1 files changed, 471 insertions, 205 deletions

diff --git a/kernel/time/posix-cpu-timers.c b/kernel/time/posix-cpu-timers.c
index 165117996ea0..50e8d04ab661 100644
--- a/kernel/time/posix-cpu-timers.c
+++ b/kernel/time/posix-cpu-timers.c
@@ -15,6 +15,7 @@
 #include <linux/workqueue.h>
 #include <linux/compat.h>
 #include <linux/sched/deadline.h>
+#include <linux/task_work.h>
 
 #include "posix-timers.h"
 
@@ -34,14 +35,20 @@ void posix_cputimers_group_init(struct posix_cputimers *pct, u64 cpu_limit)
  * tsk->signal->posix_cputimers.bases[clock].nextevt expiration cache if
  * necessary. Needs siglock protection since other code may update the
  * expiration cache as well.
+ *
+ * Returns 0 on success, -ESRCH on failure.  Can fail if the task is exiting and
+ * we cannot lock_task_sighand.  Cannot fail if task is current.
  */
-void update_rlimit_cpu(struct task_struct *task, unsigned long rlim_new)
+int update_rlimit_cpu(struct task_struct *task, unsigned long rlim_new)
 {
 	u64 nsecs = rlim_new * NSEC_PER_SEC;
+	unsigned long irq_fl;
 
-	spin_lock_irq(&task->sighand->siglock);
+	if (!lock_task_sighand(task, &irq_fl))
+		return -ESRCH;
 	set_process_cpu_timer(task, CPUCLOCK_PROF, &nsecs, NULL);
-	spin_unlock_irq(&task->sighand->siglock);
+	unlock_task_sighand(task, &irq_fl);
+	return 0;
 }
 
 /*
@@ -236,13 +243,12 @@ static void proc_sample_cputime_atomic(struct task_cputime_atomic *at,
  */
 static inline void __update_gt_cputime(atomic64_t *cputime, u64 sum_cputime)
 {
-	u64 curr_cputime;
-retry:
-	curr_cputime = atomic64_read(cputime);
-	if (sum_cputime > curr_cputime) {
-		if (atomic64_cmpxchg(cputime, curr_cputime, sum_cputime) != curr_cputime)
-			goto retry;
-	}
+	u64 curr_cputime = atomic64_read(cputime);
+
+	do {
+		if (sum_cputime <= curr_cputime)
+			return;
+	} while (!atomic64_try_cmpxchg(cputime, &curr_cputime, sum_cputime));
 }
 
 static void update_gt_cputime(struct task_cputime_atomic *cputime_atomic,
@@ -279,7 +285,7 @@ void thread_group_sample_cputime(struct task_struct *tsk, u64 *samples)
  * @tsk:	Task for which cputime needs to be started
  * @samples:	Storage for time samples
  *
- * The thread group cputime accouting is avoided when there are no posix
+ * The thread group cputime accounting is avoided when there are no posix
  * CPU timers armed. Before starting a timer it's required to check whether
  * the time accounting is active. If not, a full update of the atomic
  * accounting store needs to be done and the accounting enabled.
@@ -291,6 +297,8 @@ static void thread_group_start_cputime(struct task_struct *tsk, u64 *samples)
 	struct thread_group_cputimer *cputimer = &tsk->signal->cputimer;
 	struct posix_cputimers *pct = &tsk->signal->posix_cputimers;
 
+	lockdep_assert_task_sighand_held(tsk);
+
 	/* Check if cputimer isn't running. This is accessed without locking. */
 	if (!READ_ONCE(pct->timers_active)) {
 		struct task_cputime sum;
@@ -377,6 +385,7 @@ static int posix_cpu_clock_get(const clockid_t clock, struct timespec64 *tp)
  */
 static int posix_cpu_timer_create(struct k_itimer *new_timer)
 {
+	static struct lock_class_key posix_cpu_timers_key;
 	struct pid *pid;
 
 	rcu_read_lock();
@@ -386,6 +395,17 @@ static int posix_cpu_timer_create(struct k_itimer *new_timer)
 		return -EINVAL;
 	}
 
+	/*
+	 * If posix timer expiry is handled in task work context then
+	 * timer::it_lock can be taken without disabling interrupts as all
+	 * other locking happens in task context. This requires a separate
+	 * lock class key otherwise regular posix timer expiry would record
+	 * the lock class being taken in interrupt context and generate a
+	 * false positive warning.
+	 */
+	if (IS_ENABLED(CONFIG_POSIX_CPU_TIMERS_TASK_WORK))
+		lockdep_set_class(&new_timer->it_lock, &posix_cpu_timers_key);
+
 	new_timer->kclock = &clock_posix_cpu;
 	timerqueue_init(&new_timer->it.cpu.node);
 	new_timer->it.cpu.pid = get_pid(pid);
@@ -393,6 +413,55 @@ static int posix_cpu_timer_create(struct k_itimer *new_timer)
 	return 0;
 }
 
