diff mbox series

[net-next,7/7] bnxt_en: Use page frag RX buffers for better software GRO performance

Message ID 1640592032-8927-8-git-send-email-michael.chan@broadcom.com (mailing list archive)
State Superseded
Commit 720908e5f816d56579e098e32bd3b56bad2be8f0
Delegated to: Netdev Maintainers
Headers show
Series bnxt_en: Update for net-next | expand

Checks

Context Check Description
netdev/tree_selection success Clearly marked for net-next
netdev/fixes_present success Fixes tag not required for -next series
netdev/subject_prefix success Link
netdev/cover_letter success Series has a cover letter
netdev/patch_count success Link
netdev/header_inline success No static functions without inline keyword in header files
netdev/build_32bit success Errors and warnings before: 1 this patch: 1
netdev/cc_maintainers success CCed 3 of 3 maintainers
netdev/build_clang success Errors and warnings before: 0 this patch: 0
netdev/module_param success Was 0 now: 0
netdev/verify_signedoff success Signed-off-by tag matches author and committer
netdev/verify_fixes success No Fixes tag
netdev/build_allmodconfig_warn success Errors and warnings before: 1 this patch: 1
netdev/checkpatch success total: 0 errors, 0 warnings, 0 checks, 94 lines checked
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline warning Was 1 now: 1

Commit Message

Michael Chan Dec. 27, 2021, 8 a.m. UTC 
From: Jakub Kicinski <kuba@kernel.org>

If NETIF_F_GRO_HW is disabled, the existing driver code uses kmalloc'ed
data for RX buffers.  This causes inefficient SW GRO performance
because the GRO data is merged using the less efficient frag_list.
Use netdev_alloc_frag() and friends instead so that GRO data can be
merged into skb_shinfo(skb)->frags for better performance.

[Use skb_free_frag() - Vikas Gupta]

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Reviewed-by: Andy Gospodarek <gospo@broadcom.com>
Signed-off-by: Vikas Gupta <vikas.gupta@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
---
 drivers/net/ethernet/broadcom/bnxt/bnxt.c | 27 +++++++++++++----------
 1 file changed, 15 insertions(+), 12 deletions(-)
diff mbox series

Patch

diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
index ea5b072ba292..4d7ea62e24fb 100644
--- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c
+++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c
@@ -741,13 +741,16 @@  static struct page *__bnxt_alloc_rx_page(struct bnxt *bp, dma_addr_t *mapping,
 	return page;
 }
 
-static inline u8 *__bnxt_alloc_rx_data(struct bnxt *bp, dma_addr_t *mapping,
+static inline u8 *__bnxt_alloc_rx_frag(struct bnxt *bp, dma_addr_t *mapping,
 				       gfp_t gfp)
 {
 	u8 *data;
 	struct pci_dev *pdev = bp->pdev;
 
-	data = kmalloc(bp->rx_buf_size, gfp);
+	if (gfp == GFP_ATOMIC)
+		data = napi_alloc_frag(bp->rx_buf_size);
+	else
+		data = netdev_alloc_frag(bp->rx_buf_size);
 	if (!data)
 		return NULL;
 
@@ -756,7 +759,7 @@  static inline u8 *__bnxt_alloc_rx_data(struct bnxt *bp, dma_addr_t *mapping,
 					DMA_ATTR_WEAK_ORDERING);
 
 	if (dma_mapping_error(&pdev->dev, *mapping)) {
-		kfree(data);
+		skb_free_frag(data);
 		data = NULL;
 	}
 	return data;
@@ -779,7 +782,7 @@  int bnxt_alloc_rx_data(struct bnxt *bp, struct bnxt_rx_ring_info *rxr,
 		rx_buf->data = page;
 		rx_buf->data_ptr = page_address(page) + bp->rx_offset;
 	} else {
-		u8 *data = __bnxt_alloc_rx_data(bp, &mapping, gfp);
+		u8 *data = __bnxt_alloc_rx_frag(bp, &mapping, gfp);
 
 		if (!data)
 			return -ENOMEM;
@@ -1021,11 +1024,11 @@  static struct sk_buff *bnxt_rx_skb(struct bnxt *bp,
 		return NULL;
 	}
 
-	skb = build_skb(data, 0);
+	skb = build_skb(data, bp->rx_buf_size);
 	dma_unmap_single_attrs(&bp->pdev->dev, dma_addr, bp->rx_buf_use_size,
 			       bp->rx_dir, DMA_ATTR_WEAK_ORDERING);
 	if (!skb) {
-		kfree(data);
+		skb_free_frag(data);
 		return NULL;
 	}
 
@@ -1613,7 +1616,7 @@  static inline struct sk_buff *bnxt_tpa_end(struct bnxt *bp,
 		u8 *new_data;
 		dma_addr_t new_mapping;
 
-		new_data = __bnxt_alloc_rx_data(bp, &new_mapping, GFP_ATOMIC);
+		new_data = __bnxt_alloc_rx_frag(bp, &new_mapping, GFP_ATOMIC);
 		if (!new_data) {
 			bnxt_abort_tpa(cpr, idx, agg_bufs);
 			cpr->sw_stats.rx.rx_oom_discards += 1;
@@ -1624,13 +1627,13 @@  static inline struct sk_buff *bnxt_tpa_end(struct bnxt *bp,
 		tpa_info->data_ptr = new_data + bp->rx_offset;
 		tpa_info->mapping = new_mapping;
 
-		skb = build_skb(data, 0);
+		skb = build_skb(data, bp->rx_buf_size);
 		dma_unmap_single_attrs(&bp->pdev->dev, mapping,
 				       bp->rx_buf_use_size, bp->rx_dir,
 				       DMA_ATTR_WEAK_ORDERING);
 
 		if (!skb) {
-			kfree(data);
+			skb_free_frag(data);
 			bnxt_abort_tpa(cpr, idx, agg_bufs);
 			cpr->sw_stats.rx.rx_oom_discards += 1;
 			return NULL;
@@ -2796,7 +2799,7 @@  static void bnxt_free_one_rx_ring_skbs(struct bnxt *bp, int ring_nr)
 
 		tpa_info->data = NULL;
 
-		kfree(data);
+		skb_free_frag(data);
 	}
 
 skip_rx_tpa_free:
@@ -2822,7 +2825,7 @@  static void bnxt_free_one_rx_ring_skbs(struct bnxt *bp, int ring_nr)
 			dma_unmap_single_attrs(&pdev->dev, mapping,
 					       bp->rx_buf_use_size, bp->rx_dir,
 					       DMA_ATTR_WEAK_ORDERING);
-			kfree(data);
+			skb_free_frag(data);
 		}
 	}
 
@@ -3526,7 +3529,7 @@  static int bnxt_alloc_one_rx_ring(struct bnxt *bp, int ring_nr)
 		u8 *data;
 
 		for (i = 0; i < bp->max_tpa; i++) {
-			data = __bnxt_alloc_rx_data(bp, &mapping, GFP_KERNEL);
+			data = __bnxt_alloc_rx_frag(bp, &mapping, GFP_KERNEL);
 			if (!data)
 				return -ENOMEM;