forked from ~ljy/RK356X_SDK_RELEASE
blame | history | patch | commit | commitdiff
hc
2024-10-09 244b2c5ca8b14627e4a17755e5922221e121c771 
 kernel/kernel/time/timekeeping.c
....@@ -1,13 +1,8 @@
1
+// SPDX-License-Identifier: GPL-2.0
12 /*
2
- *  linux/kernel/time/timekeeping.c
3
- *
4
- *  Kernel timekeeping code and accessor functions
5
- *
6
- *  This code was moved from linux/kernel/timer.c.
7
- *  Please see that file for copyright and history logs.
8
- *
3
+ *  Kernel timekeeping code and accessor functions. Based on code from
4
+ *  timer.c, moved in commit 8524070b7982.
95  */
10
-
116 #include <linux/timekeeper_internal.h>
127 #include <linux/module.h>
138 #include <linux/interrupt.h>
....@@ -22,10 +17,13 @@
2217 #include <linux/clocksource.h>
2318 #include <linux/jiffies.h>
2419 #include <linux/time.h>
20
+#include <linux/timex.h>
2521 #include <linux/tick.h>
2622 #include <linux/stop_machine.h>
2723 #include <linux/pvclock_gtod.h>
2824 #include <linux/compiler.h>
25
+#include <linux/audit.h>
26
+#include <linux/random.h>
2927 
3028 #include "tick-internal.h"
3129 #include "ntp_internal.h"
....@@ -43,19 +41,23 @@
4341 	TK_ADV_FREQ
4442 };
4543 
44
+DEFINE_RAW_SPINLOCK(timekeeper_lock);
45
+
4646 /*
4747  * The most important data for readout fits into a single 64 byte
4848  * cache line.
4949  */
5050 static struct {
51
-	seqcount_t		seq;
51
+	seqcount_raw_spinlock_t	seq;
5252 	struct timekeeper	timekeeper;
5353 } tk_core ____cacheline_aligned = {
54
-	.seq = SEQCNT_ZERO(tk_core.seq),
54
+	.seq = SEQCNT_RAW_SPINLOCK_ZERO(tk_core.seq, &timekeeper_lock),
5555 };
5656 
57
-static DEFINE_RAW_SPINLOCK(timekeeper_lock);
5857 static struct timekeeper shadow_timekeeper;
58
+
59
+/* flag for if timekeeping is suspended */
60
+int __read_mostly timekeeping_suspended;
5961 
6062 /**
6163  * struct tk_fast - NMI safe timekeeper
....@@ -67,7 +69,7 @@
6769  * See @update_fast_timekeeper() below.
6870  */
6971 struct tk_fast {
70
-	seqcount_t		seq;
72
+	seqcount_latch_t	seq;
7173 	struct tk_read_base	base[2];
7274 };
7375 
....@@ -76,25 +78,41 @@
7678 
7779 static u64 dummy_clock_read(struct clocksource *cs)
7880 {
79
-	return cycles_at_suspend;
81
+	if (timekeeping_suspended)
82
+		return cycles_at_suspend;
83
+	return local_clock();
8084 }
8185 
8286 static struct clocksource dummy_clock = {
8387 	.read = dummy_clock_read,
8488 };
8589 
90
+/*
91
+ * Boot time initialization which allows local_clock() to be utilized
92
+ * during early boot when clocksources are not available. local_clock()
93
+ * returns nanoseconds already so no conversion is required, hence mult=1
94
+ * and shift=0. When the first proper clocksource is installed then
95
+ * the fast time keepers are updated with the correct values.
