]>
Commit | Line | Data |
---|---|---|
1 | /* | |
2 | * Copyright (c) 1993-2008 Apple Inc. All rights reserved. | |
3 | * | |
4 | * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ | |
5 | * | |
6 | * This file contains Original Code and/or Modifications of Original Code | |
7 | * as defined in and that are subject to the Apple Public Source License | |
8 | * Version 2.0 (the 'License'). You may not use this file except in | |
9 | * compliance with the License. The rights granted to you under the License | |
10 | * may not be used to create, or enable the creation or redistribution of, | |
11 | * unlawful or unlicensed copies of an Apple operating system, or to | |
12 | * circumvent, violate, or enable the circumvention or violation of, any | |
13 | * terms of an Apple operating system software license agreement. | |
14 | * | |
15 | * Please obtain a copy of the License at | |
16 | * http://www.opensource.apple.com/apsl/ and read it before using this file. | |
17 | * | |
18 | * The Original Code and all software distributed under the License are | |
19 | * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER | |
20 | * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, | |
21 | * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, | |
22 | * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. | |
23 | * Please see the License for the specific language governing rights and | |
24 | * limitations under the License. | |
25 | * | |
26 | * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ | |
27 | */ | |
28 | /* | |
29 | * Timer interrupt callout module. | |
30 | */ | |
31 | ||
32 | #include <mach/mach_types.h> | |
33 | ||
34 | #include <kern/clock.h> | |
35 | #include <kern/smp.h> | |
36 | #include <kern/processor.h> | |
37 | #include <kern/timer_call.h> | |
38 | #include <kern/timer_queue.h> | |
39 | #include <kern/call_entry.h> | |
40 | #include <kern/thread.h> | |
41 | #include <kern/policy_internal.h> | |
42 | ||
43 | #include <sys/kdebug.h> | |
44 | ||
45 | #if CONFIG_DTRACE | |
46 | #include <mach/sdt.h> | |
47 | #endif | |
48 | ||
49 | ||
50 | #if DEBUG | |
51 | #define TIMER_ASSERT 1 | |
52 | #endif | |
53 | ||
54 | //#define TIMER_ASSERT 1 | |
55 | //#define TIMER_DBG 1 | |
56 | ||
57 | #if TIMER_DBG | |
58 | #define DBG(x...) kprintf("DBG: " x); | |
59 | #else | |
60 | #define DBG(x...) | |
61 | #endif | |
62 | ||
63 | #if TIMER_TRACE | |
64 | #define TIMER_KDEBUG_TRACE KERNEL_DEBUG_CONSTANT_IST | |
65 | #else | |
66 | #define TIMER_KDEBUG_TRACE(x...) | |
67 | #endif | |
68 | ||
69 | ||
70 | lck_grp_t timer_call_lck_grp; | |
71 | lck_attr_t timer_call_lck_attr; | |
72 | lck_grp_attr_t timer_call_lck_grp_attr; | |
73 | ||
74 | lck_grp_t timer_longterm_lck_grp; | |
75 | lck_attr_t timer_longterm_lck_attr; | |
76 | lck_grp_attr_t timer_longterm_lck_grp_attr; | |
77 | ||
78 | /* Timer queue lock must be acquired with interrupts disabled (under splclock()) */ | |
79 | #if __SMP__ | |
80 | #define timer_queue_lock_spin(queue) \ | |
81 | lck_mtx_lock_spin_always(&queue->lock_data) | |
82 | ||
83 | #define timer_queue_unlock(queue) \ | |
84 | lck_mtx_unlock_always(&queue->lock_data) | |
85 | #else | |
86 | #define timer_queue_lock_spin(queue) (void)1 | |
87 | #define timer_queue_unlock(queue) (void)1 | |
88 | #endif | |
89 | ||
90 | #define QUEUE(x) ((queue_t)(x)) | |
91 | #define MPQUEUE(x) ((mpqueue_head_t *)(x)) | |
92 | #define TIMER_CALL(x) ((timer_call_t)(x)) | |
93 | #define TCE(x) (&(x->call_entry)) | |
94 | /* | |
95 | * The longterm timer object is a global structure holding all timers | |
96 | * beyond the short-term, local timer queue threshold. The boot processor | |
97 | * is responsible for moving each timer to its local timer queue | |
98 | * if and when that timer becomes due within the threshold. | |
99 | */ | |
100 | ||
101 | /* Sentinel for "no time set": */ | |
102 | #define TIMER_LONGTERM_NONE EndOfAllTime | |
103 | /* The default threadhold is the delta above which a timer is "long-term" */ | |
104 | #if defined(__x86_64__) | |
105 | #define TIMER_LONGTERM_THRESHOLD (1ULL * NSEC_PER_SEC) /* 1 sec */ | |
106 | #else | |
107 | #define TIMER_LONGTERM_THRESHOLD TIMER_LONGTERM_NONE /* disabled */ | |
108 | #endif | |
109 | ||
110 | /* | |
111 | * The scan_limit throttles processing of the longterm queue. | |
112 | * If the scan time exceeds this limit, we terminate, unlock | |
113 | * and defer for scan_interval. This prevents unbounded holding of | |
114 | * timer queue locks with interrupts masked. | |
115 | */ | |
116 | #define TIMER_LONGTERM_SCAN_LIMIT (100ULL * NSEC_PER_USEC) /* 100 us */ | |
117 | #define TIMER_LONGTERM_SCAN_INTERVAL (100ULL * NSEC_PER_USEC) /* 100 us */ | |
118 | /* Sentinel for "scan limit exceeded": */ | |
119 | #define TIMER_LONGTERM_SCAN_AGAIN 0 | |
120 | ||
121 | typedef struct { | |
122 | uint64_t interval; /* longterm timer interval */ | |
123 | uint64_t margin; /* fudge factor (10% of interval */ | |
124 | uint64_t deadline; /* first/soonest longterm deadline */ | |
125 | uint64_t preempted; /* sooner timer has pre-empted */ | |
126 | timer_call_t call; /* first/soonest longterm timer call */ | |
127 | uint64_t deadline_set; /* next timer set */ | |
128 | timer_call_data_t timer; /* timer used by threshold management */ | |
129 | /* Stats: */ | |
130 | uint64_t scans; /* num threshold timer scans */ | |
131 | uint64_t preempts; /* num threshold reductions */ | |
132 | uint64_t latency; /* average threshold latency */ | |
133 | uint64_t latency_min; /* minimum threshold latency */ | |
134 | uint64_t latency_max; /* maximum threshold latency */ | |
135 | } threshold_t; | |
136 | ||
137 | typedef struct { | |
138 | mpqueue_head_t queue; /* longterm timer list */ | |
139 | uint64_t enqueues; /* num timers queued */ | |
140 | uint64_t dequeues; /* num timers dequeued */ | |
141 | uint64_t escalates; /* num timers becoming shortterm */ | |
142 | uint64_t scan_time; /* last time the list was scanned */ | |
143 | threshold_t threshold; /* longterm timer threshold */ | |
144 | uint64_t scan_limit; /* maximum scan time */ | |
145 | uint64_t scan_interval; /* interval between LT "escalation" scans */ | |
146 | uint64_t scan_pauses; /* num scans exceeding time limit */ | |
147 | } timer_longterm_t; | |
148 | ||
149 | timer_longterm_t timer_longterm = { | |
150 | .scan_limit = TIMER_LONGTERM_SCAN_LIMIT, | |
151 | .scan_interval = TIMER_LONGTERM_SCAN_INTERVAL, | |
152 | }; | |
153 | ||
154 | static mpqueue_head_t *timer_longterm_queue = NULL; | |
155 | ||
156 | static void timer_longterm_init(void); | |
157 | static void timer_longterm_callout( | |
158 | timer_call_param_t p0, | |
159 | timer_call_param_t p1); | |
160 | extern void timer_longterm_scan( | |
161 | timer_longterm_t *tlp, | |
162 | uint64_t now); | |
163 | static void timer_longterm_update( | |
164 | timer_longterm_t *tlp); | |
165 | static void timer_longterm_update_locked( | |
166 | timer_longterm_t *tlp); | |
167 | static mpqueue_head_t * timer_longterm_enqueue_unlocked( | |
168 | timer_call_t call, | |
169 | uint64_t now, | |
170 | uint64_t deadline, | |
171 | mpqueue_head_t ** old_queue, | |
172 | uint64_t soft_deadline, | |
173 | uint64_t ttd, | |
174 | timer_call_param_t param1, | |
175 | uint32_t callout_flags); | |
176 | static void timer_longterm_dequeued_locked( | |
177 | timer_call_t call); | |
178 | ||
179 | uint64_t past_deadline_timers; | |
180 | uint64_t past_deadline_deltas; | |
181 | uint64_t past_deadline_longest; | |
182 | uint64_t past_deadline_shortest = ~0ULL; | |
183 | enum {PAST_DEADLINE_TIMER_ADJUSTMENT_NS = 10 * 1000}; | |
184 | ||
185 | uint64_t past_deadline_timer_adjustment; | |
186 | ||
187 | static boolean_t timer_call_enter_internal(timer_call_t call, timer_call_param_t param1, uint64_t deadline, uint64_t leeway, uint32_t flags, boolean_t ratelimited); | |
188 | boolean_t mach_timer_coalescing_enabled = TRUE; | |
189 | ||
190 | mpqueue_head_t *timer_call_enqueue_deadline_unlocked( | |
191 | timer_call_t call, | |
192 | mpqueue_head_t *queue, | |
193 | uint64_t deadline, | |
194 | uint64_t soft_deadline, | |
195 | uint64_t ttd, | |
196 | timer_call_param_t param1, | |
197 | uint32_t flags); | |
198 | ||
199 | mpqueue_head_t *timer_call_dequeue_unlocked( | |
200 | timer_call_t call); | |
201 | ||
202 | timer_coalescing_priority_params_t tcoal_prio_params; | |
203 | ||
204 | #if TCOAL_PRIO_STATS | |
205 | int32_t nc_tcl, rt_tcl, bg_tcl, kt_tcl, fp_tcl, ts_tcl, qos_tcl; | |
206 | #define TCOAL_PRIO_STAT(x) (x++) | |
207 | #else | |
208 | #define TCOAL_PRIO_STAT(x) | |
209 | #endif | |
210 | ||
211 | static void | |
212 | timer_call_init_abstime(void) | |
213 | { | |
214 | int i; | |
215 | uint64_t result; | |
216 | timer_coalescing_priority_params_ns_t * tcoal_prio_params_init = timer_call_get_priority_params(); | |
217 | nanoseconds_to_absolutetime(PAST_DEADLINE_TIMER_ADJUSTMENT_NS, &past_deadline_timer_adjustment); | |
218 | nanoseconds_to_absolutetime(tcoal_prio_params_init->idle_entry_timer_processing_hdeadline_threshold_ns, &result); | |
