bna: Enable Multi Buffer RX

The CT2 HW supports multi-buffer Rx. This patch provides the necessary changes
for bnad to use multi-buffer Rx feature. For BNAD, multi-buffer Rx is by
default enabled when MTU is > 4096. For >4096 MTU, q0 data/large buffers are of
2048 size. As the resource requirements of multi-buffer Rx are different new Rx
needs to be created to use this feature. ASIC posts multiple completions if
frame exceeds buffer size. The last completion is marked with EOP flag.
 - Separate HQ and DQ enums for resource allocations and configurations.
 - rx_config and rxq structure changes to pass the correct info from bnad.
 - DQ depth need not be same as HQ depth. So CQ depth is adjusted accordingly.
 - Rx CFG frame size is taken from configured MTU.
 - Rx q0 buffer size is configured from bnad s rx_config when multi-buffer is
   enabled.
 - Poll for entire frame completion.
 - Once EOP completion is received gather the number of vectors used by the
   frame to submit it to the stack.
 - Changed MTU to frame size wherever necessary.

Signed-off-by: Rasesh Mody <rmody@brocade.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

6 files changed, 343 insertions, 125 deletions

diff --git a/drivers/net/ethernet/brocade/bna/bfi_enet.h b/drivers/net/ethernet/brocade/bna/bfi_enet.h
index 7d10e335c27d..ae072dc5d238 100644
--- a/drivers/net/ethernet/brocade/bna/bfi_enet.h
+++ b/drivers/net/ethernet/brocade/bna/bfi_enet.h
@@ -472,7 +472,8 @@ enum bfi_enet_hds_type {
 
 struct bfi_enet_rx_cfg {
 	u8		rxq_type;
-	u8		rsvd[3];
+	u8		rsvd[1];
+	u16		frame_size;
 
 	struct {
 		u8			max_header_size;
diff --git a/drivers/net/ethernet/brocade/bna/bna_hw_defs.h b/drivers/net/ethernet/brocade/bna/bna_hw_defs.h
index af3f7bb0b3b8..2702d02e98d9 100644
--- a/drivers/net/ethernet/brocade/bna/bna_hw_defs.h
+++ b/drivers/net/ethernet/brocade/bna/bna_hw_defs.h
@@ -322,6 +322,10 @@ do {									\
 #define	BNA_CQ_EF_REMOTE	(1 << 19)
 
 #define	BNA_CQ_EF_LOCAL		(1 << 20)
+/* CAT2 ASIC does not use bit 21 as per the SPEC.
+ * Bit 31 is set in every end of frame completion
+ */
+#define BNA_CQ_EF_EOP		(1 << 31)
 
 /* Data structures */
 
diff --git a/drivers/net/ethernet/brocade/bna/bna_tx_rx.c b/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
index e02c1265e323..85e63546abe3 100644
--- a/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
+++ b/drivers/net/ethernet/brocade/bna/bna_tx_rx.c
@@ -1811,6 +1811,7 @@ bna_bfi_rx_enet_start(struct bna_rx *rx)
 	cfg_req->mh.num_entries = htons(
 		bfi_msgq_num_cmd_entries(sizeof(struct bfi_enet_rx_cfg_req)));
 
+	cfg_req->rx_cfg.frame_size = bna_enet_mtu_get(&rx->bna->enet);
 	cfg_req->num_queue_sets = rx->num_paths;
 	for (i = 0, rxp_qe = bfa_q_first(&rx->rxp_q);
 		i < rx->num_paths;
@@ -1832,8 +1833,17 @@ bna_bfi_rx_enet_start(struct bna_rx *rx)
 			/* Large/Single RxQ */
 			bfi_enet_datapath_q_init(&cfg_req->q_cfg[i].ql.q,
 						&q0->qpt);
-			q0->buffer_size =
-				bna_enet_mtu_get(&rx->bna->enet);
+			if (q0->multi_buffer)
+				/* multi-buffer is enabled by allocating
+				 * a new rx with new set of resources.
+				 * q0->buffer_size should be initialized to
+				 * fragment size.
+				 */
+				cfg_req->rx_cfg.multi_buffer =
+					BNA_STATUS_T_ENABLED;
+			else
+				q0->buffer_size =
+					bna_enet_mtu_get(&rx->bna->enet);
 			cfg_req->q_cfg[i].ql.rx_buffer_size =
 				htons((u16)q0->buffer_size);
 			break;
@@ -2383,8 +2393,8 @@ bna_rx_res_req(struct bna_rx_config *q_cfg, struct bna_res_info *res_info)
 	u32 hq_depth;
 	u32 dq_depth;
 
-	dq_depth = q_cfg->q_depth;
-	hq_depth = ((q_cfg->rxp_type == BNA_RXP_SINGLE) ? 0 : q_cfg->q_depth);
+	dq_depth = q_cfg->q0_depth;
+	hq_depth = ((q_cfg->rxp_type == BNA_RXP_SINGLE) ? 0 : q_cfg->q1_depth);
 	cq_depth = dq_depth + hq_depth;
 
