t24: a bug in ptpd_netif_linearize_rx_timestamp masked the ttrans bug

The phase was used backwards. Somehow this cancelled out the systematic
error introduced by the actual bug (wrong phase offset). When both bugs
are fixed the code seems to work every time.

lib/net.c

@@ -136,7 +136,7 @@ void ptpd_netif_linearize_rx_timestamp(wr_timestamp_t * ts, int32_t dmtd_phase,
 	// TS counter will appear
 	ts->raw_phase = dmtd_phase;
 
-	phase = clock_period - 1 - dmtd_phase;
+	phase = dmtd_phase;
 
 	// calculate the range within which falling edge timestamp is stable
 	// (no possible transitions)
@@ -148,6 +148,8 @@ void ptpd_netif_linearize_rx_timestamp(wr_timestamp_t * ts, int32_t dmtd_phase,
 	if (trip_hi >= clock_period)
 		trip_hi -= clock_period;
 
+	// pp_printf("linearize: %d %d %d %d -- %d\n", trip_lo, transition_point, trip_hi, ep_get_bitslide(), phase);
+
 	if (inside_range(trip_lo, trip_hi, phase)) {
 		// We are within +- 25% range of transition area of
 		// rising counter. Take the falling edge counter value as the
@@ -160,7 +162,7 @@ void ptpd_netif_linearize_rx_timestamp(wr_timestamp_t * ts, int32_t dmtd_phase,
 		// and eventually increase the counter by 1 to simulate a
 		// timestamp transition exactly at s->phase_transition
 		//DMTD phase value
-		if (inside_range(trip_lo, transition_point, phase))
+		if (inside_range(transition_point, trip_hi, phase))
 			ts->nsec += clock_period / 1000;
 
 	}
@@ -168,7 +170,6 @@ void ptpd_netif_linearize_rx_timestamp(wr_timestamp_t * ts, int32_t dmtd_phase,
 	ts->phase = phase - transition_point - 1;
 	if (ts->phase < 0)
 		ts->phase += clock_period;
-	ts->phase = clock_period - 1 - ts->phase;
 }
 
 /* Slow, but we don't care much... */