219 | tcoal_prio_params.idle_entry_timer_processing_hdeadline_threshold_abstime = (uint32_t)result; | |
220 | nanoseconds_to_absolutetime(tcoal_prio_params_init->interrupt_timer_coalescing_ilat_threshold_ns, &result); | |
221 | tcoal_prio_params.interrupt_timer_coalescing_ilat_threshold_abstime = (uint32_t)result; | |
222 | nanoseconds_to_absolutetime(tcoal_prio_params_init->timer_resort_threshold_ns, &result); | |
223 | tcoal_prio_params.timer_resort_threshold_abstime = (uint32_t)result; | |
224 | tcoal_prio_params.timer_coalesce_rt_shift = tcoal_prio_params_init->timer_coalesce_rt_shift; | |
225 | tcoal_prio_params.timer_coalesce_bg_shift = tcoal_prio_params_init->timer_coalesce_bg_shift; | |
226 | tcoal_prio_params.timer_coalesce_kt_shift = tcoal_prio_params_init->timer_coalesce_kt_shift; | |
227 | tcoal_prio_params.timer_coalesce_fp_shift = tcoal_prio_params_init->timer_coalesce_fp_shift; | |
228 | tcoal_prio_params.timer_coalesce_ts_shift = tcoal_prio_params_init->timer_coalesce_ts_shift; | |
229 | ||
230 | nanoseconds_to_absolutetime(tcoal_prio_params_init->timer_coalesce_rt_ns_max, | |
231 | &tcoal_prio_params.timer_coalesce_rt_abstime_max); | |
232 | nanoseconds_to_absolutetime(tcoal_prio_params_init->timer_coalesce_bg_ns_max, | |
233 | &tcoal_prio_params.timer_coalesce_bg_abstime_max); | |
234 | nanoseconds_to_absolutetime(tcoal_prio_params_init->timer_coalesce_kt_ns_max, | |
235 | &tcoal_prio_params.timer_coalesce_kt_abstime_max); | |
236 | nanoseconds_to_absolutetime(tcoal_prio_params_init->timer_coalesce_fp_ns_max, | |
237 | &tcoal_prio_params.timer_coalesce_fp_abstime_max); | |
238 | nanoseconds_to_absolutetime(tcoal_prio_params_init->timer_coalesce_ts_ns_max, | |
239 | &tcoal_prio_params.timer_coalesce_ts_abstime_max); | |
240 | ||
241 | for (i = 0; i < NUM_LATENCY_QOS_TIERS; i++) { | |
242 | tcoal_prio_params.latency_qos_scale[i] = tcoal_prio_params_init->latency_qos_scale[i]; | |
243 | nanoseconds_to_absolutetime(tcoal_prio_params_init->latency_qos_ns_max[i], | |
244 | &tcoal_prio_params.latency_qos_abstime_max[i]); | |
245 | tcoal_prio_params.latency_tier_rate_limited[i] = tcoal_prio_params_init->latency_tier_rate_limited[i]; | |
246 | } | |
247 | } | |
248 | ||
249 | ||
250 | void | |
251 | timer_call_init(void) | |
252 | { | |
253 | lck_attr_setdefault(&timer_call_lck_attr); | |
254 | lck_grp_attr_setdefault(&timer_call_lck_grp_attr); | |
255 | lck_grp_init(&timer_call_lck_grp, "timer_call", &timer_call_lck_grp_attr); | |
256 | ||
257 | timer_longterm_init(); | |
258 | timer_call_init_abstime(); | |
259 | } | |
260 | ||
261 | ||
262 | void | |
263 | timer_call_queue_init(mpqueue_head_t *queue) | |
264 | { | |
265 | DBG("timer_call_queue_init(%p)\n", queue); | |
266 | mpqueue_init(queue, &timer_call_lck_grp, &timer_call_lck_attr); | |
267 | } | |
268 | ||
269 | ||
270 | void | |
271 | timer_call_setup( | |
272 | timer_call_t call, | |
273 | timer_call_func_t func, | |
274 | timer_call_param_t param0) | |
275 | { | |
276 | DBG("timer_call_setup(%p,%p,%p)\n", call, func, param0); | |
277 | call_entry_setup(TCE(call), func, param0); | |
278 | simple_lock_init(&(call)->lock, 0); | |
279 | call->async_dequeue = FALSE; | |
280 | } | |
281 | #if TIMER_ASSERT | |
282 | static __inline__ mpqueue_head_t * | |
283 | timer_call_entry_dequeue( | |
284 | timer_call_t entry) | |
285 | { | |
286 | mpqueue_head_t *old_queue = MPQUEUE(TCE(entry)->queue); | |
287 | ||
288 | if (!hw_lock_held((hw_lock_t)&entry->lock)) { | |
289 | panic("_call_entry_dequeue() " | |
290 | "entry %p is not locked\n", entry); | |
291 | } | |
292 | /* | |
293 | * XXX The queue lock is actually a mutex in spin mode | |
294 | * but there's no way to test for it being held | |
295 | * so we pretend it's a spinlock! | |
296 | */ | |
297 | if (!hw_lock_held((hw_lock_t)&old_queue->lock_data)) { | |
298 | panic("_call_entry_dequeue() " | |
299 | "queue %p is not locked\n", old_queue); | |
300 | } | |
301 | ||
302 | call_entry_dequeue(TCE(entry)); | |
303 | old_queue->count--; | |
304 | ||
305 | return old_queue; | |
306 | } | |
307 | ||
308 | static __inline__ mpqueue_head_t * | |
309 | timer_call_entry_enqueue_deadline( | |
310 | timer_call_t entry, | |
311 | mpqueue_head_t *queue, | |
312 | uint64_t deadline) | |
313 | { | |
314 | mpqueue_head_t *old_queue = MPQUEUE(TCE(entry)->queue); | |
315 | ||
316 | if (!hw_lock_held((hw_lock_t)&entry->lock)) { | |
317 | panic("_call_entry_enqueue_deadline() " | |
318 | "entry %p is not locked\n", entry); | |
319 | } | |
320 | /* XXX More lock pretense: */ | |
321 | if (!hw_lock_held((hw_lock_t)&queue->lock_data)) { | |
322 | panic("_call_entry_enqueue_deadline() " | |
323 | "queue %p is not locked\n", queue); | |
324 | } | |
325 | if (old_queue != NULL && old_queue != queue) { | |
326 | panic("_call_entry_enqueue_deadline() " | |
327 | "old_queue %p != queue", old_queue); | |
328 | } | |
329 | ||
330 | call_entry_enqueue_deadline(TCE(entry), QUEUE(queue), deadline); | |
331 | ||
332 | /* For efficiency, track the earliest soft deadline on the queue, so that | |
333 | * fuzzy decisions can be made without lock acquisitions. | |
334 | */ | |
335 | timer_call_t thead = (timer_call_t)queue_first(&queue->head); | |
336 | ||
337 | queue->earliest_soft_deadline = thead->flags & TIMER_CALL_RATELIMITED ? TCE(thead)->deadline : thead->soft_deadline; | |
338 | ||
339 | if (old_queue) { | |
340 | old_queue->count--; | |
341 | } | |
342 | queue->count++; | |
343 | ||
344 | return old_queue; | |
345 | } | |
346 | ||
347 | #else | |
348 | ||
349 | static __inline__ mpqueue_head_t * | |
350 | timer_call_entry_dequeue( | |
351 | timer_call_t entry) | |
352 | { | |
353 | mpqueue_head_t *old_queue = MPQUEUE(TCE(entry)->queue); | |
354 | ||
355 | call_entry_dequeue(TCE(entry)); | |
356 | old_queue->count--; | |
357 | ||
358 | return old_queue; | |
359 | } | |
360 | ||
361 | static __inline__ mpqueue_head_t * | |
362 | timer_call_entry_enqueue_deadline( | |
363 | timer_call_t entry, | |
364 | mpqueue_head_t *queue, | |
365 | uint64_t deadline) | |
366 | { | |
367 | mpqueue_head_t *old_queue = MPQUEUE(TCE(entry)->queue); | |
368 | ||
369 | call_entry_enqueue_deadline(TCE(entry), QUEUE(queue), deadline); | |
370 | ||
371 | /* For efficiency, track the earliest soft deadline on the queue, | |
372 | * so that fuzzy decisions can be made without lock acquisitions. | |
373 | */ | |
374 | ||
375 | timer_call_t thead = (timer_call_t)queue_first(&queue->head); | |
376 | queue->earliest_soft_deadline = thead->flags & TIMER_CALL_RATELIMITED ? TCE(thead)->deadline : thead->soft_deadline; | |
377 | ||
378 | if (old_queue) { | |
379 | old_queue->count--; | |
380 | } | |
381 | queue->count++; | |
382 | ||
383 | return old_queue; | |
384 | } | |
385 | ||
386 | #endif | |
387 | ||
388 | static __inline__ void | |
389 | timer_call_entry_enqueue_tail( | |
390 | timer_call_t entry, | |
391 | mpqueue_head_t *queue) | |
392 | { | |
393 | call_entry_enqueue_tail(TCE(entry), QUEUE(queue)); | |
394 | queue->count++; | |
395 | return; | |
396 | } | |
397 | ||
398 | /* | |
399 | * Remove timer entry from its queue but don't change the queue pointer | |
400 | * and set the async_dequeue flag. This is locking case 2b. | |
401 | */ | |
402 | static __inline__ void | |
403 | timer_call_entry_dequeue_async( | |
404 | timer_call_t entry) | |
405 | { | |
406 | mpqueue_head_t *old_queue = MPQUEUE(TCE(entry)->queue); | |
407 | if (old_queue) { | |
408 | old_queue->count--; | |
409 | (void) remque(qe(entry)); | |
410 | entry->async_dequeue = TRUE; | |
411 | } | |
412 | return; | |
413 | } | |
414 | ||
415 | #if TIMER_ASSERT | |
416 | unsigned timer_call_enqueue_deadline_unlocked_async1; | |
417 | unsigned timer_call_enqueue_deadline_unlocked_async2; | |
418 | #endif | |
419 | /* | |
420 | * Assumes call_entry and queues unlocked, interrupts disabled. | |
421 | */ | |
422 | __inline__ mpqueue_head_t * | |
423 | timer_call_enqueue_deadline_unlocked( | |
424 | timer_call_t call, | |
425 | mpqueue_head_t *queue, | |
426 | uint64_t deadline, | |
427 | uint64_t soft_deadline, | |
428 | uint64_t ttd, | |
429 | timer_call_param_t param1, | |
430 | uint32_t callout_flags) | |
431 | { | |
432 | call_entry_t entry = TCE(call); | |
433 | mpqueue_head_t *old_queue; | |
434 | ||
435 | DBG("timer_call_enqueue_deadline_unlocked(%p,%p,)\n", call, queue); | |
436 | ||
437 | simple_lock(&call->lock, LCK_GRP_NULL); | |
438 | ||
439 | old_queue = MPQUEUE(entry->queue); | |
440 | ||
441 | if (old_queue != NULL) { | |
442 | timer_queue_lock_spin(old_queue); | |
443 | if (call->async_dequeue) { | |
444 | /* collision (1c): timer already dequeued, clear flag */ | |
445 | #if TIMER_ASSERT | |
446 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
447 | DECR_TIMER_ASYNC_DEQ | DBG_FUNC_NONE, | |
448 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
449 | call->async_dequeue, | |
450 | VM_KERNEL_UNSLIDE_OR_PERM(TCE(call)->queue), | |
451 | 0x1c, 0); | |
452 | timer_call_enqueue_deadline_unlocked_async1++; | |
453 | #endif | |
454 | call->async_dequeue = FALSE; | |
