]> git.ipfire.org Git - thirdparty/kernel/linux.git/commitdiff
timekeeping: Rework timekeeping_resume() to use shadow_timekeeper
authorAnna-Maria Behnsen <anna-maria@linutronix.de>
Wed, 9 Oct 2024 08:29:14 +0000 (10:29 +0200)
committerThomas Gleixner <tglx@linutronix.de>
Fri, 25 Oct 2024 17:49:15 +0000 (19:49 +0200)
Updates of the timekeeper can be done by operating on the shadow timekeeper
and afterwards copying the result into the real timekeeper. This has the
advantage, that the sequence count write protected region is kept as small
as possible.

While the sequence count held time is not relevant for the resume path as
there is no concurrency, there is no reason to have this function
different than all the other update sites.

Convert timekeeping_inject_offset() to use this scheme and cleanup the
variable declaration while at it.

Signed-off-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: John Stultz <jstultz@google.com>
Link: https://lore.kernel.org/all/20241009-devel-anna-maria-b4-timers-ptp-timekeeping-v2-21-554456a44a15@linutronix.de
kernel/time/timekeeping.c

index 9552bc76f3863aaa41e94307cb13190e426b91f6..94f68e7ffc9df2ba86e326ecdff6e7eb94e09096 100644 (file)
@@ -1943,12 +1943,12 @@ void timekeeping_inject_sleeptime64(const struct timespec64 *delta)
  */
 void timekeeping_resume(void)
 {
-       struct timekeeper *tk = &tk_core.timekeeper;
-       struct clocksource *clock = tk->tkr_mono.clock;
-       unsigned long flags;
+       struct timekeeper *tks = &tk_core.shadow_timekeeper;
+       struct clocksource *clock = tks->tkr_mono.clock;
        struct timespec64 ts_new, ts_delta;
-       u64 cycle_now, nsec;
        bool inject_sleeptime = false;
+       u64 cycle_now, nsec;
+       unsigned long flags;
 
        read_persistent_clock64(&ts_new);
 
@@ -1956,7 +1956,6 @@ void timekeeping_resume(void)
        clocksource_resume();
 
        raw_spin_lock_irqsave(&tk_core.lock, flags);
-       write_seqcount_begin(&tk_core.seq);
 
        /*
         * After system resumes, we need to calculate the suspended time and
@@ -1970,7 +1969,7 @@ void timekeeping_resume(void)
         * The less preferred source will only be tried if there is no better
         * usable source. The rtc part is handled separately in rtc core code.
         */
-       cycle_now = tk_clock_read(&tk->tkr_mono);
+       cycle_now = tk_clock_read(&tks->tkr_mono);
        nsec = clocksource_stop_suspend_timing(clock, cycle_now);
        if (nsec > 0) {
                ts_delta = ns_to_timespec64(nsec);
@@ -1982,17 +1981,16 @@ void timekeeping_resume(void)
 
        if (inject_sleeptime) {
                suspend_timing_needed = false;
-               __timekeeping_inject_sleeptime(tk, &ts_delta);
+               __timekeeping_inject_sleeptime(tks, &ts_delta);
        }
 
        /* Re-base the last cycle value */
-       tk->tkr_mono.cycle_last = cycle_now;
-       tk->tkr_raw.cycle_last  = cycle_now;
+       tks->tkr_mono.cycle_last = cycle_now;
+       tks->tkr_raw.cycle_last  = cycle_now;
 
-       tk->ntp_error = 0;
+       tks->ntp_error = 0;
        timekeeping_suspended = 0;
-       timekeeping_update(&tk_core, tk, TK_MIRROR | TK_CLOCK_WAS_SET);
-       write_seqcount_end(&tk_core.seq);
+       timekeeping_update_from_shadow(&tk_core, TK_CLOCK_WAS_SET);
        raw_spin_unlock_irqrestore(&tk_core.lock, flags);
 
        touch_softlockup_watchdog();