[PATCH net-next V1 0/2] net: lan743x: Fix to use multiqueue start/stop APIs

[PATCH net-next V1 0/2] net: lan743x: Fix to use multiqueue start/stop APIs

[Raju Lakkaraju]

This patch series address the fix to use multiqueue start/stop APIs and Add Rx IP and TCP checksum offload

Changes:
========
V0 -> V1:
 - Remove chip SKU check conditionals
 - Update the changes description

Raju Lakkaraju (2):
  net: lan743x: Fix to use multiqueue start/stop APIs
  net: lan743x: Add support for Rx IP & TCP checksum offload

 drivers/net/ethernet/microchip/lan743x_main.c | 64 +++++++++++--------
 drivers/net/ethernet/microchip/lan743x_main.h | 10 ++-
 2 files changed, 45 insertions(+), 29 deletions(-)

-- 
2.25.1

[PATCH net-next V1 1/2] net: lan743x: Fix to use multiqueue start/stop APIs

[Raju Lakkaraju]

- Fix to use multiqueue start/stop APIs
- Change to return NETDEV_TX_BUSY instead of holding the TX skb when busy
- Increase Tx ring size to 128 to address performance issues in some platforms
- Use NAPI_POLL_WEIGHT for Tx Napi handler instead of ring dependent value
- Use multiqueue to register 4 Rx channels

Signed-off-by: Raju Lakkaraju <Raju.Lakkaraju@microchip.com>
---
Changes:
V0 -> V1:
 - Remove chip SKU check conditionals
 - Update the changes description 

 drivers/net/ethernet/microchip/lan743x_main.c | 50 +++++++++----------
 drivers/net/ethernet/microchip/lan743x_main.h |  5 +-
 2 files changed, 27 insertions(+), 28 deletions(-)

diff --git a/drivers/net/ethernet/microchip/lan743x_main.c b/drivers/net/ethernet/microchip/lan743x_main.c
index a9a1dea6d731..3f4e1ab63f8a 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.c
+++ b/drivers/net/ethernet/microchip/lan743x_main.c
@@ -2066,11 +2066,13 @@ static netdev_tx_t lan743x_tx_xmit_frame(struct lan743x_tx *tx,
 {
 	int required_number_of_descriptors = 0;
 	unsigned int start_frame_length = 0;
+	netdev_tx_t retval = NETDEV_TX_OK;
 	unsigned int frame_length = 0;
 	unsigned int head_length = 0;
 	unsigned long irq_flags = 0;
 	bool do_timestamp = false;
 	bool ignore_sync = false;
+	struct netdev_queue *txq;
 	int nr_frags = 0;
 	bool gso = false;
 	int j;
@@ -2083,9 +2085,12 @@ static netdev_tx_t lan743x_tx_xmit_frame(struct lan743x_tx *tx,
 		if (required_number_of_descriptors > (tx->ring_size - 1)) {
 			dev_kfree_skb_irq(skb);
 		} else {
-			/* save to overflow buffer */
-			tx->overflow_skb = skb;
-			netif_stop_queue(tx->adapter->netdev);
+			/* save how many descriptors we needed to restart the queue */
+			tx->rqd_descriptors = required_number_of_descriptors;
+			retval = NETDEV_TX_BUSY;
+			txq = netdev_get_tx_queue(tx->adapter->netdev,
+						  tx->channel_number);
+			netif_tx_stop_queue(txq);
 		}
 		goto unlock;
 	}
@@ -2144,15 +2149,15 @@ static netdev_tx_t lan743x_tx_xmit_frame(struct lan743x_tx *tx,
 
 unlock:
 	spin_unlock_irqrestore(&tx->ring_lock, irq_flags);
-	return NETDEV_TX_OK;
+	return retval;
 }
 
 static int lan743x_tx_napi_poll(struct napi_struct *napi, int weight)
 {
 	struct lan743x_tx *tx = container_of(napi, struct lan743x_tx, napi);
 	struct lan743x_adapter *adapter = tx->adapter;
-	bool start_transmitter = false;
 	unsigned long irq_flags = 0;
+	struct netdev_queue *txq;
 	u32 ioc_bit = 0;
 
 	ioc_bit = DMAC_INT_BIT_TX_IOC_(tx->channel_number);
@@ -2163,24 +2168,20 @@ static int lan743x_tx_napi_poll(struct napi_struct *napi, int weight)
 