 	BNA_TO_POWER_OF_2_HIGH(cq_depth);
@@ -2501,10 +2511,10 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
 	struct bna_rxq *q0;
 	struct bna_rxq *q1;
 	struct bna_intr_info *intr_info;
-	u32 page_count;
+	struct bna_mem_descr *hqunmap_mem;
+	struct bna_mem_descr *dqunmap_mem;
 	struct bna_mem_descr *ccb_mem;
 	struct bna_mem_descr *rcb_mem;
-	struct bna_mem_descr *unmapq_mem;
 	struct bna_mem_descr *cqpt_mem;
 	struct bna_mem_descr *cswqpt_mem;
 	struct bna_mem_descr *cpage_mem;
@@ -2514,8 +2524,10 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
 	struct bna_mem_descr *dsqpt_mem;
 	struct bna_mem_descr *hpage_mem;
 	struct bna_mem_descr *dpage_mem;
-	int i;
-	int dpage_count, hpage_count, rcb_idx;
+	u32 dpage_count, hpage_count;
+	u32 hq_idx, dq_idx, rcb_idx;
+	u32 cq_depth, i;
+	u32 page_count;
 
 	if (!bna_rx_res_check(rx_mod, rx_cfg))
 		return NULL;
@@ -2523,7 +2535,8 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
 	intr_info = &res_info[BNA_RX_RES_T_INTR].res_u.intr_info;
 	ccb_mem = &res_info[BNA_RX_RES_MEM_T_CCB].res_u.mem_info.mdl[0];
 	rcb_mem = &res_info[BNA_RX_RES_MEM_T_RCB].res_u.mem_info.mdl[0];
-	unmapq_mem = &res_info[BNA_RX_RES_MEM_T_UNMAPQ].res_u.mem_info.mdl[0];
+	dqunmap_mem = &res_info[BNA_RX_RES_MEM_T_UNMAPDQ].res_u.mem_info.mdl[0];
+	hqunmap_mem = &res_info[BNA_RX_RES_MEM_T_UNMAPHQ].res_u.mem_info.mdl[0];
 	cqpt_mem = &res_info[BNA_RX_RES_MEM_T_CQPT].res_u.mem_info.mdl[0];
 	cswqpt_mem = &res_info[BNA_RX_RES_MEM_T_CSWQPT].res_u.mem_info.mdl[0];
 	cpage_mem = &res_info[BNA_RX_RES_MEM_T_CQPT_PAGE].res_u.mem_info.mdl[0];
@@ -2575,7 +2588,8 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
 	}
 
 	rx->num_paths = rx_cfg->num_paths;
-	for (i = 0, rcb_idx = 0; i < rx->num_paths; i++) {
+	for (i = 0, hq_idx = 0, dq_idx = 0, rcb_idx = 0;
+			i < rx->num_paths; i++) {
 		rxp = bna_rxp_get(rx_mod);
 		list_add_tail(&rxp->qe, &rx->rxp_q);
 		rxp->type = rx_cfg->rxp_type;
@@ -2618,9 +2632,13 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
 		q0->rxp = rxp;
 
 		q0->rcb = (struct bna_rcb *) rcb_mem[rcb_idx].kva;
-		q0->rcb->unmap_q = (void *)unmapq_mem[rcb_idx].kva;
-		rcb_idx++;
-		q0->rcb->q_depth = rx_cfg->q_depth;
+		q0->rcb->unmap_q = (void *)dqunmap_mem[dq_idx].kva;
+		rcb_idx++; dq_idx++;
+		q0->rcb->q_depth = rx_cfg->q0_depth;
+		q0->q_depth = rx_cfg->q0_depth;
+		q0->multi_buffer = rx_cfg->q0_multi_buf;
+		q0->buffer_size = rx_cfg->q0_buf_size;
+		q0->num_vecs = rx_cfg->q0_num_vecs;
 		q0->rcb->rxq = q0;
 		q0->rcb->bnad = bna->bnad;
 		q0->rcb->id = 0;
@@ -2640,15 +2658,18 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
 			q1->rxp = rxp;
 
 			q1->rcb = (struct bna_rcb *) rcb_mem[rcb_idx].kva;
-			q1->rcb->unmap_q = (void *)unmapq_mem[rcb_idx].kva;
-			rcb_idx++;
-			q1->rcb->q_depth = rx_cfg->q_depth;
+			q1->rcb->unmap_q = (void *)hqunmap_mem[hq_idx].kva;
+			rcb_idx++; hq_idx++;
+			q1->rcb->q_depth = rx_cfg->q1_depth;
+			q1->q_depth = rx_cfg->q1_depth;
+			q1->multi_buffer = BNA_STATUS_T_DISABLED;
+			q1->num_vecs = 1;
 			q1->rcb->rxq = q1;
 			q1->rcb->bnad = bna->bnad;
 			q1->rcb->id = 1;
 			q1->buffer_size = (rx_cfg->rxp_type == BNA_RXP_HDS) ?
 					rx_cfg->hds_config.forced_offset
-					: rx_cfg->small_buff_size;
+					: rx_cfg->q1_buf_size;
 			q1->rx_packets = q1->rx_bytes = 0;
 			q1->rx_packets_with_error = q1->rxbuf_alloc_failed = 0;
 
