ntp: Use freezable workqueue for RTC synchronization

The bug fixed by commit e3fab2f3 ("ntp: Fix RTC synchronization on
32-bit platforms") revealed an underlying issue: RTC synchronization may
happen anytime, even while the system is partially suspended.

On systems where the RTC is connected to an I2C bus, the I2C bus controller
may already or still be suspended, triggering a WARNING during suspend or
resume from s2ram:

    WARNING: CPU: 0 PID: 124 at drivers/i2c/i2c-core.h:54 __i2c_transfer+0x634/0x680
    i2c i2c-6: Transfer while suspended
    [...]
    Workqueue: events_power_efficient sync_hw_clock
    [...]
      (__i2c_transfer)
      (i2c_transfer)
      (regmap_i2c_read)
      ...
      (da9063_rtc_set_time)
      (rtc_set_time)
      (sync_hw_clock)
      (process_one_work)

Fix this race condition by using the freezable instead of the normal
power-efficient workqueue.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Rafael J. Wysocki <rafael@kernel.org>
Link: https://lore.kernel.org/r/20210125143039.1051912-1-geert+renesas@glider.be

kernel/time/ntp.c

@@ -502,7 +502,7 @@ static struct hrtimer sync_hrtimer;
 
 static enum hrtimer_restart sync_timer_callback(struct hrtimer *timer)
 {
-	queue_work(system_power_efficient_wq, &sync_work);
+	queue_work(system_freezable_power_efficient_wq, &sync_work);
 
 	return HRTIMER_NORESTART;
 }
@@ -668,7 +668,7 @@ void ntp_notify_cmos_timer(void)
 	 * just a pointless work scheduled.
 	 */
 	if (ntp_synced() && !hrtimer_is_queued(&sync_hrtimer))
-		queue_work(system_power_efficient_wq, &sync_work);
+		queue_work(system_freezable_power_efficient_wq, &sync_work);
 }
 
 static void __init ntp_init_cmos_sync(void)