All of lore.kernel.org
 help / color / mirror / Atom feed
From: Gerhard Engleder <gerhard@engleder-embedded.com>
To: netdev@vger.kernel.org
Cc: davem@davemloft.net, kuba@kernel.org, edumazet@google.com,
	pabeni@redhat.com, robh+dt@kernel.org,
	krzysztof.kozlowski+dt@linaro.org, devicetree@vger.kernel.org,
	Gerhard Engleder <gerhard@engleder-embedded.com>
Subject: [PATCH net-next v3 6/6] tsnep: Use page pool for RX
Date: Fri, 23 Sep 2022 22:29:11 +0200	[thread overview]
Message-ID: <20220923202911.119729-7-gerhard@engleder-embedded.com> (raw)
In-Reply-To: <20220923202911.119729-1-gerhard@engleder-embedded.com>

Use page pool for RX buffer handling. Makes RX path more efficient and
is required prework for future XDP support.

Signed-off-by: Gerhard Engleder <gerhard@engleder-embedded.com>
---
 drivers/net/ethernet/engleder/Kconfig      |   1 +
 drivers/net/ethernet/engleder/tsnep.h      |   5 +-
 drivers/net/ethernet/engleder/tsnep_main.c | 162 ++++++++++++---------
 3 files changed, 100 insertions(+), 68 deletions(-)

diff --git a/drivers/net/ethernet/engleder/Kconfig b/drivers/net/ethernet/engleder/Kconfig
index f4e2b1102d8f..3df6bf476ae7 100644
--- a/drivers/net/ethernet/engleder/Kconfig
+++ b/drivers/net/ethernet/engleder/Kconfig
@@ -21,6 +21,7 @@ config TSNEP
 	depends on HAS_IOMEM && HAS_DMA
 	depends on PTP_1588_CLOCK_OPTIONAL
 	select PHYLIB
+	select PAGE_POOL
 	help
 	  Support for the Engleder TSN endpoint Ethernet MAC IP Core.
 
diff --git a/drivers/net/ethernet/engleder/tsnep.h b/drivers/net/ethernet/engleder/tsnep.h
index 2ca34ae9b55a..09a723b827c7 100644
--- a/drivers/net/ethernet/engleder/tsnep.h
+++ b/drivers/net/ethernet/engleder/tsnep.h
@@ -96,9 +96,9 @@ struct tsnep_rx_entry {
 
 	u32 properties;
 
-	struct sk_buff *skb;
+	struct page *page;
 	size_t len;
-	DEFINE_DMA_UNMAP_ADDR(dma);
+	dma_addr_t dma;
 };
 
 struct tsnep_rx {
@@ -113,6 +113,7 @@ struct tsnep_rx {
 	int read;
 	u32 owner_counter;
 	int increment_owner_counter;
+	struct page_pool *page_pool;
 
 	u32 packets;
 	u32 bytes;
diff --git a/drivers/net/ethernet/engleder/tsnep_main.c b/drivers/net/ethernet/engleder/tsnep_main.c
index a6b81f32d76b..8a93d0aa7faa 100644
--- a/drivers/net/ethernet/engleder/tsnep_main.c
+++ b/drivers/net/ethernet/engleder/tsnep_main.c
@@ -27,10 +27,10 @@
 #include <linux/phy.h>
 #include <linux/iopoll.h>
 
-#define RX_SKB_LENGTH (round_up(TSNEP_RX_INLINE_METADATA_SIZE + ETH_HLEN + \
-				TSNEP_MAX_FRAME_SIZE + ETH_FCS_LEN, 4))
-#define RX_SKB_RESERVE ((16 - TSNEP_RX_INLINE_METADATA_SIZE) + NET_IP_ALIGN)
-#define RX_SKB_ALLOC_LENGTH (RX_SKB_RESERVE + RX_SKB_LENGTH)
+#define TSNEP_SKB_PAD (NET_SKB_PAD + NET_IP_ALIGN)
+#define TSNEP_HEADROOM ALIGN(TSNEP_SKB_PAD, 4)
+#define TSNEP_MAX_RX_BUF_SIZE (PAGE_SIZE - TSNEP_HEADROOM - \
+			       SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
 
