[PATCH net-next 00/13] bnxt_en: TX path improvements

[PATCH net-next 00/13] bnxt_en: TX path improvements

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 1983 bytes --]

All patches in this patchset are related to improving the TX path.
There are 2 areas of improvements:

1. The TX interrupt logic currently counts the number of TX completions
to determine the number of TX SKBs to free.  We now change it so that
the TX completion will now contain the hardware consumer index
information.  The driver will keep track of the latest hardware
consumer index from the last TX completion and clean up all TX SKBs
up to that index.  This scheme aligns better with future chips and
allows xmit_more code path to be more optimized.

2. The current driver logic requires an additional MSIX for each
additional MQPRIO TX ring.  This scheme uses too many MSIX vectors if
the user enables a large number of MQPRIO TCs.  We now use a new scheme
that will use the same MSIX for all the MQPRIO TX rings for each
ethtool channel.  Each ethtool TX channel can have up to 8 MQPRIO
TX rings and now they all will share the same MSIX.

Michael Chan (13):
  bnxt_en: Put the TX producer information in the TX BD opaque field
  bnxt_en: Add completion ring pointer in TX and RX ring structures
  bnxt_en: Restructure cp_ring_arr in struct bnxt_cp_ring_info
  bnxt_en: Add completion ring pointer in TX and RX ring structures
  bnxt_en: Remove BNXT_RX_HDL and BNXT_TX_HDL
  bnxt_en: Refactor bnxt_tx_int()
  bnxt_en: New encoding for the TX opaque field
  bnxt_en: Refactor bnxt_hwrm_set_coal()
  bnxt_en: Support up to 8 TX rings per MSIX
  bnxt_en: Add helper to get the number of CP rings required for TX
    rings
  bnxt_en: Add macros related to TC and TX rings
  bnxt_en: Use existing MSIX vectors for all mqprio TX rings
  bnxt_en: Optimize xmit_more TX path

 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 513 +++++++++++-------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  59 +-
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |  11 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c |  23 +-
 4 files changed, 385 insertions(+), 221 deletions(-)

-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 01/13] bnxt_en: Put the TX producer information in the TX BD opaque field

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 3665 bytes --]

Currently, the opaque field in the TX BD is only used for debugging.
The TX completion logic relies on getting one TX completion for each
packet and they always complete in order.

Improve this scheme by putting the producer information (ring index plus
number of BDs for the packet) in the opaque field.  This way, we can
handle TX completion processing by looking at the last TX completion
instead of counting the number of completions.

Since we no longer need to count the exact number of completions, we can
optimize xmit_more by disabling TX completion when the xmit_more
condition is true.  This will be done in later patches.

This patch is only initializing the opaque field in the TX BD and is
not changing the driver's TX completion logic yet.

Reviewed-by: Andy Gospodarek <gospo@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 5 +++--
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     | 7 +++++++
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c | 2 +-
 3 files changed, 11 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index d0359b569afe..669ea945d3cd 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -432,8 +432,6 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	txbd = &txr->tx_desc_ring[TX_RING(prod)][TX_IDX(prod)];
 
-	txbd->tx_bd_opaque = prod;
-
 	tx_buf = &txr->tx_buf_ring[prod];
 	tx_buf->skb = skb;
 	tx_buf->nr_frags = last_frag;
@@ -519,7 +517,9 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 		txbd->tx_bd_len_flags_type = tx_push->tx_bd_len_flags_type;
 		txbd->tx_bd_haddr = txr->data_mapping;
+		txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, prod, 2);
 		prod = NEXT_TX(prod);
+		tx_push->tx_bd_opaque = txbd->tx_bd_opaque;
 		txbd = &txr->tx_desc_ring[TX_RING(prod)][TX_IDX(prod)];
 		memcpy(txbd, tx_push1, sizeof(*txbd));
 		prod = NEXT_TX(prod);
@@ -562,6 +562,7 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		((last_frag + 2) << TX_BD_FLAGS_BD_CNT_SHIFT);
 
 	txbd->tx_bd_haddr = cpu_to_le64(mapping);
+	txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, prod, 2 + last_frag);
 
 	prod = NEXT_TX(prod);
 	txbd1 = (struct tx_bd_ext *)
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index e702dbc3e6b1..c7895e7d78d5 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -61,6 +61,13 @@ struct tx_bd {
 	__le64 tx_bd_haddr;
 } __packed;
 
+#define TX_OPAQUE_IDX_MASK	0x0000ffff
+#define TX_OPAQUE_BDS_MASK	0x00ff0000
+#define TX_OPAQUE_BDS_SHIFT	16
+
+#define SET_TX_OPAQUE(bp, idx, bds)					\
+	(((bds) << TX_OPAQUE_BDS_SHIFT) | ((idx) & (bp)->tx_ring_mask))
+
 struct tx_bd_ext {
 	__le32 tx_bd_hsize_lflags;
 	#define TX_BD_FLAGS_TCP_UDP_CHKSUM			(1 << 0)
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
index 96f5ca778c67..3e5144aafb0c 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
@@ -52,7 +52,7 @@ struct bnxt_sw_tx_bd *bnxt_xmit_bd(struct bnxt *bp,
 		((num_frags + 1) << TX_BD_FLAGS_BD_CNT_SHIFT) |
 		bnxt_lhint_arr[len >> 9];
 	txbd->tx_bd_len_flags_type = cpu_to_le32(flags);
-	txbd->tx_bd_opaque = prod;
+	txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, prod, 1 + num_frags);
 	txbd->tx_bd_haddr = cpu_to_le64(mapping);
 
 	/* now let us fill up the frags into the next buffers */
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 02/13] bnxt_en: Add completion ring pointer in TX and RX ring structures

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 9043 bytes --]

From the TX or RX ring structure, we need to find the corresponding
completion ring during initialization.  On P5 chips, we use the MSIX/napi
entry to locate the completion ring because there is only one RX/TX
ring per MSIX.  To allow multiple TX rings for each MSIX, we need
to add a direct pointer from the TX ring and RX ring structures.
This also simplifies the existing logic.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 43 +++++++++++--------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     | 11 ++++-
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c | 12 +++---
 3 files changed, 40 insertions(+), 26 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 669ea945d3cd..4dfe0b66c5f7 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -331,16 +331,16 @@ static void bnxt_sched_reset_rxr(struct bnxt *bp, struct bnxt_rx_ring_info *rxr)
 }
 
 void bnxt_sched_reset_txr(struct bnxt *bp, struct bnxt_tx_ring_info *txr,
-			  int idx)
+			  u16 curr)
 {
 	struct bnxt_napi *bnapi = txr->bnapi;
 
 	if (bnapi->tx_fault)
 		return;
 
-	netdev_err(bp->dev, "Invalid Tx completion (ring:%d tx_pkts:%d cons:%u prod:%u i:%d)",
-		   txr->txq_index, bnapi->tx_pkts,
-		   txr->tx_cons, txr->tx_prod, idx);
+	netdev_err(bp->dev, "Invalid Tx completion (ring:%d tx_hw_cons:%u cons:%u prod:%u curr:%u)",
+		   txr->txq_index, txr->tx_hw_cons,
+		   txr->tx_cons, txr->tx_prod, curr);
 	WARN_ON_ONCE(1);
 	bnapi->tx_fault = 1;
 	bnxt_queue_sp_work(bp, BNXT_RESET_TASK_SP_EVENT);
@@ -691,13 +691,13 @@ static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 {
 	struct bnxt_tx_ring_info *txr = bnapi->tx_ring;
 	struct netdev_queue *txq = netdev_get_tx_queue(bp->dev, txr->txq_index);
+	u16 hw_cons = txr->tx_hw_cons;
 	u16 cons = txr->tx_cons;
 	struct pci_dev *pdev = bp->pdev;
-	int nr_pkts = bnapi->tx_pkts;
-	int i;
 	unsigned int tx_bytes = 0;
+	int tx_pkts = 0;
 
-	for (i = 0; i < nr_pkts; i++) {
+	while (cons != hw_cons) {
 		struct bnxt_sw_tx_bd *tx_buf;
 		struct sk_buff *skb;
 		int j, last;
@@ -708,10 +708,11 @@ static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 		tx_buf->skb = NULL;
 
 		if (unlikely(!skb)) {
-			bnxt_sched_reset_txr(bp, txr, i);
+			bnxt_sched_reset_txr(bp, txr, cons);
 			return;
 		}
 
+		tx_pkts++;
 		tx_bytes += skb->len;
 
 		if (tx_buf->is_push) {
@@ -748,10 +749,10 @@ static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 		dev_consume_skb_any(skb);
 	}
 
-	bnapi->tx_pkts = 0;
+	bnapi->events &= ~BNXT_TX_CMP_EVENT;
 	WRITE_ONCE(txr->tx_cons, cons);
 
-	__netif_txq_completed_wake(txq, nr_pkts, tx_bytes,
+	__netif_txq_completed_wake(txq, tx_pkts, tx_bytes,
 				   bnxt_tx_avail(bp, txr), bp->tx_wake_thresh,
 				   READ_ONCE(txr->dev_state) == BNXT_DEV_STATE_CLOSING);
 }
@@ -2588,14 +2589,15 @@ static int __bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 {
 	struct bnxt_napi *bnapi = cpr->bnapi;
 	u32 raw_cons = cpr->cp_raw_cons;
+	struct bnxt_tx_ring_info *txr;
 	u32 cons;
-	int tx_pkts = 0;
 	int rx_pkts = 0;
 	u8 event = 0;
 	struct tx_cmp *txcmp;
 
 	cpr->has_more_work = 0;
 	cpr->had_work_done = 1;
+	txr = bnapi->tx_ring;
 	while (1) {
 		int rc;
 
@@ -2610,9 +2612,15 @@ static int __bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 		 */
 		dma_rmb();
 		if (TX_CMP_TYPE(txcmp) == CMP_TYPE_TX_L2_CMP) {
-			tx_pkts++;
+			u32 opaque = txcmp->tx_cmp_opaque;
+			u16 tx_freed;
+
+			event |= BNXT_TX_CMP_EVENT;
+			txr->tx_hw_cons = TX_OPAQUE_PROD(bp, opaque);
+			tx_freed = (txr->tx_hw_cons - txr->tx_cons) &
+				   bp->tx_ring_mask;
 			/* return full budget so NAPI will complete. */
-			if (unlikely(tx_pkts >= bp->tx_wake_thresh)) {
+			if (unlikely(tx_freed >= bp->tx_wake_thresh)) {
 				rx_pkts = budget;
 				raw_cons = NEXT_RAW_CMP(raw_cons);
 				if (budget)
@@ -2666,7 +2674,6 @@ static int __bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 	}
 
 	cpr->cp_raw_cons = raw_cons;
-	bnapi->tx_pkts += tx_pkts;
 	bnapi->events |= event;
 	return rx_pkts;
 }
@@ -2674,7 +2681,7 @@ static int __bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 static void __bnxt_poll_work_done(struct bnxt *bp, struct bnxt_napi *bnapi,
 				  int budget)
 {
-	if (bnapi->tx_pkts && !bnapi->tx_fault)
+	if ((bnapi->events & BNXT_TX_CMP_EVENT) && !bnapi->tx_fault)
 		bnapi->tx_int(bp, bnapi, budget);
 
 	if ((bnapi->events & BNXT_RX_EVENT) && !(bnapi->in_reset)) {
@@ -2687,7 +2694,7 @@ static void __bnxt_poll_work_done(struct bnxt *bp, struct bnxt_napi *bnapi,
 
 		bnxt_db_write(bp, &rxr->rx_agg_db, rxr->rx_agg_prod);
 	}
-	bnapi->events = 0;
+	bnapi->events &= BNXT_TX_CMP_EVENT;
 }
 
 static int bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
@@ -4515,6 +4522,7 @@ static void bnxt_clear_ring_indices(struct bnxt *bp)
 		if (txr) {
 			txr->tx_prod = 0;
 			txr->tx_cons = 0;
+			txr->tx_hw_cons = 0;
 		}
 
 		rxr = bnapi->rx_ring;
@@ -4524,6 +4532,7 @@ static void bnxt_clear_ring_indices(struct bnxt *bp)
 			rxr->rx_sw_agg_prod = 0;
 			rxr->rx_next_cons = 0;
 		}
+		bnapi->events = 0;
 	}
 }
 
