Date:	Fri, 02 Oct 2009 16:17:53 -0700
From:	john stultz <johnstul@...ibm.com>
To:	lkml <linux-kernel@...r.kernel.org>
Cc:	John Kacur <jkacur@...hat.com>,
	Clark Williams <williams@...hat.com>,
	Ingo Molnar <mingo@...e.hu>,
	Martin Schwidefsky <schwidefsky@...ibm.com>,
	Thomas Gleixner <tglx@...utronix.de>,
	Andrew Morton <akpm@...ux-foundation.org>
Subject: [PATCH 1/2] time: logrithmic time accumulation

Accumulating one tick at a time works well unless we're using NOHZ. Then
it can be an issue, since we may have to run through the loop a few
thousand times, which can increase timer interrupt caused latency.

The current solution was to accumulate in half-second intervals with
NOHZ. This kept the number of loops down, however it did slightly change
how we make NTP adjustments. While not an issue with NTPd users, as NTPd
makes adjustments over a longer period of time, other adjtimex() users
have noticed the half-second granularity with which we can apply
frequency changes to the clock.

For instance, if a application tries to apply a 100ppm frequency
correction for 20ms to correct a 2us offset, with NOHZ they either get
no correction, or a 50us correction.

Now, there will always be some granularity error for applying frequency
corrections. However with users sensitive to this error have seen a
50-500x increase with NOHZ compared to running without NOHZ.

So I figured I'd try another approach then just simply increasing the
interval. My approach is to consume the time interval logarithmically.
This reduces the number of times  through the loop needed keeping
latency down, while still preserving the original granularity error for
adjtimex() changes.

Further, this change allows us to remove the xtime_cache code (patch to
follow), as xtime is always within one tick of the current time, instead
of the half-second updates it saw before.

An earlier version of this patch has been shipping to x86 users in the
RedHat MRG releases for awhile without issue, but I've reworked this
version to be even more careful about avoiding possible overflows if the
shift value gets too large.

Since this is not the most trivial code, and its slightly different then
whats been tested for  awhile, it would be good to get this into some
trees for testing. Be it -tip or -mm, either would work. If there's no
problems it could be a 2.6.33 or 2.6.34 item.

Any comments or feedback would be appreciated!

Signed-off-by: John Stultz <johnstul@...ibm.com>

diff --git a/include/linux/timex.h b/include/linux/timex.h
index e6967d1..0c0ef7d 100644
--- a/include/linux/timex.h
+++ b/include/linux/timex.h
@@ -261,11 +261,7 @@ static inline int ntp_synced(void)
 
 #define NTP_SCALE_SHIFT		32
 
-#ifdef CONFIG_NO_HZ
-#define NTP_INTERVAL_FREQ  (2)
-#else
 #define NTP_INTERVAL_FREQ  (HZ)
-#endif
 #define NTP_INTERVAL_LENGTH (NSEC_PER_SEC/NTP_INTERVAL_FREQ)
 
 /* Returns how long ticks are at present, in ns / 2^NTP_SCALE_SHIFT. */
diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
index fb0f46f..4cc5656 100644
--- a/kernel/time/timekeeping.c
+++ b/kernel/time/timekeeping.c
@@ -721,6 +721,51 @@ static void timekeeping_adjust(s64 offset)
 				timekeeper.ntp_error_shift;
 }
 
+
+/**
+ * logarithmic_accumulation - shifted accumulation of cycles
+ *
+ * This functions accumulates a shifted interval of cycles into 
+ * into a shifted interval nanoseconds. Allows for O(log) accumulation
+ * loop.
+ * 
+ * Returns the unconsumed cycles.
+ */
+static cycle_t logarithmic_accumulation(cycle_t offset, int shift)
+{
+	u64 nsecps = (u64)NSEC_PER_SEC << timekeeper.shift;
+
+	/* If the offset is smaller then a shifted interval, do nothing */
+	if (offset < timekeeper.cycle_interval<<shift)
+		return offset;
+
+	/* accumulate one shifted interval */
+	offset -= timekeeper.cycle_interval << shift;
+	timekeeper.clock->cycle_last += timekeeper.cycle_interval << shift;
+	
+	timekeeper.xtime_nsec += timekeeper.xtime_interval << shift;
+	while (timekeeper.xtime_nsec >= nsecps) {
+		timekeeper.xtime_nsec -= nsecps;
+		xtime.tv_sec++;
+		second_overflow();
+	}
+	
+	/* accumulate into raw time */
+	raw_time.tv_nsec += timekeeper.raw_interval << shift;;
+	while (raw_time.tv_nsec >= NSEC_PER_SEC) {
+		raw_time.tv_nsec -= NSEC_PER_SEC;
+		raw_time.tv_sec++;
+	}
+
+	/* accumulate error between NTP and clock interval */
+	timekeeper.ntp_error += tick_length << shift;
+	timekeeper.ntp_error -= timekeeper.xtime_interval <<
+				(timekeeper.ntp_error_shift + shift);
+
+	return offset;
+}
+
+
 /**
  * update_wall_time - Uses the current clocksource to increment the wall time
  *
@@ -731,6 +776,7 @@ void update_wall_time(void)
 	struct clocksource *clock;
 	cycle_t offset;
 	u64 nsecs;
+	int shift = 0, maxshift;
 
 	/* Make sure we're fully resumed: */
 	if (unlikely(timekeeping_suspended))
@@ -744,33 +790,22 @@ void update_wall_time(void)
 #endif
 	timekeeper.xtime_nsec = (s64)xtime.tv_nsec << timekeeper.shift;
 
-	/* normally this loop will run just once, however in the
-	 * case of lost or late ticks, it will accumulate correctly.
+	/*
+	 * With NO_HZ we may have to accumulate many cycle_intervals
+	 * (think "ticks") worth of time at once. To do this efficiently,
+	 * we calculate the largest doubling multiple of cycle_intervals
+	 * that is smaller then the offset. We then accumulate that 
+	 * chunk in one go, and then try to consume the next smaller
+	 * doubled multiple.
 	 */
+	shift = ilog2(offset) - ilog2(timekeeper.cycle_interval);
+	shift = max(0, shift);
+	/* Bound shift to one less then what overflows tick_length */
+	maxshift = (8*sizeof(tick_length) - (ilog2(tick_length)+1)) - 1;
+	shift = min(shift, maxshift);	
 	while (offset >= timekeeper.cycle_interval) {
-		u64 nsecps = (u64)NSEC_PER_SEC << timekeeper.shift;
-
-		/* accumulate one interval */
-		offset -= timekeeper.cycle_interval;
-		clock->cycle_last += timekeeper.cycle_interval;
-
-		timekeeper.xtime_nsec += timekeeper.xtime_interval;
-		if (timekeeper.xtime_nsec >= nsecps) {
-			timekeeper.xtime_nsec -= nsecps;
-			xtime.tv_sec++;
-			second_overflow();
-		}
-
-		raw_time.tv_nsec += timekeeper.raw_interval;
-		if (raw_time.tv_nsec >= NSEC_PER_SEC) {
-			raw_time.tv_nsec -= NSEC_PER_SEC;
-			raw_time.tv_sec++;
-		}
-
-		/* accumulate error between NTP and clock interval */
-		timekeeper.ntp_error += tick_length;
-		timekeeper.ntp_error -= timekeeper.xtime_interval <<
-					timekeeper.ntp_error_shift;
+		offset = logarithmic_accumulation(offset, shift);
+		shift--;
 	}
 
 	/* correct the clock when NTP error is too big */


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists