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, 23 insertions, 13 deletions

diff --git a/sys/dev/nfe/if_nfe.c b/sys/dev/nfe/if_nfe.c
index cb7a163..99f75aa 100644
--- a/sys/dev/nfe/if_nfe.c
+++ b/sys/dev/nfe/if_nfe.c
@@ -93,8 +93,8 @@ static __inline void nfe_discard_rxbuf(struct nfe_softc *, int);
 static __inline void nfe_discard_jrxbuf(struct nfe_softc *, int);
 static int nfe_newbuf(struct nfe_softc *, int);
 static int nfe_jnewbuf(struct nfe_softc *, int);
-static int  nfe_rxeof(struct nfe_softc *, int);
-static int  nfe_jrxeof(struct nfe_softc *, int);
+static int  nfe_rxeof(struct nfe_softc *, int, int *);
+static int  nfe_jrxeof(struct nfe_softc *, int, int *);
 static void nfe_txeof(struct nfe_softc *);
 static int  nfe_encap(struct nfe_softc *, struct mbuf **);
 static void nfe_setmulti(struct nfe_softc *);
@@ -1551,23 +1551,24 @@ nfe_free_tx_ring(struct nfe_softc *sc, struct nfe_tx_ring *ring)
 static poll_handler_t nfe_poll;
 
 
-static void
+static int
 nfe_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
 {
 	struct nfe_softc *sc = ifp->if_softc;
 	uint32_t r;
+	int rx_npkts = 0;
 
 	NFE_LOCK(sc);
 
 	if (!(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 		NFE_UNLOCK(sc);
-		return;
+		return (rx_npkts);
 	}
 
 	if (sc->nfe_framesize > MCLBYTES - ETHER_HDR_LEN)
-		nfe_jrxeof(sc, count);
+		rx_npkts = nfe_jrxeof(sc, count, &rx_npkts);
 	else
-		nfe_rxeof(sc, count);
+		rx_npkts = nfe_rxeof(sc, count, &rx_npkts);
 	nfe_txeof(sc);
 	if (!IFQ_DRV_IS_EMPTY(&ifp->if_snd))
 		taskqueue_enqueue_fast(sc->nfe_tq, &sc->nfe_tx_task);
@@ -1575,7 +1576,7 @@ nfe_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
 	if (cmd == POLL_AND_CHECK_STATUS) {
 		if ((r = NFE_READ(sc, sc->nfe_irq_status)) == 0) {
 			NFE_UNLOCK(sc);
-			return;
+			return (rx_npkts);
 		}
 		NFE_WRITE(sc, sc->nfe_irq_status, r);
 
@@ -1586,6 +1587,7 @@ nfe_poll(struct ifnet *ifp, enum poll_cmd cmd, int count)
 		}
 	}
 	NFE_UNLOCK(sc);
+	return (rx_npkts);
 }
 #endif /* DEVICE_POLLING */
 
@@ -1826,9 +1828,9 @@ nfe_int_task(void *arg, int pending)
 	domore = 0;
 	/* check Rx ring */
 	if (sc->nfe_framesize > MCLBYTES - ETHER_HDR_LEN)
-		domore = nfe_jrxeof(sc, sc->nfe_process_limit);
+		domore = nfe_jrxeof(sc, sc->nfe_process_limit, NULL);
 	else
-		domore = nfe_rxeof(sc, sc->nfe_process_limit);
+		domore = nfe_rxeof(sc, sc->nfe_process_limit, NULL);
 	/* check Tx ring */
 	nfe_txeof(sc);
 
@@ -2015,7 +2017,7 @@ nfe_jnewbuf(struct nfe_softc *sc, int idx)
 
 
 static int
-nfe_rxeof(struct nfe_softc *sc, int count)
+nfe_rxeof(struct nfe_softc *sc, int count, int *rx_npktsp)
 {
 	struct ifnet *ifp = sc->nfe_ifp;
 	struct nfe_desc32 *desc32;
@@ -2023,9 +2025,10 @@ nfe_rxeof(struct nfe_softc *sc, int count)
 	struct nfe_rx_data *data;
 	struct mbuf *m;
 	uint16_t flags;
-	int len, prog;
+	int len, prog, rx_npkts;
 	uint32_t vtag = 0;
 
+	rx_npkts = 0;
 	NFE_LOCK_ASSERT(sc);
 
 	bus_dmamap_sync(sc->rxq.rx_desc_tag, sc->rxq.rx_desc_map,
@@ -2115,18 +2118,21 @@ nfe_rxeof(struct nfe_softc *sc, int count)
 		NFE_UNLOCK(sc);
 		(*ifp->if_input)(ifp, m);
 		NFE_LOCK(sc);
+		rx_npkts++;
 	}
 
 	if (prog > 0)
 		bus_dmamap_sync(sc->rxq.rx_desc_tag, sc->rxq.rx_desc_map,
 		    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
 
+	if (rx_npktsp != NULL)
+		*rx_npktsp = rx_npkts;
 	return (count > 0 ? 0 : EAGAIN);
 }
 
 
 static int
-nfe_jrxeof(struct nfe_softc *sc, int count)
+nfe_jrxeof(struct nfe_softc *sc, int count, int *rx_npktsp)
 {
 	struct ifnet *ifp = sc->nfe_ifp;
 	struct nfe_desc32 *desc32;
@@ -2134,9 +2140,10 @@ nfe_jrxeof(struct nfe_softc *sc, int count)
 	struct nfe_rx_data *data;
 	struct mbuf *m;
 	uint16_t flags;
-	int len, prog;
+	int len, prog, rx_npkts;
 	uint32_t vtag = 0;
 
+	rx_npkts = 0;
 	NFE_LOCK_ASSERT(sc);
 
 	bus_dmamap_sync(sc->jrxq.jrx_desc_tag, sc->jrxq.jrx_desc_map,
@@ -2227,12 +2234,15 @@ nfe_jrxeof(struct nfe_softc *sc, int count)
 		NFE_UNLOCK(sc);
 		(*ifp->if_input)(ifp, m);
 		NFE_LOCK(sc);
+		rx_npkts++;
 	}
 
 	if (prog > 0)
 		bus_dmamap_sync(sc->jrxq.jrx_desc_tag, sc->jrxq.jrx_desc_map,
 		    BUS_DMASYNC_PREREAD | BUS_DMASYNC_PREWRITE);
 
+	if (rx_npktsp != NULL)
+		*rx_npktsp = rx_npkts;
 	return (count > 0 ? 0 : EAGAIN);
 }