455 | entry->queue = NULL; | |
456 | } else if (old_queue != queue) { | |
457 | timer_call_entry_dequeue(call); | |
458 | #if TIMER_ASSERT | |
459 | timer_call_enqueue_deadline_unlocked_async2++; | |
460 | #endif | |
461 | } | |
462 | if (old_queue == timer_longterm_queue) { | |
463 | timer_longterm_dequeued_locked(call); | |
464 | } | |
465 | if (old_queue != queue) { | |
466 | timer_queue_unlock(old_queue); | |
467 | timer_queue_lock_spin(queue); | |
468 | } | |
469 | } else { | |
470 | timer_queue_lock_spin(queue); | |
471 | } | |
472 | ||
473 | call->soft_deadline = soft_deadline; | |
474 | call->flags = callout_flags; | |
475 | TCE(call)->param1 = param1; | |
476 | call->ttd = ttd; | |
477 | ||
478 | timer_call_entry_enqueue_deadline(call, queue, deadline); | |
479 | timer_queue_unlock(queue); | |
480 | simple_unlock(&call->lock); | |
481 | ||
482 | return old_queue; | |
483 | } | |
484 | ||
485 | #if TIMER_ASSERT | |
486 | unsigned timer_call_dequeue_unlocked_async1; | |
487 | unsigned timer_call_dequeue_unlocked_async2; | |
488 | #endif | |
489 | mpqueue_head_t * | |
490 | timer_call_dequeue_unlocked( | |
491 | timer_call_t call) | |
492 | { | |
493 | call_entry_t entry = TCE(call); | |
494 | mpqueue_head_t *old_queue; | |
495 | ||
496 | DBG("timer_call_dequeue_unlocked(%p)\n", call); | |
497 | ||
498 | simple_lock(&call->lock, LCK_GRP_NULL); | |
499 | old_queue = MPQUEUE(entry->queue); | |
500 | #if TIMER_ASSERT | |
501 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
502 | DECR_TIMER_ASYNC_DEQ | DBG_FUNC_NONE, | |
503 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
504 | call->async_dequeue, | |
505 | VM_KERNEL_UNSLIDE_OR_PERM(TCE(call)->queue), | |
506 | 0, 0); | |
507 | #endif | |
508 | if (old_queue != NULL) { | |
509 | timer_queue_lock_spin(old_queue); | |
510 | if (call->async_dequeue) { | |
511 | /* collision (1c): timer already dequeued, clear flag */ | |
512 | #if TIMER_ASSERT | |
513 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
514 | DECR_TIMER_ASYNC_DEQ | DBG_FUNC_NONE, | |
515 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
516 | call->async_dequeue, | |
517 | VM_KERNEL_UNSLIDE_OR_PERM(TCE(call)->queue), | |
518 | 0x1c, 0); | |
519 | timer_call_dequeue_unlocked_async1++; | |
520 | #endif | |
521 | call->async_dequeue = FALSE; | |
522 | entry->queue = NULL; | |
523 | } else { | |
524 | timer_call_entry_dequeue(call); | |
525 | } | |
526 | if (old_queue == timer_longterm_queue) { | |
527 | timer_longterm_dequeued_locked(call); | |
528 | } | |
529 | timer_queue_unlock(old_queue); | |
530 | } | |
531 | simple_unlock(&call->lock); | |
532 | return old_queue; | |
533 | } | |
534 | ||
535 | static uint64_t | |
536 | past_deadline_timer_handle(uint64_t deadline, uint64_t ctime) | |
537 | { | |
538 | uint64_t delta = (ctime - deadline); | |
539 | ||
540 | past_deadline_timers++; | |
541 | past_deadline_deltas += delta; | |
542 | if (delta > past_deadline_longest) { | |
543 | past_deadline_longest = deadline; | |
544 | } | |
545 | if (delta < past_deadline_shortest) { | |
546 | past_deadline_shortest = delta; | |
547 | } | |
548 | ||
549 | return ctime + past_deadline_timer_adjustment; | |
550 | } | |
551 | ||
552 | /* | |
553 | * Timer call entry locking model | |
554 | * ============================== | |
555 | * | |
556 | * Timer call entries are linked on per-cpu timer queues which are protected | |
557 | * by the queue lock and the call entry lock. The locking protocol is: | |
558 | * | |
559 | * 0) The canonical locking order is timer call entry followed by queue. | |
560 | * | |
561 | * 1) With only the entry lock held, entry.queue is valid: | |
562 | * 1a) NULL: the entry is not queued, or | |
563 | * 1b) non-NULL: this queue must be locked before the entry is modified. | |
564 | * After locking the queue, the call.async_dequeue flag must be checked: | |
565 | * 1c) TRUE: the entry was removed from the queue by another thread | |
566 | * and we must NULL the entry.queue and reset this flag, or | |
567 | * 1d) FALSE: (ie. queued), the entry can be manipulated. | |
568 | * | |
569 | * 2) If a queue lock is obtained first, the queue is stable: | |
570 | * 2a) If a try-lock of a queued entry succeeds, the call can be operated on | |
571 | * and dequeued. | |
572 | * 2b) If a try-lock fails, it indicates that another thread is attempting | |
573 | * to change the entry and move it to a different position in this queue | |
574 | * or to different queue. The entry can be dequeued but it should not be | |
575 | * operated upon since it is being changed. Furthermore, we don't null | |
576 | * the entry.queue pointer (protected by the entry lock we don't own). | |
577 | * Instead, we set the async_dequeue flag -- see (1c). | |
578 | * 2c) Same as 2b but occurring when a longterm timer is matured. | |
579 | * 3) A callout's parameters (deadline, flags, parameters, soft deadline &c.) | |
580 | * should be manipulated with the appropriate timer queue lock held, | |
581 | * to prevent queue traversal observations from observing inconsistent | |
582 | * updates to an in-flight callout. | |
583 | */ | |
584 | ||
585 | /* | |
586 | * Inlines timer_call_entry_dequeue() and timer_call_entry_enqueue_deadline() | |
587 | * cast between pointer types (mpqueue_head_t *) and (queue_t) so that | |
588 | * we can use the call_entry_dequeue() and call_entry_enqueue_deadline() | |
589 | * methods to operate on timer_call structs as if they are call_entry structs. | |
590 | * These structures are identical except for their queue head pointer fields. | |
591 | * | |
592 | * In the debug case, we assert that the timer call locking protocol | |
593 | * is being obeyed. | |
594 | */ | |
595 | ||
596 | static boolean_t | |
597 | timer_call_enter_internal( | |
598 | timer_call_t call, | |
599 | timer_call_param_t param1, | |
600 | uint64_t deadline, | |
601 | uint64_t leeway, | |
602 | uint32_t flags, | |
603 | boolean_t ratelimited) | |
604 | { | |
605 | mpqueue_head_t *queue = NULL; | |
606 | mpqueue_head_t *old_queue; | |
607 | spl_t s; | |
608 | uint64_t slop; | |
609 | uint32_t urgency; | |
610 | uint64_t sdeadline, ttd; | |
611 | ||
612 | assert(call->call_entry.func != NULL); | |
613 | s = splclock(); | |
614 | ||
615 | sdeadline = deadline; | |
616 | uint64_t ctime = mach_absolute_time(); | |
617 | ||
618 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
619 | DECR_TIMER_ENTER | DBG_FUNC_START, | |
620 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
621 | VM_KERNEL_ADDRHIDE(param1), deadline, flags, 0); | |
622 | ||
623 | urgency = (flags & TIMER_CALL_URGENCY_MASK); | |
624 | ||
625 | boolean_t slop_ratelimited = FALSE; | |
626 | slop = timer_call_slop(deadline, ctime, urgency, current_thread(), &slop_ratelimited); | |
627 | ||
628 | if ((flags & TIMER_CALL_LEEWAY) != 0 && leeway > slop) { | |
629 | slop = leeway; | |
630 | } | |
631 | ||
632 | if (UINT64_MAX - deadline <= slop) { | |
633 | deadline = UINT64_MAX; | |
634 | } else { | |
635 | deadline += slop; | |
636 | } | |
637 | ||
638 | if (__improbable(deadline < ctime)) { | |
639 | deadline = past_deadline_timer_handle(deadline, ctime); | |
640 | sdeadline = deadline; | |
641 | } | |
642 | ||
643 | if (ratelimited || slop_ratelimited) { | |
644 | flags |= TIMER_CALL_RATELIMITED; | |
645 | } else { | |
646 | flags &= ~TIMER_CALL_RATELIMITED; | |
647 | } | |
648 | ||
649 | ttd = sdeadline - ctime; | |
650 | #if CONFIG_DTRACE | |
651 | DTRACE_TMR7(callout__create, timer_call_func_t, TCE(call)->func, | |
652 | timer_call_param_t, TCE(call)->param0, uint32_t, flags, | |
653 | (deadline - sdeadline), | |
654 | (ttd >> 32), (unsigned) (ttd & 0xFFFFFFFF), call); | |
655 | #endif | |
656 | ||
657 | /* Program timer callout parameters under the appropriate per-CPU or | |
658 | * longterm queue lock. The callout may have been previously enqueued | |
659 | * and in-flight on this or another timer queue. | |
660 | */ | |
661 | if (!ratelimited && !slop_ratelimited) { | |
662 | queue = timer_longterm_enqueue_unlocked(call, ctime, deadline, &old_queue, sdeadline, ttd, param1, flags); | |
663 | } | |
664 | ||
665 | if (queue == NULL) { | |
666 | queue = timer_queue_assign(deadline); | |
667 | old_queue = timer_call_enqueue_deadline_unlocked(call, queue, deadline, sdeadline, ttd, param1, flags); | |
668 | } | |
669 | ||
670 | #if TIMER_TRACE | |
671 | TCE(call)->entry_time = ctime; | |
672 | #endif | |
673 | ||
674 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
675 | DECR_TIMER_ENTER | DBG_FUNC_END, | |
676 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
677 | (old_queue != NULL), deadline, queue->count, 0); | |
678 | ||
679 | splx(s); | |
680 | ||
681 | return old_queue != NULL; | |
682 | } | |
683 | ||
684 | /* | |
685 | * timer_call_*() | |
686 | * return boolean indicating whether the call was previously queued. | |
687 | */ | |
688 | boolean_t | |
689 | timer_call_enter( | |
690 | timer_call_t call, | |
691 | uint64_t deadline, | |
692 | uint32_t flags) | |
693 | { | |