@@ -9528,8 +9537,6 @@ static void bnxt_enable_napi(struct bnxt *bp)
 		cpr = &bnapi->cp_ring;
 		bnapi->in_reset = false;
 
-		bnapi->tx_pkts = 0;
-
 		if (bnapi->rx_ring) {
 			INIT_WORK(&cpr->dim.work, bnxt_dim_work);
 			cpr->dim.mode = DIM_CQ_PERIOD_MODE_START_FROM_EQE;
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index c7895e7d78d5..cf22aae91f70 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -68,6 +68,12 @@ struct tx_bd {
 #define SET_TX_OPAQUE(bp, idx, bds)					\
 	(((bds) << TX_OPAQUE_BDS_SHIFT) | ((idx) & (bp)->tx_ring_mask))
 
+#define TX_OPAQUE_IDX(opq)	((opq) & TX_OPAQUE_IDX_MASK)
+#define TX_OPAQUE_BDS(opq)	(((opq) & TX_OPAQUE_BDS_MASK) >>	\
+				 TX_OPAQUE_BDS_SHIFT)
+#define TX_OPAQUE_PROD(bp, opq)	((TX_OPAQUE_IDX(opq) + TX_OPAQUE_BDS(opq)) &\
+				 (bp)->tx_ring_mask)
+
 struct tx_bd_ext {
 	__le32 tx_bd_hsize_lflags;
 	#define TX_BD_FLAGS_TCP_UDP_CHKSUM			(1 << 0)
@@ -709,6 +715,7 @@ struct nqe_cn {
 #define BNXT_AGG_EVENT		2
 #define BNXT_TX_EVENT		4
 #define BNXT_REDIRECT_EVENT	8
+#define BNXT_TX_CMP_EVENT	0x10
 
 struct bnxt_sw_tx_bd {
 	union {
@@ -801,6 +808,7 @@ struct bnxt_tx_ring_info {
 	struct bnxt_napi	*bnapi;
 	u16			tx_prod;
 	u16			tx_cons;
+	u16			tx_hw_cons;
 	u16			txq_index;
 	u8			kick_pending;
 	struct bnxt_db_info	tx_db;
@@ -1027,7 +1035,6 @@ struct bnxt_napi {
 
 	void			(*tx_int)(struct bnxt *, struct bnxt_napi *,
 					  int budget);
-	int			tx_pkts;
 	u8			events;
 	u8			tx_fault:1;
 
@@ -2367,7 +2374,7 @@ int bnxt_reserve_rings(struct bnxt *bp, bool irq_re_init);
 void bnxt_tx_disable(struct bnxt *bp);
 void bnxt_tx_enable(struct bnxt *bp);
 void bnxt_sched_reset_txr(struct bnxt *bp, struct bnxt_tx_ring_info *txr,
-			  int idx);
+			  u16 curr);
 void bnxt_report_link(struct bnxt *bp);
 int bnxt_update_link(struct bnxt *bp, bool chng_link_state);
 int bnxt_hwrm_set_pause(struct bnxt *);
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
index 3e5144aafb0c..23476100fad2 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
@@ -129,17 +129,17 @@ void bnxt_tx_int_xdp(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 {
 	struct bnxt_tx_ring_info *txr = bnapi->tx_ring;
 	struct bnxt_rx_ring_info *rxr = bnapi->rx_ring;
+	u16 tx_hw_cons = txr->tx_hw_cons;
 	bool rx_doorbell_needed = false;
-	int nr_pkts = bnapi->tx_pkts;
 	struct bnxt_sw_tx_bd *tx_buf;
 	u16 tx_cons = txr->tx_cons;
 	u16 last_tx_cons = tx_cons;
-	int i, j, frags;
+	int j, frags;
 
 	if (!budget)
 		return;
 
-	for (i = 0; i < nr_pkts; i++) {
+	while (tx_cons != tx_hw_cons) {
 		tx_buf = &txr->tx_buf_ring[tx_cons];
 
 		if (tx_buf->action == XDP_REDIRECT) {
@@ -164,13 +164,13 @@ void bnxt_tx_int_xdp(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 				page_pool_recycle_direct(rxr->page_pool, tx_buf->page);
 			}
 		} else {
-			bnxt_sched_reset_txr(bp, txr, i);
+			bnxt_sched_reset_txr(bp, txr, tx_cons);
 			return;
 		}
 		tx_cons = NEXT_TX(tx_cons);
 	}
 
-	bnapi->tx_pkts = 0;
+	bnapi->events &= ~BNXT_TX_CMP_EVENT;
 	WRITE_ONCE(txr->tx_cons, tx_cons);
 	if (rx_doorbell_needed) {
 		tx_buf = &txr->tx_buf_ring[last_tx_cons];
@@ -275,7 +275,7 @@ bool bnxt_rx_xdp(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, u16 cons,
 	case XDP_TX:
 		rx_buf = &rxr->rx_buf_ring[cons];
 		mapping = rx_buf->mapping - bp->rx_dma_offset;
-		*event = 0;
+		*event &= BNXT_TX_CMP_EVENT;
 
 		if (unlikely(xdp_buff_has_frags(&xdp))) {
 			struct skb_shared_info *sinfo = xdp_get_shared_info_from_buff(&xdp);
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 03/13] bnxt_en: Restructure cp_ring_arr in struct bnxt_cp_ring_info

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 10653 bytes --]

The cp_ring_arr is currently a fixed array of 2 pointers for the
TX and RX completion rings.  These pointers are allocated during
ring initialization.  Currntly, we support up to 2 completion rings
for each MSIX.  In order to support more completion rings, we change
this fixed array to a pointer and allocate the required entries
during ring initialization.  This patch keeps the current scheme of
allocating only 2 entries when needed.  Later patches will expand
and allocate more entries when required.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 134 ++++++++----------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |   3 +-
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |   2 +-
 3 files changed, 66 insertions(+), 73 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 4dfe0b66c5f7..585120369935 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -2834,14 +2834,11 @@ static int __bnxt_poll_cqs(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 	struct bnxt_cp_ring_info *cpr = &bnapi->cp_ring;
 	int i, work_done = 0;
 
-	for (i = 0; i < 2; i++) {
-		struct bnxt_cp_ring_info *cpr2 = cpr->cp_ring_arr[i];
+	for (i = 0; i < cpr->cp_ring_count; i++) {
+		struct bnxt_cp_ring_info *cpr2 = &cpr->cp_ring_arr[i];
 
-		if (cpr2) {
-			work_done += __bnxt_poll_work(bp, cpr2,
-						      budget - work_done);
-			cpr->has_more_work |= cpr2->has_more_work;
-		}
+		work_done += __bnxt_poll_work(bp, cpr2, budget - work_done);
+		cpr->has_more_work |= cpr2->has_more_work;
 	}
 	return work_done;
 }
@@ -2852,11 +2849,11 @@ static void __bnxt_poll_cqs_done(struct bnxt *bp, struct bnxt_napi *bnapi,
 	struct bnxt_cp_ring_info *cpr = &bnapi->cp_ring;
 	int i;
 
-	for (i = 0; i < 2; i++) {
-		struct bnxt_cp_ring_info *cpr2 = cpr->cp_ring_arr[i];
+	for (i = 0; i < cpr->cp_ring_count; i++) {
+		struct bnxt_cp_ring_info *cpr2 = &cpr->cp_ring_arr[i];
 		struct bnxt_db_info *db;
 
-		if (cpr2 && cpr2->had_work_done) {
+		if (cpr2->had_work_done) {
 			db = &cpr2->cp_db;
 			bnxt_writeq(bp, db->db_key64 | dbr_type |
 				    RING_CMP(cpr2->cp_raw_cons), db->doorbell);
@@ -2915,7 +2912,7 @@ static int bnxt_poll_p5(struct napi_struct *napi, int budget)
 			if (budget && work_done >= budget && idx == BNXT_RX_HDL)
 				break;
 
-			cpr2 = cpr->cp_ring_arr[idx];
+			cpr2 = &cpr->cp_ring_arr[idx];
 			work_done += __bnxt_poll_work(bp, cpr2,
 						      budget - work_done);
 			cpr->has_more_work |= cpr2->has_more_work;
@@ -2930,8 +2927,8 @@ static int bnxt_poll_p5(struct napi_struct *napi, int budget)
 		BNXT_DB_NQ_P5(&cpr->cp_db, raw_cons);
 	}
 poll_done:
-	cpr_rx = cpr->cp_ring_arr[BNXT_RX_HDL];
-	if (cpr_rx && (bp->flags & BNXT_FLAG_DIM)) {
+	cpr_rx = &cpr->cp_ring_arr[BNXT_RX_HDL];
+	if (cpr_rx->bnapi && (bp->flags & BNXT_FLAG_DIM)) {
 		struct dim_sample dim_sample = {};
 
 		dim_update_sample(cpr->event_ctr,
@@ -3541,36 +3538,33 @@ static void bnxt_free_cp_rings(struct bnxt *bp)
 
 		bnxt_free_ring(bp, &ring->ring_mem);
 
-		for (j = 0; j < 2; j++) {
-			struct bnxt_cp_ring_info *cpr2 = cpr->cp_ring_arr[j];
+		if (!cpr->cp_ring_arr)
+			continue;
 
-			if (cpr2) {
-				ring = &cpr2->cp_ring_struct;
-				bnxt_free_ring(bp, &ring->ring_mem);
-				bnxt_free_cp_arrays(cpr2);
-				kfree(cpr2);
-				cpr->cp_ring_arr[j] = NULL;
-			}
+		for (j = 0; j < cpr->cp_ring_count; j++) {
+			struct bnxt_cp_ring_info *cpr2 = &cpr->cp_ring_arr[j];
+
+			ring = &cpr2->cp_ring_struct;
+			bnxt_free_ring(bp, &ring->ring_mem);
+			bnxt_free_cp_arrays(cpr2);
 		}
+		kfree(cpr->cp_ring_arr);
+		cpr->cp_ring_arr = NULL;
+		cpr->cp_ring_count = 0;
 	}
 }
 
-static struct bnxt_cp_ring_info *bnxt_alloc_cp_sub_ring(struct bnxt *bp)
+static int bnxt_alloc_cp_sub_ring(struct bnxt *bp,
+				  struct bnxt_cp_ring_info *cpr)
 {
 	struct bnxt_ring_mem_info *rmem;
 	struct bnxt_ring_struct *ring;
-	struct bnxt_cp_ring_info *cpr;
 	int rc;
 
-	cpr = kzalloc(sizeof(*cpr), GFP_KERNEL);
-	if (!cpr)
-		return NULL;
-
 	rc = bnxt_alloc_cp_arrays(cpr, bp->cp_nr_pages);
 	if (rc) {
 		bnxt_free_cp_arrays(cpr);
-		kfree(cpr);
-		return NULL;
+		return -ENOMEM;
 	}
 	ring = &cpr->cp_ring_struct;
 	rmem = &ring->ring_mem;
@@ -3583,10 +3577,8 @@ static struct bnxt_cp_ring_info *bnxt_alloc_cp_sub_ring(struct bnxt *bp)
 	if (rc) {
 		bnxt_free_ring(bp, rmem);
 		bnxt_free_cp_arrays(cpr);
-		kfree(cpr);
-		cpr = NULL;
 	}
-	return cpr;
+	return rc;
 }
 
 static int bnxt_alloc_cp_rings(struct bnxt *bp)
@@ -3598,7 +3590,7 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 	ulp_base_vec = bnxt_get_ulp_msix_base(bp);
 	for (i = 0; i < bp->cp_nr_rings; i++) {
 		struct bnxt_napi *bnapi = bp->bnapi[i];
-		struct bnxt_cp_ring_info *cpr;
+		struct bnxt_cp_ring_info *cpr, *cpr2;
 		struct bnxt_ring_struct *ring;
 
 		if (!bnapi)
@@ -3620,23 +3612,27 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 		if (!(bp->flags & BNXT_FLAG_CHIP_P5))
 			continue;
 