96
+ */
97
+#define FAST_TK_INIT						\
98
+	{							\
99
+		.clock		= &dummy_clock,			\
100
+		.mask		= CLOCKSOURCE_MASK(64),		\
101
+		.mult		= 1,				\
102
+		.shift		= 0,				\
103
+	}
104
+
86105 static struct tk_fast tk_fast_mono ____cacheline_aligned = {
87
-	.base[0] = { .clock = &dummy_clock, },
88
-	.base[1] = { .clock = &dummy_clock, },
106
+	.seq     = SEQCNT_LATCH_ZERO(tk_fast_mono.seq),
107
+	.base[0] = FAST_TK_INIT,
108
+	.base[1] = FAST_TK_INIT,
89109 };
90110 
91111 static struct tk_fast tk_fast_raw  ____cacheline_aligned = {
92
-	.base[0] = { .clock = &dummy_clock, },
93
-	.base[1] = { .clock = &dummy_clock, },
112
+	.seq     = SEQCNT_LATCH_ZERO(tk_fast_raw.seq),
113
+	.base[0] = FAST_TK_INIT,
114
+	.base[1] = FAST_TK_INIT,
94115 };
95
-
96
-/* flag for if timekeeping is suspended */
97
-int __read_mostly timekeeping_suspended;
98116 
99117 static inline void tk_normalize_xtime(struct timekeeper *tk)
100118 {
....@@ -161,7 +179,7 @@
161179  * tk_clock_read - atomic clocksource read() helper
162180  *
163181  * This helper is necessary to use in the read paths because, while the
164
- * seqlock ensures we don't return a bad value while structures are updated,
182
+ * seqcount ensures we don't return a bad value while structures are updated,
165183  * it doesn't protect from potential crashes. There is the possibility that
166184  * the tkr's clocksource may change between the read reference, and the
167185  * clock reference passed to the read function.  This can cause crashes if
....@@ -226,10 +244,10 @@
226244 	unsigned int seq;
227245 
228246 	/*
229
-	 * Since we're called holding a seqlock, the data may shift
247
+	 * Since we're called holding a seqcount, the data may shift
230248 	 * under us while we're doing the calculation. This can cause
231249 	 * false positives, since we'd note a problem but throw the
232
-	 * results away. So nest another seqlock here to atomically
250
+	 * results away. So nest another seqcount here to atomically
233251 	 * grab the points we are checking with.
234252 	 */
235253 	do {
....@@ -468,7 +486,7 @@
468486 					tk_clock_read(tkr),
469487 					tkr->cycle_last,
470488 					tkr->mask));
471
-	} while (read_seqcount_retry(&tkf->seq, seq));
489
+	} while (read_seqcount_latch_retry(&tkf->seq, seq));
472490 
473491 	return now;
474492 }
....@@ -490,7 +508,7 @@
490508  *
491509  * To keep it NMI safe since we're accessing from tracing, we're not using a
492510  * separate timekeeper with updates to monotonic clock and boot offset
493
- * protected with seqlocks. This has the following minor side effects:
511
+ * protected with seqcounts. This has the following minor side effects:
494512  *
495513  * (1) Its possible that a timestamp be taken after the boot offset is updated
496514  * but before the timekeeper is updated. If this happens, the new boot offset
....@@ -514,29 +532,29 @@
514532 }
515533 EXPORT_SYMBOL_GPL(ktime_get_boot_fast_ns);
516534 
517
-
518535 /*
519536  * See comment for __ktime_get_fast_ns() vs. timestamp ordering
520537  */
521
-static __always_inline u64 __ktime_get_real_fast_ns(struct tk_fast *tkf)
538
+static __always_inline u64 __ktime_get_real_fast(struct tk_fast *tkf, u64 *mono)
522539 {
523540 	struct tk_read_base *tkr;
541
+	u64 basem, baser, delta;
524542 	unsigned int seq;
525
-	u64 now;
526543 
527544 	do {
528545 		seq = raw_read_seqcount_latch(&tkf->seq);
529546 		tkr = tkf->base + (seq & 0x01);
530
-		now = ktime_to_ns(tkr->base_real);
547
+		basem = ktime_to_ns(tkr->base);
548
+		baser = ktime_to_ns(tkr->base_real);
531549 
532
-		now += timekeeping_delta_to_ns(tkr,
533
-				clocksource_delta(
534
-					tk_clock_read(tkr),
535
-					tkr->cycle_last,
536
-					tkr->mask));
537
-	} while (read_seqcount_retry(&tkf->seq, seq));
550
+		delta = timekeeping_delta_to_ns(tkr,
551
+				clocksource_delta(tk_clock_read(tkr),
552
+				tkr->cycle_last, tkr->mask));
553
+	} while (read_seqcount_latch_retry(&tkf->seq, seq));
538554 
539
-	return now;
555
+	if (mono)
556
+		*mono = basem + delta;
557
+	return baser + delta;
540558 }
541559 
542560 /**
....@@ -544,9 +562,63 @@
544562  */
545563 u64 ktime_get_real_fast_ns(void)
546564 {
547
-	return __ktime_get_real_fast_ns(&tk_fast_mono);
565
+	return __ktime_get_real_fast(&tk_fast_mono, NULL);
548566 }
549567 EXPORT_SYMBOL_GPL(ktime_get_real_fast_ns);
568
+
569
+/**
570
+ * ktime_get_fast_timestamps: - NMI safe timestamps
571
+ * @snapshot:	Pointer to timestamp storage
572
+ *
573
+ * Stores clock monotonic, boottime and realtime timestamps.