694 | return timer_call_enter_internal(call, NULL, deadline, 0, flags, FALSE); | |
695 | } | |
696 | ||
697 | boolean_t | |
698 | timer_call_enter1( | |
699 | timer_call_t call, | |
700 | timer_call_param_t param1, | |
701 | uint64_t deadline, | |
702 | uint32_t flags) | |
703 | { | |
704 | return timer_call_enter_internal(call, param1, deadline, 0, flags, FALSE); | |
705 | } | |
706 | ||
707 | boolean_t | |
708 | timer_call_enter_with_leeway( | |
709 | timer_call_t call, | |
710 | timer_call_param_t param1, | |
711 | uint64_t deadline, | |
712 | uint64_t leeway, | |
713 | uint32_t flags, | |
714 | boolean_t ratelimited) | |
715 | { | |
716 | return timer_call_enter_internal(call, param1, deadline, leeway, flags, ratelimited); | |
717 | } | |
718 | ||
719 | boolean_t | |
720 | timer_call_quantum_timer_enter( | |
721 | timer_call_t call, | |
722 | timer_call_param_t param1, | |
723 | uint64_t deadline, | |
724 | uint64_t ctime) | |
725 | { | |
726 | assert(call->call_entry.func != NULL); | |
727 | assert(ml_get_interrupts_enabled() == FALSE); | |
728 | ||
729 | uint32_t flags = TIMER_CALL_SYS_CRITICAL | TIMER_CALL_LOCAL; | |
730 | ||
731 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, DECR_TIMER_ENTER | DBG_FUNC_START, | |
732 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
733 | VM_KERNEL_ADDRHIDE(param1), deadline, | |
734 | flags, 0); | |
735 | ||
736 | if (__improbable(deadline < ctime)) { | |
737 | deadline = past_deadline_timer_handle(deadline, ctime); | |
738 | } | |
739 | ||
740 | uint64_t ttd = deadline - ctime; | |
741 | #if CONFIG_DTRACE | |
742 | DTRACE_TMR7(callout__create, timer_call_func_t, TCE(call)->func, | |
743 | timer_call_param_t, TCE(call)->param0, uint32_t, flags, 0, | |
744 | (ttd >> 32), (unsigned) (ttd & 0xFFFFFFFF), call); | |
745 | #endif | |
746 | ||
747 | quantum_timer_set_deadline(deadline); | |
748 | TCE(call)->deadline = deadline; | |
749 | TCE(call)->param1 = param1; | |
750 | call->ttd = ttd; | |
751 | call->flags = flags; | |
752 | ||
753 | #if TIMER_TRACE | |
754 | TCE(call)->entry_time = ctime; | |
755 | #endif | |
756 | ||
757 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, DECR_TIMER_ENTER | DBG_FUNC_END, | |
758 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
759 | 1, deadline, 0, 0); | |
760 | ||
761 | return true; | |
762 | } | |
763 | ||
764 | ||
765 | boolean_t | |
766 | timer_call_quantum_timer_cancel( | |
767 | timer_call_t call) | |
768 | { | |
769 | assert(ml_get_interrupts_enabled() == FALSE); | |
770 | ||
771 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
772 | DECR_TIMER_CANCEL | DBG_FUNC_START, | |
773 | VM_KERNEL_UNSLIDE_OR_PERM(call), TCE(call)->deadline, | |
774 | 0, call->flags, 0); | |
775 | ||
776 | TCE(call)->deadline = 0; | |
777 | quantum_timer_set_deadline(0); | |
778 | ||
779 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
780 | DECR_TIMER_CANCEL | DBG_FUNC_END, | |
781 | VM_KERNEL_UNSLIDE_OR_PERM(call), 0, | |
782 | TCE(call)->deadline - mach_absolute_time(), | |
783 | TCE(call)->deadline - TCE(call)->entry_time, 0); | |
784 | ||
785 | #if CONFIG_DTRACE | |
786 | DTRACE_TMR6(callout__cancel, timer_call_func_t, TCE(call)->func, | |
787 | timer_call_param_t, TCE(call)->param0, uint32_t, call->flags, 0, | |
788 | (call->ttd >> 32), (unsigned) (call->ttd & 0xFFFFFFFF)); | |
789 | #endif | |
790 | ||
791 | return true; | |
792 | } | |
793 | ||
794 | boolean_t | |
795 | timer_call_cancel( | |
796 | timer_call_t call) | |
797 | { | |
798 | mpqueue_head_t *old_queue; | |
799 | spl_t s; | |
800 | ||
801 | s = splclock(); | |
802 | ||
803 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
804 | DECR_TIMER_CANCEL | DBG_FUNC_START, | |
805 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
806 | TCE(call)->deadline, call->soft_deadline, call->flags, 0); | |
807 | ||
808 | old_queue = timer_call_dequeue_unlocked(call); | |
809 | ||
810 | if (old_queue != NULL) { | |
811 | timer_queue_lock_spin(old_queue); | |
812 | if (!queue_empty(&old_queue->head)) { | |
813 | timer_queue_cancel(old_queue, TCE(call)->deadline, CE(queue_first(&old_queue->head))->deadline); | |
814 | timer_call_t thead = (timer_call_t)queue_first(&old_queue->head); | |
815 | old_queue->earliest_soft_deadline = thead->flags & TIMER_CALL_RATELIMITED ? TCE(thead)->deadline : thead->soft_deadline; | |
816 | } else { | |
817 | timer_queue_cancel(old_queue, TCE(call)->deadline, UINT64_MAX); | |
818 | old_queue->earliest_soft_deadline = UINT64_MAX; | |
819 | } | |
820 | timer_queue_unlock(old_queue); | |
821 | } | |
822 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
823 | DECR_TIMER_CANCEL | DBG_FUNC_END, | |
824 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
825 | VM_KERNEL_UNSLIDE_OR_PERM(old_queue), | |
826 | TCE(call)->deadline - mach_absolute_time(), | |
827 | TCE(call)->deadline - TCE(call)->entry_time, 0); | |
828 | splx(s); | |
829 | ||
830 | #if CONFIG_DTRACE | |
831 | DTRACE_TMR6(callout__cancel, timer_call_func_t, TCE(call)->func, | |
832 | timer_call_param_t, TCE(call)->param0, uint32_t, call->flags, 0, | |
833 | (call->ttd >> 32), (unsigned) (call->ttd & 0xFFFFFFFF)); | |
834 | #endif | |
835 | ||
836 | return old_queue != NULL; | |
837 | } | |
838 | ||
839 | static uint32_t timer_queue_shutdown_lock_skips; | |
840 | static uint32_t timer_queue_shutdown_discarded; | |
841 | ||
842 | void | |
843 | timer_queue_shutdown( | |
844 | mpqueue_head_t *queue) | |
845 | { | |
846 | timer_call_t call; | |
847 | mpqueue_head_t *new_queue; | |
848 | spl_t s; | |
849 | ||
850 | ||
851 | DBG("timer_queue_shutdown(%p)\n", queue); | |
852 | ||
853 | s = splclock(); | |
854 | ||
855 | /* Note comma operator in while expression re-locking each iteration */ | |
856 | while ((void)timer_queue_lock_spin(queue), !queue_empty(&queue->head)) { | |
857 | call = TIMER_CALL(queue_first(&queue->head)); | |
858 | ||
859 | if (!simple_lock_try(&call->lock, LCK_GRP_NULL)) { | |
860 | /* | |
861 | * case (2b) lock order inversion, dequeue and skip | |
862 | * Don't change the call_entry queue back-pointer | |
863 | * but set the async_dequeue field. | |
864 | */ | |
865 | timer_queue_shutdown_lock_skips++; | |
866 | timer_call_entry_dequeue_async(call); | |
867 | #if TIMER_ASSERT | |
868 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
869 | DECR_TIMER_ASYNC_DEQ | DBG_FUNC_NONE, | |
870 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
871 | call->async_dequeue, | |
872 | VM_KERNEL_UNSLIDE_OR_PERM(TCE(call)->queue), | |
873 | 0x2b, 0); | |
874 | #endif | |
875 | timer_queue_unlock(queue); | |
876 | continue; | |
877 | } | |
878 | ||
879 | boolean_t call_local = ((call->flags & TIMER_CALL_LOCAL) != 0); | |
880 | ||
881 | /* remove entry from old queue */ | |
882 | timer_call_entry_dequeue(call); | |
883 | timer_queue_unlock(queue); | |
884 | ||
885 | if (call_local == FALSE) { | |
886 | /* and queue it on new, discarding LOCAL timers */ | |
887 | new_queue = timer_queue_assign(TCE(call)->deadline); | |
888 | timer_queue_lock_spin(new_queue); | |
889 | timer_call_entry_enqueue_deadline( | |
890 | call, new_queue, TCE(call)->deadline); | |
891 | timer_queue_unlock(new_queue); | |
892 | } else { | |
893 | timer_queue_shutdown_discarded++; | |
894 | } | |
895 | ||
896 | assert(call_local == FALSE); | |
897 | simple_unlock(&call->lock); | |
898 | } | |
899 | ||
900 | timer_queue_unlock(queue); | |
901 | splx(s); | |
902 | } | |
903 | ||
904 | ||
905 | void | |
906 | quantum_timer_expire( | |
907 | uint64_t deadline) | |
908 | { | |
909 | processor_t processor = current_processor(); | |
910 | timer_call_t call = TIMER_CALL(&(processor->quantum_timer)); | |
911 | ||
912 | if (__improbable(TCE(call)->deadline > deadline)) { | |
913 | panic("CPU quantum timer deadlin out of sync with timer call deadline"); | |
914 | } | |
915 | ||
916 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
917 | DECR_TIMER_EXPIRE | DBG_FUNC_NONE, | |
918 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
919 | TCE(call)->deadline, | |
920 | TCE(call)->deadline, | |
921 | TCE(call)->entry_time, 0); | |
922 | ||
923 | timer_call_func_t func = TCE(call)->func; | |
924 | timer_call_param_t param0 = TCE(call)->param0; | |
925 | timer_call_param_t param1 = TCE(call)->param1; | |
926 | ||
927 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
928 | DECR_TIMER_CALLOUT | DBG_FUNC_START, | |
929 | VM_KERNEL_UNSLIDE_OR_PERM(call), VM_KERNEL_UNSLIDE(func), | |
930 | VM_KERNEL_ADDRHIDE(param0), | |
931 | VM_KERNEL_ADDRHIDE(param1), | |
932 | 0); | |
933 | ||
934 | #if CONFIG_DTRACE | |
935 | DTRACE_TMR7(callout__start, timer_call_func_t, func, | |
936 | timer_call_param_t, param0, unsigned, call->flags, | |
937 | 0, (call->ttd >> 32), | |
938 | (unsigned) (call->ttd & 0xFFFFFFFF), call); | |
939 | #endif | |
940 | (*func)(param0, param1); | |
941 | ||
942 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
943 | DECR_TIMER_CALLOUT | DBG_FUNC_END, | |
944 | VM_KERNEL_UNSLIDE_OR_PERM(call), VM_KERNEL_UNSLIDE(func), | |
945 | VM_KERNEL_ADDRHIDE(param0), | |
946 | VM_KERNEL_ADDRHIDE(param1), | |
947 | 0); | |