-		if (i < bp->rx_nr_rings) {
-			struct bnxt_cp_ring_info *cpr2 =
-				bnxt_alloc_cp_sub_ring(bp);
+		cpr->cp_ring_count = 2;
+		cpr->cp_ring_arr = kcalloc(cpr->cp_ring_count, sizeof(*cpr),
+					   GFP_KERNEL);
+		if (!cpr->cp_ring_arr) {
+			cpr->cp_ring_count = 0;
+			return -ENOMEM;
+		}
 
-			cpr->cp_ring_arr[BNXT_RX_HDL] = cpr2;
-			if (!cpr2)
-				return -ENOMEM;
+		if (i < bp->rx_nr_rings) {
+			cpr2 = &cpr->cp_ring_arr[BNXT_RX_HDL];
+			rc = bnxt_alloc_cp_sub_ring(bp, cpr2);
+			if (rc)
+				return rc;
 			cpr2->bnapi = bnapi;
 		}
 		if ((sh && i < bp->tx_nr_rings) ||
 		    (!sh && i >= bp->rx_nr_rings)) {
-			struct bnxt_cp_ring_info *cpr2 =
-				bnxt_alloc_cp_sub_ring(bp);
-
-			cpr->cp_ring_arr[BNXT_TX_HDL] = cpr2;
-			if (!cpr2)
-				return -ENOMEM;
+			cpr2 = &cpr->cp_ring_arr[BNXT_TX_HDL];
+			rc = bnxt_alloc_cp_sub_ring(bp, cpr2);
+			if (rc)
+				return rc;
 			cpr2->bnapi = bnapi;
 		}
 	}
@@ -3822,11 +3818,10 @@ static void bnxt_init_cp_rings(struct bnxt *bp)
 		ring->fw_ring_id = INVALID_HW_RING_ID;
 		cpr->rx_ring_coal.coal_ticks = bp->rx_coal.coal_ticks;
 		cpr->rx_ring_coal.coal_bufs = bp->rx_coal.coal_bufs;
-		for (j = 0; j < 2; j++) {
-			struct bnxt_cp_ring_info *cpr2 = cpr->cp_ring_arr[j];
-
-			if (!cpr2)
-				continue;
+		if (!cpr->cp_ring_arr)
+			continue;
+		for (j = 0; j < cpr->cp_ring_count; j++) {
+			struct bnxt_cp_ring_info *cpr2 = &cpr->cp_ring_arr[j];
 
 			ring = &cpr2->cp_ring_struct;
 			ring->fw_ring_id = INVALID_HW_RING_ID;
@@ -5251,7 +5246,7 @@ static u16 bnxt_cp_ring_for_rx(struct bnxt *bp, struct bnxt_rx_ring_info *rxr)
 		struct bnxt_napi *bnapi = rxr->bnapi;
 		struct bnxt_cp_ring_info *cpr;
 
-		cpr = bnapi->cp_ring.cp_ring_arr[BNXT_RX_HDL];
+		cpr = &bnapi->cp_ring.cp_ring_arr[BNXT_RX_HDL];
 		return cpr->cp_ring_struct.fw_ring_id;
 	} else {
 		return bnxt_cp_ring_from_grp(bp, &rxr->rx_ring_struct);
@@ -5264,7 +5259,7 @@ static u16 bnxt_cp_ring_for_tx(struct bnxt *bp, struct bnxt_tx_ring_info *txr)
 		struct bnxt_napi *bnapi = txr->bnapi;
 		struct bnxt_cp_ring_info *cpr;
 
-		cpr = bnapi->cp_ring.cp_ring_arr[BNXT_TX_HDL];
+		cpr = &bnapi->cp_ring.cp_ring_arr[BNXT_TX_HDL];
 		return cpr->cp_ring_struct.fw_ring_id;
 	} else {
 		return bnxt_cp_ring_from_grp(bp, &txr->tx_ring_struct);
@@ -6032,7 +6027,7 @@ static int bnxt_hwrm_ring_alloc(struct bnxt *bp)
 			u32 type2 = HWRM_RING_ALLOC_CMPL;
 
 			cpr = &bnapi->cp_ring;
-			cpr2 = cpr->cp_ring_arr[BNXT_TX_HDL];
+			cpr2 = &cpr->cp_ring_arr[BNXT_TX_HDL];
 			ring = &cpr2->cp_ring_struct;
 			ring->handle = BNXT_TX_HDL;
 			map_idx = bnapi->index;
@@ -6071,7 +6066,7 @@ static int bnxt_hwrm_ring_alloc(struct bnxt *bp)
 			u32 type2 = HWRM_RING_ALLOC_CMPL;
 			struct bnxt_cp_ring_info *cpr2;
 
-			cpr2 = cpr->cp_ring_arr[BNXT_RX_HDL];
+			cpr2 = &cpr->cp_ring_arr[BNXT_RX_HDL];
 			ring = &cpr2->cp_ring_struct;
 			ring->handle = BNXT_RX_HDL;
 			rc = hwrm_ring_alloc_send_msg(bp, ring, type2, map_idx);
@@ -6218,18 +6213,16 @@ static void bnxt_hwrm_ring_free(struct bnxt *bp, bool close_path)
 		struct bnxt_ring_struct *ring;
 		int j;
 
-		for (j = 0; j < 2; j++) {
-			struct bnxt_cp_ring_info *cpr2 = cpr->cp_ring_arr[j];
-
-			if (cpr2) {
-				ring = &cpr2->cp_ring_struct;
-				if (ring->fw_ring_id == INVALID_HW_RING_ID)
-					continue;
-				hwrm_ring_free_send_msg(bp, ring,
-					RING_FREE_REQ_RING_TYPE_L2_CMPL,
-					INVALID_HW_RING_ID);
-				ring->fw_ring_id = INVALID_HW_RING_ID;
-			}
+		for (j = 0; j < cpr->cp_ring_count && cpr->cp_ring_arr; j++) {
+			struct bnxt_cp_ring_info *cpr2 = &cpr->cp_ring_arr[j];
+
+			ring = &cpr2->cp_ring_struct;
+			if (ring->fw_ring_id == INVALID_HW_RING_ID)
+				continue;
+			hwrm_ring_free_send_msg(bp, ring,
+						RING_FREE_REQ_RING_TYPE_L2_CMPL,
+						INVALID_HW_RING_ID);
+			ring->fw_ring_id = INVALID_HW_RING_ID;
 		}
 		ring = &cpr->cp_ring_struct;
 		if (ring->fw_ring_id != INVALID_HW_RING_ID) {
@@ -12005,12 +11998,11 @@ static void bnxt_chk_missed_irq(struct bnxt *bp)
 			continue;
 
 		cpr = &bnapi->cp_ring;
-		for (j = 0; j < 2; j++) {
-			struct bnxt_cp_ring_info *cpr2 = cpr->cp_ring_arr[j];
+		for (j = 0; j < cpr->cp_ring_count; j++) {
+			struct bnxt_cp_ring_info *cpr2 = &cpr->cp_ring_arr[j];
 			u32 val[2];
 
-			if (!cpr2 || cpr2->has_more_work ||
-			    !bnxt_has_work(bp, cpr2))
+			if (cpr2->has_more_work || !bnxt_has_work(bp, cpr2))
 				continue;
 
 			if (cpr2->cp_raw_cons != cpr2->last_cp_raw_cons) {
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index cf22aae91f70..429df1cf4a6a 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -1019,7 +1019,8 @@ struct bnxt_cp_ring_info {
 
 	struct bnxt_ring_struct	cp_ring_struct;
 
-	struct bnxt_cp_ring_info *cp_ring_arr[2];
+	int			cp_ring_count;
+	struct bnxt_cp_ring_info *cp_ring_arr;
 #define BNXT_RX_HDL	0
 #define BNXT_TX_HDL	1
 };
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index f3f384773ac0..675e37700289 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -3941,7 +3941,7 @@ static int bnxt_run_loopback(struct bnxt *bp)
 
 	cpr = &rxr->bnapi->cp_ring;
 	if (bp->flags & BNXT_FLAG_CHIP_P5)
-		cpr = cpr->cp_ring_arr[BNXT_RX_HDL];
+		cpr = &cpr->cp_ring_arr[BNXT_RX_HDL];
 	pkt_size = min(bp->dev->mtu + ETH_HLEN, bp->rx_copy_thresh);
 	skb = netdev_alloc_skb(bp->dev, pkt_size);
 	if (!skb)
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 04/13] bnxt_en: Add completion ring pointer in TX and RX ring structures

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 5992 bytes --]

From the TX or RX ring structure, we need to find the corresponding
completion ring during initialization.  On P5 chips, we use the MSIX/napi
entry to locate the completion ring because there is only one RX/TX
ring per MSIX.  To allow multiple TX rings for each MSIX, we need
to add a direct pointer from the TX ring and RX ring structures.
This also simplifies the existing logic.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 40 ++++++++-----------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  2 +
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |  2 +-
 3 files changed, 19 insertions(+), 25 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 585120369935..11a85cb28517 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -3584,11 +3584,11 @@ static int bnxt_alloc_cp_sub_ring(struct bnxt *bp,
 static int bnxt_alloc_cp_rings(struct bnxt *bp)
 {
 	bool sh = !!(bp->flags & BNXT_FLAG_SHARED_RINGS);
-	int i, rc, ulp_base_vec, ulp_msix;
+	int i, j, rc, ulp_base_vec, ulp_msix;
 
 	ulp_msix = bnxt_get_ulp_msix_num(bp);
 	ulp_base_vec = bnxt_get_ulp_msix_base(bp);
-	for (i = 0; i < bp->cp_nr_rings; i++) {
+	for (i = 0, j = 0; i < bp->cp_nr_rings; i++) {
 		struct bnxt_napi *bnapi = bp->bnapi[i];
 		struct bnxt_cp_ring_info *cpr, *cpr2;
 		struct bnxt_ring_struct *ring;
@@ -3626,6 +3626,7 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 			if (rc)
 				return rc;
 			cpr2->bnapi = bnapi;
+			bp->rx_ring[i].rx_cpr = cpr2;
 		}
 		if ((sh && i < bp->tx_nr_rings) ||
 		    (!sh && i >= bp->rx_nr_rings)) {
@@ -3634,6 +3635,7 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 			if (rc)
 				return rc;
 			cpr2->bnapi = bnapi;
+			bp->tx_ring[j++].tx_cpr = cpr2;
 		}
 	}
 	return 0;
@@ -4654,6 +4656,8 @@ static int bnxt_alloc_mem(struct bnxt *bp, bool irq_re_init)
 					BNXT_RMEM_RING_PTE_FLAG;
 				rxr->rx_agg_ring_struct.ring_mem.flags =
 					BNXT_RMEM_RING_PTE_FLAG;
+			} else {
+				rxr->rx_cpr =  &bp->bnapi[i]->cp_ring;
 			}
 			rxr->bnapi = bp->bnapi[i];
 			bp->bnapi[i]->rx_ring = &bp->rx_ring[i];
@@ -4682,6 +4686,8 @@ static int bnxt_alloc_mem(struct bnxt *bp, bool irq_re_init)
 			if (bp->flags & BNXT_FLAG_CHIP_P5)
 				txr->tx_ring_struct.ring_mem.flags =
 					BNXT_RMEM_RING_PTE_FLAG;
+			else
+				txr->tx_cpr =  &bp->bnapi[i]->cp_ring;
 			txr->bnapi = bp->bnapi[j];
 			bp->bnapi[j]->tx_ring = txr;
 			bp->tx_ring_map[i] = bp->tx_nr_rings_xdp + i;
@@ -5242,28 +5248,18 @@ static u16 bnxt_cp_ring_from_grp(struct bnxt *bp, struct bnxt_ring_struct *ring)
 