574
+ *
575
+ * Boot time is a racy access on 32bit systems if the sleep time injection
576
+ * happens late during resume and not in timekeeping_resume(). That could
577
+ * be avoided by expanding struct tk_read_base with boot offset for 32bit
578
+ * and adding more overhead to the update. As this is a hard to observe
579
+ * once per resume event which can be filtered with reasonable effort using
580
+ * the accurate mono/real timestamps, it's probably not worth the trouble.
581
+ *
582
+ * Aside of that it might be possible on 32 and 64 bit to observe the
583
+ * following when the sleep time injection happens late:
584
+ *
585
+ * CPU 0				CPU 1
586
+ * timekeeping_resume()
587
+ * ktime_get_fast_timestamps()
588
+ *	mono, real = __ktime_get_real_fast()
589
+ *					inject_sleep_time()
590
+ *					   update boot offset
591
+ *	boot = mono + bootoffset;
592
+ *
593
+ * That means that boot time already has the sleep time adjustment, but
594
+ * real time does not. On the next readout both are in sync again.
595
+ *
596
+ * Preventing this for 64bit is not really feasible without destroying the
597
+ * careful cache layout of the timekeeper because the sequence count and
598
+ * struct tk_read_base would then need two cache lines instead of one.
599
+ *
600
+ * Access to the time keeper clock source is disabled accross the innermost
601
+ * steps of suspend/resume. The accessors still work, but the timestamps
602
+ * are frozen until time keeping is resumed which happens very early.
603
+ *
604
+ * For regular suspend/resume there is no observable difference vs. sched
605
+ * clock, but it might affect some of the nasty low level debug printks.
606
+ *
607
+ * OTOH, access to sched clock is not guaranteed accross suspend/resume on
608
+ * all systems either so it depends on the hardware in use.
609
+ *
610
+ * If that turns out to be a real problem then this could be mitigated by
611
+ * using sched clock in a similar way as during early boot. But it's not as
612
+ * trivial as on early boot because it needs some careful protection
613
+ * against the clock monotonic timestamp jumping backwards on resume.
614
+ */
615
+void ktime_get_fast_timestamps(struct ktime_timestamps *snapshot)
616
+{
617
+	struct timekeeper *tk = &tk_core.timekeeper;
618
+
619
+	snapshot->real = __ktime_get_real_fast(&tk_fast_mono, &snapshot->mono);
620
+	snapshot->boot = snapshot->mono + ktime_to_ns(data_race(tk->offs_boot));
621
+}
550622 
551623 /**
552624  * halt_fast_timekeeper - Prevent fast timekeeper from accessing clocksource.
....@@ -730,7 +802,7 @@
730802 void ktime_get_real_ts64(struct timespec64 *ts)
731803 {
732804 	struct timekeeper *tk = &tk_core.timekeeper;
733
-	unsigned long seq;
805
+	unsigned int seq;
734806 	u64 nsecs;
735807 
736808 	WARN_ON(timekeeping_suspended);
....@@ -840,7 +912,7 @@
840912 ktime_t ktime_mono_to_any(ktime_t tmono, enum tk_offsets offs)
841913 {
842914 	ktime_t *offset = offsets[offs];
843
-	unsigned long seq;
915
+	unsigned int seq;
844916 	ktime_t tconv;
845917 
846918 	do {
....@@ -957,7 +1029,7 @@
9571029  * but without the sequence counter protect. This internal function
9581030  * is called just when timekeeping lock is already held.