@@ -2663,9 +2684,14 @@ bna_rx_create(struct bna *bna, struct bnad *bnad,
 		/* Setup CQ */
 
 		rxp->cq.ccb = (struct bna_ccb *) ccb_mem[i].kva;
-		rxp->cq.ccb->q_depth =	rx_cfg->q_depth +
-					((rx_cfg->rxp_type == BNA_RXP_SINGLE) ?
-					0 : rx_cfg->q_depth);
+		cq_depth = rx_cfg->q0_depth +
+			((rx_cfg->rxp_type == BNA_RXP_SINGLE) ?
+			 0 : rx_cfg->q1_depth);
+		/* if multi-buffer is enabled sum of q0_depth
+		 * and q1_depth need not be a power of 2
+		 */
+		BNA_TO_POWER_OF_2_HIGH(cq_depth);
+		rxp->cq.ccb->q_depth = cq_depth;
 		rxp->cq.ccb->cq = &rxp->cq;
 		rxp->cq.ccb->rcb[0] = q0->rcb;
 		q0->rcb->ccb = rxp->cq.ccb;
diff --git a/drivers/net/ethernet/brocade/bna/bna_types.h b/drivers/net/ethernet/brocade/bna/bna_types.h
index acedac2ae1c5..621547cd3504 100644
--- a/drivers/net/ethernet/brocade/bna/bna_types.h
+++ b/drivers/net/ethernet/brocade/bna/bna_types.h
@@ -109,20 +109,21 @@ enum bna_tx_res_req_type {
 enum bna_rx_mem_type {
 	BNA_RX_RES_MEM_T_CCB		= 0,	/* CQ context */
 	BNA_RX_RES_MEM_T_RCB		= 1,	/* CQ context */
-	BNA_RX_RES_MEM_T_UNMAPQ		= 2,	/* UnmapQ for RxQs */
-	BNA_RX_RES_MEM_T_CQPT		= 3,	/* CQ QPT */
-	BNA_RX_RES_MEM_T_CSWQPT		= 4,	/* S/W QPT */
-	BNA_RX_RES_MEM_T_CQPT_PAGE	= 5,	/* CQPT page */
-	BNA_RX_RES_MEM_T_HQPT		= 6,	/* RX QPT */
-	BNA_RX_RES_MEM_T_DQPT		= 7,	/* RX QPT */
-	BNA_RX_RES_MEM_T_HSWQPT		= 8,	/* RX s/w QPT */
-	BNA_RX_RES_MEM_T_DSWQPT		= 9,	/* RX s/w QPT */
-	BNA_RX_RES_MEM_T_DPAGE		= 10,	/* RX s/w QPT */
-	BNA_RX_RES_MEM_T_HPAGE		= 11,	/* RX s/w QPT */
-	BNA_RX_RES_MEM_T_IBIDX		= 12,
-	BNA_RX_RES_MEM_T_RIT		= 13,
-	BNA_RX_RES_T_INTR		= 14,	/* Rx interrupts */
-	BNA_RX_RES_T_MAX		= 15
+	BNA_RX_RES_MEM_T_UNMAPHQ	= 2,
+	BNA_RX_RES_MEM_T_UNMAPDQ	= 3,
+	BNA_RX_RES_MEM_T_CQPT		= 4,
+	BNA_RX_RES_MEM_T_CSWQPT		= 5,
+	BNA_RX_RES_MEM_T_CQPT_PAGE	= 6,
+	BNA_RX_RES_MEM_T_HQPT		= 7,
+	BNA_RX_RES_MEM_T_DQPT		= 8,
+	BNA_RX_RES_MEM_T_HSWQPT		= 9,
+	BNA_RX_RES_MEM_T_DSWQPT		= 10,
+	BNA_RX_RES_MEM_T_DPAGE		= 11,
+	BNA_RX_RES_MEM_T_HPAGE		= 12,
+	BNA_RX_RES_MEM_T_IBIDX		= 13,
+	BNA_RX_RES_MEM_T_RIT		= 14,
+	BNA_RX_RES_T_INTR		= 15,
+	BNA_RX_RES_T_MAX		= 16
 };
 
 enum bna_tx_type {
@@ -583,6 +584,8 @@ struct bna_rxq {
 
 	int			buffer_size;
 	int			q_depth;
+	u32			num_vecs;
+	enum bna_status		multi_buffer;
 
 	struct bna_qpt qpt;
 	struct bna_rcb *rcb;
@@ -632,6 +635,8 @@ struct bna_ccb {
 	struct bna_rcb *rcb[2];
 	void			*ctrl; /* For bnad */
 	struct bna_pkt_rate pkt_rate;
+	u32			pkts_una;
+	u32			bytes_per_intr;
 