 static u16 bnxt_cp_ring_for_rx(struct bnxt *bp, struct bnxt_rx_ring_info *rxr)
 {
-	if (bp->flags & BNXT_FLAG_CHIP_P5) {
-		struct bnxt_napi *bnapi = rxr->bnapi;
-		struct bnxt_cp_ring_info *cpr;
-
-		cpr = &bnapi->cp_ring.cp_ring_arr[BNXT_RX_HDL];
-		return cpr->cp_ring_struct.fw_ring_id;
-	} else {
+	if (bp->flags & BNXT_FLAG_CHIP_P5)
+		return rxr->rx_cpr->cp_ring_struct.fw_ring_id;
+	else
 		return bnxt_cp_ring_from_grp(bp, &rxr->rx_ring_struct);
-	}
 }
 
 static u16 bnxt_cp_ring_for_tx(struct bnxt *bp, struct bnxt_tx_ring_info *txr)
 {
-	if (bp->flags & BNXT_FLAG_CHIP_P5) {
-		struct bnxt_napi *bnapi = txr->bnapi;
-		struct bnxt_cp_ring_info *cpr;
-
-		cpr = &bnapi->cp_ring.cp_ring_arr[BNXT_TX_HDL];
-		return cpr->cp_ring_struct.fw_ring_id;
-	} else {
+	if (bp->flags & BNXT_FLAG_CHIP_P5)
+		return txr->tx_cpr->cp_ring_struct.fw_ring_id;
+	else
 		return bnxt_cp_ring_from_grp(bp, &txr->tx_ring_struct);
-	}
 }
 
 static int bnxt_alloc_rss_indir_tbl(struct bnxt *bp)
@@ -6022,12 +6018,10 @@ static int bnxt_hwrm_ring_alloc(struct bnxt *bp)
 		u32 map_idx;
 
 		if (bp->flags & BNXT_FLAG_CHIP_P5) {
+			struct bnxt_cp_ring_info *cpr2 = txr->tx_cpr;
 			struct bnxt_napi *bnapi = txr->bnapi;
-			struct bnxt_cp_ring_info *cpr, *cpr2;
 			u32 type2 = HWRM_RING_ALLOC_CMPL;
 
-			cpr = &bnapi->cp_ring;
-			cpr2 = &cpr->cp_ring_arr[BNXT_TX_HDL];
 			ring = &cpr2->cp_ring_struct;
 			ring->handle = BNXT_TX_HDL;
 			map_idx = bnapi->index;
@@ -6062,11 +6056,9 @@ static int bnxt_hwrm_ring_alloc(struct bnxt *bp)
 			bnxt_db_write(bp, &rxr->rx_db, rxr->rx_prod);
 		bp->grp_info[map_idx].rx_fw_ring_id = ring->fw_ring_id;
 		if (bp->flags & BNXT_FLAG_CHIP_P5) {
-			struct bnxt_cp_ring_info *cpr = &bnapi->cp_ring;
+			struct bnxt_cp_ring_info *cpr2 = rxr->rx_cpr;
 			u32 type2 = HWRM_RING_ALLOC_CMPL;
-			struct bnxt_cp_ring_info *cpr2;
 
-			cpr2 = &cpr->cp_ring_arr[BNXT_RX_HDL];
 			ring = &cpr2->cp_ring_struct;
 			ring->handle = BNXT_RX_HDL;
 			rc = hwrm_ring_alloc_send_msg(bp, ring, type2, map_idx);
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 429df1cf4a6a..c04089e7ac39 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -806,6 +806,7 @@ struct bnxt_db_info {
 
 struct bnxt_tx_ring_info {
 	struct bnxt_napi	*bnapi;
+	struct bnxt_cp_ring_info	*tx_cpr;
 	u16			tx_prod;
 	u16			tx_cons;
 	u16			tx_hw_cons;
@@ -916,6 +917,7 @@ struct bnxt_tpa_idx_map {
 
 struct bnxt_rx_ring_info {
 	struct bnxt_napi	*bnapi;
+	struct bnxt_cp_ring_info	*rx_cpr;
 	u16			rx_prod;
 	u16			rx_agg_prod;
 	u16			rx_sw_agg_prod;
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index 675e37700289..18c06158fead 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -3941,7 +3941,7 @@ static int bnxt_run_loopback(struct bnxt *bp)
 
 	cpr = &rxr->bnapi->cp_ring;
 	if (bp->flags & BNXT_FLAG_CHIP_P5)
-		cpr = &cpr->cp_ring_arr[BNXT_RX_HDL];
+		cpr = rxr->rx_cpr;
 	pkt_size = min(bp->dev->mtu + ETH_HLEN, bp->rx_copy_thresh);
 	skb = netdev_alloc_skb(bp->dev, pkt_size);
 	if (!skb)
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 05/13] bnxt_en: Remove BNXT_RX_HDL and BNXT_TX_HDL

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 5678 bytes --]

These 2 constants were used for the one RX and one TX completion
ring pointer in the cpr->cp_ring_arr fixed array.  Now that we've
changed to allocating the array for the exact number of entries to
support more TX rings, we no longer use these constants.

The array index as well as the type of completion ring (RX/TX) are
now encoded in the handle for the completion ring.  This will allow
us to locate the completion ring during NAPI for any number of
completion rings sharing the same MSIX.  In the following patches,
we'll be adding support for more TX rings associated with the same
MSIX vector.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 51 +++++++++++++----------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h | 17 +++++++-
 2 files changed, 44 insertions(+), 24 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 11a85cb28517..a4f7fa17daf8 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -2906,12 +2906,15 @@ static int bnxt_poll_p5(struct napi_struct *napi, int budget)
 
 		if (nqcmp->type == cpu_to_le16(NQ_CN_TYPE_CQ_NOTIFICATION)) {
 			u32 idx = le32_to_cpu(nqcmp->cq_handle_low);
+			u32 cq_type = BNXT_NQ_HDL_TYPE(idx);
 			struct bnxt_cp_ring_info *cpr2;
 
 			/* No more budget for RX work */
-			if (budget && work_done >= budget && idx == BNXT_RX_HDL)
+			if (budget && work_done >= budget &&
+			    cq_type == BNXT_NQ_HDL_TYPE_RX)
 				break;
 
+			idx = BNXT_NQ_HDL_IDX(idx);
 			cpr2 = &cpr->cp_ring_arr[idx];
 			work_done += __bnxt_poll_work(bp, cpr2,
 						      budget - work_done);
@@ -2927,8 +2930,9 @@ static int bnxt_poll_p5(struct napi_struct *napi, int budget)
 		BNXT_DB_NQ_P5(&cpr->cp_db, raw_cons);
 	}
 poll_done:
-	cpr_rx = &cpr->cp_ring_arr[BNXT_RX_HDL];
-	if (cpr_rx->bnapi && (bp->flags & BNXT_FLAG_DIM)) {
+	cpr_rx = &cpr->cp_ring_arr[0];
+	if (cpr_rx->cp_ring_type == BNXT_NQ_HDL_TYPE_RX &&
+	    (bp->flags & BNXT_FLAG_DIM)) {
 		struct dim_sample dim_sample = {};
 
 		dim_update_sample(cpr->event_ctr,
@@ -3592,6 +3596,7 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 		struct bnxt_napi *bnapi = bp->bnapi[i];
 		struct bnxt_cp_ring_info *cpr, *cpr2;
 		struct bnxt_ring_struct *ring;
+		int cp_count = 0, k;
 
 		if (!bnapi)
 			continue;
@@ -3612,30 +3617,32 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 		if (!(bp->flags & BNXT_FLAG_CHIP_P5))
 			continue;
 
-		cpr->cp_ring_count = 2;
-		cpr->cp_ring_arr = kcalloc(cpr->cp_ring_count, sizeof(*cpr),
+		if (i < bp->rx_nr_rings)
+			cp_count++;
+		if ((sh && i < bp->tx_nr_rings) ||
+		    (!sh && i >= bp->rx_nr_rings))
+			cp_count++;
+
+		cpr->cp_ring_arr = kcalloc(cp_count, sizeof(*cpr),
 					   GFP_KERNEL);
-		if (!cpr->cp_ring_arr) {
-			cpr->cp_ring_count = 0;
+		if (!cpr->cp_ring_arr)
 			return -ENOMEM;
-		}
+		cpr->cp_ring_count = cp_count;
 
-		if (i < bp->rx_nr_rings) {
-			cpr2 = &cpr->cp_ring_arr[BNXT_RX_HDL];
-			rc = bnxt_alloc_cp_sub_ring(bp, cpr2);
-			if (rc)
-				return rc;
-			cpr2->bnapi = bnapi;
-			bp->rx_ring[i].rx_cpr = cpr2;
-		}
-		if ((sh && i < bp->tx_nr_rings) ||
-		    (!sh && i >= bp->rx_nr_rings)) {
-			cpr2 = &cpr->cp_ring_arr[BNXT_TX_HDL];
+		for (k = 0; k < cp_count; k++) {
+			cpr2 = &cpr->cp_ring_arr[k];
 			rc = bnxt_alloc_cp_sub_ring(bp, cpr2);
 			if (rc)
 				return rc;
 			cpr2->bnapi = bnapi;
-			bp->tx_ring[j++].tx_cpr = cpr2;
+			cpr2->cp_idx = k;
+			if (!k && i < bp->rx_nr_rings) {
+				bp->rx_ring[i].rx_cpr = cpr2;
+				cpr2->cp_ring_type = BNXT_NQ_HDL_TYPE_RX;
+			} else {
+				bp->tx_ring[j++].tx_cpr = cpr2;
+				cpr2->cp_ring_type = BNXT_NQ_HDL_TYPE_TX;
+			}
 		}
 	}
 	return 0;
@@ -6023,7 +6030,7 @@ static int bnxt_hwrm_ring_alloc(struct bnxt *bp)
 			u32 type2 = HWRM_RING_ALLOC_CMPL;
 
 			ring = &cpr2->cp_ring_struct;
-			ring->handle = BNXT_TX_HDL;
+			ring->handle = BNXT_SET_NQ_HDL(cpr2);
 			map_idx = bnapi->index;
 			rc = hwrm_ring_alloc_send_msg(bp, ring, type2, map_idx);
 			if (rc)
@@ -6060,7 +6067,7 @@ static int bnxt_hwrm_ring_alloc(struct bnxt *bp)
 			u32 type2 = HWRM_RING_ALLOC_CMPL;
 
 			ring = &cpr2->cp_ring_struct;
-			ring->handle = BNXT_RX_HDL;
+			ring->handle = BNXT_SET_NQ_HDL(cpr2);
 			rc = hwrm_ring_alloc_send_msg(bp, ring, type2, map_idx);
 			if (rc)
 				goto err_out;
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index c04089e7ac39..efb0db54575b 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -543,6 +543,19 @@ struct nqe_cn {
 	__le32	cq_handle_high;
 };
 
+#define BNXT_NQ_HDL_IDX_MASK	0x00ffffff
+#define BNXT_NQ_HDL_TYPE_MASK	0xff000000
+#define BNXT_NQ_HDL_TYPE_SHIFT	24
+#define BNXT_NQ_HDL_TYPE_RX	0x00
+#define BNXT_NQ_HDL_TYPE_TX	0x01
+
+#define BNXT_NQ_HDL_IDX(hdl)	((hdl) & BNXT_NQ_HDL_IDX_MASK)
+#define BNXT_NQ_HDL_TYPE(hdl)	(((hdl) & BNXT_NQ_HDL_TYPE_MASK) >>	\
+				 BNXT_NQ_HDL_TYPE_SHIFT)
+
+#define BNXT_SET_NQ_HDL(cpr)						\
+	(((cpr)->cp_ring_type << BNXT_NQ_HDL_TYPE_SHIFT) | (cpr)->cp_idx)
+
 #define DB_IDX_MASK						0xffffff
 #define DB_IDX_VALID						(0x1 << 26)
 #define DB_IRQ_DIS						(0x1 << 27)
@@ -997,6 +1010,8 @@ struct bnxt_cp_ring_info {
 
 	u8			had_work_done:1;
 	u8			has_more_work:1;
+	u8			cp_ring_type;
+	u8			cp_idx;
 
 	u32			last_cp_raw_cons;
 
