task->thr_acts, (queue_entry_t) thr_act)); + + assert(queue_end(&task->threads, (queue_entry_t)thread)); actual = j; - size_needed = actual * sizeof(mach_port_t); + size_needed = actual * sizeof (mach_port_t); - /* can unlock task now that we've got the thr_act refs */ + /* can unlock task now that we've got the thread refs */ task_unlock(task); if (actual == 0) { - /* no thr_acts, so return null pointer and deallocate memory */ + /* no threads, so return null pointer and deallocate memory */ - *thr_act_list = 0; + *threads_out = NULL; *count = 0; if (size != 0) kfree(addr, size); - } else { + } + else { /* if we allocated too much, must copy */ if (size_needed < size) { - vm_offset_t newaddr; + void *newaddr; newaddr = kalloc(size_needed); if (newaddr == 0) { - for (i = 0; i < actual; i++) - act_deallocate(thr_acts[i]); + for (i = 0; i < actual; ++i) + thread_deallocate(thread_list[i]); kfree(addr, size); - return KERN_RESOURCE_SHORTAGE; + return (KERN_RESOURCE_SHORTAGE); } - bcopy((char *) addr, (char *) newaddr, size_needed); + bcopy(addr, newaddr, size_needed); kfree(addr, size); - thr_acts = (thread_act_t *) newaddr; + thread_list = (thread_t *)newaddr; } - *thr_act_list = thr_acts; + *threads_out = thread_list; *count = actual; /* do the conversion that Mig should handle */ - for (i = 0; i < actual; i++) - ((ipc_port_t *) thr_acts)[i] = - convert_act_to_port(thr_acts[i]); + for (i = 0; i < actual; ++i) + ((ipc_port_t *) thread_list)[i] = convert_thread_to_port(thread_list[i]); } - return KERN_SUCCESS; + return (KERN_SUCCESS); } -/* - * Routine: task_suspend - * Implement a user-level suspension on a task. - * - * Conditions: - * The caller holds a reference to the task - */ -kern_return_t -task_suspend( - register task_t task) -{ - if (task == TASK_NULL) - return (KERN_INVALID_ARGUMENT); +#define TASK_HOLD_NORMAL 0 +#define TASK_HOLD_PIDSUSPEND 1 +#define TASK_HOLD_LEGACY 2 +#define TASK_HOLD_LEGACY_ALL 3 - task_lock(task); +static kern_return_t +place_task_hold ( + register task_t task, + int mode) +{ if (!task->active) { - task_unlock(task); return (KERN_FAILURE); } - if ((task->user_stop_count)++ > 0) { + + KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, + MACHDBG_CODE(DBG_MACH_IPC,MACH_TASK_SUSPEND) | DBG_FUNC_NONE, + proc_pid(task->bsd_info), ((thread_t)queue_first(&task->threads))->thread_id, + task->user_stop_count, task->user_stop_count + 1, 0); + +#if MACH_ASSERT + current_task()->suspends_outstanding++; +#endif + + if (mode == TASK_HOLD_LEGACY) + task->legacy_stop_count++; + + if (task->user_stop_count++ > 0) { /* * If the stop count was positive, the task is * already stopped and we can exit. */ - task_unlock(task); return (KERN_SUCCESS); } @@ -1152,617 +1834,2287 @@ task_suspend( * to stop executing user code. */ task_hold_locked(task); - task_wait_locked(task); - task_unlock(task); + task_wait_locked(task, FALSE); + return (KERN_SUCCESS); } +static kern_return_t +release_task_hold ( + register task_t task, + int mode) +{ + register boolean_t release = FALSE; + + if (!task->active) { + return (KERN_FAILURE); + } + + if (mode == TASK_HOLD_PIDSUSPEND) { + if (task->pidsuspended == FALSE) { + return (KERN_FAILURE); + } + task->pidsuspended = FALSE; + } + + if (task->user_stop_count > (task->pidsuspended ? 1 : 0)) { + + KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, + MACHDBG_CODE(DBG_MACH_IPC,MACH_TASK_RESUME) | DBG_FUNC_NONE, + proc_pid(task->bsd_info), ((thread_t)queue_first(&task->threads))->thread_id, + task->user_stop_count, mode, task->legacy_stop_count); + +#if MACH_ASSERT + /* + * This is obviously not robust; if we suspend one task and then resume a different one, + * we'll fly under the radar. This is only meant to catch the common case of a crashed + * or buggy suspender. + */ + current_task()->suspends_outstanding--; +#endif + + if (mode == TASK_HOLD_LEGACY_ALL) { + if (task->legacy_stop_count >= task->user_stop_count) { + task->user_stop_count = 0; + release = TRUE; + } else { + task->user_stop_count -= task->legacy_stop_count; + } + task->legacy_stop_count = 0; + } else { + if (mode == TASK_HOLD_LEGACY && task->legacy_stop_count > 0) + task->legacy_stop_count--; + if (--task->user_stop_count == 0) + release = TRUE; + } + } + else { + return (KERN_FAILURE); + } + + /* + * Release the task if necessary. + */ + if (release) + task_release_locked(task); + + return (KERN_SUCCESS); +} + + /* - * Routine: task_resume - * Release a kernel hold on a task. - * + * task_suspend: + * + * Implement an (old-fashioned) user-level suspension on a task. + * + * Because the user isn't expecting to have to manage a suspension + * token, we'll track it for him in the kernel in the form of a naked + * send right to the task's resume port. All such send rights + * account for a single suspension against the task (unlike task_suspend2() + * where each caller gets a unique suspension count represented by a + * unique send-once right). + * * Conditions: - * The caller holds a reference to the task + * The caller holds a reference to the task */ -kern_return_t -task_resume(register task_t task) +kern_return_t +task_suspend( + register task_t task) { - register boolean_t release; + kern_return_t kr; + mach_port_t port, send, old_notify; + mach_port_name_t name; - if (task == TASK_NULL) - return(KERN_INVALID_ARGUMENT); + if (task == TASK_NULL || task == kernel_task) + return (KERN_INVALID_ARGUMENT); - release = FALSE; task_lock(task); - if (!task->active) { - task_unlock(task); - return(KERN_FAILURE); + + /* + * Claim a send right on the task resume port, and request a no-senders + * notification on that port (if none outstanding). + */ + if (task->itk_resume == IP_NULL) { + task->itk_resume = ipc_port_alloc_kernel(); + if (!IP_VALID(task->itk_resume)) + panic("failed to create resume port"); + ipc_kobject_set(task->itk_resume, (ipc_kobject_t)task, IKOT_TASK_RESUME); } - if (task->user_stop_count > 0) { - if (--(task->user_stop_count) == 0) - release = TRUE; + + port = task->itk_resume; + ip_lock(port); + assert(ip_active(port)); + + send = ipc_port_make_send_locked(port); + assert(IP_VALID(send)); + + if (port->ip_nsrequest == IP_NULL) { + ipc_port_nsrequest(port, port->ip_mscount, ipc_port_make_sonce_locked(port), &old_notify); + assert(old_notify == IP_NULL); + /* port unlocked */ + } else { + ip_unlock(port); } - else { + + /* + * place a legacy hold on the task. + */ + kr = place_task_hold(task, TASK_HOLD_LEGACY); + if (kr != KERN_SUCCESS) { task_unlock(task); - return(KERN_FAILURE); + ipc_port_release_send(send); + return kr; } + task_unlock(task); + /* - * Release the task if necessary. + * Copyout the send right into the calling task's IPC space. It won't know it is there, + * but we'll look it up when calling a traditional resume. Any IPC operations that + * deallocate the send right will auto-release the suspension. */ - if (release) - task_release_locked(task); + if ((kr = ipc_kmsg_copyout_object(current_task()->itk_space, (ipc_object_t)send, + MACH_MSG_TYPE_MOVE_SEND, &name)) != KERN_SUCCESS) { + printf("warning: %s(%d) failed to copyout suspension token for task %s(%d) with error: %d\n", + proc_name_address(current_task()->bsd_info), proc_pid(current_task()->bsd_info), + proc_name_address(task->bsd_info), proc_pid(task->bsd_info), kr); + return (kr); + } - task_unlock(task); - return(KERN_SUCCESS); + return (kr); } -kern_return_t -host_security_set_task_token( - host_security_t host_security, - task_t task, - security_token_t sec_token, - host_priv_t host_priv) +/* + * task_resume: + * Release a user hold on a task. + * + * Conditions: + * The caller holds a reference to the task + */ +kern_return_t +task_resume( + register task_t task) { kern_return_t kr; + mach_port_name_t resume_port_name; + ipc_entry_t resume_port_entry; + ipc_space_t space = current_task()->itk_space; - if (task == TASK_NULL) - return(KERN_INVALID_ARGUMENT); - - if (host_security == HOST_NULL) - return(KERN_INVALID_SECURITY); + if (task == TASK_NULL || task == kernel_task ) + return (KERN_INVALID_ARGUMENT); - task_lock(task); - task->sec_token = sec_token; - task_unlock(task); + /* release a legacy task hold */ + task_lock(task); + kr = release_task_hold(task, TASK_HOLD_LEGACY); + task_unlock(task); - if (host_priv != HOST_PRIV_NULL) { - kr = task_set_special_port(task, - TASK_HOST_PORT, - ipc_port_make_send(realhost.host_priv_self)); + is_write_lock(space); + if (is_active(space) && IP_VALID(task->itk_resume) && + ipc_hash_lookup(space, (ipc_object_t)task->itk_resume, &resume_port_name, &resume_port_entry) == TRUE) { + /* + * We found a suspension token in the caller's IPC space. Release a send right to indicate that + * we are holding one less legacy hold on the task from this caller. If the release failed, + * go ahead and drop all the rights, as someone either already released our holds or the task + * is gone. + */ + if (kr == KERN_SUCCESS) + ipc_right_dealloc(space, resume_port_name, resume_port_entry); + else + ipc_right_destroy(space, resume_port_name, resume_port_entry, FALSE, 0); + /* space unlocked */ } else { - kr = task_set_special_port(task, - TASK_HOST_PORT, - ipc_port_make_send(realhost.host_self)); + is_write_unlock(space); + if (kr == KERN_SUCCESS) + printf("warning: %s(%d) performed out-of-band resume on %s(%d)\n", + proc_name_address(current_task()->bsd_info), proc_pid(current_task()->bsd_info), + proc_name_address(task->bsd_info), proc_pid(task->bsd_info)); } - return(kr); + + return kr; } /* - * Utility routine to set a ledger + * Suspend the target task. + * Making/holding a token/reference/port is the callers responsibility. */ kern_return_t -task_set_ledger( - task_t task, - ledger_t wired, - ledger_t paged) +task_suspend_internal(task_t task) { - if (task == TASK_NULL) - return(KERN_INVALID_ARGUMENT); - - task_lock(task); - if (wired) { - ipc_port_release_send(task->wired_ledger_port); - task->wired_ledger_port = ledger_copy(wired); - } - if (paged) { - ipc_port_release_send(task->paged_ledger_port); - task->paged_ledger_port = ledger_copy(paged); - } - task_unlock(task); + kern_return_t kr; + + if (task == TASK_NULL || task == kernel_task) + return (KERN_INVALID_ARGUMENT); - return(KERN_SUCCESS); + task_lock(task); + kr = place_task_hold(task, TASK_HOLD_NORMAL); + task_unlock(task); + return (kr); } /* - * This routine was added, pretty much exclusively, for registering the - * RPC glue vector for in-kernel short circuited tasks. Rather than - * removing it completely, I have only disabled that feature (which was - * the only feature at the time). It just appears that we are going to - * want to add some user data to tasks in the future (i.e. bsd info, - * task names, etc...), so I left it in the formal task interface. + * Suspend the target task, and return a suspension token. The token + * represents a reference on the suspended task. */ kern_return_t -task_set_info( - task_t task, - task_flavor_t flavor, - task_info_t task_info_in, /* pointer to IN array */ - mach_msg_type_number_t task_info_count) +task_suspend2( + register task_t task, + task_suspension_token_t *suspend_token) { - vm_map_t map; + kern_return_t kr; + + kr = task_suspend_internal(task); + if (kr != KERN_SUCCESS) { + *suspend_token = TASK_NULL; + return (kr); + } - if (task == TASK_NULL) - return(KERN_INVALID_ARGUMENT); + /* + * Take a reference on the target task and return that to the caller + * as a "suspension token," which can be converted into an SO right to + * the now-suspended task's resume port. + */ + task_reference_internal(task); + *suspend_token = task; - switch (flavor) { - default: - return (KERN_INVALID_ARGUMENT); - } return (KERN_SUCCESS); } +/* + * Resume the task + * (reference/token/port management is caller's responsibility). + */ kern_return_t -task_info( - task_t task, - task_flavor_t flavor, - task_info_t task_info_out, - mach_msg_type_number_t *task_info_count) +task_resume_internal( + register task_suspension_token_t task) { - thread_t thread; - vm_map_t map; + kern_return_t kr; - if (task == TASK_NULL) - return(KERN_INVALID_ARGUMENT); + if (task == TASK_NULL || task == kernel_task) + return (KERN_INVALID_ARGUMENT); - switch (flavor) { + task_lock(task); + kr = release_task_hold(task, TASK_HOLD_NORMAL); + task_unlock(task); + return (kr); +} - case TASK_BASIC_INFO: - { - register task_basic_info_t basic_info; +/* + * Resume the task using a suspension token. Consumes the token's ref. + */ +kern_return_t +task_resume2( + register task_suspension_token_t task) +{ + kern_return_t kr; - if (*task_info_count < TASK_BASIC_INFO_COUNT) { - return(KERN_INVALID_ARGUMENT); - } + kr = task_resume_internal(task); + task_suspension_token_deallocate(task); - basic_info = (task_basic_info_t) task_info_out; + return (kr); +} - map = (task == kernel_task) ? kernel_map : task->map; +boolean_t +task_suspension_notify(mach_msg_header_t *request_header) +{ + ipc_port_t port = (ipc_port_t) request_header->msgh_remote_port; + task_t task = convert_port_to_task_suspension_token(port); + mach_msg_type_number_t not_count; - basic_info->virtual_size = map->size; - basic_info->resident_size = pmap_resident_count(map->pmap) - * PAGE_SIZE; + if (task == TASK_NULL || task == kernel_task) + return TRUE; /* nothing to do */ + switch (request_header->msgh_id) { + + case MACH_NOTIFY_SEND_ONCE: + /* release the hold held by this specific send-once right */ task_lock(task); - basic_info->policy = task->policy; - basic_info->suspend_count = task->user_stop_count; - basic_info->user_time.seconds - = task->total_user_time.seconds; - basic_info->user_time.microseconds - = task->total_user_time.microseconds; - basic_info->system_time.seconds - = task->total_system_time.seconds; - basic_info->system_time.microseconds - = task->total_system_time.microseconds; + release_task_hold(task, TASK_HOLD_NORMAL); task_unlock(task); - - *task_info_count = TASK_BASIC_INFO_COUNT; break; - } - - case TASK_THREAD_TIMES_INFO: - { - register task_thread_times_info_t times_info; - register thread_t thread; - register thread_act_t thr_act; - - if (*task_info_count < TASK_THREAD_TIMES_INFO_COUNT) { - return (KERN_INVALID_ARGUMENT); - } - times_info = (task_thread_times_info_t) task_info_out; - times_info->user_time.seconds = 0; - times_info->user_time.microseconds = 0; - times_info->system_time.seconds = 0; - times_info->system_time.microseconds = 0; + case MACH_NOTIFY_NO_SENDERS: + not_count = ((mach_no_senders_notification_t *)request_header)->not_count; task_lock(task); - queue_iterate(&task->thr_acts, thr_act, - thread_act_t, thr_acts) - { - time_value_t user_time, system_time; - spl_t s; - - thread = act_lock_thread(thr_act); - - /* Skip empty threads and threads that have migrated - * into this task: - */ - if (!thread || thr_act->pool_port) { - act_unlock_thread(thr_act); - continue; - } - assert(thread); /* Must have thread, if no thread_pool*/ - s = splsched(); - thread_lock(thread); + ip_lock(port); + if (port->ip_mscount == not_count) { - thread_read_times(thread, &user_time, &system_time); + /* release all the [remaining] outstanding legacy holds */ + assert(port->ip_nsrequest == IP_NULL); + ip_unlock(port); + release_task_hold(task, TASK_HOLD_LEGACY_ALL); + task_unlock(task); - thread_unlock(thread); - splx(s); - act_unlock_thread(thr_act); + } else if (port->ip_nsrequest == IP_NULL) { + ipc_port_t old_notify; - time_value_add(×_info->user_time, &user_time); - time_value_add(×_info->system_time, &system_time); + task_unlock(task); + /* new send rights, re-arm notification at current make-send count */ + ipc_port_nsrequest(port, port->ip_mscount, ipc_port_make_sonce_locked(port), &old_notify); + assert(old_notify == IP_NULL); + /* port unlocked */ + } else { + ip_unlock(port); + task_unlock(task); } - task_unlock(task); - - *task_info_count = TASK_THREAD_TIMES_INFO_COUNT; break; - } - case TASK_SCHED_FIFO_INFO: - { - register policy_fifo_base_t