 	/* clean up tx ring */
 	lan743x_tx_release_completed_descriptors(tx);
-	if (netif_queue_stopped(adapter->netdev)) {
-		if (tx->overflow_skb) {
-			if (lan743x_tx_get_desc_cnt(tx, tx->overflow_skb) <=
-				lan743x_tx_get_avail_desc(tx))
-				start_transmitter = true;
+	txq = netdev_get_tx_queue(adapter->netdev, tx->channel_number);
+	if (netif_tx_queue_stopped(txq)) {
+		if (tx->rqd_descriptors) {
+			if (tx->rqd_descriptors <=
+			    lan743x_tx_get_avail_desc(tx)) {
+				tx->rqd_descriptors = 0;
+				netif_tx_wake_queue(txq);
+			}
 		} else {
-			netif_wake_queue(adapter->netdev);
+			netif_tx_wake_queue(txq);
 		}
 	}
 	spin_unlock_irqrestore(&tx->ring_lock, irq_flags);
 
-	if (start_transmitter) {
-		/* space is now available, transmit overflow skb */
-		lan743x_tx_xmit_frame(tx, tx->overflow_skb);
-		tx->overflow_skb = NULL;
-		netif_wake_queue(adapter->netdev);
-	}
-
 	if (!napi_complete(napi))
 		goto done;
 
@@ -2304,10 +2305,7 @@ static void lan743x_tx_close(struct lan743x_tx *tx)
 
 	lan743x_tx_release_all_descriptors(tx);
 
-	if (tx->overflow_skb) {
-		dev_kfree_skb(tx->overflow_skb);
-		tx->overflow_skb = NULL;
-	}
+	tx->rqd_descriptors = 0;
 
 	lan743x_tx_ring_cleanup(tx);
 }
@@ -2387,7 +2385,7 @@ static int lan743x_tx_open(struct lan743x_tx *tx)
 							 (tx->channel_number));
 	netif_napi_add_tx_weight(adapter->netdev,
 				 &tx->napi, lan743x_tx_napi_poll,
-				 tx->ring_size - 1);
+				 NAPI_POLL_WEIGHT);
 	napi_enable(&tx->napi);
 
 	data = 0;
@@ -3347,8 +3345,10 @@ static int lan743x_pcidev_probe(struct pci_dev *pdev,
 						 PCI11X1X_USED_TX_CHANNELS,
 						 LAN743X_USED_RX_CHANNELS);
 	} else {
-		netdev = devm_alloc_etherdev(&pdev->dev,
-					     sizeof(struct lan743x_adapter));
+		netdev = devm_alloc_etherdev_mqs(&pdev->dev,
+						 sizeof(struct lan743x_adapter),
+						 LAN743X_USED_TX_CHANNELS,
+						 LAN743X_USED_RX_CHANNELS);
 	}
 
 	if (!netdev)
diff --git a/drivers/net/ethernet/microchip/lan743x_main.h b/drivers/net/ethernet/microchip/lan743x_main.h
index 72adae4f2aa0..58eb7abf976b 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.h
+++ b/drivers/net/ethernet/microchip/lan743x_main.h
@@ -954,8 +954,7 @@ struct lan743x_tx {
 
 	struct napi_struct napi;
 	u32 frame_count;
-
-	struct sk_buff *overflow_skb;
+	u32 rqd_descriptors;
 };
 
 void lan743x_tx_set_timestamping_mode(struct lan743x_tx *tx,
@@ -1110,7 +1109,7 @@ struct lan743x_tx_buffer_info {
 	unsigned int    buffer_length;
 };
 
-#define LAN743X_TX_RING_SIZE    (50)
+#define LAN743X_TX_RING_SIZE    (128)
 
 /* OWN bit is set. ie, Descs are owned by RX DMAC */
 #define RX_DESC_DATA0_OWN_                (0x00008000)
-- 
2.25.1

[PATCH net-next V1 2/2] net: lan743x: Add support for Rx IP & TCP checksum offload

[Raju Lakkaraju]

Add Rx IP and TCP checksum offload

Signed-off-by: Raju Lakkaraju <Raju.Lakkaraju@microchip.com>
---
 drivers/net/ethernet/microchip/lan743x_main.c | 14 +++++++++++++-
 drivers/net/ethernet/microchip/lan743x_main.h |  5 +++++
 2 files changed, 18 insertions(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/microchip/lan743x_main.c b/drivers/net/ethernet/microchip/lan743x_main.c
index 3f4e1ab63f8a..59441f06ff45 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.c
+++ b/drivers/net/ethernet/microchip/lan743x_main.c
@@ -1585,6 +1585,9 @@ static void lan743x_rfe_set_multicast(struct lan743x_adapter *adapter)
 			rfctl |= RFE_CTL_AM_;
 	}
 
+	if (netdev->features & NETIF_F_RXCSUM)
+		rfctl |= RFE_CTL_IP_COE_ | RFE_CTL_TCP_UDP_COE_;
+
 	memset(hash_table, 0, DP_SEL_VHF_HASH_LEN * sizeof(u32));
 	if (netdev_mc_count(netdev)) {
 		struct netdev_hw_addr *ha;
@@ -2547,6 +2550,7 @@ static int lan743x_rx_process_buffer(struct lan743x_rx *rx)
 	int result = RX_PROCESS_RESULT_NOTHING_TO_DO;
 	struct lan743x_rx_buffer_info *buffer_info;
 	int frame_length, buffer_length;
+	bool is_ice, is_tce, is_icsm;
 	int extension_index = -1;
 	bool is_last, is_first;
 	struct sk_buff *skb;
@@ -2593,6 +2597,9 @@ static int lan743x_rx_process_buffer(struct lan743x_rx *rx)
 	frame_length =
 		RX_DESC_DATA0_FRAME_LENGTH_GET_(le32_to_cpu(descriptor->data0));
 	buffer_length = buffer_info->buffer_length;
+	is_ice = descriptor->data1 & RX_DESC_DATA1_STATUS_ICE_;
+	is_tce = descriptor->data1 & RX_DESC_DATA1_STATUS_TCE_;
+	is_icsm = descriptor->data1 & RX_DESC_DATA1_STATUS_ICSM_;
 
 	netdev_dbg(netdev, "%s%schunk: %d/%d",
 		   is_first ? "first " : "      ",
@@ -2661,6 +2668,10 @@ static int lan743x_rx_process_buffer(struct lan743x_rx *rx)
 	if (is_last && rx->skb_head) {
 		rx->skb_head->protocol = eth_type_trans(rx->skb_head,
 							rx->adapter->netdev);
+		if (rx->adapter->netdev->features & NETIF_F_RXCSUM) {
+			if (!is_ice && !is_tce && !is_icsm)
+				skb->ip_summed = CHECKSUM_UNNECESSARY;
+		}
 		netdev_dbg(netdev, "sending %d byte frame to OS",
 			   rx->skb_head->len);
 		napi_gro_receive(&rx->napi, rx->skb_head);
@@ -3383,7 +3394,8 @@ static int lan743x_pcidev_probe(struct pci_dev *pdev,
 
 	adapter->netdev->netdev_ops = &lan743x_netdev_ops;
 	adapter->netdev->ethtool_ops = &lan743x_ethtool_ops;
-	adapter->netdev->features = NETIF_F_SG | NETIF_F_TSO | NETIF_F_HW_CSUM;
+	adapter->netdev->features = NETIF_F_SG | NETIF_F_TSO |
+				    NETIF_F_HW_CSUM | NETIF_F_RXCSUM;
 	adapter->netdev->hw_features = adapter->netdev->features;
 