@@ -1023,8 +1038,6 @@ struct bnxt_cp_ring_info {
 
 	int			cp_ring_count;
 	struct bnxt_cp_ring_info *cp_ring_arr;
-#define BNXT_RX_HDL	0
-#define BNXT_TX_HDL	1
 };
 
 struct bnxt_napi {
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 06/13] bnxt_en: Refactor bnxt_tx_int()

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 2207 bytes --]

bnxt_tx_int() processes the only one TX ring from the bnxt_napi pointer.
To prepare for more TX rings associated with the bnxt_napi structure,
add a new __bnxt_tx_int() function that takes the bnxt_tx_ring_info
pointer to process that one TX ring.  No functional change.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 17 ++++++++++++-----
 1 file changed, 12 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index a4f7fa17daf8..ad56ca9d3ceb 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -687,14 +687,14 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	return NETDEV_TX_OK;
 }
 
-static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
+static void __bnxt_tx_int(struct bnxt *bp, struct bnxt_tx_ring_info *txr,
+			  int budget)
 {
-	struct bnxt_tx_ring_info *txr = bnapi->tx_ring;
 	struct netdev_queue *txq = netdev_get_tx_queue(bp->dev, txr->txq_index);
-	u16 hw_cons = txr->tx_hw_cons;
-	u16 cons = txr->tx_cons;
 	struct pci_dev *pdev = bp->pdev;
+	u16 hw_cons = txr->tx_hw_cons;
 	unsigned int tx_bytes = 0;
+	u16 cons = txr->tx_cons;
 	int tx_pkts = 0;
 
 	while (cons != hw_cons) {
@@ -749,7 +749,6 @@ static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 		dev_consume_skb_any(skb);
 	}
 
-	bnapi->events &= ~BNXT_TX_CMP_EVENT;
 	WRITE_ONCE(txr->tx_cons, cons);
 
 	__netif_txq_completed_wake(txq, tx_pkts, tx_bytes,
@@ -757,6 +756,14 @@ static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 				   READ_ONCE(txr->dev_state) == BNXT_DEV_STATE_CLOSING);
 }
 
+static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
+{
+	struct bnxt_tx_ring_info *txr = bnapi->tx_ring;
+
+	__bnxt_tx_int(bp, txr, budget);
+	bnapi->events &= ~BNXT_TX_CMP_EVENT;
+}
+
 static struct page *__bnxt_alloc_rx_page(struct bnxt *bp, dma_addr_t *mapping,
 					 struct bnxt_rx_ring_info *rxr,
 					 unsigned int *offset,
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 07/13] bnxt_en: New encoding for the TX opaque field

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 3600 bytes --]

In order to support multiple TX rings on the same MSIX, we'll use the
upper byte of the TX opaque field to store the ring index in the new
tx_napi_idx field.  This tx_napi_idx field is currently always 0 until
more infrastructure is added in later patches.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     |  4 ++--
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     | 10 ++++++++--
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c |  2 +-
 3 files changed, 11 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index ad56ca9d3ceb..1a7f14d086f7 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -517,7 +517,7 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 		txbd->tx_bd_len_flags_type = tx_push->tx_bd_len_flags_type;
 		txbd->tx_bd_haddr = txr->data_mapping;
-		txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, prod, 2);
+		txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, txr, prod, 2);
 		prod = NEXT_TX(prod);
 		tx_push->tx_bd_opaque = txbd->tx_bd_opaque;
 		txbd = &txr->tx_desc_ring[TX_RING(prod)][TX_IDX(prod)];
@@ -562,7 +562,7 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		((last_frag + 2) << TX_BD_FLAGS_BD_CNT_SHIFT);
 
 	txbd->tx_bd_haddr = cpu_to_le64(mapping);
-	txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, prod, 2 + last_frag);
+	txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, txr, prod, 2 + last_frag);
 
 	prod = NEXT_TX(prod);
 	txbd1 = (struct tx_bd_ext *)
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index efb0db54575b..430538844178 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -64,11 +64,16 @@ struct tx_bd {
 #define TX_OPAQUE_IDX_MASK	0x0000ffff
 #define TX_OPAQUE_BDS_MASK	0x00ff0000
 #define TX_OPAQUE_BDS_SHIFT	16
+#define TX_OPAQUE_RING_MASK	0xff000000
+#define TX_OPAQUE_RING_SHIFT	24
 
-#define SET_TX_OPAQUE(bp, idx, bds)					\
-	(((bds) << TX_OPAQUE_BDS_SHIFT) | ((idx) & (bp)->tx_ring_mask))
+#define SET_TX_OPAQUE(bp, txr, idx, bds)				\
+	(((txr)->tx_napi_idx << TX_OPAQUE_RING_SHIFT) |			\
+	 ((bds) << TX_OPAQUE_BDS_SHIFT) | ((idx) & (bp)->tx_ring_mask))
 
 #define TX_OPAQUE_IDX(opq)	((opq) & TX_OPAQUE_IDX_MASK)
+#define TX_OPAQUE_RING(opq)	(((opq) & TX_OPAQUE_RING_MASK) >>	\
+				 TX_OPAQUE_RING_SHIFT)
 #define TX_OPAQUE_BDS(opq)	(((opq) & TX_OPAQUE_BDS_MASK) >>	\
 				 TX_OPAQUE_BDS_SHIFT)
 #define TX_OPAQUE_PROD(bp, opq)	((TX_OPAQUE_IDX(opq) + TX_OPAQUE_BDS(opq)) &\
@@ -824,6 +829,7 @@ struct bnxt_tx_ring_info {
 	u16			tx_cons;
 	u16			tx_hw_cons;
 	u16			txq_index;
+	u8			tx_napi_idx;
 	u8			kick_pending;
 	struct bnxt_db_info	tx_db;
 
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
index 23476100fad2..3515a12a6fea 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
@@ -52,7 +52,7 @@ struct bnxt_sw_tx_bd *bnxt_xmit_bd(struct bnxt *bp,
 		((num_frags + 1) << TX_BD_FLAGS_BD_CNT_SHIFT) |
 		bnxt_lhint_arr[len >> 9];
 	txbd->tx_bd_len_flags_type = cpu_to_le32(flags);
-	txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, prod, 1 + num_frags);
+	txbd->tx_bd_opaque = SET_TX_OPAQUE(bp, txr, prod, 1 + num_frags);
 	txbd->tx_bd_haddr = cpu_to_le64(mapping);
 
 	/* now let us fill up the frags into the next buffers */
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 08/13] bnxt_en: Refactor bnxt_hwrm_set_coal()

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 2606 bytes --]

Add 2 helper functions to set coalescing for each RX and TX rings.  This
will make it easier to expand the number of TX rings per MSIX in the
next patches.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 43 ++++++++++++++---------
 1 file changed, 26 insertions(+), 17 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 1a7f14d086f7..c84a72b666aa 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -6896,10 +6896,29 @@ int bnxt_hwrm_set_ring_coal(struct bnxt *bp, struct bnxt_napi *bnapi)
 	return hwrm_req_send(bp, req_rx);
 }
 
+static int
+bnxt_hwrm_set_rx_coal(struct bnxt *bp, struct bnxt_napi *bnapi,
+		      struct hwrm_ring_cmpl_ring_cfg_aggint_params_input *req)
+{
+	u16 ring_id = bnxt_cp_ring_for_rx(bp, bnapi->rx_ring);
+
+	req->ring_id = cpu_to_le16(ring_id);
+	return hwrm_req_send(bp, req);
+}
+
+static int
+bnxt_hwrm_set_tx_coal(struct bnxt *bp, struct bnxt_napi *bnapi,
+		      struct hwrm_ring_cmpl_ring_cfg_aggint_params_input *req)
+{
+	u16 ring_id = bnxt_cp_ring_for_tx(bp, bnapi->tx_ring);
+
+	req->ring_id = cpu_to_le16(ring_id);
+	return hwrm_req_send(bp, req);
+}
+
 int bnxt_hwrm_set_coal(struct bnxt *bp)
 {
-	struct hwrm_ring_cmpl_ring_cfg_aggint_params_input *req_rx, *req_tx,
-							   *req;
+	struct hwrm_ring_cmpl_ring_cfg_aggint_params_input *req_rx, *req_tx;
 	int i, rc;
 
 	rc = hwrm_req_init(bp, req_rx, HWRM_RING_CMPL_RING_CFG_AGGINT_PARAMS);
@@ -6920,18 +6939,11 @@ int bnxt_hwrm_set_coal(struct bnxt *bp)
 	for (i = 0; i < bp->cp_nr_rings; i++) {
 		struct bnxt_napi *bnapi = bp->bnapi[i];
 		struct bnxt_coal *hw_coal;
-		u16 ring_id;
 
-		req = req_rx;
-		if (!bnapi->rx_ring) {
-			ring_id = bnxt_cp_ring_for_tx(bp, bnapi->tx_ring);
-			req = req_tx;
-		} else {
-			ring_id = bnxt_cp_ring_for_rx(bp, bnapi->rx_ring);
-		}
-		req->ring_id = cpu_to_le16(ring_id);
-
-		rc = hwrm_req_send(bp, req);
+		if (!bnapi->rx_ring)
+			rc = bnxt_hwrm_set_tx_coal(bp, bnapi, req_tx);
+		else
+			rc = bnxt_hwrm_set_rx_coal(bp, bnapi, req_rx);
 		if (rc)
 			break;
 
@@ -6939,10 +6951,7 @@ int bnxt_hwrm_set_coal(struct bnxt *bp)
 			continue;
 
 		if (bnapi->rx_ring && bnapi->tx_ring) {
-			req = req_tx;
-			ring_id = bnxt_cp_ring_for_tx(bp, bnapi->tx_ring);
-			req->ring_id = cpu_to_le16(ring_id);
-			rc = hwrm_req_send(bp, req);
+			rc = bnxt_hwrm_set_tx_coal(bp, bnapi, req_tx);
 			if (rc)
 				break;
 		}
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 09/13] bnxt_en: Support up to 8 TX rings per MSIX

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 8357 bytes --]

For each mqprio TC, we allocate a set of TX rings to map to the new
hardware CoS queue.  Expand the tx_ring pointer in struct bnxt_napi
to an array of 8 to support up to 8 TX rings, one for each TC.
Only array entry 0 is used at this time.  The rest of the array
entries will be used in later patches.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 78 +++++++++++--------
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     | 12 ++-
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c |  4 +-
 3 files changed, 55 insertions(+), 39 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index c84a72b666aa..6002b834e898 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -758,9 +758,13 @@ static void __bnxt_tx_int(struct bnxt *bp, struct bnxt_tx_ring_info *txr,
 
 static void bnxt_tx_int(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 {
-	struct bnxt_tx_ring_info *txr = bnapi->tx_ring;
+	struct bnxt_tx_ring_info *txr;
+	int i;
 
-	__bnxt_tx_int(bp, txr, budget);
+	bnxt_for_each_napi_tx(i, bnapi, txr) {
+		if (txr->tx_hw_cons != txr->tx_cons)
+			__bnxt_tx_int(bp, txr, budget);
+	}
 	bnapi->events &= ~BNXT_TX_CMP_EVENT;
 }
 
@@ -2596,7 +2600,6 @@ static int __bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 {
 	struct bnxt_napi *bnapi = cpr->bnapi;
 	u32 raw_cons = cpr->cp_raw_cons;
-	struct bnxt_tx_ring_info *txr;
 	u32 cons;
 	int rx_pkts = 0;
 	u8 event = 0;
@@ -2604,7 +2607,6 @@ static int __bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 
 	cpr->has_more_work = 0;
 	cpr->had_work_done = 1;
-	txr = bnapi->tx_ring;
 	while (1) {
 		int rc;
 
@@ -2620,8 +2622,10 @@ static int __bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 		dma_rmb();
 		if (TX_CMP_TYPE(txcmp) == CMP_TYPE_TX_L2_CMP) {
 			u32 opaque = txcmp->tx_cmp_opaque;
+			struct bnxt_tx_ring_info *txr;
 			u16 tx_freed;
 