 #ifdef CONFIG_ARCH_DMA_ADDR_T_64BIT
 #define DMA_ADDR_HIGH(dma_addr) ((u32)(((dma_addr) >> 32) & 0xFFFFFFFF))
@@ -587,14 +587,15 @@ static void tsnep_rx_ring_cleanup(struct tsnep_rx *rx)
 
 	for (i = 0; i < TSNEP_RING_SIZE; i++) {
 		entry = &rx->entry[i];
-		if (dma_unmap_addr(entry, dma))
-			dma_unmap_single(dmadev, dma_unmap_addr(entry, dma),
-					 dma_unmap_len(entry, len),
-					 DMA_FROM_DEVICE);
-		if (entry->skb)
-			dev_kfree_skb(entry->skb);
+		if (entry->page)
+			page_pool_put_full_page(rx->page_pool, entry->page,
+						false);
+		entry->page = NULL;
 	}
 
+	if (rx->page_pool)
+		page_pool_destroy(rx->page_pool);
+
 	memset(rx->entry, 0, sizeof(rx->entry));
 
 	for (i = 0; i < TSNEP_RING_PAGE_COUNT; i++) {
@@ -607,31 +608,19 @@ static void tsnep_rx_ring_cleanup(struct tsnep_rx *rx)
 	}
 }
 
-static int tsnep_rx_alloc_and_map_skb(struct tsnep_rx *rx,
-				      struct tsnep_rx_entry *entry)
+static int tsnep_rx_alloc_buffer(struct tsnep_rx *rx,
+				 struct tsnep_rx_entry *entry)
 {
-	struct device *dmadev = rx->adapter->dmadev;
-	struct sk_buff *skb;
-	dma_addr_t dma;
+	struct page *page;
 
-	skb = __netdev_alloc_skb(rx->adapter->netdev, RX_SKB_ALLOC_LENGTH,
-				 GFP_ATOMIC | GFP_DMA);
-	if (!skb)
+	page = page_pool_dev_alloc_pages(rx->page_pool);
+	if (unlikely(!page))
 		return -ENOMEM;
 
-	skb_reserve(skb, RX_SKB_RESERVE);
-
-	dma = dma_map_single(dmadev, skb->data, RX_SKB_LENGTH,
-			     DMA_FROM_DEVICE);
-	if (dma_mapping_error(dmadev, dma)) {
-		dev_kfree_skb(skb);
-		return -ENOMEM;
-	}
-
-	entry->skb = skb;
-	entry->len = RX_SKB_LENGTH;
-	dma_unmap_addr_set(entry, dma, dma);
-	entry->desc->rx = __cpu_to_le64(dma);
+	entry->page = page;
+	entry->len = TSNEP_MAX_RX_BUF_SIZE;
+	entry->dma = page_pool_get_dma_addr(entry->page);
+	entry->desc->rx = __cpu_to_le64(entry->dma + TSNEP_SKB_PAD);
 
 	return 0;
 }
@@ -640,6 +629,7 @@ static int tsnep_rx_ring_init(struct tsnep_rx *rx)
 {
 	struct device *dmadev = rx->adapter->dmadev;
 	struct tsnep_rx_entry *entry;
+	struct page_pool_params pp_params = { 0 };
 	struct tsnep_rx_entry *next_entry;
 	int i, j;
 	int retval;
@@ -661,12 +651,28 @@ static int tsnep_rx_ring_init(struct tsnep_rx *rx)
 			entry->desc_dma = rx->page_dma[i] + TSNEP_DESC_SIZE * j;
 		}
 	}
+
+	pp_params.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV;
+	pp_params.order = 0;
+	pp_params.pool_size = TSNEP_RING_SIZE;
+	pp_params.nid = dev_to_node(dmadev);
+	pp_params.dev = dmadev;
+	pp_params.dma_dir = DMA_FROM_DEVICE;
+	pp_params.max_len = TSNEP_MAX_RX_BUF_SIZE;
+	pp_params.offset = TSNEP_SKB_PAD;
+	rx->page_pool = page_pool_create(&pp_params);
+	if (IS_ERR(rx->page_pool)) {
+		retval = PTR_ERR(rx->page_pool);
+		rx->page_pool = NULL;
+		goto failed;
+	}
+
 	for (i = 0; i < TSNEP_RING_SIZE; i++) {
 		entry = &rx->entry[i];
 		next_entry = &rx->entry[(i + 1) % TSNEP_RING_SIZE];
 		entry->desc->next = __cpu_to_le64(next_entry->desc_dma);
 