948 | } | |
949 | ||
950 | static uint32_t timer_queue_expire_lock_skips; | |
951 | uint64_t | |
952 | timer_queue_expire_with_options( | |
953 | mpqueue_head_t *queue, | |
954 | uint64_t deadline, | |
955 | boolean_t rescan) | |
956 | { | |
957 | timer_call_t call = NULL; | |
958 | uint32_t tc_iterations = 0; | |
959 | DBG("timer_queue_expire(%p,)\n", queue); | |
960 | ||
961 | uint64_t cur_deadline = deadline; | |
962 | timer_queue_lock_spin(queue); | |
963 | ||
964 | while (!queue_empty(&queue->head)) { | |
965 | /* Upon processing one or more timer calls, refresh the | |
966 | * deadline to account for time elapsed in the callout | |
967 | */ | |
968 | if (++tc_iterations > 1) { | |
969 | cur_deadline = mach_absolute_time(); | |
970 | } | |
971 | ||
972 | if (call == NULL) { | |
973 | call = TIMER_CALL(queue_first(&queue->head)); | |
974 | } | |
975 | ||
976 | if (call->soft_deadline <= cur_deadline) { | |
977 | timer_call_func_t func; | |
978 | timer_call_param_t param0, param1; | |
979 | ||
980 | TCOAL_DEBUG(0xDDDD0000, queue->earliest_soft_deadline, call->soft_deadline, 0, 0, 0); | |
981 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
982 | DECR_TIMER_EXPIRE | DBG_FUNC_NONE, | |
983 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
984 | call->soft_deadline, | |
985 | TCE(call)->deadline, | |
986 | TCE(call)->entry_time, 0); | |
987 | ||
988 | if ((call->flags & TIMER_CALL_RATELIMITED) && | |
989 | (TCE(call)->deadline > cur_deadline)) { | |
990 | if (rescan == FALSE) { | |
991 | break; | |
992 | } | |
993 | } | |
994 | ||
995 | if (!simple_lock_try(&call->lock, LCK_GRP_NULL)) { | |
996 | /* case (2b) lock inversion, dequeue and skip */ | |
997 | timer_queue_expire_lock_skips++; | |
998 | timer_call_entry_dequeue_async(call); | |
999 | call = NULL; | |
1000 | continue; | |
1001 | } | |
1002 | ||
1003 | timer_call_entry_dequeue(call); | |
1004 | ||
1005 | func = TCE(call)->func; | |
1006 | param0 = TCE(call)->param0; | |
1007 | param1 = TCE(call)->param1; | |
1008 | ||
1009 | simple_unlock(&call->lock); | |
1010 | timer_queue_unlock(queue); | |
1011 | ||
1012 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1013 | DECR_TIMER_CALLOUT | DBG_FUNC_START, | |
1014 | VM_KERNEL_UNSLIDE_OR_PERM(call), VM_KERNEL_UNSLIDE(func), | |
1015 | VM_KERNEL_ADDRHIDE(param0), | |
1016 | VM_KERNEL_ADDRHIDE(param1), | |
1017 | 0); | |
1018 | ||
1019 | #if CONFIG_DTRACE | |
1020 | DTRACE_TMR7(callout__start, timer_call_func_t, func, | |
1021 | timer_call_param_t, param0, unsigned, call->flags, | |
1022 | 0, (call->ttd >> 32), | |
1023 | (unsigned) (call->ttd & 0xFFFFFFFF), call); | |
1024 | #endif | |
1025 | /* Maintain time-to-deadline in per-processor data | |
1026 | * structure for thread wakeup deadline statistics. | |
1027 | */ | |
1028 | uint64_t *ttdp = &(PROCESSOR_DATA(current_processor(), timer_call_ttd)); | |
1029 | *ttdp = call->ttd; | |
1030 | (*func)(param0, param1); | |
1031 | *ttdp = 0; | |
1032 | #if CONFIG_DTRACE | |
1033 | DTRACE_TMR4(callout__end, timer_call_func_t, func, | |
1034 | param0, param1, call); | |
1035 | #endif | |
1036 | ||
1037 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1038 | DECR_TIMER_CALLOUT | DBG_FUNC_END, | |
1039 | VM_KERNEL_UNSLIDE_OR_PERM(call), VM_KERNEL_UNSLIDE(func), | |
1040 | VM_KERNEL_ADDRHIDE(param0), | |
1041 | VM_KERNEL_ADDRHIDE(param1), | |
1042 | 0); | |
1043 | call = NULL; | |
1044 | timer_queue_lock_spin(queue); | |
1045 | } else { | |
1046 | if (__probable(rescan == FALSE)) { | |
1047 | break; | |
1048 | } else { | |
1049 | int64_t skew = TCE(call)->deadline - call->soft_deadline; | |
1050 | assert(TCE(call)->deadline >= call->soft_deadline); | |
1051 | ||
1052 | /* DRK: On a latency quality-of-service level change, | |
1053 | * re-sort potentially rate-limited timers. The platform | |
1054 | * layer determines which timers require | |
1055 | * this. In the absence of the per-callout | |
1056 | * synchronization requirement, a global resort could | |
1057 | * be more efficient. The re-sort effectively | |
1058 | * annuls all timer adjustments, i.e. the "soft | |
1059 | * deadline" is the sort key. | |
1060 | */ | |
1061 | ||
1062 | if (timer_resort_threshold(skew)) { | |
1063 | if (__probable(simple_lock_try(&call->lock, LCK_GRP_NULL))) { | |
1064 | timer_call_entry_dequeue(call); | |
1065 | timer_call_entry_enqueue_deadline(call, queue, call->soft_deadline); | |
1066 | simple_unlock(&call->lock); | |
1067 | call = NULL; | |
1068 | } | |
1069 | } | |
1070 | if (call) { | |
1071 | call = TIMER_CALL(queue_next(qe(call))); | |
1072 | if (queue_end(&queue->head, qe(call))) { | |
1073 | break; | |
1074 | } | |
1075 | } | |
1076 | } | |
1077 | } | |
1078 | } | |
1079 | ||
1080 | if (!queue_empty(&queue->head)) { | |
1081 | call = TIMER_CALL(queue_first(&queue->head)); | |
1082 | cur_deadline = TCE(call)->deadline; | |
1083 | queue->earliest_soft_deadline = (call->flags & TIMER_CALL_RATELIMITED) ? TCE(call)->deadline: call->soft_deadline; | |
1084 | } else { | |
1085 | queue->earliest_soft_deadline = cur_deadline = UINT64_MAX; | |
1086 | } | |
1087 | ||
1088 | timer_queue_unlock(queue); | |
1089 | ||
1090 | return cur_deadline; | |
1091 | } | |
1092 | ||
1093 | uint64_t | |
1094 | timer_queue_expire( | |
1095 | mpqueue_head_t *queue, | |
1096 | uint64_t deadline) | |
1097 | { | |
1098 | return timer_queue_expire_with_options(queue, deadline, FALSE); | |
1099 | } | |
1100 | ||
1101 | extern int serverperfmode; | |
1102 | static uint32_t timer_queue_migrate_lock_skips; | |
1103 | /* | |
1104 | * timer_queue_migrate() is called by timer_queue_migrate_cpu() | |
1105 | * to move timer requests from the local processor (queue_from) | |
1106 | * to a target processor's (queue_to). | |
1107 | */ | |
1108 | int | |
1109 | timer_queue_migrate(mpqueue_head_t *queue_from, mpqueue_head_t *queue_to) | |
1110 | { | |
1111 | timer_call_t call; | |
1112 | timer_call_t head_to; | |
1113 | int timers_migrated = 0; | |
1114 | ||
1115 | DBG("timer_queue_migrate(%p,%p)\n", queue_from, queue_to); | |
1116 | ||
1117 | assert(!ml_get_interrupts_enabled()); | |
1118 | assert(queue_from != queue_to); | |
1119 | ||
1120 | if (serverperfmode) { | |
1121 | /* | |
1122 | * if we're running a high end server | |
1123 | * avoid migrations... they add latency | |
1124 | * and don't save us power under typical | |
1125 | * server workloads | |
1126 | */ | |
1127 | return -4; | |
1128 | } | |
1129 | ||
1130 | /* | |
1131 | * Take both local (from) and target (to) timer queue locks while | |
1132 | * moving the timers from the local queue to the target processor. | |
1133 | * We assume that the target is always the boot processor. | |
1134 | * But only move if all of the following is true: | |
1135 | * - the target queue is non-empty | |
1136 | * - the local queue is non-empty | |
1137 | * - the local queue's first deadline is later than the target's | |
1138 | * - the local queue contains no non-migrateable "local" call | |
1139 | * so that we need not have the target resync. | |
1140 | */ | |
1141 | ||
1142 | timer_queue_lock_spin(queue_to); | |
1143 | ||
1144 | head_to = TIMER_CALL(queue_first(&queue_to->head)); | |
1145 | if (queue_empty(&queue_to->head)) { | |
1146 | timers_migrated = -1; | |
1147 | goto abort1; | |
1148 | } | |
1149 | ||
1150 | timer_queue_lock_spin(queue_from); | |
1151 | ||
1152 | if (queue_empty(&queue_from->head)) { | |
1153 | timers_migrated = -2; | |
1154 | goto abort2; | |
1155 | } | |
1156 | ||
1157 | call = TIMER_CALL(queue_first(&queue_from->head)); | |
1158 | if (TCE(call)->deadline < TCE(head_to)->deadline) { | |
1159 | timers_migrated = 0; | |
1160 | goto abort2; | |
1161 | } | |
1162 | ||
1163 | /* perform scan for non-migratable timers */ | |
1164 | do { | |
1165 | if (call->flags & TIMER_CALL_LOCAL) { | |
1166 | timers_migrated = -3; | |
1167 | goto abort2; | |
1168 | } | |
1169 | call = TIMER_CALL(queue_next(qe(call))); | |
1170 | } while (!queue_end(&queue_from->head, qe(call))); | |
1171 | ||
1172 | /* migration loop itself -- both queues are locked */ | |
1173 | while (!queue_empty(&queue_from->head)) { | |
1174 | call = TIMER_CALL(queue_first(&queue_from->head)); | |
1175 | if (!simple_lock_try(&call->lock, LCK_GRP_NULL)) { | |
1176 | /* case (2b) lock order inversion, dequeue only */ | |
1177 | #ifdef TIMER_ASSERT | |
1178 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1179 | DECR_TIMER_ASYNC_DEQ | DBG_FUNC_NONE, | |
1180 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
1181 | VM_KERNEL_UNSLIDE_OR_PERM(TCE(call)->queue), | |
1182 | VM_KERNEL_UNSLIDE_OR_PERM(call->lock.interlock.lock_data), | |
1183 | 0x2b, 0); | |
1184 | #endif | |
1185 | timer_queue_migrate_lock_skips++; | |
1186 | timer_call_entry_dequeue_async(call); | |
1187 | continue; | |
1188 | } | |
1189 | timer_call_entry_dequeue(call); | |
1190 | timer_call_entry_enqueue_deadline( | |
1191 | call, queue_to, TCE(call)->deadline); | |
1192 | timers_migrated++; | |
1193 | simple_unlock(&call->lock); | |
1194 | } | |
1195 | queue_from->earliest_soft_deadline = UINT64_MAX; | |
