/*
- * Copyright (c) 2000-2010 Apple Inc. All rights reserved.
+ * Copyright (c) 2000-2012 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
#include <kern/host.h>
#include <kern/zalloc.h>
#include <kern/assert.h>
+#include <kern/exc_resource.h>
+#include <kern/telemetry.h>
+#if KPC
+#include <kern/kpc.h>
+#endif
#include <ipc/ipc_kmsg.h>
#include <ipc/ipc_port.h>
#ifdef MACH_BSD
extern void proc_exit(void *);
extern uint64_t get_dispatchqueue_offset_from_proc(void *);
+extern int proc_selfpid(void);
+extern char * proc_name_address(void *p);
#endif /* MACH_BSD */
+extern int disable_exc_resource;
+extern int audio_active;
extern int debug_task;
int thread_max = CONFIG_THREAD_MAX; /* Max number of threads */
int task_threadmax = CONFIG_THREAD_MAX;
struct _thread_ledger_indices thread_ledgers = { -1 };
static ledger_template_t thread_ledger_template = NULL;
void init_thread_ledgers(void);
+int task_disable_cpumon(task_t task);
+
+/*
+ * Level (in terms of percentage of the limit) at which the CPU usage monitor triggers telemetry.
+ *
+ * (ie when any thread's CPU consumption exceeds 70% of the limit, start taking user
+ * stacktraces, aka micro-stackshots)
+ */
+#define CPUMON_USTACKSHOTS_TRIGGER_DEFAULT_PCT 70
+
+int cpumon_ustackshots_trigger_pct; /* Percentage. Level at which we start gathering telemetry. */
+void __attribute__((noinline)) THIS_THREAD_IS_CONSUMING_TOO_MUCH_CPU__SENDING_EXC_RESOURCE(void);
+
+/*
+ * The smallest interval over which we support limiting CPU consumption is 1ms
+ */
+#define MINIMUM_CPULIMIT_INTERVAL_MS 1
void
thread_bootstrap(void)
thread_template.pending_promoter_index = 0;
thread_template.pending_promoter[0] =
thread_template.pending_promoter[1] = NULL;
+ thread_template.rwlock_count = 0;
thread_template.realtime.deadline = UINT64_MAX;
thread_template.t_dtrace_tracing = 0;
#endif /* CONFIG_DTRACE */
+#if KPC
+ thread_template.kpc_buf = NULL;
+#endif
+
thread_template.t_chud = 0;
thread_template.t_page_creation_count = 0;
thread_template.t_page_creation_time = 0;
thread_template.t_ledger = LEDGER_NULL;
thread_template.t_threadledger = LEDGER_NULL;
- thread_template.appliedstate = default_task_null_policy;
- thread_template.ext_appliedstate = default_task_null_policy;
- thread_template.policystate = default_task_proc_policy;
- thread_template.ext_policystate = default_task_proc_policy;
-#if CONFIG_EMBEDDED
- thread_template.taskwatch = NULL;
- thread_template.saved_importance = 0;
-#endif /* CONFIG_EMBEDDED */
+ thread_template.requested_policy = default_task_requested_policy;
+ thread_template.effective_policy = default_task_effective_policy;
+ thread_template.pended_policy = default_task_pended_policy;
+
+ thread_template.iotier_override = THROTTLE_LEVEL_NONE;
+
+ thread_template.thread_callout_interrupt_wakeups = thread_template.thread_callout_platform_idle_wakeups = 0;
+
+ thread_template.thread_timer_wakeups_bin_1 = thread_template.thread_timer_wakeups_bin_2 = 0;
+ thread_template.callout_woken_from_icontext = thread_template.callout_woken_from_platform_idle = 0;
+
+ thread_template.thread_tag = 0;
init_thread = thread_template;
machine_set_current_thread(&init_thread);
