cxgbe(4): Retire the allow_mbufs_in_cluster optimization.

This simplifies the driver's rx fast path as well as the bookkeeping
code that tracks various rx buffer sizes and layouts.

MFC after:	1 week
Sponsored by:	Chelsio Communications

share/man/man4/cxgbe.4

@@ -317,11 +317,6 @@ Allow the hardware to deliver multiple frames in the same receive buffer
 opportunistically.
 The default is -1 which lets the driver decide.
 0 or 1 explicitly disable or enable this feature.
-.It Va hw.cxgbe.allow_mbufs_in_cluster
-1 allows the driver to lay down one or more mbufs within the receive buffer
-opportunistically.
-This is the default.
-0 prohibits the driver from doing so.
 .It Va hw.cxgbe.largest_rx_cluster
 .It Va hw.cxgbe.safest_rx_cluster
 Sizes of rx clusters.

sys/dev/cxgbe/adapter.h

@@ -316,15 +316,6 @@ struct port_info {
 
 #define	IS_MAIN_VI(vi)		((vi) == &((vi)->pi->vi[0]))
 
-/* Where the cluster came from, how it has been carved up. */
-struct cluster_layout {
-	int8_t zidx;
-	int8_t hwidx;
-	uint16_t region1;	/* mbufs laid out within this region */
-				/* region2 is the DMA region */
-	uint16_t region3;	/* cluster_metadata within this region */
-};
-
 struct cluster_metadata {
 	uma_zone_t zone;
 	caddr_t cl;
@@ -334,7 +325,8 @@ struct cluster_metadata {
 struct fl_sdesc {
 	caddr_t cl;
 	uint16_t nmbuf;	/* # of driver originated mbufs with ref on cluster */
-	struct cluster_layout cll;
+	int16_t moff;	/* offset of metadata from cl */
+	uint8_t zidx;
 };
 
 struct tx_desc {
@@ -466,18 +458,15 @@ struct sge_eq {
 	char lockname[16];
 };
 
-struct sw_zone_info {
+struct rx_buf_info {
 	uma_zone_t zone;	/* zone that this cluster comes from */
-	int size;		/* size of cluster: 2K, 4K, 9K, 16K, etc. */
-	int type;		/* EXT_xxx type of the cluster */
-	int8_t head_hwidx;
-	int8_t tail_hwidx;
-};
-
-struct hw_buf_info {
-	int8_t zidx;		/* backpointer to zone; -ve means unused */
-	int8_t next;		/* next hwidx for this zone; -1 means no more */
-	int size;
+	uint16_t size1;		/* same as size of cluster: 2K/4K/9K/16K.
+				 * hwsize[hwidx1] = size1.  No spare. */
+	uint16_t size2;		/* hwsize[hwidx2] = size2.
+				 * spare in cluster = size1 - size2. */
+	int8_t hwidx1;		/* SGE bufsize idx for size1 */
+	int8_t hwidx2;		/* SGE bufsize idx for size2 */
+	uint8_t type;		/* EXT_xxx type of the cluster */
 };
 
 enum {
@@ -519,7 +508,8 @@ struct sge_fl {
 	struct mtx fl_lock;
 	__be64 *desc;		/* KVA of descriptor ring, ptr to addresses */
 	struct fl_sdesc *sdesc;	/* KVA of software descriptor ring */
-	struct cluster_layout cll_def;	/* default refill zone, layout */
+	uint16_t zidx;		/* refill zone idx */
+	uint16_t safe_zidx;
 	uint16_t lowat;		/* # of buffers <= this means fl needs help */
 	int flags;
 	uint16_t buf_boundary;
@@ -537,8 +527,6 @@ struct sge_fl {
 	u_int rx_offset;	/* offset in fl buf (when buffer packing) */
 	volatile uint32_t *udb;
 