1196 | abort2: | |
1197 | timer_queue_unlock(queue_from); | |
1198 | abort1: | |
1199 | timer_queue_unlock(queue_to); | |
1200 | ||
1201 | return timers_migrated; | |
1202 | } | |
1203 | ||
1204 | void | |
1205 | timer_queue_trace_cpu(int ncpu) | |
1206 | { | |
1207 | timer_call_nosync_cpu( | |
1208 | ncpu, | |
1209 | (void (*)(void *))timer_queue_trace, | |
1210 | (void*) timer_queue_cpu(ncpu)); | |
1211 | } | |
1212 | ||
1213 | void | |
1214 | timer_queue_trace( | |
1215 | mpqueue_head_t *queue) | |
1216 | { | |
1217 | timer_call_t call; | |
1218 | spl_t s; | |
1219 | ||
1220 | if (!kdebug_enable) { | |
1221 | return; | |
1222 | } | |
1223 | ||
1224 | s = splclock(); | |
1225 | timer_queue_lock_spin(queue); | |
1226 | ||
1227 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1228 | DECR_TIMER_QUEUE | DBG_FUNC_START, | |
1229 | queue->count, mach_absolute_time(), 0, 0, 0); | |
1230 | ||
1231 | if (!queue_empty(&queue->head)) { | |
1232 | call = TIMER_CALL(queue_first(&queue->head)); | |
1233 | do { | |
1234 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1235 | DECR_TIMER_QUEUE | DBG_FUNC_NONE, | |
1236 | call->soft_deadline, | |
1237 | TCE(call)->deadline, | |
1238 | TCE(call)->entry_time, | |
1239 | VM_KERNEL_UNSLIDE(TCE(call)->func), | |
1240 | 0); | |
1241 | call = TIMER_CALL(queue_next(qe(call))); | |
1242 | } while (!queue_end(&queue->head, qe(call))); | |
1243 | } | |
1244 | ||
1245 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1246 | DECR_TIMER_QUEUE | DBG_FUNC_END, | |
1247 | queue->count, mach_absolute_time(), 0, 0, 0); | |
1248 | ||
1249 | timer_queue_unlock(queue); | |
1250 | splx(s); | |
1251 | } | |
1252 | ||
1253 | void | |
1254 | timer_longterm_dequeued_locked(timer_call_t call) | |
1255 | { | |
1256 | timer_longterm_t *tlp = &timer_longterm; | |
1257 | ||
1258 | tlp->dequeues++; | |
1259 | if (call == tlp->threshold.call) { | |
1260 | tlp->threshold.call = NULL; | |
1261 | } | |
1262 | } | |
1263 | ||
1264 | /* | |
1265 | * Place a timer call in the longterm list | |
1266 | * and adjust the next timer callout deadline if the new timer is first. | |
1267 | */ | |
1268 | mpqueue_head_t * | |
1269 | timer_longterm_enqueue_unlocked(timer_call_t call, | |
1270 | uint64_t now, | |
1271 | uint64_t deadline, | |
1272 | mpqueue_head_t **old_queue, | |
1273 | uint64_t soft_deadline, | |
1274 | uint64_t ttd, | |
1275 | timer_call_param_t param1, | |
1276 | uint32_t callout_flags) | |
1277 | { | |
1278 | timer_longterm_t *tlp = &timer_longterm; | |
1279 | boolean_t update_required = FALSE; | |
1280 | uint64_t longterm_threshold; | |
1281 | ||
1282 | longterm_threshold = now + tlp->threshold.interval; | |
1283 | ||
1284 | /* | |
1285 | * Return NULL without doing anything if: | |
1286 | * - this timer is local, or | |
1287 | * - the longterm mechanism is disabled, or | |
1288 | * - this deadline is too short. | |
1289 | */ | |
1290 | if ((callout_flags & TIMER_CALL_LOCAL) != 0 || | |
1291 | (tlp->threshold.interval == TIMER_LONGTERM_NONE) || | |
1292 | (deadline <= longterm_threshold)) { | |
1293 | return NULL; | |
1294 | } | |
1295 | ||
1296 | /* | |
1297 | * Remove timer from its current queue, if any. | |
1298 | */ | |
1299 | *old_queue = timer_call_dequeue_unlocked(call); | |
1300 | ||
1301 | /* | |
1302 | * Lock the longterm queue, queue timer and determine | |
1303 | * whether an update is necessary. | |
1304 | */ | |
1305 | assert(!ml_get_interrupts_enabled()); | |
1306 | simple_lock(&call->lock, LCK_GRP_NULL); | |
1307 | timer_queue_lock_spin(timer_longterm_queue); | |
1308 | TCE(call)->deadline = deadline; | |
1309 | TCE(call)->param1 = param1; | |
1310 | call->ttd = ttd; | |
1311 | call->soft_deadline = soft_deadline; | |
1312 | call->flags = callout_flags; | |
1313 | timer_call_entry_enqueue_tail(call, timer_longterm_queue); | |
1314 | ||
1315 | tlp->enqueues++; | |
1316 | ||
1317 | /* | |
1318 | * We'll need to update the currently set threshold timer | |
1319 | * if the new deadline is sooner and no sooner update is in flight. | |
1320 | */ | |
1321 | if (deadline < tlp->threshold.deadline && | |
1322 | deadline < tlp->threshold.preempted) { | |
1323 | tlp->threshold.preempted = deadline; | |
1324 | tlp->threshold.call = call; | |
1325 | update_required = TRUE; | |
1326 | } | |
1327 | timer_queue_unlock(timer_longterm_queue); | |
1328 | simple_unlock(&call->lock); | |
1329 | ||
1330 | if (update_required) { | |
1331 | /* | |
1332 | * Note: this call expects that calling the master cpu | |
1333 | * alone does not involve locking the topo lock. | |
1334 | */ | |
1335 | timer_call_nosync_cpu( | |
1336 | master_cpu, | |
1337 | (void (*)(void *))timer_longterm_update, | |
1338 | (void *)tlp); | |
1339 | } | |
1340 | ||
1341 | return timer_longterm_queue; | |
1342 | } | |
1343 | ||
1344 | /* | |
1345 | * Scan for timers below the longterm threshold. | |
1346 | * Move these to the local timer queue (of the boot processor on which the | |
1347 | * calling thread is running). | |
1348 | * Both the local (boot) queue and the longterm queue are locked. | |
1349 | * The scan is similar to the timer migrate sequence but is performed by | |
1350 | * successively examining each timer on the longterm queue: | |
1351 | * - if within the short-term threshold | |
1352 | * - enter on the local queue (unless being deleted), | |
1353 | * - otherwise: | |
1354 | * - if sooner, deadline becomes the next threshold deadline. | |
1355 | * The total scan time is limited to TIMER_LONGTERM_SCAN_LIMIT. Should this be | |
1356 | * exceeded, we abort and reschedule again so that we don't shut others from | |
1357 | * the timer queues. Longterm timers firing late is not critical. | |
1358 | */ | |
1359 | void | |
1360 | timer_longterm_scan(timer_longterm_t *tlp, | |
1361 | uint64_t time_start) | |
1362 | { | |
1363 | queue_entry_t qe; | |
1364 | timer_call_t call; | |
1365 | uint64_t threshold; | |
1366 | uint64_t deadline; | |
1367 | uint64_t time_limit = time_start + tlp->scan_limit; | |
1368 | mpqueue_head_t *timer_master_queue; | |
1369 | ||
1370 | assert(!ml_get_interrupts_enabled()); | |
1371 | assert(cpu_number() == master_cpu); | |
1372 | ||
1373 | if (tlp->threshold.interval != TIMER_LONGTERM_NONE) { | |
1374 | threshold = time_start + tlp->threshold.interval; | |
1375 | } | |
1376 | ||
1377 | tlp->threshold.deadline = TIMER_LONGTERM_NONE; | |
1378 | tlp->threshold.call = NULL; | |
1379 | ||
1380 | if (queue_empty(&timer_longterm_queue->head)) { | |
1381 | return; | |
1382 | } | |
1383 | ||
1384 | timer_master_queue = timer_queue_cpu(master_cpu); | |
1385 | timer_queue_lock_spin(timer_master_queue); | |
1386 | ||
1387 | qe = queue_first(&timer_longterm_queue->head); | |
1388 | while (!queue_end(&timer_longterm_queue->head, qe)) { | |
1389 | call = TIMER_CALL(qe); | |
1390 | deadline = call->soft_deadline; | |
1391 | qe = queue_next(qe); | |
1392 | if (!simple_lock_try(&call->lock, LCK_GRP_NULL)) { | |
1393 | /* case (2c) lock order inversion, dequeue only */ | |
1394 | #ifdef TIMER_ASSERT | |
1395 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1396 | DECR_TIMER_ASYNC_DEQ | DBG_FUNC_NONE, | |
1397 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
1398 | VM_KERNEL_UNSLIDE_OR_PERM(TCE(call)->queue), | |
1399 | VM_KERNEL_UNSLIDE_OR_PERM(call->lock.interlock.lock_data), | |
1400 | 0x2c, 0); | |
1401 | #endif | |
1402 | timer_call_entry_dequeue_async(call); | |
1403 | continue; | |
1404 | } | |
1405 | if (deadline < threshold) { | |
1406 | /* | |
1407 | * This timer needs moving (escalating) | |
1408 | * to the local (boot) processor's queue. | |
1409 | */ | |
1410 | #ifdef TIMER_ASSERT | |
1411 | if (deadline < time_start) { | |
1412 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1413 | DECR_TIMER_OVERDUE | DBG_FUNC_NONE, | |
1414 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
1415 | deadline, | |
1416 | time_start, | |
1417 | threshold, | |
1418 | 0); | |
1419 | } | |
1420 | #endif | |
1421 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1422 | DECR_TIMER_ESCALATE | DBG_FUNC_NONE, | |
1423 | VM_KERNEL_UNSLIDE_OR_PERM(call), | |
1424 | TCE(call)->deadline, | |
1425 | TCE(call)->entry_time, | |
1426 | VM_KERNEL_UNSLIDE(TCE(call)->func), | |
1427 | 0); | |
1428 | tlp->escalates++; | |
1429 | timer_call_entry_dequeue(call); | |
1430 | timer_call_entry_enqueue_deadline( | |
1431 | call, timer_master_queue, TCE(call)->deadline); | |
1432 | /* | |
1433 | * A side-effect of the following call is to update | |
1434 | * the actual hardware deadline if required. | |
1435 | */ | |
1436 | (void) timer_queue_assign(deadline); | |
1437 | } else { | |
1438 | if (deadline < tlp->threshold.deadline) { | |
1439 | tlp->threshold.deadline = deadline; | |
1440 | tlp->threshold.call = call; | |
1441 | } | |
1442 | } | |
1443 | simple_unlock(&call->lock); | |
1444 | ||
1445 | /* Abort scan if we're taking too long. */ | |
1446 | if (mach_absolute_time() > time_limit) { | |
1447 | tlp->threshold.deadline = TIMER_LONGTERM_SCAN_AGAIN; | |
1448 | tlp->scan_pauses++; | |