 	/* Control path */
 	struct bna_cq *cq;
@@ -671,14 +676,22 @@ struct bna_rx_config {
 	int			num_paths;
 	enum bna_rxp_type rxp_type;
 	int			paused;
-	int			q_depth;
 	int			coalescing_timeo;
 	/*
 	 * Small/Large (or Header/Data) buffer size to be configured
-	 * for SLR and HDS queue type. Large buffer size comes from
-	 * enet->mtu.
+	 * for SLR and HDS queue type.
 	 */
-	int			small_buff_size;
+	u32			frame_size;
+
+	/* header or small queue */
+	u32			q1_depth;
+	u32			q1_buf_size;
+
+	/* data or large queue */
+	u32			q0_depth;
+	u32			q0_buf_size;
+	u32			q0_num_vecs;
+	enum bna_status		q0_multi_buf;
 
 	enum bna_status rss_status;
 	struct bna_rss_config rss_config;
diff --git a/drivers/net/ethernet/brocade/bna/bnad.c b/drivers/net/ethernet/brocade/bna/bnad.c
index c1357eac3500..0e9f8f5b6009 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.c
+++ b/drivers/net/ethernet/brocade/bna/bnad.c
@@ -282,27 +282,32 @@ static int
 bnad_rxq_alloc_init(struct bnad *bnad, struct bna_rcb *rcb)
 {
 	struct bnad_rx_unmap_q *unmap_q = rcb->unmap_q;
-	int mtu, order;
+	int order;
 
 	bnad_rxq_alloc_uninit(bnad, rcb);
 
-	mtu = bna_enet_mtu_get(&bnad->bna.enet);
-	order = get_order(mtu);
+	order = get_order(rcb->rxq->buffer_size);
+
+	unmap_q->type = BNAD_RXBUF_PAGE;
 
 	if (bna_is_small_rxq(rcb->id)) {
 		unmap_q->alloc_order = 0;
 		unmap_q->map_size = rcb->rxq->buffer_size;
 	} else {
-		unmap_q->alloc_order = order;
-		unmap_q->map_size =
-			(rcb->rxq->buffer_size > 2048) ?
-			PAGE_SIZE << order : 2048;
+		if (rcb->rxq->multi_buffer) {
+			unmap_q->alloc_order = 0;
+			unmap_q->map_size = rcb->rxq->buffer_size;
+			unmap_q->type = BNAD_RXBUF_MULTI_BUFF;
+		} else {
+			unmap_q->alloc_order = order;
+			unmap_q->map_size =
+				(rcb->rxq->buffer_size > 2048) ?
+				PAGE_SIZE << order : 2048;
+		}
 	}
 
 	BUG_ON(((PAGE_SIZE << order) % unmap_q->map_size));
 
-	unmap_q->type = BNAD_RXBUF_PAGE;
-
 	return 0;
 }
 
@@ -345,10 +350,10 @@ bnad_rxq_cleanup(struct bnad *bnad, struct bna_rcb *rcb)
 	for (i = 0; i < rcb->q_depth; i++) {
 		struct bnad_rx_unmap *unmap = &unmap_q->unmap[i];
 
-		if (BNAD_RXBUF_IS_PAGE(unmap_q->type))
-			bnad_rxq_cleanup_page(bnad, unmap);
-		else
+		if (BNAD_RXBUF_IS_SK_BUFF(unmap_q->type))
 			bnad_rxq_cleanup_skb(bnad, unmap);
+		else
+			bnad_rxq_cleanup_page(bnad, unmap);
 	}
 	bnad_rxq_alloc_uninit(bnad, rcb);
 }
@@ -480,10 +485,10 @@ bnad_rxq_post(struct bnad *bnad, struct bna_rcb *rcb)
 	if (!(to_alloc >> BNAD_RXQ_REFILL_THRESHOLD_SHIFT))
 		return;
 
-	if (BNAD_RXBUF_IS_PAGE(unmap_q->type))
-		bnad_rxq_refill_page(bnad, rcb, to_alloc);
-	else
+	if (BNAD_RXBUF_IS_SK_BUFF(unmap_q->type))
 		bnad_rxq_refill_skb(bnad, rcb, to_alloc);
+	else
+		bnad_rxq_refill_page(bnad, rcb, to_alloc);
 }
 
 #define flags_cksum_prot_mask (BNA_CQ_EF_IPV4 | BNA_CQ_EF_L3_CKSUM_OK | \
@@ -500,62 +505,91 @@ bnad_rxq_post(struct bnad *bnad, struct bna_rcb *rcb)
 #define flags_udp6 (BNA_CQ_EF_IPV6 | \
 				BNA_CQ_EF_UDP | BNA_CQ_EF_L4_CKSUM_OK)
 
-static inline struct sk_buff *
-bnad_cq_prepare_skb(struct bnad_rx_ctrl *rx_ctrl,
-		struct bnad_rx_unmap_q *unmap_q,
-		struct bnad_rx_unmap *unmap,
-		u32 length, u32 flags)
+static void
+bnad_cq_drop_packet(struct bnad *bnad, struct bna_rcb *rcb,
+		    u32 sop_ci, u32 nvecs)
 {
-	struct bnad *bnad = rx_ctrl->bnad;
-	struct sk_buff *skb;
+	struct bnad_rx_unmap_q *unmap_q;
+	struct bnad_rx_unmap *unmap;
+	u32 ci, vec;
 
