use FW 7.2.16

The patch integrates FW 7.2.16 HSI and implements driver
part of GRO flow.

FW 7.2.16 adds the ability to aggregate packets for GRO
(and not just LRO) and also fixes some bugs.

1. Added new aggregation mode: GRO. In this mode packets are aggregated
   such that the original packets can be reconstructed by the OS.
2. 57712 HW bug workaround - initialized all CAM TM registers to 0x32.
3. Adding the FCoE statistics structures to the BNX2X HSI.
4. Wrong configuration of TX HW input buffer size may cause theoretical
   performance effect. Performed configuration fix.
5. FCOE - Arrival of packets beyond task IO size can lead to crash.
   Fix firmware data-in flow.
6. iSCSI - In rare cases of on-chip termination the graceful termination
   timer hangs, and the termination doesn't complete. Firmware fix to MSL
   timer tolerance.
7. iSCSI - Chip hangs when target sends FIN out-of-order or with isles
   open at the initiator side. Firmware implementation corrected to drop
   FIN received out-of-order or with isles still open.
8. iSCSI - Chip hangs when in case of retransmission not aligned to 4-bytes
   from the beginning of iSCSI PDU. Firmware implementation corrected
   to support arbitrary aligned retransmissions.
9. iSCSI - Arrival of target-initiated NOP-IN during intense ISCSI traffic
   might lead to crash. Firmware fix to relevant flow.

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/broadcom/bnx2x/bnx2x.h

@@ -445,6 +445,8 @@ struct bnx2x_agg_info {
 	u16			vlan_tag;
 	u16			len_on_bd;
 	u32			rxhash;
+	u16			gro_size;
+	u16			full_page;
 };
 
 #define Q_STATS_OFFSET32(stat_name) \
@@ -473,6 +475,11 @@ struct bnx2x_fp_txdata {
 	int			txq_index;
 };
 
+enum bnx2x_tpa_mode_t {
+	TPA_MODE_LRO,
+	TPA_MODE_GRO
+};
+
 struct bnx2x_fastpath {
 	struct bnx2x		*bp; /* parent */
 
@@ -489,6 +496,8 @@ struct bnx2x_fastpath {
 
 	dma_addr_t		status_blk_mapping;
 
+	enum bnx2x_tpa_mode_t	mode;
+
 	u8			max_cos; /* actual number of active tx coses */
 	struct bnx2x_fp_txdata	txdata[BNX2X_MULTI_TX_COS];
 
@@ -1199,6 +1208,8 @@ struct bnx2x {
 #define ETH_MIN_PACKET_SIZE		60
 #define ETH_MAX_PACKET_SIZE		1500
 #define ETH_MAX_JUMBO_PACKET_SIZE	9600
+/* TCP with Timestamp Option (32) + IPv6 (40) */
+#define ETH_MAX_TPA_HEADER_SIZE		72
 
 	/* Max supported alignment is 256 (8 shift) */
 #define BNX2X_RX_ALIGN_SHIFT		min(8, L1_CACHE_SHIFT)
@@ -1269,6 +1280,7 @@ struct bnx2x {
 #define NO_MCP_FLAG			(1 << 9)
 
 #define BP_NOMCP(bp)			(bp->flags & NO_MCP_FLAG)
+#define GRO_ENABLE_FLAG			(1 << 10)
 #define MF_FUNC_DIS			(1 << 11)
 #define OWN_CNIC_IRQ			(1 << 12)
 #define NO_ISCSI_OOO_FLAG		(1 << 13)

drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c

@@ -209,13 +209,11 @@ static inline void bnx2x_update_last_max_sge(struct bnx2x_fastpath *fp,
 		fp->last_max_sge = idx;
 }
 