 	/* carrier off reporting is important to ethtool even BEFORE open */
diff --git a/drivers/net/ethernet/microchip/lan743x_main.h b/drivers/net/ethernet/microchip/lan743x_main.h
index 58eb7abf976b..67877d3b6dd9 100644
--- a/drivers/net/ethernet/microchip/lan743x_main.h
+++ b/drivers/net/ethernet/microchip/lan743x_main.h
@@ -266,6 +266,8 @@
 #define RFE_ADDR_FILT_LO(x)		(0x404 + (8 * (x)))
 
 #define RFE_CTL				(0x508)
+#define RFE_CTL_TCP_UDP_COE_		BIT(12)
+#define RFE_CTL_IP_COE_			BIT(11)
 #define RFE_CTL_AB_			BIT(10)
 #define RFE_CTL_AM_			BIT(9)
 #define RFE_CTL_AU_			BIT(8)
@@ -1121,6 +1123,9 @@ struct lan743x_tx_buffer_info {
 	(((data0) & RX_DESC_DATA0_FRAME_LENGTH_MASK_) >> 16)
 #define RX_DESC_DATA0_EXT_                (0x00004000)
 #define RX_DESC_DATA0_BUF_LENGTH_MASK_    (0x00003FFF)
+#define RX_DESC_DATA1_STATUS_ICE_         (0x00020000)
+#define RX_DESC_DATA1_STATUS_TCE_         (0x00010000)
+#define RX_DESC_DATA1_STATUS_ICSM_        (0x00000001)
 #define RX_DESC_DATA2_TS_NS_MASK_         (0x3FFFFFFF)
 
 #if ((NET_IP_ALIGN != 0) && (NET_IP_ALIGN != 2))
-- 
2.25.1

Re: [PATCH net-next V1 2/2] net: lan743x: Add support for Rx IP & TCP checksum offload

[kernel test robot]

Hi Raju,

Thank you for the patch! Perhaps something to improve:

[auto build test WARNING on net-next/master]

url:    https://github.com/intel-lab-lkp/linux/commits/Raju-Lakkaraju/net-lan743x-Fix-to-use-multiqueue-start-stop-APIs/20220907-143456
base:   https://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next.git 2786bcff28bd88955fc61adf9cb7370fbc182bad
config: sparc-randconfig-s043-20220907 (https://download.01.org/0day-ci/archive/20220908/202209080357.QlM4jMJx-lkp@intel.com/config)
compiler: sparc-linux-gcc (GCC) 12.1.0
reproduce:
        wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
        chmod +x ~/bin/make.cross
        # apt-get install sparse
        # sparse version: v0.6.4-39-gce1a6720-dirty
        # https://github.com/intel-lab-lkp/linux/commit/d965e2dd24f617d68ded6f3cbeb69aa8271d5221
        git remote add linux-review https://github.com/intel-lab-lkp/linux
        git fetch --no-tags linux-review Raju-Lakkaraju/net-lan743x-Fix-to-use-multiqueue-start-stop-APIs/20220907-143456
        git checkout d965e2dd24f617d68ded6f3cbeb69aa8271d5221
        # save the config file
        mkdir build_dir && cp config build_dir/.config
        COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-12.1.0 make.cross C=1 CF='-fdiagnostic-prefix -D__CHECK_ENDIAN__' O=build_dir ARCH=sparc SHELL=/bin/bash drivers/net/ethernet/microchip/

