enic: Use the Page Pool API for RX

The Page Pool API improves bandwidth and CPU overhead by recycling pages
instead of allocating new buffers in the driver. Make use of page pool
fragment allocation for smaller MTUs so that multiple packets can share
a page. For MTUs larger than PAGE_SIZE, adjust the 'order' page
parameter so that contiguous pages can be used to receive the larger
packets.

The RQ descriptor field 'os_buf' is repurposed to hold page pointers
allocated from page_pool instead of SKBs. When packets arrive, SKBs are
allocated and the page pointers are attached instead of preallocating SKBs.

'alloc_fail' netdev statistic is incremented when page_pool_dev_alloc()
fails.

Co-developed-by: Nelson Escobar <[email protected]>
Signed-off-by: Nelson Escobar <[email protected]>
Co-developed-by: Satish Kharat <[email protected]>
Signed-off-by: Satish Kharat <[email protected]>
Signed-off-by: John Daley <[email protected]>
Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski <[email protected]>

Author: johndale88

drivers/net/ethernet/cisco/enic/enic.h

@@ -17,6 +17,7 @@
 #include "vnic_nic.h"
 #include "vnic_rss.h"
 #include <linux/irq.h>
+#include <net/page_pool/helpers.h>
 
 #define DRV_NAME		"enic"
 #define DRV_DESCRIPTION		"Cisco VIC Ethernet NIC Driver"
@@ -158,6 +159,7 @@ struct enic_rq_stats {
 	u64 pkt_truncated;		/* truncated pkts */
 	u64 no_skb;			/* out of skbs */
 	u64 desc_skip;			/* Rx pkt went into later buffer */
+	u64 pp_alloc_fail;		/* page pool alloc failure */
 };
 
 struct enic_wq {
@@ -169,6 +171,7 @@ struct enic_wq {
 struct enic_rq {
 	struct vnic_rq vrq;
 	struct enic_rq_stats stats;
+	struct page_pool *pool;
 } ____cacheline_aligned;
 
 /* Per-instance private data structure */

drivers/net/ethernet/cisco/enic/enic_main.c

@@ -1736,6 +1736,17 @@ static int enic_open(struct net_device *netdev)
 	struct enic *enic = netdev_priv(netdev);
 	unsigned int i;
 	int err, ret;
+	unsigned int max_pkt_len = netdev->mtu + VLAN_ETH_HLEN;
+	struct page_pool_params pp_params = {
+		.order = get_order(max_pkt_len),
+		.pool_size = enic->config.rq_desc_count,
+		.nid = dev_to_node(&enic->pdev->dev),
+		.dev = &enic->pdev->dev,
+		.dma_dir = DMA_FROM_DEVICE,
+		.max_len = (max_pkt_len > PAGE_SIZE) ? max_pkt_len : PAGE_SIZE,
+		.netdev = netdev,
+		.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV,
+	};
 
 	err = enic_request_intr(enic);
 	if (err) {
@@ -1753,6 +1764,16 @@ static int enic_open(struct net_device *netdev)
 	}
 
 	for (i = 0; i < enic->rq_count; i++) {
+		/* create a page pool for each RQ */
+		pp_params.napi = &enic->napi[i];
+		pp_params.queue_idx = i;
+		enic->rq[i].pool = page_pool_create(&pp_params);
+		if (IS_ERR(enic->rq[i].pool)) {
+			err = PTR_ERR(enic->rq[i].pool);
+			enic->rq[i].pool = NULL;
+			goto err_out_free_rq;
+		}
+
 		/* enable rq before updating rq desc */
 		vnic_rq_enable(&enic->rq[i].vrq);
 		vnic_rq_fill(&enic->rq[i].vrq, enic_rq_alloc_buf);
@@ -1793,8 +1814,11 @@ static int enic_open(struct net_device *netdev)
 err_out_free_rq:
 	for (i = 0; i < enic->rq_count; i++) {
 		ret = vnic_rq_disable(&enic->rq[i].vrq);
-		if (!ret)
+		if (!ret) {
 			vnic_rq_clean(&enic->rq[i].vrq, enic_free_rq_buf);
+			page_pool_destroy(enic->rq[i].pool);
+			enic->rq[i].pool = NULL;
+		}
 	}
 	enic_dev_notify_unset(enic);
 err_out_free_intr:
@@ -1852,8 +1876,11 @@ static int enic_stop(struct net_device *netdev)
 
