X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/b0d623f7f2ae71ed96e60569f61f9a9a27016e80..4b17d6b6e417f714551ec129064745ea9919780e:/osfmk/kern/timer_call.c diff --git a/osfmk/kern/timer_call.c b/osfmk/kern/timer_call.c index 74c4534a2..5a17e057c 100644 --- a/osfmk/kern/timer_call.c +++ b/osfmk/kern/timer_call.c @@ -40,278 +40,614 @@ #include -#if CONFIG_DTRACE && (DEVELOPMENT || DEBUG ) +#if CONFIG_DTRACE #include #endif -decl_simple_lock_data(static,timer_call_lock) -#define qe(x) ((queue_entry_t)(x)) -#define TC(x) ((timer_call_t)(x)) +#if DEBUG +#define TIMER_ASSERT 1 +#endif + +//#define TIMER_ASSERT 1 +//#define TIMER_DBG 1 + +#if TIMER_DBG +#define DBG(x...) kprintf("DBG: " x); +#else +#define DBG(x...) +#endif + +lck_grp_t timer_call_lck_grp; +lck_attr_t timer_call_lck_attr; +lck_grp_attr_t timer_call_lck_grp_attr; + + +#define timer_call_lock_spin(queue) \ + lck_mtx_lock_spin_always(&queue->lock_data) + +#define timer_call_unlock(queue) \ + lck_mtx_unlock_always(&queue->lock_data) + + +#define QUEUE(x) ((queue_t)(x)) +#define MPQUEUE(x) ((mpqueue_head_t *)(x)) +#define TIMER_CALL(x) ((timer_call_t)(x)) + + +uint64_t past_deadline_timers; +uint64_t past_deadline_deltas; +uint64_t past_deadline_longest; +uint64_t past_deadline_shortest = ~0ULL; +enum {PAST_DEADLINE_TIMER_ADJUSTMENT_NS = 10 * 1000}; + +uint64_t past_deadline_timer_adjustment; + +static boolean_t timer_call_enter_internal(timer_call_t call, timer_call_param_t param1, uint64_t deadline, uint32_t flags); +boolean_t mach_timer_coalescing_enabled = TRUE; + +mpqueue_head_t *timer_call_enqueue_deadline_unlocked( + timer_call_t call, + mpqueue_head_t *queue, + uint64_t deadline); + +mpqueue_head_t *timer_call_dequeue_unlocked( + timer_call_t call); + void timer_call_initialize(void) { - simple_lock_init(&timer_call_lock, 0); + lck_attr_setdefault(&timer_call_lck_attr); + lck_grp_attr_setdefault(&timer_call_lck_grp_attr); + lck_grp_init(&timer_call_lck_grp, "timer_call", &timer_call_lck_grp_attr); + nanotime_to_absolutetime(0, PAST_DEADLINE_TIMER_ADJUSTMENT_NS, &past_deadline_timer_adjustment); +} + + +void +timer_call_initialize_queue(mpqueue_head_t *queue) +{ + DBG("timer_call_initialize_queue(%p)\n", queue); + mpqueue_init(queue, &timer_call_lck_grp, &timer_call_lck_attr); } + void timer_call_setup( timer_call_t call, timer_call_func_t func, timer_call_param_t param0) { - call_entry_setup(call, func, param0); + DBG("timer_call_setup(%p,%p,%p)\n", call, func, param0); + call_entry_setup(CE(call), func, param0); + simple_lock_init(&(call)->lock, 0); + call->async_dequeue = FALSE; } -__inline__ queue_t -call_entry_enqueue_deadline( - call_entry_t entry, - queue_t queue, - uint64_t deadline) -{ - queue_t old_queue = entry->queue; - timer_call_t current; - - if (old_queue != queue || entry->deadline < deadline) { - if (old_queue != queue) - current = TC(queue_first(queue)); - else - current = TC(queue_next(qe(entry))); - - if (old_queue != NULL) - (void)remque(qe(entry)); +/* + * Timer call entry locking model + * ============================== + * + * Timer call entries are linked on per-cpu timer queues which are protected + * by the queue lock and the call entry lock. The locking protocol is: + * + * 0) The canonical locking order is timer call entry followed by queue. + * + * 1) With