-static void bnx2x_update_sge_prod(struct bnx2x_fastpath *fp,
-				  struct eth_fast_path_rx_cqe *fp_cqe)
+static inline void bnx2x_update_sge_prod(struct bnx2x_fastpath *fp,
+					 u16 sge_len,
+					 struct eth_end_agg_rx_cqe *cqe)
 {
 	struct bnx2x *bp = fp->bp;
-	u16 sge_len = SGE_PAGE_ALIGN(le16_to_cpu(fp_cqe->pkt_len) -
-				     le16_to_cpu(fp_cqe->len_on_bd)) >>
-		      SGE_PAGE_SHIFT;
 	u16 last_max, last_elem, first_elem;
 	u16 delta = 0;
 	u16 i;
@@ -226,15 +224,15 @@ static void bnx2x_update_sge_prod(struct bnx2x_fastpath *fp,
 	/* First mark all used pages */
 	for (i = 0; i < sge_len; i++)
 		BIT_VEC64_CLEAR_BIT(fp->sge_mask,
-			RX_SGE(le16_to_cpu(fp_cqe->sgl_or_raw_data.sgl[i])));
+			RX_SGE(le16_to_cpu(cqe->sgl_or_raw_data.sgl[i])));
 
 	DP(NETIF_MSG_RX_STATUS, "fp_cqe->sgl[%d] = %d\n",
-	   sge_len - 1, le16_to_cpu(fp_cqe->sgl_or_raw_data.sgl[sge_len - 1]));
+	   sge_len - 1, le16_to_cpu(cqe->sgl_or_raw_data.sgl[sge_len - 1]));
 
 	/* Here we assume that the last SGE index is the biggest */
 	prefetch((void *)(fp->sge_mask));
 	bnx2x_update_last_max_sge(fp,
-		le16_to_cpu(fp_cqe->sgl_or_raw_data.sgl[sge_len - 1]));
+		le16_to_cpu(cqe->sgl_or_raw_data.sgl[sge_len - 1]));
 
 	last_max = RX_SGE(fp->last_max_sge);
 	last_elem = last_max >> BIT_VEC64_ELEM_SHIFT;
@@ -328,6 +326,12 @@ static void bnx2x_tpa_start(struct bnx2x_fastpath *fp, u16 queue,
 	tpa_info->len_on_bd = le16_to_cpu(cqe->len_on_bd);
 	tpa_info->placement_offset = cqe->placement_offset;
 	tpa_info->rxhash = bnx2x_get_rxhash(bp, cqe);
+	if (fp->mode == TPA_MODE_GRO) {
+		u16 gro_size = le16_to_cpu(cqe->pkt_len_or_gro_seg_len);
+		tpa_info->full_page =
+			SGE_PAGE_SIZE * PAGES_PER_SGE / gro_size * gro_size;
+		tpa_info->gro_size = gro_size;
+	}
 
 #ifdef BNX2X_STOP_ON_ERROR
 	fp->tpa_queue_used |= (1 << queue);
@@ -384,25 +388,40 @@ static inline u16 bnx2x_set_lro_mss(struct bnx2x *bp, u16 parsing_flags,
 }
 
 static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
-			       u16 queue, struct sk_buff *skb,
+			       struct bnx2x_agg_info *tpa_info,
+			       u16 pages,
+			       struct sk_buff *skb,
 			       struct eth_end_agg_rx_cqe *cqe,
 			       u16 cqe_idx)
 {
 	struct sw_rx_page *rx_pg, old_rx_pg;
-	u32 i, frag_len, frag_size, pages;
-	int err;
-	int j;
-	struct bnx2x_agg_info *tpa_info = &fp->tpa_info[queue];
+	u32 i, frag_len, frag_size;
+	int err, j, frag_id = 0;
 	u16 len_on_bd = tpa_info->len_on_bd;
+	u16 full_page = 0, gro_size = 0;
 
 	frag_size = le16_to_cpu(cqe->pkt_len) - len_on_bd;
-	pages = SGE_PAGE_ALIGN(frag_size) >> SGE_PAGE_SHIFT;
+
+	if (fp->mode == TPA_MODE_GRO) {
+		gro_size = tpa_info->gro_size;
+		full_page = tpa_info->full_page;
+	}
 
 	/* This is needed in order to enable forwarding support */
-	if (frag_size)
+	if (frag_size) {
 		skb_shinfo(skb)->gso_size = bnx2x_set_lro_mss(bp,
 					tpa_info->parsing_flags, len_on_bd);
 
+		/* set for GRO */
+		if (fp->mode == TPA_MODE_GRO)
+			skb_shinfo(skb)->gso_type =
+			    (GET_FLAG(tpa_info->parsing_flags,
+				      PARSING_FLAGS_OVER_ETHERNET_PROTOCOL) ==
+						PRS_FLAG_OVERETH_IPV6) ?
+				SKB_GSO_TCPV6 : SKB_GSO_TCPV4;
+	}
+
+
 #ifdef BNX2X_STOP_ON_ERROR
 	if (pages > min_t(u32, 8, MAX_SKB_FRAGS)*SGE_PAGE_SIZE*PAGES_PER_SGE) {
 		BNX2X_ERR("SGL length is too long: %d. CQE index is %d\n",
@@ -419,7 +438,12 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 
 		/* FW gives the indices of the SGE as if the ring is an array
 		   (meaning that "next" element will consume 2 indices) */
-		frag_len = min(frag_size, (u32)(SGE_PAGE_SIZE*PAGES_PER_SGE));
+		if (fp->mode == TPA_MODE_GRO)
+			frag_len = min_t(u32, frag_size, (u32)full_page);
+		else /* LRO */
+			frag_len = min_t(u32, frag_size,
+					 (u32)(SGE_PAGE_SIZE * PAGES_PER_SGE));
+
 		rx_pg = &fp->rx_page_ring[sge_idx];
 		old_rx_pg = *rx_pg;
 
@@ -435,9 +459,21 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 		dma_unmap_page(&bp->pdev->dev,
 			       dma_unmap_addr(&old_rx_pg, mapping),
 			       SGE_PAGE_SIZE*PAGES_PER_SGE, DMA_FROM_DEVICE);
-
 		/* Add one frag and update the appropriate fields in the skb */
-		skb_fill_page_desc(skb, j, old_rx_pg.page, 0, frag_len);
+		if (fp->mode == TPA_MODE_LRO)
+			skb_fill_page_desc(skb, j, old_rx_pg.page, 0, frag_len);
+		else { /* GRO */
+			int rem;
+			int offset = 0;
+			for (rem = frag_len; rem > 0; rem -= gro_size) {
+				int len = rem > gro_size ? gro_size : rem;
+				skb_fill_page_desc(skb, frag_id++,
+						   old_rx_pg.page, offset, len);
+				if (offset)
+					get_page(old_rx_pg.page);
+				offset += len;
+			}
+		}
 
 		skb->data_len += frag_len;
 		skb->truesize += SGE_PAGE_SIZE * PAGES_PER_SGE;
@@ -449,18 +485,17 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 	return 0;
 }
 
-static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
-			   u16 queue, struct eth_end_agg_rx_cqe *cqe,
-			   u16 cqe_idx)
+static inline void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
+				  struct bnx2x_agg_info *tpa_info,
+				  u16 pages,
+				  struct eth_end_agg_rx_cqe *cqe,
+				  u16 cqe_idx)
 {
-	struct bnx2x_agg_info *tpa_info = &fp->tpa_info[queue];
 	struct sw_rx_bd *rx_buf = &tpa_info->first_buf;
-	u32 pad = tpa_info->placement_offset;
+	u8 pad = tpa_info->placement_offset;
 	u16 len = tpa_info->len_on_bd;
 	struct sk_buff *skb = NULL;
-	u8 *data = rx_buf->data;
-	/* alloc new skb */
-	u8 *new_data;
+	u8 *new_data, *data = rx_buf->data;
 	u8 old_tpa_state = tpa_info->tpa_state;
 
 	tpa_info->tpa_state = BNX2X_TPA_STOP;
@@ -500,7 +535,8 @@ static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
 		skb->protocol = eth_type_trans(skb, bp->dev);
 		skb->ip_summed = CHECKSUM_UNNECESSARY;
 