*/
machine_thread_init();
+ if (!PE_parse_boot_argn("cpumon_ustackshots_trigger_pct", &cpumon_ustackshots_trigger_pct,
+ sizeof (cpumon_ustackshots_trigger_pct))) {
+ cpumon_ustackshots_trigger_pct = CPUMON_USTACKSHOTS_TRIGGER_DEFAULT_PCT;
+ }
+
init_thread_ledgers();
}
thread_mtx_lock(thread);
- ulock_release_all(thread);
-
ipc_thread_disable(thread);
thread_mtx_unlock(thread);
thread_policy_reset(thread);
-#if CONFIG_EMBEDDED
- thead_remove_taskwatch(thread);
-#endif /* CONFIG_EMBEDDED */
task = thread->task;
uthread_cleanup(task, thread->uthread, task->bsd_info);
thread->state |= TH_TERMINATE;
thread_mark_wait_locked(thread, THREAD_UNINT);
assert(thread->promotions == 0);
+ assert(thread->rwlock_count == 0);
thread_unlock(thread);
/* splsched */
if (thread_deallocate_internal(thread) > 0)
return;
+ if(!(thread->state & TH_TERMINATE2))
+ panic("thread_deallocate: thread not properly terminated\n");
+
+#if KPC
+ kpc_thread_destroy(thread);
+#endif
+
ipc_thread_terminate(thread);
thread_stack_daemon(void)
{
thread_t thread;
+ spl_t s;
+ s = splsched();
simple_lock(&thread_stack_lock);
while ((thread = (thread_t)dequeue_head(&thread_stack_queue)) != THREAD_NULL) {
simple_unlock(&thread_stack_lock);
+ splx(s);
+ /* allocate stack with interrupts enabled so that we can call into VM */
stack_alloc(thread);
- (void)splsched();
+ s = splsched();
thread_lock(thread);
thread_setrun(thread, SCHED_PREEMPT | SCHED_TAILQ);
thread_unlock(thread);
- (void)spllo();
simple_lock(&thread_stack_lock);
}
assert_wait((event_t)&thread_stack_queue, THREAD_UNINT);
simple_unlock(&thread_stack_lock);
+ splx(s);
thread_block((thread_continue_t)thread_stack_daemon);
/*NOTREACHED*/
lck_mtx_init(&new_thread->mutex, &thread_lck_grp, &thread_lck_attr);
ipc_thread_init(new_thread);
- queue_init(&new_thread->held_ulocks);
new_thread->continuation = continuation;
new_thread->t_chud = (TASK_PMC_FLAG == (parent_task->t_chud & TASK_PMC_FLAG)) ?
THREAD_PMC_FLAG : 0U;
#endif
+#if KPC
+ kpc_thread_create(new_thread);
+#endif
+
+ /* Only need to update policies pushed from task to thread */
+ new_thread->requested_policy.bg_iotier = parent_task->effective_policy.bg_iotier;
+ new_thread->requested_policy.terminated = parent_task->effective_policy.terminated;
/* Set the thread's scheduling parameters */
new_thread->sched_mode = SCHED(initial_thread_sched_mode)(parent_task);
new_thread->priority = (priority < 0)? parent_task->priority: priority;
if (new_thread->priority > new_thread->max_priority)
new_thread->priority = new_thread->max_priority;
-#if CONFIG_EMBEDDED
- if (new_thread->priority < MAXPRI_THROTTLE) {
- new_thread->priority = MAXPRI_THROTTLE;
- }
-#endif /* CONFIG_EMBEDDED */
- new_thread->importance =
- new_thread->priority - new_thread->task_priority;
-#if CONFIG_EMBEDDED
+ new_thread->importance = new_thread->priority - new_thread->task_priority;
new_thread->saved_importance = new_thread->importance;
- /* apple ios daemon starts all threads in darwin background */
- if (parent_task->ext_appliedstate.apptype == PROC_POLICY_IOS_APPLE_DAEMON) {
- /* Cannot use generic routines here so apply darwin bacground directly */