1449 | DBG("timer_longterm_scan() paused %llu, qlen: %llu\n", | |
1450 | time_limit, tlp->queue.count); | |
1451 | break; | |
1452 | } | |
1453 | } | |
1454 | ||
1455 | timer_queue_unlock(timer_master_queue); | |
1456 | } | |
1457 | ||
1458 | void | |
1459 | timer_longterm_callout(timer_call_param_t p0, __unused timer_call_param_t p1) | |
1460 | { | |
1461 | timer_longterm_t *tlp = (timer_longterm_t *) p0; | |
1462 | ||
1463 | timer_longterm_update(tlp); | |
1464 | } | |
1465 | ||
1466 | void | |
1467 | timer_longterm_update_locked(timer_longterm_t *tlp) | |
1468 | { | |
1469 | uint64_t latency; | |
1470 | ||
1471 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1472 | DECR_TIMER_UPDATE | DBG_FUNC_START, | |
1473 | VM_KERNEL_UNSLIDE_OR_PERM(&tlp->queue), | |
1474 | tlp->threshold.deadline, | |
1475 | tlp->threshold.preempted, | |
1476 | tlp->queue.count, 0); | |
1477 | ||
1478 | tlp->scan_time = mach_absolute_time(); | |
1479 | if (tlp->threshold.preempted != TIMER_LONGTERM_NONE) { | |
1480 | tlp->threshold.preempts++; | |
1481 | tlp->threshold.deadline = tlp->threshold.preempted; | |
1482 | tlp->threshold.preempted = TIMER_LONGTERM_NONE; | |
1483 | /* | |
1484 | * Note: in the unlikely event that a pre-empted timer has | |
1485 | * itself been cancelled, we'll simply re-scan later at the | |
1486 | * time of the preempted/cancelled timer. | |
1487 | */ | |
1488 | } else { | |
1489 | tlp->threshold.scans++; | |
1490 | ||
1491 | /* | |
1492 | * Maintain a moving average of our wakeup latency. | |
1493 | * Clamp latency to 0 and ignore above threshold interval. | |
1494 | */ | |
1495 | if (tlp->scan_time > tlp->threshold.deadline_set) { | |
1496 | latency = tlp->scan_time - tlp->threshold.deadline_set; | |
1497 | } else { | |
1498 | latency = 0; | |
1499 | } | |
1500 | if (latency < tlp->threshold.interval) { | |
1501 | tlp->threshold.latency_min = | |
1502 | MIN(tlp->threshold.latency_min, latency); | |
1503 | tlp->threshold.latency_max = | |
1504 | MAX(tlp->threshold.latency_max, latency); | |
1505 | tlp->threshold.latency = | |
1506 | (tlp->threshold.latency * 99 + latency) / 100; | |
1507 | } | |
1508 | ||
1509 | timer_longterm_scan(tlp, tlp->scan_time); | |
1510 | } | |
1511 | ||
1512 | tlp->threshold.deadline_set = tlp->threshold.deadline; | |
1513 | /* The next deadline timer to be set is adjusted */ | |
1514 | if (tlp->threshold.deadline != TIMER_LONGTERM_NONE && | |
1515 | tlp->threshold.deadline != TIMER_LONGTERM_SCAN_AGAIN) { | |
1516 | tlp->threshold.deadline_set -= tlp->threshold.margin; | |
1517 | tlp->threshold.deadline_set -= tlp->threshold.latency; | |
1518 | } | |
1519 | ||
1520 | /* Throttle next scan time */ | |
1521 | uint64_t scan_clamp = mach_absolute_time() + tlp->scan_interval; | |
1522 | if (tlp->threshold.deadline_set < scan_clamp) { | |
1523 | tlp->threshold.deadline_set = scan_clamp; | |
1524 | } | |
1525 | ||
1526 | TIMER_KDEBUG_TRACE(KDEBUG_TRACE, | |
1527 | DECR_TIMER_UPDATE | DBG_FUNC_END, | |
1528 | VM_KERNEL_UNSLIDE_OR_PERM(&tlp->queue), | |
1529 | tlp->threshold.deadline, | |
1530 | tlp->threshold.scans, | |
1531 | tlp->queue.count, 0); | |
1532 | } | |
1533 | ||
1534 | void | |
1535 | timer_longterm_update(timer_longterm_t *tlp) | |
1536 | { | |
1537 | spl_t s = splclock(); | |
1538 | ||
1539 | timer_queue_lock_spin(timer_longterm_queue); | |
1540 | ||
1541 | if (cpu_number() != master_cpu) { | |
1542 | panic("timer_longterm_update_master() on non-boot cpu"); | |
1543 | } | |
1544 | ||
1545 | timer_longterm_update_locked(tlp); | |
1546 | ||
1547 | if (tlp->threshold.deadline != TIMER_LONGTERM_NONE) { | |
1548 | timer_call_enter( | |
1549 | &tlp->threshold.timer, | |
1550 | tlp->threshold.deadline_set, | |
1551 | TIMER_CALL_LOCAL | TIMER_CALL_SYS_CRITICAL); | |
1552 | } | |
1553 | ||
1554 | timer_queue_unlock(timer_longterm_queue); | |
1555 | splx(s); | |
1556 | } | |
1557 | ||
1558 | void | |
1559 | timer_longterm_init(void) | |
1560 | { | |
1561 | uint32_t longterm; | |
1562 | timer_longterm_t *tlp = &timer_longterm; | |
1563 | ||
1564 | DBG("timer_longterm_init() tlp: %p, queue: %p\n", tlp, &tlp->queue); | |
1565 | ||
1566 | /* | |
1567 | * Set the longterm timer threshold. Defaults to TIMER_LONGTERM_THRESHOLD | |
1568 | * or TIMER_LONGTERM_NONE (disabled) for server; | |
1569 | * overridden longterm boot-arg | |
1570 | */ | |
1571 | tlp->threshold.interval = serverperfmode ? TIMER_LONGTERM_NONE | |
1572 | : TIMER_LONGTERM_THRESHOLD; | |
1573 | if (PE_parse_boot_argn("longterm", &longterm, sizeof(longterm))) { | |
1574 | tlp->threshold.interval = (longterm == 0) ? | |
1575 | TIMER_LONGTERM_NONE : | |
1576 | longterm * NSEC_PER_MSEC; | |
1577 | } | |
1578 | if (tlp->threshold.interval != TIMER_LONGTERM_NONE) { | |
1579 | printf("Longterm timer threshold: %llu ms\n", | |
1580 | tlp->threshold.interval / NSEC_PER_MSEC); | |
1581 | kprintf("Longterm timer threshold: %llu ms\n", | |
1582 | tlp->threshold.interval / NSEC_PER_MSEC); | |
1583 | nanoseconds_to_absolutetime(tlp->threshold.interval, | |
1584 | &tlp->threshold.interval); | |
1585 | tlp->threshold.margin = tlp->threshold.interval / 10; | |
1586 | tlp->threshold.latency_min = EndOfAllTime; | |
1587 | tlp->threshold.latency_max = 0; | |
1588 | } | |
1589 | ||
1590 | tlp->threshold.preempted = TIMER_LONGTERM_NONE; | |
1591 | tlp->threshold.deadline = TIMER_LONGTERM_NONE; | |
1592 | ||
1593 | lck_attr_setdefault(&timer_longterm_lck_attr); | |
1594 | lck_grp_attr_setdefault(&timer_longterm_lck_grp_attr); | |
1595 | lck_grp_init(&timer_longterm_lck_grp, | |
1596 | "timer_longterm", &timer_longterm_lck_grp_attr); | |
1597 | mpqueue_init(&tlp->queue, | |
1598 | &timer_longterm_lck_grp, &timer_longterm_lck_attr); | |
1599 | ||
1600 | timer_call_setup(&tlp->threshold.timer, | |
1601 | timer_longterm_callout, (timer_call_param_t) tlp); | |
1602 | ||
1603 | timer_longterm_queue = &tlp->queue; | |
1604 | } | |
1605 | ||
1606 | enum { | |
1607 | THRESHOLD, QCOUNT, | |
1608 | ENQUEUES, DEQUEUES, ESCALATES, SCANS, PREEMPTS, | |
1609 | LATENCY, LATENCY_MIN, LATENCY_MAX, SCAN_LIMIT, SCAN_INTERVAL, PAUSES | |
1610 | }; | |
1611 | uint64_t | |
1612 | timer_sysctl_get(int oid) | |
1613 | { | |
1614 | timer_longterm_t *tlp = &timer_longterm; | |
1615 | ||
1616 | switch (oid) { | |
1617 | case THRESHOLD: | |
1618 | return (tlp->threshold.interval == TIMER_LONGTERM_NONE) ? | |
1619 | 0 : tlp->threshold.interval / NSEC_PER_MSEC; | |
1620 | case QCOUNT: | |
1621 | return tlp->queue.count; | |
1622 | case ENQUEUES: | |
1623 | return tlp->enqueues; | |
1624 | case DEQUEUES: | |
1625 | return tlp->dequeues; | |
1626 | case ESCALATES: | |
1627 | return tlp->escalates; | |
1628 | case SCANS: | |
1629 | return tlp->threshold.scans; | |
1630 | case PREEMPTS: | |
1631 | return tlp->threshold.preempts; | |
1632 | case LATENCY: | |
1633 | return tlp->threshold.latency; | |
1634 | case LATENCY_MIN: | |
1635 | return tlp->threshold.latency_min; | |
1636 | case LATENCY_MAX: | |
1637 | return tlp->threshold.latency_max; | |
1638 | case SCAN_LIMIT: | |
1639 | return tlp->scan_limit; | |
1640 | case SCAN_INTERVAL: | |
1641 | return tlp->scan_interval; | |
1642 | case PAUSES: | |
1643 | return tlp->scan_pauses; | |
1644 | default: | |
1645 | return 0; | |
1646 | } | |
1647 | } | |
1648 | ||
1649 | /* | |
1650 | * timer_master_scan() is the inverse of timer_longterm_scan() | |
1651 | * since it un-escalates timers to the longterm queue. | |
1652 | */ | |
1653 | static void | |
1654 | timer_master_scan(timer_longterm_t *tlp, | |
1655 | uint64_t now) | |
1656 | { | |
1657 | queue_entry_t qe; | |
1658 | timer_call_t call; | |
1659 | uint64_t threshold; | |
1660 | uint64_t deadline; | |
1661 | mpqueue_head_t *timer_master_queue; | |
1662 | ||
1663 | if (tlp->threshold.interval != TIMER_LONGTERM_NONE) { | |
1664 | threshold = now + tlp->threshold.interval; | |
1665 | } else { | |
1666 | threshold = TIMER_LONGTERM_NONE; | |
1667 | } | |
1668 | ||
1669 | timer_master_queue = timer_queue_cpu(master_cpu); | |
1670 | timer_queue_lock_spin(timer_master_queue); | |
1671 | ||
1672 | qe = queue_first(&timer_master_queue->head); | |
1673 | while (!queue_end(&timer_master_queue->head, qe)) { | |
1674 | call = TIMER_CALL(qe); | |
1675 | deadline = TCE(call)->deadline; | |
1676 | qe = queue_next(qe); | |
1677 | if ((call->flags & TIMER_CALL_LOCAL) != 0) { | |
1678 | continue; | |
1679 | } | |
1680 | if (!simple_lock_try(&call->lock, LCK_GRP_NULL)) { | |
1681 | /* case (2c) lock order inversion, dequeue only */ | |
1682 | timer_call_entry_dequeue_async(call); | |
1683 | continue; | |
1684 | } | |
1685 | if (deadline > threshold) { | |
1686 | /* move from master to longterm */ | |
1687 | timer_call_entry_dequeue(call); | |
1688 | timer_call_entry_enqueue_tail(call, timer_longterm_queue); | |
1689 | if (deadline < tlp->threshold.deadline) { | |
1690 | tlp->threshold.deadline = deadline; | |
1691 | tlp->threshold.call = call; | |
1692 | } | |
1693 | } | |
1694 | simple_unlock(&call->lock); | |
1695 | } | |
1696 | timer_queue_unlock(timer_master_queue); | |