-	if (BNAD_RXBUF_IS_PAGE(unmap_q->type)) {
-		skb = napi_get_frags(&rx_ctrl->napi);
-		if (unlikely(!skb))
-			return NULL;
+	unmap_q = rcb->unmap_q;
+	for (vec = 0, ci = sop_ci; vec < nvecs; vec++) {
+		unmap = &unmap_q->unmap[ci];
+		BNA_QE_INDX_INC(ci, rcb->q_depth);
+
+		if (BNAD_RXBUF_IS_SK_BUFF(unmap_q->type))
+			bnad_rxq_cleanup_skb(bnad, unmap);
+		else
+			bnad_rxq_cleanup_page(bnad, unmap);
+	}
+}
+
+static void
+bnad_cq_setup_skb_frags(struct bna_rcb *rcb, struct sk_buff *skb,
+			u32 sop_ci, u32 nvecs, u32 last_fraglen)
+{
+	struct bnad *bnad;
+	u32 ci, vec, len, totlen = 0;
+	struct bnad_rx_unmap_q *unmap_q;
+	struct bnad_rx_unmap *unmap;
+
+	unmap_q = rcb->unmap_q;
+	bnad = rcb->bnad;
+	for (vec = 1, ci = sop_ci; vec <= nvecs; vec++) {
+		unmap = &unmap_q->unmap[ci];
+		BNA_QE_INDX_INC(ci, rcb->q_depth);
 
 		dma_unmap_page(&bnad->pcidev->dev,
 				dma_unmap_addr(&unmap->vector, dma_addr),
 				unmap->vector.len, DMA_FROM_DEVICE);
+
+		len = (vec == nvecs) ?
+			last_fraglen : unmap->vector.len;
+		totlen += len;
+
 		skb_fill_page_desc(skb, skb_shinfo(skb)->nr_frags,
-				unmap->page, unmap->page_offset, length);
-		skb->len += length;
-		skb->data_len += length;
-		skb->truesize += length;
+				unmap->page, unmap->page_offset, len);
 
 		unmap->page = NULL;
 		unmap->vector.len = 0;
-
-		return skb;
 	}
 
-	skb = unmap->skb;
-	BUG_ON(!skb);
+	skb->len += totlen;
+	skb->data_len += totlen;
+	skb->truesize += totlen;
+}
+
+static inline void
+bnad_cq_setup_skb(struct bnad *bnad, struct sk_buff *skb,
+		  struct bnad_rx_unmap *unmap, u32 len)
+{
+	prefetch(skb->data);
 
 	dma_unmap_single(&bnad->pcidev->dev,
 			dma_unmap_addr(&unmap->vector, dma_addr),
 			unmap->vector.len, DMA_FROM_DEVICE);
 
-	skb_put(skb, length);
-
+	skb_put(skb, len);
 	skb->protocol = eth_type_trans(skb, bnad->netdev);
 
 	unmap->skb = NULL;
 	unmap->vector.len = 0;
-	return skb;
 }
 
 static u32
 bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
 {
-	struct bna_cq_entry *cq, *cmpl;
+	struct bna_cq_entry *cq, *cmpl, *next_cmpl;
 	struct bna_rcb *rcb = NULL;
 	struct bnad_rx_unmap_q *unmap_q;
-	struct bnad_rx_unmap *unmap;
-	struct sk_buff *skb;
+	struct bnad_rx_unmap *unmap = NULL;
+	struct sk_buff *skb = NULL;
 	struct bna_pkt_rate *pkt_rt = &ccb->pkt_rate;
 	struct bnad_rx_ctrl *rx_ctrl = ccb->ctrl;
-	u32 packets = 0, length = 0, flags, masked_flags;
+	u32 packets = 0, len = 0, totlen = 0;
+	u32 pi, vec, sop_ci = 0, nvecs = 0;
+	u32 flags, masked_flags;
 
 	prefetch(bnad->netdev);
 
@@ -563,9 +597,6 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
 	cmpl = &cq[ccb->producer_index];
 
 	while (cmpl->valid && (packets < budget)) {
-		packets++;
-		flags = ntohl(cmpl->flags);
-		length = ntohs(cmpl->length);
 		BNA_UPDATE_PKT_CNT(pkt_rt, ntohs(cmpl->length));
 
 		if (bna_is_small_rxq(cmpl->rxq_id))
@@ -574,25 +605,68 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
 			rcb = ccb->rcb[0];
 
 		unmap_q = rcb->unmap_q;
-		unmap = &unmap_q->unmap[rcb->consumer_index];
 
-		if (unlikely(flags & (BNA_CQ_EF_MAC_ERROR |
-					BNA_CQ_EF_FCS_ERROR |
-					BNA_CQ_EF_TOO_LONG))) {
-			if (BNAD_RXBUF_IS_PAGE(unmap_q->type))
-				bnad_rxq_cleanup_page(bnad, unmap);
-			else
-				bnad_rxq_cleanup_skb(bnad, unmap);
+		/* start of packet ci */
+		sop_ci = rcb->consumer_index;
+
+		if (BNAD_RXBUF_IS_SK_BUFF(unmap_q->type)) {
+			unmap = &unmap_q->unmap[sop_ci];
+			skb = unmap->skb;
+		} else {
+			skb = napi_get_frags(&rx_ctrl->napi);
+			if (unlikely(!skb))
+				break;
+		}
+		prefetch(skb);
+
+		flags = ntohl(cmpl->flags);
+		len = ntohs(cmpl->length);
+		totlen = len;
+		nvecs = 1;
+
+		/* Check all the completions for this frame.
+		 * busy-wait doesn't help much, break here.
+		 */
+		if (BNAD_RXBUF_IS_MULTI_BUFF(unmap_q->type) &&
+		    (flags & BNA_CQ_EF_EOP) == 0) {
+			pi = ccb->producer_index;
+			do {
+				BNA_QE_INDX_INC(pi, ccb->q_depth);
+				next_cmpl = &cq[pi];
 