If you fix the issue, kindly add following tag where applicable
Reported-by: kernel test robot <lkp@intel.com>

sparse warnings: (new ones prefixed by >>)
>> drivers/net/ethernet/microchip/lan743x_main.c:2600:28: sparse: sparse: restricted __le32 degrades to integer
   drivers/net/ethernet/microchip/lan743x_main.c:2601:28: sparse: sparse: restricted __le32 degrades to integer
   drivers/net/ethernet/microchip/lan743x_main.c:2602:29: sparse: sparse: restricted __le32 degrades to integer

vim +2600 drivers/net/ethernet/microchip/lan743x_main.c

  2544	
  2545	static int lan743x_rx_process_buffer(struct lan743x_rx *rx)
  2546	{
  2547		int current_head_index = le32_to_cpu(*rx->head_cpu_ptr);
  2548		struct lan743x_rx_descriptor *descriptor, *desc_ext;
  2549		struct net_device *netdev = rx->adapter->netdev;
  2550		int result = RX_PROCESS_RESULT_NOTHING_TO_DO;
  2551		struct lan743x_rx_buffer_info *buffer_info;
  2552		int frame_length, buffer_length;
  2553		bool is_ice, is_tce, is_icsm;
  2554		int extension_index = -1;
  2555		bool is_last, is_first;
  2556		struct sk_buff *skb;
  2557	
  2558		if (current_head_index < 0 || current_head_index >= rx->ring_size)
  2559			goto done;
  2560	
  2561		if (rx->last_head < 0 || rx->last_head >= rx->ring_size)
  2562			goto done;
  2563	
  2564		if (rx->last_head == current_head_index)
  2565			goto done;
  2566	
  2567		descriptor = &rx->ring_cpu_ptr[rx->last_head];
  2568		if (le32_to_cpu(descriptor->data0) & RX_DESC_DATA0_OWN_)
  2569			goto done;
  2570		buffer_info = &rx->buffer_info[rx->last_head];
  2571	
  2572		is_last = le32_to_cpu(descriptor->data0) & RX_DESC_DATA0_LS_;
  2573		is_first = le32_to_cpu(descriptor->data0) & RX_DESC_DATA0_FS_;
  2574	
  2575		if (is_last && le32_to_cpu(descriptor->data0) & RX_DESC_DATA0_EXT_) {
  2576			/* extension is expected to follow */
  2577			int index = lan743x_rx_next_index(rx, rx->last_head);
  2578	
  2579			if (index == current_head_index)
  2580				/* extension not yet available */
  2581				goto done;
  2582			desc_ext = &rx->ring_cpu_ptr[index];
  2583			if (le32_to_cpu(desc_ext->data0) & RX_DESC_DATA0_OWN_)
  2584				/* extension not yet available */
  2585				goto done;
  2586			if (!(le32_to_cpu(desc_ext->data0) & RX_DESC_DATA0_EXT_))
  2587				goto move_forward;
  2588			extension_index = index;
  2589		}
  2590	
  2591		/* Only the last buffer in a multi-buffer frame contains the total frame
  2592		 * length. The chip occasionally sends more buffers than strictly
  2593		 * required to reach the total frame length.
  2594		 * Handle this by adding all buffers to the skb in their entirety.
  2595		 * Once the real frame length is known, trim the skb.
  2596		 */
  2597		frame_length =
  2598			RX_DESC_DATA0_FRAME_LENGTH_GET_(le32_to_cpu(descriptor->data0));
  2599		buffer_length = buffer_info->buffer_length;
> 2600		is_ice = descriptor->data1 & RX_DESC_DATA1_STATUS_ICE_;
  2601		is_tce = descriptor->data1 & RX_DESC_DATA1_STATUS_TCE_;
  2602		is_icsm = descriptor->data1 & RX_DESC_DATA1_STATUS_ICSM_;
  2603	
  2604		netdev_dbg(netdev, "%s%schunk: %d/%d",