+static struct posix_cputimer_base *timer_base(struct k_itimer *timer,
+					      struct task_struct *tsk)
+{
+	int clkidx = CPUCLOCK_WHICH(timer->it_clock);
+
+	if (CPUCLOCK_PERTHREAD(timer->it_clock))
+		return tsk->posix_cputimers.bases + clkidx;
+	else
+		return tsk->signal->posix_cputimers.bases + clkidx;
+}
+
+/*
+ * Force recalculating the base earliest expiration on the next tick.
+ * This will also re-evaluate the need to keep around the process wide
+ * cputime counter and tick dependency and eventually shut these down
+ * if necessary.
+ */
+static void trigger_base_recalc_expires(struct k_itimer *timer,
+					struct task_struct *tsk)
+{
+	struct posix_cputimer_base *base = timer_base(timer, tsk);
+
+	base->nextevt = 0;
+}
+
+/*
+ * Dequeue the timer and reset the base if it was its earliest expiration.
+ * It makes sure the next tick recalculates the base next expiration so we
+ * don't keep the costly process wide cputime counter around for a random
+ * amount of time, along with the tick dependency.
+ *
+ * If another timer gets queued between this and the next tick, its
+ * expiration will update the base next event if necessary on the next
+ * tick.
+ */
+static void disarm_timer(struct k_itimer *timer, struct task_struct *p)
+{
+	struct cpu_timer *ctmr = &timer->it.cpu;
+	struct posix_cputimer_base *base;
+
+	if (!cpu_timer_dequeue(ctmr))
+		return;
+
+	base = timer_base(timer, p);
+	if (cpu_timer_getexpires(ctmr) == base->nextevt)
+		trigger_base_recalc_expires(timer, p);
+}
+
+
 /*
  * Clean up a CPU-clock timer that is about to be destroyed.
  * This is called from timer deletion with the timer already locked.
@@ -424,19 +493,28 @@ static int posix_cpu_timer_del(struct k_itimer *timer)
 		 */
 		WARN_ON_ONCE(ctmr->head || timerqueue_node_queued(&ctmr->node));
 	} else {
-		if (timer->it.cpu.firing)
+		if (timer->it.cpu.firing) {
+			/*
+			 * Prevent signal delivery. The timer cannot be dequeued
+			 * because it is on the firing list which is not protected
+			 * by sighand->lock. The delivery path is waiting for
+			 * the timer lock. So go back, unlock and retry.
+			 */
+			timer->it.cpu.firing = false;
 			ret = TIMER_RETRY;
-		else
-			cpu_timer_dequeue(ctmr);
-
+		} else {
+			disarm_timer(timer, p);
+		}
 		unlock_task_sighand(p, &flags);
 	}
 
 out:
 	rcu_read_unlock();
-	if (!ret)
-		put_pid(ctmr->pid);
 
+	if (!ret) {
+		put_pid(ctmr->pid);
+		timer->it_status = POSIX_TIMER_DISARMED;
+	}
 	return ret;
 }
 
@@ -486,16 +564,11 @@ void posix_cpu_timers_exit_group(struct task_struct *tsk)
  */
 static void arm_timer(struct k_itimer *timer, struct task_struct *p)
 {
-	int clkidx = CPUCLOCK_WHICH(timer->it_clock);
+	struct posix_cputimer_base *base = timer_base(timer, p);
 	struct cpu_timer *ctmr = &timer->it.cpu;
 	u64 newexp = cpu_timer_getexpires(ctmr);
-	struct posix_cputimer_base *base;
-
-	if (CPUCLOCK_PERTHREAD(timer->it_clock))
-		base = p->posix_cputimers.bases + clkidx;
-	else
-		base = p->signal->posix_cputimers.bases + clkidx;
 
+	timer->it_status = POSIX_TIMER_ARMED;
 	if (!cpu_timer_enqueue(&base->tqhead, ctmr))
 		return;
 
@@ -511,7 +584,7 @@ static void arm_timer(struct k_itimer *timer, struct task_struct *p)
 	if (CPUCLOCK_PERTHREAD(timer->it_clock))
 		tick_dep_set_task(p, TICK_DEP_BIT_POSIX_TIMER);
 	else
-		tick_dep_set_signal(p->signal, TICK_DEP_BIT_POSIX_TIMER);
+		tick_dep_set_signal(p, TICK_DEP_BIT_POSIX_TIMER);
 }
 
 /*
@@ -521,36 +594,25 @@ static void cpu_timer_fire(struct k_itimer *timer)
 {
 	struct cpu_timer *ctmr = &timer->it.cpu;
 
