diff mbox series

[net-next,1/1] net: stmmac: enhance XDP ZC driver level switching performance

Message ID 20211111143949.2806049-1-boon.leong.ong@intel.com (mailing list archive)
State Accepted
Commit ac746c8520d9d056b6963ecca8ff1da9929d02f1
Delegated to: Netdev Maintainers
Headers show
Series [net-next,1/1] net: stmmac: enhance XDP ZC driver level switching performance | 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 Single patches do not need cover letters
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: 21 this patch: 21
netdev/cc_maintainers warning 6 maintainers not CCed: songliubraving@fb.com yhs@fb.com kpsingh@kernel.org andrii@kernel.org ast@kernel.org kafai@fb.com
netdev/build_clang success Errors and warnings before: 2 this patch: 2
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: 21 this patch: 21
netdev/checkpatch success total: 0 errors, 0 warnings, 0 checks, 181 lines checked
netdev/kdoc success Errors and warnings before: 0 this patch: 0
netdev/source_inline success Was 0 now: 0

Commit Message

Ong Boon Leong Nov. 11, 2021, 2:39 p.m. UTC
The previous stmmac_xdp_set_prog() implementation uses stmmac_release()
and stmmac_open() which tear down the PHY device and causes undesirable
autonegotiation which causes a delay whenever AFXDP ZC is setup.

This patch introduces two new functions that just sufficiently tear
down DMA descriptors, buffer, NAPI process, and IRQs and reestablish
them accordingly in both stmmac_xdp_release() and stammac_xdp_open().

As the results of this enhancement, we get rid of transient state
introduced by the link auto-negotiation:

$ ./xdpsock -i eth0 -t -z

 sock0@eth0:0 txonly xdp-drv
                   pps            pkts           1.00
rx                 0              0
tx                 634444         634560

 sock0@eth0:0 txonly xdp-drv
                   pps            pkts           1.00
rx                 0              0
tx                 632330         1267072

 sock0@eth0:0 txonly xdp-drv
                   pps            pkts           1.00
rx                 0              0
tx                 632438         1899584

 sock0@eth0:0 txonly xdp-drv
                   pps            pkts           1.00
rx                 0              0
tx                 632502         2532160

Reported-by: Kurt Kanzenbach <kurt@linutronix.de>
Signed-off-by: Ong Boon Leong <boon.leong.ong@intel.com>
---
 drivers/net/ethernet/stmicro/stmmac/stmmac.h  |   4 +-
 .../net/ethernet/stmicro/stmmac/stmmac_main.c | 137 +++++++++++++++++-
 .../net/ethernet/stmicro/stmmac/stmmac_xdp.c  |   4 +-
 3 files changed, 139 insertions(+), 6 deletions(-)

Comments

Kurt Kanzenbach Nov. 14, 2021, 3:11 p.m. UTC | #1
On Thu Nov 11 2021, Ong Boon Leong wrote:
> The previous stmmac_xdp_set_prog() implementation uses stmmac_release()
> and stmmac_open() which tear down the PHY device and causes undesirable
> autonegotiation which causes a delay whenever AFXDP ZC is setup.
>
> This patch introduces two new functions that just sufficiently tear
> down DMA descriptors, buffer, NAPI process, and IRQs and reestablish
> them accordingly in both stmmac_xdp_release() and stammac_xdp_open().
>
> As the results of this enhancement, we get rid of transient state
> introduced by the link auto-negotiation:
>
> $ ./xdpsock -i eth0 -t -z
>
>  sock0@eth0:0 txonly xdp-drv
>                    pps            pkts           1.00
> rx                 0              0
> tx                 634444         634560
>
>  sock0@eth0:0 txonly xdp-drv
>                    pps            pkts           1.00
> rx                 0              0
> tx                 632330         1267072
>
>  sock0@eth0:0 txonly xdp-drv
>                    pps            pkts           1.00
> rx                 0              0
> tx                 632438         1899584
>
>  sock0@eth0:0 txonly xdp-drv
>                    pps            pkts           1.00
> rx                 0              0
> tx                 632502         2532160
>
> Reported-by: Kurt Kanzenbach <kurt@linutronix.de>
> Signed-off-by: Ong Boon Leong <boon.leong.ong@intel.com>

Tested-by: Kurt Kanzenbach <kurt@linutronix.de>
patchwork-bot+netdevbpf@kernel.org Nov. 15, 2021, 1:20 p.m. UTC | #2
Hello:

This patch was applied to netdev/net-next.git (master)
by David S. Miller <davem@davemloft.net>:

On Thu, 11 Nov 2021 22:39:49 +0800 you wrote:
> The previous stmmac_xdp_set_prog() implementation uses stmmac_release()
> and stmmac_open() which tear down the PHY device and causes undesirable
> autonegotiation which causes a delay whenever AFXDP ZC is setup.
> 
> This patch introduces two new functions that just sufficiently tear
> down DMA descriptors, buffer, NAPI process, and IRQs and reestablish
> them accordingly in both stmmac_xdp_release() and stammac_xdp_open().
> 
> [...]