only the entry lock held, entry.queue is valid: + * 1a) NULL: the entry is not queued, or + * 1b) non-NULL: this queue must be locked before the entry is modified. + * After locking the queue, the call.async_dequeue flag must be checked: + * 1c) TRUE: the entry was removed from the queue by another thread + * and we must NULL the entry.queue and reset this flag, or + * 1d) FALSE: (ie. queued), the entry can be manipulated. + * + * 2) If a queue lock is obtained first, the queue is stable: + * 2a) If a try-lock of a queued entry succeeds, the call can be operated on + * and dequeued. + * 2b) If a try-lock fails, it indicates that another thread is attempting + * to change the entry and move it to a different position in this queue + * or to different queue. The entry can be dequeued but it should not be + * operated upon since it is being changed. Furthermore, we don't null + * the entry.queue pointer (protected by the entry lock we don't own). + * Instead, we set the async_dequeue flag -- see (1c). + */ - while (TRUE) { - if ( queue_end(queue, qe(current)) || - deadline < current->deadline ) { - current = TC(queue_prev(qe(current))); - break; - } +/* + * Inlines timer_call_entry_dequeue() and timer_call_entry_enqueue_deadline() + * cast between pointer types (mpqueue_head_t *) and (queue_t) so that + * we can use the call_entry_dequeue() and call_entry_enqueue_deadline() + * methods to operate on timer_call structs as if they are call_entry structs. + * These structures are identical except for their queue head pointer fields. + * + * In the debug case, we assert that the timer call locking protocol + * is being obeyed. + */ +#if TIMER_ASSERT +static __inline__ mpqueue_head_t * +timer_call_entry_dequeue( + timer_call_t entry) +{ + mpqueue_head_t *old_queue = MPQUEUE(CE(entry)->queue); + + if (!hw_lock_held((hw_lock_t)&entry->lock)) + panic("_call_entry_dequeue() " + "entry %p is not locked\n", entry); + /* + * XXX The queue lock is actually a mutex in spin mode + * but there's no way to test for it being held + * so we pretend it's a spinlock! + */ + if (!hw_lock_held((hw_lock_t)&old_queue->lock_data)) + panic("_call_entry_dequeue() " + "queue %p is not locked\n", old_queue); + + call_entry_dequeue(CE(entry)); - current = TC(queue_next(qe(current))); - } + return (old_queue); +} - insque(qe(entry), qe(current)); - } - else - if (deadline < entry->deadline) { - current = TC(queue_prev(qe(entry))); +static __inline__ mpqueue_head_t * +timer_call_entry_enqueue_deadline( + timer_call_t entry, + mpqueue_head_t *queue, + uint64_t deadline) +{ + mpqueue_head_t *old_queue = MPQUEUE(CE(entry)->queue); - (void)remque(qe(entry)); + if (!hw_lock_held((hw_lock_t)&entry->lock)) + panic("_call_entry_enqueue_deadline() " + "entry %p is not locked\n", entry); + /* XXX More lock pretense: */ + if (!hw_lock_held((hw_lock_t)&queue->lock_data)) + panic("_call_entry_enqueue_deadline() " + "queue %p is not locked\n", queue); + if (old_queue != NULL && old_queue != queue) + panic("_call_entry_enqueue_deadline() " + "old_queue %p != queue", old_queue); - while (TRUE) { - if ( queue_end(queue, qe(current)) || - current->deadline <= deadline ) { - break; - } + call_entry_enqueue_deadline(CE(entry), QUEUE(queue), deadline); - current = TC(queue_prev(qe(current))); - } + return (old_queue); +} - insque(qe(entry), qe(current)); - } +#else - entry->queue = queue; - entry->deadline = deadline; +static __inline__ mpqueue_head_t * +timer_call_entry_dequeue( + timer_call_t entry) +{ + return MPQUEUE(call_entry_dequeue(CE(entry))); +} - return (old_queue); +static __inline__ mpqueue_head_t * +timer_call_entry_enqueue_deadline( + timer_call_t entry, + mpqueue_head_t *queue, + uint64_t deadline) +{ + return MPQUEUE(call_entry_enqueue_deadline(CE(entry), + QUEUE(queue), deadline)); } -__inline__ queue_t -call_entry_enqueue_tail( - call_entry_t entry, - queue_t queue) +#endif + +#if TIMER_ASSERT +unsigned timer_call_enqueue_deadline_unlocked_async1; +unsigned timer_call_enqueue_deadline_unlocked_async2; +#endif +/* + * Assumes call_entry and queues unlocked, interrupts disabled. + */ +__inline__ mpqueue_head_t * +timer_call_enqueue_deadline_unlocked( + timer_call_t call, + mpqueue_head_t *queue, + uint64_t deadline) { - queue_t old_queue = entry->queue; + call_entry_t entry = CE(call); + mpqueue_head_t *old_queue; - if (old_queue != NULL) - (void)remque(qe(entry)); + DBG("timer_call_enqueue_deadline_unlocked(%p,%p,)\n", call, queue); - enqueue_tail(queue, qe(entry)); + simple_lock(&call->lock); + old_queue = MPQUEUE(entry->queue); + if (old_queue != NULL) { + timer_call_lock_spin(old_queue); + if (call->async_dequeue) { + /* collision (1c): null queue pointer and reset flag */ + call->async_dequeue = FALSE; + entry->queue = NULL; +#if TIMER_ASSERT + timer_call_enqueue_deadline_unlocked_async1++; +#endif + } else if (old_queue != queue) { + (void)remque(qe(entry)); + entry->queue = NULL; +#if TIMER_ASSERT + timer_call_enqueue_deadline_unlocked_async2++; +#endif + } + if (old_queue != queue) { + timer_call_unlock(old_queue); + timer_call_lock_spin(queue); + } + } else { + timer_call_lock_spin(queue); + } - entry->queue = queue; + timer_call_entry_enqueue_deadline(call, queue, deadline); + timer_call_unlock(queue); + simple_unlock(&call->lock); return (old_queue); } -__inline__ queue_t -call_entry_dequeue( - call_entry_t entry) +#if TIMER_ASSERT +unsigned timer_call_dequeue_unlocked_async1; +unsigned timer_call_dequeue_unlocked_async2; +#endif +mpqueue_head_t * +timer_call_dequeue_unlocked( + timer_call_t call) { - queue_t old_queue = entry->queue; + call_entry_t entry = CE(call); + mpqueue_head_t *old_queue; - if (old_queue != NULL) - (void)remque(qe(entry)); - - entry->queue = NULL; + DBG("timer_call_dequeue_unlocked(%p)\n", call); + simple_lock(&call->lock); + old_queue = MPQUEUE(entry->queue); + if (old_queue != NULL) { + timer_call_lock_spin(old_queue); + if (call->async_dequeue) { + /* collision (1c): null queue pointer and reset flag */ + call->async_dequeue = FALSE; +#if TIMER_ASSERT + timer_call_dequeue_unlocked_async1++; +#endif + } else { + (void)remque(qe(entry)); +#if TIMER_ASSERT + timer_call_dequeue_unlocked_async2++; +#endif + } + entry->queue = NULL; + timer_call_unlock(old_queue); + } + simple_unlock(&call->lock); return (old_queue); } -boolean_t -timer_call_enter( - timer_call_t call, - uint64_t deadline) +static boolean_t +timer_call_enter_internal( + timer_call_t call, + timer_call_param_t param1, + uint64_t deadline, + uint32_t flags) { - queue_t queue, old_queue; + mpqueue_head_t *queue; + mpqueue_head_t *old_queue; spl_t s; + uint64_t