-	if ((timer->it_sigev_notify & ~SIGEV_THREAD_ID) == SIGEV_NONE) {
-		/*
-		 * User don't want any signal.
-		 */
-		cpu_timer_setexpires(ctmr, 0);
-	} else if (unlikely(timer->sigq == NULL)) {
+	timer->it_status = POSIX_TIMER_DISARMED;
+
+	if (unlikely(ctmr->nanosleep)) {
 		/*
 		 * This a special case for clock_nanosleep,
 		 * not a normal timer from sys_timer_create.
 		 */
 		wake_up_process(timer->it_process);
 		cpu_timer_setexpires(ctmr, 0);
-	} else if (!timer->it_interval) {
-		/*
-		 * One-shot timer.  Clear it as soon as it's fired.
-		 */
-		posix_timer_event(timer, 0);
-		cpu_timer_setexpires(ctmr, 0);
-	} else if (posix_timer_event(timer, ++timer->it_requeue_pending)) {
-		/*
-		 * The signal did not get queued because the signal
-		 * was ignored, so we won't get any callback to
-		 * reload the timer.  But we need to keep it
-		 * ticking in case the signal is deliverable next time.
-		 */
-		posix_cpu_timer_rearm(timer);
-		++timer->it_requeue_pending;
+	} else {
+		posix_timer_queue_signal(timer);
+		/* Disable oneshot timers */
+		if (!timer->it_interval)
+			cpu_timer_setexpires(ctmr, 0);
 	}
 }
 
+static void __posix_cpu_timer_get(struct k_itimer *timer, struct itimerspec64 *itp, u64 now);
+
 /*
  * Guts of sys_timer_settime for CPU timers.
  * This is called with the timer locked and interrupts disabled.
@@ -560,9 +622,10 @@ static void cpu_timer_fire(struct k_itimer *timer)
 static int posix_cpu_timer_set(struct k_itimer *timer, int timer_flags,
 			       struct itimerspec64 *new, struct itimerspec64 *old)
 {
+	bool sigev_none = timer->it_sigev_notify == SIGEV_NONE;
 	clockid_t clkid = CPUCLOCK_WHICH(timer->it_clock);
-	u64 old_expires, new_expires, old_incr, val;
 	struct cpu_timer *ctmr = &timer->it.cpu;
+	u64 old_expires, new_expires, now;
 	struct sighand_struct *sighand;
 	struct task_struct *p;
 	unsigned long flags;
@@ -599,155 +662,136 @@ static int posix_cpu_timer_set(struct k_itimer *timer, int timer_flags,
 		return -ESRCH;
 	}
 
-	/*
-	 * Disarm any old timer after extracting its expiry time.
-	 */
-	old_incr = timer->it_interval;
+	/* Retrieve the current expiry time before disarming the timer */
 	old_expires = cpu_timer_getexpires(ctmr);
 
 	if (unlikely(timer->it.cpu.firing)) {
-		timer->it.cpu.firing = -1;
+		/*
+		 * Prevent signal delivery. The timer cannot be dequeued
+		 * because it is on the firing list which is not protected
+		 * by sighand->lock. The delivery path is waiting for
+		 * the timer lock. So go back, unlock and retry.
+		 */
+		timer->it.cpu.firing = false;
 		ret = TIMER_RETRY;
 	} else {
 		cpu_timer_dequeue(ctmr);
+		timer->it_status = POSIX_TIMER_DISARMED;
 	}
 
 	/*
-	 * We need to sample the current value to convert the new
-	 * value from to relative and absolute, and to convert the
-	 * old value from absolute to relative.  To set a process
-	 * timer, we need a sample to balance the thread expiry
-	 * times (in arm_timer).  With an absolute time, we must
-	 * check if it's already passed.  In short, we need a sample.
+	 * Sample the current clock for saving the previous setting
+	 * and for rearming the timer.
 	 */
 	if (CPUCLOCK_PERTHREAD(timer->it_clock))
-		val = cpu_clock_sample(clkid, p);
+		now = cpu_clock_sample(clkid, p);
 	else
-		val = cpu_clock_sample_group(clkid, p, true);
+		now = cpu_clock_sample_group(clkid, p, !sigev_none);
 
+	/* Retrieve the previous expiry value if requested. */
 	if (old) {
-		if (old_expires == 0) {
-			old->it_value.tv_sec = 0;
-			old->it_value.tv_nsec = 0;
-		} else {
-			/*
-			 * Update the timer in case it has overrun already.
-			 * If it has, we'll report it as having overrun and
-			 * with the next reloaded timer already ticking,
-			 * though we are swallowing that pending
-			 * notification here to install the new setting.
-			 */
-			u64 exp = bump_cpu_timer(timer, val);
-
-			if (val < exp) {
-				old_expires = exp - val;
-				old->it_value = ns_to_timespec64(old_expires);
-			} else {
-				old->it_value.tv_nsec = 1;
-				old->it_value.tv_sec = 0;
-			}
-		}
+		old->it_value = (struct timespec64){ };
+		if (old_expires)
+			__posix_cpu_timer_get(timer, old, now);
 	}
 