  2605			   is_first ? "first " : "      ",
  2606			   is_last  ? "last  " : "      ",
  2607			   frame_length, buffer_length);
  2608	
  2609		/* save existing skb, allocate new skb and map to dma */
  2610		skb = buffer_info->skb;
  2611		if (lan743x_rx_init_ring_element(rx, rx->last_head,
  2612						 GFP_ATOMIC | GFP_DMA)) {
  2613			/* failed to allocate next skb.
  2614			 * Memory is very low.
  2615			 * Drop this packet and reuse buffer.
  2616			 */
  2617			lan743x_rx_reuse_ring_element(rx, rx->last_head);
  2618			/* drop packet that was being assembled */
  2619			dev_kfree_skb_irq(rx->skb_head);
  2620			rx->skb_head = NULL;
  2621			goto process_extension;
  2622		}
  2623	
  2624		/* add buffers to skb via skb->frag_list */
  2625		if (is_first) {
  2626			skb_reserve(skb, RX_HEAD_PADDING);
  2627			skb_put(skb, buffer_length - RX_HEAD_PADDING);
  2628			if (rx->skb_head)
  2629				dev_kfree_skb_irq(rx->skb_head);
  2630			rx->skb_head = skb;
  2631		} else if (rx->skb_head) {
  2632			skb_put(skb, buffer_length);
  2633			if (skb_shinfo(rx->skb_head)->frag_list)
  2634				rx->skb_tail->next = skb;
  2635			else
  2636				skb_shinfo(rx->skb_head)->frag_list = skb;
  2637			rx->skb_tail = skb;
  2638			rx->skb_head->len += skb->len;
  2639			rx->skb_head->data_len += skb->len;
  2640			rx->skb_head->truesize += skb->truesize;
  2641		} else {
  2642			/* packet to assemble has already been dropped because one or
  2643			 * more of its buffers could not be allocated
  2644			 */
  2645			netdev_dbg(netdev, "drop buffer intended for dropped packet");
  2646			dev_kfree_skb_irq(skb);
  2647		}
  2648	
  2649	process_extension:
  2650		if (extension_index >= 0) {
  2651			u32 ts_sec;
  2652			u32 ts_nsec;
  2653	
  2654			ts_sec = le32_to_cpu(desc_ext->data1);
  2655			ts_nsec = (le32_to_cpu(desc_ext->data2) &
  2656				  RX_DESC_DATA2_TS_NS_MASK_);
  2657			if (rx->skb_head)
  2658				skb_hwtstamps(rx->skb_head)->hwtstamp =
  2659					ktime_set(ts_sec, ts_nsec);
  2660			lan743x_rx_reuse_ring_element(rx, extension_index);
  2661			rx->last_head = extension_index;
  2662			netdev_dbg(netdev, "process extension");
  2663		}
  2664	
  2665		if (is_last && rx->skb_head)
  2666			rx->skb_head = lan743x_rx_trim_skb(rx->skb_head, frame_length);
  2667	
  2668		if (is_last && rx->skb_head) {
  2669			rx->skb_head->protocol = eth_type_trans(rx->skb_head,
  2670								rx->adapter->netdev);
  2671			if (rx->adapter->netdev->features & NETIF_F_RXCSUM) {
  2672				if (!is_ice && !is_tce && !is_icsm)
  2673					skb->ip_summed = CHECKSUM_UNNECESSARY;
  2674			}
  2675			netdev_dbg(netdev, "sending %d byte frame to OS",
  2676				   rx->skb_head->len);
  2677			napi_gro_receive(&rx->napi, rx->skb_head);
  2678			rx->skb_head = NULL;
  2679		}
  2680	
  2681	move_forward:
  2682		/* push tail and head forward */
  2683		rx->last_tail = rx->last_head;
  2684		rx->last_head = lan743x_rx_next_index(rx, rx->last_head);
  2685		result = RX_PROCESS_RESULT_BUFFER_RECEIVED;
  2686	done:
  2687		return result;
  2688	}
  2689	

-- 
0-DAY CI Kernel Test Service
https://01.org/lkp