+				if (!next_cmpl->valid)
+					break;
+
+				len = ntohs(next_cmpl->length);
+				flags = ntohl(next_cmpl->flags);
+
+				nvecs++;
+				totlen += len;
+			} while ((flags & BNA_CQ_EF_EOP) == 0);
+
+			if (!next_cmpl->valid)
+				break;
+		}
+
+		/* TODO: BNA_CQ_EF_LOCAL ? */
+		if (unlikely(flags & (BNA_CQ_EF_MAC_ERROR |
+						BNA_CQ_EF_FCS_ERROR |
+						BNA_CQ_EF_TOO_LONG))) {
+			bnad_cq_drop_packet(bnad, rcb, sop_ci, nvecs);
 			rcb->rxq->rx_packets_with_error++;
+
 			goto next;
 		}
 
-		skb = bnad_cq_prepare_skb(ccb->ctrl, unmap_q, unmap,
-				length, flags);
+		if (BNAD_RXBUF_IS_SK_BUFF(unmap_q->type))
+			bnad_cq_setup_skb(bnad, skb, unmap, len);
+		else
+			bnad_cq_setup_skb_frags(rcb, skb, sop_ci, nvecs, len);
 
-		if (unlikely(!skb))
-			break;
+		packets++;
+		rcb->rxq->rx_packets++;
+		rcb->rxq->rx_bytes += totlen;
+		ccb->bytes_per_intr += totlen;
 
 		masked_flags = flags & flags_cksum_prot_mask;
 
@@ -606,21 +680,21 @@ bnad_cq_process(struct bnad *bnad, struct bna_ccb *ccb, int budget)
 		else
 			skb_checksum_none_assert(skb);
 
-		rcb->rxq->rx_packets++;
-		rcb->rxq->rx_bytes += length;
-
 		if (flags & BNA_CQ_EF_VLAN)
 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), ntohs(cmpl->vlan_tag));
 
-		if (BNAD_RXBUF_IS_PAGE(unmap_q->type))
-			napi_gro_frags(&rx_ctrl->napi);
-		else
+		if (BNAD_RXBUF_IS_SK_BUFF(unmap_q->type))
 			netif_receive_skb(skb);
+		else
+			napi_gro_frags(&rx_ctrl->napi);
 
 next:
-		cmpl->valid = 0;
-		BNA_QE_INDX_INC(rcb->consumer_index, rcb->q_depth);
-		BNA_QE_INDX_INC(ccb->producer_index, ccb->q_depth);
+		BNA_QE_INDX_ADD(rcb->consumer_index, nvecs, rcb->q_depth);
+		for (vec = 0; vec < nvecs; vec++) {
+			cmpl = &cq[ccb->producer_index];
+			cmpl->valid = 0;
+			BNA_QE_INDX_INC(ccb->producer_index, ccb->q_depth);
+		}
 		cmpl = &cq[ccb->producer_index];
 	}
 
@@ -1930,6 +2004,7 @@ err_return:
 static void
 bnad_init_rx_config(struct bnad *bnad, struct bna_rx_config *rx_config)
 {
+	memset(rx_config, 0, sizeof(*rx_config));
 	rx_config->rx_type = BNA_RX_T_REGULAR;
 	rx_config->num_paths = bnad->num_rxp_per_rx;
 	rx_config->coalescing_timeo = bnad->rx_coalescing_timeo;
@@ -1950,10 +2025,39 @@ bnad_init_rx_config(struct bnad *bnad, struct bna_rx_config *rx_config)
 		memset(&rx_config->rss_config, 0,
 		       sizeof(rx_config->rss_config));
 	}
+
+	rx_config->frame_size = BNAD_FRAME_SIZE(bnad->netdev->mtu);
+	rx_config->q0_multi_buf = BNA_STATUS_T_DISABLED;
+
+	/* BNA_RXP_SINGLE - one data-buffer queue
+	 * BNA_RXP_SLR - one small-buffer and one large-buffer queues
+	 * BNA_RXP_HDS - one header-buffer and one data-buffer queues
+	 */
+	/* TODO: configurable param for queue type */
 	rx_config->rxp_type = BNA_RXP_SLR;
-	rx_config->q_depth = bnad->rxq_depth;
 