+			txr = bnapi->tx_ring[TX_OPAQUE_RING(opaque)];
 			event |= BNXT_TX_CMP_EVENT;
 			txr->tx_hw_cons = TX_OPAQUE_PROD(bp, opaque);
 			tx_freed = (txr->tx_hw_cons - txr->tx_cons) &
@@ -2671,7 +2675,7 @@ static int __bnxt_poll_work(struct bnxt *bp, struct bnxt_cp_ring_info *cpr,
 		xdp_do_flush();
 
 	if (event & BNXT_TX_EVENT) {
-		struct bnxt_tx_ring_info *txr = bnapi->tx_ring;
+		struct bnxt_tx_ring_info *txr = bnapi->tx_ring[0];
 		u16 prod = txr->tx_prod;
 
 		/* Sync BD data before updating doorbell */
@@ -3657,7 +3661,7 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 
 static void bnxt_init_ring_struct(struct bnxt *bp)
 {
-	int i;
+	int i, j;
 
 	for (i = 0; i < bp->cp_nr_rings; i++) {
 		struct bnxt_napi *bnapi = bp->bnapi[i];
@@ -3702,18 +3706,16 @@ static void bnxt_init_ring_struct(struct bnxt *bp)
 		rmem->vmem = (void **)&rxr->rx_agg_ring;
 
 skip_rx:
-		txr = bnapi->tx_ring;
-		if (!txr)
-			continue;
-
-		ring = &txr->tx_ring_struct;
-		rmem = &ring->ring_mem;
-		rmem->nr_pages = bp->tx_nr_pages;
-		rmem->page_size = HW_RXBD_RING_SIZE;
-		rmem->pg_arr = (void **)txr->tx_desc_ring;
-		rmem->dma_arr = txr->tx_desc_mapping;
-		rmem->vmem_size = SW_TXBD_RING_SIZE * bp->tx_nr_pages;
-		rmem->vmem = (void **)&txr->tx_buf_ring;
+		bnxt_for_each_napi_tx(j, bnapi, txr) {
+			ring = &txr->tx_ring_struct;
+			rmem = &ring->ring_mem;
+			rmem->nr_pages = bp->tx_nr_pages;
+			rmem->page_size = HW_TXBD_RING_SIZE;
+			rmem->pg_arr = (void **)txr->tx_desc_ring;
+			rmem->dma_arr = txr->tx_desc_mapping;
+			rmem->vmem_size = SW_TXBD_RING_SIZE * bp->tx_nr_pages;
+			rmem->vmem = (void **)&txr->tx_buf_ring;
+		}
 	}
 }
 
@@ -4512,7 +4514,7 @@ static int bnxt_alloc_stats(struct bnxt *bp)
 
 static void bnxt_clear_ring_indices(struct bnxt *bp)
 {
-	int i;
+	int i, j;
 
 	if (!bp->bnapi)
 		return;
@@ -4529,8 +4531,7 @@ static void bnxt_clear_ring_indices(struct bnxt *bp)
 		cpr = &bnapi->cp_ring;
 		cpr->cp_raw_cons = 0;
 
-		txr = bnapi->tx_ring;
-		if (txr) {
+		bnxt_for_each_napi_tx(j, bnapi, txr) {
 			txr->tx_prod = 0;
 			txr->tx_cons = 0;
 			txr->tx_hw_cons = 0;
@@ -4703,7 +4704,7 @@ static int bnxt_alloc_mem(struct bnxt *bp, bool irq_re_init)
 			else
 				txr->tx_cpr =  &bp->bnapi[i]->cp_ring;
 			txr->bnapi = bp->bnapi[j];
-			bp->bnapi[j]->tx_ring = txr;
+			bp->bnapi[j]->tx_ring[0] = txr;
 			bp->tx_ring_map[i] = bp->tx_nr_rings_xdp + i;
 			if (i >= bp->tx_nr_rings_xdp) {
 				txr->txq_index = i - bp->tx_nr_rings_xdp;
@@ -6910,10 +6911,21 @@ static int
 bnxt_hwrm_set_tx_coal(struct bnxt *bp, struct bnxt_napi *bnapi,
 		      struct hwrm_ring_cmpl_ring_cfg_aggint_params_input *req)
 {
-	u16 ring_id = bnxt_cp_ring_for_tx(bp, bnapi->tx_ring);
+	struct bnxt_tx_ring_info *txr;
+	int i, rc;
 
-	req->ring_id = cpu_to_le16(ring_id);
-	return hwrm_req_send(bp, req);
+	bnxt_for_each_napi_tx(i, bnapi, txr) {
+		u16 ring_id;
+
+		ring_id = bnxt_cp_ring_for_tx(bp, txr);
+		req->ring_id = cpu_to_le16(ring_id);
+		rc = hwrm_req_send(bp, req);
+		if (rc)
+			return rc;
+		if (!(bp->flags & BNXT_FLAG_CHIP_P5))
+			return 0;
+	}
+	return 0;
 }
 
 int bnxt_hwrm_set_coal(struct bnxt *bp)
@@ -6950,7 +6962,7 @@ int bnxt_hwrm_set_coal(struct bnxt *bp)
 		if (!(bp->flags & BNXT_FLAG_CHIP_P5))
 			continue;
 
-		if (bnapi->rx_ring && bnapi->tx_ring) {
+		if (bnapi->rx_ring && bnapi->tx_ring[0]) {
 			rc = bnxt_hwrm_set_tx_coal(bp, bnapi, req_tx);
 			if (rc)
 				break;
@@ -11575,15 +11587,13 @@ static int bnxt_dbg_hwrm_ring_info_get(struct bnxt *bp, u8 ring_type,
 
 static void bnxt_dump_tx_sw_state(struct bnxt_napi *bnapi)
 {
-	struct bnxt_tx_ring_info *txr = bnapi->tx_ring;
-	int i = bnapi->index;
-
-	if (!txr)
-		return;
+	struct bnxt_tx_ring_info *txr;
+	int i = bnapi->index, j;
 
-	netdev_info(bnapi->bp->dev, "[%d]: tx{fw_ring: %d prod: %x cons: %x}\n",
-		    i, txr->tx_ring_struct.fw_ring_id, txr->tx_prod,
-		    txr->tx_cons);
+	bnxt_for_each_napi_tx(j, bnapi, txr)
+		netdev_info(bnapi->bp->dev, "[%d.%d]: tx{fw_ring: %d prod: %x cons: %x}\n",
+			    i, j, txr->tx_ring_struct.fw_ring_id, txr->tx_prod,
+			    txr->tx_cons);
 }
 
 static void bnxt_dump_rx_sw_state(struct bnxt_napi *bnapi)
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 430538844178..2028233c0561 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -1046,6 +1046,14 @@ struct bnxt_cp_ring_info {
 	struct bnxt_cp_ring_info *cp_ring_arr;
 };
 
+#define BNXT_MAX_QUEUE		8
+#define BNXT_MAX_TXR_PER_NAPI	BNXT_MAX_QUEUE
+
+#define bnxt_for_each_napi_tx(iter, bnapi, txr)		\
+	for (iter = 0, txr = (bnapi)->tx_ring[0]; txr;	\
+	     txr = (iter < BNXT_MAX_TXR_PER_NAPI - 1) ?	\
+	     (bnapi)->tx_ring[++iter] : NULL)
+
 struct bnxt_napi {
 	struct napi_struct	napi;
 	struct bnxt		*bp;
@@ -1053,7 +1061,7 @@ struct bnxt_napi {
 	int			index;
 	struct bnxt_cp_ring_info	cp_ring;
 	struct bnxt_rx_ring_info	*rx_ring;
-	struct bnxt_tx_ring_info	*tx_ring;
+	struct bnxt_tx_ring_info	*tx_ring[BNXT_MAX_TXR_PER_NAPI];
 
 	void			(*tx_int)(struct bnxt *, struct bnxt_napi *,
 					  int budget);
@@ -1391,8 +1399,6 @@ struct bnxt_link_info {
 	(PORT_PHY_CFG_REQ_FLAGS_FEC_CLAUSE74_DISABLE |		\
 	 BNXT_FEC_RS_OFF(link_info))
 
-#define BNXT_MAX_QUEUE	8
-
 struct bnxt_queue_info {
 	u8	queue_id;
 	u8	queue_profile;
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
index 3515a12a6fea..52b75108e130 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
@@ -127,7 +127,7 @@ static void __bnxt_xmit_xdp_redirect(struct bnxt *bp,
 
 void bnxt_tx_int_xdp(struct bnxt *bp, struct bnxt_napi *bnapi, int budget)
 {
-	struct bnxt_tx_ring_info *txr = bnapi->tx_ring;
+	struct bnxt_tx_ring_info *txr = bnapi->tx_ring[0];
 	struct bnxt_rx_ring_info *rxr = bnapi->rx_ring;
 	u16 tx_hw_cons = txr->tx_hw_cons;
 	bool rx_doorbell_needed = false;
@@ -249,7 +249,7 @@ bool bnxt_rx_xdp(struct bnxt *bp, struct bnxt_rx_ring_info *rxr, u16 cons,
 	pdev = bp->pdev;
 	offset = bp->rx_offset;
 
-	txr = rxr->bnapi->tx_ring;
+	txr = rxr->bnapi->tx_ring[0];
 	/* BNXT_RX_PAGE_MODE(bp) when XDP enabled */
 	orig_data = xdp.data;
 
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 10/13] bnxt_en: Add helper to get the number of CP rings required for TX rings

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 9872 bytes --]

Up until now, each TX ring always requires a completion ring/NQ/MSIX.
bnxt_trim_rings() and the assignment of bp->cp_nr_rings always make
this assumption.  This will no longer be true in the next patches, so
we refactor and add helper functions to determine the proper relationship
between TX rings and the required completion ring/NQ/MSIX.  This patch
does not change the 1:1 relationship yet.

Note that on P5 chips, each RX and TX ring still requires a completion
ring.  Only the number of NQs has been reduced.  We should no longer call
bnxt_trim_rings() to adjust the RX and TX rings on P5 chips.  Replace with
simple logic to check that RX + TX < CP and adjust accordingly.

bnxt_check_rings() should call _bnxt_get_max_rings() to get the raw
number of rings instead of bnxt_get_max_rings().  If we are about to
create TCs, bnxt_get_max_rings() would not be able to calculate the max
rings correctly.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 95 +++++++++++++++----
 drivers/net/ethernet/broadcom/bnxt/bnxt.h     |  1 +
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |  6 +-
 drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c |  5 +-
 4 files changed, 82 insertions(+), 25 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 6002b834e898..7c1a3db651f5 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -6284,7 +6284,8 @@ static int bnxt_hwrm_get_rings(struct bnxt *bp)
 			if (bp->flags & BNXT_FLAG_AGG_RINGS)
 				rx >>= 1;
 			if (cp < (rx + tx)) {
-				bnxt_trim_rings(bp, &rx, &tx, cp, false);
+				rx = cp / 2;
+				tx = rx;
 				if (bp->flags & BNXT_FLAG_AGG_RINGS)
 					rx <<= 1;
 				hw_resc->resv_rx_rings = rx;
@@ -6585,6 +6586,7 @@ static int __bnxt_reserve_rings(struct bnxt *bp)
 	int grp, rx_rings, rc;
 	int vnic = 1, stat;
 	bool sh = false;
+	int tx_cp;
 
 	if (!bnxt_need_reserve_rings(bp))
 		return 0;
@@ -6634,7 +6636,8 @@ static int __bnxt_reserve_rings(struct bnxt *bp)
 	rc = bnxt_trim_rings(bp, &rx_rings, &tx, cp, sh);
 	if (bp->flags & BNXT_FLAG_AGG_RINGS)
 		rx = rx_rings << 1;
-	cp = sh ? max_t(int, tx, rx_rings) : tx + rx_rings;
+	tx_cp = bnxt_num_tx_to_cp(bp, tx);
+	cp = sh ? max_t(int, tx_cp, rx_rings) : tx_cp + rx_rings;
 	bp->tx_nr_rings = tx;
 