fifo_base; + default: + break; + } - if (*task_info_count < POLICY_FIFO_BASE_COUNT) - return(KERN_INVALID_ARGUMENT); + task_suspension_token_deallocate(task); /* drop token reference */ + return TRUE; +} - fifo_base = (policy_fifo_base_t) task_info_out; +kern_return_t +task_pidsuspend_locked(task_t task) +{ + kern_return_t kr; - task_lock(task); - if (task->policy != POLICY_FIFO) { - task_unlock(task); - return(KERN_INVALID_POLICY); - } + if (task->pidsuspended) { + kr = KERN_FAILURE; + goto out; + } - fifo_base->base_priority = task->priority; - task_unlock(task); + task->pidsuspended = TRUE; - *task_info_count = POLICY_FIFO_BASE_COUNT; - break; - } + kr = place_task_hold(task, TASK_HOLD_PIDSUSPEND); + if (kr != KERN_SUCCESS) { + task->pidsuspended = FALSE; + } +out: + return(kr); +} - case TASK_SCHED_RR_INFO: - { - register policy_rr_base_t rr_base; - if (*task_info_count < POLICY_RR_BASE_COUNT) - return(KERN_INVALID_ARGUMENT); +/* + * task_pidsuspend: + * + * Suspends a task by placing a hold on its threads. + * + * Conditions: + * The caller holds a reference to the task + */ +kern_return_t +task_pidsuspend( + register task_t task) +{ + kern_return_t kr; + + if (task == TASK_NULL || task == kernel_task) + return (KERN_INVALID_ARGUMENT); - rr_base = (policy_rr_base_t) task_info_out; + task_lock(task); - task_lock(task); - if (task->policy != POLICY_RR) { - task_unlock(task); - return(KERN_INVALID_POLICY); - } + kr = task_pidsuspend_locked(task); - rr_base->base_priority = task->priority; - task_unlock(task); + task_unlock(task); - rr_base->quantum = (min_quantum * tick) / 1000; + return (kr); +} - *task_info_count = POLICY_RR_BASE_COUNT; - break; - } +/* If enabled, we bring all the frozen pages back in prior to resumption; otherwise, they're faulted back in on demand */ +#define THAW_ON_RESUME 1 - case TASK_SCHED_TIMESHARE_INFO: - { - register policy_timeshare_base_t ts_base; +/* + * task_pidresume: + * Resumes a previously suspended task. + * + * Conditions: + * The caller holds a reference to the task + */ +kern_return_t +task_pidresume( + register task_t task) +{ + kern_return_t kr; - if (*task_info_count < POLICY_TIMESHARE_BASE_COUNT) - return(KERN_INVALID_ARGUMENT); + if (task == TASK_NULL || task == kernel_task) + return (KERN_INVALID_ARGUMENT); - ts_base = (policy_timeshare_base_t) task_info_out; + task_lock(task); + +#if (CONFIG_FREEZE && THAW_ON_RESUME) + + while (task->changing_freeze_state) { + + assert_wait((event_t)&task->changing_freeze_state, THREAD_UNINT); + task_unlock(task); + thread_block(THREAD_CONTINUE_NULL); task_lock(task); - if (task->policy != POLICY_TIMESHARE) { - task_unlock(task); - return(KERN_INVALID_POLICY); + } + task->changing_freeze_state = TRUE; +#endif + + kr = release_task_hold(task, TASK_HOLD_PIDSUSPEND); + + task_unlock(task); + +#if (CONFIG_FREEZE && THAW_ON_RESUME) + if ((kr == KERN_SUCCESS) && (task->frozen == TRUE)) { + + if (COMPRESSED_PAGER_IS_ACTIVE || DEFAULT_FREEZER_COMPRESSED_PAGER_IS_ACTIVE) { + + kr = KERN_SUCCESS; + } else { + + kr = vm_map_thaw(task->map); } + } + task_lock(task); - ts_base->base_priority = task->priority; + if (kr == KERN_SUCCESS) + task->frozen = FALSE; + task->changing_freeze_state = FALSE; + thread_wakeup(&task->changing_freeze_state); + + task_unlock(task); +#endif + + return (kr); +} + +#if CONFIG_FREEZE + +/* + * task_freeze: + * + * Freeze a task. + * + * Conditions: + * The caller holds a reference to the task + */ +kern_return_t +task_freeze( + register task_t task, + uint32_t *purgeable_count, + uint32_t *wired_count, + uint32_t *clean_count, + uint32_t *dirty_count, + uint32_t dirty_budget, + boolean_t *shared, + boolean_t walk_only) +{ + kern_return_t kr; + + if (task == TASK_NULL || task == kernel_task) + return (KERN_INVALID_ARGUMENT); + + task_lock(task); + + while (task->changing_freeze_state) { + + assert_wait((event_t)&task->changing_freeze_state, THREAD_UNINT); + task_unlock(task); + thread_block(THREAD_CONTINUE_NULL); + + task_lock(task); + } + if (task->frozen) { + task_unlock(task); + return (KERN_FAILURE); + } + task->changing_freeze_state = TRUE; + + task_unlock(task); + + if (walk_only) { + kr = vm_map_freeze_walk(task->map, purgeable_count, wired_count, clean_count, dirty_count, dirty_budget, shared); + } else { + kr = vm_map_freeze(task->map, purgeable_count, wired_count, clean_count, dirty_count, dirty_budget, shared); + } + + task_lock(task); + + if (walk_only == FALSE && kr == KERN_SUCCESS) + task->frozen = TRUE; + task->changing_freeze_state = FALSE; + thread_wakeup(&task->changing_freeze_state); + + task_unlock(task); + + return (kr); +} + +/* + * task_thaw: + * + * Thaw a currently frozen task. + * + * Conditions: + * The caller holds a reference to the task + */ +extern void +vm_consider_waking_compactor_swapper(void); + +kern_return_t +task_thaw( + register task_t task) +{ + kern_return_t kr; + + if (task == TASK_NULL || task == kernel_task) + return (KERN_INVALID_ARGUMENT); + + task_lock(task); + + while (task->changing_freeze_state) { + + assert_wait((event_t)&task->changing_freeze_state, THREAD_UNINT); + task_unlock(task); + thread_block(THREAD_CONTINUE_NULL); + + task_lock(task); + } + if (!task->frozen) { + task_unlock(task); + return (KERN_FAILURE); + } + task->changing_freeze_state = TRUE; + + if (DEFAULT_PAGER_IS_ACTIVE || DEFAULT_FREEZER_IS_ACTIVE) { + task_unlock(task); + + kr = vm_map_thaw(task->map); + + task_lock(task); + + if (kr == KERN_SUCCESS) + task->frozen = FALSE; + } else { + task->frozen = FALSE; + kr = KERN_SUCCESS; + } + + task->changing_freeze_state = FALSE; + thread_wakeup(&task->changing_freeze_state); + + task_unlock(task); + + if (COMPRESSED_PAGER_IS_ACTIVE || DEFAULT_FREEZER_COMPRESSED_PAGER_IS_ACTIVE) { + vm_consider_waking_compactor_swapper(); + } + + return (kr); +} + +#endif /* CONFIG_FREEZE */ + +kern_return_t +host_security_set_task_token( + host_security_t host_security, + task_t task, + security_token_t sec_token, + audit_token_t audit_token, + host_priv_t host_priv) +{ + ipc_port_t host_port; + kern_return_t kr; + + if (task == TASK_NULL) + return(KERN_INVALID_ARGUMENT); + + if (host_security == HOST_NULL) + return(KERN_INVALID_SECURITY); + + task_lock(task); + task->sec_token = sec_token; + task->audit_token = audit_token; + + task_unlock(task); + + if (host_priv != HOST_PRIV_NULL) { + kr = host_get_host_priv_port(host_priv, &host_port); + } else { + kr = host_get_host_port(host_priv_self(), &host_port); + } + assert(kr == KERN_SUCCESS); + kr = task_set_special_port(task, TASK_HOST_PORT, host_port); + return(kr); +} + +kern_return_t +task_send_trace_memory( + task_t target_task, + __unused uint32_t pid, + __unused uint64_t uniqueid) +{ + kern_return_t kr = KERN_INVALID_ARGUMENT; + if (target_task == TASK_NULL) + return (KERN_INVALID_ARGUMENT); + +#if CONFIG_ATM + kr = atm_send_proc_inspect_notification(target_task, + pid, + uniqueid); + +#endif + return (kr); +} +/* + * This routine was added, pretty much exclusively, for registering the + * RPC glue vector for in-kernel short circuited tasks. Rather than + * removing it completely, I have only disabled that feature (which was + * the only feature at the time). It just appears that we are going to + * want to add some user data to tasks in the future (i.e. bsd info, + * task names, etc...), so I left it in the formal task interface. + */ +kern_return_t +task_set_info( + task_t task, + task_flavor_t flavor, + __unused task_info_t task_info_in, /* pointer to IN array */ + __unused mach_msg_type_number_t task_info_count) +{ + if (task == TASK_NULL) + return(KERN_INVALID_ARGUMENT); + + switch (flavor) { + +#if CONFIG_ATM + case TASK_TRACE_MEMORY_INFO: + { + if (task_info_count != TASK_TRACE_MEMORY_INFO_COUNT) + return (KERN_INVALID_ARGUMENT); + + assert(task_info_in != NULL); + task_trace_memory_info_t mem_info; + mem_info = (task_trace_memory_info_t) task_info_in; + kern_return_t kr = atm_register_trace_memory(task, + mem_info->user_memory_address, + mem_info->buffer_size, + mem_info->mailbox_array_size); + return kr; + break; + } + +#endif + default: + return (KERN_INVALID_ARGUMENT); + } + return (KERN_SUCCESS); +} + +kern_return_t +task_info( + task_t task, + task_flavor_t flavor, + 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; + clock_sec_t secs; + clock_usec_t usecs; + + 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 = (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; + + 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, &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; + } + + case TASK_BASIC_INFO_64: + { + task_basic_info_64_t basic_info; + vm_map_t map; + clock_sec_t secs; + clock_usec_t usecs; + + 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_64; + + 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, &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; + } + + case MACH_TASK_BASIC_INFO: + { + mach_task_basic_info_t basic_info; + vm_map_t map; + clock_sec_t secs; + clock_usec_t usecs; + + if (*task_info_count < MACH_TASK_BASIC_INFO_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + basic_info = (mach_task_basic_info_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)); + basic_info->resident_size *= PAGE_SIZE_64; + + basic_info->resident_size_max = + (mach_vm_size_t)(pmap_resident_max(map->pmap)); + basic_info->resident_size_max *= PAGE_SIZE_64; + + 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, &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 = MACH_TASK_BASIC_INFO_COUNT; + break; + } + + case TASK_THREAD_TIMES_INFO: + { + register task_thread_times_info_t times_info; + register thread_t thread; + + 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; + times_info->user_time.microseconds = 0; + times_info->system_time.seconds = 0; + times_info->system_time.microseconds = 0; + + + queue_iterate(&task->threads, thread, thread_t, task_threads) { + time_value_t user_time, system_time; + + if (thread->options & TH_OPT_IDLE_THREAD) + continue; + + thread_read_times(thread, &user_time, &system_time); + + time_value_add(×_info->user_time, &user_time); + time_value_add(×_info->system_time, &system_time); + } + + *task_info_count = TASK_THREAD_TIMES_INFO_COUNT; + break; + } + + case TASK_ABSOLUTETIME_INFO: + { + task_absolutetime_info_t info; + register thread_t thread; + + 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; + + + info->total_user = task->total_user_time; + info->total_system = task->total_system_time; + + queue_iterate(&task->threads, thread, thread_t, task_threads) { + uint64_t tval; + spl_t x; + + if (thread->options & TH_OPT_IDLE_THREAD) + continue; + + x = splsched(); + thread_lock(thread); + + tval = timer_grab(&thread->user_timer); + info->threads_user += tval; + info->total_user += tval; + + tval = timer_grab(&thread->system_timer); + if (thread->precise_user_kernel_time) { + info->threads_system += tval; + info->total_system += tval; + } else { + /* system_timer may represent either sys or user */ + info->threads_user += tval; + info->total_user += tval; + } + + thread_unlock(thread); + splx(x); + } + + + *task_info_count = TASK_ABSOLUTETIME_INFO_COUNT; + break; + } + + case TASK_DYLD_INFO: + { + task_dyld_info_t info; + + /* + * We added the format field to TASK_DYLD_INFO output. For + * temporary backward compatibility, accept the fact that + * clients may ask for the old version - distinquished by the + * size of the expected result structure. + */ +#define TASK_LEGACY_DYLD_INFO_COUNT \ + offsetof(struct task_dyld_info, all_image_info_format)/sizeof(natural_t) + + if (*task_info_count < TASK_LEGACY_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; + + /* only set format on output for those expecting it */ + if (*task_info_count >= TASK_DYLD_INFO_COUNT) { + info->all_image_info_format = task_has_64BitAddr(task) ? + TASK_DYLD_ALL_IMAGE_INFO_64 : + TASK_DYLD_ALL_IMAGE_INFO_32 ; + *task_info_count = TASK_DYLD_INFO_COUNT; + } else { + *task_info_count = TASK_LEGACY_DYLD_INFO_COUNT; + } + break; + } + + case TASK_EXTMOD_INFO: + { + task_extmod_info_t info; + void *p; + + if (*task_info_count < TASK_EXTMOD_INFO_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + info = (task_extmod_info_t)task_info_out; + + p = get_bsdtask_info(task); + if (p) { + proc_getexecutableuuid(p, info->task_uuid, sizeof(info->task_uuid)); + } else { + bzero(info->task_uuid, sizeof(info->task_uuid)); + } + info->extmod_statistics = task->extmod_statistics; + *task_info_count = TASK_EXTMOD_INFO_COUNT; + + break; + } + + case TASK_KERNELMEMORY_INFO: + { + task_kernelmemory_info_t tkm_info; + ledger_amount_t credit, debit; + + if (*task_info_count < TASK_KERNELMEMORY_INFO_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + tkm_info = (task_kernelmemory_info_t) task_info_out; + tkm_info->total_palloc = 0; + tkm_info->total_pfree = 0; + tkm_info->total_salloc = 0; + tkm_info->total_sfree = 0; + + if (task == kernel_task) { + /* + * All shared allocs/frees from other tasks count against + * the kernel private memory usage. If we are looking up + * info for the kernel task, gather from everywhere. + */ + task_unlock(task); + + /* start by accounting for all the terminated tasks against the kernel */ + tkm_info->total_palloc = tasks_tkm_private.alloc + tasks_tkm_shared.alloc; + tkm_info->total_pfree = tasks_tkm_private.free + tasks_tkm_shared.free; + + /* count all other task/thread shared alloc/free against the kernel */ + lck_mtx_lock(&tasks_threads_lock); + + /* XXX this really shouldn't be using the function parameter 'task' as a local var! */ + queue_iterate(&tasks, task, task_t, tasks) { + if (task == kernel_task) { + if (ledger_get_entries(task->ledger, + task_ledgers.tkm_private, &credit, + &debit) == KERN_SUCCESS) { + tkm_info->total_palloc += credit; + tkm_info->total_pfree += debit; + } + } + if (!ledger_get_entries(task->ledger, + task_ledgers.tkm_shared, &credit, &debit)) { + tkm_info->total_palloc += credit; + tkm_info->total_pfree += debit; + } + } + lck_mtx_unlock(&tasks_threads_lock); + } else { + if (!ledger_get_entries(task->ledger, + task_ledgers.tkm_private, &credit, &debit)) { + tkm_info->total_palloc = credit; + tkm_info->total_pfree = debit; + } + if (!ledger_get_entries(task->ledger, + task_ledgers.tkm_shared, &credit, &debit)) { + tkm_info->total_salloc = credit; + tkm_info->total_sfree = debit; + } + task_unlock(task); + } + + *task_info_count = TASK_KERNELMEMORY_INFO_COUNT; + return KERN_SUCCESS; + } + + /* OBSOLETE */ + case TASK_SCHED_FIFO_INFO: + { + + if (*task_info_count < POLICY_FIFO_BASE_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + error = KERN_INVALID_POLICY; + break; + } + + /* OBSOLETE */ + case TASK_SCHED_RR_INFO: + { + register policy_rr_base_t rr_base; + uint32_t quantum_time; + uint64_t quantum_ns; + + if (*task_info_count < POLICY_RR_BASE_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + rr_base = (policy_rr_base_t) task_info_out; + + if (task != kernel_task) { + error = KERN_INVALID_POLICY; + break; + } + + rr_base->base_priority = task->priority; + + quantum_time = SCHED(initial_quantum_size)(THREAD_NULL); + absolutetime_to_nanoseconds(quantum_time, &quantum_ns); + + rr_base->quantum = (uint32_t)(quantum_ns / 1000 / 1000); + + *task_info_count = POLICY_RR_BASE_COUNT; + break; + } + + /* OBSOLETE */ + case TASK_SCHED_TIMESHARE_INFO: + { + register policy_timeshare_base_t ts_base; + + if (*task_info_count < POLICY_TIMESHARE_BASE_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + ts_base = (policy_timeshare_base_t) task_info_out; + + if (task == kernel_task) { + error = KERN_INVALID_POLICY; + break; + } + + ts_base->base_priority = task->priority; + + *task_info_count = POLICY_TIMESHARE_BASE_COUNT; + break; + } + + case TASK_SECURITY_TOKEN: + { + register security_token_t *sec_token_p; + + if (*task_info_count < TASK_SECURITY_TOKEN_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + sec_token_p = (security_token_t *) task_info_out; + + *sec_token_p = task->sec_token; + + *task_info_count = TASK_SECURITY_TOKEN_COUNT; + break; + } + + case TASK_AUDIT_TOKEN: + { + register audit_token_t *audit_token_p; + + if (*task_info_count < TASK_AUDIT_TOKEN_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + audit_token_p = (audit_token_t *) task_info_out; + + *audit_token_p = task->audit_token; + + *task_info_count = TASK_AUDIT_TOKEN_COUNT; + break; + } + + case TASK_SCHED_INFO: + error = KERN_INVALID_ARGUMENT; + break; + + case TASK_EVENTS_INFO: + { + register task_events_info_t events_info; + register thread_t thread; + + if (*task_info_count < TASK_EVENTS_INFO_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + events_info = (task_events_info_t) task_info_out; + + + events_info->faults = task->faults; + events_info->pageins = task->pageins; + events_info->cow_faults = task->cow_faults; + events_info->messages_sent = task->messages_sent; + 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->c_switch; + + queue_iterate(&task->threads, thread, thread_t, task_threads) { + events_info->csw += thread->c_switch; + events_info->syscalls_mach += thread->syscalls_mach; + events_info->syscalls_unix += thread->syscalls_unix; + } + + + *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); + break; + } + case TASK_POWER_INFO: + { + if (*task_info_count < TASK_POWER_INFO_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + task_power_info_locked(task, (task_power_info_t)task_info_out, NULL); + break; + } + + case TASK_POWER_INFO_V2: + { + if (*task_info_count < TASK_POWER_INFO_V2_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + task_power_info_v2_t tpiv2 = (task_power_info_v2_t) task_info_out; + task_power_info_locked(task, &tpiv2->cpu_energy, &tpiv2->gpu_energy); + break; + } + + case TASK_VM_INFO: + case TASK_VM_INFO_PURGEABLE: + { + task_vm_info_t vm_info; + vm_map_t map; + + if (*task_info_count < TASK_VM_INFO_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + vm_info = (task_vm_info_t)task_info_out; + + if (task == kernel_task) { + map = kernel_map; + /* no lock */ + } else { + map = task->map; + vm_map_lock_read(map); + } + + vm_info->virtual_size = (typeof(vm_info->virtual_size))map->size; + vm_info->region_count = map->hdr.nentries; + vm_info->page_size = vm_map_page_size(map); + + vm_info->resident_size = pmap_resident_count(map->pmap); + vm_info->resident_size *= PAGE_SIZE; + vm_info->resident_size_peak = pmap_resident_max(map->pmap); + vm_info->resident_size_peak *= PAGE_SIZE; + +#define _VM_INFO(_name) \ + vm_info->_name = ((mach_vm_size_t) map->pmap->stats._name) * PAGE_SIZE + + _VM_INFO(device); + _VM_INFO(device_peak); + _VM_INFO(external); + _VM_INFO(external_peak); + _VM_INFO(internal); + _VM_INFO(internal_peak); + _VM_INFO(reusable); + _VM_INFO(reusable_peak); + _VM_INFO(compressed); + _VM_INFO(compressed_peak); + _VM_INFO(compressed_lifetime); + + vm_info->purgeable_volatile_pmap = 0; + vm_info->purgeable_volatile_resident = 0; + vm_info->purgeable_volatile_virtual = 0; + if (task == kernel_task) { + /* + * We do not maintain the detailed stats for the + * kernel_pmap, so just count everything as + * "internal"... + */ + vm_info->internal = vm_info->resident_size; + /* + * ... but since the memory held by the VM compressor + * in the kernel address space ought to be attributed + * to user-space tasks, we subtract it from "internal" + * to give memory reporting tools a more accurate idea + * of what the kernel itself is actually using, instead + * of making it look like the kernel is leaking memory + * when the system is under memory pressure. + */ + vm_info->internal -= (VM_PAGE_COMPRESSOR_COUNT * + PAGE_SIZE); + } else { + mach_vm_size_t volatile_virtual_size; + mach_vm_size_t volatile_resident_size; + mach_vm_size_t volatile_pmap_size; + kern_return_t kr; + + if (flavor == TASK_VM_INFO_PURGEABLE) { + kr = vm_map_query_volatile( + map, + &volatile_virtual_size, + &volatile_resident_size, + &volatile_pmap_size); + if (kr == KERN_SUCCESS) { + vm_info->purgeable_volatile_pmap = + volatile_pmap_size; + vm_info->purgeable_volatile_resident = + volatile_resident_size; + vm_info->purgeable_volatile_virtual = + volatile_virtual_size; + } + } + vm_map_unlock_read(map); + } + + *task_info_count = TASK_VM_INFO_COUNT; + break; + } + + case TASK_WAIT_STATE_INFO: + { + /* + * Deprecated flavor. Currently allowing some results until all users + * stop calling it. The results may not be accurate. + */ + task_wait_state_info_t wait_state_info; + uint64_t total_sfi_ledger_val = 0; + + if (*task_info_count < TASK_WAIT_STATE_INFO_COUNT) { + error = KERN_INVALID_ARGUMENT; + break; + } + + wait_state_info = (task_wait_state_info_t) task_info_out; + + wait_state_info->total_wait_state_time = 0; + bzero(wait_state_info->_reserved, sizeof(wait_state_info->_reserved)); + + int i, prev_lentry = -1; + int64_t val_credit, val_debit; + + for (i = 0; i < MAX_SFI_CLASS_ID; i++){ + val_credit =0; + /* + * checking with prev_lentry != entry ensures adjacent classes + * which share the same ledger do not add wait times twice. + * Note: Use ledger() call to get data for each individual sfi class. + */ + if (prev_lentry != task_ledgers.sfi_wait_times[i] && + KERN_SUCCESS == ledger_get_entries(task->ledger, + task_ledgers.sfi_wait_times[i], &val_credit, &val_debit)) { + total_sfi_ledger_val += val_credit; + } + prev_lentry = task_ledgers.sfi_wait_times[i]; + } + + wait_state_info->total_wait_sfi_state_time = total_sfi_ledger_val; + *task_info_count = TASK_WAIT_STATE_INFO_COUNT; + + break; + } + + default: + error = KERN_INVALID_ARGUMENT; + } + + task_unlock(task); + return (error); +} + +/* + * task_power_info + * + * Returns power stats for the task. + * Note: Called with task locked. + */ +void +task_power_info_locked( + task_t task, + task_power_info_t info, + gpu_energy_data_t ginfo) +{ + thread_t thread; + ledger_amount_t tmp; + + task_lock_assert_owned(task); + + ledger_get_entries(task->ledger, task_ledgers.interrupt_wakeups, + (ledger_amount_t *)&info->task_interrupt_wakeups, &tmp); + ledger_get_entries(task->ledger, task_ledgers.platform_idle_wakeups, + (ledger_amount_t *)&info->task_platform_idle_wakeups, &tmp); + + info->task_timer_wakeups_bin_1 = task->task_timer_wakeups_bin_1; + info->task_timer_wakeups_bin_2 = task->task_timer_wakeups_bin_2; + + info->total_user = task->total_user_time; + info->total_system = task->total_system_time; + + if (ginfo) { + ginfo->task_gpu_utilisation = task->task_gpu_ns; + } + + queue_iterate(&task->threads, thread, thread_t, task_threads) { + uint64_t tval; + spl_t x; + + if (thread->options & TH_OPT_IDLE_THREAD) + continue; + + x = splsched(); + thread_lock(thread); + + info->task_timer_wakeups_bin_1 += thread->thread_timer_wakeups_bin_1; + info->task_timer_wakeups_bin_2 += thread->thread_timer_wakeups_bin_2; + + tval = timer_grab(&thread->user_timer); + info->total_user += tval; + + tval = timer_grab(&thread->system_timer); + if (thread->precise_user_kernel_time) { + info->total_system += tval; + } else { + /* system_timer may represent either sys or user */ + info->total_user += tval; + } + + if (ginfo) { + ginfo->task_gpu_utilisation += ml_gpu_stat(thread); + } + thread_unlock(thread); + splx(x); + } +} + +/* + * task_gpu_utilisation + * + * Returns the total gpu time used by the all the threads of the task + * (both dead and alive) + */ +uint64_t +task_gpu_utilisation( + task_t task) +{ + uint64_t gpu_time = 0; + thread_t thread; + + task_lock(task); + gpu_time += task->task_gpu_ns; + + queue_iterate(&task->threads, thread, thread_t, task_threads) { + spl_t x; + x = splsched(); + thread_lock(thread); + gpu_time += ml_gpu_stat(thread); + thread_unlock(thread); + splx(x); + } + + task_unlock(task); + return gpu_time; +} + +kern_return_t +task_purgable_info( + task_t task, + task_purgable_info_t *stats) +{ + if (task == TASK_NULL || stats == NULL) + return