 	for (i = 0; i < enic->wq_count; i++)
 		vnic_wq_clean(&enic->wq[i].vwq, enic_free_wq_buf);
-	for (i = 0; i < enic->rq_count; i++)
+	for (i = 0; i < enic->rq_count; i++) {
 		vnic_rq_clean(&enic->rq[i].vrq, enic_free_rq_buf);
+		page_pool_destroy(enic->rq[i].pool);
+		enic->rq[i].pool = NULL;
+	}
 	for (i = 0; i < enic->cq_count; i++)
 		vnic_cq_clean(&enic->cq[i]);
 	for (i = 0; i < enic->intr_count; i++)
@@ -2363,6 +2390,7 @@ static void enic_get_queue_stats_rx(struct net_device *dev, int idx,
 	rxs->hw_drop_overruns = rqstats->pkt_truncated;
 	rxs->csum_unnecessary = rqstats->csum_unnecessary +
 				rqstats->csum_unnecessary_encap;
+	rxs->alloc_fail = rqstats->pp_alloc_fail;
 }
 
 static void enic_get_queue_stats_tx(struct net_device *dev, int idx,
@@ -2390,6 +2418,7 @@ static void enic_get_base_stats(struct net_device *dev,
 	rxs->hw_drops = 0;
 	rxs->hw_drop_overruns = 0;
 	rxs->csum_unnecessary = 0;
+	rxs->alloc_fail = 0;
 	txs->bytes = 0;
 	txs->packets = 0;
 	txs->csum_none = 0;

drivers/net/ethernet/cisco/enic/enic_rq.c

@@ -21,25 +21,6 @@ static void enic_intr_update_pkt_size(struct vnic_rx_bytes_counter *pkt_size,
 		pkt_size->small_pkt_bytes_cnt += pkt_len;
 }
 
-static bool enic_rxcopybreak(struct net_device *netdev, struct sk_buff **skb,
-			     struct vnic_rq_buf *buf, u16 len)
-{
-	struct enic *enic = netdev_priv(netdev);
-	struct sk_buff *new_skb;
-
-	if (len > enic->rx_copybreak)
-		return false;
-	new_skb = netdev_alloc_skb_ip_align(netdev, len);
-	if (!new_skb)
-		return false;
-	dma_sync_single_for_cpu(&enic->pdev->dev, buf->dma_addr, len,
-				DMA_FROM_DEVICE);
-	memcpy(new_skb->data, (*skb)->data, len);
-	*skb = new_skb;
-
-	return true;
-}
-
 int enic_rq_service(struct vnic_dev *vdev, struct cq_desc *cq_desc, u8 type,
 		    u16 q_number, u16 completed_index, void *opaque)
 {
@@ -142,57 +123,56 @@ int enic_rq_alloc_buf(struct vnic_rq *rq)
 {
 	struct enic *enic = vnic_dev_priv(rq->vdev);
 	struct net_device *netdev = enic->netdev;
-	struct sk_buff *skb;
+	struct enic_rq *erq = &enic->rq[rq->index];
+	struct enic_rq_stats *rqstats = &erq->stats;
+	unsigned int offset = 0;
 	unsigned int len = netdev->mtu + VLAN_ETH_HLEN;
 	unsigned int os_buf_index = 0;
 	dma_addr_t dma_addr;
 	struct vnic_rq_buf *buf = rq->to_use;
+	struct page *page;
+	unsigned int truesize = len;
 
 	if (buf->os_buf) {
 		enic_queue_rq_desc(rq, buf->os_buf, os_buf_index, buf->dma_addr,
 				   buf->len);
 
 		return 0;
 	}
-	skb = netdev_alloc_skb_ip_align(netdev, len);
-	if (!skb) {
-		enic->rq[rq->index].stats.no_skb++;
-		return -ENOMEM;
-	}
 
-	dma_addr = dma_map_single(&enic->pdev->dev, skb->data, len,
-				  DMA_FROM_DEVICE);
-	if (unlikely(enic_dma_map_check(enic, dma_addr))) {
-		dev_kfree_skb(skb);
+	page = page_pool_dev_alloc(erq->pool, &offset, &truesize);
+	if (unlikely(!page)) {
+		rqstats->pp_alloc_fail++;
 		return -ENOMEM;
 	}
-
-	enic_queue_rq_desc(rq, skb, os_buf_index, dma_addr, len);
+	buf->offset = offset;
+	buf->truesize = truesize;
+	dma_addr = page_pool_get_dma_addr(page) + offset;
+	enic_queue_rq_desc(rq, (void *)page, os_buf_index, dma_addr, len);
 
 	return 0;
 }
 
 void enic_free_rq_buf(struct vnic_rq *rq, struct vnic_rq_buf *buf)
 {
 	struct enic *enic = vnic_dev_priv(rq->vdev);
+	struct enic_rq *erq = &enic->rq[rq->index];
 
 	if (!buf->os_buf)
 		return;
 