-	rx_config->small_buff_size = BFI_SMALL_RXBUF_SIZE;
+	if (BNAD_PCI_DEV_IS_CAT2(bnad) &&
+	    rx_config->frame_size > 4096) {
+		/* though size_routing_enable is set in SLR,
+		 * small packets may get routed to same rxq.
+		 * set buf_size to 2048 instead of PAGE_SIZE.
+		 */
+		rx_config->q0_buf_size = 2048;
+		/* this should be in multiples of 2 */
+		rx_config->q0_num_vecs = 4;
+		rx_config->q0_depth = bnad->rxq_depth * rx_config->q0_num_vecs;
+		rx_config->q0_multi_buf = BNA_STATUS_T_ENABLED;
+	} else {
+		rx_config->q0_buf_size = rx_config->frame_size;
+		rx_config->q0_num_vecs = 1;
+		rx_config->q0_depth = bnad->rxq_depth;
+	}
+
+	/* initialize for q1 for BNA_RXP_SLR/BNA_RXP_HDS */
+	if (rx_config->rxp_type == BNA_RXP_SLR) {
+		rx_config->q1_depth = bnad->rxq_depth;
+		rx_config->q1_buf_size = BFI_SMALL_RXBUF_SIZE;
+	}
 
 	rx_config->vlan_strip_status = BNA_STATUS_T_ENABLED;
 }
@@ -1969,6 +2073,49 @@ bnad_rx_ctrl_init(struct bnad *bnad, u32 rx_id)
 }
 
 /* Called with mutex_lock(&bnad->conf_mutex) held */
+u32
+bnad_reinit_rx(struct bnad *bnad)
+{
+	struct net_device *netdev = bnad->netdev;
+	u32 err = 0, current_err = 0;
+	u32 rx_id = 0, count = 0;
+	unsigned long flags;
+
+	/* destroy and create new rx objects */
+	for (rx_id = 0; rx_id < bnad->num_rx; rx_id++) {
+		if (!bnad->rx_info[rx_id].rx)
+			continue;
+		bnad_destroy_rx(bnad, rx_id);
+	}
+
+	spin_lock_irqsave(&bnad->bna_lock, flags);
+	bna_enet_mtu_set(&bnad->bna.enet,
+			 BNAD_FRAME_SIZE(bnad->netdev->mtu), NULL);
+	spin_unlock_irqrestore(&bnad->bna_lock, flags);
+
+	for (rx_id = 0; rx_id < bnad->num_rx; rx_id++) {
+		count++;
+		current_err = bnad_setup_rx(bnad, rx_id);
+		if (current_err && !err) {
+			err = current_err;
+			pr_err("RXQ:%u setup failed\n", rx_id);
+		}
+	}
+
+	/* restore rx configuration */
+	if (bnad->rx_info[0].rx && !err) {
+		bnad_restore_vlans(bnad, 0);
+		bnad_enable_default_bcast(bnad);
+		spin_lock_irqsave(&bnad->bna_lock, flags);
+		bnad_mac_addr_set_locked(bnad, netdev->dev_addr);
+		spin_unlock_irqrestore(&bnad->bna_lock, flags);
+		bnad_set_rx_mode(netdev);
+	}
+
+	return count;
+}
+
+/* Called with bnad_conf_lock() held */
 void
 bnad_destroy_rx(struct bnad *bnad, u32 rx_id)
 {
@@ -2047,13 +2194,19 @@ bnad_setup_rx(struct bnad *bnad, u32 rx_id)
 	spin_unlock_irqrestore(&bnad->bna_lock, flags);
 
 	/* Fill Unmap Q memory requirements */
-	BNAD_FILL_UNMAPQ_MEM_REQ(&res_info[BNA_RX_RES_MEM_T_UNMAPQ],
-			rx_config->num_paths +
-			((rx_config->rxp_type == BNA_RXP_SINGLE) ?
-			 0 : rx_config->num_paths),
-			((bnad->rxq_depth * sizeof(struct bnad_rx_unmap)) +
-			 sizeof(struct bnad_rx_unmap_q)));
-
+	BNAD_FILL_UNMAPQ_MEM_REQ(&res_info[BNA_RX_RES_MEM_T_UNMAPDQ],
+				 rx_config->num_paths,
+			(rx_config->q0_depth *
+			 sizeof(struct bnad_rx_unmap)) +
+			 sizeof(struct bnad_rx_unmap_q));
+
+	if (rx_config->rxp_type != BNA_RXP_SINGLE) {
+		BNAD_FILL_UNMAPQ_MEM_REQ(&res_info[BNA_RX_RES_MEM_T_UNMAPHQ],
+					 rx_config->num_paths,
+				(rx_config->q1_depth *
+				 sizeof(struct bnad_rx_unmap) +
+				 sizeof(struct bnad_rx_unmap_q)));
+	}
 	/* Allocate resource */
 	err = bnad_rx_res_alloc(bnad, res_info, rx_id);
 	if (err)
@@ -2548,7 +2701,6 @@ bnad_open(struct net_device *netdev)
 	int err;
 	struct bnad *bnad = netdev_priv(netdev);
 	struct bna_pause_config pause_config;
-	int mtu;
 	unsigned long flags;
 
 	mutex_lock(&bnad->conf_mutex);
@@ -2567,10 +2719,9 @@ bnad_open(struct net_device *netdev)
 	pause_config.tx_pause = 0;
 	pause_config.rx_pause = 0;
 
