hrtimer: Handle remaining time proper for TIME_LOW_RES

commit 203cbf77de59fc8f13502dcfd11350c6d4a5c95f upstream.

If CONFIG_TIME_LOW_RES is enabled we add a jiffie to the relative timeout to
prevent short sleeps, but we do not account for that in interfaces which
retrieve the remaining time.

Helge observed that timerfd can return a remaining time larger than the
relative timeout. That's not expected and breaks userland test programs.

Store the information that the timer was armed relative and provide functions
to adjust the remaining time. To avoid bloating the hrtimer struct make state
a u8, which as a bonus results in better code on x86 at least.

Reported-and-tested-by: Helge Deller <deller@gmx.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: John Stultz <john.stultz@linaro.org>
Cc: linux-m68k@lists.linux-m68k.org
Cc: dhowells@redhat.com
Link: http://lkml.kernel.org/r/20160114164159.273328486@linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
[bwh: Backported to 3.2:
 - Use #ifdef instead of IS_ENABLED() as that doesn't work for config
   symbols that don't exist on the current architecture
 - Use KTIME_LOW_RES directly instead of hrtimer_resolution
 - Use ktime_sub() instead of modifying ktime::tv64 directly
 - Adjust filename, context]
Signed-off-by: Ben Hutchings <ben@decadent.org.uk>

2 files changed, 40 insertions, 18 deletions

diff --git a/kernel/hrtimer.c b/kernel/hrtimer.c
index d9ce3d484231..6918c031363e 100644
--- a/kernel/hrtimer.c
+++ b/kernel/hrtimer.c
@@ -910,7 +910,7 @@ static int enqueue_hrtimer(struct hrtimer *timer,
  */
 static void __remove_hrtimer(struct hrtimer *timer,
 			     struct hrtimer_clock_base *base,
-			     unsigned long newstate, int reprogram)
+			     u8 newstate, int reprogram)
 {
 	struct timerqueue_node *next_timer;
 	if (!(timer->state & HRTIMER_STATE_ENQUEUED))
@@ -944,7 +944,7 @@ static inline int
 remove_hrtimer(struct hrtimer *timer, struct hrtimer_clock_base *base)
 {
 	if (hrtimer_is_queued(timer)) {
-		unsigned long state;
+		u8 state;
 		int reprogram;
 
 		/*
@@ -970,6 +970,22 @@ remove_hrtimer(struct hrtimer *timer, struct hrtimer_clock_base *base)
 	return 0;
 }
 
+static inline ktime_t hrtimer_update_lowres(struct hrtimer *timer, ktime_t tim,
+					    const enum hrtimer_mode mode)
+{
+#ifdef CONFIG_TIME_LOW_RES
+	/*
+	 * CONFIG_TIME_LOW_RES indicates that the system has no way to return
+	 * granular time values. For relative timers we add KTIME_LOW_RES
+	 * (i.e. one jiffie) to prevent short timeouts.
+	 */
+	timer->is_rel = mode & HRTIMER_MODE_REL;
+	if (timer->is_rel)
+		tim = ktime_add_safe(tim, KTIME_LOW_RES);
+#endif
+	return tim;
+}
+
 int __hrtimer_start_range_ns(struct hrtimer *timer, ktime_t tim,
 		unsigned long delta_ns, const enum hrtimer_mode mode,
 		int wakeup)
@@ -983,19 +999,10 @@ int __hrtimer_start_range_ns(struct hrtimer *timer, ktime_t tim,
 	/* Remove an active timer from the queue: */
 	ret = remove_hrtimer(timer, base);
 
-	if (mode & HRTIMER_MODE_REL) {
+	if (mode & HRTIMER_MODE_REL)
 		tim = ktime_add_safe(tim, base->get_time());
-		/*
-		 * CONFIG_TIME_LOW_RES is a temporary way for architectures
-		 * to signal that they simply return xtime in
-		 * do_gettimeoffset(). In this case we want to round up by
-		 * resolution when starting a relative timer, to avoid short
-		 * timeouts. This will go away with the GTOD framework.
-		 */
-#ifdef CONFIG_TIME_LOW_RES
-		tim = ktime_add_safe(tim, base->resolution);
-#endif
-	}
+
+	tim = hrtimer_update_lowres(timer, tim, mode);
 
 	hrtimer_set_expires_range_ns(timer, tim, delta_ns);
 
@@ -1120,19 +1127,25 @@ EXPORT_SYMBOL_GPL(hrtimer_cancel);
 /**
  * hrtimer_get_remaining - get remaining time for the timer
  * @timer:	the timer to read
+ * @adjust:	adjust relative timers when CONFIG_TIME_LOW_RES=y
  */
-ktime_t hrtimer_get_remaining(const struct hrtimer *timer)
+ktime_t __hrtimer_get_remaining(const struct hrtimer *timer, bool adjust)
 {
 	unsigned long flags;
 	ktime_t rem;
 
 	lock_hrtimer_base(timer, &flags);
-	rem = hrtimer_expires_remaining(timer);
+#ifdef CONFIG_TIME_LOW_RES
+	if (adjust)
+		rem = hrtimer_expires_remaining_adjusted(timer);
+	else
+#endif
+		rem = hrtimer_expires_remaining(timer);
 	unlock_hrtimer_base(timer, &flags);
 
 	return rem;
 }
-EXPORT_SYMBOL_GPL(hrtimer_get_remaining);
+EXPORT_SYMBOL_GPL(__hrtimer_get_remaining);
 
 #ifdef CONFIG_NO_HZ
 /**
@@ -1249,6 +1262,15 @@ static void __run_hrtimer(struct hrtimer *timer, ktime_t *now)
 	fn = timer->function;
 
 	/*
+	 * Clear the 'is relative' flag for the TIME_LOW_RES case. If the
+	 * timer is restarted with a period then it becomes an absolute
+	 * timer. If its not restarted it does not matter.
+	 */
+#ifdef CONFIG_TIME_LOW_RES
+	timer->is_rel = false;
+#endif
+
+	/*
 	 * Because we run timers from hardirq context, there is no chance
 	 * they get migrated to another cpu, therefore its safe to unlock
 	 * the timer base.
diff --git a/kernel/time/timer_list.c b/kernel/time/timer_list.c
index 3258455549f4..8e17101216cd 100644
--- a/kernel/time/timer_list.c
+++ b/kernel/time/timer_list.c
@@ -57,7 +57,7 @@ print_timer(struct seq_file *m, struct hrtimer *taddr, struct hrtimer *timer,
 	print_name_offset(m, taddr);
 	SEQ_printf(m, ", ");
 	print_name_offset(m, timer->function);
-	SEQ_printf(m, ", S:%02lx", timer->state);
+	SEQ_printf(m, ", S:%02x", timer->state);
 #ifdef CONFIG_TIMER_STATS
 	SEQ_printf(m, ", ");
 	print_name_offset(m, timer->start_site);