-		if (!bnx2x_fill_frag_skb(bp, fp, queue, skb, cqe, cqe_idx)) {
+		if (!bnx2x_fill_frag_skb(bp, fp, tpa_info, pages,
+					 skb, cqe, cqe_idx)) {
 			if (tpa_info->parsing_flags & PARSING_FLAGS_VLAN)
 				__vlan_hwaccel_put_tag(skb, tpa_info->vlan_tag);
 			napi_gro_receive(&fp->napi, skb);
@@ -565,7 +601,7 @@ int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
 		struct eth_fast_path_rx_cqe *cqe_fp;
 		u8 cqe_fp_flags;
 		enum eth_rx_cqe_type cqe_fp_type;
-		u16 len, pad;
+		u16 len, pad, queue;
 		u8 *data;
 
 #ifdef BNX2X_STOP_ON_ERROR
@@ -586,17 +622,21 @@ int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
 		   "  queue %x  vlan %x  len %u\n", CQE_TYPE(cqe_fp_flags),
 		   cqe_fp_flags, cqe_fp->status_flags,
 		   le32_to_cpu(cqe_fp->rss_hash_result),
-		   le16_to_cpu(cqe_fp->vlan_tag), le16_to_cpu(cqe_fp->pkt_len));
+		   le16_to_cpu(cqe_fp->vlan_tag),
+		   le16_to_cpu(cqe_fp->pkt_len_or_gro_seg_len));
 
 		/* is this a slowpath msg? */
 		if (unlikely(CQE_TYPE_SLOW(cqe_fp_type))) {
 			bnx2x_sp_event(fp, cqe);
 			goto next_cqe;
 		}
+
 		rx_buf = &fp->rx_buf_ring[bd_cons];
 		data = rx_buf->data;
 
 		if (!CQE_TYPE_FAST(cqe_fp_type)) {
+			struct bnx2x_agg_info *tpa_info;
+			u16 frag_size, pages;
 #ifdef BNX2X_STOP_ON_ERROR
 			/* sanity check */
 			if (fp->disable_tpa &&
@@ -616,28 +656,38 @@ int bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
 				bnx2x_tpa_start(fp, queue,
 						bd_cons, bd_prod,
 						cqe_fp);
+
 				goto next_rx;
-			} else {
-				u16 queue =
-					cqe->end_agg_cqe.queue_index;
-				DP(NETIF_MSG_RX_STATUS,
-				   "calling tpa_stop on queue %d\n",
-				   queue);
 
-				bnx2x_tpa_stop(bp, fp, queue,
-					       &cqe->end_agg_cqe,
-					       comp_ring_cons);
+			}
+			queue = cqe->end_agg_cqe.queue_index;
+			tpa_info = &fp->tpa_info[queue];
+			DP(NETIF_MSG_RX_STATUS,
+			   "calling tpa_stop on queue %d\n",
+			   queue);
+
+			frag_size = le16_to_cpu(cqe->end_agg_cqe.pkt_len) -
+				    tpa_info->len_on_bd;
+
+			if (fp->mode == TPA_MODE_GRO)
+				pages = (frag_size + tpa_info->full_page - 1) /
+					 tpa_info->full_page;
+			else
+				pages = SGE_PAGE_ALIGN(frag_size) >>
+					SGE_PAGE_SHIFT;
+
+			bnx2x_tpa_stop(bp, fp, tpa_info, pages,
+				       &cqe->end_agg_cqe, comp_ring_cons);
 #ifdef BNX2X_STOP_ON_ERROR
-				if (bp->panic)
-					return 0;
+			if (bp->panic)
+				return 0;
 #endif
 
-				bnx2x_update_sge_prod(fp, cqe_fp);
-				goto next_cqe;
-			}
+			bnx2x_update_sge_prod(fp, pages, &cqe->end_agg_cqe);
+			goto next_cqe;
 		}
 		/* non TPA */
-		len = le16_to_cpu(cqe_fp->pkt_len);
+		len = le16_to_cpu(cqe_fp->pkt_len_or_gro_seg_len);
 		pad = cqe_fp->placement_offset;
 		dma_sync_single_for_cpu(&bp->pdev->dev,
 					dma_unmap_addr(rx_buf, mapping),
@@ -3440,13 +3490,15 @@ int bnx2x_change_mtu(struct net_device *dev, int new_mtu)
 }
 
 netdev_features_t bnx2x_fix_features(struct net_device *dev,
-	netdev_features_t features)
+				     netdev_features_t features)
 {
 	struct bnx2x *bp = netdev_priv(dev);
 
