time/timecounter: Inline timecounter_cyc2time()

New network transport protocols want NIC drivers to get hardware timestamps
of all incoming packets, and possibly all outgoing packets.

One example is the upcoming 'Swift congestion control' which is used by TCP
transport and is the primary need for timecounter_cyc2time(). This means
timecounter_cyc2time() can be called more than 100 million times per second
on a busy server.

Inlining timecounter_cyc2time() brings a 12% improvement on a UDP receive
stress test on a 100Gbit NIC.

Note that FDO, LTO, PGO are unable to magically help for this case,
presumably because NIC drivers are almost exclusively shipped as modules.

Add an unlikely() around the cc_cyc2ns_backwards() case, even if FDO (when
used) is able to take care of this optimization.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Link: https://research.google/pubs/swift-delay-is-simple-and-effective-for-congestion-control-in-the-datacenter/
Link: https://patch.msgid.link/20251129095740.3338476-1-edumazet@google.com

Author: edumazet

include/linux/timecounter.h

@@ -115,6 +115,15 @@ extern void timecounter_init(struct timecounter *tc,
  */
 extern u64 timecounter_read(struct timecounter *tc);
 
+/*
+ * This is like cyclecounter_cyc2ns(), but it is used for computing a
+ * time previous to the time stored in the cycle counter.
+ */
+static inline u64 cc_cyc2ns_backwards(const struct cyclecounter *cc, u64 cycles, u64 frac)
+{
+	return ((cycles * cc->mult) - frac) >> cc->shift;
+}
+
 /**
  * timecounter_cyc2time - convert a cycle counter to same
  *                        time base as values returned by
@@ -131,7 +140,25 @@ extern u64 timecounter_read(struct timecounter *tc);
  *
  * Returns: cycle counter converted to nanoseconds since the initial time stamp
  */
-extern u64 timecounter_cyc2time(const struct timecounter *tc,
-				u64 cycle_tstamp);
+static inline u64 timecounter_cyc2time(const struct timecounter *tc, u64 cycle_tstamp)
+{
+	const struct cyclecounter *cc = tc->cc;
+	u64 delta = (cycle_tstamp - tc->cycle_last) & cc->mask;
+	u64 nsec = tc->nsec, frac = tc->frac;
+
+	/*
+	 * Instead of always treating cycle_tstamp as more recent than
+	 * tc->cycle_last, detect when it is too far in the future and
+	 * treat it as old time stamp instead.
+	 */
+	if (unlikely(delta > cc->mask / 2)) {
+		delta = (tc->cycle_last - cycle_tstamp) & cc->mask;
+		nsec -= cc_cyc2ns_backwards(cc, delta, frac);
+	} else {
+		nsec += cyclecounter_cyc2ns(cc, delta, tc->mask, &frac);
+	}
+
+	return nsec;
+}
 
 #endif

kernel/time/timecounter.c

@@ -62,38 +62,3 @@ u64 timecounter_read(struct timecounter *tc)
 }
 EXPORT_SYMBOL_GPL(timecounter_read);
 
-/*
- * This is like cyclecounter_cyc2ns(), but it is used for computing a
- * time previous to the time stored in the cycle counter.
- */
-static u64 cc_cyc2ns_backwards(const struct cyclecounter *cc,
-			       u64 cycles, u64 mask, u64 frac)
-{
-	u64 ns = (u64) cycles;
-
-	ns = ((ns * cc->mult) - frac) >> cc->shift;
-
-	return ns;
-}
-
-u64 timecounter_cyc2time(const struct timecounter *tc,
-			 u64 cycle_tstamp)
-{
-	u64 delta = (cycle_tstamp - tc->cycle_last) & tc->cc->mask;
-	u64 nsec = tc->nsec, frac = tc->frac;
-
-	/*
-	 * Instead of always treating cycle_tstamp as more recent
-	 * than tc->cycle_last, detect when it is too far in the
-	 * future and treat it as old time stamp instead.
-	 */
-	if (delta > tc->cc->mask / 2) {
-		delta = (tc->cycle_last - cycle_tstamp) & tc->cc->mask;
-		nsec -= cc_cyc2ns_backwards(tc->cc, delta, tc->mask, frac);
-	} else {
-		nsec += cyclecounter_cyc2ns(tc->cc, delta, tc->mask, &frac);
-	}
-
-	return nsec;
-}
-EXPORT_SYMBOL_GPL(timecounter_cyc2time);