[lttng-dev] [PATCH] Fix: measure UST clock offset with a median of samples

[Mathieu Desnoyers]

Fixes #729

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>
---
 src/bin/lttng-sessiond/ust-metadata.c |   52 +++++++++++++++++++++++++++------
 1 file changed, 43 insertions(+), 9 deletions(-)

diff --git a/src/bin/lttng-sessiond/ust-metadata.c b/src/bin/lttng-sessiond/ust-metadata.c
index b0f83d2..f511253 100644
--- a/src/bin/lttng-sessiond/ust-metadata.c
+++ b/src/bin/lttng-sessiond/ust-metadata.c
@@ -24,6 +24,7 @@
 #include <string.h>
 #include <stdarg.h>
 #include <stdio.h>
+#include <stdlib.h>
 #include <limits.h>
 #include <unistd.h>
 #include <inttypes.h>
@@ -37,6 +38,8 @@
 #define max_t(type, a, b)	((type) ((a) > (b) ? (a) : (b)))
 #endif
 
+#define NR_CLOCK_OFFSET_SAMPLES		20
+
 static inline
 int fls(unsigned int x)
 {
@@ -490,14 +493,21 @@ int _lttng_event_header_declare(struct ust_registry_session *session)
 	);
 }
 
-/*
- * Approximation of NTP time of day to clock monotonic correlation,
- * taken at start of trace.
- * Yes, this is only an approximation. Yes, we can (and will) do better
- * in future versions.
- */
 static
-uint64_t measure_clock_offset(void)
+int sort_uint64(const void *_a, const void *_b)
+{
+	const uint64_t *a = _a, *b = _b;
+
+	if (a < b)
+		return -1;
+	else if (a > b)
+		return 1;
+	else
+		return 0;
+}
+
+static
+int measure_single_clock_offset(uint64_t *offset_sample)
 {
 	uint64_t offset, monotonic[2], realtime;
 	struct timespec rts = { 0, 0 };
@@ -506,15 +516,39 @@ uint64_t measure_clock_offset(void)
 	monotonic[0] = trace_clock_read64();
 	ret = clock_gettime(CLOCK_REALTIME, &rts);
 	if (ret < 0)
-		return 0;
+		return ret;
 	monotonic[1] = trace_clock_read64();
 	offset = (monotonic[0] + monotonic[1]) >> 1;
 	realtime = (uint64_t) rts.tv_sec * 1000000000ULL;
 	realtime += rts.tv_nsec;
 	offset = realtime - offset;
-	return offset;
+	*offset_sample = offset;
+	return 0;
 }
 
+/*
+ * Approximation of NTP time of day to clock monotonic correlation,
+ * taken at start of trace.
+ * Take the median of many measurements to remove uncertainty caused by
+ * preemption.
+ */
+static
+uint64_t measure_clock_offset(void)
+{
+	uint64_t offset_samples[NR_CLOCK_OFFSET_SAMPLES];
+	int i;
+
+	for (i = 0; i < NR_CLOCK_OFFSET_SAMPLES; i++) {
+		if (measure_single_clock_offset(&offset_samples[i])) {
+			return 0;
+		}
+	}
+
+	/* Find median (using qsort() which is already implemented) */
+	qsort(offset_samples, NR_CLOCK_OFFSET_SAMPLES,
+		sizeof(*offset_samples), sort_uint64);
+	return offset_samples[NR_CLOCK_OFFSET_SAMPLES >> 1];
+}
 
 /*
  * Should be called with session registry mutex held.
-- 
1.7.10.4