9591031  */
960
-time64_t __ktime_get_real_seconds(void)
1032
+noinstr time64_t __ktime_get_real_seconds(void)
9611033 {
9621034 	struct timekeeper *tk = &tk_core.timekeeper;
9631035 
....@@ -971,7 +1043,7 @@
9711043 void ktime_get_snapshot(struct system_time_snapshot *systime_snapshot)
9721044 {
9731045 	struct timekeeper *tk = &tk_core.timekeeper;
974
-	unsigned long seq;
1046
+	unsigned int seq;
9751047 	ktime_t base_raw;
9761048 	ktime_t base_real;
9771049 	u64 nsec_raw;
....@@ -1132,7 +1204,7 @@
11321204 	ktime_t base_real, base_raw;
11331205 	u64 nsec_real, nsec_raw;
11341206 	u8 cs_was_changed_seq;
1135
-	unsigned long seq;
1207
+	unsigned int seq;
11361208 	bool do_interp;
11371209 	int ret;
11381210 
....@@ -1258,6 +1330,11 @@
12581330 
12591331 	/* signal hrtimers about time change */
12601332 	clock_was_set();
1333
+
1334
+	if (!ret) {
1335
+		audit_tk_injoffset(ts_delta);
1336
+		add_device_randomness(ts, sizeof(*ts));
1337
+	}
12611338 
12621339 	return ret;
12631340 }
....@@ -1418,7 +1495,7 @@
14181495 void ktime_get_raw_ts64(struct timespec64 *ts)
14191496 {
14201497 	struct timekeeper *tk = &tk_core.timekeeper;
1421
-	unsigned long seq;
1498
+	unsigned int seq;
14221499 	u64 nsecs;
14231500 
14241501 	do {
....@@ -1440,7 +1517,7 @@
14401517 int timekeeping_valid_for_hres(void)
14411518 {
14421519 	struct timekeeper *tk = &tk_core.timekeeper;
1443
-	unsigned long seq;
1520
+	unsigned int seq;
14441521 	int ret;
14451522 
14461523 	do {
....@@ -1459,7 +1536,7 @@
14591536 u64 timekeeping_max_deferment(void)
14601537 {
14611538 	struct timekeeper *tk = &tk_core.timekeeper;
1462
-	unsigned long seq;
1539
+	unsigned int seq;
14631540 	u64 ret;
14641541 
14651542 	do {
....@@ -1473,7 +1550,7 @@
14731550 }
14741551 
14751552 /**
1476
- * read_persistent_clock -  Return time from the persistent clock.
1553
+ * read_persistent_clock64 -  Return time from the persistent clock.
14771554  *
14781555  * Weak dummy function for arches that do not yet support it.
14791556  * Reads the time from the battery backed persistent clock.
....@@ -1481,18 +1558,10 @@
14811558  *
14821559  *  XXX - Do be sure to remove it once all arches implement it.
14831560  */
1484
-void __weak read_persistent_clock(struct timespec *ts)
1561
+void __weak read_persistent_clock64(struct timespec64 *ts)
14851562 {
14861563 	ts->tv_sec = 0;
14871564 	ts->tv_nsec = 0;
1488
-}
1489
-
1490
-void __weak read_persistent_clock64(struct timespec64 *ts64)
1491
-{
1492
-	struct timespec ts;
1493
-
1494
-	read_persistent_clock(&ts);
1495
-	*ts64 = timespec_to_timespec64(ts);
14961565 }
14971566 
14981567 /**
....@@ -2009,7 +2078,7 @@
20092078  * logarithmic_accumulation - shifted accumulation of cycles
20102079  *
20112080  * This functions accumulates a shifted interval of cycles into
2012
- * into a shifted interval nanoseconds. Allows for O(log) accumulation
2081
+ * a shifted interval nanoseconds. Allows for O(log) accumulation
20132082  * loop.
20142083  *
20152084  * Returns the unconsumed cycles.
....@@ -2167,7 +2236,7 @@
21672236 void ktime_get_coarse_real_ts64(struct timespec64 *ts)
21682237 {
21692238 	struct timekeeper *tk = &tk_core.timekeeper;
2170
-	unsigned long seq;
2239
+	unsigned int seq;
21712240 
21722241 	do {
21732242 		seq = read_seqcount_begin(&tk_core.seq);
....@@ -2181,7 +2250,7 @@
21812250 {
21822251 	struct timekeeper *tk = &tk_core.timekeeper;
21832252 	struct timespec64 now, mono;
2184
-	unsigned long seq;
2253
+	unsigned int seq;
21852254 
21862255 	do {
21872256 		seq = read_seqcount_begin(&tk_core.seq);
....@@ -2201,7 +2270,7 @@
22012270 void do_timer(unsigned long ticks)
22022271 {
22032272 	jiffies_64 += ticks;
2204
-	calc_global_load(ticks);
2273
+	calc_global_load();
22052274 }
22062275 
22072276 /**
....@@ -2251,7 +2320,7 @@
22512320 /**
22522321  * timekeeping_validate_timex - Ensures the timex is ok for use in do_adjtimex
22532322  */
2254
-static int timekeeping_validate_timex(const struct timex *txc)
2323
+static int timekeeping_validate_timex(const struct __kernel_timex *txc)
22552324 {
22562325 	if (txc->modes & ADJ_ADJTIME) {
22572326 		/* singleshot must not be used with any other mode bits */
....@@ -2313,13 +2382,28 @@
23132382 	return 0;
23142383 }
23152384 
2385
+/**
2386
+ * random_get_entropy_fallback - Returns the raw clock source value,
2387
+ * used by random.c for platforms with no valid random_get_entropy().
2388
+ */
2389
+unsigned long random_get_entropy_fallback(void)
2390
+{
2391
+	struct tk_read_base *tkr = &tk_core.timekeeper.tkr_mono;
2392
+	struct clocksource *clock = READ_ONCE(tkr->clock);
2393
+
2394
+	if (unlikely(timekeeping_suspended || !clock))
2395
+		return 0;
2396
+	return clock->read(clock);
2397
+}
2398
+EXPORT_SYMBOL_GPL(random_get_entropy_fallback);
23162399 
23172400 /**
23182401  * do_adjtimex() - Accessor function to NTP __do_adjtimex function
23192402  */
2320
-int do_adjtimex(struct timex *txc)
2403
+int do_adjtimex(struct __kernel_timex *txc)
23212404 {
23222405 	struct timekeeper *tk = &tk_core.timekeeper;
2406
+	struct audit_ntp_data ad;
23232407 	unsigned long flags;
23242408 	struct timespec64 ts;
23252409 	s32 orig_tai, tai;
....@@ -2329,6 +2413,7 @@
23292413 	ret = timekeeping_validate_timex(txc);
23302414 	if (ret)
23312415 		return ret;
2416
+	add_device_randomness(txc, sizeof(*txc));
23322417 
23332418 	if (txc->modes & ADJ_SETOFFSET) {
23342419 		struct timespec64 delta;
....@@ -2339,15 +2424,20 @@
23392424 		ret = timekeeping_inject_offset(&delta);
23402425 		if (ret)
23412426 			return ret;
2427
+
2428
+		audit_tk_injoffset(delta);
23422429 	}
23432430 
2431
+	audit_ntp_init(&ad);
2432
+
23442433 	ktime_get_real_ts64(&ts);
2434
+	add_device_randomness(&ts, sizeof(ts));
23452435 
23462436 	raw_spin_lock_irqsave(&timekeeper_lock, flags);
23472437 	write_seqcount_begin(&tk_core.seq);
23482438 
23492439 	orig_tai = tai = tk->tai_offset;
2350
-	ret = __do_adjtimex(txc, &ts, &tai);
2440
+	ret = __do_adjtimex(txc, &ts, &tai, &ad);
23512441 
23522442 	if (tai != orig_tai) {
23532443 		__timekeeping_set_tai_offset(tk, tai);
....@@ -2357,6 +2447,8 @@
23572447 
23582448 	write_seqcount_end(&tk_core.seq);
23592449 	raw_spin_unlock_irqrestore(&timekeeper_lock, flags);
2450
+
2451
+	audit_ntp_log(&ad);
23602452 
23612453 	/* Update the multiplier immediately if frequency was set directly */
23622454 	if (txc->modes & (ADJ_FREQUENCY | ADJ_TICK))
....@@ -2397,8 +2489,10 @@
23972489  */
23982490 void xtime_update(unsigned long ticks)
23992491 {
2400
-	write_seqlock(&jiffies_lock);
2492
+	raw_spin_lock(&jiffies_lock);
2493
+	write_seqcount_begin(&jiffies_seq);
24012494 	do_timer(ticks);
2402
-	write_sequnlock(&jiffies_lock);
2495
+	write_seqcount_end(&jiffies_seq);
2496
+	raw_spin_unlock(&jiffies_lock);
24032497 	update_wall_time();
24042498 }