- new_thread->policystate.hw_bg = TASK_POLICY_BACKGROUND_ATTRIBUTE_ALL;
- /* set thread self backgrounding */
- new_thread->appliedstate.hw_bg = new_thread->policystate.hw_bg;
- /* priority will get recomputed suitably bit later */
- new_thread->importance = INT_MIN;
- /* to avoid changes to many pri compute routines, set the effect of those here */
- new_thread->priority = MAXPRI_THROTTLE;
- }
-#endif /* CONFIG_EMBEDDED */
#if defined(CONFIG_SCHED_TRADITIONAL)
new_thread->sched_stamp = sched_tick;
stack_alloc(thread);
assert(thread->kernel_stack != 0);
-#if CONFIG_EMBEDDED
- if (priority > BASEPRI_KERNEL)
-#endif
thread->reserved_stack = thread->kernel_stack;
thread->parameter = parameter;
return kernel_thread_start_priority(continuation, parameter, -1, new_thread);
}
-#if defined(__i386__)
-
-thread_t
-kernel_thread(
- task_t task,
- void (*start)(void))
-{
- kern_return_t result;
- thread_t thread;
-
- if (task != kernel_task)
- panic("kernel_thread");
-
- result = kernel_thread_start_priority((thread_continue_t)start, NULL, -1, &thread);
- if (result != KERN_SUCCESS)
- return (THREAD_NULL);
-
- thread_deallocate(thread);
-
- return (thread);
-}
-
-#endif /* defined(__i386__) */
kern_return_t
thread_info_internal(
POLICY_TIMESHARE: POLICY_RR);
flags = 0;
- if (thread->bound_processor != PROCESSOR_NULL && thread->bound_processor->idle_thread == thread)
+ if (thread->options & TH_OPT_IDLE_THREAD)
flags |= TH_FLAGS_IDLE;
if (!thread->kernel_stack)
return (thread_wire_internal(host_priv, thread, wired, NULL));
}
+
+/*
+ * XXX assuming current thread only, for now...
+ */
+void
+thread_guard_violation(thread_t thread, unsigned type)
+{
+ assert(thread == current_thread());
+
+ spl_t s = splsched();
+ /*
+ * Use the saved state area of the thread structure
+ * to store all info required to handle the AST when
+ * returning to userspace
+ */
+ thread->guard_exc_info.type = type;
+ thread_ast_set(thread, AST_GUARD);
+ ast_propagate(thread->ast);
+
+ splx(s);
+}
+
+/*
+ * guard_ast:
+ *
+ * Handle AST_GUARD for a thread. This routine looks at the
+ * state saved in the thread structure to determine the cause
+ * of this exception. Based on this value, it invokes the
+ * appropriate routine which determines other exception related
+ * info and raises the exception.
+ */
+void
+guard_ast(thread_t thread)
+{
+ if (thread->guard_exc_info.type == GUARD_TYPE_MACH_PORT)
+ mach_port_guard_ast(thread);
+ else
+ fd_guard_ast(thread);
+}
+
static void
-thread_resource_exception(const void *arg0, __unused const void *arg1)
+thread_cputime_callback(int warning, __unused const void *arg0, __unused const void *arg1)
{
- thread_t thread = current_thread();
- int code = (int)((uintptr_t)arg0 & ((int)-1));
-
+ if (warning == LEDGER_WARNING_ROSE_ABOVE) {
+#if CONFIG_TELEMETRY
+ /*
+ * This thread is in danger of violating the CPU usage monitor. Enable telemetry
+ * on the entire task so there are micro-stackshots available if and when
+ * EXC_RESOURCE is triggered. We could have chosen to enable micro-stackshots
+ * for this thread only; but now that this task is suspect, knowing what all of
+ * its threads are up to will be useful.