KERN_INVALID_ARGUMENT; + /* Take task reference */ + task_reference(task); + vm_purgeable_stats((vm_purgeable_info_t)stats, task); + /* Drop task reference */ + task_deallocate(task); + return KERN_SUCCESS; +} + +void +task_vtimer_set( + task_t task, + integer_t which) +{ + thread_t thread; + spl_t x; + + /* 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) { + x = splsched(); + thread_lock(thread); + if (thread->precise_user_kernel_time) + thread->vtimer_user_save = timer_grab(&thread->user_timer); + else + thread->vtimer_user_save = timer_grab(&thread->system_timer); + thread_unlock(thread); + splx(x); + } + break; + + case TASK_VTIMER_PROF: + queue_iterate(&task->threads, thread, thread_t, task_threads) { + x = splsched(); + thread_lock(thread); + thread->vtimer_prof_save = timer_grab(&thread->user_timer); + thread->vtimer_prof_save += timer_grab(&thread->system_timer); + thread_unlock(thread); + splx(x); + } + break; + + case TASK_VTIMER_RLIM: + queue_iterate(&task->threads, thread, thread_t, task_threads) { + x = splsched(); + thread_lock(thread); + thread->vtimer_rlim_save = timer_grab(&thread->user_timer); + thread->vtimer_rlim_save += timer_grab(&thread->system_timer); + thread_unlock(thread); + splx(x); + } + 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: + if (thread->precise_user_kernel_time) { + tdelt = (uint32_t)timer_delta(&thread->user_timer, + &thread->vtimer_user_save); + } else { + tdelt = (uint32_t)timer_delta(&thread->system_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; + } + +} + +/* + * task_assign: + * + * Change the assigned processor set for the task + */ +kern_return_t +task_assign( + __unused task_t task, + __unused processor_set_t new_pset, + __unused boolean_t assign_threads) +{ + return(KERN_FAILURE); +} + +/* + * task_assign_default: + * + * Version of task_assign to assign to default processor set. + */ +kern_return_t +task_assign_default( + task_t task, + boolean_t assign_threads) +{ + return (task_assign(task, &pset0, assign_threads)); +} + +/* + * task_get_assignment + * + * Return name of processor set that task is assigned to. + */ +kern_return_t +task_get_assignment( + task_t task, + processor_set_t *pset) +{ + if (!task->active) + return(KERN_FAILURE); + + *pset = &pset0; + + return (KERN_SUCCESS); +} + + +/* + * task_policy + * + * Set scheduling policy and parameters, both base and limit, for + * the given task. Policy must be a policy which is enabled for the + * processor set. Change contained threads if requested. + */ +kern_return_t +task_policy( + __unused task_t task, + __unused policy_t policy_id, + __unused policy_base_t base, + __unused mach_msg_type_number_t count, + __unused boolean_t set_limit, + __unused boolean_t change) +{ + return(KERN_FAILURE); +} + +/* + * task_set_policy + * + * Set scheduling policy and parameters, both base and limit, for + * the given task. Policy can be any policy implemented by the + * processor set, whether enabled or not. Change contained threads + * if requested. + */ +kern_return_t +task_set_policy( + __unused task_t task, + __unused processor_set_t pset, + __unused policy_t policy_id, + __unused policy_base_t base, + __unused mach_msg_type_number_t base_count, + __unused policy_limit_t limit, + __unused mach_msg_type_number_t limit_count, + __unused boolean_t change) +{ + return(KERN_FAILURE); +} + +kern_return_t +task_set_ras_pc( + __unused task_t task, + __unused vm_offset_t pc, + __unused vm_offset_t endpc) +{ + return KERN_FAILURE; +} + +void +task_synchronizer_destroy_all(task_t task) +{ + semaphore_t semaphore; + + /* + * Destroy owned semaphores + */ + + while (!queue_empty(&task->semaphore_list)) { + semaphore = (semaphore_t) queue_first(&task->semaphore_list); + (void) semaphore_destroy(task, semaphore); + } +} + +/* + * 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; +} + +#if CONFIG_JETSAM +#define HWM_USERCORE_MINSPACE 250 // free space (in MB) required *after* core file creation + +void __attribute__((noinline)) +THIS_PROCESS_CROSSED_HIGH_WATERMARK__SENDING_EXC_RESOURCE(int max_footprint_mb) +{ + task_t task = current_task(); + int pid = 0; + char *procname = (char *) "unknown"; + mach_exception_data_type_t code[EXCEPTION_CODE_MAX]; + +#ifdef MACH_BSD + pid = proc_selfpid(); + + if (pid == 1) { + /* + * Cannot have ReportCrash analyzing + * a suspended initproc. + */ + return; + } + + if (task->bsd_info != NULL) + procname = proc_name_address(current_task()->bsd_info); +#endif + + if (hwm_user_cores) { + int error; + uint64_t starttime, end; + clock_sec_t secs = 0; + uint32_t microsecs = 0; + + starttime = mach_absolute_time(); + /* + * Trigger a coredump of this process. Don't proceed unless we know we won't + * be filling up the disk; and ignore the core size resource limit for this + * core file. + */ + if ((error = coredump(current_task()->bsd_info, HWM_USERCORE_MINSPACE, 1)) != 0) { + printf("couldn't take coredump of %s[%d]: %d\n", procname, pid, error); + } + /* + * coredump() leaves the task suspended. + */ + task_resume_internal(current_task()); + + end = mach_absolute_time(); + absolutetime_to_microtime(end - starttime, &secs, µsecs); + printf("coredump of %s[%d] taken in %d secs %d microsecs\n", + proc_name_address(current_task()->bsd_info), pid, (int)secs, microsecs); + } + + if (disable_exc_resource) { + printf("process %s[%d] crossed memory high watermark (%d MB); EXC_RESOURCE " + "supressed by a boot-arg.\n", procname, pid, max_footprint_mb); + return; + } + + printf("process %s[%d] crossed memory high watermark (%d MB); sending " + "EXC_RESOURCE.\n", procname, pid, max_footprint_mb); + + code[0] = code[1] = 0; + EXC_RESOURCE_ENCODE_TYPE(code[0], RESOURCE_TYPE_MEMORY); + EXC_RESOURCE_ENCODE_FLAVOR(code[0], FLAVOR_HIGH_WATERMARK); + EXC_RESOURCE_HWM_ENCODE_LIMIT(code[0], max_footprint_mb); + + /* + * Use the _internal_ variant so that no user-space + * process can resume our task from under us. + */ + task_suspend_internal(task); + exception_triage(EXC_RESOURCE, code, EXCEPTION_CODE_MAX); + task_resume_internal(task); +} + +/* + * Callback invoked when a task exceeds its physical footprint limit. + */ +void +task_footprint_exceeded(int warning, __unused const void *param0, __unused const void *param1) +{ + ledger_amount_t max_footprint, max_footprint_mb; + ledger_amount_t footprint_after_purge; + task_t task; + + if (warning == LEDGER_WARNING_DIPPED_BELOW) { + /* + * Task memory limits only provide a warning on the way up. + */ + return; + } + + task = current_task(); + + ledger_get_limit(task->ledger, task_ledgers.phys_footprint, &max_footprint); + max_footprint_mb = max_footprint >> 20; + + /* + * Try and purge all "volatile" memory in that task first. + */ + (void) task_purge_volatile_memory(task); + /* are we still over the limit ? */ + ledger_get_balance(task->ledger, + task_ledgers.phys_footprint, + &footprint_after_purge); + if ((!warning && + footprint_after_purge <= max_footprint) || + (warning && + footprint_after_purge <= ((max_footprint * + PHYS_FOOTPRINT_WARNING_LEVEL) / 100))) { + /* all better now */ + ledger_reset_callback_state(task->ledger, + task_ledgers.phys_footprint); + return; + } + /* still over the limit after purging... */ - *task_info_count = POLICY_TIMESHARE_BASE_COUNT; - break; - } + /* + * If this an actual violation (not a warning), + * generate a non-fatal high watermark EXC_RESOURCE. + */ + if ((warning == 0) && (task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_PHYS_FOOTPRINT_EXCEPTION)) { + THIS_PROCESS_CROSSED_HIGH_WATERMARK__SENDING_EXC_RESOURCE((int)max_footprint_mb); + } - case TASK_SECURITY_TOKEN: - { - register security_token_t *sec_token_p; + memorystatus_on_ledger_footprint_exceeded((warning == LEDGER_WARNING_ROSE_ABOVE) ? TRUE : FALSE, + (int)max_footprint_mb); +} - if (*task_info_count < TASK_SECURITY_TOKEN_COUNT) { - return(KERN_INVALID_ARGUMENT); - } +extern int proc_check_footprint_priv(void); - sec_token_p = (security_token_t *) task_info_out; +kern_return_t +task_set_phys_footprint_limit( + task_t task, + int new_limit_mb, + int *old_limit_mb) +{ + kern_return_t error; - task_lock(task); - *sec_token_p = task->sec_token; - task_unlock(task); + if ((error = proc_check_footprint_priv())) { + return (KERN_NO_ACCESS); + } - *task_info_count = TASK_SECURITY_TOKEN_COUNT; - break; - } - - case TASK_SCHED_INFO: - return(KERN_INVALID_ARGUMENT); + return task_set_phys_footprint_limit_internal(task, new_limit_mb, old_limit_mb, FALSE); +} - case TASK_EVENTS_INFO: - { - register task_events_info_t events_info; +kern_return_t +task_set_phys_footprint_limit_internal( + task_t task, + int new_limit_mb, + int *old_limit_mb, + boolean_t trigger_exception) +{ + ledger_amount_t old; - if (*task_info_count < TASK_EVENTS_INFO_COUNT) { - return(KERN_INVALID_ARGUMENT); - } + ledger_get_limit(task->ledger, task_ledgers.phys_footprint, &old); + + if (old_limit_mb) { + *old_limit_mb = old >> 20; + } - events_info = (task_events_info_t) task_info_out; + if (new_limit_mb == -1) { + /* + * Caller wishes to remove the limit. + */ + ledger_set_limit(task->ledger, task_ledgers.phys_footprint, + max_task_footprint ? max_task_footprint : LEDGER_LIMIT_INFINITY, + max_task_footprint ? PHYS_FOOTPRINT_WARNING_LEVEL : 0); + return (KERN_SUCCESS); + } - task_lock(task); - events_info->faults = task->faults; - events_info->pageins = task->pageins; - events_info->cow_faults = task->cow_faults; - events_info->messages_sent = task->messages_sent; - 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); +#ifdef CONFIG_NOMONITORS + return (KERN_SUCCESS); +#endif /* CONFIG_NOMONITORS */ - *task_info_count = TASK_EVENTS_INFO_COUNT; - break; - } + task_lock(task); - default: - return (KERN_INVALID_ARGUMENT); + if (trigger_exception) { + task->rusage_cpu_flags |= TASK_RUSECPU_FLAGS_PHYS_FOOTPRINT_EXCEPTION; + } else { + task->rusage_cpu_flags &= ~TASK_RUSECPU_FLAGS_PHYS_FOOTPRINT_EXCEPTION; } - return(KERN_SUCCESS); + ledger_set_limit(task->ledger, task_ledgers.phys_footprint, + (ledger_amount_t)new_limit_mb << 20, PHYS_FOOTPRINT_WARNING_LEVEL); + + task_unlock(task); + + return (KERN_SUCCESS); } -/* - * task_assign: - * - * Change the assigned processor set for the task - */ kern_return_t -task_assign( - task_t task, - processor_set_t new_pset, - boolean_t assign_threads) +task_get_phys_footprint_limit( + task_t task, + int *limit_mb) { -#ifdef lint - task++; new_pset++; assign_threads++; -#endif /* lint */ - return(KERN_FAILURE); + ledger_amount_t limit; + + ledger_get_limit(task->ledger, task_ledgers.phys_footprint, &limit); + *limit_mb = limit >> 20; + + return (KERN_SUCCESS); +} +#else /* CONFIG_JETSAM */ +kern_return_t +task_set_phys_footprint_limit( + __unused task_t task, + __unused int new_limit_mb, + __unused int *old_limit_mb) +{ + return (KERN_FAILURE); } -/* - * task_assign_default: - * - * Version of task_assign to assign to default processor set. - */ kern_return_t -task_assign_default( - task_t task, - boolean_t assign_threads) +task_get_phys_footprint_limit( + __unused task_t task, + __unused int *limit_mb) { - return (task_assign(task, &default_pset, assign_threads)); + return (KERN_FAILURE); } +#endif /* CONFIG_JETSAM */ /* - * task_get_assignment - * - * Return name of processor set that task is assigned to. + * We need to export some functions to other components that + * are currently implemented in macros within the osfmk + * component. Just export them as functions of the same name. */ -kern_return_t -task_get_assignment( - task_t task, - processor_set_t *pset) +boolean_t is_kerneltask(task_t t) { - if (!task->active) - return(KERN_FAILURE); + if (t == kernel_task) + return (TRUE); - *pset = task->processor_set; - pset_reference(*pset); - return(KERN_SUCCESS); + return (FALSE); } +int +check_for_tasksuspend(task_t task) +{ -/* - * task_policy - * - * Set scheduling policy and parameters, both base and limit, for - * the given task. Policy must be a policy which is enabled for the - * processor set. Change contained threads if requested. + if (task == TASK_NULL) + return (0); + + return (task->suspend_count > 0); +} + +#undef current_task +task_t current_task(void); +task_t current_task(void) +{ + return (current_task_fast()); +} + +#undef task_reference +void task_reference(task_t task); +void +task_reference( + task_t task) +{ + if (task != TASK_NULL) + task_reference_internal(task); +} + +/* + * This routine is called always with task lock held. + * And it returns a thread handle without reference as the caller + * operates on it under the task lock held. */ -kern_return_t -task_policy( - task_t task, - policy_t policy_id, - policy_base_t base, - mach_msg_type_number_t count, - boolean_t set_limit, - boolean_t change) +thread_t +task_findtid(task_t task, uint64_t tid) { - return(KERN_FAILURE); + thread_t thread= THREAD_NULL; + + queue_iterate(&task->threads, thread, thread_t, task_threads) { + if (thread->thread_id == tid) + return(thread); + } + return(THREAD_NULL); } /* - * task_set_policy - * - * Set scheduling policy and parameters, both base and limit, for - * the given task. Policy can be any policy implemented by the - * processor set, whether enabled or not. Change contained threads - * if requested. + * Control the CPU usage monitor for a task. */ kern_return_t -task_set_policy( - task_t task, - processor_set_t pset, - policy_t policy_id, - policy_base_t base, - mach_msg_type_number_t base_count, - policy_limit_t limit, - mach_msg_type_number_t limit_count, - boolean_t change) +task_cpu_usage_monitor_ctl(task_t task, uint32_t *flags) { - return(KERN_FAILURE); + int error = KERN_SUCCESS; + + if (*flags & CPUMON_MAKE_FATAL) { + task->rusage_cpu_flags |= TASK_RUSECPU_FLAGS_FATAL_CPUMON; + } else { + error = KERN_INVALID_ARGUMENT; + } + + return error; } /* - * task_collect_scan: - * - * Attempt to free resources owned by tasks. + * Control the wakeups monitor for a task. */ - -void -task_collect_scan(void) +kern_return_t +task_wakeups_monitor_ctl(task_t task, uint32_t *flags, int32_t *rate_hz) { - register task_t task, prev_task; - processor_set_t pset = &default_pset; + ledger_t ledger = task->ledger; - prev_task = TASK_NULL; - - pset_lock(pset); - pset->ref_count++; - task = (task_t) queue_first(&pset->tasks); - while (!queue_end(&pset->tasks, (queue_entry_t) task)) { - task_reference(task); - pset_unlock(pset); + task_lock(task); + if (*flags & WAKEMON_GET_PARAMS) { + ledger_amount_t limit; + uint64_t period; - pmap_collect(task->map->pmap); + ledger_get_limit(ledger, task_ledgers.interrupt_wakeups, &limit); + ledger_get_period(ledger, task_ledgers.interrupt_wakeups, &period); - if (prev_task != TASK_NULL) - task_deallocate(prev_task); - prev_task = task; + if (limit != LEDGER_LIMIT_INFINITY) { + /* + * An active limit means the wakeups monitor is enabled. + */ + *rate_hz = (int32_t)(limit / (int64_t)(period / NSEC_PER_SEC)); + *flags = WAKEMON_ENABLE; + if (task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_FATAL_WAKEUPSMON) { + *flags |= WAKEMON_MAKE_FATAL; + } + } else { + *flags = WAKEMON_DISABLE; + *rate_hz = -1; + } - pset_lock(pset); - task = (task_t) queue_next(&task->pset_tasks); + /* + * If WAKEMON_GET_PARAMS is present in flags, all other flags are ignored. + */ + task_unlock(task); + return KERN_SUCCESS; } - pset_unlock(pset); - - pset_deallocate(pset); - - if (prev_task != TASK_NULL) - task_deallocate(prev_task); -} - -boolean_t task_collect_allowed = FALSE; -unsigned task_collect_last_tick = 0; -unsigned task_collect_max_rate = 0; /* in ticks */ -/* - * consider_task_collect: - * - * Called by the pageout daemon when the system needs more free pages. - */ + if (*flags & WAKEMON_ENABLE) { + if (*flags & WAKEMON_SET_DEFAULTS) { + *rate_hz = task_wakeups_monitor_rate; + } -void -consider_task_collect(void) -{ - /* - * By default, don't attempt task collection more frequently - * than once per second. - */ +#ifndef CONFIG_NOMONITORS + if (*flags & WAKEMON_MAKE_FATAL) { + task->rusage_cpu_flags |= TASK_RUSECPU_FLAGS_FATAL_WAKEUPSMON; + } +#endif /* CONFIG_NOMONITORS */ - if (task_collect_max_rate == 0) - task_collect_max_rate = (2 << SCHED_TICK_SHIFT); + if (*rate_hz < 0) { + task_unlock(task); + return KERN_INVALID_ARGUMENT; + } - if (task_collect_allowed && - (sched_tick > (task_collect_last_tick + task_collect_max_rate))) { - task_collect_last_tick = sched_tick; - task_collect_scan(); +#ifndef CONFIG_NOMONITORS + ledger_set_limit(ledger, task_ledgers.interrupt_wakeups, *rate_hz * task_wakeups_monitor_interval, + task_wakeups_monitor_ustackshots_trigger_pct); + ledger_set_period(ledger, task_ledgers.interrupt_wakeups, task_wakeups_monitor_interval * NSEC_PER_SEC); + ledger_enable_callback(ledger, task_ledgers.interrupt_wakeups); +#endif /* CONFIG_NOMONITORS */ + } else if (*flags & WAKEMON_DISABLE) { + /* + * Caller wishes to disable wakeups monitor on the task. + * + * Disable telemetry if it was triggered by the wakeups monitor, and + * remove the limit & callback on the wakeups ledger entry. + */ +#if CONFIG_TELEMETRY + telemetry_task_ctl_locked(current_task(), TF_WAKEMON_WARNING, 0); +#endif + ledger_disable_refill(ledger, task_ledgers.interrupt_wakeups); + ledger_disable_callback(ledger, task_ledgers.interrupt_wakeups); } -} -kern_return_t -task_set_ras_pc( - task_t task, - vm_offset_t pc, - vm_offset_t endpc) -{ -#if FAST_TAS - extern int fast_tas_debug; - - if (fast_tas_debug) { - printf("task 0x%x: setting fast_tas to [0x%x, 0x%x]\n", - task, pc, endpc); - } - task_lock(task); - task->fast_tas_base = pc; - task->fast_tas_end = endpc; task_unlock(task); return KERN_SUCCESS; - -#else /* FAST_TAS */ -#ifdef lint - task++; - pc++; - endpc++; -#endif /* lint */ - - return KERN_FAILURE; - -#endif /* FAST_TAS */ } void -task_synchronizer_destroy_all(task_t task) +task_wakeups_rate_exceeded(int warning, __unused const void *param0, __unused const void *param1) { - semaphore_t semaphore; - lock_set_t lock_set; + if (warning == LEDGER_WARNING_ROSE_ABOVE) { +#if CONFIG_TELEMETRY + /* + * This task is in danger of violating the wakeups monitor. Enable telemetry on this task + * so there are micro-stackshots available if and when EXC_RESOURCE is triggered. + */ + telemetry_task_ctl(current_task(), TF_WAKEMON_WARNING, 1); +#endif + return; + } +#if CONFIG_TELEMETRY /* - * Destroy owned semaphores + * If the balance has dipped below the warning level (LEDGER_WARNING_DIPPED_BELOW) or + * exceeded the limit, turn telemetry off for the task. */ + telemetry_task_ctl(current_task(), TF_WAKEMON_WARNING, 0); +#endif - while (!queue_empty(&task->semaphore_list)) { - semaphore = (semaphore_t) queue_first(&task->semaphore_list); - (void) semaphore_destroy(task, semaphore); + if (warning == 0) { + THIS_PROCESS_IS_CAUSING_TOO_MANY_WAKEUPS__SENDING_EXC_RESOURCE(); } +} + +void __attribute__((noinline)) +THIS_PROCESS_IS_CAUSING_TOO_MANY_WAKEUPS__SENDING_EXC_RESOURCE(void) +{ + task_t task = current_task(); + int pid = 0; + char *procname = (char *) "unknown"; + uint64_t observed_wakeups_rate; + uint64_t permitted_wakeups_rate; + uint64_t observation_interval; + mach_exception_data_type_t code[EXCEPTION_CODE_MAX]; + struct ledger_entry_info lei; + +#ifdef MACH_BSD + pid = proc_selfpid(); + if (task->bsd_info != NULL) + procname = proc_name_address(current_task()->bsd_info); +#endif + + ledger_get_entry_info(task->ledger, task_ledgers.interrupt_wakeups, &lei); /* - * Destroy owned lock sets + * Disable the exception notification so we don't overwhelm + * the listener with an endless stream of redundant exceptions. */ + uint32_t flags = WAKEMON_DISABLE; + task_wakeups_monitor_ctl(task, &flags, NULL); + + observed_wakeups_rate = (lei.lei_balance * (int64_t)NSEC_PER_SEC) / lei.lei_last_refill; + permitted_wakeups_rate = lei.lei_limit / task_wakeups_monitor_interval; + observation_interval = lei.lei_refill_period / NSEC_PER_SEC; - while (!queue_empty(&task->lock_set_list)) { - lock_set = (lock_set_t) queue_first(&task->lock_set_list); - (void) lock_set_destroy(task, lock_set); + if (disable_exc_resource) { + printf("process %s[%d] caught causing excessive wakeups. EXC_RESOURCE " + "supressed by a boot-arg\n", procname, pid); + return; + } + if (audio_active) { + printf("process %s[%d] caught causing excessive wakeups. EXC_RESOURCE " + "supressed due to audio playback\n", procname, pid); + return; + } + printf("process %s[%d] caught causing excessive wakeups. Observed wakeups rate " + "(per sec): %lld; Maximum permitted wakeups rate (per sec): %lld; Observation " + "period: %lld seconds; Task lifetime number of wakeups: %lld\n", + procname, pid, observed_wakeups_rate, permitted_wakeups_rate, + observation_interval, lei.lei_credit); + + code[0] = code[1] = 0; + EXC_RESOURCE_ENCODE_TYPE(code[0], RESOURCE_TYPE_WAKEUPS); + EXC_RESOURCE_ENCODE_FLAVOR(code[0], FLAVOR_WAKEUPS_MONITOR); + EXC_RESOURCE_CPUMONITOR_ENCODE_WAKEUPS_PERMITTED(code[0], task_wakeups_monitor_rate); + EXC_RESOURCE_CPUMONITOR_ENCODE_OBSERVATION_INTERVAL(code[0], observation_interval); + EXC_RESOURCE_CPUMONITOR_ENCODE_WAKEUPS_OBSERVED(code[1], lei.lei_balance * (int64_t)NSEC_PER_SEC / lei.lei_last_refill); + exception_triage(EXC_RESOURCE, code, EXCEPTION_CODE_MAX); + + if (task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_FATAL_WAKEUPSMON) { + task_terminate_internal(task); } } -void -task_subsystem_destroy_all(task_t task) +kern_return_t +task_purge_volatile_memory( + task_t task) { - subsystem_t subsystem; + vm_map_t map; + int num_object_purged; - /* - * Destroy owned subsystems - */ + if (task == TASK_NULL) + return KERN_INVALID_TASK; + + task_lock(task); - while (!queue_empty(&task->subsystem_list)) { - subsystem = (subsystem_t) queue_first(&task->subsystem_list); - subsystem_deallocate(subsystem); + if (!task->active) { + task_unlock(task); + return KERN_INVALID_TASK; + } + map = task->map; + if (map == VM_MAP_NULL) { + task_unlock(task); + return KERN_INVALID_TASK; } + vm_map_reference(task->map); + + task_unlock(task); + + num_object_purged = vm_map_purge(map); + vm_map_deallocate(map); + + return KERN_SUCCESS; } -/* - * task_set_port_space: - * - * Set port name space of task to specified size. - */ +/* Placeholders for the task set/get voucher interfaces */ +kern_return_t +task_get_mach_voucher( + task_t task, + mach_voucher_selector_t __unused which, + ipc_voucher_t *voucher) +{ + if (TASK_NULL == task) + return KERN_INVALID_TASK; + + *voucher = NULL; + return KERN_SUCCESS; +} + +kern_return_t +task_set_mach_voucher( + task_t task, + ipc_voucher_t __unused voucher) +{ + if (TASK_NULL == task) + return KERN_INVALID_TASK; + + return KERN_SUCCESS; +} kern_return_t -task_set_port_space( - task_t task, - int table_entries) +task_swap_mach_voucher( + task_t task, + ipc_voucher_t new_voucher, + ipc_voucher_t *in_out_old_voucher) { - kern_return_t kr; - - is_write_lock(task->itk_space); - kr = ipc_entry_grow_table(task->itk_space, table_entries); - if (kr == KERN_SUCCESS) - is_write_unlock(task->itk_space); - return kr; + if (TASK_NULL == task) + return KERN_INVALID_TASK; + + *in_out_old_voucher = new_voucher; + return KERN_SUCCESS; } -/* - * We need to export some functions to other components that - * are currently implemented in macros within the osfmk - * component. Just export them as functions of the same name. - */ -boolean_t is_kerneltask(task_t t) +void task_set_gpu_denied(task_t task, boolean_t denied) { - if (t == kernel_task) - return(TRUE); - else - return((t->kernel_loaded)); + task_lock(task); + + if (denied) { + task->t_flags |= TF_GPU_DENIED; + } else { + task->t_flags &= ~TF_GPU_DENIED; + } + + task_unlock(task); } -#undef current_task -task_t current_task() +boolean_t task_is_gpu_denied(task_t task) { - return (current_task_fast()); + /* We don't need the lock to read this flag */ + return (task->t_flags & TF_GPU_DENIED) ? TRUE : FALSE; }