+	/* Retry if the timer expiry is running concurrently */
 	if (unlikely(ret)) {
-		/*
-		 * We are colliding with the timer actually firing.
-		 * Punt after filling in the timer's old value, and
-		 * disable this firing since we are already reporting
-		 * it as an overrun (thanks to bump_cpu_timer above).
-		 */
 		unlock_task_sighand(p, &flags);
 		goto out;
 	}
 
-	if (new_expires != 0 && !(timer_flags & TIMER_ABSTIME)) {
-		new_expires += val;
-	}
+	/* Convert relative expiry time to absolute */
+	if (new_expires && !(timer_flags & TIMER_ABSTIME))
+		new_expires += now;
+
+	/* Set the new expiry time (might be 0) */
+	cpu_timer_setexpires(ctmr, new_expires);
 
 	/*
-	 * Install the new expiry time (or zero).
-	 * For a timer with no notification action, we don't actually
-	 * arm the timer (we'll just fake it for timer_gettime).
+	 * Arm the timer if it is not disabled, the new expiry value has
+	 * not yet expired and the timer requires signal delivery.
+	 * SIGEV_NONE timers are never armed. In case the timer is not
+	 * armed, enforce the reevaluation of the timer base so that the
+	 * process wide cputime counter can be disabled eventually.
 	 */
-	cpu_timer_setexpires(ctmr, new_expires);
-	if (new_expires != 0 && val < new_expires) {
-		arm_timer(timer, p);
+	if (likely(!sigev_none)) {
+		if (new_expires && now < new_expires)
+			arm_timer(timer, p);
+		else
+			trigger_base_recalc_expires(timer, p);
 	}
 
 	unlock_task_sighand(p, &flags);
-	/*
-	 * Install the new reload setting, and
-	 * set up the signal and overrun bookkeeping.
-	 */
-	timer->it_interval = timespec64_to_ktime(new->it_interval);
+
+	posix_timer_set_common(timer, new);
 
 	/*
-	 * This acts as a modification timestamp for the timer,
-	 * so any automatic reload attempt will punt on seeing
-	 * that we have reset the timer manually.
+	 * If the new expiry time was already in the past the timer was not
+	 * queued. Fire it immediately even if the thread never runs to
+	 * accumulate more time on this clock.
 	 */
-	timer->it_requeue_pending = (timer->it_requeue_pending + 2) &
-		~REQUEUE_PENDING;
-	timer->it_overrun_last = 0;
-	timer->it_overrun = -1;
-
-	if (new_expires != 0 && !(val < new_expires)) {
-		/*
-		 * The designated time already passed, so we notify
-		 * immediately, even if the thread never runs to
-		 * accumulate more time on this clock.
-		 */
+	if (!sigev_none && new_expires && now >= new_expires)
 		cpu_timer_fire(timer);
-	}
-
-	ret = 0;
- out:
+out:
 	rcu_read_unlock();
-	if (old)
-		old->it_interval = ns_to_timespec64(old_incr);
-
 	return ret;
 }
 
-static void posix_cpu_timer_get(struct k_itimer *timer, struct itimerspec64 *itp)
+static void __posix_cpu_timer_get(struct k_itimer *timer, struct itimerspec64 *itp, u64 now)
 {
-	clockid_t clkid = CPUCLOCK_WHICH(timer->it_clock);
-	struct cpu_timer *ctmr = &timer->it.cpu;
-	u64 now, expires = cpu_timer_getexpires(ctmr);
-	struct task_struct *p;
-
-	rcu_read_lock();
-	p = cpu_timer_task_rcu(timer);
-	if (!p)
-		goto out;
+	bool sigev_none = timer->it_sigev_notify == SIGEV_NONE;
+	u64 expires, iv = timer->it_interval;
 
 	/*
-	 * Easy part: convert the reload time.
+	 * Make sure that interval timers are moved forward for the
+	 * following cases:
+	 *  - SIGEV_NONE timers which are never armed
+	 *  - Timers which expired, but the signal has not yet been
+	 *    delivered
 	 */
-	itp->it_interval = ktime_to_timespec64(timer->it_interval);
-
-	if (!expires)
-		goto out;
+	if (iv && timer->it_status != POSIX_TIMER_ARMED)
+		expires = bump_cpu_timer(timer, now);
+	else
+		expires = cpu_timer_getexpires(&timer->it.cpu);
 
