Resuming timekeeping is a clock-was-set event and uses the clock-was-set
notification mechanism. This is in the way of making the clock-was-set
update for hrtimers selective so unnecessary IPIs are avoided when a CPU
base does not have timers queued which are affected by the clock setting.

Distangle it by invoking hrtimer_resume() on each unfreezing CPU and invoke
the new timerfd_resume() function from timekeeping_resume() which is the
only place where this is needed.

Rename hrtimer_resume() to hrtimer_resume_local() to reflect the change.

With this the clock_was_set*() functions are not longer required to IPI all
CPUs unconditionally and can get some smarts to avoid them.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20210713135158.488853478@linutronix.de

Some SoCs have two per-cpu timer implementations where the timer with the
higher rating stops in deep idle (i.e. suffers from CLOCK_EVT_FEAT_C3STOP)
but is otherwise preferable to the timer with the lower rating. In such a
design, selecting the higher rated devices relies on a global broadcast
timer and IPIs to wake up from deep idle states.

To avoid the reliance on a global broadcast timer and also to reduce the
overhead associated with the IPI wakeups, extend
tick_install_broadcast_device() to manage per-cpu wakeup timers separately
from the broadcast device.

For now, these timers remain unused.

Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20210524221818.15850-4-will@kernel.org

The function tick_check_replacement() is the combination of
tick_check_percpu() and tick_check_preferred(), but tick_check_new_device()
has the same logic open coded.

Use the helper to simplify the code.

[ tglx: Massage changelog ]

Signed-off-by: Wang Wensheng <wangwensheng4@huawei.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20210326022328.3266-1-wangwensheng4@huawei.com

tick_handover_do_timer() which is invoked when a CPU is unplugged has a
check for cpumask_first(cpu_online_mask) when it tries to hand over the
tick update duty.

Checking the result of cpumask_first() there is pointless because if the
online mask is empty at this point, then this would be the last CPU in the
system going offline, which is impossible. There is always at least one CPU
remaining. If online mask would be really empty then the timer duty would
be the least of the resulting problems.

Remove the well meant check simply because it is pointless and confusing.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Link: https://lore.kernel.org/r/20201206212002.582579516@linutronix.de

The variable tick_period is initialized to NSEC_PER_TICK / HZ during boot
and never updated again.

If NSEC_PER_TICK is not an integer multiple of HZ this computation is less
accurate than TICK_NSEC which has proper rounding in place.

Aside of the inaccuracy there is no reason for having this variable at
all. It's just a pointless indirection and all usage sites can just use the
TICK_NSEC constant.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20201117132006.766643526@linutronix.de

The protection rules for tick_next_period and last_jiffies_update are blury
at best. Clarify this.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20201117132006.197713794@linutronix.de

seqlock consists of a sequence counter and a spinlock_t which is used to
serialize the writers. spinlock_t is substituted by a "sleeping" spinlock
on PREEMPT_RT enabled kernels which breaks the usage in the timekeeping
code as the writers are executed in hard interrupt and therefore
non-preemptible context even on PREEMPT_RT.

The spinlock in seqlock cannot be unconditionally replaced by a
raw_spinlock_t as many seqlock users have nesting spinlock sections or
other code which is not suitable to run in truly atomic context on RT.

Instead of providing a raw_seqlock API for a single use case, open code the
seqlock for the jiffies use case and implement it with a raw_spinlock_t and
a sequence counter.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lkml.kernel.org/r/20200321113242.120587764@linutronix.de

Suspend to IDLE invokes tick_unfreeze() on resume. tick_unfreeze() on the
first resuming CPU resumes timekeeping, which also has the side effect of
resetting the softlockup watchdog on this CPU.

But on the secondary CPUs the watchdog is not reset in the resume /
unfreeze() path, which can result in false softlockup warnings on those
CPUs depending on the time spent in suspend.

Prevent this by clearing the softlock watchdog in the unfreeze path also
on the secondary resuming CPUs.

[ tglx: Massaged changelog ]

Signed-off-by: Chunyan Zhang <chunyan.zhang@unisoc.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://lore.kernel.org/r/20200110083902.27276-1-chunyan.zhang@unisoc.com

Pull timer updates from Ingo Molnar:
 "This cycle had the following changes:

   - Timer tracing improvements (Anna-Maria Gleixner)

   - Continued tasklet reduction work: remove the hrtimer_tasklet
     (Thomas Gleixner)

   - Fix CPU hotplug remove race in the tick-broadcast mask handling
     code (Thomas Gleixner)

   - Force upper bound for setting CLOCK_REALTIME, to fix ABI
     inconsistencies with handling values that are close to the maximum
     supported and the vagueness of when uptime related wraparound might
     occur. Make the consistent maximum the year 2232 across all
     relevant ABIs and APIs. (Thomas Gleixner)

   - various cleanups and smaller fixes"

* 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  tick: Fix typos in comments
  tick/broadcast: Fix warning about undefined tick_broadcast_oneshot_offline()
  timekeeping: Force upper bound for setting CLOCK_REALTIME
  timer/trace: Improve timer tracing
  timer/trace: Replace deprecated vsprintf pointer extension %pf by %ps
  timer: Move trace point to get proper index
  tick/sched: Update tick_sched struct documentation
  tick: Remove outgoing CPU from broadcast masks
  timekeeping: Consistently use unsigned int for seqcount snapshot
  softirq: Remove tasklet_hrtimer
  xfrm: Replace hrtimer tasklet with softirq hrtimer
  mac80211_hwsim: Replace hrtimer tasklet with softirq hrtimer

Allow the boot CPU/CPU0 to be nohz_full. Have the boot CPU take the
do_timer duty during boot until a housekeeping CPU can take over.

This is supported when CONFIG_PM_SLEEP_SMP is not configured, or when
it is configured and the arch allows suspend on non-zero CPUs.

nohz_full has been trialed at a large supercomputer site and found to
significantly reduce jitter. In order to deploy it in production, they
need CPU0 to be nohz_full because their job control system requires
the application CPUs to start from 0, and the housekeeping CPUs are
placed higher. An equivalent job scheduling that uses CPU0 for
housekeeping could be achieved by modifying their system, but it is
preferable if nohz_full can support their environment without
modification.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rafael J . Wysocki <rafael.j.wysocki@intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linuxppc-dev@lists.ozlabs.org
Link: https://lkml.kernel.org/r/20190411033448.20842-6-npiggin@gmail.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>