+ */
+ telemetry_task_ctl(current_task(), TF_CPUMON_WARNING, 1);
+#endif
+ return;
+ }
+
+#if CONFIG_TELEMETRY
+ /*
+ * 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_CPUMON_WARNING, 0);
+#endif
+
+ if (warning == 0) {
+ THIS_THREAD_IS_CONSUMING_TOO_MUCH_CPU__SENDING_EXC_RESOURCE();
+ }
+}
+
+void __attribute__((noinline))
+THIS_THREAD_IS_CONSUMING_TOO_MUCH_CPU__SENDING_EXC_RESOURCE(void)
+{
+ int pid = 0;
+ task_t task = current_task();
+ thread_t thread = current_thread();
+ uint64_t tid = thread->thread_id;
+ char *procname = (char *) "unknown";
+ time_value_t thread_total_time = {0, 0};
+ time_value_t thread_system_time;
+ time_value_t thread_user_time;
+ int action;
+ uint8_t percentage;
+ uint32_t limit_percent;
+ uint32_t usage_percent;
+ uint32_t interval_sec;
+ uint64_t interval_ns;
+ uint64_t balance_ns;
+ boolean_t fatal = FALSE;
+
+ mach_exception_data_type_t code[EXCEPTION_CODE_MAX];
+ struct ledger_entry_info lei;
+
assert(thread->t_threadledger != LEDGER_NULL);
/*
- * Disable the exception notification so we don't overwhelm
- * the listener with an endless stream of redundant exceptions.
+ * Now that a thread has tripped the monitor, disable it for the entire task.
*/
- ledger_set_action(thread->t_threadledger, thread_ledgers.cpu_time,
- LEDGER_ACTION_IGNORE);
- ledger_disable_callback(thread->t_threadledger, thread_ledgers.cpu_time);
+ task_lock(task);
+ if ((task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_PERTHR_LIMIT) == 0) {
+ /*
+ * The CPU usage monitor has been disabled on our task, so some other
+ * thread must have gotten here first. We only send one exception per
+ * task lifetime, so there's nothing left for us to do here.
+ */
+ task_unlock(task);
+ return;
+ }
+ if (task->rusage_cpu_flags & TASK_RUSECPU_FLAGS_FATAL_CPUMON) {
+ fatal = TRUE;
+ }
+ task_disable_cpumon(task);
+ task_unlock(task);
+
+#ifdef MACH_BSD
+ pid = proc_selfpid();
+ if (task->bsd_info != NULL)
+ procname = proc_name_address(task->bsd_info);
+#endif
+
+ thread_get_cpulimit(&action, &percentage, &interval_ns);
+
+ interval_sec = (uint32_t)(interval_ns / NSEC_PER_SEC);
+
+ thread_read_times(thread, &thread_user_time, &thread_system_time);
+ time_value_add(&thread_total_time, &thread_user_time);
+ time_value_add(&thread_total_time, &thread_system_time);
+
+ ledger_get_entry_info(thread->t_threadledger, thread_ledgers.cpu_time, &lei);
+
+ absolutetime_to_nanoseconds(lei.lei_balance, &balance_ns);
+ usage_percent = (uint32_t) ((balance_ns * 100ULL) / lei.lei_last_refill);
- /* XXX code should eventually be a user-exported namespace of resources */
- (void) task_exception_notify(EXC_RESOURCE, code, 0);
+ /* Show refill period in the same units as balance, limit, etc */
+ nanoseconds_to_absolutetime(lei.lei_refill_period, &lei.lei_refill_period);
+
+ limit_percent = (uint32_t) ((lei.lei_limit * 100ULL) / lei.lei_refill_period);
+
+ /* TODO: show task total runtime as well? see TASK_ABSOLUTETIME_INFO */
+
+ if (disable_exc_resource) {
+ printf("process %s[%d] thread %llu caught burning CPU!; EXC_RESOURCE "
+ "supressed by a boot-arg\n", procname, pid, tid);
+ return;
+ }
+
+ if (audio_active) {
+ printf("process %s[%d] thread %llu caught burning CPU!; EXC_RESOURCE "
+ "supressed due to audio playback\n", procname, pid, tid);
+ return;
+ }
+ printf("process %s[%d] thread %llu caught burning CPU! "