 	/*
-	 * Sample the clock to take the difference with the expiry time.
+	 * Expired interval timers cannot have a remaining time <= 0.
+	 * The kernel has to move them forward so that the next
+	 * timer expiry is > @now.
 	 */
-	if (CPUCLOCK_PERTHREAD(timer->it_clock))
-		now = cpu_clock_sample(clkid, p);
-	else
-		now = cpu_clock_sample_group(clkid, p, false);
-
 	if (now < expires) {
 		itp->it_value = ns_to_timespec64(expires - now);
 	} else {
 		/*
-		 * The timer should have expired already, but the firing
-		 * hasn't taken place yet.  Say it's just about to expire.
+		 * A single shot SIGEV_NONE timer must return 0, when it is
+		 * expired! Timers which have a real signal delivery mode
+		 * must return a remaining time greater than 0 because the
+		 * signal has not yet been delivered.
 		 */
-		itp->it_value.tv_nsec = 1;
-		itp->it_value.tv_sec = 0;
+		if (!sigev_none)
+			itp->it_value.tv_nsec = 1;
+	}
+}
+
+static void posix_cpu_timer_get(struct k_itimer *timer, struct itimerspec64 *itp)
+{
+	clockid_t clkid = CPUCLOCK_WHICH(timer->it_clock);
+	struct task_struct *p;
+	u64 now;
+
+	rcu_read_lock();
+	p = cpu_timer_task_rcu(timer);
+	if (p && cpu_timer_getexpires(&timer->it.cpu)) {
+		itp->it_interval = ktime_to_timespec64(timer->it_interval);
+
+		if (CPUCLOCK_PERTHREAD(timer->it_clock))
+			now = cpu_clock_sample(clkid, p);
+		else
+			now = cpu_clock_sample_group(clkid, p, false);
+
+		__posix_cpu_timer_get(timer, itp, now);
 	}
-out:
 	rcu_read_unlock();
 }
 
@@ -769,7 +813,9 @@ static u64 collect_timerqueue(struct timerqueue_head *head,
 		if (++i == MAX_COLLECTED || now < expires)
 			return expires;
 
-		ctmr->firing = 1;
+		ctmr->firing = true;
+		/* See posix_cpu_timer_wait_running() */
+		rcu_assign_pointer(ctmr->handling, current);
 		cpu_timer_dequeue(ctmr);
 		list_add_tail(&ctmr->elist, firing);
 	}
@@ -793,7 +839,7 @@ static inline void check_dl_overrun(struct task_struct *tsk)
 {
 	if (tsk->dl.dl_overrun) {
 		tsk->dl.dl_overrun = 0;
-		__group_send_sig_info(SIGXCPU, SEND_SIG_PRIV, tsk);
+		send_signal_locked(SIGXCPU, SEND_SIG_PRIV, tsk, PIDTYPE_TGID);
 	}
 }
 
@@ -807,7 +853,7 @@ static bool check_rlimit(u64 time, u64 limit, int signo, bool rt, bool hard)
 			rt ? "RT" : "CPU", hard ? "hard" : "soft",
 			current->comm, task_pid_nr(current));
 	}
-	__group_send_sig_info(signo, SEND_SIG_PRIV, current);
+	send_signal_locked(signo, SEND_SIG_PRIV, current, PIDTYPE_TGID);
 	return true;
 }
 
@@ -881,7 +927,7 @@ static void check_cpu_itimer(struct task_struct *tsk, struct cpu_itimer *it,
 		trace_itimer_expire(signo == SIGPROF ?
 				    ITIMER_PROF : ITIMER_VIRTUAL,
 				    task_tgid(tsk), cur_time);
-		__group_send_sig_info(signo, SEND_SIG_PRIV, tsk);
+		send_signal_locked(signo, SEND_SIG_PRIV, tsk, PIDTYPE_TGID);
 	}
 
 	if (it->expires && it->expires < *expires)
@@ -979,6 +1025,11 @@ static void posix_cpu_timer_rearm(struct k_itimer *timer)
 	if (!p)
 		goto out;
 
+	/* Protect timer list r/w in arm_timer() */
+	sighand = lock_task_sighand(p, &flags);
+	if (unlikely(sighand == NULL))
+		goto out;
+
 	/*
 	 * Fetch the current sample and update the timer's expiry time.
 	 */
@@ -989,11 +1040,6 @@ static void posix_cpu_timer_rearm(struct k_itimer *timer)
 
 	bump_cpu_timer(timer, now);
 
-	/* Protect timer list r/w in arm_timer() */
-	sighand = lock_task_sighand(p, &flags);
-	if (unlikely(sighand == NULL))
-		goto out;
-
 	/*
 	 * Now re-arm for the new expiry time.
 	 */
@@ -1080,43 +1126,233 @@ static inline bool fastpath_timer_check(struct task_struct *tsk)
 	return false;
 }
 