 	/* If we cannot reserve all the RX rings, reset the RSS map only
@@ -9061,8 +9064,8 @@ static int bnxt_set_real_num_queues(struct bnxt *bp)
 	return rc;
 }
 
-static int bnxt_trim_rings(struct bnxt *bp, int *rx, int *tx, int max,
-			   bool shared)
+static int __bnxt_trim_rings(struct bnxt *bp, int *rx, int *tx, int max,
+			     bool shared)
 {
 	int _rx = *rx, _tx = *tx;
 
@@ -9085,6 +9088,46 @@ static int bnxt_trim_rings(struct bnxt *bp, int *rx, int *tx, int max,
 	return 0;
 }
 
+static int __bnxt_num_tx_to_cp(struct bnxt *bp, int tx, int tx_sets, int tx_xdp)
+{
+	return tx;
+}
+
+int bnxt_num_tx_to_cp(struct bnxt *bp, int tx)
+{
+	int tcs = netdev_get_num_tc(bp->dev);
+
+	if (!tcs)
+		tcs = 1;
+	return __bnxt_num_tx_to_cp(bp, tx, tcs, bp->tx_nr_rings_xdp);
+}
+
+static int bnxt_num_cp_to_tx(struct bnxt *bp, int tx_cp)
+{
+	int tcs = netdev_get_num_tc(bp->dev);
+
+	return (tx_cp - bp->tx_nr_rings_xdp) * tcs +
+	       bp->tx_nr_rings_xdp;
+}
+
+static int bnxt_trim_rings(struct bnxt *bp, int *rx, int *tx, int max,
+			   bool sh)
+{
+	int tx_cp = bnxt_num_tx_to_cp(bp, *tx);
+
+	if (tx_cp != *tx) {
+		int tx_saved = tx_cp, rc;
+
+		rc = __bnxt_trim_rings(bp, rx, &tx_cp, max, sh);
+		if (rc)
+			return rc;
+		if (tx_cp != tx_saved)
+			*tx = bnxt_num_cp_to_tx(bp, tx_cp);
+		return 0;
+	}
+	return __bnxt_trim_rings(bp, rx, tx, max, sh);
+}
+
 static void bnxt_setup_msix(struct bnxt *bp)
 {
 	const int len = sizeof(bp->irq_tbl[0].name);
@@ -9247,7 +9290,7 @@ static int bnxt_get_num_msix(struct bnxt *bp)
 
 static int bnxt_init_msix(struct bnxt *bp)
 {
-	int i, total_vecs, max, rc = 0, min = 1, ulp_msix;
+	int i, total_vecs, max, rc = 0, min = 1, ulp_msix, tx_cp;
 	struct msix_entry *msix_ent;
 
 	total_vecs = bnxt_get_num_msix(bp);
@@ -9289,9 +9332,10 @@ static int bnxt_init_msix(struct bnxt *bp)
 		if (rc)
 			goto msix_setup_exit;
 
+		tx_cp = bnxt_num_tx_to_cp(bp, bp->tx_nr_rings);
 		bp->cp_nr_rings = (min == 1) ?
-				  max_t(int, bp->tx_nr_rings, bp->rx_nr_rings) :
-				  bp->tx_nr_rings + bp->rx_nr_rings;
+				  max_t(int, tx_cp, bp->rx_nr_rings) :
+				  tx_cp + bp->rx_nr_rings;
 
 	} else {
 		rc = -ENOMEM;
@@ -12186,23 +12230,27 @@ static void bnxt_sp_task(struct work_struct *work)
 	clear_bit(BNXT_STATE_IN_SP_TASK, &bp->state);
 }
 
+static void _bnxt_get_max_rings(struct bnxt *bp, int *max_rx, int *max_tx,
+				int *max_cp);
+
 /* Under rtnl_lock */
 int bnxt_check_rings(struct bnxt *bp, int tx, int rx, bool sh, int tcs,
 		     int tx_xdp)
 {
-	int max_rx, max_tx, tx_sets = 1;
+	int max_rx, max_tx, max_cp, tx_sets = 1, tx_cp;
 	int tx_rings_needed, stats;
 	int rx_rings = rx;
-	int cp, vnics, rc;
+	int cp, vnics;
 
 	if (tcs)
 		tx_sets = tcs;
 
-	rc = bnxt_get_max_rings(bp, &max_rx, &max_tx, sh);
-	if (rc)
-		return rc;
+	if (bp->flags & BNXT_FLAG_AGG_RINGS)
+		rx_rings <<= 1;
 
-	if (max_rx < rx)
+	_bnxt_get_max_rings(bp, &max_rx, &max_tx, &max_cp);
+
+	if (max_rx < rx_rings)
 		return -ENOMEM;
 
 	tx_rings_needed = tx * tx_sets + tx_xdp;
@@ -12211,11 +12259,12 @@ int bnxt_check_rings(struct bnxt *bp, int tx, int rx, bool sh, int tcs,
 
 	vnics = 1;
 	if ((bp->flags & (BNXT_FLAG_RFS | BNXT_FLAG_CHIP_P5)) == BNXT_FLAG_RFS)
-		vnics += rx_rings;
+		vnics += rx;
 
-	if (bp->flags & BNXT_FLAG_AGG_RINGS)
-		rx_rings <<= 1;
-	cp = sh ? max_t(int, tx_rings_needed, rx) : tx_rings_needed + rx;
+	tx_cp = __bnxt_num_tx_to_cp(bp, tx_rings_needed, tx_sets, tx_xdp);
+	cp = sh ? max_t(int, tx_cp, rx) : tx_cp + rx;
+	if (max_cp < cp)
+		return -ENOMEM;
 	stats = cp;
 	if (BNXT_NEW_RM(bp)) {
 		cp += bnxt_get_ulp_msix_num(bp);
@@ -12849,7 +12898,7 @@ int bnxt_setup_mq_tc(struct net_device *dev, u8 tc)
 {
 	struct bnxt *bp = netdev_priv(dev);
 	bool sh = false;
-	int rc;
+	int rc, tx_cp;
 
 	if (tc > bp->max_tc) {
 		netdev_err(dev, "Too many traffic classes requested: %d. Max supported is %d.\n",
@@ -12880,8 +12929,9 @@ int bnxt_setup_mq_tc(struct net_device *dev, u8 tc)
 		netdev_reset_tc(dev);
 	}
 	bp->tx_nr_rings += bp->tx_nr_rings_xdp;
-	bp->cp_nr_rings = sh ? max_t(int, bp->tx_nr_rings, bp->rx_nr_rings) :
-			       bp->tx_nr_rings + bp->rx_nr_rings;
+	tx_cp = bnxt_num_tx_to_cp(bp, bp->tx_nr_rings);
+	bp->cp_nr_rings = sh ? max_t(int, tx_cp, bp->rx_nr_rings) :
+			       tx_cp + bp->rx_nr_rings;
 
 	if (netif_running(bp->dev))
 		return bnxt_open_nic(bp, true, false);
@@ -13360,7 +13410,10 @@ static void _bnxt_get_max_rings(struct bnxt *bp, int *max_rx, int *max_tx,
 	if (bp->flags & BNXT_FLAG_AGG_RINGS)
 		*max_rx >>= 1;
 	if (bp->flags & BNXT_FLAG_CHIP_P5) {
-		bnxt_trim_rings(bp, max_rx, max_tx, *max_cp, false);
+		if (*max_cp < (*max_rx + *max_tx)) {
+			*max_rx = *max_cp / 2;
+			*max_tx = *max_rx;
+		}
 		/* On P5 chips, max_cp output param should be available NQs */
 		*max_cp = max_irq;
 	}
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.h b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
index 2028233c0561..4ce993943924 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.h
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.h
@@ -2393,6 +2393,7 @@ int __bnxt_hwrm_get_tx_rings(struct bnxt *bp, u16 fid, int *tx_rings);
 int bnxt_nq_rings_in_use(struct bnxt *bp);
 int bnxt_hwrm_set_coal(struct bnxt *);
 void bnxt_free_ctx_mem(struct bnxt *bp);
+int bnxt_num_tx_to_cp(struct bnxt *bp, int tx);
 unsigned int bnxt_get_max_func_stat_ctxs(struct bnxt *bp);
 unsigned int bnxt_get_avail_stat_ctxs_for_en(struct bnxt *bp);
 unsigned int bnxt_get_max_func_cp_rings(struct bnxt *bp);
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index 18c06158fead..76f2eab52ce7 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -923,6 +923,7 @@ static int bnxt_set_channels(struct net_device *dev,
 	bool sh = false;
 	int tx_xdp = 0;
 	int rc = 0;
+	int tx_cp;
 
 	if (channel->other_count)
 		return -EINVAL;
@@ -994,8 +995,9 @@ static int bnxt_set_channels(struct net_device *dev,
 	if (tcs > 1)
 		bp->tx_nr_rings = bp->tx_nr_rings_per_tc * tcs + tx_xdp;
 
-	bp->cp_nr_rings = sh ? max_t(int, bp->tx_nr_rings, bp->rx_nr_rings) :
-			       bp->tx_nr_rings + bp->rx_nr_rings;
+	tx_cp = bnxt_num_tx_to_cp(bp, bp->tx_nr_rings);
+	bp->cp_nr_rings = sh ? max_t(int, tx_cp, bp->rx_nr_rings) :
+			       tx_cp + bp->rx_nr_rings;
 
 	/* After changing number of rx channels, update NTUPLE feature. */
 	netdev_update_features(dev);
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
index 52b75108e130..9d428eb3fdb9 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c
@@ -398,7 +398,7 @@ int bnxt_xdp_xmit(struct net_device *dev, int num_frames,
 static int bnxt_xdp_set(struct bnxt *bp, struct bpf_prog *prog)
 {
 	struct net_device *dev = bp->dev;
-	int tx_xdp = 0, rc, tc;
+	int tx_xdp = 0, tx_cp, rc, tc;
 	struct bpf_prog *old;
 
 	if (prog && !prog->aux->xdp_has_frags &&
@@ -446,7 +446,8 @@ static int bnxt_xdp_set(struct bnxt *bp, struct bpf_prog *prog)
 	}
 	bp->tx_nr_rings_xdp = tx_xdp;
 	bp->tx_nr_rings = bp->tx_nr_rings_per_tc * tc + tx_xdp;
-	bp->cp_nr_rings = max_t(int, bp->tx_nr_rings, bp->rx_nr_rings);
+	tx_cp = bnxt_num_tx_to_cp(bp, bp->tx_nr_rings);
+	bp->cp_nr_rings = max_t(int, tx_cp, bp->rx_nr_rings);
 	bnxt_set_tpa_flags(bp);
 	bnxt_set_ring_params(bp);
 
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 11/13] bnxt_en: Add macros related to TC and TX rings

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 1587 bytes --]

Add 3 macros that handle to conversions between TC numbers and TX
ring numbers.  These will help to clarify the existing logic and the
new logic in the next patch.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index 7c1a3db651f5..d0eca7648927 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -3415,6 +3415,15 @@ static void bnxt_free_tx_rings(struct bnxt *bp)
 	}
 }
 
+#define BNXT_TC_TO_RING_BASE(bp, tc)	\
+	((tc) * (bp)->tx_nr_rings_per_tc)
+
+#define BNXT_RING_TO_TC_OFF(bp, tx)	\
+	((tx) % (bp)->tx_nr_rings_per_tc)
+
+#define BNXT_RING_TO_TC(bp, tx)		\
+	((tx) / (bp)->tx_nr_rings_per_tc)
+
 static int bnxt_alloc_tx_rings(struct bnxt *bp)
 {
 	int i, j, rc;
@@ -3470,7 +3479,7 @@ static int bnxt_alloc_tx_rings(struct bnxt *bp)
 		spin_lock_init(&txr->xdp_tx_lock);
 		if (i < bp->tx_nr_rings_xdp)
 			continue;
-		if (i % bp->tx_nr_rings_per_tc == (bp->tx_nr_rings_per_tc - 1))
+		if (BNXT_RING_TO_TC_OFF(bp, i) == (bp->tx_nr_rings_per_tc - 1))
 			j++;
 	}
 	return 0;
@@ -9140,7 +9149,7 @@ static void bnxt_setup_msix(struct bnxt *bp)
 
 		for (i = 0; i < tcs; i++) {
 			count = bp->tx_nr_rings_per_tc;
-			off = i * count;
+			off = BNXT_TC_TO_RING_BASE(bp, i);
 			netdev_set_tc_queue(dev, i, count, off);
 		}
 	}
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 12/13] bnxt_en: Use existing MSIX vectors for all mqprio TX rings

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 5321 bytes --]

We can now fully support sharing the same MSIX for all mqprio TX rings
belonging to the same ethtool channel with the new infrastructure:

1. Allocate the proper entries for cp_ring_arr in struct bnxt_cp_ring_info
to support the additional TX rings.

2. Populate the tx_ring array in struct bnxt_napi for all TX rings
sharing the same NAPI.

3. bnxt_num_tx_to_cp() returns the proper NQ/completion rings to support
the TX rings in the input.

4. Adjust bnxt_get_num_ring_stats() for the reduced number of ring
counters with the new scheme.

Reviewed-by: Andy Gospodarek <andrew.gospodarek@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c     | 56 ++++++++++++++-----
 .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c |  3 +-
 2 files changed, 43 insertions(+), 16 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index d0eca7648927..d1af1d2ff800 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -3609,7 +3609,10 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 {
 	bool sh = !!(bp->flags & BNXT_FLAG_SHARED_RINGS);
 	int i, j, rc, ulp_base_vec, ulp_msix;
+	int tcs = netdev_get_num_tc(bp->dev);
 
+	if (!tcs)
+		tcs = 1;
 	ulp_msix = bnxt_get_ulp_msix_num(bp);
 	ulp_base_vec = bnxt_get_ulp_msix_base(bp);
 	for (i = 0, j = 0; i < bp->cp_nr_rings; i++) {
@@ -3617,6 +3620,7 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 		struct bnxt_cp_ring_info *cpr, *cpr2;
 		struct bnxt_ring_struct *ring;
 		int cp_count = 0, k;
+		int rx = 0, tx = 0;
 
 		if (!bnapi)
 			continue;
@@ -3637,11 +3641,18 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 		if (!(bp->flags & BNXT_FLAG_CHIP_P5))
 			continue;
 
-		if (i < bp->rx_nr_rings)
+		if (i < bp->rx_nr_rings) {
 			cp_count++;
-		if ((sh && i < bp->tx_nr_rings) ||
-		    (!sh && i >= bp->rx_nr_rings))
+			rx = 1;
+		}
+		if (i < bp->tx_nr_rings_xdp) {
 			cp_count++;
+			tx = 1;
+		} else if ((sh && i < bp->tx_nr_rings) ||
+			 (!sh && i >= bp->rx_nr_rings)) {
+			cp_count += tcs;
+			tx = 1;
+		}
 
 		cpr->cp_ring_arr = kcalloc(cp_count, sizeof(*cpr),
 					   GFP_KERNEL);
@@ -3656,14 +3667,19 @@ static int bnxt_alloc_cp_rings(struct bnxt *bp)
 				return rc;
 			cpr2->bnapi = bnapi;
 			cpr2->cp_idx = k;
-			if (!k && i < bp->rx_nr_rings) {
+			if (!k && rx) {
 				bp->rx_ring[i].rx_cpr = cpr2;
 				cpr2->cp_ring_type = BNXT_NQ_HDL_TYPE_RX;
 			} else {
-				bp->tx_ring[j++].tx_cpr = cpr2;
+				int n, tc = k - rx;
+
+				n = BNXT_TC_TO_RING_BASE(bp, tc) + j;
+				bp->tx_ring[n].tx_cpr = cpr2;
 				cpr2->cp_ring_type = BNXT_NQ_HDL_TYPE_TX;
 			}
 		}
+		if (tx)
+			j++;
 	}
 	return 0;
 }
@@ -4704,24 +4720,33 @@ static int bnxt_alloc_mem(struct bnxt *bp, bool irq_re_init)
 		else
 			j = bp->rx_nr_rings;
 
-		for (i = 0; i < bp->tx_nr_rings; i++, j++) {
+		for (i = 0; i < bp->tx_nr_rings; i++) {
 			struct bnxt_tx_ring_info *txr = &bp->tx_ring[i];
+			struct bnxt_napi *bnapi2;
 
 			if (bp->flags & BNXT_FLAG_CHIP_P5)
 				txr->tx_ring_struct.ring_mem.flags =
 					BNXT_RMEM_RING_PTE_FLAG;
-			else
-				txr->tx_cpr =  &bp->bnapi[i]->cp_ring;
-			txr->bnapi = bp->bnapi[j];
-			bp->bnapi[j]->tx_ring[0] = txr;
 			bp->tx_ring_map[i] = bp->tx_nr_rings_xdp + i;
 			if (i >= bp->tx_nr_rings_xdp) {
+				int k = j + BNXT_RING_TO_TC_OFF(bp, i);
+
+				bnapi2 = bp->bnapi[k];
 				txr->txq_index = i - bp->tx_nr_rings_xdp;
-				bp->bnapi[j]->tx_int = bnxt_tx_int;
+				txr->tx_napi_idx =
+					BNXT_RING_TO_TC(bp, txr->txq_index);
+				bnapi2->tx_ring[txr->tx_napi_idx] = txr;
+				bnapi2->tx_int = bnxt_tx_int;
 			} else {
-				bp->bnapi[j]->flags |= BNXT_NAPI_FLAG_XDP;
-				bp->bnapi[j]->tx_int = bnxt_tx_int_xdp;
+				bnapi2 = bp->bnapi[j];
+				bnapi2->flags |= BNXT_NAPI_FLAG_XDP;
+				bnapi2->tx_ring[0] = txr;
+				bnapi2->tx_int = bnxt_tx_int_xdp;
+				j++;
 			}
+			txr->bnapi = bnapi2;
+			if (!(bp->flags & BNXT_FLAG_CHIP_P5))
+				txr->tx_cpr = &bnapi2->cp_ring;
 		}
 
 		rc = bnxt_alloc_stats(bp);
@@ -9099,7 +9124,7 @@ static int __bnxt_trim_rings(struct bnxt *bp, int *rx, int *tx, int max,
 
 static int __bnxt_num_tx_to_cp(struct bnxt *bp, int tx, int tx_sets, int tx_xdp)
 {
-	return tx;
+	return (tx - tx_xdp) / tx_sets + tx_xdp;
 }
 
 int bnxt_num_tx_to_cp(struct bnxt *bp, int tx)
@@ -13723,7 +13748,8 @@ static int bnxt_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
 	}
 
 	max_irqs = bnxt_get_max_irq(pdev);
-	dev = alloc_etherdev_mq(sizeof(*bp), max_irqs);
+	dev = alloc_etherdev_mqs(sizeof(*bp), max_irqs * BNXT_MAX_QUEUE,
+				 max_irqs);
 	if (!dev)
 		return -ENOMEM;
 
diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
index 76f2eab52ce7..585044310141 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt_ethtool.c
@@ -528,7 +528,8 @@ static int bnxt_get_num_ring_stats(struct bnxt *bp)
 	     bnxt_get_num_tpa_ring_stats(bp);
 	tx = NUM_RING_TX_HW_STATS;
 	cmn = NUM_RING_CMN_SW_STATS;
-	return rx * bp->rx_nr_rings + tx * bp->tx_nr_rings +
+	return rx * bp->rx_nr_rings +
+	       tx * (bp->tx_nr_rings_xdp + bp->tx_nr_rings_per_tc) +
 	       cmn * bp->cp_nr_rings;
 }
 
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

[PATCH net-next 13/13] bnxt_en: Optimize xmit_more TX path

[Michael Chan]

[-- Attachment #1: Type: text/plain, Size: 2546 bytes --]

Now that we use the cumulative consumer index scheme for TX completion,
we don't need to have one TX completion per TX packet in the xmit_more
code path.  Set the TX_BD_FLAGS_NO_CMPL flag if xmit_more is true.
Fallback to one interrupt per packet if the ring is filled beyond
bp->tx_wake_thresh.

Also, move the wmb() to bnxt_txr_db_kick().  When xmit_more is true,
we'll skip the bnxt_txr_db_kick() call and there is no need to call
wmb() to sync. the TX BD data.

Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index d1af1d2ff800..e6ac1bd21bb3 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -381,6 +381,8 @@ static u16 bnxt_xmit_get_cfa_action(struct sk_buff *skb)
 static void bnxt_txr_db_kick(struct bnxt *bp, struct bnxt_tx_ring_info *txr,
 			     u16 prod)
 {
+	/* Sync BD data before updating doorbell */
+	wmb();
 	bnxt_db_write(bp, &txr->tx_db, prod);
 	txr->kick_pending = 0;
 }
@@ -388,7 +390,7 @@ static void bnxt_txr_db_kick(struct bnxt *bp, struct bnxt_tx_ring_info *txr,
 static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 {
 	struct bnxt *bp = netdev_priv(dev);
-	struct tx_bd *txbd;
+	struct tx_bd *txbd, *txbd0;
 	struct tx_bd_ext *txbd1;
 	struct netdev_queue *txq;
 	int i;
@@ -602,6 +604,7 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	txbd1->tx_bd_cfa_meta = cpu_to_le32(vlan_tag_flags);
 	txbd1->tx_bd_cfa_action =
 			cpu_to_le32(cfa_action << TX_BD_CFA_ACTION_SHIFT);
+	txbd0 = txbd;
 	for (i = 0; i < last_frag; i++) {
 		skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
 
@@ -633,16 +636,17 @@ static netdev_tx_t bnxt_start_xmit(struct sk_buff *skb, struct net_device *dev)
 
 	skb_tx_timestamp(skb);
 
-	/* Sync BD data before updating doorbell */
-	wmb();
-
 	prod = NEXT_TX(prod);
 	WRITE_ONCE(txr->tx_prod, prod);
 
-	if (!netdev_xmit_more() || netif_xmit_stopped(txq))
+	if (!netdev_xmit_more() || netif_xmit_stopped(txq)) {
 		bnxt_txr_db_kick(bp, txr, prod);
-	else
+	} else {
+		if (free_size >= bp->tx_wake_thresh)
+			txbd0->tx_bd_len_flags_type |=
+				cpu_to_le32(TX_BD_FLAGS_NO_CMPL);
 		txr->kick_pending = 1;
+	}
 
 tx_done:
 
-- 
2.30.1


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4209 bytes --]

Re: [PATCH net-next 00/13] bnxt_en: TX path improvements

[Jakub Kicinski]

On Fri, 27 Oct 2023 16:22:39 -0700 Michael Chan wrote:
> All patches in this patchset are related to improving the TX path.
> There are 2 areas of improvements:
> 
> 1. The TX interrupt logic currently counts the number of TX completions
> to determine the number of TX SKBs to free.  We now change it so that
> the TX completion will now contain the hardware consumer index
> information.  The driver will keep track of the latest hardware
> consumer index from the last TX completion and clean up all TX SKBs
> up to that index.  This scheme aligns better with future chips and
> allows xmit_more code path to be more optimized.
> 
> 2. The current driver logic requires an additional MSIX for each
> additional MQPRIO TX ring.  This scheme uses too many MSIX vectors if
> the user enables a large number of MQPRIO TCs.  We now use a new scheme
> that will use the same MSIX for all the MQPRIO TX rings for each
> ethtool channel.  Each ethtool TX channel can have up to 8 MQPRIO
> TX rings and now they all will share the same MSIX.

## Form letter - net-next-closed

The merge window for v6.7 has begun and we have already posted our pull
request. Therefore net-next is closed for new drivers, features, code
refactoring and optimizations. We are currently accepting bug fixes only.

Please repost when net-next reopens after Nov 12th.

RFC patches sent for review only are obviously welcome at any time.

See: https://www.kernel.org/doc/html/next/process/maintainer-netdev.html#development-cycle
-- 
pw-bot: defer