1697 | } | |
1698 | ||
1699 | static void | |
1700 | timer_sysctl_set_threshold(uint64_t value) | |
1701 | { | |
1702 | timer_longterm_t *tlp = &timer_longterm; | |
1703 | spl_t s = splclock(); | |
1704 | boolean_t threshold_increase; | |
1705 | ||
1706 | timer_queue_lock_spin(timer_longterm_queue); | |
1707 | ||
1708 | timer_call_cancel(&tlp->threshold.timer); | |
1709 | ||
1710 | /* | |
1711 | * Set the new threshold and note whther it's increasing. | |
1712 | */ | |
1713 | if (value == 0) { | |
1714 | tlp->threshold.interval = TIMER_LONGTERM_NONE; | |
1715 | threshold_increase = TRUE; | |
1716 | timer_call_cancel(&tlp->threshold.timer); | |
1717 | } else { | |
1718 | uint64_t old_interval = tlp->threshold.interval; | |
1719 | tlp->threshold.interval = value * NSEC_PER_MSEC; | |
1720 | nanoseconds_to_absolutetime(tlp->threshold.interval, | |
1721 | &tlp->threshold.interval); | |
1722 | tlp->threshold.margin = tlp->threshold.interval / 10; | |
1723 | if (old_interval == TIMER_LONGTERM_NONE) { | |
1724 | threshold_increase = FALSE; | |
1725 | } else { | |
1726 | threshold_increase = (tlp->threshold.interval > old_interval); | |
1727 | } | |
1728 | } | |
1729 | ||
1730 | if (threshold_increase /* or removal */) { | |
1731 | /* Escalate timers from the longterm queue */ | |
1732 | timer_longterm_scan(tlp, mach_absolute_time()); | |
1733 | } else { /* decrease or addition */ | |
1734 | /* | |
1735 | * We scan the local/master queue for timers now longterm. | |
1736 | * To be strictly correct, we should scan all processor queues | |
1737 | * but timer migration results in most timers gravitating to the | |
1738 | * master processor in any case. | |
1739 | */ | |
1740 | timer_master_scan(tlp, mach_absolute_time()); | |
1741 | } | |
1742 | ||
1743 | /* Set new timer accordingly */ | |
1744 | tlp->threshold.deadline_set = tlp->threshold.deadline; | |
1745 | if (tlp->threshold.deadline != TIMER_LONGTERM_NONE) { | |
1746 | tlp->threshold.deadline_set -= tlp->threshold.margin; | |
1747 | tlp->threshold.deadline_set -= tlp->threshold.latency; | |
1748 | timer_call_enter( | |
1749 | &tlp->threshold.timer, | |
1750 | tlp->threshold.deadline_set, | |
1751 | TIMER_CALL_LOCAL | TIMER_CALL_SYS_CRITICAL); | |
1752 | } | |
1753 | ||
1754 | /* Reset stats */ | |
1755 | tlp->enqueues = 0; | |
1756 | tlp->dequeues = 0; | |
1757 | tlp->escalates = 0; | |
1758 | tlp->scan_pauses = 0; | |
1759 | tlp->threshold.scans = 0; | |
1760 | tlp->threshold.preempts = 0; | |
1761 | tlp->threshold.latency = 0; | |
1762 | tlp->threshold.latency_min = EndOfAllTime; | |
1763 | tlp->threshold.latency_max = 0; | |
1764 | ||
1765 | timer_queue_unlock(timer_longterm_queue); | |
1766 | splx(s); | |
1767 | } | |
1768 | ||
1769 | int | |
1770 | timer_sysctl_set(int oid, uint64_t value) | |
1771 | { | |
1772 | switch (oid) { | |
1773 | case THRESHOLD: | |
1774 | timer_call_cpu( | |
1775 | master_cpu, | |
1776 | (void (*)(void *))timer_sysctl_set_threshold, | |
1777 | (void *) value); | |
1778 | return KERN_SUCCESS; | |
1779 | case SCAN_LIMIT: | |
1780 | timer_longterm.scan_limit = value; | |
1781 | return KERN_SUCCESS; | |
1782 | case SCAN_INTERVAL: | |
1783 | timer_longterm.scan_interval = value; | |
1784 | return KERN_SUCCESS; | |
1785 | default: | |
1786 | return KERN_INVALID_ARGUMENT; | |
1787 | } | |
1788 | } | |
1789 | ||
1790 | ||
1791 | /* Select timer coalescing window based on per-task quality-of-service hints */ | |
1792 | static boolean_t | |
1793 | tcoal_qos_adjust(thread_t t, int32_t *tshift, uint64_t *tmax_abstime, boolean_t *pratelimited) | |
1794 | { | |
1795 | uint32_t latency_qos; | |
1796 | boolean_t adjusted = FALSE; | |
1797 | task_t ctask = t->task; | |
1798 | ||
1799 | if (ctask) { | |
1800 | latency_qos = proc_get_effective_thread_policy(t, TASK_POLICY_LATENCY_QOS); | |
1801 | ||
1802 | assert(latency_qos <= NUM_LATENCY_QOS_TIERS); | |
1803 | ||
1804 | if (latency_qos) { | |
1805 | *tshift = tcoal_prio_params.latency_qos_scale[latency_qos - 1]; | |
1806 | *tmax_abstime = tcoal_prio_params.latency_qos_abstime_max[latency_qos - 1]; | |
1807 | *pratelimited = tcoal_prio_params.latency_tier_rate_limited[latency_qos - 1]; | |
1808 | adjusted = TRUE; | |
1809 | } | |
1810 | } | |
1811 | return adjusted; | |
1812 | } | |
1813 | ||
1814 | ||
1815 | /* Adjust timer deadlines based on priority of the thread and the | |
1816 | * urgency value provided at timeout establishment. With this mechanism, | |
1817 | * timers are no longer necessarily sorted in order of soft deadline | |
1818 | * on a given timer queue, i.e. they may be differentially skewed. | |
1819 | * In the current scheme, this could lead to fewer pending timers | |
1820 | * processed than is technically possible when the HW deadline arrives. | |
1821 | */ | |
1822 | static void | |
1823 | timer_compute_leeway(thread_t cthread, int32_t urgency, int32_t *tshift, uint64_t *tmax_abstime, boolean_t *pratelimited) | |
1824 | { | |
1825 | int16_t tpri = cthread->sched_pri; | |
1826 | if ((urgency & TIMER_CALL_USER_MASK) != 0) { | |
1827 | if (tpri >= BASEPRI_RTQUEUES || | |
1828 | urgency == TIMER_CALL_USER_CRITICAL) { | |
1829 | *tshift = tcoal_prio_params.timer_coalesce_rt_shift; | |
1830 | *tmax_abstime = tcoal_prio_params.timer_coalesce_rt_abstime_max; | |
1831 | TCOAL_PRIO_STAT(rt_tcl); | |
1832 | } else if (proc_get_effective_thread_policy(cthread, TASK_POLICY_DARWIN_BG) || | |
1833 | (urgency == TIMER_CALL_USER_BACKGROUND)) { | |
1834 | /* Determine if timer should be subjected to a lower QoS */ | |
1835 | if (tcoal_qos_adjust(cthread, tshift, tmax_abstime, pratelimited)) { | |
1836 | if (*tmax_abstime > tcoal_prio_params.timer_coalesce_bg_abstime_max) { | |
1837 | return; | |
1838 | } else { | |
1839 | *pratelimited = FALSE; | |
1840 | } | |
1841 | } | |
1842 | *tshift = tcoal_prio_params.timer_coalesce_bg_shift; | |
1843 | *tmax_abstime = tcoal_prio_params.timer_coalesce_bg_abstime_max; | |
1844 | TCOAL_PRIO_STAT(bg_tcl); | |
1845 | } else if (tpri >= MINPRI_KERNEL) { | |
1846 | *tshift = tcoal_prio_params.timer_coalesce_kt_shift; | |
1847 | *tmax_abstime = tcoal_prio_params.timer_coalesce_kt_abstime_max; | |
1848 | TCOAL_PRIO_STAT(kt_tcl); | |
1849 | } else if (cthread->sched_mode == TH_MODE_FIXED) { | |
1850 | *tshift = tcoal_prio_params.timer_coalesce_fp_shift; | |
1851 | *tmax_abstime = tcoal_prio_params.timer_coalesce_fp_abstime_max; | |
1852 | TCOAL_PRIO_STAT(fp_tcl); | |
1853 | } else if (tcoal_qos_adjust(cthread, tshift, tmax_abstime, pratelimited)) { | |
1854 | TCOAL_PRIO_STAT(qos_tcl); | |
1855 | } else if (cthread->sched_mode == TH_MODE_TIMESHARE) { | |
1856 | *tshift = tcoal_prio_params.timer_coalesce_ts_shift; | |
1857 | *tmax_abstime = tcoal_prio_params.timer_coalesce_ts_abstime_max; | |
1858 | TCOAL_PRIO_STAT(ts_tcl); | |
1859 | } else { | |
1860 | TCOAL_PRIO_STAT(nc_tcl); | |
1861 | } | |
1862 | } else if (urgency == TIMER_CALL_SYS_BACKGROUND) { | |
1863 | *tshift = tcoal_prio_params.timer_coalesce_bg_shift; | |
1864 | *tmax_abstime = tcoal_prio_params.timer_coalesce_bg_abstime_max; | |
1865 | TCOAL_PRIO_STAT(bg_tcl); | |
1866 | } else { | |
1867 | *tshift = tcoal_prio_params.timer_coalesce_kt_shift; | |
1868 | *tmax_abstime = tcoal_prio_params.timer_coalesce_kt_abstime_max; | |
1869 | TCOAL_PRIO_STAT(kt_tcl); | |
1870 | } | |
1871 | } | |
1872 | ||
1873 | ||
1874 | int timer_user_idle_level; | |
1875 | ||
1876 | uint64_t | |
1877 | timer_call_slop(uint64_t deadline, uint64_t now, uint32_t flags, thread_t cthread, boolean_t *pratelimited) | |
1878 | { | |
1879 | int32_t tcs_shift = 0; | |
1880 | uint64_t tcs_max_abstime = 0; | |
1881 | uint64_t adjval; | |
1882 | uint32_t urgency = (flags & TIMER_CALL_URGENCY_MASK); | |
1883 | ||
1884 | if (mach_timer_coalescing_enabled && | |
1885 | (deadline > now) && (urgency != TIMER_CALL_SYS_CRITICAL)) { | |
1886 | timer_compute_leeway(cthread, urgency, &tcs_shift, &tcs_max_abstime, pratelimited); | |
1887 | ||
1888 | if (tcs_shift >= 0) { | |
1889 | adjval = MIN((deadline - now) >> tcs_shift, tcs_max_abstime); | |
1890 | } else { | |
1891 | adjval = MIN((deadline - now) << (-tcs_shift), tcs_max_abstime); | |
1892 | } | |
1893 | /* Apply adjustments derived from "user idle level" heuristic */ | |
1894 | adjval += (adjval * timer_user_idle_level) >> 7; | |
1895 | return adjval; | |
1896 | } else { | |
1897 | return 0; | |
1898 | } | |
1899 | } | |
1900 | ||
1901 | int | |
1902 | timer_get_user_idle_level(void) | |
1903 | { | |
1904 | return timer_user_idle_level; | |
1905 | } | |
1906 | ||
1907 | kern_return_t | |
1908 | timer_set_user_idle_level(int ilevel) | |
1909 | { | |
1910 | boolean_t do_reeval = FALSE; | |
1911 | ||
1912 | if ((ilevel < 0) || (ilevel > 128)) { | |
1913 | return KERN_INVALID_ARGUMENT; | |
1914 | } | |
1915 | ||
1916 | if (ilevel < timer_user_idle_level) { | |
1917 | do_reeval = TRUE; | |
1918 | } | |
1919 | ||
1920 | timer_user_idle_level = ilevel; | |
1921 | ||
1922 | if (do_reeval) { | |
1923 | ml_timer_evaluate(); | |
1924 | } | |
1925 | ||
1926 | return KERN_SUCCESS; | |
1927 | } |