+static void handle_posix_cpu_timers(struct task_struct *tsk);
+
+#ifdef CONFIG_POSIX_CPU_TIMERS_TASK_WORK
+static void posix_cpu_timers_work(struct callback_head *work)
+{
+	struct posix_cputimers_work *cw = container_of(work, typeof(*cw), work);
+
+	mutex_lock(&cw->mutex);
+	handle_posix_cpu_timers(current);
+	mutex_unlock(&cw->mutex);
+}
+
 /*
- * This is called from the timer interrupt handler.  The irq handler has
- * already updated our counts.  We need to check if any timers fire now.
- * Interrupts are disabled.
+ * Invoked from the posix-timer core when a cancel operation failed because
+ * the timer is marked firing. The caller holds rcu_read_lock(), which
+ * protects the timer and the task which is expiring it from being freed.
  */
-void run_posix_cpu_timers(void)
+static void posix_cpu_timer_wait_running(struct k_itimer *timr)
 {
-	struct task_struct *tsk = current;
-	struct k_itimer *timer, *next;
-	unsigned long flags;
-	LIST_HEAD(firing);
+	struct task_struct *tsk = rcu_dereference(timr->it.cpu.handling);
 
-	lockdep_assert_irqs_disabled();
+	/* Has the handling task completed expiry already? */
+	if (!tsk)
+		return;
 
+	/* Ensure that the task cannot go away */
+	get_task_struct(tsk);
+	/* Now drop the RCU protection so the mutex can be locked */
+	rcu_read_unlock();
+	/* Wait on the expiry mutex */
+	mutex_lock(&tsk->posix_cputimers_work.mutex);
+	/* Release it immediately again. */
+	mutex_unlock(&tsk->posix_cputimers_work.mutex);
+	/* Drop the task reference. */
+	put_task_struct(tsk);
+	/* Relock RCU so the callsite is balanced */
+	rcu_read_lock();
+}
+
+static void posix_cpu_timer_wait_running_nsleep(struct k_itimer *timr)
+{
+	/* Ensure that timr->it.cpu.handling task cannot go away */
+	rcu_read_lock();
+	spin_unlock_irq(&timr->it_lock);
+	posix_cpu_timer_wait_running(timr);
+	rcu_read_unlock();
+	/* @timr is on stack and is valid */
+	spin_lock_irq(&timr->it_lock);
+}
+
+/*
+ * Clear existing posix CPU timers task work.
+ */
+void clear_posix_cputimers_work(struct task_struct *p)
+{
 	/*
-	 * The fast path checks that there are no expired thread or thread
-	 * group timers.  If that's so, just return.
+	 * A copied work entry from the old task is not meaningful, clear it.
+	 * N.B. init_task_work will not do this.
 	 */
-	if (!fastpath_timer_check(tsk))
-		return;
+	memset(&p->posix_cputimers_work.work, 0,
+	       sizeof(p->posix_cputimers_work.work));
+	init_task_work(&p->posix_cputimers_work.work,
+		       posix_cpu_timers_work);
+	mutex_init(&p->posix_cputimers_work.mutex);
+	p->posix_cputimers_work.scheduled = false;
+}
 
-	lockdep_posixtimer_enter();
-	if (!lock_task_sighand(tsk, &flags)) {
-		lockdep_posixtimer_exit();
+/*
+ * Initialize posix CPU timers task work in init task. Out of line to
+ * keep the callback static and to avoid header recursion hell.
+ */
+void __init posix_cputimers_init_work(void)
+{
+	clear_posix_cputimers_work(current);
+}
+
+/*
+ * Note: All operations on tsk->posix_cputimer_work.scheduled happen either
+ * in hard interrupt context or in task context with interrupts
+ * disabled. Aside of that the writer/reader interaction is always in the
+ * context of the current task, which means they are strict per CPU.
+ */
+static inline bool posix_cpu_timers_work_scheduled(struct task_struct *tsk)
+{
+	return tsk->posix_cputimers_work.scheduled;
+}
+
+static inline void __run_posix_cpu_timers(struct task_struct *tsk)
+{
+	if (WARN_ON_ONCE(tsk->posix_cputimers_work.scheduled))
 		return;
+
+	/* Schedule task work to actually expire the timers */
+	tsk->posix_cputimers_work.scheduled = true;
+	task_work_add(tsk, &tsk->posix_cputimers_work.work, TWA_RESUME);
+}
+
+static inline bool posix_cpu_timers_enable_work(struct task_struct *tsk,
+						unsigned long start)
+{
+	bool ret = true;
+
+	/*
+	 * On !RT kernels interrupts are disabled while collecting expired
+	 * timers, so no tick can happen and the fast path check can be
+	 * reenabled without further checks.
+	 */
+	if (!IS_ENABLED(CONFIG_PREEMPT_RT)) {
+		tsk->posix_cputimers_work.scheduled = false;
+		return true;
 	}
+
 	/*
-	 * Here we take off tsk->signal->cpu_timers[N] and
-	 * tsk->cpu_timers[N] all the timers that are firing, and
-	 * put them on the firing list.
+	 * On RT enabled kernels ticks can happen while the expired timers
+	 * are collected under sighand lock. But any tick which observes
+	 * the CPUTIMERS_WORK_SCHEDULED bit set, does not run the fastpath
+	 * checks. So reenabling the tick work has do be done carefully:
+	 *
+	 * Disable interrupts and run the fast path check if jiffies have
+	 * advanced since the collecting of expired timers started. If
+	 * jiffies have not advanced or the fast path check did not find
+	 * newly expired timers, reenable the fast path check in the timer
+	 * interrupt. If there are newly expired timers, return false and
+	 * let the collection loop repeat.
 	 */
-	check_thread_timers(tsk, &firing);
+	local_irq_disable();
+	if (start != jiffies && fastpath_timer_check(tsk))
+		ret = false;
+	else
+		tsk->posix_cputimers_work.scheduled = false;
+	local_irq_enable();
+
+	return ret;
+}
+#else /* CONFIG_POSIX_CPU_TIMERS_TASK_WORK */
+static inline void __run_posix_cpu_timers(struct task_struct *tsk)
+{
+	lockdep_posixtimer_enter();
+	handle_posix_cpu_timers(tsk);
+	lockdep_posixtimer_exit();
+}
+
+static void posix_cpu_timer_wait_running(struct k_itimer *timr)
+{
+	cpu_relax();
+}
 