-		retval = tsnep_rx_alloc_and_map_skb(rx, entry);
+		retval = tsnep_rx_alloc_buffer(rx, entry);
 		if (retval)
 			goto failed;
 	}
@@ -682,7 +688,7 @@ static void tsnep_rx_activate(struct tsnep_rx *rx, int index)
 {
 	struct tsnep_rx_entry *entry = &rx->entry[index];
 
-	/* RX_SKB_LENGTH is a multiple of 4 */
+	/* TSNEP_MAX_RX_BUF_SIZE is a multiple of 4 */
 	entry->properties = entry->len & TSNEP_DESC_LENGTH_MASK;
 	entry->properties |= TSNEP_DESC_INTERRUPT_FLAG;
 	if (index == rx->increment_owner_counter) {
@@ -705,19 +711,52 @@ static void tsnep_rx_activate(struct tsnep_rx *rx, int index)
 	entry->desc->properties = __cpu_to_le32(entry->properties);
 }
 
+static struct sk_buff *tsnep_build_skb(struct tsnep_rx *rx, struct page *page,
+				       int length)
+{
+	struct sk_buff *skb;
+
+	skb = napi_build_skb(page_address(page), PAGE_SIZE);
+	if (unlikely(!skb))
+		return NULL;
+
+	/* update pointers within the skb to store the data */
+	skb_reserve(skb, TSNEP_SKB_PAD + TSNEP_RX_INLINE_METADATA_SIZE);
+	__skb_put(skb, length - TSNEP_RX_INLINE_METADATA_SIZE - ETH_FCS_LEN);
+
+	if (rx->adapter->hwtstamp_config.rx_filter == HWTSTAMP_FILTER_ALL) {
+		struct skb_shared_hwtstamps *hwtstamps = skb_hwtstamps(skb);
+		struct tsnep_rx_inline *rx_inline =
+			(struct tsnep_rx_inline *)(page_address(page) +
+						   TSNEP_SKB_PAD);
+
+		skb_shinfo(skb)->tx_flags |=
+			SKBTX_HW_TSTAMP_NETDEV;
+		memset(hwtstamps, 0, sizeof(*hwtstamps));
+		hwtstamps->netdev_data = rx_inline;
+	}
+
+	skb_record_rx_queue(skb, rx->queue_index);
+	skb->protocol = eth_type_trans(skb, rx->adapter->netdev);
+
+	return skb;
+}
+
 static int tsnep_rx_poll(struct tsnep_rx *rx, struct napi_struct *napi,
 			 int budget)
 {
 	struct device *dmadev = rx->adapter->dmadev;
 	int done = 0;
+	enum dma_data_direction dma_dir;
 	struct tsnep_rx_entry *entry;
+	struct page *page;
 	struct sk_buff *skb;
-	size_t len;
-	dma_addr_t dma;
 	int length;
 	bool enable = false;
 	int retval;
 
+	dma_dir = page_pool_get_dma_dir(rx->page_pool);
+
 	while (likely(done < budget)) {
 		entry = &rx->entry[rx->read];
 		if ((__le32_to_cpu(entry->desc_wb->properties) &
@@ -730,43 +769,34 @@ static int tsnep_rx_poll(struct tsnep_rx *rx, struct napi_struct *napi,
 		 */
 		dma_rmb();
 
-		skb = entry->skb;
-		len = dma_unmap_len(entry, len);
-		dma = dma_unmap_addr(entry, dma);
+		prefetch(page_address(entry->page) + TSNEP_SKB_PAD);
+		length = __le32_to_cpu(entry->desc_wb->properties) &
+			 TSNEP_DESC_LENGTH_MASK;
+		dma_sync_single_range_for_cpu(dmadev, entry->dma, TSNEP_SKB_PAD,
+					      length, dma_dir);
+		page = entry->page;
 