+ "It used more than %d%% CPU (Actual recent usage: %d%%) over %d seconds. "
+ "thread lifetime cpu usage %d.%06d seconds, (%d.%06d user, %d.%06d system) "
+ "ledger info: balance: %lld credit: %lld debit: %lld limit: %llu (%d%%) "
+ "period: %llu time since last refill (ns): %llu \n",
+ procname, pid, tid,
+ percentage, usage_percent, interval_sec,
+ thread_total_time.seconds, thread_total_time.microseconds,
+ thread_user_time.seconds, thread_user_time.microseconds,
+ thread_system_time.seconds, thread_system_time.microseconds,
+ lei.lei_balance,
+ lei.lei_credit, lei.lei_debit,
+ lei.lei_limit, limit_percent,
+ lei.lei_refill_period, lei.lei_last_refill);
+
+
+ code[0] = code[1] = 0;
+ EXC_RESOURCE_ENCODE_TYPE(code[0], RESOURCE_TYPE_CPU);
+ EXC_RESOURCE_ENCODE_FLAVOR(code[0], FLAVOR_CPU_MONITOR);
+ EXC_RESOURCE_CPUMONITOR_ENCODE_INTERVAL(code[0], interval_sec);
+ EXC_RESOURCE_CPUMONITOR_ENCODE_PERCENTAGE(code[0], limit_percent);
+ EXC_RESOURCE_CPUMONITOR_ENCODE_PERCENTAGE(code[1], usage_percent);
+ exception_triage(EXC_RESOURCE, code, EXCEPTION_CODE_MAX);
+
+ if (fatal) {
+ task_terminate_internal(task);
+ }
}
void
panic("couldn't create cpu_time entry for thread ledger template");
}
- if (ledger_set_callback(t, idx, thread_resource_exception,
- (void *)(uintptr_t)idx, NULL) < 0) {
+ if (ledger_set_callback(t, idx, thread_cputime_callback, NULL, NULL) < 0) {
panic("couldn't set thread ledger callback for cpu_time entry");
}
thread_ledger_template = t;
}
+/*
+ * Returns currently applied CPU usage limit, or 0/0 if none is applied.
+ */
+int
+thread_get_cpulimit(int *action, uint8_t *percentage, uint64_t *interval_ns)
+{
+ int64_t abstime = 0;
+ uint64_t limittime = 0;
+ thread_t thread = current_thread();
+
+ *percentage = 0;
+ *interval_ns = 0;
+ *action = 0;
+
+ if (thread->t_threadledger == LEDGER_NULL) {
+ /*
+ * This thread has no per-thread ledger, so it can't possibly
+ * have a CPU limit applied.
+ */
+ return (KERN_SUCCESS);
+ }
+
+ ledger_get_period(thread->t_threadledger, thread_ledgers.cpu_time, interval_ns);
+ ledger_get_limit(thread->t_threadledger, thread_ledgers.cpu_time, &abstime);
+
+ if ((abstime == LEDGER_LIMIT_INFINITY) || (*interval_ns == 0)) {
+ /*
+ * This thread's CPU time ledger has no period or limit; so it
+ * doesn't have a CPU limit applied.
+ */
+ return (KERN_SUCCESS);
+ }
+
+ /*
+ * This calculation is the converse to the one in thread_set_cpulimit().
+ */
+ absolutetime_to_nanoseconds(abstime, &limittime);
+ *percentage = (limittime * 100ULL) / *interval_ns;
+ assert(*percentage <= 100);
+
+ if (thread->options & TH_OPT_PROC_CPULIMIT) {
+ assert((thread->options & TH_OPT_PRVT_CPULIMIT) == 0);
+
+ *action = THREAD_CPULIMIT_BLOCK;
+ } else if (thread->options & TH_OPT_PRVT_CPULIMIT) {
+ assert((thread->options & TH_OPT_PROC_CPULIMIT) == 0);
+
+ *action = THREAD_CPULIMIT_EXCEPTION;
+ } else {
+ *action = THREAD_CPULIMIT_DISABLE;
+ }
+
+ return (KERN_SUCCESS);
+}
+
/*
* Set CPU usage limit on a thread.
*
* Calling with percentage of 0 will unset the limit for this thread.
*/
-
int
thread_set_cpulimit(int action, uint8_t percentage, uint64_t interval_ns)
{
assert(percentage <= 100);
- if (percentage == 0) {
+ if (action == THREAD_CPULIMIT_DISABLE) {
/*
* Remove CPU limit, if any exists.
*/
if (thread->t_threadledger != LEDGER_NULL) {
+ l = thread->t_threadledger;
/*
* The only way to get a per-thread ledger is via CPU limits.