-	check_process_timers(tsk, &firing);
+static void posix_cpu_timer_wait_running_nsleep(struct k_itimer *timr)
+{
+	spin_unlock_irq(&timr->it_lock);
+	cpu_relax();
+	spin_lock_irq(&timr->it_lock);
+}
+
+static inline bool posix_cpu_timers_work_scheduled(struct task_struct *tsk)
+{
+	return false;
+}
+
+static inline bool posix_cpu_timers_enable_work(struct task_struct *tsk,
+						unsigned long start)
+{
+	return true;
+}
+#endif /* CONFIG_POSIX_CPU_TIMERS_TASK_WORK */
+
+static void handle_posix_cpu_timers(struct task_struct *tsk)
+{
+	struct k_itimer *timer, *next;
+	unsigned long flags, start;
+	LIST_HEAD(firing);
+
+	if (!lock_task_sighand(tsk, &flags))
+		return;
+
+	do {
+		/*
+		 * On RT locking sighand lock does not disable interrupts,
+		 * so this needs to be careful vs. ticks. Store the current
+		 * jiffies value.
+		 */
+		start = READ_ONCE(jiffies);
+		barrier();
+
+		/*
+		 * Here we take off tsk->signal->cpu_timers[N] and
+		 * tsk->cpu_timers[N] all the timers that are firing, and
+		 * put them on the firing list.
+		 */
+		check_thread_timers(tsk, &firing);
+
+		check_process_timers(tsk, &firing);
+
+		/*
+		 * The above timer checks have updated the expiry cache and
+		 * because nothing can have queued or modified timers after
+		 * sighand lock was taken above it is guaranteed to be
+		 * consistent. So the next timer interrupt fastpath check
+		 * will find valid data.
+		 *
+		 * If timer expiry runs in the timer interrupt context then
+		 * the loop is not relevant as timers will be directly
+		 * expired in interrupt context. The stub function below
+		 * returns always true which allows the compiler to
+		 * optimize the loop out.
+		 *
+		 * If timer expiry is deferred to task work context then
+		 * the following rules apply:
+		 *
+		 * - On !RT kernels no tick can have happened on this CPU
+		 *   after sighand lock was acquired because interrupts are
+		 *   disabled. So reenabling task work before dropping
+		 *   sighand lock and reenabling interrupts is race free.
+		 *
+		 * - On RT kernels ticks might have happened but the tick
+		 *   work ignored posix CPU timer handling because the
+		 *   CPUTIMERS_WORK_SCHEDULED bit is set. Reenabling work
+		 *   must be done very carefully including a check whether
+		 *   ticks have happened since the start of the timer
+		 *   expiry checks. posix_cpu_timers_enable_work() takes
+		 *   care of that and eventually lets the expiry checks
+		 *   run again.
+		 */
+	} while (!posix_cpu_timers_enable_work(tsk, start));
 
 	/*
-	 * We must release these locks before taking any timer's lock.
+	 * We must release sighand lock before taking any timer's lock.
 	 * There is a potential race with timer deletion here, as the
 	 * siglock now protects our private firing list.  We have set
 	 * the firing flag in each timer, so that a deletion attempt
@@ -1132,22 +1368,58 @@ void run_posix_cpu_timers(void)
 	 * timer call will interfere.
 	 */
 	list_for_each_entry_safe(timer, next, &firing, it.cpu.elist) {
-		int cpu_firing;
+		bool cpu_firing;
 