 		/* forward skb only if allocation is successful, otherwise
-		 * skb is reused and frame dropped
+		 * page is reused and frame dropped
 		 */
-		retval = tsnep_rx_alloc_and_map_skb(rx, entry);
+		retval = tsnep_rx_alloc_buffer(rx, entry);
 		if (!retval) {
-			dma_unmap_single(dmadev, dma, len, DMA_FROM_DEVICE);
-
-			length = __le32_to_cpu(entry->desc_wb->properties) &
-				 TSNEP_DESC_LENGTH_MASK;
-			skb_put(skb, length - ETH_FCS_LEN);
-			if (rx->adapter->hwtstamp_config.rx_filter ==
-			    HWTSTAMP_FILTER_ALL) {
-				struct skb_shared_hwtstamps *hwtstamps =
-					skb_hwtstamps(skb);
-				struct tsnep_rx_inline *rx_inline =
-					(struct tsnep_rx_inline *)skb->data;
-
-				skb_shinfo(skb)->tx_flags |=
-					SKBTX_HW_TSTAMP_NETDEV;
-				memset(hwtstamps, 0, sizeof(*hwtstamps));
-				hwtstamps->netdev_data = rx_inline;
-			}
-			skb_pull(skb, TSNEP_RX_INLINE_METADATA_SIZE);
-			skb_record_rx_queue(skb, rx->queue_index);
-			skb->protocol = eth_type_trans(skb,
-						       rx->adapter->netdev);
+			skb = tsnep_build_skb(rx, page, length);
+			if (skb) {
+				page_pool_release_page(rx->page_pool, page);
+
+				rx->packets++;
+				rx->bytes += length -
+					     TSNEP_RX_INLINE_METADATA_SIZE;
+				if (skb->pkt_type == PACKET_MULTICAST)
+					rx->multicast++;
 
-			rx->packets++;
-			rx->bytes += length - TSNEP_RX_INLINE_METADATA_SIZE;
-			if (skb->pkt_type == PACKET_MULTICAST)
-				rx->multicast++;
+				napi_gro_receive(napi, skb);
+			} else {
+				page_pool_recycle_direct(rx->page_pool, page);
 
-			napi_gro_receive(napi, skb);
+				rx->dropped++;
+			}
 			done++;
 		} else {
 			rx->dropped++;
-- 
2.30.2


      parent reply	other threads:[~2022-09-23 20:37 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-09-23 20:29 [PATCH net-next v3 0/6] tsnep: multi queue support and some other improvements Gerhard Engleder
2022-09-23 20:29 ` [PATCH net-next v3 1/6] dt-bindings: net: tsnep: Allow dma-coherent Gerhard Engleder
2022-09-24  9:15   ` Krzysztof Kozlowski
2022-09-24 18:11     ` Gerhard Engleder
2022-09-25  7:41       ` Krzysztof Kozlowski
2022-09-25  8:14         ` Gerhard Engleder
2022-09-26  8:55           ` Krzysztof Kozlowski
2022-09-26 17:46             ` Gerhard Engleder
2022-09-28  7:17               ` Krzysztof Kozlowski
2022-09-23 20:29 ` [PATCH net-next v3 2/6] dt-bindings: net: tsnep: Allow additional interrupts Gerhard Engleder
2022-09-24  9:17   ` Krzysztof Kozlowski
2022-09-24 19:06     ` Gerhard Engleder
2022-09-23 20:29 ` [PATCH net-next v3 3/6] tsnep: Move interrupt from device to queue Gerhard Engleder
2022-09-23 20:29 ` [PATCH net-next v3 4/6] tsnep: Support multiple TX/RX queue pairs Gerhard Engleder
2022-09-23 20:29 ` [PATCH net-next v3 5/6] tsnep: Add EtherType RX flow classification support Gerhard Engleder
2022-09-23 20:29 ` Gerhard Engleder [this message]

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220923202911.119729-7-gerhard@engleder-embedded.com \
    --to=gerhard@engleder-embedded.com \
    --cc=davem@davemloft.net \
    --cc=devicetree@vger.kernel.org \
    --cc=edumazet@google.com \
    --cc=krzysztof.kozlowski+dt@linaro.org \
    --cc=kuba@kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=robh+dt@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body. 
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.