slop = 0; s = splclock(); - simple_lock(&timer_call_lock); + + call->soft_deadline = deadline; + call->flags = flags; + + if ((flags & TIMER_CALL_CRITICAL) == 0 && + mach_timer_coalescing_enabled) { + slop = timer_call_slop(deadline); + deadline += slop; + } + +#if defined(__i386__) || defined(__x86_64__) + uint64_t ctime = mach_absolute_time(); + if (__improbable(deadline < ctime)) { + uint64_t delta = (ctime - deadline); + + past_deadline_timers++; + past_deadline_deltas += delta; + if (delta > past_deadline_longest) + past_deadline_longest = deadline; + if (delta < past_deadline_shortest) + past_deadline_shortest = delta; + + deadline = ctime + past_deadline_timer_adjustment; + call->soft_deadline = deadline; + } +#endif + call->ttd = call->soft_deadline - ctime; + +#if CONFIG_DTRACE + DTRACE_TMR6(callout__create, timer_call_func_t, CE(call)->func, + timer_call_param_t, CE(call)->param0, uint32_t, call->flags, + (deadline - call->soft_deadline), + (call->ttd >> 32), (unsigned) (call->ttd & 0xFFFFFFFF)); +#endif queue = timer_queue_assign(deadline); - old_queue = call_entry_enqueue_deadline(call, queue, deadline); + old_queue = timer_call_enqueue_deadline_unlocked(call, queue, deadline); - call->param1 = NULL; + CE(call)->param1 = param1; - simple_unlock(&timer_call_lock); splx(s); return (old_queue != NULL); } +boolean_t +timer_call_enter( + timer_call_t call, + uint64_t deadline, + uint32_t flags) +{ + return timer_call_enter_internal(call, NULL, deadline, flags); +} + boolean_t timer_call_enter1( timer_call_t call, timer_call_param_t param1, - uint64_t deadline) + uint64_t deadline, + uint32_t flags) { - queue_t queue, old_queue; - spl_t s; - - s = splclock(); - simple_lock(&timer_call_lock); - - queue = timer_queue_assign(deadline); - - old_queue = call_entry_enqueue_deadline(call, queue, deadline); - - call->param1 = param1; - - simple_unlock(&timer_call_lock); - splx(s); - - return (old_queue != NULL); + return timer_call_enter_internal(call, param1, deadline, flags); } boolean_t timer_call_cancel( timer_call_t call) { - queue_t old_queue; + mpqueue_head_t *old_queue; spl_t s; s = splclock(); - simple_lock(&timer_call_lock); - old_queue = call_entry_dequeue(call); + old_queue = timer_call_dequeue_unlocked(call); if (old_queue != NULL) { - if (!queue_empty(old_queue)) - timer_queue_cancel(old_queue, call->deadline, TC(queue_first(old_queue))->deadline); + timer_call_lock_spin(old_queue); + if (!queue_empty(&old_queue->head)) + timer_queue_cancel(old_queue, CE(call)->deadline, CE(queue_first(&old_queue->head))->deadline); else - timer_queue_cancel(old_queue, call->deadline, UINT64_MAX); + timer_queue_cancel(old_queue, CE(call)->deadline, UINT64_MAX); + timer_call_unlock(old_queue); } - - simple_unlock(&timer_call_lock); splx(s); +#if CONFIG_DTRACE + DTRACE_TMR6(callout__cancel, timer_call_func_t, CE(call)->func, + timer_call_param_t, CE(call)->param0, uint32_t, call->flags, 0, + (call->ttd >> 32), (unsigned) (call->ttd & 0xFFFFFFFF)); +#endif + return (old_queue != NULL); } +uint32_t timer_queue_shutdown_lock_skips; void timer_queue_shutdown( - queue_t queue) + mpqueue_head_t *queue) { - timer_call_t call; - queue_t new_queue; + timer_call_t call; + mpqueue_head_t *new_queue; spl_t s; + DBG("timer_queue_shutdown(%p)\n", queue); + s = splclock(); - simple_lock(&timer_call_lock); - call = TC(queue_first(queue)); + /* Note comma operator in while expression re-locking each iteration */ + while (timer_call_lock_spin(queue), !queue_empty(&queue->head)) { + call = TIMER_CALL(queue_first(&queue->head)); + if (!simple_lock_try(&call->lock)) { + /* + * case (2b) lock order inversion, dequeue and skip + * Don't change the call_entry queue back-pointer + * but set the async_dequeue field. + */ + timer_queue_shutdown_lock_skips++; + (void) remque(qe(call)); + call->async_dequeue = TRUE; + timer_call_unlock(queue); + continue; + } - while (!queue_end(queue, qe(call))) { - new_queue = timer_queue_assign(call->deadline); + /* remove entry from old queue */ + timer_call_entry_dequeue(call); + timer_call_unlock(queue); - call_entry_enqueue_deadline(call, new_queue, call->deadline); + /* and queue it on new */ + new_queue = timer_queue_assign(CE(call)->deadline); + timer_call_lock_spin(new_queue); + timer_call_entry_enqueue_deadline( + call, new_queue, CE(call)->deadline); + timer_call_unlock(new_queue); - call = TC(queue_first(queue)); + simple_unlock(&call->lock); } - simple_unlock(&timer_call_lock); + timer_call_unlock(queue); splx(s); } +uint32_t timer_queue_expire_lock_skips; uint64_t timer_queue_expire( - queue_t queue, + mpqueue_head_t *queue, uint64_t deadline) { timer_call_t call; - simple_lock(&timer_call_lock); + DBG("timer_queue_expire(%p,)\n", queue); + + timer_call_lock_spin(queue); - call = TC(queue_first(queue)); + while (!queue_empty(&queue->head)) { + call = TIMER_CALL(queue_first(&queue->head)); - while (!queue_end(queue, qe(call))) { - if (call->deadline <= deadline) { + if (call->soft_deadline <= deadline) { timer_call_func_t func; timer_call_param_t param0, param1; - call_entry_dequeue(call); + if (!simple_lock_try(&call->lock)) { + /* case (2b) lock inversion, dequeue and skip */ + timer_queue_expire_lock_skips++; + (void) remque(qe(call)); + call->async_dequeue = TRUE; + continue; + } - func = call->func; - param0 = call->param0; - param1 = call->param1; + timer_call_entry_dequeue(call); - simple_unlock(&timer_call_lock); + func = CE(call)->func; + param0 = CE(call)->param0; + param1 = CE(call)->param1; - KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_EXCP_DECI, - 2) - | DBG_FUNC_START, - func, - param0, - param1, 0, 0); + simple_unlock(&call->lock); + timer_call_unlock(queue); -#if CONFIG_DTRACE && (DEVELOPMENT || DEBUG ) - DTRACE_TMR3(callout__start, timer_call_func_t, func, - timer_call_param_t, param0, - timer_call_param_t, param1); + KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, + DECR_TIMER_CALLOUT | DBG_FUNC_START, + VM_KERNEL_UNSLIDE(func), param0, param1, 0, 0); + +#if CONFIG_DTRACE + DTRACE_TMR6(callout__start, timer_call_func_t, func, + timer_call_param_t, param0, unsigned, call->flags, + 0, (call->ttd >> 32), + (unsigned) (call->ttd & 0xFFFFFFFF)); #endif + /* Maintain time-to-deadline in per-processor data + * structure for thread wakeup deadline statistics. + */ + uint64_t *ttdp = &(PROCESSOR_DATA(current_processor(), timer_call_ttd)); + *ttdp = call->ttd; (*func)(param0, param1); + *ttdp = 0; -#if CONFIG_DTRACE && (DEVELOPMENT || DEBUG ) - DTRACE_TMR3(callout__end, timer_call_func_t, func, - timer_call_param_t, param0, - timer_call_param_t, param1); +#if CONFIG_DTRACE + DTRACE_TMR3(callout__end, timer_call_func_t, func, + timer_call_param_t, param0, timer_call_param_t, + param1); #endif - KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_EXCP_DECI, - 2) - | DBG_FUNC_END, - func, - param0, - param1, 0, 0); + KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE, + DECR_TIMER_CALLOUT | DBG_FUNC_END, + VM_KERNEL_UNSLIDE(func), param0, param1, 0, 0); - simple_lock(&timer_call_lock); + timer_call_lock_spin(queue); } else break; - - call = TC(queue_first(queue)); } - if (!queue_end(queue, qe(call))) - deadline = call->deadline; + if (!queue_empty(&queue->head)) + deadline = CE(call)->deadline; else deadline = UINT64_MAX; - simple_unlock(&timer_call_lock); + timer_call_unlock(queue); return (deadline); } + + +extern int serverperfmode; +uint32_t timer_queue_migrate_lock_skips; +/* + * timer_queue_migrate() is called by etimer_queue_migrate() + * to move timer requests from the local processor (queue_from) + * to a target processor's (queue_to). + */ +int +timer_queue_migrate(mpqueue_head_t *queue_from, mpqueue_head_t *queue_to) +{ + timer_call_t call; + timer_call_t head_to; + int timers_migrated = 0; + + DBG("timer_queue_migrate(%p,%p)\n", queue_from, queue_to); + + assert(!ml_get_interrupts_enabled()); + assert(queue_from != queue_to); + + if (serverperfmode) { + /* + * if we're running a high end server + * avoid migrations... they add latency + * and don't save us power under typical + * server workloads + */ + return -4; + } + + /* + * Take both local (from) and target (to) timer queue locks while + * moving the timers from the local queue to the target processor. + * We assume that the target is always the boot processor. + * But only move if all of the following is true: + * - the target queue is non-empty + * - the local queue is non-empty + * - the local queue's first deadline is later than the target's + * - the local queue contains no non-migrateable "local" call + * so that we need not have the target resync. + */ + + timer_call_lock_spin(queue_to); + + head_to = TIMER_CALL(queue_first(&queue_to->head)); + if (queue_empty(&queue_to->head)) { + timers_migrated = -1; + goto abort1; + } + + timer_call_lock_spin(queue_from); + + if (queue_empty(&queue_from->head)) { + timers_migrated = -2; + goto abort2; + } + + call = TIMER_CALL(queue_first(&queue_from->head)); + if (CE(call)->deadline < CE(head_to)->deadline) { + timers_migrated = 0; + goto abort2; + } + + /* perform scan for non-migratable timers */ + do { + if (call->flags & TIMER_CALL_LOCAL) { + timers_migrated = -3; + goto abort2; + } + call = TIMER_CALL(queue_next(qe(call))); + } while (!queue_end(&queue_from->head, qe(call))); + + /* migration loop itself -- both queues are locked */ + while (!queue_empty(&queue_from->head)) { + call = TIMER_CALL(queue_first(&queue_from->head)); + if (!simple_lock_try(&call->lock)) { + /* case (2b) lock order inversion, dequeue only */ + timer_queue_migrate_lock_skips++; + (void) remque(qe(call)); + call->async_dequeue = TRUE; + continue; + } + timer_call_entry_dequeue(call); + timer_call_entry_enqueue_deadline( + call, queue_to, CE(call)->deadline); + timers_migrated++; + simple_unlock(&call->lock); + } + +abort2: + timer_call_unlock(queue_from); +abort1: + timer_call_unlock(queue_to); + + return timers_migrated; +}