+		/*
+		 * spin_lock() is sufficient here even independent of the
+		 * expiry context. If expiry happens in hard interrupt
+		 * context it's obvious. For task work context it's safe
+		 * because all other operations on timer::it_lock happen in
+		 * task context (syscall or exit).
+		 */
 		spin_lock(&timer->it_lock);
 		list_del_init(&timer->it.cpu.elist);
 		cpu_firing = timer->it.cpu.firing;
-		timer->it.cpu.firing = 0;
+		timer->it.cpu.firing = false;
 		/*
-		 * The firing flag is -1 if we collided with a reset
-		 * of the timer, which already reported this
-		 * almost-firing as an overrun.  So don't generate an event.
+		 * If the firing flag is cleared then this raced with a
+		 * timer rearm/delete operation. So don't generate an
+		 * event.
 		 */
-		if (likely(cpu_firing >= 0))
+		if (likely(cpu_firing))
 			cpu_timer_fire(timer);
+		/* See posix_cpu_timer_wait_running() */
+		rcu_assign_pointer(timer->it.cpu.handling, NULL);
 		spin_unlock(&timer->it_lock);
 	}
-	lockdep_posixtimer_exit();
+}
+
+/*
+ * This is called from the timer interrupt handler.  The irq handler has
+ * already updated our counts.  We need to check if any timers fire now.
+ * Interrupts are disabled.
+ */
+void run_posix_cpu_timers(void)
+{
+	struct task_struct *tsk = current;
+
+	lockdep_assert_irqs_disabled();
+
+	/*
+	 * If the actual expiry is deferred to task work context and the
+	 * work is already scheduled there is no point to do anything here.
+	 */
+	if (posix_cpu_timers_work_scheduled(tsk))
+		return;
+
+	/*
+	 * The fast path checks that there are no expired thread or thread
+	 * group timers.  If that's so, just return.
+	 */
+	if (!fastpath_timer_check(tsk))
+		return;
+
+	__run_posix_cpu_timers(tsk);
 }
 
 /*
@@ -1180,9 +1452,8 @@ void set_process_cpu_timer(struct task_struct *tsk, unsigned int clkid,
 			}
 		}
 
-		if (!*newval)
-			return;
-		*newval += now;
+		if (*newval)
+			*newval += now;
 	}
 
 	/*
@@ -1192,7 +1463,7 @@ void set_process_cpu_timer(struct task_struct *tsk, unsigned int clkid,
 	if (*newval < *nextevt)
 		*nextevt = *newval;
 
-	tick_dep_set_signal(tsk->signal, TICK_DEP_BIT_POSIX_TIMER);
+	tick_dep_set_signal(tsk, TICK_DEP_BIT_POSIX_TIMER);
 }
 
 static int do_cpu_nanosleep(const clockid_t which_clock, int flags,
@@ -1212,6 +1483,7 @@ static int do_cpu_nanosleep(const clockid_t which_clock, int flags,
 	timer.it_overrun = -1;
 	error = posix_cpu_timer_create(&timer);
 	timer.it_process = current;
+	timer.it.cpu.nanosleep = true;
 
 	if (!error) {
 		static struct itimerspec64 zero_it;
@@ -1253,23 +1525,16 @@ static int do_cpu_nanosleep(const clockid_t which_clock, int flags,
 		expires = cpu_timer_getexpires(&timer.it.cpu);
 		error = posix_cpu_timer_set(&timer, 0, &zero_it, &it);
 		if (!error) {
-			/*
-			 * Timer is now unarmed, deletion can not fail.
-			 */
+			/* Timer is now unarmed, deletion can not fail. */
 			posix_cpu_timer_del(&timer);
+		} else {
+			while (error == TIMER_RETRY) {
+				posix_cpu_timer_wait_running_nsleep(&timer);
+				error = posix_cpu_timer_del(&timer);
+			}
 		}
-		spin_unlock_irq(&timer.it_lock);
 
-		while (error == TIMER_RETRY) {
-			/*
-			 * We need to handle case when timer was or is in the
-			 * middle of firing. In other cases we already freed
-			 * resources.
-			 */
-			spin_lock_irq(&timer.it_lock);
-			error = posix_cpu_timer_del(&timer);
-			spin_unlock_irq(&timer.it_lock);
-		}
+		spin_unlock_irq(&timer.it_lock);
 
 		if ((it.it_value.tv_sec | it.it_value.tv_nsec) == 0) {
 			/*
@@ -1314,8 +1579,8 @@ static int posix_cpu_nsleep(const clockid_t which_clock, int flags,
 		if (flags & TIMER_ABSTIME)
 			return -ERESTARTNOHAND;
 
-		restart_block->fn = posix_cpu_nsleep_restart;
 		restart_block->nanosleep.clockid = which_clock;
+		set_restart_fn(restart_block, posix_cpu_nsleep_restart);
 	}
 	return error;
 }
@@ -1379,6 +1644,7 @@ const struct k_clock clock_posix_cpu = {
 	.timer_del		= posix_cpu_timer_del,
 	.timer_get		= posix_cpu_timer_get,
 	.timer_rearm		= posix_cpu_timer_rearm,
+	.timer_wait_running	= posix_cpu_timer_wait_running,
 };
 
 const struct k_clock clock_process = {