-	dma_unmap_single(&enic->pdev->dev, buf->dma_addr, buf->len,
-			 DMA_FROM_DEVICE);
-	dev_kfree_skb_any(buf->os_buf);
+	page_pool_put_full_page(erq->pool, (struct page *)buf->os_buf, true);
 	buf->os_buf = NULL;
 }
 
 void enic_rq_indicate_buf(struct vnic_rq *rq, struct cq_desc *cq_desc,
 			  struct vnic_rq_buf *buf, int skipped, void *opaque)
 {
 	struct enic *enic = vnic_dev_priv(rq->vdev);
-	struct net_device *netdev = enic->netdev;
 	struct sk_buff *skb;
 	struct vnic_cq *cq = &enic->cq[enic_cq_rq(enic, rq->index)];
 	struct enic_rq_stats *rqstats = &enic->rq[rq->index].stats;
+	struct napi_struct *napi;
 
 	u8 type, color, eop, sop, ingress_port, vlan_stripped;
 	u8 fcoe, fcoe_sof, fcoe_fc_crc_ok, fcoe_enc_error, fcoe_eof;
@@ -208,8 +188,6 @@ void enic_rq_indicate_buf(struct vnic_rq *rq, struct cq_desc *cq_desc,
 		return;
 	}
 
-	skb = buf->os_buf;
-
 	cq_enet_rq_desc_dec((struct cq_enet_rq_desc *)cq_desc, &type, &color,
 			    &q_number, &completed_index, &ingress_port, &fcoe,
 			    &eop, &sop, &rss_type, &csum_not_calc, &rss_hash,
@@ -219,48 +197,46 @@ void enic_rq_indicate_buf(struct vnic_rq *rq, struct cq_desc *cq_desc,
 			    &tcp, &ipv4_csum_ok, &ipv6, &ipv4, &ipv4_fragment,
 			    &fcs_ok);
 
-	if (enic_rq_pkt_error(rq, packet_error, fcs_ok, bytes_written)) {
-		dma_unmap_single(&enic->pdev->dev, buf->dma_addr, buf->len,
-				 DMA_FROM_DEVICE);
-		dev_kfree_skb_any(skb);
-		buf->os_buf = NULL;
-
+	if (enic_rq_pkt_error(rq, packet_error, fcs_ok, bytes_written))
 		return;
-	}
 
 	if (eop && bytes_written > 0) {
 		/* Good receive
 		 */
 		rqstats->bytes += bytes_written;
-		if (!enic_rxcopybreak(netdev, &skb, buf, bytes_written)) {
-			buf->os_buf = NULL;
-			dma_unmap_single(&enic->pdev->dev, buf->dma_addr,
-					 buf->len, DMA_FROM_DEVICE);
+		napi = &enic->napi[rq->index];
+		skb = napi_get_frags(napi);
+		if (unlikely(!skb)) {
+			net_warn_ratelimited("%s: skb alloc error rq[%d], desc[%d]\n",
+					     enic->netdev->name, rq->index,
+					     completed_index);
+			rqstats->no_skb++;
+			return;
 		}
+
 		prefetch(skb->data - NET_IP_ALIGN);
 
-		skb_put(skb, bytes_written);
-		skb->protocol = eth_type_trans(skb, netdev);
+		dma_sync_single_for_cpu(&enic->pdev->dev, buf->dma_addr,
+					bytes_written, DMA_FROM_DEVICE);
+		skb_add_rx_frag(skb, skb_shinfo(skb)->nr_frags,
+				(struct page *)buf->os_buf, buf->offset,
+				bytes_written, buf->truesize);
 		skb_record_rx_queue(skb, q_number);
 		enic_rq_set_skb_flags(rq, type, rss_hash, rss_type, fcoe,
 				      fcoe_fc_crc_ok, vlan_stripped,
 				      csum_not_calc, tcp_udp_csum_ok, ipv6,
 				      ipv4_csum_ok, vlan_tci, skb);
-		skb_mark_napi_id(skb, &enic->napi[rq->index]);
-		if (!(netdev->features & NETIF_F_GRO))
-			netif_receive_skb(skb);
-		else
-			napi_gro_receive(&enic->napi[q_number], skb);
+		skb_mark_for_recycle(skb);
+		napi_gro_frags(napi);
 		if (enic->rx_coalesce_setting.use_adaptive_rx_coalesce)
 			enic_intr_update_pkt_size(&cq->pkt_size_counter,
 						  bytes_written);
+		buf->os_buf = NULL;
+		buf->dma_addr = 0;
+		buf = buf->next;
 	} else {
 		/* Buffer overflow
 		 */
 		rqstats->pkt_truncated++;
-		dma_unmap_single(&enic->pdev->dev, buf->dma_addr, buf->len,
-				 DMA_FROM_DEVICE);
-		dev_kfree_skb_any(skb);
-		buf->os_buf = NULL;
 	}
 }

drivers/net/ethernet/cisco/enic/vnic_rq.h

@@ -61,6 +61,8 @@ struct vnic_rq_buf {
 	unsigned int index;
 	void *desc;
 	uint64_t wr_id;
+	unsigned int offset;
+	unsigned int truesize;
 };
 
 enum enic_poll_state {