-	uint64_t mbuf_allocated;/* # of mbuf allocated from zone_mbuf */
-	uint64_t mbuf_inlined;	/* # of mbuf created within clusters */
 	uint64_t cl_allocated;	/* # of clusters allocated */
 	uint64_t cl_recycled;	/* # of clusters recycled */
 	uint64_t cl_fast_recycled; /* # of clusters recycled (fast) */
@@ -555,7 +543,6 @@ struct sge_fl {
 	bus_dmamap_t desc_map;
 	char lockname[16];
 	bus_addr_t ba;		/* bus address of descriptor ring */
-	struct cluster_layout cll_alt;	/* alternate refill zone, layout */
 };
 
 struct mp_ring;
@@ -781,10 +768,8 @@ struct sge {
 	struct sge_iq **iqmap;	/* iq->cntxt_id to iq mapping */
 	struct sge_eq **eqmap;	/* eq->cntxt_id to eq mapping */
 
-	int8_t safe_hwidx1;	/* may not have room for metadata */
-	int8_t safe_hwidx2;	/* with room for metadata and maybe more */
-	struct sw_zone_info sw_zone_info[SW_ZONE_SIZES];
-	struct hw_buf_info hw_buf_info[SGE_FLBUF_SIZES];
+	int8_t safe_zidx;
+	struct rx_buf_info rx_buf_info[SW_ZONE_SIZES];
 };
 
 struct devnames {

sys/dev/cxgbe/t4_main.c

@@ -10291,8 +10291,6 @@ clear_stats(struct adapter *sc, u_int port_id)
 				rxq->rxcsum = 0;
 				rxq->vlan_extraction = 0;
 
-				rxq->fl.mbuf_allocated = 0;
-				rxq->fl.mbuf_inlined = 0;
 				rxq->fl.cl_allocated = 0;
 				rxq->fl.cl_recycled = 0;
 				rxq->fl.cl_fast_recycled = 0;
@@ -10334,8 +10332,6 @@ clear_stats(struct adapter *sc, u_int port_id)
 #endif
 #ifdef TCP_OFFLOAD
 			for_each_ofld_rxq(vi, i, ofld_rxq) {
-				ofld_rxq->fl.mbuf_allocated = 0;
-				ofld_rxq->fl.mbuf_inlined = 0;
 				ofld_rxq->fl.cl_allocated = 0;
 				ofld_rxq->fl.cl_recycled = 0;
 				ofld_rxq->fl.cl_fast_recycled = 0;

sys/dev/cxgbe/t4_netmap.c

@@ -345,7 +345,7 @@ cxgbe_netmap_on(struct adapter *sc, struct vi_info *vi, struct ifnet *ifp,
 	struct sge_nm_rxq *nm_rxq;
 	struct sge_nm_txq *nm_txq;
 	int rc, i, j, hwidx, defq, nrssq;
-	struct hw_buf_info *hwb;
+	struct rx_buf_info *rxb;
 
 	ASSERT_SYNCHRONIZED_OP(sc);
 
@@ -353,17 +353,22 @@ cxgbe_netmap_on(struct adapter *sc, struct vi_info *vi, struct ifnet *ifp,
 	    (ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
 		return (EAGAIN);
 
-	hwb = &sc->sge.hw_buf_info[0];
-	for (i = 0; i < SGE_FLBUF_SIZES; i++, hwb++) {
-		if (hwb->size == NETMAP_BUF_SIZE(na))
+	rxb = &sc->sge.rx_buf_info[0];
+	for (i = 0; i < SW_ZONE_SIZES; i++, rxb++) {
+		if (rxb->size1 == NETMAP_BUF_SIZE(na)) {
+			hwidx = rxb->hwidx1;
 			break;
+		}
+		if (rxb->size2 == NETMAP_BUF_SIZE(na)) {
+			hwidx = rxb->hwidx2;
+			break;
+		}
 	}
-	if (i >= SGE_FLBUF_SIZES) {
+	if (i >= SW_ZONE_SIZES) {
 		if_printf(ifp, "no hwidx for netmap buffer size %d.\n",
 		    NETMAP_BUF_SIZE(na));
 		return (ENXIO);
 	}
-	hwidx = i;
 
 	/* Must set caps before calling netmap_reset */
 	nm_set_native_flags(na);