Here is the summary with links:
  - [net-next,1/1] net: stmmac: enhance XDP ZC driver level switching performance
    https://git.kernel.org/netdev/net-next/c/ac746c8520d9

You are awesome, thank you!
diff mbox series

Patch

diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac.h b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
index 43eead726886..dd7adf9b2537 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac.h
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac.h
@@ -316,8 +316,8 @@  void stmmac_set_ethtool_ops(struct net_device *netdev);
 
 void stmmac_ptp_register(struct stmmac_priv *priv);
 void stmmac_ptp_unregister(struct stmmac_priv *priv);
-int stmmac_open(struct net_device *dev);
-int stmmac_release(struct net_device *dev);
+int stmmac_xdp_open(struct net_device *dev);
+void stmmac_xdp_release(struct net_device *dev);
 int stmmac_resume(struct device *dev);
 int stmmac_suspend(struct device *dev);
 int stmmac_dvr_remove(struct device *dev);
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
index d3f350c25b9b..033c35c09a54 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_main.c
@@ -3643,7 +3643,7 @@  static int stmmac_request_irq(struct net_device *dev)
  *  0 on success and an appropriate (-)ve integer as defined in errno.h
  *  file on failure.
  */
-int stmmac_open(struct net_device *dev)
+static int stmmac_open(struct net_device *dev)
 {
 	struct stmmac_priv *priv = netdev_priv(dev);
 	int mode = priv->plat->phy_interface;
@@ -3767,7 +3767,7 @@  static void stmmac_fpe_stop_wq(struct stmmac_priv *priv)
  *  Description:
  *  This is the stop entry point of the driver.
  */
-int stmmac_release(struct net_device *dev)
+static int stmmac_release(struct net_device *dev)
 {
 	struct stmmac_priv *priv = netdev_priv(dev);
 	u32 chan;
@@ -6429,6 +6429,139 @@  void stmmac_enable_tx_queue(struct stmmac_priv *priv, u32 queue)
 	spin_unlock_irqrestore(&ch->lock, flags);
 }
 