*/
assert(thread->options & (TH_OPT_PROC_CPULIMIT | TH_OPT_PRVT_CPULIMIT));
- ledger_dereference(thread->t_threadledger);
- thread->t_threadledger = LEDGER_NULL;
+ thread->t_threadledger = NULL;
+ ledger_dereference(l);
thread->options &= ~(TH_OPT_PROC_CPULIMIT | TH_OPT_PRVT_CPULIMIT);
}
return (0);
}
+ if (interval_ns < MINIMUM_CPULIMIT_INTERVAL_MS * NSEC_PER_MSEC) {
+ return (KERN_INVALID_ARGUMENT);
+ }
+
l = thread->t_threadledger;
if (l == LEDGER_NULL) {
/*
*/
limittime = (interval_ns * percentage) / 100;
nanoseconds_to_absolutetime(limittime, &abstime);
- ledger_set_limit(l, thread_ledgers.cpu_time, abstime);
+ ledger_set_limit(l, thread_ledgers.cpu_time, abstime, cpumon_ustackshots_trigger_pct);
/*
* Refill the thread's allotted CPU time every interval_ns nanoseconds.
*/
ledger_set_period(l, thread_ledgers.cpu_time, interval_ns);
- /*
- * Ledgers supports multiple actions for one ledger entry, so we do too.
- */
if (action == THREAD_CPULIMIT_EXCEPTION) {
+ /*
+ * We don't support programming the CPU usage monitor on a task if any of its
+ * threads have a per-thread blocking CPU limit configured.
+ */
+ if (thread->options & TH_OPT_PRVT_CPULIMIT) {
+ panic("CPU usage monitor activated, but blocking thread limit exists");
+ }
+
+ /*
+ * Make a note that this thread's CPU limit is being used for the task-wide CPU
+ * usage monitor. We don't have to arm the callback which will trigger the
+ * exception, because that was done for us in ledger_instantiate (because the
+ * ledger template used has a default callback).
+ */
thread->options |= TH_OPT_PROC_CPULIMIT;
- ledger_set_action(l, thread_ledgers.cpu_time, LEDGER_ACTION_EXCEPTION);
- }
+ } else {
+ /*
+ * We deliberately override any CPU limit imposed by a task-wide limit (eg
+ * CPU usage monitor).
+ */
+ thread->options &= ~TH_OPT_PROC_CPULIMIT;
- if (action == THREAD_CPULIMIT_BLOCK) {
thread->options |= TH_OPT_PRVT_CPULIMIT;
/* The per-thread ledger template by default has a callback for CPU time */
ledger_disable_callback(l, thread_ledgers.cpu_time);
ledger_set_action(l, thread_ledgers.cpu_time, LEDGER_ACTION_BLOCK);
}
- thread->t_threadledger = l;
return (0);
}
return (thread != THREAD_NULL? thread->thread_id: 0);
}
-uint16_t
-thread_set_tag(thread_t th, uint16_t tag) {
+uint16_t thread_set_tag(thread_t th, uint16_t tag) {
return thread_set_tag_internal(th, tag);
}
-uint16_t
-thread_get_tag(thread_t th) {
+uint16_t thread_get_tag(thread_t th) {
return thread_get_tag_internal(th);
}
void dtrace_thread_bootstrap(void)
{
task_t task = current_task();
- if(task->thread_count == 1) {
+
+ if (task->thread_count == 1) {
+ thread_t thread = current_thread();
+ if (thread->t_dtrace_flags & TH_DTRACE_EXECSUCCESS) {
+ thread->t_dtrace_flags &= ~TH_DTRACE_EXECSUCCESS;
+ DTRACE_PROC(exec__success);
+ }
DTRACE_PROC(start);
}
DTRACE_PROC(lwp__start);
}
+
+void
+dtrace_thread_didexec(thread_t thread)
+{
+ thread->t_dtrace_flags |= TH_DTRACE_EXECSUCCESS;
+}
#endif /* CONFIG_DTRACE */
projects / apple / xnu.git / blobdiff