 	/* TPA requires Rx CSUM offloading */
-	if (!(features & NETIF_F_RXCSUM) || bp->disable_tpa)
+	if (!(features & NETIF_F_RXCSUM) || bp->disable_tpa) {
 		features &= ~NETIF_F_LRO;
+		features &= ~NETIF_F_GRO;
+	}
 
 	return features;
 }
@@ -3462,6 +3514,11 @@ int bnx2x_set_features(struct net_device *dev, netdev_features_t features)
 	else
 		flags &= ~TPA_ENABLE_FLAG;
 
+	if (features & NETIF_F_GRO)
+		flags |= GRO_ENABLE_FLAG;
+	else
+		flags &= ~GRO_ENABLE_FLAG;
+
 	if (features & NETIF_F_LOOPBACK) {
 		if (bp->link_params.loopback_mode != LOOPBACK_BMAC) {
 			bp->link_params.loopback_mode = LOOPBACK_BMAC;

drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.h

@@ -534,8 +534,9 @@ int bnx2x_change_mtu(struct net_device *dev, int new_mtu);
  */
 int bnx2x_fcoe_get_wwn(struct net_device *dev, u64 *wwn, int type);
 #endif
+
 netdev_features_t bnx2x_fix_features(struct net_device *dev,
-	netdev_features_t features);
+				     netdev_features_t features);
 int bnx2x_set_features(struct net_device *dev, netdev_features_t features);
 
 /**
@@ -1491,6 +1492,18 @@ static inline u16 bnx2x_extract_max_cfg(struct bnx2x *bp, u32 mf_cfg)
 	return max_cfg;
 }
 
+/* checks if HW supports GRO for given MTU */
+static inline bool bnx2x_mtu_allows_gro(int mtu)
+{
+	/* gro frags per page */
+	int fpp = SGE_PAGE_SIZE / (mtu - ETH_MAX_TPA_HEADER_SIZE);
+
+	/*
+	 * 1. number of frags should not grow above MAX_SKB_FRAGS
+	 * 2. frag must fit the page
+	 */
+	return mtu <= SGE_PAGE_SIZE && (U_ETH_SGL_SIZE * fpp) <= MAX_SKB_FRAGS;
+}
 /**
  * bnx2x_bz_fp - zero content of the fastpath structure.
  *
@@ -1556,7 +1569,14 @@ static inline void bnx2x_bz_fp(struct bnx2x *bp, int index)
 	 * set the tpa flag for each queue. The tpa flag determines the queue
 	 * minimal size so it must be set prior to queue memory allocation
 	 */
-	fp->disable_tpa = (bp->flags & TPA_ENABLE_FLAG) == 0;
+	fp->disable_tpa = !(bp->flags & TPA_ENABLE_FLAG ||
+				  (bp->flags & GRO_ENABLE_FLAG &&
+				   bnx2x_mtu_allows_gro(bp->dev->mtu)));
+	if (bp->flags & TPA_ENABLE_FLAG)
+		fp->mode = TPA_MODE_LRO;
+	else if (bp->flags & GRO_ENABLE_FLAG)
+		fp->mode = TPA_MODE_GRO;
+
 #ifdef BCM_CNIC
 	/* We don't want TPA on an FCoE L2 ring */
 	if (IS_FCOE_FP(fp))

drivers/net/ethernet/broadcom/bnx2x/bnx2x_ethtool.c

@@ -1931,7 +1931,7 @@ static int bnx2x_run_loopback(struct bnx2x *bp, int loopback_mode)
 	if (!CQE_TYPE_FAST(cqe_fp_type) || (cqe_fp_flags & ETH_RX_ERROR_FALGS))
 		goto test_loopback_rx_exit;
 
-	len = le16_to_cpu(cqe->fast_path_cqe.pkt_len);
+	len = le16_to_cpu(cqe->fast_path_cqe.pkt_len_or_gro_seg_len);
 	if (len != pkt_size)
 		goto test_loopback_rx_exit;