[PATCH net-next v5 0/2] net: ethernet: mediatek: add enhancements to RX path

From: Sean Wang <sean.wang@mediatek.com>

Changes since v1:
- fix message typos and add coverletter
  
Changes since v2:
- split from the previous series for submitting add enhancements as 
a series targeting 'net-next' and add indents before comments.

Changes since v3:
- merge the patch using PDMA RX path
- fixed the input of mtk_poll_rx is with the remaining budget

Changes since v4:
- save one wmb and register update when no packet is being handled
inside mtk_poll_rx call
- fixed incorrect return packet count from mtk_napi_rx

Sean Wang (2):
  net: ethernet: mediatek: enhance RX path by reducing the frequency of
    the memory barrier used
  net: ethernet: mediatek: enhance RX path by aggregating more SKBs into
    NAPI

 drivers/net/ethernet/mediatek/mtk_eth_soc.c | 26 +++++++++++++++-----------
 1 file changed, 15 insertions(+), 11 deletions(-)

-- 
1.9.1

From: Sean Wang <sean.wang@mediatek.com>

The patch makes move wmb() to outside the loop that could help
RX path handling more faster although that RX descriptors aren't
freed for DMA to use as soon as possible, but based on my experiment
and the result shows it still can reach about 943Mbpis without
performance drop that is tested based on the setup with one port
using Giga PHY and 256 RX descriptors for DMA to move.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c | 11 ++++++-----
 1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index 431b851..a1bdb53 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c

@@ -890,13 +890,14 @@ release_desc:
 		rxd->rxd2 = RX_DMA_PLEN0(ring->buf_size);
 
 		ring->calc_idx = idx;
-		/* make sure that all changes to the dma ring are flushed before
-		 * we continue
-		 */
-		wmb();
-		mtk_w32(eth, ring->calc_idx, MTK_PRX_CRX_IDX0);
+
 		done++;
 	}
+	/* make sure that all changes to the dma ring are flushed before
+	 * we continue
+	 */
+	wmb();
+	mtk_w32(eth, ring->calc_idx, MTK_PRX_CRX_IDX0);
 
 	if (done < budget)
 		mtk_w32(eth, MTK_RX_DONE_INT, MTK_PDMA_INT_STATUS);

-- 
1.9.1

From: Sean Wang <sean.wang@mediatek.com>

The patch adds support for aggregating more SKBs feed into NAPI in
order to get more benefits from generic receive offload (GRO) by
peeking at the RX ring status and moving more packets right before
returning from NAPI RX polling handler if NAPI budgets are still
available and some packets already present in hardware.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>
---
 drivers/net/ethernet/mediatek/mtk_eth_soc.c | 31 ++++++++++++++++-------------
 1 file changed, 17 insertions(+), 14 deletions(-)

diff --git a/drivers/net/ethernet/mediatek/mtk_eth_soc.c b/drivers/net/ethernet/mediatek/mtk_eth_soc.c
index a1bdb53..66fd45a 100644
--- a/drivers/net/ethernet/mediatek/mtk_eth_soc.c
+++ b/drivers/net/ethernet/mediatek/mtk_eth_soc.c

@@ -893,14 +893,14 @@ release_desc:
 
 		done++;
 	}
-	/* make sure that all changes to the dma ring are flushed before
-	 * we continue
-	 */
-	wmb();
-	mtk_w32(eth, ring->calc_idx, MTK_PRX_CRX_IDX0);
 
-	if (done < budget)
-		mtk_w32(eth, MTK_RX_DONE_INT, MTK_PDMA_INT_STATUS);
+	if (done) {
+		/* make sure that all changes to the dma ring are flushed before
+		 * we continue
+		 */
+		wmb();
+		mtk_w32(eth, ring->calc_idx, MTK_PRX_CRX_IDX0);
+	}
 
 	return done;
 }

@@ -1020,10 +1020,13 @@ static int mtk_napi_rx(struct napi_struct *napi, int budget)
 	struct mtk_eth *eth = container_of(napi, struct mtk_eth, rx_napi);
 	u32 status, mask;
 	int rx_done = 0;
+	int remain_budget = budget;
 
 	mtk_handle_status_irq(eth);
+
+poll_again:
 	mtk_w32(eth, MTK_RX_DONE_INT, MTK_PDMA_INT_STATUS);
-	rx_done = mtk_poll_rx(napi, budget, eth);
+	rx_done = mtk_poll_rx(napi, remain_budget, eth);
 
 	if (unlikely(netif_msg_intr(eth))) {
 		status = mtk_r32(eth, MTK_PDMA_INT_STATUS);

@@ -1032,18 +1035,18 @@ static int mtk_napi_rx(struct napi_struct *napi, int budget)
 			 "done rx %d, intr 0x%08x/0x%x\n",
 			 rx_done, status, mask);
 	}
-
-	if (rx_done == budget)
+	if (rx_done == remain_budget)
 		return budget;
 
 	status = mtk_r32(eth, MTK_PDMA_INT_STATUS);
-	if (status & MTK_RX_DONE_INT)
-		return budget;
-
+	if (status & MTK_RX_DONE_INT) {
+		remain_budget -= rx_done;
+		goto poll_again;
+	}
 	napi_complete(napi);
 	mtk_irq_enable(eth, MTK_PDMA_INT_MASK, MTK_RX_DONE_INT);
 
-	return rx_done;
+	return rx_done + budget - remain_budget;
 }
 
 static int mtk_tx_alloc(struct mtk_eth *eth)

-- 
1.9.1

On 03/09/2016 11:59, sean.wang@mediatek.com wrote:

From: Sean Wang <sean.wang@mediatek.com>

The patch makes move wmb() to outside the loop that could help
RX path handling more faster although that RX descriptors aren't
freed for DMA to use as soon as possible, but based on my experiment
and the result shows it still can reach about 943Mbpis without
performance drop that is tested based on the setup with one port
using Giga PHY and 256 RX descriptors for DMA to move.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>

gave it a quick spin on my hardware and it works well even when using
more than 1 gmac.

Acked-by: John Crispin <john@phrozen.org>

On 03/09/2016 11:59, sean.wang@mediatek.com wrote:

From: Sean Wang <sean.wang@mediatek.com>

The patch adds support for aggregating more SKBs feed into NAPI in
order to get more benefits from generic receive offload (GRO) by
peeking at the RX ring status and moving more packets right before
returning from NAPI RX polling handler if NAPI budgets are still
available and some packets already present in hardware.

Signed-off-by: Sean Wang <sean.wang@mediatek.com>

this looks ok to me, gave it a quick try on my HW and it works well. i
am just not sure if this is something that is commonly done. someone
with more overview of how NAPI should be used might want to comment on this.

Tested-by: John Crispin <john@phrozen.org>