When user_frac in the polling subsystem is low it is going to busy the

CPU for too long period than necessary.  Additively, interfaces are kept
polled (in the tick) even if no more packets are available.
In order to avoid such situations a new generic mechanism can be
implemented in proactive way, keeping track of the time spent on any
packet and fragmenting the time for any tick, stopping the processing
as soon as possible.

In order to implement such mechanism, the polling handler needs to
change, returning the number of packets processed.
While the intended logic is not part of this patch, the polling KPI is
broken by this commit, adding an int return value and the new flag
IFCAP_POLLING_NOCOUNT (which will signal that the return value is
meaningless for the installed handler and checking should be skipped).

Bump __FreeBSD_version in order to signal such situation.

Reviewed by:	emaste
Sponsored by:	Sandvine Incorporated

1 files changed, 14 insertions, 6 deletions

diff --git a/sys/dev/tsec/if_tsec.c b/sys/dev/tsec/if_tsec.c
index 7c11229..3676cfc 100644
--- a/sys/dev/tsec/if_tsec.c
+++ b/sys/dev/tsec/if_tsec.c
@@ -97,7 +97,7 @@ static int	tsec_sysctl_ic_time(SYSCTL_HANDLER_ARGS);
 static int	tsec_sysctl_ic_count(SYSCTL_HANDLER_ARGS);
 static void	tsec_set_rxic(struct tsec_softc *sc);
 static void	tsec_set_txic(struct tsec_softc *sc);
-static void	tsec_receive_intr_locked(struct tsec_softc *sc, int count);
+static int	tsec_receive_intr_locked(struct tsec_softc *sc, int count);
 static void	tsec_transmit_intr_locked(struct tsec_softc *sc);
 static void	tsec_error_intr_locked(struct tsec_softc *sc, int count);
 static void	tsec_offload_setup(struct tsec_softc *sc);
@@ -857,16 +857,19 @@ tsec_setfilter(struct tsec_softc *sc)
 #ifdef DEVICE_POLLING
 static poll_handler_t tsec_poll;
 
-static void
+static int
 tsec_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
 {
 	uint32_t ie;
 	struct tsec_softc *sc = ifp->if_softc;
+	int rx_npkts;
+
+	rx_npkts = 0;
 
 	TSEC_GLOBAL_LOCK(sc);
 	if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 		TSEC_GLOBAL_UNLOCK(sc);
-		return;
+		return (rx_npkts);
 	}
 
 	if (cmd == POLL_AND_CHECK_STATUS) {
@@ -881,9 +884,11 @@ tsec_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
 
 	TSEC_GLOBAL_TO_RECEIVE_LOCK(sc);
 
-	tsec_receive_intr_locked(sc, count);
+	rx_npkts = tsec_receive_intr_locked(sc, count);
 
 	TSEC_RECEIVE_UNLOCK(sc);
+
+	return (rx_npkts);
 }
 #endif /* DEVICE_POLLING */
 
@@ -1248,7 +1253,7 @@ tsec_tick(void *arg)
  *
  *  Loops at most count times if count is > 0, or until done if count < 0.
  */
-static void
+static int
 tsec_receive_intr_locked(struct tsec_softc *sc, int count)
 {
 	struct tsec_desc *rx_desc;
@@ -1257,7 +1262,7 @@ tsec_receive_intr_locked(struct tsec_softc *sc, int count)
 	struct mbuf *m;
 	device_t dev;
 	uint32_t i;
-	int c;
+	int c, rx_npkts;
 	uint16_t flags;
 
 	TSEC_RECEIVE_LOCK_ASSERT(sc);
@@ -1265,6 +1270,7 @@ tsec_receive_intr_locked(struct tsec_softc *sc, int count)
 	ifp = sc->tsec_ifp;
 	rx_data = sc->rx_data;
 	dev = sc->dev;
+	rx_npkts = 0;
 
 	bus_dmamap_sync(sc->tsec_rx_dtag, sc->tsec_rx_dmap,
 	    BUS_DMASYNC_POSTREAD | BUS_DMASYNC_POSTWRITE);
@@ -1358,6 +1364,7 @@ tsec_receive_intr_locked(struct tsec_softc *sc, int count)
 			TSEC_RECEIVE_UNLOCK(sc);
 			(*ifp->if_input)(ifp, m);
 			TSEC_RECEIVE_LOCK(sc);
+			rx_npkts++;
 		}
 	}
 
@@ -1373,6 +1380,7 @@ tsec_receive_intr_locked(struct tsec_softc *sc, int count)
 	 * halted, and is harmless if already running.
 	 */
 	TSEC_WRITE(sc, TSEC_REG_RSTAT, TSEC_RSTAT_QHLT);
+	return (rx_npkts);
 }
 
 void