+void stmmac_xdp_release(struct net_device *dev)
+{
+	struct stmmac_priv *priv = netdev_priv(dev);
+	u32 chan;
+
+	/* Disable NAPI process */
+	stmmac_disable_all_queues(priv);
+
+	for (chan = 0; chan < priv->plat->tx_queues_to_use; chan++)
+		hrtimer_cancel(&priv->tx_queue[chan].txtimer);
+
+	/* Free the IRQ lines */
+	stmmac_free_irq(dev, REQ_IRQ_ERR_ALL, 0);
+
+	/* Stop TX/RX DMA channels */
+	stmmac_stop_all_dma(priv);
+
+	/* Release and free the Rx/Tx resources */
+	free_dma_desc_resources(priv);
+
+	/* Disable the MAC Rx/Tx */
+	stmmac_mac_set(priv, priv->ioaddr, false);
+
+	/* set trans_start so we don't get spurious
+	 * watchdogs during reset
+	 */
+	netif_trans_update(dev);
+	netif_carrier_off(dev);
+}
+
+int stmmac_xdp_open(struct net_device *dev)
+{
+	struct stmmac_priv *priv = netdev_priv(dev);
+	u32 rx_cnt = priv->plat->rx_queues_to_use;
+	u32 tx_cnt = priv->plat->tx_queues_to_use;
+	u32 dma_csr_ch = max(rx_cnt, tx_cnt);
+	struct stmmac_rx_queue *rx_q;
+	struct stmmac_tx_queue *tx_q;
+	u32 buf_size;
+	bool sph_en;
+	u32 chan;
+	int ret;
+
+	ret = alloc_dma_desc_resources(priv);
+	if (ret < 0) {
+		netdev_err(dev, "%s: DMA descriptors allocation failed\n",
+			   __func__);
+		goto dma_desc_error;
+	}
+
+	ret = init_dma_desc_rings(dev, GFP_KERNEL);
+	if (ret < 0) {
+		netdev_err(dev, "%s: DMA descriptors initialization failed\n",
+			   __func__);
+		goto init_error;
+	}
+
+	/* DMA CSR Channel configuration */
+	for (chan = 0; chan < dma_csr_ch; chan++)
+		stmmac_init_chan(priv, priv->ioaddr, priv->plat->dma_cfg, chan);
+
+	/* Adjust Split header */
+	sph_en = (priv->hw->rx_csum > 0) && priv->sph;
+
+	/* DMA RX Channel Configuration */
+	for (chan = 0; chan < rx_cnt; chan++) {
+		rx_q = &priv->rx_queue[chan];
+
+		stmmac_init_rx_chan(priv, priv->ioaddr, priv->plat->dma_cfg,
+				    rx_q->dma_rx_phy, chan);
+
+		rx_q->rx_tail_addr = rx_q->dma_rx_phy +
+				     (rx_q->buf_alloc_num *
+				      sizeof(struct dma_desc));
+		stmmac_set_rx_tail_ptr(priv, priv->ioaddr,
+				       rx_q->rx_tail_addr, chan);
+
+		if (rx_q->xsk_pool && rx_q->buf_alloc_num) {
+			buf_size = xsk_pool_get_rx_frame_size(rx_q->xsk_pool);
+			stmmac_set_dma_bfsize(priv, priv->ioaddr,
+					      buf_size,
+					      rx_q->queue_index);
+		} else {
+			stmmac_set_dma_bfsize(priv, priv->ioaddr,
+					      priv->dma_buf_sz,
+					      rx_q->queue_index);
+		}
+
+		stmmac_enable_sph(priv, priv->ioaddr, sph_en, chan);
+	}
+
+	/* DMA TX Channel Configuration */
+	for (chan = 0; chan < tx_cnt; chan++) {
+		tx_q = &priv->tx_queue[chan];
+
+		stmmac_init_tx_chan(priv, priv->ioaddr, priv->plat->dma_cfg,
+				    tx_q->dma_tx_phy, chan);
+
+		tx_q->tx_tail_addr = tx_q->dma_tx_phy;
+		stmmac_set_tx_tail_ptr(priv, priv->ioaddr,
+				       tx_q->tx_tail_addr, chan);
+	}
+
+	/* Enable the MAC Rx/Tx */
+	stmmac_mac_set(priv, priv->ioaddr, true);
+
+	/* Start Rx & Tx DMA Channels */
+	stmmac_start_all_dma(priv);
+
+	stmmac_init_coalesce(priv);
+
+	ret = stmmac_request_irq(dev);
+	if (ret)
+		goto irq_error;
+
+	/* Enable NAPI process*/
+	stmmac_enable_all_queues(priv);
+	netif_carrier_on(dev);
+	netif_tx_start_all_queues(dev);
+
+	return 0;
+
+irq_error:
+	for (chan = 0; chan < priv->plat->tx_queues_to_use; chan++)
+		hrtimer_cancel(&priv->tx_queue[chan].txtimer);
+
+	stmmac_hw_teardown(dev);
+init_error:
+	free_dma_desc_resources(priv);
+dma_desc_error:
+	return ret;
+}
+
 int stmmac_xsk_wakeup(struct net_device *dev, u32 queue, u32 flags)
 {
 	struct stmmac_priv *priv = netdev_priv(dev);
diff --git a/drivers/net/ethernet/stmicro/stmmac/stmmac_xdp.c b/drivers/net/ethernet/stmicro/stmmac/stmmac_xdp.c
index 2a616c6f7cd0..9d4d8c3dad0a 100644
--- a/drivers/net/ethernet/stmicro/stmmac/stmmac_xdp.c
+++ b/drivers/net/ethernet/stmicro/stmmac/stmmac_xdp.c
@@ -119,7 +119,7 @@  int stmmac_xdp_set_prog(struct stmmac_priv *priv, struct bpf_prog *prog,
 
 	need_update = !!priv->xdp_prog != !!prog;
 	if (if_running && need_update)
-		stmmac_release(dev);
+		stmmac_xdp_release(dev);
 
 	old_prog = xchg(&priv->xdp_prog, prog);
 	if (old_prog)
@@ -129,7 +129,7 @@  int stmmac_xdp_set_prog(struct stmmac_priv *priv, struct bpf_prog *prog,
 	priv->sph = priv->sph_cap && !stmmac_xdp_is_enabled(priv);
 
 	if (if_running && need_update)
-		stmmac_open(dev);
+		stmmac_xdp_open(dev);
 
 	return 0;
 }