-	mtu = ETH_HLEN + VLAN_HLEN + bnad->netdev->mtu + ETH_FCS_LEN;
-
 	spin_lock_irqsave(&bnad->bna_lock, flags);
-	bna_enet_mtu_set(&bnad->bna.enet, mtu, NULL);
+	bna_enet_mtu_set(&bnad->bna.enet,
+			 BNAD_FRAME_SIZE(bnad->netdev->mtu), NULL);
 	bna_enet_pause_config(&bnad->bna.enet, &pause_config, NULL);
 	bna_enet_enable(&bnad->bna.enet);
 	spin_unlock_irqrestore(&bnad->bna_lock, flags);
@@ -3092,14 +3243,14 @@ bnad_set_mac_address(struct net_device *netdev, void *mac_addr)
 }
 
 static int
-bnad_mtu_set(struct bnad *bnad, int mtu)
+bnad_mtu_set(struct bnad *bnad, int frame_size)
 {
 	unsigned long flags;
 
 	init_completion(&bnad->bnad_completions.mtu_comp);
 
 	spin_lock_irqsave(&bnad->bna_lock, flags);
-	bna_enet_mtu_set(&bnad->bna.enet, mtu, bnad_cb_enet_mtu_set);
+	bna_enet_mtu_set(&bnad->bna.enet, frame_size, bnad_cb_enet_mtu_set);
 	spin_unlock_irqrestore(&bnad->bna_lock, flags);
 
 	wait_for_completion(&bnad->bnad_completions.mtu_comp);
@@ -3110,18 +3261,34 @@ bnad_mtu_set(struct bnad *bnad, int mtu)
 static int
 bnad_change_mtu(struct net_device *netdev, int new_mtu)
 {
-	int err, mtu = netdev->mtu;
+	int err, mtu;
 	struct bnad *bnad = netdev_priv(netdev);
+	u32 rx_count = 0, frame, new_frame;
 
 	if (new_mtu + ETH_HLEN < ETH_ZLEN || new_mtu > BNAD_JUMBO_MTU)
 		return -EINVAL;
 
 	mutex_lock(&bnad->conf_mutex);
 
+	mtu = netdev->mtu;
 	netdev->mtu = new_mtu;
 
-	mtu = ETH_HLEN + VLAN_HLEN + new_mtu + ETH_FCS_LEN;
-	err = bnad_mtu_set(bnad, mtu);
+	frame = BNAD_FRAME_SIZE(mtu);
+	new_frame = BNAD_FRAME_SIZE(new_mtu);
+
+	/* check if multi-buffer needs to be enabled */
+	if (BNAD_PCI_DEV_IS_CAT2(bnad) &&
+	    netif_running(bnad->netdev)) {
+		/* only when transition is over 4K */
+		if ((frame <= 4096 && new_frame > 4096) ||
+		    (frame > 4096 && new_frame <= 4096))
+			rx_count = bnad_reinit_rx(bnad);
+	}
+
+	/* rx_count > 0 - new rx created
+	 *	- Linux set err = 0 and return
+	 */
+	err = bnad_mtu_set(bnad, new_frame);
 	if (err)
 		err = -EBUSY;
 
diff --git a/drivers/net/ethernet/brocade/bna/bnad.h b/drivers/net/ethernet/brocade/bna/bnad.h
index 9d10c69f5889..2e6b943262a2 100644
--- a/drivers/net/ethernet/brocade/bna/bnad.h
+++ b/drivers/net/ethernet/brocade/bna/bnad.h
@@ -105,6 +105,9 @@ struct bnad_rx_ctrl {
 #define BNAD_NUM_TXQ			(bnad->num_tx * bnad->num_txq_per_tx)
 #define BNAD_NUM_RXP			(bnad->num_rx * bnad->num_rxp_per_rx)
 
+#define BNAD_FRAME_SIZE(_mtu) \
+	(ETH_HLEN + VLAN_HLEN + (_mtu) + ETH_FCS_LEN)
+
 /*
  * DATA STRUCTURES
  */
@@ -241,12 +244,13 @@ struct bnad_rx_unmap {
 
 enum bnad_rxbuf_type {
 	BNAD_RXBUF_NONE		= 0,
-	BNAD_RXBUF_SKB		= 1,
+	BNAD_RXBUF_SK_BUFF	= 1,
 	BNAD_RXBUF_PAGE		= 2,
-	BNAD_RXBUF_MULTI	= 3
+	BNAD_RXBUF_MULTI_BUFF	= 3
 };
 
-#define BNAD_RXBUF_IS_PAGE(_type)	((_type) == BNAD_RXBUF_PAGE)
+#define BNAD_RXBUF_IS_SK_BUFF(_type)	((_type) == BNAD_RXBUF_SK_BUFF)
+#define BNAD_RXBUF_IS_MULTI_BUFF(_type)	((_type) == BNAD_RXBUF_MULTI_BUFF)
 
 struct bnad_rx_unmap_q {
 	int			reuse_pi;
@@ -256,6 +260,9 @@ struct bnad_rx_unmap_q {
 	struct bnad_rx_unmap	unmap[0];
 };
 
+#define BNAD_PCI_DEV_IS_CAT2(_bnad) \
+	((_bnad)->pcidev->device == BFA_PCI_DEVICE_ID_CT2)
+
 /* Bit mask values for bnad->cfg_flags */
 #define	BNAD_CF_DIM_ENABLED		0x01	/* DIM */
 #define	BNAD_CF_PROMISC			0x02